Datasets:
J0nasW
/
paperswithcode

Dataset card Data Studio Files
xet
Community
1
paperID
stringlengths
36
36
pwc_id
stringlengths
8
47
arxiv_id
stringlengths
6
16
⌀
nips_id
float64
url_abs
stringlengths
18
329
url_pdf
stringlengths
18
742
title
stringlengths
8
325
abstract
stringlengths
1
7.27k
⌀
authors
stringlengths
2
7.06k
published
stringlengths
10
10
⌀
conference
stringlengths
12
47
⌀
conference_url_abs
stringlengths
16
198
⌀
conference_url_pdf
stringlengths
27
199
⌀
proceeding
stringlengths
6
47
⌀
taskID
stringlengths
7
1.44k
areaID
stringclasses
688 values
embedding
stringlengths
9.26k
12.5k
umap_embedding
stringlengths
29
44
34ac490c-e281-431e-a875-06981ace4392
rethinking-translation-memory-augmented
2306.06948
null
https://arxiv.org/abs/2306.06948v1
https://arxiv.org/pdf/2306.06948v1.pdf
Rethinking Translation Memory Augmented Neural Machine Translation
This paper rethinks translation memory augmented neural machine translation (TM-augmented NMT) from two perspectives, i.e., a probabilistic view of retrieval and the variance-bias decomposition principle. The finding demonstrates that TM-augmented NMT is good at the ability of fitting data (i.e., lower bias) but is more sensitive to the fluctuations in the training data (i.e., higher variance), which provides an explanation to a recently reported contradictory phenomenon on the same translation task: TM-augmented NMT substantially advances vanilla NMT under the high-resource scenario whereas it fails under the low-resource scenario. Then we propose a simple yet effective TM-augmented NMT model to promote the variance and address the contradictory phenomenon. Extensive experiments show that the proposed TM-augmented NMT achieves consistent gains over both conventional NMT and existing TM-augmented NMT under two variance-preferable (low-resource and plug-and-play) scenarios as well as the high-resource scenario.
['Rui Wang', 'Shuming Shi', 'Zhirui Zhang', 'Lemao Liu', 'Guoping Huang', 'Hongkun Hao']
2023-06-12
null
null
null
null
['nmt']
['computer-code']
[ 4.05098855e-01 -2.66616583e-01 -6.81865811e-01 -5.21741398e-02 -1.21208763e+00 -2.04908416e-01 9.49769676e-01 -5.90770841e-01 -4.03900445e-01 8.06268930e-01 2.85619497e-01 -8.20697665e-01 -1.16103590e-01 -4.33498591e-01 -9.10238385e-01 -7.55554795e-01 4.42228019e-01 7.76880383e-01 -2.03159779e-01 -5.17095685e-01 2.24035606e-01 1.95180383e-02 -1.11836934e+00 3.03756386e-01 1.13110554e+00 4.09108073e-01 6.52184725e-01 2.16808736e-01 -2.38711149e-01 4.15782809e-01 -4.45016384e-01 -3.95708174e-01 5.48833489e-01 -5.02743542e-01 -4.32451397e-01 -3.49617690e-01 1.83764547e-01 -2.62187123e-01 -3.19039464e-01 1.00662959e+00 6.23663485e-01 3.22618723e-01 5.36410391e-01 -1.04677582e+00 -1.33647072e+00 8.10185313e-01 -8.44433784e-01 4.52890694e-01 -9.31546018e-02 -2.42821183e-02 8.24746847e-01 -1.38637161e+00 4.96751100e-01 1.39677036e+00 3.98402572e-01 7.27188170e-01 -1.12105119e+00 -5.77740133e-01 2.28002757e-01 -9.78047848e-02 -1.16962779e+00 -6.29701912e-01 4.79685128e-01 -6.69845864e-02 1.07397521e+00 3.80717367e-01 1.06693439e-01 1.46476364e+00 4.72013474e-01 9.32682812e-01 1.53638756e+00 -7.77093291e-01 3.32174711e-02 3.01493615e-01 1.48526102e-01 2.26069912e-01 4.06868339e-01 2.91787922e-01 -8.80066335e-01 -3.66266012e-01 5.63024998e-01 2.39791870e-02 -2.19402447e-01 -1.59780785e-01 -1.46011865e+00 6.43688381e-01 -2.98915170e-02 3.81315529e-01 -4.61766630e-01 1.98742688e-01 2.78252363e-01 5.57924926e-01 8.13643396e-01 4.23764169e-01 -8.59640837e-01 -3.51966560e-01 -1.15979207e+00 1.07404597e-01 2.98691303e-01 1.25079203e+00 5.20572543e-01 3.43971670e-01 -5.43323517e-01 1.07574654e+00 3.54399085e-01 1.03003681e+00 9.39091623e-01 -4.59957153e-01 9.39413786e-01 8.24730173e-02 1.20157979e-01 -6.32487357e-01 -1.12081885e-01 -8.32863271e-01 -8.73624325e-01 -3.94702017e-01 1.19304955e-01 -1.06710248e-01 -1.15677667e+00 2.37584019e+00 6.83665499e-02 2.37471592e-02 2.05465451e-01 9.84469116e-01 4.34779942e-01 7.67958283e-01 -2.25330219e-01 -6.62961960e-01 1.29029429e+00 -1.25494206e+00 -1.22767234e+00 -4.96207297e-01 7.87519097e-01 -9.46853399e-01 1.57807136e+00 -6.14289790e-02 -1.09172595e+00 -2.45146856e-01 -9.61800694e-01 4.36082631e-02 -3.00572962e-01 3.41387361e-01 6.42569423e-01 8.00769389e-01 -9.73745763e-01 4.19471055e-01 -8.62868607e-01 -4.34822649e-01 1.27421662e-01 2.85049140e-01 -3.38280238e-02 -1.58595026e-01 -1.38822985e+00 1.18741536e+00 5.02591059e-02 2.35063881e-01 -6.99875236e-01 -5.04081070e-01 -2.89707363e-01 2.12581381e-02 4.77234006e-01 -9.90129411e-01 1.34223413e+00 -9.60748911e-01 -1.72884965e+00 3.21121037e-01 -6.45293593e-01 -2.53354877e-01 5.56030393e-01 -5.24919868e-01 -2.74036825e-01 -4.61276323e-01 -7.56848603e-03 2.95628637e-01 8.84525299e-01 -1.15018678e+00 -1.51315302e-01 -6.03869617e-01 -4.70453262e-01 4.75188136e-01 -6.23970151e-01 -2.07917951e-02 -4.61066246e-01 -9.90553141e-01 6.54332876e-01 -1.22578263e+00 4.03364599e-02 -6.19988680e-01 -4.35702473e-01 -3.39002721e-02 6.14619076e-01 -5.48579097e-01 1.27879739e+00 -1.85785306e+00 2.60674983e-01 -5.35074212e-02 8.43448564e-02 2.06689924e-01 -3.56087863e-01 6.15738630e-01 -8.14467520e-02 3.45761299e-01 5.56626879e-02 -5.87508380e-01 1.89613551e-02 3.45827550e-01 -3.44928622e-01 2.28722766e-01 6.40532151e-02 1.01144683e+00 -7.50651181e-01 -3.08840573e-01 -3.12036842e-01 2.73994029e-01 -3.44664454e-01 7.95990676e-02 -2.83761442e-01 3.18754703e-01 -3.59892607e-01 7.55324364e-01 7.54811168e-01 -2.94681758e-01 3.77291769e-01 3.27951983e-02 -2.73649246e-02 3.80490512e-01 -6.42793238e-01 1.80365169e+00 -4.30121183e-01 6.39661014e-01 -3.43541652e-01 -5.40415764e-01 8.36365342e-01 4.32408601e-01 2.06426576e-01 -9.69258606e-01 -3.31489253e-03 5.36319375e-01 1.16799138e-01 -2.97243714e-01 1.08042765e+00 -2.60858744e-01 2.97429085e-01 9.22595561e-01 -1.27569988e-01 5.72067142e-01 -4.11365300e-01 2.81215578e-01 7.91066051e-01 4.45409089e-01 3.44326794e-02 -4.93596613e-01 -5.39249256e-02 -5.66580966e-02 5.90491652e-01 9.14953828e-01 -2.01961994e-01 2.77313381e-01 7.37975985e-02 -1.49750993e-01 -1.22757459e+00 -7.16505110e-01 4.46580797e-02 1.53437853e+00 1.82236508e-01 -3.22825104e-01 -6.96503103e-01 -4.70548838e-01 -2.85266221e-01 7.76917398e-01 -5.53892791e-01 -3.89082670e-01 -5.76378167e-01 -1.37817442e+00 5.79553306e-01 3.90300542e-01 5.59722304e-01 -4.87191916e-01 -3.14055800e-01 -1.49658695e-01 -6.61612272e-01 -9.38297868e-01 -6.88108563e-01 2.85919756e-01 -1.29012644e+00 -1.68285090e-02 -7.53187418e-01 -4.52446967e-01 6.38063133e-01 8.66121292e-01 1.01985991e+00 -1.33880839e-01 5.84137261e-01 -1.39875621e-01 -3.86270761e-01 -5.59819520e-01 -4.04546410e-01 2.56702930e-01 4.29862350e-01 -1.28844500e-01 5.24187624e-01 -5.64331889e-01 -4.21691090e-01 3.29462171e-01 -8.11887145e-01 3.12014490e-01 1.13241494e+00 1.36681366e+00 5.24990797e-01 -2.70504862e-01 6.63718402e-01 -7.00879812e-01 1.07932997e+00 -5.41396797e-01 -1.17092080e-01 5.81196010e-01 -1.38454556e+00 2.36679927e-01 3.55970234e-01 -1.06870973e+00 -1.08011007e+00 -5.27985632e-01 3.80191982e-01 -6.94482923e-01 6.46221936e-01 7.15451598e-01 -1.48341849e-01 2.52304286e-01 4.93126243e-01 5.99880755e-01 3.27495001e-02 -6.41740024e-01 4.99839216e-01 7.93472052e-01 1.58072472e-01 -8.72760475e-01 8.17944646e-01 3.86688523e-02 -1.18619755e-01 -3.63297760e-01 -3.63259196e-01 -5.32599837e-02 -3.10945809e-01 2.51068398e-02 3.36971790e-01 -1.01226735e+00 -1.12009970e-02 2.55915612e-01 -1.09933305e+00 -1.00953653e-01 6.04244359e-02 7.44616449e-01 -6.06850088e-01 2.72071183e-01 -7.76537240e-01 -1.08451080e+00 -7.54289091e-01 -1.44285595e+00 1.04186404e+00 -1.90043136e-01 -5.14351390e-02 -6.69601023e-01 -5.17241471e-03 5.82330287e-01 8.39099705e-01 -4.15617049e-01 1.38394773e+00 -8.63117754e-01 -5.34604847e-01 3.08560245e-02 -3.46708655e-01 1.34195924e-01 -2.45813038e-02 -2.01138794e-01 -9.26432312e-01 -5.77627957e-01 1.46244079e-01 -6.55667782e-02 6.16018474e-01 3.21701854e-01 4.70874608e-01 -4.41693157e-01 -1.89443544e-01 3.62862051e-01 1.27906275e+00 1.86172649e-01 5.42284250e-01 5.27236700e-01 4.98443246e-01 2.82826424e-01 6.64575934e-01 1.40267506e-01 2.02550247e-01 1.04492283e+00 2.26665080e-01 -1.78422675e-01 -5.68145812e-02 -3.80647600e-01 6.10820591e-01 1.49092114e+00 -1.94043797e-02 -5.17681479e-01 -7.51048803e-01 2.75509387e-01 -2.28283167e+00 -5.43469250e-01 -5.67495003e-02 2.41913056e+00 9.14469719e-01 8.53190199e-02 -5.29677458e-02 -2.60636330e-01 6.51177526e-01 3.33541855e-02 -5.03433824e-01 -5.46857595e-01 -4.25462991e-01 -9.68147144e-02 5.40984571e-01 2.87695736e-01 -4.30192471e-01 1.10704827e+00 6.75864172e+00 1.10998905e+00 -9.65910494e-01 7.58000016e-01 5.42504430e-01 -3.08181077e-01 -4.89895076e-01 8.76475647e-02 -4.56769705e-01 6.29412234e-01 1.32942379e+00 -3.87873769e-01 5.85107386e-01 6.37416065e-01 2.92296767e-01 2.43227303e-01 -1.15281689e+00 9.51838851e-01 1.53406009e-01 -1.22093129e+00 5.73958695e-01 4.89124954e-01 7.08360016e-01 2.37608641e-01 4.85665143e-01 5.60055792e-01 1.38742417e-01 -7.79384911e-01 8.70001614e-01 2.84012109e-01 8.05718601e-01 -4.75851238e-01 7.89861977e-01 7.36805201e-01 -5.39488733e-01 -2.63140742e-02 -4.40918267e-01 5.30162491e-02 -1.79800138e-01 3.53465348e-01 -4.63747025e-01 7.86222219e-01 3.47128749e-01 1.40575811e-01 -2.51653165e-01 3.40945691e-01 3.28657389e-01 6.87401295e-01 -1.87782347e-01 -6.69519082e-02 2.97687501e-01 -3.12186509e-01 6.86920762e-01 9.82913613e-01 6.39319956e-01 -1.63462758e-01 -5.47822863e-02 8.72553051e-01 -1.98719263e-01 3.36595535e-01 -5.64045131e-01 -1.46806180e-01 7.40908563e-01 6.98755264e-01 -3.35307628e-01 -4.15051162e-01 -2.77727842e-01 1.09061337e+00 2.95208246e-01 7.26667464e-01 -8.56399000e-01 4.10146087e-01 5.48104227e-01 -1.09619968e-01 5.18675074e-02 -2.55420744e-01 -4.70020831e-01 -1.40542495e+00 2.04742014e-01 -1.13364255e+00 -1.37446029e-02 -8.83265853e-01 -1.03967345e+00 8.29812527e-01 3.74024436e-02 -1.30962753e+00 -1.91951677e-01 -4.35269475e-01 -1.02332875e-01 1.34353328e+00 -1.51459408e+00 -1.25289106e+00 2.87948787e-01 3.31453025e-01 8.06595504e-01 -3.63736063e-01 9.69745994e-01 4.91990983e-01 -9.54244792e-01 1.07321787e+00 7.84129500e-01 -3.22699726e-01 9.31782424e-01 -7.08951592e-01 3.58307987e-01 1.02871835e+00 2.07689807e-01 1.28544104e+00 8.02568197e-01 -6.38379753e-01 -2.06183314e+00 -9.67294931e-01 1.09412384e+00 -7.33326852e-01 3.38525534e-01 -3.84161979e-01 -7.70322323e-01 6.52414203e-01 2.37665445e-01 -5.74435294e-01 7.33849883e-01 2.24752545e-01 -6.70340717e-01 1.11205049e-01 -9.71323788e-01 9.85357583e-01 9.45050657e-01 -6.97024643e-01 -6.20462656e-01 5.21769524e-01 1.04649079e+00 -4.20083344e-01 -5.73977649e-01 4.99412060e-01 6.85937166e-01 -5.11818171e-01 6.80423856e-01 -9.79348361e-01 5.97227693e-01 -1.33280829e-03 -8.17011952e-01 -1.41491997e+00 -5.68399489e-01 -9.62108374e-01 -4.10051018e-01 9.54754770e-01 7.19757318e-01 -8.82050097e-01 5.79642713e-01 4.52716172e-01 -1.34992108e-01 -1.10219598e+00 -1.21979105e+00 -1.16654062e+00 4.26119864e-01 -3.91336679e-01 8.63162875e-01 1.12186038e+00 -1.25856623e-01 5.76129913e-01 -9.02041972e-01 2.07189452e-02 3.50880861e-01 3.93172987e-02 6.90558493e-01 -7.56019473e-01 -4.86485958e-01 -3.96380424e-01 2.44965926e-01 -1.43745661e+00 -1.15553811e-01 -7.98386157e-01 -3.32335122e-02 -1.30261934e+00 7.06829429e-01 -3.71242285e-01 -5.66923141e-01 3.27021182e-01 -5.44662356e-01 6.90883547e-02 2.86948711e-01 9.85506296e-01 -2.06314892e-01 6.54745042e-01 1.20103788e+00 -4.65100370e-02 -1.82138801e-01 -1.42825752e-01 -9.27343249e-01 4.38913107e-02 7.08988309e-01 -5.73052168e-01 -5.93761206e-01 -8.80262136e-01 3.51477742e-01 1.43158920e-02 -2.13475391e-01 -3.69194955e-01 4.64442447e-02 -3.80776405e-01 -2.56495085e-03 -6.73591971e-01 5.06192446e-01 -8.40146542e-01 1.37673810e-01 2.09425762e-01 -5.29760778e-01 7.42351890e-01 1.49806336e-01 7.41467178e-01 2.19138950e-01 2.77108818e-01 3.18992138e-01 -1.37984633e-01 -1.75687417e-01 1.10362940e-01 -3.60730886e-01 -1.71949118e-01 4.54051882e-01 -2.25169107e-01 -9.16888118e-01 -5.22495747e-01 -1.63263604e-01 6.60835505e-02 2.87690341e-01 7.53934622e-01 5.34156263e-01 -1.52074623e+00 -7.96058178e-01 2.82324016e-01 2.40446776e-01 -6.13864720e-01 2.46554986e-02 1.36277163e+00 2.17550933e-01 6.65233135e-01 -2.00767647e-02 -6.23011053e-01 -1.16101444e+00 5.96518636e-01 2.17075139e-01 -4.23223227e-01 -3.52422655e-01 6.19226456e-01 1.32215545e-01 -5.50766110e-01 1.54793456e-01 -2.39853144e-01 4.13076878e-01 -3.06195021e-01 2.80746430e-01 2.95544714e-01 2.89507240e-01 -6.66897893e-01 -1.79211572e-01 2.12992132e-01 -4.30379897e-01 -4.29832608e-01 8.41779411e-01 -3.93127918e-01 -2.63753533e-01 6.58653378e-01 7.33185053e-01 -1.81831911e-01 -4.65932220e-01 -8.62718821e-01 3.22402120e-02 -7.07597196e-01 2.73266256e-01 -1.12845778e+00 -7.89904892e-01 7.95007229e-01 6.51450753e-01 -2.24857032e-01 1.06221497e+00 -4.95377392e-01 9.77189243e-01 5.65559328e-01 5.19169271e-01 -1.05043519e+00 -3.20704211e-03 5.72340250e-01 5.36536694e-01 -1.12831295e+00 1.90029055e-01 -2.47497812e-01 -6.40700340e-01 7.28523433e-01 5.69070995e-01 5.92346311e-01 3.23926985e-01 8.19382966e-02 2.92055219e-01 6.30143890e-03 -1.35282731e+00 1.81766421e-01 2.46689722e-01 4.27376539e-01 6.03008628e-01 2.10503802e-01 -6.42258883e-01 5.08138061e-01 2.38517355e-02 -2.74482071e-01 3.09007198e-01 9.51163113e-01 -1.07479379e-01 -1.11227989e+00 -4.42985415e-01 4.19039994e-01 -5.05761504e-01 -6.15022123e-01 -6.67279541e-01 8.20417106e-01 -2.08025798e-01 1.02119172e+00 -2.22710267e-01 -9.07937884e-01 8.45756531e-02 4.96612340e-01 3.57807279e-01 -4.78617042e-01 -6.12864316e-01 5.63326895e-01 5.65748811e-02 -3.49017888e-01 -3.92824411e-01 -3.70010108e-01 -6.59473598e-01 -5.49867094e-01 -8.58799636e-01 2.85423994e-01 9.98749793e-01 9.85549569e-01 8.20939720e-01 3.73647422e-01 6.17572069e-01 -6.21533215e-01 -1.28344929e+00 -1.40049386e+00 -2.54614979e-01 2.23201036e-01 2.08238572e-01 -6.05186403e-01 -3.47027838e-01 -4.49048996e-01]
[11.634973526000977, 10.160614013671875]
b9a2e528-8768-4ab1-98cf-3e66202a265b
convolutional-tensor-train-lstm-for-long-term
null
null
https://openreview.net/forum?id=Hkee1JBKwB
https://openreview.net/pdf?id=Hkee1JBKwB
Convolutional Tensor-Train LSTM for Long-Term Video Prediction
Long-term video prediction is highly challenging since it entails simultaneously capturing spatial and temporal information across a long range of image frames.Standard recurrent models are ineffective since they are prone to error propagation and cannot effectively capture higher-order correlations. A potential solution is to extend to higher-order spatio-temporal recurrent models. However, such a model requires a large number of parameters and operations, making it intractable to learn in practice and is prone to overfitting. In this work, we propose convolutional tensor-train LSTM (Conv-TT-LSTM), which learns higher-orderConvolutional LSTM (ConvLSTM) efficiently using convolutional tensor-train decomposition (CTTD). Our proposed model naturally incorporates higher-order spatio-temporal information at a small cost of memory and computation by using efficient low-rank tensor representations. We evaluate our model on Moving-MNIST and KTH datasets and show improvements over standard ConvLSTM and better/comparable results to other ConvLSTM-based approaches, but with much fewer parameters.
['Animashree Anandkumar', 'Jan Kautz', 'Furong Huang', 'Wonmin Byeon', 'Jiahao Su']
2019-09-25
null
null
null
null
['video-prediction']
['computer-vision']
[-1.92173824e-01 -6.20735288e-01 -2.58885264e-01 -1.19569853e-01 -7.10438430e-01 -2.89086908e-01 5.73053360e-01 -1.98014110e-01 -3.16687465e-01 3.94700557e-01 5.29568970e-01 -3.41806740e-01 -3.19526136e-01 -5.12686253e-01 -7.55235434e-01 -6.28679037e-01 -5.13665795e-01 2.01229677e-01 3.21635306e-01 -4.81081344e-02 8.44796672e-02 3.66468817e-01 -1.18175936e+00 9.15965855e-01 4.62668896e-01 1.15460181e+00 2.44517773e-01 9.92468655e-01 1.15286019e-02 1.57608187e+00 -2.91243754e-02 -3.87001425e-01 -1.05142072e-01 -1.97247371e-01 -1.06313837e+00 1.43440142e-01 6.80514812e-01 -5.56807756e-01 -7.27244318e-01 6.38206005e-01 6.53785691e-02 2.93701977e-01 2.61119902e-01 -8.37313473e-01 -5.77415228e-01 4.78635132e-01 -4.95895028e-01 6.27811909e-01 1.16880797e-02 -6.84697703e-02 1.33437419e+00 -9.12216425e-01 4.71296251e-01 1.09026897e+00 9.43460405e-01 1.44114956e-01 -1.38744915e+00 -3.83018792e-01 3.15149844e-01 5.16990006e-01 -1.08423960e+00 -2.34721974e-01 7.62909532e-01 -5.24956584e-01 1.34578180e+00 2.24275708e-01 8.34730983e-01 1.34818542e+00 3.66026819e-01 1.06450045e+00 8.37861478e-01 3.17971148e-02 -1.44263268e-01 -6.35072231e-01 7.81695917e-02 7.92279065e-01 -2.80604064e-01 8.11497420e-02 -7.63556659e-01 -2.46519431e-01 1.05695534e+00 6.51269913e-01 -1.77646410e-02 -1.70476064e-01 -1.69443083e+00 6.59360766e-01 6.13340020e-01 6.55360043e-01 -7.10915029e-01 7.95533121e-01 6.78971529e-01 4.01651144e-01 7.72721112e-01 2.68066704e-01 -5.43300331e-01 -5.22851765e-01 -1.18562734e+00 1.92854807e-01 2.19588518e-01 5.89491248e-01 6.66840851e-01 2.76246279e-01 -2.59154290e-01 9.26270187e-01 7.40515068e-02 2.38624394e-01 6.11521482e-01 -1.07592583e+00 6.06953263e-01 4.02185261e-01 -3.84244584e-02 -1.25353241e+00 -4.28127319e-01 -6.68696642e-01 -1.21397328e+00 -3.26642334e-01 1.13240980e-01 -2.47285832e-02 -8.85450900e-01 1.33054221e+00 -1.04352064e-01 6.88778102e-01 -9.02156159e-02 9.64439452e-01 4.33374584e-01 8.91105056e-01 -7.08454773e-02 -1.89347774e-01 1.15883362e+00 -1.17191088e+00 -7.11122274e-01 7.59376660e-02 1.05271530e+00 -8.27628732e-01 9.25562203e-01 3.21088105e-01 -9.43890870e-01 -3.99802387e-01 -5.94868243e-01 -2.45610550e-01 2.85956915e-02 5.26763380e-01 9.83993411e-01 -1.59366667e-01 -1.24550676e+00 9.06841815e-01 -1.32685578e+00 -3.32197696e-01 2.30088368e-01 2.82282680e-01 -1.89580202e-01 -7.43693858e-02 -8.85847807e-01 4.78900820e-01 9.09895822e-02 4.67360437e-01 -9.67652977e-01 -7.46452510e-01 -6.38577998e-01 -8.80901739e-02 1.54768333e-01 -6.57419622e-01 1.22315228e+00 -5.54680765e-01 -1.40452373e+00 1.09379791e-01 -4.75157559e-01 -6.01576269e-01 3.26677233e-01 -4.56075728e-01 -2.27025166e-01 1.98240265e-01 -9.34851244e-02 4.62454498e-01 8.55798721e-01 -7.18551695e-01 -4.01013970e-01 -2.60149896e-01 3.31400782e-01 2.01608285e-01 -6.26451492e-01 4.05390151e-02 -3.95385027e-01 -1.11518061e+00 3.63237381e-01 -1.33478606e+00 -4.94720578e-01 -1.49358124e-01 -3.29713643e-01 -1.26039639e-01 1.08661819e+00 -5.61771572e-01 1.23459136e+00 -1.96263552e+00 3.86218429e-01 -2.68407404e-01 3.61979514e-01 3.43067259e-01 -3.07347506e-01 5.91916621e-01 -7.02399909e-02 2.96295285e-02 1.92281470e-01 -7.44257569e-01 -2.51645058e-01 5.49490392e-01 -6.64678276e-01 2.43829295e-01 1.98343426e-01 1.09067154e+00 -9.76041377e-01 -2.71528274e-01 3.45191091e-01 7.96528220e-01 -7.15763807e-01 -1.63409095e-02 -3.38294178e-01 4.25653189e-01 -4.78357077e-01 4.65698063e-01 2.05036700e-01 -7.47165740e-01 6.96503669e-02 -5.34547091e-01 -2.78902143e-01 4.37350333e-01 -6.41501606e-01 1.84360600e+00 -6.29182816e-01 9.15193319e-01 -6.65315986e-01 -1.05021513e+00 4.89571482e-01 4.97682780e-01 7.92673051e-01 -5.14747202e-01 -1.63491532e-01 1.43269002e-01 -1.71004742e-01 -5.67783237e-01 7.89835095e-01 -4.85545956e-02 1.96495965e-01 6.22497201e-01 1.01933941e-01 5.21793783e-01 1.48607522e-01 3.16178471e-01 1.25563753e+00 1.26617491e-01 -4.42810416e-01 -1.28584041e-04 2.26332590e-01 -1.28466845e-01 5.00018179e-01 3.68658841e-01 2.37835586e-01 4.90480840e-01 4.01513129e-01 -9.77083027e-01 -1.19066751e+00 -6.30349040e-01 2.03775465e-01 1.24541712e+00 -4.46819335e-01 -6.58540845e-01 -2.29270607e-01 -3.79490167e-01 -3.37780833e-01 2.53983468e-01 -7.28408754e-01 3.06089446e-02 -8.95177722e-01 -8.80254328e-01 6.43984556e-01 7.17612684e-01 4.96912420e-01 -5.74447572e-01 -4.93042290e-01 3.33196253e-01 -6.13582492e-01 -1.48780489e+00 -5.40260911e-01 -1.00843675e-01 -1.53609467e+00 -6.66886449e-01 -9.32084501e-01 -5.39161444e-01 3.92706841e-01 6.25000775e-01 9.40407336e-01 6.67362437e-02 -6.60002753e-02 9.46652703e-03 -4.71894562e-01 3.51678073e-01 2.34471858e-01 1.42194897e-01 2.51396996e-04 3.21752518e-01 1.78823903e-01 -8.69539738e-01 -6.18008971e-01 3.09324533e-01 -7.88176775e-01 1.69889972e-01 6.49231672e-01 1.09672296e+00 5.74703097e-01 -2.97350138e-02 -1.42486334e-01 -6.03742003e-01 3.48225147e-01 -2.31121078e-01 -3.83915812e-01 1.96288511e-01 -1.53024375e-01 2.58798867e-01 8.37159932e-01 -6.90766811e-01 -8.40693414e-01 -2.80192997e-02 1.73286840e-01 -1.18381906e+00 3.56258154e-01 9.40747440e-01 8.21947336e-01 -1.48498714e-01 2.18038693e-01 4.67438936e-01 -3.55179340e-01 -7.49627352e-01 1.20637722e-01 -6.85204715e-02 1.91204533e-01 -4.63123679e-01 6.03700995e-01 7.80113995e-01 1.54900402e-01 -1.04886651e+00 -1.06665814e+00 -3.41704458e-01 -8.55892718e-01 -1.52819112e-01 7.23245263e-01 -1.04308164e+00 -7.68241763e-01 4.69733000e-01 -1.33007467e+00 -4.60273147e-01 5.27649187e-03 9.59722340e-01 -3.34941804e-01 3.40958655e-01 -1.22484207e+00 -5.94660699e-01 -1.79426089e-01 -1.06135809e+00 1.06636441e+00 -5.12945831e-01 2.69603003e-02 -1.25160980e+00 1.62227929e-01 3.44151258e-01 5.35256684e-01 1.97941363e-01 6.97672784e-01 5.11431172e-02 -1.02985287e+00 -2.06046581e-01 -2.84302831e-01 1.68455556e-01 -1.65166110e-01 1.47327662e-01 -6.48538411e-01 -3.51701081e-01 -1.27617523e-01 -4.71022129e-01 1.33200264e+00 5.59706628e-01 1.18793440e+00 -5.43666303e-01 -1.28116906e-01 8.71345341e-01 1.18049598e+00 -3.14023614e-01 3.99550438e-01 9.35545415e-02 1.26124108e+00 3.61472249e-01 3.21939141e-01 3.96280169e-01 7.54668832e-01 8.00773382e-01 1.62623078e-01 -2.00843751e-01 -9.61618870e-02 -2.19664693e-01 4.98220354e-01 1.57414424e+00 -7.65283167e-01 6.52945563e-02 -9.72341120e-01 6.46360338e-01 -2.20047808e+00 -1.17718768e+00 -4.54396665e-01 1.66833735e+00 4.20342833e-01 -7.37000853e-02 2.58435309e-01 1.93789467e-01 1.83232561e-01 5.80967307e-01 -3.46925616e-01 -2.53713042e-01 -9.51636303e-03 -1.56016991e-01 5.02853155e-01 3.04632634e-01 -1.08613670e+00 1.05225706e+00 6.44403458e+00 5.87708712e-01 -1.54579711e+00 3.98533821e-01 5.37621796e-01 -4.32537377e-01 -2.86632597e-01 3.30236256e-02 -4.61725771e-01 7.95453712e-02 1.09790409e+00 2.48402238e-01 4.47152704e-01 5.23814023e-01 3.67848307e-01 1.89499483e-01 -9.76734579e-01 1.06099796e+00 -2.17715763e-02 -1.60992992e+00 3.45889330e-01 2.25313574e-01 8.18278730e-01 6.22011840e-01 3.23079854e-01 2.16347829e-01 4.00712669e-01 -8.93316388e-01 5.61751246e-01 6.53956652e-01 5.04393995e-01 -5.42263925e-01 5.68466127e-01 2.28628382e-01 -1.54463482e+00 -1.68711826e-01 -6.35492742e-01 -2.92312652e-01 2.14391813e-01 6.51231289e-01 -4.58670229e-01 3.61451656e-01 9.64803755e-01 1.40847504e+00 -4.03369606e-01 7.89285898e-01 1.57199770e-01 6.83423042e-01 -2.72928923e-01 1.11162309e-02 8.73277366e-01 -4.64970767e-02 4.97984976e-01 1.17346823e+00 5.17314255e-01 1.21082939e-01 3.16389233e-01 4.73558992e-01 3.60925794e-02 -3.04919034e-01 -5.30775368e-01 -4.41370368e-01 -1.07651301e-01 9.53755915e-01 -4.95502830e-01 -4.05803859e-01 -5.00474095e-01 1.08874190e+00 6.03498757e-01 7.11682916e-01 -7.75469482e-01 2.29665846e-01 7.56119728e-01 -1.93781424e-02 5.25311410e-01 -8.10586214e-01 -8.02638009e-02 -1.67794085e+00 2.52452582e-01 -5.37861049e-01 5.68681419e-01 -6.29611194e-01 -1.17215657e+00 9.81313109e-01 -1.55403167e-01 -1.53961062e+00 -4.27058458e-01 -4.92542714e-01 -2.20485091e-01 3.94646227e-01 -1.70612013e+00 -1.64743698e+00 -1.17482029e-01 8.72846186e-01 7.16008663e-01 -2.72940770e-02 7.97375083e-01 6.24539077e-01 -6.47925198e-01 4.77856219e-01 1.93014368e-01 2.06625864e-01 2.03022957e-01 -9.58785594e-01 3.60426933e-01 7.72205830e-01 3.40859234e-01 7.55441666e-01 5.38843811e-01 -4.11752075e-01 -1.66982949e+00 -1.39601946e+00 1.13300836e+00 -2.76473194e-01 1.21137905e+00 -2.59037793e-01 -8.06113958e-01 1.05796206e+00 -3.69638726e-02 5.27119637e-01 6.20760322e-01 4.95819092e-01 -6.19639575e-01 -1.60746232e-01 -2.26585537e-01 6.95677519e-01 1.08149672e+00 -1.01097786e+00 -2.36077875e-01 5.18534005e-01 7.62535274e-01 -1.56408906e-01 -1.10377133e+00 4.14769083e-01 6.74418151e-01 -1.02060020e+00 9.86459553e-01 -5.90941310e-01 6.77048802e-01 -1.70566425e-01 -2.29777947e-01 -1.14388931e+00 -5.83626270e-01 -7.80805647e-01 -4.18604225e-01 6.30121827e-01 4.05042589e-01 -2.88935393e-01 7.95043647e-01 3.27794492e-01 -1.59894511e-01 -1.01388526e+00 -9.27974224e-01 -8.79700303e-01 -8.98260623e-02 -8.55219185e-01 2.88655698e-01 1.21525884e+00 -1.48718210e-03 3.34112972e-01 -1.02584350e+00 2.25768611e-01 5.50284505e-01 2.76698112e-01 6.13729656e-01 -7.47690499e-01 -3.07535559e-01 -2.66125321e-01 -5.31926036e-01 -1.50134528e+00 2.89314650e-02 -7.09175646e-01 -2.69987732e-01 -1.36988235e+00 3.01710010e-01 -3.19021463e-01 -5.62612951e-01 5.60600281e-01 2.28217125e-01 5.78478634e-01 1.63994461e-01 6.86220586e-01 -8.71270597e-01 8.09808969e-01 1.33862424e+00 -1.58005849e-01 2.05940157e-02 -2.21174106e-01 1.33542389e-01 7.57624865e-01 4.26837027e-01 -3.89287323e-01 -4.12972420e-01 -1.09974682e+00 3.55894417e-01 4.69825476e-01 6.66019857e-01 -9.75282073e-01 4.35580909e-01 -1.11672677e-01 2.85878003e-01 -9.13553059e-01 7.48611569e-01 -5.13994336e-01 2.42199361e-01 3.62909198e-01 -5.31666577e-01 5.11246622e-01 3.15237530e-02 5.90219736e-01 -4.97120529e-01 4.32290971e-01 3.23987097e-01 -2.44052082e-01 -4.73313242e-01 7.32061923e-01 -5.61904132e-01 -4.50137198e-01 6.28265440e-01 -4.16504219e-02 -2.03694910e-01 -3.71005952e-01 -7.90135562e-01 3.19660269e-02 2.56121755e-01 4.63104576e-01 8.10307920e-01 -1.62811422e+00 -5.82175195e-01 -6.47918880e-02 -2.25871354e-02 -2.40619689e-01 5.66785872e-01 1.24797440e+00 -4.90365714e-01 7.50097632e-01 4.57401909e-02 -9.53658998e-01 -1.21628344e+00 4.50492322e-01 2.20570758e-01 -6.72480583e-01 -9.26537037e-01 8.85134876e-01 8.84886235e-02 -2.57188946e-01 1.04639612e-01 -7.04421222e-01 -2.76877105e-01 -1.71732940e-02 5.44273973e-01 2.75978982e-01 -2.60564629e-02 -9.32481885e-01 -2.19303146e-01 7.72747338e-01 -2.65879363e-01 -4.36538868e-02 1.62548542e+00 -1.30703794e-02 -3.33105505e-01 8.45647395e-01 1.60236037e+00 -6.14263952e-01 -1.19225872e+00 -6.35353208e-01 2.44239405e-01 -5.78930199e-01 5.61260760e-01 -6.79554269e-02 -1.37522662e+00 1.26569128e+00 2.69242644e-01 2.90252678e-02 9.65745687e-01 -2.42414594e-01 1.29394770e+00 7.56878555e-01 3.42238098e-01 -8.17423105e-01 7.42591739e-01 9.58283663e-01 8.06847811e-01 -1.00381422e+00 4.36689332e-02 -6.73220605e-02 -6.40243113e-01 1.36059487e+00 2.78797984e-01 -4.63772684e-01 8.08757722e-01 -8.49580243e-02 -1.32018492e-01 -4.61239755e-01 -1.45564306e+00 -1.29768103e-01 6.44444168e-01 3.32230963e-02 6.92655802e-01 -1.49273667e-02 1.06355987e-01 -2.16410801e-01 -5.14541976e-02 1.41948238e-01 3.19249123e-01 7.58349061e-01 1.06818371e-01 -9.67055857e-01 -3.21687907e-02 5.46501517e-01 -6.89602494e-01 -2.31813386e-01 6.17225729e-02 1.51139721e-01 -1.91526204e-01 7.51048386e-01 2.28369638e-01 -7.90710747e-01 -8.84065181e-02 -4.04604226e-01 3.05198222e-01 -2.76570529e-01 -3.54700476e-01 4.65767711e-01 1.24596663e-01 -1.02478588e+00 -8.13117743e-01 -7.70390987e-01 -8.00776422e-01 -5.39319515e-01 -1.88587040e-01 1.07238358e-02 4.44472015e-01 1.07219803e+00 5.10782838e-01 6.72070026e-01 5.03945589e-01 -1.08772755e+00 -2.49786124e-01 -1.00305057e+00 -3.23917955e-01 3.29485983e-01 7.16121554e-01 -5.85844517e-01 -2.13860683e-02 7.90909827e-02]
[8.842601776123047, 0.4941946864128113]
bb2fc990-de09-4d08-8876-d45998698769
fourier-transformer-fast-long-range-modeling
2305.15099
null
https://arxiv.org/abs/2305.15099v1
https://arxiv.org/pdf/2305.15099v1.pdf
Fourier Transformer: Fast Long Range Modeling by Removing Sequence Redundancy with FFT Operator
The transformer model is known to be computationally demanding, and prohibitively costly for long sequences, as the self-attention module uses a quadratic time and space complexity with respect to sequence length. Many researchers have focused on designing new forms of self-attention or introducing new parameters to overcome this limitation, however a large portion of them prohibits the model to inherit weights from large pretrained models. In this work, the transformer's inefficiency has been taken care of from another perspective. We propose Fourier Transformer, a simple yet effective approach by progressively removing redundancies in hidden sequence using the ready-made Fast Fourier Transform (FFT) operator to perform Discrete Cosine Transformation (DCT). Fourier Transformer is able to significantly reduce computational costs while retain the ability to inherit from various large pretrained models. Experiments show that our model achieves state-of-the-art performances among all transformer-based models on the long-range modeling benchmark LRA with significant improvement in both speed and space. For generative seq-to-seq tasks including CNN/DailyMail and ELI5, by inheriting the BART weights our model outperforms the standard BART and other efficient models. \footnote{Our code is publicly available at \url{https://github.com/LUMIA-Group/FourierTransformer}}
['Zhouhan Lin', 'Jingwen Leng', 'Xinbing Wang', 'Jingcheng Yin', 'Minwei Feng', 'Meng Yang', 'Ziwei He']
2023-05-24
null
null
null
null
['abstractive-text-summarization', 'long-range-modeling', 'open-domain-question-answering', 'document-summarization']
['natural-language-processing', 'natural-language-processing', 'natural-language-processing', 'natural-language-processing']
[ 2.97485501e-01 -1.01497367e-01 2.24270746e-01 -3.83987904e-01 -1.00099266e+00 -5.50934315e-01 5.00044405e-01 -2.74853706e-01 -5.81170619e-01 7.02982545e-01 2.74333149e-01 -4.40607101e-01 6.22423925e-02 -6.85783863e-01 -9.45894003e-01 -8.76611650e-01 7.40862042e-02 4.42164361e-01 1.38454810e-01 -2.60044128e-01 1.07752033e-01 2.73165405e-01 -1.41157436e+00 2.85829455e-01 7.63107836e-01 8.58877182e-01 6.25678241e-01 8.01579356e-01 -1.37581900e-01 8.11130106e-01 -3.98801684e-01 -6.10068798e-01 2.35679880e-01 -6.79301798e-01 -8.45318735e-01 -4.08560634e-01 3.87187272e-01 -2.56743997e-01 -3.44152570e-01 9.09051001e-01 1.01948977e+00 2.16139317e-01 2.79295057e-01 -1.11654580e+00 -7.98541903e-01 9.19071198e-01 -3.42508465e-01 4.94006038e-01 2.36511193e-02 1.86625734e-01 1.22876096e+00 -1.03056526e+00 3.99004817e-01 1.01488304e+00 1.00820327e+00 4.89971906e-01 -1.04491675e+00 -7.85785258e-01 8.63031447e-02 4.54508543e-01 -1.53822339e+00 -5.02917290e-01 5.94079316e-01 -1.83579996e-01 1.60211718e+00 5.06728232e-01 4.77242082e-01 1.24301755e+00 1.99978665e-01 8.27759624e-01 9.22191679e-01 -3.23355705e-01 -8.63490030e-02 -1.01582825e-01 1.00085497e-01 7.56768763e-01 -2.45898694e-01 3.75816273e-03 -4.58706647e-01 -4.07123230e-02 6.73698187e-01 2.57413294e-02 -3.85394931e-01 -3.96557003e-02 -1.14357376e+00 7.73340225e-01 3.85990679e-01 4.73204553e-01 -2.83481985e-01 3.48084778e-01 4.69066054e-01 3.83630425e-01 6.43748820e-01 2.59276092e-01 -7.25807428e-01 -5.37361622e-01 -9.02500987e-01 1.22521006e-01 5.67640066e-01 1.14331889e+00 6.63134456e-01 2.50108808e-01 -1.78604200e-01 1.00468946e+00 7.47018785e-04 3.11514378e-01 9.25472081e-01 -6.12298906e-01 3.34846258e-01 2.34827027e-01 -4.17371124e-01 -6.79476142e-01 -2.47882158e-01 -8.55267644e-01 -1.18618560e+00 -3.07407945e-01 1.10224076e-01 7.14584067e-02 -9.19755518e-01 1.71765411e+00 3.45998406e-02 4.43370432e-01 -1.24120213e-01 6.79752052e-01 6.67129278e-01 9.00637031e-01 -1.09944664e-01 -1.26387671e-01 1.27789342e+00 -1.23041403e+00 -5.16959548e-01 -2.28913888e-01 7.52502561e-01 -8.38346243e-01 1.28342164e+00 2.29443833e-01 -1.18015492e+00 -7.73520052e-01 -9.17490959e-01 -4.79667217e-01 -3.13410521e-01 9.98022705e-02 4.37658191e-01 6.33890867e-01 -1.17955399e+00 7.91099787e-01 -8.88391316e-01 -2.87680447e-01 3.83147001e-01 3.45779568e-01 -3.54559660e-01 1.25724664e-02 -1.20965433e+00 8.30589294e-01 3.53747219e-01 3.00041232e-02 -7.23495603e-01 -1.04201400e+00 -8.17617893e-01 4.23487693e-01 2.33472317e-01 -7.50995278e-01 1.22200608e+00 -8.76941919e-01 -1.64443076e+00 5.50361276e-01 -3.04948658e-01 -8.59975219e-01 4.59666252e-01 -5.95519543e-01 -1.27201453e-01 -2.58590192e-01 -1.55111939e-01 6.34117126e-01 6.14040017e-01 -6.70323491e-01 -3.61115307e-01 -2.03000873e-01 -1.29403740e-01 3.15964110e-02 -5.05240500e-01 6.71233162e-02 -5.18001854e-01 -9.07002151e-01 -4.08521265e-01 -1.17623734e+00 -1.27980426e-01 -4.69795972e-01 -3.49952459e-01 -2.38265425e-01 9.46507335e-01 -7.80164063e-01 1.40774632e+00 -2.03430820e+00 1.07759871e-01 -1.65217698e-01 -6.84441701e-02 5.55486083e-01 -3.62160265e-01 6.27401531e-01 -3.21111590e-01 1.88176632e-01 -3.56591344e-01 -4.37698036e-01 -4.07902636e-02 3.31133664e-01 -3.89352769e-01 3.33002687e-01 1.36084527e-01 1.10127389e+00 -7.55539834e-01 -2.43326411e-01 1.21344380e-01 7.82027960e-01 -7.87478507e-01 2.21739545e-01 -1.69778168e-01 2.74940997e-01 -5.06845117e-02 1.77740023e-01 5.86321771e-01 -5.71688175e-01 4.78039011e-02 -9.54544470e-02 -8.16485956e-02 6.82060122e-01 -7.44904459e-01 1.79559326e+00 -5.58246136e-01 6.10809147e-01 -3.22823495e-01 -1.15428698e+00 7.76380599e-01 3.49466473e-01 3.33741397e-01 -6.54521883e-01 2.49814168e-01 2.06691295e-01 3.84875685e-02 -3.11333179e-01 3.40157747e-01 -6.66261166e-02 1.80468380e-01 3.90351087e-01 2.35477358e-01 2.24226430e-01 1.94855049e-01 1.51760727e-01 1.08073783e+00 3.51925671e-01 3.24676007e-01 -3.54543924e-01 5.21725655e-01 -3.40078890e-01 4.43860263e-01 6.07316434e-01 3.11205000e-01 6.82549477e-01 2.23442718e-01 -5.84448159e-01 -1.30148721e+00 -7.15152919e-01 2.36186996e-01 1.36621726e+00 -3.95403892e-01 -5.67946792e-01 -7.88104355e-01 -6.25433087e-01 -2.47830629e-01 8.35712850e-01 -5.60940444e-01 -1.69606701e-01 -9.64257061e-01 -7.39744186e-01 7.27932751e-01 7.76963055e-01 5.41942060e-01 -1.03015566e+00 -6.25736833e-01 2.39993334e-01 -5.04591525e-01 -1.00967586e+00 -8.95228624e-01 3.60019088e-01 -7.97928095e-01 -5.28220534e-01 -9.82887745e-01 -5.85152626e-01 3.48533303e-01 2.18031093e-01 1.06091714e+00 1.17279381e-01 -3.64842653e-01 -1.00884266e-01 -4.79138732e-01 -4.36354637e-01 -2.62937278e-01 6.52375221e-01 -2.78952599e-01 -1.31144688e-01 2.56796330e-01 -8.39283407e-01 -5.28998196e-01 3.04298550e-01 -8.41814399e-01 2.21511856e-01 6.88039482e-01 1.04040110e+00 5.23765624e-01 -9.10782292e-02 5.12107790e-01 -8.57696295e-01 4.80386376e-01 -3.54941398e-01 -5.06054878e-01 3.78344655e-02 -5.02932310e-01 2.40040734e-01 8.39644015e-01 -3.28042358e-01 -8.82661223e-01 -4.49575447e-02 -6.91572249e-01 -7.34157085e-01 1.51513264e-01 4.75463361e-01 -4.22745943e-02 1.56509057e-01 4.12831068e-01 7.27332652e-01 2.83987988e-02 -7.58355618e-01 1.76472142e-01 4.86340523e-01 2.92458504e-01 -2.61593848e-01 6.43183112e-01 2.02881798e-01 -2.81578124e-01 -8.11928511e-01 -9.50661957e-01 -3.93599927e-01 -5.58212936e-01 1.24556087e-01 7.25518107e-01 -8.83810341e-01 -6.65359259e-01 4.64624763e-01 -1.16525245e+00 -4.93859470e-01 -1.98338255e-01 4.10786629e-01 -6.14827335e-01 4.19137359e-01 -7.66783595e-01 -5.70062757e-01 -7.20984459e-01 -1.10727239e+00 1.00317621e+00 -1.47337809e-01 -1.74150258e-01 -8.38692188e-01 6.01109304e-02 3.29403251e-01 7.92393982e-01 -1.60871342e-01 9.58155751e-01 -7.13459969e-01 -6.56746805e-01 -4.04026918e-02 -1.11069605e-01 6.07172608e-01 -1.51960552e-02 -2.83309191e-01 -1.08877838e+00 -4.67939585e-01 3.93352732e-02 -2.00352818e-01 1.04808998e+00 2.13915497e-01 1.45573449e+00 -4.26269054e-01 -1.89157560e-01 8.86257470e-01 1.34396517e+00 1.87587872e-01 8.02302122e-01 1.36048526e-01 7.34087646e-01 1.25690952e-01 2.85214245e-01 4.80593085e-01 3.14322948e-01 6.72603548e-01 1.89889103e-01 -1.01957865e-01 -2.32294008e-01 -2.76098669e-01 4.78133291e-01 1.24959028e+00 -3.43884706e-01 -5.43436110e-01 -8.11532140e-01 5.25691450e-01 -1.83179986e+00 -1.27346480e+00 7.56761953e-02 2.12852383e+00 9.64823365e-01 1.43395094e-02 2.91767959e-02 9.38381329e-02 3.62091094e-01 2.12840363e-01 -3.33139837e-01 -3.94211233e-01 -1.08720578e-01 4.10559177e-01 6.15056753e-01 5.00196874e-01 -9.55486000e-01 9.94080067e-01 5.93931103e+00 1.19412935e+00 -1.20970786e+00 3.40050131e-01 5.60782671e-01 -1.89957380e-01 -1.94697648e-01 -2.03474954e-01 -9.42520797e-01 5.21070421e-01 1.36761725e+00 -1.98043197e-01 4.56428826e-01 8.56895745e-01 1.77497957e-02 3.63151789e-01 -1.08586419e+00 7.84032524e-01 1.63825631e-01 -1.38327825e+00 1.03727832e-01 9.58703011e-02 3.87926877e-01 3.83603096e-01 7.90954456e-02 5.17879009e-01 2.67994463e-01 -1.10860693e+00 6.89769745e-01 3.49999011e-01 7.22695649e-01 -7.83477306e-01 8.69763911e-01 4.59467411e-01 -1.23287487e+00 2.39743479e-02 -4.37287837e-01 -8.04002136e-02 2.53750771e-01 6.03688717e-01 -8.96141052e-01 5.72641730e-01 8.46044421e-01 6.20121300e-01 -3.86626750e-01 7.35848486e-01 2.21401185e-01 7.92339087e-01 -3.54938537e-01 -5.38917296e-02 4.37798351e-01 -9.30674523e-02 4.31985885e-01 1.59817696e+00 7.47404099e-01 -5.91362491e-02 -9.29353014e-02 6.20003402e-01 -3.93756181e-02 1.02210484e-01 -4.86721218e-01 -1.09441712e-01 2.01503322e-01 9.77050543e-01 -4.85375017e-01 -4.74509358e-01 -4.51895803e-01 1.19766808e+00 3.87005806e-01 2.18582883e-01 -1.25632060e+00 -4.23018128e-01 5.91434419e-01 2.50213802e-01 8.01498771e-01 -3.02804559e-01 -3.16656418e-02 -1.08034861e+00 1.09905235e-01 -1.02531528e+00 3.68116021e-01 -7.57519364e-01 -1.08636165e+00 9.93104637e-01 -1.15151502e-01 -1.05971313e+00 -2.27372512e-01 -3.71231824e-01 -3.23557615e-01 9.80431736e-01 -1.57438612e+00 -1.25559723e+00 -2.31450096e-01 5.89631855e-01 9.24202204e-01 -5.95354103e-02 9.13974822e-01 7.22540438e-01 -4.98381317e-01 8.72250021e-01 1.45840704e-01 1.24124840e-01 5.66641629e-01 -1.10536671e+00 8.73790145e-01 6.44523025e-01 2.83599168e-01 7.51399279e-01 5.78254819e-01 -4.27556902e-01 -1.46971929e+00 -1.11512280e+00 1.13737142e+00 -2.48858437e-01 5.89700282e-01 -5.77789247e-01 -1.09650433e+00 1.00595868e+00 5.80488622e-01 2.79059205e-02 8.47114801e-01 -6.03781119e-02 -4.73072886e-01 -8.17053169e-02 -5.69281697e-01 4.88470674e-01 1.37472796e+00 -4.22425032e-01 -3.61074358e-01 2.78799146e-01 8.90230417e-01 -3.14166248e-01 -7.80802250e-01 3.91971856e-01 5.99474311e-01 -1.14092934e+00 1.00743425e+00 -3.22727591e-01 3.66520733e-01 -1.18319429e-01 -4.18127328e-02 -1.33650029e+00 -6.24326289e-01 -8.49907577e-01 2.88028332e-05 1.25381351e+00 4.73701537e-01 -8.69066060e-01 6.75154090e-01 4.92979363e-02 -4.01204735e-01 -9.36883032e-01 -9.41527665e-01 -1.13231277e+00 7.05863163e-02 -4.48158890e-01 6.46917522e-01 9.90503788e-01 -2.45384559e-01 6.33307815e-01 -7.52335131e-01 -6.95696771e-02 2.92932779e-01 6.81725517e-02 7.83277512e-01 -8.29595029e-01 -5.95126390e-01 -4.22476321e-01 -1.33284897e-01 -1.45452332e+00 -3.33704688e-02 -1.07939529e+00 -1.97122376e-02 -1.29233015e+00 1.94018289e-01 -2.73018956e-01 -3.68288070e-01 6.42496169e-01 -1.77340180e-01 3.41177136e-01 3.48213583e-01 7.49161001e-03 -2.77435750e-01 7.35822022e-01 1.09516573e+00 2.52142940e-02 1.29984051e-01 -3.36991437e-02 -5.31537473e-01 5.57321012e-01 8.92550826e-01 -5.72494507e-01 -5.26230335e-01 -7.05213547e-01 1.68556184e-01 -1.08626649e-01 2.89087623e-01 -9.84981239e-01 1.91888481e-01 1.87769875e-01 1.65509045e-01 -6.75880432e-01 4.95434046e-01 -7.14948535e-01 4.05040443e-01 5.32674491e-01 -3.78201216e-01 5.38099468e-01 3.48439455e-01 3.82089466e-01 -1.87944219e-01 -1.43829331e-01 8.06163192e-01 -2.94498384e-01 -4.32611674e-01 3.27216238e-01 -3.50042015e-01 -3.23499106e-02 7.83628047e-01 -7.58662447e-02 -1.47434324e-01 -3.50795060e-01 -5.13820589e-01 -7.43311569e-02 2.40086317e-01 5.81791461e-01 4.49057668e-01 -1.10782981e+00 -9.04564798e-01 3.63804877e-01 -2.34736830e-01 -2.02335954e-01 5.66064537e-01 1.05790985e+00 -4.91578281e-01 7.08530724e-01 -1.09121501e-01 -5.20374477e-01 -1.57264078e+00 7.65881658e-01 1.37518927e-01 -5.97873211e-01 -7.49394119e-01 1.21365249e+00 3.39725137e-01 -3.56880367e-01 1.49990663e-01 -3.92669499e-01 2.14248538e-01 -1.01136498e-01 4.90335882e-01 3.37886423e-01 2.02536300e-01 -5.71139276e-01 -3.54435980e-01 7.05352128e-01 -2.73350894e-01 2.27452442e-01 1.52375627e+00 -3.23052406e-02 1.91676058e-02 2.61883110e-01 1.57397258e+00 -6.08676225e-02 -1.08935893e+00 -3.00367683e-01 -1.78593472e-01 -4.23353851e-01 -5.87407313e-02 -6.86071396e-01 -1.13963032e+00 1.13958871e+00 3.24658096e-01 1.04615450e-01 1.23075473e+00 -1.01938218e-01 1.21351194e+00 2.69665897e-01 2.03056380e-01 -7.04839110e-01 -5.12469672e-02 8.06207776e-01 1.03143811e+00 -9.12899435e-01 -1.66955546e-01 -2.88858384e-01 -6.36816204e-01 8.25000465e-01 4.18900937e-01 -6.87159896e-02 5.32320619e-01 6.66423321e-01 -8.43344554e-02 1.09298192e-01 -1.04914761e+00 -5.43286279e-02 1.27450407e-01 2.34922647e-01 7.18814313e-01 -1.50821269e-01 -2.04881415e-01 4.57059801e-01 -5.07985771e-01 2.86690388e-02 1.80935517e-01 8.20564330e-01 -3.53123575e-01 -1.19749546e+00 5.20989932e-02 2.79211402e-01 -7.13318646e-01 -6.05351567e-01 -5.86331710e-02 6.57831669e-01 8.17272887e-02 4.44593638e-01 1.27765954e-01 -3.35711718e-01 1.56986982e-01 3.41315478e-01 3.96388859e-01 -3.20447594e-01 -8.42309952e-01 3.45481485e-01 6.40250966e-02 -6.14839554e-01 -2.43565068e-01 -4.97681111e-01 -1.03527653e+00 -3.66777569e-01 -2.99530834e-01 3.16863656e-02 4.25136507e-01 7.59697676e-01 7.32532620e-01 7.49567389e-01 2.93164998e-01 -7.75304794e-01 -6.22107804e-01 -1.22390974e+00 -1.17572576e-01 2.47507319e-01 3.35364640e-01 -3.65811437e-01 -1.42275795e-01 2.58556426e-01]
[10.889935493469238, 6.582707405090332]
48dcb424-d409-44eb-a55e-ab59ce665d97
exploiting-social-information-in-grounded
null
null
https://aclanthology.org/P12-1093
https://aclanthology.org/P12-1093.pdf
Exploiting Social Information in Grounded Language Learning via Grammatical Reduction
null
['Mark Johnson', 'Michael Frank', 'Katherine Demuth']
2012-07-01
null
null
null
acl-2012-7
['grounded-language-learning']
['natural-language-processing']
[-8.63703638e-02 1.71006292e-01 -6.22772932e-01 -4.08054382e-01 -8.41685571e-03 -9.08429027e-01 6.55310392e-01 -6.53472245e-01 -2.85945535e-01 1.06888819e+00 -4.63127941e-02 -1.01159286e+00 -3.91567826e-01 -9.63214397e-01 -4.95059669e-01 -6.31337762e-01 -9.79754329e-01 7.25764990e-01 3.30370307e-01 -6.93831444e-01 7.03166842e-01 7.88774848e-01 -1.68942046e+00 7.18545914e-01 7.04417467e-01 8.52217197e-01 2.49141872e-01 1.14950800e+00 -1.95044339e-01 1.55633950e+00 -7.48382092e-01 -5.46825826e-01 3.13719302e-01 -1.23176083e-01 -7.22945035e-01 -1.01074085e-01 9.28529128e-02 -8.59008506e-02 -2.09758401e-01 9.22211111e-01 5.37373662e-01 4.49454933e-02 1.08379531e+00 -1.42548037e+00 -5.91619551e-01 6.10313773e-01 -4.01565880e-02 1.21627934e-01 1.03678203e+00 -5.39447069e-01 1.19919395e+00 -1.13026452e+00 7.20913768e-01 1.26888943e+00 8.66221786e-01 5.44149756e-01 -1.22286928e+00 -1.94712028e-01 -3.26822817e-01 -9.51717794e-02 -1.46558487e+00 -3.25250506e-01 4.25783843e-02 -2.08119690e-01 1.66093647e+00 1.26596653e+00 1.20609856e+00 1.01401424e+00 1.26658809e+00 8.34431887e-01 1.04267764e+00 -5.13792276e-01 3.35295945e-01 3.66983831e-01 1.54683650e-01 6.33519173e-01 8.40953708e-01 5.26628852e-01 -7.06372619e-01 -9.13127720e-01 9.33553874e-01 -2.94925272e-01 1.71355158e-01 -5.05680561e-01 -9.05919552e-01 6.91228509e-01 1.78732842e-01 3.83959889e-01 -1.39880210e-01 9.89067405e-02 1.26390755e-01 5.30987144e-01 -2.58292928e-02 6.47037446e-01 -9.11868811e-01 -1.33165747e-01 -8.71728659e-01 5.10332465e-01 1.25398111e+00 1.52653182e+00 1.24482810e-01 2.94908643e-01 -9.34252143e-02 3.17179203e-01 8.92314315e-01 1.01808000e+00 4.28362608e-01 -1.36146402e+00 -6.87414408e-02 1.72361732e-01 5.01781464e-01 -8.52631688e-01 -6.33224547e-01 -9.64177120e-03 -8.93263519e-01 4.49267089e-01 3.49161088e-01 4.57367361e-01 -8.02827001e-01 5.07305264e-01 4.33481112e-02 -2.34125629e-01 4.53833073e-01 5.55570945e-02 4.99930978e-01 3.76208365e-01 -1.34477139e-01 -5.73289394e-01 1.06082785e+00 -1.36716676e+00 -1.35299087e+00 2.33215362e-01 9.05734658e-01 -1.07320261e+00 4.35900748e-01 5.33875942e-01 -1.55548143e+00 -1.37560293e-01 -1.08699942e+00 1.78573877e-01 -7.27255583e-01 -3.14239264e-01 8.57801437e-01 1.43120694e+00 -1.60129595e+00 9.73287821e-01 -4.91727620e-01 6.59165755e-02 1.20568443e-02 8.24621081e-01 -2.64718989e-03 4.62812334e-01 -1.33193445e+00 1.08501506e+00 2.22979754e-01 -1.21242590e-01 -1.65216476e-01 -2.13068098e-01 -8.23704481e-01 -5.63443303e-01 -4.78693932e-01 -5.29636025e-01 1.44139910e+00 -2.59346128e-01 -1.65295815e+00 9.71794367e-01 -1.42069459e-01 -1.97814897e-01 6.14786744e-01 -1.28011424e-02 -8.31891418e-01 2.42498964e-01 -1.89849049e-01 5.76383233e-01 9.28263724e-01 -1.35132408e+00 -7.59897232e-01 -1.67359829e-01 -1.23336017e-01 2.66287565e-01 -1.25510961e-01 1.89734384e-01 2.11616129e-01 -1.12999000e-01 3.27147305e-01 -7.20919967e-01 -2.53068686e-01 -5.32041907e-01 -1.46512717e-01 -7.10518599e-01 7.70373225e-01 -4.51523662e-01 1.83705616e+00 -1.67618537e+00 -2.06720144e-01 3.98590982e-01 3.57815564e-01 -1.24705513e-03 2.40583986e-01 1.08380008e+00 -2.76906848e-01 7.32199550e-01 4.11965609e-01 -1.10722095e-01 2.24991128e-01 4.85861301e-01 -4.16602850e-01 3.05609167e-01 -1.29282743e-01 1.15307164e+00 -1.16605783e+00 -5.23096442e-01 4.11106765e-01 9.42391157e-02 -4.58732933e-01 4.79237735e-01 2.99364805e-01 1.47170946e-01 -3.56553018e-01 1.39399457e+00 1.15709066e+00 -1.31984919e-01 1.45911396e-01 5.30878425e-01 -3.79135728e-01 3.55090618e-01 -6.96863770e-01 1.05554795e+00 7.08333924e-02 5.00173986e-01 1.02364108e-01 -7.94621468e-01 3.33247900e-01 8.61540735e-01 4.37155962e-01 -9.67555881e-01 -2.26398129e-02 6.92409754e-01 1.12803578e-01 -6.07703328e-01 7.58228302e-01 3.94563079e-02 -3.64872098e-01 6.40070081e-01 -2.37588286e-01 -6.59476995e-01 9.64643434e-02 3.08100313e-01 6.22585893e-01 -5.18246442e-02 5.62923312e-01 -1.05613089e+00 7.86340594e-01 -1.82965681e-01 -1.81299388e-01 1.03415680e+00 -3.09923887e-01 3.19085121e-01 2.99841821e-01 -6.65102363e-01 -6.45341039e-01 -1.12307119e+00 -4.89381433e-01 1.30636716e+00 3.24267983e-01 -4.39044595e-01 -9.54439282e-01 -2.49762803e-01 1.77620783e-01 6.89606130e-01 -5.90509653e-01 3.84124845e-01 -5.03739953e-01 -8.32535863e-01 7.39044368e-01 3.45434904e-01 -5.07752821e-02 -1.33414865e+00 -6.58416986e-01 1.25490099e-01 -2.20292807e-01 -6.63697243e-01 -6.23428151e-02 4.48765576e-01 -1.35989368e+00 -5.18594682e-01 -6.66252747e-02 -8.20914626e-01 5.87345481e-01 2.46782884e-01 1.27047324e+00 5.39230824e-01 -2.31483161e-01 4.26904231e-01 -1.21292919e-01 -4.95818377e-01 -4.59671497e-01 -8.00336525e-02 5.28869390e-01 -5.87835789e-01 5.19427478e-01 -2.50617653e-01 -7.29350567e-01 5.37953973e-01 -6.88540697e-01 1.62748516e-01 1.79803044e-01 1.04410267e+00 1.35816500e-01 -9.34035778e-02 1.22507080e-01 -6.38007045e-01 8.72274399e-01 -1.69219792e-01 -3.78732830e-01 5.77745810e-02 -6.77108407e-01 -3.74140263e-01 3.21430594e-01 -3.25342178e-01 -1.01981449e+00 -4.87835288e-01 -9.82677937e-02 2.45538145e-01 1.11353043e-02 -1.46784872e-01 6.47139177e-02 -5.24923325e-01 8.02199244e-01 9.25758183e-02 1.99174434e-02 -6.80815242e-03 3.01039815e-01 7.09525108e-01 -6.82967342e-03 -6.68678164e-01 8.44880998e-01 4.91470337e-01 7.98524171e-02 -9.57177758e-01 -1.52186140e-01 -2.60129690e-01 -9.51962709e-01 -6.54426932e-01 6.56643391e-01 -6.78531289e-01 -9.10833478e-01 3.91110867e-01 -9.38691139e-01 -3.38627815e-01 -3.91645581e-01 4.25431967e-01 -1.01278400e+00 2.75717527e-02 -3.90154392e-01 -1.27895141e+00 -5.10977268e-01 -1.02017939e+00 9.43384409e-01 5.30070923e-02 -5.10597289e-01 -1.26927447e+00 5.87685481e-02 2.71537274e-01 1.81734428e-01 -1.73075795e-01 6.90226793e-01 -2.38256708e-01 -4.24233019e-01 -1.53791070e-01 2.34436691e-02 -1.39755070e-01 1.70832314e-02 4.95917559e-01 -9.81751978e-01 -5.31145096e-01 6.65065646e-02 -1.92070693e-01 -1.08835101e-01 6.52520418e-01 5.91872573e-01 -2.29931593e-01 -8.56000841e-01 5.40386558e-01 1.38545322e+00 3.85070026e-01 5.32770038e-01 7.28214979e-01 1.41836226e-01 5.53460240e-01 9.17806149e-01 4.63203549e-01 1.30579369e-02 3.28798652e-01 2.40537539e-01 1.49327129e-01 1.11720070e-01 -1.54819340e-01 3.77893507e-01 1.16112018e+00 -8.18235934e-01 -2.69281328e-01 -5.07867396e-01 4.42987174e-01 -1.72482407e+00 -1.40330648e+00 -4.32368398e-01 6.90478683e-01 6.25676990e-01 1.56016424e-01 -1.48347050e-01 3.35214496e-01 4.99015123e-01 -2.03574806e-01 -1.19133167e-01 -1.06291151e+00 -1.43546045e-01 3.15233678e-01 7.37729073e-01 1.00061214e+00 -7.20721722e-01 1.03317809e+00 1.29781246e+01 1.02230716e+00 2.21112028e-01 1.03134915e-01 5.16071796e-01 3.48020852e-01 -4.36954498e-01 -4.56139445e-02 -1.04416132e+00 2.72933897e-02 1.38140702e+00 -4.30666685e-01 6.85999811e-01 5.44219851e-01 3.44648361e-01 -4.23268199e-01 -1.26188684e+00 5.26221812e-01 9.73738134e-02 -1.40886843e+00 -2.83300440e-04 6.85225725e-01 7.73699820e-01 -5.08050561e-01 6.22419357e-01 3.24184299e-01 6.09259963e-01 -1.14389277e+00 8.60300779e-01 2.53660440e-01 1.03040910e+00 -6.05088234e-01 5.67372203e-01 1.68872893e-01 -1.14389896e+00 -2.20873043e-01 -8.77727985e-01 -1.00755692e+00 3.93533185e-02 -1.81779593e-01 -4.29956943e-01 3.48861217e-01 9.58353162e-01 2.99398601e-01 -3.93658698e-01 9.95779395e-01 -4.78694476e-02 1.04875881e-02 -2.71853864e-01 -4.48467314e-01 4.83122796e-01 -3.54241252e-01 4.66730654e-01 1.00164843e+00 2.48499006e-01 3.51035744e-01 -9.84472036e-02 4.01770771e-01 5.45058846e-01 3.29446048e-02 -1.19659424e+00 -1.78908288e-01 2.83276141e-01 9.16795909e-01 -4.83487815e-01 -4.22520459e-01 -2.00212970e-01 8.62069130e-01 -3.55488248e-02 5.01107454e-01 -6.11489356e-01 -4.35615242e-01 9.72222984e-01 -1.27327025e-01 -1.14700586e-01 -3.48497719e-01 -6.23769283e-01 -7.30352640e-01 -5.89872956e-01 -4.54965204e-01 5.93606755e-02 -5.53365827e-01 -1.39813089e+00 5.79277515e-01 -2.27688253e-02 -1.40553558e+00 -6.99901402e-01 -1.27676582e+00 -4.76714373e-01 4.92853165e-01 -1.11898029e+00 -1.10984349e+00 2.50124663e-01 4.52870727e-01 1.64141744e-01 -5.34416080e-01 1.39563632e+00 3.57715860e-02 1.00637585e-01 9.24474537e-01 6.69434488e-01 -7.37814724e-01 5.56605101e-01 -1.27867436e+00 5.68737745e-01 -1.37897313e-01 -4.31265175e-01 9.05828118e-01 6.28349900e-01 -5.39804697e-01 -1.41196322e+00 -3.66917729e-01 1.08350635e+00 -9.83769417e-01 6.55218959e-01 -3.86345625e-01 4.23767231e-02 7.88592756e-01 7.15902448e-01 -6.18741751e-01 8.21781039e-01 -1.83753878e-01 1.80774391e-01 5.75296998e-01 -1.39248300e+00 6.12354755e-01 1.66275799e+00 -4.63594139e-01 -6.25784039e-01 7.60327101e-01 8.13696027e-01 -6.94087505e-01 -1.30082703e+00 3.34633321e-01 8.65424156e-01 -8.75409484e-01 1.61978090e+00 -1.32660246e+00 -4.43697497e-02 2.81152606e-01 -2.61993498e-01 -9.32519078e-01 -5.96193194e-01 -1.23518765e+00 -5.33532679e-01 -5.83747849e-02 5.96577883e-01 -1.13057327e+00 3.42365682e-01 8.74560475e-01 -2.82833427e-01 -6.42737269e-01 -1.06996536e+00 -1.32016802e+00 -3.35779637e-02 -1.45572275e-01 4.90409225e-01 7.63798356e-01 6.83744550e-01 1.09839931e-01 -6.36873543e-02 -1.00294888e-01 5.33176839e-01 9.55312885e-03 4.41501856e-01 -1.34294486e+00 3.86843324e-01 -5.75816095e-01 -3.07655483e-01 -9.45992947e-01 -8.85957032e-02 -8.12076271e-01 -6.53862000e-01 -1.28511906e+00 -8.32044985e-03 -1.91056758e-01 -1.11109078e-01 -1.65725678e-01 3.67937148e-01 2.13746816e-01 1.20859891e-02 1.03788137e-01 -3.71160030e-01 6.18435517e-02 1.29639816e+00 6.91750320e-05 -1.62315920e-01 4.85058486e-01 -4.81304944e-01 7.84440815e-01 8.58408585e-02 -2.99253196e-01 -6.78878546e-01 6.11881316e-02 6.69384480e-01 4.61409837e-02 3.24159935e-02 -7.42885649e-01 5.37211418e-01 -3.75702560e-01 4.78586555e-01 -1.32223868e+00 1.30741090e-01 -9.61415648e-01 6.95283338e-02 9.40189242e-01 2.74610907e-01 1.20424610e-02 8.66204947e-02 5.39500564e-02 -1.44871444e-01 -5.70943117e-01 9.21121240e-01 -4.07591403e-01 -4.92852688e-01 -5.20386267e-03 -1.03226590e+00 8.97834301e-02 9.92593169e-01 -7.84614205e-01 -3.59281451e-01 -4.20183957e-01 -8.29068601e-01 -1.95836127e-02 6.50830388e-01 3.03609259e-02 7.30431557e-01 -1.51530886e+00 -2.30721906e-01 7.18729138e-01 -3.22939813e-01 -3.74400020e-01 -1.70157343e-01 6.58265352e-01 -1.32361674e+00 1.02442718e+00 -5.41665435e-01 -4.55340147e-01 -1.14228773e+00 4.78126436e-01 4.28307921e-01 -2.41845414e-01 -2.15481281e-01 1.11954463e+00 2.71224789e-02 -8.23025763e-01 1.85185194e-01 -9.21545625e-02 -7.54407048e-01 3.55081353e-03 6.88606799e-01 1.05194807e+00 -2.91290224e-01 -6.04341030e-01 -4.56784427e-01 6.65885091e-01 2.32151806e-01 -2.87484169e-01 9.17833567e-01 -2.19243199e-01 -9.89108324e-01 4.28274393e-01 8.38715494e-01 -1.36269778e-01 -3.69319022e-02 4.16855574e-01 1.25943512e-01 -8.02164078e-01 -4.32406247e-01 -3.55811834e-01 -1.85641110e-01 5.54822803e-01 5.34874737e-01 8.99602413e-01 8.57008278e-01 -3.02566767e-01 8.18335712e-01 9.66778398e-01 5.72402716e-01 -1.68019545e+00 -2.13140488e-01 6.89524531e-01 9.12339568e-01 -9.28350806e-01 5.44190466e-01 -7.27165341e-01 -4.14997995e-01 1.32979155e+00 4.68304873e-01 -1.55325383e-01 1.27306652e+00 5.74917436e-01 1.14069022e-02 -3.70670199e-01 -9.44949508e-01 1.12705544e-01 3.75366658e-01 1.11147714e+00 5.06513238e-01 5.07374525e-01 -9.66500878e-01 3.21953118e-01 -7.28706717e-01 -2.34555230e-01 4.90474731e-01 1.41972518e+00 -6.43810987e-01 -1.20391917e+00 -7.23931909e-01 4.61561680e-01 -5.49773455e-01 -1.16372630e-01 -5.28106689e-01 8.46754074e-01 -5.76629937e-02 1.49448860e+00 -1.97535474e-03 -5.03491640e-01 4.11356747e-01 1.41089618e-01 7.21762300e-01 -1.23501487e-01 -9.04846430e-01 4.15413082e-01 3.84890139e-01 -1.23056793e+00 -8.58632207e-01 -1.05834293e+00 -1.40667629e+00 -1.19437599e+00 -5.12782812e-01 1.89310342e-01 3.83317530e-01 3.90289724e-01 -2.06836104e-01 2.85260603e-02 9.80917513e-01 -1.07949340e+00 -3.90341938e-01 -9.39418614e-01 -1.00026262e+00 -7.84516707e-02 2.89751232e-01 -8.00943017e-01 -7.83523321e-01 2.83909619e-01]
[-7.320960521697998, 3.691803455352783]
eb2184a5-1f4f-4c5b-8c26-165c9beed047
coactseg-learning-from-heterogeneous-data-for
2307.04513
null
https://arxiv.org/abs/2307.04513v1
https://arxiv.org/pdf/2307.04513v1.pdf
CoactSeg: Learning from Heterogeneous Data for New Multiple Sclerosis Lesion Segmentation
New lesion segmentation is essential to estimate the disease progression and therapeutic effects during multiple sclerosis (MS) clinical treatments. However, the expensive data acquisition and expert annotation restrict the feasibility of applying large-scale deep learning models. Since single-time-point samples with all-lesion labels are relatively easy to collect, exploiting them to train deep models is highly desirable to improve new lesion segmentation. Therefore, we proposed a coaction segmentation (CoactSeg) framework to exploit the heterogeneous data (i.e., new-lesion annotated two-time-point data and all-lesion annotated single-time-point data) for new MS lesion segmentation. The CoactSeg model is designed as a unified model, with the same three inputs (the baseline, follow-up, and their longitudinal brain differences) and the same three outputs (the corresponding all-lesion and new-lesion predictions), no matter which type of heterogeneous data is being used. Moreover, a simple and effective relation regularization is proposed to ensure the longitudinal relations among the three outputs to improve the model learning. Extensive experiments demonstrate that utilizing the heterogeneous data and the proposed longitudinal relation constraint can significantly improve the performance for both new-lesion and all-lesion segmentation tasks. Meanwhile, we also introduce an in-house MS-23v1 dataset, including 38 Oceania single-time-point samples with all-lesion labels. Codes and the dataset are released at https://github.com/ycwu1997/CoactSeg.
['Jianfei Cai', 'Winston Chong', 'Bjoern Picker', 'Hengcan Shi', 'Zhonghua Wu', 'Yicheng Wu']
2023-07-10
null
null
null
null
['lesion-segmentation']
['medical']
[ 1.17400460e-01 -3.62469614e-01 -6.82352662e-01 -6.49776399e-01 -1.02753055e+00 -3.11932564e-01 4.54381913e-01 -9.48925018e-02 -4.96788204e-01 8.59713674e-01 -7.12250471e-02 -2.92020351e-01 -2.80031711e-01 -4.41718638e-01 -5.78735054e-01 -9.03935254e-01 -1.37474313e-01 7.66987085e-01 3.67765576e-01 1.08123176e-01 3.62032056e-02 5.15583456e-01 -9.68774736e-01 4.47859168e-01 1.23642933e+00 9.51009989e-01 5.72882533e-01 2.74676651e-01 -1.14383604e-02 1.77553594e-01 -2.65275061e-01 2.16793180e-01 2.93431252e-01 -3.05691093e-01 -7.22566307e-01 1.90872774e-01 2.61647731e-01 -4.93735939e-01 -3.16028535e-01 8.65851223e-01 8.49114418e-01 -3.76635008e-02 6.48318708e-01 -1.22896469e+00 -5.04055679e-01 3.62259001e-01 -8.82483304e-01 6.00971937e-01 -4.37709779e-01 7.29292512e-01 5.32929242e-01 -6.78317010e-01 8.42287540e-01 9.95001733e-01 6.56297743e-01 4.83066857e-01 -1.22344673e+00 -5.89878261e-01 2.70231992e-01 4.78858769e-01 -1.20391464e+00 -2.32737511e-01 3.78284097e-01 -7.95490444e-01 6.77747190e-01 2.81838596e-01 7.03280449e-01 1.14693642e+00 2.27759942e-01 9.66253340e-01 1.34680271e+00 4.81790490e-02 -5.90491891e-02 -2.81236589e-01 5.65600514e-01 4.20661509e-01 -3.34752351e-01 -2.81458301e-03 8.85728225e-02 5.24213612e-02 9.70377326e-01 2.24994346e-01 -3.91599268e-01 -2.54284710e-01 -1.53306985e+00 7.32254982e-01 6.80435658e-01 5.47406495e-01 -4.66143101e-01 -1.51869416e-01 4.37032133e-01 -8.96537453e-02 6.26714826e-01 -1.10702880e-01 -5.79949677e-01 2.25050583e-01 -1.13294399e+00 2.67720699e-01 -1.01983696e-01 6.68100417e-01 3.71660441e-01 -9.21417326e-02 -5.92033386e-01 1.22147214e+00 -1.77228913e-01 5.40629685e-01 7.50313520e-01 -7.78352439e-01 6.37912035e-01 7.69543350e-01 -2.34583318e-01 -5.98641694e-01 -9.20224071e-01 -8.06090117e-01 -1.04898763e+00 3.62674531e-04 4.72941577e-01 -3.43406409e-01 -1.43896580e+00 1.76700842e+00 3.69476378e-01 2.36838624e-01 -5.38683355e-01 1.09420848e+00 4.88280058e-01 4.45842594e-01 -6.67077005e-02 -4.81280476e-01 1.26426852e+00 -9.37220216e-01 -6.92985058e-01 -2.20672131e-01 9.24853086e-01 -3.00197989e-01 1.19287837e+00 1.27536848e-01 -9.53697503e-01 -2.46836856e-01 -6.63500905e-01 -6.94616735e-02 -3.54476124e-01 5.05758107e-01 5.97676039e-01 5.12721278e-02 -8.36254060e-01 4.78417754e-01 -1.12825036e+00 -1.93703726e-01 8.27722371e-01 5.32385230e-01 -3.51442724e-01 -9.12071615e-02 -1.26250625e+00 9.87387359e-01 3.99767101e-01 3.58535767e-01 -7.82501400e-01 -1.08730412e+00 -1.07733130e-01 -3.52339685e-01 3.64722729e-01 -5.74225247e-01 9.97469127e-01 -6.12050474e-01 -1.04342985e+00 9.85352278e-01 -3.22920799e-01 -1.72139555e-01 7.32215285e-01 1.56135052e-01 -5.34333527e-01 2.26015165e-01 3.20234954e-01 7.02771962e-01 4.78216290e-01 -1.03479946e+00 -6.36101723e-01 -9.16022837e-01 -2.59302497e-01 5.11396676e-02 -1.36198983e-01 1.17554508e-01 -4.64202017e-01 -7.12421358e-01 2.35751674e-01 -9.55653727e-01 -4.01516676e-01 7.42285922e-02 -6.66664600e-01 -2.09657237e-01 6.77464843e-01 -1.17572236e+00 1.20662284e+00 -1.70295846e+00 1.10964365e-01 1.68156326e-02 5.22716582e-01 3.55561614e-01 -2.39750251e-01 -2.12306440e-01 -5.05617023e-01 2.29352206e-01 -5.30475080e-01 -7.48296157e-02 -1.96307480e-01 -1.46153763e-01 3.25041980e-01 4.61193949e-01 1.80756435e-01 1.17275262e+00 -5.83045363e-01 -4.47878182e-01 1.22021556e-01 3.02440703e-01 -2.96479911e-01 -1.68253541e-01 -9.65052396e-02 1.00255334e+00 -7.51612902e-01 6.44330621e-01 6.94745600e-01 -3.51111978e-01 -1.06425025e-01 -4.45303649e-01 -7.32530579e-02 -1.77881867e-01 -8.49221528e-01 1.58616745e+00 -1.04871169e-01 2.97020674e-01 3.16534527e-02 -1.32267547e+00 5.28906584e-01 3.58275712e-01 1.12534726e+00 -8.67395401e-01 2.54906654e-01 4.11149144e-01 2.63250917e-01 -7.38367915e-01 -3.13959181e-01 -2.53213465e-01 2.74320364e-01 4.72713888e-01 -1.29016325e-01 3.13094676e-01 4.54271466e-01 -1.98359340e-02 9.59387720e-01 -8.17202255e-02 5.93532342e-03 -9.46317986e-02 4.24876720e-01 9.45805460e-02 9.82545674e-01 5.20606399e-01 -6.08006001e-01 7.14711845e-01 7.27826595e-01 -1.63244396e-01 -1.02763557e+00 -1.08330595e+00 -6.20006144e-01 6.88356578e-01 -1.85410544e-01 2.60965079e-01 -7.00012863e-01 -8.49426687e-01 3.04539483e-02 4.23695594e-01 -6.89560771e-01 -5.08355647e-02 -7.38498747e-01 -1.33575916e+00 4.02720779e-01 6.70940042e-01 4.23757643e-01 -8.02393854e-01 -9.97213572e-02 2.30184287e-01 -4.51098621e-01 -8.53611469e-01 -7.07603276e-01 3.50242145e-02 -1.15810561e+00 -1.15121365e+00 -1.33785200e+00 -7.51615524e-01 7.26314604e-01 4.33432795e-02 4.28765178e-01 -1.04039341e-01 -3.54368895e-01 -2.10180715e-01 -1.08532190e-01 -1.16208084e-01 6.73711747e-02 3.85076180e-02 -3.41636799e-02 1.47914067e-01 2.46687949e-01 -5.61616004e-01 -7.49958336e-01 2.10823312e-01 -8.90541375e-01 3.25428218e-01 7.68136561e-01 8.62286866e-01 9.95829403e-01 -1.67520344e-01 1.04573870e+00 -6.74469531e-01 4.85723585e-01 -5.16170442e-01 -1.80521980e-01 6.17554247e-01 -5.20736516e-01 -4.01556760e-01 2.55785346e-01 -7.17577934e-01 -9.15976584e-01 -4.14986536e-02 -1.14956290e-01 -3.80361646e-01 -1.67679757e-01 7.81343579e-01 -2.03760520e-01 1.77743271e-01 4.95259523e-01 1.69872791e-01 1.76303297e-01 -6.26120925e-01 3.51996124e-01 9.44292605e-01 5.22990763e-01 -3.18868965e-01 2.27855250e-01 4.85172719e-01 -4.34925482e-02 -5.31026304e-01 -7.74021208e-01 -6.30173564e-01 -1.06717062e+00 -2.51770735e-01 8.50271285e-01 -6.35247588e-01 -2.67247528e-01 1.14587379e+00 -9.55125332e-01 -6.21319592e-01 -4.73104082e-02 8.15475523e-01 -6.82777166e-01 2.64923960e-01 -5.59179902e-01 -2.67006904e-01 -6.58819973e-01 -1.54833806e+00 9.49626386e-01 8.33014473e-02 1.80712745e-01 -8.69489551e-01 -2.05488175e-01 5.90053260e-01 1.99569777e-01 3.63032401e-01 1.30728054e+00 -7.48911202e-01 -3.01517487e-01 -3.71490091e-01 -5.04553258e-01 2.42115602e-01 3.51751119e-01 -3.44411731e-01 -3.33938360e-01 -3.05859089e-01 1.25390962e-01 -1.31251082e-01 9.23668623e-01 1.11265111e+00 1.27604735e+00 2.09506691e-01 -4.70730126e-01 6.60394013e-01 1.09945524e+00 4.11531866e-01 5.62990785e-01 4.69539762e-01 7.13148952e-01 5.18863142e-01 4.60821331e-01 2.60251462e-01 2.75785565e-01 9.25134659e-01 -7.99960792e-02 -2.33193040e-01 -3.95243108e-01 4.23880994e-01 3.92735824e-02 8.33431005e-01 -6.46624789e-02 2.91876849e-02 -1.20456100e+00 6.28789961e-01 -1.87463725e+00 -8.28248918e-01 -6.12395942e-01 2.06742334e+00 1.25373781e+00 -4.98915091e-03 1.40154898e-01 -1.84584618e-01 1.13498902e+00 -4.35836241e-02 -1.15412569e+00 2.79999971e-01 -3.13112885e-01 1.13849752e-01 2.98033804e-01 3.56020212e-01 -1.05768776e+00 7.68292308e-01 5.27454281e+00 1.16731584e+00 -1.41172111e+00 5.21269679e-01 6.93920910e-01 -4.45257723e-01 1.70885008e-02 -1.95021123e-01 -7.93046594e-01 6.85321033e-01 6.36113226e-01 -1.13322563e-01 3.82505566e-01 1.36353090e-01 7.68413603e-01 -9.97908339e-02 -8.23150456e-01 6.35110199e-01 -2.37553149e-01 -1.25058734e+00 -8.30120891e-02 -7.69422278e-02 5.67096949e-01 2.89458066e-01 5.86746074e-02 3.11016053e-01 -1.77961677e-01 -7.29295552e-01 3.54096234e-01 8.80390227e-01 1.00378501e+00 -3.87732446e-01 6.07036591e-01 5.39193690e-01 -8.37126195e-01 5.40275835e-02 6.13157004e-02 5.85100949e-01 4.70745236e-01 9.13242221e-01 -6.24442101e-01 8.10912669e-01 6.04017794e-01 9.05370116e-01 -7.97119021e-01 1.38749576e+00 -5.71547709e-02 5.51096261e-01 -3.14472020e-01 5.84844053e-01 2.12210834e-01 -3.19345057e-01 6.81780457e-01 8.48896563e-01 3.46428990e-01 1.44669935e-01 2.65687436e-01 1.08027709e+00 3.68837595e-01 1.31195575e-01 8.69797990e-02 -1.64884523e-01 1.02505311e-01 1.20047510e+00 -6.74822569e-01 -2.82415956e-01 -4.40484732e-01 5.51820576e-01 1.57793447e-01 7.39314377e-01 -7.49139369e-01 -2.42813155e-01 1.61433965e-01 2.03827485e-01 -1.17781207e-01 -1.61007419e-01 -5.59534848e-01 -1.28394461e+00 6.36034235e-02 -8.69559884e-01 4.29814786e-01 -8.42023790e-01 -1.46717346e+00 4.66814876e-01 -5.83547074e-03 -1.13271320e+00 3.14002335e-02 -4.16811973e-01 -6.02432430e-01 1.26975214e+00 -1.27107298e+00 -1.48764813e+00 -6.52495548e-02 7.48297274e-01 3.09040368e-01 -1.97265267e-01 4.32580322e-01 5.86172462e-01 -1.23186815e+00 4.17853177e-01 3.25982779e-01 1.06108703e-01 7.77029097e-01 -8.63670290e-01 5.81726693e-02 6.48742735e-01 -5.66991389e-01 6.58769608e-01 5.96774481e-02 -1.15952253e+00 -7.63861895e-01 -1.19707155e+00 6.56796992e-01 1.26237841e-02 9.44281518e-01 -5.53989112e-02 -1.23129642e+00 7.97967255e-01 -3.59408468e-01 9.06283185e-02 4.00552064e-01 -7.70509392e-02 2.15288714e-01 5.78172654e-02 -1.18473530e+00 6.60208344e-01 1.05278385e+00 -3.51909846e-01 -5.64441144e-01 6.14486158e-01 7.38978863e-01 -4.97226417e-01 -1.17987895e+00 9.34043884e-01 3.83317918e-01 -3.17640811e-01 9.56864953e-01 -9.29772019e-01 2.51533747e-01 -9.93515402e-02 5.98329753e-02 -1.30528426e+00 -3.12477618e-01 7.72233680e-02 -1.22979477e-01 1.12930155e+00 3.70738953e-01 -8.16761076e-01 7.39123523e-01 5.28131783e-01 -5.58111846e-01 -1.32370031e+00 -9.79019940e-01 -9.01092112e-01 5.75338483e-01 -2.63075531e-01 4.89029437e-01 1.11051583e+00 -3.79025221e-01 3.37997861e-02 -1.50783241e-01 1.18971802e-01 6.08340323e-01 8.66857916e-02 1.59892648e-01 -1.20243776e+00 1.23220749e-01 -6.13530219e-01 -2.00157166e-01 -6.63168490e-01 1.15032487e-01 -1.51173651e+00 -1.51393369e-01 -1.87168086e+00 4.60954398e-01 -6.19591177e-01 -6.30975842e-01 7.45813787e-01 -3.67848366e-01 8.08880373e-04 -5.21401651e-02 5.12830377e-01 -1.98484495e-01 5.07354736e-01 1.86014867e+00 -2.13053942e-01 -3.52337807e-01 2.66067207e-01 -3.83851737e-01 6.00668967e-01 9.37188625e-01 -3.81253600e-01 -3.70449424e-01 -4.64163452e-01 -5.96058309e-01 4.31695394e-02 6.62497282e-01 -8.82188737e-01 1.98177114e-01 -4.96297061e-01 6.00255132e-01 -9.29672301e-01 1.53445378e-01 -4.55100745e-01 3.92147638e-02 4.81237590e-01 -4.11157906e-01 -4.79949087e-01 1.23309866e-01 3.76694977e-01 -5.21953590e-02 -2.66949773e-01 9.59290981e-01 -6.54014274e-02 -6.38647377e-01 1.01672029e+00 -1.26493961e-01 1.93105534e-01 1.16627991e+00 -1.88720226e-01 -5.92491865e-01 4.68572043e-02 -1.35134554e+00 7.21601963e-01 -3.83053608e-02 4.98069495e-01 4.10262823e-01 -1.55139506e+00 -8.39204311e-01 -4.15646732e-02 -1.21207513e-01 7.39017222e-03 8.35429728e-01 1.71147466e+00 -2.11720973e-01 6.33004129e-01 -4.53953475e-01 -9.23249483e-01 -9.71259058e-01 3.43459994e-01 5.36614299e-01 -5.24668932e-01 -8.07693541e-01 5.85300505e-01 4.38284010e-01 -7.23185956e-01 2.09092468e-01 -2.43987784e-01 -2.07650527e-01 2.23282918e-01 4.53531235e-01 4.12605554e-01 2.71432966e-01 -4.85736489e-01 -2.12141573e-01 5.12499630e-01 -3.68294448e-01 -2.23108642e-02 1.45864773e+00 -2.40591437e-01 -2.73429960e-01 5.75573087e-01 1.20954144e+00 -6.95186019e-01 -1.16675007e+00 -4.75340009e-01 7.49413148e-02 -2.03753993e-01 3.36557120e-01 -1.19226015e+00 -1.41079640e+00 1.10305214e+00 1.04236984e+00 -3.93005788e-01 1.09419060e+00 -1.06091693e-01 1.16019142e+00 4.63203490e-02 4.17592257e-01 -1.03296185e+00 -1.99646443e-01 3.21832418e-01 1.23003983e+00 -1.19157958e+00 -8.34991559e-02 -3.28397512e-01 -7.22981036e-01 8.30358565e-01 6.78361356e-01 2.50293314e-01 5.87064922e-01 -9.20074210e-02 7.47412890e-02 -2.10066274e-01 -5.07013798e-01 -2.29351476e-01 4.33231235e-01 8.15574050e-01 1.83148399e-01 2.71887720e-01 -7.24860311e-01 1.16159153e+00 3.26827079e-01 4.32916164e-01 1.86421514e-01 6.90582454e-01 -3.32209915e-01 -1.18883121e+00 -1.44697040e-01 1.15985763e+00 -2.77975053e-01 5.07302321e-02 -1.66481271e-01 8.10606241e-01 3.88647825e-01 5.84648073e-01 -1.56817868e-01 -2.31483057e-01 4.94174331e-01 1.61533684e-01 4.23956990e-01 -5.80120802e-01 -4.78113592e-01 4.00283545e-01 -1.83249414e-01 -4.29634184e-01 -3.13466072e-01 -8.39678347e-01 -1.93959284e+00 -1.10417791e-01 -4.17976975e-01 -2.72832036e-01 5.50975025e-01 1.15492058e+00 4.02529836e-01 7.86013544e-01 6.60591066e-01 -6.75221026e-01 -5.96427858e-01 -1.12507451e+00 -8.58090460e-01 2.25701660e-01 1.39548510e-01 -9.40477431e-01 -8.20275098e-02 1.98463872e-01]
[14.434042930603027, -2.0264604091644287]
0c29cf14-f038-4ee3-88aa-0996d33d5938
frame-wise-cross-modal-match-for-video-moment
2009.10434
null
https://arxiv.org/abs/2009.10434v2
https://arxiv.org/pdf/2009.10434v2.pdf
Frame-wise Cross-modal Matching for Video Moment Retrieval
Video moment retrieval targets at retrieving a moment in a video for a given language query. The challenges of this task include 1) the requirement of localizing the relevant moment in an untrimmed video, and 2) bridging the semantic gap between textual query and video contents. To tackle those problems, early approaches adopt the sliding window or uniform sampling to collect video clips first and then match each clip with the query. Obviously, these strategies are time-consuming and often lead to unsatisfied accuracy in localization due to the unpredictable length of the golden moment. To avoid the limitations, researchers recently attempt to directly predict the relevant moment boundaries without the requirement to generate video clips first. One mainstream approach is to generate a multimodal feature vector for the target query and video frames (e.g., concatenation) and then use a regression approach upon the multimodal feature vector for boundary detection. Although some progress has been achieved by this approach, we argue that those methods have not well captured the cross-modal interactions between the query and video frames. In this paper, we propose an Attentive Cross-modal Relevance Matching (ACRM) model which predicts the temporal boundaries based on an interaction modeling. In addition, an attention module is introduced to assign higher weights to query words with richer semantic cues, which are considered to be more important for finding relevant video contents. Another contribution is that we propose an additional predictor to utilize the internal frames in the model training to improve the localization accuracy. Extensive experiments on two datasets TACoS and Charades-STA demonstrate the superiority of our method over several state-of-the-art methods. Ablation studies have been also conducted to examine the effectiveness of different modules in our ACRM model.
['IEEE', 'Meng Liu', 'Jihua Zhu', 'Haoyu Tang', 'Zhiyong Cheng', 'Zan Gao', 'Member']
2020-09-22
null
null
null
null
['moment-retrieval']
['computer-vision']
[ 2.75719345e-01 -4.67247158e-01 -3.96076769e-01 -1.06777877e-01 -9.74544585e-01 -3.03263903e-01 6.74543560e-01 4.77794483e-02 -4.37167972e-01 3.12261194e-01 2.76780218e-01 2.37893492e-01 -1.09359026e-01 -4.23620641e-01 -6.02239668e-01 -5.61841488e-01 -4.33307998e-02 3.67382611e-03 5.78992844e-01 -1.08122148e-01 5.33982813e-01 1.57604545e-01 -1.78001750e+00 5.32766104e-01 7.55527020e-01 1.25867665e+00 6.18412077e-01 4.15273279e-01 -2.60485619e-01 6.94425046e-01 -5.03003716e-01 -2.16699570e-01 1.05143435e-01 -5.40806830e-01 -6.55622959e-01 1.26206860e-01 2.75184840e-01 -3.56983304e-01 -4.54369634e-01 1.04446793e+00 5.22457957e-01 3.21030706e-01 4.80512679e-01 -1.35521364e+00 -5.27109385e-01 4.60142255e-01 -7.57691860e-01 4.97347265e-01 8.97574604e-01 -2.23677248e-01 9.90487397e-01 -1.06778538e+00 6.73996270e-01 1.07069373e+00 4.01406705e-01 2.65369952e-01 -5.85724592e-01 -6.75914466e-01 3.74358088e-01 9.08360124e-01 -1.64052904e+00 -4.97111827e-01 1.13547742e+00 -3.24288070e-01 6.09287143e-01 3.58532041e-01 6.96337700e-01 1.18149114e+00 1.35770589e-01 9.44561839e-01 5.28768897e-01 -4.40359354e-01 -1.02781482e-01 5.28540611e-02 -7.60113001e-02 4.40249175e-01 -2.93175876e-01 -1.53894484e-01 -6.71970308e-01 1.22775370e-02 5.77947378e-01 2.12170288e-01 -4.55791622e-01 -1.16146930e-01 -1.40932691e+00 6.88885450e-01 1.85866967e-01 7.01114595e-01 -4.48658019e-01 -6.43213019e-02 5.57367384e-01 8.90638158e-02 3.81984860e-01 1.87711403e-01 -6.92833140e-02 -1.19876109e-01 -1.09543657e+00 1.46023110e-01 4.52765405e-01 9.86368954e-01 7.11325824e-01 -3.23944807e-01 -4.17817682e-01 7.74320364e-01 2.93349326e-01 2.29695857e-01 7.11374700e-01 -7.00632334e-01 6.75932050e-01 5.93846619e-01 1.42746404e-01 -1.55954707e+00 -1.12700924e-01 -7.44760409e-02 -4.27419275e-01 -3.93915832e-01 2.27691859e-01 1.58369109e-01 -7.42984354e-01 1.65713847e+00 3.50584596e-01 4.62186962e-01 -2.92561557e-02 1.28406680e+00 8.94011974e-01 9.68302369e-01 1.48350015e-01 -4.10774827e-01 1.46782196e+00 -9.51824009e-01 -8.03346157e-01 -1.05936579e-01 3.57339472e-01 -1.05933273e+00 9.78585541e-01 1.53759301e-01 -1.00062537e+00 -8.08118641e-01 -9.27984238e-01 -4.34275381e-02 -4.31154370e-01 2.53176659e-01 4.14713740e-01 1.33586869e-01 -6.99331105e-01 2.95731634e-01 -5.45372128e-01 -5.36589026e-01 -7.87608996e-02 1.60046935e-01 -3.65231365e-01 -1.53706754e-02 -1.60277677e+00 6.05592430e-01 6.21928811e-01 2.65314639e-01 -7.31615067e-01 -3.38860691e-01 -8.10531318e-01 6.84808046e-02 6.58898294e-01 -4.92309481e-01 9.33286667e-01 -1.37105560e+00 -1.22690666e+00 6.01447880e-01 -3.53063017e-01 -2.76279628e-01 3.35998178e-01 -2.48842746e-01 -5.62338650e-01 6.59429252e-01 2.62067139e-01 7.88390398e-01 1.00564873e+00 -1.07468820e+00 -9.95728612e-01 -1.34384766e-01 3.20528269e-01 5.62366486e-01 -5.32549977e-01 3.08938116e-01 -1.19002330e+00 -8.53904188e-01 2.72630632e-01 -8.40990424e-01 2.57658549e-02 -3.72018337e-01 -1.24927193e-01 -3.54222447e-01 9.84405637e-01 -8.20000708e-01 1.69331551e+00 -2.14096498e+00 2.51622736e-01 -5.72786806e-03 -1.56443864e-01 1.53255403e-01 -1.87086865e-01 4.93761301e-01 -4.43130881e-02 3.35857123e-02 1.39797986e-01 -1.68626070e-01 -2.21255973e-01 -8.48421231e-02 -4.15757537e-01 3.57349396e-01 1.04742022e-02 7.44831562e-01 -8.28854918e-01 -1.05427659e+00 1.75928473e-01 5.16852617e-01 -4.30330932e-01 3.48238647e-01 -1.53973356e-01 3.64326417e-01 -5.38134992e-01 7.89566159e-01 3.94685537e-01 -8.86341780e-02 -1.41597807e-01 -6.53481603e-01 -1.69027805e-01 -1.27045780e-01 -1.23840010e+00 1.93201685e+00 -2.16629192e-01 5.97058415e-01 -7.94334039e-02 -1.07789195e+00 6.43276513e-01 5.98918021e-01 7.78748453e-01 -7.34970212e-01 1.20430991e-01 8.78313854e-02 -2.88841218e-01 -1.05273187e+00 7.55516529e-01 1.61981821e-01 -1.11510560e-01 1.11587653e-02 -1.30518988e-01 2.96864122e-01 2.84940064e-01 8.70799199e-02 7.12237656e-01 3.55112195e-01 2.31376767e-01 1.77642897e-01 9.53660369e-01 -6.86872154e-02 6.14496112e-01 5.77387989e-01 -3.85322630e-01 7.54936337e-01 3.07396591e-01 -2.55370468e-01 -7.57035613e-01 -5.43782473e-01 1.82049751e-01 1.14670002e+00 7.64775276e-01 -5.68264544e-01 -6.81284547e-01 -6.15604699e-01 -4.40219551e-01 3.07262421e-01 -4.64740157e-01 -2.20585480e-01 -6.63038075e-01 -4.37349379e-01 3.64315301e-01 3.94268662e-01 5.73837578e-01 -1.00414479e+00 -5.76009870e-01 1.14699557e-01 -8.69261861e-01 -1.21976233e+00 -9.24158871e-01 -4.58526850e-01 -5.70528626e-01 -8.83066475e-01 -7.51694918e-01 -9.40989196e-01 4.70275521e-01 6.52327657e-01 6.37459934e-01 1.12677954e-01 -2.43602246e-01 4.99873549e-01 -7.32626379e-01 5.21781072e-02 1.01363264e-01 1.46680683e-01 -4.63360809e-02 4.61018682e-01 4.15100992e-01 -2.82777011e-01 -8.74000371e-01 6.13440990e-01 -1.14450598e+00 9.80171040e-02 5.63997686e-01 7.45974004e-01 5.93152165e-01 2.08774265e-02 5.79590142e-01 -2.60349184e-01 4.98843282e-01 -7.28926301e-01 -3.49656284e-01 4.43365544e-01 -3.26117396e-01 -1.70525700e-01 5.49500585e-01 -7.52317905e-01 -1.06811976e+00 8.46632570e-02 -1.50229940e-02 -8.16935599e-01 -1.84218697e-02 7.17433691e-01 -1.95743725e-01 1.21640541e-01 1.52988002e-01 3.80735457e-01 -2.43855447e-01 -2.82553703e-01 1.22103691e-01 7.06772029e-01 4.29974735e-01 -4.77841228e-01 5.44032514e-01 4.91052628e-01 -1.95545271e-01 -8.21290851e-01 -6.46236718e-01 -8.54060113e-01 -4.25128222e-01 -5.62019885e-01 1.01055419e+00 -9.49707150e-01 -6.02396071e-01 2.04323158e-01 -1.20441759e+00 3.31141055e-01 2.91146785e-01 7.57916927e-01 -5.22813380e-01 8.85372221e-01 -5.00286102e-01 -7.44991541e-01 -2.32601538e-01 -1.24745762e+00 1.19508052e+00 4.20769751e-01 -1.20566927e-01 -6.78671539e-01 -1.50107980e-01 5.16450226e-01 3.22478354e-01 -4.18606475e-02 5.47978044e-01 -6.36944711e-01 -7.92542577e-01 -4.31284606e-01 -3.06737185e-01 -1.68812677e-01 1.18231103e-01 1.23767391e-01 -6.94527566e-01 -1.06257886e-01 1.01048209e-01 -1.19343735e-01 7.26085961e-01 3.21485043e-01 1.17727900e+00 -1.82772040e-01 -5.38049340e-01 4.88804251e-01 1.30454516e+00 5.44915855e-01 6.92285836e-01 4.83019531e-01 5.04936159e-01 6.67187929e-01 1.27896285e+00 4.66879517e-01 3.32712263e-01 9.71937776e-01 4.33364630e-01 1.47922933e-01 5.69409691e-02 -3.63725245e-01 5.95591366e-01 8.12518954e-01 -1.38773695e-01 -3.40222239e-01 -6.81709170e-01 6.62840009e-01 -2.12070608e+00 -1.39088297e+00 1.87063575e-01 2.05868244e+00 6.10723615e-01 1.91336847e-04 -1.15255155e-02 -9.99056455e-03 9.95641708e-01 2.88242817e-01 -2.70773262e-01 1.38993099e-01 -6.83286041e-02 -3.57626200e-01 9.70125496e-02 3.58416855e-01 -1.33584738e+00 9.47159648e-01 5.27829123e+00 1.21278822e+00 -1.28128254e+00 5.95582426e-02 4.42499250e-01 -1.16435744e-01 -5.27893193e-02 1.59685180e-01 -8.16919684e-01 6.90125048e-01 6.80081606e-01 -2.82526780e-02 3.54936570e-01 7.42653072e-01 4.70096678e-01 -2.99601018e-01 -9.51804221e-01 1.22965205e+00 5.12806654e-01 -1.04679251e+00 1.29927173e-01 -3.12285602e-01 4.38225657e-01 -4.15559083e-01 -3.25108357e-02 3.56615096e-01 -7.50701249e-01 -6.35356188e-01 8.55003893e-01 8.60825360e-01 5.71111321e-01 -6.77594304e-01 7.55052328e-01 2.97393382e-01 -1.60559785e+00 -1.22810856e-01 -4.17424679e-01 1.88746989e-01 2.32508957e-01 1.44033924e-01 -6.33114874e-01 7.52029181e-01 8.59764993e-01 7.29963541e-01 -6.53144538e-01 1.22707105e+00 1.08018368e-01 2.49530271e-01 -2.64317989e-01 7.45019093e-02 1.62803397e-01 -2.25271627e-01 7.16525912e-01 1.23044121e+00 6.14623904e-01 -3.53766815e-03 3.55305910e-01 6.47238910e-01 1.95186988e-01 4.63511348e-01 -6.11084163e-01 -5.76197095e-02 4.15838599e-01 1.27100134e+00 -8.70873034e-01 -2.88556516e-01 -6.81748748e-01 1.13741839e+00 5.91309695e-03 5.41247129e-01 -1.25544775e+00 -5.32793939e-01 3.25261265e-01 6.58659860e-02 2.89662838e-01 -1.31513670e-01 2.39337862e-01 -1.27466083e+00 2.74067581e-01 -7.64405906e-01 4.84197974e-01 -8.80616248e-01 -1.03910756e+00 5.27230918e-01 2.16389269e-01 -1.70904779e+00 -3.35347116e-01 -8.66877288e-02 -4.58921939e-01 7.09646702e-01 -1.43311465e+00 -1.17669141e+00 -4.11531776e-01 7.70517945e-01 9.02664483e-01 -5.39791677e-03 2.84645230e-01 7.16669142e-01 -6.72132909e-01 4.88825023e-01 -2.02392578e-01 6.83094561e-02 9.95481312e-01 -5.96285403e-01 -2.82927036e-01 9.67498899e-01 8.12014490e-02 6.38753653e-01 5.96348822e-01 -7.28446066e-01 -1.40280032e+00 -8.48019838e-01 9.77521718e-01 -2.15739876e-01 6.17598891e-01 -5.24151474e-02 -8.25792789e-01 5.39138854e-01 2.79061586e-01 -1.45267203e-01 4.84983653e-01 -3.90494168e-01 -1.21211950e-02 -3.13749522e-01 -6.42338276e-01 7.94777453e-01 8.82815182e-01 -6.18402719e-01 -7.67707705e-01 3.67235802e-02 7.61983633e-01 -2.46771738e-01 -7.40750194e-01 4.93831515e-01 6.62068486e-01 -1.03797638e+00 1.06538320e+00 -2.56928533e-01 4.16806936e-01 -5.94590068e-01 -2.66639113e-01 -7.87347078e-01 -4.57394160e-02 -5.02883255e-01 -6.38794452e-02 1.54202116e+00 8.63385499e-02 -1.18827924e-01 5.32748580e-01 5.56439817e-01 -8.50827768e-02 -6.80247009e-01 -9.15103257e-01 -4.98343915e-01 -5.39703012e-01 -5.47781706e-01 4.17207569e-01 7.88026810e-01 2.48790309e-01 3.14017028e-01 -6.45172298e-01 1.82774499e-01 8.46805423e-02 3.58201385e-01 5.69075048e-01 -8.61973047e-01 -8.07071701e-02 -4.91592050e-01 -4.46911663e-01 -1.32249260e+00 1.65921032e-01 -5.25703847e-01 1.26100883e-01 -1.32550442e+00 3.79516691e-01 -3.96300405e-02 -3.79587561e-01 4.27065827e-02 -3.82779926e-01 2.26221308e-01 2.86204129e-01 5.48149586e-01 -1.07958925e+00 5.37334561e-01 1.13246977e+00 -1.25277326e-01 -3.35178286e-01 -2.26547550e-02 -3.61506432e-01 6.48961604e-01 5.23664832e-01 -2.26616934e-01 -5.99590898e-01 -3.39623064e-01 1.73236385e-01 4.64156061e-01 3.70096236e-01 -1.02446735e+00 5.37336707e-01 -2.44837567e-01 3.29485774e-01 -9.43598688e-01 5.90683281e-01 -1.06072724e+00 2.37045959e-01 8.90167579e-02 -3.27490538e-01 3.21707010e-01 -1.16642984e-02 7.31478333e-01 -6.90568388e-01 -4.23990548e-01 4.37635988e-01 -1.00892097e-01 -1.13640511e+00 3.01189035e-01 -3.24979812e-01 -1.74771950e-01 1.31251490e+00 -4.46758181e-01 -7.71651193e-02 -5.67155838e-01 -5.97241998e-01 2.82754630e-01 4.02118236e-01 7.10491121e-01 7.68767416e-01 -1.42001510e+00 -3.51191998e-01 3.58173624e-02 1.91835359e-01 -4.49740946e-01 4.76553321e-01 1.06957936e+00 -3.18584949e-01 6.26066029e-01 -3.55469473e-02 -6.62060857e-01 -1.28228879e+00 9.45831180e-01 1.52389362e-01 -1.23926528e-01 -3.87959689e-01 5.56886554e-01 3.12368900e-01 3.81888241e-01 4.72875059e-01 -2.33323619e-01 -6.72794342e-01 5.17025650e-01 5.78851759e-01 1.59397557e-01 -2.06259802e-01 -1.01777935e+00 -3.70891482e-01 8.89097393e-01 -5.67739876e-03 -1.00647129e-01 8.36614788e-01 -6.23804688e-01 -7.40183964e-02 4.35966700e-01 1.31169772e+00 -6.55060634e-02 -1.03331745e+00 -2.58010179e-01 1.16471231e-01 -5.35341501e-01 -1.23668849e-01 -2.98615009e-01 -1.02922761e+00 7.31052279e-01 6.34731114e-01 1.79978654e-01 1.39674866e+00 4.10696398e-03 9.53734517e-01 3.36131863e-02 2.42093354e-01 -1.18072677e+00 2.09407970e-01 2.44634449e-01 8.58803749e-01 -1.35551035e+00 -1.65502533e-01 -3.33818167e-01 -5.87284684e-01 1.21945608e+00 6.64596140e-01 1.91469207e-01 5.20730555e-01 -2.73716331e-01 -6.74016625e-02 4.47428301e-02 -6.41214728e-01 -4.10110325e-01 5.67823589e-01 2.01997757e-01 3.93465459e-01 -4.06940460e-01 -7.41643906e-01 7.25931466e-01 2.90507138e-01 3.87461223e-02 1.82197541e-01 8.04633796e-01 -3.55129480e-01 -1.00640345e+00 -6.51334941e-01 7.43796155e-02 -6.97857797e-01 -6.40984178e-02 -1.67748213e-01 6.96763873e-01 1.66727781e-01 1.14228034e+00 -2.82937605e-02 -5.42681634e-01 2.49519646e-01 1.81912333e-01 1.49611235e-01 -2.21601292e-01 -3.89228314e-01 5.70333898e-01 -2.24000812e-01 -6.40213430e-01 -8.14455628e-01 -6.58578753e-01 -1.11737549e+00 8.14944729e-02 -3.60206157e-01 3.04657042e-01 3.93866211e-01 8.58813584e-01 4.44954157e-01 2.71210700e-01 5.53561866e-01 -1.04252577e+00 -1.89095423e-01 -8.63969624e-01 -2.62851894e-01 6.77697659e-01 1.65679500e-01 -7.74367630e-01 -3.78888667e-01 2.20495716e-01]
[10.189934730529785, 0.7134169340133667]
c3b3cb0d-5003-497d-a25d-d69a0c0273f3
tree-stack-lstm-in-transition-based
null
null
https://aclanthology.org/K18-2012
https://aclanthology.org/K18-2012.pdf
Tree-Stack LSTM in Transition Based Dependency Parsing
We introduce tree-stack LSTM to model state of a transition based parser with recurrent neural networks. Tree-stack LSTM does not use any parse tree based or hand-crafted features, yet performs better than models with these features. We also develop new set of embeddings from raw features to enhance the performance. There are 4 main components of this model: stack{'}s σ-LSTM, buffer{'}s β-LSTM, actions{'} LSTM and tree-RNN. All LSTMs use continuous dense feature vectors (embeddings) as an input. Tree-RNN updates these embeddings based on transitions. We show that our model improves performance with low resource languages compared with its predecessors. We participate in CoNLL 2018 UD Shared Task as the {``}KParse{''} team and ranked 16th in LAS, 15th in BLAS and BLEX metrics, of 27 participants parsing 82 test sets from 57 languages.
['Erenay Dayan{\\i}k', '{\\"O}mer K{\\i}rnap', 'Deniz Yuret']
2018-10-01
null
null
null
conll-2018-10
['transition-based-dependency-parsing', 'morphological-tagging']
['natural-language-processing', 'natural-language-processing']
[ 2.35906720e-01 3.82725924e-01 -1.14058010e-01 -5.89922786e-01 -1.14841211e+00 -6.72977507e-01 2.13985294e-01 2.18706608e-01 -7.84040391e-01 4.16675031e-01 7.04175770e-01 -9.95349228e-01 5.46635687e-01 -8.32465172e-01 -9.92044866e-01 -1.58791304e-01 -2.69667268e-01 3.95251572e-01 5.23006201e-01 -2.50055969e-01 6.13395870e-02 -4.88712415e-02 -9.71470296e-01 6.22524738e-01 5.23445129e-01 7.07094073e-01 4.86166388e-01 1.26801229e+00 -7.03974366e-01 1.20637012e+00 -4.35665369e-01 -5.04249334e-01 -6.19813278e-02 -1.73376858e-01 -1.19201982e+00 -8.77010107e-01 3.30827326e-01 -1.04337826e-01 -3.23406398e-01 9.34647381e-01 3.83596212e-01 7.07032159e-02 1.11016110e-01 -5.09799242e-01 -8.52692187e-01 1.54224443e+00 -1.17573254e-01 6.21029496e-01 2.97127455e-01 7.85845742e-02 1.27024519e+00 -7.68527031e-01 7.17811048e-01 1.48048794e+00 9.26189303e-01 7.45191157e-01 -9.53375578e-01 -4.75296259e-01 6.03257656e-01 7.52284080e-02 -3.88305426e-01 -4.82704908e-01 3.53604585e-01 -1.17068037e-01 1.90985990e+00 9.45467055e-02 3.13650161e-01 1.52877653e+00 4.23014641e-01 1.18111682e+00 1.10105586e+00 -7.46672928e-01 -4.71928269e-02 -2.74321467e-01 8.82569790e-01 1.13816869e+00 -1.05226487e-01 4.73502129e-02 -6.52897894e-01 -3.66793424e-02 3.66581917e-01 -2.50745088e-01 2.80797809e-01 7.00996816e-02 -1.07856023e+00 9.85640883e-01 3.67363125e-01 3.78936380e-01 -3.84269878e-02 4.88303095e-01 8.70219588e-01 4.49107647e-01 5.71415484e-01 2.75451064e-01 -1.04418278e+00 -5.03730237e-01 -5.02103090e-01 1.66003662e-03 8.03844631e-01 1.04479849e+00 5.32567859e-01 1.21501036e-01 -4.07214940e-01 8.93395126e-01 2.88929462e-01 3.53563070e-01 7.85076022e-01 -1.11247897e+00 8.93425524e-01 3.97897482e-01 -1.63169563e-01 -1.43131644e-01 -5.74344456e-01 -1.99837595e-01 -5.96977651e-01 -2.77714193e-01 2.39747107e-01 -3.00064266e-01 -1.35881484e+00 1.88560808e+00 -1.30615875e-01 2.75050895e-03 2.05183581e-01 2.40563169e-01 1.03411758e+00 7.81667292e-01 4.50637788e-01 1.15422502e-01 1.52626729e+00 -1.38121557e+00 -6.35068059e-01 -5.51325321e-01 1.20475888e+00 -5.85786343e-01 1.17371070e+00 2.54581153e-01 -1.28102183e+00 -4.65491146e-01 -9.27441418e-01 -5.20984173e-01 -4.21642035e-01 2.51993060e-01 9.54386950e-01 7.96458185e-01 -1.57747996e+00 8.77551019e-01 -1.37548316e+00 -4.68233615e-01 3.96599993e-02 4.12564486e-01 -3.91244382e-01 7.61201233e-02 -1.26963365e+00 1.05704319e+00 4.77435589e-01 3.80068749e-01 -7.83220053e-01 -4.13339078e-01 -1.26931190e+00 -2.06388216e-02 2.70418264e-02 -5.59561312e-01 1.88706446e+00 -4.62790221e-01 -1.97551286e+00 9.78023708e-01 -4.68053371e-01 -8.93545449e-01 7.64624178e-02 -6.18063986e-01 -2.46528253e-01 -4.62308437e-01 1.19344354e-01 5.35181344e-01 2.87920445e-01 -5.15642107e-01 -7.88288057e-01 -1.53683364e-01 -6.56753108e-02 -1.56049401e-01 1.16343331e-02 4.12205905e-01 -1.94491714e-01 -3.33447933e-01 4.28501695e-01 -8.60953152e-01 -4.29276407e-01 -9.59691465e-01 -6.84064567e-01 -6.57230973e-01 4.56541270e-01 -9.17609036e-01 1.19736671e+00 -1.84014475e+00 1.93313360e-02 -3.93271297e-01 1.32625595e-01 5.10266006e-01 -5.14993548e-01 5.20328283e-01 -1.78116560e-01 5.23308396e-01 -2.29572594e-01 -8.08718920e-01 3.82278264e-02 3.77848029e-01 -5.23163915e-01 -5.42677147e-03 5.03294230e-01 8.45538437e-01 -9.75194871e-01 -2.74500579e-01 -7.13916942e-02 2.47626558e-01 -4.87535506e-01 3.12506527e-01 -4.77599561e-01 1.19683124e-01 -4.70018595e-01 5.20254612e-01 3.76395136e-01 8.29314440e-02 3.25108320e-01 1.92264080e-01 -5.23158073e-01 1.23118472e+00 -6.16456866e-01 2.05937886e+00 -6.05015755e-01 5.17297745e-01 -2.68621117e-01 -7.36965418e-01 9.38560724e-01 2.91746348e-01 -2.34661669e-01 -7.26874650e-01 -9.99888964e-03 3.49235386e-01 1.68228567e-01 -3.02200705e-01 6.85033202e-01 1.61177754e-01 -5.07505357e-01 3.20604861e-01 4.92487997e-01 2.94610590e-01 2.23940328e-01 2.28367120e-01 2.03635120e+00 5.46578050e-01 2.50584912e-02 -2.46384531e-01 3.56399149e-01 -2.46483371e-01 8.88153255e-01 1.02130675e+00 -3.02177966e-01 3.01115513e-01 8.24527681e-01 -7.77699292e-01 -1.05082285e+00 -1.19692445e+00 7.78326467e-02 1.98092079e+00 -5.72679877e-01 -8.19301963e-01 -7.46935248e-01 -7.41787612e-01 -4.94492412e-01 8.79382730e-01 -8.42563868e-01 1.04272269e-01 -1.29738092e+00 -5.87464690e-01 9.11626816e-01 9.60880876e-01 4.91362244e-01 -1.58309209e+00 -6.79256558e-01 7.10854948e-01 -1.07360139e-01 -1.10780895e+00 -3.31261009e-01 9.38598275e-01 -1.04319811e+00 -5.76414287e-01 -2.32802376e-01 -1.03497756e+00 1.15863122e-01 -5.52846491e-01 1.34997678e+00 -1.91499308e-01 1.51856821e-02 -1.20180771e-02 -6.96504653e-01 -4.13773477e-01 -7.33446836e-01 7.87445128e-01 -2.55200595e-01 -7.77177393e-01 4.17945594e-01 -4.29748684e-01 -1.63149670e-01 -3.89055550e-01 -3.40734363e-01 4.17241342e-02 6.44572020e-01 8.32054496e-01 2.94592291e-01 -7.54919350e-01 8.21610838e-02 -1.35872185e+00 6.45142078e-01 -2.47378439e-01 -5.00833213e-01 1.94736108e-01 -2.96420842e-01 6.36271060e-01 8.84887040e-01 -6.58387691e-02 -1.10751843e+00 9.42181423e-02 -5.04737914e-01 1.62913904e-01 4.77633812e-02 3.00559878e-01 -1.07248873e-01 5.73109627e-01 4.27270263e-01 2.50253435e-02 -6.12813294e-01 -9.45649326e-01 5.29270828e-01 4.68224317e-01 7.28560090e-01 -8.56084526e-01 3.23016971e-01 -1.68907657e-01 -3.83920133e-01 -4.08122540e-01 -1.02056539e+00 -2.13166140e-02 -6.51211441e-01 5.47259927e-01 1.03830361e+00 -7.93098986e-01 -7.32718527e-01 5.90373218e-01 -1.65569150e+00 -6.64070368e-01 -1.86392099e-01 3.07704419e-01 -4.20301527e-01 1.74236819e-01 -1.49913979e+00 -7.46657133e-01 -7.93845892e-01 -1.24381661e+00 9.15449977e-01 2.79505789e-01 -1.10116452e-01 -9.44736958e-01 4.83708829e-01 -1.04722455e-02 4.31057096e-01 7.62709081e-02 1.23117709e+00 -8.97036135e-01 -4.52829421e-01 7.23211542e-02 -8.41748342e-02 4.37725395e-01 -1.79711193e-01 1.60262644e-01 -1.17931402e+00 -2.15610310e-01 -3.99374336e-01 -3.57592672e-01 1.52356851e+00 6.06001675e-01 9.65149820e-01 -8.92425850e-02 -3.88795942e-01 6.86079443e-01 1.06320810e+00 1.57781124e-01 4.15413290e-01 6.12107217e-01 6.55340195e-01 2.86160380e-01 2.10300773e-01 -2.89765801e-02 8.29590976e-01 9.82117876e-02 3.43833625e-01 2.87883461e-01 1.50318462e-02 -6.13810956e-01 1.20127833e+00 1.46259904e+00 -5.75041920e-02 -3.48113567e-01 -1.09584153e+00 5.88739395e-01 -1.76438975e+00 -5.10744154e-01 -3.39251012e-01 1.94762123e+00 7.94542134e-01 5.88596404e-01 2.46415031e-04 -1.72170192e-01 6.10983610e-01 4.45825875e-01 -4.29243863e-01 -1.56340218e+00 -1.91819407e-02 9.62828815e-01 7.28371978e-01 6.89721048e-01 -1.28150713e+00 1.94791722e+00 6.14887619e+00 1.71429783e-01 -8.92385900e-01 5.08980751e-01 4.28601861e-01 2.28228524e-01 -3.69721055e-01 4.13720548e-01 -1.22544360e+00 2.68685043e-01 1.79717505e+00 2.05944926e-01 1.60543919e-01 7.87074089e-01 -3.26137930e-01 -4.62322533e-02 -1.31524897e+00 3.12836289e-01 -3.79888028e-01 -1.33488178e+00 -3.34045589e-01 -1.82959393e-01 1.05968907e-01 9.13243651e-01 3.64154913e-02 1.14862728e+00 1.36709356e+00 -1.24489713e+00 7.64872909e-01 3.44257385e-01 7.42978334e-01 -5.34442961e-01 7.79019296e-01 1.43978104e-01 -1.39937103e+00 -1.15482748e-01 -5.13950527e-01 -2.32493877e-01 2.76660502e-01 3.79494309e-01 -7.77787387e-01 3.68078023e-01 9.26312447e-01 5.87233722e-01 -6.86038733e-01 4.60359186e-01 -4.63846266e-01 1.21547544e+00 -4.00599033e-01 -2.13625923e-01 5.68951666e-01 1.26003310e-01 3.97926569e-01 1.76480770e+00 7.70078972e-03 -8.16003010e-02 1.04445577e-01 5.64784825e-01 -2.90566564e-01 -8.80415663e-02 -5.65165877e-01 -2.50218123e-01 5.19970596e-01 1.09926069e+00 -5.19716918e-01 -4.06408399e-01 -1.71357110e-01 9.93750215e-01 8.54518771e-01 8.66965428e-02 -5.08009911e-01 -4.10688698e-01 5.15973508e-01 -3.74505669e-01 3.86895061e-01 -4.61348057e-01 3.96576151e-02 -1.07918024e+00 -5.13505712e-02 -5.97433269e-01 5.85174799e-01 -4.67983216e-01 -9.88007665e-01 1.11402762e+00 -2.58145273e-01 -3.28606069e-01 -5.07634401e-01 -9.40158188e-01 -7.45900929e-01 9.44438696e-01 -1.55558968e+00 -1.30813634e+00 3.30167264e-01 -3.09948879e-03 7.46097744e-01 -2.16000468e-01 1.33457136e+00 -2.97408029e-02 -7.92248130e-01 6.75957382e-01 7.28883967e-02 6.11199081e-01 4.54859078e-01 -1.73728204e+00 1.53224313e+00 9.65309203e-01 1.65862128e-01 8.22506368e-01 5.62045038e-01 -5.29168904e-01 -1.43117476e+00 -1.14091730e+00 1.35538626e+00 -8.11643004e-01 8.57392967e-01 -7.75300026e-01 -8.56499553e-01 1.62336552e+00 3.95638645e-01 2.29944810e-01 4.79523689e-01 5.21039665e-01 -5.12184501e-01 2.10629910e-01 -6.81422472e-01 4.06823725e-01 1.32968199e+00 -5.48817933e-01 -8.71109188e-01 9.50118899e-02 1.29266918e+00 -8.67379546e-01 -8.71231556e-01 3.68170172e-01 5.25527358e-01 -9.42942083e-01 6.10459685e-01 -1.01543045e+00 4.51706827e-01 2.99522340e-01 -3.19728404e-01 -1.36463094e+00 -4.72717047e-01 -6.27190530e-01 -2.30549723e-02 1.39807737e+00 8.62718105e-01 -7.46707797e-01 7.37546206e-01 2.83869833e-01 -5.75151324e-01 -6.57654703e-01 -1.17768204e+00 -7.17928469e-01 7.48613894e-01 -7.93791652e-01 4.81008351e-01 3.72782946e-01 -2.32769519e-01 5.58852494e-01 -7.78499991e-02 1.14221431e-01 2.73599386e-01 -2.71418989e-01 5.06032944e-01 -1.01519287e+00 -3.47470939e-01 -2.79421926e-01 -1.36210963e-01 -1.08417022e+00 5.62000751e-01 -1.20316184e+00 1.07561812e-01 -1.53438318e+00 1.44820347e-01 -6.11005843e-01 -6.66683078e-01 9.78404880e-01 -1.89173326e-01 -4.66977865e-01 3.06364238e-01 -2.85270959e-01 -7.57383168e-01 3.06780845e-01 6.69965982e-01 -2.92313993e-02 -1.48757145e-01 -9.62017924e-02 -5.45056880e-01 8.37039590e-01 7.64150798e-01 -7.91727781e-01 1.65214777e-01 -1.03163302e+00 4.27589804e-01 2.21580952e-01 -1.36707708e-01 -8.23289633e-01 1.12759978e-01 3.04021239e-01 1.92606241e-01 -6.27121150e-01 2.17113703e-01 4.30403128e-02 -3.50995511e-01 5.79222500e-01 -5.26134253e-01 6.99903488e-01 4.57681060e-01 2.22879112e-01 1.62433103e-01 -4.43669021e-01 4.33417588e-01 -6.28719389e-01 -6.57498598e-01 9.16062146e-02 -5.86038172e-01 3.01807672e-01 1.10392772e-01 1.56555548e-01 -5.20540118e-01 5.94306700e-02 -8.00080240e-01 2.23294169e-01 -6.83101872e-03 6.74256682e-01 2.35899627e-01 -8.75902236e-01 -6.19323552e-01 1.22332148e-01 -1.35188133e-01 1.94728136e-01 6.31275401e-02 3.78658384e-01 -6.27575040e-01 5.38211763e-01 -3.68299112e-02 -4.13520366e-01 -1.07273114e+00 5.15516475e-02 1.89875081e-01 -1.05979335e+00 -8.52487981e-01 1.45825434e+00 -1.97643846e-01 -1.15372312e+00 1.26157388e-01 -1.16800606e+00 -6.97821304e-02 -2.76212513e-01 4.25331682e-01 1.50274172e-01 1.69048786e-01 -2.19166130e-01 -5.73982060e-01 3.31973612e-01 -5.48433661e-01 -2.39772737e-01 1.57517111e+00 4.01608378e-01 -1.38647914e-01 8.68401706e-01 1.16582310e+00 -6.79306611e-02 -9.91852462e-01 -4.21953112e-01 7.31112778e-01 3.89370143e-01 -8.55880007e-02 -9.21944320e-01 -6.96497142e-01 1.19920266e+00 5.60088873e-01 -1.17736742e-01 5.22311985e-01 -9.72289070e-02 1.37243950e+00 6.01401389e-01 3.42905223e-01 -1.08164656e+00 -1.85968310e-01 1.32927155e+00 3.74779016e-01 -1.01050436e+00 -5.95455229e-01 7.24268928e-02 -3.27802896e-01 1.25551474e+00 6.03639364e-01 -3.24979395e-01 3.97276133e-01 5.92375100e-01 2.12616306e-02 1.33181944e-01 -1.30211258e+00 -3.30179751e-01 -3.41690809e-01 5.51619053e-01 9.51857626e-01 4.19326097e-01 -2.58219361e-01 9.97635782e-01 -6.20239854e-01 -1.54462457e-01 5.53474367e-01 1.06063497e+00 -5.73281586e-01 -1.49864972e+00 7.85636380e-02 3.99760872e-01 -6.86157405e-01 -5.68270266e-01 -8.01254138e-02 7.75473416e-01 5.76855503e-02 7.56426811e-01 1.89128801e-01 -5.30708313e-01 3.52036655e-01 4.70356643e-01 6.01766765e-01 -1.04815018e+00 -9.89367783e-01 -3.53609055e-01 4.66687202e-01 -7.25336790e-01 2.71733552e-01 -7.57165432e-01 -1.55400443e+00 -1.66254878e-01 -1.03361242e-01 9.86374766e-02 7.90264308e-01 9.75341141e-01 2.81016290e-01 7.80621946e-01 3.19075286e-01 -5.61751485e-01 -6.92775130e-01 -1.28994358e+00 -2.46598288e-01 -8.15466866e-02 2.50123680e-01 -2.27221027e-01 -1.23189598e-01 -1.87647596e-01]
[10.377205848693848, 9.549481391906738]
4b54bc51-e941-432d-9407-d02be092acb7
a-survey-of-deep-graph-clustering-taxonomy
2211.12875
null
https://arxiv.org/abs/2211.12875v2
https://arxiv.org/pdf/2211.12875v2.pdf
A Survey of Deep Graph Clustering: Taxonomy, Challenge, and Application
Graph clustering, which aims to divide the nodes in the graph into several distinct clusters, is a fundamental and challenging task. In recent years, deep graph clustering methods have been increasingly proposed and achieved promising performance. However, the corresponding survey paper is scarce and it is imminent to make a summary in this field. From this motivation, this paper makes the first comprehensive survey of deep graph clustering. Firstly, the detailed definition of deep graph clustering and the important baseline methods are introduced. Besides, the taxonomy of deep graph clustering methods is proposed based on four different criteria including graph type, network architecture, learning paradigm, and clustering method. In addition, through the careful analysis of the existing works, the challenges and opportunities from five perspectives are summarized. At last, the applications of deep graph clustering in four domains are presented. It is worth mentioning that a collection of state-of-the-art deep graph clustering methods including papers, codes, and datasets is available on GitHub. We hope this work will serve as a quick guide and help researchers to overcome challenges in this vibrant field.
['Xinwang Liu', 'Stan Z. Li', 'Wenxuan Tu', 'Ke Liang', 'Xihong Yang', 'Xifeng Guo', 'Siwei Wang', 'Sihang Zhou', 'Jun Xia', 'Yue Liu']
2022-11-23
null
null
null
null
['graph-clustering']
['graphs']
[-5.13954878e-01 3.19843516e-02 -2.23467097e-01 -1.20669603e-01 -2.06011385e-01 -5.06438076e-01 2.07005903e-01 3.43446344e-01 1.57944039e-02 1.68337569e-01 -1.07585818e-01 -1.37378454e-01 -3.18082929e-01 -8.02922070e-01 -1.28170177e-01 -1.03056204e+00 -4.66575235e-01 5.81089914e-01 -1.27635943e-02 8.88948701e-03 1.52013242e-01 3.83734018e-01 -1.33359456e+00 -5.43250889e-02 8.72316539e-01 5.85365355e-01 1.45633191e-01 1.71433970e-01 -1.54563278e-01 7.62737930e-01 -6.31264865e-01 -6.41585946e-01 -2.86853351e-02 -6.03821516e-01 -8.27906907e-01 3.97332937e-01 2.48091742e-01 2.03767810e-02 -8.24907720e-01 1.30555856e+00 7.51658618e-01 2.63296753e-01 6.91441596e-01 -1.61122453e+00 -8.97181988e-01 9.09250915e-01 -7.63889611e-01 1.06488839e-01 3.03752609e-02 -7.37298504e-02 1.21268880e+00 -5.87589920e-01 3.64045650e-01 9.51049685e-01 6.38636351e-01 3.46577227e-01 -6.71474099e-01 -6.04233861e-01 5.59087873e-01 5.38153589e-01 -1.77288949e+00 1.73540320e-02 1.12369514e+00 -4.16477442e-01 8.44054043e-01 -1.02901002e-02 8.34806859e-01 8.02076221e-01 -4.42146920e-02 8.79226327e-01 5.97941756e-01 -1.21700875e-01 2.86410272e-01 -4.57090765e-01 3.78103316e-01 6.74210906e-01 4.16642427e-01 -3.71272117e-01 -7.94330463e-02 8.49241465e-02 4.68444467e-01 1.19267866e-01 -1.19264401e-01 -7.50129104e-01 -1.02869296e+00 8.56891870e-01 9.13143933e-01 6.40856743e-01 -1.47550568e-01 1.25052840e-01 5.56247473e-01 6.57533631e-02 3.53032082e-01 1.47748813e-01 1.32585153e-01 1.28969744e-01 -9.60233688e-01 2.00548135e-02 6.89623356e-01 1.22887909e+00 7.01857746e-01 1.60633191e-01 1.53947875e-01 8.74871910e-01 3.95531386e-01 4.41694558e-02 1.38079807e-01 -7.89160728e-01 2.78735101e-01 9.49185848e-01 -7.62982070e-01 -1.62328577e+00 -5.76208949e-01 -5.85404098e-01 -1.54905665e+00 -3.17372292e-01 3.60776149e-02 -1.57424271e-01 -8.27766180e-01 1.26363397e+00 2.65589565e-01 2.21388102e-01 -3.74139875e-01 9.20451105e-01 1.72265863e+00 4.53107089e-01 -1.10786945e-01 -1.68198019e-01 1.24296618e+00 -1.16261935e+00 -9.96474445e-01 4.88988534e-02 5.30423164e-01 -5.87560952e-01 7.66162395e-01 2.14170411e-01 -9.82700586e-01 -5.22938907e-01 -8.46667230e-01 -7.24497512e-02 -6.21186852e-01 6.14815727e-02 1.08115637e+00 6.09060705e-01 -1.58939147e+00 5.36110938e-01 -1.06127703e+00 -9.41492319e-01 4.86499608e-01 4.93161708e-01 -1.05328999e-01 -1.80295646e-01 -1.01350772e+00 2.03698352e-01 4.24839526e-01 1.59250855e-01 -7.42979586e-01 -1.02597572e-01 -8.75913024e-01 1.56203091e-01 4.58409965e-01 -6.76485240e-01 7.04252601e-01 -5.19677699e-01 -1.14220583e+00 1.06166375e+00 8.21983144e-02 -2.39093587e-01 2.16607258e-01 3.72281224e-01 -4.63299274e-01 3.67034137e-01 5.64506948e-02 4.78066981e-01 4.71845448e-01 -1.28223157e+00 -3.53973657e-01 -5.47650456e-01 -8.00566282e-03 3.67078871e-01 -5.77697814e-01 5.38376579e-03 -1.22026360e+00 -6.40580118e-01 1.69124901e-02 -9.72729981e-01 -4.97860968e-01 -6.09558880e-01 -8.60091925e-01 -6.37578547e-01 7.21279383e-01 -2.27771699e-01 1.88107371e+00 -2.19221902e+00 3.29702586e-01 2.32109591e-01 1.07032967e+00 8.53661150e-02 -1.01490365e-02 9.31529582e-01 -1.46865621e-01 2.64297098e-01 -4.73159373e-01 -4.12706316e-01 1.66440949e-01 -4.04430702e-02 1.31775692e-01 8.02633464e-01 -3.06588411e-01 1.14801264e+00 -8.90108705e-01 -5.21563947e-01 5.48984349e-01 5.35252690e-01 -2.12786585e-01 1.08143762e-01 2.60261536e-01 4.19005990e-01 -2.86748260e-01 6.97714150e-01 8.46732616e-01 -7.52627671e-01 6.72026634e-01 -1.51301533e-01 1.21353105e-01 2.41422262e-02 -1.05015290e+00 1.64409840e+00 2.92971879e-01 7.08645880e-01 3.53653491e-01 -1.56755614e+00 1.00798929e+00 1.01298891e-01 8.85797381e-01 -4.20891464e-01 3.62605959e-01 -2.99206339e-02 3.63439113e-01 -5.50462939e-02 3.08608174e-01 3.56821001e-01 -2.10253552e-01 3.49624664e-01 -3.04713361e-02 2.14423053e-02 6.60245657e-01 7.01052666e-01 1.22819340e+00 -4.43777263e-01 2.67150521e-01 -4.27113712e-01 5.01225591e-01 -7.34393448e-02 3.06190729e-01 5.42797506e-01 -5.75237572e-01 6.66759670e-01 5.99789739e-01 -5.39365292e-01 -5.79478800e-01 -9.74087119e-01 1.07762814e-01 7.82402337e-01 4.27335054e-01 -9.05969083e-01 -8.67016673e-01 -7.11242616e-01 -8.56093317e-02 4.32263277e-02 -7.41724730e-01 -1.80753157e-01 -5.95764697e-01 -1.10053158e+00 2.87643671e-01 5.33201516e-01 7.17768133e-01 -1.03221679e+00 3.54855269e-01 -1.28293172e-01 -2.39797890e-01 -1.14255846e+00 -2.56049305e-01 -8.35307389e-02 -9.26363170e-01 -1.36464417e+00 -7.31239498e-01 -1.56009293e+00 8.10797453e-01 8.68173122e-01 1.47790337e+00 8.31103742e-01 -1.76549107e-01 3.65410030e-01 -6.44925594e-01 2.67070115e-01 -1.83491707e-02 4.40383255e-01 -9.66783464e-02 -2.29450062e-01 7.32510090e-01 -6.86176062e-01 -8.01038623e-01 2.29713932e-01 -6.71392560e-01 -1.83567241e-01 4.30977225e-01 3.55614722e-01 6.80702984e-01 6.27554059e-01 3.46154213e-01 -9.33578134e-01 7.70943642e-01 -5.40078878e-01 -4.54056114e-01 -5.13215512e-02 -6.67014897e-01 -4.85152870e-01 8.15505147e-01 2.12327078e-01 -3.57366025e-01 -8.37142617e-02 -1.47064611e-01 -3.25409174e-01 -2.06640720e-01 7.24845588e-01 -3.62224638e-01 -1.49869574e-02 2.81081915e-01 5.82783185e-02 -3.08924615e-02 -5.26285112e-01 5.64934134e-01 4.24658656e-01 3.64421785e-01 -2.20748395e-01 7.35968471e-01 6.48236692e-01 1.01226335e-02 -9.53230679e-01 -5.61624765e-01 -7.91660190e-01 -1.00127482e+00 -4.81757760e-01 9.16079044e-01 -8.95672500e-01 -7.79709339e-01 6.48222804e-01 -8.98757160e-01 -5.96402228e-01 2.67621547e-01 1.95704356e-01 -2.46924832e-01 1.05675685e+00 -9.65312958e-01 -2.55454838e-01 -2.97476292e-01 -1.33770812e+00 8.42404068e-01 1.33375376e-01 -1.84433647e-02 -1.57619023e+00 -1.42271489e-01 3.28015387e-01 3.60938497e-02 5.38948298e-01 7.67446876e-01 -6.17882550e-01 -6.46902561e-01 -1.68545723e-01 -3.31159621e-01 4.91942763e-02 3.58566493e-01 3.96998137e-01 -5.22953093e-01 -6.88250601e-01 -3.18239331e-01 -2.61514261e-03 1.06880772e+00 7.55377233e-01 1.49024415e+00 1.54091800e-02 -7.74134040e-01 8.59511852e-01 1.44052231e+00 1.95721909e-01 4.66901481e-01 1.55446991e-01 1.42559171e+00 5.30726552e-01 2.40472034e-01 1.78915590e-01 8.78417730e-01 5.48059464e-01 5.88900626e-01 -2.23319918e-01 -3.40105057e-01 1.51959270e-01 2.65675019e-02 1.54314029e+00 -1.09104209e-01 -6.73084974e-01 -1.31352794e+00 5.32538414e-01 -1.99989188e+00 -8.89404178e-01 -7.12637424e-01 1.81199908e+00 1.59637243e-01 -6.87078834e-02 5.37250757e-01 2.35109448e-01 1.36045778e+00 3.84782672e-01 -5.45752704e-01 1.67289704e-01 -1.26793459e-01 -1.23229660e-01 2.31747940e-01 4.23173994e-01 -1.45972538e+00 1.20194995e+00 6.40880585e+00 7.62939334e-01 -8.30121338e-01 -1.70448229e-01 5.46896696e-01 2.76715606e-01 -8.77248570e-02 -9.58458558e-02 -3.14971387e-01 5.50477028e-01 6.54357016e-01 -3.42521340e-01 5.93479574e-01 8.20423126e-01 9.50073008e-04 2.58679599e-01 -9.69919264e-01 1.32079041e+00 1.81400239e-01 -1.31134951e+00 6.00791313e-02 3.38614553e-01 7.45970011e-01 3.29843253e-01 1.66559920e-01 4.43812639e-01 6.77038908e-01 -1.08098948e+00 -4.91755158e-02 5.60039133e-02 5.92507780e-01 -1.11015618e+00 7.30723202e-01 1.01287551e-01 -1.89606476e+00 3.19444500e-02 -4.51391011e-01 -2.32477546e-01 -8.25896626e-04 8.82861972e-01 -3.03852111e-01 1.03587270e+00 9.87331033e-01 1.48930430e+00 -6.97878599e-01 1.11326039e+00 -2.19153032e-01 6.56855762e-01 1.38906345e-01 -5.38166985e-02 3.14099729e-01 -6.27471507e-01 2.46861532e-01 1.25799644e+00 1.07505709e-01 -2.85771303e-02 5.93903005e-01 7.18509495e-01 -3.03936809e-01 1.78551316e-01 -8.02851021e-01 -3.87251824e-01 6.50289834e-01 1.58865654e+00 -1.47609377e+00 -3.68516624e-01 -4.85079408e-01 8.44383359e-01 4.77144420e-01 4.50148910e-01 -8.28849733e-01 -7.70891190e-01 6.12592161e-01 -1.88404784e-01 1.21330641e-01 -6.38059735e-01 -3.65149349e-01 -1.07223237e+00 -3.04093480e-01 -8.53254616e-01 7.84510672e-01 -3.79877299e-01 -1.49634683e+00 7.66921878e-01 -4.35721762e-02 -1.12229681e+00 2.25654975e-01 -4.32738692e-01 -7.24633217e-01 3.74001354e-01 -1.03133857e+00 -8.51288557e-01 -7.68388867e-01 6.26060188e-01 2.72850275e-01 -3.78406912e-01 4.58761692e-01 6.04617834e-01 -9.39902604e-01 5.92003167e-01 3.21498007e-01 8.06019545e-01 4.57425743e-01 -1.49136138e+00 8.47457409e-01 9.60839093e-01 7.69453645e-02 7.65759051e-01 2.91936785e-01 -7.29848802e-01 -1.48142076e+00 -1.33719361e+00 5.23026645e-01 -4.31994945e-01 7.52763450e-01 -6.93048775e-01 -9.57187116e-01 5.89969873e-01 6.75717473e-01 -1.63018793e-01 8.27353477e-01 1.77037209e-01 3.83763341e-03 -1.82481915e-01 -7.20767081e-01 7.69880116e-01 1.25133276e+00 -3.00505966e-01 2.25635767e-02 6.80494666e-01 7.71210194e-01 -3.01684797e-01 -9.11597669e-01 2.48217747e-01 -8.08452815e-02 -1.26898360e+00 8.42679083e-01 -1.87070489e-01 1.39474124e-01 -4.63659704e-01 2.21712455e-01 -1.27735853e+00 -9.29729581e-01 -7.48809576e-01 -1.93874270e-01 1.25878298e+00 -1.80094019e-01 -4.52496886e-01 1.05949295e+00 4.95866276e-02 -4.74574745e-01 -8.12856257e-01 -5.47165751e-01 -7.46717036e-01 3.12749833e-01 -2.09138319e-01 6.16250813e-01 1.26055586e+00 1.43112987e-01 7.68534362e-01 -8.09615180e-02 1.37128174e-01 9.28535700e-01 2.69585013e-01 1.01987350e+00 -1.33655536e+00 3.52107674e-01 -8.74688804e-01 -5.90129673e-01 -1.09672332e+00 4.00703698e-01 -1.40525997e+00 -3.37713450e-01 -2.51874757e+00 4.15063143e-01 -1.72593921e-01 -2.48689383e-01 2.94140190e-01 -2.51535267e-01 3.35950464e-01 1.28815323e-01 4.12517101e-01 -1.12290311e+00 5.89263082e-01 1.34615707e+00 -3.46860141e-01 -6.26914576e-02 -8.68419856e-02 -1.00012040e+00 5.57777107e-01 1.12634349e+00 -1.45062432e-01 -6.62562072e-01 -5.06726980e-01 1.78082660e-01 -2.84201384e-01 7.60245845e-02 -1.04823601e+00 5.02807379e-01 2.12675542e-01 -4.48461249e-02 -1.03860652e+00 1.86931007e-02 -7.86189675e-01 -2.54428480e-02 4.77524817e-01 9.59099364e-03 4.47176963e-01 -6.27133250e-02 7.89322197e-01 -5.31548202e-01 3.91560137e-01 8.19801807e-01 -2.69719511e-01 -9.71128821e-01 8.76891971e-01 -3.90067667e-01 2.62456656e-01 1.18019307e+00 -2.59809762e-01 -2.36767307e-01 -5.73725641e-01 -7.30029881e-01 6.42156184e-01 6.07494771e-01 3.71640652e-01 6.53467894e-01 -1.51947439e+00 -6.11486971e-01 -2.89490938e-01 4.35075983e-02 2.77401268e-01 5.73636949e-01 9.10839498e-01 -6.57133818e-01 3.50150466e-01 -7.59439450e-03 -7.56552041e-01 -1.08997548e+00 1.11557031e+00 3.26112986e-01 -7.48730525e-02 -7.62334108e-01 8.29892397e-01 5.16850293e-01 -5.18887997e-01 5.92012465e-01 -9.73780453e-02 -4.65175450e-01 7.23791867e-03 1.99972391e-01 4.58984792e-01 1.54204726e-01 -7.93938577e-01 -8.24907303e-01 6.59963667e-01 -3.63534987e-02 6.51443958e-01 1.19470990e+00 -4.72218066e-01 -6.52036190e-01 2.79373437e-01 1.30690372e+00 -2.11981162e-01 -7.70884216e-01 9.02247988e-03 -3.66602950e-02 -5.04117832e-02 -9.01915953e-02 -3.37157935e-01 -1.73957109e+00 1.13374424e+00 1.74075499e-01 7.68041193e-01 1.25846791e+00 3.42867136e-01 8.03180337e-01 1.52438432e-01 2.48038068e-01 -1.11802447e+00 2.89256930e-01 6.37656808e-01 6.21839881e-01 -1.29246259e+00 1.63854986e-01 -6.12919152e-01 -6.06737554e-01 1.00254190e+00 7.33011067e-01 -3.31169516e-01 9.93732929e-01 1.12366430e-01 1.25637844e-01 -7.83721209e-01 -3.23204041e-01 -5.43209434e-01 2.03162596e-01 1.03583348e+00 8.60979080e-01 2.10762680e-01 -2.36416206e-01 3.75690192e-01 -3.00845504e-01 -7.26923823e-01 4.22976643e-01 3.42756063e-01 -2.45394811e-01 -1.11585963e+00 7.40389004e-02 4.30410922e-01 -2.68533200e-01 -1.44924060e-01 -9.85221148e-01 1.00240779e+00 -1.91579998e-01 1.36186767e+00 1.51683211e-01 -6.37316763e-01 4.70736884e-02 -6.52654231e-01 1.89158767e-01 -7.07919002e-01 -3.99619848e-01 5.46887890e-02 -2.50381976e-01 -5.27968228e-01 -5.95142543e-01 -3.11761379e-01 -1.39620066e+00 -8.62478614e-01 -7.42629394e-02 4.88937587e-01 2.10445300e-01 5.77458978e-01 5.67848623e-01 6.87563002e-01 5.28017581e-01 -8.54387403e-01 2.57164270e-01 -6.92882180e-01 -9.27646577e-01 3.55240881e-01 5.66427670e-02 -7.50068069e-01 -4.47094887e-01 -2.50210702e-01]
[7.293485641479492, 5.956362247467041]
fd404f10-05fe-45f9-bbcf-cb79865ec6ae
live-video-comment-generation-based-on
1808.04091
null
http://arxiv.org/abs/1808.04091v1
http://arxiv.org/pdf/1808.04091v1.pdf
Live Video Comment Generation Based on Surrounding Frames and Live Comments
In this paper, we propose the task of live comment generation. Live comments are a new form of comments on videos, which can be regarded as a mixture of comments and chats. A high-quality live comment should be not only relevant to the video, but also interactive with other users. In this work, we first construct a new dataset for live comment generation. Then, we propose a novel end-to-end model to generate the human-like live comments by referring to the video and the other users' comments. Finally, we evaluate our model on the constructed dataset. Experimental results show that our method can significantly outperform the baselines.
['Damai Dai']
2018-08-13
null
null
null
null
['comment-generation']
['natural-language-processing']
[ 1.74239695e-01 7.46287405e-02 1.12168908e-01 -5.04412055e-01 -1.01373613e+00 -4.85516816e-01 8.23276281e-01 -2.07955658e-01 -4.59228419e-02 7.75570631e-01 1.02555525e+00 -2.99790762e-02 8.49548340e-01 -2.27284744e-01 -3.39223981e-01 -5.66023827e-01 4.05703366e-01 1.48837734e-02 3.52381527e-01 -6.82929456e-02 3.19975257e-01 -3.94520104e-01 -1.23049998e+00 8.63249421e-01 6.50810778e-01 9.08510983e-01 3.92110467e-01 9.59908187e-01 7.12301349e-03 1.33323026e+00 -8.59705925e-01 -5.81081033e-01 -4.50621396e-01 -9.51197028e-01 -7.17951655e-01 4.58364487e-01 3.19276243e-01 -7.68590391e-01 -2.48097166e-01 8.97595823e-01 6.43288612e-01 2.71265388e-01 6.33473158e-01 -1.35128331e+00 -8.62157166e-01 8.36901665e-01 -2.68926799e-01 8.43376666e-03 9.54152703e-01 1.04097441e-01 1.03654921e+00 -1.37211728e+00 8.53970706e-01 1.20064592e+00 6.56130239e-02 8.74360383e-01 -5.15608311e-01 -6.63575530e-01 5.16894460e-01 1.19500563e-01 -9.99063313e-01 -7.25111783e-01 7.49674380e-01 -5.36138833e-01 3.38377804e-01 4.74143118e-01 5.77079117e-01 1.39296281e+00 -1.92913204e-01 1.29174042e+00 2.73251861e-01 -1.54530704e-01 -1.09380260e-01 9.98004749e-02 -3.36938113e-01 2.26348177e-01 -4.62077171e-01 -4.11769301e-01 -6.57932699e-01 -2.52049357e-01 4.76794034e-01 1.08874336e-01 -6.08658433e-01 2.20642835e-01 -1.64129758e+00 8.69647026e-01 1.80424005e-01 3.60610969e-02 -1.89823717e-01 -3.84364533e-03 6.66305304e-01 2.61612743e-01 9.86110330e-01 7.66476095e-02 -4.80773393e-03 -5.18189073e-01 -7.02906370e-01 4.92771208e-01 8.84886205e-01 1.29638922e+00 4.24959719e-01 -2.76297778e-01 -7.66834378e-01 7.85175204e-01 6.04681075e-01 5.30620098e-01 3.30681771e-01 -7.86860645e-01 9.17339087e-01 4.35507178e-01 6.15571976e-01 -1.09541893e+00 7.84526914e-02 6.62000990e-03 -9.14615989e-01 -5.01797736e-01 -3.85633744e-02 -6.62669420e-01 -2.53908575e-01 1.12352240e+00 1.10763557e-01 4.80393708e-01 -2.50750054e-02 9.10977781e-01 1.44840169e+00 1.32297099e+00 -2.48066470e-01 -7.04003036e-01 7.13083148e-01 -1.72069132e+00 -1.18348479e+00 -1.11457799e-02 6.03747010e-01 -1.04837716e+00 1.11680937e+00 1.81587249e-01 -1.30405664e+00 -6.22034848e-01 -6.09088421e-01 -8.54830816e-02 5.46426415e-01 4.75394547e-01 1.72789201e-01 -6.12708889e-02 -1.06985343e+00 6.50797784e-02 -4.29721594e-01 -1.78781599e-01 1.33407071e-01 -1.91820502e-01 -1.51738748e-01 1.39318639e-02 -1.06784356e+00 1.83868036e-01 9.98226926e-02 1.93337157e-01 -7.50768960e-01 -5.70447981e-01 -8.74935508e-01 -3.66895199e-02 3.44253957e-01 -4.88711804e-01 1.94229496e+00 -1.49927425e+00 -1.63185799e+00 7.29989409e-01 -7.61763871e-01 -2.61982203e-01 6.35718703e-01 -3.31905186e-01 -5.34788251e-01 3.65713328e-01 6.79169372e-02 7.10232794e-01 8.14448178e-01 -1.34576440e+00 -1.13776791e+00 4.09132689e-01 4.64583069e-01 3.22959840e-01 -3.19786191e-01 5.13187468e-01 -6.71143889e-01 -6.77253246e-01 -4.34206396e-01 -1.06215584e+00 -4.22580168e-02 -7.02986792e-02 -5.38660228e-01 -3.27115417e-01 1.10932171e+00 -6.40644193e-01 1.54860497e+00 -2.40382767e+00 6.15327060e-02 -2.80056268e-01 2.98746854e-01 2.94894755e-01 -2.12687328e-01 8.12297940e-01 6.38681650e-02 3.15708518e-01 -3.54638807e-02 -6.79863155e-01 -9.72877815e-02 -2.66154170e-01 -6.41641438e-01 1.85221329e-01 7.72240907e-02 1.00490344e+00 -1.13839841e+00 -4.43470687e-01 -1.79821715e-01 3.80653679e-01 -5.64000547e-01 9.12254930e-01 -4.94741559e-01 8.25691879e-01 -4.68332320e-01 1.14503920e-01 4.74030823e-01 -5.38755953e-01 -5.31138293e-02 9.81069449e-03 -8.73150304e-02 4.48157668e-01 -6.28753960e-01 1.79843068e+00 -5.01297355e-01 8.18669200e-01 -1.05507016e-01 -4.78209376e-01 9.55925941e-01 8.52463722e-01 1.73354894e-01 -3.19170803e-01 1.62680730e-01 5.85448630e-02 -2.38125637e-01 -8.70688498e-01 7.35208511e-01 1.14267163e-01 -9.53947157e-02 1.03656685e+00 -2.22448915e-01 3.52543257e-02 4.95811760e-01 6.46250963e-01 6.63498759e-01 -2.23651771e-02 2.27929115e-01 3.11866999e-01 8.20269406e-01 -3.22848231e-01 3.54975432e-01 4.01260436e-01 -7.94665739e-02 1.08501327e+00 7.11530268e-01 -1.53465092e-01 -8.24484468e-01 -7.10354447e-01 5.35055578e-01 1.13745582e+00 2.28709817e-01 -7.13019311e-01 -5.05025506e-01 -1.00912142e+00 -5.71386993e-01 3.61217022e-01 -5.62431455e-01 -2.32415423e-02 -5.04011273e-01 -4.27035131e-02 2.97331847e-02 4.28673029e-01 4.06446904e-01 -1.25887609e+00 -1.80528134e-01 1.56917870e-01 -1.02362144e+00 -1.05708647e+00 -1.34478569e+00 -8.23231757e-01 -5.85468650e-01 -9.90823209e-01 -1.22579312e+00 -1.13183630e+00 8.67978692e-01 7.29200184e-01 1.21701694e+00 7.32797682e-01 6.52245641e-01 1.61395222e-01 -1.24319077e+00 -3.78142625e-01 -5.70488453e-01 1.06504269e-01 -4.57749158e-01 5.61831951e-01 -7.64068291e-02 -3.01387012e-01 -5.82441986e-01 3.86828661e-01 -9.97568548e-01 5.93269527e-01 2.30265275e-01 8.19077551e-01 3.33117157e-01 -4.85840827e-01 7.92618752e-01 -1.10343575e+00 9.58905518e-01 -6.48899794e-01 1.18597530e-01 9.83896405e-02 2.85837233e-01 -4.25926030e-01 6.64882779e-01 -4.63239461e-01 -1.43284678e+00 6.45132959e-02 -4.57883924e-01 -2.12701395e-01 -5.13113327e-02 8.84077370e-01 -2.98514128e-01 4.48721707e-01 1.98410958e-01 6.33191764e-02 -1.64189458e-01 -3.26188117e-01 9.88959074e-02 1.14130247e+00 4.34389144e-01 -2.94877768e-01 7.07793117e-01 4.76351947e-01 -7.79322624e-01 -6.27385139e-01 -1.13811541e+00 -6.86438143e-01 -5.03014922e-01 -6.65490210e-01 7.86019862e-01 -1.08710778e+00 -6.08572960e-01 5.49390793e-01 -1.74503303e+00 -4.63413537e-01 1.66623577e-01 2.62093574e-01 -5.11953473e-01 3.85290951e-01 -6.60516143e-01 -8.26850653e-01 -4.33967292e-01 -1.05267310e+00 1.16780365e+00 5.22275805e-01 -2.06528261e-01 -1.18846750e+00 6.26734048e-02 2.88275003e-01 2.34909102e-01 1.52011529e-01 2.89913714e-02 -6.71053290e-01 -4.12498564e-01 -3.48323882e-01 -4.08763587e-01 3.48203033e-01 1.25618637e-01 4.46623325e-01 -7.33714938e-01 -2.88848042e-01 -2.05140516e-01 -7.34039545e-01 5.98480821e-01 -1.49963975e-01 1.08969486e+00 -8.15027237e-01 -8.27144012e-02 3.09829473e-01 8.85442555e-01 3.18501890e-01 7.71906435e-01 -1.72570020e-01 7.77454495e-01 6.16122901e-01 1.07900572e+00 8.31544816e-01 7.16081798e-01 4.56054509e-01 6.24761760e-01 -4.74558473e-02 2.27321759e-01 -6.61162853e-01 6.28409326e-01 1.60259843e+00 1.23604052e-01 -7.99645901e-01 -6.24647915e-01 7.15412498e-01 -2.29377246e+00 -1.29007185e+00 -5.95744550e-01 1.73288429e+00 6.95719719e-01 9.36358720e-02 2.22984403e-01 -1.93330199e-01 8.35125983e-01 5.38612247e-01 -3.02606344e-01 -1.19522840e-01 1.86930135e-01 -3.36647749e-01 -2.85787523e-01 3.80350411e-01 -9.56036329e-01 6.03537023e-01 6.05542707e+00 4.15747434e-01 -1.20560896e+00 1.57433823e-01 6.73439324e-01 -1.92371637e-01 -5.91027141e-01 -1.18532859e-01 -7.21947730e-01 8.56077492e-01 9.01545286e-01 -8.18484068e-01 -3.25602442e-02 7.37054169e-01 7.99289525e-01 2.73134671e-02 -1.20914447e+00 1.00020528e+00 7.12373793e-01 -1.52229702e+00 3.36627066e-01 -3.36004734e-01 1.05407274e+00 -4.57652420e-01 -1.59287080e-01 2.22147226e-01 1.55971453e-01 -7.85133958e-01 9.81305420e-01 4.30438608e-01 9.53777671e-01 -6.37014806e-01 9.23907757e-01 5.80286801e-01 -1.40079331e+00 1.64941147e-01 -2.01180547e-01 -3.32762897e-01 9.12438512e-01 4.91843373e-01 -5.42796016e-01 4.85030323e-01 3.61912578e-01 1.59544110e+00 -4.11685646e-01 1.20681274e+00 -5.98355293e-01 7.54618645e-01 4.37039167e-01 -4.45535034e-01 1.70526728e-01 2.67607253e-02 2.58986861e-01 1.39311635e+00 4.66920078e-01 3.20927590e-01 4.67292160e-01 4.59606171e-01 -4.12784576e-01 2.00217381e-01 -3.18226486e-01 -2.08564386e-01 5.05050778e-01 1.34397745e+00 -2.06082866e-01 -6.26888990e-01 -6.35877430e-01 1.22143114e+00 1.41773254e-01 5.67376435e-01 -9.58233833e-01 -4.80122626e-01 3.91931027e-01 1.12818792e-01 2.24649850e-02 3.82425115e-02 4.71269846e-01 -1.60615909e+00 2.27945581e-01 -8.60807836e-01 1.01228885e-01 -1.32814813e+00 -1.35368395e+00 6.48849487e-01 -3.69874030e-01 -1.73922586e+00 -3.92784625e-01 -8.56051780e-03 -1.24788821e+00 8.35902512e-01 -1.28326917e+00 -1.15274775e+00 -6.62075818e-01 6.25849009e-01 1.02344191e+00 1.41695857e-01 6.03048563e-01 4.78269398e-01 -2.12895900e-01 4.10292268e-01 -1.07230842e-01 3.85818273e-01 1.11381757e+00 -1.05715466e+00 5.58548987e-01 9.41388905e-01 -9.57769528e-03 3.73671770e-01 5.79609156e-01 -6.24677479e-01 -7.77732968e-01 -1.35571635e+00 1.34194076e+00 -1.95312932e-01 7.67566323e-01 -4.11084086e-01 -7.61319637e-01 8.44415486e-01 6.39220655e-01 -2.76739687e-01 9.42880988e-01 -3.02929491e-01 5.04845195e-02 1.98898003e-01 -4.70636666e-01 7.12863326e-01 1.04138541e+00 -6.90689504e-01 -5.19602001e-01 6.62164390e-01 1.28991199e+00 -7.33643532e-01 -4.21592653e-01 -1.23140842e-01 4.79024470e-01 -1.04544437e+00 3.83176565e-01 -5.69765449e-01 1.16808629e+00 -2.13589713e-01 3.60196978e-01 -1.58507621e+00 -1.55225933e-01 -9.00740743e-01 -7.50377774e-03 1.68504775e+00 5.23655236e-01 -1.40615866e-01 5.90374708e-01 3.30098152e-01 -4.82187331e-01 -4.88299400e-01 -3.18556368e-01 -2.43926361e-01 -3.95629853e-01 -1.98204249e-01 4.75773931e-01 8.60845804e-01 3.99944544e-01 6.56185448e-01 -9.25611317e-01 -3.50163788e-01 -8.85115191e-02 1.89113114e-02 1.21271133e+00 -1.16601884e+00 -2.51727492e-01 -9.35589373e-02 9.19122249e-02 -1.75581765e+00 2.75518835e-01 -6.06142342e-01 6.26657665e-01 -1.87332320e+00 5.70102930e-01 -7.22039789e-02 2.68459737e-01 -1.60116449e-01 -5.56581259e-01 1.56985447e-01 3.83120567e-01 5.19777775e-01 -1.09245455e+00 5.35211742e-01 1.68972635e+00 -3.26409191e-02 -1.20688237e-01 2.62730747e-01 -7.68759966e-01 5.90485215e-01 6.08975589e-01 -2.41744742e-01 -5.00031829e-01 -5.94629467e-01 6.53169572e-01 4.68350738e-01 4.05778959e-02 -6.39199495e-01 1.50581494e-01 -2.21386358e-01 -2.78726012e-01 -8.80324006e-01 1.81018099e-01 -5.58063805e-01 -8.97864103e-02 1.98067576e-02 -9.24363554e-01 1.18141502e-01 -3.83897364e-01 6.42119229e-01 -7.75675774e-01 -1.63469553e-01 5.33787489e-01 -8.88294280e-02 -3.18553209e-01 6.24000490e-01 -5.51267803e-01 2.09032834e-01 1.04764426e+00 2.35384867e-01 -6.12845838e-01 -1.25273776e+00 -7.53170490e-01 5.33930480e-01 5.66005468e-01 7.70437062e-01 7.68591106e-01 -1.58579648e+00 -1.05342674e+00 -2.88978904e-01 3.88707548e-01 2.62822062e-01 3.24694782e-01 7.33945549e-01 -5.26037276e-01 2.27131754e-01 1.16503991e-01 -4.33980674e-01 -1.58055973e+00 5.14173090e-01 -2.86354143e-02 -2.85946310e-01 -4.85864222e-01 1.06970620e+00 4.38233644e-01 -1.67698711e-01 4.68627632e-01 -2.12060452e-01 -5.95748663e-01 3.03075761e-01 1.37569284e+00 5.54494634e-02 -4.21992958e-01 -9.02998507e-01 -4.01271321e-03 -3.50383855e-02 -1.32215440e-01 -1.56877056e-01 1.14002693e+00 -5.81853569e-01 -4.29702289e-02 8.80612671e-01 1.45479953e+00 2.39778891e-01 -1.38675392e+00 -2.23761886e-01 -2.55726904e-01 -7.23696411e-01 -6.37403667e-01 -3.69206578e-01 -1.27360630e+00 1.06424785e+00 -2.26775602e-01 3.81928951e-01 8.88971388e-01 3.51994694e-03 1.29428267e+00 8.58497992e-02 7.60273337e-02 -9.32424605e-01 6.87936425e-01 6.01685822e-01 1.23618710e+00 -1.26618814e+00 -3.02039802e-01 -4.74471539e-01 -1.28528810e+00 1.03517842e+00 7.62144089e-01 -7.54399002e-02 5.22669256e-01 -6.84887320e-02 4.91671026e-01 1.99537873e-01 -1.39073062e+00 1.61304255e-03 1.95023164e-01 6.72323167e-01 8.69117975e-01 -2.31741026e-01 -4.62926954e-01 7.40319788e-01 5.38037606e-02 3.19011271e-01 1.20456433e+00 8.94787371e-01 -5.17153144e-01 -1.03682303e+00 -1.27739131e-01 4.41990733e-01 -4.03459251e-01 -1.06632531e-01 -7.68903673e-01 6.38988167e-02 -1.98049635e-01 1.52552307e+00 2.16690242e-01 -5.32826483e-01 2.92288750e-01 -1.70361489e-01 -3.24555993e-01 -1.12489200e+00 -7.48238444e-01 1.03445180e-01 4.06270623e-01 -2.77971566e-01 -8.89822602e-01 -5.75639725e-01 -1.27391720e+00 -4.17492211e-01 -2.88167328e-01 5.42723894e-01 3.39465141e-01 1.01649666e+00 2.91806698e-01 4.09060508e-01 1.01714885e+00 -1.00525057e+00 6.05136752e-02 -1.16493154e+00 -4.37860906e-01 7.20873594e-01 4.16733593e-01 -1.18831918e-01 -5.34806430e-01 9.30768132e-01]
[10.741358757019043, 0.6763034462928772]
490dc13f-8cb5-438e-945c-10e20b35a370
synthetic-sample-selection-for-generalized
2304.02846
null
https://arxiv.org/abs/2304.02846v1
https://arxiv.org/pdf/2304.02846v1.pdf
Synthetic Sample Selection for Generalized Zero-Shot Learning
Generalized Zero-Shot Learning (GZSL) has emerged as a pivotal research domain in computer vision, owing to its capability to recognize objects that have not been seen during training. Despite the significant progress achieved by generative techniques in converting traditional GZSL to fully supervised learning, they tend to generate a large number of synthetic features that are often redundant, thereby increasing training time and decreasing accuracy. To address this issue, this paper proposes a novel approach for synthetic feature selection using reinforcement learning. In particular, we propose a transformer-based selector that is trained through proximal policy optimization (PPO) to select synthetic features based on the validation classification accuracy of the seen classes, which serves as a reward. The proposed method is model-agnostic and data-agnostic, making it applicable to both images and videos and versatile for diverse applications. Our experimental results demonstrate the superiority of our approach over existing feature-generating methods, yielding improved overall performance on multiple benchmarks.
['Shreyank N Gowda']
2023-04-06
null
null
null
null
['zero-shot-action-recognition', 'generalized-zero-shot-learning', 'generalized-zero-shot-learning']
['computer-vision', 'computer-vision', 'methodology']
[ 4.60728586e-01 -1.32944867e-01 -2.61721611e-01 -3.10324520e-01 -9.36690331e-01 -2.71817386e-01 8.57997537e-01 -9.90609825e-02 -1.61030844e-01 7.33979940e-01 -1.73273340e-01 1.51994780e-01 -1.11621797e-01 -8.27101231e-01 -6.09897316e-01 -9.49656367e-01 2.29864776e-01 3.02578509e-01 2.42888376e-01 -5.60562424e-02 2.71674007e-01 4.59819078e-01 -1.83342707e+00 6.03917520e-03 1.05262649e+00 1.11513412e+00 3.73252481e-01 2.36212075e-01 1.00403488e-01 7.68998086e-01 -5.04522741e-01 -3.00914794e-01 3.15008849e-01 -5.96475780e-01 -4.81699198e-01 5.59164762e-01 3.27271223e-01 -9.32849199e-02 -2.59891570e-01 8.48176777e-01 5.29910088e-01 2.91868418e-01 6.12122655e-01 -1.51575744e+00 -5.55615366e-01 3.43299896e-01 -4.06507939e-01 3.39094885e-02 7.57140070e-02 2.73495674e-01 1.04857516e+00 -1.15323806e+00 5.79979539e-01 1.21097827e+00 3.57619613e-01 7.42839217e-01 -1.34132493e+00 -6.10071659e-01 5.67946695e-02 4.20387983e-01 -1.28616154e+00 -4.44006622e-01 1.06197011e+00 -3.43667328e-01 5.95777273e-01 1.86290979e-01 7.82562554e-01 1.14724290e+00 3.22939188e-04 1.11213684e+00 1.23571122e+00 -5.07928312e-01 5.36673427e-01 2.66373873e-01 -3.52014452e-01 7.84724832e-01 2.57293701e-01 2.05005571e-01 -6.54013216e-01 -2.25358963e-01 7.21354067e-01 2.69343793e-01 -2.75011510e-01 -1.06377292e+00 -1.12934947e+00 1.06387675e+00 3.83668870e-01 -9.80770439e-02 -3.59142274e-01 6.14660000e-03 3.36300880e-01 3.05345654e-01 4.55806762e-01 5.74649751e-01 -2.83518225e-01 -4.62890007e-02 -7.63268173e-01 3.34797382e-01 3.95597607e-01 9.17416275e-01 8.47746611e-01 4.34746027e-01 -4.06690687e-01 1.02670276e+00 9.72103029e-02 4.00447518e-01 7.64101148e-01 -8.39848876e-01 2.03187585e-01 7.77207851e-01 8.90129134e-02 -7.92811513e-01 -3.22909541e-02 -6.89496636e-01 -7.77361393e-01 4.31227446e-01 1.26015451e-02 7.39546493e-03 -9.89129305e-01 1.77195823e+00 3.73643696e-01 1.96190670e-01 1.53860345e-01 7.52455890e-01 5.63859463e-01 5.36422074e-01 -6.08066469e-02 -1.76064715e-01 8.02138507e-01 -1.09696305e+00 -2.69479841e-01 -1.52831286e-01 2.99678653e-01 -5.73528111e-01 1.17790556e+00 2.86239684e-01 -7.42684603e-01 -6.12610936e-01 -1.05492997e+00 5.25529385e-01 -7.29306415e-02 8.33264217e-02 6.24716640e-01 5.10196269e-01 -6.94298506e-01 7.13927507e-01 -7.80868292e-01 -2.64716536e-01 7.48921216e-01 2.16353029e-01 -1.82753175e-01 -8.67285579e-02 -8.50499749e-01 5.71844876e-01 4.93278027e-01 -3.39706033e-01 -1.11083829e+00 -5.96938193e-01 -9.42974448e-01 2.51923561e-01 6.46375716e-01 -6.72686994e-01 1.23356664e+00 -1.06869626e+00 -1.63416219e+00 4.12890315e-01 4.87623774e-02 -6.12797320e-01 7.02224314e-01 -8.47446024e-02 -1.63517177e-01 2.26751655e-01 2.94570833e-01 6.39657021e-01 1.43259668e+00 -1.21784365e+00 -7.87814319e-01 -1.90734982e-01 -7.69596547e-02 3.52576435e-01 -4.24840003e-01 -3.57568085e-01 -2.34729692e-01 -9.17592466e-01 1.46808010e-02 -1.00740135e+00 -3.87578577e-01 1.11464307e-01 -2.87240326e-01 -3.10702443e-01 9.69517827e-01 9.92959812e-02 7.78126538e-01 -2.13322783e+00 6.44412115e-02 1.18800782e-01 7.45465085e-02 4.53738958e-01 -1.03016540e-01 5.46246946e-01 1.59190759e-01 -2.21540675e-01 -3.20262700e-01 -2.44187549e-01 -2.42545202e-01 1.63538888e-01 -3.18460882e-01 2.74906278e-01 6.40193880e-01 8.91147792e-01 -1.20341003e+00 -4.79305238e-01 4.16040331e-01 3.31961602e-01 -4.72016841e-01 3.37445915e-01 -3.15838605e-01 4.85057354e-01 -7.17014253e-01 7.60292113e-01 3.08178186e-01 -4.70952541e-01 1.46341706e-02 1.12850055e-01 1.47396386e-01 -2.80038804e-01 -1.02741647e+00 1.25859058e+00 -5.00452280e-01 4.24444765e-01 -5.02311707e-01 -1.17899215e+00 1.14605975e+00 2.02495947e-01 5.70190191e-01 -5.00249267e-01 -1.11912256e-02 1.68168455e-01 -1.64066672e-01 -3.28731000e-01 5.70868477e-02 -8.47686082e-02 2.30948105e-02 2.66667545e-01 3.16303879e-01 -2.39501357e-01 3.70063931e-01 1.12323433e-01 9.80472386e-01 3.50784570e-01 4.92437690e-01 -4.69685569e-02 5.35938382e-01 1.49436191e-01 8.37140501e-01 8.14941883e-01 -3.03907096e-02 5.06628573e-01 2.57895410e-01 -4.31992680e-01 -1.01015055e+00 -1.07241213e+00 2.63418779e-02 8.61637056e-01 2.45582744e-01 -2.26676553e-01 -5.22945106e-01 -7.86918104e-01 1.34127131e-02 6.17523491e-01 -6.20441496e-01 -5.54291725e-01 -2.62347341e-01 -5.56154013e-01 8.62517208e-02 4.28034246e-01 5.01341581e-01 -1.23301530e+00 -9.96477365e-01 3.43635112e-01 4.14993837e-02 -9.25599813e-01 -2.51268238e-01 1.82597056e-01 -9.87792552e-01 -1.21820736e+00 -8.86354923e-01 -6.98777437e-01 7.79197872e-01 5.76175272e-01 1.02283978e+00 -1.57846957e-01 -4.99332547e-01 4.48767036e-01 -7.30692446e-01 -5.39989531e-01 -3.18380833e-01 -4.69127893e-02 2.66951285e-02 4.82188672e-01 1.20900027e-01 -3.80613357e-01 -6.91035271e-01 2.69155741e-01 -8.29749405e-01 3.41508053e-02 7.92045593e-01 1.52580047e+00 7.26074874e-01 -2.58958433e-02 9.11130667e-01 -9.96089280e-01 5.67222834e-01 -3.15210074e-01 -6.40787363e-01 2.67985284e-01 -9.10618603e-01 1.34890601e-01 1.00462210e+00 -4.24100488e-01 -1.04067159e+00 3.19548786e-01 3.24513704e-01 -6.84607983e-01 4.76881899e-02 2.32000351e-01 -9.90845710e-02 -3.16878885e-01 6.79198265e-01 6.29546583e-01 2.13401318e-01 -2.72163272e-01 2.72343457e-01 3.54290217e-01 2.87472099e-01 -3.27750325e-01 9.31438148e-01 4.18783277e-01 3.09603754e-02 -8.28152359e-01 -9.81794059e-01 -4.28401738e-01 -3.16340387e-01 -2.04061523e-01 3.49938929e-01 -8.01639676e-01 -3.61973315e-01 3.52010667e-01 -5.39634049e-01 2.07001008e-02 -6.27207279e-01 4.82672393e-01 -8.93067300e-01 1.95523605e-01 -8.24874043e-02 -7.95496762e-01 -4.93138433e-01 -1.28756130e+00 9.33711231e-01 4.11578923e-01 9.09396708e-02 -6.05152667e-01 -2.41986606e-02 1.13078490e-01 4.07052726e-01 4.77321237e-01 8.45824897e-01 -5.46073139e-01 -7.13447034e-01 -3.08264792e-01 -5.00568151e-02 4.52613860e-01 3.75475496e-01 -5.57524115e-02 -8.82723451e-01 -5.58258474e-01 -1.94412604e-01 -7.47769833e-01 9.12554085e-01 3.06556255e-01 1.19378304e+00 -1.74007535e-01 -2.73009121e-01 4.19534981e-01 1.54766464e+00 2.82555938e-01 2.25672513e-01 4.58064377e-01 4.87069309e-01 3.07283133e-01 9.57184732e-01 7.52487838e-01 7.27779120e-02 7.38837898e-01 6.15383804e-01 -1.25612933e-02 -1.81793883e-01 -3.17559749e-01 1.60033062e-01 3.24257284e-01 4.16864548e-03 -3.53373513e-02 -5.63287258e-01 7.00957358e-01 -2.00080371e+00 -1.06913149e+00 3.71024549e-01 2.26783514e+00 6.92347109e-01 1.42616734e-01 1.94558799e-01 2.27248579e-01 6.20708525e-01 1.84797108e-01 -9.33917463e-01 -1.19853616e-01 3.47121023e-02 3.92779559e-01 1.99750602e-01 -3.22921425e-02 -1.12437546e+00 9.38772917e-01 5.80198288e+00 1.03644276e+00 -1.31916606e+00 -9.46876854e-02 6.24115586e-01 -1.52303949e-01 -1.13958150e-01 -1.13830455e-01 -5.89696407e-01 5.04263878e-01 4.40697223e-01 -4.56151366e-01 2.40224585e-01 1.34980536e+00 1.35946035e-01 -2.06559002e-02 -1.02760720e+00 1.13365674e+00 1.18451416e-01 -1.40804744e+00 3.54654044e-01 -2.07647681e-01 9.59749699e-01 -1.70652255e-01 2.75546134e-01 4.05457646e-01 1.35922506e-01 -7.92090833e-01 6.13154829e-01 3.28131974e-01 7.09890485e-01 -1.01777101e+00 5.54475725e-01 3.76021892e-01 -1.02514362e+00 -3.34632516e-01 -5.44384122e-01 1.89432070e-01 -9.50613841e-02 4.61952657e-01 -1.03560209e+00 4.73298848e-01 5.83617508e-01 6.14143312e-01 -5.14000177e-01 1.31295788e+00 -2.09239602e-01 6.73700213e-01 -5.63658122e-03 -2.26537824e-01 3.23094666e-01 -9.93517488e-02 6.20351672e-01 7.15121388e-01 3.53022069e-01 -3.04221570e-01 4.65574533e-01 6.86500371e-01 -4.77277860e-02 2.08798945e-01 -8.01228762e-01 4.12033014e-02 5.01435757e-01 1.31062341e+00 -7.22104013e-01 -3.33121300e-01 -3.92638385e-01 8.78439665e-01 3.06125790e-01 1.72406703e-01 -6.27479374e-01 -4.89437371e-01 4.27815557e-01 -4.75308858e-02 7.04798579e-01 1.73322652e-02 -1.47646628e-02 -1.23015177e+00 1.07937232e-01 -8.16682100e-01 1.68429732e-01 -3.53151292e-01 -1.25367212e+00 8.30704808e-01 -1.27692416e-01 -1.87538755e+00 -5.30899048e-01 -2.58088976e-01 -5.17238140e-01 5.12487650e-01 -1.55152977e+00 -1.20322001e+00 -4.22743529e-01 5.70536494e-01 1.00405669e+00 -6.54101849e-01 8.81596386e-01 -1.46992679e-03 -4.34858859e-01 6.42827272e-01 4.31091636e-01 -1.31041065e-01 5.62269330e-01 -1.28849781e+00 3.29191476e-01 6.78742230e-01 4.57425147e-01 2.55540520e-01 6.61079824e-01 -4.00989801e-01 -1.35535419e+00 -1.33318293e+00 2.82958627e-01 -2.44508125e-02 3.68208319e-01 -2.42101476e-01 -6.03524148e-01 1.95038199e-01 -1.67434394e-01 3.07182491e-01 5.94050050e-01 -2.64444202e-01 -3.21735084e-01 -3.75104070e-01 -1.20448172e+00 6.74217641e-01 8.79178464e-01 -1.49285033e-01 -3.91420543e-01 2.59782135e-01 4.64989662e-01 -1.19829461e-01 -5.41681647e-01 5.92192054e-01 4.64550972e-01 -9.74685848e-01 8.22828948e-01 -4.86994803e-01 4.60137188e-01 -2.41578132e-01 3.16395028e-03 -1.69829106e+00 -4.03600544e-01 -4.74798262e-01 -3.66698474e-01 1.06513429e+00 2.67815590e-01 -5.12745440e-01 1.04664624e+00 1.87339395e-01 2.42213812e-02 -1.21988964e+00 -8.20385396e-01 -1.03903735e+00 -5.03748775e-01 -4.35500145e-02 4.62721646e-01 7.54015267e-01 -4.02136266e-01 4.41768289e-01 -6.35343909e-01 -2.76722640e-01 9.85664904e-01 5.47089040e-01 8.21807325e-01 -1.35459793e+00 -4.73857790e-01 -2.66275316e-01 -6.31131768e-01 -7.46980011e-01 1.34056294e-02 -7.94009805e-01 1.81916282e-01 -1.28729880e+00 2.25138441e-01 -5.92829227e-01 -4.93995816e-01 5.22227943e-01 -2.64337063e-01 4.38752741e-01 2.71878928e-01 2.32862741e-01 -6.01573408e-01 1.00850856e+00 1.33884215e+00 -2.55701780e-01 -1.96405828e-01 3.41746807e-01 -6.98163033e-01 6.10811651e-01 8.22410822e-01 -4.88443702e-01 -7.95333862e-01 1.17054367e-02 -2.65705854e-01 -2.47362838e-03 3.15199733e-01 -1.18907285e+00 -1.85885221e-01 -2.81602383e-01 5.93967378e-01 -3.58819425e-01 4.89695638e-01 -8.35289419e-01 -1.57995865e-01 5.90105236e-01 -3.21904421e-01 -2.42693126e-01 -2.74686754e-01 8.94402564e-01 -3.81462872e-01 -2.86991805e-01 1.00501037e+00 -2.10525319e-01 -9.83921945e-01 5.46512306e-01 -1.12212591e-01 -2.90192962e-02 1.42773247e+00 -2.73569226e-01 -3.53029580e-03 -4.21752602e-01 -4.97613639e-01 1.66689456e-01 5.26451528e-01 5.23852944e-01 9.65846360e-01 -1.42245722e+00 -6.20236576e-01 3.70618403e-01 6.24399483e-01 -2.13406864e-03 1.15081117e-01 4.62956369e-01 -1.89019471e-01 2.34068766e-01 -4.14878309e-01 -6.95248306e-01 -1.12875688e+00 5.97674012e-01 -1.20702777e-02 -1.98744118e-01 -8.78344953e-01 7.01420605e-01 9.09610614e-02 -1.80466980e-01 2.35795751e-01 1.70660749e-01 -1.69119909e-01 -1.46690503e-01 4.73767638e-01 3.59157115e-01 -4.87931352e-03 -4.16027665e-01 -9.53967869e-02 3.68387371e-01 -2.46022493e-01 1.40785575e-01 1.54282308e+00 1.17006794e-01 4.31416005e-01 4.38300639e-01 9.52613115e-01 -3.24750870e-01 -1.63496876e+00 -4.50087070e-01 1.10449702e-01 -8.05838287e-01 -6.89070672e-02 -5.36686361e-01 -1.05092502e+00 5.74952602e-01 6.56079292e-01 2.74053752e-01 1.07434678e+00 -6.53651804e-02 7.25504875e-01 5.21728218e-01 6.80185676e-01 -1.12265110e+00 4.95812476e-01 2.17365891e-01 7.94301152e-01 -1.51766658e+00 -1.11261725e-01 -3.72384846e-01 -9.61369634e-01 1.02270329e+00 7.40136743e-01 -4.31445241e-01 3.20850700e-01 -2.06278294e-01 1.66841466e-02 6.79580122e-02 -8.16604555e-01 -3.35585833e-01 1.59806207e-01 7.06397831e-01 1.31361082e-01 -2.85912119e-02 -3.38493496e-01 5.89619875e-02 1.17712796e-01 6.34795502e-02 3.45452726e-01 1.19557393e+00 -5.43348551e-01 -1.33298814e+00 -1.60183921e-01 8.86039913e-01 -2.47854292e-01 1.92374438e-01 -1.51712641e-01 5.88364542e-01 -1.20610952e-01 8.09639513e-01 -2.17730388e-01 -3.12630236e-01 2.00234339e-01 -7.53363445e-02 4.96501565e-01 -8.27706397e-01 -1.87094226e-01 6.03395700e-02 -2.15615600e-01 -4.39642251e-01 -4.86372828e-01 -6.03330970e-01 -9.07550275e-01 3.06831360e-01 -5.52412271e-01 3.13121259e-01 3.87864977e-01 8.90233934e-01 4.72511739e-01 4.81975287e-01 1.26976943e+00 -9.40517008e-01 -9.98291016e-01 -7.55869448e-01 -5.73089421e-01 5.09562016e-01 2.61126220e-01 -1.19785368e+00 -6.51320117e-03 -3.74036729e-02]
[9.937841415405273, 2.9018869400024414]
323de0ae-33ee-4324-87a1-dde329e3fcd0
a-comparative-study-of-machine-learning-6
2307.00361
null
https://arxiv.org/abs/2307.00361v1
https://arxiv.org/pdf/2307.00361v1.pdf
A Comparative Study of Machine Learning Algorithms for Anomaly Detection in Industrial Environments: Performance and Environmental Impact
In the context of Industry 4.0, the use of artificial intelligence (AI) and machine learning for anomaly detection is being hampered by high computational requirements and associated environmental effects. This study seeks to address the demands of high-performance machine learning models with environmental sustainability, contributing to the emerging discourse on 'Green AI.' An extensive variety of machine learning algorithms, coupled with various Multilayer Perceptron (MLP) configurations, were meticulously evaluated. Our investigation encapsulated a comprehensive suite of evaluation metrics, comprising Accuracy, Area Under the Curve (AUC), Recall, Precision, F1 Score, Kappa Statistic, Matthews Correlation Coefficient (MCC), and F1 Macro. Simultaneously, the environmental footprint of these models was gauged through considerations of time duration, CO2 equivalent, and energy consumption during the training, cross-validation, and inference phases. Traditional machine learning algorithms, such as Decision Trees and Random Forests, demonstrate robust efficiency and performance. However, superior outcomes were obtained with optimised MLP configurations, albeit with a commensurate increase in resource consumption. The study incorporated a multi-objective optimisation approach, invoking Pareto optimality principles, to highlight the trade-offs between a model's performance and its environmental impact. The insights derived underscore the imperative of striking a balance between model performance, complexity, and environmental implications, thus offering valuable directions for future work in the development of environmentally conscious machine learning models for industrial applications.
['Alejandro Echeverría Rey', 'Rubén García Maezo', 'Carlos Martí-González', 'Álvaro Huertas-García']
2023-07-01
null
null
null
null
['anomaly-detection']
['methodology']
[ 6.46433175e-01 -7.64787942e-02 -3.39379907e-01 1.32968843e-01 -2.67452508e-01 -5.30258119e-01 6.38889372e-01 4.35758859e-01 -2.87827432e-01 4.96409267e-01 -3.33846509e-01 -8.66786182e-01 -7.14594781e-01 -6.45805836e-01 -2.37169147e-01 -9.70382452e-01 -1.49625421e-01 -1.03762120e-01 -3.42829794e-01 3.12258273e-01 4.92799848e-01 6.82942092e-01 -1.49002123e+00 -1.75429896e-01 9.37501669e-01 1.39222980e+00 -4.96225655e-02 5.65733910e-01 1.75841570e-01 7.45337069e-01 -4.79078561e-01 -4.35143560e-01 2.63171226e-01 -1.51101604e-01 -3.61951053e-01 -8.62862989e-02 -1.84455484e-01 1.87024772e-01 3.30210477e-01 6.33935928e-01 3.74469519e-01 -3.10251117e-02 7.31179178e-01 -1.50846326e+00 -5.14076769e-01 5.99029288e-02 -2.91873008e-01 1.76172361e-01 -2.61900872e-01 9.91851807e-01 9.31649268e-01 -3.82431507e-01 6.42798841e-03 6.65755987e-01 6.09425604e-01 -7.84807093e-03 -1.36658204e+00 -6.41725779e-01 9.86783877e-02 2.10822299e-01 -1.03161764e+00 -4.78774011e-01 5.96096694e-01 -6.57505572e-01 1.42116618e+00 6.04568899e-01 8.63490105e-01 6.10142708e-01 8.38332236e-01 2.02544164e-02 1.27994621e+00 -5.92920601e-01 5.71921170e-01 4.48494256e-01 -3.84768635e-01 3.45226884e-01 5.96625447e-01 6.29011691e-01 -2.21599832e-01 1.00953886e-02 3.64748299e-01 -1.92318801e-02 3.52278531e-01 -4.22210485e-01 -8.27871382e-01 6.35696769e-01 4.53233160e-02 2.79548205e-02 -5.75744033e-01 1.02163842e-02 5.20287931e-01 2.86070347e-01 3.35499555e-01 1.11443818e+00 -7.59955883e-01 -1.80734210e-02 -5.67853749e-01 -1.20353454e-03 7.86269665e-01 3.70537311e-01 4.24026877e-01 3.47513705e-01 -3.95046920e-02 6.57334685e-01 6.07265890e-01 3.06244731e-01 1.03081152e-01 -1.23633456e+00 2.64489830e-01 7.81426013e-01 1.98040843e-01 -1.05700481e+00 -4.42129940e-01 -8.23393524e-01 -6.54969335e-01 5.56052625e-01 1.41681761e-01 -2.05009982e-01 -6.13976717e-01 1.15821373e+00 1.90591127e-01 -3.92036021e-01 1.30944565e-01 4.41141844e-01 -1.31428525e-01 4.07725990e-01 7.52485573e-01 -5.37522912e-01 7.60811448e-01 -7.06540287e-01 -4.18581784e-01 -3.36716682e-01 6.11022592e-01 -4.06405538e-01 8.41113210e-01 5.10766983e-01 -8.91171873e-01 -3.00245881e-01 -1.05888426e+00 6.19621336e-01 -5.34696460e-01 -1.87767774e-01 6.78064585e-01 1.04972100e+00 -3.18953931e-01 8.30134153e-01 -6.59798503e-01 -1.09963588e-01 3.97188663e-01 4.42418873e-01 1.88443869e-01 1.65500760e-01 -6.50229752e-01 1.42604160e+00 3.61511409e-01 4.11933720e-01 -6.00096703e-01 -7.35195279e-01 -5.25819540e-01 -3.59245613e-02 2.46586889e-01 -5.09451330e-01 1.09824872e+00 -6.62975490e-01 -1.25741899e+00 4.31512892e-01 4.61866230e-01 -4.66062874e-01 4.72104728e-01 -2.02858761e-01 -9.37106133e-01 -3.85334402e-01 -2.28097677e-01 1.84824437e-01 2.49931857e-01 -1.10100257e+00 -8.79323304e-01 -5.86862803e-01 -3.13650846e-01 4.05817740e-02 -2.12938905e-01 5.22144549e-02 5.23180246e-01 -1.71303794e-01 -2.04403594e-01 -9.48776424e-01 -5.33782780e-01 -2.88195878e-01 -2.75591195e-01 -8.81880969e-02 5.78252494e-01 -6.74197972e-01 1.49104619e+00 -1.86031985e+00 -3.74933124e-01 5.19060194e-01 -2.74979383e-01 3.51989001e-01 1.91780344e-01 4.14164305e-01 -8.01220536e-02 2.84849674e-01 -3.65217388e-01 5.63213527e-01 1.24022067e-01 2.63553392e-02 1.72515318e-01 4.73891109e-01 7.33398914e-01 6.00639701e-01 -9.32890773e-01 -1.82628155e-01 6.70629799e-01 2.18121931e-01 -1.26361653e-01 -1.51521802e-01 -2.31793866e-01 3.15571755e-01 -1.93962619e-01 1.07266355e+00 -6.59052208e-02 6.88010454e-02 3.68614018e-01 8.68152827e-03 -6.06900930e-01 1.29977629e-01 -9.85077441e-01 9.64132428e-01 -6.59179032e-01 3.71816039e-01 6.50277436e-02 -8.54634464e-01 1.10371649e+00 1.85712114e-01 6.23510540e-01 -1.17422497e+00 2.95406561e-02 4.63941634e-01 2.32514113e-01 -7.82226801e-01 2.78297216e-02 -1.63384393e-01 2.47368887e-01 4.03444655e-02 -3.34573090e-01 -8.27181488e-02 -2.39718035e-02 -8.21649969e-01 1.09947038e+00 4.55365151e-01 3.78006309e-01 -4.15528864e-01 3.18997979e-01 3.80259365e-01 2.94154018e-01 1.96519718e-01 -3.13482881e-01 -1.76747099e-01 3.23041737e-01 -4.01597291e-01 -1.02681017e+00 -8.88889194e-01 -3.43876362e-01 9.63935912e-01 -7.89484605e-02 2.38724813e-01 -2.69195348e-01 -2.85744309e-01 3.06614220e-01 1.31476951e+00 -1.33776709e-01 -4.82059628e-01 -3.59561801e-01 -1.25795841e+00 2.64670402e-01 2.95388252e-01 2.62967110e-01 -8.08319628e-01 -1.39794171e+00 4.99240279e-01 5.03840566e-01 -5.74542522e-01 4.17764038e-01 7.37749755e-01 -1.05546975e+00 -1.30678546e+00 -1.08335823e-01 -2.16622248e-01 2.37720504e-01 -1.15572140e-01 1.16085303e+00 -2.95310676e-01 -4.46062028e-01 1.43748060e-01 1.03868857e-01 -1.16166317e+00 -6.04140580e-01 -1.91656556e-02 -1.61538702e-02 -5.28458297e-01 5.67153215e-01 -6.81946397e-01 -7.61888802e-01 4.30273235e-01 -6.87482655e-01 -3.29776376e-01 1.17273319e+00 5.30805409e-01 5.02554476e-01 4.88098711e-01 1.00399649e+00 -5.42642593e-01 5.90169489e-01 -7.02172220e-01 -5.13329446e-01 2.46528253e-01 -1.80471075e+00 -1.00711502e-01 4.19202060e-01 -2.75114298e-01 -9.00232315e-01 -2.33850088e-02 1.80541039e-01 -6.32281005e-02 -3.13508481e-01 4.89686340e-01 -2.31307983e-01 5.74368611e-02 7.77285993e-01 -1.63422614e-01 2.95860231e-01 -3.95629823e-01 1.90160926e-02 6.03994012e-01 5.08030236e-01 -4.07562137e-01 5.83594382e-01 -1.73399061e-01 3.37139517e-01 -6.80157542e-01 -3.29891890e-01 -3.44326854e-01 -4.87560123e-01 -5.64405501e-01 4.24791962e-01 -6.59372985e-01 -6.70007885e-01 2.18218178e-01 -4.60558236e-01 -2.04086035e-01 -3.88665169e-01 5.16036749e-01 -3.45249861e-01 -1.02977090e-01 4.15790007e-02 -1.42141104e+00 -5.03665805e-01 -8.12411606e-01 3.44869912e-01 8.31976011e-02 -7.06012785e-01 -8.54512274e-01 -3.77640426e-02 5.04472554e-01 5.31683624e-01 9.32251394e-01 1.45120120e+00 -7.41757691e-01 -3.57905865e-01 -3.79510581e-01 -4.29521650e-02 6.71336949e-01 1.33006051e-01 2.99862772e-01 -8.79765153e-01 -8.62170011e-02 -1.48018867e-01 6.01573288e-02 3.67026716e-01 2.00757429e-01 1.04349566e+00 -7.12761283e-01 -2.27217421e-01 5.57977073e-02 1.71053720e+00 8.66059482e-01 4.48007405e-01 8.60473096e-01 4.29605514e-01 8.27863693e-01 6.08004212e-01 3.25934768e-01 9.17147007e-03 2.21773356e-01 9.75834846e-01 8.14539865e-02 3.21144491e-01 1.05595112e-01 3.47109586e-01 4.20945913e-01 -2.70490259e-01 -7.68127963e-02 -1.08497679e+00 5.30914605e-01 -1.41163754e+00 -7.99908996e-01 -1.68233722e-01 2.36600780e+00 4.73500401e-01 5.14568388e-01 3.62598717e-01 7.30079114e-01 4.81956840e-01 -6.34264797e-02 -8.19647074e-01 -1.06938267e+00 2.86644787e-01 1.61207747e-02 8.52479637e-01 -2.19956681e-01 -8.56487334e-01 -3.50018851e-02 6.74095058e+00 2.94142723e-01 -1.04857528e+00 -3.04882050e-01 8.60165477e-01 -4.69206274e-01 -2.12899700e-01 -1.85738608e-01 -3.89908314e-01 6.52145147e-01 1.67089844e+00 -2.48255581e-01 5.08429885e-01 8.89165223e-01 5.19305587e-01 -2.89699197e-01 -1.04832757e+00 1.23893790e-01 -4.35701996e-01 -9.52331781e-01 -4.21216756e-01 3.49954367e-01 5.59503496e-01 6.01113401e-02 1.16835542e-01 2.00878173e-01 2.51864254e-01 -1.09723318e+00 7.24965334e-01 4.40836608e-01 6.43107593e-01 -1.01306939e+00 6.98004067e-01 7.09047168e-02 -6.31455481e-01 -9.02047336e-01 1.64537847e-01 -4.55225527e-01 -3.48933995e-01 8.35719943e-01 -9.06274915e-01 4.64897871e-01 7.20901549e-01 -3.07507999e-02 -3.65591109e-01 1.02353847e+00 2.92897046e-01 6.50905132e-01 -2.28237242e-01 -2.87822664e-01 2.45478645e-01 -4.06893671e-01 3.22271168e-01 1.25046885e+00 2.99455255e-01 -4.12126750e-01 -3.04892212e-01 8.56129646e-01 5.55020630e-01 1.69098720e-01 -6.53025866e-01 -3.27363908e-01 7.13804364e-01 1.03918970e+00 -5.03544569e-01 3.27900827e-01 -4.79124010e-01 6.64845631e-02 -4.76512551e-01 1.78738698e-01 -6.46303713e-01 -2.94459701e-01 7.07374811e-01 2.69463897e-01 -1.60691977e-01 -9.62747782e-02 -1.09039843e+00 -2.55464166e-02 8.40532333e-02 -7.24531651e-01 2.66537160e-01 -3.71677071e-01 -1.09218776e+00 -1.72781378e-01 6.84199035e-02 -9.88968492e-01 -1.63011268e-01 -7.93828011e-01 -6.12424552e-01 8.93037260e-01 -1.46610916e+00 -8.25551927e-01 -5.87402657e-02 -3.02221507e-01 3.13902736e-01 -2.30996311e-01 9.36268091e-01 1.56919599e-01 -9.16854262e-01 1.15166955e-01 5.85592330e-01 -6.29616737e-01 1.05862312e-01 -1.12609959e+00 -4.59211320e-03 5.13107300e-01 -5.63409269e-01 2.12534383e-01 8.02426338e-01 -5.61144054e-01 -1.38051939e+00 -1.32105446e+00 6.04797184e-01 -3.39473903e-01 7.10316777e-01 4.25130546e-01 -3.70203227e-01 2.83119917e-01 -8.45990255e-02 -5.90102434e-01 9.87358332e-01 2.15145469e-01 1.36129364e-01 -4.68773693e-01 -1.37095284e+00 5.09099305e-01 7.22653687e-01 -1.93456680e-01 -7.16204150e-03 2.91178450e-02 3.88662934e-01 1.25269353e-01 -1.48100376e+00 8.36608052e-01 9.79719460e-01 -8.73307645e-01 9.37206447e-01 -6.79710388e-01 5.59166908e-01 -8.02963786e-03 -1.79097548e-01 -1.14017367e+00 -5.06711602e-01 -3.55722070e-01 -4.36427832e-01 1.30716932e+00 7.66656637e-01 -6.35571778e-01 5.82233667e-01 1.14691579e+00 -3.20092171e-01 -1.22956848e+00 -9.02320087e-01 -8.80135536e-01 1.35058969e-01 -4.63400543e-01 5.76423168e-01 7.57025123e-01 -5.44517078e-02 1.34571016e-01 9.20478553e-02 1.31516367e-01 5.72505295e-01 -1.97062269e-01 2.82280266e-01 -1.53672481e+00 1.18560873e-01 -7.16143787e-01 -3.30637038e-01 4.00515079e-01 -3.48532975e-01 -5.40259957e-01 -2.35629544e-01 -1.34399748e+00 -9.66512561e-02 -5.19338012e-01 -9.60668445e-01 3.87301981e-01 1.41913285e-02 -1.68983281e-01 1.65790558e-01 2.30626777e-01 -8.77103731e-02 6.96936473e-02 9.02568579e-01 -2.31147870e-01 -5.31186938e-01 3.59352410e-01 -9.67634916e-01 6.33350313e-01 1.29735529e+00 -3.76083404e-01 -3.63224268e-01 -2.51891255e-01 3.31028044e-01 -2.96936572e-01 5.23522437e-01 -1.15420806e+00 -8.02819431e-02 -7.96340287e-01 7.82864451e-01 -1.93162113e-01 -4.15523313e-02 -1.23825443e+00 8.83160889e-01 7.28666008e-01 -3.73511612e-01 3.19732726e-01 3.04113537e-01 5.74454963e-01 2.91118473e-01 -1.32706687e-01 6.70327485e-01 -1.59936100e-01 -5.76593816e-01 -2.68758804e-01 -4.34908688e-01 -5.22290528e-01 1.35591877e+00 -7.19075441e-01 -2.06614763e-01 4.84577209e-01 -5.16610622e-01 3.31412017e-01 4.72857475e-01 3.60506207e-01 -6.77669235e-03 -9.18544233e-01 -3.63251299e-01 -1.01773776e-01 1.83119282e-01 7.96352513e-03 -8.53505480e-05 8.04678082e-01 -3.61838698e-01 5.26918411e-01 -2.76747376e-01 -3.72762650e-01 -8.44246745e-01 5.28922856e-01 4.71088886e-01 -2.81683892e-01 -1.91361919e-01 2.41054341e-01 -4.20668602e-01 -2.73964435e-01 3.00627708e-01 -1.35633036e-01 1.69485658e-01 2.74032392e-02 -1.40604461e-02 1.36420929e+00 3.00639153e-01 -2.87159793e-02 -3.63676995e-01 1.19880438e-01 5.57509720e-01 1.76799804e-01 1.35098493e+00 -4.53662910e-02 4.74649668e-02 5.08236885e-01 7.71391153e-01 -3.97566050e-01 -1.25816190e+00 1.66921526e-01 6.31024957e-01 -2.78518498e-01 3.65406334e-01 -1.52900243e+00 -7.60369122e-01 4.57540661e-01 8.40914428e-01 3.92989844e-01 1.29516447e+00 -5.12386978e-01 3.78313273e-01 3.34459394e-01 -4.53938767e-02 -1.53583419e+00 -2.22229719e-01 -2.53239512e-01 7.19401836e-01 -1.00794959e+00 3.42866212e-01 -4.17692512e-02 -5.78229725e-01 9.49585080e-01 6.56139553e-01 4.89521444e-01 3.86048019e-01 3.22615594e-01 -7.14237615e-02 -9.16474313e-02 -8.26231420e-01 1.21054247e-01 -1.07546389e-01 6.02017403e-01 1.71942711e-01 3.24023426e-01 -9.93652716e-02 4.62596804e-01 -5.14288917e-02 1.06200024e-01 -1.23099886e-01 9.45980728e-01 -5.53566933e-01 -6.93492711e-01 -2.75981903e-01 9.57877398e-01 -7.06130207e-01 1.65388823e-01 -3.71111959e-01 8.55411708e-01 4.13734823e-01 1.20944047e+00 1.99715681e-02 -5.47575653e-01 6.10995412e-01 4.52249348e-01 4.30822521e-02 -2.39065200e-01 -7.43062794e-01 -1.90495864e-01 5.48646450e-01 -2.12542102e-01 -2.68197507e-01 -7.62151420e-01 -6.64280891e-01 -2.97120929e-01 -5.13975024e-01 -4.14609276e-02 1.28650689e+00 8.71586323e-01 4.30577189e-01 9.11127508e-01 9.09183681e-01 -4.58990395e-01 -8.35889339e-01 -7.95443594e-01 -4.69655126e-01 -2.52587616e-01 1.80656597e-01 -3.86585563e-01 -4.18294430e-01 -1.38022959e-01]
[6.250392436981201, 3.634669303894043]
3f38662a-4d14-4df9-8d48-4a5f7eb9ef93
co-salient-object-detection-with-co
2303.07670
null
https://arxiv.org/abs/2303.07670v1
https://arxiv.org/pdf/2303.07670v1.pdf
Co-Salient Object Detection with Co-Representation Purification
Co-salient object detection (Co-SOD) aims at discovering the common objects in a group of relevant images. Mining a co-representation is essential for locating co-salient objects. Unfortunately, the current Co-SOD method does not pay enough attention that the information not related to the co-salient object is included in the co-representation. Such irrelevant information in the co-representation interferes with its locating of co-salient objects. In this paper, we propose a Co-Representation Purification (CoRP) method aiming at searching noise-free co-representation. We search a few pixel-wise embeddings probably belonging to co-salient regions. These embeddings constitute our co-representation and guide our prediction. For obtaining purer co-representation, we use the prediction to iteratively reduce irrelevant embeddings in our co-representation. Experiments on three datasets demonstrate that our CoRP achieves state-of-the-art performances on the benchmark datasets. Our source code is available at https://github.com/ZZY816/CoRP.
['Ming-Ming Cheng', 'Xing Sun', 'Zheng Lin', 'Zhao Zhang', 'Ziyue Zhu']
2023-03-14
null
null
null
null
['co-saliency-detection']
['computer-vision']
[-7.08431331e-03 -1.20904967e-01 -2.64517635e-01 4.52877879e-02 -7.67502904e-01 -5.11027239e-02 4.65956837e-01 6.70750618e-01 -1.57021657e-01 2.22650915e-01 5.35012662e-01 2.10747421e-01 -2.75834471e-01 -7.47372746e-01 -5.93780220e-01 -6.52996778e-01 -1.40326977e-01 -1.23862728e-01 4.95310992e-01 3.59006487e-02 4.99745160e-01 3.90289754e-01 -1.96483707e+00 4.08465266e-01 8.55709374e-01 9.48076427e-01 8.16092372e-01 1.61032513e-01 -2.01699212e-01 5.88884175e-01 -3.88476402e-01 7.29608610e-02 3.18386883e-01 -4.06995505e-01 -7.01759100e-01 1.24660604e-01 2.85977364e-01 1.25264794e-01 -5.11243232e-02 1.38997412e+00 3.05946559e-01 1.00697972e-01 5.77132046e-01 -1.14339650e+00 -9.54264939e-01 3.91445369e-01 -8.75068247e-01 6.80654943e-01 -8.83693248e-02 -6.39182255e-02 1.30690300e+00 -1.43659461e+00 4.92710769e-01 1.07271147e+00 2.94675142e-01 2.24423051e-01 -8.88584495e-01 -4.34858471e-01 2.64941305e-01 6.28762126e-01 -1.86884606e+00 -3.17452669e-01 1.28615439e+00 -1.12948105e-01 8.12397718e-01 4.12949562e-01 6.98377430e-01 4.32029098e-01 -2.23770011e-02 8.94901931e-01 7.33671188e-01 -6.67484581e-01 1.23449974e-01 2.77109981e-01 9.96272713e-02 6.69276297e-01 4.13866490e-01 -3.10295999e-01 -5.11883557e-01 -1.81576461e-01 5.33661902e-01 5.79317689e-01 -2.69596726e-01 -5.17682612e-01 -1.28476381e+00 7.04427183e-01 7.53173709e-01 6.93807602e-01 -5.05154252e-01 -7.18778148e-02 2.65050113e-01 1.80361960e-02 5.95386028e-01 6.67125404e-01 -1.87566087e-01 9.69419554e-02 -5.86038768e-01 4.83753681e-02 9.01662633e-02 8.45323145e-01 1.13493097e+00 -3.61664861e-01 -2.35237375e-01 9.94148433e-01 1.59696817e-01 1.89816412e-02 9.04986203e-01 -4.90366131e-01 5.12157619e-01 1.17606628e+00 6.43603280e-02 -1.42107856e+00 -1.29152671e-01 -5.38765013e-01 -5.69941461e-01 2.18726378e-02 -1.07322522e-02 1.81102604e-01 -6.54701054e-01 1.32825589e+00 5.46840966e-01 2.80088127e-01 3.99270421e-03 1.09404969e+00 7.60967374e-01 7.34930396e-01 -2.87189633e-02 -1.37604028e-01 1.47991419e+00 -1.16231930e+00 -6.11477077e-01 -3.51093948e-01 6.50238752e-01 -8.80290270e-01 1.05533469e+00 -8.30585063e-02 -6.81942761e-01 -6.58058763e-01 -1.07948458e+00 -1.71900108e-01 -5.39824486e-01 4.15587664e-01 4.60974604e-01 -6.51289523e-03 -5.17158628e-01 5.96111476e-01 -4.27194655e-01 -4.15797383e-01 6.13308012e-01 1.47836924e-01 -4.29944873e-01 -9.96563286e-02 -9.35394585e-01 8.84347618e-01 3.55298162e-01 -1.35871947e-01 -6.84516609e-01 -6.28964126e-01 -8.00233901e-01 6.59816116e-02 5.83880723e-01 -2.69032776e-01 6.72016442e-01 -9.77761507e-01 -5.82873762e-01 8.36641729e-01 -5.16834497e-01 -3.23439091e-01 9.53575969e-02 -4.80954349e-01 -7.30068386e-01 3.33928823e-01 3.97785962e-01 6.60703838e-01 1.06556654e+00 -1.39725769e+00 -8.60637784e-01 -3.18165779e-01 -3.20317060e-01 5.44010878e-01 -7.38039255e-01 2.53981620e-01 -5.93442380e-01 -8.31891239e-01 4.28504109e-01 -5.46806097e-01 -2.26549700e-01 2.47596040e-01 -4.50328022e-01 -4.35697585e-01 1.11297524e+00 -1.60311565e-01 1.52040684e+00 -2.38047123e+00 -2.55675912e-01 -4.16624248e-02 3.95295590e-01 3.67303073e-01 -2.45394304e-01 5.44598937e-01 -3.24499488e-01 2.12087616e-01 -4.10755016e-02 -3.48561734e-01 -2.68941283e-01 -1.66540630e-02 -2.36064598e-01 5.17776430e-01 5.59869468e-01 7.06151783e-01 -9.65174913e-01 -8.39534581e-01 3.30347002e-01 2.20880046e-01 -3.23560476e-01 1.92446187e-01 7.61943087e-02 -8.69101733e-02 -7.25937009e-01 7.80375838e-01 7.82007873e-01 -3.03405046e-01 -2.36138389e-01 -4.20447558e-01 -3.58986378e-01 1.52631685e-01 -1.16714454e+00 1.41057742e+00 -1.65109988e-02 4.49162304e-01 -3.42507094e-01 -1.11913025e+00 1.21008146e+00 -2.61167884e-01 6.44505858e-01 -7.38642275e-01 3.03608775e-02 1.85332566e-01 2.59052087e-02 -5.88539362e-01 7.68766940e-01 8.71242881e-02 3.20463210e-01 4.19042438e-01 -6.02638237e-02 3.89172226e-01 8.51593763e-02 1.48506910e-01 8.30663085e-01 -3.58116031e-01 6.03466392e-01 -4.88689184e-01 6.20917737e-01 1.12382270e-01 6.35228336e-01 4.65990514e-01 -5.22977412e-01 9.04872417e-01 2.12134302e-01 -4.66461539e-01 -8.82039130e-01 -8.01216662e-01 -1.23838663e-01 8.37422788e-01 6.55962169e-01 -7.40110159e-01 -3.10193717e-01 -8.72355402e-01 8.40655714e-02 3.61866772e-01 -9.83046114e-01 -3.44061911e-01 -4.40306127e-01 -5.19833982e-01 -1.08462639e-01 3.13627452e-01 2.73278475e-01 -1.19578779e+00 -7.06608295e-01 2.55091414e-02 -3.40860337e-02 -6.78391874e-01 -6.64996982e-01 3.21587622e-01 -8.56907725e-01 -1.17201376e+00 -8.43333304e-01 -1.28933525e+00 1.11723483e+00 1.12950242e+00 8.26609552e-01 2.48467401e-01 -4.78767604e-01 -1.50873497e-01 -5.74774086e-01 -4.16683048e-01 2.13536263e-01 -2.62416303e-01 -4.00585262e-03 2.12278843e-01 9.88221645e-01 -3.44050199e-01 -8.71573508e-01 3.82905364e-01 -6.66668653e-01 7.29525462e-03 5.29611409e-01 6.18594944e-01 1.18159473e+00 1.79777026e-01 6.29629433e-01 -5.24283826e-01 5.42249441e-01 -7.64237225e-01 -2.84210205e-01 1.74549773e-01 -4.95888084e-01 6.60956651e-02 5.96565187e-01 -5.66304147e-01 -6.11187696e-01 7.93082416e-02 2.63424218e-01 -8.83056998e-01 -1.31590560e-01 3.99354666e-01 -1.40014440e-01 3.32646072e-01 5.65962970e-01 4.64456797e-01 -1.94568828e-01 -7.53007054e-01 3.07384610e-01 6.13271892e-01 2.99336314e-01 -1.88490182e-01 7.60022640e-01 6.59547031e-01 -2.35410839e-01 -6.29394829e-01 -9.84956741e-01 -9.92509007e-01 -5.38049340e-01 5.45961857e-02 4.56303030e-01 -9.28679407e-01 1.56382278e-01 -1.58470660e-01 -8.41614664e-01 5.06485105e-01 -6.45028949e-01 4.22685176e-01 -2.53635973e-01 4.03540164e-01 1.62457004e-01 -8.41715515e-01 -3.63500148e-01 -9.93302882e-01 8.76406968e-01 4.14485633e-01 -1.55054867e-01 -4.68585998e-01 1.15133822e-01 -8.08355883e-02 1.42276362e-01 1.10077867e-02 6.24593914e-01 -8.05698931e-01 -4.95334357e-01 -1.98501006e-01 -4.33876485e-01 1.15028039e-01 6.61898255e-01 -1.95750162e-01 -8.81912172e-01 -1.03083886e-02 -9.47758630e-02 -1.81202851e-02 1.09934354e+00 2.81781346e-01 1.57091105e+00 -3.92126501e-01 -5.79328716e-01 4.46481705e-01 1.54675341e+00 1.49292201e-01 5.73561788e-01 4.69433397e-01 6.90963447e-01 7.04101980e-01 1.22916687e+00 5.64818859e-01 2.63414472e-01 6.50028586e-01 5.87602973e-01 -9.31015387e-02 -1.37646630e-01 -5.11398494e-01 1.83942243e-01 7.52599180e-01 2.80461580e-01 2.11641178e-01 -7.83203423e-01 1.34199286e+00 -1.98456144e+00 -9.54839110e-01 -3.93977046e-01 1.96050358e+00 9.11735594e-01 1.64493732e-02 2.11517781e-01 2.54989862e-01 1.03415680e+00 3.29205483e-01 -4.15894210e-01 -1.42119184e-01 -1.20653413e-01 -1.16433688e-01 2.83482194e-01 6.89203739e-02 -1.42024279e+00 8.04375410e-01 4.94436502e+00 1.09423292e+00 -8.43395352e-01 3.03042293e-01 5.58778584e-01 -1.41567111e-01 -3.55489671e-01 2.65294481e-02 -8.54043603e-01 5.67552924e-01 4.00782734e-01 -4.17008609e-01 -1.91133052e-01 1.14749467e+00 2.16242999e-01 -2.56791174e-01 -7.93940246e-01 1.07894409e+00 3.76047581e-01 -1.41521513e+00 1.07196286e-01 -9.86786261e-02 8.95110965e-01 -1.39082596e-01 2.35636026e-01 4.91878539e-02 -1.38323158e-01 -7.56955862e-01 7.23577440e-01 3.55709821e-01 5.73130906e-01 -9.50403810e-01 6.22653604e-01 3.35415632e-01 -1.46239007e+00 -2.09200054e-01 -8.75525653e-01 2.84689609e-02 -2.35598162e-01 8.07373405e-01 -7.09588885e-01 4.37687516e-01 9.36578751e-01 1.04747367e+00 -7.21723139e-01 1.33050680e+00 -4.52742204e-02 3.12017351e-01 -1.51078999e-01 -1.95129797e-01 1.94959626e-01 -3.29673216e-02 6.06238484e-01 1.18829978e+00 3.31185579e-01 8.78571719e-03 1.59498006e-02 9.83206272e-01 -4.78566661e-02 4.07247543e-01 -6.28506362e-01 -1.94041245e-02 7.18389690e-01 1.17368376e+00 -8.20949316e-01 -3.32296610e-01 -3.54513437e-01 9.49487507e-01 5.03570974e-01 3.66621204e-02 -6.52648449e-01 -6.79116786e-01 9.85072553e-01 1.33638889e-01 7.47777402e-01 8.99228081e-02 -3.69948536e-01 -9.89687204e-01 2.74617046e-01 -5.10272503e-01 5.33200502e-01 -6.41967356e-01 -1.32904518e+00 4.64809120e-01 -8.58331174e-02 -1.94051266e+00 3.22175622e-01 -6.81182593e-02 -7.36837149e-01 9.35067832e-01 -1.83614612e+00 -9.88530576e-01 -2.83597022e-01 5.04582584e-01 7.57800043e-01 -4.19814140e-02 7.21492052e-01 1.77619487e-01 -5.55859208e-01 4.05337393e-01 8.49281996e-02 -1.28489882e-01 5.73115587e-01 -1.03015244e+00 1.49062276e-01 1.12716365e+00 3.85134935e-01 8.60673368e-01 5.64917922e-01 -6.76113188e-01 -1.20924413e+00 -1.33515763e+00 1.09048200e+00 -2.43789315e-01 6.64592922e-01 -2.21807763e-01 -1.09144545e+00 2.78295845e-01 9.64772888e-03 6.71361148e-01 5.29252172e-01 -5.12025207e-02 -2.79234290e-01 -2.33024508e-01 -1.05896187e+00 6.43887937e-01 1.05981302e+00 -6.29421234e-01 -9.04231131e-01 1.73335537e-01 7.56170750e-01 1.34710923e-01 -3.99596870e-01 1.89056516e-01 2.11946979e-01 -7.29756594e-01 1.17145848e+00 -4.51395869e-01 5.88293731e-01 -8.31839561e-01 -1.53746620e-01 -1.21323454e+00 -5.65100908e-01 -3.36828768e-01 -4.56925511e-01 1.29200602e+00 4.94626164e-02 -1.88703135e-01 6.52164340e-01 -5.27298227e-02 -1.00738637e-01 -1.04489434e+00 -7.82626748e-01 -8.52929652e-01 -3.26243252e-01 -2.38167912e-01 6.96413040e-01 9.95520651e-01 1.01324432e-01 -7.80089051e-02 -2.74629831e-01 3.94669294e-01 6.47908449e-01 5.98537266e-01 5.09659767e-01 -1.07894945e+00 1.13333799e-01 -4.16108698e-01 -4.57191169e-01 -1.11923671e+00 -2.36621588e-01 -8.54669750e-01 5.58836348e-02 -1.67170024e+00 4.98106211e-01 -5.95552325e-01 -9.31940734e-01 6.76747978e-01 -4.99669522e-01 2.79812247e-01 2.41191804e-01 6.19926155e-01 -9.26171541e-01 8.33818316e-01 1.25308728e+00 -2.09165275e-01 -2.61395752e-01 -4.73713487e-01 -1.10239744e+00 7.21438468e-01 1.06453681e+00 -7.80623317e-01 -2.93777943e-01 -3.09396684e-01 -2.63020486e-01 -7.41578043e-01 2.60542274e-01 -1.17980874e+00 3.28735918e-01 -2.24149004e-01 4.59156096e-01 -8.47621739e-01 2.86969453e-01 -9.91546154e-01 -1.32206142e-01 4.60680991e-01 -3.08490217e-01 8.89179930e-02 1.25747770e-01 5.82844734e-01 -4.58627194e-01 -4.88220185e-01 6.89904690e-01 -1.31635442e-01 -1.28581488e+00 8.76014456e-02 2.47503873e-02 -5.07707410e-02 1.27179921e+00 -4.41705704e-01 -4.97850180e-01 9.43803266e-02 -5.05555511e-01 1.40037462e-01 4.64094609e-01 7.11412787e-01 1.13335383e+00 -1.48628736e+00 -4.06590372e-01 2.81125456e-01 8.27391446e-01 -1.21904597e-01 2.26668522e-01 5.83754003e-01 -2.10765570e-01 3.04095298e-01 -1.31223232e-01 -4.31220770e-01 -1.43523645e+00 8.64522040e-01 7.12164864e-02 -3.34301256e-02 -8.28323126e-01 9.31154490e-01 3.73529762e-01 1.31509781e-01 6.71816468e-02 -3.56733322e-01 -6.57298803e-01 2.00179562e-01 8.77496064e-01 1.26438916e-01 1.27872080e-03 -8.94505501e-01 -6.94069445e-01 6.50421262e-01 -2.62311310e-01 3.87210429e-01 1.46254814e+00 -2.61611909e-01 -1.02171294e-01 4.65420812e-01 1.53485703e+00 1.08604990e-01 -1.18409467e+00 -3.96126330e-01 1.52852103e-01 -1.10664749e+00 2.21717656e-01 -3.92870307e-01 -1.15729392e+00 6.75812423e-01 7.47621417e-01 2.27063037e-02 1.22990704e+00 4.13673580e-01 5.45384049e-01 2.10391134e-01 2.39214778e-01 -1.03320098e+00 2.92708516e-01 1.84398204e-01 1.14797437e+00 -1.32744420e+00 2.33472019e-01 -4.56793547e-01 -8.71909916e-01 7.41347909e-01 8.58631730e-01 -4.25517052e-01 8.84797752e-01 -1.86891943e-01 -1.65013313e-01 -4.65854764e-01 -7.57118762e-01 -7.38677382e-01 5.05650818e-01 5.21428823e-01 8.10000598e-02 1.01299681e-01 -4.03600603e-01 5.05898297e-01 1.04641214e-01 -3.68625164e-01 3.02424580e-01 1.06375623e+00 -9.52297390e-01 -8.30184579e-01 -4.53033745e-01 5.85257411e-01 -3.57854486e-01 -9.58428010e-02 -3.82321477e-01 4.11009103e-01 4.26777989e-01 9.95274425e-01 3.00125092e-01 -4.47977394e-01 2.50684321e-01 -2.41563857e-01 -3.93261388e-02 -8.44455063e-01 -3.48152667e-01 9.05137807e-02 -2.66155869e-01 -5.86873114e-01 -4.07313645e-01 -6.02581441e-01 -1.24493456e+00 1.39059812e-01 -4.95974362e-01 2.41266862e-01 2.81759799e-01 4.06574398e-01 6.83345973e-01 4.16032553e-01 8.45682323e-01 -6.10282660e-01 -1.18692786e-01 -8.63865018e-01 -6.42586231e-01 7.21483409e-01 4.65018094e-01 -8.16543281e-01 -3.00080359e-01 -7.37530738e-02]
[9.753531455993652, -0.1449088752269745]
f2da0e59-9e91-492f-9649-3c124f4cff26
exploring-and-exploiting-multi-granularity
2208.08750
null
https://arxiv.org/abs/2208.08750v1
https://arxiv.org/pdf/2208.08750v1.pdf
Exploring and Exploiting Multi-Granularity Representations for Machine Reading Comprehension
Recently, the attention-enhanced multi-layer encoder, such as Transformer, has been extensively studied in Machine Reading Comprehension (MRC). To predict the answer, it is common practice to employ a predictor to draw information only from the final encoder layer which generates the coarse-grained representations of the source sequences, i.e., passage and question. The analysis shows that the representation of source sequence becomes more coarse-grained from finegrained as the encoding layer increases. It is generally believed that with the growing number of layers in deep neural networks, the encoding process will gather relevant information for each location increasingly, resulting in more coarse-grained representations, which adds the likelihood of similarity to other locations (referring to homogeneity). Such phenomenon will mislead the model to make wrong judgement and degrade the performance. In this paper, we argue that it would be better if the predictor could exploit representations of different granularity from the encoder, providing different views of the source sequences, such that the expressive power of the model could be fully utilized. To this end, we propose a novel approach called Adaptive Bidirectional Attention-Capsule Network (ABA-Net), which adaptively exploits the source representations of different levels to the predictor. Furthermore, due to the better representations are at the core for boosting MRC performance, the capsule network and self-attention module are carefully designed as the building blocks of our encoders, which provides the capability to explore the local and global representations, respectively. Experimental results on three benchmark datasets, i.e., SQuAD 1.0, SQuAD 2.0 and COQA, demonstrate the effectiveness of our approach. In particular, we set the new state-of-the-art performance on the SQuAD 1.0 dataset
['Chenyu You', 'Nuo Chen']
2022-08-18
null
null
null
null
['machine-reading-comprehension']
['natural-language-processing']
[ 2.14367434e-01 1.90845668e-01 -5.60062751e-02 -2.62331665e-01 -5.31950772e-01 -3.50344777e-01 2.88232207e-01 3.30683023e-01 -2.12128267e-01 5.76557934e-01 6.43184900e-01 -3.76507640e-01 8.58137831e-02 -9.93038356e-01 -9.04437840e-01 -6.67928517e-01 2.03260511e-01 4.37933244e-02 3.63994092e-01 -4.81354356e-01 1.97844759e-01 -1.05828501e-01 -1.68393362e+00 7.51514792e-01 1.18670607e+00 1.04595828e+00 5.99211156e-01 4.36781704e-01 -1.20227791e-01 1.26323533e+00 -6.16043508e-01 -6.13834977e-01 -2.45830148e-01 -6.60017729e-01 -1.03431404e+00 -2.39559755e-01 1.12946197e-01 -5.00580251e-01 -3.80812466e-01 1.05922985e+00 3.43692183e-01 1.80216298e-01 4.33197677e-01 -7.28268206e-01 -9.83734131e-01 1.00183976e+00 -5.02015948e-01 3.44920039e-01 4.17553395e-01 1.85365543e-01 1.28302753e+00 -7.72847414e-01 3.61351371e-01 1.33436346e+00 2.36900032e-01 3.13403398e-01 -6.64388478e-01 -5.21495581e-01 5.17638445e-01 6.80637062e-01 -1.22453153e+00 -2.74594486e-01 8.83758366e-01 -1.23228185e-01 6.84937656e-01 3.65481853e-01 4.59951580e-01 1.38564456e+00 2.60230899e-01 1.20289636e+00 1.07627594e+00 -3.81959468e-01 7.58451074e-02 5.97490408e-02 3.59388709e-01 5.80592096e-01 -5.61699048e-02 -1.09674647e-01 -4.88372117e-01 9.45789590e-02 5.45822144e-01 1.12633683e-01 -8.69615376e-01 -5.67382015e-02 -1.08421886e+00 8.32234621e-01 7.49903738e-01 4.30609882e-01 -6.02614999e-01 -2.83487111e-01 3.99329752e-01 2.32987389e-01 1.76187664e-01 3.00363988e-01 -5.44538558e-01 -1.76417112e-01 -6.81093276e-01 1.29289076e-01 5.85702121e-01 9.81993079e-01 5.22905946e-01 -3.11150432e-01 -5.96046448e-01 7.75581181e-01 2.40763381e-01 -3.05068213e-02 8.21231663e-01 -5.39652824e-01 7.76342928e-01 8.04755807e-01 -1.02389313e-01 -1.17406309e+00 -2.03804195e-01 -1.09323263e+00 -1.15618730e+00 -3.15807045e-01 3.12623292e-01 1.14046447e-01 -6.83404505e-01 1.84906161e+00 1.11231253e-01 1.91858783e-02 1.77833572e-01 1.04762185e+00 7.92041779e-01 7.68479109e-01 -4.89208810e-02 5.77848265e-03 1.51101828e+00 -1.26210964e+00 -8.18824768e-01 -2.74873197e-01 4.46704507e-01 -4.89460945e-01 1.32535183e+00 2.80148029e-01 -1.23475003e+00 -1.07504427e+00 -1.18357670e+00 -2.56821603e-01 -1.76117599e-01 -8.72891769e-02 2.30105653e-01 4.06352840e-02 -6.64780617e-01 6.07468665e-01 -4.93197799e-01 -3.33866943e-03 2.98991293e-01 -2.20750153e-01 4.69187535e-02 -2.44620070e-01 -1.77134514e+00 8.50375712e-01 5.67261159e-01 2.15298176e-01 -8.65061939e-01 -4.90485400e-01 -6.27438962e-01 6.11816347e-01 3.32294136e-01 -6.10621154e-01 1.18330920e+00 -1.20275295e+00 -1.30745900e+00 4.38541889e-01 -2.79771298e-01 -4.10004705e-01 3.58932555e-01 -4.16604161e-01 -3.14926654e-01 2.48053297e-01 2.73263138e-02 5.11020660e-01 8.76999319e-01 -1.11536860e+00 -6.26338005e-01 -2.03145891e-01 4.77233410e-01 2.54132152e-01 -2.86551803e-01 -1.82146728e-01 -2.78181046e-01 -6.72802031e-01 -4.30943742e-02 -5.44902980e-01 1.28787264e-01 -4.05177206e-01 -4.04277802e-01 -4.72312242e-01 4.73111063e-01 -9.73141730e-01 1.51991773e+00 -2.28390288e+00 5.00658929e-01 -5.81323542e-02 3.03875476e-01 2.74478555e-01 -1.63392305e-01 4.08748180e-01 -1.00170843e-01 1.98766008e-01 -8.60564783e-02 -4.17733826e-02 -1.87804520e-01 1.18701346e-01 -4.31504160e-01 1.19646065e-01 3.34577680e-01 1.01406717e+00 -1.09087193e+00 -2.41073474e-01 -2.88079947e-01 3.32807571e-01 -6.08076870e-01 5.09212315e-01 -3.34101081e-01 4.12501127e-01 -5.75768471e-01 3.67153168e-01 6.21857643e-01 -6.76452875e-01 -5.08181602e-02 -2.81449288e-01 -4.89254296e-03 6.34128213e-01 -7.75294125e-01 1.62926757e+00 -4.40525800e-01 3.46063912e-01 -2.03993335e-01 -9.32385504e-01 9.63647485e-01 3.98793310e-01 -1.79584876e-01 -9.83902156e-01 1.00790404e-01 1.26097873e-01 2.91496933e-01 -6.73146307e-01 5.27739704e-01 1.91670001e-01 2.70017069e-02 2.51210123e-01 -8.29280987e-02 6.27488136e-01 1.18448073e-02 2.69860089e-01 9.43257570e-01 1.71009824e-01 2.96014607e-01 -1.13638096e-01 7.83899903e-01 -2.46490225e-01 4.16075259e-01 6.03098631e-01 1.57338344e-02 4.88199025e-01 6.96896076e-01 -2.51323193e-01 -8.39955389e-01 -7.80140281e-01 5.29047810e-02 1.33350265e+00 2.77741432e-01 -4.26463246e-01 -8.27539206e-01 -7.45366633e-01 -2.61051446e-01 8.49628210e-01 -8.16911757e-01 -6.28769517e-01 -7.41125047e-01 -3.80005002e-01 4.01872039e-01 6.80441916e-01 7.41989195e-01 -1.18307889e+00 -8.42541158e-01 4.30017471e-01 -6.65954590e-01 -8.58747065e-01 -3.72538358e-01 1.32499069e-01 -6.75710678e-01 -9.25785005e-01 -8.47221196e-01 -7.35636652e-01 4.62563723e-01 2.09172398e-01 1.29654002e+00 4.47235674e-01 3.83615524e-01 -1.76368743e-01 -9.61612165e-01 -1.42892778e-01 -3.12192947e-01 3.45186710e-01 -4.49618697e-01 3.06651592e-01 3.26002926e-01 -4.21113759e-01 -7.78322935e-01 2.55851954e-01 -9.05091882e-01 4.09400642e-01 8.09856772e-01 9.61678922e-01 3.81958783e-01 1.66111378e-04 8.30796301e-01 -6.84911370e-01 8.33763659e-01 -8.96450400e-01 -9.03183967e-02 4.26624089e-01 -1.29717708e-01 1.77095279e-01 1.13688421e+00 -4.70002860e-01 -1.18891680e+00 -6.42925203e-01 -3.86185288e-01 -3.78066540e-01 -5.87123074e-02 8.52651238e-01 -4.62618828e-01 4.57521170e-01 3.98621649e-01 6.53318286e-01 -3.02909970e-01 -6.98290884e-01 4.03012782e-01 8.10234785e-01 3.82925719e-01 -5.24533272e-01 3.41352880e-01 -5.03215753e-02 -5.85970044e-01 -3.41250420e-01 -1.18563688e+00 -1.63409218e-01 -4.28528756e-01 -9.74898413e-02 9.33238983e-01 -7.67470300e-01 -4.92193609e-01 3.95069420e-01 -1.38571656e+00 5.24648391e-02 -1.03662901e-01 2.09006116e-01 -2.97413856e-01 1.84833094e-01 -6.21307909e-01 -4.59607303e-01 -2.52537727e-01 -1.44190741e+00 8.37638855e-01 4.76164818e-01 -2.71348685e-01 -8.53650868e-01 -2.15144455e-01 4.02609646e-01 5.54946184e-01 3.74479555e-02 1.40050387e+00 -8.84789884e-01 -7.26255178e-01 2.62053281e-01 -3.13844949e-01 3.66024286e-01 -5.90786599e-02 -3.42465490e-01 -1.14837027e+00 -2.76099980e-01 3.95983189e-01 -4.32032675e-01 9.71616209e-01 -5.59059381e-02 1.65706491e+00 -5.44742644e-01 -8.39038938e-02 4.83362228e-01 1.20027268e+00 2.02734798e-01 8.91278863e-01 3.78550828e-01 6.18032157e-01 7.41189420e-01 5.14141500e-01 2.16852054e-01 6.87286019e-01 4.81198579e-01 6.37936532e-01 3.34952027e-02 -1.84039816e-01 -6.08949900e-01 3.04011494e-01 1.35244477e+00 2.37888904e-04 -5.15238702e-01 -6.36941493e-01 6.20880187e-01 -1.78661346e+00 -8.77891541e-01 3.18900421e-02 1.82741320e+00 9.71919000e-01 1.03581473e-01 -3.03971022e-01 2.05710873e-01 6.34784341e-01 3.18476796e-01 -6.59080327e-01 -4.17443305e-01 -2.40783375e-02 1.66072994e-01 -4.39245217e-02 1.64480582e-01 -7.80983388e-01 5.53229511e-01 4.87713051e+00 9.63275850e-01 -9.81346846e-01 1.24043420e-01 9.17254627e-01 2.22431093e-01 -4.57379699e-01 -1.48173511e-01 -7.82795191e-01 8.10128212e-01 1.00992811e+00 -9.81829166e-02 2.03499660e-01 5.38799226e-01 2.13217512e-02 1.25778601e-01 -1.14660490e+00 4.41337228e-01 3.79092135e-02 -1.11379659e+00 2.13900432e-01 -8.21907818e-02 5.36669672e-01 -1.31396145e-01 4.94928136e-02 5.32323062e-01 1.09505869e-01 -1.07010150e+00 8.53172302e-01 7.06819773e-01 4.08614278e-01 -8.69714558e-01 9.13518012e-01 7.40451634e-01 -1.13694620e+00 -2.43584722e-01 -6.07367396e-01 -5.82733750e-02 1.67789146e-01 4.70710963e-01 -2.90551901e-01 8.06956530e-01 7.50494182e-01 6.70288026e-01 -6.47533834e-01 8.42858672e-01 -5.30864775e-01 7.86730170e-01 2.21615031e-01 -7.74604380e-02 4.14345771e-01 1.49438426e-01 2.34293506e-01 1.02470529e+00 3.23141128e-01 1.82674661e-01 3.17818783e-02 8.31056416e-01 -3.27227056e-01 1.02876015e-01 -1.60156414e-01 2.12031081e-01 4.98908788e-01 9.35639560e-01 -1.81535453e-01 -5.19350946e-01 -6.06532812e-01 8.86407912e-01 8.16680551e-01 4.03251737e-01 -9.83865678e-01 -4.56833333e-01 4.23399538e-01 1.14834316e-01 6.04684770e-01 1.65481538e-01 -1.11700103e-01 -1.27190256e+00 1.04693763e-01 -1.39609480e+00 4.64118958e-01 -7.57501841e-01 -1.29544532e+00 8.52499902e-01 -2.04126254e-01 -1.05343091e+00 -1.17354982e-01 -2.95356929e-01 -6.56336546e-01 1.12435400e+00 -1.80703151e+00 -8.84816349e-01 -3.32103342e-01 4.69063848e-01 7.93619156e-01 6.88142702e-02 5.77957749e-01 2.55602449e-01 -4.96061802e-01 7.75765479e-01 -2.79085469e-02 2.20051005e-01 3.53748292e-01 -9.79383469e-01 2.15627328e-01 8.67679358e-01 4.26441878e-02 8.36496890e-01 3.84181529e-01 -3.71646017e-01 -1.11770904e+00 -9.49007988e-01 9.62789536e-01 -8.07543471e-02 5.94481468e-01 -1.91138119e-01 -1.61618340e+00 5.71225882e-01 6.11032546e-01 -3.30131799e-01 6.29951835e-01 3.22874412e-02 -4.12295789e-01 -1.11938499e-01 -7.84392357e-01 5.07941782e-01 8.96535099e-01 -6.19318128e-01 -9.70845819e-01 -1.07521554e-02 1.15965450e+00 -3.35331559e-01 -8.92658234e-01 3.69713098e-01 3.79248708e-01 -1.19545305e+00 8.21202874e-01 -5.38171589e-01 8.86682093e-01 -2.00135455e-01 -1.67205468e-01 -1.52903676e+00 -7.05050647e-01 -1.66866302e-01 -4.75808024e-01 1.38624239e+00 3.38817239e-01 -5.18753290e-01 2.31279850e-01 1.24753430e-01 -2.74535626e-01 -1.22705162e+00 -8.29614103e-01 -3.70288163e-01 2.29477063e-01 -5.95786460e-02 1.05230927e+00 6.89422965e-01 -4.85511050e-02 5.49292743e-01 -2.94714361e-01 1.95156708e-01 1.87457755e-01 3.48054200e-01 2.44850203e-01 -9.71908331e-01 -5.44561863e-01 -4.42997247e-01 4.06457521e-02 -1.80125415e+00 -8.01532250e-03 -8.77678275e-01 -4.59371842e-02 -1.37223566e+00 2.37377599e-01 -3.37656230e-01 -6.62433863e-01 3.29175979e-01 -7.66961157e-01 -3.44868004e-01 2.48150438e-01 3.42331707e-01 -5.26918054e-01 7.91956007e-01 1.66352534e+00 -1.70215502e-01 2.09605068e-01 -1.44022167e-01 -1.09574044e+00 6.68949127e-01 8.72277021e-01 -1.99546471e-01 -6.27211630e-01 -8.70675981e-01 2.85151720e-01 1.04428820e-01 4.05162215e-01 -7.24374831e-01 1.69018134e-01 1.52519405e-01 4.81625110e-01 -6.12119734e-01 1.01724058e-01 -6.80371702e-01 -2.41407827e-01 3.23672980e-01 -7.59263277e-01 1.94486275e-01 -2.89288871e-02 4.71748650e-01 -4.71689492e-01 -3.97895128e-01 6.46458387e-01 -3.22216570e-01 -5.80295980e-01 1.18056156e-01 -1.06020778e-01 2.85351306e-01 6.92334652e-01 9.17812064e-02 -5.08499920e-01 -4.31361705e-01 -4.78726298e-01 4.21173960e-01 3.01578939e-01 7.07681656e-01 7.34414518e-01 -1.31160223e+00 -8.98999453e-01 2.83613414e-01 1.69526696e-01 2.00240806e-01 4.64335084e-01 6.85155571e-01 4.70427237e-03 4.83257324e-01 -6.61197454e-02 -5.17946661e-01 -8.32126260e-01 8.07807863e-01 2.34789237e-01 -5.93353093e-01 -6.20083988e-01 9.41492140e-01 5.48259854e-01 -7.34162033e-02 2.80713230e-01 -5.90461373e-01 -6.98476434e-01 2.05968156e-01 8.53039801e-01 1.71395734e-01 -4.31698412e-02 -5.55229366e-01 -1.29132926e-01 3.84310275e-01 -5.05484045e-01 3.89786631e-01 1.13001442e+00 -3.02041978e-01 -1.74841300e-01 4.24049586e-01 1.29005158e+00 -1.12150244e-01 -1.29300225e+00 -3.99641812e-01 -3.51303928e-02 -3.15119535e-01 -8.28498825e-02 -8.66317213e-01 -9.85502422e-01 1.23221242e+00 1.80033684e-01 2.91834980e-01 1.29317033e+00 1.03416592e-01 1.03810310e+00 -1.94626711e-02 2.76675671e-01 -6.52497172e-01 1.50243476e-01 5.40411472e-01 1.05974054e+00 -8.65694821e-01 -4.83988553e-01 -1.13844283e-01 -7.80638516e-01 1.05771554e+00 8.34427238e-01 2.10267473e-02 1.81001857e-01 -3.28157991e-02 -2.13295594e-01 -4.88653444e-02 -1.00092232e+00 1.85693670e-02 3.47126931e-01 1.98370546e-01 5.42856574e-01 8.12078938e-02 -9.15189013e-02 1.06068230e+00 -2.71817714e-01 -1.27261624e-01 4.40184504e-01 7.53009737e-01 -6.98413849e-01 -9.21949744e-01 -2.57269472e-01 4.20851886e-01 -5.03156841e-01 -2.98869729e-01 -1.12397105e-01 4.27131623e-01 3.08644295e-01 9.74691749e-01 2.85622906e-02 -2.61849135e-01 3.88730496e-01 1.22334890e-01 1.67790949e-01 -3.47211480e-01 -6.59277976e-01 -2.38335907e-01 -4.38326709e-02 -5.80577075e-01 -2.82773137e-01 -3.86642009e-01 -1.14078617e+00 -1.51432022e-01 -3.33723277e-01 3.01970989e-01 6.85447007e-02 1.04414213e+00 4.91464317e-01 1.08095467e+00 6.44339621e-01 -2.07424700e-01 -8.07870448e-01 -1.10435581e+00 -1.15591429e-01 3.95762801e-01 5.23155153e-01 -4.99314219e-01 -2.61685878e-01 -1.58326909e-01]
[11.314998626708984, 8.199069023132324]
06650a30-6363-4b9c-a647-fe97289a7cc9
hybrid-code-networks-using-a-convolutional
1907.12162
null
https://arxiv.org/abs/1907.12162v1
https://arxiv.org/pdf/1907.12162v1.pdf
Hybrid Code Networks using a convolutional neural network as an input layer achieves higher turn accuracy
The dialogue management is a task of conversational artificial intelligence. The goal of the dialogue manager is to select the appropriate response to the conversational partner conditioned by the input message and recent dialogue state. Hybrid Code Networks is one of the models of dialogue managers, which uses an average of word embeddings and bag-of-words as input features. We perform experiments on Dialogue bAbI Task 6 and Alquist Conversational Dataset. The experiments show that the convolutional neural network used as an input layer of the Hybrid Code Network improves the model's turn accuracy.
['Petr Marek']
2019-07-28
null
null
null
null
['dialogue-management']
['natural-language-processing']
[-3.28233778e-01 5.68770349e-01 -1.06849736e-02 -8.11664581e-01 1.62325241e-02 -3.44831765e-01 9.42881286e-01 -1.66641325e-01 -4.34363186e-01 7.34470844e-01 7.63698995e-01 -4.45842326e-01 2.47008249e-01 -8.14323008e-01 2.16731861e-01 -3.72936070e-01 -1.08272485e-01 9.01036680e-01 9.47914869e-02 -9.30432141e-01 3.32153261e-01 1.04793288e-01 -9.42803085e-01 7.08977222e-01 4.91543531e-01 9.05818820e-01 1.27416745e-01 1.44903064e+00 -7.99104273e-01 1.44757450e+00 -8.97667885e-01 -3.37119579e-01 -1.32864907e-01 -7.35978305e-01 -1.47423089e+00 -2.73236930e-01 -4.79004860e-01 -4.94691700e-01 -8.03165555e-01 6.13754272e-01 6.02401376e-01 5.63761175e-01 6.58815563e-01 -1.30356872e+00 -3.12854946e-01 6.98656976e-01 4.49652702e-01 4.38800961e-01 7.33680546e-01 3.31898540e-01 1.00869930e+00 -8.29628468e-01 5.34774303e-01 1.80120850e+00 4.69959944e-01 1.09937167e+00 -1.02532113e+00 -3.97007048e-01 -1.38337359e-01 2.24028111e-01 -5.07718682e-01 -6.51915193e-01 5.63462913e-01 -3.88429850e-01 1.49796951e+00 1.38197333e-01 5.19203007e-01 1.20606589e+00 4.09583151e-01 9.59981620e-01 7.18142211e-01 -3.81755561e-01 5.77246286e-02 4.32159483e-01 5.64245760e-01 6.96409106e-01 -9.76903260e-01 -5.94168305e-02 -5.53924620e-01 -6.54173434e-01 3.88486862e-01 -1.94779024e-01 -1.61973268e-01 8.07251558e-02 -8.60938907e-01 1.53445721e+00 4.90533352e-01 2.84792215e-01 -2.86219060e-01 -2.12637663e-01 1.01962626e+00 7.12717652e-01 4.01019841e-01 7.11611748e-01 -4.81251001e-01 -8.35687459e-01 7.97867104e-02 4.29747164e-01 1.72534287e+00 6.40235364e-01 4.56369877e-01 -7.51657113e-02 -3.41198176e-01 1.60814834e+00 4.09874648e-01 1.86515413e-02 6.33524299e-01 -7.11029172e-01 5.05766869e-01 9.82398450e-01 -1.18536480e-01 -8.88338208e-01 -6.73771977e-01 4.76289570e-01 -8.33994031e-01 9.68592316e-02 3.47646981e-01 -7.98846781e-01 -3.97983611e-01 1.41931832e+00 2.44601697e-01 -4.87734407e-01 7.33897746e-01 7.02844441e-01 1.37648606e+00 1.06912601e+00 -2.11665183e-01 1.42720342e-01 1.27131438e+00 -1.50364470e+00 -1.09472334e+00 -2.82496125e-01 7.68452346e-01 -7.38682568e-01 7.97765434e-01 6.46469891e-02 -8.06213498e-01 -6.18956566e-01 -9.29589272e-01 -1.63871676e-01 -5.20518005e-01 -1.65317088e-01 7.76515841e-01 3.55857044e-01 -8.81018341e-01 2.24571869e-01 -1.71873465e-01 -6.14965022e-01 -1.62585184e-01 2.60044426e-01 -4.29185212e-01 3.41153532e-01 -1.74187124e+00 1.43843496e+00 4.76505607e-01 7.26982430e-02 -5.11913896e-01 4.05682288e-02 -9.70418334e-01 1.41024455e-01 -1.49953827e-01 -2.81376064e-01 1.80813229e+00 -9.96669888e-01 -2.20404911e+00 6.84531212e-01 1.45738095e-01 -8.37970853e-01 3.15868080e-01 -2.09211353e-02 -3.93044323e-01 5.76832294e-02 -5.13508618e-01 7.75105298e-01 2.56365567e-01 -7.75889456e-01 -7.78528512e-01 9.60636288e-02 4.92508173e-01 6.52652085e-01 -2.29486562e-02 -6.18836209e-02 9.11508724e-02 6.26066029e-02 -2.32911959e-01 -1.01505685e+00 -1.04400620e-01 -4.33835447e-01 -2.72630423e-01 -9.60974574e-01 9.42398965e-01 -6.37783945e-01 1.21079385e+00 -1.99731445e+00 2.14379072e-01 -2.22138330e-01 1.17882900e-01 4.80379522e-01 -9.18847695e-02 9.19357181e-01 7.90742040e-02 -1.40417874e-01 1.99554846e-01 -4.94496748e-02 2.86940653e-02 2.81283200e-01 -2.12534577e-01 -3.51996757e-02 1.95118815e-01 6.13511384e-01 -7.09708869e-01 -1.37761369e-01 3.33941311e-01 -4.00499161e-03 -4.60815459e-01 1.30749011e+00 -3.47794861e-01 2.97656000e-01 -3.00993830e-01 -1.48079187e-01 2.25540474e-01 9.16572586e-02 2.72191882e-01 1.04395568e-01 1.79031193e-02 1.12116587e+00 -5.39795399e-01 1.42814887e+00 -6.39324248e-01 1.03826392e+00 2.22346649e-01 -6.94653094e-01 1.38693523e+00 5.08464634e-01 -1.80339497e-02 -4.16221917e-01 3.30349237e-01 -2.87346333e-01 7.86444485e-01 -9.63263631e-01 5.96984029e-01 3.74564342e-02 -5.05575120e-01 8.01901937e-01 3.43395442e-01 -2.12488919e-01 -1.46087617e-01 3.71991605e-01 1.18776202e+00 -5.19731939e-01 4.42342609e-01 -2.18845502e-01 7.65764117e-01 4.88015525e-02 3.87888163e-01 5.94693184e-01 -3.61744940e-01 1.03691578e-01 1.09109652e+00 -9.65387642e-01 -1.05209577e+00 -5.46823919e-01 2.60085333e-02 1.50980473e+00 -4.06880319e-01 -3.07731688e-01 -6.80121124e-01 -8.52258146e-01 -2.63574952e-03 8.39641392e-01 -6.12820923e-01 -1.45770505e-01 -5.82068443e-01 -4.70963538e-01 6.41987145e-01 3.39897811e-01 8.60233605e-01 -1.43263435e+00 -3.11549723e-01 5.85558414e-01 -2.53016710e-01 -7.33972669e-01 -4.07723963e-01 2.78909385e-01 -1.87336177e-01 -1.25924039e+00 -3.11654508e-02 -1.01710165e+00 1.22450128e-01 -1.28392845e-01 9.18333054e-01 1.37260333e-01 8.50762799e-02 1.97130561e-01 -6.22933507e-01 -1.80664465e-01 -1.15254676e+00 3.80497873e-01 1.64451338e-02 -4.40949537e-02 5.73734283e-01 -1.73773974e-01 -3.88468027e-01 2.69668728e-01 -3.13187391e-01 4.54971254e-01 -8.11656043e-02 1.48311865e+00 -9.48992133e-01 -5.39705157e-01 8.94952238e-01 -1.02951074e+00 1.59776223e+00 -4.49399889e-01 9.25037079e-03 1.56605393e-01 -1.86241284e-01 1.54284582e-01 4.67511982e-01 -3.65079820e-01 -1.25792158e+00 -2.66162336e-01 -5.74030876e-01 5.27420819e-01 -8.89863912e-03 4.41206187e-01 -1.79678649e-01 1.98330641e-01 5.49571574e-01 1.46136194e-01 5.79003036e-01 -3.56818825e-01 3.16190511e-01 1.57831085e+00 1.64674953e-01 -4.27524835e-01 -3.15986782e-01 -3.37366104e-01 -7.87750185e-01 -8.64654005e-01 -2.91671813e-01 -4.56750661e-01 -6.06747091e-01 -5.27595460e-01 1.02392554e+00 -5.23931503e-01 -1.23489189e+00 6.34100735e-01 -1.67018020e+00 -4.74032223e-01 4.19365823e-01 2.69486994e-01 -5.84443510e-01 -2.03023981e-02 -7.96240389e-01 -1.00876784e+00 -4.77938354e-01 -1.21143758e+00 7.57129714e-02 4.20046538e-01 -7.92127073e-01 -1.00820792e+00 2.72586942e-01 4.48173106e-01 6.79892182e-01 -1.34991780e-01 1.27887917e+00 -1.79133499e+00 1.96781710e-01 -2.90583879e-01 -1.12709470e-01 5.62149346e-01 1.73087016e-01 2.47068908e-02 -1.07736242e+00 1.24536790e-01 1.19399711e-01 -6.94416821e-01 5.53783953e-01 -9.99832992e-03 5.79065859e-01 -6.26741707e-01 -5.76693602e-02 -3.59349512e-02 6.93338394e-01 9.19170737e-01 4.67767000e-01 -2.95297336e-02 3.96204740e-01 9.58219886e-01 4.61406291e-01 4.44163412e-01 6.79876387e-01 5.29928625e-01 1.75002307e-01 2.65300900e-01 2.61142761e-01 -1.12987399e-01 3.90015095e-01 1.20079505e+00 5.48829615e-01 -2.70460755e-01 -1.32981277e+00 3.73272866e-01 -2.11806297e+00 -9.49671865e-01 -2.53030360e-02 1.55963051e+00 1.03137934e+00 2.74314553e-01 8.97700414e-02 -2.77843922e-01 7.09818423e-01 5.09805977e-01 -4.65844780e-01 -1.49217451e+00 2.43473798e-01 -3.09079409e-01 -1.26983285e-01 9.18154120e-01 -9.61962342e-01 9.51528490e-01 6.65546417e+00 3.12931806e-01 -7.49808431e-01 1.03700437e-01 7.11937249e-01 1.25857964e-01 2.76508540e-01 -2.92268135e-02 -7.68367290e-01 4.09337372e-01 1.39153779e+00 -2.09969178e-01 6.03626609e-01 8.20692062e-01 -4.24601249e-02 -3.54183108e-01 -1.30626690e+00 7.12030292e-01 1.41381294e-01 -1.31577420e+00 -2.62553453e-01 -2.57211983e-01 2.05724269e-01 -5.00360169e-02 -3.96299362e-01 1.04088616e+00 9.40388620e-01 -1.16817260e+00 -1.06949143e-01 6.69015348e-01 1.84904650e-01 -9.75835979e-01 1.08162856e+00 4.99365538e-01 -6.82945371e-01 -2.66724437e-01 -4.10579294e-01 -4.45043325e-01 8.23904648e-02 -1.51774406e-01 -1.73813820e+00 -3.09299141e-01 3.87505263e-01 1.87665984e-01 -2.15300918e-03 5.09187758e-01 -2.82663833e-02 5.54111958e-01 5.56388982e-02 -7.21084714e-01 5.96178710e-01 -1.85530439e-01 4.32476372e-01 1.50827277e+00 -4.82581377e-01 3.38275105e-01 4.37824875e-01 4.75147694e-01 -1.45759419e-01 1.39030322e-01 -9.04765129e-01 -3.09479803e-01 5.32824099e-01 1.12319982e+00 4.33174782e-02 -3.56939107e-01 -3.64626944e-01 7.35655248e-01 5.46789587e-01 1.38444051e-01 -3.65821511e-01 -9.23912168e-01 1.04047775e+00 -6.40778124e-01 -1.47851124e-01 -1.44406900e-01 -1.18178025e-01 -6.10830486e-01 -5.20633340e-01 -1.18209052e+00 2.85106987e-01 -7.20513880e-01 -1.28239369e+00 1.00230551e+00 -2.83416897e-01 -7.86337197e-01 -8.06086898e-01 -7.53060758e-01 -1.06063914e+00 1.26047492e+00 -8.02153885e-01 -7.16762006e-01 -1.44351855e-01 2.23012447e-01 1.17314339e+00 -9.75346267e-01 1.60730004e+00 -3.69576454e-01 -5.06934106e-01 3.40006679e-01 1.05316862e-01 9.17316854e-01 7.37332046e-01 -1.34802663e+00 7.09758818e-01 -2.83694297e-01 -5.95745921e-01 4.85470980e-01 7.05537975e-01 -3.29749346e-01 -1.14941013e+00 -4.34168667e-01 1.17947376e+00 -2.82513797e-01 7.62800515e-01 -6.00818574e-01 -8.83768380e-01 4.43239123e-01 1.09780967e+00 -6.74737155e-01 9.76026654e-01 3.97916883e-01 1.28524154e-02 4.38695885e-02 -1.25519693e+00 5.29038548e-01 3.20291728e-01 -8.26828420e-01 -1.14510262e+00 3.97246897e-01 9.05358851e-01 -6.45448983e-01 -9.88832355e-01 -3.71224359e-02 5.16590536e-01 -1.06555581e+00 5.03332734e-01 -1.14319813e+00 6.61806941e-01 5.36315799e-01 2.78510880e-02 -1.78144455e+00 -1.01035684e-02 -6.26079142e-01 9.54873413e-02 1.25759256e+00 5.00976324e-01 -6.29688561e-01 3.28163892e-01 9.68411148e-01 9.54165235e-02 -7.18675494e-01 -1.03986084e+00 -7.10438937e-03 8.02671239e-02 9.97209400e-02 6.85961604e-01 9.04103398e-01 8.47933769e-01 1.19872642e+00 -3.23499113e-01 -6.23046279e-01 -4.57949013e-01 -4.33922894e-02 1.17611063e+00 -1.11778092e+00 -7.11006448e-02 -4.08350021e-01 -3.29400748e-01 -1.41369820e+00 4.73178208e-01 -7.13831782e-01 3.56064111e-01 -1.44536018e+00 -1.98418617e-01 -2.24537164e-01 2.04678267e-01 3.24426144e-01 2.08541169e-03 -7.71329522e-01 9.54337269e-02 -1.30847484e-01 -2.37083301e-01 7.89327800e-01 9.59983528e-01 -4.45230484e-01 -4.91476983e-01 5.16879499e-01 -2.79380262e-01 6.33846164e-01 1.03430212e+00 -2.25464433e-01 -4.61761624e-01 -2.70721819e-02 3.40991765e-02 7.03043520e-01 -1.20483413e-01 -5.69375992e-01 5.27611852e-01 -1.25220075e-01 4.83882427e-02 -5.80386341e-01 6.34004176e-01 -5.20334601e-01 -4.60423380e-01 3.98197502e-01 -1.13222969e+00 3.13460022e-01 1.65570661e-01 3.39001805e-01 -2.81718373e-01 -5.82691252e-01 8.26683462e-01 -3.16612333e-01 -5.51022887e-01 -2.64749676e-01 -1.06709659e+00 -1.04823394e-03 6.14439607e-01 2.37482756e-01 -4.36313480e-01 -8.76901448e-01 -6.83366358e-01 6.82308137e-01 -7.30508491e-02 9.45035875e-01 6.83635831e-01 -1.32701933e+00 -6.87249601e-01 2.15451702e-01 1.73461571e-01 -2.95683891e-01 4.18692827e-04 3.63738507e-01 -7.91643798e-01 5.74283183e-01 -4.62727606e-01 -3.39561462e-01 -1.47289252e+00 -1.56828299e-01 8.08114409e-01 -2.05013439e-01 -4.69531864e-01 1.06355870e+00 -1.87775820e-01 -1.26125014e+00 7.38261104e-01 -1.17592014e-01 -8.80714118e-01 2.63767511e-01 8.02282274e-01 2.17181757e-01 -1.92721292e-01 -5.28981626e-01 -9.74832997e-02 -6.42872274e-01 -5.98174751e-01 -3.73161584e-01 1.31391621e+00 -5.08781299e-02 -4.09986705e-01 9.85663116e-01 1.34511304e+00 -6.76743209e-01 -1.01635230e+00 -5.17559111e-01 1.45205975e-01 -2.31315956e-01 -2.07582071e-01 -9.08725798e-01 -5.10175586e-01 1.07726455e+00 5.37676752e-01 7.50002563e-01 2.29365543e-01 -4.62083817e-01 7.57261753e-01 8.32096577e-01 1.10742502e-01 -1.40921795e+00 1.29595622e-01 1.43895793e+00 1.33398855e+00 -1.41540337e+00 -4.17661935e-01 3.09445381e-01 -1.31862068e+00 1.58243334e+00 1.16342330e+00 -9.73076671e-02 8.65033269e-01 2.06576094e-01 5.82714498e-01 -1.59330130e-01 -1.44049680e+00 1.77617654e-01 -2.01798707e-01 5.62882662e-01 6.68219447e-01 1.70193151e-01 -4.76448327e-01 6.53131604e-01 -4.55896437e-01 -5.26347697e-01 6.09480441e-01 7.86834419e-01 -7.14657843e-01 -9.94589508e-01 -4.15204763e-02 6.06543660e-01 -4.93861251e-02 1.18612582e-02 -1.05178452e+00 4.26570684e-01 -4.89459187e-01 1.44530082e+00 3.07921320e-01 -9.03015435e-01 3.61604631e-01 8.50826144e-01 -2.58054495e-01 -8.72060359e-01 -1.38844490e+00 -6.51862144e-01 9.13806736e-01 -2.01783285e-01 -1.89714149e-01 -1.77304968e-01 -1.48198879e+00 -6.82255447e-01 -3.21274430e-01 5.88900208e-01 6.43466294e-01 9.76099551e-01 1.84443906e-01 4.81748730e-01 1.10523558e+00 -5.67333758e-01 -7.44551122e-01 -1.83114326e+00 -2.95499325e-01 3.02637666e-01 3.80125374e-01 -4.53051180e-01 -2.68608660e-01 -3.33858639e-01]
[12.891552925109863, 7.916103363037109]
e497bdcf-2245-4fcc-b2b1-05b631351541
synthetic-data-generation-method-for-data
2301.04338
null
https://arxiv.org/abs/2301.04338v2
https://arxiv.org/pdf/2301.04338v2.pdf
Synthetic data generation method for data-free knowledge distillation in regression neural networks
Knowledge distillation is the technique of compressing a larger neural network, known as the teacher, into a smaller neural network, known as the student, while still trying to maintain the performance of the larger neural network as much as possible. Existing methods of knowledge distillation are mostly applicable for classification tasks. Many of them also require access to the data used to train the teacher model. To address the problem of knowledge distillation for regression tasks under the absence of original training data, previous work has proposed a data-free knowledge distillation method where synthetic data are generated using a generator model trained adversarially against the student model. These synthetic data and their labels predicted by the teacher model are then used to train the student model. In this study, we investigate the behavior of various synthetic data generation methods and propose a new synthetic data generation strategy that directly optimizes for a large but bounded difference between the student and teacher model. Our results on benchmark and case study experiments demonstrate that the proposed strategy allows the student model to learn better and emulate the performance of the teacher model more closely.
['Keng-Hwee Chiam', 'Tianxun Zhou']
2023-01-11
null
null
null
null
['synthetic-data-generation', 'synthetic-data-generation']
['medical', 'miscellaneous']
[ 5.44382513e-01 7.52823591e-01 -1.29116982e-01 -1.99929982e-01 -4.13860530e-01 -6.71505332e-01 4.86930758e-01 9.93251652e-02 -5.06545901e-01 1.16179490e+00 -3.04395795e-01 -2.97135890e-01 2.71068722e-01 -1.21432304e+00 -1.06527066e+00 -8.63397956e-01 3.74863386e-01 7.30626285e-01 1.39684588e-01 -1.39270425e-01 -6.17752038e-02 5.28696120e-01 -1.49457443e+00 7.34646544e-02 1.31867194e+00 8.30672801e-01 1.39620425e-02 7.93801546e-01 2.74215192e-01 9.84183669e-01 -1.27525651e+00 -4.26087946e-01 5.77398539e-01 -6.40330076e-01 -8.41080904e-01 -2.97045738e-01 4.59162295e-01 -8.12737942e-02 -4.73389357e-01 9.69374776e-01 4.04100090e-01 4.59057331e-01 5.80306709e-01 -1.41472948e+00 -7.36770928e-01 7.98508763e-01 -1.25959078e-02 -3.17915864e-02 -9.75677222e-02 1.39338344e-01 5.01335263e-01 -3.21842700e-01 4.27611440e-01 1.00314748e+00 4.37893361e-01 9.32982564e-01 -1.54915595e+00 -1.02143538e+00 -2.93161690e-01 -2.88410373e-02 -1.48069215e+00 -2.50410110e-01 7.92646646e-01 -2.29761943e-01 7.09396362e-01 1.69274583e-01 6.36846721e-01 1.00903583e+00 -1.17438786e-01 7.84870684e-01 1.17167473e+00 -4.86322373e-01 2.58625418e-01 6.61455870e-01 -1.33372143e-01 7.58031964e-01 4.44267750e-01 3.70149642e-01 -2.42923155e-01 -3.12187642e-01 5.80864429e-01 -2.17448816e-01 -4.99299169e-01 -3.98797870e-01 -9.41375315e-01 8.96999657e-01 4.79555726e-01 -4.35255393e-02 -3.75142843e-01 1.62743956e-01 2.12686971e-01 6.44298732e-01 3.81542414e-01 1.00009489e+00 -6.66940093e-01 1.14831880e-01 -1.21982419e+00 5.26460469e-01 1.11372662e+00 8.10737193e-01 9.22025919e-01 5.63231707e-01 -3.64119381e-01 5.55304646e-01 -8.57001841e-02 4.83412325e-01 1.04659283e+00 -7.56999791e-01 4.51372683e-01 7.47938275e-01 -1.28315896e-01 -6.70236349e-01 2.19633073e-01 -5.91973960e-01 -7.37836242e-01 4.01712567e-01 4.03096318e-01 -4.98180389e-01 -1.09706175e+00 1.94821608e+00 3.79189014e-01 7.61972129e-01 6.31521523e-01 2.92520165e-01 6.15591168e-01 7.22221434e-01 -1.08037740e-01 -9.79674235e-02 5.02529025e-01 -1.16108143e+00 -1.68034747e-01 -7.70214200e-02 6.90669417e-01 -3.80558491e-01 8.52062881e-01 4.72114712e-01 -1.33174956e+00 -7.45325327e-01 -1.25240934e+00 2.50924408e-01 -6.55113280e-01 2.78124064e-02 4.74326089e-02 7.74763107e-01 -9.85506535e-01 8.35714638e-01 -7.16614664e-01 2.48319373e-01 5.24566352e-01 5.18038750e-01 -1.31043464e-01 3.26584578e-02 -1.35018849e+00 1.01419330e+00 1.09829712e+00 -2.51769453e-01 -1.26000643e+00 -9.36424196e-01 -8.94055188e-01 2.95817345e-01 3.48428011e-01 -6.31055892e-01 1.43937361e+00 -1.23100650e+00 -1.44970894e+00 5.25873005e-01 3.16338211e-01 -8.10888588e-01 5.94351411e-01 -2.06866451e-02 -1.23201050e-01 -1.36429191e-01 -2.72095233e-01 7.83169985e-01 9.30671096e-01 -1.40101445e+00 -4.83670086e-01 8.56988728e-02 -8.12749490e-02 7.74780959e-02 -2.51136810e-01 -7.03563094e-01 -4.51303758e-02 -9.62126374e-01 -4.14710343e-01 -1.01042175e+00 -1.62402406e-01 -4.91802245e-01 -6.47542059e-01 -1.45558909e-01 1.14636075e+00 -4.75453079e-01 1.15836561e+00 -1.95886922e+00 1.75644815e-01 5.48435509e-01 1.53604627e-01 9.47331071e-01 -3.63634229e-01 1.36734217e-01 -5.41824877e-01 1.15149572e-01 -4.12129372e-01 -1.35701820e-01 -2.90248781e-01 6.66576624e-01 -5.29594123e-01 1.46611601e-01 3.03597182e-01 1.01368916e+00 -1.07516384e+00 -3.00990880e-01 -1.98248342e-01 2.63204068e-01 -7.31699049e-01 7.28929758e-01 -4.62908208e-01 3.95019680e-01 -3.48594517e-01 2.52095073e-01 3.89626205e-01 -1.31241277e-01 -1.04500592e-01 4.56490628e-02 4.38214064e-01 -4.30471003e-02 -1.06959856e+00 1.24465311e+00 -4.26120311e-01 5.68300843e-01 -2.63302565e-01 -1.34528577e+00 1.15945065e+00 4.04305249e-01 1.89516470e-01 -4.61301714e-01 3.20713036e-02 1.27028719e-01 2.30283648e-01 -2.50997901e-01 4.55608070e-01 -3.99971366e-01 5.93355931e-02 5.84004104e-01 1.27695054e-01 -5.93324661e-01 1.40445724e-01 1.53186455e-01 1.05663013e+00 6.38038367e-02 1.23229913e-01 2.10234597e-01 5.78650117e-01 1.73273131e-01 5.68335950e-01 9.51506615e-01 1.02797695e-01 3.15300733e-01 3.92717928e-01 -3.20934743e-01 -1.20886648e+00 -1.09593308e+00 2.40716800e-01 7.62161076e-01 -2.14924634e-01 -1.09595574e-01 -1.02901757e+00 -1.07592523e+00 2.03809857e-01 1.16650748e+00 -8.93616438e-01 -8.46430779e-01 -5.23343801e-01 -4.50584769e-01 1.08248329e+00 4.65456963e-01 6.94643617e-01 -1.10182810e+00 -4.79687214e-01 1.45757526e-01 7.60518163e-02 -8.77357304e-01 -2.45697454e-01 2.48811945e-01 -9.10127223e-01 -1.27075255e+00 -6.24898851e-01 -7.72409022e-01 1.05693066e+00 -2.35685334e-01 9.90995526e-01 1.49763435e-01 -5.69949448e-02 1.05863273e-01 -2.16866896e-01 -6.61496818e-01 -1.15096045e+00 2.42793694e-01 1.42496809e-01 -7.62456805e-02 2.02250063e-01 -4.67632055e-01 -1.66896060e-01 1.03678606e-01 -1.32118237e+00 1.91368088e-01 5.37352443e-01 1.14014888e+00 6.27080858e-01 4.86354619e-01 8.43694627e-01 -1.18712032e+00 9.34253752e-01 -5.59160411e-01 -5.04949987e-01 4.68497753e-01 -9.75359023e-01 6.65594101e-01 1.14761829e+00 -9.91864085e-01 -1.02897048e+00 1.77451484e-02 1.23313867e-01 -8.46201658e-01 3.03667020e-02 3.55999380e-01 -1.12459563e-01 -2.16052473e-01 9.90840197e-01 4.63694125e-01 1.27845064e-01 -3.07552725e-01 4.59742516e-01 4.72221226e-01 8.15298676e-01 -8.18927526e-01 1.17874253e+00 -2.12206878e-02 -4.37892824e-02 -3.60584021e-01 -9.38161790e-01 2.13693097e-01 -4.53077585e-01 1.96272463e-01 3.98132384e-01 -5.85285604e-01 -3.01865667e-01 5.25500238e-01 -9.70928013e-01 -7.74528801e-01 -1.02171648e+00 2.27874890e-01 -5.15291214e-01 -1.07929215e-01 -3.46781015e-02 -3.82012725e-01 -4.59196478e-01 -1.18518817e+00 4.86874223e-01 4.30573106e-01 -4.21733335e-02 -1.18601537e+00 3.78661007e-01 2.55459100e-01 4.90394890e-01 6.64891124e-01 1.12507129e+00 -1.31688452e+00 -2.25377262e-01 -3.98732126e-01 1.30928308e-01 1.07576466e+00 1.91554099e-01 -1.56377345e-01 -9.75545406e-01 -3.78390610e-01 -1.31269300e-03 -7.45359182e-01 8.80285442e-01 -1.87043697e-01 1.40844035e+00 -8.40973973e-01 -2.53991097e-01 6.56748772e-01 1.30508709e+00 3.34294528e-01 6.21560752e-01 1.56457126e-01 7.23822534e-01 3.16509694e-01 2.83977270e-01 3.13448906e-03 4.94097359e-02 2.76264369e-01 2.73923159e-01 -4.06301171e-02 -4.42756377e-02 -4.80054647e-01 2.45104373e-01 5.89624286e-01 5.97599372e-02 -2.38774613e-01 -8.54895771e-01 5.48587441e-01 -1.55735934e+00 -8.09366465e-01 5.11001229e-01 2.33865762e+00 1.48355627e+00 2.35127553e-01 -1.52033478e-01 2.85384625e-01 4.39048409e-01 -1.63173690e-01 -7.57183433e-01 -6.29480183e-01 1.61802888e-01 8.49213719e-01 5.79778910e-01 4.46144372e-01 -7.46034086e-01 9.79525924e-01 6.36475277e+00 8.81259978e-01 -1.33379972e+00 -7.06736818e-02 6.86946094e-01 -2.38004535e-01 -1.81858972e-01 -1.15567006e-01 -5.68255067e-01 4.75371212e-01 1.24527800e+00 -5.95887959e-01 6.41707003e-01 1.05005980e+00 6.23014905e-02 -6.29551932e-02 -1.38957810e+00 5.77312171e-01 1.37910292e-01 -1.27526653e+00 3.10235322e-01 -1.15323283e-01 1.25717783e+00 -2.10196018e-01 4.18798141e-02 7.56934166e-01 5.51514149e-01 -1.34105706e+00 4.13776308e-01 5.93519866e-01 7.52243876e-01 -9.65421736e-01 6.65704906e-01 6.83214366e-01 -6.38827741e-01 -6.90149935e-03 -1.61593199e-01 1.26151711e-01 -4.36999470e-01 2.13699341e-01 -1.50721323e+00 2.56479770e-01 1.62115470e-01 1.60961479e-01 -8.26031983e-01 8.71607482e-01 -4.75094199e-01 9.73760724e-01 -2.14958638e-01 1.43335834e-01 2.18103185e-01 1.18755654e-01 3.88722062e-01 1.00304377e+00 1.30015135e-01 -7.53147006e-02 2.82848120e-01 1.01680028e+00 -4.38683659e-01 -2.03858867e-01 -8.14813077e-01 -4.23543632e-01 5.53692877e-01 8.22635591e-01 -2.28374526e-01 -6.99229658e-01 -1.12914719e-01 7.94867158e-01 6.06932402e-01 4.27892745e-01 -6.85497046e-01 -7.45598257e-01 3.32855761e-01 4.44642901e-02 2.08741680e-01 -1.70296840e-02 -1.36632517e-01 -9.32563007e-01 -1.52225584e-01 -1.12892675e+00 3.10533047e-01 -6.61225617e-01 -9.14183855e-01 5.85371017e-01 1.46926522e-01 -9.79126573e-01 -7.02169180e-01 -3.36697787e-01 -7.61035562e-01 1.22060084e+00 -1.53038871e+00 -8.04491222e-01 -3.45412761e-01 7.25412548e-01 3.28938216e-01 -5.98193288e-01 9.99545693e-01 -1.15852810e-01 -5.78874350e-01 1.01965582e+00 3.98350567e-01 3.90897542e-01 5.35734057e-01 -1.27082479e+00 2.34307319e-01 6.77095294e-01 -5.83065823e-02 4.71496731e-01 7.46517479e-01 -7.28943229e-01 -9.79457319e-01 -1.42370462e+00 6.26629889e-01 -3.20006222e-01 4.43213731e-01 -1.45191997e-02 -1.16493082e+00 6.67828143e-01 5.41669875e-02 -7.16595575e-02 6.33062124e-01 -4.21412736e-01 -3.89591932e-01 -1.92905217e-01 -1.65054846e+00 3.67257059e-01 8.35754946e-02 -2.17539743e-01 -9.34653342e-01 1.54454127e-01 8.56749654e-01 -6.04709208e-01 -9.44928646e-01 4.16332752e-01 3.75759989e-01 -4.02988464e-01 8.17063749e-01 -1.03437054e+00 5.21679103e-01 -4.64948826e-02 2.11938709e-01 -2.01915884e+00 2.89549410e-01 -5.81520677e-01 -7.07644224e-01 9.65302408e-01 5.29110193e-01 -6.63319886e-01 1.19259882e+00 7.16645062e-01 1.94718748e-01 -9.46217179e-01 -7.57098615e-01 -9.64922905e-01 3.44354987e-01 -5.08962497e-02 7.52531886e-01 1.23513496e+00 -3.81818354e-01 3.01844239e-01 -2.91071266e-01 2.70130602e-03 4.82313544e-01 -3.15271839e-02 8.49826157e-01 -1.14049947e+00 -4.98835534e-01 -1.98904246e-01 -3.52899224e-01 -6.45890772e-01 3.75307471e-01 -1.11425805e+00 1.59664843e-02 -1.08114433e+00 -1.62451431e-01 -7.30858922e-01 -2.79815853e-01 8.18702221e-01 -3.78390759e-01 6.14760108e-02 1.33057922e-01 -2.52599698e-02 8.77114236e-02 5.33346474e-01 1.53032875e+00 -1.99246317e-01 -4.46657538e-01 2.69933730e-01 -8.20877016e-01 4.79270935e-01 1.00717950e+00 -8.21692407e-01 -8.81754398e-01 -2.19291031e-01 1.20233092e-03 -7.22536370e-02 2.32577756e-01 -1.05580664e+00 2.83206850e-01 -3.14219028e-01 4.90646780e-01 -1.38168201e-01 1.89397112e-01 -7.89227128e-01 -7.98792765e-02 6.14688337e-01 -5.47531188e-01 -2.79502440e-02 3.28229040e-01 5.28848171e-01 -2.43495584e-01 -6.71472490e-01 1.06288743e+00 -7.81589225e-02 -2.40082026e-01 1.88604593e-01 -3.08630839e-02 4.51142788e-01 1.30692601e+00 -1.26612186e-01 -3.84152532e-01 -2.66759217e-01 -8.13269138e-01 4.18214619e-01 3.43853176e-01 3.19638640e-01 6.36185944e-01 -1.22518861e+00 -8.21460068e-01 5.00757277e-01 -2.56857306e-01 5.67964971e-01 -3.04817349e-01 1.65782779e-01 -5.04168928e-01 3.42015892e-01 -1.61133021e-01 -4.23712619e-02 -1.00761127e+00 7.88578570e-01 6.55170083e-01 -5.16328156e-01 -1.73156917e-01 7.55481899e-01 -5.70447519e-02 -7.23107696e-01 3.60947013e-01 -3.88944060e-01 -3.39579582e-02 -2.31600612e-01 4.77017373e-01 4.24348354e-01 -1.70688052e-02 -2.39390314e-01 2.65338272e-01 -2.10774377e-01 -2.76377171e-01 3.11438236e-02 1.15374422e+00 6.74089432e-01 2.31277257e-01 2.37141028e-02 1.14061940e+00 -1.83229446e-01 -1.10306418e+00 -5.47727227e-01 -3.24232459e-01 -2.83220917e-01 -1.02015115e-01 -9.49374318e-01 -1.30857944e+00 6.88448668e-01 4.12454665e-01 3.59714597e-01 1.27594352e+00 -3.02073568e-01 7.17988729e-01 7.83924520e-01 -1.34545639e-02 -1.12731206e+00 2.11182982e-01 4.01968598e-01 6.74082220e-01 -1.01380277e+00 -1.43442139e-01 -3.61818969e-02 -6.36893213e-01 9.07532573e-01 1.08746791e+00 -2.92334408e-01 5.08987546e-01 2.47121558e-01 -7.57078156e-02 1.53838426e-01 -8.41680288e-01 1.68264061e-01 2.87252694e-01 7.12472022e-01 -7.72602782e-02 -9.45905782e-03 1.29932106e-01 4.55223233e-01 -5.69599271e-01 1.08157538e-01 5.02871454e-01 1.03314006e+00 -4.31499243e-01 -1.34748912e+00 -4.10458505e-01 6.57595634e-01 -2.87215829e-01 -1.16636537e-01 -5.14289021e-01 8.07596982e-01 4.01524037e-01 6.28816545e-01 -1.30200833e-02 -4.81664449e-01 3.92370075e-01 3.92722338e-01 3.99228901e-01 -8.79042149e-01 -8.06580544e-01 -6.66214108e-01 6.57829107e-04 -1.42608345e-01 -1.18601985e-01 -1.87150016e-01 -1.16334963e+00 -1.87221453e-01 -3.19837064e-01 6.90379083e-01 5.30768931e-01 9.91497815e-01 9.83739570e-02 6.24181271e-01 7.16599047e-01 -3.93684566e-01 -1.01930904e+00 -8.93919945e-01 -4.32306498e-01 3.78458649e-01 4.16902751e-01 -4.25903946e-01 -2.49264359e-01 3.01859856e-01]
[9.559248924255371, 3.362666130065918]
90ce70d0-623e-4b4a-8335-3b444b54269a
improved-deep-learning-baselines-for-ubuntu
1510.03753
null
http://arxiv.org/abs/1510.03753v2
http://arxiv.org/pdf/1510.03753v2.pdf
Improved Deep Learning Baselines for Ubuntu Corpus Dialogs
This paper presents results of our experiments for the next utterance ranking on the Ubuntu Dialog Corpus -- the largest publicly available multi-turn dialog corpus. First, we use an in-house implementation of previously reported models to do an independent evaluation using the same data. Second, we evaluate the performances of various LSTMs, Bi-LSTMs and CNNs on the dataset. Third, we create an ensemble by averaging predictions of multiple models. The ensemble further improves the performance and it achieves a state-of-the-art result for the next utterance ranking on this dataset. Finally, we discuss our future plans using this corpus.
['Jan Kleindienst', 'Rudolf Kadlec', 'Martin Schmid']
2015-10-13
null
null
null
null
['conversational-response-selection']
['natural-language-processing']
[-2.20260099e-01 4.55690682e-01 -4.66554165e-02 -9.53582823e-01 -1.12478387e+00 -5.03232300e-01 6.51399791e-01 2.31394339e-02 -5.87194383e-01 9.41229522e-01 8.53166759e-01 -5.18175781e-01 2.07861021e-01 -2.25053743e-01 -2.43126005e-01 -8.56952667e-02 -2.24737704e-01 1.08114529e+00 5.74072361e-01 -9.21834648e-01 3.05792958e-01 -7.49518769e-03 -1.23498070e+00 1.10295868e+00 5.36976814e-01 1.16433001e+00 5.09057604e-02 9.69576061e-01 8.88862088e-02 1.54604793e+00 -6.88721001e-01 -6.00827694e-01 -7.35685155e-02 -4.00766172e-02 -1.75911212e+00 -5.30412972e-01 6.97474122e-01 -9.50827956e-01 -2.92650253e-01 3.48328322e-01 6.96332633e-01 7.86418796e-01 4.31104660e-01 -6.23309851e-01 -4.87868398e-01 1.02613544e+00 4.29252416e-01 2.53978103e-01 6.27947330e-01 -1.55175462e-01 1.36422408e+00 -6.70611799e-01 7.71386743e-01 1.53259075e+00 5.05152166e-01 9.91453469e-01 -9.54447389e-01 -1.28343388e-01 1.47259608e-02 3.48979622e-01 -4.88572687e-01 -9.45003033e-01 5.65883040e-01 -1.63882405e-01 1.69260168e+00 4.40586984e-01 -8.78010597e-03 1.24982321e+00 1.85634494e-01 1.21286249e+00 1.01034570e+00 -4.30103153e-01 -1.57384127e-01 3.89163122e-02 6.91414654e-01 7.90130675e-01 -1.20859110e+00 -1.40436798e-01 -4.09846067e-01 -3.40277821e-01 8.40133503e-02 -2.68280655e-01 -1.10407710e-01 -2.16651373e-02 -1.03920901e+00 9.44598973e-01 6.13334119e-01 3.68018806e-01 -1.64538547e-01 -1.24137647e-01 8.39893579e-01 5.77318728e-01 9.26002085e-01 5.23570180e-01 -8.44596863e-01 -4.85206902e-01 -4.48249698e-01 3.66213024e-01 1.25194490e+00 6.48013473e-01 3.09309036e-01 -3.24013621e-01 -5.48145652e-01 1.44724810e+00 1.95717767e-01 -7.75794089e-02 5.59683859e-01 -1.26224864e+00 7.98160255e-01 2.29864255e-01 2.55390614e-01 -2.84276664e-01 -7.01780975e-01 3.03824395e-01 -3.21362495e-01 -1.98739216e-01 6.65377557e-01 -5.07211804e-01 -6.05560064e-01 1.66898310e+00 -7.62676224e-02 -2.61515439e-01 3.68956476e-01 6.99099243e-01 1.41431475e+00 5.70713460e-01 2.03756779e-01 -1.33405551e-01 1.22068012e+00 -1.70321190e+00 -8.68572414e-01 1.11309133e-01 8.93057466e-01 -6.76421583e-01 8.50555062e-01 1.69507995e-01 -1.14094818e+00 -5.34969211e-01 -7.45331168e-01 -4.91177320e-01 -5.80407739e-01 2.50792921e-01 6.70587659e-01 3.43816638e-01 -1.50613320e+00 8.94925892e-01 -5.00494957e-01 -5.02046347e-01 -1.79800615e-01 6.80404425e-01 -7.01543465e-02 4.50007200e-01 -1.66651905e+00 1.70605540e+00 4.51465577e-01 2.34355554e-01 -7.52805352e-01 -2.70250618e-01 -7.54114509e-01 2.24331282e-02 1.46307349e-01 -1.14706554e-01 2.33201671e+00 -3.91239613e-01 -2.37889051e+00 7.76548982e-01 -3.99430037e-01 -1.00417948e+00 4.17322904e-01 -3.85868728e-01 -2.84597337e-01 -2.32586816e-01 -3.14757228e-01 1.06891727e+00 8.09355546e-03 -1.09464622e+00 -9.60134387e-01 -6.55349642e-02 6.05793953e-01 4.73112017e-01 -1.38647601e-01 1.26024023e-01 7.34710991e-02 7.93351755e-02 -1.16676651e-01 -1.04523706e+00 -3.02791148e-01 -1.16645586e+00 -5.22192299e-01 -8.36836159e-01 9.10257816e-01 -8.67034376e-01 1.11941767e+00 -1.79638469e+00 1.22318238e-01 -4.76634264e-01 1.39059648e-01 2.94419378e-01 -4.99648079e-02 5.89351714e-01 1.98588699e-01 -7.78475776e-02 3.98152530e-01 -1.00584328e+00 8.60327098e-04 1.83693498e-01 -5.01718998e-01 -1.40655488e-01 -2.68339962e-01 9.76451159e-01 -8.95019054e-01 -5.29091299e-01 6.01684034e-01 4.38581361e-03 -2.62462705e-01 4.56415057e-01 -4.95170057e-01 5.03174901e-01 -2.87498951e-01 2.14475408e-01 8.25239122e-02 4.04035151e-02 4.22412425e-01 -1.11670479e-01 -1.20127462e-01 1.11165106e+00 -8.05337057e-02 1.88793647e+00 -6.52370572e-01 8.83572161e-01 -2.63615340e-01 -5.26900172e-01 8.48535657e-01 7.02084601e-01 2.25308478e-01 -5.92527688e-01 3.41317713e-01 3.73542964e-01 2.44620696e-01 -1.98279023e-01 1.10852647e+00 1.12530410e-01 -1.83703735e-01 5.09036720e-01 5.27776599e-01 1.07220955e-01 3.07084382e-01 2.41880342e-01 9.62634504e-01 6.05100160e-03 1.97596744e-01 -2.62332559e-01 6.37950301e-01 -8.22014883e-02 -9.99241769e-02 8.00609469e-01 -4.23676074e-01 3.97956163e-01 3.80196363e-01 -9.99979258e-01 -7.76957750e-01 -8.90104830e-01 4.07877006e-03 1.95857346e+00 -3.40283573e-01 -1.37909800e-01 -7.69592524e-01 -1.17569458e+00 -4.35016304e-01 1.08651245e+00 -7.03989327e-01 3.65008980e-01 -7.72774935e-01 -6.10806406e-01 6.33849323e-01 4.22736257e-01 7.54139066e-01 -1.71818864e+00 -4.83182847e-01 3.59998405e-01 -7.49115765e-01 -1.34799373e+00 -1.67339548e-01 3.69956911e-01 -8.95578861e-01 -8.42455089e-01 -3.17074656e-01 -9.41429257e-01 -2.24147979e-02 -1.15653671e-01 1.60265756e+00 5.07320017e-02 5.79200327e-01 1.94280088e-01 -5.49169242e-01 -2.96470046e-01 -7.71738648e-01 5.25851369e-01 -8.38865265e-02 -4.58596796e-01 5.41835845e-01 -1.38237879e-01 -1.57730252e-01 2.57037610e-01 -8.97647217e-02 2.58729651e-05 1.22475401e-02 1.01589417e+00 -1.83193505e-01 -5.74734271e-01 6.86058819e-01 -9.83038187e-01 1.24310052e+00 3.39860958e-03 -3.20149422e-01 4.65879828e-01 -1.15314268e-01 -1.58127211e-03 5.57003677e-01 -8.65504667e-02 -1.52335107e+00 1.22231752e-01 -4.75620419e-01 7.57965520e-02 -3.73136401e-01 2.57411927e-01 3.92502099e-01 1.98414996e-01 4.52885628e-01 -1.87582299e-01 -2.65230417e-01 -5.19864023e-01 6.24823868e-01 1.15250456e+00 3.71424705e-01 -4.89657462e-01 -3.19193989e-01 -5.97007398e-04 -5.61720312e-01 -8.10606539e-01 -1.21755707e+00 -4.19005215e-01 -1.00738847e+00 -4.15354520e-01 1.04660082e+00 -6.18513465e-01 -1.19489634e+00 4.18313086e-01 -1.66648555e+00 -9.91465569e-01 3.54862392e-01 2.29565620e-01 -6.21643603e-01 1.40091345e-01 -1.44352925e+00 -9.89467919e-01 -7.57805407e-01 -1.37534332e+00 8.63931000e-01 5.67661747e-02 -5.00764489e-01 -1.30834734e+00 3.71756345e-01 7.11194575e-01 3.89118612e-01 -2.96015918e-01 6.38689458e-01 -1.44959307e+00 -1.71979904e-01 -1.71330608e-02 2.42656190e-02 3.54230076e-01 1.14702992e-02 -1.62807867e-01 -1.41718233e+00 -1.01720966e-01 -1.03973620e-01 -1.01526058e+00 9.98547256e-01 5.50786316e-01 7.76749611e-01 -2.36049488e-01 -2.77714103e-01 -4.06817436e-01 6.43995106e-01 5.81778109e-01 5.68703353e-01 3.94382149e-01 4.68052000e-01 7.98871458e-01 7.00053513e-01 6.67045936e-02 8.76961946e-01 8.96570504e-01 2.48293236e-01 6.04243539e-02 3.29325706e-01 4.10975516e-02 2.52855927e-01 1.02051175e+00 -1.89924374e-01 -4.13336813e-01 -8.89336169e-01 4.68952447e-01 -2.24138618e+00 -9.61776137e-01 -3.61772552e-02 1.83375025e+00 7.32915640e-01 1.75752237e-01 2.62815177e-01 -4.89398628e-01 5.29403806e-01 4.66910392e-01 -1.02615222e-01 -9.64758515e-01 3.28148276e-01 7.60160163e-02 1.89547881e-01 1.05735600e+00 -1.48897231e+00 1.64551544e+00 7.68846655e+00 5.57897210e-01 -9.53654647e-01 4.82975453e-01 1.08675730e+00 -1.06533952e-01 3.69346112e-01 -1.98537380e-01 -9.58088636e-01 1.59146115e-01 1.49059486e+00 2.78277636e-01 4.12402451e-01 1.02791214e+00 8.37200731e-02 -3.03620070e-01 -1.39693236e+00 1.77712873e-01 7.24431798e-02 -1.32891190e+00 -2.67866701e-01 -1.86582148e-01 5.76960623e-01 5.59342623e-01 -2.07446322e-01 9.58649933e-01 9.92325664e-01 -8.17825139e-01 4.57208186e-01 1.34242997e-01 3.17346364e-01 -5.93470931e-01 1.30146837e+00 1.98415950e-01 -8.12096059e-01 -1.13818757e-01 -4.75510150e-01 -3.19210678e-01 3.73154759e-01 -5.78317035e-04 -1.51401603e+00 4.45345342e-01 6.24415874e-01 4.53286290e-01 -5.10442317e-01 3.60776335e-01 -1.38464198e-01 1.54769063e-01 -7.91889653e-02 -4.99637753e-01 6.67884111e-01 1.74935535e-02 1.78835571e-01 1.45406497e+00 -2.77309746e-01 2.64091492e-01 5.11318207e-01 9.02051032e-02 -2.70732552e-01 -1.12836957e-01 -6.40079558e-01 4.28216368e-01 4.48374569e-01 1.25504100e+00 -2.88857937e-01 -7.01971292e-01 -2.22385049e-01 7.71679103e-01 1.00595558e+00 3.41122389e-01 -6.70598805e-01 -1.93206593e-03 5.37242472e-01 -7.82499611e-01 5.09718619e-02 4.00690548e-02 7.37422286e-03 -9.35372174e-01 -3.41549456e-01 -8.26694310e-01 6.47465110e-01 -5.89135468e-01 -1.16700518e+00 1.30540359e+00 1.42838255e-01 -8.32747161e-01 -1.01702237e+00 -7.55895197e-01 -7.66853452e-01 5.87656796e-01 -1.07348514e+00 -1.04026365e+00 2.72741318e-01 2.29195192e-01 1.10581732e+00 -4.05921519e-01 1.31575835e+00 1.34565800e-01 -4.65048432e-01 4.58872318e-01 8.88960287e-02 4.19422776e-01 8.52615833e-01 -1.52494729e+00 6.17711663e-01 1.26633033e-01 9.47031844e-03 6.08815253e-01 6.75735831e-01 -3.76249880e-01 -5.47631919e-01 -5.99881709e-01 1.21836090e+00 -1.00423956e+00 8.01288843e-01 -2.96360463e-01 -5.12785137e-01 1.33983076e+00 8.68493974e-01 -6.84461892e-01 7.38688231e-01 8.57442617e-01 1.12994790e-01 2.98917800e-01 -1.08698702e+00 6.43288732e-01 6.45822287e-01 -5.90894222e-01 -1.03744483e+00 5.38710713e-01 9.54488277e-01 -9.87364233e-01 -1.08571637e+00 5.67816913e-01 8.61488402e-01 -1.36573029e+00 8.59406590e-01 -9.33492541e-01 8.89065862e-01 5.74142933e-01 -3.02653819e-01 -1.62912774e+00 4.82004546e-02 -5.21714747e-01 -1.78435937e-01 1.04558933e+00 9.34882462e-01 -4.92370009e-01 7.04685211e-01 7.20312238e-01 -5.35269916e-01 -9.64335620e-01 -1.14647293e+00 -3.05994481e-01 1.45466462e-01 -3.18283468e-01 4.76804197e-01 6.43293679e-01 6.13595724e-01 1.00549996e+00 -7.21154153e-01 -5.41093528e-01 6.21535629e-02 2.49719527e-02 8.76653314e-01 -1.08278322e+00 3.44593748e-02 -4.11214173e-01 1.10618994e-01 -1.46008420e+00 6.95359707e-01 -5.75822115e-01 4.63012874e-01 -1.66474628e+00 7.26314355e-03 -4.95492555e-02 -3.14186424e-01 5.60535610e-01 -1.60065159e-01 -3.61045673e-02 3.86825949e-01 2.07014512e-02 -1.15755320e+00 6.58136249e-01 9.83031392e-01 -7.40469396e-02 -5.08556426e-01 2.29182467e-01 -9.08970907e-02 7.06841111e-01 9.26561654e-01 -1.15832940e-01 -1.47185892e-01 -5.43929577e-01 -6.56770051e-01 5.64734280e-01 -1.73262700e-01 -8.87177765e-01 8.79000500e-02 3.13924998e-01 7.93097094e-02 -1.12768614e+00 9.21466172e-01 -2.64919907e-01 -7.43075430e-01 3.14415783e-01 -9.57842708e-01 1.33143067e-01 1.89946964e-01 -5.49128652e-03 -2.50486076e-01 -3.54523838e-01 5.45864165e-01 -4.23835129e-01 -8.17940772e-01 -2.07173556e-01 -6.11380398e-01 -1.76196739e-01 4.23657417e-01 4.38960075e-01 -7.83007085e-01 -9.17226493e-01 -8.90389323e-01 5.80254078e-01 -1.18755817e-01 8.31572950e-01 3.90905619e-01 -1.05854940e+00 -5.19595861e-01 -4.21085835e-01 -1.25087783e-01 -5.42644322e-01 1.70050085e-01 6.73148036e-01 -2.89767832e-01 1.06988180e+00 -1.99249387e-01 -4.42292899e-01 -1.64338040e+00 6.88292980e-02 4.73858535e-01 -9.54361975e-01 -2.35279158e-01 9.49828982e-01 -1.81837186e-01 -1.32401288e+00 4.03443605e-01 -2.85851449e-01 -7.92427123e-01 1.64867520e-01 4.51522410e-01 3.15315783e-01 1.17375739e-01 -6.47664428e-01 -1.89358890e-01 -2.87647694e-01 -5.11540771e-01 -5.28933644e-01 1.12635744e+00 -4.30335701e-02 -2.74769664e-01 8.87717903e-01 1.27731740e+00 -3.73854220e-01 -7.83631027e-01 -1.29037499e-01 3.31608355e-01 2.69236043e-02 -2.33750612e-01 -1.03453827e+00 -5.50130725e-01 7.80334353e-01 4.41008240e-01 6.43070579e-01 5.09618044e-01 -1.23025760e-01 8.88207614e-01 1.17743587e+00 4.37003016e-01 -1.24027097e+00 4.42068800e-02 1.64270103e+00 8.50923479e-01 -1.57202911e+00 -1.77830353e-01 -8.46277624e-02 -1.04408264e+00 1.19815338e+00 1.01001728e+00 2.90630069e-02 4.52861398e-01 -3.92280191e-01 6.74088240e-01 -2.66765356e-01 -1.16905344e+00 -1.93554595e-01 3.21851403e-01 2.85849482e-01 1.02024317e+00 1.90709740e-01 -2.16749415e-01 1.96084350e-01 -5.41808605e-01 -2.50583291e-01 5.31940997e-01 5.94275475e-01 -3.91787380e-01 -1.18016875e+00 4.34993915e-02 5.42117417e-01 -5.24437249e-01 -3.61379027e-01 -9.39184546e-01 7.28747904e-01 -5.80220699e-01 1.54640973e+00 1.04087349e-02 -9.62235034e-01 4.35342163e-01 5.18730819e-01 2.31354460e-01 -7.68505275e-01 -1.03012228e+00 -1.80137530e-01 1.10031164e+00 -5.75870454e-01 -4.28543180e-01 -3.46520096e-01 -1.06163061e+00 -3.69446814e-01 -4.50770020e-01 3.59297812e-01 5.19411147e-01 1.29403937e+00 -1.64494999e-02 6.71154797e-01 5.65765023e-01 -1.34373856e+00 -7.83495128e-01 -1.88813603e+00 -5.40184267e-02 1.94642052e-01 1.40960768e-01 -6.42685533e-01 -1.49863124e-01 -4.55818653e-01]
[12.77357006072998, 7.891060829162598]
d0786ad2-589e-4e3e-8095-ee20c3082dc1
exploiting-abstract-meaning-representation
2305.17050
null
https://arxiv.org/abs/2305.17050v1
https://arxiv.org/pdf/2305.17050v1.pdf
Exploiting Abstract Meaning Representation for Open-Domain Question Answering
The Open-Domain Question Answering (ODQA) task involves retrieving and subsequently generating answers from fine-grained relevant passages within a database. Current systems leverage Pretrained Language Models (PLMs) to model the relationship between questions and passages. However, the diversity in surface form expressions can hinder the model's ability to capture accurate correlations, especially within complex contexts. Therefore, we utilize Abstract Meaning Representation (AMR) graphs to assist the model in understanding complex semantic information. We introduce a method known as Graph-as-Token (GST) to incorporate AMRs into PLMs. Results from Natural Questions (NQ) and TriviaQA (TQ) demonstrate that our GST method can significantly improve performance, resulting in up to 2.44/3.17 Exact Match score improvements on NQ/TQ respectively. Furthermore, our method enhances robustness and outperforms alternative Graph Neural Network (GNN) methods for integrating AMRs. To the best of our knowledge, we are the first to employ semantic graphs in ODQA.
['Yue Zhang', 'Zheng Zhang', 'Xuefeng Bai', 'Xiangkun Hu', 'Qipeng Guo', 'Zhikun Xu', 'Cunxiang Wang']
2023-05-26
null
null
null
null
['natural-questions', 'triviaqa', 'open-domain-question-answering']
['miscellaneous', 'miscellaneous', 'natural-language-processing']
[ 6.56453073e-02 2.76142389e-01 1.73814237e-01 -2.34942704e-01 -1.34374321e+00 -7.07087517e-01 5.73533356e-01 5.48627019e-01 -2.63898790e-01 5.38760960e-01 6.14750922e-01 -5.05889535e-01 -9.86162573e-02 -9.50244427e-01 -6.30604982e-01 2.12780654e-01 1.50159404e-01 5.35156727e-01 5.01366973e-01 -8.49932194e-01 3.11014950e-01 8.54767188e-02 -1.07481945e+00 7.08188117e-01 1.27861702e+00 1.02509034e+00 1.25741914e-01 8.47700715e-01 -9.32411253e-01 1.20727277e+00 -8.70571375e-01 -9.31962550e-01 -1.54884011e-01 -5.70061862e-01 -1.35841155e+00 -3.84158999e-01 6.41703367e-01 -1.35769486e-01 -4.72507715e-01 7.40828216e-01 3.16562116e-01 6.44352615e-01 7.30562389e-01 -9.77965355e-01 -1.35769796e+00 6.13158464e-01 -1.26879096e-01 3.86432260e-01 1.02810836e+00 -6.66635856e-02 1.72636580e+00 -9.37336862e-01 5.78508973e-01 1.44251406e+00 5.84778786e-01 4.98495042e-01 -8.37742627e-01 -2.29576170e-01 2.50854850e-01 4.89953727e-01 -1.25072503e+00 -1.25424400e-01 6.80413365e-01 9.82984602e-02 1.58869326e+00 1.49307042e-01 3.71347338e-01 8.11584115e-01 -5.17834909e-03 9.04288292e-01 6.86682403e-01 -5.04285395e-01 1.83704216e-02 -3.11170578e-01 4.29743290e-01 8.48964512e-01 2.17450455e-01 -6.38946414e-01 -5.33570647e-01 -2.08790272e-01 5.10118544e-01 -2.84648687e-01 -1.16048351e-01 1.99178427e-01 -8.71107876e-01 9.80389237e-01 7.27676332e-01 2.22180218e-01 -4.54791009e-01 1.72859088e-01 2.72123635e-01 6.51017249e-01 3.52556765e-01 1.09505785e+00 -2.59225905e-01 -4.22494002e-02 -5.42338550e-01 5.86166143e-01 1.18452847e+00 9.42513049e-01 7.03570724e-01 -1.34403631e-01 -5.27000844e-01 1.11065888e+00 3.08956355e-01 4.03513759e-01 4.25253481e-01 -8.24791610e-01 7.89606810e-01 9.29740965e-01 -7.40861371e-02 -1.19560564e+00 -2.85341948e-01 -3.55965525e-01 -2.21850902e-01 -7.79562593e-01 4.05383378e-01 5.32396771e-02 -8.29417229e-01 1.54257572e+00 8.75012279e-02 -1.91764440e-02 3.80273283e-01 7.49212027e-01 1.46572375e+00 8.88340890e-01 2.96015918e-01 3.56836081e-01 1.54670227e+00 -1.03825879e+00 -8.33341599e-01 -5.44232786e-01 8.63499939e-01 -7.43067026e-01 1.58560359e+00 1.09844282e-01 -9.76784468e-01 -5.10177016e-01 -8.20253909e-01 -6.97994828e-01 -5.96183598e-01 -4.00148690e-01 4.88991797e-01 2.79931962e-01 -1.21516621e+00 3.00396174e-01 -2.75783718e-01 -3.80725473e-01 4.06385213e-01 1.88532844e-02 -1.57387704e-01 -7.23363578e-01 -1.70726049e+00 9.90023851e-01 1.90851822e-01 -2.13049620e-01 -5.87006629e-01 -7.35537946e-01 -1.03684676e+00 1.18610919e-01 5.11489511e-01 -1.13645196e+00 1.38600922e+00 -5.05756974e-01 -1.21394062e+00 7.18464613e-01 -4.47928220e-01 -5.66221595e-01 -1.70631424e-01 -4.36999559e-01 -5.59972048e-01 5.77651858e-01 2.94154942e-01 6.87409639e-01 4.94108289e-01 -1.06854057e+00 -2.20445231e-01 -1.93509072e-01 6.35078967e-01 3.56660485e-01 -2.54285634e-01 1.58848405e-01 -4.08031136e-01 -6.61752164e-01 8.18007961e-02 -5.66954851e-01 -1.00965433e-01 -3.17513883e-01 -3.72415125e-01 -8.98247600e-01 4.43663657e-01 -1.19863701e+00 1.48650932e+00 -1.73064148e+00 -8.02796036e-02 1.51296589e-03 3.06258976e-01 2.39485502e-01 -5.86218715e-01 9.53026950e-01 4.96885598e-01 2.11606815e-01 -1.85372725e-01 -3.37099165e-01 2.63422787e-01 4.98039842e-01 -4.84694958e-01 -3.76117676e-01 5.09695709e-01 1.58426082e+00 -1.05457735e+00 -4.97819781e-01 -4.55094725e-01 2.88238585e-01 -7.22589135e-01 4.49185818e-01 -7.46478736e-01 5.57778887e-02 -7.00247705e-01 5.78234792e-01 2.14941755e-01 -5.60688376e-01 -9.50069129e-02 -7.96079710e-02 6.14473760e-01 9.81486380e-01 -6.27320945e-01 1.86526835e+00 -6.99078202e-01 4.85926718e-01 -4.65505064e-01 -5.86780071e-01 1.15286052e+00 2.06866190e-01 -1.19302697e-01 -1.05930054e+00 -2.38399252e-01 1.32951945e-01 -1.49221301e-01 -6.50382280e-01 1.06715655e+00 -1.66123033e-01 -2.01698154e-01 5.44101238e-01 1.86352462e-01 -5.13335824e-01 3.48432451e-01 7.98635244e-01 1.30527878e+00 -1.08600803e-01 8.59868526e-02 6.93530729e-03 4.50461596e-01 2.00520813e-01 2.87288707e-02 9.67350066e-01 -1.77797973e-02 5.56703210e-01 5.86507797e-01 -6.20108061e-02 -6.19687617e-01 -1.24181271e+00 4.91598278e-01 1.25948727e+00 4.31745201e-02 -6.16779327e-01 -6.29545629e-01 -8.13277543e-01 2.00110123e-01 1.17243123e+00 -3.03884596e-01 -2.44909674e-01 -6.69358253e-01 -4.32316028e-02 8.27862918e-01 7.23892927e-01 3.58515441e-01 -1.09912431e+00 1.24860421e-01 2.90449888e-01 -6.71262860e-01 -1.50622380e+00 -5.18464208e-01 -3.92712176e-01 -8.45661461e-01 -1.04766715e+00 -4.53299403e-01 -7.64307857e-01 3.92726064e-01 4.23052073e-01 1.70895648e+00 2.80412614e-01 4.88796383e-02 7.35484362e-01 -8.52333784e-01 -3.24534953e-01 -3.94550890e-01 1.30684659e-01 -4.26446795e-01 -2.83368468e-01 6.12229466e-01 -3.05746675e-01 -5.87906897e-01 3.66499834e-02 -9.75146055e-01 -2.98860580e-01 3.71280134e-01 6.71244860e-01 4.18869227e-01 -2.69117326e-01 1.17156839e+00 -8.91625524e-01 1.41899121e+00 -8.23619127e-01 -1.32124543e-01 5.26794016e-01 -5.07833064e-01 2.11197227e-01 7.13023722e-01 -9.47953537e-02 -1.23738980e+00 -7.92909622e-01 -3.37225378e-01 -5.57427816e-02 1.62159298e-02 8.39589417e-01 5.85134663e-02 -2.47406233e-02 8.13818336e-01 -1.18148269e-03 -1.15107879e-01 -4.13389564e-01 8.11393857e-01 5.10521650e-01 4.37981725e-01 -8.16653430e-01 6.31988108e-01 1.43623143e-01 -1.79785267e-01 -6.64447963e-01 -1.30448151e+00 -7.05851257e-01 -1.62040561e-01 -4.34169024e-02 9.22915578e-01 -9.56735909e-01 -3.35792482e-01 -4.07202082e-04 -1.28773952e+00 -2.74063885e-01 -2.30703712e-01 1.61197349e-01 -3.51910591e-01 5.59586167e-01 -9.12274182e-01 -5.15414000e-01 -5.79670250e-01 -6.09478056e-01 9.59768295e-01 2.29872078e-01 -5.03004789e-01 -1.44358659e+00 -1.23921238e-01 1.09513104e+00 6.76634252e-01 1.61712021e-01 1.14323378e+00 -1.03742993e+00 -5.72192371e-01 -2.84377843e-01 -4.43108350e-01 3.03425789e-01 1.40029609e-01 -2.99669862e-01 -7.77462900e-01 5.52698597e-02 -2.27321088e-01 -6.33683085e-01 7.72178531e-01 -2.48076562e-02 9.21109676e-01 -5.24945617e-01 1.24049991e-01 -3.70938666e-02 1.18038440e+00 -1.22261040e-01 6.54857278e-01 1.99445114e-01 9.03472483e-01 8.27861667e-01 5.50202250e-01 1.39116421e-01 1.06167412e+00 3.56454462e-01 2.66193330e-01 1.86938137e-01 -5.92493474e-01 -7.84892738e-01 4.02607083e-01 1.12060416e+00 3.15958679e-01 -5.01303613e-01 -1.08200502e+00 7.88272619e-01 -1.49326181e+00 -5.87352276e-01 -3.30471963e-01 1.61991322e+00 8.62374365e-01 1.11865383e-02 -1.01436190e-01 -3.25387448e-01 4.12945449e-01 2.80757427e-01 -4.47925061e-01 -6.15863979e-01 -1.14413761e-01 6.52515173e-01 -1.08143119e-02 5.86449981e-01 -4.70203340e-01 1.20943546e+00 5.83150673e+00 8.55700850e-01 -3.54806483e-01 -5.60188927e-02 2.75050402e-01 3.14861685e-01 -8.27245176e-01 4.78044041e-02 -6.85990334e-01 -1.17136143e-01 1.13892317e+00 -2.51141906e-01 4.17729676e-01 6.03316128e-01 -1.00494832e-01 1.50690854e-01 -8.67248237e-01 7.43585110e-01 4.64564770e-01 -1.36463201e+00 8.31244051e-01 -5.75435519e-01 7.34956324e-01 -5.86550310e-02 -1.77835792e-01 6.63759112e-01 4.98039812e-01 -1.19767189e+00 2.72251695e-01 5.90290546e-01 4.46871281e-01 -7.50810564e-01 7.69741297e-01 1.19317256e-01 -1.22103298e+00 -1.03789782e-02 -4.61316139e-01 -1.82706177e-01 2.28491291e-01 2.81660467e-01 -1.08909333e+00 1.00808740e+00 3.86985958e-01 6.11449361e-01 -8.73398662e-01 6.50608122e-01 -6.91158116e-01 7.97451854e-01 4.17537689e-02 -5.72264731e-01 4.92362827e-01 8.13805759e-02 3.86418968e-01 1.20569849e+00 1.30940408e-01 2.46367782e-01 4.70609078e-03 9.58930910e-01 -3.59871686e-01 3.68353188e-01 -4.90696579e-01 -4.84960169e-01 6.02617562e-01 8.87403965e-01 -4.17167842e-01 -4.26990777e-01 -7.35260129e-01 9.33398128e-01 7.51775324e-01 5.16501248e-01 -3.64123315e-01 -5.95557928e-01 6.05140269e-01 -1.27047282e-02 1.61677927e-01 -3.52459788e-01 -1.90400615e-01 -1.22280884e+00 1.60915956e-01 -1.08143103e+00 1.04368174e+00 -1.09291470e+00 -1.92678583e+00 7.04153717e-01 -5.85381575e-02 -7.31612206e-01 -2.78989494e-01 -6.15079641e-01 -5.67248821e-01 9.91380215e-01 -1.89922750e+00 -1.23111224e+00 -2.01558307e-01 7.06448972e-01 6.60129130e-01 2.71754619e-02 9.25285995e-01 1.20669276e-01 -8.04207399e-02 6.54949784e-01 -4.66171980e-01 3.24784219e-01 5.41075230e-01 -1.37491333e+00 8.22164893e-01 7.30811119e-01 5.13404012e-01 8.84348035e-01 4.71146494e-01 -6.70704484e-01 -1.53015614e+00 -1.08557713e+00 1.35290110e+00 -7.83586204e-01 9.66377378e-01 -1.33690059e-01 -1.44343233e+00 6.03063464e-01 3.79470170e-01 -2.07124084e-01 8.91582072e-01 2.38407642e-01 -8.96621585e-01 1.42348781e-01 -9.71184254e-01 6.58731580e-01 1.10640180e+00 -1.04700112e+00 -1.31078482e+00 3.85830492e-01 1.41361511e+00 -3.43603194e-01 -1.14684618e+00 2.07287043e-01 1.88706741e-02 -5.39947987e-01 1.06721640e+00 -9.91073310e-01 5.34520388e-01 -1.59667641e-01 -3.60127896e-01 -1.35528827e+00 -1.51462913e-01 -5.81176221e-01 -4.30350035e-01 1.16078794e+00 6.41728938e-01 -8.28677833e-01 4.20036733e-01 7.02207565e-01 -4.03160423e-01 -7.71914959e-01 -7.47394323e-01 -7.72014260e-01 3.26312304e-01 -5.69986582e-01 7.63981938e-01 9.95845377e-01 7.00898916e-02 5.96235752e-01 2.84857422e-01 2.65997082e-01 2.12432116e-01 7.07142577e-02 5.08314252e-01 -9.88794148e-01 -2.87700534e-01 -4.16339070e-01 -2.05725402e-01 -1.29155302e+00 3.79786462e-01 -1.14589691e+00 -2.52767265e-01 -2.34128475e+00 -1.68026850e-01 -2.82214880e-01 -3.35477173e-01 3.44346195e-01 -7.26345003e-01 4.38950248e-02 2.34492719e-01 -1.33333758e-01 -1.04677129e+00 6.11209571e-01 1.48485506e+00 1.06734526e-03 6.46073967e-02 -2.91921437e-01 -1.15827429e+00 2.67570704e-01 1.02406299e+00 -3.00660789e-01 -7.15668917e-01 -7.52575457e-01 5.58075309e-01 1.40352413e-01 4.01436925e-01 -6.40695393e-01 2.45071873e-01 3.97256017e-02 3.95640684e-03 -3.24566364e-01 2.59772897e-01 -1.68869972e-01 -4.44956481e-01 1.32419154e-01 -7.00209975e-01 4.74048972e-01 3.18623662e-01 6.27262712e-01 -6.73979759e-01 -1.94324970e-01 7.13271275e-02 -2.78658330e-01 -8.53703558e-01 1.69171929e-01 -1.66089535e-01 8.76936078e-01 2.33288944e-01 1.22085316e-02 -7.48855352e-01 -1.00771213e+00 -3.23435873e-01 5.45478463e-01 -2.63078362e-02 8.29124033e-01 9.30475533e-01 -1.35390091e+00 -8.76459479e-01 -2.45351627e-01 5.56484818e-01 1.80958256e-01 4.23507899e-01 4.07089233e-01 -5.17096281e-01 6.97053790e-01 3.34481537e-01 -1.14745155e-01 -9.16441977e-01 3.88198227e-01 1.63017482e-01 -6.58047259e-01 -4.57395315e-01 1.08768010e+00 -3.50341238e-02 -6.72612190e-01 -1.13162436e-01 -3.90247405e-01 -5.58375299e-01 -5.99093102e-02 5.12692511e-01 3.80734712e-01 1.87675759e-01 -4.47089940e-01 -1.50341848e-02 3.71366858e-01 -1.80514768e-01 1.22256507e-03 1.09954834e+00 -2.03200474e-01 -2.93544710e-01 2.33887643e-01 1.29092121e+00 -6.38841093e-02 -7.34242141e-01 -6.49753034e-01 4.32680517e-01 -2.43610442e-01 -2.06641495e-01 -9.02480662e-01 -4.59315300e-01 7.67817378e-01 -1.68498918e-01 4.20881182e-01 9.61826742e-01 3.36824715e-01 1.46972191e+00 7.75099754e-01 1.97718605e-01 -8.80287111e-01 6.74369693e-01 9.34377372e-01 1.15150857e+00 -9.87882435e-01 -3.86421978e-01 -4.26292062e-01 -9.03715432e-01 1.14568710e+00 7.42339194e-01 -1.25164568e-01 8.94145295e-02 -4.09375995e-01 2.15847522e-01 -4.91943300e-01 -9.05210853e-01 -4.79077816e-01 7.37313986e-01 5.61039269e-01 4.21987236e-01 -8.40739831e-02 -1.05909526e-01 5.87959588e-01 -4.95518237e-01 -3.68745685e-01 4.45807040e-01 8.42623234e-01 -5.56172073e-01 -1.01733649e+00 -1.05727114e-01 6.02106690e-01 -4.69097525e-01 -5.61157227e-01 -6.61064863e-01 6.06180847e-01 -5.77641487e-01 1.54185557e+00 -2.15615816e-02 -1.81523412e-01 6.49823427e-01 4.09408599e-01 3.95412177e-01 -8.42985570e-01 -6.78793728e-01 -6.55477464e-01 6.78613305e-01 -6.10869408e-01 -1.98928893e-01 -1.86423495e-01 -1.62823546e+00 -1.29763484e-01 -1.56889588e-01 4.54068303e-01 3.47474635e-01 1.01092327e+00 6.35333180e-01 6.12374067e-01 2.56317914e-01 2.06787005e-01 -5.88516653e-01 -9.71031368e-01 -3.26128811e-01 7.05007732e-01 2.63171792e-01 -3.33721250e-01 -2.55694181e-01 -2.31595084e-01]
[10.902263641357422, 7.998457908630371]
25df81e4-72fc-46b6-b448-09aafb914b3a
non-local-spatial-propagation-network-for
2007.10042
null
https://arxiv.org/abs/2007.10042v1
https://arxiv.org/pdf/2007.10042v1.pdf
Non-Local Spatial Propagation Network for Depth Completion
In this paper, we propose a robust and efficient end-to-end non-local spatial propagation network for depth completion. The proposed network takes RGB and sparse depth images as inputs and estimates non-local neighbors and their affinities of each pixel, as well as an initial depth map with pixel-wise confidences. The initial depth prediction is then iteratively refined by its confidence and non-local spatial propagation procedure based on the predicted non-local neighbors and corresponding affinities. Unlike previous algorithms that utilize fixed-local neighbors, the proposed algorithm effectively avoids irrelevant local neighbors and concentrates on relevant non-local neighbors during propagation. In addition, we introduce a learnable affinity normalization to better learn the affinity combinations compared to conventional methods. The proposed algorithm is inherently robust to the mixed-depth problem on depth boundaries, which is one of the major issues for existing depth estimation/completion algorithms. Experimental results on indoor and outdoor datasets demonstrate that the proposed algorithm is superior to conventional algorithms in terms of depth completion accuracy and robustness to the mixed-depth problem. Our implementation is publicly available on the project page.
['Chi-Kuei Liu', 'Kyungdon Joo', 'Jinsun Park', 'In So Kweon', 'Zhe Hu']
2020-07-20
null
https://www.ecva.net/papers/eccv_2020/papers_ECCV/html/1810_ECCV_2020_paper.php
https://www.ecva.net/papers/eccv_2020/papers_ECCV/papers/123580120.pdf
eccv-2020-8
['stereo-lidar-fusion']
['computer-vision']
[ 2.64235169e-01 6.35313764e-02 -3.00499558e-01 -7.13844538e-01 -6.36268735e-01 -2.17907622e-01 2.65026629e-01 2.19205216e-01 -6.32818103e-01 7.30042040e-01 2.71165282e-01 2.73615181e-01 -2.09047735e-01 -9.49315608e-01 -5.97163498e-01 -7.19958961e-01 9.46049287e-04 4.40358669e-01 6.88528359e-01 1.33924097e-01 4.89695549e-01 5.21176279e-01 -1.56681347e+00 1.80106819e-01 8.94843340e-01 1.35160398e+00 3.55995715e-01 7.12845922e-01 -4.67091464e-02 8.38547885e-01 -9.74799693e-02 3.06105483e-02 2.94098347e-01 -1.82557642e-01 -6.74192607e-01 -9.62026194e-02 7.95607686e-01 -6.92688823e-01 -3.82044256e-01 1.10753465e+00 5.64599216e-01 3.80727589e-01 4.30098653e-01 -8.82945180e-01 -3.52758855e-01 2.91434526e-01 -8.28795731e-01 1.42480835e-01 2.39267871e-01 -1.45811111e-01 7.80417085e-01 -1.12935734e+00 4.93587524e-01 1.03706980e+00 7.10061193e-01 4.07658756e-01 -8.31674397e-01 -5.01984715e-01 4.74511683e-01 1.44561931e-01 -1.67841506e+00 -2.17560545e-01 9.02800977e-01 -2.46418417e-01 8.41638803e-01 -5.09300828e-02 7.09510028e-01 5.06069064e-01 9.57407653e-02 7.26868331e-01 7.68646181e-01 -3.34838271e-01 5.28536856e-01 -1.20479271e-01 2.95952279e-02 7.84768999e-01 -8.84878507e-04 -2.65469458e-02 -7.86152840e-01 6.12176321e-02 1.16752088e+00 -6.20482080e-02 -4.19429541e-01 -5.98469496e-01 -1.26394916e+00 4.98785734e-01 8.99583578e-01 1.00154459e-01 -4.46566552e-01 3.89037460e-01 -2.27446213e-01 -3.10759813e-01 4.82056350e-01 -1.27125889e-01 -3.35203856e-01 -1.00503631e-01 -1.10608625e+00 1.12485744e-01 3.93059105e-01 8.51601541e-01 1.34776676e+00 -4.17335898e-01 -1.39720291e-02 7.92080581e-01 4.81513768e-01 2.36199811e-01 3.15806538e-01 -1.26281095e+00 6.51767015e-01 5.93002439e-01 2.24864885e-01 -1.13517547e+00 -4.21346724e-01 -3.39391023e-01 -9.74464357e-01 4.07028466e-01 5.01687586e-01 8.42780173e-02 -1.02981019e+00 1.56350684e+00 5.54192662e-01 5.50932705e-01 -4.84744385e-02 1.21698773e+00 8.67903709e-01 5.81971705e-01 -2.45468199e-01 1.63526647e-02 5.23422360e-01 -1.17924631e+00 -5.91443181e-01 -6.18269742e-01 1.92527741e-01 -5.35559475e-01 8.42269361e-01 2.69951403e-01 -1.04323745e+00 -5.91771901e-01 -1.11087859e+00 -4.01571780e-01 -1.00581788e-01 3.09174601e-02 8.26728940e-01 3.64112854e-01 -1.37553728e+00 5.83895683e-01 -9.29407954e-01 -2.97061950e-01 2.66550303e-01 6.05379045e-01 -2.17328101e-01 -4.83130485e-01 -8.31336737e-01 3.20373535e-01 3.32240522e-01 3.68774027e-01 -5.15144050e-01 -5.65974832e-01 -9.27743614e-01 -2.27418974e-01 -1.89615060e-02 -7.57883132e-01 1.04720366e+00 -9.54938412e-01 -1.49485457e+00 5.05347431e-01 -5.58067679e-01 2.20321640e-02 4.79002684e-01 -1.17210597e-01 3.20113562e-02 3.10976595e-01 3.86381626e-01 1.10983288e+00 4.98451203e-01 -1.22531259e+00 -1.11663437e+00 -6.41060114e-01 1.53548032e-01 6.19274437e-01 -2.90682822e-01 -6.83330178e-01 -1.07312751e+00 -4.21217650e-01 1.02996004e+00 -6.31308734e-01 -4.14522290e-01 5.99517107e-01 -2.69446075e-01 1.96438074e-01 5.62011838e-01 -1.86955199e-01 9.75940526e-01 -2.13336444e+00 1.77396685e-01 5.79494596e-01 3.71745229e-01 -4.15677547e-01 3.09353713e-02 1.37279779e-01 3.16148937e-01 -3.78496289e-01 -3.80996019e-01 -6.90014064e-01 -2.02175364e-01 1.72308892e-01 2.47775227e-01 6.69303596e-01 -1.68427005e-01 4.54218537e-01 -1.15187216e+00 -4.44297940e-01 5.90267420e-01 9.48450089e-01 -5.16877115e-01 7.97947720e-02 8.39629769e-02 5.69597304e-01 -4.75312144e-01 9.27673817e-01 9.49413359e-01 -1.67150036e-01 -2.33456075e-01 -3.56848389e-01 -2.37463012e-01 8.21388811e-02 -1.40086937e+00 2.03430223e+00 -3.91661912e-01 4.34381336e-01 7.31758326e-02 -4.60795313e-01 9.17770684e-01 -1.34482756e-01 4.53113377e-01 -5.77118754e-01 -9.03969109e-02 3.19000423e-01 -4.01039988e-01 -1.29397377e-01 7.06920803e-01 1.43413454e-01 3.56672198e-01 1.84929103e-01 -2.24579394e-01 -9.85886604e-02 -3.05946507e-02 1.61391109e-01 1.05632830e+00 2.02106908e-01 1.93574015e-04 -7.61460960e-02 6.04261994e-01 -1.39799237e-01 8.38203371e-01 6.27360642e-01 -3.41714591e-01 1.14899623e+00 7.41826072e-02 -5.15109837e-01 -6.00922525e-01 -1.08754909e+00 -1.57397702e-01 7.76737750e-01 9.95624304e-01 -3.05736929e-01 -6.25619829e-01 -4.03034389e-01 -1.07572138e-01 1.87085316e-01 -6.79578304e-01 1.88889191e-01 -3.40748161e-01 -5.04719019e-01 2.39280835e-01 7.59414375e-01 9.08934593e-01 -7.05900490e-01 -4.19862300e-01 2.47714087e-01 -3.22888076e-01 -1.08054411e+00 -4.58131224e-01 2.42536753e-01 -1.16741800e+00 -8.19459200e-01 -8.85964513e-01 -9.72245216e-01 1.09562051e+00 3.76001686e-01 7.02646255e-01 -5.05052805e-02 4.68024872e-02 2.45805144e-01 -1.74328178e-01 1.54258668e-01 3.61920089e-01 4.81998622e-02 -1.44446045e-01 6.81846142e-02 3.44658762e-01 -7.88907945e-01 -1.28930426e+00 2.32263446e-01 -7.23114550e-01 5.08182086e-02 6.06968164e-01 6.55071557e-01 1.22309530e+00 2.85130054e-01 7.46642202e-02 -6.43436253e-01 1.86498269e-01 -2.64804453e-01 -6.67023838e-01 1.01597188e-02 -5.33869982e-01 2.01155897e-02 2.50140935e-01 -1.48075417e-01 -1.13809776e+00 6.64588571e-01 -1.67981699e-01 -1.76827803e-01 -1.23353399e-01 3.06865096e-01 -2.25804806e-01 -5.18803716e-01 4.04265106e-01 1.39096230e-01 -4.97540534e-01 -2.85879105e-01 1.93362370e-01 2.64721602e-01 7.35022545e-01 -3.80848318e-01 6.93374634e-01 8.11526775e-01 -8.29134975e-03 -6.49490654e-01 -6.17427647e-01 -6.44671500e-01 -9.40941095e-01 -3.31979334e-01 8.19461167e-01 -1.14131999e+00 -4.91137117e-01 7.93124020e-01 -1.12359715e+00 -3.76824647e-01 -1.25478998e-01 6.73340738e-01 -3.30206782e-01 5.35618365e-01 -8.17182004e-01 -8.12522948e-01 -9.92268547e-02 -1.01772821e+00 1.15811694e+00 5.05338669e-01 -4.05823253e-02 -1.30065668e+00 1.22204788e-01 2.28181005e-01 3.43190104e-01 1.38339147e-01 5.19195020e-01 2.93336689e-01 -1.12057316e+00 -1.42301813e-01 -3.90882581e-01 1.02632508e-01 1.20296016e-01 -4.74913642e-02 -1.17838871e+00 -1.57941073e-01 -1.74476147e-01 -1.70581847e-01 1.02547693e+00 7.51215100e-01 9.78088140e-01 1.04504943e-01 -3.49300921e-01 9.80234921e-01 1.65434778e+00 -5.46668731e-02 6.48710668e-01 4.77467328e-01 9.72490370e-01 5.34225404e-01 8.50220859e-01 7.09129274e-01 8.06630850e-01 2.75353283e-01 8.11008632e-01 -3.46818268e-01 -7.35447556e-02 -3.31955403e-01 1.83148123e-03 5.86347222e-01 -8.71722922e-02 -2.83858269e-01 -7.25911438e-01 6.45264626e-01 -1.98214507e+00 -6.73797071e-01 -2.43317544e-01 2.28899860e+00 6.38191104e-01 2.38779441e-01 -2.51438707e-01 1.48671284e-01 7.13496685e-01 2.28103921e-01 -6.54546559e-01 5.29971868e-02 -3.21629554e-01 1.90569952e-01 7.70826161e-01 1.01520443e+00 -1.08240628e+00 8.70251954e-01 5.95400524e+00 3.98071557e-01 -9.70203638e-01 -5.26408888e-02 7.83871055e-01 -2.51262218e-01 -5.15500247e-01 -3.57435197e-02 -8.33542287e-01 1.75938070e-01 5.44009991e-02 3.87255967e-01 3.02656054e-01 8.78747225e-01 2.08093137e-01 -8.40404809e-01 -1.04135442e+00 1.23051584e+00 3.61743160e-02 -1.17038858e+00 -1.95523679e-01 -1.20161451e-01 1.08250511e+00 1.45941839e-01 3.71077284e-02 -2.63819665e-01 -4.29606773e-02 -7.38029897e-01 7.36568213e-01 4.65184063e-01 5.58169544e-01 -8.90111804e-01 9.08808947e-01 3.39806646e-01 -1.40417421e+00 -4.37259078e-02 -5.70787489e-01 -3.60155225e-01 7.59480000e-02 8.68004203e-01 -3.00734431e-01 1.88429177e-01 9.61602926e-01 1.01743758e+00 -4.59399521e-01 1.17430568e+00 -4.03843045e-01 -1.84959231e-03 -5.96790493e-01 8.29564631e-02 3.12848717e-01 -1.04597025e-01 -2.51685362e-02 9.70228851e-01 3.76113325e-01 2.48112738e-01 1.53802121e-02 7.56446838e-01 8.18369612e-02 8.05122331e-02 -2.13472977e-01 6.30092919e-01 6.07845426e-01 1.10211551e+00 -1.05383933e+00 -2.39436805e-01 -4.01442856e-01 1.34642434e+00 4.95386869e-01 5.49956441e-01 -7.12847352e-01 -5.57374120e-01 6.60886168e-01 2.29813606e-01 2.71520466e-01 -1.39435276e-01 -4.87885594e-01 -9.61556017e-01 2.29857907e-01 -1.11479655e-01 2.22478256e-01 -7.73595333e-01 -1.14475799e+00 5.64522922e-01 -2.05170304e-01 -1.08202672e+00 -5.44688329e-02 -3.66960794e-01 -5.22678077e-01 9.10647869e-01 -1.97191370e+00 -1.03863823e+00 -9.56003010e-01 8.82579088e-01 3.66775781e-01 4.85052586e-01 4.92273301e-01 4.89249259e-01 -3.84567916e-01 4.69350755e-01 -8.36488605e-03 -3.01141888e-02 6.44909918e-01 -1.13524580e+00 1.74668759e-01 1.01136661e+00 -2.02715844e-01 4.08612877e-01 3.62093836e-01 -7.84867167e-01 -1.02656662e+00 -9.94373560e-01 9.82632995e-01 -6.01452924e-02 2.52933562e-01 -2.17455283e-01 -7.10842371e-01 4.87276942e-01 -2.08682910e-01 3.40924740e-01 3.03154826e-01 -2.26571634e-01 -9.33111161e-02 -3.70177954e-01 -1.24490619e+00 3.65497977e-01 1.23747838e+00 -5.15172362e-01 -3.52096632e-02 2.97599226e-01 4.67142463e-01 -7.79375136e-01 -6.04430854e-01 5.38638175e-01 5.56534767e-01 -1.56503606e+00 9.60225165e-01 6.82427347e-01 3.58971506e-01 -6.50145292e-01 -3.73250782e-01 -7.70323157e-01 -4.82869774e-01 -1.41278401e-01 7.37061873e-02 1.06240821e+00 3.91533911e-01 -4.31183487e-01 1.37894619e+00 8.13430786e-01 -2.26687014e-01 -7.41897106e-01 -9.94978130e-01 -2.67996073e-01 -4.47498590e-01 -6.23521447e-01 4.12501454e-01 7.88794816e-01 -2.52984554e-01 5.03332261e-03 -2.89147258e-01 7.81383216e-01 9.33091879e-01 8.53478163e-02 7.02313423e-01 -9.05281544e-01 -3.85074615e-01 -2.39289626e-01 -7.28694737e-01 -1.73196685e+00 -6.56564459e-02 -4.83931571e-01 4.10978824e-01 -1.99133372e+00 1.26132131e-01 -7.08232582e-01 -3.74565125e-01 3.74143124e-01 -1.30063131e-01 7.16219425e-01 -4.55354184e-01 2.18495235e-01 -6.68677032e-01 6.18959606e-01 1.25648594e+00 -1.78474918e-01 -4.57840443e-01 -4.63189632e-02 -2.60386288e-01 8.23368013e-01 6.72525525e-01 -3.91343921e-01 -7.76017785e-01 -7.04095304e-01 1.49230897e-01 -4.98983152e-02 2.74460852e-01 -1.42618906e+00 6.48595691e-01 -2.11051747e-01 7.06589997e-01 -9.93770123e-01 8.23777497e-01 -7.89063454e-01 -6.64345995e-02 1.91543058e-01 -1.63157046e-01 -5.67255355e-02 -1.35467947e-02 7.34285295e-01 -4.25993323e-01 -8.15845951e-02 7.18478262e-01 -1.49649829e-01 -1.18659401e+00 4.96782720e-01 -1.48226321e-01 -2.03109547e-01 9.92070973e-01 -8.91878307e-01 -1.09790757e-01 -6.71589851e-01 -6.02739513e-01 1.94517225e-01 7.68894136e-01 3.39047797e-02 1.10830903e+00 -1.20418167e+00 -3.20778519e-01 4.03780639e-01 2.51403570e-01 5.83916306e-01 3.86894286e-01 7.33071923e-01 -9.59542930e-01 7.64755979e-02 -1.77655101e-01 -8.10983479e-01 -9.35174346e-01 5.99554479e-02 5.38061559e-01 2.29726493e-01 -5.00191152e-01 1.46827126e+00 4.25034136e-01 -3.27201009e-01 6.81532443e-01 -6.05977535e-01 5.97316846e-02 -4.30791736e-01 5.38363039e-01 2.82757044e-01 -4.01430391e-02 -6.44521892e-01 -6.23755932e-01 1.09868133e+00 5.11658378e-02 -2.05256373e-01 1.19880688e+00 -5.77666461e-01 -2.53957301e-01 5.08685589e-01 1.16210163e+00 2.93995608e-02 -1.60766017e+00 -2.41000026e-01 -4.17528629e-01 -8.88710260e-01 4.57792103e-01 -5.95571399e-01 -1.37836552e+00 7.38031924e-01 7.92072475e-01 -5.75107813e-01 1.30357420e+00 -8.52081031e-02 8.17933500e-01 2.12472081e-01 5.32336295e-01 -1.04574513e+00 2.00885341e-01 5.51377177e-01 3.72103184e-01 -1.42451942e+00 1.61995813e-01 -6.34349167e-01 -2.10054189e-01 1.00842631e+00 1.11434066e+00 -2.23755045e-03 7.65627503e-01 2.52938926e-01 3.06288272e-01 -4.77351062e-02 -2.64235526e-01 -3.80076133e-02 2.69715879e-02 8.05969954e-01 3.67540061e-01 -2.82721639e-01 2.27596564e-03 3.59221734e-02 4.29943316e-02 1.00391852e-02 3.73319268e-01 1.01400995e+00 -5.79456747e-01 -9.71137643e-01 -4.64966953e-01 1.28413402e-02 -4.80072126e-02 -2.42629707e-01 -3.19924235e-01 6.77709758e-01 4.42623377e-01 8.54161382e-01 3.35207105e-01 -5.35508096e-01 1.03197813e-01 -5.16245246e-01 5.22803903e-01 -4.68394399e-01 -2.11800039e-01 1.33469105e-01 -2.61927515e-01 -8.70222390e-01 -6.48789287e-01 -4.14631426e-01 -1.76030314e+00 -3.22359353e-01 -2.05477759e-01 -1.64861619e-01 7.31100798e-01 6.99793279e-01 3.27707976e-01 2.44498640e-01 4.70406443e-01 -1.27290583e+00 1.85758784e-01 -6.03755653e-01 -6.41760945e-01 1.58498287e-01 4.66997772e-01 -5.11361659e-01 -5.02570808e-01 -2.54858851e-01]
[8.85837459564209, -2.480748176574707]
d8ef0f89-2573-45ad-a854-847100c88ad6
semi-supervised-text-classification-via-self
2109.15300
null
https://arxiv.org/abs/2109.15300v1
https://arxiv.org/pdf/2109.15300v1.pdf
Semi-Supervised Text Classification via Self-Pretraining
We present a neural semi-supervised learning model termed Self-Pretraining. Our model is inspired by the classic self-training algorithm. However, as opposed to self-training, Self-Pretraining is threshold-free, it can potentially update its belief about previously labeled documents, and can cope with the semantic drift problem. Self-Pretraining is iterative and consists of two classifiers. In each iteration, one classifier draws a random set of unlabeled documents and labels them. This set is used to initialize the second classifier, to be further trained by the set of labeled documents. The algorithm proceeds to the next iteration and the classifiers' roles are reversed. To improve the flow of information across the iterations and also to cope with the semantic drift problem, Self-Pretraining employs an iterative distillation process, transfers hypotheses across the iterations, utilizes a two-stage training model, uses an efficient learning rate schedule, and employs a pseudo-label transformation heuristic. We have evaluated our model in three publicly available social media datasets. Our experiments show that Self-Pretraining outperforms the existing state-of-the-art semi-supervised classifiers across multiple settings. Our code is available at https://github.com/p-karisani/self_pretraining.
['Negin Karisani', 'Payam Karisani']
2021-09-30
null
null
null
null
['semi-supervised-text-classification-1']
['natural-language-processing']
[ 2.96914309e-01 2.81353980e-01 -6.03052914e-01 -7.21662521e-01 -4.61932302e-01 -4.02640402e-01 6.61909878e-01 3.45890492e-01 -5.81662834e-01 9.64633465e-01 -4.71892320e-02 -1.84713319e-01 1.90522879e-01 -6.23666465e-01 -7.47611105e-01 -7.77489603e-01 -5.56930155e-02 9.27581906e-01 4.08948451e-01 1.22010134e-01 1.57207102e-01 -6.49221390e-02 -1.67094445e+00 4.44868654e-01 9.67909455e-01 8.14416885e-01 1.81065558e-03 4.99860674e-01 -3.16380471e-01 7.76173174e-01 -4.00705725e-01 -2.58502752e-01 3.34024355e-02 -5.27506173e-01 -1.09378445e+00 2.82176465e-01 1.10033654e-01 -1.41629413e-01 6.33618385e-02 1.09354687e+00 3.04653198e-01 6.31184876e-02 5.56091487e-01 -1.23707163e+00 -5.47483087e-01 1.03472507e+00 -4.48205948e-01 9.34956148e-02 4.19380143e-03 -1.37912989e-01 6.59533858e-01 -9.43416059e-01 5.30753076e-01 1.12337637e+00 7.36661851e-01 8.67644310e-01 -1.15225387e+00 -8.90631735e-01 3.82045746e-01 2.54962202e-02 -1.19064939e+00 -3.95943165e-01 6.85919821e-01 -2.72793323e-01 4.93851364e-01 -7.82462023e-03 6.55511796e-01 1.14654851e+00 -1.47292629e-01 1.08744371e+00 1.32934451e+00 -7.75690556e-01 6.35564029e-01 5.80165923e-01 6.37441337e-01 7.12362349e-01 9.71608758e-02 1.05833054e-01 -6.05806828e-01 -2.78573811e-01 2.72363007e-01 1.86275184e-01 -3.03855222e-02 -4.35652971e-01 -9.14637387e-01 6.92397177e-01 3.65121812e-01 4.16314632e-01 -1.88635722e-01 -5.65325506e-02 3.63645524e-01 3.97457093e-01 8.49349618e-01 1.08886793e-01 -6.43741608e-01 -1.01797931e-01 -1.12859869e+00 -1.19600385e-01 8.75977635e-01 7.42402554e-01 1.08496785e+00 -4.31310713e-01 -9.52460095e-02 1.04358900e+00 3.49995106e-01 1.93236351e-01 1.06039202e+00 -7.58275568e-01 2.51811177e-01 8.08898330e-01 -3.90831046e-02 -1.64424285e-01 -3.65958601e-01 -6.10793114e-01 -7.56128967e-01 -8.22735131e-02 1.55715495e-01 -3.95094007e-01 -1.24547052e+00 1.68500376e+00 4.47365999e-01 4.78127301e-01 1.17724843e-01 5.28395593e-01 6.73586369e-01 4.81213242e-01 1.78588033e-01 -4.09599483e-01 1.01101267e+00 -1.24659264e+00 -6.08396709e-01 -5.27977526e-01 8.51994097e-01 -2.62412548e-01 9.05091286e-01 3.01871896e-01 -1.00150871e+00 -5.61359227e-01 -9.99066949e-01 4.10642236e-01 -4.67967749e-01 1.51307657e-01 5.32496154e-01 6.38060689e-01 -1.18153834e+00 8.03175330e-01 -9.61094797e-01 -5.06677747e-01 6.72900915e-01 3.36565137e-01 -2.22978964e-02 -7.82639682e-02 -1.14962935e+00 6.22566164e-01 8.14692795e-01 -1.49342582e-01 -7.75100470e-01 -4.19049263e-01 -6.96578979e-01 -7.93485418e-02 3.86506617e-01 -5.29491901e-01 1.58822918e+00 -1.37228441e+00 -1.53500652e+00 9.38173294e-01 -4.45935547e-01 -4.02462810e-01 6.67919993e-01 -2.48747885e-01 -3.20396304e-01 -1.68112963e-01 2.01356977e-01 8.54571044e-01 8.39152038e-01 -1.53678060e+00 -8.91649008e-01 -2.18906641e-01 -4.14046824e-01 3.33669573e-01 -6.42135084e-01 -1.35053664e-01 -7.63998449e-01 -3.71951044e-01 2.59824395e-01 -1.18131626e+00 -1.56199202e-01 -4.08573836e-01 -3.71151775e-01 -3.48827809e-01 7.83583820e-01 5.97049203e-03 1.24107695e+00 -2.32278371e+00 -2.23819584e-01 3.04182470e-01 1.44882530e-01 3.47293645e-01 -2.39307014e-03 2.37113476e-01 -1.52294576e-01 9.92337987e-02 -4.99294132e-01 -8.48580658e-01 -3.71801287e-01 2.65127569e-01 -1.36645272e-01 3.51787090e-01 3.63035128e-02 5.98909259e-01 -1.20598149e+00 -5.77593744e-01 -1.16245754e-01 2.17825890e-01 -4.13653463e-01 1.75038159e-01 -4.08742368e-01 2.73528397e-01 -3.96079570e-01 4.39965367e-01 5.56776226e-01 -7.59369433e-01 2.61356562e-01 4.08817261e-01 7.46348361e-03 3.03639472e-01 -1.17576349e+00 1.47149968e+00 -2.01299265e-01 3.46637219e-01 -3.03183198e-01 -1.03067684e+00 1.05558443e+00 1.67139485e-01 4.63927269e-01 -2.86256164e-01 2.67606050e-01 1.89173326e-01 -2.08672598e-01 -1.83331922e-01 2.33748645e-01 7.22333938e-02 2.67391920e-01 9.40783739e-01 2.19704211e-01 2.44083375e-01 4.85610783e-01 4.17384803e-01 9.34376478e-01 2.95620203e-01 1.06214903e-01 -7.75776803e-02 5.01337230e-01 9.17430967e-02 6.33792162e-01 9.16574240e-01 -1.63228676e-01 3.63432318e-01 2.40158677e-01 -4.44471806e-01 -7.28691936e-01 -9.60082650e-01 -2.14554489e-01 1.43200529e+00 2.95129735e-02 -3.70213836e-01 -8.00155699e-01 -1.17464101e+00 1.71536446e-01 7.96584010e-01 -9.03960586e-01 -4.10336167e-01 -1.72988951e-01 -8.76035571e-01 1.26942709e-01 4.01269972e-01 7.94400930e-01 -1.17786455e+00 -3.54421705e-01 3.41345847e-01 3.98259936e-03 -6.71574891e-01 -2.60655850e-01 6.47081912e-01 -1.15153503e+00 -9.73743677e-01 -4.68177974e-01 -9.75706816e-01 1.17202842e+00 2.08573446e-01 9.34793234e-01 1.88011795e-01 1.61223531e-01 9.03270319e-02 -5.82708418e-01 -6.88301623e-01 -8.00345302e-01 3.60736281e-01 1.67247891e-01 2.68747211e-01 6.69250965e-01 -4.91127551e-01 -4.15121436e-01 3.49843949e-01 -7.55311131e-01 9.57913622e-02 5.25274754e-01 9.38700974e-01 6.44018412e-01 1.73836052e-01 7.35770524e-01 -1.55972707e+00 6.09197915e-01 -7.55669832e-01 -4.35747892e-01 3.10780615e-01 -1.09202051e+00 1.48556277e-01 5.64359605e-01 -7.18702674e-01 -1.21345890e+00 2.32650205e-01 2.02737555e-01 -1.98095500e-01 -2.20622763e-01 5.52563071e-01 3.01459849e-01 2.93810576e-01 8.62094879e-01 3.05225819e-01 4.74123806e-02 -3.98314238e-01 3.65232140e-01 1.14651418e+00 4.92617756e-01 -2.99450904e-01 6.89700484e-01 4.82730031e-01 -7.04349399e-01 -4.35814083e-01 -1.36778879e+00 -5.55153072e-01 -7.36050725e-01 -3.91617656e-01 2.94598192e-01 -9.51921821e-01 -3.45924795e-01 6.49675310e-01 -6.69528782e-01 -7.19519913e-01 -4.90940273e-01 2.65835762e-01 -2.17224628e-01 1.65551141e-01 -5.99867523e-01 -8.80462050e-01 -4.45155233e-01 -6.64449036e-01 5.28132021e-01 5.11388063e-01 -2.84375757e-01 -1.10873067e+00 2.90908217e-01 2.36632437e-01 3.03956777e-01 -2.55623817e-01 4.19606894e-01 -1.13902080e+00 -3.44537012e-02 -3.35290968e-01 5.80334701e-02 4.70552772e-01 2.93233454e-01 -2.05881789e-01 -1.02296567e+00 -6.16723955e-01 -5.98311387e-02 -7.07056642e-01 1.17562556e+00 2.20072821e-01 1.14280927e+00 -2.43329644e-01 -6.90699816e-01 4.38640326e-01 1.04192340e+00 2.70246178e-01 2.74182409e-01 6.67767763e-01 4.58829939e-01 4.00237054e-01 5.57835281e-01 4.54237819e-01 6.23020053e-01 -8.96609295e-03 1.37105718e-01 -2.11207002e-01 1.21076912e-01 -3.23589325e-01 2.96365798e-01 8.02899480e-01 3.62690777e-01 -2.05249369e-01 -1.02088439e+00 5.37685752e-01 -2.18433189e+00 -6.81307256e-01 1.80777267e-01 2.14074230e+00 1.29458296e+00 4.46739018e-01 1.94201097e-01 3.33354980e-01 9.13364828e-01 -1.70471311e-01 -9.23138738e-01 -1.65996462e-01 4.26010601e-02 -8.96458328e-02 5.13882399e-01 3.40776592e-01 -1.23348236e+00 1.27905798e+00 6.16889143e+00 2.60073811e-01 -1.08303905e+00 2.60327458e-01 7.83638656e-01 -1.81854486e-01 -3.38372849e-02 1.23121358e-01 -9.78524864e-01 5.97797990e-01 1.02245855e+00 -1.24434315e-01 3.42789114e-01 9.33440566e-01 -1.85118571e-01 -3.71204287e-01 -1.21725965e+00 5.80999911e-01 1.92655161e-01 -1.22670233e+00 -1.03935905e-01 -2.68562168e-01 1.00554144e+00 4.22933012e-01 -1.17225014e-01 4.80304927e-01 9.27189410e-01 -4.18304741e-01 6.79464459e-01 2.22093150e-01 5.93106270e-01 -8.28465998e-01 7.82094240e-01 6.87692463e-01 -7.28754103e-01 -4.03630227e-01 -1.14413738e-01 8.94872844e-02 -1.30672812e-01 8.15685987e-01 -9.81777132e-01 1.93506077e-01 9.35916483e-01 7.82576680e-01 -8.25960636e-01 9.94472504e-01 -4.95269209e-01 1.05695784e+00 -2.59418935e-01 -1.49850100e-01 1.60290539e-01 -2.83260047e-02 1.21541090e-01 1.35514963e+00 -1.42483711e-01 -1.14316382e-02 3.48608971e-01 4.43633527e-01 -1.63634434e-01 1.29303932e-01 -1.83387533e-01 9.50879008e-02 8.29116702e-01 1.03970408e+00 -8.74249339e-01 -8.54902744e-01 -1.78101256e-01 1.01248908e+00 7.41097867e-01 4.37187731e-01 -4.58053768e-01 -2.98173755e-01 -4.45304550e-02 1.32026047e-01 1.93786070e-01 2.14403346e-01 -1.72158554e-01 -1.10413706e+00 -1.14672877e-01 -5.71267724e-01 6.66964054e-01 -6.21601522e-01 -1.20508993e+00 6.28903389e-01 -1.37187336e-02 -9.48737323e-01 -4.50697362e-01 -1.40714660e-01 -6.07737243e-01 5.25823176e-01 -1.63029444e+00 -8.90487313e-01 -3.80510956e-01 5.74266672e-01 4.44272667e-01 -2.25068346e-01 7.72119939e-01 1.70556642e-02 -8.32476199e-01 5.93818605e-01 4.17988151e-01 3.03761929e-01 1.06813192e+00 -1.35631514e+00 2.80222207e-01 6.40307188e-01 -2.14094706e-02 6.15060866e-01 4.23444450e-01 -8.93543124e-01 -7.82278836e-01 -1.32397163e+00 1.00491333e+00 -1.77084923e-01 5.15303612e-01 -4.26339954e-01 -1.13665009e+00 8.44821513e-01 7.12360963e-02 -7.10089952e-02 9.96267617e-01 2.69596636e-01 -2.24331975e-01 -1.82906270e-01 -1.01867986e+00 4.09548610e-01 9.62092221e-01 -1.78532094e-01 -6.56037807e-01 6.00970387e-01 4.66937989e-01 -2.89023727e-01 -5.25574088e-01 2.64687836e-01 2.61709720e-01 -8.37307990e-01 4.67653215e-01 -4.38205808e-01 2.41807148e-01 -1.02418430e-01 4.57667887e-01 -1.53175497e+00 -4.96452808e-01 -6.38565540e-01 -2.98941523e-01 1.44355488e+00 6.98503792e-01 -8.94952595e-01 9.98511016e-01 4.66692179e-01 5.89320511e-02 -6.97614312e-01 -6.69853508e-01 -9.15500641e-01 -8.72733966e-02 -2.66002178e-01 4.53694493e-01 1.18358231e+00 2.23684043e-01 4.36494499e-01 -8.99005160e-02 -8.26332420e-02 8.32291842e-01 1.55765954e-02 6.12402380e-01 -1.57920349e+00 -1.57588542e-01 -2.52138644e-01 2.75241643e-01 -9.63435829e-01 3.23700309e-01 -1.18348539e+00 2.87789226e-01 -1.44005573e+00 4.06875253e-01 -7.97825515e-01 -5.80925882e-01 1.11164391e+00 -3.63524199e-01 1.16904773e-01 -4.00693789e-02 6.09878182e-01 -9.15102482e-01 3.85635525e-01 8.85307789e-01 -6.35161847e-02 -7.19665289e-01 1.24296144e-01 -7.87575006e-01 6.86423123e-01 1.12784266e+00 -8.95549476e-01 -4.26192462e-01 -2.32764155e-01 2.42467336e-02 -3.34386498e-01 -1.60101548e-01 -9.87296999e-01 5.35347402e-01 1.33665195e-02 6.43604517e-01 -5.43319464e-01 -1.07066721e-01 -4.63346750e-01 -1.34385899e-01 7.56688416e-01 -6.70489311e-01 -2.30424866e-01 7.50356242e-02 5.31421900e-01 9.46301594e-03 -3.99449855e-01 9.92486060e-01 -1.61559895e-01 -4.46881086e-01 2.94294477e-01 -4.76721525e-01 8.43539983e-02 1.07307017e+00 -2.52326936e-01 -1.76766038e-01 -2.50684589e-01 -7.95940161e-01 6.96789145e-01 5.43266237e-01 5.68377793e-01 3.52612227e-01 -1.24107742e+00 -6.50258720e-01 2.90198237e-01 2.57077157e-01 3.22656572e-01 -8.91483948e-02 3.89280826e-01 7.18123419e-03 4.90749776e-02 1.35845423e-01 -8.32585037e-01 -1.14367974e+00 6.43172085e-01 1.32977962e-01 -2.99652278e-01 -3.52964252e-01 8.44898403e-01 -2.02713341e-01 -5.56958616e-01 5.01427829e-01 -6.99943900e-02 -3.33981216e-01 7.29391128e-02 5.70672870e-01 1.22021444e-01 7.10479042e-04 -2.59890825e-01 -3.76267046e-01 2.08705634e-01 -6.20787799e-01 -2.89967626e-01 1.35021782e+00 -1.50387749e-01 -2.94351857e-02 8.03512692e-01 1.03093016e+00 -4.24446404e-01 -1.38320231e+00 -8.48717630e-01 2.04930410e-01 -8.98948759e-02 1.16845503e-01 -9.93179679e-01 -8.84728134e-01 3.33269715e-01 5.36652923e-01 1.56123742e-01 1.04436159e+00 1.72719300e-01 4.23704743e-01 5.37992179e-01 2.26228893e-01 -1.46815145e+00 2.83550650e-01 7.04679251e-01 4.31829989e-01 -1.32104528e+00 -3.27536464e-02 -2.22128227e-01 -7.22015500e-01 8.16801429e-01 8.58716905e-01 5.09618595e-02 8.60221326e-01 3.36946368e-01 4.08817619e-01 1.30584121e-01 -1.00222147e+00 -1.04415141e-01 -5.05649783e-02 2.92595476e-01 3.91850859e-01 -9.77483615e-02 -1.79616556e-01 4.23805535e-01 -1.94881856e-01 3.68589520e-01 3.35544080e-01 1.33106339e+00 -6.21567726e-01 -1.33375919e+00 -3.60040784e-01 5.49748838e-01 -8.51594955e-02 1.59703437e-02 -5.46093702e-01 3.98953408e-01 2.84125507e-01 1.00288045e+00 3.08927089e-01 -4.58680838e-01 5.10857515e-02 3.82604390e-01 9.71428752e-02 -9.33733463e-01 -6.66320860e-01 3.07406317e-02 -4.45362041e-03 -2.16233999e-01 -5.63460469e-01 -8.72814596e-01 -1.41316676e+00 -4.19938490e-02 -5.52479446e-01 4.78777975e-01 7.42997527e-01 9.78295147e-01 3.38891208e-01 4.11607474e-01 9.89399433e-01 -6.34385109e-01 -4.85731840e-01 -1.19680846e+00 -2.04038456e-01 4.30239111e-01 2.36431181e-01 -6.49952233e-01 -4.99597698e-01 2.17012599e-01]
[9.665337562561035, 3.902024745941162]
7b0eb972-a6ce-4163-8ab7-501d3fd1f7c9
generative-adversarial-networks-a-survey-and
1906.01529
null
https://arxiv.org/abs/1906.01529v6
https://arxiv.org/pdf/1906.01529v6.pdf
Generative Adversarial Networks in Computer Vision: A Survey and Taxonomy
Generative adversarial networks (GANs) have been extensively studied in the past few years. Arguably their most significant impact has been in the area of computer vision where great advances have been made in challenges such as plausible image generation, image-to-image translation, facial attribute manipulation and similar domains. Despite the significant successes achieved to date, applying GANs to real-world problems still poses significant challenges, three of which we focus on here. These are: (1) the generation of high quality images, (2) diversity of image generation, and (3) stable training. Focusing on the degree to which popular GAN technologies have made progress against these challenges, we provide a detailed review of the state of the art in GAN-related research in the published scientific literature. We further structure this review through a convenient taxonomy we have adopted based on variations in GAN architectures and loss functions. While several reviews for GANs have been presented to date, none have considered the status of this field based on their progress towards addressing practical challenges relevant to computer vision. Accordingly, we review and critically discuss the most popular architecture-variant, and loss-variant GANs, for tackling these challenges. Our objective is to provide an overview as well as a critical analysis of the status of GAN research in terms of relevant progress towards important computer vision application requirements. As we do this we also discuss the most compelling applications in computer vision in which GANs have demonstrated considerable success along with some suggestions for future research directions. Code related to GAN-variants studied in this work is summarized on https://github.com/sheqi/GAN_Review.
['Zhengwei Wang', 'Tomas E. Ward', 'Qi She']
2019-06-04
null
null
null
null
['image-quality-estimation']
['computer-vision']
[ 8.46843064e-01 2.71799684e-01 -4.90291696e-03 -2.96661317e-01 -8.99388790e-01 -4.42217320e-01 7.34301150e-01 -5.59108675e-01 -2.82466203e-01 8.57420564e-01 -7.20583946e-02 -4.92257625e-02 1.40788078e-01 -6.28490090e-01 -4.16980475e-01 -9.22775209e-01 2.99811244e-01 1.74753085e-01 -3.58097553e-01 -2.68053532e-01 -6.01290772e-03 5.18403292e-01 -1.68247521e+00 7.59099424e-02 8.23665917e-01 9.01484668e-01 -9.51396078e-02 7.75232673e-01 1.15572684e-01 6.29422128e-01 -9.36092496e-01 -8.75723302e-01 3.11532021e-01 -1.16791141e+00 -5.20661950e-01 3.35944831e-01 6.50484264e-01 -4.07614768e-01 -1.85612291e-01 8.83605063e-01 9.15425479e-01 -7.92959854e-02 6.88668847e-01 -1.45912015e+00 -8.31339955e-01 9.55497324e-02 -6.16825700e-01 1.10803112e-01 6.85296655e-02 3.36473972e-01 5.63273013e-01 -6.70842052e-01 4.84169960e-01 9.53402579e-01 5.79292417e-01 1.26136982e+00 -9.59363103e-01 -7.10043907e-01 1.35665759e-01 2.69521624e-01 -1.16648579e+00 -6.34356797e-01 8.61944497e-01 -2.63885945e-01 8.40565920e-01 4.21677738e-01 7.26840198e-01 1.27277446e+00 2.18597755e-01 7.58700609e-01 1.20303798e+00 -6.28645778e-01 1.15245461e-01 2.53099413e-03 -5.57030439e-01 5.09123564e-01 1.76699050e-02 2.40373388e-01 -4.95030850e-01 1.21977299e-01 9.04999435e-01 -4.53986317e-01 -1.64207324e-01 -6.34840056e-02 -8.65529418e-01 9.79195654e-01 4.63296294e-01 3.76053989e-01 -3.81623685e-01 2.75504291e-01 2.84318537e-01 3.53751421e-01 5.51149547e-01 3.27368826e-01 1.71819981e-03 -7.80611262e-02 -9.65823650e-01 3.48075300e-01 3.53627414e-01 1.05325687e+00 4.71205205e-01 9.12520111e-01 -6.83210120e-02 9.91527379e-01 1.17671400e-01 4.48409706e-01 4.38902050e-01 -9.51437294e-01 1.09266333e-01 -8.80038142e-02 -3.42932016e-01 -8.08758020e-01 1.06030209e-02 -2.85207450e-01 -1.15346491e+00 6.94311619e-01 1.26830012e-01 -2.30065241e-01 -1.15278053e+00 1.81812394e+00 1.27979308e-01 1.58387065e-01 -3.85845974e-02 5.34257114e-01 1.14726925e+00 4.50684518e-01 2.04322726e-01 -1.64637059e-01 1.26176929e+00 -1.06838393e+00 -7.76893616e-01 -3.96372586e-01 -1.37115732e-01 -1.10102439e+00 6.98984027e-01 2.60356873e-01 -1.49974442e+00 -6.81858003e-01 -1.05559766e+00 -1.00787021e-01 -2.68397719e-01 1.20615944e-01 7.74532437e-01 1.14069343e+00 -1.45106542e+00 2.84888119e-01 -8.67894292e-01 -7.43194044e-01 7.44253635e-01 3.07601362e-01 -2.97687918e-01 -4.39547934e-02 -8.63977730e-01 9.30687189e-01 -4.83970195e-02 2.53127873e-01 -9.23645854e-01 -5.53757966e-01 -7.71851718e-01 -4.23438042e-01 2.29700387e-01 -1.29239225e+00 1.24526107e+00 -1.44991052e+00 -1.72419262e+00 1.34129941e+00 -1.78352311e-01 -5.05982161e-01 7.17540741e-01 -6.57711923e-02 -4.50955242e-01 9.09549892e-02 -1.41091481e-01 1.23839855e+00 1.13304269e+00 -1.23665774e+00 -6.79281771e-01 -1.97112396e-01 -1.12288781e-02 2.79327661e-01 -1.22604810e-01 3.21049184e-01 -4.43691552e-01 -1.12731743e+00 -4.39254701e-01 -1.11494911e+00 -1.76408306e-01 2.20767155e-01 -2.28861853e-01 -9.33487490e-02 7.26725340e-01 -4.24060524e-01 9.56833959e-01 -1.97656989e+00 1.28853813e-01 -4.05247808e-01 1.38799638e-01 6.07812405e-01 -3.00194085e-01 5.71331382e-01 -8.89492258e-02 1.42268598e-01 -6.44815803e-01 -6.22213840e-01 -2.87583053e-01 7.99949393e-02 -4.92040962e-01 3.75782490e-01 4.19450223e-01 1.41187620e+00 -5.83871603e-01 -2.62515306e-01 4.91210371e-01 9.96043384e-01 -2.20272288e-01 3.61257702e-01 1.20849656e-02 7.96728492e-01 -1.27388373e-01 8.09318244e-01 7.23337829e-01 2.07469299e-01 -9.56433937e-02 -1.27890214e-01 1.10011742e-01 -3.08855474e-02 -5.68690896e-01 1.49648511e+00 -4.01254416e-01 8.54339480e-01 2.86484718e-01 -1.02924931e+00 9.07235622e-01 5.11615217e-01 3.77810419e-01 -4.96578813e-01 4.28177491e-02 1.87160254e-01 2.57629575e-03 -1.25631630e-01 2.28527889e-01 -4.56603318e-01 3.07561934e-01 1.81898713e-01 1.47010118e-01 -6.35556340e-01 6.02020435e-02 -4.87271324e-03 7.10268259e-01 4.61431593e-01 5.39835751e-01 1.01703353e-01 6.60495818e-01 -1.30016133e-01 3.28467697e-01 6.87132418e-01 -4.43423629e-01 1.03402853e+00 1.99572697e-01 -3.22370559e-01 -1.05458391e+00 -1.03339231e+00 1.19798966e-02 6.95230484e-01 -2.86907464e-01 -2.75151640e-01 -1.03605652e+00 -5.32556057e-01 -4.74299580e-01 5.91015935e-01 -8.48227680e-01 3.51666026e-02 -6.55570984e-01 -1.01715028e+00 7.88374722e-01 5.24802089e-01 7.47030675e-01 -1.54564488e+00 -5.96544087e-01 -3.40043195e-02 -5.46557605e-02 -1.06683123e+00 -1.68303564e-01 -3.04311574e-01 -1.02691674e+00 -8.37061048e-01 -1.13963294e+00 -8.72319460e-01 7.66314805e-01 2.36315608e-01 1.30582047e+00 2.50854075e-01 -4.14756656e-01 7.58864641e-01 -3.42519492e-01 -7.83161461e-01 -5.78736067e-01 -8.97748545e-02 -2.84862787e-01 -1.76542521e-01 2.24673435e-01 -6.65241301e-01 -6.93120658e-01 2.82849014e-01 -1.27159441e+00 1.14227042e-01 6.33748651e-01 9.52311695e-01 6.95071876e-01 -1.01442434e-01 8.82661641e-01 -9.68387663e-01 6.71042979e-01 -2.89104611e-01 -4.86285985e-01 6.64605722e-02 -7.60322988e-01 -5.48958302e-01 4.64965254e-01 2.24154312e-02 -1.10519433e+00 -8.88973624e-02 -7.19186306e-01 -1.22652672e-01 -4.62289274e-01 1.49655685e-01 -1.42996222e-01 -3.89005214e-01 6.29490137e-01 3.85709673e-01 3.64472330e-01 -1.17534116e-01 5.30173063e-01 5.66837192e-01 6.21233165e-01 -4.66569245e-01 8.98815155e-01 6.32649779e-01 1.55195951e-01 -1.02785969e+00 -5.38970828e-01 2.04162821e-02 -4.91991907e-01 -2.49453977e-01 9.07947898e-01 -7.21912563e-01 -3.05054188e-01 1.00250959e+00 -8.69449139e-01 -4.46452796e-01 -5.08520842e-01 5.75276501e-02 -1.00773346e+00 1.90098539e-01 -6.60605967e-01 -7.69227266e-01 -7.46817768e-01 -1.47647130e+00 8.99606407e-01 5.18227816e-01 -2.07925618e-01 -1.30676365e+00 -7.93780610e-02 6.43312156e-01 7.77696133e-01 7.00969756e-01 7.41687953e-01 2.37781890e-02 -6.52737081e-01 -7.52358660e-02 8.81815180e-02 6.79919541e-01 4.11823273e-01 1.51167855e-01 -1.29898548e+00 -3.96993130e-01 1.45120218e-01 -4.53757167e-01 7.61861861e-01 7.52871990e-01 1.04340947e+00 -7.76599050e-02 -1.43342599e-01 9.12896216e-01 1.31913626e+00 4.45802391e-01 1.01052606e+00 3.14310819e-01 5.82107604e-01 6.36552274e-01 5.76296747e-01 -2.02244967e-02 7.47953802e-02 8.04693222e-01 4.74689960e-01 -3.83250803e-01 -8.13298702e-01 -1.68292210e-01 2.74426699e-01 6.38963580e-01 -5.31626821e-01 -3.98521066e-01 -3.97088766e-01 4.63767618e-01 -1.50933158e+00 -9.77166235e-01 1.40669599e-01 2.07990599e+00 6.02656186e-01 -2.78945208e-01 3.18111420e-01 -2.91811544e-02 7.09714592e-01 3.21109474e-01 -7.17009366e-01 -5.39521813e-01 -4.46270764e-01 5.95058858e-01 1.01436555e-01 4.31974232e-01 -1.08960426e+00 9.58522081e-01 7.10427332e+00 9.30561721e-01 -1.35590589e+00 6.00041524e-02 9.63164151e-01 -3.30698974e-02 -2.64947116e-01 -2.39768133e-01 -5.78404427e-01 3.35976958e-01 6.61795557e-01 -1.13910131e-01 5.16902745e-01 7.61393666e-01 -1.20526850e-01 -4.02514562e-02 -9.41226482e-01 1.12026286e+00 4.87563491e-01 -1.39781296e+00 1.58892289e-01 1.47635981e-01 1.01658750e+00 -1.17319385e-02 6.12226725e-01 3.42940465e-02 6.47307336e-02 -1.45830417e+00 5.63121676e-01 2.06698045e-01 1.26912415e+00 -8.00719023e-01 5.39357066e-01 -6.82088584e-02 -9.21026051e-01 3.97077769e-01 -3.56318295e-01 1.17574260e-01 2.21690953e-01 4.24603552e-01 -5.59615016e-01 6.51250005e-01 5.38670421e-01 8.35466385e-01 -3.18333298e-01 8.59547436e-01 -6.48798287e-01 6.42251670e-01 -5.65142892e-02 2.82923311e-01 1.42435312e-01 -4.09299374e-01 5.05121350e-01 1.00657201e+00 4.04162258e-01 4.33928594e-02 -3.97898555e-01 9.05643702e-01 -7.79872760e-02 -1.02074832e-01 -6.10393882e-01 -1.95575114e-02 3.15413654e-01 1.29703748e+00 -6.93530560e-01 -3.86658795e-02 -5.83474457e-01 1.24333906e+00 1.68009952e-01 4.05641258e-01 -8.12792063e-01 -3.21687579e-01 9.07247603e-01 -2.13668924e-02 1.52971178e-01 -9.91315618e-02 -3.98983598e-01 -8.21646810e-01 -7.23830098e-03 -1.22629452e+00 4.94646430e-01 -8.52022171e-01 -1.32610714e+00 1.04618454e+00 2.76144617e-03 -1.10056710e+00 -5.54581404e-01 -4.60195214e-01 -6.62983835e-01 1.10069132e+00 -1.50062633e+00 -1.51250517e+00 -4.79606688e-01 4.06939685e-01 8.36052358e-01 -4.27246273e-01 1.18049169e+00 1.05277784e-01 -5.83203077e-01 8.56266618e-01 -1.17446274e-01 -1.39391005e-01 7.52270162e-01 -9.13498044e-01 9.26119089e-01 9.25403059e-01 2.43105650e-01 4.90374923e-01 5.36686242e-01 -4.80997086e-01 -1.15048635e+00 -1.00483704e+00 5.51232755e-01 -5.17930388e-01 2.06548616e-01 -1.90188363e-01 -5.19212902e-01 9.30453777e-01 6.43177569e-01 -1.12337686e-01 7.03488886e-01 -4.79172081e-01 -1.95329115e-01 -1.81068242e-01 -1.40239656e+00 6.98245704e-01 9.46159422e-01 -2.77282774e-01 -3.47028226e-02 6.92522228e-02 6.73210993e-02 -6.06405139e-01 -6.00510180e-01 3.97547096e-01 5.97509801e-01 -1.35657609e+00 9.84726369e-01 -3.93796057e-01 6.45761132e-01 -3.02257866e-01 1.98124513e-01 -1.44203317e+00 -3.02118361e-01 -9.51983333e-01 9.83530134e-02 1.36361742e+00 -5.55342361e-02 -7.92921543e-01 9.14341688e-01 5.56623340e-01 -1.95193842e-01 -1.11968219e+00 -9.37218606e-01 -5.93130231e-01 3.91216695e-01 -2.95641541e-01 4.95266050e-01 6.06610715e-01 -6.30712032e-01 1.88156515e-01 -7.01264858e-01 -4.56409663e-01 6.59196734e-01 -1.18382849e-01 1.04701316e+00 -6.94691718e-01 -4.05454300e-02 -6.42632484e-01 -6.98732615e-01 -8.69437993e-01 1.29176661e-01 -7.22049356e-01 -1.76316306e-01 -1.75374341e+00 1.27422705e-01 -2.42241845e-01 -4.91916984e-02 4.39254820e-01 -3.13038528e-01 9.29488003e-01 3.81603539e-01 2.12818936e-01 7.27889836e-02 4.15948033e-01 1.22661138e+00 -4.55847271e-02 1.39609039e-01 1.51333988e-01 -1.05193484e+00 5.08251727e-01 1.04771709e+00 -7.66972080e-02 -6.77142084e-01 -5.50135970e-01 6.64402265e-03 -3.09899062e-01 3.85692418e-01 -1.00949550e+00 -1.33521408e-01 -1.30817175e-01 3.97854477e-01 -1.44882232e-01 6.60257697e-01 -7.36328483e-01 4.47737306e-01 2.83595532e-01 -1.01204716e-01 2.10785002e-01 3.67292106e-01 2.44475648e-01 -3.78241211e-01 -2.34994292e-01 1.18971038e+00 -2.36608222e-01 -8.05950522e-01 4.59898144e-01 -3.96416247e-01 4.02651392e-02 1.32646012e+00 -3.73048604e-01 -1.80909604e-01 -8.86580646e-01 -5.19730508e-01 -3.06068093e-01 6.29266620e-01 5.29862583e-01 6.09108627e-01 -1.23734152e+00 -1.04535365e+00 2.40841776e-01 1.13054372e-01 -1.05320461e-01 4.66409743e-01 5.91179609e-01 -5.39375484e-01 3.91376793e-01 -5.78884184e-01 -4.16835368e-01 -1.44065571e+00 3.62965763e-01 4.10747528e-01 3.11605111e-02 -5.16660690e-01 1.06830549e+00 4.77323741e-01 2.92032026e-02 3.21272761e-02 3.52195323e-01 -7.77639523e-02 -2.20842764e-01 4.89953250e-01 3.96405458e-01 2.82920867e-01 -7.48015225e-01 -2.64119089e-01 7.41139829e-01 7.11582508e-03 -1.19058765e-01 1.25195742e+00 -2.13449553e-01 -2.35176250e-01 -1.34562433e-04 8.02878320e-01 -6.34751320e-02 -1.39904642e+00 2.08958626e-01 -7.32209682e-01 -4.11368161e-01 -1.50490463e-01 -9.30494964e-01 -1.62992096e+00 1.00242460e+00 5.71458519e-01 -3.99641395e-02 1.76538026e+00 -1.03315888e-02 7.70911455e-01 -4.10746187e-01 3.61764461e-01 -6.82647169e-01 6.17537722e-02 2.80599117e-01 1.24518895e+00 -9.91757095e-01 5.73788770e-02 -5.30319631e-01 -7.17012048e-01 1.00018334e+00 6.23266101e-01 -1.51775435e-01 5.03150344e-01 2.99198538e-01 5.39766073e-01 8.43982324e-02 -3.33043814e-01 -5.77035323e-02 2.78326511e-01 1.39308989e+00 7.86405981e-01 -5.47829308e-02 -3.33701819e-01 -7.55944625e-02 -5.98339319e-01 -6.88067600e-02 3.89752358e-01 7.70117879e-01 1.48950115e-01 -1.54183197e+00 -3.27059060e-01 4.48329031e-01 -6.78956807e-01 -1.71001390e-01 -4.86741990e-01 6.69241250e-01 1.48056656e-01 1.11094666e+00 -1.80534776e-02 -1.31634161e-01 1.35683388e-01 -2.02087581e-01 9.25489366e-01 -5.06046057e-01 -6.11592889e-01 -6.42341748e-02 -1.45121366e-02 -3.48548323e-01 -7.74431586e-01 -5.81015050e-01 -6.20586991e-01 -4.53847826e-01 3.25560309e-02 -1.78607047e-01 8.53859246e-01 6.85253143e-01 4.04683292e-01 5.95593452e-01 3.17414701e-01 -9.09671009e-01 -2.55058795e-01 -8.40449095e-01 -4.13817346e-01 3.10037255e-01 2.05660313e-01 -3.99530321e-01 -2.25725472e-01 3.18598688e-01]
[11.81147289276123, -0.3007112741470337]
ed6dea4f-2423-41f2-862e-58caa7418708
hierarchical-attention-learning-of-scene-flow
2010.05762
null
https://arxiv.org/abs/2010.05762v1
https://arxiv.org/pdf/2010.05762v1.pdf
Hierarchical Attention Learning of Scene Flow in 3D Point Clouds
Scene flow represents the 3D motion of every point in the dynamic environments. Like the optical flow that represents the motion of pixels in 2D images, 3D motion representation of scene flow benefits many applications, such as autonomous driving and service robot. This paper studies the problem of scene flow estimation from two consecutive 3D point clouds. In this paper, a novel hierarchical neural network with double attention is proposed for learning the correlation of point features in adjacent frames and refining scene flow from coarse to fine layer by layer. The proposed network has a new more-for-less hierarchical architecture. The more-for-less means that the number of input points is greater than the number of output points for scene flow estimation, which brings more input information and balances the precision and resource consumption. In this hierarchical architecture, scene flow of different levels is generated and supervised respectively. A novel attentive embedding module is introduced to aggregate the features of adjacent points using a double attention method in a patch-to-patch manner. The proper layers for flow embedding and flow supervision are carefully considered in our network designment. Experiments show that the proposed network outperforms the state-of-the-art performance of 3D scene flow estimation on the FlyingThings3D and KITTI Scene Flow 2015 datasets. We also apply the proposed network to realistic LiDAR odometry task, which is an key problem in autonomous driving. The experiment results demonstrate that our proposed network can outperform the ICP-based method and shows the good practical application ability.
['Hesheng Wang', 'Zhe Liu', 'Xinrui Wu', 'Guangming Wang']
2020-10-12
null
null
null
null
['scene-flow-estimation']
['computer-vision']
[-3.03391218e-01 -4.37499076e-01 -9.63205546e-02 -3.82073134e-01 1.75629705e-01 -2.87188794e-02 3.31883460e-01 -1.64779916e-01 -6.91088617e-01 5.35670877e-01 2.05803975e-01 -1.80487722e-01 -9.24366191e-02 -9.92189109e-01 -5.85632503e-01 -7.20507085e-01 6.91710552e-03 2.37999722e-01 5.18129766e-01 -3.66268724e-01 3.60057712e-01 7.84253180e-01 -1.70550740e+00 -1.35993987e-01 7.51773834e-01 8.95605624e-01 4.30465400e-01 6.22292697e-01 -5.38682044e-01 1.05482757e+00 -4.35922474e-01 5.96879497e-02 5.24160445e-01 -1.71909750e-01 -5.83278656e-01 1.55856973e-02 8.26347291e-01 -5.52945077e-01 -7.90696204e-01 9.09991026e-01 4.38411087e-01 4.10679102e-01 5.39011776e-01 -1.53402841e+00 -4.87531155e-01 5.07380394e-03 -8.21689665e-01 6.65489912e-01 7.39739537e-02 3.07971716e-01 6.93564713e-01 -9.03744400e-01 4.43641871e-01 1.59347129e+00 3.23654920e-01 3.39612782e-01 -7.80536234e-01 -8.11333060e-01 3.40376586e-01 5.97345591e-01 -1.26307905e+00 -2.72497274e-02 8.80372941e-01 -7.50019073e-01 1.00661635e+00 -2.82195657e-01 9.48746443e-01 4.24621493e-01 3.45528096e-01 7.44226098e-01 5.53628325e-01 6.84083477e-02 -1.05888091e-01 -2.11934261e-02 2.02754438e-01 9.25972521e-01 4.33122814e-01 2.62348413e-01 -3.70996624e-01 4.32708144e-01 9.63525414e-01 3.63569111e-01 -2.29939193e-01 -4.68700290e-01 -1.31094766e+00 9.72680390e-01 1.06720984e+00 1.31762072e-01 -2.10277528e-01 3.67386758e-01 4.22411084e-01 3.59489881e-02 3.14218849e-01 9.31425765e-02 -2.48666316e-01 -7.25970268e-02 -7.32199669e-01 2.05946058e-01 4.96749163e-01 1.09059680e+00 1.28665924e+00 4.11924124e-01 4.41130884e-02 4.62623537e-01 3.75926226e-01 5.30911982e-01 4.45556641e-01 -1.17412567e+00 6.31406724e-01 9.24493253e-01 -7.09750578e-02 -1.54282105e+00 -4.54554170e-01 -2.07549587e-01 -1.04551661e+00 3.93846840e-01 1.48159578e-01 -4.00030147e-03 -7.29849696e-01 1.51229358e+00 6.01757228e-01 6.74813986e-01 7.27124792e-03 1.19473696e+00 1.04969597e+00 1.10401285e+00 -1.00096703e-01 -1.06169702e-02 1.09191823e+00 -1.09720457e+00 -8.53757918e-01 -2.97836721e-01 5.81698060e-01 -4.87032443e-01 7.59802282e-01 -1.53545782e-01 -8.44280303e-01 -1.22703087e+00 -1.18300033e+00 -5.33504844e-01 -4.78704751e-01 -6.52598292e-02 6.83169961e-01 2.78121710e-01 -8.22072804e-01 2.89036036e-01 -7.13760793e-01 -2.48668045e-01 5.37191868e-01 2.46971592e-01 -5.03857136e-01 -1.69173211e-01 -1.13435757e+00 8.80671620e-01 5.43534398e-01 5.35714328e-01 -7.77373672e-01 -8.94937277e-01 -1.28075707e+00 7.08190873e-02 2.17041522e-02 -9.59932745e-01 7.49136865e-01 -1.87164277e-01 -1.37843359e+00 5.37998855e-01 -3.58666927e-01 -3.36846441e-01 5.11403561e-01 -3.43034387e-01 -9.83301923e-02 2.13599041e-01 2.42665678e-01 1.02809155e+00 5.69704592e-01 -9.90010023e-01 -1.09145951e+00 -2.86046624e-01 2.07736596e-01 4.42890495e-01 -1.15473025e-01 -4.98216391e-01 -3.37935448e-01 -1.06390946e-01 1.38616934e-01 -8.04815590e-01 -3.82481545e-01 2.71256089e-01 -1.00473706e-02 -4.16043729e-01 1.37692118e+00 -1.39075950e-01 1.01556027e+00 -2.11311793e+00 1.09533690e-01 -2.88811892e-01 3.66799861e-01 1.79742277e-01 8.23506862e-02 1.54618584e-02 1.21461779e-01 -1.37017608e-01 -4.44690526e-01 -3.08272749e-01 -2.13739052e-01 3.45966607e-01 -1.87013313e-01 6.08915210e-01 4.87157464e-01 8.22754085e-01 -1.06819391e+00 -7.55290031e-01 9.22354281e-01 6.90616786e-01 -8.12037885e-01 2.22423017e-01 2.12848261e-01 5.91693759e-01 -5.16862154e-01 2.52336830e-01 9.32986200e-01 -1.40382424e-01 -6.80124819e-01 -4.12200540e-01 -3.53632838e-01 1.07092880e-01 -1.38755298e+00 1.95396459e+00 -5.82291603e-01 1.00833595e+00 -2.29649276e-01 -8.98727417e-01 1.18595719e+00 -1.63099989e-01 5.95295668e-01 -6.52215540e-01 2.58371890e-01 3.55891101e-02 -6.13917224e-02 -7.37925053e-01 7.25981772e-01 6.80047497e-02 -1.11022100e-01 5.02551496e-02 7.29855448e-02 -5.57753503e-01 3.67194116e-01 2.19734803e-01 7.88579226e-01 1.88607454e-01 7.06250519e-02 -3.50613236e-01 1.09689415e+00 2.21045658e-01 8.64933252e-01 3.47032547e-01 -6.49017334e-01 3.17787081e-01 1.52853206e-01 -6.95918560e-01 -9.37099218e-01 -7.30403066e-01 -1.93189234e-01 3.57780010e-01 8.97866964e-01 -1.55864745e-01 -3.29213321e-01 -5.81314504e-01 3.37646633e-01 3.28596056e-01 -5.75307846e-01 -2.31511399e-01 -8.83116663e-01 -4.50762719e-01 1.39821783e-01 5.78765631e-01 1.05376852e+00 -1.12680197e+00 -8.87284040e-01 1.65515035e-01 -2.63742238e-01 -1.54403651e+00 -5.17984986e-01 -1.92780778e-01 -9.03129637e-01 -1.00503445e+00 -4.04656559e-01 -9.98368621e-01 5.60483336e-01 7.50760138e-01 8.01684022e-01 -9.15619079e-03 -3.63598764e-01 9.90861729e-02 -1.60913572e-01 -2.24754289e-01 7.68201873e-02 3.09309959e-02 6.27348274e-02 4.86547463e-02 6.08199716e-01 -5.61602473e-01 -9.10551548e-01 3.09896171e-01 -7.81722188e-01 1.44143421e-02 4.11326855e-01 7.68605530e-01 3.55212092e-01 1.74942881e-01 1.33614272e-01 -4.78351504e-01 4.58896965e-01 -2.86637366e-01 -6.91702127e-01 -3.89518738e-01 -1.94556326e-01 8.30664337e-02 6.05456173e-01 -6.42211884e-02 -1.03782272e+00 2.31343150e-01 -5.66100180e-02 -7.73200512e-01 -2.65055209e-01 8.46496969e-02 -1.79172643e-02 -2.11303622e-01 4.22599435e-01 7.64165744e-02 5.85853588e-03 -8.12232196e-02 5.66435039e-01 4.24720854e-01 6.37285173e-01 -1.66025862e-01 9.55215335e-01 7.43611634e-01 3.66981268e-01 -9.09830272e-01 -6.39007390e-01 -7.82208979e-01 -9.53920484e-01 -4.51699883e-01 1.20724928e+00 -1.04982841e+00 -1.16650462e+00 4.75061536e-01 -1.44616318e+00 -1.16408616e-01 -2.99364358e-01 8.15195858e-01 -6.45358264e-01 4.09112036e-01 -6.92183554e-01 -5.80239296e-01 -1.39908463e-01 -1.49245083e+00 1.01799488e+00 5.44711232e-01 2.38608733e-01 -1.00446582e+00 9.29118916e-02 2.90549427e-01 1.27608746e-01 3.46590400e-01 4.84742314e-01 2.92796046e-01 -9.16399240e-01 4.38261740e-02 -5.72224379e-01 2.88001716e-01 2.58872896e-01 1.14867054e-01 -8.99964809e-01 -4.49283868e-02 1.76155880e-01 -7.93609768e-02 1.00318193e+00 5.01999795e-01 9.84103143e-01 1.02790728e-01 -1.56324923e-01 1.04301131e+00 1.56856310e+00 2.87594527e-01 6.39679849e-01 4.45748955e-01 1.25601721e+00 9.04881775e-01 7.05296874e-01 1.50558949e-01 6.22504234e-01 3.87186408e-01 7.19370246e-01 -2.33391434e-01 -1.84480041e-01 -3.54546010e-01 2.53776908e-01 9.76439536e-01 9.05099660e-02 -8.38566199e-02 -7.48697340e-01 6.93257928e-01 -1.96486533e+00 -1.08978570e+00 -3.99508923e-01 1.82781339e+00 2.88032591e-01 2.08758831e-01 -1.57510251e-01 2.47004837e-01 7.49740124e-01 4.11911458e-01 -5.73018372e-01 -4.35902774e-01 4.64777015e-02 -9.92869288e-02 5.69582403e-01 8.16251159e-01 -1.03817117e+00 1.14610744e+00 4.95209980e+00 4.89944696e-01 -1.20167530e+00 2.36510765e-03 3.00600082e-01 3.46322395e-02 7.75482580e-02 -7.89345708e-03 -1.16522443e+00 5.92724204e-01 4.48572576e-01 2.49254964e-02 -7.21482486e-02 7.08387196e-01 4.69090551e-01 -9.58514437e-02 -8.65996897e-01 1.28530252e+00 -8.10710788e-02 -1.51440060e+00 2.55734742e-01 6.44488037e-02 6.24173284e-01 1.41091391e-01 -2.28761435e-01 1.51580304e-01 2.34318614e-01 -6.45238936e-01 6.23878539e-01 3.85038942e-01 4.72130686e-01 -7.50740886e-01 7.96561956e-01 4.10374254e-01 -1.58315587e+00 -2.06699148e-01 -5.95750034e-01 -2.98950613e-01 7.56150186e-01 5.96045971e-01 -5.60322464e-01 6.60263240e-01 9.53957558e-01 1.45970678e+00 -4.33573425e-01 1.23991561e+00 -2.16595978e-01 7.24801794e-02 -2.46069297e-01 -9.88545343e-02 6.48148000e-01 -4.84604895e-01 5.27804434e-01 1.25412679e+00 3.04971457e-01 1.01804040e-01 2.69952804e-01 9.49616492e-01 1.86178878e-01 -9.13557857e-02 -9.27484453e-01 5.09218752e-01 3.58493894e-01 1.30617213e+00 -5.12214184e-01 -3.89021546e-01 -4.06784505e-01 6.02765143e-01 2.52202868e-01 2.64379472e-01 -6.97116196e-01 -7.45730042e-01 9.03660655e-01 5.03637567e-02 2.55685300e-01 -7.04417169e-01 -9.53681618e-02 -1.04437780e+00 -1.30847827e-01 2.00912859e-02 2.46904030e-01 -8.80792499e-01 -1.13197720e+00 5.71739256e-01 5.80540635e-02 -1.52112973e+00 -1.48014929e-02 -7.55313993e-01 -6.97573364e-01 8.96424174e-01 -2.23158145e+00 -6.64496899e-01 -1.10010326e+00 7.97911465e-01 6.42910302e-01 1.38187528e-01 1.27144173e-01 6.00346029e-01 -7.68882334e-01 2.30616648e-02 -4.42654848e-01 3.06714833e-01 5.12641370e-01 -1.04542184e+00 4.87050891e-01 1.02340198e+00 -9.11589712e-02 2.48564467e-01 3.15198034e-01 -4.24351484e-01 -1.05927610e+00 -1.18075716e+00 8.24377179e-01 -3.42127055e-01 4.54410791e-01 -2.98308343e-01 -8.31745505e-01 3.96720529e-01 1.35960206e-01 2.68519282e-01 1.47101805e-01 -4.85998988e-01 -9.72442422e-03 -5.83635747e-01 -1.00882936e+00 3.72802824e-01 1.25744164e+00 -4.11084503e-01 -5.22921085e-01 5.32603413e-02 1.12123036e+00 -4.43095744e-01 -7.50102997e-01 6.47671640e-01 2.98963577e-01 -1.07308674e+00 1.05995131e+00 -2.85334826e-01 5.59038401e-01 -9.11388397e-01 -3.18620130e-02 -1.16170490e+00 -5.32627106e-01 -4.02519464e-01 -9.91189331e-02 8.56660247e-01 2.50420347e-03 -6.40385985e-01 9.57949400e-01 2.19436720e-01 -4.61549580e-01 -4.91243690e-01 -8.32676113e-01 -5.34578562e-01 3.10162958e-02 -2.42541611e-01 6.53370917e-01 9.03043628e-01 -4.69748199e-01 6.53503239e-01 -2.18573213e-01 2.17490599e-01 6.18950188e-01 -6.12988696e-03 1.13685095e+00 -1.30610752e+00 2.89310694e-01 -5.72781980e-01 -1.09287989e+00 -1.63714302e+00 2.88763344e-01 -8.36224079e-01 9.82994288e-02 -1.64230835e+00 -1.90784827e-01 -4.36862260e-01 -9.34823602e-02 4.07057218e-02 -3.21750969e-01 1.16870552e-01 4.22965080e-01 2.06448078e-01 -2.86850542e-01 9.64948237e-01 1.81448948e+00 -3.84075254e-01 -3.99961799e-01 -2.38706276e-01 -2.31716812e-01 4.90689725e-01 6.67514563e-01 -2.14866385e-01 -6.80856347e-01 -7.63995051e-01 -1.14578046e-01 -4.16918099e-02 4.18246537e-01 -1.27455068e+00 6.48252666e-01 -7.62601346e-02 3.22101206e-01 -1.02834940e+00 4.89479065e-01 -1.06634593e+00 -2.73917198e-01 6.04800165e-01 3.89859229e-02 3.75684470e-01 3.21289927e-01 6.90483928e-01 -5.72056234e-01 -1.30088837e-03 7.67157257e-01 -2.57844388e-01 -1.26034176e+00 9.20677662e-01 -3.14545408e-02 8.68319049e-02 1.09012938e+00 -5.63540876e-01 -4.67455745e-01 -1.87553734e-01 -2.00052649e-01 5.90130687e-01 7.12437034e-02 7.38923907e-01 8.41551602e-01 -1.42537642e+00 -6.65822446e-01 4.32011038e-01 1.51496336e-01 7.89059997e-01 4.08826798e-01 7.31284261e-01 -1.06189072e+00 3.99634063e-01 -5.40367663e-01 -1.16738784e+00 -7.81251907e-01 4.20048296e-01 4.55283552e-01 7.02472329e-02 -7.74086297e-01 8.08897316e-01 6.19610488e-01 -5.16173184e-01 5.51968329e-02 -4.98957604e-01 -5.18363059e-01 -2.75504384e-02 3.57928395e-01 6.09161198e-01 -2.31540278e-01 -1.00625169e+00 -3.13369751e-01 1.14982247e+00 1.22124024e-01 1.86245754e-01 1.08353829e+00 -3.18368822e-01 -1.55585870e-01 5.11031926e-01 1.51577175e+00 -4.64463204e-01 -1.67649090e+00 -1.59104869e-01 -4.71841216e-01 -8.39382052e-01 3.23451370e-01 1.30185008e-01 -1.43780565e+00 1.48006177e+00 6.23420238e-01 4.58497182e-02 9.16471064e-01 -4.12374109e-01 1.05630243e+00 2.31402010e-01 3.25686723e-01 -6.70057356e-01 1.09387085e-01 6.87557161e-01 5.44362366e-01 -1.47925293e+00 5.20364055e-03 -4.75481749e-01 -4.39507931e-01 1.13208759e+00 1.06371319e+00 -4.54653263e-01 7.75012374e-01 3.99137707e-03 -1.00876042e-03 -2.25051373e-01 -5.38583755e-01 -4.06048298e-01 9.98865589e-02 5.86320877e-01 -9.48430691e-03 -2.59799510e-01 3.12144607e-02 -2.94791818e-01 -3.45106304e-01 -4.76669334e-02 5.47117054e-01 7.68900275e-01 -6.17261648e-01 -5.87107658e-01 -4.61093746e-02 1.74716979e-01 1.94721416e-01 1.71200424e-01 9.13168490e-02 9.71706569e-01 4.32645887e-01 9.84349549e-01 6.06793165e-01 -3.87324899e-01 6.11677587e-01 -4.08602327e-01 3.89722258e-01 -3.39944631e-01 -2.32190415e-01 -4.11198378e-01 -3.97777736e-01 -7.21317232e-01 -8.75537694e-01 -4.99161363e-01 -1.41047823e+00 -5.25103152e-01 -2.23779082e-01 6.72232360e-02 6.17963910e-01 6.13524497e-01 4.00089949e-01 6.83912694e-01 8.84748220e-01 -1.08865356e+00 1.61601558e-01 -7.39854515e-01 -3.32673371e-01 4.10610735e-01 7.84592986e-01 -9.76686776e-01 -6.03052020e-01 8.16333517e-02]
[8.573638916015625, -2.0935323238372803]
bc50d0ad-50b9-46d6-9381-f2e50d8bf9da
towards-open-world-object-detection
2103.02603
null
https://arxiv.org/abs/2103.02603v2
https://arxiv.org/pdf/2103.02603v2.pdf
Towards Open World Object Detection
Humans have a natural instinct to identify unknown object instances in their environments. The intrinsic curiosity about these unknown instances aids in learning about them, when the corresponding knowledge is eventually available. This motivates us to propose a novel computer vision problem called: `Open World Object Detection', where a model is tasked to: 1) identify objects that have not been introduced to it as `unknown', without explicit supervision to do so, and 2) incrementally learn these identified unknown categories without forgetting previously learned classes, when the corresponding labels are progressively received. We formulate the problem, introduce a strong evaluation protocol and provide a novel solution, which we call ORE: Open World Object Detector, based on contrastive clustering and energy based unknown identification. Our experimental evaluation and ablation studies analyze the efficacy of ORE in achieving Open World objectives. As an interesting by-product, we find that identifying and characterizing unknown instances helps to reduce confusion in an incremental object detection setting, where we achieve state-of-the-art performance, with no extra methodological effort. We hope that our work will attract further research into this newly identified, yet crucial research direction.
['Vineeth N Balasubramanian', 'Fahad Shahbaz Khan', 'Salman Khan', 'K J Joseph']
2021-03-03
null
http://openaccess.thecvf.com//content/CVPR2021/html/Joseph_Towards_Open_World_Object_Detection_CVPR_2021_paper.html
http://openaccess.thecvf.com//content/CVPR2021/papers/Joseph_Towards_Open_World_Object_Detection_CVPR_2021_paper.pdf
cvpr-2021-1
['open-world-object-detection']
['computer-vision']
[ 3.11716437e-01 3.37683588e-01 -1.49294995e-02 -2.54313648e-01 -7.81432331e-01 -8.56779099e-01 6.03874564e-01 3.29732150e-01 -6.48271561e-01 6.08043313e-01 -2.41080478e-01 1.38864502e-01 -1.73952132e-01 -4.39918995e-01 -9.36818898e-01 -7.41162002e-01 -3.14767212e-01 7.49839425e-01 3.67860019e-01 2.34731779e-01 2.00428218e-01 2.51584530e-01 -1.88893402e+00 9.80782136e-02 8.84845138e-01 8.49009871e-01 4.34309661e-01 6.57006562e-01 1.87474862e-01 8.74816895e-01 -1.48283184e-01 -5.48628926e-01 2.88291365e-01 -2.22226247e-01 -1.10228562e+00 2.63945192e-01 6.39745474e-01 -1.93995923e-01 6.10638633e-02 1.16390872e+00 3.44608426e-01 2.29473814e-01 7.44593561e-01 -1.20107424e+00 -9.36895669e-01 5.98072588e-01 -5.44065177e-01 3.71460259e-01 -1.22688785e-01 3.97168100e-01 1.28044140e+00 -7.95369267e-01 7.43499637e-01 1.02766657e+00 7.96493351e-01 7.06290722e-01 -1.49277508e+00 -3.82826895e-01 6.49078250e-01 5.65903842e-01 -1.61131275e+00 -4.54685509e-01 7.01741695e-01 -4.27689314e-01 5.76472461e-01 3.03247929e-01 2.96554118e-01 1.08447564e+00 -2.23272786e-01 1.10976374e+00 1.13497674e+00 -4.08320427e-01 4.62572068e-01 7.57592201e-01 6.34448647e-01 8.41621220e-01 4.93361056e-01 3.01775247e-01 -4.08900976e-01 6.26810566e-02 1.48364320e-01 6.56422302e-02 -2.65470445e-01 -6.77002192e-01 -1.12132430e+00 6.73495293e-01 8.11052561e-01 2.55078495e-01 -2.84754544e-01 3.02509051e-02 5.26279435e-02 3.61431748e-01 5.35152018e-01 7.63444543e-01 -6.18354499e-01 8.63808170e-02 -5.25081575e-01 -4.34345566e-02 7.24597692e-01 8.27686250e-01 1.00628161e+00 -5.51754236e-01 4.00607893e-03 7.81828284e-01 8.91292989e-02 3.59796047e-01 5.13616562e-01 -9.54730630e-01 -9.74410959e-03 6.42021418e-01 1.88167974e-01 -7.45464981e-01 -4.35806066e-01 -7.22205341e-01 -4.61240083e-01 2.09848180e-01 4.28596795e-01 -7.51249120e-02 -9.39593315e-01 1.94353759e+00 5.07939398e-01 4.54376847e-01 2.12563679e-01 8.67971241e-01 4.81704980e-01 3.26884985e-01 1.34799302e-01 -1.88606903e-01 1.34401727e+00 -1.01950276e+00 -3.25732678e-01 -5.14440060e-01 4.82079178e-01 -3.51541638e-01 9.29156661e-01 4.32693928e-01 -6.89228356e-01 -6.06175244e-01 -9.24343467e-01 5.62914610e-02 -4.73483652e-01 4.85471822e-02 7.82189608e-01 5.27296722e-01 -9.46755230e-01 6.19024694e-01 -8.77479494e-01 -5.67657769e-01 8.78859639e-01 2.72144288e-01 -2.33121425e-01 -4.99138646e-02 -6.64468646e-01 8.84920776e-01 5.89217424e-01 -1.12203717e-01 -1.23011744e+00 -6.94936097e-01 -5.37575841e-01 9.75181833e-02 9.66711223e-01 -8.38592649e-01 1.40772498e+00 -1.22718537e+00 -7.02701092e-01 1.16927302e+00 -1.59794539e-01 -6.74406409e-01 5.79990745e-01 -2.70748913e-01 -2.34639063e-01 5.05009443e-02 3.55101347e-01 8.53729367e-01 1.08513105e+00 -1.74578404e+00 -1.07419503e+00 -7.75780141e-01 2.12794065e-01 3.09836090e-01 -5.24260402e-01 -2.88175106e-01 -4.39716786e-01 -4.07663733e-01 3.74149054e-01 -1.06061304e+00 -3.30083072e-01 2.60761708e-01 -4.24163550e-01 -4.01299506e-01 6.37184381e-01 -1.39620885e-01 8.07421088e-01 -2.10482907e+00 1.11753404e-01 8.87884051e-02 7.56019950e-01 1.86936855e-01 -1.24469645e-01 -2.02291042e-01 1.17596455e-01 -3.08082271e-02 -5.12004554e-01 -2.38528699e-01 4.03389186e-02 1.72939431e-02 -3.69668365e-01 4.09366161e-01 2.92482555e-01 9.78101552e-01 -1.09754395e+00 -4.09497768e-01 1.88863397e-01 2.41463706e-01 -4.64102119e-01 1.65398896e-01 -3.93907726e-01 2.71171778e-01 -4.32020456e-01 5.93887329e-01 5.58413923e-01 -6.35179698e-01 -1.92196816e-02 -5.65456413e-02 2.42317662e-01 -3.35903972e-01 -1.35816371e+00 1.31306958e+00 -2.62517840e-01 6.05496883e-01 4.70338985e-02 -1.18849385e+00 4.16827440e-01 -1.67902231e-01 1.48839220e-01 -3.14507484e-01 1.14530392e-01 6.89238608e-02 2.24665622e-03 -6.27393365e-01 2.43147105e-01 -1.56083122e-01 2.38266751e-01 4.20061141e-01 2.66558766e-01 3.08103442e-01 1.11473761e-01 5.34477293e-01 1.24228728e+00 -1.81700811e-01 1.79526761e-01 -1.87149182e-01 2.14903563e-01 6.50215298e-02 2.77148426e-01 1.43998218e+00 -5.61141789e-01 5.27964830e-01 1.16239347e-01 -5.57903528e-01 -6.59398317e-01 -1.43870378e+00 -2.83086032e-01 1.23751581e+00 4.22004282e-01 3.50408852e-02 -7.15214074e-01 -1.10937679e+00 6.47877529e-02 5.46282589e-01 -8.29013169e-01 -3.51726800e-01 -1.36927605e-01 -6.66477323e-01 5.52066006e-02 4.98704344e-01 3.94999325e-01 -1.12243533e+00 -7.82738268e-01 1.38433307e-01 -2.81937629e-01 -1.06493878e+00 -2.53248930e-01 5.68151653e-01 -7.95603812e-01 -1.43728888e+00 -6.32178485e-01 -9.46842134e-01 1.00032628e+00 4.11643356e-01 1.20686018e+00 1.94434240e-01 -5.66128194e-01 8.59281480e-01 -3.82168978e-01 -5.79859376e-01 -1.43366873e-01 -1.30124018e-01 3.53164881e-01 5.10710716e-01 5.64997315e-01 -4.02259350e-01 -5.81528425e-01 2.15582684e-01 -6.78350329e-01 -1.52076349e-01 8.57472956e-01 8.60900104e-01 5.49408913e-01 2.81659186e-01 5.60088694e-01 -1.20541394e+00 3.21785092e-01 -5.41255951e-01 -5.83217263e-01 4.22903210e-01 -6.20244861e-01 2.42698833e-01 5.45136988e-01 -7.30858564e-01 -1.09975767e+00 2.73641258e-01 2.38543808e-01 -3.72380465e-01 -4.33808982e-01 2.28328496e-01 8.46751500e-03 -1.34041816e-01 1.02845192e+00 3.24801147e-01 -3.44622344e-01 -4.69244897e-01 6.55888140e-01 7.01111674e-01 8.17825377e-01 -5.82897961e-01 1.01069987e+00 8.22120428e-01 -4.69004154e-01 -7.24519908e-01 -1.35060501e+00 -8.22984815e-01 -7.85845041e-01 -1.50240257e-01 6.55462682e-01 -9.86151457e-01 -7.62916505e-01 2.86645234e-01 -1.03250432e+00 -1.92666888e-01 -6.01479352e-01 3.20576012e-01 -4.85466033e-01 3.63821477e-01 -1.88689992e-01 -1.15027070e+00 1.39703192e-02 -8.25243175e-01 9.59928870e-01 3.67131650e-01 -1.75388366e-01 -8.44710231e-01 -1.46659344e-01 5.50942540e-01 1.11981302e-01 -9.47885364e-02 6.93203032e-01 -9.03911948e-01 -9.86119509e-01 -9.80396643e-02 -4.21288133e-01 2.71442562e-01 3.18071730e-02 -5.08211136e-01 -1.42780447e+00 -4.95229512e-01 9.80789363e-02 -4.99792814e-01 1.39388883e+00 1.94382563e-01 1.28382766e+00 -2.81232834e-01 -7.50240326e-01 4.43710059e-01 1.35767078e+00 -1.13607325e-01 3.68191116e-02 2.30968624e-01 6.06932282e-01 6.69635177e-01 4.58612382e-01 4.53680605e-01 3.20842683e-01 4.78635520e-01 4.18723047e-01 -4.50049043e-02 -2.55248338e-01 -1.08169138e-01 1.18511014e-01 2.02464759e-01 -3.64786722e-02 -2.26557702e-02 -8.70039165e-01 9.94357407e-01 -1.84340334e+00 -9.17102993e-01 1.42967358e-01 2.27657843e+00 8.89403045e-01 3.04181695e-01 -1.04913734e-01 -5.61537854e-02 8.48124564e-01 -3.42438877e-01 -9.64277565e-01 2.48686105e-01 7.39154667e-02 9.34590921e-02 3.55052501e-01 3.29921871e-01 -1.51719701e+00 9.77238953e-01 5.83622551e+00 7.95030296e-01 -7.97100782e-01 2.28030980e-01 8.55313420e-01 -3.69291455e-02 -2.97254347e-03 -1.53523367e-02 -9.32806849e-01 3.92830193e-01 7.24527776e-01 -1.75290525e-01 5.76960802e-01 1.10937917e+00 -3.14678341e-01 -3.01170975e-01 -1.66554046e+00 1.00390518e+00 3.64663243e-01 -1.12195039e+00 5.79985697e-03 -1.76843286e-01 7.10739017e-01 1.71963766e-01 1.93337440e-01 4.34080511e-01 6.09633267e-01 -5.75935006e-01 5.47170699e-01 3.64190280e-01 3.77056211e-01 -4.69453156e-01 4.22990650e-01 5.22796810e-01 -9.49500442e-01 -6.07440472e-01 -4.20492321e-01 -9.39318724e-03 -2.18295917e-01 5.09063959e-01 -9.16653097e-01 1.73735455e-01 9.13205981e-01 6.08969212e-01 -9.35705364e-01 1.39498222e+00 -3.09711158e-01 7.16406047e-01 -4.21763420e-01 -4.62773517e-02 1.42256230e-01 1.53656602e-01 7.94728935e-01 7.68205345e-01 -1.41039208e-01 2.85648495e-01 2.04819903e-01 1.15473068e+00 -2.67136514e-01 -1.46144718e-01 -6.08136177e-01 1.39189631e-01 3.10926020e-01 1.29339778e+00 -1.14116240e+00 -4.93565917e-01 -2.64065892e-01 1.17083752e+00 6.82906151e-01 2.76204437e-01 -6.00162268e-01 -3.28896344e-01 4.25972998e-01 -2.09203765e-01 4.59638298e-01 3.21418911e-01 -1.61705554e-01 -1.28732407e+00 1.83346197e-01 -5.70409060e-01 7.61164367e-01 -5.98770738e-01 -1.83847749e+00 4.20504034e-01 -2.05032185e-01 -1.00410390e+00 1.05748124e-01 -6.54481530e-01 -1.98060617e-01 2.34426185e-01 -1.66779447e+00 -1.04538751e+00 -2.40117669e-01 4.87031668e-01 7.52710462e-01 3.07134204e-02 6.04148626e-01 1.86138183e-01 -4.03818667e-01 6.03492260e-01 3.37151796e-01 8.52331296e-02 5.75336814e-01 -1.46355438e+00 1.78291395e-01 9.24607754e-01 6.05607271e-01 5.16102374e-01 6.95247650e-01 -5.37875712e-01 -1.26595366e+00 -1.32945263e+00 5.82928658e-01 -1.25735748e+00 5.68950474e-01 -5.31120598e-01 -1.07326853e+00 8.78650308e-01 -3.31837147e-01 3.82196605e-01 4.01983261e-01 5.47343254e-01 -4.64800686e-01 -8.66949037e-02 -1.22979999e+00 5.13319075e-01 1.24929643e+00 -2.97994196e-01 -9.70183134e-01 5.44021726e-01 8.80282402e-01 -7.60438368e-02 -2.34954268e-01 2.44404897e-01 2.84099370e-01 -7.19082832e-01 1.06512558e+00 -8.14723372e-01 1.35313958e-01 -3.43443781e-01 -2.84716804e-02 -1.31266570e+00 -3.65187347e-01 -4.14959103e-01 -3.51993471e-01 1.12954545e+00 5.24998963e-01 -6.79871798e-01 8.36102068e-01 8.47250998e-01 -7.86655303e-03 -3.48981142e-01 -9.40711915e-01 -1.11452818e+00 -1.56847909e-01 -4.77945149e-01 9.80597734e-02 1.17637491e+00 -1.64347291e-01 5.68987787e-01 -2.02329263e-01 5.79522669e-01 9.73280549e-01 2.68002838e-01 5.88814557e-01 -1.64005685e+00 -3.36453199e-01 -2.24556938e-01 -5.21624267e-01 -9.45548534e-01 6.52688816e-02 -9.81654465e-01 2.97996491e-01 -1.01081443e+00 9.84517217e-01 -7.03660309e-01 -6.95322692e-01 6.53200328e-01 -5.73394239e-01 4.66784239e-01 1.91807479e-01 1.55866712e-01 -1.42097652e+00 4.00776923e-01 8.80519152e-01 -5.31869411e-01 -2.84986645e-01 7.35513940e-02 -1.20048106e+00 8.41242850e-01 3.42908204e-01 -6.24596655e-01 -2.58442074e-01 -3.10440719e-01 -1.17762938e-01 -6.16412997e-01 7.05753624e-01 -1.27125180e+00 3.29522908e-01 8.50091688e-03 4.49367970e-01 -2.80277520e-01 2.52125323e-01 -1.09589589e+00 -2.20607057e-01 5.06079257e-01 -4.40470606e-01 -6.94268763e-01 1.45882577e-01 1.03456247e+00 1.89633220e-01 -2.68718451e-01 8.78926396e-01 -2.83262968e-01 -1.32842457e+00 3.97771657e-01 -3.21264602e-02 2.98565060e-01 1.36776102e+00 -2.16846228e-01 -2.88161099e-01 -8.93298313e-02 -1.04095316e+00 4.55136865e-01 3.50056529e-01 5.57097316e-01 4.99751925e-01 -8.29231799e-01 -7.08968818e-01 8.04896206e-02 5.24576843e-01 -1.32065192e-01 2.24617049e-01 5.03130078e-01 2.24311687e-02 1.96543530e-01 1.74976274e-01 -9.03884292e-01 -1.25218391e+00 1.05299175e+00 3.00250322e-01 8.39831121e-03 -6.74536705e-01 1.19453979e+00 6.61111176e-01 -5.90528131e-01 5.55860400e-01 7.87543412e-03 -1.72109291e-01 2.78008059e-02 6.07691109e-01 2.81638473e-01 1.00271277e-01 -3.25734019e-01 -2.99332708e-01 3.80063474e-01 -4.13752645e-01 2.23532125e-01 1.37635827e+00 -2.92845100e-01 1.14643559e-01 5.12739599e-01 1.04903984e+00 -4.64301974e-01 -1.35427034e+00 -5.51854491e-01 1.85386524e-01 -6.30611122e-01 4.88981605e-02 -1.18983710e+00 -7.26862848e-01 6.13042176e-01 1.22579646e+00 1.46154612e-01 9.82639015e-01 6.20406270e-01 4.87274498e-01 9.42020297e-01 5.63877523e-01 -1.07118845e+00 2.25216702e-01 3.48170012e-01 5.27317703e-01 -1.85072708e+00 -1.71876028e-01 -3.29215974e-01 -4.25268084e-01 5.53123653e-01 8.34910095e-01 2.43510008e-01 4.27659780e-01 -1.27634808e-01 2.41148025e-02 -3.06031257e-01 -8.01135004e-01 -7.06273377e-01 1.97914720e-01 8.47493708e-01 -2.67276049e-01 8.95835832e-02 2.53601044e-01 5.87694287e-01 1.82774328e-02 -1.86316192e-01 3.65879685e-01 7.42571652e-01 -7.94071317e-01 -7.00917363e-01 -4.65634972e-01 7.78836727e-01 -2.30243087e-01 -1.93524271e-01 -6.69845998e-01 5.39882243e-01 3.61360431e-01 9.28532302e-01 6.84078932e-02 -1.04034014e-01 1.58391610e-01 8.26760978e-02 4.34730500e-01 -7.40617871e-01 -2.20299646e-01 -4.79296446e-01 -2.43878067e-01 -3.38063896e-01 -2.14747682e-01 -7.01263249e-01 -9.66257513e-01 1.38723627e-01 -6.73461139e-01 2.28628889e-01 4.26502287e-01 9.68895435e-01 5.27124405e-01 3.83092821e-01 7.21529543e-01 -6.32662058e-01 -6.40832365e-01 -5.92606962e-01 -4.83259857e-01 7.04148889e-01 6.56881332e-01 -8.11684370e-01 -5.72349727e-01 3.46924752e-01]
[9.65168285369873, 1.9541254043579102]
8efe4af2-9649-4a17-ad9f-661a0c20bfab
non-intrusive-load-monitoring-based-on-self
2210.04176
null
https://arxiv.org/abs/2210.04176v1
https://arxiv.org/pdf/2210.04176v1.pdf
Non-intrusive Load Monitoring based on Self-supervised Learning
Deep learning models for non-intrusive load monitoring (NILM) tend to require a large amount of labeled data for training. However, it is difficult to generalize the trained models to unseen sites due to different load characteristics and operating patterns of appliances between data sets. For addressing such problems, self-supervised learning (SSL) is proposed in this paper, where labeled appliance-level data from the target data set or house is not required. Initially, only the aggregate power readings from target data set are required to pre-train a general network via a self-supervised pretext task to map aggregate power sequences to derived representatives. Then, supervised downstream tasks are carried out for each appliance category to fine-tune the pre-trained network, where the features learned in the pretext task are transferred. Utilizing labeled source data sets enables the downstream tasks to learn how each load is disaggregated, by mapping the aggregate to labels. Finally, the fine-tuned network is applied to load disaggregation for the target sites. For validation, multiple experimental cases are designed based on three publicly accessible REDD, UK-DALE, and REFIT data sets. Besides, state-of-the-art neural networks are employed to perform NILM task in the experiments. Based on the NILM results in various cases, SSL generally outperforms zero-shot learning in improving load disaggregation performance without any sub-metering data from the target data sets.
['Yixin Yu', 'Wenpeng Luan', 'Mingjun Zhong', 'Bochao Zhao', 'Shuyi Chen']
2022-10-09
null
null
null
null
['non-intrusive-load-monitoring', 'non-intrusive-load-monitoring', 'non-intrusive-load-monitoring']
['knowledge-base', 'miscellaneous', 'time-series']
[ 2.07580477e-02 3.98856513e-02 -3.98538142e-01 -7.38344193e-01 -8.25404584e-01 -3.08779925e-01 3.94896626e-01 -1.66952521e-01 1.33392587e-01 8.76148164e-01 1.81498945e-01 -1.82483330e-01 -2.53330264e-02 -1.00215495e+00 -4.89763051e-01 -8.49878848e-01 -1.87961251e-01 6.13690615e-01 -3.06651741e-01 -1.05933426e-02 -4.62042570e-01 1.95982039e-01 -1.21368694e+00 1.79642916e-01 9.58634675e-01 1.30252361e+00 2.60234654e-01 2.91907161e-01 -8.46319497e-02 1.07348871e+00 -9.64698434e-01 1.71740532e-01 2.23126009e-01 -4.63397264e-01 -6.38379991e-01 3.17418933e-01 -1.49420947e-01 -7.79872596e-01 -2.94098943e-01 9.26981628e-01 7.18124330e-01 1.56318411e-01 6.62864029e-01 -1.58562195e+00 -7.90321589e-01 1.10579014e+00 -5.32323599e-01 3.19853008e-01 1.61403343e-01 2.71816969e-01 9.73675191e-01 -1.08204477e-01 -1.79442838e-01 8.09664369e-01 5.79192400e-01 3.25549781e-01 -1.36378193e+00 -1.02515626e+00 7.52437860e-02 3.40018421e-01 -1.30813491e+00 -2.61154771e-01 1.05554557e+00 -1.90709636e-01 1.32194924e+00 2.58193463e-01 4.01652753e-01 1.37773705e+00 4.51871082e-02 8.54872942e-01 8.92843664e-01 -1.30833670e-01 7.03665793e-01 1.98309213e-01 2.89035767e-01 1.64020583e-01 1.59124374e-01 9.72454846e-02 -9.45477467e-03 -1.26458198e-01 2.03421846e-01 2.34121516e-01 -1.76553220e-01 -1.64742731e-02 -6.56142652e-01 8.01881552e-01 5.83689451e-01 5.36035180e-01 -6.58954263e-01 -2.74181813e-01 7.01810241e-01 2.90447325e-01 6.64938331e-01 1.84632033e-01 -7.73250163e-01 -6.93592653e-02 -1.03637981e+00 -3.76337260e-01 1.06487954e+00 1.01884902e+00 9.09848809e-01 4.58804339e-01 -3.16801012e-01 7.55738020e-01 1.62519991e-01 4.64402288e-01 8.54447722e-01 -5.86661279e-01 6.96850598e-01 8.52262676e-01 6.19329736e-02 -3.39509159e-01 -8.28471363e-01 -3.02181572e-01 -1.30632734e+00 -1.34111205e-02 -1.95090339e-01 -5.80237746e-01 -8.58701825e-01 1.66823447e+00 7.94447288e-02 3.07117760e-01 2.38850653e-01 5.04561365e-01 6.33809745e-01 1.03471613e+00 4.39079285e-01 -5.46262383e-01 1.10940170e+00 -7.14013577e-01 -7.92717457e-01 -7.17840791e-02 7.45675147e-01 -5.69565371e-02 1.10832822e+00 1.42725363e-01 -6.14009678e-01 -6.83873832e-01 -1.18981493e+00 3.30804437e-01 -6.10051334e-01 -1.98658789e-03 2.85036683e-01 4.31677431e-01 -4.96898949e-01 6.26278639e-01 -8.42440307e-01 -6.09590113e-01 8.49160969e-01 3.28136802e-01 2.61034042e-01 2.41594017e-01 -1.42612767e+00 7.90389776e-01 9.11790907e-01 4.57718372e-02 -1.02088320e+00 -8.19401741e-01 -9.30497766e-01 4.71773118e-01 1.41266778e-01 -2.15031922e-01 1.35363889e+00 -7.41160989e-01 -1.45088851e+00 4.00029778e-01 5.65613687e-01 -6.31367266e-01 3.37598294e-01 1.02687045e-04 -1.04081929e+00 -3.71999145e-01 2.27769300e-01 2.62774825e-01 7.48825312e-01 -1.26420105e+00 -9.54639018e-01 -2.42603928e-01 -1.58502489e-01 -9.52619687e-02 -4.83044207e-01 -5.45922458e-01 2.09973127e-01 -4.74860311e-01 -4.73807603e-01 -5.77524900e-01 1.95497155e-01 -9.42240894e-01 -8.04019153e-01 -6.38284683e-01 1.29084456e+00 -7.51228631e-01 1.35237169e+00 -2.30898976e+00 -4.51015085e-01 1.78518638e-01 -4.51252013e-02 3.67347747e-01 -8.76137465e-02 3.38161200e-01 -4.78413910e-01 -2.82213122e-01 -1.79047137e-01 -3.73641700e-01 6.69104934e-01 5.86649001e-01 -3.75476956e-01 4.05793846e-01 -1.50213325e-02 1.21661818e+00 -9.44363832e-01 -2.75976986e-01 8.60345721e-01 1.49293125e-01 1.07649341e-01 5.93261838e-01 -4.42269981e-01 4.97579664e-01 -3.53055149e-01 5.28065681e-01 4.78333980e-01 -3.42446506e-01 2.13290751e-01 -5.66269636e-01 3.02603811e-01 3.15742970e-01 -8.88165534e-01 1.46503305e+00 -9.07669604e-01 2.39565328e-01 -4.72931206e-01 -1.36708629e+00 8.79714251e-01 4.98502046e-01 9.12261128e-01 -9.13413346e-01 3.91849488e-01 -1.82166710e-01 -1.28733203e-01 -6.08915687e-01 -1.45361677e-01 -8.63102376e-02 -5.43611944e-01 6.72356844e-01 1.92990631e-01 1.34680539e-01 2.26158395e-01 -3.60928804e-01 1.14392102e+00 -8.53329599e-02 3.15516204e-01 -2.11894408e-01 3.84566814e-01 -1.82166636e-01 8.36142302e-01 4.85289812e-01 -1.92576066e-01 2.34506885e-03 6.19385317e-02 -7.05359519e-01 -9.54488099e-01 -1.21617496e+00 -1.73717871e-01 1.37142193e+00 -2.62811512e-01 -1.07296780e-01 -8.15170825e-01 -8.75345230e-01 -8.22734262e-04 1.54865992e+00 -6.79741442e-01 -3.45311761e-01 -3.01745445e-01 -1.24673629e+00 2.79984772e-01 8.92668247e-01 9.91105676e-01 -1.54204917e+00 -6.78279102e-01 6.37857676e-01 5.07534444e-02 -9.53455687e-01 -4.63838547e-01 8.83911014e-01 -4.85909492e-01 -1.10498440e+00 -2.02473491e-01 -6.94463015e-01 4.70559061e-01 -2.24911585e-01 1.35534656e+00 -4.46248323e-01 -1.60107777e-01 -1.89419482e-02 -2.24136308e-01 -5.47784805e-01 -5.42017698e-01 7.42459536e-01 2.25503415e-01 1.61334500e-01 1.13322210e+00 -1.12307048e+00 -4.23029065e-01 2.75806934e-01 -7.72903562e-01 -3.01066935e-01 3.30237627e-01 5.60893357e-01 4.98599827e-01 6.78577185e-01 1.11148572e+00 -8.29016209e-01 6.48512602e-01 -7.71592200e-01 -7.42399514e-01 2.40596771e-01 -7.92040229e-01 -1.47574320e-01 1.18208206e+00 -5.50607145e-01 -1.00882101e+00 -4.58831228e-02 5.22161322e-03 -4.07745212e-01 -4.25717205e-01 6.00418933e-02 -6.56464338e-01 8.78424823e-01 5.82231462e-01 2.34869555e-01 -4.86815780e-01 -6.15624368e-01 2.35187873e-01 1.02812111e+00 6.51610494e-01 -2.17160553e-01 9.24184680e-01 4.32582833e-02 -4.07045275e-01 -4.38467234e-01 -1.29813039e+00 -1.83414951e-01 -6.05875134e-01 2.11803794e-01 1.12354362e+00 -9.01014447e-01 -7.36139774e-01 6.15364015e-01 -6.52171433e-01 -8.22414041e-01 -9.30368245e-01 2.39596367e-01 -6.64647400e-01 -1.97348222e-01 -6.62571609e-01 -7.95932174e-01 -7.30401814e-01 -9.82668817e-01 9.82001483e-01 2.67215282e-01 -4.01988149e-01 -1.27283359e+00 9.70168635e-02 1.11859798e-01 4.87479895e-01 2.96851277e-01 1.39434850e+00 -1.05932391e+00 -2.28526667e-02 -2.92552412e-01 -8.12971294e-02 6.88020587e-01 1.04285872e+00 -5.40267229e-01 -1.24338996e+00 -6.16247475e-01 2.24073812e-01 -5.03558874e-01 2.95599371e-01 3.54318798e-01 1.47155392e+00 -4.77405787e-01 -2.79283315e-01 6.34250879e-01 1.37490392e+00 4.49059665e-01 4.19505179e-01 3.10606509e-01 6.71802521e-01 -3.32316896e-03 1.10603049e-01 6.34536743e-01 6.52919650e-01 2.37486035e-01 3.29079360e-01 -3.21694165e-01 1.48644507e-01 -4.56142098e-01 3.69035304e-01 9.51173961e-01 4.77036148e-01 -2.68704057e-01 -4.76718187e-01 3.42108786e-01 -1.89264286e+00 -9.22138035e-01 4.38482434e-01 1.96914732e+00 8.82901132e-01 2.15080366e-01 2.69036025e-01 3.85639131e-01 6.69312537e-01 3.01214010e-01 -1.37049568e+00 -1.25525877e-01 5.22722192e-02 4.07117605e-02 5.16372740e-01 1.90463573e-01 -1.31526613e+00 3.97956997e-01 5.55931473e+00 7.61401713e-01 -9.43299472e-01 5.09110928e-01 8.03714871e-01 -1.18551187e-01 2.34957919e-01 -6.83211029e-01 -7.02103257e-01 9.77543414e-01 1.32898629e+00 -2.92784363e-01 7.71224976e-01 9.94404793e-01 4.21569943e-01 3.76692675e-02 -1.58888292e+00 1.09762371e+00 -1.59388736e-01 -8.06603193e-01 -3.05669576e-01 -4.41190898e-02 1.03341794e+00 3.77400368e-01 -2.92773902e-01 9.71365809e-01 8.63201082e-01 -8.28280151e-01 3.86691429e-02 4.48020041e-01 6.65041685e-01 -8.65823269e-01 9.54874873e-01 5.61239362e-01 -1.37585211e+00 -5.36518097e-01 -3.30515116e-01 1.13598645e-01 2.16757983e-01 7.52752781e-01 -7.05413938e-01 6.62548959e-01 8.82405698e-01 5.95490038e-01 -4.54953134e-01 3.96238387e-01 -2.04482272e-01 9.01757061e-01 -4.72103029e-01 3.06236625e-01 1.51894987e-01 -2.11072862e-01 -2.08905518e-01 9.46064830e-01 1.66294470e-01 -4.53764722e-02 6.52139723e-01 9.86809671e-01 -2.79430300e-01 -2.42968425e-01 -4.43149269e-01 2.17333868e-01 5.08126140e-01 1.42750931e+00 -3.33403617e-01 -6.10851645e-01 -5.51879764e-01 1.12714291e+00 1.47735700e-01 6.48039460e-01 -1.06344545e+00 -3.27107579e-01 6.08802199e-01 -9.53216180e-02 2.99394429e-01 3.16477031e-01 -1.11552618e-01 -1.03450370e+00 -2.14699134e-01 -4.78110284e-01 5.76055288e-01 -7.24086761e-01 -2.07562876e+00 4.67987478e-01 7.55089968e-02 -1.12796509e+00 -6.73913002e-01 -4.32980768e-02 -1.15128100e+00 8.96621525e-01 -1.50142229e+00 -1.05023181e+00 -5.93919396e-01 7.73560107e-01 7.36350775e-01 -2.91355371e-01 9.87327814e-01 4.24411237e-01 -1.02462065e+00 6.84724152e-01 3.91265392e-01 4.62908596e-01 2.90783823e-01 -1.56527293e+00 4.85072225e-01 3.16390157e-01 -1.40105680e-01 -2.32752278e-01 3.80090743e-01 -4.66556728e-01 -1.02997005e+00 -1.96744215e+00 3.65067631e-01 -3.51734698e-01 8.27612460e-01 -5.34796000e-01 -1.04702640e+00 1.07222259e+00 3.96383077e-01 8.71586502e-02 9.20346320e-01 -9.49565023e-02 3.38530205e-02 -6.63866520e-01 -1.39892733e+00 -3.00687016e-03 6.76506460e-01 -5.79321325e-01 -8.47662270e-01 6.85646474e-01 5.02197087e-01 -1.80603508e-02 -1.25940979e+00 2.64619380e-01 -7.48044327e-02 -6.25480115e-01 4.82072890e-01 -5.35300612e-01 -1.36049315e-01 -2.76448548e-01 -4.82194871e-01 -1.84045696e+00 -5.95339537e-01 -5.03289223e-01 -6.89284086e-01 1.78685403e+00 3.24515790e-01 -7.86403894e-01 5.57993770e-01 6.06081486e-01 -8.06924701e-02 -4.99890894e-01 -7.97548592e-01 -8.28324378e-01 -5.62336156e-03 -3.27225298e-01 1.41342020e+00 1.00772619e+00 5.56669198e-03 7.47181892e-01 -2.06919178e-01 3.66979331e-01 6.97843254e-01 3.32232058e-01 4.94739562e-01 -1.19639122e+00 -2.10115135e-01 -1.55331925e-01 -1.59766972e-01 -7.46406615e-01 5.23685873e-01 -1.03293836e+00 2.58222111e-02 -1.53376567e+00 1.45913213e-01 -3.42995733e-01 -7.33127952e-01 9.08361912e-01 -7.87252001e-03 2.34058663e-01 2.44368985e-02 6.01105392e-02 -5.56606770e-01 8.12920928e-01 6.65433407e-01 -6.66940510e-01 -5.07043719e-01 2.65552014e-01 -5.08197725e-01 6.50797188e-01 1.35934079e+00 -2.33637914e-01 -8.21068823e-01 -2.57212639e-01 -4.63344544e-01 -2.51687646e-01 1.02617294e-01 -1.14039183e+00 -1.88754573e-01 4.29774784e-02 9.55980182e-01 -8.31373870e-01 2.59438269e-02 -1.14204359e+00 2.82062382e-01 3.70714962e-01 -8.21719170e-02 -7.43931457e-02 5.51747307e-02 2.72810370e-01 2.85050005e-01 -1.58495203e-01 7.57985950e-01 -7.05307126e-02 -7.05106258e-01 5.68837702e-01 -1.10638395e-01 -3.45383286e-02 1.15348220e+00 -6.40639570e-03 -9.08321962e-02 -3.06280583e-01 -9.50129271e-01 6.09756649e-01 3.48199345e-02 5.80957651e-01 -2.16957718e-01 -1.74980521e+00 -5.91376245e-01 4.36946332e-01 9.27026868e-02 1.36612251e-01 2.48534799e-01 3.93482864e-01 3.01169217e-01 3.30848992e-01 1.16452090e-01 -6.62545145e-01 -3.51617217e-01 1.00862682e+00 5.96271873e-01 -4.03652698e-01 -8.24087203e-01 -3.51346545e-02 1.65515780e-01 -6.10615790e-01 4.19770211e-01 -7.48680949e-01 7.80244842e-02 3.10291052e-01 5.29344738e-01 4.87577885e-01 3.76896799e-01 -3.02323908e-01 -1.42798722e-01 1.40964121e-01 1.29312286e-02 6.85798407e-01 1.60670662e+00 -2.86685765e-01 1.32938892e-01 8.42255652e-01 1.55928326e+00 -8.66155744e-01 -1.45700610e+00 -2.92383462e-01 3.26133706e-02 2.28582293e-01 1.11691169e-01 -1.09193933e+00 -1.49319184e+00 5.63290060e-01 1.03431499e+00 7.14849591e-01 1.54716611e+00 1.30679488e-01 1.05504167e+00 3.34794283e-01 4.96177107e-01 -1.35038483e+00 -4.00524229e-01 9.12012085e-02 5.18620908e-01 -1.35785043e+00 -3.84185731e-01 3.74146342e-01 -2.52502531e-01 7.27525294e-01 7.92311907e-01 2.68776398e-02 8.20638359e-01 4.79454190e-01 1.04563676e-01 -9.96904001e-02 -5.36581755e-01 -7.13827014e-02 -2.16854140e-01 8.75167310e-01 -4.01103795e-02 4.49189216e-01 3.25842202e-01 9.47981000e-01 -3.95630389e-01 1.82764769e-01 9.48626176e-02 8.14717710e-01 -1.06423996e-01 -7.18327522e-01 -2.22984284e-01 1.04132199e+00 -9.06059444e-02 1.69488415e-01 1.88738778e-01 6.16014540e-01 3.05139959e-01 1.14207482e+00 3.68864983e-01 -4.12315100e-01 7.09693432e-01 3.93428892e-01 6.56579956e-02 -5.60868621e-01 -5.36116302e-01 -2.04362795e-01 -1.67419821e-01 -4.14052337e-01 -1.47322029e-01 -4.16317523e-01 -1.21547031e+00 -3.28389049e-01 -2.11317465e-01 6.54341653e-02 1.45332679e-01 1.09214485e+00 2.21081123e-01 1.03481257e+00 1.40202785e+00 -1.07693446e+00 -9.62029517e-01 -1.53585756e+00 -1.05507505e+00 7.98524499e-01 2.95863271e-01 -5.14641702e-01 -4.54149425e-01 6.85524046e-02]
[16.05364418029785, 7.570494651794434]
7c6717b7-e891-44f8-8309-2c0fa0337f54
modelling-high-level-mathematical-reasoning
2006.09265
null
https://arxiv.org/abs/2006.09265v2
https://arxiv.org/pdf/2006.09265v2.pdf
IsarStep: a Benchmark for High-level Mathematical Reasoning
A well-defined benchmark is essential for measuring and accelerating research progress of machine learning models. In this paper, we present a benchmark for high-level mathematical reasoning and study the reasoning capabilities of neural sequence-to-sequence models. We build a non-synthetic dataset from the largest repository of proofs written by human experts in a theorem prover. The dataset has a broad coverage of undergraduate and research-level mathematical and computer science theorems. In our defined task, a model is required to fill in a missing intermediate proposition given surrounding proofs. This task provides a starting point for the long-term goal of having machines generate human-readable proofs automatically. Our experiments and analysis reveal that while the task is challenging, neural models can capture non-trivial mathematical reasoning. We further design a hierarchical transformer that outperforms the transformer baseline.
['Lei Yu', 'Wenda Li', 'Yuhuai Wu', 'Lawrence C. Paulson']
2020-06-13
isarstep-a-benchmark-for-high-level
https://openreview.net/forum?id=Pzj6fzU6wkj
https://openreview.net/pdf?id=Pzj6fzU6wkj
iclr-2021-1
['mathematical-proofs', 'mathematical-reasoning']
['miscellaneous', 'natural-language-processing']
[ 3.53985697e-01 4.06994164e-01 -2.55978435e-01 -7.20516816e-02 -8.04414332e-01 -9.57181811e-01 8.37574899e-01 1.67680845e-01 1.99908078e-01 9.80552375e-01 -3.26000266e-02 -1.40054023e+00 -1.61025763e-01 -1.09756386e+00 -1.47946429e+00 1.05029270e-01 -8.93481821e-02 3.18123758e-01 6.78829178e-02 -1.77316695e-01 3.99671257e-01 2.16635987e-01 -1.32546115e+00 6.31519735e-01 1.20848453e+00 4.11393613e-01 9.87537131e-02 1.03544497e+00 -1.92799211e-01 1.58835411e+00 -5.15361249e-01 -1.01631391e+00 4.03956324e-02 -4.95936960e-01 -1.24480200e+00 -7.69250274e-01 1.01927602e+00 -4.33804333e-01 -3.73888254e-01 1.26248407e+00 -1.99934635e-02 -2.67576694e-01 7.44913936e-01 -1.71926618e+00 -8.26138020e-01 1.44979453e+00 1.17980078e-01 2.34553412e-01 6.26494348e-01 2.14071810e-01 1.31719518e+00 -4.45496082e-01 7.42318809e-01 1.33301485e+00 9.14133549e-01 9.04224932e-01 -1.43456459e+00 -6.26292586e-01 -1.62773475e-01 6.06813908e-01 -8.27108979e-01 -3.07382017e-01 8.02926719e-01 -3.80871177e-01 8.81422400e-01 3.41256887e-01 6.23248637e-01 1.17389643e+00 4.06958342e-01 8.99562299e-01 9.39857602e-01 -5.34619153e-01 -1.73153002e-02 -7.09765255e-02 4.00720030e-01 1.26983178e+00 3.99329394e-01 -1.57356903e-01 -3.78655225e-01 -1.19408928e-01 5.76555371e-01 -3.21459025e-01 -2.06347108e-01 -5.30901551e-01 -1.56735373e+00 6.66604042e-01 1.97496355e-01 2.69831240e-01 8.95252824e-02 6.69696391e-01 5.30941486e-01 6.56665385e-01 -2.34779790e-01 1.01319242e+00 -8.04966569e-01 -5.72764099e-01 -1.01398015e+00 8.52967620e-01 1.26033473e+00 9.67075050e-01 3.54582779e-02 -1.70426160e-01 -3.35232675e-01 2.22610474e-01 -1.37541234e-01 3.54963869e-01 2.90970635e-02 -1.53319538e+00 5.58390200e-01 5.70273817e-01 5.80639243e-02 -7.77505398e-01 -7.99437333e-03 -5.30391991e-01 -8.59897137e-01 -5.55041917e-02 7.74692953e-01 2.33929947e-01 -3.06458026e-01 1.99783051e+00 -5.02954721e-02 3.98928106e-01 3.08971047e-01 3.10267746e-01 1.10248137e+00 7.02654481e-01 -2.65728474e-01 -7.83092752e-02 1.23146951e+00 -1.04707789e+00 -4.29522991e-01 8.14766511e-02 9.11059499e-01 -3.45086485e-01 1.10012436e+00 6.67423844e-01 -1.64332759e+00 -2.87072361e-01 -1.03595495e+00 -4.71752644e-01 -3.55765909e-01 -2.24217121e-02 1.01229489e+00 1.26414821e-01 -1.11186862e+00 7.53355384e-01 -3.24773103e-01 2.90850729e-01 5.83844900e-01 -1.23411305e-01 -2.34669060e-01 -3.89701605e-01 -1.27878380e+00 1.09233201e+00 4.87260759e-01 -1.95614204e-01 -1.18007541e+00 -1.52836919e+00 -1.07605207e+00 3.54556322e-01 4.22401696e-01 -1.14727807e+00 1.94313562e+00 -4.38386947e-01 -1.05602276e+00 9.44005430e-01 -1.62807494e-01 -1.07446718e+00 8.49733591e-01 9.30339172e-02 9.80527252e-02 7.64915049e-02 1.69150874e-01 6.64199173e-01 3.52548182e-01 -1.14255893e+00 -3.59928161e-01 1.41418586e-02 6.49519503e-01 -6.37234211e-01 1.01466216e-01 -9.34652612e-02 1.34484351e-01 -5.00576437e-01 -2.49351218e-01 -4.92655933e-01 2.15572178e-01 -5.33449017e-02 -4.37051296e-01 -7.63144791e-01 5.47944427e-01 -7.14725435e-01 1.06949985e+00 -1.49276257e+00 3.24924201e-01 -1.82003647e-01 5.80013871e-01 1.19445175e-01 -1.10299148e-01 2.53319502e-01 -2.47161165e-01 4.85929072e-01 -8.16942230e-02 3.39103676e-02 6.99942112e-01 -6.46474808e-02 -9.02710915e-01 -1.73956603e-01 3.02551091e-01 1.55404699e+00 -1.32846367e+00 -8.27466369e-01 -7.86760449e-03 -1.04317196e-01 -8.08112681e-01 2.24123448e-01 -1.01496279e+00 -6.97704852e-02 -1.51943460e-01 6.01279795e-01 5.12094319e-01 -5.54986596e-01 4.87172455e-02 1.58301488e-01 2.98807234e-01 9.49459910e-01 -5.85590184e-01 1.91600525e+00 -5.36733091e-01 9.77147877e-01 2.64483877e-02 -1.13170695e+00 3.82669657e-01 1.50658861e-01 -1.03292733e-01 -6.17573917e-01 2.29441058e-02 1.18758366e-01 3.34161133e-01 -3.60502392e-01 3.66744637e-01 -9.79133025e-02 -6.96959496e-02 7.85010695e-01 -7.69347772e-02 -6.10027134e-01 8.32148135e-01 7.66037524e-01 1.54094195e+00 4.56712842e-01 -1.99108347e-02 -2.06890181e-01 6.18478715e-01 1.75116435e-01 3.12994927e-01 1.20354342e+00 9.63237416e-03 -4.86976467e-02 1.10800672e+00 -6.67981684e-01 -1.38323426e+00 -1.12016201e+00 1.20382063e-01 8.90121460e-01 -3.34163606e-01 -8.61322403e-01 -8.00493479e-01 -7.99964666e-01 4.97987010e-02 1.14160466e+00 -4.21729118e-01 -3.79417688e-01 -6.75410032e-01 8.27628523e-02 1.08172500e+00 5.28567672e-01 5.15158355e-01 -1.08566701e+00 -4.94514704e-01 9.01810452e-03 -5.99355936e-01 -1.31691563e+00 5.61809428e-02 -4.47176173e-02 -8.27143073e-01 -1.37651408e+00 -2.81022429e-01 -9.60613906e-01 5.68904877e-01 5.89617193e-02 1.76192033e+00 5.82849741e-01 -2.46535957e-01 3.34062517e-01 1.49659708e-01 -6.25734866e-01 -9.98613477e-01 1.22003436e-01 -2.55995303e-01 -1.00337744e+00 2.64498860e-01 -6.74454868e-01 6.24727421e-02 -3.35864037e-01 -5.84154427e-01 5.94779074e-01 4.61474657e-01 8.76306891e-01 -4.22933362e-02 2.03663707e-01 3.34547549e-01 -5.65285563e-01 1.00857699e+00 -1.50405347e-01 -9.38143194e-01 5.85810184e-01 -3.73282611e-01 5.50048828e-01 1.21956730e+00 -4.27576244e-01 -5.27236581e-01 -6.24728739e-01 3.13058347e-01 -4.36209381e-01 8.73522013e-02 5.11124611e-01 2.54347362e-02 2.53748477e-01 6.40123308e-01 5.99082053e-01 -4.70518284e-02 8.21099505e-02 4.81074929e-01 -1.29128294e-02 9.65895593e-01 -1.62801886e+00 1.33139801e+00 -3.14938962e-01 6.30516469e-01 -7.60951787e-02 -1.21926832e+00 2.84297615e-01 -5.31582832e-01 5.06530814e-02 2.02776566e-01 -7.20025182e-01 -1.44605458e+00 -9.94768832e-03 -1.68063486e+00 -9.89715755e-01 -1.09838448e-01 4.01194952e-02 -8.27490032e-01 4.77196425e-01 -7.91185200e-01 -6.84947073e-01 -4.85754400e-01 -1.11184669e+00 7.25428343e-01 -1.30751312e-01 -5.45181394e-01 -1.10640717e+00 -7.88216442e-02 5.50233662e-01 1.03293054e-01 2.37068355e-01 1.62048841e+00 -4.70675379e-01 -1.03582799e+00 6.51024133e-02 -6.16751611e-01 4.87197101e-01 -3.38111430e-01 9.39996317e-02 -7.26800978e-01 1.56409934e-01 -3.54101241e-01 -7.67314911e-01 8.89915705e-01 5.23002855e-02 1.68674576e+00 -7.15091228e-01 -1.37596294e-01 3.31169546e-01 9.11876857e-01 -4.22982514e-01 6.42305195e-01 4.29793030e-01 3.49636853e-01 1.74316570e-01 2.81663328e-01 -2.08216354e-01 6.25524879e-01 -6.15577176e-02 2.99236149e-01 3.77317220e-01 6.87228981e-03 -8.28329742e-01 3.40269208e-01 4.16688502e-01 2.11431775e-02 1.41821414e-01 -1.30165660e+00 6.39661312e-01 -1.70385742e+00 -1.59568202e+00 -7.61608109e-02 1.77448106e+00 1.64558804e+00 5.78151047e-01 -8.07124190e-03 5.18115222e-01 2.29467452e-01 -2.78169900e-01 -2.24647239e-01 -7.76901066e-01 2.51771003e-01 5.18953621e-01 4.33647726e-03 6.90996945e-01 -9.29641843e-01 8.44328284e-01 6.65825844e+00 8.16417515e-01 -6.46196783e-01 -4.09418195e-01 1.66909263e-01 8.17262288e-03 -7.05120504e-01 5.91743477e-02 -4.67734873e-01 2.64393151e-01 1.22014332e+00 -6.07331157e-01 8.65147054e-01 8.93028796e-01 -1.01520330e-01 2.97185481e-01 -1.96849072e+00 7.08117187e-01 9.14703459e-02 -1.96189690e+00 3.88114929e-01 -1.80559754e-01 5.28690755e-01 -5.34115851e-01 -1.71365272e-02 8.06837261e-01 7.66586304e-01 -1.53907347e+00 9.56822634e-01 4.99582797e-01 7.37657309e-01 -6.78370416e-01 4.88461435e-01 7.15263128e-01 -6.74564302e-01 -2.68683046e-01 -8.04408938e-02 -7.39891768e-01 -4.62770313e-01 3.20777148e-01 -1.24034917e+00 4.03783619e-01 2.36022577e-01 9.10668015e-01 -6.81977689e-01 7.79562593e-01 -7.94041097e-01 6.96431756e-01 1.76012330e-02 -4.69679385e-01 3.44109982e-01 6.39530048e-02 4.15757656e-01 1.36663389e+00 -6.89146072e-02 9.24788043e-02 -2.57287711e-01 1.83199275e+00 -5.69366157e-01 -6.20624781e-01 -9.32314396e-01 -4.35705066e-01 4.78728145e-01 1.17873132e+00 -2.26647764e-01 -6.80547237e-01 -3.26607339e-02 5.79856277e-01 6.19945467e-01 1.34608626e-01 -1.17551303e+00 -4.01554167e-01 3.89135182e-01 -5.22733070e-02 -6.15627989e-02 -4.29341346e-01 -5.66115797e-01 -1.31236219e+00 4.18260992e-01 -1.32044935e+00 1.09875232e-01 -1.31285477e+00 -9.62743223e-01 -2.49476209e-01 1.48760289e-01 -5.22167206e-01 -4.99687463e-01 -8.70966494e-01 -8.49717915e-01 7.65059471e-01 -1.69123900e+00 -1.31088078e+00 -1.24241523e-01 2.44945034e-01 4.23015237e-01 2.25141132e-03 9.63517964e-01 -8.17817673e-02 -3.30422461e-01 7.16832399e-01 -3.81269127e-01 6.84013247e-01 2.14105830e-01 -1.56975222e+00 4.93978560e-01 1.06206894e+00 2.21569270e-01 1.24739301e+00 9.13935542e-01 -3.85662645e-01 -1.97756314e+00 -1.06587911e+00 1.30260432e+00 -9.57685351e-01 1.19893277e+00 -4.45056379e-01 -6.37972951e-01 1.02726185e+00 1.04854032e-01 -3.65214407e-01 1.43516049e-01 2.34046608e-01 -9.09844756e-01 8.30127969e-02 -9.31796193e-01 8.91462028e-01 1.26443791e+00 -8.68294179e-01 -1.39507210e+00 5.36922812e-01 9.27775800e-01 -5.61169744e-01 -8.94473910e-01 3.00841719e-01 7.33620822e-01 -6.74837649e-01 1.17401397e+00 -1.20513213e+00 1.63748872e+00 -9.84199792e-02 1.31295174e-01 -1.09457004e+00 -1.75052926e-01 -9.03572381e-01 -6.79380059e-01 1.04441893e+00 5.26079774e-01 -1.45923451e-01 6.13358557e-01 6.89986348e-01 -1.01713888e-01 -8.47582638e-01 -5.76559186e-01 -7.86824942e-01 8.36184502e-01 -6.87116206e-01 5.56101680e-01 9.78550553e-01 4.96825516e-01 6.46925569e-01 3.24432760e-01 -1.24375083e-01 9.44638133e-01 5.64805925e-01 1.09341693e+00 -1.41716230e+00 -3.95520590e-02 -9.98147309e-01 -7.37231150e-02 -9.90638793e-01 1.15459692e+00 -1.57545245e+00 -1.92586213e-01 -1.92822778e+00 6.30461216e-01 -1.07621931e-01 -4.46153656e-02 7.00587332e-01 2.38392898e-03 -1.84901088e-01 -6.00140169e-03 -5.62111616e-01 -8.09907377e-01 2.11889654e-01 1.42036378e+00 -7.00826645e-01 5.60041428e-01 -2.62350082e-01 -7.62031555e-01 6.76906765e-01 7.13559210e-01 -1.29913539e-01 -3.98577690e-01 -4.66464728e-01 8.67430329e-01 1.78209275e-01 1.19090641e+00 -6.92063868e-01 3.30778539e-01 -5.08308709e-01 1.52344406e-01 -5.98147392e-01 -1.01549834e-01 -5.15729487e-01 -4.97911662e-01 6.93030357e-01 -8.34264278e-01 2.05996156e-01 3.71874481e-01 3.71632203e-02 3.63371931e-02 -1.34696573e-01 4.45850402e-01 -4.73635226e-01 -3.92417669e-01 -3.43153365e-02 -6.92681894e-02 5.46956956e-01 7.32938766e-01 1.06989600e-01 -5.89040339e-01 -2.41690889e-01 -1.62695333e-01 4.45987552e-01 4.17366266e-01 3.42122823e-01 7.27259338e-01 -1.07136595e+00 -8.50747943e-01 -1.78323984e-01 5.74970245e-03 1.47482455e-01 -4.03177857e-01 6.51183009e-01 -5.99131107e-01 8.81213129e-01 -2.16944933e-01 -3.02254349e-01 -1.37657845e+00 7.81294048e-01 4.80327904e-01 -5.46977758e-01 -5.88002503e-01 7.24228084e-01 -2.79370546e-01 -8.00075471e-01 3.81920636e-01 -1.10649192e+00 3.38566631e-01 -6.94591463e-01 7.14517593e-01 2.37543970e-01 -2.39021763e-01 3.51442665e-01 -1.85016245e-01 6.32447079e-02 1.12456672e-01 1.18389063e-01 1.36436582e+00 6.91654384e-01 -7.01171219e-01 4.84324038e-01 6.89071357e-01 -1.42725214e-01 -3.61088306e-01 -8.56795460e-02 1.17007956e-01 1.17217287e-01 -2.50779361e-01 -1.02467287e+00 -1.83947742e-01 1.11409986e+00 -6.36133850e-01 2.26632342e-01 5.34605980e-01 -1.16047785e-01 7.64347434e-01 1.20480573e+00 4.36964363e-01 -6.01298511e-01 9.36732292e-02 9.62834775e-01 7.76493907e-01 -1.02058363e+00 -6.55213669e-02 -2.09168389e-01 6.39612451e-02 1.33530056e+00 6.19164228e-01 7.79587254e-02 -1.24210194e-01 5.43795586e-01 -6.39105618e-01 -2.38938327e-03 -1.44462669e+00 4.19516951e-01 1.63595304e-01 3.09806645e-01 5.28685391e-01 -4.14821170e-02 7.16928914e-02 6.26183093e-01 -9.85559285e-01 5.71221709e-01 7.37287700e-01 8.55169415e-01 -1.56240061e-01 -9.35296178e-01 -3.37563545e-01 3.09463441e-01 -4.72622812e-01 -5.42119920e-01 -7.12526560e-01 7.79490173e-01 -1.42196581e-01 7.52579749e-01 -3.59618187e-01 4.61991616e-02 -1.79663584e-01 4.39750433e-01 1.31699991e+00 -3.27341288e-01 -2.91433424e-01 -1.10114741e+00 4.00099546e-01 -3.42724800e-01 -9.07862857e-02 -3.69677722e-01 -1.31189525e+00 -9.91064429e-01 1.48677260e-01 4.94569153e-01 3.55130017e-01 9.08288777e-01 6.16791695e-02 8.43725026e-01 6.68050647e-02 -1.87707931e-01 -9.84342337e-01 -8.30433130e-01 1.68519169e-01 3.83156508e-01 5.90810895e-01 -2.47310966e-01 -3.54522705e-01 3.95177305e-01]
[9.072474479675293, 7.111479759216309]
49982ba5-ab7b-4a2f-bd65-2f17d7d6f4f8
cluster-based-mention-typing-for-named-entity
2109.11389
null
https://arxiv.org/abs/2109.11389v1
https://arxiv.org/pdf/2109.11389v1.pdf
Cluster-based Mention Typing for Named Entity Disambiguation
An entity mention in text such as "Washington" may correspond to many different named entities such as the city "Washington D.C." or the newspaper "Washington Post." The goal of named entity disambiguation is to identify the mentioned named entity correctly among all possible candidates. If the type (e.g. location or person) of a mentioned entity can be correctly predicted from the context, it may increase the chance of selecting the right candidate by assigning low probability to the unlikely ones. This paper proposes cluster-based mention typing for named entity disambiguation. The aim of mention typing is to predict the type of a given mention based on its context. Generally, manually curated type taxonomies such as Wikipedia categories are used. We introduce cluster-based mention typing, where named entities are clustered based on their contextual similarities and the cluster ids are assigned as types. The hyperlinked mentions and their context in Wikipedia are used in order to obtain these cluster-based types. Then, mention typing models are trained on these mentions, which have been labeled with their cluster-based types through distant supervision. At the named entity disambiguation phase, first the cluster-based types of a given mention are predicted and then, these types are used as features in a ranking model to select the best entity among the candidates. We represent entities at multiple contextual levels and obtain different clusterings (and thus typing models) based on each level. As each clustering breaks the entity space differently, mention typing based on each clustering discriminates the mention differently. When predictions from all typing models are used together, our system achieves better or comparable results based on randomization tests with respect to the state-of-the-art levels on four defacto test sets.
['Arzucan Özgür', 'Arda Çelebi']
2021-09-23
null
null
null
null
['entity-disambiguation']
['natural-language-processing']
[-5.30172586e-01 2.70639986e-01 -4.67555791e-01 -5.21480918e-01 -5.12126863e-01 -7.40380108e-01 6.92980170e-01 8.34995627e-01 -6.69870436e-01 9.11427796e-01 3.73654664e-01 -2.69215614e-01 1.74524565e-03 -1.19607842e+00 -5.80605268e-01 -5.16219079e-01 -2.17783883e-01 9.33882833e-01 5.21594524e-01 -8.17485899e-02 2.65844375e-01 3.29908580e-01 -1.37996662e+00 3.86473387e-01 1.41947246e+00 5.55589795e-01 4.29446310e-01 2.05036090e-03 -7.34544873e-01 4.74015594e-01 -8.73267174e-01 -5.92548728e-01 -4.41359319e-02 -2.16448843e-01 -1.13806295e+00 -3.05826724e-01 3.09475720e-01 3.58513415e-01 1.41359806e-01 1.18582106e+00 1.49511069e-01 2.63410866e-01 9.00509953e-01 -8.48238587e-01 -5.63152254e-01 1.11755645e+00 -2.57302433e-01 -1.23498499e-01 5.52651703e-01 -4.65330124e-01 1.59888184e+00 -8.84946108e-01 1.03373396e+00 1.06678021e+00 4.97056514e-01 6.01256371e-01 -1.19415855e+00 -4.41460252e-01 4.92615104e-01 3.38629931e-01 -1.57483673e+00 -5.56146167e-02 3.67951512e-01 -6.19576871e-01 7.82758415e-01 2.04700172e-01 2.35054597e-01 3.98925424e-01 -1.81089163e-01 6.22396827e-01 8.93450618e-01 -4.88073587e-01 4.24995840e-01 4.37066317e-01 6.84019208e-01 4.77049053e-01 5.95095098e-01 -3.98522228e-01 -2.52085298e-01 -3.08192998e-01 6.44180030e-02 -2.06847385e-01 -2.18629017e-01 -1.60313979e-01 -1.26853371e+00 6.90416574e-01 9.13490534e-01 7.06947863e-01 -4.63071972e-01 -3.85382384e-01 2.21757755e-01 -1.58252016e-01 3.31168413e-01 9.36361432e-01 -8.36254776e-01 4.67657059e-01 -8.16275179e-01 8.99397116e-03 1.01555693e+00 1.02810121e+00 1.19895720e+00 -5.88736236e-01 -2.15949520e-01 1.12532365e+00 5.35056233e-01 3.51776242e-01 4.85533625e-01 -2.51349002e-01 5.57965577e-01 1.11477709e+00 4.42058027e-01 -9.91492271e-01 -7.26741254e-01 -1.70241758e-01 -5.89723051e-01 -2.11048517e-02 4.43730950e-01 -3.09536427e-01 -1.08004558e+00 1.77901125e+00 6.33910954e-01 3.88230264e-01 3.46491247e-01 8.43449831e-01 1.05398035e+00 5.58633089e-01 5.64455748e-01 -1.41041979e-01 1.63771558e+00 -5.07575214e-01 -4.76628035e-01 -1.49575606e-01 8.90534163e-01 -5.72855592e-01 5.09511054e-01 -2.70155221e-01 -3.79867166e-01 -2.97132492e-01 -7.37806618e-01 4.26973641e-01 -9.24828649e-01 2.47222558e-01 2.55559921e-01 4.81987536e-01 -6.35244966e-01 6.95484817e-01 -7.45814323e-01 -5.61937988e-01 -1.83818921e-01 2.93032289e-01 -2.84056604e-01 1.87303618e-01 -1.72625375e+00 1.02720332e+00 7.57374644e-01 -1.44461423e-01 -1.72892109e-01 -4.65317070e-01 -9.23594117e-01 9.66885611e-02 4.46539730e-01 -6.64101720e-01 6.98265851e-01 -6.36730373e-01 -6.29249871e-01 8.80310118e-01 -4.71706837e-01 -5.15724182e-01 -4.16482836e-02 1.35890633e-01 -9.58135009e-01 3.38890478e-02 7.17613697e-01 4.13648963e-01 1.44878194e-01 -1.46960843e+00 -1.24231660e+00 -1.31264850e-01 3.50789726e-01 1.66193485e-01 1.26449903e-02 -4.09062505e-02 -3.82996678e-01 -4.92913425e-01 4.34414446e-01 -1.03309953e+00 -3.37260574e-01 -8.74156892e-01 -8.29901993e-01 -7.31305778e-01 3.34244758e-01 -4.46153253e-01 1.50700557e+00 -1.78130114e+00 -2.19394863e-02 4.66791302e-01 3.15210134e-01 5.53829409e-02 2.72944242e-01 3.07953566e-01 -1.73232898e-01 6.18055463e-01 1.09316126e-01 -8.45665187e-02 8.20840523e-02 8.95277858e-02 4.14018966e-02 1.80032462e-01 1.23215392e-01 2.82569706e-01 -1.06443405e+00 -5.96845329e-01 -8.54936987e-02 3.48998085e-02 -4.88979429e-01 1.11007400e-01 -4.07178491e-01 2.19310969e-01 -7.11406887e-01 4.64580238e-01 4.69290763e-01 -1.98038846e-01 6.66122496e-01 -4.33761716e-01 -3.43000859e-01 8.27309906e-01 -1.45140982e+00 9.25982773e-01 -5.28324485e-01 1.25165343e-01 -3.17572027e-01 -7.56078362e-01 1.02390838e+00 4.66736823e-01 3.99733514e-01 -1.28043042e-02 -3.82343918e-01 5.30718386e-01 3.49488147e-02 -4.04934555e-01 7.74344385e-01 1.07969707e-02 -5.04213274e-01 7.34465495e-02 5.05635031e-02 5.96648335e-01 6.27470434e-01 4.98859167e-01 9.32838321e-01 -5.89866750e-02 5.51314414e-01 -2.34576181e-01 5.16521394e-01 3.45888317e-01 1.07150447e+00 7.47081101e-01 6.38274625e-02 3.68111640e-01 3.99006993e-01 -2.25140020e-01 -7.93975830e-01 -9.49878633e-01 -3.52236658e-01 9.98245716e-01 4.26286399e-01 -6.57366097e-01 -4.65038538e-01 -1.15984964e+00 5.46485707e-02 8.54053438e-01 -6.89017713e-01 2.84096837e-01 -4.84622240e-01 -7.21163332e-01 3.59599292e-01 2.37129360e-01 3.08418393e-01 -1.13836312e+00 -7.33599365e-02 2.94328421e-01 -4.95543569e-01 -8.63741457e-01 -1.38638675e-01 4.68418449e-01 -1.98824480e-01 -1.25096416e+00 -4.07340974e-01 -9.57809031e-01 9.03074801e-01 -3.85039002e-02 1.12841618e+00 3.64962220e-01 4.96117532e-01 -1.71356782e-01 -7.73187637e-01 1.01664953e-01 -3.75218153e-01 4.09016043e-01 4.17022258e-01 6.17897213e-02 5.60668945e-01 -2.23557636e-01 -3.54218751e-01 4.91818428e-01 -5.39970875e-01 -4.05902892e-01 4.82108444e-01 8.65045190e-01 4.91145283e-01 2.19898537e-01 5.84886491e-01 -1.64410210e+00 3.23263705e-02 -9.51504052e-01 -3.46351951e-01 5.43410480e-01 -6.59971952e-01 3.32499593e-01 8.36988151e-01 -5.05846918e-01 -1.21057284e+00 2.29847223e-01 -9.86756459e-02 1.08611107e-01 -5.04237771e-01 1.17765129e+00 -6.41109824e-01 4.30948973e-01 7.47221172e-01 -1.78531975e-01 -9.89595175e-01 -5.54426789e-01 2.56847382e-01 6.32046521e-01 4.18300301e-01 -8.86946201e-01 7.43907571e-01 -1.58867881e-01 -3.97049218e-01 -4.73328322e-01 -9.33313489e-01 -8.36253762e-01 -1.11165929e+00 -1.43613890e-01 1.03171110e+00 -8.75716090e-01 -3.37137312e-01 1.26544774e-01 -1.27693355e+00 1.94007978e-01 1.12826861e-01 6.23389244e-01 1.19016998e-01 3.35505784e-01 -4.16726351e-01 -5.45178711e-01 1.30006358e-01 -1.00184751e+00 7.47297883e-01 5.48267543e-01 -4.45463657e-01 -1.21776414e+00 7.95147195e-02 8.55771676e-02 -1.70373529e-01 2.28931129e-01 1.25743461e+00 -1.58870673e+00 -5.16240418e-01 -2.88360626e-01 2.53517460e-02 -3.90156299e-01 2.40536973e-01 1.94207355e-01 -6.27389371e-01 1.90640427e-02 -6.68408692e-01 3.56234729e-01 6.85439050e-01 1.25959530e-01 1.84386477e-01 -6.47622883e-01 -1.05460727e+00 3.10574114e-01 1.50381231e+00 2.83469558e-01 4.29586172e-01 7.77324557e-01 9.35553789e-01 5.33951700e-01 9.26257312e-01 2.13483229e-01 5.55173278e-01 7.60206342e-01 9.49873254e-02 9.50503722e-02 1.34497583e-01 -2.85255700e-01 2.18175828e-01 5.97434223e-01 5.30787632e-02 -2.84634709e-01 -1.24335873e+00 8.14048707e-01 -1.80021620e+00 -1.21768022e+00 -5.91283619e-01 2.43399000e+00 1.17286623e+00 1.44847170e-01 1.95673257e-01 -2.51610279e-01 1.35425770e+00 -7.65402690e-02 -3.41146886e-01 5.44914268e-02 -1.65224910e-01 -1.51549309e-01 5.28007984e-01 6.58549368e-01 -1.30139887e+00 1.20127022e+00 4.76013660e+00 7.77555823e-01 -9.44337308e-01 -1.85613379e-01 3.52699548e-01 5.27543426e-01 -2.73530930e-01 4.84582752e-01 -1.48134816e+00 7.12368548e-01 7.22249210e-01 -4.06304777e-01 -1.91824883e-02 8.61125767e-01 -3.24860215e-02 -3.37110341e-01 -9.70527470e-01 1.80935085e-01 -2.34783337e-01 -1.17146206e+00 8.45451578e-02 -5.91106080e-02 6.88157558e-01 3.52346227e-02 -4.80596006e-01 5.22466123e-01 8.12463880e-01 -5.28958559e-01 7.36387432e-01 4.20581073e-01 5.44944763e-01 -8.12801540e-01 9.79923069e-01 4.09593612e-01 -1.53191650e+00 6.99897781e-02 -4.54804420e-01 4.34656590e-01 1.34201899e-01 8.24902952e-01 -1.29232895e+00 9.32338297e-01 5.73266625e-01 5.76589942e-01 -4.93029028e-01 1.28279459e+00 -6.45158350e-01 6.81784987e-01 -3.99937481e-01 -3.09907615e-01 8.94139856e-02 -9.41805765e-02 6.63933516e-01 1.41327477e+00 2.90415227e-01 3.76410812e-01 6.16738498e-01 6.75101161e-01 -8.87997448e-02 3.85066122e-01 -1.91970959e-01 4.21823084e-01 1.20030808e+00 1.32040930e+00 -9.11357582e-01 -7.32923031e-01 -3.79359871e-01 5.29603839e-01 6.31159902e-01 3.20525438e-01 -6.69511199e-01 -6.69586241e-01 6.71778679e-01 1.14502124e-01 4.38755989e-01 4.97153290e-02 -3.91621888e-01 -1.37341321e+00 -1.30128160e-01 -4.22634661e-01 8.51179600e-01 -4.55645800e-01 -1.48057246e+00 8.16083610e-01 -1.29222088e-02 -1.38812101e+00 -3.24578643e-01 -3.62038881e-01 -6.06739581e-01 9.40651953e-01 -1.31974292e+00 -9.68461573e-01 -1.76796168e-02 2.58437574e-01 5.78257851e-02 -1.03475362e-01 1.07156575e+00 4.03574049e-01 -6.71294570e-01 7.29122698e-01 1.07567124e-01 8.75202477e-01 9.23174739e-01 -1.77350080e+00 1.88127235e-01 1.00670838e+00 2.11961046e-01 1.03279376e+00 9.02453780e-01 -1.13256979e+00 -4.68158066e-01 -1.32567787e+00 1.60751522e+00 -4.36811805e-01 8.25298548e-01 -5.90082556e-02 -1.17389238e+00 6.79321527e-01 -2.57572904e-02 -2.83973157e-01 7.30124176e-01 7.20389962e-01 -5.96203268e-01 2.18063220e-01 -1.19941080e+00 5.44188261e-01 6.17368639e-01 -2.64279187e-01 -9.39285934e-01 3.59454483e-01 6.38116002e-01 -2.85235494e-01 -1.20399904e+00 5.62743433e-02 4.14681397e-02 -4.43021655e-01 8.19505215e-01 -4.74685192e-01 2.30374038e-01 -7.60190964e-01 9.81520955e-03 -1.78887415e+00 -5.54789245e-01 -1.05766132e-01 2.66074628e-01 1.92739654e+00 1.08551824e+00 -5.92945397e-01 5.90216100e-01 8.88051510e-01 -1.30089685e-01 -3.33110452e-01 -8.58793259e-01 -6.77287996e-01 3.28312181e-02 2.35375300e-01 8.30540121e-01 1.62870395e+00 6.88333511e-01 6.89460397e-01 1.09723791e-01 7.88264573e-01 4.50909317e-01 4.56966400e-01 4.53844458e-01 -1.58286154e+00 -3.76486708e-03 -3.63178551e-01 -5.46447754e-01 -8.68819535e-01 3.91968131e-01 -1.16777611e+00 2.29043007e-01 -1.68248332e+00 7.65107945e-02 -1.28235412e+00 -2.67611355e-01 9.68136787e-01 -7.41318643e-01 -1.82579845e-01 2.11646885e-01 5.75935006e-01 -6.53899789e-01 -1.28568457e-02 5.16471386e-01 -2.95400351e-01 -3.85580450e-01 1.57016426e-01 -5.94578743e-01 7.07980931e-01 6.32946551e-01 -7.23131895e-01 3.15728843e-01 -1.39923155e-01 1.84921041e-01 9.32075381e-02 -1.09724894e-01 -8.89797211e-01 5.84474981e-01 -3.53009731e-01 2.16661230e-01 -5.90404749e-01 -2.47124415e-02 -7.79640377e-01 2.46520758e-01 2.10511938e-01 -4.88061994e-01 -2.20596254e-01 -3.18725824e-01 3.73000056e-01 -2.82927573e-01 -5.53637981e-01 7.80330062e-01 -1.25921756e-01 -9.91499007e-01 1.18966833e-01 -2.00276226e-01 2.08870754e-01 8.95536363e-01 6.93890871e-03 -4.60151553e-01 5.34000285e-02 -1.16163981e+00 2.75358111e-01 6.97022378e-01 2.51994938e-01 1.06239252e-01 -1.36279905e+00 -7.11468637e-01 -3.39510053e-01 3.21519017e-01 3.55782211e-02 -6.92178234e-02 5.00838280e-01 -4.72753718e-02 1.94056571e-01 7.95540139e-02 -3.75879437e-01 -1.32140148e+00 5.54610312e-01 3.36431623e-01 -5.37519217e-01 -3.09957325e-01 6.60623133e-01 9.91050154e-02 -6.71857357e-01 -2.88811401e-02 -1.13208897e-01 -9.11702156e-01 4.48203206e-01 4.16482002e-01 -1.52997062e-01 1.13807492e-01 -1.01498961e+00 -6.81160152e-01 3.58611703e-01 -2.13938594e-01 7.97576979e-02 1.25543416e+00 -2.13294197e-02 -3.67123187e-01 4.46218163e-01 9.21503007e-01 5.81121862e-01 -4.73163515e-01 -3.66585433e-01 7.45753229e-01 -2.79582173e-01 -4.68007565e-01 -9.59556401e-01 -1.03639376e+00 1.62963957e-01 1.27529114e-01 5.78395307e-01 6.24938607e-01 2.32897475e-01 4.24002886e-01 3.76696885e-01 6.35361552e-01 -9.56056952e-01 -7.23121166e-01 9.87253845e-01 1.10269174e-01 -9.47572529e-01 -6.14627823e-02 -7.36766756e-01 -7.63980329e-01 8.38022530e-01 7.45331526e-01 -6.04237951e-02 5.95504940e-01 1.42287239e-01 2.12691441e-01 -1.04290433e-01 -6.51203215e-01 -5.92983782e-01 4.75972027e-01 6.53487325e-01 7.98439622e-01 4.64576393e-01 -7.38813460e-01 7.63874352e-01 -2.64646322e-01 -6.20515764e-01 5.82283795e-01 5.20508707e-01 -7.05409348e-01 -1.26602662e+00 -5.40827990e-01 6.24505401e-01 -4.96276945e-01 -3.68096590e-01 -3.82807523e-01 5.71095765e-01 3.34841996e-01 1.10629559e+00 2.96522398e-02 -5.15003324e-01 3.09707284e-01 1.03611372e-01 -2.18281552e-01 -1.17867684e+00 -8.88510466e-01 -2.80773729e-01 6.18464470e-01 6.56976923e-02 -5.68091094e-01 -7.64959514e-01 -1.81729841e+00 -2.68640488e-01 -7.97124326e-01 9.42985356e-01 3.44070405e-01 1.13953531e+00 3.54927957e-01 1.52787447e-01 7.24580348e-01 -4.71577853e-01 -8.98401905e-03 -8.05803061e-01 -6.35613918e-01 8.63446176e-01 -6.90876991e-02 -9.75697815e-01 -3.68485540e-01 7.06919432e-02]
[9.58000373840332, 9.182506561279297]
0845e47c-37e3-4fea-bc57-30619ea62a33
tag2pix-line-art-colorization-using-text-tag
1908.05840
null
https://arxiv.org/abs/1908.05840v1
https://arxiv.org/pdf/1908.05840v1.pdf
Tag2Pix: Line Art Colorization Using Text Tag With SECat and Changing Loss
Line art colorization is expensive and challenging to automate. A GAN approach is proposed, called Tag2Pix, of line art colorization which takes as input a grayscale line art and color tag information and produces a quality colored image. First, we present the Tag2Pix line art colorization dataset. A generator network is proposed which consists of convolutional layers to transform the input line art, a pre-trained semantic extraction network, and an encoder for input color information. The discriminator is based on an auxiliary classifier GAN to classify the tag information as well as genuineness. In addition, we propose a novel network structure called SECat, which makes the generator properly colorize even small features such as eyes, and also suggest a novel two-step training method where the generator and discriminator first learn the notion of object and shape and then, based on the learned notion, learn colorization, such as where and how to place which color. We present both quantitative and qualitative evaluations which prove the effectiveness of the proposed method.
['Ho Young Jhoo', 'Hyunsu Kim', 'Sungjoo Yoo', 'Eunhyeok Park']
2019-08-16
tag2pix-line-art-colorization-using-text-tag-1
http://openaccess.thecvf.com/content_ICCV_2019/html/Kim_Tag2Pix_Line_Art_Colorization_Using_Text_Tag_With_SECat_and_ICCV_2019_paper.html
http://openaccess.thecvf.com/content_ICCV_2019/papers/Kim_Tag2Pix_Line_Art_Colorization_Using_Text_Tag_With_SECat_and_ICCV_2019_paper.pdf
iccv-2019-10
['line-art-colorization']
['computer-vision']
[ 3.32422942e-01 1.24304570e-01 1.62127867e-01 -3.83470684e-01 -6.66977048e-01 -9.88549531e-01 5.43165028e-01 -3.92033935e-01 -1.83539540e-01 6.29685104e-01 -2.30137169e-01 -1.48496762e-01 3.25251073e-01 -1.11254525e+00 -9.43716586e-01 -6.98249280e-01 7.17820168e-01 5.32601416e-01 1.90853775e-02 1.12732120e-01 2.45385945e-01 4.03908223e-01 -1.17522323e+00 1.95373505e-01 1.04851866e+00 1.39672470e+00 -9.53986570e-02 6.82041168e-01 -3.65236491e-01 8.04369807e-01 -5.78938544e-01 -7.88891077e-01 5.92267394e-01 -1.11792612e+00 -7.20885932e-01 4.44024920e-01 4.57947582e-01 -3.46768320e-01 4.44297621e-04 1.12509525e+00 3.00216794e-01 3.46547598e-03 7.77230322e-01 -1.58745694e+00 -1.10024786e+00 6.71404600e-01 -6.25433207e-01 -5.19742131e-01 2.29118392e-02 2.65339375e-01 8.82608831e-01 -5.79551637e-01 6.02933943e-01 9.67490256e-01 5.66342413e-01 8.03834379e-01 -1.11714470e+00 -6.74998641e-01 -1.10405430e-01 -1.17025651e-01 -9.95596886e-01 -2.49532275e-02 1.16306698e+00 -5.23600817e-01 -1.97939605e-01 1.02339178e-01 1.03188872e+00 1.11078238e+00 -1.25345603e-01 8.67732704e-01 1.42848516e+00 -5.76978147e-01 6.62326574e-01 1.90897912e-01 -2.75900990e-01 9.55061674e-01 2.57958680e-01 -6.89019356e-03 -8.29847381e-02 2.10632503e-01 1.10098124e+00 1.77360922e-01 -4.95881438e-02 -4.77330416e-01 -9.81258452e-01 6.74995124e-01 7.65513062e-01 1.52022153e-01 -3.86766434e-01 5.79683721e-01 -9.62756425e-02 -7.45023489e-02 4.24588859e-01 5.01345217e-01 -5.53598590e-02 -1.33812845e-01 -1.00922787e+00 -1.81532249e-01 5.66588879e-01 9.71013486e-01 1.17250681e+00 8.78978670e-02 -3.62258911e-01 6.66909754e-01 1.05490498e-01 8.11932147e-01 1.02142759e-01 -9.20068622e-01 3.08005184e-01 8.87423754e-01 1.32706389e-01 -6.89731538e-01 3.87267843e-02 -1.61219224e-01 -9.04343843e-01 6.21518672e-01 4.16854709e-01 -3.64688993e-01 -1.30680716e+00 1.68769646e+00 9.87603143e-02 4.47054915e-02 8.43992233e-02 1.05250061e+00 4.27087188e-01 4.18168813e-01 -2.59957053e-02 4.89044935e-01 1.33873177e+00 -1.29947591e+00 -5.14348924e-01 -2.24433586e-01 2.74973750e-01 -7.43780017e-01 1.27235878e+00 3.08397114e-01 -1.20934188e+00 -5.41830003e-01 -1.05732107e+00 -1.62770092e-01 -7.01959908e-01 8.30286384e-01 8.11263323e-01 7.46576071e-01 -1.12365806e+00 4.34955388e-01 -6.77557051e-01 -3.28154266e-01 6.54204071e-01 3.15240733e-02 -2.27829963e-01 -8.73483270e-02 -8.40326071e-01 5.14653921e-01 5.48701286e-01 2.36726463e-01 -7.46717334e-01 -5.15996695e-01 -6.94851816e-01 3.60822938e-02 1.44810647e-01 -9.49660242e-01 8.73796105e-01 -1.78586638e+00 -2.04981160e+00 8.08562219e-01 3.51787925e-01 -2.00422615e-01 8.55636835e-01 -7.80863240e-02 -3.48484695e-01 3.87602746e-01 7.86719695e-02 8.20979476e-01 1.02676344e+00 -1.60662055e+00 -8.77880037e-01 -4.05227914e-02 2.97229558e-01 2.21785698e-02 -1.33803770e-01 -4.49960351e-01 -9.41456258e-01 -8.37288916e-01 -4.55305651e-02 -9.18294489e-01 9.41948146e-02 2.62814105e-01 -9.52260613e-01 1.54623821e-01 6.56746745e-01 -6.04376078e-01 8.83055151e-01 -2.22210264e+00 -3.81162986e-02 6.68905914e-01 2.98973620e-01 -1.03560023e-01 -2.59738922e-01 2.15392843e-01 3.00889974e-03 1.57129124e-01 -5.45276165e-01 -2.72496432e-01 1.29702836e-01 7.83577785e-02 -2.09509239e-01 2.72833705e-01 4.78126794e-01 1.05950439e+00 -8.87593329e-01 -3.11441630e-01 2.57334977e-01 4.95457172e-01 -5.83160996e-01 4.90158617e-01 -4.86872613e-01 4.00358796e-01 -5.40640473e-01 6.27067566e-01 9.45933461e-01 -3.68729383e-01 -1.58824157e-02 -5.54164827e-01 -9.45868194e-02 -3.45266521e-01 -9.32058334e-01 1.76044023e+00 -4.72736806e-01 5.01501262e-01 -2.70119220e-01 -6.79799020e-01 1.18300486e+00 -9.50856283e-02 3.78234118e-01 -7.30649352e-01 5.12618601e-01 2.07726583e-01 -4.94500846e-01 -1.55383110e-01 4.74545121e-01 3.22779007e-02 -3.09402019e-01 7.14768171e-01 -1.15927570e-01 -4.44488972e-01 1.14859581e-01 2.81218946e-01 8.98975015e-01 5.79985261e-01 -3.03536385e-01 4.53482866e-02 4.34246421e-01 4.57572788e-02 2.98056185e-01 3.75245124e-01 4.27350789e-01 9.44631457e-01 8.00257146e-01 -3.76322687e-01 -1.23023105e+00 -9.68223929e-01 6.20894492e-01 7.97066808e-01 4.83780265e-01 -6.15175031e-02 -1.25991714e+00 -1.00185239e+00 -7.46200308e-02 8.34394574e-01 -1.08991337e+00 -2.79046208e-01 -3.53949249e-01 -3.05199116e-01 4.13489014e-01 8.15276206e-01 1.00035644e+00 -9.79999661e-01 -3.83193612e-01 -2.25157708e-01 6.73148483e-02 -1.04658508e+00 -8.51526618e-01 1.88862309e-01 -4.40065742e-01 -1.28311038e+00 -7.07295418e-01 -7.85960138e-01 1.17945552e+00 -9.95534733e-02 1.03236854e+00 1.07235037e-01 -7.63976425e-02 3.17747444e-01 -4.88105267e-01 -2.13522986e-01 -3.87967259e-01 2.13519651e-02 -6.10984623e-01 6.81361258e-01 1.65244527e-02 -2.83568978e-01 -9.12496567e-01 1.00563299e-02 -1.08337808e+00 6.12230837e-01 1.04158354e+00 7.88281679e-01 7.97497630e-01 -1.39358357e-01 4.30894867e-02 -1.41087341e+00 3.39208812e-01 -8.88306275e-03 -8.00008357e-01 6.23104393e-01 -6.20557785e-01 3.35900486e-01 7.78109610e-01 -1.61472872e-01 -1.02504551e+00 4.41376150e-01 1.09389849e-01 -5.60970008e-01 -7.06239138e-03 2.65502892e-02 -3.88558865e-01 -1.43306702e-01 4.70336556e-01 2.31569842e-01 -1.57825500e-01 -4.31213975e-01 9.99097705e-01 3.67878795e-01 8.68899286e-01 -6.25446558e-01 1.18470716e+00 6.39017224e-01 -2.24744305e-01 -2.28341565e-01 -6.64336503e-01 1.22443639e-01 -7.37859011e-01 -1.61667332e-01 1.24782741e+00 -5.34346521e-01 -6.36463881e-01 5.84773064e-01 -9.80858862e-01 -7.94548631e-01 -6.00087643e-01 1.01796784e-01 -7.14716077e-01 -1.37720898e-01 -4.54818994e-01 -6.83126330e-01 -4.08747494e-01 -8.22675288e-01 1.23894060e+00 5.50101221e-01 4.97360021e-01 -1.03226078e+00 -1.81506455e-01 1.15438864e-01 3.86560768e-01 6.68773711e-01 9.78638709e-01 -2.98634082e-01 -7.20257044e-01 -2.52215952e-01 -7.09269106e-01 5.59299886e-01 3.09156746e-01 1.99732170e-01 -8.18686128e-01 -5.23720160e-02 -5.48336446e-01 -3.07938516e-01 9.08439279e-01 1.05216047e-02 1.40390098e+00 -3.91944528e-01 -1.61535814e-01 9.76917624e-01 1.78020620e+00 3.62230003e-01 7.56196678e-01 2.07131714e-01 1.09573054e+00 1.47442639e-01 1.85681477e-01 2.57192075e-01 3.99274856e-01 3.85471404e-01 4.00197923e-01 -8.35880160e-01 -5.64646900e-01 -7.20422864e-01 1.76775753e-01 4.79183823e-01 -1.73755944e-01 -3.54125082e-01 -4.11111027e-01 1.64073721e-01 -1.50022507e+00 -6.64568245e-01 3.29249114e-01 1.87994969e+00 8.67148638e-01 -5.66678010e-02 1.61618531e-01 8.28103349e-02 8.29532266e-01 -2.54583001e-01 -6.73627257e-01 -1.96384534e-01 -2.39912853e-01 4.28262711e-01 6.88681543e-01 3.10233057e-01 -9.66689765e-01 1.27366173e+00 5.23831272e+00 6.86937928e-01 -1.65139723e+00 -1.19891716e-02 8.50881577e-01 5.96732676e-01 -6.89189792e-01 1.37045383e-01 -1.79252252e-01 5.92482328e-01 9.86427143e-02 1.91171974e-01 8.29040289e-01 8.42930377e-01 -1.12224311e-01 4.38737171e-03 -1.08640826e+00 1.05230641e+00 3.03654224e-01 -1.27847219e+00 3.85854512e-01 -2.46824235e-01 9.20301974e-01 -6.40607715e-01 2.09201962e-01 -2.27318015e-02 6.22363925e-01 -6.89876437e-01 1.16791677e+00 9.96572614e-01 1.32597220e+00 -7.06391275e-01 4.99513775e-01 -2.71231711e-01 -1.15283954e+00 8.52277949e-02 -1.25422448e-01 5.03094018e-01 -2.27852985e-01 3.08008194e-01 -5.77450991e-01 7.37117350e-01 3.26504737e-01 7.08060563e-01 -9.04026151e-01 9.88918006e-01 -7.57020414e-01 6.07752502e-01 -9.29900855e-02 5.61129302e-02 2.92623490e-01 -6.53002679e-01 -2.72065639e-01 1.01883137e+00 4.67400253e-01 -8.58263224e-02 1.47096589e-01 1.30963981e+00 -3.77135932e-01 -1.45686522e-01 -1.56955421e-01 -4.39390779e-01 3.70774627e-01 1.58354104e+00 -1.18610048e+00 -4.90016997e-01 -1.42454490e-01 1.68426704e+00 2.07616165e-01 7.41685271e-01 -9.44797635e-01 -7.38241255e-01 1.58467263e-01 -8.08746740e-02 4.85704124e-01 4.61537391e-02 -5.08922696e-01 -1.18556857e+00 -1.36112437e-01 -4.02726024e-01 2.42611855e-01 -1.21964121e+00 -1.30610168e+00 7.94081151e-01 -4.57045257e-01 -1.34997237e+00 -1.64754838e-01 -7.20686913e-01 -8.91062975e-01 7.32964754e-01 -1.42986155e+00 -1.81053865e+00 -7.90248990e-01 6.94300711e-01 4.82178703e-02 -1.42637625e-01 5.56961715e-01 2.06186801e-01 -5.65223157e-01 8.64027560e-01 6.99567720e-02 6.18007004e-01 5.75062513e-01 -1.61154056e+00 2.54749656e-01 8.39042068e-01 1.54492155e-01 2.15105832e-01 2.46664301e-01 -4.77782667e-01 -1.52968240e+00 -1.50067592e+00 2.46740907e-01 -2.40624830e-01 3.97733361e-01 -6.08106196e-01 -3.32219779e-01 5.51592171e-01 1.43411592e-01 1.44999921e-01 3.15094054e-01 -5.37152290e-01 -4.96975958e-01 -4.66988832e-01 -1.20198750e+00 4.55367088e-01 8.26443136e-01 -4.66638863e-01 -2.51724660e-01 1.64896414e-01 5.75491726e-01 -3.89442652e-01 -4.09288436e-01 -1.05733104e-01 6.23601615e-01 -9.59738970e-01 6.07448936e-01 -2.82397062e-01 6.58535004e-01 -5.09751558e-01 1.08732142e-01 -1.37564826e+00 -2.15774208e-01 -4.51495588e-01 3.29167366e-01 1.43037033e+00 4.91082698e-01 -4.70953345e-01 9.28941727e-01 5.67076564e-01 -2.62143165e-01 -5.24399459e-01 -3.80332828e-01 -5.87975562e-01 -1.24809816e-01 -1.11560091e-01 9.35820639e-01 9.17930484e-01 -6.74718976e-01 2.65397817e-01 -3.52471858e-01 -6.69001937e-02 6.02659762e-01 5.53904951e-01 8.23946893e-01 -9.86316741e-01 -1.60383075e-01 -5.21473885e-01 -3.42027724e-01 -7.13713109e-01 -7.46277049e-02 -8.95378947e-01 1.09590165e-01 -1.69644988e+00 6.99988008e-02 -5.56497812e-01 -2.78224021e-01 7.87136793e-01 -2.02164546e-01 7.77324021e-01 2.16389626e-01 -5.35443351e-02 -6.00184798e-01 6.35480165e-01 1.47087836e+00 -3.94998044e-01 -1.54465735e-01 -1.36924833e-01 -8.62971544e-01 4.96851236e-01 6.64892614e-01 -6.37011975e-02 -3.59481066e-01 -5.78919530e-01 2.95966893e-01 -1.98533237e-01 5.71500242e-01 -1.09490407e+00 1.14380039e-01 -1.62723869e-01 8.50373387e-01 -2.16776267e-01 1.74115673e-01 -1.05651140e+00 1.56389803e-01 3.50940168e-01 -3.71064007e-01 -8.33686963e-02 -1.64472193e-01 3.86714131e-01 -8.37533101e-02 1.41371593e-01 9.47122276e-01 5.41238226e-02 -6.73025727e-01 5.13361037e-01 3.62397939e-01 6.44489098e-03 1.06177759e+00 -3.35672855e-01 -2.86274642e-01 -3.75448674e-01 -5.16265929e-01 4.50773444e-03 7.93741882e-01 2.99202293e-01 4.31096524e-01 -1.73085093e+00 -4.91122872e-01 4.41531122e-01 2.72111863e-01 -1.64023370e-01 7.54330978e-02 3.14553261e-01 -8.31059992e-01 -1.08149774e-01 -5.40846109e-01 -2.10387871e-01 -5.24816930e-01 6.70330048e-01 3.10890138e-01 2.36378133e-01 -6.05819643e-01 6.76451623e-01 3.18878770e-01 6.02109218e-03 -1.37589440e-01 -4.98343766e-01 -6.56240480e-03 -8.58343393e-02 1.86667472e-01 7.21499249e-02 -1.77613333e-01 -3.28930497e-01 -3.79339755e-02 9.12043512e-01 3.73445809e-01 -3.59100044e-01 1.16978824e+00 3.92939476e-03 -3.14414829e-01 2.44862169e-01 1.07526290e+00 7.68353492e-02 -1.60608554e+00 -3.28731239e-02 -1.81014165e-01 -5.07881403e-01 -1.67747587e-01 -1.19542825e+00 -1.78308988e+00 8.69151175e-01 7.48127997e-01 2.19031364e-01 1.47693717e+00 1.12257162e-02 8.82708549e-01 -1.17415851e-02 2.05353960e-01 -1.22040284e+00 3.24961066e-01 1.00224182e-01 7.36876249e-01 -9.46386218e-01 -4.57246631e-01 -3.11546028e-01 -7.56948471e-01 1.37165618e+00 7.15931356e-01 -3.76852065e-01 3.23063642e-01 1.70863166e-01 2.55530238e-01 -2.12472692e-01 -1.17409401e-01 -3.96713257e-01 4.21012133e-01 5.37114084e-01 1.56638011e-01 2.33750343e-01 -1.81541219e-01 7.75020480e-01 -5.00569582e-01 6.34919927e-02 2.38136873e-01 5.43225288e-01 -7.35929757e-02 -1.05834401e+00 -1.32021770e-01 3.07047874e-01 9.89782345e-03 2.53419057e-02 -7.63467729e-01 5.93804598e-01 5.80623746e-01 5.12994766e-01 4.19397563e-01 -7.34024465e-01 3.01756680e-01 -9.78084654e-02 3.91407311e-01 -3.20591718e-01 -4.10941482e-01 -8.35475922e-02 -2.81474769e-01 -6.53795600e-01 -3.51058036e-01 -1.25293515e-03 -1.16560030e+00 -1.76191553e-01 -1.26696795e-01 3.34989399e-01 9.14876640e-01 7.86788940e-01 2.25920647e-01 6.98174357e-01 9.81271148e-01 -6.22800946e-01 4.42779623e-02 -8.00463796e-01 -7.75208712e-01 7.76615441e-01 1.39741972e-01 -4.23773319e-01 -2.06963763e-01 4.65653718e-01]
[11.487502098083496, -0.8639295101165771]
df5f04ce-25e9-4f7a-9992-ea80312b0998
kronecker-decomposition-for-knowledge-graph
2205.06560
null
https://arxiv.org/abs/2205.06560v1
https://arxiv.org/pdf/2205.06560v1.pdf
Kronecker Decomposition for Knowledge Graph Embeddings
Knowledge graph embedding research has mainly focused on learning continuous representations of entities and relations tailored towards the link prediction problem. Recent results indicate an ever increasing predictive ability of current approaches on benchmark datasets. However, this effectiveness often comes with the cost of over-parameterization and increased computationally complexity. The former induces extensive hyperparameter optimization to mitigate malicious overfitting. The latter magnifies the importance of winning the hardware lottery. Here, we investigate a remedy for the first problem. We propose a technique based on Kronecker decomposition to reduce the number of parameters in a knowledge graph embedding model, while retaining its expressiveness. Through Kronecker decomposition, large embedding matrices are split into smaller embedding matrices during the training process. Hence, embeddings of knowledge graphs are not plainly retrieved but reconstructed on the fly. The decomposition ensures that elementwise interactions between three embedding vectors are extended with interactions within each embedding vector. This implicitly reduces redundancy in embedding vectors and encourages feature reuse. To quantify the impact of applying Kronecker decomposition on embedding matrices, we conduct a series of experiments on benchmark datasets. Our experiments suggest that applying Kronecker decomposition on embedding matrices leads to an improved parameter efficiency on all benchmark datasets. Moreover, empirical evidence suggests that reconstructed embeddings entail robustness against noise in the input knowledge graph. To foster reproducible research, we provide an open-source implementation of our approach, including training and evaluation scripts as well as pre-trained models in our knowledge graph embedding framework (https://github.com/dice-group/dice-embeddings).
['Axel-Cyrille Ngonga Ngomo', 'Julian Lienen', 'Caglar Demir']
2022-05-13
null
null
null
null
['knowledge-graph-embeddings', 'knowledge-graph-embeddings']
['graphs', 'methodology']
[-7.92744979e-02 2.40450457e-01 -2.70686746e-01 5.14527969e-02 -1.23531498e-01 -6.92147136e-01 6.01874352e-01 3.10498506e-01 -4.05527145e-01 4.55328554e-01 2.57913709e-01 -4.75952923e-01 -4.05741036e-01 -1.04966688e+00 -6.10630870e-01 -5.30206621e-01 -3.14665854e-01 1.03638679e-01 8.17634761e-02 -1.08114794e-01 1.60432503e-01 3.98427725e-01 -1.20483351e+00 -1.37501448e-01 6.22078061e-01 7.07703412e-01 -2.07140282e-01 4.99418437e-01 -1.26483003e-02 5.80385923e-01 -3.80030423e-01 -1.03355265e+00 3.49102169e-01 2.73321159e-02 -8.05115163e-01 -1.19041800e-01 -2.75884252e-02 -2.12882087e-01 -7.19905317e-01 9.43237185e-01 4.20650691e-01 7.24183097e-02 6.02926195e-01 -1.41455567e+00 -9.59434271e-01 5.80389380e-01 -2.66759545e-01 2.14204583e-02 2.07819238e-01 1.53482780e-01 1.50564146e+00 -9.08914864e-01 7.32430577e-01 7.27835536e-01 6.88370526e-01 1.00878984e-01 -1.53595400e+00 -3.78933549e-01 -1.38217241e-01 2.81471401e-01 -1.63103712e+00 -3.67342651e-01 7.82597125e-01 -4.36051786e-01 1.17941201e+00 4.09940451e-01 7.58340955e-01 9.27898586e-01 -1.09264694e-01 3.25531304e-01 4.17401552e-01 -3.93908799e-01 1.88063979e-01 4.70380366e-01 2.70860583e-01 8.49684000e-01 7.95191705e-01 -9.06580966e-03 -3.22600812e-01 -4.95238364e-01 4.44048673e-01 -5.59386052e-03 -4.30611402e-01 -9.37451363e-01 -1.02168989e+00 9.54199314e-01 5.70495844e-01 2.70464212e-01 -2.71058559e-01 3.03223372e-01 5.01248538e-01 4.68145549e-01 2.42208079e-01 5.89965641e-01 -3.02493244e-01 -1.46918997e-01 -3.81426901e-01 -1.07891329e-01 1.10034347e+00 7.53913760e-01 8.80588472e-01 -2.94272125e-01 1.98824733e-01 8.19981575e-01 2.24341109e-01 1.42964631e-01 3.02186042e-01 -6.46057904e-01 5.01295567e-01 1.04338920e+00 -1.18756719e-01 -1.56047690e+00 -1.41530246e-01 -4.15223897e-01 -6.63879395e-01 -2.19007105e-01 2.87479520e-01 9.64134857e-02 -4.30645764e-01 1.58012247e+00 4.59618688e-01 1.10802382e-01 1.22694470e-01 7.30768800e-01 3.97945255e-01 5.16087294e-01 -1.42886704e-02 2.08237261e-01 1.49516487e+00 -8.87345076e-01 -4.42756295e-01 2.10243687e-02 1.10185170e+00 -5.86355686e-01 1.03955019e+00 4.30103503e-02 -7.29183853e-01 -1.00209460e-01 -1.15840042e+00 -1.49176553e-01 -7.56461978e-01 2.26558462e-01 7.88841367e-01 7.64499366e-01 -8.97322536e-01 7.00411081e-01 -7.15498865e-01 -3.05333227e-01 3.18869263e-01 4.81914192e-01 -7.79737473e-01 -8.57498422e-02 -1.27221608e+00 9.11732256e-01 4.54375714e-01 6.22099526e-02 -2.42975101e-01 -7.83686757e-01 -8.77135634e-01 4.89085436e-01 4.18782979e-01 -7.65248179e-01 4.36123848e-01 -4.78551954e-01 -1.27940726e+00 5.07855773e-01 2.56453395e-01 -3.87496561e-01 2.05587521e-01 -7.07935914e-02 -4.20149267e-01 1.72270536e-01 -2.78018177e-01 2.23165870e-01 7.60736942e-01 -1.04475415e+00 5.67463040e-02 -2.16901034e-01 2.76868671e-01 -1.46205723e-01 -1.04468656e+00 -5.18062115e-01 -5.38774133e-01 -4.96201336e-01 -1.73097059e-01 -1.00539231e+00 5.18589579e-02 -2.59599593e-02 -4.24232036e-01 -1.20571554e-01 5.02958536e-01 -4.44541633e-01 1.64159966e+00 -2.31982374e+00 1.88744128e-01 6.12784743e-01 4.36234057e-01 4.73264098e-01 -3.19676697e-01 9.60988104e-01 -2.46984735e-01 3.05668086e-01 -8.35065916e-02 -1.42612681e-01 2.71099299e-01 3.11184883e-01 -3.77760261e-01 5.41460395e-01 2.73194432e-01 9.92729247e-01 -8.50402057e-01 -3.46394241e-01 1.44832358e-01 8.35541546e-01 -8.02059650e-01 -2.91465390e-02 4.98426333e-02 -3.75104338e-01 -3.58395308e-01 3.60679507e-01 4.96877909e-01 -6.57802045e-01 7.47533560e-01 -5.82309663e-01 2.21806288e-01 3.71840894e-01 -1.36019194e+00 1.33943176e+00 -5.17949283e-01 3.71132046e-01 -2.75653809e-01 -1.15720534e+00 7.38939464e-01 2.17115864e-01 3.97522658e-01 -4.31666225e-01 1.78399459e-01 1.28993884e-01 8.08475539e-03 -3.87611032e-01 5.99132180e-01 1.94458842e-01 1.83609322e-01 6.24957085e-01 8.59322622e-02 3.22172403e-01 2.82003790e-01 5.95641017e-01 1.35127449e+00 -1.41710073e-01 2.65836060e-01 -8.84037092e-02 4.36582565e-01 -1.38900384e-01 2.26768836e-01 2.50315279e-01 5.97931296e-02 9.11037922e-02 7.86053360e-01 -3.98450434e-01 -1.05224752e+00 -9.31817055e-01 1.76458675e-02 8.69080365e-01 5.90478396e-03 -1.05343568e+00 -4.97030258e-01 -7.59839594e-01 3.69078308e-01 6.22206628e-01 -7.84222484e-01 -4.94768649e-01 -3.32094759e-01 -6.29535794e-01 7.49904633e-01 4.56881016e-01 -3.36134504e-03 -5.50890028e-01 -3.97924930e-01 4.91231792e-02 9.56896394e-02 -8.98715019e-01 -4.79567498e-01 1.44563988e-01 -7.75692523e-01 -1.25712860e+00 -2.35266954e-01 -5.74140787e-01 9.51720774e-01 3.37322950e-01 9.13970768e-01 4.66885775e-01 -4.17027712e-01 5.52358270e-01 -3.85785848e-01 1.58791423e-01 -1.54128253e-01 2.51585424e-01 -6.80050552e-02 -1.22558521e-02 4.18631881e-01 -8.21835816e-01 -6.78288221e-01 2.30472982e-01 -1.05307329e+00 -2.21173048e-01 5.40031016e-01 1.12929702e+00 3.14493537e-01 1.45674840e-01 2.60833263e-01 -8.19143474e-01 7.94136882e-01 -6.62976801e-01 -7.07946658e-01 2.75820673e-01 -9.61047769e-01 4.75970596e-01 6.56776369e-01 -2.94136137e-01 -3.79989684e-01 -1.51397035e-01 2.32309848e-01 -4.98869270e-01 5.01263440e-01 8.42709422e-01 5.74678108e-02 -1.77209914e-01 4.97971267e-01 1.83042243e-01 2.31549472e-01 -3.20226610e-01 7.92474091e-01 4.93872106e-01 3.94393541e-02 -5.08032799e-01 9.40510213e-01 3.89369041e-01 -7.15297367e-03 -7.07443535e-01 -2.76455134e-01 -3.54950935e-01 -3.96697253e-01 1.14327729e-01 3.36796522e-01 -6.93883717e-01 -9.43415701e-01 -3.34657818e-01 -8.75826240e-01 -4.03343812e-02 -2.20833004e-01 4.58947062e-01 -1.61307305e-01 6.55930161e-01 -4.33361232e-01 -5.51058948e-01 -3.40363055e-01 -9.83057261e-01 6.11082137e-01 -2.18583107e-01 -4.95256186e-01 -1.21237516e+00 2.65694022e-01 3.35001588e-01 3.04979593e-01 -7.80484155e-02 1.26825845e+00 -6.71773672e-01 -6.72443330e-01 -4.80051488e-01 -3.12353402e-01 3.43946755e-01 4.90074083e-02 1.40588224e-01 -7.39484429e-01 -3.16910237e-01 -5.04176974e-01 -3.90583500e-02 7.46215761e-01 -3.13864619e-01 8.78563464e-01 -5.31327009e-01 -3.48321825e-01 5.89395463e-01 1.55008554e+00 -3.17245573e-01 5.64669073e-01 4.57747042e-01 7.35656857e-01 5.08278251e-01 1.77845255e-01 4.82398391e-01 5.64599037e-01 5.88648319e-01 3.49229634e-01 2.46176198e-01 5.85564077e-02 -5.02419412e-01 3.11732739e-01 9.45376813e-01 -1.54577002e-01 -2.39446476e-01 -8.63013506e-01 7.55354047e-01 -1.66228735e+00 -9.28389251e-01 5.41444607e-02 2.32986355e+00 8.72529864e-01 5.63656092e-02 3.89742441e-02 3.26745629e-01 3.41566503e-01 1.79584250e-01 -2.09900096e-01 -4.41398978e-01 1.19318724e-01 1.91723615e-01 6.33331716e-01 5.71804821e-01 -7.41631687e-01 7.84108043e-01 5.09886074e+00 6.57700062e-01 -1.06162584e+00 7.77565464e-02 3.32689546e-02 -1.52610779e-01 -6.46652102e-01 2.86268651e-01 -3.80014122e-01 3.90833050e-01 9.24885511e-01 -5.00815034e-01 7.06207156e-01 7.94783771e-01 -3.48258942e-01 1.73881218e-01 -1.15152872e+00 8.60392392e-01 -2.01453343e-01 -1.48063803e+00 3.05024032e-02 3.82835984e-01 3.66307199e-01 -4.54197749e-02 3.65819149e-02 3.46179485e-01 1.85888574e-01 -9.48324263e-01 2.16480479e-01 2.39244714e-01 5.87212622e-01 -7.82255828e-01 6.78490460e-01 -1.14920072e-01 -1.20705736e+00 -3.84874037e-03 -5.50439477e-01 1.05212824e-02 -5.87287173e-02 7.92699039e-01 -1.09930110e+00 7.56809056e-01 2.46444747e-01 6.53590918e-01 -6.64366186e-01 6.14928484e-01 -2.87571222e-01 5.26442945e-01 -3.99218172e-01 1.16473138e-02 8.52356106e-02 -4.36870247e-01 3.73385221e-01 1.12408888e+00 1.48094550e-01 -1.32220149e-01 -2.70205379e-01 8.37400496e-01 -2.70876080e-01 2.48881012e-01 -8.36947262e-01 -6.83460653e-01 7.62287855e-01 1.39206040e+00 -5.62190115e-01 -1.25488967e-01 -5.66513896e-01 9.28204417e-01 7.34837890e-01 3.97771060e-01 -9.32054281e-01 -6.17551267e-01 8.81180406e-01 1.61547244e-01 6.33198023e-01 -3.42567652e-01 -1.76809430e-01 -1.18454063e+00 2.88470834e-01 -5.84011734e-01 4.36846137e-01 -2.91159511e-01 -1.19474566e+00 3.53484720e-01 -2.62018085e-01 -9.48559642e-01 -1.38162272e-02 -5.38893938e-01 -3.66311938e-01 5.85725129e-01 -1.44232380e+00 -9.68439698e-01 -1.51133597e-01 3.94559801e-01 -3.87066811e-01 -9.97014623e-03 9.68945324e-01 6.06380820e-01 -8.22323799e-01 9.88306463e-01 1.74561813e-01 2.57630885e-01 5.46730161e-01 -1.06694567e+00 2.58877665e-01 6.39128625e-01 3.65303844e-01 1.17951655e+00 5.34742713e-01 -4.87188041e-01 -1.90785241e+00 -9.96138811e-01 1.05795944e+00 -5.10164857e-01 1.13599718e+00 -4.11759049e-01 -9.99354184e-01 8.01392794e-01 -9.96806473e-02 1.85712606e-01 9.85792398e-01 5.03421247e-01 -7.46768653e-01 -9.56918746e-02 -8.18903327e-01 7.67318010e-01 1.02359998e+00 -8.54061782e-01 -4.35369343e-01 2.12184295e-01 5.85579634e-01 9.28841606e-02 -1.51340032e+00 1.92486674e-01 7.00360596e-01 -7.19073951e-01 1.12758350e+00 -5.71839571e-01 3.99434447e-01 -3.39672118e-01 -1.91040918e-01 -1.17339838e+00 -4.11426902e-01 -5.33113122e-01 -5.83140910e-01 1.15597379e+00 4.89733785e-01 -9.54769492e-01 7.62639880e-01 7.34506905e-01 3.33603054e-01 -1.03197241e+00 -7.34610200e-01 -9.05595958e-01 -1.62920028e-01 -1.71086088e-01 6.92022979e-01 1.21823490e+00 4.83604431e-01 4.81875032e-01 -2.97757268e-01 3.38210136e-01 4.98528898e-01 1.42329633e-01 1.00408185e+00 -9.75306332e-01 -3.73581201e-01 -4.12780821e-01 -7.76258767e-01 -6.15572810e-01 1.71484485e-01 -1.32497764e+00 -7.10147381e-01 -1.36827564e+00 1.22560091e-01 -4.50964242e-01 -4.11189228e-01 7.06639171e-01 -1.35808125e-01 1.62235528e-01 1.60360396e-01 2.63947159e-01 -3.92362028e-01 7.83417881e-01 8.36370230e-01 1.71729624e-02 -8.38576779e-02 -5.10002673e-01 -7.89611638e-01 3.59069914e-01 8.30719173e-01 -5.23316443e-01 -6.37066722e-01 -3.58515382e-01 5.88559747e-01 -3.96098137e-01 5.51628053e-01 -5.62869370e-01 2.66174763e-01 1.85734898e-01 -6.49209023e-02 -1.31433625e-02 3.32450449e-01 -9.61338878e-01 4.68636483e-01 4.25002843e-01 -3.82964164e-02 -5.72635904e-02 6.98615834e-02 6.77569807e-01 -1.59886852e-01 -2.51723498e-01 4.77440804e-01 4.13337618e-01 -5.42802274e-01 2.54968882e-01 1.14320278e-01 -1.07079916e-01 1.01913822e+00 -2.61611938e-01 -4.34708089e-01 -4.77986187e-02 -6.62224829e-01 1.20449275e-01 6.13461196e-01 3.42274904e-01 3.97244692e-01 -1.31889331e+00 -1.71817735e-01 2.72174895e-01 3.27540100e-01 -4.97014105e-01 8.16081688e-02 1.09855115e+00 -4.54941571e-01 4.75998521e-01 4.25587669e-02 -4.30116951e-02 -1.18043923e+00 7.62667060e-01 4.90945503e-02 -4.24799114e-01 -6.90977216e-01 6.83909357e-01 -1.65097043e-01 -4.21811610e-01 1.62029669e-01 -9.47908983e-02 9.19668563e-03 2.19999447e-01 2.25462630e-01 5.64937234e-01 2.65407562e-02 -2.75211543e-01 -5.44439077e-01 3.43821138e-01 -1.91594541e-01 2.28750288e-01 1.51824641e+00 -5.39981935e-04 -2.11712837e-01 4.97780405e-02 1.50980496e+00 3.04606736e-01 -8.47311676e-01 -2.18880519e-01 1.07637107e-01 -5.26446283e-01 1.29766434e-01 -4.32267606e-01 -1.11919796e+00 7.31903315e-01 1.49401739e-01 4.22222793e-01 8.25772047e-01 -6.46298826e-02 6.87221348e-01 6.13101423e-01 2.52082974e-01 -1.05805075e+00 -1.58338863e-02 2.67963231e-01 6.96184933e-01 -8.97664905e-01 3.20054144e-01 -6.00560188e-01 -5.97097576e-01 1.02446425e+00 2.63412267e-01 -2.09947452e-01 8.76235485e-01 -4.58377898e-02 -2.78126270e-01 -4.67892170e-01 -8.63720894e-01 -9.20189172e-02 9.18017700e-02 3.27171415e-01 3.95816177e-01 1.25825852e-01 -4.29299176e-01 6.73231184e-01 -2.22536787e-01 -3.41110677e-02 3.85317355e-01 8.54297876e-01 -9.16359797e-02 -1.40259182e+00 3.09321657e-02 3.53232831e-01 -1.07998021e-01 -2.01030269e-01 -3.71265292e-01 8.66713464e-01 -1.50196895e-01 6.28887653e-01 -1.48373932e-01 -6.19703948e-01 4.42715049e-01 1.74590185e-01 3.82460058e-01 -4.31677252e-01 -4.43408459e-01 -4.58077848e-01 3.30114514e-01 -5.57826996e-01 3.70181836e-02 -2.83818901e-01 -1.18013644e+00 -6.36580110e-01 -5.20808637e-01 3.34608465e-01 5.88877797e-01 4.50221390e-01 1.00471687e+00 4.42613333e-01 4.74755257e-01 -3.62334907e-01 -8.70144188e-01 -5.86272418e-01 -5.63159406e-01 5.59930801e-01 -4.92755473e-02 -8.12002838e-01 -5.17351687e-01 -1.04426332e-01]
[8.64687728881836, 7.7732439041137695]
8ab7e956-9674-479e-b7d9-66b9a450886f
imprecise-bayesian-neural-networks
2302.09656
null
https://arxiv.org/abs/2302.09656v2
https://arxiv.org/pdf/2302.09656v2.pdf
Imprecise Bayesian Neural Networks
Uncertainty quantification and robustness to distribution shifts are important goals in machine learning and artificial intelligence. Although Bayesian neural networks (BNNs) allow for uncertainty in the predictions to be assessed, different sources of uncertainty are indistinguishable. We present imprecise Bayesian neural networks (IBNNs); they generalize and overcome some of the drawbacks of standard BNNs. These latter are trained using a single prior and likelihood distributions, whereas IBNNs are trained using credal prior and likelihood sets. They allow to distinguish between aleatoric and epistemic uncertainties, and to quantify them. In addition, IBNNs are robust in the sense of Bayesian sensitivity analysis, and are more robust than BNNs to distribution shift. They can also be used to compute sets of outcomes that enjoy PAC-like properties. We apply IBNNs to two case studies. One, to model blood glucose and insulin dynamics for artificial pancreas control, and two, for motion prediction in autonomous driving scenarios. We show that IBNNs performs better when compared to an ensemble of BNNs benchmark.
['Insup Lee', 'Oleg Sokolsky', 'Radoslav Ivanov', 'Vivian Lin', 'Kuk Jin Jang', 'Souradeep Dutta', 'Michele Caprio']
2023-02-19
null
null
null
null
['motion-prediction']
['computer-vision']
[ 1.94659859e-01 5.99145234e-01 -1.63054168e-01 -6.81272805e-01 -5.21168053e-01 -4.29377556e-01 9.54416990e-01 2.90655136e-01 -5.18698394e-01 1.28533554e+00 5.02911722e-03 -3.07941645e-01 -8.56832504e-01 -9.72618222e-01 -9.71404910e-01 -8.93970072e-01 -2.63529629e-01 7.56553352e-01 3.15424532e-01 2.02120915e-01 -1.17515638e-01 4.66962695e-01 -1.36969793e+00 -3.26568663e-01 7.97382355e-01 1.38402689e+00 -5.11317968e-01 3.67091775e-01 2.06704170e-01 8.50215793e-01 -4.25481856e-01 -2.82415628e-01 1.39175743e-01 5.12802936e-02 -3.77624869e-01 -8.07283282e-01 -4.76120971e-02 -3.81171286e-01 -2.62779534e-01 1.36392260e+00 3.67649138e-01 2.22863033e-01 1.25968587e+00 -1.55059755e+00 -1.31984934e-01 1.10564864e+00 4.37625311e-02 -3.03354170e-02 -1.16224572e-01 1.32890865e-01 4.43647265e-01 -2.05677971e-01 1.03014827e-01 1.40818894e+00 9.55087483e-01 6.45569980e-01 -1.20208168e+00 -3.98022085e-01 -5.31473197e-02 -6.04763255e-02 -1.18546867e+00 -4.84981656e-01 2.11757496e-01 -6.52716756e-01 3.58549297e-01 7.88234025e-02 4.45850581e-01 1.52621615e+00 6.91883564e-01 6.12377167e-01 1.15392649e+00 -2.99771070e-01 1.00357354e+00 1.05334312e-01 2.94968486e-01 2.40055263e-01 8.18831325e-01 9.75502431e-01 -5.12035906e-01 -3.31957638e-01 6.37346268e-01 -1.89715639e-01 -2.86242872e-01 -6.24730110e-01 -1.04312789e+00 9.93570924e-01 1.44260794e-01 -1.94757313e-01 -3.22134703e-01 7.34584212e-01 3.01552206e-01 -1.54656306e-01 2.32537150e-01 2.80624151e-01 -4.89603788e-01 -1.12755813e-01 -5.75251341e-01 5.37106574e-01 1.23930550e+00 9.06009495e-01 2.21393242e-01 6.43988550e-02 -2.95243710e-01 4.47805822e-01 9.15988743e-01 6.79704130e-01 2.57762939e-01 -1.33033693e+00 1.60711855e-02 -1.53114080e-01 5.00307441e-01 -7.16189981e-01 -6.96668923e-01 -1.92670077e-01 -9.86992180e-01 5.84680915e-01 7.96761632e-01 -3.86558652e-01 -1.24376690e+00 1.92287731e+00 -1.55555420e-02 1.81214914e-01 4.48645562e-01 6.97330296e-01 6.28685713e-01 4.88510877e-01 2.34081775e-01 -7.50059262e-02 1.12874615e+00 -9.02819261e-02 -8.40780973e-01 -1.45990819e-01 1.42590344e-01 7.56984651e-02 1.84543848e-01 7.10668027e-01 -1.08945429e+00 -8.68905708e-02 -1.34883881e+00 4.54376101e-01 -5.12008429e-01 -5.89733481e-01 5.55330038e-01 1.16913700e+00 -7.98232377e-01 1.01850402e+00 -1.22935271e+00 -6.23192452e-02 3.29446882e-01 2.49267906e-01 9.84316170e-02 1.15826286e-01 -1.76770020e+00 1.35508263e+00 9.00488317e-01 4.56015408e-01 -1.18089914e+00 -7.15816855e-01 -1.09303021e+00 -8.83937553e-02 3.28662902e-01 -6.22587025e-01 1.59360194e+00 -5.20625770e-01 -1.82779145e+00 6.98855370e-02 5.98353446e-01 -9.76615429e-01 8.17200840e-01 -4.30577807e-02 -5.15866041e-01 -1.18558317e-01 -3.00799668e-01 8.01769316e-01 5.90631604e-01 -1.12244391e+00 -5.49280465e-01 -1.92374244e-01 -3.67867872e-02 -2.17584178e-01 3.37241232e-01 -3.25900257e-01 -1.50825968e-02 -3.90838861e-01 3.42680812e-01 -8.42269003e-01 -3.91571701e-01 6.98210746e-02 -5.98169327e-01 -5.54875843e-02 1.43545002e-01 -2.25454181e-01 7.04202056e-01 -1.61562085e+00 8.06280039e-03 6.76027536e-01 3.81097011e-02 -1.37023643e-01 2.76007742e-01 -1.36250660e-01 2.07879737e-01 2.09005162e-01 -6.87533557e-01 1.08285226e-01 6.95048094e-01 7.01893091e-01 -1.92554310e-01 5.94771266e-01 2.04574049e-01 5.76349676e-01 -8.89073670e-01 -3.75566334e-01 2.28388235e-01 5.20345867e-01 -3.89384548e-03 -2.99196005e-01 -5.13994455e-01 2.05329582e-01 -4.21469301e-01 5.10880828e-01 6.41241789e-01 1.21724822e-01 1.69967800e-01 -1.63510576e-01 1.59812525e-01 1.46646341e-02 -1.41110992e+00 1.19506335e+00 -7.83708394e-02 4.23504770e-01 -2.12299600e-01 -8.43383789e-01 8.92338455e-01 2.27105975e-01 1.58122450e-01 -3.38195980e-01 5.16956925e-01 1.59696102e-01 4.38451953e-02 -1.67164803e-01 2.21058115e-01 -4.85824019e-01 -4.39091831e-01 3.30884933e-01 3.28793496e-01 -4.49677020e-01 2.17504308e-01 -3.55059318e-02 7.64461577e-01 5.57532728e-01 2.91852206e-01 -6.04405284e-01 1.38959557e-01 -3.32275957e-01 1.03094566e+00 1.20884597e+00 -4.88112837e-01 4.11470473e-01 8.62984598e-01 -5.23633420e-01 -6.46183670e-01 -1.62360919e+00 -8.40696752e-01 4.75314826e-01 2.49634817e-01 2.53110468e-01 -5.89915991e-01 -4.72012430e-01 3.71813506e-01 1.17315948e+00 -7.15166152e-01 -4.05363917e-01 1.80398375e-02 -1.23547649e+00 8.05534959e-01 8.50330353e-01 2.53832400e-01 -5.69887400e-01 -8.60720813e-01 2.83252597e-01 4.19962779e-02 -8.45440805e-01 3.43162656e-01 6.81680262e-01 -7.51980305e-01 -9.32889402e-01 -5.05032003e-01 1.99153483e-01 1.07366353e-01 -9.91901338e-01 1.25495744e+00 -7.83336997e-01 7.07890093e-02 5.41523337e-01 8.75290409e-02 -1.22154725e+00 -7.12011158e-01 -4.73181427e-01 6.64324701e-01 -3.76759350e-01 4.44142044e-01 -5.30173838e-01 -3.52013379e-01 5.50483406e-01 -1.08800948e+00 -4.98890489e-01 3.65552396e-01 7.11064935e-01 5.71376264e-01 1.44718096e-01 7.80702770e-01 -6.48524404e-01 5.91345131e-01 -5.83285570e-01 -1.24783933e+00 2.73463577e-01 -8.57520342e-01 4.67946768e-01 -6.15664583e-04 -2.32744083e-01 -1.26103675e+00 1.09793087e-02 9.84145552e-02 -1.61109596e-01 -3.09219509e-01 7.41604984e-01 -1.34325847e-01 1.22594930e-01 8.86343360e-01 -5.01064301e-01 1.29565209e-01 -1.93359450e-01 2.96923012e-01 4.04846847e-01 7.31884062e-01 -1.04837930e+00 2.40514621e-01 2.52170086e-01 6.10961616e-01 -3.70911002e-01 -5.88188171e-01 3.51731032e-01 -4.72170740e-01 -3.40139091e-01 8.22600961e-01 -6.05497897e-01 -9.73977208e-01 4.41972494e-01 -8.52929950e-01 -3.94493997e-01 -6.57462597e-01 9.52965975e-01 -9.42305267e-01 2.04455450e-01 -6.22235656e-01 -1.17646086e+00 1.14373267e-01 -1.19799364e+00 4.46984708e-01 3.95349175e-01 -1.17458455e-01 -1.22859550e+00 1.71333134e-01 -3.69442105e-01 3.62199903e-01 6.36600316e-01 8.17707241e-01 -9.79784608e-01 -3.38256001e-01 -1.50949851e-01 -2.04459742e-01 4.62104321e-01 -2.16600969e-01 2.23162949e-01 -1.11469030e+00 -1.29501140e-02 2.59770416e-02 -1.80726051e-01 1.06736124e+00 1.06198502e+00 6.89326704e-01 -4.23763186e-01 -5.18284380e-01 3.66076648e-01 1.28204298e+00 4.12803620e-01 5.49546778e-01 3.15627903e-01 -1.17069287e-02 8.96959901e-01 2.64845341e-01 4.43798095e-01 1.81706309e-01 3.21698904e-01 9.48989630e-01 7.15236783e-01 5.09307086e-01 4.84535694e-02 2.02928856e-01 2.05272704e-01 -1.16861768e-01 -3.31674010e-01 -1.15060902e+00 3.27554554e-01 -2.06617475e+00 -7.16215909e-01 -1.43696830e-01 2.34204507e+00 8.17691088e-01 4.70897764e-01 8.79175067e-02 -1.21651411e-01 6.97373152e-01 -2.18202621e-01 -8.62435997e-01 -3.67020994e-01 -1.17069654e-01 -9.14920345e-02 8.77279758e-01 4.72280204e-01 -1.28991699e+00 1.24150939e-01 7.34519482e+00 5.90179622e-01 -3.96427780e-01 -1.07265674e-01 7.48822391e-01 -6.79645240e-02 -2.39789128e-01 -1.58045292e-01 -7.12756574e-01 5.75649619e-01 1.53838098e+00 6.24053106e-02 6.76247329e-02 7.59123445e-01 -1.49362132e-01 -4.90052074e-01 -1.47615552e+00 6.34160042e-01 -3.24601531e-01 -1.11913168e+00 -3.00096363e-01 -1.62100837e-01 6.56956434e-01 2.52587736e-01 -1.29570218e-03 3.98182809e-01 1.18464291e+00 -1.38037014e+00 9.26551342e-01 1.29553032e+00 3.25030565e-01 -9.76744115e-01 1.44646788e+00 2.25083753e-01 -3.38245034e-01 -1.14585571e-01 -5.69437444e-01 1.70028508e-01 3.44880641e-01 1.07245541e+00 -5.71869016e-01 6.06798291e-01 8.83834064e-01 2.98864454e-01 -4.17188220e-02 1.23890150e+00 -2.05470413e-01 5.42698920e-01 -9.68525410e-01 -2.70136446e-01 2.02334329e-01 -2.47177050e-01 6.10535681e-01 1.01615012e+00 5.15069544e-01 -1.57008514e-01 -3.40991259e-01 1.36520636e+00 3.32092762e-01 -7.52324104e-01 -5.04460335e-01 1.89546887e-02 4.82830465e-01 7.94450641e-01 -5.82270622e-01 -1.77539259e-01 1.19476698e-01 -2.14424878e-02 -3.70675653e-01 4.90848333e-01 -8.58604968e-01 -5.92422247e-01 6.55313909e-01 -5.58233202e-01 3.80640924e-02 1.90842465e-01 -2.23528638e-01 -7.12855101e-01 -2.80873477e-01 -4.68996137e-01 6.38001621e-01 -8.51756036e-01 -1.67875004e+00 4.59186256e-01 7.45158434e-01 -1.01926303e+00 -7.98047245e-01 -1.23375392e+00 -1.48289338e-01 8.31683278e-01 -1.49168706e+00 -7.35698819e-01 7.20174834e-02 2.00232342e-01 -2.61911452e-01 1.10934831e-01 7.63401389e-01 -2.89119869e-01 -5.54264963e-01 1.66913658e-01 4.94262040e-01 2.20775474e-02 5.22984862e-01 -1.54792690e+00 -6.58058673e-02 6.74220204e-01 -4.96819019e-01 4.46622878e-01 1.19826674e+00 -6.80362642e-01 -9.61332858e-01 -1.18098569e+00 1.21235073e-01 -7.05985725e-01 6.98917389e-01 1.08547583e-01 -5.51070750e-01 8.72050405e-01 -6.64031506e-02 1.35287091e-01 4.57839519e-01 9.92540643e-02 -4.04949158e-01 -9.00496319e-02 -1.59503579e+00 4.44481730e-01 5.01776755e-01 -4.87903953e-02 -8.24219882e-01 1.75194383e-01 5.45449555e-01 -5.45623362e-01 -1.28199577e+00 8.72340143e-01 1.10646808e+00 -1.14520121e+00 9.77562129e-01 -6.57333016e-01 2.12865755e-01 -4.93852526e-01 -5.94095886e-01 -1.50823879e+00 5.13874292e-02 -6.52520359e-01 -3.50408792e-01 9.44000244e-01 5.00028253e-01 -1.06649685e+00 4.19459939e-01 1.13010144e+00 -1.34124279e-01 -3.23278189e-01 -1.32213521e+00 -1.19332743e+00 5.62388539e-01 -8.96968961e-01 9.50242758e-01 5.30461490e-01 1.06545672e-01 -2.84217149e-01 -1.12431675e-01 4.28527266e-01 1.11520135e+00 -5.84095836e-01 1.07332561e-02 -1.85328388e+00 -1.75280198e-01 -7.27360606e-01 -6.84641659e-01 -4.64677125e-01 -2.73882016e-03 -4.64181244e-01 6.45723045e-01 -1.36955547e+00 -4.10522744e-02 -5.33967555e-01 -6.03361845e-01 4.20994550e-01 4.05756503e-01 8.42270777e-02 -1.73763171e-01 -2.39513174e-01 -3.42765242e-01 4.09476340e-01 5.19102097e-01 -2.94731021e-01 -2.96194367e-02 3.44583362e-01 -3.92502695e-01 1.21152234e+00 7.79878318e-01 -5.41206479e-01 -3.67099911e-01 8.26676786e-02 7.56031513e-01 2.04460591e-01 6.93011522e-01 -1.10081470e+00 3.04750293e-01 -3.15914929e-01 5.38416862e-01 -6.57887101e-01 3.18021953e-01 -7.98178136e-01 4.65839654e-01 3.88336390e-01 -4.27717149e-01 -4.49966431e-01 3.51593345e-01 1.05570829e+00 5.88361584e-02 -8.70868862e-01 7.66052663e-01 -1.93621427e-01 -3.90808761e-01 1.58169821e-01 -7.59618104e-01 -2.15376824e-01 7.89703131e-01 -7.87662864e-02 -5.43960512e-01 -4.52356070e-01 -1.03893101e+00 2.83181548e-01 3.86509821e-02 -1.53702665e-02 3.67252082e-01 -1.29173219e+00 -4.99384880e-01 -2.63287257e-02 1.04575463e-01 3.72580945e-01 2.88329363e-01 7.82285273e-01 -4.06726390e-01 5.87470055e-01 -2.05046594e-01 -9.60787356e-01 -1.66562751e-01 3.94009531e-01 7.79339254e-01 6.50386885e-02 -2.34687701e-01 7.01038897e-01 -3.01133958e-04 -8.44054282e-01 7.01108396e-01 -9.24538374e-01 -1.95521638e-01 -7.53099769e-02 4.99608427e-01 6.16430044e-01 6.92863157e-03 -2.08509684e-01 -2.54313022e-01 2.25828454e-01 3.76603037e-01 -4.33266997e-01 1.15095341e+00 -1.36442989e-01 -5.76106980e-02 7.07250237e-01 3.33459109e-01 -6.48701906e-01 -1.57753706e+00 -1.12688415e-01 3.76355678e-01 1.06857479e-01 2.61960059e-01 -1.38657653e+00 -6.92923009e-01 7.72629142e-01 7.48619556e-01 4.20628458e-01 7.80894041e-01 -2.56698966e-01 -4.02694894e-03 6.36912823e-01 3.90742272e-01 -1.15359676e+00 -5.62020957e-01 4.50001031e-01 6.96876824e-01 -1.09868395e+00 7.55456090e-02 1.49443150e-01 -4.08502162e-01 1.30753267e+00 2.51491159e-01 2.07628042e-01 9.62621450e-01 6.94923222e-01 -4.25031520e-02 1.27022922e-01 -6.54637575e-01 1.12144433e-01 2.73844957e-01 9.86088455e-01 -2.11152926e-01 1.41566187e-01 8.83196592e-02 1.07658923e+00 1.45103307e-02 2.38302112e-01 7.20453084e-01 8.34879935e-01 -3.82410020e-01 -5.35284519e-01 -5.17802358e-01 5.13666213e-01 -3.81090730e-01 2.00103939e-01 2.39340320e-01 8.78256977e-01 -3.39361914e-02 9.99305606e-01 2.19777405e-01 -3.22289094e-02 1.16792582e-01 3.45803618e-01 4.81611490e-01 2.63081733e-02 2.51359075e-01 -3.03693354e-01 4.73891705e-01 -5.80622315e-01 -5.00743747e-01 -7.23966777e-01 -9.71752167e-01 -1.06182612e-01 -3.77544522e-01 2.14877486e-01 1.01671648e+00 1.01778662e+00 8.51866207e-05 5.89477062e-01 -1.65947359e-02 -8.49459529e-01 -1.20198143e+00 -1.04006422e+00 -9.95348036e-01 -3.56630504e-01 3.84781033e-01 -1.01429939e+00 -7.26382554e-01 -1.51178703e-01]
[7.415647506713867, 3.855130195617676]
a3bddcea-5eac-4ef1-a619-9b7689f75201
the-role-of-learning-regime-architecture-and
null
null
https://openreview.net/forum?id=3r034NfDKnL
https://openreview.net/pdf?id=3r034NfDKnL
The Role of Learning Regime, Architecture and Dataset Structure on Systematic Generalization in Simple Neural Networks
Humans often systematically generalize in situations where standard deep neural networks do not. Empirical studies have shown that the learning procedure and network architecture can influence systematicity in deep networks, but the underlying reasons for this influence remain unclear. Here we theoretically study the acquisition of systematic knowledge by simple neural networks. We introduce a minimal space of datasets with systematic and non-systematic features in both the input and output. For shallow and deep linear networks, we derive learning trajectories for all datasets in this space. The solutions reveal that both shallow and deep networks rely on non-systematic inputs to the same extent throughout learning, such that even with early stopping, no networks learn a fully systematic mapping. Turning to the impact of architecture, we show that modularity improves extraction of systematic structure, but only achieves perfect systematicity in the trivial setting where systematic mappings are fully segregated from non-systematic information. Finally, we analyze iterated learning, a procedure in which generations of networks learn from languages generated by earlier learners. Here we find that networks with output modularity successfully converge over generations to a fully systematic `language’ starting from any dataset in our space. Our results contribute to clarifying the role of learning regime, architecture, and dataset structure in promoting systematic generalization, and provide theoretical support for empirical observations that iterated learning can improve systematicity.
['Andrew M Saxe', 'Benjamin Rosman', 'Richard Klein', 'Devon Jarvis']
2021-09-29
null
null
null
null
['systematic-generalization']
['reasoning']
[ 2.27220505e-01 3.45125616e-01 -6.71536475e-02 -1.00259587e-01 2.52241582e-01 -9.77527738e-01 5.70817292e-01 -1.25159817e-02 -5.28572083e-01 5.41486979e-01 9.36502293e-02 -3.46735984e-01 -7.17053592e-01 -9.71248448e-01 -9.99845564e-01 -6.76904023e-01 -2.92586088e-01 4.10558939e-01 8.49334300e-02 -2.35291362e-01 1.04379952e-01 2.67647445e-01 -1.40641415e+00 1.18497573e-01 1.04468906e+00 7.37610683e-02 2.84782529e-01 7.54752100e-01 4.35776375e-02 5.27667463e-01 -3.33432019e-01 -5.80087245e-01 5.75884163e-01 -8.41823876e-01 -9.36723113e-01 -1.47062615e-01 5.19481599e-01 -9.49223340e-02 -3.00601661e-01 1.12328684e+00 3.31463695e-01 -2.04266667e-01 9.11732614e-01 -9.19226885e-01 -1.21620166e+00 1.43954158e+00 -2.90417612e-01 5.31411767e-01 -2.06179574e-01 5.62659085e-01 1.26763964e+00 -5.59138179e-01 6.48056567e-01 1.17033017e+00 1.05114365e+00 6.66653097e-01 -1.69037259e+00 -4.60670412e-01 2.05491900e-01 -4.15368497e-01 -1.12497938e+00 -3.39575112e-01 6.08478904e-01 -8.54682326e-01 8.20658147e-01 -1.97568282e-01 1.28863931e+00 9.59292114e-01 2.62120217e-01 2.90275604e-01 1.04719973e+00 -5.54570436e-01 2.46344674e-02 8.73322338e-02 3.75028074e-01 8.49060655e-01 9.08797503e-01 5.26461899e-01 -7.34349370e-01 3.09214801e-01 1.17149639e+00 -1.27607405e-01 -2.60394961e-01 -4.93918359e-01 -1.16523087e+00 6.98849618e-01 7.36777365e-01 5.79477429e-01 -2.52095401e-01 3.46496940e-01 4.65596877e-02 9.87104535e-01 -1.10740341e-01 1.26628399e+00 -6.98618352e-01 1.23315804e-01 -5.42951524e-01 6.88150451e-02 6.79943919e-01 9.54144716e-01 1.17174375e+00 4.36252579e-02 3.37341726e-01 5.53204656e-01 2.04886943e-02 3.72716188e-01 6.51081026e-01 -1.30869973e+00 2.99430281e-01 8.99060845e-01 -5.20793259e-01 -6.51983142e-01 -6.69085145e-01 -1.13866580e+00 -9.03838515e-01 -1.03947721e-01 7.75995374e-01 -5.18541634e-01 -4.70990002e-01 2.50640678e+00 -2.63174057e-01 -6.19996905e-01 4.38438356e-02 5.31994700e-01 3.90131921e-01 2.30534300e-01 -1.54225782e-01 -1.09327219e-01 8.83348823e-01 -5.29760957e-01 -1.48329481e-01 -4.39479411e-01 8.65290523e-01 -1.43678635e-01 1.24298167e+00 3.08049738e-01 -1.43028259e+00 -5.05513251e-01 -8.85317981e-01 1.14678800e-01 -2.76416332e-01 -3.23262960e-01 8.50819409e-01 9.07872081e-01 -1.66363299e+00 1.03591740e+00 -5.34807682e-01 -5.84619045e-01 7.23324656e-01 6.36739969e-01 -1.98329598e-01 2.22254083e-01 -1.19712031e+00 9.27549303e-01 5.14494538e-01 2.15639751e-02 -9.91283178e-01 -9.18541312e-01 -4.24277753e-01 1.54022068e-01 2.13530332e-01 -1.24323094e+00 1.15126383e+00 -1.66771805e+00 -1.18409646e+00 7.25888252e-01 2.21234217e-01 -4.63566273e-01 7.85944834e-02 4.63950336e-02 3.21586698e-01 -3.91619019e-02 -1.17990598e-01 9.10605431e-01 4.90435034e-01 -1.15693545e+00 -1.03670612e-01 -3.85837287e-01 4.46562558e-01 2.84446359e-01 -6.22399569e-01 -3.07966262e-01 1.24689259e-01 -3.95594329e-01 6.90753758e-02 -1.03880692e+00 -1.24513060e-01 4.15565410e-05 -1.55866638e-01 -7.67149553e-02 -3.45164806e-01 -2.96462566e-01 1.00589907e+00 -1.90810478e+00 6.57751203e-01 2.72871584e-01 5.95970988e-01 -2.06984296e-01 -4.25604999e-01 4.19211596e-01 -1.46258280e-01 5.46014249e-01 -3.48341376e-01 -1.28308818e-01 -2.98563857e-02 3.18935335e-01 -4.78897840e-02 3.21070910e-01 2.07439169e-01 1.41954267e+00 -9.77566779e-01 -6.31542280e-02 -5.49168169e-01 6.60300255e-02 -9.05661583e-01 -1.64761558e-01 -1.21343106e-01 2.80251980e-01 1.89857453e-01 1.00579515e-01 2.29640588e-01 -5.19908428e-01 3.90038371e-01 3.73921990e-01 -1.24087542e-01 3.57927054e-01 -7.49499679e-01 1.55876505e+00 -3.60654891e-01 9.03683662e-01 -1.66070178e-01 -8.85623634e-01 5.35612166e-01 6.09345920e-03 -2.35794753e-01 -6.21101677e-01 1.40245527e-01 4.80128467e-01 1.25550282e+00 -2.65893251e-01 3.96258198e-02 -4.50766832e-01 1.92371383e-01 1.02642095e+00 6.67809546e-01 -3.33482325e-02 3.41995060e-01 4.02019203e-01 1.08521461e+00 -2.74278849e-01 1.85361892e-01 -9.00971651e-01 1.74210683e-01 -1.39550656e-01 5.90891838e-01 1.17810214e+00 6.52699322e-02 3.17138940e-01 1.05693722e+00 -3.81680250e-01 -1.33989882e+00 -1.16155469e+00 -3.33998688e-02 1.29573441e+00 -5.02346009e-02 -1.62344158e-01 -8.69654059e-01 -4.35899585e-01 6.27451614e-02 5.23480296e-01 -1.09703398e+00 -6.88217759e-01 -6.31440461e-01 -6.58812642e-01 6.50824606e-01 3.94506991e-01 4.97660041e-01 -1.14457572e+00 -6.04086936e-01 -1.99366689e-01 4.18774426e-01 -3.54147106e-01 -4.97283697e-01 4.49666172e-01 -1.52944982e+00 -9.66637254e-01 -5.92244387e-01 -8.07358384e-01 1.09608924e+00 1.55190170e-01 1.13037217e+00 6.14824712e-01 -5.36776818e-02 4.28671300e-01 1.51828349e-01 -4.42569749e-03 -6.29110754e-01 8.39730680e-01 3.35481346e-01 -5.27075648e-01 -6.66909069e-02 -1.23181295e+00 -4.84990209e-01 2.36957818e-01 -9.02888656e-01 1.42393261e-01 8.00236166e-01 6.84508085e-01 -2.62671828e-01 7.17557892e-02 1.03177357e+00 -9.43292260e-01 9.11434054e-01 -5.70846140e-01 -4.65893447e-01 3.01615238e-01 -7.86920786e-01 6.61864996e-01 6.84560537e-01 -5.61053932e-01 -9.54208553e-01 -3.17284822e-01 3.82235438e-01 3.17075044e-01 1.37254059e-01 7.65882969e-01 -3.72654013e-02 -3.18832904e-01 1.24184752e+00 4.57539231e-01 1.96258470e-01 -4.21155393e-01 4.41925794e-01 1.87965557e-02 1.49003327e-01 -7.66003728e-01 1.05092013e+00 1.68317750e-01 1.34556249e-01 -6.41226232e-01 -7.62601554e-01 3.89561981e-01 -1.18510544e+00 -1.35330230e-01 4.66488481e-01 -8.26118290e-01 -5.64491391e-01 5.97465873e-01 -1.11514771e+00 -8.88798952e-01 -6.62820399e-01 2.54583240e-01 -3.82848531e-01 8.62530097e-02 -5.43191254e-01 -4.86575127e-01 -5.14685586e-02 -8.37566972e-01 8.56539086e-02 3.17957610e-01 -3.83874416e-01 -1.49254131e+00 3.92905623e-01 -3.15508604e-01 6.80336416e-01 -4.17584658e-01 1.31215584e+00 -6.01476908e-01 -8.76633286e-01 4.58440959e-01 2.12263893e-02 2.71324933e-01 2.22935200e-01 2.00913306e-02 -5.71181834e-01 -1.99492544e-01 -1.63412794e-01 -3.40148509e-01 1.21373641e+00 2.55091637e-01 5.09241760e-01 -4.77395058e-01 5.50622828e-02 8.65552604e-01 1.30069172e+00 -7.96709035e-04 2.52637535e-01 2.80985266e-01 5.92263341e-01 9.71241474e-01 -5.90612173e-01 -1.97238654e-01 3.45061690e-01 6.68599978e-02 1.64720640e-02 4.29107010e-01 -2.00858071e-01 -2.07523584e-01 5.00897527e-01 9.87194300e-01 -2.84179688e-01 9.26438421e-02 -1.20288777e+00 9.71442223e-01 -1.50928462e+00 -1.00748217e+00 1.28299564e-01 2.16716456e+00 1.51089382e+00 3.34351838e-01 2.98009932e-01 -2.40008142e-02 5.36659658e-01 -2.88856864e-01 -5.44144809e-01 -4.79248494e-01 -5.66146076e-01 1.94664940e-01 4.37173337e-01 6.51466668e-01 -4.09158170e-01 8.59500408e-01 6.91935778e+00 2.18834192e-01 -9.46793020e-01 7.24657699e-02 3.61691058e-01 -3.40289593e-01 -8.81572962e-01 -1.24024339e-02 -6.55549645e-01 1.44726381e-01 1.10689306e+00 -4.49646562e-01 8.30543935e-01 3.51431787e-01 -2.99933076e-01 -2.27627549e-02 -1.49060678e+00 3.95649701e-01 -1.70741290e-01 -1.42717457e+00 1.38708174e-01 2.15726212e-01 1.37694645e+00 -2.88087912e-02 3.65033031e-01 2.89527804e-01 7.18164504e-01 -1.08745539e+00 8.19225729e-01 4.97308493e-01 4.62192237e-01 -7.22241640e-01 3.06876749e-01 6.38370454e-01 -5.83519936e-01 -5.01191735e-01 -3.63751590e-01 -5.48091710e-01 -2.10520655e-01 4.95773375e-01 -5.62025487e-01 -8.00535530e-02 2.96602160e-01 6.44265473e-01 -1.15640366e+00 9.80416417e-01 -5.10904610e-01 6.78270698e-01 -1.45795658e-01 6.33360492e-03 -8.75692815e-02 1.08131334e-01 3.63409996e-01 8.18610191e-01 2.11680695e-01 3.34126912e-02 -6.46639228e-01 1.38671303e+00 -3.45821679e-01 -2.70153582e-01 -9.54981089e-01 -3.23855370e-01 4.87047285e-01 7.76371539e-01 -9.21384513e-01 1.45281106e-01 -1.22586519e-01 6.66619480e-01 7.39162922e-01 4.44949597e-01 -1.89321697e-01 -2.17811510e-01 5.01514375e-01 1.65209576e-01 3.23573411e-01 -5.11051595e-01 -9.16255653e-01 -1.06134975e+00 -2.69146740e-01 -6.78335726e-01 8.06494430e-03 -5.01852512e-01 -1.14733613e+00 2.79181719e-01 2.14789379e-02 -3.91151905e-01 -2.24962339e-01 -5.16801059e-01 -6.45642519e-01 8.20711911e-01 -1.00136781e+00 -5.91474771e-01 -5.88202886e-02 3.38683188e-01 1.53402478e-01 -3.89841676e-01 4.92361963e-01 -1.25683442e-01 -6.86720490e-01 7.13821352e-01 5.44026792e-02 1.48748979e-01 2.46980950e-01 -1.36648273e+00 8.76575172e-01 8.57411206e-01 3.94350201e-01 1.59069049e+00 3.62907112e-01 -8.68300498e-01 -1.32190979e+00 -7.61341453e-01 7.52916753e-01 -6.69957042e-01 7.16997147e-01 -6.73391581e-01 -8.62547576e-01 8.76406610e-01 4.73180592e-01 -4.32898670e-01 5.86297870e-01 5.14766335e-01 -5.10965884e-01 -2.97274273e-02 -7.18776524e-01 9.02824402e-01 1.86688066e+00 -6.05862200e-01 -7.79376984e-01 -1.01188265e-01 1.11128879e+00 2.92808205e-01 -6.18616045e-01 9.08711255e-02 4.88872319e-01 -1.08807743e+00 7.54338801e-01 -7.07198560e-01 8.40535045e-01 6.46798238e-02 2.78473526e-01 -1.62636709e+00 -6.40224338e-01 -7.12356091e-01 4.69636396e-02 9.71519589e-01 8.51805687e-01 -9.44810092e-01 6.05341077e-01 2.98826128e-01 2.84606200e-02 -5.50665557e-01 -7.44401634e-01 -9.25282896e-01 7.99118280e-01 1.82938855e-02 5.06944418e-01 8.69580805e-01 4.24730405e-03 5.88992178e-01 1.98458806e-01 -2.00695932e-01 7.49595463e-01 6.80965558e-03 4.66345429e-01 -1.52074504e+00 -4.87883866e-01 -9.38691974e-01 -2.17365861e-01 -9.55874622e-01 1.37146696e-01 -1.26362276e+00 -1.59241885e-01 -1.29264116e+00 5.67603469e-01 -3.37896258e-01 -2.16053128e-02 2.91213930e-01 -1.11787640e-01 4.68661860e-02 2.37867504e-01 3.81402075e-01 -4.57935125e-01 3.69188964e-01 1.34506822e+00 1.96706146e-01 -2.55787909e-01 -1.17857538e-01 -1.43257535e+00 7.62462199e-01 9.51156437e-01 -5.66376328e-01 -7.22415626e-01 -8.67236674e-01 1.22998118e+00 -5.45177817e-01 4.76159602e-01 -8.76196742e-01 3.56139272e-01 -2.09301308e-01 4.28659320e-01 7.37560168e-02 -3.87930185e-01 -5.92894733e-01 -3.23184137e-03 8.35632265e-01 -8.16982090e-01 2.43931711e-01 1.39676526e-01 2.63159156e-01 4.76846993e-01 -5.60110509e-01 7.75135815e-01 -3.65210533e-01 -2.10306540e-01 2.74299067e-02 -5.93563795e-01 5.91283381e-01 6.52706683e-01 -2.94187516e-01 -6.49666786e-01 -1.75762117e-01 -7.26799965e-01 3.53599638e-02 7.20069885e-01 5.75127676e-02 3.58850062e-01 -1.24810433e+00 -6.74337566e-01 3.25238675e-01 -3.09659421e-01 -5.46063744e-02 7.67063498e-02 9.22218978e-01 -4.77304786e-01 3.51903647e-01 -4.17813390e-01 -5.21668315e-01 -6.20962441e-01 4.19798166e-01 7.76183844e-01 6.56401366e-02 -1.46060795e-01 1.31124794e+00 6.95247352e-01 -7.27170765e-01 1.76955219e-02 -4.56990898e-01 -4.05949838e-02 2.41344407e-01 1.96116194e-01 1.82233781e-01 -4.23621982e-01 -1.54251948e-01 -8.52347761e-02 6.78518474e-01 -2.51572639e-01 -3.23470414e-01 1.41158473e+00 -1.15279987e-01 -5.38351238e-01 8.11892748e-01 9.12957311e-01 -1.24853492e-01 -1.31855702e+00 -4.18917209e-01 1.65369306e-02 -4.45807949e-02 -4.70666409e-01 -6.64254248e-01 -1.25370908e+00 1.07817042e+00 4.88841422e-02 3.03568006e-01 8.74250650e-01 1.71177626e-01 5.37382551e-02 7.71061540e-01 3.86751831e-01 -7.60522068e-01 4.73230660e-01 9.73426819e-01 9.41666603e-01 -8.26164722e-01 1.74201056e-02 1.69203132e-01 -3.59505773e-01 1.16372812e+00 7.60529459e-01 -3.25545609e-01 5.99326074e-01 -1.17397189e-01 -5.05120993e-01 -2.88958430e-01 -1.07772088e+00 -2.26271570e-01 1.83107987e-01 6.86743379e-01 4.05015916e-01 -2.56000519e-01 -4.39501293e-02 8.14451635e-01 -7.62599528e-01 -1.98383108e-01 7.10050523e-01 6.01494789e-01 -6.98212683e-01 -9.54380691e-01 -2.89845634e-02 4.11546916e-01 -1.01857744e-01 -5.20989299e-01 -7.87732244e-01 8.52704823e-01 3.92088860e-01 2.31809542e-01 2.40378618e-01 -3.23174417e-01 8.60515703e-03 1.86770678e-01 1.06586647e+00 -9.98843610e-01 -1.04089642e+00 -4.63322729e-01 -1.58335075e-01 4.13639918e-02 -1.56142041e-01 -1.05974567e+00 -1.04323089e+00 -5.31838417e-01 -2.19160244e-01 -1.43118843e-01 6.23556636e-02 7.82694995e-01 2.47845739e-01 5.98776996e-01 2.82374829e-01 -5.28224647e-01 -6.26453876e-01 -8.55577826e-01 -5.14925301e-01 1.09472387e-01 5.42110801e-01 -3.74881804e-01 -6.91700220e-01 3.63220423e-02]
[8.22217082977295, 3.4949724674224854]
21129f3e-5b34-4190-bc3e-2c008d74c1f8
menuai-restaurant-food-recommendation-system
2210.08266
null
https://arxiv.org/abs/2210.08266v1
https://arxiv.org/pdf/2210.08266v1.pdf
MenuAI: Restaurant Food Recommendation System via a Transformer-based Deep Learning Model
Food recommendation system has proven as an effective technology to provide guidance on dietary choices, and this is especially important for patients suffering from chronic diseases. Unlike other multimedia recommendations, such as books and movies, food recommendation task is highly relied on the context at the moment, since users' food preference can be highly dynamic over time. For example, individuals tend to eat more calories earlier in the day and eat a little less at dinner. However, there are still limited research works trying to incorporate both current context and nutritional knowledge for food recommendation. Thus, a novel restaurant food recommendation system is proposed in this paper to recommend food dishes to users according to their special nutritional needs. Our proposed system utilises Optical Character Recognition (OCR) technology and a transformer-based deep learning model, Learning to Rank (LTR) model, to conduct food recommendation. Given a single RGB image of the menu, the system is then able to rank the food dishes in terms of the input search key (e.g., calorie, protein level). Due to the property of the transformer, our system can also rank unseen food dishes. Comprehensive experiments are conducted to validate our methods on a self-constructed menu dataset, known as MenuRank dataset. The promising results, with accuracy ranging from 77.2% to 99.5%, have demonstrated the great potential of LTR model in addressing food recommendation problems.
['Benny Lo', 'Jiachuan Peng', 'Peilun Shi', 'Jianing Qiu', 'Frank Po Wen Lo', 'Xinwei Ju']
2022-10-15
null
null
null
null
['food-recommendation']
['miscellaneous']
[-1.97332539e-02 -5.41652262e-01 -5.20125568e-01 -4.80519384e-01 -1.75340965e-01 -5.49274743e-01 4.39879075e-02 7.36541927e-01 -2.11720407e-01 1.00911528e-01 4.86557156e-01 2.26772204e-02 -1.31202966e-01 -1.14304197e+00 -3.91397327e-01 -6.88728333e-01 -3.10537927e-02 1.19894922e-01 1.40206710e-01 -3.43306959e-01 2.71104425e-01 3.03620882e-02 -1.77141595e+00 5.22738516e-01 7.72870898e-01 1.02802312e+00 6.18455768e-01 4.63711739e-01 -2.49583915e-01 3.97355080e-01 -4.27492108e-04 -2.73154583e-02 1.83399454e-01 -3.28890264e-01 -1.67378113e-01 1.59735069e-01 2.61014313e-01 -7.61348128e-01 -3.10863495e-01 1.13896859e+00 5.16511440e-01 3.34204197e-01 5.53481042e-01 -6.15895212e-01 -1.15595114e+00 9.59732652e-01 -1.93150654e-01 1.60436593e-02 6.28295660e-01 3.22175324e-02 9.14429128e-01 -5.64516604e-01 6.74374700e-02 9.97947752e-01 3.15856218e-01 4.94271815e-01 -7.68795907e-01 -5.75037122e-01 4.70663011e-01 4.20552433e-01 -1.08371806e+00 -8.52919966e-02 7.89169908e-01 -4.33001578e-01 6.15117073e-01 2.37577617e-01 1.08963966e+00 8.98617923e-01 1.92652002e-01 8.12023163e-01 9.88747418e-01 -3.05903405e-01 2.22475111e-01 1.53759211e-01 2.46759042e-01 6.59294665e-01 3.28046292e-01 9.03498381e-02 -3.47571343e-01 4.10490215e-01 6.38519466e-01 6.54919565e-01 -2.06629261e-01 -4.08667356e-01 -1.42794180e+00 9.36411381e-01 8.10860097e-01 2.39565685e-01 -3.82829964e-01 -2.34399423e-01 3.92859638e-01 2.55441844e-01 4.14961129e-01 1.13522179e-01 -3.77951592e-01 4.02907223e-01 -6.55456960e-01 1.20303139e-01 7.53008664e-01 7.64418185e-01 3.08164984e-01 -2.49198973e-01 -8.69273692e-02 9.22327161e-01 8.36321294e-01 8.59815478e-01 6.91068649e-01 -4.68157917e-01 1.13854654e-01 5.78862011e-01 3.30197185e-01 -1.51394260e+00 -6.60726488e-01 -2.04155177e-01 -8.77523661e-01 -9.17115659e-02 3.48367155e-01 1.90799087e-01 -7.82849908e-01 1.40924609e+00 4.50991213e-01 -2.75164843e-01 1.57476086e-02 1.35335088e+00 1.35488296e+00 7.38884389e-01 3.24265063e-01 -1.56520620e-01 1.53585088e+00 -8.24290454e-01 -6.46992326e-01 7.95197859e-02 4.45663661e-01 -8.51518631e-01 1.15856278e+00 6.35283411e-01 -6.04710042e-01 -7.45528877e-01 -1.01285362e+00 -1.20606549e-01 -5.49330413e-01 3.60150456e-01 8.47763360e-01 5.87768674e-01 -7.47100115e-01 6.82499707e-01 -7.32885063e-01 -8.66418540e-01 -7.41657391e-02 3.61259103e-01 6.87367097e-02 -4.47934002e-01 -1.17774510e+00 3.84658664e-01 4.46500152e-01 2.50764221e-01 -5.30494630e-01 -4.45537567e-01 -7.48735428e-01 -5.55666201e-02 2.59907871e-01 -6.21856570e-01 1.20498788e+00 -7.50007331e-01 -1.80434477e+00 5.00897765e-01 3.97977680e-01 -1.34640217e-01 1.80682987e-01 -3.79167259e-01 -7.98544407e-01 1.33776128e-01 -4.30536233e-02 7.46071458e-01 4.84447777e-01 -8.22842836e-01 -9.32801604e-01 -4.31005538e-01 3.76814276e-01 3.47057283e-01 -4.52752799e-01 -6.42879233e-02 -3.50863695e-01 -6.02544785e-01 2.86539733e-01 -7.38099992e-01 -1.55271664e-01 3.13035578e-01 -1.41532958e-01 -3.14833522e-01 3.52693647e-01 -8.01197171e-01 1.25660884e+00 -2.18823600e+00 -1.40146673e-01 1.31421193e-01 -1.78773984e-01 1.79254994e-01 -3.54553126e-02 1.35809109e-01 2.50199109e-01 -1.37879193e-01 2.36903057e-01 4.15924937e-01 2.86721230e-01 8.63255039e-02 1.13797337e-01 4.34528768e-01 -3.72792006e-01 5.99097133e-01 -1.21873724e+00 -4.12417531e-01 6.32372737e-01 6.81131184e-01 -4.74136412e-01 5.95969222e-02 -6.02892458e-01 2.22492293e-01 -7.35184312e-01 7.66439378e-01 5.12371957e-01 -2.73776352e-01 2.90592045e-01 -7.38122523e-01 -3.67922544e-01 2.16671735e-01 -1.21426749e+00 1.81428790e+00 -2.49760821e-01 -8.52648765e-02 -2.91738153e-01 -6.65090144e-01 8.91379654e-01 2.85646558e-01 7.18843162e-01 -1.11296105e+00 2.97695041e-01 2.03318119e-01 -8.13607946e-02 -8.41388822e-01 4.86280233e-01 2.83249110e-01 9.80390012e-02 3.65977854e-01 -4.73448843e-01 4.21556562e-01 4.89122212e-01 -1.72213659e-01 5.63788652e-01 4.67358917e-01 3.84714067e-01 -4.53830421e-01 5.21543145e-01 1.10598706e-01 1.58644870e-01 2.86350816e-01 -2.05423683e-01 3.23312700e-01 -5.16778648e-01 -6.93619490e-01 -9.62485373e-01 -9.14212823e-01 -6.37406185e-02 1.35058916e+00 4.30132776e-01 -2.23932266e-01 -2.93412924e-01 -4.97950464e-01 -9.16076079e-02 5.51680446e-01 -3.88858855e-01 -1.62964135e-01 -5.24416566e-01 -7.24459350e-01 -4.13607985e-01 5.23986161e-01 8.32281768e-01 -1.13247168e+00 -8.53618443e-01 5.29131472e-01 -4.35022384e-01 -6.50052309e-01 -9.13573563e-01 1.38550242e-02 -8.46785665e-01 -1.10631073e+00 -9.00290012e-01 -1.09404790e+00 5.46216846e-01 8.26819241e-01 1.10039890e+00 1.31137341e-01 -7.35670105e-02 4.44386542e-01 -8.28468025e-01 -4.28271145e-02 -2.40711853e-01 -1.07703647e-02 1.76820345e-02 -1.02670491e-01 7.70199180e-01 -1.85087606e-01 -1.45984960e+00 4.88817185e-01 -9.64466751e-01 1.78122044e-01 4.76905376e-01 4.73291904e-01 9.28668797e-01 3.50247502e-01 3.15113455e-01 -5.88395178e-01 3.24620575e-01 -6.26121700e-01 -5.09880781e-01 2.72018701e-01 -5.80749810e-01 -9.97298658e-02 6.81214809e-01 -8.44781816e-01 -7.12968707e-01 2.80491441e-01 -2.90664196e-01 1.29130796e-01 -2.89841712e-01 5.60196757e-01 -1.54039875e-01 2.09506080e-01 4.41249669e-01 2.11070590e-02 -4.42566901e-01 -6.79907441e-01 3.93914580e-01 7.01257169e-01 4.99752194e-01 -4.06454772e-01 1.82334393e-01 1.58414215e-01 -2.81718880e-01 -6.48138463e-01 -8.40136647e-01 -7.12984025e-01 -4.27328765e-01 -3.01377863e-01 9.43180501e-01 -8.41175854e-01 -9.77827847e-01 3.11432213e-01 -4.37309980e-01 -2.26551801e-01 9.88866612e-02 9.65135455e-01 -2.27957442e-01 3.36361557e-01 -6.94864631e-01 -4.70533937e-01 -7.06120670e-01 -9.96576726e-01 7.49404311e-01 4.07086492e-01 -2.48221587e-02 -9.14553702e-01 -1.41914010e-01 2.22842425e-01 4.40273881e-01 3.01048279e-01 9.50739563e-01 -3.68204713e-01 -4.09405977e-01 7.96455052e-03 -3.23203623e-01 -1.36934116e-01 5.00675857e-01 -5.46238199e-02 -5.53457618e-01 -3.04333001e-01 -1.51968375e-01 -2.68044844e-02 8.07791948e-01 6.13187969e-01 8.68009627e-01 -2.56003082e-01 -2.90384531e-01 3.53241384e-01 1.53340375e+00 4.92043436e-01 4.26193625e-01 5.27101219e-01 7.68693924e-01 5.17700851e-01 7.73867548e-01 4.42305148e-01 7.25647986e-01 7.96251297e-01 4.81593817e-01 -4.04855385e-02 -3.91173154e-01 -2.28685632e-01 5.70740879e-01 1.00614941e+00 2.13845760e-01 -3.88511419e-01 -2.34607548e-01 3.05774242e-01 -1.72642577e+00 -7.64276028e-01 -1.92285821e-01 2.42212820e+00 6.95668578e-01 -2.37442940e-01 3.37812603e-01 1.59916312e-01 6.25564814e-01 -2.88099915e-01 -9.34696913e-01 -1.48906454e-01 1.48284674e-01 -1.53671801e-01 4.29000974e-01 3.83475423e-02 -1.38687813e+00 5.67778766e-01 5.60363483e+00 2.74711192e-01 -1.30364442e+00 -2.21721217e-01 4.46928889e-01 -1.83590185e-02 -2.28515953e-01 -7.98852503e-01 -7.19061732e-01 6.61700070e-01 7.04336822e-01 2.11868078e-01 6.84648037e-01 7.40250885e-01 2.27125913e-01 -9.82261598e-02 -1.15187275e+00 9.84108090e-01 1.86550885e-01 -6.90433681e-01 -1.10165037e-01 -2.11182296e-01 4.50119197e-01 -1.09728396e-01 3.57730538e-01 1.49703115e-01 5.17663181e-01 -6.53935194e-01 8.10852051e-01 4.49494332e-01 6.53994799e-01 -6.41777992e-01 5.52859247e-01 1.47165328e-01 -1.57737851e+00 -1.48359001e-01 -7.71528780e-01 1.67317942e-01 -1.75677374e-01 7.30249345e-01 -4.82440412e-01 4.48781312e-01 1.09935558e+00 1.03426898e+00 -3.39132935e-01 1.27747083e+00 -1.43672615e-01 2.97299385e-01 -5.91022313e-01 -2.75677979e-01 -1.13196895e-02 -3.45882773e-01 4.48343344e-03 9.69804227e-01 8.68936181e-01 6.12981558e-01 5.90088606e-01 3.95584166e-01 1.60986319e-01 8.60089421e-01 -1.67257383e-01 -8.89371932e-02 -1.78076074e-01 1.23921597e+00 -1.17738259e+00 -1.06288865e-01 -5.55756092e-01 1.02338338e+00 -1.61808193e-01 1.53237283e-02 -6.01056993e-01 1.48221061e-01 3.95803928e-01 -2.66869813e-02 5.69053650e-01 -1.64761588e-01 3.03302467e-01 -1.13490617e+00 -2.70822108e-01 -8.50243390e-01 6.50472045e-01 -7.09346056e-01 -1.38231206e+00 3.07376236e-01 -2.11188003e-01 -1.25004423e+00 9.23821852e-02 -6.51048124e-01 -5.13656624e-02 4.23366040e-01 -1.65934193e+00 -1.15108633e+00 -4.00917470e-01 6.68988109e-01 7.39629328e-01 2.20951468e-01 8.80342066e-01 6.70940697e-01 -5.27290225e-01 5.12460291e-01 4.15090114e-01 -1.24794640e-01 7.14259088e-01 -1.20532334e+00 -1.18652157e-01 3.54876041e-01 1.01803407e-01 4.92609799e-01 7.32162237e-01 -7.81733453e-01 -1.72733057e+00 -1.12185812e+00 7.30988383e-01 -1.89875718e-03 3.24693888e-01 5.90088107e-02 -6.82794452e-01 1.77269489e-01 -3.70514132e-02 -3.33837718e-01 9.90494907e-01 -4.79479171e-02 -3.06008697e-01 -3.63526285e-01 -1.36630869e+00 5.31429648e-01 8.93750846e-01 1.67301055e-02 -2.48689145e-01 5.91341317e-01 6.87781453e-01 -3.48676741e-01 -1.04258442e+00 2.72775501e-01 1.02277684e+00 -7.83310413e-01 1.16699052e+00 -1.27691612e-01 3.65616232e-01 -5.12334287e-01 -5.49310267e-01 -1.33370030e+00 -7.72694707e-01 6.09628260e-02 -1.73890777e-02 1.08544421e+00 8.90737474e-02 -2.68410772e-01 6.15072012e-01 4.52180088e-01 -1.43392667e-01 -4.51811910e-01 -6.66684285e-02 -2.52650917e-01 -3.38819802e-01 -3.07795126e-02 1.03559792e+00 7.51779377e-01 4.32751402e-02 1.36521921e-01 -5.39096951e-01 4.08515036e-01 5.27651191e-01 5.29564500e-01 2.85449713e-01 -1.29434383e+00 -2.61368424e-01 -4.63678360e-01 -1.61601901e-02 -1.25790989e+00 -5.80028892e-01 -9.54793215e-01 3.20983410e-01 -1.93382645e+00 2.67838806e-01 -4.66856927e-01 -8.81525397e-01 3.77244323e-01 -7.92251974e-02 3.84670228e-01 8.62271264e-02 1.56165510e-01 -6.38507068e-01 1.32036313e-01 1.47442901e+00 -5.24094701e-01 -4.15045857e-01 2.55198088e-02 -8.38646233e-01 5.89947343e-01 9.35437858e-01 -2.27457270e-01 -5.44447362e-01 -5.08137822e-01 4.81914610e-01 -1.79468445e-03 -4.16530818e-02 -1.01902890e+00 6.95181713e-02 -2.37705052e-01 6.08717680e-01 -7.53660679e-01 -6.35689273e-02 -1.13883710e+00 3.94568563e-01 7.13111341e-01 -5.50001502e-01 -1.44501403e-01 -4.94809039e-02 5.41138232e-01 3.31347257e-01 -6.02079593e-02 3.92518401e-01 -2.60602742e-01 -1.02020550e+00 4.93316531e-01 -2.11038485e-01 -7.02665150e-01 7.30126679e-01 -1.15772530e-01 -1.22171707e-01 -2.54636407e-02 -8.67203355e-01 2.96462655e-01 3.44651103e-01 5.50636947e-01 6.10963821e-01 -1.57420242e+00 -6.66238844e-01 2.28906751e-01 4.07812834e-01 -3.69457245e-01 4.82453644e-01 5.01202285e-01 -5.38545489e-01 4.53305036e-01 -2.47170761e-01 -5.11992157e-01 -1.06998646e+00 1.09043932e+00 6.73932731e-02 -9.45466906e-02 -1.08343220e+00 5.09661913e-01 2.08915830e-01 -2.73349881e-01 4.18946236e-01 -9.64476883e-01 -8.90725374e-01 7.18132267e-03 9.28398848e-01 2.75591284e-01 4.19111289e-02 -5.24955153e-01 -6.33798838e-02 9.40629482e-01 -9.54490975e-02 5.83393514e-01 1.34774148e+00 -6.15165889e-01 2.55209506e-01 3.70269001e-01 7.51340032e-01 -3.39852273e-01 -1.13713539e+00 -3.38523805e-01 -3.03446323e-01 -5.00198424e-01 1.99964896e-01 -1.09371114e+00 -1.41618037e+00 6.57442927e-01 1.32800937e+00 2.60772586e-01 1.32096696e+00 -1.73828080e-01 1.26937425e+00 3.87343973e-01 5.12918711e-01 -8.18648934e-01 8.83191153e-02 1.89504512e-02 5.02239168e-01 -1.48078763e+00 -7.40946131e-03 -3.77865106e-01 -2.53348589e-01 1.31588733e+00 2.23678783e-01 -5.58802672e-02 8.57327878e-01 -2.16167271e-01 2.64268667e-01 -6.73600510e-02 -1.84468076e-01 -1.94358096e-01 3.41156960e-01 6.06044352e-01 6.78895772e-01 4.77102637e-01 -3.58895510e-01 5.71665287e-01 -2.37340808e-01 4.04665172e-01 -1.51979998e-02 4.97525692e-01 -8.34890902e-01 -1.09029162e+00 -3.48723859e-01 5.08576572e-01 -2.14498758e-01 -2.04956066e-02 3.06046247e-01 2.53152817e-01 3.76142859e-01 1.36848772e+00 -1.08209975e-01 -4.65248466e-01 3.54927003e-01 -2.81318992e-01 5.29484928e-01 -4.84791130e-01 -6.18974745e-01 4.29495007e-01 -1.84286997e-01 -4.54615027e-01 -7.46980190e-01 -6.59438074e-01 -1.29745030e+00 -5.15220404e-01 -2.08314553e-01 -6.24006428e-02 6.67805731e-01 8.42971325e-01 -1.00412302e-01 5.52757144e-01 6.22916281e-01 -6.93288445e-01 -3.48945767e-01 -5.57407260e-01 -7.26827562e-01 6.93719566e-01 8.15517157e-02 -6.94427550e-01 7.00071305e-02 1.43877879e-01]
[11.552659034729004, 4.422238826751709]
201a1647-fc6a-42b7-b29c-98b8199055f9
douzero-mastering-doudizhu-with-self-play
2106.06135
null
https://arxiv.org/abs/2106.06135v1
https://arxiv.org/pdf/2106.06135v1.pdf
DouZero: Mastering DouDizhu with Self-Play Deep Reinforcement Learning
Games are abstractions of the real world, where artificial agents learn to compete and cooperate with other agents. While significant achievements have been made in various perfect- and imperfect-information games, DouDizhu (a.k.a. Fighting the Landlord), a three-player card game, is still unsolved. DouDizhu is a very challenging domain with competition, collaboration, imperfect information, large state space, and particularly a massive set of possible actions where the legal actions vary significantly from turn to turn. Unfortunately, modern reinforcement learning algorithms mainly focus on simple and small action spaces, and not surprisingly, are shown not to make satisfactory progress in DouDizhu. In this work, we propose a conceptually simple yet effective DouDizhu AI system, namely DouZero, which enhances traditional Monte-Carlo methods with deep neural networks, action encoding, and parallel actors. Starting from scratch in a single server with four GPUs, DouZero outperformed all the existing DouDizhu AI programs in days of training and was ranked the first in the Botzone leaderboard among 344 AI agents. Through building DouZero, we show that classic Monte-Carlo methods can be made to deliver strong results in a hard domain with a complex action space. The code and an online demo are released at https://github.com/kwai/DouZero with the hope that this insight could motivate future work.
['Ji Liu', 'Xia Hu', 'Xiangru Lian', 'Sheng Zhang', 'Wenye Ma', 'Jingru Xie', 'Daochen Zha']
2021-06-11
null
null
null
null
['game-of-poker', 'card-games']
['playing-games', 'playing-games']
[-4.80268747e-01 -2.83650696e-01 1.00835465e-01 3.50964725e-01 -6.02375984e-01 -7.40064502e-01 8.52207899e-01 -2.52976328e-01 -5.49292922e-01 1.40659857e+00 -1.24158219e-01 -4.71392542e-01 -1.55982941e-01 -1.23574483e+00 -7.01023817e-01 -8.09023082e-01 -4.29705203e-01 1.16752219e+00 4.05416846e-01 -9.37190533e-01 1.04983866e-01 4.67830617e-03 -1.37128961e+00 1.95360538e-02 6.86800122e-01 6.35953784e-01 -1.33092150e-01 1.02902031e+00 4.91223216e-01 1.33328426e+00 -8.98551762e-01 -4.16604519e-01 7.29130089e-01 -5.93923151e-01 -8.05696845e-01 -4.64651823e-01 -1.27830833e-01 -8.29633057e-01 -7.50986397e-01 8.16188395e-01 7.82412589e-01 2.20992997e-01 2.93597758e-01 -1.53787434e+00 -2.02654079e-01 9.46109772e-01 -3.14896286e-01 1.41035631e-01 3.54841560e-01 1.06101012e+00 1.05131352e+00 -1.84130758e-01 5.99361062e-01 9.68338132e-01 5.00583231e-01 7.23491192e-01 -7.44164884e-01 -7.55259275e-01 -2.74291515e-01 6.02426112e-01 -1.06428599e+00 2.01910421e-01 2.79187948e-01 -1.84409142e-01 1.27142608e+00 1.96525320e-01 1.17737031e+00 1.18652284e+00 3.35607171e-01 1.14860618e+00 1.24912298e+00 -9.18443650e-02 7.63392568e-01 -8.98032784e-01 -4.03396875e-01 7.85999119e-01 2.01463401e-01 7.13203609e-01 -2.54806012e-01 -4.34591562e-01 7.14897692e-01 -1.16126373e-01 2.95429409e-01 -3.29577714e-01 -1.23382020e+00 1.10236561e+00 4.16395903e-01 2.16361791e-01 -5.32244802e-01 7.86828339e-01 4.94747639e-01 4.14699376e-01 -4.66869362e-02 7.66989172e-01 -3.11115324e-01 -8.30177188e-01 -5.76295912e-01 1.09617293e+00 1.07371569e+00 4.54154104e-01 5.92136741e-01 2.89678991e-01 -1.11030951e-01 2.09531691e-02 -2.65713930e-01 4.30103272e-01 8.24108124e-02 -1.50418127e+00 3.42166722e-01 2.52936929e-01 4.40876395e-01 -5.14506936e-01 -5.83198905e-01 -1.97191656e-01 -8.69119763e-01 8.19556713e-01 8.16724300e-01 -9.45400059e-01 -5.48100710e-01 1.51046646e+00 4.28096324e-01 3.54350865e-01 1.89901114e-01 1.05835581e+00 2.85027206e-01 8.64999533e-01 -2.29872093e-01 2.87214696e-01 1.13997984e+00 -1.16177917e+00 -2.48803616e-01 -2.93318719e-01 8.97895575e-01 -3.20891410e-01 7.05127835e-01 7.99652398e-01 -1.32308853e+00 1.02132089e-01 -1.01054633e+00 3.55856478e-01 -2.66293585e-01 -2.61130273e-01 1.43798351e+00 6.42128170e-01 -1.10832167e+00 6.88076556e-01 -1.04858232e+00 3.23289260e-02 5.42273223e-01 4.72673863e-01 -7.17596114e-02 -6.70947134e-02 -1.71215284e+00 9.22042310e-01 6.08553052e-01 -1.31467938e-01 -1.36184752e+00 -3.62094551e-01 -4.35141742e-01 9.42280516e-02 1.11174130e+00 -7.34926403e-01 1.59293485e+00 -5.76379716e-01 -2.05656695e+00 3.75039220e-01 7.98466742e-01 -8.88854086e-01 9.08101559e-01 1.41791195e-01 1.73143059e-01 -1.21410623e-01 -7.10563064e-02 4.61869836e-01 1.94853529e-01 -5.74836671e-01 -8.22947562e-01 -1.95931286e-01 8.14159334e-01 4.66000438e-01 4.72070485e-01 1.13565344e-02 9.71964151e-02 -2.06715420e-01 -6.42926633e-01 -1.09213889e+00 -8.06292295e-01 -5.58030486e-01 -3.71937156e-02 -6.21386766e-01 1.34233981e-01 -1.91906616e-01 8.11695576e-01 -1.50500381e+00 1.07633218e-01 -1.44904237e-02 4.77026820e-01 4.19778764e-01 -1.50094301e-01 9.35612798e-01 4.80439216e-01 -2.05893740e-02 -1.05161682e-01 3.04662704e-01 5.17315090e-01 4.47531074e-01 3.13637704e-02 4.41687316e-01 -4.98785190e-02 1.14872921e+00 -1.38132584e+00 -9.64994952e-02 2.24154577e-01 -2.57508695e-01 -1.01727402e+00 1.00200931e-02 -5.86115003e-01 4.29693043e-01 -7.31527150e-01 4.37803596e-01 5.27941227e-01 -7.65704438e-02 3.02921534e-01 8.24485540e-01 -2.51395047e-01 2.66320825e-01 -1.22133231e+00 1.54694295e+00 -2.25582555e-01 5.07671475e-01 1.83674365e-01 -1.07001853e+00 4.66411501e-01 2.40601510e-01 6.84911609e-01 -8.69867802e-01 6.11419678e-01 4.09493387e-01 6.04359031e-01 -8.83510783e-02 5.68764865e-01 -7.30647356e-04 -5.75676382e-01 6.56432509e-01 -4.17765230e-02 -7.79310346e-01 8.75774622e-01 3.46754253e-01 1.76395679e+00 1.81250229e-01 4.87616479e-01 -3.71057093e-02 -6.29967451e-02 6.35080159e-01 4.84187335e-01 1.49832940e+00 -6.02683485e-01 2.35362545e-01 9.42680538e-01 -7.58619547e-01 -1.18601584e+00 -8.80180895e-01 3.41279149e-01 1.17468512e+00 2.70093113e-01 -5.08839369e-01 -9.75100517e-01 -4.27189380e-01 -5.43345697e-02 6.79138422e-01 -6.32433057e-01 1.42645454e-02 -6.55233860e-01 -7.43606567e-01 8.52832139e-01 2.31373429e-01 8.56860340e-01 -1.41050053e+00 -1.01214480e+00 5.19194245e-01 8.92991200e-02 -6.06526554e-01 3.66527587e-02 1.67488784e-01 -2.37987921e-01 -1.35957956e+00 -4.90892142e-01 -1.91670582e-01 -1.70305982e-01 8.65281522e-02 1.31327069e+00 1.37661487e-01 -3.30905080e-01 1.62821829e-01 -4.03818280e-01 -4.51189995e-01 -5.04803836e-01 1.01146497e-01 6.02295250e-02 -7.90099084e-01 3.86090845e-01 -6.22364163e-01 -6.77883089e-01 4.14785236e-01 -7.60430813e-01 2.59127617e-01 4.11994070e-01 1.32992613e+00 -6.93309829e-02 1.55819967e-01 2.18487218e-01 -6.74605012e-01 6.91949427e-01 -6.67228937e-01 -1.06305993e+00 -1.54306069e-01 -1.86291020e-02 -2.02573121e-01 9.48300838e-01 -2.62164384e-01 -7.10500240e-01 -2.05024123e-01 -3.63228738e-01 -2.69302782e-02 -1.43501803e-01 4.15202290e-01 3.10423017e-01 1.03053778e-01 9.73973751e-01 1.29130185e-01 -8.84078145e-02 2.04914585e-01 3.40049297e-01 4.08870608e-01 2.01898620e-01 -9.46135223e-01 5.03008783e-01 2.52011955e-01 -2.11158041e-02 -3.75271857e-01 -3.63238662e-01 4.09146622e-02 -1.35646224e-01 -3.01538974e-01 6.09073937e-01 -8.50908279e-01 -1.65153539e+00 9.85769808e-01 -1.01587820e+00 -1.09590268e+00 -5.00137031e-01 4.68464166e-01 -8.99163485e-01 1.04886279e-01 -8.79643381e-01 -8.39999616e-01 1.00325882e-01 -1.21197498e+00 6.32985950e-01 4.38281178e-01 1.06729776e-01 -6.55579329e-01 6.05452299e-01 5.78076005e-01 5.90583861e-01 1.66571200e-01 3.65850061e-01 -4.96611536e-01 -8.84692013e-01 -4.18401480e-01 2.35618308e-01 -1.04151154e-02 -3.96111876e-01 3.43745947e-02 -3.02548140e-01 -3.98188531e-01 -2.04929486e-01 -9.30183649e-01 5.13276041e-01 5.36828578e-01 8.48337233e-01 -3.15102816e-01 8.40253457e-02 3.17379594e-01 1.05164957e+00 6.61796927e-01 8.04192066e-01 5.81451893e-01 2.74032950e-01 1.51451081e-02 4.91310388e-01 1.12553513e+00 5.89818358e-01 7.84125626e-01 9.59746301e-01 2.27771938e-01 2.87561208e-01 -9.18936729e-02 4.82459307e-01 4.27224636e-01 -6.28669977e-01 -4.30720240e-01 -1.15114415e+00 3.79488111e-01 -2.13567662e+00 -1.40771401e+00 -7.98114575e-03 1.88700306e+00 8.21674705e-01 2.22860277e-01 4.22148794e-01 -1.99133858e-01 3.19522321e-01 2.59055972e-01 -7.31674790e-01 -4.96791154e-01 -2.01337129e-01 4.57029194e-01 7.44641662e-01 4.01963413e-01 -9.88714635e-01 1.33598578e+00 5.90200472e+00 1.35492444e+00 -6.93693042e-01 2.63051540e-01 5.10824859e-01 -4.45965707e-01 1.79070845e-01 2.96034634e-01 -3.46945077e-01 5.05472124e-01 8.00514877e-01 -4.29542392e-01 1.12317860e+00 8.19056273e-01 1.17615134e-01 -6.00408435e-01 -7.30066299e-01 8.46541226e-01 -4.53341186e-01 -1.64623606e+00 -5.39805412e-01 2.08492741e-01 1.08686256e+00 5.45626163e-01 -1.01123743e-01 8.78760636e-01 1.72041678e+00 -1.18027556e+00 6.23742342e-01 -5.32031385e-03 1.76222071e-01 -9.94623661e-01 8.64122450e-01 7.32453346e-01 -7.40842819e-01 -3.07543516e-01 -4.64619845e-01 -1.03667665e+00 -8.00797790e-02 -2.44704597e-02 -7.32779503e-01 5.20511270e-01 6.31295860e-01 3.10644060e-01 2.55744415e-03 1.08226490e+00 -4.28574085e-01 6.88306987e-01 -6.68454230e-01 -7.31229424e-01 1.10186076e+00 -4.71993983e-01 5.75203836e-01 3.89427274e-01 1.94658652e-01 6.33073211e-01 4.78393495e-01 6.69195056e-01 8.74366425e-03 -2.73367494e-01 -7.25399554e-01 -1.33222684e-01 1.61938921e-01 9.98281896e-01 -5.08464217e-01 -3.46580714e-01 -8.94324556e-02 8.74686241e-01 3.07519883e-01 7.84702525e-02 -1.13548088e+00 -1.49137154e-01 8.84883285e-01 -2.83285707e-01 7.73083121e-02 -4.99352485e-01 -1.05952270e-01 -1.30174363e+00 -3.38351279e-01 -1.30504537e+00 3.88053149e-01 -6.50113344e-01 -1.05230713e+00 3.86869460e-01 -1.54039070e-01 -1.01346195e+00 -7.59141147e-01 -6.78055644e-01 -7.87639141e-01 4.11153615e-01 -9.09748435e-01 -7.67765820e-01 1.90443136e-02 6.00292206e-01 5.37816823e-01 -4.22162980e-01 6.69121563e-01 4.44925502e-02 -5.66286266e-01 2.67007142e-01 6.30701661e-01 9.37236100e-02 -3.06829643e-02 -1.30879247e+00 8.13813984e-01 5.76179266e-01 1.29351288e-01 -4.79038581e-02 7.18834579e-01 -5.61547041e-01 -1.78368819e+00 -3.62431169e-01 6.89079165e-02 -4.25468087e-01 1.16981030e+00 -4.03658092e-01 -1.43559590e-01 5.27400255e-01 4.33965683e-01 -4.25295234e-02 1.24650702e-01 -5.55968173e-02 1.38487563e-01 3.97103906e-01 -8.43045533e-01 1.03319204e+00 1.16946590e+00 -1.56110808e-01 -1.93272069e-01 6.77537203e-01 1.78597137e-01 -8.26216936e-01 -2.50388592e-01 -2.27645189e-01 4.14087683e-01 -1.24766588e+00 6.83757782e-01 -9.20710623e-01 6.51584625e-01 -1.28806651e-01 1.24173820e-01 -1.78410089e+00 -1.42575100e-01 -1.16988003e+00 1.29334360e-01 3.18358749e-01 -2.13327818e-02 -7.73194373e-01 1.18190134e+00 4.01490539e-01 3.13162170e-02 -7.53892601e-01 -1.29215705e+00 -8.60303223e-01 5.93516231e-01 -4.53466982e-01 7.72926152e-01 7.35555172e-01 2.54057944e-01 2.34822854e-02 -9.43253577e-01 -2.45421514e-01 6.08339906e-01 -4.69147302e-02 1.13777971e+00 -8.15260768e-01 -7.12628841e-01 -6.32681251e-01 -4.37999696e-01 -9.61225688e-01 -1.92552097e-02 -6.87250614e-01 1.51714012e-01 -1.47682297e+00 1.34872481e-01 -5.74303985e-01 6.85748309e-02 6.74555838e-01 1.64627098e-02 1.86891347e-01 3.93409461e-01 1.97727233e-02 -8.98313999e-01 7.41210103e-01 1.47741842e+00 -3.37806791e-01 -7.46200606e-02 2.03760743e-01 -2.70803243e-01 5.62060356e-01 1.03986454e+00 -4.20457095e-01 -8.55097547e-02 -3.94962907e-01 5.15828073e-01 4.87458140e-01 2.81213045e-01 -1.37454510e+00 3.93190771e-01 -6.74517453e-01 -1.90162718e-01 -2.40338683e-01 2.93968558e-01 -4.99218434e-01 2.98284978e-01 1.01082802e+00 4.86894092e-03 1.14948720e-01 5.11943072e-04 1.98146746e-01 -9.80390161e-02 -2.40626618e-01 5.41707814e-01 -4.59220946e-01 -8.57967556e-01 4.02150452e-01 -9.58199084e-01 3.69390696e-01 1.39948797e+00 -6.52585402e-02 -6.44190967e-01 -7.25775421e-01 -4.45638150e-01 7.73524702e-01 4.60912198e-01 -2.31122468e-02 4.89293411e-02 -9.71753240e-01 -1.08902919e+00 -1.95660412e-01 -3.16067994e-01 -1.64087251e-01 7.04547942e-01 5.84573090e-01 -1.19646704e+00 2.35567957e-01 -7.37912416e-01 -5.61548918e-02 -7.79480100e-01 3.02372068e-01 6.68232679e-01 -9.09533799e-01 -4.87059534e-01 7.13997841e-01 1.45348124e-02 -7.39047587e-01 -1.16899990e-01 -8.34616199e-02 1.02389053e-01 -1.09844692e-01 4.30821985e-01 5.50086498e-01 -1.15334772e-01 -1.40830457e-01 -2.78908938e-01 -1.17534123e-01 1.31704673e-01 -2.86195904e-01 1.38833940e+00 3.86298925e-01 1.23004988e-01 -2.24128813e-02 3.77334535e-01 -4.32492256e-01 -1.59071314e+00 8.60093236e-02 -3.66513342e-01 -4.25842911e-01 -2.96062659e-02 -9.49502707e-01 -1.05354762e+00 9.18511808e-01 1.93342954e-01 4.80437517e-01 7.78841376e-01 -3.14885169e-01 9.01787043e-01 8.88419330e-01 1.11881387e+00 -1.18176174e+00 4.78696823e-03 1.28003657e+00 5.50350130e-01 -1.08809102e+00 -1.55433014e-01 2.97986686e-01 -9.65658307e-01 8.68914306e-01 7.51941979e-01 -5.49201787e-01 1.26999378e-01 5.24176896e-01 -1.88231856e-01 -2.77942330e-01 -1.12883186e+00 -5.18733203e-01 -7.15889454e-01 5.60694575e-01 -2.50094742e-01 4.32189226e-01 -4.52795446e-01 4.89093333e-01 -4.08201754e-01 3.77387911e-01 8.94203246e-01 1.10595500e+00 -2.56891489e-01 -1.20047021e+00 -2.68992126e-01 3.43815118e-01 -3.64225537e-01 -1.95625350e-02 -3.12242448e-01 9.37756419e-01 9.93648991e-02 7.52557635e-01 7.08903596e-02 -2.21324608e-01 1.89509392e-01 -4.63813841e-01 6.75049245e-01 -2.49638960e-01 -1.06655991e+00 -4.37264353e-01 2.15153918e-01 -8.31298649e-01 1.40875220e-01 -4.52484131e-01 -1.17887211e+00 -1.20602024e+00 1.30798988e-04 4.07783836e-01 3.87789428e-01 9.46863174e-01 1.64737299e-01 3.62125635e-01 5.60742915e-01 -1.26588106e+00 -9.55415845e-01 -6.29774511e-01 -8.34941506e-01 -6.08189665e-02 -1.47427216e-01 -8.24959934e-01 -1.93054363e-01 -8.57478380e-01]
[3.5195438861846924, 1.484937071800232]
f889c07b-f3a4-4da5-bce1-278debc43727
sepp-similarity-estimation-of-predicted
2110.05748
null
https://arxiv.org/abs/2110.05748v2
https://arxiv.org/pdf/2110.05748v2.pdf
SEPP: Similarity Estimation of Predicted Probabilities for Defending and Detecting Adversarial Text
There are two cases describing how a classifier processes input text, namely, misclassification and correct classification. In terms of misclassified texts, a classifier handles the texts with both incorrect predictions and adversarial texts, which are generated to fool the classifier, which is called a victim. Both types are misunderstood by the victim, but they can still be recognized by other classifiers. This induces large gaps in predicted probabilities between the victim and the other classifiers. In contrast, text correctly classified by the victim is often successfully predicted by the others and induces small gaps. In this paper, we propose an ensemble model based on similarity estimation of predicted probabilities (SEPP) to exploit the large gaps in the misclassified predictions in contrast to small gaps in the correct classification. SEPP then corrects the incorrect predictions of the misclassified texts. We demonstrate the resilience of SEPP in defending and detecting adversarial texts through different types of victim classifiers, classification tasks, and adversarial attacks.
['Shinsaku Kiyomoto', 'Kazuhide Fukushima', 'Seira Hidano', 'Hoang-Quoc Nguyen-Son']
2021-10-12
null
null
null
null
['adversarial-text']
['adversarial']
[ 5.97781718e-01 3.71638834e-01 2.71080732e-01 -1.64604068e-01 -4.66354817e-01 -1.01007867e+00 6.98847651e-01 5.72320044e-01 3.19971927e-02 8.07888150e-01 -1.86587617e-01 -3.18028301e-01 1.83541477e-01 -1.09514511e+00 -7.11643517e-01 -7.89910734e-01 1.03639193e-01 5.08076906e-01 4.82298642e-01 -3.29500943e-01 5.59073627e-01 2.24149004e-01 -1.69698131e+00 1.06146169e+00 1.05366135e+00 7.69568205e-01 -3.09762001e-01 8.16046476e-01 -2.54359752e-01 1.26148283e+00 -1.29404318e+00 -9.87278283e-01 2.27432862e-01 -5.02348065e-01 -7.99170613e-01 -4.39655781e-01 3.80130559e-01 -1.59569845e-01 -2.43294835e-01 1.30816901e+00 3.97701003e-02 -1.79213762e-01 9.37536776e-01 -1.77552068e+00 -5.47512650e-01 1.09758675e+00 -2.49977231e-01 -5.22448588e-03 6.24192655e-01 -2.13360786e-02 6.75223947e-01 -6.35511339e-01 2.58352906e-01 1.38341224e+00 7.76560128e-01 8.50188732e-01 -7.95063972e-01 -9.75154757e-01 1.85770586e-01 1.42781585e-01 -9.97360945e-01 -1.61705166e-01 5.54572046e-01 -4.12598699e-01 3.68354082e-01 6.92445993e-01 6.37279600e-02 1.49324203e+00 7.61731923e-01 4.96068984e-01 1.09375453e+00 -2.94875950e-01 2.98905075e-01 6.27239883e-01 4.87156779e-01 2.70761728e-01 2.65983224e-01 4.70692158e-01 -3.32648128e-01 -6.74705029e-01 -4.10676032e-01 2.93721288e-01 -4.76576298e-01 3.16460222e-01 -7.49924242e-01 8.67423832e-01 1.44805714e-01 9.68796015e-02 -1.19461559e-01 -4.42075938e-01 4.62788790e-01 5.91808319e-01 5.65403283e-01 4.19000983e-01 -4.89749849e-01 2.95424342e-01 -6.71349287e-01 1.97016418e-01 1.29433727e+00 7.92319298e-01 4.13750231e-01 -4.19039838e-02 -8.78096521e-02 6.05621874e-01 7.24195614e-02 8.79176676e-01 6.13501489e-01 -3.52313258e-02 6.82409644e-01 7.97123194e-01 1.86418563e-01 -1.49652243e+00 -8.92855301e-02 -4.00908589e-01 -8.01843524e-01 5.08692503e-01 7.04827249e-01 -2.21092142e-02 -7.96220303e-01 1.46340954e+00 1.62633732e-01 1.76531538e-01 5.49045801e-01 6.03128731e-01 4.10810083e-01 6.71961367e-01 3.38544369e-01 -1.70344085e-01 9.95438993e-01 -8.91790926e-01 -6.12253189e-01 -1.99364990e-01 7.31900215e-01 -9.24425006e-01 5.93564689e-01 5.13283491e-01 -8.44273269e-01 -5.38921058e-01 -1.11565733e+00 6.41773522e-01 -8.03460896e-01 -1.44376770e-01 -3.60786855e-01 9.37272906e-01 -2.93532819e-01 9.87634480e-01 -3.35872740e-01 8.61511528e-02 2.88968682e-01 1.97878137e-01 -2.13624179e-01 -3.46849859e-02 -1.64466798e+00 1.00841856e+00 4.70383376e-01 -1.51048809e-01 -8.85570765e-01 -7.39912868e-01 -3.65173787e-01 7.46003091e-02 1.35245934e-01 -7.15350211e-02 9.96049583e-01 -1.45938551e+00 -7.28567898e-01 7.43825614e-01 9.78178307e-02 -6.04047894e-01 1.16878676e+00 -8.54023099e-02 -7.78270423e-01 -1.33526936e-01 8.76749828e-02 -2.58266956e-01 1.34703183e+00 -1.27312911e+00 -1.09361351e+00 -5.90432703e-01 -1.32136032e-01 1.16289537e-02 -3.14210981e-01 -1.42302588e-01 6.88205361e-01 -7.70814002e-01 1.70260578e-01 -6.53589904e-01 1.79205388e-01 -3.65286857e-01 -7.38030910e-01 -1.25614434e-01 1.32154858e+00 -4.89407361e-01 1.36935437e+00 -2.10547876e+00 -2.83428133e-01 3.45331848e-01 4.26344037e-01 5.67559242e-01 1.49425521e-01 5.63938975e-01 -3.10675383e-01 3.57757330e-01 -2.71316767e-01 -4.52818498e-02 -2.00663388e-01 3.04049462e-01 -1.41514182e+00 3.46608847e-01 9.26037580e-02 2.85373926e-01 -1.08351386e+00 5.56686260e-02 -1.30386259e-02 2.10776068e-02 3.49780209e-02 4.53029215e-01 5.96110150e-02 9.78256464e-02 -3.96792084e-01 2.72654831e-01 1.07799160e+00 2.93668509e-01 2.26601541e-01 3.04949079e-02 1.86505079e-01 3.55187505e-01 -1.24997115e+00 -7.99760148e-02 -2.12885216e-01 3.49277347e-01 -7.04968348e-02 -1.16884172e+00 1.13607419e+00 2.59133428e-01 -3.52449924e-01 7.84692839e-02 1.30577505e-01 5.33991456e-01 -4.36725505e-02 -3.54805887e-01 5.01170695e-01 -1.26591578e-01 -1.53146297e-01 8.58060002e-01 -2.94704765e-01 2.27135584e-01 -2.03940019e-01 3.83189410e-01 1.15142882e+00 -4.01476324e-01 2.16610044e-01 -1.00320578e-01 1.02005708e+00 1.90272972e-01 5.46430588e-01 1.34031439e+00 -2.11699113e-01 4.10848141e-01 5.47480106e-01 -7.01308370e-01 -9.94221449e-01 -1.10807812e+00 1.07181082e-02 1.20872796e+00 4.50634569e-01 -2.63139516e-01 -8.37582409e-01 -1.60493541e+00 2.66482621e-01 9.84292984e-01 -7.12685883e-01 -6.83011293e-01 -4.74689811e-01 -5.91324747e-01 1.07899654e+00 4.08245832e-01 3.98207098e-01 -1.09674811e+00 -3.25677067e-01 3.26987386e-01 -1.39832526e-01 -8.13147604e-01 -3.90251309e-01 3.24117392e-01 -4.40561146e-01 -1.60409987e+00 1.31862415e-02 -4.55529749e-01 9.38059747e-01 2.41209865e-01 8.80532265e-01 5.61105192e-01 8.55680332e-02 -7.07297400e-02 -7.52880752e-01 -9.11019504e-01 -1.56450987e+00 -2.23529905e-01 3.47790867e-01 3.29062313e-01 3.25666279e-01 -1.37130827e-01 3.78633365e-02 6.47720337e-01 -9.48718905e-01 -3.15558851e-01 1.84271634e-01 1.00304711e+00 8.36547986e-02 4.32678133e-01 7.79588878e-01 -1.45569873e+00 5.14040232e-01 -9.51502204e-01 -2.64581293e-01 5.53267717e-01 -5.91794789e-01 -4.37737964e-02 1.55454528e+00 -1.00086486e+00 -9.39431190e-01 -3.22093278e-01 -1.09462224e-01 -2.94801712e-01 -4.08062667e-01 1.42722055e-01 -1.87114969e-01 2.81376578e-02 9.00912046e-01 3.77173454e-01 2.40878128e-02 -1.97884127e-01 -1.99855864e-01 1.14048707e+00 3.71451885e-01 -1.79605320e-01 1.19300759e+00 3.14521194e-01 -3.81966591e-01 -4.23761249e-01 -1.19658744e+00 6.54358044e-02 -4.03651714e-01 -2.49420494e-01 2.37174869e-01 -3.95427406e-01 -3.95225763e-01 8.76121700e-01 -1.24586928e+00 1.59102306e-01 -1.66367203e-01 -2.24565268e-02 6.63093105e-02 5.86168110e-01 -3.23174089e-01 -1.17903376e+00 -4.83758271e-01 -1.09395862e+00 6.56319797e-01 2.02485606e-01 -4.23501223e-01 -9.59888101e-01 -3.62591088e-01 2.43195117e-01 3.80577445e-01 2.43023679e-01 1.16096163e+00 -1.88816202e+00 1.13570608e-01 -7.07815289e-01 2.65762299e-01 5.75513005e-01 -8.95399153e-02 3.03685725e-01 -1.12860453e+00 -3.42887878e-01 3.16516221e-01 -1.63551793e-01 5.90647340e-01 -4.16961253e-01 9.79515970e-01 -7.56256819e-01 -6.62322402e-01 1.89723074e-01 8.75404060e-01 5.56253433e-01 5.85169554e-01 1.59829646e-01 4.51074034e-01 8.22789609e-01 6.04421020e-01 1.23227127e-01 -1.01340443e-01 1.63718104e-01 4.67260629e-01 2.19396040e-01 3.74304742e-01 -4.58653271e-01 5.46760023e-01 3.63389701e-01 3.58632982e-01 -5.84223449e-01 -9.66789067e-01 9.53649431e-02 -1.65112746e+00 -1.26944137e+00 -5.28499663e-01 2.37881494e+00 6.97723985e-01 4.19792205e-01 -3.53195965e-01 7.72034526e-01 1.19866633e+00 3.39875259e-02 -4.82002914e-01 -7.38868773e-01 -2.43310899e-01 1.16780631e-01 3.23496670e-01 5.38059652e-01 -1.21961379e+00 6.79507375e-01 5.88098574e+00 1.02230942e+00 -9.32528794e-01 9.68925562e-03 7.20400870e-01 2.39397228e-01 -1.92236960e-01 9.19657946e-02 -9.75730062e-01 9.66446459e-01 1.03294492e+00 -3.85554790e-01 1.61284849e-01 1.14994323e+00 -2.61511177e-01 5.62895946e-02 -1.14066899e+00 3.40072721e-01 3.47428739e-01 -1.10503149e+00 3.22005987e-01 -2.67089397e-01 5.34473002e-01 -5.24743617e-01 1.34293055e-02 5.28786421e-01 5.52034080e-01 -9.31870639e-01 9.50290322e-01 3.44768763e-01 3.08932751e-01 -8.88019502e-01 1.05583084e+00 1.00459325e+00 -8.20170820e-01 -5.78244090e-01 -3.66918892e-01 -1.25393033e-01 -5.49266338e-01 5.49067020e-01 -6.43271923e-01 5.22166133e-01 6.02411270e-01 3.91211748e-01 -4.46061611e-01 4.33855057e-01 -4.24229026e-01 8.74441147e-01 4.66893464e-02 -8.16729665e-02 -6.36320561e-02 7.28723854e-02 9.02067959e-01 9.16137397e-01 2.16898061e-02 1.24608748e-01 3.48351985e-01 8.18189442e-01 -9.50625241e-02 -9.37673170e-03 -6.87141359e-01 2.03405291e-01 8.79806876e-01 9.31774437e-01 -4.54261214e-01 -6.53919518e-01 6.87738359e-02 8.84192824e-01 2.99054682e-01 8.30944553e-02 -8.81957531e-01 -5.72424829e-01 4.76282775e-01 5.34311272e-02 -1.93125859e-01 5.89829266e-01 -4.46860909e-01 -1.02696931e+00 8.53359699e-02 -1.16297209e+00 6.48874164e-01 -4.64333385e-01 -1.92069387e+00 9.87267673e-01 -3.91585469e-01 -1.47121334e+00 5.57934046e-02 -5.52187860e-01 -1.09757340e+00 1.07741845e+00 -1.03929877e+00 -9.44324434e-01 -3.30077350e-01 5.00188231e-01 4.87221211e-01 -4.06313151e-01 7.84182906e-01 -2.52273917e-01 -5.92943370e-01 9.07403827e-01 1.67199463e-01 3.97049457e-01 8.48332465e-01 -1.22832763e+00 2.19159842e-01 8.82422030e-01 -2.81582236e-01 2.82785147e-01 7.56357193e-01 -1.01075983e+00 -6.88192248e-01 -1.44365323e+00 1.25566435e+00 -7.78248310e-01 6.68883264e-01 -3.13378483e-01 -1.37421346e+00 5.18953443e-01 -1.80365533e-01 -3.35545033e-01 7.83575773e-01 -3.75688344e-01 -7.95468211e-01 1.34282380e-01 -1.60990226e+00 4.70320433e-01 5.53873241e-01 -3.59829277e-01 -9.93120193e-01 5.09632826e-01 3.41165870e-01 -3.71417016e-01 -4.92124736e-01 2.73846835e-01 5.12820542e-01 -1.05944085e+00 7.43839264e-01 -9.21355903e-01 6.34346902e-01 -1.44253254e-01 -2.91172247e-02 -1.36336684e+00 3.86219211e-02 -3.45867217e-01 -1.00717269e-01 1.24953306e+00 4.68912244e-01 -1.08459973e+00 4.79030758e-01 6.10546768e-01 7.74876475e-02 -6.20162427e-01 -8.48899007e-01 -7.06133842e-01 5.31696498e-01 -3.09216291e-01 7.06618369e-01 1.25780547e+00 1.38135090e-01 -1.30240500e-01 -4.40037876e-01 5.69252729e-01 7.43759513e-01 2.04380631e-01 5.47989190e-01 -1.31335783e+00 5.39235258e-03 -2.80725956e-01 -3.02398115e-01 -4.62816834e-01 2.48356953e-01 -1.05763018e+00 4.23754416e-02 -6.09371662e-01 -6.29904773e-03 -6.87631369e-01 7.11385086e-02 5.53476512e-01 -7.75035143e-01 2.31130585e-01 3.57089013e-01 5.85612297e-01 -1.02809794e-01 1.28837228e-01 5.24457455e-01 -2.80036241e-01 -3.92054208e-02 4.30976301e-01 -6.48502350e-01 9.52210605e-01 9.55201864e-01 -1.14943659e+00 -8.69905204e-03 1.48518950e-01 -9.28654149e-02 1.66862532e-02 4.20788527e-01 -9.12564814e-01 2.47960612e-01 -1.30131602e-01 5.70995212e-01 -5.96818984e-01 -3.79159451e-01 -8.81337821e-01 7.96998441e-02 9.51186776e-01 -7.94191658e-01 -7.14548603e-02 -1.56708077e-01 8.00654590e-01 -4.23389897e-02 -4.81142968e-01 9.77282584e-01 -8.46381560e-02 1.39460061e-02 3.29099298e-02 -8.51182640e-01 1.28107592e-01 1.39845192e+00 -3.00905645e-01 -8.77282917e-01 -2.36799896e-01 -9.28933620e-01 1.97446719e-01 2.81813323e-01 4.60535794e-01 6.79594517e-01 -9.97692943e-01 -8.84815156e-01 5.97471833e-01 -2.00532116e-02 -5.46390951e-01 2.01022357e-01 5.22267640e-01 -1.51530996e-01 -1.90255210e-01 -4.82267961e-02 -2.10789740e-01 -1.67350841e+00 7.28162706e-01 5.96393526e-01 -4.41004485e-01 -2.48906717e-01 6.43380880e-01 1.42741174e-01 -6.09660387e-01 2.25014716e-01 2.28330359e-01 -2.82884926e-01 3.83609086e-02 9.68484223e-01 5.00207961e-01 1.47602130e-02 -7.49754786e-01 -2.86571473e-01 7.53558651e-02 -5.44793069e-01 4.89252418e-01 7.41501629e-01 1.38620928e-01 -2.06562757e-01 2.81994015e-01 8.34154546e-01 3.13038826e-01 -6.41127110e-01 -1.18896320e-01 1.69730354e-02 -6.01423800e-01 -3.92842591e-01 -1.19595325e+00 -7.79554427e-01 9.52366352e-01 2.63692826e-01 8.86355162e-01 9.73821580e-01 -3.20856303e-01 6.55758798e-01 9.12168473e-02 3.99348468e-01 -8.69874716e-01 -1.95688698e-02 7.35830545e-01 7.94742703e-01 -9.90080953e-01 -4.12647545e-01 -6.88610315e-01 -1.02628481e+00 1.51832545e+00 7.53398478e-01 -1.84195355e-01 4.96090353e-01 3.40287298e-01 3.04985881e-01 3.60103935e-01 -8.93727660e-01 6.06995165e-01 5.49769513e-02 7.27010906e-01 -4.96613905e-02 1.64951727e-01 -2.87169755e-01 9.75270629e-01 -2.95545489e-01 -7.62293994e-01 8.25654149e-01 9.19774413e-01 -5.64200521e-01 -1.01031613e+00 -9.34475839e-01 6.58700705e-01 -5.88621199e-01 -1.05627462e-01 -8.56870234e-01 3.88969272e-01 3.83137256e-01 1.45128787e+00 1.45705089e-01 -1.04148507e+00 3.32276702e-01 4.18002248e-01 -3.07975322e-01 -5.09621322e-01 -1.27183175e+00 -6.67898834e-01 3.79860820e-03 -2.22156450e-01 1.65608436e-01 -4.34753716e-01 -1.15032494e+00 -7.21460402e-01 -6.73861802e-01 4.46374297e-01 2.07479864e-01 1.07610381e+00 2.56351024e-01 3.32027733e-01 1.26668668e+00 -3.14181864e-01 -1.33973610e+00 -9.41379130e-01 -5.43399274e-01 8.93830717e-01 3.57316524e-01 -5.73602855e-01 -1.34086943e+00 4.14988846e-02]
[5.858665466308594, 7.955933570861816]
2d5e476a-efaa-4171-8380-d07c90fede86
soa-nlp-lt-edi-acl2022-an-ensemble-model-for
null
null
https://aclanthology.org/2022.ltedi-1.31
https://aclanthology.org/2022.ltedi-1.31.pdf
SOA_NLP@LT-EDI-ACL2022: An Ensemble Model for Hope Speech Detection from YouTube Comments
Language should be accommodating of equality and diversity as a fundamental aspect of communication. The language of internet users has a big impact on peer users all over the world. On virtual platforms such as Facebook, Twitter, and YouTube, people express their opinions in different languages. People respect others’ accomplishments, pray for their well-being, and cheer them on when they fail. Such motivational remarks are hope speech remarks. Simultaneously, a group of users encourages discrimination against women, people of color, people with disabilities, and other minorities based on gender, race, sexual orientation, and other factors. To recognize hope speech from YouTube comments, the current study offers an ensemble approach that combines a support vector machine, logistic regression, and random forest classifiers. Extensive testing was carried out to discover the best features for the aforementioned classifiers. In the support vector machine and logistic regression classifiers, char-level TF-IDF features were used, whereas in the random forest classifier, word-level features were used. The proposed ensemble model performed significantly well among English, Spanish, Tamil, Malayalam, and Kannada YouTube comments.
['Pradeep Roy', 'Sunil Saumya', 'Abhinav Kumar']
null
null
null
null
ltedi-acl-2022-5
['hope-speech-detection']
['natural-language-processing']
[-5.46092153e-01 1.03359595e-02 -1.01637459e+00 -3.23918879e-01 -1.31158620e-01 -2.33166352e-01 6.13463700e-01 5.00895739e-01 -5.92525721e-01 9.34519351e-01 6.76559210e-01 -4.94950145e-01 -9.58964601e-02 -6.56122267e-01 3.78099263e-01 -4.17501599e-01 2.91815728e-01 -3.54419611e-02 -2.93463141e-01 -5.68970919e-01 9.15737331e-01 2.81397611e-01 -1.31333220e+00 -1.76828783e-02 1.27290833e+00 6.59158349e-01 -5.49871385e-01 2.21729308e-01 -5.28352082e-01 9.85067308e-01 -3.43914121e-01 -1.11271477e+00 -1.73211619e-01 -2.48834595e-01 -4.49253976e-01 -2.07621336e-01 1.65631205e-01 -3.42053145e-01 -1.05110504e-01 1.01825058e+00 5.99449217e-01 -1.21706381e-01 6.81837261e-01 -1.43720651e+00 -7.45554864e-01 7.66460538e-01 -5.66808999e-01 -4.97749858e-02 6.75915003e-01 -4.40540016e-01 6.78320646e-01 -9.31563020e-01 5.69860518e-01 1.31086457e+00 5.94993412e-01 3.61682862e-01 -5.99716425e-01 -1.03147662e+00 -2.99980063e-02 1.12610675e-01 -1.30862308e+00 -5.48318207e-01 7.42229819e-01 -6.11790955e-01 5.07592499e-01 4.90018904e-01 8.54436636e-01 1.01478624e+00 4.65953410e-01 2.95373440e-01 1.20488334e+00 -3.17844093e-01 1.45747110e-01 8.54376733e-01 7.54537340e-03 5.87966442e-01 1.76413879e-01 -5.37961841e-01 -7.21685648e-01 -5.99663973e-01 -1.04131848e-02 2.92536467e-01 -7.19563514e-02 1.47577807e-01 -8.98065031e-01 1.22698438e+00 1.10538863e-01 4.08129156e-01 -1.02465712e-01 -5.95229685e-01 5.09615362e-01 1.30337521e-01 7.80113339e-01 -1.31076365e-03 3.51787470e-02 -6.61666989e-01 -9.04661000e-01 -9.78175923e-02 1.04041708e+00 2.25080147e-01 3.11281830e-01 -4.27056253e-02 9.20486227e-02 1.02540576e+00 4.75556403e-01 4.68255669e-01 8.28852832e-01 -6.11041665e-01 3.50778103e-01 8.60873520e-01 -6.34197146e-02 -1.58785307e+00 -8.83952826e-02 -5.61896682e-01 -9.33831275e-01 -7.20940605e-02 2.62732089e-01 -3.26311320e-01 -2.29576305e-01 1.30405033e+00 2.75302976e-01 -3.64672691e-01 2.85640750e-02 6.41898870e-01 1.07477784e+00 3.44197482e-01 3.49662066e-01 -5.59527457e-01 8.77726018e-01 -5.78009367e-01 -8.82474840e-01 2.40947261e-01 5.73880851e-01 -9.52417433e-01 6.54901803e-01 2.89281577e-01 -9.15819764e-01 -3.11510742e-01 -7.33528495e-01 8.91438723e-02 -3.68557066e-01 2.13716865e-01 6.82654679e-01 1.28533268e+00 -6.58557951e-01 6.36792481e-01 -2.73316413e-01 -5.65596461e-01 5.56289017e-01 1.87577084e-01 -4.35340941e-01 2.69560099e-01 -9.95692134e-01 6.75554514e-01 -1.77819595e-01 -7.85833299e-02 2.46152058e-01 -1.89673588e-01 -7.04570711e-01 -1.47635579e-01 -9.81936529e-02 -1.67779446e-01 5.38152218e-01 -1.16529369e+00 -1.27768075e+00 9.18229520e-01 -2.29997486e-01 -9.35039371e-02 7.49355674e-01 -4.75645484e-03 -8.09091568e-01 -1.91139087e-01 2.56664008e-01 3.24705154e-01 7.74496436e-01 -6.53990686e-01 -6.13798082e-01 -6.62308395e-01 -1.04169389e-02 2.90904492e-01 -8.94487321e-01 4.24394041e-01 3.20841163e-01 -4.23315823e-01 1.56084344e-01 -6.69244528e-01 1.14596106e-01 -1.76751167e-02 -5.14489889e-01 -2.84694046e-01 1.01597512e+00 -8.24375927e-01 1.63116884e+00 -2.25652504e+00 -1.77584976e-01 2.40549773e-01 3.63041282e-01 1.27387166e-01 5.20567834e-01 6.88912570e-01 3.57296079e-01 7.10141957e-01 4.53217387e-01 2.54012883e-01 -1.19865902e-01 1.70590337e-02 7.77636543e-02 4.36837107e-01 -1.86562806e-01 3.04679096e-01 -6.10592008e-01 -7.12809265e-01 -1.47103146e-01 5.10967791e-01 -5.64930856e-01 -3.03451449e-01 5.98370254e-01 2.82216638e-01 -5.71404874e-01 7.75672674e-01 5.89290917e-01 -1.10225730e-01 3.57195705e-01 2.24462658e-01 -6.30312860e-01 1.59584641e-01 -9.01619732e-01 7.06002653e-01 -3.26840967e-01 7.16822088e-01 1.04206473e-01 -8.08923721e-01 1.25333571e+00 6.48089796e-02 3.13076198e-01 -6.04664803e-01 4.11896616e-01 5.46253920e-01 3.37249696e-01 -6.62143767e-01 6.53862715e-01 5.98882660e-02 6.11503944e-02 1.76532388e-01 -4.13570344e-01 2.37879083e-01 8.69561285e-02 3.26441199e-01 1.89846948e-01 -1.48604989e-01 7.23504424e-01 -3.99817705e-01 7.68381298e-01 -4.11130071e-01 5.18682957e-01 2.86864519e-01 -4.54287708e-01 9.04509202e-02 7.35795319e-01 -4.03166026e-01 -6.66414022e-01 -7.04940140e-01 -2.81057209e-01 1.19802594e+00 9.98214707e-02 -2.91225404e-01 -3.53493929e-01 -4.32707071e-01 1.12869874e-01 7.26758957e-01 -2.16676310e-01 8.09332728e-02 -2.96200463e-03 -5.03071845e-01 3.42974722e-01 -2.02465758e-01 8.94883215e-01 -3.75725597e-01 -2.44380713e-01 4.12998982e-02 -1.78078666e-01 -7.60198414e-01 -1.45402819e-01 -5.33734679e-01 -6.89538300e-01 -8.50316048e-01 -9.42846119e-01 -5.15085459e-01 4.63104635e-01 1.62524641e-01 4.40801591e-01 2.48709455e-01 8.63032639e-02 9.90035459e-02 -4.57741082e-01 -5.58844507e-01 -4.12211150e-01 1.45176902e-01 1.16462789e-01 1.48059413e-01 3.10578108e-01 -5.59080064e-01 -3.47527742e-01 4.01524752e-01 -2.77234316e-01 -2.16784934e-03 8.75225104e-03 4.59779084e-01 -3.06020647e-01 -1.07186213e-02 7.29439616e-01 -9.05094683e-01 8.93101275e-01 -1.00428665e+00 3.08805019e-01 5.23441993e-02 -6.69345438e-01 -6.43528700e-01 3.21923971e-01 -3.59161794e-01 -8.71796727e-01 -5.30047476e-01 -3.29966128e-01 5.53542733e-01 2.40409702e-01 5.92070460e-01 4.70789745e-02 -1.28854617e-01 5.26778162e-01 1.16952799e-01 3.84967923e-01 -1.27870113e-01 -2.67875552e-01 1.44333017e+00 -1.36128321e-01 -1.60788119e-01 3.89789104e-01 3.18697333e-01 -3.63847971e-01 -1.30108750e+00 -5.18003762e-01 -3.20103824e-01 -2.76720554e-01 -6.85532093e-01 6.69784963e-01 -9.05163229e-01 -7.15394795e-01 7.33124614e-01 -8.16407681e-01 4.32587534e-01 4.21619654e-01 7.67393827e-01 2.31765613e-01 3.02290142e-01 -2.53702044e-01 -1.23398888e+00 -3.31254333e-01 -7.48076379e-01 3.97661030e-01 7.74711013e-01 -5.69137454e-01 -1.01585543e+00 -4.50467318e-01 6.91190124e-01 6.73514068e-01 3.13629091e-01 8.59294057e-01 -1.07008564e+00 3.36334825e-01 -3.35432231e-01 -3.30320537e-01 1.61219686e-01 2.68453062e-01 4.73542392e-01 -4.56766278e-01 1.85039006e-02 -1.73179224e-01 -2.55040377e-01 3.51008862e-01 1.76114216e-01 8.87742877e-01 -8.29198897e-01 -2.69894063e-01 3.94923724e-02 1.07380986e+00 1.67806149e-01 5.05902708e-01 3.84676605e-01 3.21246415e-01 7.14090526e-01 3.70921910e-01 9.70085979e-01 6.86365068e-01 2.40980700e-01 3.76494288e-01 1.75722405e-01 3.12795162e-01 -3.21631461e-01 4.04078603e-01 7.97092557e-01 -3.89713585e-01 -5.86939715e-02 -8.87877762e-01 3.13599139e-01 -1.43012142e+00 -1.07005107e+00 -3.35881054e-01 2.21875858e+00 4.61939245e-01 2.18294829e-01 4.36259925e-01 5.06126046e-01 7.58572340e-01 2.08102286e-01 -3.25160921e-01 -9.64188218e-01 -2.23097786e-01 -8.38588327e-02 2.53735155e-01 4.39995736e-01 -7.92190015e-01 5.29849648e-01 4.90463543e+00 8.57616663e-01 -1.64060032e+00 -5.08424193e-02 1.17331755e+00 5.02912030e-02 -6.68813527e-01 -8.70212093e-02 -8.34093988e-01 5.04128098e-01 6.72412217e-01 -7.80929625e-01 1.59130305e-01 8.87059093e-01 4.45544064e-01 -3.24094296e-01 -1.96591765e-01 1.09105575e+00 3.94360483e-01 -1.27999711e+00 -3.21237445e-01 2.78546214e-01 5.91473937e-01 -2.24313706e-01 4.08274978e-01 2.53214240e-01 -2.15081573e-01 -9.97807622e-01 4.74516392e-01 3.16732764e-01 5.52718699e-01 -6.30226076e-01 7.06095934e-01 6.26285255e-01 -4.13557678e-01 -2.03717172e-01 -1.51826650e-01 -7.02488303e-01 -7.94921890e-02 6.53429091e-01 -4.22615021e-01 3.14515948e-01 4.72381920e-01 8.28499317e-01 -5.40046573e-01 7.27585912e-01 4.39469935e-03 5.91090202e-01 -2.29565501e-01 -6.46592379e-01 1.55260250e-01 -2.26495758e-01 4.42486167e-01 9.71390963e-01 6.95002437e-01 -8.35172459e-02 -1.49611966e-03 1.02509014e-01 2.66274065e-02 1.07435596e+00 -6.41289771e-01 -4.73538816e-01 4.86485362e-01 1.29102576e+00 -6.48354173e-01 -8.72753784e-02 -4.85131770e-01 4.98936743e-01 7.96390548e-02 6.22506291e-02 -5.96854687e-01 -2.45331869e-01 5.54384291e-01 6.21417880e-01 -4.14749533e-01 -2.15813175e-01 -3.96479905e-01 -1.13285172e+00 -3.62817556e-01 -8.77990842e-01 1.70939967e-01 -2.21882924e-01 -1.04068887e+00 3.60259354e-01 -4.05194759e-01 -1.15828156e+00 -1.98161513e-01 -3.94905746e-01 -7.64126480e-01 7.86194921e-01 -1.09297228e+00 -7.36462951e-01 -1.27025783e-01 2.86521107e-01 2.81200707e-01 -6.36859417e-01 6.37818217e-01 3.05955619e-01 -7.11557448e-01 6.45611525e-01 1.72218680e-01 2.02450037e-01 6.46844089e-01 -4.46871012e-01 -5.28378427e-01 2.02583984e-01 -1.06334880e-01 6.36451662e-01 6.23576164e-01 -6.68919683e-01 -9.92863417e-01 -3.73370618e-01 1.77588344e+00 3.61909837e-01 6.84240282e-01 -3.42395413e-03 -2.84285635e-01 1.09941624e-01 3.19461137e-01 -3.18931043e-01 1.37731922e+00 3.94160390e-01 -1.75403118e-01 1.14053287e-01 -1.47008348e+00 7.61585951e-01 6.53018236e-01 -3.64255637e-01 -6.01842478e-02 4.37743932e-01 -6.27507716e-02 -1.75197005e-01 -8.43503714e-01 -9.36696455e-02 1.08518422e+00 -1.49702775e+00 5.32206178e-01 -7.13290691e-01 8.48738194e-01 4.86487627e-01 -2.09114805e-01 -1.03384614e+00 -2.27682907e-02 -5.16116083e-01 4.30402398e-01 1.53161514e+00 5.02748311e-01 -9.52649832e-01 8.18231761e-01 8.96659672e-01 4.52158540e-01 -9.23259854e-01 -1.00445235e+00 -2.38742188e-01 2.14285418e-01 -2.47385189e-01 3.74996752e-01 1.26380110e+00 3.96071315e-01 2.63697773e-01 -5.75826287e-01 -4.03220534e-01 3.16724300e-01 -1.48541972e-01 8.25272560e-01 -1.36945426e+00 1.00316934e-01 -6.05440974e-01 -4.83781636e-01 -4.50544924e-01 1.69539556e-01 -7.21662819e-01 -1.05617809e+00 -1.38675535e+00 3.07628512e-01 -4.96246457e-01 1.88263401e-01 1.68497473e-01 2.30672378e-02 9.94599089e-02 3.20348531e-01 2.69791454e-01 -1.46221995e-01 3.38726074e-01 1.13934302e+00 -1.65591627e-01 -3.62260669e-01 4.20635790e-01 -1.22394609e+00 8.89131606e-01 1.04645860e+00 -1.98486537e-01 -2.06008747e-01 1.08371869e-01 6.98809087e-01 2.32283324e-01 9.62009355e-02 -7.45435894e-01 8.71715024e-02 -5.14455795e-01 3.48167539e-01 -2.31379896e-01 3.39551449e-01 -5.33808231e-01 -2.57321205e-02 7.34918654e-01 -3.08382958e-01 -1.23350359e-01 -2.36642987e-01 1.63505902e-03 -4.05748039e-01 -1.85282990e-01 5.08657038e-01 1.94166496e-01 -2.38752276e-01 1.23282649e-01 -4.11438912e-01 3.68979461e-02 1.07657206e+00 -6.84396088e-01 -4.43988174e-01 -1.07062793e+00 -6.49710894e-01 1.00302763e-01 3.70775729e-01 5.99074185e-01 5.71363509e-01 -1.14166260e+00 -1.10100853e+00 1.77683339e-01 -3.86758447e-02 -9.37519908e-01 2.34747946e-01 1.11853898e+00 -3.40899318e-01 2.11218875e-02 -4.69354659e-01 -1.14126384e-01 -1.54983401e+00 -1.41882941e-01 -1.46967798e-01 7.49052316e-02 1.47439595e-02 6.73270226e-01 -5.79793096e-01 -3.60693485e-01 1.14628457e-01 4.27723497e-01 -8.07934105e-01 6.59606278e-01 4.03021246e-01 8.41230989e-01 -3.20085704e-01 -1.16836774e+00 -4.09618616e-01 3.55647534e-01 -9.80905294e-02 -4.13269885e-02 1.06776869e+00 -1.54232800e-01 -3.61569345e-01 3.58782947e-01 1.17562985e+00 8.89209986e-01 -1.16249688e-01 2.21819773e-01 -2.05014274e-01 -9.32391226e-01 -1.31847829e-01 -4.54001516e-01 -7.83135295e-01 7.26511657e-01 1.89125299e-01 5.75559556e-01 8.23245645e-01 -3.56355548e-01 4.99802142e-01 -1.94505900e-01 3.06073278e-01 -1.25101686e+00 -2.41163164e-01 4.04855341e-01 8.11709702e-01 -1.42549491e+00 2.84633607e-01 -4.66621161e-01 -8.40828776e-01 1.27618122e+00 4.05902028e-01 1.68264821e-01 1.04972303e+00 -2.03497380e-01 1.78885326e-01 4.98564154e-01 -4.94566888e-01 2.23029092e-01 2.28790250e-02 5.71075022e-01 1.03334963e+00 2.07791939e-01 -1.38583493e+00 6.46364808e-01 -5.31070411e-01 3.18151116e-02 7.01339841e-01 4.15167093e-01 -6.61002755e-01 -1.17268920e+00 -2.06450447e-01 6.95387244e-01 -9.79469121e-01 -6.63464293e-02 -5.38806140e-01 6.95690989e-01 2.35610068e-01 1.25693941e+00 -3.07527632e-02 -4.59984541e-01 -2.35455200e-01 -7.62232067e-03 -2.32332632e-01 -1.45152763e-01 -7.26355672e-01 -9.16861147e-02 5.84422708e-01 -1.70828141e-02 -4.40187544e-01 -7.69133329e-01 -1.04362810e+00 -8.78595769e-01 -3.68117616e-02 3.97124380e-01 1.06316984e+00 8.41575742e-01 1.65100425e-01 -2.78162092e-01 9.86987054e-01 -2.54847944e-01 -4.42983210e-01 -8.58407915e-01 -5.53894281e-01 -6.22813404e-03 -9.25979689e-02 -3.99472654e-01 -4.71603662e-01 -6.21940613e-01]
[8.910080909729004, 10.671156883239746]
5f38c9bd-daa4-4f04-81a8-fded043da9c7
exploring-anisotropy-and-outliers-in
2306.00458
null
https://arxiv.org/abs/2306.00458v2
https://arxiv.org/pdf/2306.00458v2.pdf
Exploring Anisotropy and Outliers in Multilingual Language Models for Cross-Lingual Semantic Sentence Similarity
Previous work has shown that the representations output by contextual language models are more anisotropic than static type embeddings, and typically display outlier dimensions. This seems to be true for both monolingual and multilingual models, although much less work has been done on the multilingual context. Why these outliers occur and how they affect the representations is still an active area of research. We investigate outlier dimensions and their relationship to anisotropy in multiple pre-trained multilingual language models. We focus on cross-lingual semantic similarity tasks, as these are natural tasks for evaluating multilingual representations. Specifically, we examine sentence representations. Sentence transformers which are fine-tuned on parallel resources (that are not always available) perform better on this task, and we show that their representations are more isotropic. However, we aim to improve multilingual representations in general. We investigate how much of the performance difference can be made up by only transforming the embedding space without fine-tuning, and visualise the resulting spaces. We test different operations: Removing individual outlier dimensions, cluster-based isotropy enhancement, and ZCA whitening. We publish our code for reproducibility.
['Alexander Fraser', 'Jindřich Libovický', 'Alina Fastowski', 'Katharina Hämmerl']
2023-06-01
null
null
null
null
['semantic-textual-similarity', 'semantic-similarity']
['natural-language-processing', 'natural-language-processing']
[-2.42291495e-01 1.12192713e-01 -4.18872833e-02 -3.58831644e-01 -7.80013084e-01 -7.68238842e-01 1.03180099e+00 3.26229572e-01 -7.00616479e-01 4.30329114e-01 9.43094909e-01 -5.32082438e-01 1.91755164e-02 -5.16565442e-01 -6.43142223e-01 -6.41110599e-01 1.12256585e-02 4.67995286e-01 9.57987383e-02 -3.50439221e-01 2.40028724e-01 2.62082100e-01 -1.19503284e+00 2.93516397e-01 9.20499206e-01 1.38488606e-01 6.95241243e-03 6.11406088e-01 -2.98741609e-01 3.94901603e-01 -5.53235769e-01 -5.27352452e-01 1.00035168e-01 -3.65021735e-01 -9.58231807e-01 -2.65423030e-01 7.21440792e-01 2.48464301e-01 4.68207672e-02 1.15350413e+00 5.26911080e-01 -3.29164751e-02 9.39271390e-01 -7.92935371e-01 -1.14714086e+00 9.13036823e-01 -4.33750600e-01 4.94678468e-01 2.74032027e-01 -1.05632260e-01 1.18832636e+00 -9.60821211e-01 8.60815585e-01 1.58088565e+00 7.31247962e-01 1.68181106e-01 -1.53332508e+00 -3.19750577e-01 3.15852761e-01 2.08731458e-01 -1.29822671e+00 -5.35630107e-01 6.38063848e-01 -5.45999467e-01 1.18393242e+00 1.91954657e-01 3.28673095e-01 1.37217975e+00 2.55379945e-01 4.56408679e-01 1.39157808e+00 -6.50673091e-01 -1.07434891e-01 4.26715404e-01 2.86890358e-01 3.18744630e-01 3.95411730e-01 -1.39718786e-01 -2.13821545e-01 -1.48275018e-01 5.77581584e-01 -3.07712078e-01 -2.60700822e-01 -6.31308436e-01 -1.65188384e+00 1.13211858e+00 4.06392455e-01 9.42705870e-01 -3.01370412e-01 -9.60031599e-02 5.66698730e-01 6.29499435e-01 7.56062031e-01 8.34967852e-01 -5.17198265e-01 -2.79978186e-01 -8.17811430e-01 8.19444656e-02 6.15827620e-01 5.57841122e-01 6.45825803e-01 7.98580796e-03 1.88566029e-01 1.24223137e+00 4.30972613e-02 4.18028057e-01 9.08864260e-01 -7.20220923e-01 4.89803076e-01 3.01761717e-01 -3.39341640e-01 -1.13866174e+00 -5.41492581e-01 -3.38822633e-01 -5.95173240e-01 4.81876284e-02 5.71364880e-01 3.44706625e-02 -5.42992532e-01 1.89099348e+00 -5.40392585e-02 -2.52478510e-01 3.27553004e-01 8.46017361e-01 4.08964217e-01 5.30241847e-01 2.31141448e-01 1.91310540e-01 1.36210918e+00 -8.26301754e-01 -5.76074243e-01 -3.18678588e-01 1.08335102e+00 -1.29874563e+00 1.48022163e+00 5.00661926e-03 -9.54521239e-01 -4.49834228e-01 -1.15682757e+00 -3.74597698e-01 -7.63668001e-01 -4.35170271e-02 6.55053079e-01 7.73489535e-01 -1.08716738e+00 5.93186080e-01 -9.42244172e-01 -8.52121413e-01 -1.52834594e-01 6.85820878e-02 -6.75731838e-01 -1.79019906e-02 -1.25688636e+00 1.50345623e+00 2.25328088e-01 -2.13351578e-01 -8.68078601e-03 -5.96999764e-01 -1.06202781e+00 -1.59933567e-01 -8.00653696e-02 -5.42539954e-01 8.65194798e-01 -1.04554427e+00 -1.06866837e+00 1.01088727e+00 -1.88902393e-01 -2.30282888e-01 2.82515049e-01 1.05663696e-02 -4.70327049e-01 -2.79837161e-01 1.84449092e-01 6.42854333e-01 4.90065485e-01 -1.14571977e+00 -2.44370431e-01 -5.90387940e-01 1.13222815e-01 2.79475808e-01 -5.41380465e-01 3.67579967e-01 -1.13834515e-01 -1.11928952e+00 2.68916756e-01 -1.14247036e+00 -7.69413561e-02 -4.62990612e-01 -5.33240885e-02 -2.63534904e-01 4.26298350e-01 -8.97305489e-01 1.15733075e+00 -2.08251476e+00 2.30734527e-01 1.91865176e-01 -1.53061122e-01 -9.87481624e-02 -4.37322646e-01 4.66216356e-01 -2.78280735e-01 5.62766850e-01 -1.72003791e-01 -2.68356174e-01 2.39444092e-01 4.11006093e-01 -3.12803417e-01 5.50084054e-01 1.65974289e-01 7.41214335e-01 -7.48986721e-01 -4.01665956e-01 1.17457271e-01 7.87286401e-01 -8.27760100e-01 -3.61658663e-01 2.51025885e-01 2.04579636e-01 9.22127143e-02 3.16005260e-01 5.58275700e-01 2.89764673e-01 3.20271462e-01 -2.01476842e-01 -1.92470253e-01 6.34588301e-01 -1.10401344e+00 1.74811614e+00 -9.01417732e-01 8.93639922e-01 -1.14692040e-01 -7.84724474e-01 8.18473339e-01 2.25467473e-01 8.45174678e-03 -8.30406547e-01 -2.10067123e-01 5.60169160e-01 4.39569682e-01 -3.82227242e-01 8.63162577e-01 -3.16199005e-01 -1.29198343e-01 6.96571946e-01 1.36734799e-01 -8.43456537e-02 2.44517684e-01 9.25886706e-02 7.72679448e-01 2.25250274e-02 1.20122448e-01 -6.44980967e-01 3.26945186e-01 -1.66756615e-01 2.90146172e-01 4.86200929e-01 -6.19474761e-02 7.22912133e-01 5.95667481e-01 -3.28277797e-01 -1.33705080e+00 -1.18207192e+00 -5.49927533e-01 1.29546630e+00 -2.17293769e-01 -5.73686779e-01 -6.82011306e-01 -5.96488297e-01 -1.08573727e-01 8.84179592e-01 -6.39725566e-01 -7.05505610e-02 -7.11445928e-01 -1.05893207e+00 6.12374187e-01 6.45530343e-01 -1.72408894e-01 -9.86627758e-01 -2.08354443e-01 9.71401930e-02 -7.17115849e-02 -9.86570895e-01 -4.97300714e-01 2.19623432e-01 -9.11407173e-01 -8.92391980e-01 -7.55739987e-01 -7.73686528e-01 5.09023964e-01 -3.53380516e-02 1.26399457e+00 -1.57798931e-01 -1.91036239e-02 3.75649929e-01 -3.57042432e-01 -1.89792097e-01 -6.41380787e-01 4.23834682e-01 3.10069799e-01 -3.92777234e-01 7.02837765e-01 -5.95752537e-01 -2.61169523e-01 1.99729651e-01 -9.73072290e-01 -3.78113061e-01 5.69520235e-01 8.51758540e-01 3.10421497e-01 -2.37791210e-01 3.83850247e-01 -9.71541703e-01 1.24728262e+00 -3.92354906e-01 -1.11822449e-01 1.65773153e-01 -5.63458681e-01 5.32567978e-01 5.04843414e-01 -3.61580223e-01 -7.81971216e-01 -5.01477480e-01 -2.48395190e-01 -2.42668856e-02 -2.98817933e-01 5.28698444e-01 5.46556674e-02 2.08188832e-01 8.86229277e-01 -2.43935913e-01 -6.05917815e-03 -6.58017039e-01 6.85360551e-01 7.54419088e-01 2.57168680e-01 -9.32866335e-01 5.71599007e-01 1.98668838e-01 -5.97599924e-01 -1.08212137e+00 -3.92414600e-01 -3.15611452e-01 -7.80764043e-01 3.22265953e-01 9.80700850e-01 -7.44455636e-01 9.38327014e-02 -1.13064051e-01 -1.26520872e+00 -1.38285741e-01 -2.51732498e-01 8.24476242e-01 -3.05933386e-01 4.45130676e-01 -5.72170854e-01 -3.25183094e-01 1.95916779e-02 -1.54092503e+00 9.59250748e-01 -3.47328901e-01 -7.16845334e-01 -1.56450927e+00 4.84137625e-01 1.01827487e-01 7.79892623e-01 -1.13294333e-01 1.36034262e+00 -9.24137831e-01 6.68395730e-03 3.31884660e-02 -9.09302160e-02 3.85025561e-01 6.14283308e-02 5.27301133e-02 -1.08123195e+00 -3.82040113e-01 -9.73436609e-02 -2.09756065e-02 8.99714053e-01 2.17947096e-01 6.98479533e-01 -1.58496127e-01 -2.35871579e-02 4.69657034e-01 1.14320230e+00 -3.09403569e-01 4.04436082e-01 7.44596779e-01 6.52420998e-01 8.76154006e-01 5.67885190e-02 -2.71210194e-01 6.21411502e-01 9.33394194e-01 -6.85704723e-02 -1.32730946e-01 -2.05516711e-01 -1.11899890e-01 6.78957164e-01 1.54373670e+00 -1.39811590e-01 2.79222786e-01 -1.09441864e+00 8.14558864e-01 -1.57215166e+00 -9.32304978e-01 -2.05769196e-01 2.22037482e+00 7.02915668e-01 6.03864565e-02 1.23906210e-01 9.93340164e-02 5.49680054e-01 4.23800260e-01 1.06181763e-02 -1.10044563e+00 -3.52392226e-01 2.49341428e-01 4.95325804e-01 7.50838041e-01 -1.06220222e+00 1.04435337e+00 6.52222919e+00 5.17679513e-01 -1.26313412e+00 2.79809803e-01 3.75683665e-01 5.49324118e-02 -7.27436960e-01 6.05651625e-02 -4.04310703e-01 2.18307897e-01 1.15058184e+00 -5.59554994e-02 2.83363402e-01 6.94492280e-01 4.11960296e-02 1.88104406e-01 -1.20990944e+00 7.58103788e-01 3.58871609e-01 -7.91750431e-01 2.10831501e-02 1.40881374e-01 6.92684472e-01 3.76767486e-01 1.27523661e-01 5.65622091e-01 3.06365818e-01 -1.08068526e+00 6.22494996e-01 3.20046932e-01 3.38739544e-01 -7.63717353e-01 8.90167236e-01 -1.50649726e-01 -9.89661574e-01 1.80789977e-01 -7.07909942e-01 1.46974683e-01 4.16762941e-03 5.43118715e-01 -5.90831935e-01 4.36290443e-01 6.69250846e-01 6.53209269e-01 -1.15464151e+00 8.12439263e-01 -1.82437927e-01 5.48841059e-01 -2.95783907e-01 1.37081489e-01 2.80734956e-01 -4.31472182e-01 4.90675360e-01 1.51918530e+00 4.31195289e-01 -8.12677205e-01 -3.84969413e-02 6.75284624e-01 1.47393152e-01 6.40283525e-01 -8.22146952e-01 -1.50764465e-01 1.88459545e-01 1.00465727e+00 -6.00026369e-01 -1.58530384e-01 -6.54938281e-01 9.47127759e-01 5.06581128e-01 3.55491549e-01 -5.72997928e-01 -3.34289998e-01 9.13125277e-01 6.12246618e-02 1.17231645e-01 -3.90933543e-01 -2.68224388e-01 -1.39981055e+00 5.12583852e-02 -1.06846857e+00 3.09283793e-01 -7.02747762e-01 -1.54845917e+00 7.33573258e-01 -2.28621345e-02 -7.83789754e-01 -4.92310435e-01 -8.12267900e-01 -4.82358396e-01 1.03800845e+00 -1.36488974e+00 -9.98853445e-01 2.82063633e-01 3.87182683e-01 3.62188369e-01 -1.32889181e-01 1.13201702e+00 4.26401258e-01 -3.49981189e-01 6.81557596e-01 1.82239801e-01 1.16827160e-01 1.13311064e+00 -1.45709050e+00 5.29109120e-01 7.77757883e-01 6.29609525e-01 1.05413413e+00 8.07448447e-01 -2.81289250e-01 -1.02409732e+00 -8.00538540e-01 1.29882598e+00 -7.50651121e-01 1.16171670e+00 -3.82694572e-01 -1.15250695e+00 9.01426077e-01 6.06991708e-01 -9.20545161e-02 7.56075203e-01 7.81109273e-01 -7.28117764e-01 2.59921163e-01 -6.19082808e-01 8.65086079e-01 9.22611356e-01 -9.40493703e-01 -9.27157998e-01 2.49167502e-01 6.46830738e-01 2.94582509e-02 -1.08381486e+00 1.89035907e-01 4.99371529e-01 -9.88580644e-01 1.06305993e+00 -6.43174291e-01 2.23499373e-01 -1.64604977e-01 -2.82869548e-01 -1.90058100e+00 -3.07124048e-01 -3.94158438e-02 5.91795504e-01 1.38342118e+00 8.13654125e-01 -9.86079395e-01 3.04978698e-01 1.98368609e-01 -1.27743185e-01 -4.64467674e-01 -8.30862522e-01 -9.86166239e-01 9.27590489e-01 -5.03114402e-01 6.52996898e-01 1.41509223e+00 1.25612557e-01 4.30713743e-01 1.00915574e-01 -6.67213202e-02 2.34655619e-01 -5.74980602e-02 5.80923855e-01 -1.05604208e+00 -1.35340825e-01 -9.34045732e-01 -5.55216372e-01 -5.71909130e-01 3.94846141e-01 -1.40165532e+00 -3.23569834e-01 -1.41146505e+00 4.40255478e-02 -4.86626714e-01 -1.93532646e-01 3.05561155e-01 -2.26058751e-01 2.64803678e-01 3.01045030e-01 2.91422635e-01 -2.26602033e-01 3.16290379e-01 8.83984208e-01 -3.39350514e-02 -2.33645272e-02 -4.79109347e-01 -7.99185216e-01 8.07247281e-01 1.00682151e+00 -3.96755576e-01 -2.63674706e-01 -8.61928403e-01 2.93838680e-01 -6.99980319e-01 1.44011974e-01 -9.04875755e-01 -1.83458224e-01 1.93117559e-01 2.63596088e-01 5.55283390e-03 1.92685440e-01 -7.12186038e-01 -1.33524105e-01 5.63878603e-02 -5.10756075e-01 7.72338569e-01 1.01760626e-01 9.96424928e-02 -4.52109277e-01 -3.10462475e-01 5.62702954e-01 -9.50891972e-02 -4.09332424e-01 -1.83271140e-01 -3.89968663e-01 2.00477391e-01 6.19071245e-01 6.56540319e-03 -3.65499467e-01 -1.99896887e-01 -8.07165086e-01 -2.03972936e-01 8.76608253e-01 8.97520125e-01 3.74060348e-02 -1.59010172e+00 -8.26013625e-01 2.80765563e-01 2.90069640e-01 -6.57526016e-01 -1.35837913e-01 1.07423949e+00 -4.33367491e-01 5.45147002e-01 -1.09648660e-01 -5.99295259e-01 -1.04416704e+00 4.96029407e-01 2.72278279e-01 -2.50622123e-01 -5.56576252e-01 4.69806761e-01 1.27623320e-01 -1.00722778e+00 -5.05753653e-03 -5.29666781e-01 -3.33106846e-01 4.76292342e-01 2.22368151e-01 1.82746857e-01 3.35770428e-01 -1.07685101e+00 -3.95419061e-01 8.23265016e-01 -1.87889323e-01 -3.33779246e-01 1.29612410e+00 1.03041867e-03 -2.28244945e-01 6.89977527e-01 1.52888215e+00 4.77792442e-01 -4.85552430e-01 -8.26889724e-02 2.79555678e-01 -2.45156258e-01 -7.81403016e-03 -3.79114389e-01 -8.84430885e-01 1.15840375e+00 7.25629151e-01 4.36680555e-01 5.20606339e-01 8.82913694e-02 4.15452868e-01 1.60803631e-01 6.99999854e-02 -1.17808628e+00 -3.34044665e-01 7.45716989e-01 1.10948396e+00 -1.17566657e+00 -3.92325781e-02 5.76558374e-02 -6.70930624e-01 1.06449664e+00 2.96931714e-01 -2.89686590e-01 5.76995015e-01 1.41843081e-01 4.06828254e-01 -1.69309571e-01 -4.91102427e-01 -1.86613545e-01 2.32012644e-01 4.77280766e-01 1.20858383e+00 2.17521504e-01 -5.82180619e-01 3.11841935e-01 -8.49744856e-01 -6.23947263e-01 4.15121794e-01 7.34211743e-01 -1.15156256e-01 -1.46330059e+00 -7.35949874e-01 3.11300635e-01 -5.93293071e-01 -4.43826556e-01 -4.71679479e-01 1.01000202e+00 -2.78464030e-03 7.28388548e-01 3.17627013e-01 -1.48461536e-01 4.46364939e-01 4.60830003e-01 5.52194238e-01 -6.57959998e-01 -4.96541649e-01 3.48546566e-03 2.24165931e-01 -3.22532326e-01 -3.79720360e-01 -9.00431275e-01 -9.19570804e-01 -3.14545602e-01 -1.43669978e-01 2.33919978e-01 1.02318943e+00 7.79026806e-01 2.82965004e-01 4.88129199e-01 1.17034189e-01 -8.64513040e-01 -3.98236185e-01 -1.26103139e+00 -2.07574457e-01 7.54415214e-01 6.90751746e-02 -6.55670583e-01 -6.96665406e-01 -1.77659020e-01]
[10.861401557922363, 9.865848541259766]
16dbcf51-bb73-45fe-91b5-0114684fd4aa
rada-robust-adversarial-data-augmentation-for
2112.02469
null
https://arxiv.org/abs/2112.02469v1
https://arxiv.org/pdf/2112.02469v1.pdf
RADA: Robust Adversarial Data Augmentation for Camera Localization in Challenging Weather
Camera localization is a fundamental and crucial problem for many robotic applications. In recent years, using deep-learning for camera-based localization has become a popular research direction. However, they lack robustness to large domain shifts, which can be caused by seasonal or illumination changes between training and testing data sets. Data augmentation is an attractive approach to tackle this problem, as it does not require additional data to be provided. However, existing augmentation methods blindly perturb all pixels and therefore cannot achieve satisfactory performance. To overcome this issue, we proposed RADA, a system whose aim is to concentrate on perturbing the geometrically informative parts of the image. As a result, it learns to generate minimal image perturbations that are still capable of perplexing the network. We show that when these examples are utilized as augmentation, it greatly improves robustness. We show that our method outperforms previous augmentation techniques and achieves up to two times higher accuracy than the SOTA localization models (e.g., AtLoc and MapNet) when tested on `unseen' challenging weather conditions.
['Andrew Markham', 'Niki Trigon', 'Chris Xiaoxuan Lu', 'Muhamad Risqi U. Saputra', 'Jialu Wang']
2021-12-05
null
null
null
null
['camera-localization']
['computer-vision']
[-2.21416652e-02 -2.21624196e-01 -1.39476508e-02 -1.67747006e-01 -6.23562276e-01 -8.37221682e-01 7.98782647e-01 -1.65615916e-01 -7.08637476e-01 7.31928289e-01 6.49176538e-02 -1.76107988e-01 3.77648622e-01 -4.63831335e-01 -1.06187940e+00 -9.54237461e-01 2.52057910e-01 2.59965926e-01 2.95912683e-01 -2.65554130e-01 1.18877485e-01 7.18028724e-01 -1.36142111e+00 -2.42734447e-01 8.48442197e-01 5.64787745e-01 2.90497631e-01 5.18847466e-01 -1.27580175e-02 6.86584175e-01 -7.49968171e-01 -4.87509221e-02 3.45407665e-01 -3.43397290e-01 -5.24329960e-01 2.21002072e-01 6.58721328e-01 -2.82708734e-01 -5.62976420e-01 1.29365480e+00 4.92208302e-01 2.46843874e-01 4.42395926e-01 -1.24110913e+00 -5.50405383e-01 2.88805932e-01 -6.61794722e-01 1.87300414e-01 1.85976461e-01 1.25331476e-01 4.50606287e-01 -8.09139192e-01 4.06527460e-01 1.11589324e+00 8.29554856e-01 6.65061355e-01 -1.35645795e+00 -6.05380237e-01 3.11766922e-01 1.87208056e-01 -1.42687583e+00 -5.19042373e-01 9.16165650e-01 -3.18130493e-01 5.71407199e-01 3.20268311e-02 2.66734779e-01 1.32572722e+00 -6.84638321e-02 8.57752979e-01 1.01407802e+00 -4.96938527e-01 3.97527665e-01 1.74741641e-01 -2.95565546e-01 4.21634883e-01 2.78945535e-01 4.93873917e-02 -3.71892691e-01 3.21313702e-02 8.49147081e-01 -3.33289579e-02 -6.05296791e-01 -7.17867434e-01 -1.51761675e+00 6.36738896e-01 8.24259639e-01 2.97916204e-01 -4.18759942e-01 3.63694787e-01 1.31260589e-01 7.46821761e-02 3.54843497e-01 8.05107713e-01 -4.67561334e-01 7.27754906e-02 -7.72693872e-01 9.07037854e-02 5.11248887e-01 1.01466155e+00 8.57074142e-01 2.98153281e-01 2.92737424e-01 7.02493072e-01 4.89861928e-02 8.57486665e-01 6.64943457e-01 -8.57502937e-01 4.21426952e-01 3.86678398e-01 4.41201538e-01 -1.15461433e+00 -5.68267763e-01 -4.87267256e-01 -1.08665061e+00 2.38580272e-01 5.76057911e-01 -2.90251732e-01 -1.25822055e+00 1.99133623e+00 2.04396620e-01 3.20030957e-01 2.59319603e-01 1.08469415e+00 5.22903502e-01 6.07483387e-01 -1.48394763e-01 6.42283484e-02 1.03250837e+00 -1.04541683e+00 -8.26146245e-01 -8.51470828e-01 6.47200406e-01 -7.77978718e-01 1.11431134e+00 2.63250560e-01 -5.98721623e-01 -5.94934344e-01 -1.16899705e+00 2.40750417e-01 -6.23694241e-01 3.19782913e-01 5.22862554e-01 4.76485401e-01 -1.01491308e+00 2.43528605e-01 -9.45749402e-01 -8.12413037e-01 2.94127941e-01 3.19291085e-01 -7.63718247e-01 -2.76931375e-01 -1.01713693e+00 1.12418747e+00 3.57787549e-01 1.59523949e-01 -8.40114832e-01 -4.57565337e-01 -1.01293099e+00 -2.05357030e-01 4.00280684e-01 -3.00332785e-01 1.26382911e+00 -9.47103083e-01 -1.38321090e+00 6.15974545e-01 -1.19873643e-01 -5.93837321e-01 5.30155122e-01 -4.75802004e-01 -2.90159225e-01 2.45942418e-02 1.17005996e-01 1.00462782e+00 9.38655436e-01 -1.44235826e+00 -4.87614781e-01 -2.25376159e-01 8.41960777e-03 2.78312266e-01 -4.46928829e-01 -2.82495111e-01 -7.81396806e-01 -8.90032232e-01 3.61486673e-01 -1.12357056e+00 -5.03519058e-01 -1.77526772e-02 -4.35539573e-01 3.66953552e-01 9.99803483e-01 -4.99711871e-01 6.58911645e-01 -2.22744036e+00 1.35835201e-01 8.00550357e-02 -1.25106215e-01 5.43774664e-01 -3.07155073e-01 1.52744129e-01 -1.12061143e-01 4.17489633e-02 -3.97973686e-01 -3.37641299e-01 -2.26993367e-01 3.84777218e-01 -2.88445920e-01 7.76834607e-01 2.67253190e-01 6.84841216e-01 -7.75935650e-01 -4.54128571e-02 5.77713251e-01 4.29952741e-01 -4.24561739e-01 2.34911472e-01 -3.38347554e-01 9.47737038e-01 5.93198929e-03 5.23183346e-01 8.49766135e-01 6.42387941e-02 -1.37654111e-01 -1.39630631e-01 7.24029392e-02 -7.52512440e-02 -1.19868875e+00 1.77960229e+00 -4.85332876e-01 8.96376729e-01 4.59023453e-02 -1.00951958e+00 8.23010325e-01 6.68061376e-02 1.92622110e-01 -7.40851045e-01 1.39437184e-01 1.65730059e-01 -1.60077140e-02 -3.59946400e-01 4.21116233e-01 1.89996183e-01 -1.84380874e-01 5.42156734e-02 -3.94799002e-02 -2.40464270e-01 3.16284597e-02 6.58286959e-02 1.04223788e+00 7.39112720e-02 1.83009416e-01 -5.76863736e-02 5.41651607e-01 2.44453281e-01 5.95160306e-01 7.59254098e-01 -2.03434244e-01 7.92256892e-01 2.28754491e-01 -5.35481513e-01 -1.18838823e+00 -7.06194997e-01 1.35077283e-01 7.17363596e-01 3.86236310e-01 1.33052887e-02 -8.63251388e-01 -6.44254923e-01 -1.05137095e-01 5.33312798e-01 -5.48510611e-01 -2.60294199e-01 -5.70304036e-01 -8.57574105e-01 6.13769293e-01 6.98964417e-01 9.06450689e-01 -8.95102501e-01 -3.08806568e-01 1.26759693e-01 -4.02535200e-01 -1.46572852e+00 -3.50534499e-01 3.18796128e-01 -6.29021168e-01 -1.06444955e+00 -9.23382580e-01 -8.57901394e-01 1.08276796e+00 6.41452193e-01 7.85460651e-01 -1.27883404e-01 -5.17267808e-02 3.81292701e-01 -3.73750806e-01 -3.68660361e-01 -3.24167311e-01 2.12813392e-01 4.46399897e-01 3.88757908e-03 3.45182925e-01 -4.83644307e-01 -3.73351246e-01 4.72770363e-01 -1.07787204e+00 -1.93888426e-01 8.03241253e-01 9.01572168e-01 5.26372492e-01 6.43776283e-02 2.32802510e-01 -6.61952317e-01 3.53109330e-01 -1.87061921e-01 -9.83355820e-01 4.59705479e-02 -2.44848043e-01 7.85344690e-02 7.38821805e-01 -7.10161626e-01 -8.79244626e-01 6.25575006e-01 -7.68446922e-02 -6.02829635e-01 -5.46584427e-01 4.51544851e-01 -3.50554258e-01 -5.38470030e-01 1.03730667e+00 2.77641922e-01 -6.96861446e-02 -4.14592683e-01 4.60917085e-01 5.03579617e-01 8.89691472e-01 -2.81411588e-01 1.36511064e+00 5.72446167e-01 -1.31422235e-02 -1.08344162e+00 -7.22229838e-01 -6.10064685e-01 -8.44829082e-01 -9.96672809e-02 6.91094756e-01 -1.11639249e+00 -2.81565458e-01 6.80172622e-01 -1.26118815e+00 -3.53397518e-01 -6.19163271e-03 6.73230171e-01 -4.04189914e-01 3.76001537e-01 -2.17716366e-01 -6.03628933e-01 1.45742863e-01 -1.19164503e+00 8.16270530e-01 3.20298314e-01 1.01901487e-01 -9.69184458e-01 9.66029465e-02 3.22409086e-02 4.51652944e-01 3.37129027e-01 4.21315044e-01 -5.58797181e-01 -6.28124654e-01 -4.19813126e-01 -1.31264627e-01 5.20787001e-01 3.38440984e-01 -2.33948365e-01 -1.05085981e+00 -5.02404213e-01 -9.12316516e-02 -3.40230703e-01 8.78417552e-01 2.08694860e-01 1.08707762e+00 -2.40423411e-01 -2.45546088e-01 8.32123816e-01 1.31462264e+00 1.77548232e-03 7.02322185e-01 6.75574660e-01 7.93259144e-01 4.37129408e-01 6.53137207e-01 9.45334956e-02 1.47639886e-01 6.62652493e-01 8.66303086e-01 -5.72179198e-01 -9.68731120e-02 -2.28589579e-01 2.95565218e-01 2.22894937e-01 1.78508699e-01 -3.23019207e-01 -1.09441924e+00 7.21703827e-01 -1.94368005e+00 -7.28644013e-01 -1.43487066e-01 2.22933531e+00 4.98591214e-01 -6.22382015e-02 -5.19542508e-02 7.12922215e-02 8.20825458e-01 1.90124407e-01 -6.72191858e-01 2.05593094e-01 -3.61875921e-01 -2.29355112e-01 9.46489275e-01 3.89554620e-01 -1.52279127e+00 1.38061297e+00 6.16221428e+00 2.18272462e-01 -1.24655843e+00 -1.12207100e-01 2.32771620e-01 3.01223874e-01 1.68267608e-01 -2.21334904e-01 -6.20984375e-01 4.05318797e-01 5.65598786e-01 2.80779004e-01 4.14438814e-01 1.05714762e+00 1.30327463e-01 -2.32377216e-01 -8.98932457e-01 1.16625464e+00 4.30161387e-01 -1.10781085e+00 -9.59702805e-02 -4.64794785e-02 8.55430245e-01 2.23529711e-01 1.92649856e-01 2.07194448e-01 4.27252859e-01 -8.37843001e-01 4.28531468e-01 2.06930086e-01 5.55670857e-01 -7.06283092e-01 1.11908412e+00 5.20815730e-01 -6.75883949e-01 1.82014946e-02 -6.85553432e-01 1.68113075e-02 3.17378007e-02 4.81655985e-01 -9.16265666e-01 2.74729073e-01 7.80607522e-01 4.84222412e-01 -7.09273458e-01 1.38708639e+00 -6.14056408e-01 5.24655819e-01 -5.52250028e-01 2.42973834e-01 3.55649829e-01 1.62252448e-02 5.17154813e-01 9.99438643e-01 3.06078821e-01 -1.21108018e-01 2.53450245e-01 5.63900292e-01 -2.51476347e-01 -1.74499944e-01 -9.75951850e-01 1.73394144e-01 4.24677134e-01 1.11989260e+00 -5.91110706e-01 -6.06040135e-02 -3.90113264e-01 1.41463149e+00 3.03117394e-01 5.84017217e-01 -9.80103374e-01 -3.88645589e-01 8.07088375e-01 -1.54752478e-01 3.69362295e-01 -5.29895723e-01 2.76974551e-02 -1.33188272e+00 4.90993299e-02 -8.48998606e-01 5.46551608e-02 -8.84931028e-01 -9.86298442e-01 5.17368078e-01 -2.39812955e-01 -1.30761206e+00 -1.93162948e-01 -6.35268033e-01 -3.88981730e-01 8.22841525e-01 -1.76074016e+00 -1.31737709e+00 -6.88013613e-01 6.88659728e-01 3.84069979e-01 -1.55055314e-01 8.10702145e-01 3.54252726e-01 -5.93865514e-01 6.11281514e-01 3.43537241e-01 3.82709026e-01 1.10222757e+00 -1.27661467e+00 5.84223628e-01 1.46873569e+00 3.21136981e-01 3.93970907e-01 9.01563585e-01 -4.68740493e-01 -1.27890134e+00 -1.44988084e+00 5.00230312e-01 -5.10244429e-01 5.62466860e-01 -4.78562295e-01 -9.54155087e-01 9.14052904e-01 1.10674389e-01 2.77923167e-01 3.21126461e-01 -1.50910348e-01 -3.60305488e-01 -2.51578540e-01 -1.03849208e+00 7.15518236e-01 6.76761270e-01 -3.92017394e-01 -4.78384852e-01 4.68720347e-01 6.51466787e-01 -7.19298482e-01 -3.12380046e-01 3.73430252e-01 -3.24138962e-02 -8.15057814e-01 9.21363413e-01 -4.57570136e-01 -2.79350504e-02 -7.26461172e-01 -2.27027386e-01 -1.67959273e+00 -3.15481722e-01 -5.47731519e-01 3.77240255e-02 1.26222563e+00 4.15482312e-01 -6.56855881e-01 9.81981754e-01 3.38369578e-01 -9.55202430e-03 2.09396202e-02 -7.43126333e-01 -9.94674325e-01 2.89303456e-02 -3.00910860e-01 4.16434944e-01 1.14082396e+00 -4.16958213e-01 1.93709731e-01 -6.93522811e-01 8.11418831e-01 5.43243349e-01 -3.35110933e-01 1.22391582e+00 -1.03180170e+00 1.02175340e-01 -7.89429843e-02 -6.71107590e-01 -1.12446249e+00 2.14761302e-01 -3.90764654e-01 4.67685223e-01 -1.46490479e+00 -2.37789646e-01 -3.85929286e-01 -1.28206372e-01 6.85821593e-01 -1.71995789e-01 5.80906510e-01 6.92480654e-02 3.02586645e-01 -4.79990929e-01 5.62176466e-01 8.56695235e-01 -1.34985477e-01 -2.78458744e-01 1.01413041e-01 -5.50748169e-01 9.74446177e-01 1.14381075e+00 -2.99701571e-01 -2.62175232e-01 -7.29023755e-01 -2.64622755e-02 -4.93120879e-01 4.65318561e-01 -1.35768569e+00 4.79073256e-01 -5.20182662e-02 5.74752152e-01 -5.89383483e-01 3.43031228e-01 -1.16811383e+00 -7.72113353e-02 4.54679370e-01 5.11791930e-03 1.34181470e-01 4.97039139e-01 6.66307986e-01 -3.44488025e-01 -1.83626022e-02 1.09669459e+00 -5.86801507e-02 -1.04516363e+00 1.80976927e-01 -5.19431531e-01 -2.23392293e-01 1.10855651e+00 8.48834366e-02 -3.37438822e-01 -5.76969445e-01 -4.65116918e-01 2.62259603e-01 7.70634115e-01 4.60158736e-01 4.58305299e-01 -1.35823059e+00 -3.98287237e-01 2.35165432e-01 3.33037823e-01 3.67443889e-01 4.16914038e-02 7.85938919e-01 -7.96361506e-01 4.31740522e-01 -8.58785361e-02 -8.11167002e-01 -1.11841702e+00 7.70571589e-01 4.55116123e-01 6.34621233e-02 -4.91439760e-01 8.08863342e-01 1.48054987e-01 -6.91695035e-01 5.51023841e-01 -3.54990155e-01 -6.43095225e-02 -1.57665595e-01 5.62179863e-01 -3.06299166e-03 1.98506549e-01 -6.05577290e-01 -4.33256269e-01 5.81974924e-01 -1.04811795e-01 -7.31157064e-02 1.26819313e+00 -2.00675935e-01 3.69084589e-02 7.63300955e-02 8.87500048e-01 -6.98769419e-03 -1.43685853e+00 -4.79244441e-01 -2.58597340e-02 -4.85660911e-01 1.42240793e-01 -7.89615989e-01 -1.05997658e+00 9.11541462e-01 6.60726488e-01 -6.10586666e-02 1.07534504e+00 -1.31468832e-01 5.01759291e-01 8.38481843e-01 3.18244964e-01 -1.06577277e+00 1.69310227e-01 5.54911613e-01 8.91598463e-01 -1.64340150e+00 -1.62822321e-01 -2.41762534e-01 -5.30961871e-01 8.81940961e-01 9.18582201e-01 1.95040703e-02 2.09776565e-01 1.25864103e-01 4.71192449e-01 2.02827349e-01 -1.36152282e-01 -4.89326119e-01 -1.96373016e-02 1.00169837e+00 -6.32224679e-02 -2.94423759e-01 3.61812055e-01 6.29366040e-02 -7.44862035e-02 -4.98144180e-01 5.79977512e-01 8.60071123e-01 -3.88509721e-01 -9.78331864e-01 -8.72453630e-01 -9.63200256e-02 -2.57035017e-01 5.80010712e-02 -5.98070204e-01 1.15265381e+00 7.67973289e-02 7.93159187e-01 -4.93866317e-02 -3.79722118e-01 4.38531041e-01 -1.36699989e-01 3.17722023e-01 -4.78724569e-01 -3.99181023e-02 2.55604871e-02 -2.78294146e-01 -5.74458480e-01 -5.46033144e-01 -5.96792936e-01 -9.23337638e-01 -3.39593023e-01 -4.99593914e-01 8.07644576e-02 9.48326826e-01 8.49787891e-01 2.73199499e-01 5.09332180e-01 6.29423738e-01 -1.15704918e+00 -4.47519153e-01 -9.56687450e-01 -3.03581119e-01 3.59716386e-01 6.71660542e-01 -6.69179678e-01 -5.77303708e-01 1.42391711e-01]
[7.877261161804199, -2.0593552589416504]
7d6d0135-9246-419b-902b-302c665dbfe5
a-comparative-study-of-western-and-chinese
2002.09021
null
https://arxiv.org/abs/2002.09021v1
https://arxiv.org/pdf/2002.09021v1.pdf
A Comparative Study of Western and Chinese Classical Music based on Soundscape Models
Whether literally or suggestively, the concept of soundscape is alluded in both modern and ancient music. In this study, we examine whether we can analyze and compare Western and Chinese classical music based on soundscape models. We addressed this question through a comparative study. Specifically, corpora of Western classical music excerpts (WCMED) and Chinese classical music excerpts (CCMED) were curated and annotated with emotional valence and arousal through a crowdsourcing experiment. We used a sound event detection (SED) and soundscape emotion recognition (SER) models with transfer learning to predict the perceived emotion of WCMED and CCMED. The results show that both SER and SED models could be used to analyze Chinese and Western classical music. The fact that SER and SED work better on Chinese classical music emotion recognition provides evidence that certain similarities exist between Chinese classical music and soundscape recordings, which permits transferability between machine learning models.
['Yi-Hsuan Yang', 'Jianyu Fan', 'Philippe Pasquier', 'Kui Dong']
2020-02-20
null
null
null
null
['music-emotion-recognition']
['music']
[-5.37221655e-02 -4.57280487e-01 2.14139134e-01 -2.25142002e-01 -6.80260241e-01 -9.62953627e-01 4.51039493e-01 1.41792176e-02 -6.54514730e-01 4.33501184e-01 3.66826504e-01 2.00999796e-01 -1.24407001e-01 -4.32786047e-01 -2.60554969e-01 -3.39499056e-01 -9.63444486e-02 -1.94683038e-02 -8.95967036e-02 -3.91379327e-01 4.47117895e-01 2.27579474e-02 -1.78431964e+00 7.12187827e-01 5.89140952e-01 1.02791882e+00 2.53784209e-01 9.37275171e-01 -1.06866151e-01 7.44829655e-01 -9.83506203e-01 -4.65209872e-01 4.23440076e-02 -8.75321805e-01 -6.69368386e-01 -4.77212697e-01 1.24204688e-01 4.79558676e-01 1.77675202e-01 9.88741696e-01 8.53103936e-01 2.84913927e-01 4.25696731e-01 -1.28632450e+00 -8.95200670e-01 1.06140149e+00 -1.37970015e-01 -5.31483181e-02 6.30441785e-01 -2.28045523e-01 1.06245673e+00 -7.91074991e-01 4.91059601e-01 8.79452825e-01 9.84709918e-01 6.18425190e-01 -1.12117553e+00 -9.28196073e-01 -2.59897530e-01 5.59305191e-01 -1.46340311e+00 -1.80739611e-01 1.05211127e+00 -5.88818073e-01 6.52713120e-01 7.04140127e-01 1.07054210e+00 1.15636730e+00 -4.81115803e-02 5.11045218e-01 1.44523501e+00 -5.28671861e-01 7.06609368e-01 5.19144237e-01 -1.09557688e-01 -9.27977860e-02 -4.16123867e-01 3.60976681e-02 -1.06879771e+00 -2.87128776e-01 2.95207292e-01 -5.94917834e-01 -2.08745062e-01 6.82565808e-01 -1.12030756e+00 6.63696110e-01 6.88414574e-02 9.86511052e-01 -3.97808164e-01 1.67427808e-01 8.30173492e-01 4.98635799e-01 3.58393222e-01 9.07754362e-01 -6.08564436e-01 -8.56218874e-01 -1.06918812e+00 5.02555519e-02 9.38198745e-01 4.75667715e-01 4.67066973e-01 3.32365811e-01 1.10984981e-01 1.05833352e+00 -6.28777668e-02 5.84994793e-01 8.64414513e-01 -9.81967747e-01 -3.97418797e-01 2.39948377e-01 -9.30036455e-02 -1.10647929e+00 -3.00633401e-01 1.66940056e-02 -3.13281476e-01 2.73771107e-01 3.92743610e-02 -9.95045975e-02 -6.81897625e-02 1.69937277e+00 -1.28719017e-01 1.52241766e-01 4.67217900e-03 1.10085750e+00 9.34623778e-01 6.18143380e-01 3.78436655e-01 -2.74543077e-01 1.43387187e+00 -4.34748530e-01 -8.74178886e-01 -7.68822506e-02 2.69354939e-01 -1.06603384e+00 1.67544186e+00 8.41527283e-01 -9.17902708e-01 -9.42133367e-01 -9.68092680e-01 6.19102344e-02 -4.12957340e-01 2.76801109e-01 4.13252473e-01 9.57228720e-01 -6.28286541e-01 8.62842917e-01 -5.28282702e-01 -4.55579698e-01 3.25017497e-02 -1.65368319e-02 -5.23683205e-02 9.42033768e-01 -1.17324066e+00 6.96118355e-01 3.97603244e-01 -6.78314567e-02 -5.02898812e-01 -4.89620000e-01 -3.57076377e-01 -2.00592041e-01 -1.07350051e-01 1.34034604e-01 1.34698737e+00 -1.93957281e+00 -1.94011724e+00 1.17020023e+00 3.95129085e-01 -1.81855440e-01 -3.78300771e-02 -4.07639176e-01 -1.14647663e+00 2.85653234e-01 -6.79319799e-02 2.74800301e-01 6.56403482e-01 -1.14427543e+00 -3.93332511e-01 -8.20920318e-02 -5.90192676e-01 -1.38156757e-01 -5.71964085e-01 5.96288860e-01 -6.06031679e-02 -1.00080478e+00 -1.96994320e-01 -9.82639313e-01 2.56786078e-01 -2.14730278e-01 -3.45019996e-01 -9.03209224e-02 4.25938517e-01 -6.41598165e-01 1.48856699e+00 -2.58994079e+00 -1.13473469e-02 2.28563040e-01 -2.77567208e-01 1.56617045e-01 -2.27657974e-01 6.42781079e-01 -4.23595197e-02 2.83723418e-02 -1.58668339e-01 1.39078274e-01 4.53528672e-01 3.15061629e-01 -4.40351009e-01 3.12368691e-01 -1.26907915e-01 8.10747266e-01 -8.26737821e-01 -5.74613869e-01 -2.64913794e-02 5.00992715e-01 -5.58589041e-01 5.69550693e-02 -2.50681043e-01 6.24211669e-01 -3.67761254e-01 4.52783167e-01 2.58973241e-01 3.35876942e-01 1.64697066e-01 -1.63274333e-01 -5.16902149e-01 1.80101737e-01 -1.08940649e+00 1.85299444e+00 -4.70272332e-01 9.35895562e-01 -2.49645691e-02 -5.34807146e-01 1.30658698e+00 5.88549376e-01 4.09151077e-01 -9.86582994e-01 3.42705280e-01 3.90868992e-01 5.41432947e-02 -8.08170557e-01 7.61509955e-01 -8.47770810e-01 -3.65093499e-01 5.06173015e-01 4.34172526e-02 -3.77489954e-01 -1.30874440e-01 -3.15397233e-01 6.69343948e-01 4.45558220e-01 3.20017070e-01 -2.37975240e-01 5.40979207e-01 9.67065319e-02 6.06293440e-01 3.75343472e-01 -1.90405026e-01 4.76511121e-01 2.33831450e-01 -1.50885984e-01 -6.63450420e-01 -9.79297757e-01 4.15923297e-02 1.65572894e+00 -1.15493618e-01 -9.62849140e-01 -7.07474053e-01 5.98911233e-02 -3.97110224e-01 9.01763439e-01 -5.93926013e-01 -1.77676305e-01 -3.49304914e-01 -3.40910912e-01 1.14737535e+00 5.86078227e-01 5.77489706e-03 -1.54846740e+00 -9.51516807e-01 2.65730441e-01 -2.59746462e-01 -8.38326573e-01 -2.52641857e-01 3.03807855e-01 -3.86578172e-01 -8.81268919e-01 -4.05091554e-01 -7.70535648e-01 -4.66472924e-01 -3.24124008e-01 1.25255358e+00 -2.66022384e-01 -4.18390572e-01 7.91728139e-01 -8.62507582e-01 -1.06803608e+00 -5.90773582e-01 -3.61682147e-01 2.65061349e-01 3.09561454e-02 8.75814021e-01 -8.98088694e-01 -4.02156621e-01 2.16294795e-01 -9.55266058e-01 -2.89396554e-01 2.83284038e-01 1.40515774e-01 5.85730374e-01 -2.91068614e-01 8.91640544e-01 -3.68178993e-01 7.90530026e-01 -2.26989701e-01 -6.00529015e-02 -6.83940277e-02 -5.58845937e-01 -4.13890809e-01 6.08159125e-01 -8.48746657e-01 -9.64931369e-01 -9.46748257e-02 -3.08078229e-01 -3.68306637e-01 -2.83248395e-01 4.44166392e-01 7.96745941e-02 1.97356671e-01 1.16332722e+00 1.09123345e-02 -4.23508644e-01 -5.94478071e-01 5.16110063e-01 1.17835617e+00 9.60515320e-01 -7.89838850e-01 4.73997831e-01 3.97942096e-01 -4.00328249e-01 -1.09643924e+00 -5.70327938e-01 -3.60488057e-01 -4.50021178e-01 -6.26822054e-01 1.23521316e+00 -9.03773665e-01 -9.84529734e-01 2.03283094e-02 -7.78564274e-01 -3.21065545e-01 -6.62710547e-01 9.25602198e-01 -9.12135005e-01 3.26187164e-01 -7.94935703e-01 -1.17252231e+00 -4.61745590e-01 -2.73130655e-01 1.02166522e+00 8.92503634e-02 -9.99531150e-01 -7.73936689e-01 8.01649332e-01 -1.08370058e-01 2.84898818e-01 3.66229057e-01 7.64055490e-01 -9.30237472e-01 5.13693988e-01 -1.78657696e-01 4.20915663e-01 5.71156979e-01 -1.59509927e-02 4.23297882e-02 -1.52583468e+00 2.43130922e-01 3.38967741e-01 -6.51150227e-01 2.78046519e-01 7.26340413e-02 8.72473061e-01 -4.42345738e-02 5.15888810e-01 3.18005115e-01 1.39519501e+00 4.81369257e-01 8.31579804e-01 4.82887149e-01 1.57981321e-01 5.92357457e-01 6.28600419e-01 8.89451087e-01 -5.07992283e-02 5.70549667e-01 -1.13916092e-01 3.06397736e-01 1.39173642e-01 -3.20520937e-01 8.36091220e-01 1.37181652e+00 -3.94460857e-01 1.82742566e-01 -6.89965785e-01 2.50630736e-01 -1.69410443e+00 -1.26455247e+00 -2.82482415e-01 2.04317021e+00 1.08086419e+00 -3.25765729e-01 4.04386848e-01 5.57152092e-01 5.95292985e-01 -1.46372259e-01 4.21722308e-02 -8.81972373e-01 -4.80172455e-01 8.33610356e-01 -2.90501177e-01 -6.72978684e-02 -9.70102310e-01 7.93551743e-01 6.28920031e+00 7.79145241e-01 -1.33226442e+00 2.00769991e-01 -2.15505376e-01 -4.53276008e-01 -9.49944034e-02 -1.07462965e-01 1.76407024e-01 4.15488482e-01 1.33653247e+00 -1.51187181e-01 7.29871929e-01 8.12705100e-01 2.36396894e-01 -6.84188632e-03 -1.24925995e+00 1.39568317e+00 2.54544616e-01 -8.92572284e-01 -4.13552165e-01 -4.53012496e-01 5.76503634e-01 -1.08916640e-01 -6.42069802e-02 5.57718933e-01 -3.31967354e-01 -7.85352349e-01 1.09502065e+00 8.15591395e-01 7.18956172e-01 -5.69717109e-01 4.39728498e-01 -7.83373788e-03 -1.13876820e+00 -7.17204660e-02 -3.02151412e-01 -3.57827544e-01 1.00781061e-01 1.33074047e-02 -3.73299897e-01 4.41580504e-01 9.64122951e-01 6.71833277e-01 -6.06379032e-01 8.15039337e-01 -9.71527398e-02 1.07493031e+00 -9.95125249e-02 -2.92835802e-01 -3.32571939e-03 -3.02180469e-01 5.36738038e-01 1.61639357e+00 5.78103960e-01 -1.22496627e-01 -7.04674199e-02 1.14103997e+00 4.05000836e-01 7.80054927e-01 -3.43207605e-02 -6.30399168e-01 4.25993204e-01 1.40854812e+00 -8.85518670e-01 -2.86718398e-01 -4.50823963e-01 1.16268981e+00 -3.03791791e-01 7.60990679e-02 -1.01232862e+00 -3.88398290e-01 4.70452279e-01 -2.80548155e-01 1.60704911e-01 1.95661351e-01 -4.35077369e-01 -9.56457019e-01 -3.05983543e-01 -1.00174904e+00 3.67674708e-01 -1.19417286e+00 -1.75154150e+00 7.86178172e-01 -4.55861628e-01 -1.67875397e+00 2.04658136e-02 -6.42196059e-01 -8.12189579e-01 6.46363735e-01 -9.32044804e-01 -8.61147940e-01 -2.84330547e-01 8.45860124e-01 2.33415231e-01 -1.71765894e-01 1.34975290e+00 3.83957773e-01 -1.98765457e-01 2.02581331e-01 -1.27969846e-01 1.19545579e-01 1.18743587e+00 -1.43367946e+00 -3.09542686e-01 6.47823215e-02 8.54779661e-01 2.00125903e-01 9.30442393e-01 -4.00855452e-01 -1.27140915e+00 -8.02851379e-01 7.27257609e-01 -4.07734573e-01 8.43172073e-01 -2.11572126e-01 -7.89937258e-01 2.93203324e-01 4.11754698e-01 -4.42450613e-01 1.57291198e+00 2.60352343e-01 -5.68394423e-01 9.90213007e-02 -6.45729423e-01 5.50111592e-01 5.26614785e-01 -1.00446439e+00 -1.04825127e+00 -8.33582878e-02 3.64345402e-01 2.63148487e-01 -1.11170828e+00 -6.67911768e-02 9.06121135e-01 -1.00433099e+00 6.00492656e-01 -6.10400498e-01 4.38162148e-01 -3.52645963e-01 -5.78920722e-01 -1.30704117e+00 -5.28067499e-02 -7.11882174e-01 3.40242803e-01 1.39443767e+00 2.68556178e-01 -4.81040552e-02 1.67067811e-01 2.22918779e-01 -3.38124692e-01 1.64299265e-01 -8.52567315e-01 -9.82857525e-01 3.22559983e-01 -1.12730086e+00 3.02514791e-01 1.35974813e+00 6.24164164e-01 4.54943061e-01 -3.18378419e-01 -3.16555947e-01 -2.29215190e-01 3.17212522e-01 5.53535938e-01 -1.56820238e+00 -5.30998290e-01 -6.38730288e-01 -4.37602222e-01 2.67280608e-01 2.81248420e-01 -1.02488303e+00 2.80576944e-01 -9.74075556e-01 9.65196267e-03 1.81970373e-01 -6.19510412e-01 3.87081563e-01 1.02345303e-01 7.23115385e-01 4.52397943e-01 1.33504212e-01 -5.15775681e-01 4.48664367e-01 7.66008973e-01 2.80884832e-01 -6.43167377e-01 -2.06823558e-01 -5.73157132e-01 9.94596183e-01 9.06490564e-01 -5.47600806e-01 -1.92555171e-02 1.62450552e-01 9.07557845e-01 -1.78362951e-01 5.60376585e-01 -1.15777218e+00 6.36251345e-02 5.36929406e-02 2.80235499e-01 -4.66117971e-02 4.69499856e-01 -7.38767684e-01 3.37359607e-01 2.18998462e-01 -5.59403360e-01 -1.16041787e-01 4.84762162e-01 4.35328424e-01 -4.41441745e-01 -1.67881191e-01 6.65381849e-01 -3.95251475e-02 -8.40000272e-01 -5.01530051e-01 -7.49836743e-01 1.84869081e-01 7.48903573e-01 -2.93342143e-01 2.98569687e-02 -4.94535923e-01 -1.04381704e+00 -5.53084075e-01 1.62375599e-01 6.38225198e-01 4.57895935e-01 -1.52549720e+00 -7.70029187e-01 -7.02021122e-02 4.81433332e-01 -1.19509411e+00 3.56249154e-01 1.04528034e+00 -3.17078501e-01 -5.10880873e-02 -5.02203226e-01 -3.11543643e-01 -1.21782243e+00 8.22967529e-01 9.88746807e-02 6.33730769e-01 -5.49139798e-01 5.06931186e-01 -3.53539109e-01 -3.85892600e-01 2.58838058e-01 -1.85973838e-01 -3.00000638e-01 3.34180444e-01 6.49237931e-01 5.43941259e-01 -2.58987784e-01 -7.98765838e-01 -2.79793471e-01 5.90460718e-01 8.45831513e-01 -6.96302652e-01 1.26755285e+00 3.08362097e-02 -3.97030622e-01 1.41558349e+00 9.46652889e-01 6.48195088e-01 -3.92302036e-01 1.31205231e-01 4.52871203e-01 -1.02546319e-01 -1.62348107e-01 -1.15675414e+00 -5.06955266e-01 7.98810244e-01 9.53401506e-01 4.63325232e-01 1.47073758e+00 2.41238810e-02 6.30178690e-01 4.04794872e-01 1.85914278e-01 -1.74490845e+00 6.91104382e-02 2.17446819e-01 9.59531307e-01 -6.85031414e-01 -4.48307574e-01 1.43031552e-01 -1.27914298e+00 1.12820756e+00 2.33476356e-01 -2.57915616e-01 7.23365128e-01 3.33463043e-01 3.88729662e-01 -1.84179142e-01 -6.65548980e-01 -3.94873083e-01 5.59613943e-01 4.73496079e-01 9.98879969e-01 2.95032769e-01 -6.87628567e-01 1.67227840e+00 -8.59256387e-01 1.62935704e-01 4.30278093e-01 8.63288224e-01 -5.73821008e-01 -1.04942834e+00 -5.97531796e-01 -1.35088295e-01 -6.96070433e-01 -1.30225420e-01 -1.12422204e+00 7.11594820e-01 6.23194456e-01 1.30760884e+00 1.57109633e-01 -9.01711524e-01 3.27895105e-01 5.91588616e-01 4.10017431e-01 -4.79013324e-01 -1.39035130e+00 6.03952885e-01 7.73988366e-02 -5.39620340e-01 -8.31424236e-01 -2.86617249e-01 -1.38700175e+00 1.31917596e-01 -1.50618643e-01 5.16987681e-01 7.94562638e-01 7.36813009e-01 1.43305182e-01 6.65531397e-01 6.15947485e-01 -6.73860669e-01 -1.40695468e-01 -1.21168149e+00 -1.14145029e+00 5.66961646e-01 -2.33664036e-01 -2.47811042e-02 -3.68402839e-01 4.53607500e-01]
[15.856063842773438, 5.188298225402832]
7bca421b-40af-443e-9fa8-9f8885f7f247
sts-ccl-spatial-temporal-synchronous
2307.02507
null
https://arxiv.org/abs/2307.02507v1
https://arxiv.org/pdf/2307.02507v1.pdf
STS-CCL: Spatial-Temporal Synchronous Contextual Contrastive Learning for Urban Traffic Forecasting
Efficiently capturing the complex spatiotemporal representations from large-scale unlabeled traffic data remains to be a challenging task. In considering of the dilemma, this work employs the advanced contrastive learning and proposes a novel Spatial-Temporal Synchronous Contextual Contrastive Learning (STS-CCL) model. First, we elaborate the basic and strong augmentation methods for spatiotemporal graph data, which not only perturb the data in terms of graph structure and temporal characteristics, but also employ a learning-based dynamic graph view generator for adaptive augmentation. Second, we introduce a Spatial-Temporal Synchronous Contrastive Module (STS-CM) to simultaneously capture the decent spatial-temporal dependencies and realize graph-level contrasting. To further discriminate node individuals in negative filtering, a Semantic Contextual Contrastive method is designed based on semantic features and spatial heterogeneity, achieving node-level contrastive learning along with negative filtering. Finally, we present a hard mutual-view contrastive training scheme and extend the classic contrastive loss to an integrated objective function, yielding better performance. Extensive experiments and evaluations demonstrate that building a predictor upon STS-CCL contrastive learning model gains superior performance than existing traffic forecasting benchmarks. The proposed STS-CCL is highly suitable for large datasets with only a few labeled data and other spatiotemporal tasks with data scarcity issue.
['Jichao Bi', 'Fengji Luo', 'Kaixiang Yang', 'Lincan Li']
2023-07-05
null
null
null
null
['contrastive-learning', 'contrastive-learning']
['computer-vision', 'methodology']
[ 1.54194802e-01 -3.78300786e-01 -3.63851875e-01 -3.96197170e-01 -3.70988041e-01 -2.75491416e-01 7.30482876e-01 -5.26443906e-02 -1.63750604e-01 6.60248637e-01 1.32285625e-01 -5.24728417e-01 -2.14081571e-01 -8.65002275e-01 -5.87916911e-01 -7.63567090e-01 -2.37546653e-01 1.98987275e-01 6.32911026e-01 -5.93005896e-01 3.49282771e-02 6.73002422e-01 -1.40873098e+00 1.38306022e-01 1.20821095e+00 1.18851995e+00 2.47597307e-01 4.91088867e-01 -3.77288043e-01 9.89398599e-01 -4.38726842e-01 -4.24139798e-01 2.28257298e-01 -4.35603917e-01 -4.65459198e-01 1.07385494e-01 5.57712734e-01 2.51532346e-01 -7.18611896e-01 9.46551323e-01 4.71685559e-01 4.34659004e-01 4.87599134e-01 -1.83538151e+00 -8.51952851e-01 2.16700763e-01 -9.25362945e-01 8.48495066e-01 -4.11383919e-02 4.92530912e-01 9.95414019e-01 -7.97002196e-01 3.52809757e-01 1.33466113e+00 6.14535511e-01 2.98682541e-01 -1.09592426e+00 -8.38246822e-01 9.89186883e-01 5.88835716e-01 -1.40194952e+00 -1.34897411e-01 1.35834873e+00 -3.27796012e-01 5.77097476e-01 4.10296082e-01 6.11798227e-01 1.09845448e+00 -2.14872397e-02 7.83795357e-01 1.17736673e+00 -3.34271751e-02 -6.72351345e-02 -2.46260744e-02 8.09467658e-02 9.27128851e-01 -9.32824835e-02 2.47877061e-01 -3.95348638e-01 3.69050689e-02 5.83963752e-01 1.27910480e-01 -2.06358824e-03 -4.05007362e-01 -1.13285971e+00 6.46262228e-01 8.59009624e-01 1.90789327e-01 -4.20869559e-01 -1.00110546e-02 5.48278153e-01 3.79846424e-01 8.41671646e-01 -1.19923629e-01 -2.85455287e-01 2.83301681e-01 -8.16214919e-01 1.05646521e-01 2.22523510e-01 1.05753314e+00 8.90841663e-01 5.28415203e-01 -6.13719761e-01 6.59712017e-01 1.54280379e-01 5.90880990e-01 1.91379860e-01 -4.90069002e-01 8.66406560e-01 7.96172202e-01 -4.19354975e-01 -1.47445107e+00 -4.52491641e-01 -9.21820998e-01 -1.17368042e+00 -5.95375784e-02 -7.45001435e-03 -6.98466823e-02 -9.66220796e-01 2.10845685e+00 3.93776953e-01 1.05726755e+00 -2.34781295e-01 8.24292421e-01 8.92190456e-01 7.32848942e-01 6.51795447e-01 -4.04903263e-01 1.11861122e+00 -1.02516890e+00 -7.93493807e-01 -2.21906275e-01 6.87190652e-01 -2.34466165e-01 1.35415316e+00 -1.17286235e-01 -8.70394111e-01 -8.37192655e-01 -9.19660985e-01 2.24231064e-01 -7.56076336e-01 -1.35164037e-02 7.41474807e-01 5.24769664e-01 -9.82838333e-01 1.35402068e-01 -3.75146806e-01 -2.14593008e-01 6.54128134e-01 1.84002310e-01 -2.34900668e-01 -2.85578668e-02 -1.62135398e+00 4.58673954e-01 2.78440386e-01 3.04480761e-01 -5.24000108e-01 -8.01612914e-01 -9.06276643e-01 1.45162065e-02 6.27270579e-01 -6.23122633e-01 3.35942328e-01 -7.37403095e-01 -9.91617680e-01 6.78697348e-01 -1.62661165e-01 -2.31512293e-01 6.08007729e-01 3.32405478e-01 -1.16574204e+00 6.21204898e-02 3.26530606e-01 4.86004859e-01 9.40762699e-01 -1.36038399e+00 -6.84675634e-01 -3.99491131e-01 5.26186377e-02 3.23908269e-01 -5.66826880e-01 -1.93579733e-01 -6.05465353e-01 -1.15949214e+00 -4.56312522e-02 -7.64893651e-01 -3.48127365e-01 1.49427177e-02 -2.70570248e-01 -3.12140673e-01 1.19479299e+00 -5.68129361e-01 1.54521728e+00 -1.94828856e+00 1.69759374e-02 4.30528730e-01 5.85728347e-01 4.85956907e-01 -5.21613538e-01 2.37606302e-01 -2.35413864e-01 3.91051434e-02 -3.59816194e-01 -1.72873825e-01 -1.38483077e-01 1.34914935e-01 -1.83449060e-01 2.62117803e-01 5.65844357e-01 1.27523530e+00 -1.09015727e+00 -7.68820226e-01 3.60976100e-01 3.81991774e-01 -4.32609648e-01 1.89946234e-01 -1.64820552e-01 6.99988663e-01 -6.35760188e-01 6.46946549e-01 1.01058280e+00 -4.80656385e-01 -1.11446753e-01 -5.20905972e-01 9.07726660e-02 -2.08374113e-01 -9.87127841e-01 1.47459733e+00 -3.96460742e-01 3.07243317e-01 1.65295918e-02 -1.50795960e+00 1.08316982e+00 -1.27788693e-01 6.79304481e-01 -1.29025233e+00 -1.53860927e-01 -3.81418280e-02 -3.08794498e-01 -5.86865366e-01 3.70212048e-01 -6.01662844e-02 -5.28644212e-02 9.62057710e-02 -7.62567073e-02 2.89761722e-01 7.66427144e-02 5.56690872e-01 1.08174324e+00 1.32342026e-01 -2.65057802e-01 -3.00601870e-01 9.98713732e-01 -2.20540956e-01 7.48753011e-01 4.32031929e-01 -5.54391921e-01 3.22690368e-01 5.65274298e-01 -4.30383742e-01 -5.88663638e-01 -1.14750516e+00 2.86242694e-01 1.28092182e+00 4.51807380e-01 -2.88401902e-01 -2.44996533e-01 -1.25513148e+00 -2.69430373e-02 4.97869879e-01 -7.09122777e-01 -4.88159299e-01 -8.89288008e-01 -8.78799617e-01 3.36260706e-01 6.89977109e-01 8.22114229e-01 -8.32969725e-01 4.08385515e-01 6.92178831e-02 -2.56915420e-01 -1.26727533e+00 -8.06719959e-01 -1.49985120e-01 -6.06330335e-01 -8.93354595e-01 -5.42394102e-01 -9.87816870e-01 5.00066996e-01 7.13996112e-01 1.25227904e+00 3.19211781e-01 -1.69696495e-01 2.72886366e-01 -4.16573375e-01 -8.65808316e-03 6.20613173e-02 1.71093062e-01 -2.63304532e-01 5.70886731e-01 2.54405946e-01 -1.02804506e+00 -7.87920415e-01 4.88811642e-01 -7.92999387e-01 4.75352928e-02 7.08213329e-01 8.08110178e-01 5.47386348e-01 -5.06087625e-03 1.03988075e+00 -8.70955348e-01 7.69994199e-01 -9.09717679e-01 -3.01596671e-01 4.37015146e-01 -6.81980968e-01 -1.85865626e-01 7.64169693e-01 -5.97988963e-01 -1.21474004e+00 -1.58099577e-01 -5.08690700e-02 -8.24375570e-01 3.14139314e-02 5.80876470e-01 -3.31460297e-01 -1.67475745e-01 5.01247644e-01 5.69114327e-01 -5.01397178e-02 -1.23709641e-01 6.84370339e-01 1.35796949e-01 5.48536539e-01 -6.92873597e-01 1.13373029e+00 5.25393903e-01 3.21340382e-01 -7.60153174e-01 -8.58887672e-01 -3.63201410e-01 -7.10657835e-01 -5.15080035e-01 8.30287158e-01 -9.23229933e-01 -5.78201294e-01 3.67549717e-01 -7.06829488e-01 -3.57295275e-01 -3.35699826e-01 3.21006417e-01 -3.60654205e-01 5.84103286e-01 -4.39152062e-01 -6.03391051e-01 8.42277333e-03 -7.83668935e-01 1.07301342e+00 7.31662512e-02 6.79714262e-01 -1.31359577e+00 -4.83602509e-02 3.03852201e-01 5.06638050e-01 4.30106312e-01 8.87618482e-01 -5.45845747e-01 -6.53204262e-01 -2.67736558e-02 -6.34256661e-01 2.92929471e-01 1.71792284e-01 -1.87638447e-01 -6.42542422e-01 -2.40439460e-01 -4.41334218e-01 -1.96968660e-01 1.08074045e+00 1.80332527e-01 1.60476637e+00 -3.20149302e-01 -4.76421058e-01 7.35180795e-01 1.15940094e+00 -1.05136722e-01 5.42625308e-01 -4.56415340e-02 1.29572809e+00 8.25782537e-01 7.62180865e-01 1.79710865e-01 7.89930880e-01 6.32069170e-01 4.62527573e-01 -4.66289848e-01 -4.48802620e-01 -4.30537641e-01 3.33217951e-03 9.34659421e-01 -1.13454007e-01 -4.74530399e-01 -6.49902403e-01 5.61102509e-01 -2.03467894e+00 -1.21256018e+00 -3.32532674e-01 1.75469685e+00 2.30376855e-01 3.38190556e-01 4.47102904e-01 4.90256250e-02 1.01337087e+00 8.04972529e-01 -5.92300475e-01 9.43434313e-02 -4.32341158e-01 -3.03082779e-04 4.55834806e-01 3.31835032e-01 -1.23006570e+00 1.03818607e+00 5.37170076e+00 1.29339862e+00 -1.28998280e+00 2.25819886e-01 8.01743269e-01 1.58677384e-01 -6.60614371e-01 -1.59907982e-01 -3.67243350e-01 8.95265222e-01 7.39647388e-01 7.10695460e-02 4.52024132e-01 4.99875844e-01 3.87074709e-01 6.12200737e-01 -5.00044346e-01 1.05809700e+00 7.03748390e-02 -1.32921374e+00 3.99272084e-01 6.71084679e-04 5.74789286e-01 -5.91682550e-03 2.45491415e-01 6.92136765e-01 1.80072010e-01 -7.94564128e-01 5.28237045e-01 5.84138989e-01 7.19558835e-01 -6.73120558e-01 3.22736859e-01 2.42462039e-01 -2.15336132e+00 -2.72192836e-01 -7.56299719e-02 2.29234204e-01 4.56314027e-01 4.22771960e-01 -7.38359429e-03 1.02192366e+00 5.14497161e-01 1.15768266e+00 -1.05090582e+00 7.39560246e-01 9.43320543e-02 6.90955758e-01 -2.43549943e-02 1.71406701e-01 3.96323085e-01 -3.96838158e-01 6.05828702e-01 1.29720891e+00 1.07876204e-01 1.75540701e-01 6.16112769e-01 8.83312583e-01 -5.23381047e-02 2.28698999e-01 -6.61940277e-01 1.91616654e-01 6.92864776e-01 1.41250312e+00 -5.71763635e-01 -4.98117983e-01 -6.73141241e-01 8.15512121e-01 5.65513372e-01 6.26716077e-01 -1.22409332e+00 -1.67529777e-01 2.92598933e-01 3.29419553e-01 2.30177149e-01 -2.00152963e-01 -1.05232395e-01 -1.20756006e+00 3.12787741e-01 -6.43429458e-01 7.76787043e-01 -6.77691281e-01 -1.72306776e+00 6.92723453e-01 1.62736878e-01 -1.55153406e+00 2.87964582e-01 -3.85864437e-01 -1.07589507e+00 7.68482745e-01 -1.97212422e+00 -1.92653728e+00 -7.12279797e-01 1.15629816e+00 4.86977607e-01 -4.49680924e-01 1.09044388e-01 7.49797821e-01 -7.96662807e-01 7.02592432e-01 -5.57394505e-01 -1.26470581e-01 4.33524042e-01 -1.09904253e+00 4.64068264e-01 8.67685914e-01 -5.27219288e-02 2.39501938e-01 2.34979555e-01 -7.77491093e-01 -1.11277509e+00 -1.72941768e+00 6.00699782e-01 -2.36623272e-01 7.64544845e-01 -3.76270324e-01 -1.12010920e+00 6.39505208e-01 5.70285879e-02 7.60989487e-01 2.85416365e-01 -4.57875356e-02 -4.92381930e-01 -6.30855262e-01 -1.04042697e+00 6.48407102e-01 1.74631333e+00 -7.04565644e-01 -1.78185835e-01 4.35258210e-01 9.55245435e-01 4.79449034e-02 -6.96098387e-01 8.39321196e-01 1.19476333e-01 -7.37066686e-01 1.12648702e+00 -8.62373471e-01 -2.46289283e-01 -4.87142324e-01 5.20838164e-02 -1.14914405e+00 -7.66055524e-01 -6.51814520e-01 -2.81421632e-01 1.45909965e+00 3.46184880e-01 -9.09684598e-01 9.69951630e-01 6.45438433e-02 -3.65982413e-01 -7.44599342e-01 -7.76646316e-01 -9.49943066e-01 1.34189008e-02 -3.80710334e-01 7.64161289e-01 1.31628394e+00 -4.53337878e-01 5.49390495e-01 -5.43383420e-01 2.11215481e-01 5.76221406e-01 1.27879875e-02 6.83463633e-01 -1.35630953e+00 6.60246983e-02 -5.32360733e-01 -5.29352963e-01 -1.20954406e+00 4.55180556e-01 -1.00239336e+00 -3.44036669e-01 -1.26187265e+00 -1.89832486e-02 -6.90381169e-01 -7.30413020e-01 9.18282941e-02 -6.08623862e-01 2.25467429e-01 2.02079434e-02 1.46651104e-01 -9.19551194e-01 9.48422492e-01 1.46247399e+00 -4.38708723e-01 -9.19463113e-02 8.18088204e-02 -7.10211217e-01 3.12242389e-01 7.48176515e-01 -7.27899373e-02 -1.05548596e+00 -1.30543798e-01 -8.42631832e-02 -6.86167851e-02 5.55427194e-01 -9.47206020e-01 2.65871376e-01 -4.95270371e-01 1.80672452e-01 -8.26056778e-01 2.16706499e-01 -7.64240384e-01 -1.32646307e-01 1.37330621e-01 -3.39734375e-01 4.91604894e-01 1.24718688e-01 1.16896343e+00 -2.50595540e-01 6.52554870e-01 6.13364339e-01 -1.06840141e-01 -1.16220009e+00 1.03714001e+00 -3.48681509e-02 4.01622176e-01 1.26606274e+00 -3.71666729e-01 -4.70666200e-01 -3.60387504e-01 -9.03188705e-01 6.33335054e-01 1.68101732e-02 7.28090167e-01 6.41947150e-01 -1.87368739e+00 -7.96675026e-01 3.16461533e-01 4.69940901e-01 -3.57408613e-01 7.35813320e-01 1.02406383e+00 -1.57727420e-01 8.29705074e-02 -8.90172347e-02 -7.50478983e-01 -9.63305116e-01 9.94113088e-01 2.67747760e-01 -4.83173490e-01 -5.38851023e-01 7.54125118e-01 4.84165043e-01 -6.47069037e-01 4.74267974e-02 -2.99270507e-02 -5.25531173e-01 3.72110642e-02 1.50157809e-01 4.34802264e-01 6.22154251e-02 -1.03954649e+00 -4.53541696e-01 6.53684139e-01 2.79089391e-01 2.86631227e-01 1.17848682e+00 -4.60615933e-01 1.01765893e-01 4.74592373e-02 1.24733746e+00 3.52024995e-02 -1.29681230e+00 -5.20816624e-01 -1.42691448e-01 -5.78316689e-01 -9.49412510e-02 -4.54229355e-01 -1.60489929e+00 9.97602880e-01 7.39150107e-01 5.16820014e-01 1.30497217e+00 -1.25799999e-01 9.01027501e-01 6.97018728e-02 3.78139317e-02 -1.01667070e+00 5.58297157e-01 2.64667332e-01 8.17737520e-01 -1.38483799e+00 -1.82821035e-01 -8.06299686e-01 -7.72832453e-01 5.77126145e-01 1.09355128e+00 -2.17429176e-01 1.09270537e+00 -1.08967349e-01 -2.25724623e-01 -5.33534229e-01 -8.31821263e-01 -6.10921860e-01 5.73432148e-01 8.46829534e-01 6.27028421e-02 4.69445549e-02 -3.26131135e-01 3.44063550e-01 2.91049868e-01 -4.09591228e-01 -1.93140537e-01 7.85325825e-01 -1.23481184e-01 -8.59947145e-01 1.54249877e-01 5.63516796e-01 -7.54121272e-03 -4.89671156e-02 -2.40485266e-01 7.67587364e-01 3.47716987e-01 9.73242223e-01 1.57308340e-01 -8.30152452e-01 5.58477104e-01 -2.96686500e-01 9.64379385e-02 -2.95165569e-01 -4.09967631e-01 -1.01203442e-01 1.20947853e-01 -6.51164055e-01 -5.52639604e-01 -3.57262760e-01 -9.51567650e-01 -4.32386249e-01 -2.86464393e-01 1.76225066e-01 1.11123219e-01 8.51180613e-01 6.02840722e-01 8.41206968e-01 1.05204344e+00 -7.35924125e-01 -2.20593154e-01 -6.98404074e-01 -5.03082752e-01 7.68065453e-01 3.77720594e-01 -8.69144738e-01 -4.02301043e-01 -1.12373248e-01]
[6.499320983886719, 2.0869951248168945]
683cde65-f00c-496c-889b-9ba6c3eba2ce
regularized-training-of-nearest-neighbor
2109.08249
null
https://arxiv.org/abs/2109.08249v1
https://arxiv.org/pdf/2109.08249v1.pdf
Regularized Training of Nearest Neighbor Language Models
Including memory banks in a natural language processing architecture increases model capacity by equipping it with additional data at inference time. In this paper, we build upon $k$NN-LM \citep{khandelwal20generalization}, which uses a pre-trained language model together with an exhaustive $k$NN search through the training data (memory bank) to achieve state-of-the-art results. We investigate whether we can improve the $k$NN-LM performance by instead training a LM with the knowledge that we will be using a $k$NN post-hoc. We achieved significant improvement using our method on language modeling tasks on \texttt{WIKI-2} and \texttt{WIKI-103}. The main phenomenon that we encounter is that adding a simple L2 regularization on the activations (not weights) of the model, a transformer, improves the post-hoc $k$NN classification performance. We explore some possible reasons for this improvement. In particular, we find that the added L2 regularization seems to improve the performance for high-frequency words without deteriorating the performance for low frequency ones.
['Josh Susskind', 'Shuangfei Zhai', 'Walter Talbott', 'Jean-Francois Ton']
2021-09-16
null
https://aclanthology.org/2022.naacl-srw.4
https://aclanthology.org/2022.naacl-srw.4.pdf
naacl-acl-2022-7
['l2-regularization']
['methodology']
[ 2.32174490e-02 2.72794217e-01 -2.59245280e-02 -4.30444568e-01 -9.22675908e-01 -4.64148164e-01 4.15942311e-01 2.48868003e-01 -1.13203537e+00 3.76101017e-01 5.18540777e-02 -7.85734296e-01 3.69466953e-02 -6.60036087e-01 -8.30869079e-01 -5.15284479e-01 -2.28725113e-02 4.34121221e-01 2.79068649e-01 -1.34334788e-01 2.74534255e-01 1.70488089e-01 -1.19583642e+00 6.92690313e-01 5.05023062e-01 8.95236790e-01 5.84225297e-01 5.39352834e-01 -4.40707445e-01 6.14793062e-01 -5.09315968e-01 -4.08728033e-01 1.86174020e-01 -3.52303147e-01 -9.38235700e-01 -3.76863897e-01 4.51969892e-01 4.72676121e-02 -2.01140776e-01 8.09127152e-01 3.56284291e-01 3.50915700e-01 6.77882552e-01 -4.40798044e-01 -4.52712655e-01 1.16522944e+00 -4.96350646e-01 4.93457496e-01 -1.28914088e-01 1.16537668e-01 1.12042773e+00 -1.20333815e+00 4.32723165e-01 1.25727296e+00 6.79135382e-01 5.13789654e-01 -1.11665654e+00 -6.12624526e-01 5.03383279e-01 1.57913014e-01 -1.62724996e+00 -7.29518414e-01 4.90954399e-01 -1.81735411e-01 1.69213510e+00 7.97945820e-03 1.62613735e-01 5.42409778e-01 -3.88194597e-03 8.29281390e-01 8.76599967e-01 -1.01346040e+00 -7.44753703e-03 3.20486933e-01 4.67981070e-01 7.78460979e-01 1.32110808e-02 -9.16855335e-02 -5.39727151e-01 -4.08226103e-02 4.54548627e-01 -5.20677626e-01 1.75450724e-02 2.51320601e-01 -8.74027789e-01 8.51918817e-01 2.98220843e-01 7.37933993e-01 -1.24417812e-01 3.53283942e-01 4.69520301e-01 4.00976360e-01 4.33494955e-01 6.53181434e-01 -8.67019653e-01 -5.21362759e-02 -1.13244343e+00 -7.41723701e-02 5.56884646e-01 7.01722741e-01 7.14208722e-01 2.46266931e-01 -8.76535103e-02 1.19588768e+00 3.67735356e-01 3.30161989e-01 8.37249339e-01 -6.03352368e-01 7.14420259e-01 4.27311718e-01 -2.61678427e-01 -6.48992658e-01 -6.34293795e-01 -4.77629453e-01 -5.78293562e-01 -1.79819673e-01 5.52266717e-01 -2.76651114e-01 -9.80449021e-01 2.07833385e+00 -4.70683247e-01 -2.20455930e-01 3.70250456e-02 4.36629564e-01 6.60895824e-01 8.62991035e-01 4.93425161e-01 -2.92711318e-01 1.14840746e+00 -1.06924725e+00 -4.71132040e-01 -9.12038088e-01 1.34832764e+00 -7.65411615e-01 1.49712026e+00 4.94809091e-01 -1.28164887e+00 -7.02726841e-01 -1.04508746e+00 -4.68598716e-02 -6.92941904e-01 4.05353904e-01 4.98756081e-01 6.04956985e-01 -1.20959532e+00 4.72894996e-01 -5.72615266e-01 -3.79178286e-01 3.53150405e-02 4.90910381e-01 -1.56041548e-01 -9.18678194e-02 -1.48257911e+00 1.30735362e+00 5.93884468e-01 6.38110042e-02 -4.31603819e-01 -4.83612418e-01 -8.57598424e-01 1.80693701e-01 2.70706117e-01 -2.74399757e-01 1.15391695e+00 -8.18081260e-01 -1.34927034e+00 1.16285920e+00 -5.33618569e-01 -5.19268453e-01 1.31828887e-02 -1.75257057e-01 -4.49801832e-01 -2.51795679e-01 -3.22577387e-01 8.27488720e-01 6.18411899e-01 -1.07984519e+00 -5.31307399e-01 -2.59458452e-01 -1.55576497e-01 1.00907879e-02 -6.84984148e-01 6.17067982e-03 -5.10180771e-01 -8.13526750e-01 1.34306118e-01 -9.69150245e-01 -1.15166768e-01 -6.12286866e-01 -1.11412801e-01 -5.27808368e-01 2.86662698e-01 -6.73909128e-01 1.86975193e+00 -2.14321375e+00 -1.01182766e-01 4.29023802e-01 -2.80631483e-01 5.95615327e-01 -5.04321933e-01 3.59924644e-01 -1.49107978e-01 3.88817519e-01 -1.67599663e-01 -5.60119808e-01 -4.73612100e-02 1.20922171e-01 -2.82249659e-01 2.00086772e-01 2.16374189e-01 9.91608083e-01 -5.06543517e-01 -3.81128579e-01 -4.88111712e-02 2.38193870e-01 -5.92232347e-01 -6.24824353e-02 -3.51181060e-01 -1.91739231e-01 6.87843859e-02 1.05102099e-01 4.68246251e-01 -2.54696846e-01 3.07157546e-01 -2.97765359e-02 -1.95262551e-01 8.72572958e-01 -1.02714705e+00 1.74870658e+00 -8.05171490e-01 6.95769191e-01 -9.54001173e-02 -1.15597427e+00 7.37574577e-01 2.63451576e-01 -3.77280340e-02 -8.95313025e-01 7.53635392e-02 3.81683499e-01 5.38402975e-01 -2.12779582e-01 3.68260235e-01 -2.60362029e-01 -7.59396479e-02 6.09537780e-01 2.34456658e-01 -6.82382062e-02 3.05851787e-01 6.51599243e-02 9.99918401e-01 -1.69829726e-01 6.61869571e-02 -7.08405674e-01 6.20549381e-01 -1.09858565e-01 8.63097608e-02 1.03654361e+00 2.67971754e-01 3.42297941e-01 2.98359543e-01 -1.30909428e-01 -8.43844533e-01 -9.36060309e-01 -2.48398110e-01 1.62903583e+00 -5.68315804e-01 -6.14714742e-01 -6.35898113e-01 -3.82996649e-01 -2.40358070e-01 1.23820817e+00 -4.61782306e-01 -5.33758163e-01 -9.20836091e-01 -1.09330487e+00 7.86182344e-01 6.63133204e-01 4.81833607e-01 -1.22915828e+00 -3.34648341e-01 3.10359031e-01 -6.33503050e-02 -9.47958052e-01 -2.86176473e-01 8.94565225e-01 -9.74067688e-01 -3.20407838e-01 -5.75981677e-01 -1.14650536e+00 7.60020435e-01 -1.03554666e-01 1.24333096e+00 3.16324741e-01 3.50803323e-02 3.70277907e-03 -2.00209185e-01 -4.65231091e-01 -3.41071427e-01 5.71661472e-01 1.17883407e-01 -4.71974164e-01 6.33080244e-01 -2.86724806e-01 -2.16305479e-01 5.58672026e-02 -8.93775105e-01 -1.14420287e-01 8.42501998e-01 7.95220733e-01 3.67246240e-01 2.57415920e-01 6.66553378e-01 -1.07169437e+00 5.71519494e-01 -1.91592321e-01 -3.00403923e-01 1.55693501e-01 -6.87622905e-01 4.56292450e-01 7.13063717e-01 -6.43498480e-01 -8.54938447e-01 -1.59748316e-01 -4.64552164e-01 -3.11097130e-03 1.44693792e-01 9.82517540e-01 -4.06646468e-02 2.82153264e-02 7.49544442e-01 2.62428105e-01 -4.47056890e-01 -6.71277285e-01 2.98178375e-01 5.05244017e-01 1.10979393e-01 -5.52870512e-01 5.93033731e-01 -7.09619257e-04 -4.18417007e-01 -9.59606588e-01 -9.51135278e-01 -3.71885002e-01 -6.16240680e-01 2.04069063e-01 5.57127833e-01 -7.82397211e-01 -5.10752738e-01 4.74036902e-01 -1.22836614e+00 -9.15147841e-01 -3.31362709e-02 5.02781510e-01 -2.77783394e-01 6.82706609e-02 -8.36318254e-01 -8.26704085e-01 -2.14678675e-01 -8.75701666e-01 5.78850508e-01 -1.82060957e-01 -5.64497471e-01 -1.15759647e+00 -1.55000001e-01 5.66736907e-02 6.76118731e-01 -8.24081004e-01 1.53890657e+00 -9.10922885e-01 -3.65943834e-02 -2.32288390e-01 -1.75932825e-01 6.07084215e-01 -1.57319665e-01 -3.63503993e-01 -1.23473275e+00 -2.34450430e-01 8.09532851e-02 -3.36432427e-01 1.28914022e+00 3.54998827e-01 1.38983810e+00 -3.02290857e-01 -2.33523503e-01 3.02162409e-01 1.36050630e+00 2.06589907e-01 7.53581405e-01 2.99201727e-01 4.67683405e-01 5.82406878e-01 2.22463816e-01 5.17051928e-02 2.17677906e-01 6.61545575e-01 -1.10202335e-01 -9.74315256e-02 9.49313492e-03 -2.01038420e-01 5.70408404e-01 1.06753838e+00 2.59686083e-01 -4.63997632e-01 -1.22035229e+00 7.11299598e-01 -1.67905819e+00 -5.72630942e-01 7.82304034e-02 2.24873137e+00 1.17188668e+00 6.85880184e-01 -2.10975349e-01 1.49971172e-01 3.18723291e-01 7.43476301e-02 -2.43011490e-01 -7.85253644e-01 -3.28620255e-01 5.32646239e-01 6.03126049e-01 8.86504352e-01 -8.28571498e-01 1.42826045e+00 6.38805294e+00 1.16057003e+00 -1.33561647e+00 1.43934220e-01 7.94647455e-01 -2.76755095e-01 -2.28788197e-01 -1.43097639e-01 -1.12200093e+00 3.08544636e-01 1.56076050e+00 1.33994892e-01 4.20718193e-01 5.59820294e-01 2.06149146e-01 -3.64331365e-01 -1.16991401e+00 8.79294693e-01 1.76348090e-01 -1.30453217e+00 3.33356798e-01 -9.83279869e-02 3.82551968e-01 1.49499133e-01 1.59075618e-01 7.09842026e-01 1.83326527e-01 -1.32691777e+00 5.41223466e-01 3.15031916e-01 6.04766309e-01 -6.67198658e-01 6.90408826e-01 6.37729347e-01 -9.54565525e-01 6.21512905e-03 -5.15120506e-01 -2.24631324e-01 9.30998772e-02 6.18429482e-01 -6.63517892e-01 -7.25335702e-02 6.04608893e-01 1.45132571e-01 -7.49395251e-01 6.69349790e-01 -1.89171210e-01 9.44455683e-01 -5.02366722e-01 -2.04181537e-01 3.82556885e-01 2.07922623e-01 1.03671864e-01 1.68916512e+00 2.21570402e-01 2.09258407e-01 -7.25021213e-02 6.39416218e-01 -6.57374933e-02 3.63080531e-01 -4.55799997e-01 -2.34279618e-01 2.94100195e-01 8.64590108e-01 -6.38851464e-01 -5.87957740e-01 -3.93343866e-01 8.13632190e-01 7.18415737e-01 3.74282449e-01 -5.83214521e-01 -3.87637466e-01 2.10957110e-01 3.17491114e-01 4.25261825e-01 -5.04952908e-01 -5.48853517e-01 -9.28768575e-01 -8.30187723e-02 -6.47471726e-01 4.00009155e-01 -7.14936614e-01 -9.98225212e-01 6.69161916e-01 -1.55030601e-02 -4.14562911e-01 -3.34941894e-01 -9.25708592e-01 -3.72596681e-01 1.14719093e+00 -1.33591521e+00 -8.00301433e-01 4.13529158e-01 3.98725450e-01 5.32765329e-01 -1.65191710e-01 1.05954051e+00 5.23836315e-01 -5.31612873e-01 9.88279045e-01 -1.19705331e-02 2.61509061e-01 7.16280639e-01 -9.41363215e-01 2.96670288e-01 8.02763522e-01 3.51281971e-01 1.06805098e+00 5.90356171e-01 -3.93896878e-01 -1.11524224e+00 -9.31730866e-01 1.53158593e+00 -4.25615788e-01 6.40962660e-01 -5.45972586e-01 -1.05511761e+00 8.47898066e-01 2.11503774e-01 -2.75422573e-01 7.78981090e-01 2.53079802e-01 -3.18483025e-01 4.00104839e-03 -7.87247360e-01 7.44861782e-01 9.71983969e-01 -7.04752564e-01 -7.59603560e-01 2.28711322e-01 7.81487107e-01 -5.23187295e-02 -5.41552901e-01 3.49006116e-01 2.33705595e-01 -3.93865168e-01 8.01440895e-01 -7.69486547e-01 2.24631742e-01 -3.39426063e-02 -2.59862959e-01 -1.32757509e+00 -3.78816128e-01 -2.01470315e-01 5.00913262e-02 1.11756837e+00 1.18511546e+00 -5.63299596e-01 6.12857044e-01 7.59920180e-01 -1.91405624e-01 -6.24519944e-01 -8.94715011e-01 -7.32097149e-01 7.30961263e-01 -8.68095100e-01 -6.26278147e-02 6.92169964e-01 3.98908675e-01 4.96110886e-01 -1.85105354e-01 -1.42282933e-01 1.10304222e-01 -4.26899701e-01 2.26270229e-01 -8.66762340e-01 -4.25064832e-01 -6.21694982e-01 1.17197879e-01 -1.40470481e+00 4.29809898e-01 -1.06631184e+00 2.17893645e-01 -1.29270482e+00 -4.11639251e-02 -6.08876705e-01 -5.01844883e-01 8.69162142e-01 -7.74664283e-02 2.19736412e-01 2.28094101e-01 6.42124331e-03 -4.81652647e-01 1.85035691e-01 7.31617570e-01 -3.82441431e-02 -2.59477407e-01 -1.41440883e-01 -7.28919029e-01 8.11406076e-01 8.45965385e-01 -5.01284361e-01 -2.97330558e-01 -1.01308906e+00 4.71214980e-01 -3.69346172e-01 -2.94315387e-02 -1.00789320e+00 4.28063542e-01 1.71892315e-01 3.17586839e-01 -3.44181746e-01 4.02660221e-01 -6.02356017e-01 -3.79330337e-01 5.37130475e-01 -7.23739803e-01 2.07947671e-01 8.36528897e-01 -2.97656450e-02 -1.03891425e-01 -6.74463034e-01 9.61134076e-01 -4.36655432e-01 -7.73919761e-01 -1.88079149e-01 -7.74200976e-01 1.21160306e-01 3.24692428e-01 1.74696967e-01 -2.41218761e-01 -1.86374515e-01 -8.43429625e-01 1.04487643e-01 6.80427700e-02 3.60460937e-01 2.85224438e-01 -9.46764946e-01 -5.88205457e-01 3.64974529e-01 6.23035505e-02 -1.93375349e-01 -3.04758195e-02 7.86827922e-01 -1.85394868e-01 8.74218524e-01 2.38180891e-01 -1.82487383e-01 -9.13549006e-01 3.15140784e-01 4.84008998e-01 -5.40225983e-01 -1.34089187e-01 1.20510828e+00 1.05229825e-01 -5.17759562e-01 4.41112190e-01 -5.36022365e-01 -8.62978920e-02 2.22244829e-01 4.81131911e-01 6.64191619e-02 5.09567738e-01 -3.66204947e-01 -5.44057071e-01 5.10715067e-01 -3.72152388e-01 -4.06390190e-01 1.21319604e+00 -8.39683488e-02 -1.62611172e-01 8.46695721e-01 1.29561365e+00 1.78718325e-02 -5.55229425e-01 -4.87775236e-01 3.38506848e-01 2.88639888e-02 1.99993074e-01 -1.04509544e+00 -1.07003367e+00 1.05615366e+00 4.03146863e-01 -1.61987543e-01 8.34775150e-01 5.57885617e-02 6.60374105e-01 7.91157424e-01 3.14062268e-01 -1.49971569e+00 -8.09004679e-02 1.16448653e+00 6.75188005e-01 -1.22207701e+00 -1.65470660e-01 -3.88497598e-02 -3.75324249e-01 1.05663455e+00 7.98274755e-01 -1.72359824e-01 9.96674895e-01 2.95945793e-01 1.58834010e-01 -2.20696218e-02 -9.18234527e-01 -1.35935768e-01 2.46778041e-01 1.21485166e-01 9.09719229e-01 -1.60786033e-01 -6.32450223e-01 9.50629294e-01 -3.07946652e-01 -2.19670519e-01 2.12409988e-01 8.39946330e-01 -7.16457665e-01 -1.21943355e+00 -1.13168441e-01 6.62637949e-01 -4.91189033e-01 -8.27195406e-01 -3.00227016e-01 7.54053473e-01 4.13127653e-02 9.50648129e-01 1.11279972e-01 -3.09744090e-01 2.80777127e-01 7.65830040e-01 2.56921440e-01 -8.97677183e-01 -7.66760170e-01 1.37177944e-01 1.74049944e-01 -1.93616584e-01 -5.47951832e-03 -4.57344025e-01 -1.61351812e+00 -2.23083004e-01 -2.83653796e-01 2.16586694e-01 6.43548310e-01 1.08537126e+00 1.43753320e-01 4.60016787e-01 3.11001576e-02 -5.76173484e-01 -5.09458244e-01 -1.21240211e+00 -4.78898734e-01 1.83154508e-01 -4.15819325e-02 -4.27367330e-01 -6.06432557e-01 3.39023280e-03]
[10.716266632080078, 8.797856330871582]
bda18350-bff9-4204-9caf-6acca65effe7
local-guided-global-paired-similarity
null
null
http://openaccess.thecvf.com//content/CVPR2023/html/Choi_Local-Guided_Global_Paired_Similarity_Representation_for_Visual_Reinforcement_Learning_CVPR_2023_paper.html
http://openaccess.thecvf.com//content/CVPR2023/papers/Choi_Local-Guided_Global_Paired_Similarity_Representation_for_Visual_Reinforcement_Learning_CVPR_2023_paper.pdf
Local-Guided Global: Paired Similarity Representation for Visual Reinforcement Learning
Recent vision-based reinforcement learning (RL) methods have found extracting high-level features from raw pixels with self-supervised learning to be effective in learning policies. However, these methods focus on learning global representations of images, and disregard local spatial structures present in the consecutively stacked frames. In this paper, we propose a novel approach, termed self-supervised Paired Similarity Representation Learning (PSRL) for effectively encoding spatial structures in an unsupervised manner. Given the input frames, the latent volumes are first generated individually using an encoder, and they are used to capture the variance in terms of local spatial structures, i.e., correspondence maps among multiple frames. This enables for providing plenty of fine-grained samples for training the encoder of deep RL. We further attempt to learn the global semantic representations in the global prediction module that predicts future state representations using action vector as a medium. The proposed method imposes similarity constraints on the three latent volumes; transformed query representations by estimated pixel-wise correspondence, predicted query representations from the global prediction model, and target representations of future state, guiding global prediction with locality-inherent volume. Experimental results on complex tasks in Atari Games and DeepMind Control Suite demonstrate that the RL methods are significantly boosted by the proposed self-supervised learning of structured representations.
['Dongbo Min', 'Kwanghoon Sohn', 'Sangryul Jeon', 'Wonil Song', 'Hunsang Lee', 'Hyesong Choi']
2023-01-01
null
null
null
cvpr-2023-1
['atari-games']
['playing-games']
[ 1.79649353e-01 -7.54972994e-02 -5.23990393e-01 -3.82227391e-01 -5.53182185e-01 -1.65592562e-02 8.95342469e-01 -1.09098613e-01 -2.48395070e-01 6.91919506e-01 3.23099911e-01 2.30539873e-01 -1.81973517e-01 -7.85777271e-01 -9.13360715e-01 -8.55541170e-01 -2.78021336e-01 2.66058981e-01 4.98474807e-01 -1.98840633e-01 6.12409115e-01 4.30122375e-01 -1.78042102e+00 5.71960151e-01 6.46639347e-01 1.23248279e+00 5.82855225e-01 4.93079007e-01 -7.24277422e-02 1.59590042e+00 -3.47366393e-01 4.94232982e-01 3.18060011e-01 -7.03711867e-01 -7.11054802e-01 2.95743078e-01 4.89231229e-01 -6.11456096e-01 -6.46801829e-01 7.70565510e-01 5.61363585e-02 6.12395883e-01 7.69686162e-01 -1.14321280e+00 -5.45809686e-01 1.03018701e-01 -6.50833905e-01 5.93001544e-01 1.85429618e-01 5.83426774e-01 1.04148710e+00 -7.67831981e-01 7.67340899e-01 1.38695872e+00 2.62188315e-01 6.16858006e-01 -1.19366670e+00 -3.92627448e-01 3.87097716e-01 5.60734689e-01 -1.20070934e+00 -4.28782791e-01 1.01604664e+00 -5.14307320e-01 1.25768411e+00 -1.22239247e-01 7.30696559e-01 9.21999276e-01 7.79149950e-01 8.47927332e-01 1.45092463e+00 -2.17139646e-01 5.94510853e-01 -6.44283369e-02 -1.82107955e-01 9.70340192e-01 -3.60466719e-01 6.23171389e-01 -8.82950127e-01 1.29933253e-01 1.21160960e+00 1.58097759e-01 -7.41228536e-02 -9.33041275e-01 -1.12993109e+00 1.02010357e+00 6.48039401e-01 2.72606611e-01 -6.13837302e-01 4.76309836e-01 2.94632703e-01 1.08094491e-01 5.52573681e-01 1.42761543e-01 -3.43584061e-01 9.43170413e-02 -9.52104867e-01 3.22461903e-01 1.38222903e-01 6.72897160e-01 1.15290952e+00 3.89962673e-01 -5.82612038e-01 5.64084113e-01 5.02367258e-01 4.49177563e-01 1.13116920e+00 -1.10725570e+00 4.14488077e-01 5.09727061e-01 -2.07373910e-02 -1.34027493e+00 -1.85724035e-01 -2.35356033e-01 -6.90249503e-01 5.79858720e-01 1.08317830e-01 1.83734953e-01 -1.07310200e+00 1.71551120e+00 2.55340874e-01 6.57533586e-01 2.21055120e-01 9.23918903e-01 4.53627795e-01 9.69306231e-01 1.98984295e-01 -3.87973785e-01 6.69917881e-01 -1.15416598e+00 -6.35346651e-01 -4.35788393e-01 5.53832412e-01 -2.40035951e-01 8.09795737e-01 -2.67917123e-02 -1.05193162e+00 -1.05074131e+00 -8.81804585e-01 1.52953625e-01 -2.54301697e-01 8.33143666e-02 5.34661293e-01 -6.12010732e-02 -1.12252820e+00 1.02072394e+00 -9.77878094e-01 -1.97595119e-01 5.84606707e-01 2.92896122e-01 -1.41115010e-01 2.21374512e-01 -1.09932041e+00 1.00098896e+00 4.72648412e-01 -5.60705736e-02 -1.36179924e+00 -6.21362984e-01 -1.04467142e+00 -9.59156603e-02 2.23363668e-01 -3.76874149e-01 8.85613859e-01 -1.25154686e+00 -1.67402840e+00 5.04155397e-01 -1.57708243e-01 -8.05541635e-01 1.30192444e-01 -8.62390995e-02 1.14202090e-01 4.48714584e-01 3.75302464e-01 9.59917307e-01 1.26161087e+00 -1.29266429e+00 -7.58673251e-01 -4.97535378e-01 -8.04989710e-02 6.46731794e-01 -6.96675256e-02 -4.91586417e-01 -7.52625316e-02 -5.49067855e-01 1.53290436e-01 -7.21112192e-01 -5.54527104e-01 7.52392858e-02 8.67012367e-02 -2.70924449e-01 1.02368319e+00 -7.01700628e-01 8.54884863e-01 -1.99570084e+00 3.18453640e-01 2.04559237e-01 7.30420351e-02 1.81846902e-01 -2.26607203e-01 1.75326224e-02 -7.50238895e-02 -5.43301105e-01 -9.63848755e-02 -1.78947911e-01 -2.93861896e-01 5.68625450e-01 -6.40363455e-01 4.30609345e-01 3.25313210e-01 1.18867755e+00 -1.09535992e+00 -6.77523136e-01 7.78243542e-01 2.76072443e-01 -4.50702637e-01 5.32591462e-01 -3.61988574e-01 7.31623471e-01 -6.84176385e-01 2.35268801e-01 2.28459567e-01 -9.50850025e-02 5.17640188e-02 -1.79727927e-01 -8.14988017e-02 6.80439994e-02 -9.83237624e-01 1.99542022e+00 -4.17722970e-01 4.68363881e-01 -4.37032491e-01 -1.50392091e+00 1.21312523e+00 -1.77072376e-01 7.25899458e-01 -1.20389783e+00 -1.41832177e-02 -2.70363748e-01 -4.05431122e-01 -4.01013792e-01 2.86208600e-01 -1.04741938e-01 9.81120542e-02 2.67418563e-01 3.85497242e-01 -1.04055807e-01 -4.45152558e-02 1.12370640e-01 7.51427948e-01 7.15969145e-01 3.12144518e-01 -1.60257339e-01 5.18173754e-01 9.59120020e-02 5.33758402e-01 6.45715177e-01 -5.07293224e-01 4.43590283e-01 2.10473090e-01 -6.17366672e-01 -1.00679672e+00 -1.24190140e+00 2.66041636e-01 9.44180369e-01 3.56256187e-01 -3.77173036e-01 -4.75272536e-01 -6.84751570e-01 -2.10953727e-01 5.92012227e-01 -7.74859548e-01 -5.42016208e-01 -5.85805237e-01 -2.05645695e-01 -1.05544403e-01 6.96556091e-01 7.09455729e-01 -1.32118213e+00 -1.06557989e+00 2.62590826e-01 1.84387844e-02 -1.01477277e+00 -2.60203212e-01 2.78730929e-01 -1.27578282e+00 -9.57796752e-01 -5.37867963e-01 -7.87908792e-01 4.93782192e-01 3.83524030e-01 8.09205770e-01 -2.41296053e-01 -4.57564890e-01 6.49551570e-01 -2.36528918e-01 1.03596918e-01 -1.22186989e-01 -5.55054426e-01 1.56238481e-01 2.60237694e-01 1.46717921e-01 -5.26169896e-01 -7.56765842e-01 8.98047686e-02 -6.05228543e-01 1.78340718e-01 6.11305475e-01 1.09630442e+00 9.70686555e-01 -6.81432933e-02 3.67675364e-01 -5.84524691e-01 3.22295725e-01 -3.73140872e-01 -5.04360974e-01 1.93837389e-01 -5.73695719e-01 3.52663964e-01 7.21432388e-01 -3.00337166e-01 -1.28749299e+00 3.20964038e-01 3.07023853e-01 -9.32110131e-01 -3.12512010e-01 1.77547097e-01 2.33662277e-01 -7.09522516e-02 5.34034431e-01 8.22728157e-01 2.83218443e-01 3.30993533e-03 6.39777005e-01 1.82747722e-01 3.68417889e-01 -5.18540978e-01 6.11810446e-01 5.62439382e-01 5.19152358e-02 -6.44477010e-01 -7.47838795e-01 -5.11650681e-01 -8.21435094e-01 -4.08741415e-01 1.09964442e+00 -9.91812944e-01 -4.31252956e-01 3.06344092e-01 -7.80392170e-01 -4.70833689e-01 -7.83728182e-01 5.15028358e-01 -1.35327852e+00 4.30848271e-01 -3.36174905e-01 -8.06352139e-01 -1.24362990e-01 -1.14230132e+00 1.05250251e+00 3.06448758e-01 -4.59070690e-02 -1.19588220e+00 3.81821781e-01 2.73438513e-01 2.25410849e-01 1.50538549e-01 6.93186343e-01 -2.93948561e-01 -8.25809717e-01 2.87423790e-01 1.45081757e-02 4.52438921e-01 2.44313061e-01 -2.91086465e-01 -8.99867177e-01 -2.92694569e-01 9.79289412e-02 -7.15664625e-01 9.94382203e-01 7.15323210e-01 1.38696396e+00 -2.77760655e-01 -2.61138529e-01 5.08892298e-01 1.40346789e+00 3.64076495e-01 7.47570992e-01 3.44880790e-01 7.11417556e-01 7.06434488e-01 8.74569714e-01 6.53932631e-01 2.20206514e-01 6.89209998e-01 5.91804206e-01 -6.98198378e-02 -1.46825179e-01 -5.32716930e-01 5.44298649e-01 5.20785809e-01 -1.59340113e-01 4.57111418e-01 -5.28290868e-01 4.56383586e-01 -2.27106333e+00 -1.37343693e+00 4.04626995e-01 1.98443389e+00 6.46822751e-01 -3.55746262e-02 -1.32735908e-01 -1.44302845e-01 5.56947410e-01 6.15516186e-01 -8.12795639e-01 -2.52802461e-01 7.95539171e-02 2.93319881e-01 3.24814051e-01 4.37741190e-01 -1.21155834e+00 1.27122998e+00 6.05923986e+00 9.82846200e-01 -9.95576203e-01 -5.18895648e-02 9.77325082e-01 -2.36390755e-02 -1.51376594e-02 1.06646910e-01 -6.88679874e-01 3.16747695e-01 9.14260805e-01 3.72190624e-02 4.34543848e-01 9.66593206e-01 5.13224900e-01 -2.93123007e-01 -1.02266490e+00 1.12949932e+00 2.19561204e-01 -1.66360760e+00 4.22625870e-01 -6.64141402e-02 9.94565308e-01 -2.08146214e-01 3.88240993e-01 4.77422029e-01 2.03939959e-01 -9.61602151e-01 5.59600472e-01 9.30702686e-01 4.30427194e-01 -7.20215797e-01 2.84804374e-01 3.93709421e-01 -1.26271689e+00 -3.66681516e-01 -6.94512904e-01 -1.29437104e-01 -2.28189215e-01 -8.54417905e-02 -5.06650507e-01 2.90594667e-01 6.74089611e-01 1.23744047e+00 -4.93660867e-01 5.39445400e-01 5.24119660e-02 2.31805742e-01 2.01701343e-01 1.05898321e-01 6.05032742e-01 -2.50664413e-01 2.08182335e-01 6.85351849e-01 1.07304109e-02 6.22079857e-02 4.60664511e-01 1.00755620e+00 4.84761804e-01 -5.06813489e-02 -8.42418611e-01 2.43122697e-01 -4.07453068e-02 1.05705464e+00 -5.88451028e-01 -3.82852554e-01 -3.35629880e-01 1.03292966e+00 5.77024996e-01 4.65823561e-01 -8.52081776e-01 2.41644710e-01 5.80112457e-01 -5.89850545e-02 4.74376768e-01 -2.55240619e-01 -2.77363113e-03 -1.12252891e+00 -2.74857879e-01 -5.57106018e-01 2.07128108e-01 -8.63570988e-01 -1.13768935e+00 4.06641811e-01 1.16268218e-01 -1.55257976e+00 -7.95047641e-01 -5.76376200e-01 -7.46815145e-01 7.06132412e-01 -1.74528801e+00 -1.10684884e+00 -2.11471736e-01 1.12376094e+00 1.03801703e+00 -6.13557458e-01 6.24222338e-01 -4.49850291e-01 -3.10937554e-01 2.53194541e-01 2.40454733e-01 -5.57754412e-02 3.71872067e-01 -1.10761213e+00 -6.11878820e-02 6.13244593e-01 4.47310418e-01 2.81261116e-01 4.04280752e-01 -6.09632969e-01 -1.29454684e+00 -1.36913478e+00 4.19350922e-01 -2.26005703e-01 5.90956986e-01 -3.44147533e-02 -6.67454481e-01 5.48810005e-01 1.04490042e-01 3.99858266e-01 2.95340151e-01 -4.23828840e-01 -1.56959936e-01 -2.41616011e-01 -8.99561167e-01 4.96487916e-01 8.54353249e-01 -7.89148152e-01 -6.81579351e-01 4.63192195e-01 4.36554819e-01 -2.06297725e-01 -6.16619170e-01 1.64409846e-01 3.40300024e-01 -1.14008522e+00 1.18003488e+00 -6.55045688e-01 6.69085026e-01 -2.81791419e-01 -3.03602546e-01 -1.30539811e+00 -5.75396240e-01 -1.64662048e-01 -3.77667069e-01 8.29030097e-01 -6.63360134e-02 -1.84902072e-01 1.06170511e+00 3.17699730e-01 -9.93383750e-02 -9.13160563e-01 -9.76917803e-01 -7.08800972e-01 -6.08876767e-03 -1.82171300e-01 6.47016764e-02 6.91394806e-01 -5.54739460e-02 2.77824908e-01 -5.14399469e-01 6.98291957e-02 6.81409597e-01 2.76990414e-01 5.83931923e-01 -6.89272583e-01 -3.98338228e-01 -1.50238127e-01 -8.11995685e-01 -1.25146651e+00 4.82466519e-01 -7.36230969e-01 1.84496060e-01 -1.52729917e+00 2.75668293e-01 -2.18129829e-01 -4.95877624e-01 5.02060235e-01 4.85535115e-02 4.45060022e-02 3.66360962e-01 5.08235335e-01 -8.32952917e-01 1.10344374e+00 1.44837272e+00 -3.10853660e-01 -2.94811696e-01 -1.14000581e-01 -7.78574795e-02 7.02390134e-01 8.39420378e-01 -1.15133576e-01 -9.51094806e-01 -1.44344956e-01 -5.77595770e-01 2.99487770e-01 7.06718504e-01 -1.30804801e+00 1.55596718e-01 -6.39862180e-01 8.17338943e-01 -6.56127155e-01 5.53796887e-01 -8.53620827e-01 -3.71158570e-01 6.18409038e-01 -7.17001557e-01 -1.60136059e-01 -7.94211682e-03 8.86199832e-01 -3.75663221e-01 1.94940006e-03 9.08975244e-01 -3.97104025e-01 -1.44422710e+00 5.22594988e-01 -2.82754421e-01 -4.56769168e-02 1.34113634e+00 -5.56727648e-01 -2.23722681e-02 -3.39701384e-01 -8.56660604e-01 6.35663643e-02 -1.21035529e-02 4.84422088e-01 1.28817201e+00 -1.49434364e+00 -3.15995932e-01 4.19326544e-01 6.38310239e-02 -1.74301177e-01 5.10164917e-01 4.49828863e-01 -3.04298729e-01 4.46340621e-01 -8.70629013e-01 -8.91046226e-01 -8.85716498e-01 6.44598842e-01 4.18015301e-01 -4.07272905e-01 -7.73666263e-01 7.40700483e-01 6.91684902e-01 -1.36753529e-01 1.32923707e-01 -3.13749194e-01 -5.65880358e-01 -1.69270560e-01 5.64922988e-01 1.25511810e-01 -3.57127964e-01 -9.25857604e-01 -1.16727792e-01 7.76892126e-01 -1.68610111e-01 -9.40446109e-02 1.40577209e+00 -2.69755870e-01 2.15868294e-01 7.14152515e-01 1.35683537e+00 -7.69406855e-01 -2.08873534e+00 -3.93597782e-01 -1.66482002e-01 -6.72855556e-01 2.97673255e-01 -3.42608094e-01 -1.28408921e+00 9.36452329e-01 1.03940916e+00 -4.25505906e-01 1.08341467e+00 8.54152590e-02 5.15863597e-01 3.07931811e-01 5.49737394e-01 -1.31206048e+00 7.61950493e-01 5.49984097e-01 8.75267148e-01 -1.27307558e+00 -5.18573895e-02 7.75233209e-02 -9.96227205e-01 1.09623849e+00 1.13444400e+00 -6.48087740e-01 5.69879889e-01 -2.43707001e-01 -2.53595203e-01 -1.91629156e-01 -9.04054761e-01 -2.21689492e-01 3.87940228e-01 9.86732423e-01 6.35902062e-02 -2.05474988e-01 1.22697823e-01 2.22739413e-01 2.44357556e-01 -2.15036601e-01 2.18748525e-01 9.01854634e-01 -7.87762225e-01 -6.52023494e-01 -2.08331928e-01 4.69753593e-01 2.01021165e-01 9.02389064e-02 9.44666266e-02 5.82519054e-01 1.39319316e-01 6.25049055e-01 4.52133596e-01 -4.85646427e-01 1.31066784e-01 -2.41426930e-01 6.60084128e-01 -5.61332941e-01 -2.01445520e-01 6.55830279e-02 -4.44265097e-01 -1.12497032e+00 -5.50741076e-01 -6.25123680e-01 -1.51723588e+00 1.54795170e-01 1.76401347e-01 -3.22264098e-02 3.22523147e-01 1.15791345e+00 2.14751914e-01 5.83886445e-01 9.76145446e-01 -1.20675397e+00 -7.66696930e-01 -6.35774791e-01 -6.41103625e-01 6.25295401e-01 3.25747758e-01 -9.75017130e-01 -1.30986392e-01 3.34654093e-01]
[8.615937232971191, 0.4875110983848572]
dbe069bd-7930-4c6d-81b1-600510c30501
contrastive-identity-aware-learning-for-multi
2211.12712
null
https://arxiv.org/abs/2211.12712v2
https://arxiv.org/pdf/2211.12712v2.pdf
Contrastive Identity-Aware Learning for Multi-Agent Value Decomposition
Value Decomposition (VD) aims to deduce the contributions of agents for decentralized policies in the presence of only global rewards, and has recently emerged as a powerful credit assignment paradigm for tackling cooperative Multi-Agent Reinforcement Learning (MARL) problems. One of the main challenges in VD is to promote diverse behaviors among agents, while existing methods directly encourage the diversity of learned agent networks with various strategies. However, we argue that these dedicated designs for agent networks are still limited by the indistinguishable VD network, leading to homogeneous agent behaviors and thus downgrading the cooperation capability. In this paper, we propose a novel Contrastive Identity-Aware learning (CIA) method, explicitly boosting the credit-level distinguishability of the VD network to break the bottleneck of multi-agent diversity. Specifically, our approach leverages contrastive learning to maximize the mutual information between the temporal credits and identity representations of different agents, encouraging the full expressiveness of credit assignment and further the emergence of individualities. The algorithm implementation of the proposed CIA module is simple yet effective that can be readily incorporated into various VD architectures. Experiments on the SMAC benchmarks and across different VD backbones demonstrate that the proposed method yields results superior to the state-of-the-art counterparts. Our code is available at https://github.com/liushunyu/CIA.
['Mingli Song', 'Zunlei Feng', 'Tongtian Zhu', 'KaiXuan Chen', 'Tongya Zheng', 'Jie Song', 'Yihe Zhou', 'Shunyu Liu']
2022-11-23
null
null
null
null
['smac-1', 'smac']
['playing-games', 'playing-games']
[-4.65744972e-01 1.43921033e-01 -6.92700505e-01 6.10507689e-02 -3.58948976e-01 -5.70481479e-01 8.90241802e-01 1.02123983e-01 -3.98546785e-01 9.75222707e-01 2.38933533e-01 -1.17513128e-02 -5.23821115e-01 -6.89474881e-01 -4.07050073e-01 -1.11810124e+00 -3.62459928e-01 6.93235397e-01 1.20364301e-01 -5.97957671e-01 1.49504706e-01 3.04588705e-01 -1.37527585e+00 -7.33828619e-02 1.15562201e+00 9.09050882e-01 2.29055062e-03 3.91031861e-01 3.41487974e-02 1.45262623e+00 -6.64578795e-01 -2.30255231e-01 4.73774999e-01 -5.42511761e-01 -6.02607310e-01 6.03430234e-02 -1.86048783e-02 -6.06285453e-01 -4.65068370e-01 9.56443906e-01 4.05637980e-01 2.06915975e-01 7.19368279e-01 -1.91570449e+00 -8.09944928e-01 1.16270173e+00 -8.79966319e-01 9.26752537e-02 -1.24594286e-01 2.24774957e-01 1.44140494e+00 -2.93307394e-01 5.59507310e-01 1.12694430e+00 9.74189490e-02 7.62974739e-01 -1.20499253e+00 -6.80728018e-01 6.52309597e-01 3.95706505e-01 -8.60196054e-01 -2.60041475e-01 8.91854942e-01 -3.18439126e-01 7.12225676e-01 -2.14656945e-02 6.01684213e-01 1.18694842e+00 -6.15840182e-02 1.13324976e+00 1.19872820e+00 -7.93188065e-02 5.42998135e-01 -6.47917464e-02 4.77703214e-02 7.44269013e-01 4.05112535e-01 3.84168416e-01 -6.71826005e-01 -4.46111977e-01 9.44825232e-01 3.31844166e-02 -6.71230629e-02 -8.12823892e-01 -1.19407618e+00 1.03588450e+00 5.06470501e-01 8.65890980e-02 -6.74250662e-01 2.28358120e-01 4.48445737e-01 5.99610806e-01 1.88844413e-01 4.71280247e-01 -1.91266671e-01 6.63704984e-03 -4.00139898e-01 4.73574638e-01 7.42029786e-01 6.99183345e-01 6.63896561e-01 3.28640729e-01 -2.83025235e-01 5.39491534e-01 3.17521960e-01 3.99921834e-01 3.04603785e-01 -1.62113786e+00 3.29951346e-01 6.55792117e-01 1.40855685e-01 -8.13889325e-01 -4.30178732e-01 -6.01237118e-01 -1.00867867e+00 6.83424115e-01 5.44290602e-01 -5.61040878e-01 -2.80416340e-01 2.11969543e+00 3.45383346e-01 2.70263612e-01 2.60567427e-01 9.00949597e-01 3.79105389e-01 4.85776424e-01 4.14681584e-02 -3.35789561e-01 8.70830894e-01 -1.19468355e+00 -3.73599917e-01 -3.34842466e-02 4.63052839e-01 -6.18270300e-02 4.33134854e-01 1.85921326e-01 -1.09964573e+00 -6.99792281e-02 -9.86569166e-01 5.36927223e-01 1.47073299e-01 -2.98460037e-01 8.53581965e-01 3.37859422e-01 -1.08283377e+00 5.53024411e-01 -7.22554624e-01 -8.23828764e-03 6.02369249e-01 5.27935386e-01 -1.69598032e-02 3.18751067e-01 -1.12784016e+00 6.85971200e-01 1.42668873e-01 -2.66036719e-01 -1.17892563e+00 -5.81395805e-01 -4.83746409e-01 1.51939735e-01 9.58523870e-01 -6.92129493e-01 1.23178542e+00 -1.23744619e+00 -1.72009611e+00 2.41789863e-01 4.41122502e-01 -5.96545696e-01 7.23985553e-01 2.28096202e-01 1.08281672e-02 2.11279199e-01 1.91275313e-01 6.57953680e-01 8.16078722e-01 -1.59931302e+00 -1.02305281e+00 -1.41795456e-01 5.95697045e-01 5.22053301e-01 -4.53528792e-01 -2.39811689e-01 2.91374147e-01 -5.40818870e-01 -5.71056724e-01 -9.63547528e-01 -4.75020379e-01 -1.79485634e-01 -5.82648218e-02 -6.52209103e-01 4.34505194e-01 -8.71783346e-02 9.46316123e-01 -1.98322153e+00 5.36480725e-01 2.65393615e-01 6.79835796e-01 1.16464935e-01 -4.97408658e-01 4.88829792e-01 4.59170401e-01 -1.67492107e-01 4.25356850e-02 -3.11469704e-01 3.86836320e-01 3.99064511e-01 -1.50197923e-01 4.96277124e-01 4.78821173e-02 7.21364796e-01 -1.08824897e+00 -3.62412006e-01 -1.07083842e-01 1.03002355e-01 -6.73258781e-01 2.76446521e-01 -3.65099043e-01 5.60039699e-01 -6.86766088e-01 6.30095363e-01 5.15605211e-01 -3.13274860e-01 7.70769656e-01 3.25322568e-01 -1.58698019e-02 5.24699949e-02 -1.29705215e+00 1.31146133e+00 -2.64337100e-03 1.76993415e-01 3.86827767e-01 -1.05700815e+00 8.39230418e-01 2.67207533e-01 9.54301238e-01 -8.26531112e-01 1.55726001e-01 3.17482948e-01 3.62502038e-01 -3.95054519e-02 4.01254624e-01 2.29720950e-01 -1.06609702e-01 8.15752208e-01 -2.45693102e-02 4.54803735e-01 5.26464820e-01 5.99135458e-01 9.72395539e-01 4.61048372e-02 3.07721078e-01 -4.13644016e-01 3.81621093e-01 -2.62212195e-02 9.68838453e-01 9.87953603e-01 -7.09713697e-01 -2.05456823e-01 9.15965080e-01 -2.94837892e-01 -1.08805430e+00 -1.11068547e+00 4.52421069e-01 1.43246639e+00 4.68694270e-01 -1.44630358e-01 -5.04410684e-01 -9.20528650e-01 3.66927832e-01 2.49016523e-01 -6.67283952e-01 3.34971584e-02 -7.70628691e-01 -7.09128678e-01 4.45763022e-01 4.95076984e-01 5.18327713e-01 -1.09230351e+00 -6.48346305e-01 3.06958377e-01 -1.53994471e-01 -7.32750475e-01 -6.45931184e-01 1.24529935e-01 -5.04097223e-01 -1.19761324e+00 -6.38423443e-01 -7.93969691e-01 4.81199831e-01 3.93188149e-01 1.00179255e+00 1.91578552e-01 1.70057297e-01 4.87776041e-01 -3.74304444e-01 -1.52066238e-02 -5.09941876e-01 3.10304284e-01 4.57079411e-01 5.89968860e-02 8.18942860e-02 -8.48511517e-01 -8.20209026e-01 3.22354317e-01 -6.08446121e-01 -9.20135677e-02 6.37492537e-01 1.03292823e+00 1.01670742e-01 -1.28475800e-01 1.29111099e+00 -7.18185782e-01 8.72673690e-01 -7.88566232e-01 -6.54185951e-01 3.13917369e-01 -8.70391786e-01 2.20247105e-01 8.89632821e-01 -6.50255263e-01 -1.10546517e+00 -3.23695779e-01 4.85791296e-01 -1.51506782e-01 1.89744994e-01 3.40069920e-01 1.08920969e-01 -2.48201936e-02 4.92043316e-01 1.07686616e-01 5.04853427e-01 -4.43963148e-02 5.31772137e-01 4.10576910e-01 2.46884391e-01 -1.05034232e+00 6.11334324e-01 4.17257816e-01 1.01014368e-01 -3.05334270e-01 -3.55736285e-01 -1.53042585e-01 -8.66321102e-02 -3.87955487e-01 3.14060688e-01 -1.00111973e+00 -1.35007668e+00 6.95560157e-01 -8.22671413e-01 -4.59673107e-01 -3.80166203e-01 3.45197350e-01 -6.03400826e-01 3.67097229e-01 -9.28426206e-01 -7.38494933e-01 -2.15978533e-01 -1.04880548e+00 1.71712503e-01 5.71911871e-01 1.41300336e-01 -9.84733582e-01 3.73895496e-01 3.04296136e-01 6.16654634e-01 1.31424114e-01 8.86728287e-01 -7.35186815e-01 -8.47235441e-01 5.83350420e-01 -6.42480031e-02 4.64197882e-02 1.42357543e-01 -1.61033139e-01 -7.42538571e-01 -6.42670691e-01 -3.59265089e-01 -7.29196727e-01 8.69303286e-01 3.63145411e-01 4.13490891e-01 -6.82871521e-01 -7.30517805e-02 1.23588257e-01 1.28268874e+00 2.11761028e-01 2.66631395e-01 6.40408337e-01 6.02931499e-01 7.09103882e-01 4.38775092e-01 1.07166326e+00 9.04327393e-01 6.20877922e-01 8.36069882e-01 1.53277829e-01 1.49462461e-01 -7.56153464e-02 7.06499696e-01 7.54579544e-01 -3.30243230e-01 -9.43679810e-02 -4.49276000e-01 4.71186429e-01 -2.36923242e+00 -1.31966817e+00 1.74018249e-01 1.90642917e+00 9.62249100e-01 -1.27658248e-01 7.26851046e-01 -3.32106471e-01 5.98691106e-01 4.14308697e-01 -9.96081829e-01 -3.23203474e-01 -3.25444102e-01 -4.09215659e-01 4.80692327e-01 4.57904994e-01 -9.15897310e-01 6.76860750e-01 5.65339184e+00 8.67112815e-01 -7.71047890e-01 9.64594334e-02 5.27270615e-01 -4.02684093e-01 -2.60135114e-01 -6.03620969e-02 -6.88917398e-01 3.82299215e-01 6.06780231e-01 -4.87482935e-01 8.54616940e-01 7.37876356e-01 2.45592892e-01 4.98648733e-02 -1.03005803e+00 6.06249154e-01 -1.41574204e-01 -1.41019630e+00 -1.22088611e-01 2.32053339e-01 1.12698817e+00 1.56103820e-01 1.93291113e-01 5.43631375e-01 1.07765436e+00 -6.81762040e-01 8.26816857e-01 1.92779794e-01 1.81452230e-01 -9.29850399e-01 4.31183517e-01 4.19045359e-01 -1.07120061e+00 -5.07900774e-01 -1.78838208e-01 -2.37786949e-01 -1.43044427e-01 6.38296753e-02 -4.68229651e-01 6.64385498e-01 4.56087023e-01 7.97569275e-01 -4.36312079e-01 6.59760296e-01 -1.92399755e-01 2.05248028e-01 1.34745780e-02 -1.86010450e-01 5.95367014e-01 -3.32475066e-01 6.11380100e-01 6.38195038e-01 -1.18558444e-01 -8.92171711e-02 6.86532855e-01 8.09305966e-01 -4.00427341e-01 -1.17405005e-01 -3.69062394e-01 2.15597674e-02 7.75390148e-01 1.30838966e+00 -4.54569936e-01 -2.19377294e-01 -3.65128785e-01 6.00784779e-01 8.26759279e-01 4.45973545e-01 -8.87619793e-01 3.72499153e-02 9.82334435e-01 -2.49358848e-01 3.19973648e-01 -1.77271858e-01 -8.31213966e-02 -1.15211749e+00 -1.08586855e-01 -1.28951633e+00 5.52797318e-01 -2.69925036e-02 -1.71175766e+00 4.90775883e-01 -1.98857963e-01 -1.16193056e+00 -5.55145681e-01 -3.04616183e-01 -6.20309472e-01 4.24111068e-01 -1.85322392e+00 -1.21166503e+00 -6.08482212e-02 7.82989740e-01 3.21618229e-01 -6.67805195e-01 5.79266548e-01 8.85754675e-02 -6.88291848e-01 6.31942391e-01 5.69352746e-01 4.63869330e-03 6.65831983e-01 -1.34395993e+00 -1.85186952e-01 5.81196785e-01 -1.14248373e-01 3.26016426e-01 4.15328056e-01 -4.62952793e-01 -1.26393902e+00 -8.49511802e-01 3.20546508e-01 -3.25584747e-02 9.15981710e-01 -9.52057615e-02 -7.39058614e-01 4.40741122e-01 6.36772037e-01 -2.13577941e-01 5.46659350e-01 6.99499100e-02 -5.42216361e-01 -2.66437352e-01 -9.63092089e-01 7.87698269e-01 9.10011470e-01 -1.89394787e-01 -1.20815277e-01 -4.54299618e-03 6.19764686e-01 3.03780790e-02 -8.55814517e-01 3.02216271e-04 3.85754466e-01 -1.16082835e+00 8.50334048e-01 -6.38677835e-01 4.85002607e-01 -3.60759228e-01 -1.15433112e-01 -1.51643574e+00 -5.75099885e-01 -7.95730293e-01 -3.71469975e-01 1.34435391e+00 2.89742976e-01 -9.41287398e-01 8.57903659e-01 4.61918473e-01 1.55027673e-01 -7.58694947e-01 -9.91378367e-01 -9.81436670e-01 4.35652673e-01 4.78445143e-01 6.65900826e-01 1.19742799e+00 4.41143334e-01 1.99736610e-01 -6.81747437e-01 8.83423910e-02 1.16433382e+00 2.83338517e-01 5.74261844e-01 -1.17431092e+00 -7.31015623e-01 -9.45682466e-01 -3.60070392e-02 -9.09498692e-01 4.78064597e-01 -9.20200109e-01 -2.65193492e-01 -1.39210796e+00 5.59232354e-01 -8.86741102e-01 -5.27484834e-01 4.98567343e-01 -5.61315902e-02 -1.71193793e-01 6.48970425e-01 6.24863088e-01 -1.05704772e+00 8.30677211e-01 1.37937641e+00 -3.60941768e-01 -3.46471906e-01 -3.77026163e-02 -1.00021803e+00 4.89140511e-01 1.13416648e+00 -5.33274829e-01 -6.08637393e-01 -3.33791912e-01 1.67910486e-01 1.66305125e-01 3.91951889e-01 -4.98393893e-01 6.05660319e-01 -6.75387502e-01 -1.20851435e-01 -2.17059366e-02 4.01786678e-02 -5.48763216e-01 -6.66435063e-02 7.31843114e-01 -6.42700315e-01 1.79554317e-02 -3.51777673e-01 6.87762320e-01 1.66877229e-02 -7.22592697e-02 7.71015823e-01 -1.71585858e-01 -6.31561100e-01 3.84844780e-01 -3.53106171e-01 3.67223531e-01 1.20625818e+00 3.19097102e-01 -8.93800259e-01 -5.04753172e-01 -1.99490830e-01 8.59764516e-01 4.60288167e-01 3.25984329e-01 3.85353774e-01 -1.41139138e+00 -1.03134024e+00 -1.11913763e-01 -1.38836339e-01 -3.11053634e-01 3.59865010e-01 8.92692864e-01 -5.68643324e-02 2.76395865e-02 -7.25647986e-01 -3.36703658e-01 -8.55154514e-01 4.97538686e-01 5.28005779e-01 -6.46469474e-01 -4.10116166e-01 3.63594711e-01 2.23526537e-01 -3.71217906e-01 3.64447504e-01 2.35726237e-01 -4.60298508e-01 2.17746422e-01 3.77174050e-01 7.42233515e-01 -6.20102465e-01 -4.04110700e-01 -2.41136551e-01 6.52351975e-02 -5.41053236e-01 -1.68736547e-01 1.45367920e+00 -2.20020041e-01 -6.90661892e-02 -1.09400228e-02 5.34097970e-01 -2.01667994e-01 -1.89241338e+00 -5.00259936e-01 -2.16129608e-02 -3.09906602e-01 -1.83512688e-01 -7.65730917e-01 -1.35674238e+00 2.99986571e-01 1.77329332e-01 4.48347986e-01 9.39987600e-01 3.30325626e-02 5.12562454e-01 3.16353321e-01 7.39895165e-01 -1.32181478e+00 4.78514850e-01 4.50696379e-01 5.69848239e-01 -1.30336678e+00 -3.44809331e-02 1.33529335e-01 -1.18732083e+00 8.26347947e-01 8.01719904e-01 -3.46830040e-01 3.14376265e-01 2.04270482e-01 -4.53041084e-02 3.47646959e-02 -1.13052034e+00 -2.13670462e-01 -3.78875643e-01 8.56016934e-01 -1.80461090e-02 3.13456982e-01 -2.65787393e-01 7.27846801e-01 4.36132699e-01 -2.69162327e-01 9.85551894e-01 9.55497622e-01 -4.74304855e-01 -1.42907357e+00 -8.65882337e-02 2.76499152e-01 -2.53220826e-01 1.48112610e-01 -3.21688503e-01 5.59737325e-01 -1.21946774e-01 1.16354930e+00 -1.28252208e-01 -1.41913071e-01 9.79626998e-02 -5.87727308e-01 4.38473701e-01 -6.22246303e-02 -9.11443055e-01 6.39436692e-02 -1.05870441e-02 -4.34614271e-01 -7.09883749e-01 -7.95178413e-01 -1.34449792e+00 -8.42502296e-01 3.17792483e-02 1.83547586e-01 6.46583587e-02 6.25975490e-01 5.42671382e-01 4.56297040e-01 1.12687063e+00 -6.09284103e-01 -1.17434752e+00 -5.21488011e-01 -8.48886907e-01 3.92003536e-01 4.34966028e-01 -9.67810929e-01 -4.15007055e-01 -4.03155267e-01]
[3.7517197132110596, 2.057478904724121]
b466df66-4e4d-4360-bde5-99a309f18a7c
image-as-a-foreign-language-beit-pretraining
2208.10442
null
https://arxiv.org/abs/2208.10442v2
https://arxiv.org/pdf/2208.10442v2.pdf
Image as a Foreign Language: BEiT Pretraining for All Vision and Vision-Language Tasks
A big convergence of language, vision, and multimodal pretraining is emerging. In this work, we introduce a general-purpose multimodal foundation model BEiT-3, which achieves state-of-the-art transfer performance on both vision and vision-language tasks. Specifically, we advance the big convergence from three aspects: backbone architecture, pretraining task, and model scaling up. We introduce Multiway Transformers for general-purpose modeling, where the modular architecture enables both deep fusion and modality-specific encoding. Based on the shared backbone, we perform masked "language" modeling on images (Imglish), texts (English), and image-text pairs ("parallel sentences") in a unified manner. Experimental results show that BEiT-3 obtains state-of-the-art performance on object detection (COCO), semantic segmentation (ADE20K), image classification (ImageNet), visual reasoning (NLVR2), visual question answering (VQAv2), image captioning (COCO), and cross-modal retrieval (Flickr30K, COCO).
['Furu Wei', 'Subhojit Som', 'Saksham Singhal', 'Owais Khan Mohammed', 'Kriti Aggarwal', 'Qiang Liu', 'Zhiliang Peng', 'Johan Bjorck', 'Li Dong', 'Hangbo Bao', 'Wenhui Wang']
2022-08-22
null
null
null
null
['visual-reasoning', 'zero-shot-cross-modal-retrieval', 'visual-reasoning']
['computer-vision', 'miscellaneous', 'reasoning']
[-6.93727583e-02 4.30830643e-02 -1.64008029e-02 -4.31578666e-01 -1.14818752e+00 -6.43954992e-01 7.98942983e-01 -4.19756807e-02 -6.42772257e-01 1.70912966e-01 1.96652450e-02 -4.53497142e-01 4.18145567e-01 -3.52500439e-01 -1.06979394e+00 -2.97024608e-01 3.37307304e-01 6.45031035e-01 1.17333077e-01 -3.52166504e-01 3.24721262e-02 1.18028626e-01 -1.40899575e+00 9.54079270e-01 7.10745692e-01 1.17975724e+00 3.20390135e-01 9.37190950e-01 -3.26331735e-01 1.10339034e+00 -2.73412019e-02 -8.66683424e-01 -3.06296647e-01 -4.92839664e-02 -1.35883188e+00 2.39508450e-01 9.30376828e-01 -4.27231818e-01 -4.64436442e-01 9.70480800e-01 4.63068277e-01 -1.76123112e-01 7.23675370e-01 -1.48384523e+00 -1.29730856e+00 3.97918820e-01 -6.38773918e-01 -2.13894993e-01 2.27874905e-01 5.71834087e-01 1.05496418e+00 -1.33976305e+00 6.28218651e-01 1.68552709e+00 4.70587343e-01 7.69008517e-01 -1.15084875e+00 -2.36814946e-01 2.56235093e-01 4.50220346e-01 -1.25486195e+00 -4.43581611e-01 3.76829326e-01 -5.13125002e-01 1.03476465e+00 2.55233437e-01 4.92414474e-01 1.32174802e+00 -8.49784464e-02 1.55667639e+00 1.02652419e+00 -4.09186840e-01 -1.78119801e-02 3.37910533e-01 2.58624524e-01 8.79560411e-01 -2.30853736e-01 -3.42575878e-01 -5.64730346e-01 3.34023178e-01 6.49074614e-01 -1.95082545e-01 -3.00627679e-01 -2.82680333e-01 -1.37485600e+00 6.70680344e-01 6.39710665e-01 1.53831020e-01 -2.31312126e-01 4.10680413e-01 5.72328210e-01 2.96102613e-01 1.27618149e-01 1.00988880e-01 -4.05778170e-01 2.44725615e-01 -7.79690921e-01 -2.00520679e-02 5.82633436e-01 1.11222959e+00 7.60462165e-01 -8.39176178e-02 -6.75752878e-01 1.03411257e+00 7.08180487e-01 1.02754772e+00 4.04542565e-01 -1.05510426e+00 5.71114123e-01 5.89196980e-01 -7.37700537e-02 -5.85153699e-01 -2.21886113e-01 -4.22960147e-02 -9.36594069e-01 -1.95525691e-01 1.54045537e-01 -3.47939925e-03 -1.46007311e+00 1.72430098e+00 -2.06089914e-02 -1.82182059e-01 3.56429845e-01 1.07058585e+00 1.61066651e+00 8.25198412e-01 5.73299050e-01 2.74867535e-01 1.85586977e+00 -1.55173028e+00 -4.90793407e-01 -5.43449879e-01 3.75947356e-01 -7.26462364e-01 1.43476081e+00 2.60355081e-02 -1.33572614e+00 -9.17400301e-01 -4.80241865e-01 -5.50338089e-01 -6.35110736e-01 2.38156989e-01 4.83871520e-01 2.31394544e-01 -1.61631250e+00 -4.18142229e-01 -3.92642885e-01 -8.29790711e-01 5.17897725e-01 -9.69107728e-03 -5.62876225e-01 -4.55700666e-01 -1.12294841e+00 8.87041152e-01 3.41310740e-01 1.59825772e-01 -1.49700928e+00 -7.05385923e-01 -1.01595664e+00 -6.17265403e-02 3.07173222e-01 -1.24594748e+00 1.34967494e+00 -1.03000593e+00 -1.14594793e+00 1.65358377e+00 -1.75719574e-01 -4.57194954e-01 4.46145922e-01 -1.86713681e-01 -2.94620126e-01 7.23300457e-01 3.73391085e-03 1.64954960e+00 9.22862768e-01 -1.57073295e+00 -4.33620244e-01 -3.96505862e-01 1.67092264e-01 4.26312357e-01 -2.89137185e-01 -1.67753175e-02 -1.21735823e+00 -2.44296059e-01 -3.01515073e-01 -7.54188657e-01 -5.64062968e-02 -2.22178381e-02 -6.00877523e-01 -4.18791711e-01 6.53868318e-01 -9.38415527e-01 6.31377578e-01 -2.25197411e+00 2.79849082e-01 -3.03375334e-01 2.45999038e-01 3.31835568e-01 -8.18635046e-01 4.70498860e-01 1.84947848e-01 9.50076710e-03 -2.13547528e-01 -6.85790658e-01 2.70445675e-01 2.47572586e-01 -3.96708041e-01 3.47255841e-02 3.87475342e-01 1.86506033e+00 -6.49758279e-01 -9.19479012e-01 3.44921976e-01 5.36248505e-01 -4.50272441e-01 3.79678816e-01 -6.85005784e-01 2.88977742e-01 -2.73000628e-01 1.02018762e+00 5.64902306e-01 -7.30431914e-01 -1.03832878e-01 -6.83303654e-01 1.27176195e-01 -4.75293130e-01 -4.87584889e-01 2.05970478e+00 -4.75213796e-01 7.57067919e-01 3.99533778e-01 -7.53946781e-01 4.68856543e-01 2.29146525e-01 6.66372851e-02 -1.05597782e+00 1.51314512e-01 -1.50794715e-01 -5.78849912e-01 -9.73809063e-01 5.74517787e-01 3.86149496e-01 5.48845949e-03 7.28397071e-02 5.37624598e-01 -1.72317967e-01 2.76086092e-01 5.16493320e-01 6.02366388e-01 2.87615284e-02 -2.29988188e-01 3.72807607e-02 6.22452855e-01 2.37258509e-01 -2.98267871e-01 9.58790183e-01 -1.68998748e-01 7.56573319e-01 4.05785769e-01 -6.64659888e-02 -8.97991419e-01 -1.24841213e+00 1.38059184e-01 1.48485386e+00 2.54784375e-01 -2.80138940e-01 -9.20167089e-01 -5.72244883e-01 8.91328081e-02 7.01588809e-01 -7.32620418e-01 -2.12697964e-02 -1.82238117e-01 -5.63305557e-01 7.81482220e-01 6.39878988e-01 8.19796205e-01 -1.34523308e+00 -3.49231839e-01 -3.15863550e-01 -5.33921897e-01 -1.70236933e+00 -4.31092501e-01 -2.14651436e-01 -5.63701451e-01 -9.31887448e-01 -1.00592828e+00 -1.11772025e+00 5.21322787e-01 5.23774922e-01 1.50531244e+00 -1.23118579e-01 -3.68538976e-01 1.38784385e+00 -1.94925442e-01 -1.31995678e-01 -3.14443260e-01 -2.05571294e-01 -5.11553943e-01 1.60613909e-01 1.01460099e-01 1.71882715e-02 -6.38165832e-01 5.94810769e-02 -1.19767416e+00 3.69839996e-01 9.10120308e-01 7.80357480e-01 7.16821730e-01 -7.18971729e-01 3.29980761e-01 -4.16594654e-01 4.87051785e-01 -3.47530305e-01 -3.42210412e-01 9.70249534e-01 -2.64905602e-01 -2.40671560e-01 3.30803469e-02 -4.34797019e-01 -1.07244873e+00 -9.93341282e-02 -1.55357957e-01 -8.80854666e-01 -4.74270076e-01 5.69753289e-01 -2.98006058e-01 -9.51363356e-04 3.31254900e-01 7.07951009e-01 6.42489716e-02 -4.43099946e-01 1.20003009e+00 6.65771067e-01 9.87013221e-01 -5.68925261e-01 3.82620692e-01 5.93677580e-01 -4.09409225e-01 -1.01592338e+00 -8.17819536e-01 -3.49799752e-01 -5.60276985e-01 -3.12847316e-01 1.60333538e+00 -1.39779913e+00 -1.01574707e+00 7.02788532e-01 -1.43175328e+00 -5.78850985e-01 -2.88858619e-02 8.08777288e-02 -4.12859321e-01 4.07363325e-01 -8.07265580e-01 -6.50692761e-01 -7.47019112e-01 -1.19742429e+00 1.44024360e+00 2.99465179e-01 3.53982627e-01 -9.23055828e-01 -2.18311638e-01 1.17036009e+00 3.64219159e-01 -1.55751988e-01 9.93678451e-01 -5.56094944e-01 -7.68187463e-01 9.62756947e-02 -1.05838263e+00 5.91516912e-01 -6.93394721e-01 -1.59920856e-01 -1.18968725e+00 -2.56842554e-01 -4.36587155e-01 -1.10385025e+00 1.20911252e+00 3.94948602e-01 1.13866282e+00 2.86571104e-02 -1.94623545e-01 5.86911678e-01 1.40319419e+00 -2.29584441e-01 7.33691096e-01 2.75686830e-02 1.12474680e+00 5.56142509e-01 3.09910476e-01 -1.11579083e-01 1.20117056e+00 3.82119358e-01 7.62436986e-01 -5.25131047e-01 -4.58452940e-01 -1.88877150e-01 5.16605616e-01 8.31644535e-01 2.78399318e-01 -2.78461844e-01 -1.17248130e+00 8.07349503e-01 -2.01214337e+00 -7.10757613e-01 -2.11134151e-01 1.53018904e+00 6.92980945e-01 -3.42297792e-01 -9.09697562e-02 -8.29244494e-01 5.67947567e-01 8.95384327e-02 -5.73262274e-01 -3.07649583e-01 -3.18652838e-01 -2.59378523e-01 1.68503478e-01 3.29579949e-01 -1.19726539e+00 1.43185711e+00 6.41368198e+00 9.35267150e-01 -1.10406232e+00 3.09193671e-01 9.22203183e-01 2.72670776e-01 -4.07368511e-01 -2.84681618e-01 -7.18322754e-01 1.11007638e-01 6.55018866e-01 5.24072826e-01 3.42184097e-01 6.63418174e-01 -3.19621772e-01 -9.51517448e-02 -1.04759896e+00 1.51575112e+00 5.41178167e-01 -1.51448190e+00 6.74243748e-01 -1.91406772e-01 6.23060942e-01 5.95146775e-01 3.54779571e-01 6.58486664e-01 1.09541759e-01 -1.16441250e+00 8.53716254e-01 7.42283821e-01 7.87132740e-01 -2.63198107e-01 5.78624189e-01 8.94224793e-02 -9.36630666e-01 4.45768945e-02 -9.89932716e-02 5.41422546e-01 2.52354354e-01 1.22213483e-01 -4.75754440e-01 7.25435555e-01 9.64853525e-01 6.45431638e-01 -1.04773128e+00 6.72366619e-01 -1.69013068e-01 4.69174355e-01 -1.67659670e-02 1.04601376e-01 5.64465582e-01 -7.59254862e-03 2.77837545e-01 1.58526409e+00 -2.85943538e-01 -1.66928440e-01 3.23920161e-01 9.37636137e-01 -4.20179456e-01 4.57974821e-02 -3.06375742e-01 -2.15658382e-01 2.06775993e-01 1.45465744e+00 -5.23922861e-01 -6.15552425e-01 -6.27567649e-01 1.25655973e+00 3.66275787e-01 8.05441916e-01 -9.34843659e-01 -3.11136246e-02 3.89977157e-01 -2.87063241e-01 5.29420376e-01 -6.05958961e-02 -4.27698344e-02 -1.34977877e+00 -1.11507818e-01 -9.37345266e-01 5.01701593e-01 -1.46844566e+00 -1.56397009e+00 6.36481047e-01 4.56941910e-02 -5.45827508e-01 4.44546230e-02 -8.73415112e-01 -2.73941815e-01 6.45142555e-01 -1.75862634e+00 -1.93477738e+00 -5.39789259e-01 9.86916780e-01 6.27405524e-01 -1.98895469e-01 5.70258379e-01 4.61310685e-01 -5.28517127e-01 6.16839945e-01 -2.91331410e-01 4.89827394e-01 7.21596837e-01 -1.09222138e+00 2.90282607e-01 5.05563080e-01 3.00907165e-01 3.03342581e-01 1.62464038e-01 -3.02206933e-01 -1.84141302e+00 -1.32144487e+00 6.96991086e-01 -6.94609940e-01 8.37267578e-01 -3.19259256e-01 -6.99301600e-01 6.96609855e-01 7.98543513e-01 3.89218479e-02 4.32258010e-01 -1.62150830e-01 -8.15186620e-01 9.55598429e-02 -9.01752412e-01 7.90358067e-01 7.55012095e-01 -1.01301360e+00 -6.03516400e-01 4.29065496e-01 1.21733356e+00 -2.10467026e-01 -7.30482757e-01 3.96197945e-01 4.90557432e-01 -6.70846641e-01 1.30695963e+00 -9.38208520e-01 6.37029946e-01 -2.87447780e-01 -4.76518869e-01 -7.44593322e-01 -4.90096072e-03 -2.81832039e-01 -4.51904088e-02 1.29288197e+00 3.44876081e-01 -1.64783344e-01 1.79593608e-01 2.50026345e-01 -3.07003677e-01 -6.13402188e-01 -6.72229648e-01 -3.44526857e-01 8.75838548e-02 -6.40958726e-01 2.12591022e-01 9.86792564e-01 -4.12079871e-01 8.58196318e-01 -2.96644241e-01 1.22790925e-01 6.44197404e-01 1.26370892e-01 8.98337126e-01 -6.99210346e-01 -2.89295197e-01 -5.33133566e-01 -1.35119423e-01 -1.25434923e+00 2.12176025e-01 -1.04676914e+00 1.22761514e-04 -2.10329294e+00 6.11421168e-01 3.08509439e-01 -2.61742413e-01 7.92071283e-01 -1.21828161e-01 5.62155187e-01 5.78943193e-01 1.35338679e-01 -1.40698850e+00 6.30719662e-01 1.52796888e+00 -4.90098357e-01 1.84897691e-01 -6.78047717e-01 -6.65574074e-01 6.22251570e-01 3.56807828e-01 3.46857011e-01 -3.67824227e-01 -1.10240090e+00 2.06538856e-01 1.10244758e-01 1.08485532e+00 -7.02412665e-01 2.26514086e-01 1.23282239e-01 3.84450108e-01 -7.99854875e-01 5.29876113e-01 -6.56205297e-01 -3.98812532e-01 2.92556942e-01 -4.65423971e-01 5.16843051e-02 4.40909803e-01 4.97562945e-01 -4.30150598e-01 9.97040421e-02 6.62642777e-01 -2.12528527e-01 -1.20544994e+00 2.67435700e-01 -1.71977773e-01 3.01098049e-01 7.08891630e-01 1.67659461e-01 -9.01772976e-01 -4.49839383e-01 -8.03118765e-01 8.25428724e-01 2.34974742e-01 6.15576744e-01 8.58959913e-01 -1.16767728e+00 -9.15653646e-01 -1.44461870e-01 6.46655023e-01 -1.53384164e-01 6.37872875e-01 1.18484366e+00 -4.63966906e-01 7.07542777e-01 -8.02027434e-02 -1.01910615e+00 -1.08583331e+00 7.28602588e-01 3.10823262e-01 -6.07739054e-02 -3.44018161e-01 1.09652817e+00 4.92820442e-01 -5.99440873e-01 4.71982360e-01 -2.68994987e-01 -1.81292474e-01 2.51354665e-01 4.26115870e-01 -1.16667069e-01 -1.80380300e-01 -7.24967957e-01 -4.13818032e-01 6.45827293e-01 -1.94537595e-01 -2.36415669e-01 8.43966424e-01 -3.77896994e-01 -5.07232785e-01 3.52721363e-01 1.22206247e+00 -6.91899776e-01 -1.02144063e+00 -3.26645970e-01 -2.82427907e-01 2.15975851e-01 -3.61832753e-02 -1.15986943e+00 -1.06563115e+00 1.31182110e+00 7.51191616e-01 3.95574793e-02 1.01228583e+00 5.45337021e-01 9.28810656e-01 5.35864472e-01 6.40011579e-02 -9.74956930e-01 4.21454042e-01 7.74289191e-01 9.79169548e-01 -1.54477561e+00 -5.28767228e-01 -1.00209668e-01 -1.10606539e+00 6.60561919e-01 7.70233274e-01 2.88518101e-01 4.49713320e-01 -3.56977522e-01 3.41624141e-01 -3.08017403e-01 -8.56917799e-01 -7.24492252e-01 7.34663725e-01 6.03952527e-01 2.00583190e-01 2.40964424e-02 3.00169051e-01 4.62681264e-01 4.11140174e-01 -2.64676750e-01 -6.06011264e-02 6.33422852e-01 -3.75182122e-01 -5.61175644e-01 -3.31963927e-01 -1.10899610e-02 -9.05379578e-02 -5.53409219e-01 -5.22170782e-01 9.08023357e-01 1.54556446e-02 8.92556310e-01 1.25114828e-01 -2.14123353e-01 2.73743451e-01 2.69262284e-01 6.33492768e-01 -2.78485090e-01 -6.21318579e-01 1.27168566e-01 1.17526408e-02 -7.09765077e-01 -4.41722423e-01 -1.52321532e-01 -1.14850938e+00 5.63686192e-02 1.83151484e-01 -7.38250986e-02 8.66494179e-01 1.03316569e+00 5.60787380e-01 5.44413507e-01 -3.60485911e-02 -7.10846305e-01 -2.25101873e-01 -8.51587415e-01 -1.04319453e-01 4.38221186e-01 3.13917279e-01 -1.31160274e-01 7.60842720e-03 4.22125220e-01]
[10.848350524902344, 1.6063101291656494]
a76e69da-8ab0-43a2-8340-0fdd8af1f777
eagle-large-scale-dataset-for-vehicle
2007.06124
null
https://arxiv.org/abs/2007.06124v3
https://arxiv.org/pdf/2007.06124v3.pdf
EAGLE: Large-scale Vehicle Detection Dataset in Real-World Scenarios using Aerial Imagery
Multi-class vehicle detection from airborne imagery with orientation estimation is an important task in the near and remote vision domains with applications in traffic monitoring and disaster management. In the last decade, we have witnessed significant progress in object detection in ground imagery, but it is still in its infancy in airborne imagery, mostly due to the scarcity of diverse and large-scale datasets. Despite being a useful tool for different applications, current airborne datasets only partially reflect the challenges of real-world scenarios. To address this issue, we introduce EAGLE (oriEnted vehicle detection using Aerial imaGery in real-worLd scEnarios), a large-scale dataset for multi-class vehicle detection with object orientation information in aerial imagery. It features high-resolution aerial images composed of different real-world situations with a wide variety of camera sensor, resolution, flight altitude, weather, illumination, haze, shadow, time, city, country, occlusion, and camera angle. The annotation was done by airborne imagery experts with small- and large-vehicle classes. EAGLE contains 215,986 instances annotated with oriented bounding boxes defined by four points and orientation, making it by far the largest dataset to date in this task. It also supports researches on the haze and shadow removal as well as super-resolution and in-painting applications. We define three tasks: detection by (1) horizontal bounding boxes, (2) rotated bounding boxes, and (3) oriented bounding boxes. We carried out several experiments to evaluate several state-of-the-art methods in object detection on our dataset to form a baseline. Experiments show that the EAGLE dataset accurately reflects real-world situations and correspondingly challenging applications.
['Reza Bahmanyar', 'Seyed Majid Azimi', 'Franz Kurz', 'Corenin Henry']
2020-07-12
null
null
null
null
['shadow-removal']
['computer-vision']
[ 3.07810664e-01 -6.71849251e-01 1.69885457e-01 -4.60500658e-01 -4.59902495e-01 -6.61824167e-01 4.53844875e-01 -1.72502056e-01 -4.80209947e-01 6.09745920e-01 -2.94509232e-01 -2.37097472e-01 -7.95984417e-02 -9.07224357e-01 -4.66263413e-01 -8.69625747e-01 -2.74248809e-01 5.83518386e-01 8.50090623e-01 -4.68876183e-01 1.01031281e-01 8.75251532e-01 -1.97378659e+00 4.17396575e-01 6.68159425e-01 8.75150681e-01 4.30725932e-01 6.82877898e-01 2.48551205e-01 4.94702309e-01 -9.21179831e-01 -1.21912807e-01 6.39707208e-01 1.39049813e-01 -2.88646281e-01 6.30624473e-01 1.05950117e+00 -6.18967593e-01 -2.45646894e-01 1.12035823e+00 3.98221284e-01 5.50346868e-03 5.51550210e-01 -1.59175479e+00 -3.98714751e-01 5.24551934e-03 -9.30115104e-01 7.23902106e-01 -8.01463947e-02 2.72518277e-01 4.44257855e-01 -1.08326173e+00 4.87556845e-01 1.24421048e+00 7.45872855e-01 2.01908752e-01 -8.36422384e-01 -9.30383801e-01 3.88519704e-01 3.34683150e-01 -1.93301284e+00 -2.06714645e-01 3.70853782e-01 -6.05217695e-01 7.45042801e-01 5.82129717e-01 7.27423787e-01 4.90076452e-01 -2.47352431e-03 5.81368864e-01 1.06433856e+00 -2.16667891e-01 -4.39663455e-02 4.50974584e-01 1.32516205e-01 6.13888144e-01 5.28570116e-01 3.04479003e-01 5.11384755e-02 -1.21430121e-01 5.20222366e-01 4.45708722e-01 -3.63448262e-01 -2.68364429e-01 -1.18430030e+00 8.52383137e-01 6.36600971e-01 -1.88693516e-02 -1.52281314e-01 -1.53812945e-01 5.98196052e-02 1.59920663e-01 5.31389058e-01 5.19612245e-02 -4.38055038e-01 8.18470299e-01 -1.01301467e+00 6.05071008e-01 1.71680063e-01 1.19571388e+00 9.37135994e-01 3.44587058e-01 9.70578287e-04 7.96721220e-01 1.34480476e-01 1.01766443e+00 -2.67502759e-02 -5.76728165e-01 5.89109004e-01 6.41521931e-01 4.73596841e-01 -1.11750758e+00 -5.53431809e-01 -2.66746432e-01 -8.03935885e-01 6.50169969e-01 3.36025298e-01 1.50859533e-02 -1.11016405e+00 1.05609274e+00 6.68260455e-01 1.78460762e-01 -2.91471686e-02 1.23718894e+00 1.08192933e+00 9.37935174e-01 -1.08182117e-01 -1.13247186e-01 1.73622966e+00 -8.20582092e-01 -6.56162977e-01 -6.65638268e-01 4.91488278e-01 -9.05842662e-01 6.00701213e-01 5.74838042e-01 -4.34252053e-01 -7.31661737e-01 -1.02967954e+00 4.04625744e-01 -8.90436351e-01 4.89825398e-01 4.96188164e-01 7.75455534e-01 -7.68858671e-01 -1.14002414e-02 -2.07505405e-01 -4.78067160e-01 4.60544467e-01 6.31315112e-02 -4.03230458e-01 -3.94685179e-01 -1.13030565e+00 8.59698772e-01 5.32629132e-01 4.04808700e-01 -1.22631252e+00 -5.33283174e-01 -6.41920030e-01 -4.08784419e-01 6.65262997e-01 -4.19796914e-01 6.09084785e-01 -8.79550457e-01 -5.13551116e-01 9.59560037e-01 2.41710097e-01 -5.40988386e-01 5.01980305e-01 -3.29178572e-01 -9.71308291e-01 -1.05613008e-01 2.84553796e-01 6.73205435e-01 1.19163060e+00 -1.30744648e+00 -1.21481371e+00 -6.40285313e-01 1.65216282e-01 1.72412694e-01 -1.86481059e-01 3.65977108e-01 -2.43641540e-01 -6.91564500e-01 -1.26201540e-01 -9.63639855e-01 -1.76740602e-01 1.56724215e-01 -4.29779977e-01 -4.34882976e-02 1.33797729e+00 -4.24439073e-01 1.18919683e+00 -2.05010009e+00 -2.30503350e-01 -1.05730906e-01 5.04791504e-03 5.22255719e-01 -7.96079710e-02 3.27480316e-01 -5.25664585e-03 1.36712372e-01 -3.32625180e-01 2.29801744e-01 -4.31871206e-01 1.21379010e-01 -6.65463448e-01 7.35336602e-01 2.17896387e-01 3.53215277e-01 -6.45152330e-01 -7.66945004e-01 6.90409422e-01 4.95052844e-01 -1.37617230e-01 1.80606484e-01 1.07340820e-01 9.11141634e-02 -3.06998461e-01 1.32976294e+00 1.17894506e+00 4.64538008e-01 -3.48613441e-01 -3.28867972e-01 -5.37687778e-01 -6.25367880e-01 -1.78638208e+00 8.16633582e-01 -1.59235656e-01 9.83021557e-01 1.89152330e-01 -6.15505278e-01 8.84509206e-01 1.02865204e-01 1.71174407e-02 -3.10046434e-01 -3.42922769e-02 5.25020137e-02 -2.51738071e-01 -6.77970707e-01 9.16053772e-01 3.34718451e-02 9.20856744e-02 -2.06767514e-01 -3.34927469e-01 -5.55424154e-01 4.15336162e-01 1.14491500e-01 6.58435881e-01 -3.51931721e-01 5.45457244e-01 -1.76243916e-01 3.96399170e-01 5.02656639e-01 4.88518655e-01 8.31013560e-01 -3.85155529e-01 7.08339989e-01 -9.85943973e-02 -1.17057657e+00 -1.01415646e+00 -8.57420981e-01 -5.49534798e-01 9.42276061e-01 3.98407429e-01 -1.06542617e-01 -5.58007360e-01 -3.77107501e-01 9.93472934e-02 5.61204255e-01 -6.64438307e-01 2.50282794e-01 -5.13545275e-01 -1.29729974e+00 5.24445474e-01 3.62081707e-01 7.64163613e-01 -8.61556768e-01 -9.03663933e-01 1.64888129e-02 -2.85277307e-01 -1.62200248e+00 -2.40738466e-01 -9.48587880e-02 -7.72675097e-01 -1.37108862e+00 -3.96498293e-01 -5.21389067e-01 5.56039870e-01 1.25266087e+00 1.14101994e+00 3.07498962e-01 -1.03436947e+00 4.73532230e-02 -3.43599409e-01 -9.12590623e-01 1.28960162e-01 -4.63706195e-01 6.94145933e-02 1.82353437e-01 3.11339021e-01 9.50803980e-02 -6.45572722e-01 9.16376293e-01 -1.08870053e+00 -2.31856242e-01 5.82858443e-01 5.82418501e-01 5.45766175e-01 6.82523966e-01 -6.06120229e-02 -7.46369958e-01 -1.45114422e-01 -4.55042720e-01 -1.18099010e+00 3.29541191e-02 -1.34057999e-01 -8.68472159e-01 2.34062821e-01 -4.15720373e-01 -9.61699247e-01 2.88779944e-01 3.06550413e-01 -2.62221247e-01 -8.40573430e-01 -8.24558958e-02 -8.69360641e-02 -3.12722266e-01 8.53621840e-01 1.18226871e-01 -6.46617115e-01 -1.35001317e-01 1.70543447e-01 1.03333735e+00 5.52508652e-01 -7.26118386e-02 1.18343294e+00 1.09399474e+00 -1.03603052e-02 -1.50740790e+00 -1.10553253e+00 -9.91354048e-01 -9.70856190e-01 -3.91493589e-01 9.84430492e-01 -1.38183248e+00 -2.05869094e-01 3.77782583e-01 -1.18337846e+00 -5.55385416e-03 5.04778214e-02 3.08199286e-01 6.58343197e-04 2.82032222e-01 1.25232218e-02 -1.29521918e+00 -2.63690501e-01 -9.71850455e-01 1.45157313e+00 -1.70352741e-03 4.41510856e-01 -5.01859665e-01 -3.01182300e-01 4.82020825e-01 1.84812471e-01 5.25856972e-01 5.08937657e-01 -7.08926916e-02 -1.02611876e+00 -3.93082529e-01 -5.13749599e-01 3.87189507e-01 -1.10057749e-01 2.75863320e-01 -9.48740542e-01 -3.49344492e-01 -3.21110696e-01 -1.22495731e-02 1.15782058e+00 3.65981817e-01 1.03389955e+00 5.51159345e-02 -5.76355278e-01 7.37379372e-01 1.60042369e+00 2.22207874e-01 6.82212710e-01 5.59202135e-01 6.69084370e-01 7.50886321e-01 1.38161314e+00 5.20967603e-01 3.07252482e-02 9.95603621e-01 1.12345278e+00 -4.06701863e-01 -1.03142314e-01 3.38936061e-01 2.55056918e-01 -2.62033373e-01 -4.19086158e-01 -4.48483735e-01 -1.02019012e+00 6.81251824e-01 -1.59304190e+00 -1.30897927e+00 -6.98038816e-01 2.15853286e+00 5.22697493e-02 7.20730498e-02 2.28719205e-01 1.18926600e-01 8.77404094e-01 3.04102123e-01 -2.02616483e-01 2.03341573e-01 -5.42541862e-01 -4.00897890e-01 9.35544550e-01 2.52066374e-01 -1.87310123e+00 1.07793641e+00 6.07444525e+00 8.44210982e-01 -9.03984129e-01 1.69734508e-01 1.89458176e-01 2.32673436e-01 4.01955426e-01 -1.92294121e-01 -1.46543324e+00 2.49611378e-01 5.27976274e-01 5.52007020e-01 2.18157262e-01 9.98989522e-01 1.16975814e-01 -1.50780171e-01 -4.10238951e-01 9.20896947e-01 4.03364927e-01 -1.05559158e+00 1.41612723e-01 2.28024670e-03 7.49032497e-01 2.77217686e-01 4.60872194e-03 1.50963262e-01 4.02899757e-02 -7.67283201e-01 9.72606838e-01 2.58823894e-02 8.81309867e-01 -5.96017659e-01 1.02225435e+00 4.90862787e-01 -1.65330923e+00 -5.79743803e-01 -8.97194088e-01 -6.37966907e-03 1.79615945e-01 4.63374108e-01 -9.60197330e-01 3.47673625e-01 1.16053891e+00 5.16075969e-01 -7.86856771e-01 1.20730042e+00 -5.12761101e-02 4.13028121e-01 -3.90733331e-01 3.08889806e-01 3.37821513e-01 -3.11918288e-01 6.48190677e-01 1.39906204e+00 4.04480636e-01 4.24486697e-01 5.69816470e-01 5.43101311e-01 3.87898415e-01 -1.44688845e-01 -1.00465345e+00 2.41824836e-01 5.11117816e-01 1.70748079e+00 -9.22134161e-01 -4.02631044e-01 -4.98190612e-01 6.01817310e-01 -1.19183153e-01 4.45081741e-01 -1.18390203e+00 -3.85017663e-01 9.17689383e-01 5.72363257e-01 4.40618128e-01 -5.15350223e-01 2.79538423e-01 -8.71061862e-01 -2.34240443e-01 -6.50951028e-01 5.01899302e-01 -9.66739357e-01 -8.68048072e-01 9.33827758e-01 3.55796456e-01 -1.52567649e+00 2.95179486e-01 -1.06426120e+00 -4.45178807e-01 5.13459802e-01 -1.90127242e+00 -1.40752029e+00 -1.05321062e+00 5.57542205e-01 9.87260044e-01 -2.78012544e-01 6.01419985e-01 5.45405805e-01 -7.08094656e-01 -9.33335721e-02 8.35689381e-02 1.43582746e-01 7.45508134e-01 -9.91152108e-01 3.25974286e-01 9.78050053e-01 2.37538025e-01 2.11103708e-01 7.63239920e-01 -5.97202718e-01 -1.19508159e+00 -1.66087413e+00 3.45565885e-01 -8.17311466e-01 5.46047032e-01 -3.72054398e-01 -7.26890266e-01 8.66456270e-01 -1.26114249e-01 4.67372417e-01 2.98846066e-01 -2.85165876e-01 -2.27635130e-01 -5.25624335e-01 -9.87630069e-01 2.40366891e-01 8.66292417e-01 -8.12097862e-02 -3.87334228e-01 8.45963418e-01 4.44202155e-01 -5.07845938e-01 -4.62657362e-01 6.75179541e-01 3.82059783e-01 -1.03965676e+00 1.52126479e+00 -4.31654930e-01 -1.62397385e-01 -9.04361904e-01 -4.66771603e-01 -1.02147317e+00 -4.25425917e-01 -5.03135994e-02 3.40657711e-01 1.19101524e+00 7.38573447e-02 -4.53427136e-01 5.66669285e-01 5.49747795e-02 -4.31033283e-01 -3.51923764e-01 -7.56781518e-01 -8.77567172e-01 -3.92970324e-01 -4.03628588e-01 8.04033458e-01 6.44538283e-01 -1.06820369e+00 -2.87806690e-02 -7.43842065e-01 9.15468335e-01 8.35385025e-01 4.82377350e-01 1.22495735e+00 -1.63175511e+00 3.15938652e-01 -5.06105386e-02 -6.52610004e-01 -5.91086566e-01 -1.72173083e-01 -2.98477113e-01 8.05519223e-02 -1.38912868e+00 2.52469122e-01 -4.75079119e-01 1.36874840e-01 3.41105491e-01 -1.68521851e-01 8.86701882e-01 1.99152365e-01 3.63783121e-01 -6.12046599e-01 1.98537663e-01 7.83876717e-01 -4.66688991e-01 1.80180281e-01 1.83429644e-01 -2.47838229e-01 1.05117333e+00 6.37638628e-01 -6.50511146e-01 -1.37634069e-01 -5.32233894e-01 -6.87081516e-02 -2.89157629e-01 7.48563647e-01 -1.16020083e+00 5.39907366e-02 -5.44645369e-01 6.48974419e-01 -1.21690118e+00 7.13562906e-01 -1.20599782e+00 3.00530702e-01 5.50029218e-01 3.33015263e-01 2.10569128e-01 2.08661705e-01 7.24751055e-01 -3.09826761e-01 -1.72784194e-01 9.24700618e-01 -3.38129431e-01 -1.43802142e+00 4.48221087e-01 -3.88826311e-01 -4.11982127e-02 1.38107896e+00 -5.92963994e-01 -6.54264271e-01 2.61737057e-03 -3.28682661e-01 2.29651824e-01 3.79885316e-01 5.20973861e-01 8.79101157e-01 -1.01332176e+00 -1.02123547e+00 3.77006352e-01 5.80299139e-01 1.39938667e-02 4.82853234e-01 6.81605339e-01 -8.36658657e-01 5.26013970e-01 -2.79636949e-01 -8.73513758e-01 -1.75826168e+00 7.76976407e-01 2.81488746e-01 2.42317066e-01 -4.55352038e-01 7.47512758e-01 5.37637651e-01 -3.59623015e-01 9.90530550e-02 -2.07598597e-01 -4.04406309e-01 2.23898023e-01 1.03271949e+00 5.86819351e-01 1.63624421e-01 -1.04328954e+00 -4.48635876e-01 1.00226676e+00 2.07421035e-01 4.49946135e-01 9.95419681e-01 -2.33206972e-01 3.98102850e-02 2.26980299e-01 4.47529435e-01 -1.67197794e-01 -1.01697016e+00 -1.28595740e-01 -4.11847383e-01 -9.63522851e-01 2.27702111e-01 -5.20601094e-01 -1.01488769e+00 1.00392592e+00 1.06259882e+00 3.03463012e-01 1.04561222e+00 -2.41790518e-01 5.05613744e-01 5.38103938e-01 5.13767004e-01 -9.93794799e-01 -1.76044554e-02 3.24762940e-01 1.01134300e+00 -1.74055803e+00 5.16662657e-01 -8.66977870e-01 -7.30273724e-01 1.18708003e+00 8.17659199e-01 1.71353072e-01 5.93495965e-01 1.71665639e-01 3.53027403e-01 -3.53342980e-01 -3.54526579e-01 -6.20326459e-01 2.94305116e-01 8.47894967e-01 7.40461126e-02 3.13790947e-01 2.16419280e-01 6.42362908e-02 -4.25517038e-02 -4.98796433e-01 5.72738051e-01 7.47327566e-01 -1.03914475e+00 -4.62004513e-01 -1.12399971e+00 2.87787884e-01 -1.64899111e-01 -6.34031557e-03 -3.42955858e-01 1.43826950e+00 7.03697383e-01 1.20378566e+00 7.94923306e-02 -2.43157834e-01 5.73479593e-01 -5.72871983e-01 2.41516501e-01 -6.79425478e-01 -3.24030668e-01 -1.21991843e-01 7.59870633e-02 -4.20333445e-01 -3.95976931e-01 -6.99293554e-01 -8.61469090e-01 -1.84623122e-01 -8.24651003e-01 -1.02000140e-01 7.59145081e-01 6.04647934e-01 -3.32174934e-02 4.87822384e-01 5.14238715e-01 -1.24229407e+00 -2.42463544e-01 -9.50078607e-01 -1.00283897e+00 2.93257028e-01 5.49261391e-01 -1.06889296e+00 -4.88935679e-01 4.72381525e-02]
[8.710943222045898, -0.7993441224098206]
465c2d9d-eaee-4c11-91bd-7ecd1651e439
overlap-guided-gaussian-mixture-models-for
2210.09836
null
https://arxiv.org/abs/2210.09836v1
https://arxiv.org/pdf/2210.09836v1.pdf
Overlap-guided Gaussian Mixture Models for Point Cloud Registration
Probabilistic 3D point cloud registration methods have shown competitive performance in overcoming noise, outliers, and density variations. However, registering point cloud pairs in the case of partial overlap is still a challenge. This paper proposes a novel overlap-guided probabilistic registration approach that computes the optimal transformation from matched Gaussian Mixture Model (GMM) parameters. We reformulate the registration problem as the problem of aligning two Gaussian mixtures such that a statistical discrepancy measure between the two corresponding mixtures is minimized. We introduce a Transformer-based detection module to detect overlapping regions, and represent the input point clouds using GMMs by guiding their alignment through overlap scores computed by this detection module. Experiments show that our method achieves superior registration accuracy and efficiency than state-of-the-art methods when handling point clouds with partial overlap and different densities on synthetic and real-world datasets. https://github.com/gfmei/ogmm
['Nicu Sebe', 'Elisa Ricci', 'Jian Zhang', 'Cristiano Saltori', 'Fabio Poiesi', 'Guofeng Mei']
2022-10-17
null
null
null
null
['point-cloud-registration']
['computer-vision']
[-3.17283481e-01 -3.72655332e-01 1.33489162e-01 -2.22826988e-01 -1.23112714e+00 -4.98007357e-01 6.84458375e-01 1.14145234e-01 -2.40033790e-01 9.20556635e-02 -3.59258831e-01 3.44147384e-02 -6.40489310e-02 -7.53555715e-01 -5.79339325e-01 -7.22945273e-01 1.12191945e-01 1.17914474e+00 5.12844205e-01 2.48119459e-01 4.53622282e-01 9.92543817e-01 -1.61133671e+00 -1.82286993e-01 1.25993776e+00 6.76624358e-01 4.31824028e-01 2.93559909e-01 -3.20900857e-01 -3.82689029e-01 -4.93317008e-01 -8.45348611e-02 5.93896508e-01 2.71553516e-01 -4.78644133e-01 1.39638916e-01 5.76660335e-01 1.12690717e-01 -5.04053980e-02 1.30635703e+00 3.95048738e-01 3.71541530e-01 9.69667733e-01 -1.53091908e+00 -2.22831443e-01 -1.29484400e-01 -9.11557257e-01 -7.63626173e-02 3.14207882e-01 -9.69682410e-02 3.55280250e-01 -1.21986079e+00 3.36023599e-01 1.24978840e+00 7.29366302e-01 8.93779099e-02 -1.39279735e+00 -8.90711844e-01 -4.21023443e-02 -2.33470976e-01 -2.00447130e+00 -1.13805994e-01 5.76680362e-01 -8.43555331e-01 8.02384496e-01 1.91462353e-01 3.81231278e-01 3.47922504e-01 9.22905058e-02 1.83123291e-01 8.44748199e-01 -1.56482399e-01 1.54668719e-01 5.99621935e-03 -1.27928719e-01 3.21943671e-01 3.75417590e-01 -5.20614199e-02 -1.74667075e-01 -9.03418481e-01 9.39470589e-01 2.76351511e-01 -2.27178290e-01 -7.99535573e-01 -1.24898779e+00 6.83714628e-01 1.92993373e-01 2.99638450e-01 -4.97410715e-01 -1.76399752e-01 -2.80432820e-01 -2.04696208e-01 7.86707222e-01 5.60985208e-02 -1.95145562e-01 1.12017216e-02 -1.18952560e+00 4.24468547e-01 5.40346742e-01 1.40237248e+00 1.08124363e+00 -2.77402312e-01 2.06432983e-01 9.65591609e-01 7.65405297e-01 9.62508440e-01 7.33219758e-02 -9.82212067e-01 3.51907462e-01 6.28010929e-01 1.83879197e-01 -1.06666315e+00 -8.51830542e-02 -8.09496269e-02 -6.91449881e-01 3.64675492e-01 1.89689189e-01 3.29771161e-01 -9.22634780e-01 1.30866218e+00 7.93498755e-01 8.65371644e-01 -1.35726139e-01 7.95727193e-01 6.78044558e-01 6.27167284e-01 -1.71308249e-01 1.04516717e-02 1.22010922e+00 -2.38767907e-01 -2.41531372e-01 1.07124280e-02 3.00689399e-01 -1.37589002e+00 5.61610460e-01 1.37434313e-02 -1.03860795e+00 -4.36414033e-01 -8.62739384e-01 2.37058401e-01 -5.19864522e-02 5.79671487e-02 9.01280269e-02 4.28466707e-01 -9.73072946e-01 6.40352666e-01 -1.17440271e+00 -4.00183231e-01 3.09519559e-01 6.14488959e-01 -4.84955698e-01 -1.97203886e-02 -4.07181889e-01 7.65320182e-01 1.52060539e-01 -1.19846202e-01 -3.74611676e-01 -8.69791448e-01 -8.53500605e-01 -2.47211337e-01 -5.67279160e-02 -8.27433348e-01 9.84580517e-01 -9.90347564e-02 -1.24933720e+00 1.09908116e+00 -5.72452128e-01 -1.99186523e-02 4.09583688e-01 -2.03152522e-01 -7.27604330e-02 -2.02816129e-01 3.48278642e-01 5.43476403e-01 5.86601734e-01 -1.76251721e+00 -6.05099082e-01 -5.62569618e-01 -6.94879889e-01 2.72951037e-01 2.85454154e-01 2.39420757e-01 -9.10668969e-01 -3.45356703e-01 8.99238229e-01 -1.20098186e+00 -4.68593001e-01 -1.92464530e-01 -3.43738765e-01 -2.38870844e-01 9.99621272e-01 -3.65882963e-01 6.82941616e-01 -2.27994180e+00 -1.58708692e-01 7.25801229e-01 1.19766459e-01 -4.61594202e-02 5.01811244e-02 1.67579159e-01 1.08213268e-01 -1.03396073e-01 -4.59840357e-01 -8.06010127e-01 9.11159292e-02 1.42722026e-01 -2.13158607e-01 9.33368325e-01 2.91896611e-02 3.01984817e-01 -8.63990724e-01 -5.85705638e-01 6.55319929e-01 7.93892682e-01 -3.35333616e-01 3.39997768e-01 8.83268416e-02 7.18643785e-01 -2.57816702e-01 7.41121888e-01 1.45794022e+00 -1.14065632e-02 -1.18959993e-01 -2.25026727e-01 -9.13746580e-02 6.01396989e-03 -1.72202098e+00 1.54958761e+00 -1.08695202e-01 2.64976948e-01 1.78212106e-01 -4.57473099e-01 1.40937960e+00 2.22258240e-01 1.05425286e+00 9.03074294e-02 1.70514151e-01 4.03465241e-01 -2.77418882e-01 1.37980655e-01 7.37159014e-01 -1.09437183e-01 4.96303551e-02 1.50675282e-01 -3.43438983e-02 -7.82190502e-01 -2.86607414e-01 1.64933391e-02 7.26178706e-01 1.65694222e-01 1.53297588e-01 -2.15464696e-01 4.31265265e-01 2.96463519e-02 5.65032959e-01 4.39409375e-01 -2.60525905e-02 1.22125590e+00 -2.55146831e-01 5.69375828e-02 -9.65031385e-01 -1.48368180e+00 -3.30916822e-01 2.20393583e-01 5.72453082e-01 -4.10650313e-01 -6.22612357e-01 -2.46885553e-01 3.14826280e-01 5.59450746e-01 -1.56682968e-01 1.35665298e-01 -5.44889688e-01 -8.43070030e-01 1.00321442e-01 2.89981872e-01 1.47270858e-01 -3.45653474e-01 -1.54781029e-01 2.69973632e-02 -1.49584308e-01 -1.46768618e+00 -5.87535739e-01 -3.05043519e-01 -1.28799093e+00 -1.06888247e+00 -6.26414835e-01 -7.00015485e-01 1.01398826e+00 7.15198100e-01 1.18071699e+00 -2.46123709e-02 -2.42400244e-02 4.29763585e-01 -1.72029883e-02 -3.95169139e-01 -4.74986672e-01 -8.23994130e-02 3.97846043e-01 -2.17121020e-01 6.49041653e-01 -8.21387649e-01 -4.02857721e-01 8.36108208e-01 -6.74050272e-01 -2.73373365e-01 1.99511096e-01 2.96902895e-01 1.11441910e+00 -1.02144599e-01 -2.54793853e-01 -3.95215750e-01 4.31624800e-01 -5.71123838e-01 -1.06867361e+00 -6.87117577e-02 -3.63514841e-01 -9.86230895e-02 -7.94289038e-02 -4.42815095e-01 -7.40037262e-01 4.36555207e-01 2.82656122e-02 -1.00291300e+00 -4.99364406e-01 4.85773245e-03 -3.72643381e-01 -3.64678621e-01 4.52619314e-01 3.76938470e-02 1.05403975e-01 -6.26913011e-01 3.29194158e-01 7.65269995e-01 7.89248586e-01 -9.03504431e-01 1.28128052e+00 8.02272141e-01 3.59784476e-02 -6.90481782e-01 -1.35871723e-01 -1.07921386e+00 -1.02407539e+00 -2.04839483e-01 6.41094387e-01 -9.49096262e-01 -5.53762078e-01 3.72460246e-01 -1.40576804e+00 1.34064317e-01 -2.05203831e-01 7.49194145e-01 -5.23550510e-01 5.73712885e-01 -2.36407518e-01 -8.37629676e-01 -2.66563296e-01 -1.52125859e+00 1.50843048e+00 2.57484078e-01 -1.27207845e-01 -7.44810700e-01 5.45768261e-01 1.93376824e-01 1.02249712e-01 3.47148538e-01 3.58266503e-01 -6.25864863e-01 -8.75453472e-01 -5.35242379e-01 -1.01266354e-01 6.82270993e-03 3.23890120e-01 4.27403301e-01 -9.67086375e-01 -2.97009259e-01 1.01419538e-01 6.91202879e-01 3.16538930e-01 4.62569028e-01 7.81878114e-01 2.33124688e-01 -7.30170250e-01 8.30997467e-01 1.37993622e+00 9.63233933e-02 6.39645576e-01 2.22131386e-01 8.04026842e-01 4.19371068e-01 8.44904363e-01 3.56696308e-01 5.20439506e-01 9.18434441e-01 5.05724013e-01 2.45278105e-02 2.02769607e-01 1.15965577e-02 7.68937841e-02 9.23647344e-01 -2.18979955e-01 2.35231116e-01 -1.27906597e+00 4.98427451e-01 -1.97018576e+00 -7.30251968e-01 -4.97019410e-01 2.71428132e+00 3.95326883e-01 -1.65128291e-01 5.32304496e-02 -3.38912427e-01 1.12967348e+00 -2.00484589e-01 -7.06230700e-02 2.39198923e-01 -2.26850379e-02 1.13564827e-01 6.28631234e-01 6.52095914e-01 -1.24607325e+00 9.84405696e-01 5.55258799e+00 8.78964782e-01 -8.86869609e-01 3.37098539e-01 1.78456694e-01 1.10042624e-01 -1.10668153e-01 1.44642964e-01 -8.85186255e-01 5.87624073e-01 6.49902523e-01 -1.23899534e-01 -6.94767237e-02 9.51777697e-01 1.46383911e-01 -2.52145439e-01 -8.41055214e-01 1.32250619e+00 1.10933101e-02 -1.15887320e+00 -1.79545015e-01 4.52739269e-01 8.82204950e-01 5.60352206e-01 3.14625874e-02 -1.74752295e-01 4.98472929e-01 -7.32699156e-01 5.11750460e-01 6.56747580e-01 5.40043712e-01 -6.33595526e-01 8.51733565e-01 3.93224359e-01 -1.47922623e+00 7.10976243e-01 -6.82432353e-01 5.09754658e-01 3.22201580e-01 7.92962492e-01 -9.97992575e-01 7.95290053e-01 6.91662669e-01 3.77776444e-01 -2.68958747e-01 1.64213669e+00 1.10637039e-01 9.44799185e-02 -8.83373499e-01 6.27561152e-01 -3.14461380e-01 -8.28902841e-01 9.95884717e-01 1.13118935e+00 9.48051155e-01 5.34272157e-02 3.95149112e-01 1.09130979e+00 1.13323957e-01 1.59420684e-01 -6.77344084e-01 6.87350512e-01 9.89591777e-01 1.27552569e+00 -9.77052927e-01 -2.46269718e-01 -2.31344774e-01 7.82340527e-01 8.31340849e-02 1.08893923e-01 -7.65905321e-01 2.38357872e-01 1.10873544e+00 2.90630698e-01 4.67813537e-02 -6.59366369e-01 -4.33858514e-01 -1.03936732e+00 1.31218165e-01 -4.95795906e-01 2.03419000e-01 -7.81883299e-01 -1.33103263e+00 7.10722744e-01 3.62192422e-01 -1.81376326e+00 5.63113429e-02 -2.83494025e-01 -7.86780059e-01 1.12042022e+00 -1.09728360e+00 -1.04395974e+00 -6.42145574e-01 4.67244089e-01 1.38483077e-01 7.30478540e-02 7.10805833e-01 4.70645070e-01 -2.93616384e-01 2.48127833e-01 2.16964141e-01 -1.27771199e-01 7.38919377e-01 -1.18201387e+00 5.60681999e-01 8.19407761e-01 1.93707526e-01 7.42463648e-01 6.76039219e-01 -8.78175139e-01 -1.05687451e+00 -1.13062131e+00 5.35946071e-01 -6.98310971e-01 3.19228560e-01 -3.35498780e-01 -1.16618168e+00 4.66141641e-01 -1.87372118e-01 7.47352466e-02 6.02514863e-01 -1.88150316e-01 -1.62660882e-01 2.04168141e-01 -1.41005135e+00 3.82202566e-01 8.09114337e-01 -4.08749759e-01 -4.41453964e-01 3.98243666e-01 3.37512761e-01 -8.14231932e-01 -1.21194875e+00 7.06885755e-01 1.86413944e-01 -6.85466588e-01 1.28374290e+00 1.48252353e-01 -3.75687808e-01 -9.29401219e-01 -3.18310708e-01 -1.29670036e+00 -2.08204657e-01 -5.74419320e-01 1.48511648e-01 1.29022074e+00 1.13609500e-01 -6.82956696e-01 9.40199375e-01 7.17155576e-01 -2.55492628e-01 -4.24200594e-01 -1.23736143e+00 -9.63364482e-01 -3.63651365e-02 -5.96783757e-01 8.65513742e-01 1.02479649e+00 -3.39812458e-01 -2.99988627e-01 1.48547634e-01 9.19621825e-01 9.42366302e-01 2.07153231e-01 1.33738518e+00 -1.59964418e+00 2.31000647e-01 -4.80676889e-01 -8.35477233e-01 -9.13652480e-01 1.54025123e-01 -8.84962201e-01 3.63217562e-01 -1.59354079e+00 1.64014027e-01 -8.15364122e-01 1.44318968e-01 3.48434895e-02 -2.77899444e-01 2.15799689e-01 3.17155987e-01 8.40902150e-01 -3.25590223e-01 5.46387434e-01 6.79637849e-01 8.30437392e-02 -3.36303055e-01 3.21315527e-01 -1.17539048e-01 8.47122550e-01 9.02674675e-01 -7.92013466e-01 1.02519855e-01 -2.69044399e-01 -3.63695383e-01 -2.73685217e-01 3.08672428e-01 -1.33805680e+00 4.99236494e-01 -2.68943429e-01 1.16705984e-01 -1.18602109e+00 6.76943779e-01 -1.07529581e+00 7.91356206e-01 5.73551431e-02 6.10214651e-01 4.26395893e-01 3.97107482e-01 3.67416799e-01 -1.89331144e-01 -9.66200307e-02 9.15305078e-01 2.10735798e-01 -2.92195708e-01 6.95185006e-01 -4.43858877e-02 -2.66552597e-01 1.06363499e+00 -4.91193563e-01 -2.54668985e-02 -1.35165721e-01 -5.96245408e-01 9.51153114e-02 1.08973408e+00 3.29406291e-01 7.65438914e-01 -1.44058609e+00 -7.81910181e-01 2.77457684e-01 2.09303528e-01 5.81337690e-01 1.44455209e-01 9.58335519e-01 -5.45724213e-01 -3.28069963e-02 2.00049609e-01 -1.31042886e+00 -1.54728174e+00 2.04727083e-01 5.49740672e-01 -1.43831382e-02 -4.34347749e-01 7.55013764e-01 8.90664086e-02 -9.84146178e-01 -1.51208982e-01 -4.23700720e-01 1.91464096e-01 -3.24336827e-01 1.33525804e-01 4.13204253e-01 2.70521671e-01 -1.19410539e+00 -5.42059660e-01 1.09416831e+00 2.28110194e-01 -2.43345276e-01 1.11580634e+00 2.21566949e-02 -2.31688097e-01 2.00794160e-01 1.04300010e+00 2.40217522e-01 -1.10377526e+00 -4.35059100e-01 1.22949094e-01 -9.43697691e-01 -1.41898796e-01 9.88073945e-02 -1.02927184e+00 5.36831141e-01 8.91398489e-01 -7.95294195e-02 6.51559055e-01 4.08428133e-01 6.08126283e-01 -1.04402594e-01 5.82717001e-01 -7.49068856e-01 -4.72560734e-01 5.98745883e-01 6.67332351e-01 -1.13431633e+00 1.94128916e-01 -8.99754703e-01 -2.39191413e-01 8.12796175e-01 5.74410081e-01 -4.63600516e-01 9.25266087e-01 4.54856038e-01 1.11361258e-01 -3.92733008e-01 -1.24536753e-01 -2.54689157e-01 5.21596611e-01 7.64832675e-01 1.81412235e-01 1.88023627e-01 1.26607105e-01 1.98658794e-01 -3.68914425e-01 -3.33981931e-01 7.42598763e-03 9.11402047e-01 -4.45755422e-01 -1.30335820e+00 -1.16981518e+00 3.39438498e-01 -7.91917220e-02 2.78919172e-02 -1.01640508e-01 6.81862473e-01 6.34228736e-02 7.79329479e-01 5.74335277e-01 -4.46183950e-01 5.40955842e-01 2.12813970e-02 3.65764976e-01 -8.02147925e-01 -3.77010584e-01 5.19275665e-01 -4.91967440e-01 -6.07517302e-01 -5.32045722e-01 -9.70678687e-01 -1.29434919e+00 -3.52440864e-01 -7.02814519e-01 3.39854479e-01 1.02635777e+00 6.39768183e-01 7.34722733e-01 5.51116932e-03 6.27018869e-01 -1.42328048e+00 -3.31052661e-01 -8.28257442e-01 -4.93432939e-01 4.44294870e-01 -4.16295640e-02 -1.03440249e+00 -4.28647429e-01 -1.36705026e-01]
[7.7567458152771, -2.9714651107788086]
dc5b5e55-470c-4a31-9cfc-1fb88e3880bf
pokemonchat-auditing-chatgpt-for-pokemon
2306.03024
null
https://arxiv.org/abs/2306.03024v1
https://arxiv.org/pdf/2306.03024v1.pdf
PokemonChat: Auditing ChatGPT for Pokémon Universe Knowledge
The recently released ChatGPT model demonstrates unprecedented capabilities in zero-shot question-answering. In this work, we probe ChatGPT for its conversational understanding and introduce a conversational framework (protocol) that can be adopted in future studies. The Pok\'emon universe serves as an ideal testing ground for auditing ChatGPT's reasoning capabilities due to its closed world assumption. After bringing ChatGPT's background knowledge (on the Pok\'emon universe) to light, we test its reasoning process when using these concepts in battle scenarios. We then evaluate its ability to acquire new knowledge and include it in its reasoning process. Our ultimate goal is to assess ChatGPT's ability to generalize, combine features, and to acquire and reason over newly introduced knowledge from human feedback. We find that ChatGPT has prior knowledge of the Pokemon universe, which can reason upon in battle scenarios to a great extent, even when new information is introduced. The model performs better with collaborative feedback and if there is an initial phase of information retrieval, but also hallucinates occasionally and is susceptible to adversarial attacks.
['Ilias Chalkidis', 'Jiaang Li', 'Laura Cabello']
2023-06-05
null
null
null
null
['information-retrieval']
['natural-language-processing']
[-1.31388620e-01 3.95024329e-01 3.19639206e-01 -1.17690139e-01 -6.77751184e-01 -9.11487818e-01 7.16090024e-01 4.57825810e-02 -4.23257440e-01 9.07639086e-01 4.08192247e-01 -6.22540176e-01 -3.53334486e-01 -8.95856261e-01 -4.84960824e-02 -2.63048381e-01 -3.69154245e-01 7.88370132e-01 2.52264142e-01 -1.02065861e+00 3.44752729e-01 1.36057019e-01 -1.06248379e+00 6.45180225e-01 5.86227655e-01 3.68321985e-01 -8.69914815e-02 1.27519000e+00 8.57107341e-02 1.70648348e+00 -1.30232763e+00 -9.30732906e-01 2.72510260e-01 -3.45868707e-01 -1.80333686e+00 -3.39092433e-01 -9.86788496e-02 -5.75321317e-01 -6.20935440e-01 7.19737113e-01 5.95115900e-01 6.02430761e-01 3.12958807e-01 -1.33693409e+00 -4.27821249e-01 7.62620032e-01 3.85390729e-01 5.50322771e-01 1.03156269e+00 7.24272192e-01 1.06394529e+00 -1.03562847e-01 8.73817861e-01 1.19290996e+00 9.24976468e-01 8.88109386e-01 -7.70104527e-01 -4.49379027e-01 -1.19945817e-01 6.04825914e-01 -9.68960464e-01 -2.58819938e-01 4.93841380e-01 -9.82619673e-02 1.01736367e+00 7.46112943e-01 4.14327383e-01 1.44776654e+00 1.26612008e-01 7.97613800e-01 1.10191727e+00 -2.98937827e-01 5.12657166e-01 3.42193604e-01 2.08818182e-01 4.33306187e-01 -4.89040911e-01 4.19739485e-02 -9.02213573e-01 -6.45752728e-01 1.62066609e-01 -5.11109233e-01 -3.65467459e-01 3.79689991e-01 -7.83332944e-01 8.59828413e-01 2.72556871e-01 3.61955285e-01 -3.89585376e-01 -9.75849554e-02 5.26640654e-01 1.03612316e+00 1.75962910e-01 1.17288435e+00 -4.12563324e-01 -1.06240058e+00 -1.50183618e-01 2.91330040e-01 1.60465729e+00 5.43944657e-01 5.02684176e-01 -3.92299652e-01 -1.98284805e-01 5.80945134e-01 -3.78879875e-01 2.10765406e-01 4.40738559e-01 -1.45789874e+00 5.12783110e-01 5.36288798e-01 2.76735067e-01 -9.71709669e-01 -4.19954687e-01 -7.04906508e-03 -1.83598191e-01 -2.11666543e-02 4.31498617e-01 -6.61409378e-01 -4.33048569e-02 1.70881438e+00 3.03491652e-01 -7.56385103e-02 6.49821460e-01 8.46062899e-01 8.96903098e-01 4.19704109e-01 -2.93322257e-03 5.77429915e-03 1.30577397e+00 -6.62140608e-01 -5.31310022e-01 -4.78022039e-01 9.39254284e-01 -5.18756211e-01 9.45809543e-01 5.43801308e-01 -9.26365674e-01 -2.43639648e-01 -6.98483527e-01 2.52020329e-01 -3.30085188e-01 -8.67781222e-01 1.07585764e+00 9.63249147e-01 -1.05715549e+00 6.06946051e-01 -2.93893486e-01 -6.65535212e-01 6.01615496e-02 1.51056170e-01 -2.85285503e-01 -1.95089981e-01 -2.03742552e+00 1.48037720e+00 2.86169171e-01 -1.90871209e-02 -8.13902020e-01 -2.56251544e-01 -5.97092330e-01 2.92039439e-02 7.89660335e-01 -7.13990331e-01 1.74617553e+00 -3.72520655e-01 -1.76423538e+00 4.01666224e-01 1.97402105e-01 -5.74763358e-01 6.83241904e-01 7.57487491e-02 -3.20832491e-01 6.34335160e-01 1.25883799e-02 2.14666575e-01 3.76209974e-01 -9.87662792e-01 -3.39966148e-01 -1.77530497e-01 1.03727436e+00 5.70943356e-01 -1.68754041e-01 -2.06856765e-02 6.22262545e-02 -5.97022697e-02 -3.15485328e-01 -9.17213440e-01 -8.33342150e-02 -4.64226961e-01 -2.89769143e-01 -1.17625140e-01 8.41234028e-01 -6.28733695e-01 1.02033770e+00 -1.72873282e+00 -1.70530573e-01 3.14595222e-01 3.15187007e-01 5.39554119e-01 -3.17538269e-02 1.33952284e+00 1.64307892e-01 2.00842366e-01 6.83578774e-02 -1.07045867e-01 6.13660328e-02 4.26723778e-01 -1.96071982e-01 -6.81962445e-02 -1.10408731e-01 9.78308201e-01 -1.12869442e+00 -2.52021641e-01 1.15842558e-01 -1.12363979e-01 -7.14849055e-01 3.20604980e-01 -2.83199638e-01 5.78096449e-01 -5.73912024e-01 3.73522520e-01 2.07047760e-01 -1.65890262e-01 3.20594400e-01 4.06802863e-01 3.31797242e-01 4.27662909e-01 -6.00480378e-01 1.45366430e+00 -4.87096161e-01 7.82472372e-01 1.01721838e-01 -8.66967440e-01 8.60679209e-01 5.39923608e-01 -3.19474116e-02 -3.90799582e-01 5.58974802e-01 -3.43758613e-01 3.67723584e-01 -9.98117685e-01 7.41801798e-01 -4.66330498e-01 -2.98165590e-01 1.05427504e+00 1.10012867e-01 -3.70624006e-01 -5.77084310e-02 8.45140338e-01 1.73655057e+00 -7.30197489e-01 1.69338167e-01 -2.33436096e-03 4.07540679e-01 3.60701799e-01 -1.93499327e-02 1.41973293e+00 -4.35545295e-01 5.94325364e-02 5.87392390e-01 -4.62019682e-01 -4.38902229e-01 -8.90805185e-01 2.92557091e-01 1.18583047e+00 1.62799805e-01 -6.35404527e-01 -6.45137608e-01 -7.26192236e-01 -3.62509161e-01 1.33358896e+00 -7.04399347e-01 -5.86590469e-01 -1.27757892e-01 -3.88015300e-01 1.11912906e+00 2.44879559e-01 9.66543317e-01 -1.23829114e+00 -7.02604413e-01 2.06808478e-01 -9.43230748e-01 -9.83233273e-01 1.03854313e-02 4.23702709e-02 -5.15844285e-01 -1.41786075e+00 3.85564193e-02 -2.07256049e-01 1.42234936e-02 6.35936856e-02 9.76960659e-01 5.00513852e-01 -1.29907146e-01 1.27150905e+00 -8.63173544e-01 -4.18623149e-01 -9.94633615e-01 3.16039957e-02 -2.16221642e-02 -4.04136360e-01 5.83584368e-01 -6.26129270e-01 -4.15817142e-01 6.39380991e-01 -8.87157023e-01 -4.42426085e-01 8.41169506e-02 9.06007946e-01 -9.21916127e-01 2.81427979e-01 5.45333028e-01 -1.20553005e+00 1.39359331e+00 -6.31826937e-01 2.51593709e-01 3.53139609e-01 -1.46194980e-01 -3.19475204e-01 5.08512080e-01 -4.38230246e-01 -1.45925510e+00 -9.97592151e-01 -2.35717133e-01 2.14691716e-03 -7.70316273e-02 7.48567164e-01 1.47323430e-01 -2.13863209e-01 1.30942023e+00 5.70323095e-02 1.76823184e-01 -1.50078773e-01 3.97871137e-01 8.38565588e-01 6.38435543e-01 -9.35087800e-01 9.16249335e-01 3.26129854e-01 -5.98235488e-01 -8.75324428e-01 -7.04709649e-01 -4.14982289e-01 -2.65973926e-01 -5.58418155e-01 5.83861411e-01 -5.46532512e-01 -1.62966442e+00 4.95009869e-01 -1.31327367e+00 -6.52429104e-01 -1.78537846e-01 3.69040996e-01 -4.85602975e-01 7.77764261e-01 -9.71012712e-01 -1.29009032e+00 -2.81013101e-01 -5.49978614e-01 2.10494533e-01 1.52038574e-01 -9.60902631e-01 -1.30617821e+00 1.74314752e-01 1.11339915e+00 5.30524015e-01 -1.46501005e-01 7.37254024e-01 -1.36912823e+00 -1.68640956e-01 -4.41731364e-01 1.97984532e-01 4.25277382e-01 -6.45499751e-02 -5.11384070e-01 -1.23027289e+00 -3.36623490e-01 6.17875338e-01 -6.95521176e-01 2.38639414e-01 -5.77960730e-01 5.48903525e-01 -6.21858537e-01 1.89421505e-01 -1.06902905e-01 5.95490515e-01 3.78598720e-01 8.55687916e-01 5.70019662e-01 -1.69005960e-01 9.74833012e-01 5.26468456e-01 5.01773417e-01 4.39128339e-01 2.33806670e-01 2.87282079e-01 6.23465300e-01 3.45415652e-01 -2.35560104e-01 4.78908360e-01 7.45589137e-01 -2.40574867e-01 -4.26033288e-01 -1.08330584e+00 2.32895285e-01 -1.76263249e+00 -1.25344193e+00 2.67207861e-01 1.63585269e+00 8.77881527e-01 3.98389488e-01 -2.48591945e-01 4.49150056e-02 5.55741549e-01 -3.08113266e-03 -4.10752982e-01 -9.50958967e-01 9.22205970e-02 3.90162379e-01 -1.32702634e-01 6.81013823e-01 -4.65341508e-01 9.02073443e-01 6.96672869e+00 4.76085871e-01 -4.67198819e-01 9.83506367e-02 2.83947766e-01 2.24911094e-01 -1.46975651e-01 3.22643161e-01 -2.39882153e-02 2.43555829e-01 9.30650353e-01 -4.63385075e-01 6.57824457e-01 6.75281525e-01 -3.64441037e-01 -5.41674078e-01 -9.23074841e-01 5.57854116e-01 2.52269238e-01 -1.19341886e+00 -1.34701774e-01 -2.67844886e-01 2.23201394e-01 -1.57768413e-01 -2.22264916e-01 1.01026845e+00 9.27597046e-01 -7.97646463e-01 6.00813702e-02 3.53802562e-01 2.65532702e-01 -5.90500474e-01 1.15259266e+00 9.08863604e-01 -4.69503164e-01 -4.08351064e-01 -1.16929777e-01 -9.79834735e-01 -3.34182801e-03 -3.58739734e-01 -1.62585497e+00 7.43576288e-01 6.11945212e-01 7.02813193e-02 -5.37699759e-01 7.18656480e-01 -3.89573932e-01 6.70982778e-01 -3.21759135e-01 -4.04785693e-01 2.69142389e-01 5.95706478e-02 8.29923391e-01 8.30876291e-01 -1.45390615e-01 8.69999349e-01 3.28662246e-01 4.94362175e-01 6.69301450e-02 -2.56070167e-01 -6.43842876e-01 -3.39433216e-02 6.26979589e-01 8.75030458e-01 -2.30429590e-01 -4.52944338e-01 2.67221015e-02 1.02414799e+00 1.97228342e-01 4.39211845e-01 -4.75533575e-01 -3.41681600e-01 5.82767248e-01 -3.68235290e-01 -1.66438296e-01 7.10562691e-02 1.42002314e-01 -1.11199725e+00 -2.17049226e-01 -1.17884874e+00 8.25433910e-01 -1.23583591e+00 -1.51180041e+00 8.01500261e-01 1.67559326e-01 -8.42956722e-01 -6.28428996e-01 -3.58912557e-01 -1.17817664e+00 6.90355778e-01 -8.67815495e-01 -7.76031017e-01 -3.47278476e-01 8.93963754e-01 3.75089586e-01 -2.60181367e-01 1.28141093e+00 -2.25092456e-01 -1.23279788e-01 4.88974661e-01 -3.24489325e-01 4.87480491e-01 6.68060422e-01 -9.73923862e-01 4.48319614e-01 3.44714522e-01 -1.21313937e-01 1.03331721e+00 1.14909863e+00 -7.20045567e-01 -1.34194553e+00 -4.32618558e-01 5.66345572e-01 -1.17212820e+00 1.12986887e+00 -2.38818645e-01 -1.05131292e+00 9.43334460e-01 2.31300816e-01 -7.77826488e-01 1.04067469e+00 7.21292675e-01 -4.65843111e-01 3.05015683e-01 -1.34801960e+00 7.84102321e-01 9.25275683e-01 -1.06935382e+00 -1.56804740e+00 5.07662237e-01 5.94831109e-01 -4.97355759e-01 -7.93077528e-01 -8.22709203e-02 2.79152244e-01 -1.22686934e+00 6.64505363e-01 -1.12888181e+00 1.90694481e-01 2.10251689e-01 -3.98167968e-03 -1.54199529e+00 1.52870595e-01 -1.30897021e+00 3.28163475e-01 9.68861699e-01 2.88743258e-01 -1.02431989e+00 8.36297214e-01 1.25211596e+00 9.81786177e-02 -2.48552158e-01 -1.00308347e+00 -7.21951008e-01 3.06006614e-02 -8.36974323e-01 3.51683915e-01 1.10554898e+00 9.20026839e-01 5.07097125e-01 -5.94939411e-01 6.00622855e-02 2.64687032e-01 -2.79139757e-01 1.21494818e+00 -1.10835862e+00 -6.26552045e-01 3.75974961e-02 -4.25555170e-01 -8.47592831e-01 -3.93220969e-02 -6.85696185e-01 2.16265898e-02 -1.27450466e+00 1.57701760e-03 -1.96596473e-01 2.09086314e-01 4.29701507e-01 -1.88641742e-01 -2.35214643e-02 4.85005289e-01 1.35231212e-01 -8.41096878e-01 3.47732931e-01 1.28234279e+00 -1.33306026e-01 -1.07898697e-01 2.11024672e-01 -8.95136714e-01 7.11691976e-01 9.10173655e-01 -3.27077985e-01 -5.72508872e-01 -3.29801664e-02 3.07385206e-01 5.09105682e-01 4.13229048e-01 -9.86954331e-01 7.57291317e-01 -5.56798019e-02 -1.21110342e-01 -7.41662038e-03 8.31432939e-01 -5.53703308e-01 4.44445051e-02 3.86119068e-01 -4.77728665e-01 -1.64633945e-01 3.27969372e-01 5.88780165e-01 -1.59638464e-01 -5.40431559e-01 1.25268891e-01 -5.46729982e-01 -7.13847756e-01 -1.38646156e-01 -1.18836677e+00 3.70900780e-01 8.00297201e-01 -3.07477921e-01 -7.84373999e-01 -1.40576303e+00 -1.01177621e+00 6.35772884e-01 2.87700534e-01 4.05020803e-01 5.26539266e-01 -7.38127887e-01 -6.33823931e-01 -1.22459747e-01 2.12777525e-01 -5.31508327e-01 6.85903072e-01 5.99306941e-01 -5.48968136e-01 1.15444638e-01 -2.31194034e-01 -1.82188198e-01 -1.23014319e+00 4.38824087e-01 3.42328817e-01 -3.06768000e-01 -5.43238997e-01 9.19523895e-01 -1.41028672e-01 -7.54172206e-01 1.88050613e-01 2.47097135e-01 -2.34690100e-01 -6.66680187e-02 7.74158835e-01 2.08869830e-01 -1.81757342e-02 -4.44799140e-02 -1.55294478e-01 -3.02510440e-01 -3.00252140e-01 -3.50302488e-01 9.96087015e-01 -1.99547172e-01 1.72140878e-02 5.65844178e-01 9.36576307e-01 2.69098743e-03 -5.82337797e-01 -2.60429114e-01 -5.93718588e-02 -6.73472762e-01 -5.22049546e-01 -1.18045592e+00 -2.88287938e-01 6.90282881e-01 -7.66351521e-02 6.48259759e-01 5.95579326e-01 -5.12916073e-02 8.25644612e-01 1.39597714e+00 7.55930781e-01 -1.05001497e+00 3.39037955e-01 8.55259001e-01 7.43864238e-01 -1.06788278e+00 -4.14423198e-02 -1.67445645e-01 -1.16856122e+00 1.12575614e+00 6.47824585e-01 2.50950903e-01 3.76437753e-01 -8.02654549e-02 5.36853909e-01 -5.95253766e-01 -1.22329688e+00 -5.87377325e-02 -3.18891406e-01 8.30178142e-01 -3.43491316e-01 -8.77401084e-02 9.69083756e-02 6.09643757e-01 -8.47309351e-01 -1.87064305e-01 1.04339349e+00 1.09617531e+00 -3.96418482e-01 -8.88635278e-01 -4.46064532e-01 3.40115905e-01 -1.79740682e-01 -1.50073484e-01 -9.97111857e-01 9.30704176e-01 -3.47671568e-01 1.63437426e+00 -3.23269427e-01 -8.05847585e-01 3.10206980e-01 4.43056852e-01 1.17702425e-01 -6.22184873e-01 -1.18688381e+00 -8.49347234e-01 8.86969090e-01 -4.51350510e-01 -2.86932409e-01 -5.28851211e-01 -7.96602726e-01 -8.15005004e-01 -4.21391100e-01 1.02644205e+00 1.71480387e-01 1.20637560e+00 -2.33810097e-02 1.78195029e-01 5.55221915e-01 -2.36815259e-01 -7.78964877e-01 -1.23410511e+00 -2.91886747e-01 6.81904614e-01 -1.53700516e-01 -4.93889093e-01 -7.52034307e-01 -4.68830854e-01]
[12.636487007141113, 8.050992965698242]
7c3a2a79-002a-4931-b30d-79e93c04a48a
stock-price-prediction-using-dynamic-neural
2306.12969
null
https://arxiv.org/abs/2306.12969v1
https://arxiv.org/pdf/2306.12969v1.pdf
Stock Price Prediction using Dynamic Neural Networks
This paper will analyze and implement a time series dynamic neural network to predict daily closing stock prices. Neural networks possess unsurpassed abilities in identifying underlying patterns in chaotic, non-linear, and seemingly random data, thus providing a mechanism to predict stock price movements much more precisely than many current techniques. Contemporary methods for stock analysis, including fundamental, technical, and regression techniques, are conversed and paralleled with the performance of neural networks. Also, the Efficient Market Hypothesis (EMH) is presented and contrasted with Chaos theory using neural networks. This paper will refute the EMH and support Chaos theory. Finally, recommendations for using neural networks in stock price prediction will be presented.
['David Noel']
2023-06-18
null
null
null
null
['stock-price-prediction']
['time-series']
[-8.42554629e-01 -6.67747080e-01 -1.85960412e-01 -7.46255666e-02 5.11917353e-01 -7.50414193e-01 5.80606997e-01 -4.33743179e-01 -3.09866369e-01 8.81826699e-01 2.22133636e-01 -7.76074052e-01 -4.04395103e-01 -9.58715558e-01 -1.35243936e-02 -6.43397927e-01 -5.70341349e-01 1.08887926e-01 -1.72690824e-01 -7.27132082e-01 7.55245209e-01 7.95700192e-01 -1.41054535e+00 -2.59824753e-01 3.70736927e-01 1.34301138e+00 -1.40786484e-01 4.90044504e-01 -3.33167493e-01 1.29354346e+00 -7.20897675e-01 -2.12570965e-01 7.40302980e-01 -5.08490145e-01 1.32497177e-01 -5.64343929e-01 -3.29016894e-01 -4.96133387e-01 -7.07740188e-01 1.05359423e+00 2.11960152e-01 6.34018406e-02 5.46039045e-01 -9.96220648e-01 -1.21656728e+00 8.22989702e-01 -1.12120129e-01 9.46852744e-01 -1.23192050e-01 4.91062403e-02 7.09350646e-01 -8.60284209e-01 2.47564629e-01 6.22038484e-01 1.16875720e+00 1.49053887e-01 -9.32003200e-01 -8.64203811e-01 -2.14150578e-01 1.05617523e-01 -1.06466746e+00 -2.46598765e-01 1.01155150e+00 -4.23093557e-01 1.08069706e+00 3.43510747e-01 1.39559388e+00 3.24009359e-01 7.29432583e-01 5.47592938e-01 1.27021253e+00 -2.84768075e-01 2.16931328e-01 5.07213734e-02 3.74340624e-01 2.18906224e-01 6.83049023e-01 8.34458113e-01 -9.51288790e-02 -2.00611085e-01 1.05839741e+00 2.77302593e-01 -1.20877028e-01 3.15977573e-01 -8.32933426e-01 1.01613879e+00 1.50469005e-01 9.46261168e-01 -8.72891307e-01 9.84919295e-02 3.85009944e-01 8.99984539e-01 4.93450671e-01 7.69438863e-01 -6.71739697e-01 -3.73994023e-01 -1.30142856e+00 4.94177401e-01 1.28449452e+00 2.93694407e-01 2.13946745e-01 1.16861510e+00 3.33028525e-01 1.98766425e-01 7.58427829e-02 6.18399739e-01 1.29137921e+00 -7.80512869e-01 -8.27807412e-02 3.25997353e-01 1.90702289e-01 -1.39368534e+00 -6.41120195e-01 -6.26699865e-01 -1.03796458e+00 5.20247459e-01 1.93658099e-01 -3.58771801e-01 -4.66636598e-01 1.06077111e+00 -5.44756353e-01 3.64104807e-01 4.68459904e-01 2.82912433e-01 3.18508297e-01 9.37812328e-01 -3.74893785e-01 -7.82279730e-01 8.08013499e-01 -4.78237391e-01 -1.07314229e+00 2.21584737e-01 8.88310075e-02 -5.99146426e-01 3.51783574e-01 2.31448382e-01 -1.19148600e+00 -3.49892557e-01 -9.74400938e-01 6.10121310e-01 -6.29583240e-01 -4.81762558e-01 6.24193490e-01 6.10615015e-01 -1.09715986e+00 1.07980597e+00 -6.28675878e-01 4.50688064e-01 -2.84317881e-01 3.97219747e-01 4.23407257e-01 1.06987381e+00 -1.39386320e+00 1.34920371e+00 7.25452900e-01 4.77902681e-01 3.85849655e-01 -5.39548755e-01 -6.12814248e-01 9.76377651e-02 -4.30214375e-01 -1.41851455e-01 1.19183636e+00 -6.88079357e-01 -1.42799258e+00 1.54578671e-01 -5.76417055e-03 -1.23374534e+00 4.21235740e-01 3.08751374e-01 -1.04547513e+00 2.67672203e-02 -2.05156580e-01 -8.47676471e-02 5.66005588e-01 -4.96433228e-01 -5.16194224e-01 -2.41971742e-02 -8.01064134e-01 -1.86065763e-01 -1.48875058e-01 -1.16635919e-01 7.69192755e-01 -1.17586207e+00 3.91471207e-01 -6.53366446e-01 -1.38283551e-01 -5.40805280e-01 2.63964534e-01 -3.18086088e-01 9.57445562e-01 -8.34294736e-01 1.59666407e+00 -1.95266259e+00 -8.47245038e-01 8.47422183e-01 1.41144574e-01 2.05119073e-01 3.00742656e-01 6.43451691e-01 -4.65152979e-01 6.12628385e-02 4.53791469e-02 4.31816578e-01 1.35824814e-01 2.21006140e-01 -1.02181923e+00 3.22367907e-01 -1.45157993e-01 1.44911516e+00 -3.49909395e-01 1.84107929e-01 3.24347436e-01 -7.27389380e-02 8.38286206e-02 -5.32035887e-01 1.54621795e-01 -2.35381007e-01 -2.14091375e-01 6.39503658e-01 5.65949380e-01 -2.88548648e-01 -1.87435091e-01 3.81566346e-01 -8.53616178e-01 1.30614296e-01 -1.05574918e+00 2.11149126e-01 1.23433070e-02 1.25107110e+00 -5.70759356e-01 -1.04179657e+00 1.52548778e+00 6.37924552e-01 6.33893907e-01 -9.39915061e-01 1.11871041e-01 8.22217345e-01 3.80388975e-01 -4.15071428e-01 5.56104004e-01 -4.48107392e-01 1.56728670e-01 8.78360212e-01 -5.12930810e-01 1.95713690e-03 1.52234927e-01 -5.62844813e-01 5.95624089e-01 -3.69330496e-01 2.97990918e-01 -4.95354146e-01 2.40376592e-01 1.56651318e-01 5.75743556e-01 6.83912218e-01 -3.33723307e-01 7.14031756e-02 3.66301060e-01 -1.32217014e+00 -1.18464804e+00 -7.37049997e-01 -3.13404828e-01 2.94376552e-01 -7.53267631e-02 4.07181621e-01 -3.47995847e-01 3.21038723e-01 5.05409002e-01 8.80743921e-01 -6.97606564e-01 1.34654105e-01 -5.89077830e-01 -9.19787288e-01 5.68169653e-01 5.02532482e-01 6.69737160e-01 -1.63207936e+00 -7.03983426e-01 7.98989058e-01 3.10257137e-01 -7.13681877e-01 -1.57305703e-01 2.92519450e-01 -1.19243765e+00 -7.95924783e-01 -8.84070158e-01 -8.58197749e-01 1.84622407e-01 1.94245815e-01 1.00617373e+00 2.91267753e-01 1.54938236e-01 8.83522555e-02 9.89965275e-02 -7.96630979e-01 -5.17002940e-01 -2.56695300e-01 5.41553736e-01 -3.63397390e-01 8.24438095e-01 -8.63075793e-01 -6.41586661e-01 1.90837130e-01 -6.95477068e-01 -4.28692549e-01 3.09044123e-01 8.23876321e-01 1.36067569e-01 6.26419127e-01 1.01514161e+00 -8.27167407e-02 1.36567140e+00 -5.06617010e-01 -1.18281806e+00 1.72910616e-01 -1.41613150e+00 -5.46934782e-03 5.76840103e-01 -6.21998608e-01 -5.56304693e-01 -3.75047982e-01 1.21653220e-02 -4.10197198e-01 3.39566797e-01 8.68976831e-01 1.06272531e+00 -2.65159607e-01 5.15923381e-01 9.92018640e-01 5.81702650e-01 -2.03059077e-01 -3.75010937e-01 3.85828793e-01 7.13481843e-01 1.61538988e-01 1.07746089e+00 2.10548103e-01 -1.89260602e-01 -9.13558125e-01 3.39929201e-02 -1.25413269e-01 -5.21478713e-01 -2.63330936e-01 2.36225039e-01 -7.31912196e-01 -7.79285014e-01 8.88626277e-01 -8.36905301e-01 -1.09594036e-02 -6.90774143e-01 8.01934242e-01 -5.83818614e-01 -1.79234855e-02 -1.09116149e+00 -1.39661562e+00 -5.74456990e-01 -4.67584401e-01 -1.32818967e-01 4.25342023e-01 -2.98492968e-01 -1.45759749e+00 4.23911124e-01 -5.60116351e-01 9.26448703e-01 4.97924924e-01 5.83191276e-01 -1.02755940e+00 -2.28186443e-01 -8.22520137e-01 1.68209851e-01 3.42029333e-01 9.83315706e-02 2.62364239e-01 -5.30551553e-01 1.11850053e-01 7.24725723e-01 3.32178414e-01 6.01552308e-01 9.01873708e-01 2.96699405e-01 -7.38468945e-01 1.66124687e-01 5.94934046e-01 1.46355069e+00 1.11111188e+00 7.01808751e-01 9.75450575e-01 -3.03962350e-01 5.10640860e-01 -1.64387345e-01 4.55659986e-01 1.04554698e-01 -2.73304224e-01 -1.46955401e-01 1.66875884e-01 7.86209106e-01 -2.53393114e-01 2.54542768e-01 1.32839715e+00 -2.16000035e-01 2.68395126e-01 -7.70322978e-01 2.52855778e-01 -1.60527909e+00 -1.80977786e+00 -2.37922445e-01 1.37260735e+00 5.30175567e-01 5.26170313e-01 5.14634788e-01 5.41436076e-01 6.89727783e-01 1.58765540e-01 -7.73031414e-01 -4.01257426e-01 -7.01977968e-01 1.58430219e-01 7.43777156e-01 2.13826001e-01 -8.85520339e-01 5.81169784e-01 8.11522388e+00 3.09500903e-01 -1.41725218e+00 -5.34661055e-01 7.15017617e-01 2.30000257e-01 -3.47593993e-01 -3.59912843e-01 -5.50461590e-01 8.46787453e-01 1.18027246e+00 -1.06516838e+00 4.54414189e-01 8.84576082e-01 4.83010381e-01 1.99454889e-01 -1.23669274e-01 9.25975204e-01 -4.66500878e-01 -1.90303528e+00 -1.88304901e-01 2.67888397e-01 8.02872002e-01 1.98599637e-01 4.31624204e-01 3.84624720e-01 1.38445184e-01 -7.11537421e-01 8.43688428e-01 7.93934464e-01 1.07023910e-01 -7.74395704e-01 9.09730196e-01 4.65747416e-01 -1.28909016e+00 -5.00162959e-01 -5.53311110e-01 -7.73485601e-01 1.89010680e-01 2.72729307e-01 -6.83454573e-02 2.29664240e-02 5.61448216e-01 7.47885823e-01 -9.25615430e-02 1.18475747e+00 2.66197801e-01 4.08596069e-01 -5.06675899e-01 -7.87587106e-01 5.58205783e-01 -6.48354173e-01 4.69795406e-01 7.87266493e-01 6.74091816e-01 4.72273082e-01 -5.63714802e-01 8.93329263e-01 6.54625654e-01 -6.40217811e-02 -8.58120322e-01 -3.73152822e-01 5.43450475e-01 5.54154038e-01 -9.01885867e-01 -4.33725476e-01 -5.21901131e-01 4.73661065e-01 -7.28631556e-01 4.26093996e-01 -2.98440605e-01 -5.71908057e-01 5.95671713e-01 1.40902758e-01 1.56967744e-01 -4.23390329e-01 -9.15147066e-01 -1.17026889e+00 8.81263390e-02 -4.91370291e-01 1.77902952e-01 -6.00167334e-01 -1.50115991e+00 7.60031700e-01 -3.59667130e-02 -1.66109598e+00 -1.09242737e+00 -9.58475590e-01 -1.12629616e+00 1.02636564e+00 -1.39954388e+00 -1.06059723e-01 4.27789897e-01 6.48369014e-01 3.59997660e-01 -1.01979136e+00 8.28339398e-01 -2.07906187e-01 -3.40426058e-01 2.57948656e-02 8.62929404e-01 5.18880546e-01 -3.60673785e-01 -1.28041685e+00 1.04398155e+00 6.45295620e-01 9.18321237e-02 5.10722876e-01 7.67200053e-01 -8.75955224e-01 -9.42809045e-01 -5.73776186e-01 1.06786919e+00 2.29163393e-02 1.31636393e+00 2.72758245e-01 -1.06966567e+00 4.56900924e-01 2.43506998e-01 -3.49130213e-01 4.86658454e-01 -3.52052003e-01 1.69569984e-01 -1.37636494e-02 -1.10194647e+00 6.57433987e-01 2.59446830e-01 -2.85119534e-01 -1.11966705e+00 -7.61313438e-02 3.84920835e-01 -9.58997663e-03 -7.30877101e-01 2.24470019e-01 9.98259187e-01 -1.13544571e+00 8.13073575e-01 -3.44232619e-01 5.17322170e-03 5.07037379e-02 1.38127968e-01 -1.04738581e+00 -5.70833623e-01 -9.68282580e-01 -3.64733756e-01 4.05971467e-01 6.00662827e-01 -1.41290712e+00 9.65141237e-01 9.45380211e-01 2.45943904e-01 -5.15751481e-01 -1.04625964e+00 -1.31479144e+00 3.20350140e-01 -2.54115105e-01 8.43327463e-01 1.18941915e+00 2.31407657e-01 -3.44189435e-01 -2.48306736e-01 -2.63499349e-01 5.25312960e-01 2.33653992e-01 6.00279719e-02 -1.57527709e+00 9.82451290e-02 -1.19104135e+00 -4.67891991e-01 -6.07760429e-01 1.53280929e-01 -3.66490602e-01 -4.15709406e-01 -9.88580287e-01 -7.45532513e-01 -1.70602828e-01 -6.50316775e-01 -5.88434413e-02 3.38834882e-01 2.10295275e-01 3.28059286e-01 9.96316016e-01 5.59537947e-01 2.75888413e-01 8.24927747e-01 1.02131389e-01 -7.57516861e-01 5.01333773e-01 -5.06292880e-01 8.03088546e-01 1.28048193e+00 -2.02844143e-01 -1.79797009e-01 2.61588186e-01 3.77863765e-01 3.89794588e-01 6.78535551e-02 -1.02162969e+00 5.00796020e-01 -1.92149624e-01 8.73854041e-01 -1.05576181e+00 3.36963385e-02 -8.22019041e-01 4.77004349e-01 1.23371792e+00 -4.55575213e-02 1.23071289e+00 4.06252861e-01 2.27179170e-01 -5.59272587e-01 -4.15458083e-01 5.13790846e-01 -3.36300164e-01 -8.00824285e-01 6.70061335e-02 -7.90266216e-01 -2.72175252e-01 1.04699314e+00 -9.00034010e-01 -2.45519295e-01 -7.68434346e-01 -8.07446718e-01 7.78236762e-02 1.73419714e-02 3.70836616e-01 7.42096066e-01 -1.55619669e+00 -6.44037783e-01 6.06803060e-01 -6.45363390e-01 -1.04527164e+00 -1.04472384e-01 4.99712884e-01 -1.03553367e+00 7.39329338e-01 -4.68158633e-01 -1.65739972e-02 -4.51798350e-01 6.55878365e-01 8.18808496e-01 -1.00444600e-01 -8.58870506e-01 1.92936599e-01 -6.05141938e-01 1.96808174e-01 5.77315725e-02 -4.10582006e-01 -3.71385157e-01 4.33139890e-01 8.57594550e-01 4.55670029e-01 -2.26974741e-01 -4.23894465e-01 8.70739445e-02 5.92455387e-01 2.21204847e-01 -3.51097256e-01 1.74674535e+00 7.62588307e-02 -3.07757646e-01 9.51992095e-01 1.00099421e+00 -4.85613227e-01 -8.60130250e-01 -6.79388419e-02 6.84276581e-01 1.14597894e-01 -1.59759670e-01 -5.48268199e-01 -9.07080889e-01 2.58996427e-01 6.17041171e-01 1.32436419e+00 1.15165019e+00 -7.05070436e-01 1.00160408e+00 7.19651163e-01 7.76370913e-02 -1.38917434e+00 -4.78348553e-01 7.05296457e-01 8.79095018e-01 -9.35271561e-01 -2.83386678e-01 4.96124059e-01 -4.23934400e-01 1.79935026e+00 -1.38564482e-01 -8.82798791e-01 1.46685600e+00 6.95320606e-01 2.88655519e-01 -9.17170718e-02 -8.96766007e-01 5.31313792e-02 -5.30598313e-03 2.44372338e-01 1.09302603e-01 2.53055021e-02 -3.97100359e-01 4.25461054e-01 -8.54249239e-01 2.47243807e-01 5.94521761e-01 1.00849807e+00 -7.25078583e-01 -5.28844953e-01 -6.85592473e-01 6.76663101e-01 -7.68144429e-01 -3.77292722e-01 -3.02857727e-01 1.27950966e+00 -5.84210753e-01 5.48563778e-01 7.82968879e-01 -5.44618011e-01 2.85060316e-01 2.94039398e-01 -3.41132641e-01 2.34104618e-01 -8.06395173e-01 2.71375030e-01 -5.12638986e-01 1.34841859e-01 -1.78031832e-01 -8.13743234e-01 -9.73443329e-01 -8.16085279e-01 -1.81666464e-01 5.40000081e-01 6.23467386e-01 7.38703847e-01 -8.34814832e-02 -3.53642888e-02 1.08728278e+00 -6.62166238e-01 -9.12999451e-01 -8.32200110e-01 -1.14551270e+00 -3.10523063e-01 6.52237415e-01 -1.89301744e-01 -7.43750453e-01 -1.59642965e-01]
[4.499557018280029, 4.214079856872559]
1863a131-51ed-4579-8f66-841d5e06af99
juewu-mc-playing-minecraft-with-sample
2112.04907
null
https://arxiv.org/abs/2112.04907v1
https://arxiv.org/pdf/2112.04907v1.pdf
JueWu-MC: Playing Minecraft with Sample-efficient Hierarchical Reinforcement Learning
Learning rational behaviors in open-world games like Minecraft remains to be challenging for Reinforcement Learning (RL) research due to the compound challenge of partial observability, high-dimensional visual perception and delayed reward. To address this, we propose JueWu-MC, a sample-efficient hierarchical RL approach equipped with representation learning and imitation learning to deal with perception and exploration. Specifically, our approach includes two levels of hierarchy, where the high-level controller learns a policy to control over options and the low-level workers learn to solve each sub-task. To boost the learning of sub-tasks, we propose a combination of techniques including 1) action-aware representation learning which captures underlying relations between action and representation, 2) discriminator-based self-imitation learning for efficient exploration, and 3) ensemble behavior cloning with consistency filtering for policy robustness. Extensive experiments show that JueWu-MC significantly improves sample efficiency and outperforms a set of baselines by a large margin. Notably, we won the championship of the NeurIPS MineRL 2021 research competition and achieved the highest performance score ever.
['Wei Yang', 'Qiang Fu', 'Deheng Ye', 'Jianing Shi', 'Junyou Li', 'Zichuan Lin']
2021-12-07
null
null
null
null
['hierarchical-reinforcement-learning']
['methodology']
[ 5.76186739e-02 2.20619753e-01 -5.29215991e-01 2.39104986e-01 -9.11953092e-01 -6.15250826e-01 5.70977867e-01 -2.48297989e-01 -6.97374582e-01 9.56285834e-01 3.19219679e-01 -2.79095292e-01 -1.88695356e-01 -4.45562422e-01 -8.70534301e-01 -7.51113296e-01 -2.31712222e-01 4.99169916e-01 2.08500475e-01 -4.24098670e-01 2.83440143e-01 1.80568844e-01 -1.48593175e+00 1.70581505e-01 9.71909106e-01 1.29690313e+00 4.92893517e-01 7.71367192e-01 4.19257194e-01 1.44421995e+00 -2.21169367e-01 1.13169253e-01 6.05029404e-01 -4.23913181e-01 -7.56636143e-01 -4.91151586e-02 -1.10727504e-01 -5.44936478e-01 -3.33966941e-01 9.55697775e-01 6.44542873e-01 6.29954755e-01 4.63616818e-01 -1.31873560e+00 -5.04014730e-01 7.43257403e-01 -4.77909863e-01 -5.96370101e-02 -3.86632234e-02 8.66988182e-01 1.19263327e+00 -2.17382580e-01 6.68405175e-01 1.65823460e+00 1.46404892e-01 1.05992568e+00 -1.38957095e+00 -6.46480560e-01 6.41233683e-01 1.13537408e-01 -6.90863550e-01 -3.12376678e-01 2.09619433e-01 -2.10071579e-01 1.06762111e+00 -2.52082348e-01 7.13433385e-01 1.53132451e+00 1.43360049e-01 1.14583611e+00 1.43765378e+00 3.00383158e-02 7.12831259e-01 -4.27814454e-01 -4.99336153e-01 8.98029029e-01 -3.52582633e-02 7.72303402e-01 -3.68718237e-01 -1.69599205e-01 1.19971132e+00 -1.44221440e-01 -8.82941559e-02 -8.29644740e-01 -1.21125257e+00 9.39783871e-01 6.78940177e-01 -1.34637877e-01 -5.59272647e-01 7.31647968e-01 3.71103525e-01 5.87672353e-01 -2.10146353e-01 1.08219957e+00 -6.90436959e-01 -3.39853406e-01 -2.92481899e-01 7.80370116e-01 6.13950491e-01 7.99839437e-01 6.19305074e-01 4.35215980e-01 -6.22597992e-01 6.77281260e-01 1.22636490e-01 2.07265586e-01 7.42482483e-01 -1.74338937e+00 6.01788104e-01 4.10671115e-01 3.44945490e-01 -3.87318224e-01 -4.83359188e-01 -4.12154824e-01 -5.15195906e-01 8.37139368e-01 4.15739387e-01 -4.82121468e-01 -8.77680898e-01 2.16240716e+00 2.40775019e-01 1.35471717e-01 3.41909647e-01 1.04873276e+00 2.54692078e-01 3.87531489e-01 1.00573793e-01 -1.80759594e-01 1.06897402e+00 -1.34130335e+00 -3.17111433e-01 -3.66813689e-01 4.26410139e-01 1.05103357e-02 1.14653158e+00 3.99264008e-01 -1.28633833e+00 -3.43846142e-01 -9.44729388e-01 2.80025285e-02 -7.39003941e-02 4.43201847e-02 5.80994129e-01 1.59218279e-03 -9.87803996e-01 8.83108556e-01 -8.95886004e-01 -1.36260511e-02 7.14238465e-01 4.79601562e-01 -2.80159544e-02 2.05681667e-01 -9.80894089e-01 8.20345521e-01 5.32153368e-01 -4.18765575e-01 -1.72690988e+00 -5.20889342e-01 -6.29266322e-01 9.46111009e-02 1.20632231e+00 -7.23667681e-01 1.72367108e+00 -8.19852531e-01 -2.17565608e+00 3.61823440e-01 4.30393547e-01 -8.23734820e-01 7.18463123e-01 -3.48808497e-01 1.89207003e-01 -8.02214742e-02 1.17263235e-01 9.18843389e-01 9.61246252e-01 -1.18564820e+00 -8.89103889e-01 -3.49201232e-01 4.43317145e-01 5.97489417e-01 1.48880541e-01 -5.93153179e-01 -3.30285221e-01 -6.77354455e-01 -3.73982847e-01 -1.19481599e+00 -7.57059574e-01 -1.29738703e-01 -1.06622852e-01 -3.91068965e-01 3.65706533e-01 -3.13442975e-01 8.58760774e-01 -1.83424306e+00 5.83505332e-01 -4.70657833e-03 2.49468341e-01 1.79136783e-01 -5.31272590e-01 1.48553714e-01 3.44716430e-01 -5.26158288e-02 9.14613008e-02 -1.87026054e-01 2.15892330e-01 4.30067211e-01 -5.97221315e-01 2.52256632e-01 7.87409618e-02 1.20422721e+00 -1.18520224e+00 -1.14450030e-01 -9.23643187e-02 -9.24113318e-02 -8.34342718e-01 4.73857850e-01 -8.91377687e-01 8.60184133e-01 -8.25194061e-01 5.86502254e-01 5.37944818e-03 -1.97566107e-01 3.55354697e-01 4.14796621e-01 -1.72145009e-01 4.20490682e-01 -1.27476001e+00 1.84011459e+00 -5.33515096e-01 1.16388192e-02 3.80020082e-01 -8.56303871e-01 6.27621412e-01 1.44729614e-01 3.31648678e-01 -1.17768455e+00 1.63461044e-01 1.14787944e-01 1.79001749e-01 -2.59709567e-01 5.43258607e-01 1.48740739e-01 -3.09821784e-01 5.74707389e-01 1.78313866e-01 -2.65459716e-01 1.73903063e-01 6.49829879e-02 1.27258837e+00 7.99678743e-01 4.36322689e-01 -1.41035348e-01 1.41761139e-01 -4.40499932e-02 9.02526438e-01 1.15378892e+00 -4.37782496e-01 1.00149453e-01 9.45301712e-01 -1.78445399e-01 -8.84613037e-01 -9.35296774e-01 6.19235098e-01 1.59780788e+00 1.34066567e-01 -2.04833329e-01 -5.77760756e-01 -7.48688519e-01 3.34450483e-01 6.04872406e-01 -7.51743197e-01 -4.52378839e-01 -6.45143270e-01 -2.79922187e-01 3.69813025e-01 6.82774246e-01 5.64391971e-01 -1.62105596e+00 -1.24835694e+00 2.47661844e-01 1.62755206e-01 -9.19221997e-01 -5.20863533e-01 4.85298187e-01 -8.34849954e-01 -1.10480070e+00 -7.09924459e-01 -4.68820184e-01 2.07745045e-01 -1.99159812e-02 1.07852638e+00 -1.88050002e-01 -2.88761169e-01 5.16968071e-01 -2.46211469e-01 -2.63953656e-01 -2.45375112e-01 9.20319483e-02 3.14355761e-01 -2.63688356e-01 -4.40481395e-01 -5.94687164e-01 -8.12466800e-01 2.68191934e-01 -4.47138160e-01 -5.24161644e-02 9.67089593e-01 1.02429676e+00 6.17051482e-01 -3.84155184e-01 6.95595503e-01 -5.78710973e-01 9.47682202e-01 -4.42503870e-01 -1.08071899e+00 2.24277601e-01 -7.28297651e-01 6.29718542e-01 7.37558424e-01 -7.21849978e-01 -9.75243688e-01 1.19944140e-01 3.21517259e-01 -6.97088659e-01 7.74535015e-02 3.01488629e-03 1.45263067e-02 7.19038472e-02 7.27364242e-01 3.73415470e-01 2.87703574e-01 -3.16733450e-01 6.99152648e-01 1.52561918e-01 6.39605761e-01 -9.10394669e-01 5.53564250e-01 2.88548976e-01 5.68351559e-02 -2.88003594e-01 -7.34138727e-01 -3.25007774e-02 -8.43512192e-02 -9.34180543e-02 9.46277559e-01 -1.16207206e+00 -1.43921804e+00 2.01554954e-01 -6.97578609e-01 -1.24663103e+00 -7.20024347e-01 3.81150305e-01 -1.32251191e+00 -5.35566211e-02 -5.49656212e-01 -1.10559952e+00 -1.20679103e-01 -1.36343205e+00 8.59101295e-01 3.93523216e-01 1.60932809e-01 -5.02519906e-01 2.93927938e-01 2.15991989e-01 4.46729392e-01 1.10832036e-01 7.85535693e-01 -3.41412038e-01 -8.42398524e-01 5.63305080e-01 -2.33418494e-02 8.94728005e-02 -2.88887501e-01 -5.79475701e-01 -6.65630281e-01 -6.04750037e-01 -2.24540040e-01 -1.33466327e+00 1.21386218e+00 4.24256235e-01 1.40958011e+00 -4.97970968e-01 -7.14463368e-02 6.88525677e-01 1.12635148e+00 2.92074353e-01 4.07843828e-01 5.49758613e-01 4.22884077e-01 4.32636082e-01 7.28195131e-01 8.63618851e-01 5.50675452e-01 6.25216722e-01 1.02927077e+00 3.22086006e-01 1.55480117e-01 -6.49128020e-01 5.74298203e-01 -5.13234362e-03 -3.13501030e-01 6.76007047e-02 -2.99962044e-01 2.52878666e-01 -2.46022773e+00 -9.99983728e-01 7.90065587e-01 2.14986038e+00 8.45431268e-01 8.84391218e-02 6.96723282e-01 -4.39866751e-01 1.50274500e-01 2.72890002e-01 -1.40306711e+00 -2.87112534e-01 3.06242574e-02 1.42337801e-02 6.05276227e-01 2.66636729e-01 -9.88896966e-01 1.36207199e+00 5.86126471e+00 9.29839253e-01 -8.58433008e-01 1.89254116e-02 5.44618607e-01 -4.13918614e-01 1.92731041e-02 -6.56430051e-02 -6.76519394e-01 2.22614512e-01 6.07406616e-01 9.27708521e-02 1.18434727e+00 1.07579029e+00 3.03160958e-02 -1.34816930e-01 -9.13675785e-01 7.46295989e-01 -4.35549676e-01 -1.33048165e+00 -2.27642521e-01 3.37422639e-01 1.08656561e+00 4.00802642e-01 3.93752515e-01 9.80517745e-01 1.28397417e+00 -1.16222787e+00 8.81177306e-01 5.77476740e-01 6.18362606e-01 -9.15852904e-01 4.62053753e-02 6.22447491e-01 -1.08421671e+00 -8.71689618e-01 -4.73280221e-01 -2.43669122e-01 -2.98554480e-01 -4.09967124e-01 -3.55939180e-01 2.80483514e-01 5.98812044e-01 7.11679757e-01 -1.06828280e-01 7.10906684e-01 -4.95614082e-01 2.40717322e-01 6.83870465e-02 -2.96746463e-01 7.87834227e-01 -2.73198515e-01 5.67605138e-01 5.39345562e-01 2.98248548e-02 8.58656168e-02 5.58780313e-01 9.76579010e-01 -2.03241318e-01 -2.98929870e-01 -5.57294309e-01 -2.82035291e-01 3.46547961e-01 1.20927310e+00 -3.38242024e-01 -1.30691841e-01 -2.75982097e-02 9.38275516e-01 9.59317446e-01 5.04625261e-01 -7.13294744e-01 1.52379796e-01 9.91017282e-01 -3.48787099e-01 7.18153834e-01 -2.34532371e-01 6.34404048e-02 -1.11516559e+00 -3.54645103e-01 -1.22162497e+00 4.67178643e-01 -4.59486812e-01 -1.05851865e+00 3.02565753e-01 -2.63991207e-01 -1.07473302e+00 -6.32772803e-01 -6.38244271e-01 -4.97567683e-01 3.76013577e-01 -1.55705929e+00 -8.58338237e-01 1.24134593e-01 7.33566642e-01 6.74527824e-01 -6.66511595e-01 6.49573028e-01 -4.84406531e-01 -6.92935824e-01 5.51021218e-01 2.97522813e-01 -2.02681303e-01 4.37029511e-01 -1.48341346e+00 2.34862387e-01 2.97521979e-01 -4.63729613e-02 2.22873017e-01 4.49739128e-01 -4.70424235e-01 -1.73360813e+00 -1.10075760e+00 -8.11005160e-02 -2.57670820e-01 8.15309107e-01 -3.45166951e-01 -4.41796124e-01 6.00753427e-01 2.85074767e-02 3.13897952e-02 7.79510513e-02 -3.48875597e-02 -4.42172676e-01 -1.60365794e-02 -8.57869029e-01 1.03079164e+00 1.17488670e+00 -3.01947564e-01 -6.01190567e-01 1.56745493e-01 7.88195133e-01 -6.08420551e-01 -4.58606184e-01 1.75607607e-01 6.43684208e-01 -8.82291079e-01 1.00415564e+00 -1.02624536e+00 5.33074200e-01 -7.51278996e-02 -1.58149332e-01 -1.44016612e+00 -3.52667332e-01 -1.18984377e+00 -5.22152662e-01 6.13481343e-01 2.50982761e-01 -4.41163272e-01 6.48636341e-01 2.06060126e-01 -2.32963506e-02 -1.10886157e+00 -9.47955072e-01 -9.95519817e-01 3.20096642e-01 -1.70518309e-01 3.70592684e-01 4.33246285e-01 1.47952437e-01 3.94808084e-01 -8.11277807e-01 -2.46392563e-01 5.97522795e-01 3.20501477e-01 8.71859193e-01 -8.69084954e-01 -8.86306286e-01 -7.59294510e-01 3.24055910e-01 -1.43396401e+00 4.00715858e-01 -7.53032029e-01 3.36130589e-01 -1.32958710e+00 8.61019418e-02 -2.92016774e-01 -4.80727196e-01 5.80238700e-01 -9.51686278e-02 -2.65074492e-01 5.62893808e-01 2.84744173e-01 -1.29653943e+00 1.01934302e+00 1.66205549e+00 -3.42821865e-03 -5.53780258e-01 1.17140509e-01 -9.24200714e-01 6.25622272e-01 8.24162841e-01 -2.85094500e-01 -4.85775799e-01 -2.50937581e-01 1.41918838e-01 4.52867925e-01 4.36480731e-01 -9.44123149e-01 7.54788294e-02 -5.79977036e-01 1.72831491e-01 -6.77962080e-02 3.47243935e-01 -4.07288134e-01 -6.55765295e-01 9.78222191e-01 -8.73560429e-01 5.29502258e-02 3.20178568e-02 9.81751144e-01 4.21111584e-01 2.01790303e-01 9.23569679e-01 -4.82844949e-01 -8.49601567e-01 5.49847364e-01 -6.12736404e-01 5.70845187e-01 1.02188909e+00 2.48081207e-01 -3.89376044e-01 -5.23971558e-01 -6.05893135e-01 9.97260511e-01 3.16872478e-01 4.61418748e-01 4.23930943e-01 -1.10243404e+00 -5.06283104e-01 -9.00624767e-02 -4.04397957e-02 -8.39541852e-02 1.67432398e-01 6.35526836e-01 1.27491429e-01 2.31448516e-01 -6.46507919e-01 -1.15300372e-01 -7.06015348e-01 5.75460672e-01 4.05997694e-01 -8.59393775e-01 -6.06086612e-01 8.35567653e-01 3.36492926e-01 -6.44052267e-01 4.93371040e-01 -3.55508834e-01 -3.49980712e-01 -1.46450847e-01 3.72756034e-01 5.76949716e-01 -6.49619341e-01 -2.29493663e-01 -6.25612438e-02 2.89690882e-01 -1.41881421e-01 -4.41564947e-01 1.17750466e+00 1.23128489e-01 4.14624691e-01 2.93719351e-01 6.34695709e-01 -6.84598386e-01 -2.22626996e+00 -3.08705151e-01 1.07326455e-01 -1.17777288e-01 6.18329830e-02 -1.09769166e+00 -8.65149021e-01 6.76014304e-01 4.49854851e-01 -2.12183312e-01 8.84607852e-01 -7.07147643e-02 6.46804333e-01 8.53513837e-01 5.89286387e-01 -1.52469134e+00 7.79293060e-01 8.75853539e-01 1.08104718e+00 -1.33201349e+00 -1.90883830e-01 5.90930998e-01 -1.20293820e+00 7.89732814e-01 1.02714157e+00 -6.10840619e-01 9.68781039e-02 4.56314608e-02 -2.49474183e-01 7.02630877e-02 -1.46215391e+00 -7.05292761e-01 1.67612463e-01 7.03460813e-01 -2.04846889e-01 2.79013868e-02 6.25315681e-02 6.62244678e-01 9.64644402e-02 -6.92500621e-02 1.28457695e-01 9.85403240e-01 -6.27642155e-01 -1.03101003e+00 2.98308581e-02 4.55254465e-01 -2.53208160e-01 2.48983681e-01 -2.60833114e-01 7.41817296e-01 -8.20740759e-02 7.44917154e-01 -8.98747072e-02 -3.71226758e-01 3.74873608e-01 -3.32055688e-01 6.11429930e-01 -5.26069999e-01 -8.07971716e-01 2.22762212e-01 -2.19965190e-01 -1.39946699e+00 -3.92018221e-02 -6.52703464e-01 -1.41578078e+00 1.15051925e-01 2.39578277e-01 -1.15758091e-01 2.77283013e-01 9.62835908e-01 5.74708939e-01 7.36626208e-01 6.85449779e-01 -1.10160160e+00 -1.49047005e+00 -7.25723743e-01 -6.96097434e-01 1.37555942e-01 6.50482476e-01 -9.63500619e-01 -1.24292381e-01 -5.31336486e-01]
[4.134886741638184, 1.633744716644287]
9f5cda5c-1e56-447c-a82f-3a04f11f5753
syllable-discovery-and-cross-lingual
2305.11435
null
https://arxiv.org/abs/2305.11435v1
https://arxiv.org/pdf/2305.11435v1.pdf
Syllable Discovery and Cross-Lingual Generalization in a Visually Grounded, Self-Supervised Speech Mode
In this paper, we show that representations capturing syllabic units emerge when training a self-supervised speech model with a visually-grounded training objective. We demonstrate that a nearly identical model architecture (HuBERT) trained with a masked language modeling loss does not exhibit this same ability, suggesting that the visual grounding objective is responsible for the emergence of this phenomenon. We propose the use of a minimum cut algorithm to automatically predict syllable boundaries in speech, followed by a 2-stage clustering method to group identical syllables together. We show that our model not only outperforms a state-of-the-art syllabic segmentation method on the language it was trained on (English), but also generalizes in a zero-shot fashion to Estonian. Finally, we show that the same model is capable of zero-shot generalization for a word segmentation task on 4 other languages from the Zerospeech Challenge, in some cases beating the previous state-of-the-art.
['David Harwath', 'Abdelrahman Mohamed', 'Okko Räsänen', 'Shang-Wen Li', 'Puyuan Peng']
2023-05-19
null
null
null
null
['visual-grounding']
['computer-vision']
[ 3.59924406e-01 6.11737251e-01 4.08836603e-02 -2.38713816e-01 -7.10155189e-01 -8.26646745e-01 6.77225649e-01 1.70950681e-01 -3.06551099e-01 2.88162857e-01 1.95653960e-01 -6.83946013e-01 4.55956250e-01 -3.93209130e-01 -7.43914127e-01 -5.87870419e-01 -1.79275319e-01 7.96226561e-01 4.35631305e-01 -2.19665334e-01 8.92398730e-02 1.24059670e-01 -1.58984065e+00 5.23962796e-01 7.46112049e-01 4.12816882e-01 4.28938776e-01 8.42085779e-01 -1.91508040e-01 3.44157428e-01 -5.24576306e-01 -2.66947508e-01 1.75139725e-01 -8.51408958e-01 -1.07746148e+00 5.76724350e-01 6.73666418e-01 2.50602037e-01 1.53658520e-02 8.98769498e-01 2.86745667e-01 2.05844223e-01 9.07666445e-01 -7.52746284e-01 -9.26077902e-01 8.25727463e-01 -3.92604113e-01 2.67347664e-01 9.60966945e-02 1.20893098e-01 1.26842868e+00 -1.13173461e+00 8.87049615e-01 1.35211897e+00 5.53118289e-01 8.10820222e-01 -1.52431631e+00 -2.08961219e-01 3.86138648e-01 -1.57705620e-02 -1.40209746e+00 -5.04069865e-01 6.09717190e-01 -6.17602766e-01 1.27481842e+00 1.01115979e-01 6.04668021e-01 1.06289613e+00 -2.39312842e-01 1.02434385e+00 1.16995811e+00 -9.02458072e-01 4.73906785e-01 1.95751116e-01 1.62332445e-01 9.21033204e-01 -8.06762502e-02 1.41375139e-01 -4.72264409e-01 3.93253505e-01 4.25470024e-01 -6.74975693e-01 -1.91204712e-01 -6.20232761e-01 -9.69286919e-01 9.18352246e-01 4.27324414e-01 6.10688984e-01 1.86581118e-03 -1.08798988e-01 3.48916173e-01 1.21043168e-01 6.27488673e-01 3.10632408e-01 -2.59149373e-01 1.80045087e-02 -1.30321801e+00 -3.20672721e-01 7.81790555e-01 8.84810865e-01 7.21882105e-01 6.00457907e-01 -1.51319047e-02 8.01563025e-01 2.41555616e-01 2.94684395e-02 7.63191342e-01 -7.39519298e-01 3.55968803e-01 2.08516568e-01 -2.87960023e-01 -2.33197704e-01 -3.93426627e-01 -4.71422583e-01 -4.11052376e-01 4.87248391e-01 5.00282228e-01 -1.76610842e-01 -1.47196352e+00 1.78319478e+00 -7.08976761e-02 1.43708393e-01 2.93936282e-01 8.77428770e-01 6.10541463e-01 8.07878435e-01 2.34004125e-01 -1.00488804e-01 1.40550113e+00 -1.02420914e+00 -5.12258410e-01 -5.46537638e-01 6.74311459e-01 -5.35744309e-01 1.45995402e+00 5.08924186e-01 -1.11344051e+00 -7.96048760e-01 -1.36839366e+00 -7.48339714e-03 -5.97623050e-01 -2.47489549e-02 5.09808779e-01 1.00374770e+00 -1.31289423e+00 6.37612283e-01 -8.66306067e-01 -8.06052804e-01 2.04211205e-01 -1.79371191e-03 -1.89781502e-01 3.03881675e-01 -8.51498842e-01 9.37302887e-01 6.56477153e-01 -3.03466886e-01 -1.04432571e+00 -3.69079113e-01 -1.02833617e+00 1.97806492e-01 2.06136584e-01 -3.62927228e-01 1.37668455e+00 -1.40909028e+00 -1.48463154e+00 1.39964008e+00 -2.72757351e-01 -6.92173600e-01 2.16609389e-01 5.32479174e-02 -4.27916139e-01 4.19646204e-01 1.66345313e-02 1.06873667e+00 9.54448640e-01 -1.75356281e+00 -5.99836946e-01 -3.07854563e-01 -2.98366755e-01 1.63735911e-01 -1.62063777e-01 -1.04968891e-01 -4.59324479e-01 -7.90412545e-01 3.85638475e-02 -8.63929272e-01 -9.04526114e-02 -3.79945695e-01 -4.87877160e-01 -2.57230759e-01 6.79385245e-01 -7.56037474e-01 1.02658761e+00 -2.28948689e+00 4.23625886e-01 1.20345719e-01 1.07920736e-01 2.48619333e-01 -2.03607440e-01 4.14423972e-01 -2.13570505e-01 4.59959924e-01 -8.67235065e-01 -9.67771709e-01 1.31222293e-01 4.03269738e-01 -3.88370872e-01 4.16724592e-01 5.35415351e-01 8.21252108e-01 -7.44482756e-01 -5.02301037e-01 4.32137437e-02 3.65329802e-01 -4.94079292e-01 -2.43193079e-02 -3.19838703e-01 1.79731280e-01 4.11034554e-01 4.58012730e-01 3.47091854e-01 -3.99524681e-02 4.19136852e-01 3.13315749e-01 -2.73369104e-01 4.04101610e-01 -8.57896209e-01 1.90246129e+00 -3.87672454e-01 9.01809871e-01 1.06506094e-01 -1.09409022e+00 8.32226753e-01 4.58933264e-01 -1.49094909e-01 -6.75054431e-01 -5.94543070e-02 2.69535989e-01 3.22112322e-01 -1.75103351e-01 3.95763814e-01 -3.66624326e-01 -2.98637934e-02 3.58391464e-01 6.24531090e-01 -3.31697971e-01 4.20444548e-01 3.45658749e-01 7.20770836e-01 1.74502030e-01 1.21745802e-01 -5.85210502e-01 7.61325210e-02 2.46018693e-01 2.34680846e-01 8.27594042e-01 -2.65510023e-01 1.00161278e+00 3.94308954e-01 -9.41200629e-02 -1.20419610e+00 -1.45600200e+00 -1.19256467e-01 1.34618175e+00 -2.30486482e-01 -3.60656440e-01 -1.18210936e+00 -6.47113979e-01 -3.22123140e-01 1.09290838e+00 -7.06766367e-01 -1.12402633e-01 -5.13842523e-01 -4.67261195e-01 5.86072922e-01 6.50722504e-01 5.05159199e-02 -1.39944315e+00 -6.00611091e-01 2.68479407e-01 8.30384269e-02 -1.17351294e+00 -4.32358682e-01 7.89265275e-01 -7.34028697e-01 -5.16267478e-01 -9.34222043e-01 -1.43248987e+00 5.67278802e-01 -2.03583017e-02 1.26075816e+00 4.84114140e-02 -3.23468059e-01 5.06879270e-01 -4.06875700e-01 -3.22102666e-01 -7.36780524e-01 2.83537567e-01 -4.73527461e-02 -7.71280304e-02 3.24579686e-01 -2.75392741e-01 -1.68865681e-01 -4.48089540e-02 -7.83201277e-01 7.70929549e-03 3.57703120e-01 6.81333721e-01 3.98694962e-01 -2.75922060e-01 5.59354961e-01 -7.60785878e-01 5.32240152e-01 -2.25817412e-01 -4.71158087e-01 1.96415037e-01 -4.35503930e-01 1.83377817e-01 5.64243674e-01 -3.38973761e-01 -9.43130553e-01 4.25371975e-01 -1.72815785e-01 -2.93994755e-01 -5.52806675e-01 4.08317566e-01 1.53165916e-02 2.96050608e-01 5.48656881e-01 3.29840869e-01 -1.57460079e-01 -6.57417178e-01 8.30528736e-01 5.93112051e-01 7.95101881e-01 -3.54011506e-01 8.01990390e-01 4.59271878e-01 -3.71041626e-01 -1.23451173e+00 -5.88142633e-01 -4.85162854e-01 -8.89478028e-01 9.05248895e-02 1.34826005e+00 -8.28972340e-01 -2.40647852e-01 1.87616035e-01 -1.20633340e+00 -7.06591070e-01 -4.94053632e-01 2.29600340e-01 -6.55807495e-01 4.32082772e-01 -7.54440367e-01 -9.64416683e-01 1.12840254e-02 -8.26680303e-01 9.33267593e-01 7.04343691e-02 -2.80840337e-01 -9.86850560e-01 3.43259156e-01 1.06664941e-01 -1.11309625e-02 -2.49915063e-01 1.10352552e+00 -9.32212770e-01 -2.84760416e-01 3.36608708e-01 -6.15453124e-02 3.37161273e-01 -7.42548183e-02 1.76915199e-01 -1.28169703e+00 -3.29154223e-01 -1.50801688e-01 -3.26684177e-01 1.31789231e+00 3.56220454e-01 6.11669004e-01 -1.41689152e-01 -1.46179050e-01 5.84857821e-01 1.17456329e+00 3.46563727e-01 4.39066142e-01 1.77842215e-01 6.06084704e-01 9.56221282e-01 3.68447378e-02 -1.72726586e-01 3.36321533e-01 6.07986629e-01 3.04542691e-01 -4.91836727e-01 -4.73510355e-01 -4.94903594e-01 3.98697525e-01 1.00584567e+00 1.33503631e-01 -3.35024536e-01 -1.20397413e+00 1.05211306e+00 -1.74257135e+00 -9.00046587e-01 1.49469972e-01 2.07005978e+00 8.31301928e-01 5.46830356e-01 3.85681391e-01 2.33724356e-01 6.94469631e-01 1.46453574e-01 -9.15276110e-02 -9.64418352e-01 -3.81186396e-01 6.10220253e-01 2.19646469e-01 8.80848825e-01 -1.15654314e+00 1.62129605e+00 7.12215233e+00 6.84718549e-01 -1.06896460e+00 1.35145456e-01 6.87928975e-01 3.51039678e-01 -2.87972122e-01 9.90136415e-02 -6.09536946e-01 1.18513659e-01 1.09261096e+00 1.74550474e-01 4.22455072e-01 8.38501215e-01 1.08775057e-01 -3.26925120e-03 -9.60312963e-01 5.54024935e-01 4.66651022e-01 -1.28267467e+00 1.01527506e-02 3.59471478e-02 6.25915706e-01 1.02596246e-01 1.43799618e-01 4.34227616e-01 4.41504627e-01 -1.22732604e+00 1.06145608e+00 -5.59602641e-02 7.11785734e-01 -6.28746986e-01 1.57819893e-02 3.15385550e-01 -1.31925166e+00 1.90678924e-01 -2.58090764e-01 -1.02577947e-01 3.12832624e-01 2.03930177e-02 -1.19039857e+00 2.11814836e-01 5.84815502e-01 2.93983400e-01 -8.74022067e-01 9.15754735e-01 -3.67584318e-01 9.74166155e-01 -6.16145656e-02 1.50996316e-02 6.00299418e-01 -6.16100654e-02 6.16330564e-01 1.83817685e+00 1.19416274e-01 -1.92454651e-01 3.10730934e-01 8.49638104e-01 1.22288592e-01 2.66888678e-01 -6.18034840e-01 -1.74244598e-01 2.70448953e-01 1.02048981e+00 -1.23670375e+00 -5.16742289e-01 -4.80137676e-01 1.34116518e+00 4.83133376e-01 5.71153641e-01 -4.57909197e-01 -4.77223039e-01 3.55873525e-01 3.43194306e-02 9.66997325e-01 -4.88436460e-01 -4.04626459e-01 -9.38475609e-01 -2.01664835e-01 -7.93863535e-01 1.14130385e-01 -9.31644499e-01 -1.07914305e+00 9.23083782e-01 -2.53963232e-01 -9.26318347e-01 -5.43273926e-01 -8.85739386e-01 -8.23660612e-01 6.77566767e-01 -1.28403580e+00 -1.06450939e+00 2.39445508e-01 4.69648868e-01 9.52429116e-01 -3.15623999e-01 9.40062225e-01 -2.27716938e-01 -6.10361338e-01 3.40614945e-01 -1.51603684e-01 3.18263233e-01 4.40391898e-01 -1.72460592e+00 8.62351179e-01 1.26663709e+00 1.04849708e+00 4.77791637e-01 8.75817001e-01 -4.69183743e-01 -7.32264638e-01 -8.01033914e-01 1.03113449e+00 -3.49472523e-01 7.29587376e-01 -8.83227170e-01 -1.14002883e+00 8.93217623e-01 8.82440507e-01 -2.86764741e-01 6.16000593e-01 9.95534882e-02 -5.25619149e-01 4.49831992e-01 -7.70465195e-01 7.94996917e-01 1.04593444e+00 -7.20184863e-01 -1.22207022e+00 1.09885581e-01 1.00351918e+00 -5.89202829e-02 -1.26495242e-01 -3.42546888e-02 1.08445205e-01 -1.04528260e+00 7.41976202e-01 -8.34488511e-01 2.27760956e-01 -2.08227277e-01 -1.99558079e-01 -1.56822526e+00 -3.53433728e-01 -8.14836681e-01 9.32136774e-02 1.26082945e+00 7.98575401e-01 -3.25121254e-01 7.29708731e-01 -1.55227423e-01 -7.65413404e-01 -3.27562988e-01 -1.21680677e+00 -1.03176522e+00 3.97131145e-01 -4.82168525e-01 -4.15178090e-02 8.01658809e-01 2.28433058e-01 7.24029541e-01 1.12369023e-02 7.89198205e-02 4.96586293e-01 -4.37507927e-02 3.36036831e-01 -1.19792747e+00 -4.92473036e-01 -7.03668416e-01 -3.26798260e-01 -1.09774351e+00 3.70445758e-01 -9.86348987e-01 4.32174772e-01 -1.50474453e+00 -6.82043731e-02 1.09730907e-01 -2.13024840e-01 4.96955216e-01 1.40105132e-02 3.48686576e-01 5.66076458e-01 9.61402953e-02 -4.23596144e-01 3.12746555e-01 8.39446247e-01 -1.20700829e-01 -2.74916083e-01 -2.43634492e-01 -6.41120493e-01 8.23494256e-01 7.97662020e-01 -3.93856615e-01 -3.67872894e-01 -3.87681365e-01 -1.41424879e-01 -3.83727729e-01 3.04490864e-01 -1.16520512e+00 9.52742621e-02 2.92580545e-01 1.74985021e-01 -4.48722303e-01 4.47011739e-01 -4.32581872e-01 -5.29334486e-01 5.06293058e-01 -2.42019340e-01 3.36693265e-02 5.27309000e-01 3.05263162e-01 -2.48886734e-01 -4.34642106e-01 8.80959153e-01 -2.46080637e-01 -9.71130073e-01 -2.55631298e-01 -9.57591593e-01 3.48658711e-01 7.09371507e-01 -3.99029404e-01 -2.60096699e-01 -2.81710237e-01 -1.08065450e+00 -1.09843433e-01 4.16681170e-01 5.24822772e-01 4.81788993e-01 -1.00597620e+00 -8.06038797e-01 3.97651136e-01 1.74832880e-01 -5.23650408e-01 -8.79395977e-02 4.51570272e-01 -5.08054912e-01 4.07338053e-01 -8.77370313e-02 -8.46554995e-01 -1.07175314e+00 7.04327106e-01 1.68256104e-01 -8.06406960e-02 -5.81656814e-01 9.99050915e-01 4.99774754e-01 -3.63145381e-01 3.84530008e-01 -4.12728935e-01 -4.10406105e-02 2.79138356e-01 1.80996791e-01 -1.78909302e-01 1.23055302e-01 -8.01626921e-01 -4.77628976e-01 6.39400542e-01 1.56861916e-01 -8.01461637e-01 1.02291107e+00 -3.01400781e-01 5.01360357e-01 8.53242338e-01 9.02413905e-01 1.35612518e-01 -1.46737933e+00 -3.49497423e-04 3.21555555e-01 1.60268933e-01 -1.51888996e-01 -8.13023627e-01 -8.41859460e-01 1.29467142e+00 6.06231570e-01 5.09535789e-01 8.43728781e-01 3.53758335e-01 4.66748625e-01 2.43713066e-01 7.98662379e-03 -1.24261010e+00 1.80981070e-01 7.05663443e-01 8.20290565e-01 -1.29482687e+00 -5.70720196e-01 -2.61644661e-01 -1.20479822e+00 9.79622960e-01 3.12087536e-01 -2.24914506e-01 5.19505739e-01 2.86914676e-01 3.17391008e-01 -1.99786071e-02 -6.62963271e-01 -8.52758884e-01 4.44206119e-01 1.06196034e+00 5.44717610e-01 4.74699587e-02 -8.72123465e-02 4.96046305e-01 -4.62034583e-01 -5.71093023e-01 4.36623514e-01 8.07340682e-01 -8.40794146e-01 -8.84252250e-01 -3.06357443e-01 -3.33045006e-01 -2.22220808e-01 -6.00516319e-01 -8.68015647e-01 1.16725791e+00 1.54594287e-01 1.00670803e+00 4.95658994e-01 -2.37698242e-01 2.06056774e-01 7.51229048e-01 4.27073330e-01 -1.10099399e+00 -6.46804810e-01 4.66150224e-01 6.81853592e-02 -1.52376801e-01 -1.27159268e-01 -6.41218066e-01 -1.44442677e+00 2.44453564e-01 2.59222277e-02 3.42810638e-02 6.17584944e-01 9.30885255e-01 1.24040544e-01 5.52048862e-01 3.00929338e-01 -1.01868176e+00 -2.24155635e-01 -1.00593674e+00 -7.17633605e-01 4.86543536e-01 4.54471171e-01 -4.12386209e-01 -6.15067780e-01 4.81771559e-01]
[14.468788146972656, 6.609226703643799]
e699f86b-a0f6-4260-87c0-559cac01585d
echocardiography-segmentation-using-neural
2306.09687
null
https://arxiv.org/abs/2306.09687v1
https://arxiv.org/pdf/2306.09687v1.pdf
Echocardiography Segmentation Using Neural ODE-based Diffeomorphic Registration Field
Convolutional neural networks (CNNs) have recently proven their excellent ability to segment 2D cardiac ultrasound images. However, the majority of attempts to perform full-sequence segmentation of cardiac ultrasound videos either rely on models trained only on keyframe images or fail to maintain the topology over time. To address these issues, in this work, we consider segmentation of ultrasound video as a registration estimation problem and present a novel method for diffeomorphic image registration using neural ordinary differential equations (Neural ODE). In particular, we consider the registration field vector field between frames as a continuous trajectory ODE. The estimated registration field is then applied to the segmentation mask of the first frame to obtain a segment for the whole cardiac cycle. The proposed method, Echo-ODE, introduces several key improvements compared to the previous state-of-the-art. Firstly, by solving a continuous ODE, the proposed method achieves smoother segmentation, preserving the topology of segmentation maps over the whole sequence (Hausdorff distance: 3.7-4.4). Secondly, it maintains temporal consistency between frames without explicitly optimizing for temporal consistency attributes, achieving temporal consistency in 91% of the videos in the dataset. Lastly, the proposed method is able to maintain the clinical accuracy of the segmentation maps (MAE of the LVEF: 2.7-3.1). The results show that our method surpasses the previous state-of-the-art in multiple aspects, demonstrating the importance of spatial-temporal data processing for the implementation of Neural ODEs in medical imaging applications. These findings open up new research directions for solving echocardiography segmentation tasks.
['Long Tran Quoc', 'Hieu Pham Huy', 'Phi Nguyen Van']
2023-06-16
null
null
null
null
['image-registration']
['computer-vision']
[-8.65696277e-03 1.87207699e-01 1.98910862e-01 -3.09516311e-01 -5.41671693e-01 -5.19334078e-01 1.17859073e-01 1.56370297e-01 -5.08422494e-01 4.47247714e-01 -3.06280792e-01 -2.56952345e-01 -3.03066373e-01 -4.31571275e-01 -6.69445753e-01 -7.09653676e-01 -5.86416245e-01 3.38914841e-01 2.92262137e-01 3.67963687e-02 -1.69742163e-02 7.46642888e-01 -8.71991336e-01 -2.79127479e-01 9.78615224e-01 1.03193259e+00 -6.95883408e-02 8.03284466e-01 1.55758321e-01 3.98569077e-01 -2.28917927e-01 -1.67027682e-01 3.39393824e-01 -7.46973932e-01 -9.73157883e-01 -5.14039304e-04 5.64124703e-01 -5.01722038e-01 -2.80796856e-01 9.78051364e-01 7.06743062e-01 2.15373605e-01 4.34642643e-01 -6.66381657e-01 -3.88517976e-01 2.26912081e-01 -3.00799727e-01 4.12056983e-01 -1.16738714e-01 -1.30639017e-01 4.30947423e-01 -6.50903821e-01 1.03539550e+00 4.69346285e-01 1.17663133e+00 4.42689687e-01 -1.16265547e+00 -3.11926961e-01 -3.27762514e-01 9.32258740e-03 -1.28302348e+00 -1.05351910e-01 7.16490328e-01 -7.26271391e-01 5.53393662e-01 3.21054369e-01 9.18061912e-01 2.58488119e-01 6.38197184e-01 2.83538550e-01 9.67928588e-01 -2.04921573e-01 6.83356971e-02 -2.98312187e-01 1.52380049e-01 9.20168877e-01 5.79796210e-02 -1.26481876e-01 2.14644726e-02 3.19933929e-02 1.22156894e+00 -1.26896366e-01 -5.88594437e-01 -5.23046613e-01 -1.49351501e+00 5.37572801e-01 3.43610525e-01 9.52453136e-01 -5.01609504e-01 1.73768010e-02 5.41512311e-01 2.19784811e-01 5.83442152e-01 4.58183169e-01 -1.01524055e-01 -2.03773662e-01 -1.28880358e+00 6.06465414e-02 7.31594801e-01 6.40409708e-01 4.38661456e-01 4.86659706e-02 -1.29145056e-01 5.47860861e-01 -1.05608413e-02 2.01092646e-01 5.27023375e-01 -1.29380405e+00 -2.78367437e-02 2.79295117e-01 -1.48204207e-01 -1.22962499e+00 -7.77317166e-01 -4.44619805e-01 -1.26789880e+00 -6.69727772e-02 6.59763098e-01 -2.46870250e-01 -8.02540243e-01 1.51712418e+00 5.17929494e-01 6.24437571e-01 -8.61863196e-02 1.02839196e+00 8.03059459e-01 3.50119859e-01 -9.08850878e-02 -5.16566455e-01 1.22737432e+00 -5.86803257e-01 -9.89332080e-01 4.14793730e-01 7.24454582e-01 -6.87962770e-01 3.90129924e-01 1.19378343e-01 -1.52933967e+00 -7.30599523e-01 -1.03230500e+00 3.00797254e-01 1.29154921e-01 1.25080422e-01 3.49666834e-01 4.75288182e-01 -1.42490602e+00 1.19468462e+00 -1.37411571e+00 -1.20229952e-01 4.61117953e-01 5.38509429e-01 -4.85195041e-01 2.20243067e-01 -1.16810691e+00 8.31238091e-01 2.72947758e-01 4.15554166e-01 -5.10173321e-01 -1.22406375e+00 -8.81793261e-01 -2.64167786e-02 -2.02470608e-02 -5.40769160e-01 9.43458200e-01 -7.83383548e-01 -1.58935666e+00 1.04682600e+00 1.05848469e-01 -5.03126621e-01 8.74879837e-01 -4.47794981e-02 -2.00389519e-01 5.90508878e-01 -8.79166871e-02 4.97546792e-01 5.35941303e-01 -9.61273491e-01 -2.15785205e-01 -1.27576634e-01 -1.62537888e-01 -4.89877500e-02 -1.27019197e-01 -8.17102864e-02 -6.43616736e-01 -8.02973807e-01 4.92478609e-01 -1.22004616e+00 -3.21276814e-01 2.47245431e-01 -8.88937637e-02 2.41691142e-01 6.31460547e-01 -1.16601980e+00 1.18200481e+00 -2.16731191e+00 3.49127531e-01 1.92505881e-01 6.04672432e-01 5.53812683e-01 1.59799367e-01 -2.06348047e-01 -2.85418034e-01 4.66943085e-02 -6.56259894e-01 -2.62382448e-01 -4.67501611e-01 4.20825779e-02 2.48514131e-01 1.03128934e+00 -1.41358040e-02 1.10638750e+00 -7.97943711e-01 -6.54867589e-01 3.99772704e-01 7.28136420e-01 -4.76657689e-01 2.89909005e-01 3.52867961e-01 1.09884059e+00 -2.69878626e-01 1.82180256e-01 8.99652362e-01 -1.04126461e-01 4.04849410e-01 -4.46096331e-01 -2.19245300e-01 -3.11304718e-01 -1.06376350e+00 1.96195638e+00 -2.68811345e-01 7.67199039e-01 2.39756942e-01 -1.41246974e+00 7.60745287e-01 7.04491198e-01 1.39040220e+00 -4.61242974e-01 2.93780923e-01 5.03994584e-01 2.23892361e-01 -8.00161421e-01 -3.20265777e-02 -2.11672336e-01 3.36549491e-01 3.99264038e-01 1.79186478e-01 -1.87568590e-01 3.14145237e-01 -1.78627536e-01 7.73536921e-01 1.88387424e-01 8.61669052e-03 -5.59000432e-01 7.62243569e-01 -1.37595937e-01 6.71120465e-01 6.17928624e-01 -4.98104602e-01 1.00019205e+00 5.31532049e-01 -7.73088038e-01 -1.21427858e+00 -8.25877070e-01 -4.94363934e-01 2.95760483e-02 2.89165556e-01 6.16368279e-02 -1.25542259e+00 -6.15030229e-01 -3.57789516e-01 7.34202713e-02 -5.44850528e-01 6.24931157e-02 -1.29075527e+00 -6.45229399e-01 7.18398511e-01 4.73906815e-01 6.31269336e-01 -9.25922394e-01 -8.78736854e-01 4.81983900e-01 -3.81615967e-01 -1.40149283e+00 -6.09544218e-01 -3.70675415e-01 -1.37731504e+00 -1.13409781e+00 -1.19092023e+00 -9.21730816e-01 7.47275710e-01 -2.71122009e-01 1.00134325e+00 3.04435402e-01 -3.68708611e-01 3.43690038e-01 -6.20918348e-02 2.28876069e-01 -7.05750525e-01 -3.19025619e-03 8.30366183e-03 1.07736595e-01 -2.21185550e-01 -6.95255339e-01 -8.27057719e-01 1.99803561e-01 -8.54543030e-01 -1.14960922e-02 1.33258492e-01 7.71509886e-01 6.96423531e-01 -1.67672366e-01 2.54096955e-01 -6.91580832e-01 4.28093731e-01 -2.10630536e-01 -6.03213489e-01 1.46469399e-01 -4.64160174e-01 -1.58955559e-01 5.29922605e-01 -5.18374979e-01 -7.10155785e-01 9.50958878e-02 -2.95330226e-01 -6.41115308e-01 -1.70161635e-01 4.28571194e-01 6.06640339e-01 -4.94093776e-01 3.41929376e-01 2.64612854e-01 5.36536336e-01 -2.66154468e-01 -6.99764341e-02 1.93806157e-01 7.44180381e-01 -3.70609134e-01 4.66432124e-01 7.18666196e-01 4.80631024e-01 -7.92634308e-01 -3.61668319e-01 -4.01143044e-01 -1.19042289e+00 -4.53288585e-01 1.27277684e+00 -3.71336281e-01 -8.08376253e-01 6.04261577e-01 -1.22623086e+00 -2.92375833e-01 -4.50500727e-01 9.18968081e-01 -7.54708946e-01 8.59863579e-01 -8.95681083e-01 -4.34228808e-01 -4.91685092e-01 -1.41239417e+00 7.45956302e-01 1.23659208e-01 -1.46548837e-01 -1.35708046e+00 1.85008511e-01 -4.70224991e-02 5.73951840e-01 8.16706240e-01 7.37130821e-01 -4.73891586e-01 -3.90938342e-01 -1.77485675e-01 -6.97222874e-02 5.78055561e-01 1.89464763e-01 -7.75371939e-02 -5.49747050e-01 -3.34807634e-01 4.58099753e-01 2.49049321e-01 4.41347569e-01 1.13223124e+00 1.16443360e+00 1.13059536e-01 -1.16648845e-01 9.57764983e-01 1.26647782e+00 4.18048412e-01 5.73472381e-01 1.59459077e-02 5.78085303e-01 6.01841033e-01 4.34318513e-01 1.42485410e-01 1.48555055e-01 5.82470000e-01 2.45443955e-01 -6.65272772e-01 -1.94658995e-01 3.83990914e-01 -1.97976589e-01 1.18966877e+00 -5.52196741e-01 3.59726399e-01 -9.12922323e-01 6.07735336e-01 -1.74642646e+00 -6.50178492e-01 -2.94414192e-01 2.22855806e+00 8.02098155e-01 -1.46003917e-01 -1.32085174e-01 -3.19053344e-02 7.58662581e-01 4.07333253e-03 -2.96415508e-01 -3.85599226e-01 1.42660514e-01 4.00650173e-01 3.27357113e-01 5.50655484e-01 -1.40486372e+00 3.93437952e-01 6.28247261e+00 2.92914331e-01 -1.40030241e+00 2.79992133e-01 7.41768539e-01 3.27138603e-01 2.29187131e-01 -1.26314089e-01 -1.73802182e-01 3.49923432e-01 7.97868729e-01 9.86510366e-02 1.23305008e-01 3.04334104e-01 3.74613136e-01 4.95351888e-02 -1.01110911e+00 8.72435212e-01 6.71381056e-02 -1.68515801e+00 -4.54786658e-01 2.04382706e-02 7.81917334e-01 -1.78090140e-01 -1.67792961e-01 -1.89498484e-01 -5.37751615e-01 -1.01144290e+00 3.69136333e-01 8.66578877e-01 1.06326830e+00 -4.79360282e-01 1.04647195e+00 1.10722028e-01 -1.12952912e+00 5.19729555e-01 4.91249338e-02 3.21944177e-01 4.66287434e-01 5.43296099e-01 -6.28543139e-01 7.17735827e-01 6.37855947e-01 9.00615096e-01 -3.38836342e-01 1.11695766e+00 3.72932613e-01 4.36928242e-01 -2.25072786e-01 5.42100430e-01 2.80775100e-01 -6.02830172e-01 7.27649331e-01 1.27797806e+00 4.87720162e-01 2.42961302e-01 9.89764333e-02 9.43007350e-01 1.02743447e-01 1.81758061e-01 -3.62305880e-01 1.01569690e-01 -8.88395384e-02 1.19929790e+00 -1.02161908e+00 -4.26781237e-01 -2.69212127e-01 8.13024938e-01 -1.81200460e-01 3.49102139e-01 -7.98916757e-01 -4.02968556e-01 3.44638646e-01 2.45624274e-01 1.92340434e-01 -3.65495235e-01 -2.29512364e-01 -1.11781180e+00 2.20188007e-01 -6.66479349e-01 3.34074467e-01 -3.83887917e-01 -7.74273872e-01 7.69630253e-01 5.65535501e-02 -1.26140380e+00 -3.48825037e-01 -3.80614489e-01 -5.00914872e-01 7.83107996e-01 -1.31437922e+00 -8.40781748e-01 -3.14401656e-01 2.37207934e-01 4.09045160e-01 1.10554427e-01 7.34808862e-01 5.60143530e-01 -1.65213853e-01 3.07107121e-01 6.43872693e-02 4.25878316e-01 5.07430375e-01 -1.19098377e+00 3.19156170e-01 9.46139216e-01 -3.33690643e-01 7.74417102e-01 4.13940340e-01 -6.14401162e-01 -1.04754090e+00 -7.88485229e-01 8.66480589e-01 -2.79396512e-02 4.05825406e-01 8.33458528e-02 -1.18781960e+00 5.84130883e-01 1.99157402e-01 5.49715519e-01 3.58184695e-01 -6.15196943e-01 3.75704139e-01 -4.07593660e-02 -1.24728858e+00 3.35946977e-01 7.02391803e-01 -1.39733523e-01 -2.73451775e-01 1.53467253e-01 4.87317264e-01 -1.04556262e+00 -1.55080616e+00 7.66262770e-01 6.00257993e-01 -9.22097147e-01 1.03656936e+00 -4.34719026e-01 4.54815894e-01 -1.92966312e-01 4.30492967e-01 -1.01391184e+00 -1.11687511e-01 -9.30935502e-01 -1.75817132e-01 7.80050755e-01 -6.90550879e-02 -6.38416588e-01 4.94501531e-01 5.87281764e-01 -3.02826583e-01 -9.71468449e-01 -1.30754268e+00 -4.31619942e-01 2.71540284e-01 -3.29990059e-01 2.25606412e-01 1.27754343e+00 -2.51461029e-01 -4.37406361e-01 -1.52577162e-01 4.90490943e-02 7.29386628e-01 -4.38240394e-02 1.71676368e-01 -1.24137914e+00 -1.60962325e-02 -4.94832605e-01 -5.85579574e-01 -8.31494987e-01 1.98457181e-01 -9.14271116e-01 6.01389818e-02 -1.33334208e+00 -8.59589204e-02 -6.17081404e-01 -2.20414370e-01 1.11424848e-01 2.90941466e-02 5.56182086e-01 1.06939465e-01 3.04934770e-01 -2.04266474e-01 2.00435147e-02 1.81483614e+00 1.12952597e-01 -3.51150334e-01 3.32293212e-02 -3.21561564e-03 6.66243732e-01 7.21581161e-01 -3.21786702e-01 -4.07089218e-02 -1.83940500e-01 -2.54070401e-01 4.66136724e-01 3.90379816e-01 -1.10349274e+00 2.87825167e-01 2.77991712e-01 1.74669534e-01 -3.82678151e-01 -4.74320678e-03 -7.93428957e-01 3.36360693e-01 8.26843917e-01 -2.73542553e-01 1.63571030e-01 2.35730037e-01 4.13028561e-02 -3.29920352e-01 -3.56828570e-01 1.12024534e+00 -1.62232697e-01 -3.03641081e-01 4.35571581e-01 -4.58908141e-01 2.48254880e-01 9.52510417e-01 -3.11755389e-01 3.93927574e-01 -1.60325021e-01 -1.19139969e+00 -1.11090630e-01 1.09449431e-01 1.44319952e-01 5.89837134e-01 -1.17037392e+00 -7.97232211e-01 4.23136741e-01 -5.37735701e-01 1.84966430e-01 5.53577363e-01 1.71532714e+00 -1.26591945e+00 3.63645107e-01 -2.58260429e-01 -1.16123950e+00 -1.17710865e+00 2.25028560e-01 1.06234443e+00 -3.74114871e-01 -1.09041917e+00 4.73366171e-01 -4.37733866e-02 -4.11554247e-01 1.20675765e-01 -6.03622258e-01 -2.45429114e-01 -1.58332333e-01 1.92828476e-01 3.57783884e-01 2.70638674e-01 -8.89519751e-01 -2.99815685e-01 1.07951665e+00 2.33770743e-01 7.10639954e-02 1.42612553e+00 -1.35830075e-01 -4.90233511e-01 1.50505453e-01 1.38629401e+00 -2.73663223e-01 -1.32792985e+00 -4.06807885e-02 -1.71578169e-01 -2.60455728e-01 1.47893652e-01 -3.46490949e-01 -1.49499190e+00 8.13195586e-01 9.70245183e-01 2.74156660e-01 1.13172615e+00 -3.17526639e-01 1.06843936e+00 -1.60270587e-01 -1.33882582e-01 -8.89737070e-01 -1.89358562e-01 3.70248467e-01 6.54295027e-01 -9.71155107e-01 -6.23327456e-02 -5.07147551e-01 -4.77238417e-01 1.43575203e+00 2.02862337e-01 -2.83476263e-01 8.48502815e-01 3.55331182e-01 2.96428561e-01 -1.81162849e-01 1.36764258e-01 2.05649242e-01 6.32961094e-01 5.18810749e-01 7.25405931e-01 -2.61172086e-01 -7.63836443e-01 9.96578410e-02 1.73405290e-01 2.42470279e-01 4.99736816e-01 7.60581970e-01 1.09028868e-01 -8.34978700e-01 -2.16680139e-01 1.59163341e-01 -8.95101368e-01 9.43121910e-02 2.78165966e-01 9.13751066e-01 1.59601346e-01 4.47117060e-01 1.81761369e-01 1.20442018e-01 3.55679721e-01 2.16376800e-02 5.58749437e-01 -1.75540835e-01 -7.65335977e-01 3.18965048e-01 -3.19057375e-01 -6.61930025e-01 -7.81343699e-01 -7.19006419e-01 -1.47546899e+00 -4.92839739e-02 -9.86156613e-02 1.46939591e-01 7.46070385e-01 9.96228278e-01 3.32449466e-01 7.94087052e-01 4.92525071e-01 -9.77385521e-01 -3.26277167e-01 -6.70224965e-01 -5.17009795e-01 5.75351954e-01 5.15605748e-01 -4.12880063e-01 -1.66623324e-01 4.29653674e-01]
[13.998424530029297, -2.5073139667510986]
d11bc8aa-8b88-46c9-ad5b-fcbd89287778
lapscore-language-guided-person-search-via
null
null
http://openaccess.thecvf.com//content/ICCV2021/html/Wu_LapsCore_Language-Guided_Person_Search_via_Color_Reasoning_ICCV_2021_paper.html
http://openaccess.thecvf.com//content/ICCV2021/papers/Wu_LapsCore_Language-Guided_Person_Search_via_Color_Reasoning_ICCV_2021_paper.pdf
LapsCore: Language-Guided Person Search via Color Reasoning
The key point of language-guided person search is to construct the cross-modal association between visual and textual input. Existing methods focus on designing multimodal attention mechanisms and novel cross-modal loss functions to learn such association implicitly. We propose a representation learning method for language-guided person search based on color reasoning (LapsCore). It can explicitly build a fine-grained cross-modal association bidirectionally. Specifically, a pair of dual sub-tasks, image colorization and text completion, is designed. In the former task, rich text information is learned to colorize gray images, and the latter one requests the model to understand the image and complete color word vacancies in the captions. The two sub-tasks enable models to learn correct alignments between text phrases and image regions, so that rich multimodal representations can be learned. Extensive experiments on multiple datasets demonstrate the effectiveness and superiority of the proposed method.
['Shuguang Cui', 'Changqing Zou', 'Guanbin Li', 'Xiaoguang Han', 'Zizheng Yan', 'Yushuang Wu']
2021-01-01
null
null
null
iccv-2021-1
['person-search']
['computer-vision']
[ 4.30597961e-02 -1.63697556e-01 -2.36261383e-01 -5.13865709e-01 -9.17320549e-01 -5.53644419e-01 7.43415236e-01 -2.45598584e-01 -4.82877553e-01 4.29082721e-01 3.96155685e-01 -2.59270728e-01 2.18990013e-01 -5.32712221e-01 -6.87861562e-01 -5.54007351e-01 4.32773083e-01 5.42814076e-01 -4.00114685e-01 4.68303896e-02 2.53320128e-01 1.29235521e-01 -1.30532479e+00 5.89669108e-01 1.00659847e+00 8.87946486e-01 2.53139853e-01 6.44933879e-01 -5.98857880e-01 7.58697510e-01 -2.34326094e-01 -7.10839033e-01 -2.59541571e-02 -4.32439506e-01 -9.51312900e-01 2.68803984e-01 5.55947304e-01 -2.68931150e-01 -4.77506340e-01 9.93447542e-01 4.34600085e-01 3.02840561e-01 8.13262463e-01 -1.22912371e+00 -1.44255245e+00 4.31824893e-01 -9.69425082e-01 -1.76215038e-01 6.97646499e-01 4.31612790e-01 1.12568903e+00 -1.09283626e+00 2.99503148e-01 1.71154821e+00 1.34667322e-01 9.22141016e-01 -1.25780833e+00 -6.56066716e-01 7.74186492e-01 3.06188315e-01 -1.27252638e+00 -8.95186961e-02 9.54399228e-01 -4.93901819e-01 4.86161351e-01 4.32669938e-01 8.23471308e-01 1.22875547e+00 -4.09411162e-01 1.24245381e+00 1.33719659e+00 -5.72824597e-01 -1.70219541e-01 3.95133108e-01 5.07290624e-02 1.25942111e+00 -1.19831972e-01 1.51419774e-01 -6.74495935e-01 -1.18525982e-01 9.67235446e-01 2.73726195e-01 -1.87838107e-01 -3.22129637e-01 -1.31113374e+00 6.93099201e-01 6.66467071e-01 3.52121815e-02 -1.62383080e-01 5.16684890e-01 3.35572362e-01 -1.68862388e-01 1.81698158e-01 1.88232198e-01 -5.10973204e-03 1.62600622e-01 -6.09696090e-01 1.86100066e-01 1.94440335e-01 1.02947760e+00 8.34634602e-01 -3.25805068e-01 -8.44452977e-01 1.06232524e+00 6.47389174e-01 8.22623670e-01 2.96899676e-01 -6.94119215e-01 8.95963013e-01 8.77254486e-01 1.77545473e-01 -9.65457022e-01 -8.52403324e-03 -6.23760484e-02 -9.20621276e-01 -2.70398264e-03 3.72699916e-01 8.62291735e-03 -1.29196262e+00 1.90882945e+00 8.97290409e-02 -1.62731782e-01 -6.29677549e-02 1.30445230e+00 8.96203041e-01 6.76499426e-01 5.36431670e-01 3.31846714e-01 1.77079368e+00 -1.35162032e+00 -6.39028311e-01 -5.55785954e-01 -1.36881880e-02 -7.09975958e-01 1.65563059e+00 8.66651684e-02 -1.26434839e+00 -6.73668027e-01 -7.19471753e-01 -5.40994585e-01 -5.23288250e-01 5.88665366e-01 7.58288622e-01 4.80960369e-01 -9.24820006e-01 -1.83960259e-01 -2.70695478e-01 -2.59909928e-01 4.01992887e-01 6.23845905e-02 -1.51500791e-01 -4.21190828e-01 -1.21757770e+00 5.33509195e-01 2.15206727e-01 3.18944871e-01 -8.51641059e-01 -3.73316377e-01 -7.16915011e-01 1.20837912e-01 2.13974506e-01 -1.03943574e+00 8.37893069e-01 -1.30772448e+00 -1.25179458e+00 1.35363030e+00 -2.76246399e-01 1.30674466e-01 6.02569342e-01 -3.35740924e-01 -1.54254407e-01 2.70262092e-01 1.47921622e-01 1.07230544e+00 1.12996709e+00 -1.75087500e+00 -7.26705015e-01 -4.47819948e-01 1.79789558e-01 6.41866446e-01 -4.28364068e-01 -8.81482214e-02 -1.50032079e+00 -8.66970956e-01 -2.43591517e-01 -7.84300923e-01 -1.43565342e-01 2.41228819e-01 -8.06361318e-01 -4.48199123e-01 3.08201760e-01 -8.93522143e-01 9.38812256e-01 -1.87076545e+00 5.48692226e-01 4.75889683e-01 1.33939028e-01 -2.50168979e-01 -4.70581472e-01 2.03167602e-01 -2.17551775e-02 1.61267683e-01 4.54059988e-02 -5.27163088e-01 3.04420382e-01 4.61279750e-02 -4.03757364e-01 1.05681740e-01 1.51454628e-01 1.36458468e+00 -8.63536417e-01 -6.76851690e-01 5.98550476e-02 4.90099460e-01 -2.74944276e-01 5.16359568e-01 -2.99780369e-01 3.34459990e-01 -5.15406132e-01 8.03028822e-01 6.02719128e-01 -4.59604651e-01 4.92309183e-02 -4.90158737e-01 1.53079659e-01 -2.34428197e-01 -7.71469712e-01 2.08118725e+00 -4.82963353e-01 4.47294384e-01 -9.36750919e-02 -6.50775731e-01 6.96684778e-01 -1.00725204e-01 1.10601569e-02 -1.20352554e+00 -9.60285440e-02 -3.50576848e-01 -7.62116313e-01 -7.71731675e-01 5.29080808e-01 1.63309246e-01 -2.56441295e-01 7.62393117e-01 -3.03162396e-01 2.88044959e-01 2.58201594e-03 4.45700049e-01 2.45861620e-01 3.38818818e-01 -3.24658543e-01 -2.32882309e-03 8.36024463e-01 -1.67779222e-01 5.95696941e-02 7.05078959e-01 1.28547132e-01 4.46592063e-01 3.41325074e-01 -4.13120627e-01 -9.54672456e-01 -1.07468581e+00 5.45606434e-01 1.64320910e+00 6.79262578e-01 -2.25977555e-01 -7.17021942e-01 -8.36153269e-01 -1.89110823e-02 5.03426790e-01 -1.00467873e+00 -6.77600503e-02 -4.23575819e-01 -5.55587113e-01 4.45580184e-01 6.76195800e-01 6.36280596e-01 -1.12062562e+00 -3.23636293e-01 -5.25835276e-01 -5.67987740e-01 -8.40924144e-01 -1.23580241e+00 -3.02401811e-01 -3.69690657e-01 -1.14176261e+00 -1.15019095e+00 -1.24570632e+00 1.26116240e+00 3.93541962e-01 1.20864856e+00 5.17598212e-01 -6.52454853e-01 9.73612130e-01 -2.89269567e-01 1.85954317e-01 2.84496218e-01 -2.40864322e-01 -4.56977218e-01 3.32725048e-01 3.71779472e-01 1.16768412e-01 -1.02093613e+00 1.29596606e-01 -7.47098804e-01 5.09638309e-01 9.49880123e-01 1.04977298e+00 6.17487669e-01 -4.35210019e-01 1.52292131e-02 -6.21717155e-01 1.01971364e+00 -1.54356271e-01 -3.72311711e-01 1.03562903e+00 -3.96104068e-01 3.98733288e-01 1.74786255e-01 -4.36301321e-01 -1.39878285e+00 1.25361517e-01 2.25457817e-01 -4.87726778e-01 -8.36464167e-02 3.06308568e-01 -2.01931730e-01 1.65370077e-01 1.16479449e-01 5.73248625e-01 -2.14990214e-01 -3.43205929e-01 1.03926873e+00 3.87717068e-01 7.27045774e-01 -1.15892673e+00 7.80962527e-01 5.22763431e-01 -5.18013895e-01 -4.58024502e-01 -6.90237045e-01 -4.22444940e-01 -4.78903592e-01 -2.89542973e-01 1.41049206e+00 -9.86266077e-01 -1.07982481e+00 2.17645898e-01 -1.25191092e+00 -3.76267403e-01 9.65058059e-02 1.25142395e-01 -4.48861957e-01 5.09239376e-01 -4.31404829e-01 -1.02567899e+00 -4.69015688e-01 -9.80956852e-01 1.33912122e+00 5.53699732e-01 3.27999830e-01 -9.45528626e-01 -6.56283051e-02 7.23131537e-01 -9.13813338e-03 -1.26146197e-01 1.40160012e+00 -1.09538302e-01 -8.49411190e-01 -2.90663505e-04 -9.14153636e-01 -1.51517943e-01 -1.34444907e-01 -1.99517250e-01 -8.21939409e-01 -2.69919455e-01 -7.09166825e-01 -7.75134981e-01 1.12954056e+00 4.84854542e-02 1.61383796e+00 -2.96289444e-01 -3.07119787e-01 6.48733675e-01 1.35118687e+00 2.80593522e-03 7.51760781e-01 1.08015746e-01 9.54318047e-01 7.34164357e-01 5.88217974e-01 2.56256998e-01 7.65951693e-01 6.38232887e-01 2.63421535e-01 -6.04055285e-01 -2.21420184e-01 -7.04607785e-01 1.46464571e-01 3.20557207e-01 -1.57681987e-01 -7.15659112e-02 -7.02457726e-01 3.52396309e-01 -2.03249574e+00 -1.00676465e+00 2.94329971e-01 1.95900071e+00 9.75787282e-01 -3.87326777e-01 1.78089619e-01 -5.81824243e-01 7.53371596e-01 4.73265611e-02 -6.57812476e-01 -1.06555760e-01 -6.71145767e-02 -1.44861981e-01 6.10566959e-02 4.70864475e-01 -1.05335796e+00 1.14883912e+00 5.61395979e+00 7.78950334e-01 -8.12051833e-01 -7.66633153e-02 9.81907666e-01 1.02162361e-02 -7.88487196e-01 -1.59563601e-01 -5.61888218e-01 3.42133582e-01 -1.51588246e-01 2.53830612e-01 7.38531590e-01 5.19589186e-01 -2.25511584e-02 -4.50294875e-02 -1.13085604e+00 1.67302799e+00 4.72299367e-01 -1.17091572e+00 6.55159593e-01 -2.06531718e-01 6.49924219e-01 -7.58706033e-01 5.89662433e-01 3.69018435e-01 3.82290334e-01 -1.33727336e+00 8.01772296e-01 1.15875614e+00 1.24684215e+00 -8.16745102e-01 1.39252961e-01 -4.79386970e-02 -1.57464635e+00 -2.44142711e-01 -1.65274739e-01 5.12013853e-01 1.44414365e-01 -9.69968736e-02 -1.96403086e-01 2.49238357e-01 7.57059038e-01 5.02613425e-01 -1.01161635e+00 8.12130272e-01 -3.37581784e-01 -7.06702890e-03 2.39605010e-01 -2.60364145e-01 2.54110038e-01 -4.79205012e-01 4.94957268e-02 1.38072050e+00 -1.22114727e-02 2.59184893e-02 2.88346052e-01 1.36869848e+00 -2.27650464e-01 1.42240435e-01 -1.52203694e-01 -2.58352220e-01 2.65298992e-01 1.30954921e+00 -3.75762135e-01 -4.17429358e-01 -5.79665780e-01 1.58095801e+00 7.74138391e-01 1.09561276e+00 -9.69119489e-01 -2.55524457e-01 4.51272011e-01 -2.44745612e-01 7.85300434e-02 -5.36174141e-02 -3.95592541e-01 -1.20962286e+00 4.06887718e-02 -9.11055088e-01 6.69701219e-01 -1.18071747e+00 -1.65158629e+00 4.46644187e-01 -2.52717286e-01 -7.72415757e-01 1.01626784e-01 -6.02512062e-01 -6.50596082e-01 1.20932555e+00 -1.54531240e+00 -1.88868833e+00 -6.42170668e-01 1.13237202e+00 6.96760774e-01 -2.50137568e-01 6.84942544e-01 3.11106712e-01 -5.81340313e-01 9.31210518e-01 -3.76971483e-01 4.08399701e-01 6.79620385e-01 -1.34826803e+00 6.86084703e-02 7.00608909e-01 2.60157943e-01 8.55365217e-01 1.81235641e-01 -7.19929874e-01 -1.60825872e+00 -9.65759277e-01 4.47572142e-01 -2.97355443e-01 3.70450079e-01 -4.93400604e-01 -5.80464005e-01 4.35295910e-01 5.26151061e-01 -3.69446874e-01 5.77079654e-01 2.19614252e-01 -8.30501854e-01 -1.02388456e-01 -5.35672307e-01 8.55673194e-01 8.78541946e-01 -9.45533931e-01 -4.97199059e-01 3.99891973e-01 5.06030500e-01 -2.08935305e-01 -2.44106755e-01 -1.54657543e-01 7.10994184e-01 -6.28824055e-01 1.39054799e+00 -9.40679669e-01 4.03135121e-01 -2.81285554e-01 2.13746424e-03 -9.70602572e-01 -2.79514611e-01 -3.55273634e-01 1.48220778e-01 1.26461542e+00 5.29240549e-01 -1.63307413e-01 6.95733547e-01 9.73188043e-01 4.01563257e-01 -6.58034205e-01 -4.93560702e-01 -1.70247436e-01 -9.30276886e-03 -4.99500632e-02 5.15051067e-01 8.42693746e-01 3.67052741e-02 4.77243394e-01 -7.69572914e-01 1.25469118e-01 7.20321476e-01 5.30452549e-01 7.77726233e-01 -7.94368029e-01 -2.26517230e-01 -7.04407394e-01 2.94163257e-01 -1.51010966e+00 1.84533641e-01 -8.19496512e-01 6.83798715e-02 -1.76337874e+00 9.15071487e-01 -5.42547166e-01 -3.71721953e-01 5.74169338e-01 -6.63719356e-01 1.73717380e-01 2.47157201e-01 1.47886127e-01 -1.04564667e+00 7.52449751e-01 1.31819069e+00 -6.51729465e-01 -7.80477747e-02 -3.35202336e-01 -8.71010363e-01 4.81693029e-01 4.70311970e-01 2.20177069e-01 -5.68392456e-01 -8.37629616e-01 4.23096001e-01 3.58255319e-02 8.49842072e-01 -4.78860319e-01 3.91609579e-01 -4.46071386e-01 9.59543765e-01 -7.47124851e-01 5.66227257e-01 -7.59494662e-01 -3.34871083e-01 2.86133587e-01 -9.01990712e-01 3.30928922e-01 -1.13971628e-01 9.06599104e-01 -1.45914257e-01 1.99349061e-01 4.67385054e-01 -3.66228610e-01 -9.42937195e-01 5.49175560e-01 7.86978528e-02 1.27026320e-01 9.12417829e-01 1.72339641e-02 -3.71521503e-01 -5.01239061e-01 -7.90635228e-01 6.70636177e-01 2.65644521e-01 7.24258006e-01 1.18735170e+00 -1.77376246e+00 -5.20720541e-01 1.61133945e-01 3.82632405e-01 -3.35971534e-01 4.74957913e-01 5.80202520e-01 -3.91213834e-01 4.61299717e-01 -1.10017963e-01 -4.35442984e-01 -1.31910598e+00 8.44873726e-01 4.49696094e-01 -9.61470753e-02 -4.18441981e-01 1.18232441e+00 6.65006399e-01 -2.64190346e-01 6.87490225e-01 7.30134845e-02 -3.31989944e-01 -7.31004253e-02 6.05062127e-01 -1.31231453e-02 -7.49239743e-01 -5.87104023e-01 -1.96813136e-01 9.05696154e-01 -2.19806060e-01 -4.86133307e-01 8.17808032e-01 -5.66030681e-01 -3.14685106e-01 2.03488171e-01 1.05760241e+00 -1.61801323e-01 -1.29569232e+00 -2.96931863e-01 -2.31571972e-01 -6.12630069e-01 -3.19398999e-01 -1.10909629e+00 -1.17149031e+00 1.18548656e+00 8.65720570e-01 -2.91798025e-01 1.25399017e+00 2.37951532e-01 6.30263448e-01 4.17139947e-01 -1.14321224e-01 -1.31923532e+00 8.12966764e-01 2.79038370e-01 1.04052901e+00 -1.27664423e+00 -1.02456711e-01 -1.95047423e-01 -1.07795203e+00 9.91928995e-01 1.03499126e+00 1.92361131e-01 9.52538773e-02 -4.66238797e-01 6.93203658e-02 -2.98670739e-01 -6.12327099e-01 -4.70117211e-01 9.24969614e-01 6.98107839e-01 3.33144844e-01 6.39618933e-02 -7.12403134e-02 6.34625196e-01 3.13902676e-01 -5.38903296e-01 -3.81886631e-01 5.82625568e-01 -2.84303933e-01 -1.02350175e+00 -5.08009315e-01 2.15965018e-01 8.25499371e-02 -3.79987448e-01 -8.39712858e-01 5.24774909e-01 1.02205582e-01 8.55306685e-01 7.46918544e-02 -3.03072870e-01 1.33348584e-01 1.13003202e-01 5.07857442e-01 -1.48988694e-01 -2.79441655e-01 6.02201782e-02 -8.60766843e-02 -6.34947062e-01 -2.74332166e-01 -2.28520423e-01 -1.12321126e+00 -1.64092965e-02 1.56012937e-01 2.08853647e-01 5.35256267e-01 7.66778886e-01 1.86315238e-01 3.55538994e-01 4.29303855e-01 -7.10548997e-01 -5.49776293e-02 -5.23374319e-01 -3.64943713e-01 8.68532419e-01 3.14534158e-01 -5.96918583e-01 7.78757930e-02 3.22785646e-01]
[10.879855155944824, 1.4269344806671143]
d86d5687-db24-42e4-b729-3b63ab672f31
end-to-end-offline-speech-translation-system
null
null
https://aclanthology.org/2020.iwslt-1.7
https://aclanthology.org/2020.iwslt-1.7.pdf
End-to-End Offline Speech Translation System for IWSLT 2020 using Modality Agnostic Meta-Learning
In this paper, we describe the system submitted to the IWSLT 2020 Offline Speech Translation Task. We adopt the Transformer architecture coupled with the meta-learning approach to build our end-to-end Speech-to-Text Translation (ST) system. Our meta-learning approach tackles the data scarcity of the ST task by leveraging the data available from Automatic Speech Recognition (ASR) and Machine Translation (MT) tasks. The meta-learning approach combined with synthetic data augmentation techniques improves the model performance significantly and achieves BLEU scores of 24.58, 27.51, and 27.61 on IWSLT test 2015, MuST-C test, and Europarl-ST test sets respectively.
['Sangha Kim', 'Sathish Reddy Indurthi', 'Mohd Abbas Zaidi', 'Nikhil Kumar Lakumarapu', 'Hou Jeung Han', 'Beomseok Lee']
2020-07-01
null
null
null
ws-2020-7
['speech-to-text-translation']
['natural-language-processing']
[ 3.51893216e-01 2.84681439e-01 -2.37827122e-01 -3.40927631e-01 -1.77986455e+00 -5.44985950e-01 9.94211376e-01 -1.48768902e-01 -4.08185065e-01 9.64308739e-01 4.40895915e-01 -9.85819161e-01 5.47187746e-01 -3.85363307e-03 -7.74640203e-01 -2.32562676e-01 4.87864673e-01 9.65202212e-01 -1.14101827e-01 -5.39938807e-01 -2.62829334e-01 -2.37948537e-01 -7.51510918e-01 9.30074394e-01 1.21335077e+00 7.60935903e-01 2.59961218e-01 8.33154678e-01 -2.18714282e-01 6.32650912e-01 -6.93250179e-01 -8.36444557e-01 2.84910560e-01 -5.26703417e-01 -9.30584192e-01 -1.47865012e-01 3.11152399e-01 3.36968601e-02 -4.44741040e-01 5.24324894e-01 7.38313437e-01 -9.43806171e-02 3.93431216e-01 -9.85333383e-01 -7.18732119e-01 9.80517149e-01 -1.78285941e-01 1.15871131e-01 2.79339343e-01 1.17407575e-01 9.05115008e-01 -1.42660964e+00 5.87910593e-01 1.10649312e+00 4.23448950e-01 7.96077073e-01 -9.30987835e-01 -4.63069767e-01 -1.99022800e-01 2.84449190e-01 -1.12217748e+00 -1.05175567e+00 3.61783177e-01 -1.07371069e-01 1.50883591e+00 3.77995640e-01 3.77875865e-01 1.79420578e+00 -1.94376022e-01 1.17642927e+00 1.27679658e+00 -8.41851413e-01 2.08471566e-01 1.32009536e-01 -5.54980397e-01 3.74532461e-01 -3.77298981e-01 5.87379634e-02 -8.88893008e-01 -1.55413104e-02 6.82666525e-02 -6.86944962e-01 1.71115294e-01 2.97715336e-01 -1.63144195e+00 4.34838742e-01 -2.31605738e-01 2.44801849e-01 -4.89176184e-01 -1.85608268e-01 7.24406540e-01 7.23219335e-01 1.04166913e+00 3.20633501e-01 -1.05997860e+00 -6.39094293e-01 -1.07987583e+00 -2.51852959e-01 7.15589464e-01 1.29993486e+00 2.31970415e-01 5.94393909e-01 -4.62116599e-01 1.34220517e+00 2.50179023e-01 1.02920783e+00 7.15079427e-01 -7.44103730e-01 1.38830531e+00 2.21120566e-02 7.90598467e-02 -4.73685153e-02 2.97655433e-01 -6.78378999e-01 -6.06096864e-01 -6.84487522e-01 -4.22291877e-03 -3.27502936e-01 -1.17277229e+00 1.59843755e+00 -1.28181893e-02 8.90730619e-02 5.82501769e-01 4.25749272e-01 8.70445490e-01 1.13123977e+00 -5.78350760e-02 -4.71174985e-01 8.68711889e-01 -1.60522604e+00 -9.10187900e-01 -5.27509451e-01 1.00331819e+00 -1.15769243e+00 1.21526742e+00 1.47972375e-01 -1.33521044e+00 -5.67813158e-01 -8.41706395e-01 1.29675478e-01 -1.40267581e-01 5.50306439e-01 -4.05948464e-04 5.55209756e-01 -1.22397041e+00 4.81986552e-02 -7.97122359e-01 -3.66200894e-01 1.29335567e-01 1.14836343e-01 -3.06794256e-01 -2.04983465e-02 -1.25839436e+00 1.08722341e+00 3.41141373e-01 -2.73127928e-02 -1.05249846e+00 -5.08680761e-01 -5.67673683e-01 -2.87000358e-01 1.95585519e-01 -5.51471412e-01 1.79346585e+00 -1.05480266e+00 -1.90749145e+00 8.37402344e-01 -4.00727689e-01 -8.37049782e-01 6.55875504e-01 -1.77755639e-01 -7.23872125e-01 -7.09411129e-02 -8.47164094e-02 6.09734356e-01 6.73460066e-01 -8.98341119e-01 -6.11014843e-01 -9.44208503e-02 -5.71828008e-01 3.22160125e-01 -1.61916137e-01 4.30324435e-01 -3.34161192e-01 -6.51335418e-01 -8.86026844e-02 -1.03678310e+00 -3.94219644e-02 -9.70605493e-01 -4.73796427e-01 -7.78765082e-02 7.15889513e-01 -1.35494947e+00 1.16689217e+00 -1.87510622e+00 3.52780342e-01 -3.61633599e-01 -5.06658018e-01 7.39028513e-01 -5.74104667e-01 8.05967093e-01 -2.19841339e-02 7.74367005e-02 -1.82645589e-01 -9.33792770e-01 -1.44162551e-01 2.27835611e-01 -4.50955749e-01 -8.76755416e-02 9.41529274e-02 1.24386704e+00 -7.78821766e-01 -3.61644298e-01 2.70297587e-01 9.01689529e-02 -9.53520387e-02 3.55165303e-01 -3.36844891e-01 6.43531799e-01 -1.32603288e-01 8.44831705e-01 2.60747999e-01 2.29660824e-01 1.57056063e-01 1.91487476e-01 -1.80399626e-01 1.07918584e+00 -1.34148812e-02 2.01777077e+00 -8.94350052e-01 5.90765238e-01 -1.05972685e-01 -8.32580328e-01 9.52768385e-01 9.76661921e-01 2.04028174e-01 -1.11025620e+00 -1.24145977e-01 8.21253657e-01 8.62390697e-02 -3.78152043e-01 4.80821729e-01 -2.64613152e-01 -1.20186247e-01 4.27387863e-01 4.09797609e-01 -2.31693849e-01 -6.90870807e-02 7.90243000e-02 9.92902160e-01 1.38778865e-01 1.63485538e-02 2.38344111e-02 4.88060623e-01 2.13809714e-01 4.53260750e-01 3.89460385e-01 -9.43908319e-02 5.73414207e-01 -1.92757964e-01 -3.27507943e-01 -1.55710530e+00 -9.00973082e-01 4.01853472e-01 9.93088186e-01 -7.58362353e-01 -4.71206784e-01 -9.04661000e-01 -8.58997524e-01 -3.91452909e-01 1.13437390e+00 -1.86423674e-01 -2.87413180e-01 -7.02115774e-01 -5.94684243e-01 9.19055045e-01 2.08228797e-01 5.22306383e-01 -8.87362778e-01 4.19079751e-01 4.52059805e-01 -9.48675931e-01 -1.61288369e+00 -6.89174294e-01 1.41354948e-02 -8.06684494e-01 -2.20111743e-01 -8.79860222e-01 -7.86613047e-01 1.19949207e-01 2.75807023e-01 9.95394647e-01 -4.17957366e-01 4.23052758e-01 2.32124049e-02 -7.21511960e-01 -2.55960554e-01 -1.28538096e+00 5.98427534e-01 3.11417222e-01 2.70772036e-02 9.80697223e-04 -3.72275770e-01 -8.19520950e-02 3.27357352e-01 -3.13456655e-01 5.75377285e-01 8.27127635e-01 9.26061332e-01 5.13711810e-01 -7.97360957e-01 8.86348546e-01 -3.30437958e-01 4.16472286e-01 -3.78429741e-01 -3.07198673e-01 7.36414194e-01 -7.13275671e-01 -1.40546381e-01 6.35358214e-01 -4.61983204e-01 -1.05940807e+00 -2.16211423e-01 -4.42386359e-01 -3.77038002e-01 2.87306815e-01 6.50637686e-01 -2.38203049e-01 2.26898193e-01 4.93914664e-01 7.09684253e-01 -1.81906760e-01 -6.20715618e-01 3.74930590e-01 1.50032556e+00 5.17466426e-01 -4.74809319e-01 7.65332222e-01 -2.66283005e-01 -5.54317117e-01 -6.94456279e-01 -9.43469524e-01 -2.25482732e-01 -5.07364690e-01 -8.15155059e-02 6.55473769e-01 -1.21820509e+00 1.07941017e-01 5.22857726e-01 -1.30703616e+00 -8.53103459e-01 -1.83654338e-01 5.50915360e-01 -8.16834629e-01 1.75081745e-01 -6.11971319e-01 -9.11745906e-01 -9.18431103e-01 -1.24240053e+00 1.20359170e+00 -5.31943977e-01 -1.33746579e-01 -7.64832556e-01 3.07845175e-01 9.58993912e-01 6.12076879e-01 -3.00088525e-01 6.78105891e-01 -9.26969051e-01 -2.34684616e-01 -2.42112979e-01 1.73333287e-01 7.13853300e-01 9.08718705e-02 -7.64733180e-02 -8.48391891e-01 -4.91120428e-01 -2.64789581e-01 -5.22291839e-01 4.60912049e-01 5.16518429e-02 6.62773550e-01 -7.49466181e-01 2.38171518e-01 4.12167758e-01 8.49997640e-01 2.04997405e-01 5.17336488e-01 4.49876457e-01 5.19969463e-01 5.10039866e-01 6.67760909e-01 1.14502788e-01 5.94981909e-01 1.16249132e+00 -8.47347528e-02 1.08647957e-01 -5.23770511e-01 -7.01052785e-01 8.55029166e-01 1.87324798e+00 3.18371236e-01 -5.89928627e-01 -1.26249635e+00 6.41566157e-01 -1.72152925e+00 -7.36997843e-01 -8.18729699e-02 2.13771129e+00 1.03935099e+00 4.74262647e-02 4.92761806e-02 -3.37508917e-02 6.07111335e-01 1.64848194e-02 -5.23265079e-02 -7.97137439e-01 -4.05969203e-01 2.16729343e-01 4.35622633e-01 5.98216534e-01 -6.54467583e-01 1.54679394e+00 6.47290707e+00 1.11375487e+00 -1.03228700e+00 8.02246392e-01 7.64185369e-01 -5.18664718e-02 -3.75102252e-01 5.20909652e-02 -5.72108150e-01 5.10585308e-01 1.93667793e+00 -2.63268799e-01 6.90194070e-01 4.53992933e-01 3.39838415e-01 5.71965337e-01 -7.39367783e-01 9.14417207e-01 -5.04358485e-03 -1.34107506e+00 2.47777283e-01 2.15856079e-02 9.71317947e-01 8.68800461e-01 1.99105456e-01 7.72378564e-01 3.42382550e-01 -9.14086342e-01 9.84649658e-01 3.90726849e-02 1.22147846e+00 -6.01714432e-01 7.58793950e-01 6.51856542e-01 -7.51934290e-01 6.33954182e-02 -9.43738595e-02 3.06030482e-01 2.73591548e-01 3.46004158e-01 -1.46774554e+00 8.89389873e-01 1.55530438e-01 4.32375252e-01 -4.16872293e-01 4.85972464e-01 -3.45146477e-01 1.26991320e+00 -1.80606604e-01 4.09528799e-02 4.34790879e-01 -5.01923040e-02 6.98462427e-01 1.22309732e+00 5.62178433e-01 -4.12040949e-01 -5.38334697e-02 3.44032496e-01 -4.26227272e-01 3.38596255e-01 -5.01143038e-01 -4.65543360e-01 7.40260839e-01 9.39140439e-01 1.25019550e-01 -5.40643156e-01 -2.89253205e-01 1.22611606e+00 2.64194787e-01 4.64453816e-01 -4.87187207e-01 1.14322111e-01 3.46771985e-01 -2.49434799e-01 4.17461619e-02 -5.78321576e-01 -3.33606303e-01 -1.36992157e+00 2.10861593e-01 -1.43751919e+00 -1.07461691e-01 -9.06609297e-01 -7.96238422e-01 1.18601692e+00 -1.91621736e-01 -1.34594584e+00 -7.46416926e-01 -1.56171456e-01 -4.29444611e-01 1.13290966e+00 -1.40570343e+00 -1.68036139e+00 3.48414809e-01 1.72788695e-01 1.20986712e+00 -9.03536260e-01 8.99419308e-01 3.24729711e-01 -5.10044575e-01 1.08596945e+00 4.29118007e-01 4.57369350e-02 7.27680683e-01 -9.92339432e-01 1.35229087e+00 8.19687307e-01 4.12553936e-01 3.23052168e-01 7.05412149e-01 -6.63471997e-01 -1.59924400e+00 -1.26791251e+00 1.56594527e+00 -6.97905660e-01 7.18182921e-01 -6.05744123e-01 -5.57589829e-01 6.13829195e-01 5.16957402e-01 -2.20251158e-01 6.12004757e-01 -2.67285528e-03 -3.36092740e-01 -1.18052863e-01 -9.94237423e-01 6.48361504e-01 8.76501739e-01 -8.71936738e-01 -5.94184697e-01 5.54888904e-01 1.11441457e+00 -4.58350509e-01 -8.34533751e-01 4.85131294e-01 5.03447413e-01 -1.18794635e-01 7.38682449e-01 -9.21539605e-01 3.11284930e-01 1.68015823e-01 -7.45300531e-01 -1.84342313e+00 2.03263640e-01 -1.13685274e+00 -2.77197838e-01 1.32687843e+00 1.14899492e+00 -4.62523878e-01 4.67225730e-01 -4.52866144e-02 -7.13456869e-01 -6.96548223e-01 -1.63077283e+00 -1.13538492e+00 3.59147578e-01 -4.31952089e-01 6.36544645e-01 8.88856888e-01 -3.72884702e-03 6.08601093e-01 -6.16517782e-01 -1.71770006e-01 4.53188956e-01 -3.77557755e-01 9.03797328e-01 -6.15185142e-01 -4.43522006e-01 -1.69980004e-01 1.85492665e-01 -9.25057530e-01 3.17785800e-01 -1.20317268e+00 6.79274695e-03 -1.46280956e+00 -9.60892905e-03 -1.09326042e-01 -7.13207275e-02 5.85408330e-01 4.33184244e-02 4.90883552e-02 7.30050206e-02 4.60740298e-01 -3.78960401e-01 9.55893755e-01 1.02250671e+00 -2.30541646e-01 3.04137226e-02 1.13082595e-01 -3.10082376e-01 -8.61937739e-03 1.14337456e+00 -4.50378627e-01 -1.85450703e-01 -9.65323269e-01 -5.57094589e-02 5.19047022e-01 -2.20610917e-01 -6.12017155e-01 -1.07473195e-01 -3.71524133e-02 -1.41021788e-01 -7.10748553e-01 5.11562467e-01 -4.47466850e-01 -3.09076137e-03 3.27787459e-01 -5.33743978e-01 4.51935023e-01 3.57111841e-01 1.09496728e-01 -2.89959788e-01 2.33801693e-01 5.71313918e-01 -1.05549723e-01 -1.68249048e-02 2.10868955e-01 -4.86512601e-01 1.83943436e-01 3.76046062e-01 1.02840766e-01 -4.53101367e-01 -6.03548229e-01 -5.20664573e-01 1.66098699e-01 2.35391587e-01 8.14628065e-01 5.33475995e-01 -1.53302634e+00 -1.47375524e+00 3.01081240e-01 4.36562479e-01 -5.06485760e-01 -8.47008452e-02 1.09525120e+00 -1.26642749e-01 7.61279285e-01 1.32925391e-01 -5.23916602e-01 -1.09350228e+00 2.25574449e-02 3.15890074e-01 -3.38122904e-01 -3.19809645e-01 6.39966130e-01 -5.51597774e-01 -1.03696132e+00 1.40571803e-01 3.68543193e-02 5.45839071e-01 -3.94478530e-01 2.69433856e-01 3.37548524e-01 6.26299322e-01 -8.64941359e-01 -2.29216844e-01 -6.13412932e-02 -2.52318352e-01 -1.02512670e+00 1.24880874e+00 -4.68084037e-01 2.49248445e-01 5.45086443e-01 1.14797616e+00 5.86065697e-03 -6.82524979e-01 -6.38940394e-01 3.44348848e-01 -1.22221760e-01 5.25943823e-02 -1.50119495e+00 -5.74521124e-01 1.04775476e+00 3.20904464e-01 -2.11841822e-01 8.02764893e-01 -1.20649785e-01 1.32271075e+00 6.31440639e-01 5.14171720e-01 -1.42106700e+00 1.07883019e-02 9.36410129e-01 9.95777488e-01 -1.28736246e+00 -6.19261563e-01 -1.24593927e-02 -8.10694158e-01 7.52317548e-01 2.85522074e-01 6.20641291e-01 -4.94841300e-02 1.61848277e-01 3.56506467e-01 4.99547452e-01 -1.44652605e+00 3.10498942e-02 4.27233338e-01 5.08123338e-01 4.15965945e-01 2.70019889e-01 -1.37038320e-01 3.51070821e-01 -3.23559761e-01 -1.74749643e-01 2.21733883e-01 7.10379481e-01 -4.07750905e-01 -1.45452189e+00 4.80477139e-02 -4.54008859e-03 -5.16340554e-01 -6.25420570e-01 -9.31133330e-01 3.54046106e-01 -5.27569830e-01 1.37540901e+00 -4.14975524e-01 -8.41194630e-01 3.93867344e-01 5.95693111e-01 4.28981066e-01 -6.88557029e-01 -7.27716506e-01 2.66050339e-01 6.96232855e-01 -1.68736309e-01 -1.51605189e-01 -7.28189647e-01 -7.01139867e-01 -2.60111183e-01 -2.05508113e-01 5.43086350e-01 1.20281959e+00 1.14349782e+00 5.33039570e-01 2.77151644e-01 1.05975413e+00 -3.45062733e-01 -8.29140961e-01 -1.71544969e+00 2.72551239e-01 -1.83263406e-01 2.97684699e-01 4.66549881e-02 -2.65674382e-01 -1.13615379e-01]
[14.48514461517334, 7.18217134475708]
dda72c64-a1e6-4247-8962-3709b5f434b5
measuring-annotator-agreement-generally
2212.09503
null
https://arxiv.org/abs/2212.09503v1
https://arxiv.org/pdf/2212.09503v1.pdf
Measuring Annotator Agreement Generally across Complex Structured, Multi-object, and Free-text Annotation Tasks
When annotators label data, a key metric for quality assurance is inter-annotator agreement (IAA): the extent to which annotators agree on their labels. Though many IAA measures exist for simple categorical and ordinal labeling tasks, relatively little work has considered more complex labeling tasks, such as structured, multi-object, and free-text annotations. Krippendorff's alpha, best known for use with simpler labeling tasks, does have a distance-based formulation with broader applicability, but little work has studied its efficacy and consistency across complex annotation tasks. We investigate the design and evaluation of IAA measures for complex annotation tasks, with evaluation spanning seven diverse tasks: image bounding boxes, image keypoints, text sequence tagging, ranked lists, free text translations, numeric vectors, and syntax trees. We identify the difficulty of interpretability and the complexity of choosing a distance function as key obstacles in applying Krippendorff's alpha generally across these tasks. We propose two novel, more interpretable measures, showing they yield more consistent IAA measures across tasks and annotation distance functions.
['Matthew Lease', 'Omar Alonso', 'Alexander Braylan']
2022-12-15
null
null
null
null
['text-annotation']
['natural-language-processing']
[ 2.51447737e-01 2.62314767e-01 -3.16890150e-01 -9.06782329e-01 -9.90083992e-01 -1.18683267e+00 5.55424631e-01 8.48825514e-01 -7.05058157e-01 6.25963807e-01 4.44351196e-01 -2.68282175e-01 -5.03924370e-01 8.52883682e-02 -1.03108607e-01 -2.96430349e-01 1.08685650e-01 7.63392866e-01 5.71088791e-02 1.25468701e-01 3.24527919e-01 3.80693853e-01 -1.43235433e+00 3.10586900e-01 7.45915592e-01 1.04023242e+00 -1.37911201e-01 5.37814140e-01 -2.38937929e-01 7.64075875e-01 -7.95522809e-01 -7.24684775e-01 9.14875567e-02 -1.17650330e-01 -1.48252630e+00 -7.95170839e-04 7.15940654e-01 2.19738573e-01 3.57322991e-01 1.03504777e+00 3.88214320e-01 -7.86404610e-02 1.04303205e+00 -1.40380859e+00 -9.03485894e-01 5.84353030e-01 -4.22125578e-01 3.01228277e-02 7.43940949e-01 -2.36949455e-02 1.61598134e+00 -8.14051449e-01 6.33328676e-01 1.32553291e+00 1.22284544e+00 3.30756903e-01 -1.30471134e+00 -2.30743736e-01 2.04076860e-02 -1.52225360e-01 -1.37459850e+00 -3.08343731e-02 9.69368443e-02 -8.70328665e-01 7.22567081e-01 4.86997128e-01 6.47176132e-02 6.63743079e-01 -6.03278130e-02 5.57560325e-01 1.44030356e+00 -6.41660273e-01 9.03781652e-02 2.18507290e-01 5.20973086e-01 8.09081733e-01 3.23053211e-01 -6.65359080e-01 -3.94979924e-01 -5.55068374e-01 4.21440095e-01 -3.43144566e-01 -5.64441718e-02 -3.87601167e-01 -1.59303463e+00 7.44832695e-01 8.69212672e-02 5.68093240e-01 1.25758452e-02 1.34373069e-01 9.34379160e-01 3.76716316e-01 5.82193613e-01 1.09044826e+00 -7.93935478e-01 -1.37131780e-01 -6.40964389e-01 3.81582409e-01 7.69981444e-01 1.02576685e+00 7.49450386e-01 -6.56623065e-01 -5.15444279e-01 1.05447853e+00 3.34150851e-01 1.85883939e-01 3.34739178e-01 -1.19054961e+00 3.61836940e-01 6.65564477e-01 3.96640360e-01 -1.06088400e+00 -8.27759564e-01 -1.11199431e-01 -5.42747378e-01 1.13335751e-01 6.78347468e-01 1.12263300e-01 -6.86320722e-01 1.70310855e+00 1.72812045e-01 -7.61846066e-01 -8.12171474e-02 8.78364980e-01 7.56373286e-01 1.65829137e-01 5.77699423e-01 -2.56557554e-01 1.82011092e+00 -7.23692954e-01 -8.27671528e-01 -1.18993238e-01 1.25357914e+00 -1.10173798e+00 1.38869858e+00 2.02756971e-01 -9.40377414e-01 -5.86478055e-01 -6.97755694e-01 -5.35032511e-01 -5.79930246e-01 3.63895774e-01 7.91638792e-01 7.38429546e-01 -1.11671364e+00 4.66388524e-01 -3.74910533e-01 -5.12294531e-01 2.50930190e-01 4.45798069e-01 -7.71388829e-01 8.04052502e-02 -1.09682608e+00 1.27290547e+00 4.29248542e-01 -2.14462072e-01 -2.67563403e-01 -6.23494864e-01 -8.12362850e-01 -1.49240419e-02 2.62801677e-01 -7.32349396e-01 1.45777416e+00 -8.65092337e-01 -7.97043443e-01 1.42845011e+00 4.05812450e-02 -8.46169740e-02 3.66710037e-01 -1.52904853e-01 -1.93939939e-01 -1.89541519e-01 4.44225490e-01 1.02607799e+00 3.23750436e-01 -1.16749239e+00 -6.38850749e-01 -4.27926958e-01 1.56689167e-01 4.23554301e-01 -3.71429414e-01 3.60230714e-01 -3.12251523e-02 -7.71301270e-01 -4.82940339e-02 -1.08925486e+00 -1.96123734e-01 3.09082985e-01 -1.73484385e-01 -8.84975374e-01 4.45524931e-01 -4.21097189e-01 1.47382224e+00 -2.27929807e+00 -1.37601383e-02 1.79957837e-01 4.38799292e-01 3.48642357e-02 -1.80386871e-01 3.12441230e-01 -2.09445804e-01 5.90636373e-01 -3.09306473e-01 -3.27666819e-01 3.38214070e-01 3.33733439e-01 1.03790991e-01 3.26197505e-01 6.07615747e-02 7.21588790e-01 -1.19323051e+00 -8.91452074e-01 -4.54144888e-02 5.65591976e-02 -2.34102845e-01 1.52132651e-02 -1.59545332e-01 1.44400656e-01 -2.79889137e-01 6.67646348e-01 1.96242496e-01 -5.41272640e-01 2.55219102e-01 -5.97920597e-01 -2.35601410e-01 7.12376088e-02 -1.03506219e+00 1.70264745e+00 -3.43299568e-01 7.58213878e-01 -1.12953596e-01 -6.04829252e-01 9.24328864e-01 3.79757583e-01 6.15796328e-01 -3.33558738e-01 3.91073450e-02 3.91896039e-01 -1.09395541e-01 -5.81995308e-01 6.34078085e-01 -1.08926699e-01 -5.00135601e-01 6.37191594e-01 2.10699305e-01 -3.72883290e-01 4.93431002e-01 1.37775123e-01 1.19553983e+00 -9.43521261e-02 7.28356063e-01 -6.68641210e-01 3.54319066e-01 1.52999625e-01 1.77365616e-01 7.66202807e-01 -4.31104958e-01 6.20787919e-01 4.77589846e-01 -7.36365020e-01 -1.39647913e+00 -8.25037062e-01 -3.48294586e-01 1.50743973e+00 -8.51981640e-02 -7.66197145e-01 -7.17161238e-01 -1.05214608e+00 3.56131792e-02 3.79214108e-01 -7.89736092e-01 2.92699814e-01 -1.86366990e-01 -5.68932414e-01 8.72698069e-01 4.85877275e-01 1.91583544e-01 -9.04463708e-01 -5.35473585e-01 -9.55588594e-02 -3.10171247e-01 -9.82502282e-01 -7.83772290e-01 3.52617353e-01 -7.10888207e-01 -1.11665821e+00 -6.87291861e-01 -9.56770837e-01 6.97871327e-01 1.26222651e-02 1.46102345e+00 6.17070347e-02 -1.93406686e-01 6.38855517e-01 -3.76411647e-01 -3.91630441e-01 -5.87932765e-01 1.89523295e-01 9.43913758e-02 -3.91155988e-01 4.12425905e-01 6.76902607e-02 -2.52487808e-01 6.92778587e-01 -1.07986546e+00 -1.45738736e-01 5.99902153e-01 8.04779053e-01 6.23346627e-01 -2.52614737e-01 2.97176957e-01 -1.15219390e+00 9.17260289e-01 -9.47110653e-02 -1.78762570e-01 6.01066291e-01 -8.23226988e-01 1.92856550e-01 2.87589043e-01 -3.41009885e-01 -7.61783183e-01 2.69411951e-01 -1.66606475e-02 -7.45421797e-02 -2.52829432e-01 4.28457081e-01 8.13545659e-02 -1.49881616e-01 1.09035647e+00 -5.09825826e-01 -1.24391720e-01 -3.19882929e-01 6.91737890e-01 7.17560112e-01 4.85609621e-01 -7.23312557e-01 5.48160791e-01 3.28370661e-01 -1.53053049e-02 -5.02371311e-01 -1.43349028e+00 -9.02139544e-01 -9.82062936e-01 -1.02855280e-01 1.19991612e+00 -7.26231813e-01 -8.52959871e-01 1.65955387e-02 -1.32228494e+00 2.66106892e-02 -4.68100637e-01 3.39190573e-01 -6.31469190e-01 6.40732646e-01 -4.89339113e-01 -5.56581974e-01 -2.42985308e-01 -1.13239241e+00 1.51048863e+00 -2.81076759e-01 -1.21748865e+00 -1.26625776e+00 2.30906636e-01 5.15236795e-01 1.64760724e-01 2.45605260e-01 1.13946414e+00 -9.01465178e-01 1.67079255e-01 -2.13997319e-01 -4.29026306e-01 3.36486489e-01 1.95881382e-01 -5.63550889e-02 -7.41538942e-01 -4.39168438e-02 -2.47628421e-01 -6.96340084e-01 4.80106026e-01 1.88117877e-01 1.16461754e+00 -6.43433988e-01 -3.54068398e-01 3.17693092e-02 1.10674107e+00 -3.40448320e-03 3.85546744e-01 3.77338022e-01 8.55412483e-01 8.41735721e-01 8.18866551e-01 2.03447029e-01 3.55518281e-01 7.97835648e-01 1.17151447e-01 -1.56197175e-01 9.47711915e-02 1.06946021e-01 -5.24035320e-02 9.48720336e-01 -1.22027971e-01 -2.83396244e-01 -1.23052764e+00 3.58720571e-01 -1.84738255e+00 -8.07404578e-01 -3.47187012e-01 1.84005833e+00 1.06499851e+00 1.46700099e-01 4.45380472e-02 2.57474273e-01 6.60572588e-01 1.34906784e-01 -1.10127166e-01 -4.40759003e-01 -7.40917996e-02 -2.01065674e-01 5.37798226e-01 4.53752309e-01 -1.38069201e+00 5.87547243e-01 7.65881205e+00 9.32055771e-01 -4.04728711e-01 2.85531193e-01 7.54989684e-01 5.19573510e-01 -4.01914716e-01 4.10105996e-02 -6.57872021e-01 2.82624185e-01 6.99389398e-01 1.79997638e-01 -1.82911605e-01 1.03319788e+00 -1.16321690e-01 6.38063848e-02 -1.54728997e+00 1.08083606e+00 1.13870673e-01 -9.73218024e-01 -8.71250108e-02 -2.66102284e-01 8.09957623e-01 -3.32754701e-01 -8.53460431e-02 8.29426199e-02 4.39233899e-01 -1.06013346e+00 8.93702984e-01 3.93785566e-01 1.01889956e+00 -2.50584930e-01 9.23155367e-01 -1.13028400e-01 -1.01592445e+00 6.51458502e-02 -1.76535055e-01 1.94900774e-03 5.17980196e-02 6.74833477e-01 -8.96956623e-01 1.99907660e-01 6.29879236e-01 5.83023548e-01 -8.88478756e-01 7.43196785e-01 1.01437099e-01 3.32967430e-01 -9.21792835e-02 -1.19974136e-01 4.16102052e-01 4.14940044e-02 1.73465922e-01 1.63824463e+00 1.12986147e-01 -3.64222974e-02 4.36953962e-01 3.16739202e-01 -2.93997359e-02 3.36199164e-01 -6.18870974e-01 -9.20036584e-02 5.44122815e-01 1.30704653e+00 -1.10749185e+00 -3.04559350e-01 -3.72646183e-01 7.18278170e-01 2.85157263e-01 -2.92650294e-02 -7.40928769e-01 -2.67726302e-01 4.24676836e-01 4.17770408e-02 -2.58389026e-01 -2.49113813e-01 -6.75496459e-01 -8.26444745e-01 3.32412380e-03 -9.26605821e-01 7.94961035e-01 -1.00606275e+00 -1.70182264e+00 5.41816473e-01 2.75517344e-01 -1.28853822e+00 -2.63062809e-02 -8.29346776e-01 1.18920811e-01 5.51640332e-01 -7.40313768e-01 -1.14932370e+00 -4.58863914e-01 3.33261251e-01 4.25070852e-01 7.68585056e-02 1.05782044e+00 3.95652622e-01 3.13555705e-04 5.34260571e-01 -1.34640262e-01 1.44755885e-01 1.05360401e+00 -1.64684701e+00 2.29551241e-01 1.48189411e-01 2.00594947e-01 6.47054195e-01 7.91027069e-01 -5.00222147e-01 -8.49250972e-01 -7.71198809e-01 1.14266503e+00 -1.05965793e+00 8.01407039e-01 -1.86782822e-01 -9.78213668e-01 7.05035388e-01 8.64659995e-02 -1.25050858e-01 9.05661404e-01 4.25576806e-01 -7.94420719e-01 7.86785558e-02 -1.08043766e+00 4.18240815e-01 1.09633517e+00 -7.41787076e-01 -5.11502028e-01 8.03453267e-01 7.73944199e-01 -2.77783006e-01 -1.44244790e+00 6.13644063e-01 6.46987021e-01 -6.42284930e-01 9.19668555e-01 -6.12014651e-01 1.98154822e-01 -5.42099595e-01 -4.67096344e-02 -1.21353710e+00 -6.02423310e-01 -3.79100978e-01 5.36523461e-01 1.22681069e+00 6.10856354e-01 -2.97933936e-01 3.41903359e-01 8.15516174e-01 -3.12947571e-01 -5.93368530e-01 -7.79025078e-01 -7.58729458e-01 -9.41918641e-02 -2.07708284e-01 3.63871992e-01 1.50565505e+00 1.96407989e-01 7.22433209e-01 -1.73442349e-01 -1.78575456e-01 2.99641609e-01 -3.96107465e-01 5.56309938e-01 -1.62502968e+00 6.03856482e-02 -6.01464033e-01 -6.36893511e-01 -5.75922549e-01 1.63395658e-01 -9.66556430e-01 1.91606298e-01 -1.63410604e+00 2.69991189e-01 -7.17123508e-01 6.88197985e-02 7.75727689e-01 -2.09761277e-01 4.21710223e-01 1.96566910e-01 6.50528371e-01 -1.19677889e+00 -1.98580548e-01 1.01312888e+00 -2.77594298e-01 1.89007744e-01 -3.30360174e-01 -5.18749774e-01 9.64665592e-01 5.50869763e-01 -6.56299531e-01 -3.11730981e-01 -6.43092930e-01 7.08752394e-01 -2.26235136e-01 1.97625026e-01 -8.21078777e-01 1.35229170e-01 -1.20309450e-01 3.31644237e-01 -2.46447101e-01 -9.37324464e-02 -8.94166291e-01 3.11152786e-01 1.80389717e-01 -1.00913298e+00 7.08589673e-01 -1.34012267e-01 3.64375651e-01 -3.10915560e-01 -6.18428528e-01 6.77957833e-01 -2.76538372e-01 -6.80056453e-01 6.53172284e-02 -2.84830213e-01 3.48397642e-01 9.85863507e-01 -1.96472585e-01 -3.50259006e-01 -6.82783872e-02 -9.94941533e-01 1.92337990e-01 6.00563467e-01 3.91315132e-01 8.48284513e-02 -1.40997374e+00 -6.27391756e-01 -5.52326560e-01 6.21183455e-01 -1.73583984e-01 -1.32925943e-01 8.13305318e-01 -6.53865635e-01 5.01443028e-01 -3.64342518e-02 -8.32638800e-01 -1.47511303e+00 5.32079875e-01 5.28128035e-02 -6.21352136e-01 -8.70979875e-02 6.29608274e-01 2.58053929e-01 -4.65846449e-01 3.40445757e-01 -4.48166788e-01 -3.04419488e-01 4.14387017e-01 3.09605271e-01 4.04727727e-01 9.14265588e-02 -7.25546181e-01 -3.87303442e-01 7.17727065e-01 7.59928441e-03 -3.19898203e-02 7.91022480e-01 -2.51069993e-01 -3.14572126e-01 7.81516433e-01 1.20148885e+00 -1.65753141e-01 -6.83977544e-01 -8.99120718e-02 5.86376131e-01 -3.76570076e-01 -5.48871636e-01 -8.74685287e-01 -2.57536680e-01 6.95568800e-01 5.73091209e-01 6.90979540e-01 8.80717456e-01 2.82198906e-01 2.64069438e-01 5.75210512e-01 3.70775580e-01 -1.14982271e+00 3.45922142e-01 5.38498163e-01 8.26385617e-01 -1.27350223e+00 2.48859733e-01 -4.36387062e-01 -9.23453689e-01 1.06037760e+00 4.43290681e-01 5.35809159e-01 4.14971858e-01 1.95789665e-01 2.35630542e-01 -4.85903412e-01 -3.26746106e-01 -1.67095438e-01 6.15396082e-01 5.28697073e-01 9.93441463e-01 1.39582232e-01 -5.98791718e-01 2.69063503e-01 -1.66248396e-01 -2.44221091e-01 2.74419487e-01 8.08072805e-01 -5.01217127e-01 -1.05719900e+00 -2.77741075e-01 6.37363434e-01 -6.87969267e-01 -4.94139791e-02 -5.61412930e-01 8.51974308e-01 3.02152902e-01 6.66336954e-01 2.84473039e-02 -1.93663314e-01 3.25984806e-01 2.63852984e-01 3.66451293e-01 -9.13937986e-01 -8.80888641e-01 -3.61756057e-01 4.09536868e-01 -3.70081067e-01 -1.07738614e+00 -6.57957077e-01 -9.77024615e-01 -1.87805314e-02 -4.74166334e-01 5.19447446e-01 6.28484368e-01 1.07288527e+00 6.45269081e-02 9.25851315e-02 1.98662356e-01 -5.28557837e-01 -4.35304880e-01 -1.05341041e+00 -3.14396948e-01 1.01285744e+00 1.71518981e-01 -6.35813773e-01 -3.47761631e-01 3.65069330e-01]
[9.624165534973145, 4.702820777893066]
0abcd2e9-4e66-4a81-8856-0565c527d5b7
pseudo-label-generation-evaluation-framework
null
null
https://ieeexplore.ieee.org/document/9506549
https://ieeexplore.ieee.org/iel7/9506008/9506009/09506549.pdf
Pseudo-Label Generation-Evaluation Framework For Cross Domain Weakly Supervised Object Detection
Cross domain weakly supervised object detection (CDWSOD), where we can get access to instance-level annotations in the source domain while only image-level annotations are available in the target domain, adapts object detectors from label-rich to label-poor domains. It usually generates pseudo labels in the target domain and utilize them to finetune the detector pretrained in the source domain. In this paper, we propose a new pseudo-label generation-evaluation framework for CDWSOD task. In particular, an evaluator is introduced for the generated pseudo labels in the target domain and the transferring process involves two players: the detector to generate instance-level pseudo labels and the evaluator to judge the quality of pseudo labels. Only high-quality pseudo labels selected by the evaluator are utilized to finetune the detector. Experiments on three representative datasets demonstrate the effectiveness of our framework in various domains.
['Yingming Li', 'Kejie Lyu', 'Xinglu Wang', 'Shengxiong Ouyang']
2021-08-23
null
null
null
ieee-international-conference-on-image-3
['weakly-supervised-object-detection']
['computer-vision']
[ 4.78941947e-01 3.02787393e-01 -2.63817757e-01 -5.56156874e-01 -1.16159546e+00 -7.48941422e-01 3.47805738e-01 3.45346425e-03 -6.89912021e-01 6.94195509e-01 -3.18608493e-01 2.77044833e-01 4.93993580e-01 -6.35568917e-01 -7.16638982e-01 -7.61687160e-01 4.45836544e-01 6.46519899e-01 8.65752995e-01 1.77305207e-01 -1.73679829e-01 8.32590833e-02 -1.40746045e+00 5.78973830e-01 7.39723802e-01 1.15359044e+00 5.54275751e-01 4.44903791e-01 -6.70760572e-02 8.15196335e-01 -4.80204523e-01 -3.56015056e-01 4.62473273e-01 -3.49510223e-01 -7.37608016e-01 4.84136254e-01 3.54159504e-01 -2.37012252e-01 -5.25787212e-02 1.42400968e+00 4.54262167e-01 -1.96065381e-01 7.84418941e-01 -1.35491025e+00 -5.87763608e-01 6.58342957e-01 -6.57579362e-01 8.48559886e-02 -6.74797222e-02 5.10168612e-01 1.00698018e+00 -1.18828309e+00 8.22952449e-01 1.18875480e+00 1.98527426e-01 8.28534722e-01 -1.26260448e+00 -9.34463859e-01 4.01631147e-01 4.54387888e-02 -1.27639270e+00 -4.28320915e-01 8.48352969e-01 -5.80451727e-01 2.63221890e-01 -1.86979815e-01 4.98583131e-02 9.33916807e-01 -5.58981836e-01 1.00110996e+00 1.13736224e+00 -5.71485937e-01 3.51154089e-01 5.75085282e-01 1.65137231e-01 6.88139081e-01 1.62491500e-01 1.37475103e-01 -5.03083348e-01 -1.38154328e-01 4.43934172e-01 -2.76913613e-01 -8.80261436e-02 -6.18171215e-01 -1.27941597e+00 5.47392666e-01 5.44492900e-01 1.12482503e-01 -4.05029267e-01 -1.42658278e-01 4.40510839e-01 2.25929886e-01 5.53497136e-01 3.55191737e-01 -6.86153471e-01 4.14328516e-01 -8.19334447e-01 2.39703543e-02 3.87102187e-01 1.20181739e+00 9.65793252e-01 -3.79593134e-01 -6.72098041e-01 1.11537719e+00 2.61559963e-01 4.49306041e-01 2.88733363e-01 -9.96569633e-01 6.33560836e-01 1.04810238e+00 4.95498985e-01 -3.38173568e-01 -4.58590426e-02 -3.32262635e-01 -3.74984771e-01 3.85542482e-01 5.61418056e-01 -1.62554830e-01 -1.18415022e+00 1.96752429e+00 7.47279048e-01 1.59295410e-01 7.07611367e-02 1.18234003e+00 9.49021757e-01 5.88003695e-01 4.17157263e-01 -1.28951415e-01 1.41328251e+00 -1.31576324e+00 -2.90896982e-01 -4.50496435e-01 6.80847347e-01 -5.85204840e-01 1.31705821e+00 1.05613016e-01 -1.00919402e+00 -8.81877124e-01 -9.97730970e-01 6.50534257e-02 -2.42814571e-01 7.78916240e-01 -2.21999824e-01 1.59753740e-01 -7.36080050e-01 1.93118840e-01 -5.37009418e-01 -2.20162585e-01 7.66661227e-01 2.52202183e-01 -3.48457664e-01 -3.80617678e-01 -1.28285623e+00 6.64358914e-01 9.21440959e-01 -3.06868166e-01 -1.37592745e+00 -5.14874458e-01 -6.33846164e-01 -1.67019546e-01 7.04401553e-01 -4.65119988e-01 1.51687586e+00 -1.51609957e+00 -1.16689193e+00 1.56182456e+00 1.39128834e-01 -2.69437023e-02 7.88750589e-01 2.98211217e-01 -3.87981385e-01 2.84002095e-01 6.02248311e-01 9.00956869e-01 9.92821157e-01 -1.58064282e+00 -1.28352511e+00 -2.60106623e-01 1.88834772e-01 3.59695017e-01 -2.33393162e-01 -2.20928699e-01 -9.48105037e-01 -4.60909337e-01 -3.83071192e-02 -9.56291616e-01 -1.04094192e-01 3.45800221e-01 -5.45749605e-01 -5.54926932e-01 8.05664062e-01 -1.91873744e-01 7.75204003e-01 -2.31826091e+00 -8.49433541e-02 3.71361934e-02 3.00983727e-01 5.46005905e-01 -5.13405919e-01 -1.76248148e-01 -1.16602801e-01 -4.16671038e-01 -3.87140453e-01 -3.20852846e-01 4.33932357e-02 1.83972329e-01 2.02261582e-02 3.82943451e-01 5.96530855e-01 8.38999748e-01 -1.19728076e+00 -8.54209840e-01 -3.64626981e-02 -2.45030839e-02 -1.37621149e-01 6.15592360e-01 -5.35807908e-01 3.28420997e-01 -7.63447344e-01 5.83213389e-01 8.15189064e-01 -5.45093775e-01 4.53503616e-02 -3.78100455e-01 1.84748888e-01 -5.68033196e-02 -1.41740119e+00 1.42853737e+00 -2.99250394e-01 8.74420553e-02 1.19166888e-01 -8.76398623e-01 6.87601388e-01 1.09762013e-01 1.71629727e-01 -7.21549928e-01 7.12699369e-02 3.00389737e-01 -2.51330465e-01 -6.29604876e-01 1.09295711e-01 -2.53789306e-01 -2.56867468e-01 6.76250756e-01 3.50615561e-01 2.05708683e-01 4.59700227e-01 3.96817714e-01 1.11793339e+00 2.60811120e-01 2.19243735e-01 -1.24084838e-01 8.91239405e-01 4.34143305e-01 7.84751952e-01 7.18186855e-01 -4.25891906e-01 5.58472276e-01 4.62063015e-01 -2.36922204e-01 -1.24279618e+00 -1.07528567e+00 -5.64833209e-02 1.83781159e+00 4.27081525e-01 1.01386793e-01 -9.32281315e-01 -1.46293855e+00 3.27789970e-02 3.90990198e-01 -7.92167068e-01 -3.22856307e-01 -2.38838539e-01 -3.34359229e-01 4.77970093e-01 5.33935606e-01 6.71206534e-01 -1.30040491e+00 -4.02076304e-01 1.94243908e-01 -2.45604455e-01 -1.35049522e+00 -8.57848465e-01 2.92969376e-01 -4.27556038e-01 -1.24316072e+00 -7.55707383e-01 -1.12392378e+00 1.06203079e+00 2.41239026e-01 1.13796294e+00 -1.17531963e-01 -2.12795094e-01 -5.09851202e-02 -4.53958690e-01 -2.29806706e-01 -7.97080398e-01 -2.97645256e-02 -1.15635440e-01 2.66802490e-01 5.51477432e-01 2.09869906e-01 -4.48265314e-01 7.28157938e-01 -1.02136683e+00 3.22556011e-02 7.01209784e-01 8.88667166e-01 7.99914360e-01 1.04436673e-01 6.47858739e-01 -1.56763518e+00 2.14401126e-01 -4.09830898e-01 -8.85555804e-01 3.96501392e-01 -4.79515374e-01 1.36284187e-01 4.76444453e-01 -9.03737426e-01 -1.36512661e+00 6.60299182e-01 1.61609173e-01 -3.62541437e-01 -3.30936879e-01 1.85199887e-01 -6.32746100e-01 1.46383509e-01 1.03927886e+00 2.65905783e-02 -3.69110614e-01 -4.56479251e-01 4.40529674e-01 8.42514038e-01 7.39257812e-01 -5.59457779e-01 9.86497343e-01 4.75424707e-01 -5.62023878e-01 2.09547896e-02 -1.44445062e+00 -6.39329255e-01 -7.49746799e-01 -2.24469423e-01 8.87787223e-01 -1.19075143e+00 -3.82132046e-02 5.32950878e-01 -1.00813711e+00 -5.40317476e-01 -6.90214872e-01 1.84656233e-01 -3.65918756e-01 -1.24631040e-01 -3.60794753e-01 -5.36911547e-01 -1.50934175e-01 -1.23095822e+00 1.33725274e+00 2.55222410e-01 2.56330550e-01 -6.36558354e-01 -1.47107989e-01 3.53116304e-01 -1.62535504e-01 1.16436526e-01 8.31957996e-01 -7.34204292e-01 -4.41855848e-01 -2.51518697e-01 -8.03086221e-01 6.65225625e-01 2.85981093e-02 -3.94102663e-01 -1.25394630e+00 -1.49478883e-01 -3.56306255e-01 -7.55216658e-01 8.74359965e-01 1.30481198e-01 1.00430167e+00 -1.41243502e-01 -4.98564929e-01 1.34831712e-01 1.38754714e+00 -4.55088578e-02 1.33700268e-02 1.33508280e-01 6.55970871e-01 7.93650210e-01 1.03563368e+00 2.87118137e-01 4.20110822e-01 6.25468731e-01 2.55561233e-01 -4.06183004e-01 -5.89839458e-01 -4.18459058e-01 5.69038868e-01 1.20415844e-01 4.75081474e-01 -3.03029805e-01 -7.31459558e-01 7.16036439e-01 -1.83472085e+00 -5.36687732e-01 -7.38492608e-02 1.86022115e+00 1.39322197e+00 4.59928691e-01 3.36453438e-01 -2.18337551e-01 1.27969670e+00 -1.41385362e-01 -1.06930816e+00 3.20485681e-01 -7.53046246e-03 -1.27796426e-01 5.01076102e-01 3.15999568e-01 -1.26838517e+00 1.32859504e+00 5.15442276e+00 8.72725070e-01 -8.58978033e-01 5.93054414e-01 6.13633871e-01 2.53612492e-02 6.13965690e-02 -5.83155900e-02 -8.92480433e-01 5.89226544e-01 3.83106768e-01 -2.30071619e-02 -1.29411612e-02 1.27246046e+00 1.05251186e-01 -2.39941940e-01 -1.28274477e+00 9.09241736e-01 -8.25619549e-02 -9.90127921e-01 -9.38736424e-02 -2.40382805e-01 9.56603527e-01 2.35638544e-01 -6.31069094e-02 3.29993695e-01 7.60806084e-01 -3.62636685e-01 9.18949902e-01 2.35686794e-01 1.23303509e+00 -5.62372863e-01 7.50006735e-01 6.98001742e-01 -1.20431852e+00 -1.54359728e-01 -4.65440691e-01 4.04226333e-01 1.85000636e-02 7.88627982e-01 -1.07802486e+00 1.66464765e-02 5.51340520e-01 4.50170904e-01 -6.98897541e-01 7.28023291e-01 -7.16053724e-01 6.43641889e-01 -9.22713876e-02 4.02421802e-01 2.96264172e-01 1.72325179e-01 2.25829825e-01 1.33613443e+00 -4.95975614e-02 1.91421002e-01 5.95934451e-01 8.46467853e-01 -5.19108772e-01 -1.21965915e-01 -1.48479477e-01 1.07124180e-01 5.07054627e-01 1.49720550e+00 -6.80475354e-01 -7.06363380e-01 -2.85558224e-01 1.08287084e+00 4.26366121e-01 4.94460732e-01 -6.79920435e-01 -1.20544076e-01 2.29064167e-01 3.37536305e-01 2.18385145e-01 3.49603623e-01 -1.29222289e-01 -1.07019424e+00 6.68818802e-02 -5.87870061e-01 7.63397574e-01 -9.65095460e-01 -1.52834702e+00 6.54146612e-01 -1.28420174e-01 -1.29487145e+00 -7.80964717e-02 -5.12726784e-01 -2.63824105e-01 9.79614913e-01 -1.75681376e+00 -1.33986926e+00 -6.36679173e-01 7.54448354e-01 6.45337522e-01 -1.32526845e-01 3.97491127e-01 5.09357989e-01 -5.00276864e-01 6.30375862e-01 -2.11986244e-01 3.78554702e-01 8.75811756e-01 -1.35185337e+00 1.84933379e-01 8.50630939e-01 -5.76294260e-03 -3.33852172e-02 5.68827689e-01 -6.33075416e-01 -5.55608571e-01 -1.59794450e+00 7.06096888e-01 -3.39900523e-01 4.60073113e-01 -4.79519993e-01 -1.02018380e+00 4.05104011e-01 -2.70551682e-01 6.53427780e-01 3.37690026e-01 -3.89418423e-01 -5.16750753e-01 -1.48548067e-01 -1.38954306e+00 1.31418899e-01 1.06787694e+00 -6.13590598e-01 -5.65015614e-01 5.30151546e-01 6.20450437e-01 -3.54925215e-01 -3.14334065e-01 3.97444367e-01 2.12334350e-01 -3.36349219e-01 7.66510010e-01 -5.82057655e-01 2.57866323e-01 -8.43592882e-01 7.53437579e-02 -1.22959971e+00 -3.21429372e-01 8.16064999e-02 1.29868537e-01 1.51555490e+00 6.57068014e-01 -2.25457519e-01 8.06571960e-01 5.79163790e-01 4.79720384e-02 -3.64763588e-01 -7.79372096e-01 -7.54699707e-01 -2.82097787e-01 -3.44577044e-01 4.49407458e-01 8.56890082e-01 -4.15161967e-01 4.80218649e-01 -2.78452020e-02 4.09395665e-01 7.74690151e-01 2.76514620e-01 7.49834478e-01 -1.19663334e+00 -3.55065823e-01 -5.38831344e-03 -3.29570509e-02 -9.01118934e-01 4.35766757e-01 -1.05906379e+00 6.52500272e-01 -1.34058046e+00 6.44378781e-01 -8.43087316e-01 -6.01280868e-01 7.97401607e-01 -5.14465094e-01 6.70583844e-01 2.65253633e-02 3.27897877e-01 -1.10983217e+00 2.53577858e-01 1.35225129e+00 -3.39479774e-01 -3.07154600e-02 1.80882469e-01 -7.36969829e-01 8.46577764e-01 5.02321184e-01 -9.33396876e-01 -5.31394958e-01 -3.32878560e-01 -3.39517295e-02 -2.80354083e-01 4.49137181e-01 -8.10419798e-01 1.40933827e-01 -2.41563693e-01 3.76788706e-01 -3.64555955e-01 -2.23050304e-02 -7.69868433e-01 -1.55722216e-01 2.38071352e-01 -7.13512063e-01 -5.23961842e-01 -1.78011596e-01 6.45416021e-01 -1.22785702e-01 -4.21937883e-01 1.55200112e+00 -1.65333703e-01 -1.11117923e+00 5.33704996e-01 1.79754376e-01 3.83113414e-01 1.33712780e+00 -1.33179165e-02 -2.55179137e-01 1.14964359e-01 -9.99378324e-01 4.82820868e-01 5.60714662e-01 3.43484461e-01 4.34161752e-01 -1.40499318e+00 -8.59711051e-01 3.02245431e-02 9.34165716e-01 1.88369855e-01 1.05069242e-01 2.99500942e-01 1.31769106e-03 -1.98461935e-01 6.31497428e-02 -6.28039122e-01 -1.15196848e+00 8.08625758e-01 3.53950918e-01 -2.50805914e-01 -2.64511526e-01 1.22364140e+00 6.92348838e-01 -5.39009869e-01 3.47460181e-01 -1.05837546e-01 -8.10818002e-02 2.99730182e-01 6.21139705e-01 3.92648131e-02 -6.62958845e-02 -7.59638369e-01 -4.26789463e-01 3.65733981e-01 -1.65624008e-01 -1.49963260e-01 1.16401577e+00 -2.14996710e-01 9.99843031e-02 4.05755103e-01 1.07548690e+00 -4.87092167e-01 -1.84214032e+00 -8.72015655e-01 1.65190428e-01 -2.62063891e-01 8.63742903e-02 -1.21536064e+00 -1.13475442e+00 7.06499696e-01 8.03828657e-01 -1.42738894e-01 1.32958817e+00 5.45169890e-01 6.64352655e-01 1.00784823e-01 3.29794079e-01 -1.59387255e+00 5.08481622e-01 1.08951569e-01 5.24353385e-01 -1.49012232e+00 -3.44000965e-01 -4.38903213e-01 -1.08013892e+00 6.11595035e-01 1.10350168e+00 -7.34175742e-02 3.96094888e-01 2.22201213e-01 1.97378263e-01 -8.49512294e-02 -6.60944581e-01 -6.01425588e-01 3.28079939e-01 6.81864500e-01 9.39979684e-03 1.12183467e-01 -1.05536699e-01 5.83109438e-01 5.58945179e-01 1.64511666e-01 2.11768389e-01 7.14878619e-01 -7.84991562e-01 -1.27633834e+00 -4.07779455e-01 4.20462728e-01 -1.38133228e-01 6.61270991e-02 -4.53566581e-01 3.32195699e-01 7.63555944e-01 8.83500814e-01 -1.39478520e-01 -1.55459926e-01 5.83373070e-01 9.04961750e-02 1.25075877e-01 -1.23665476e+00 -4.68687534e-01 6.98774755e-02 4.03003395e-02 -4.70359832e-01 -5.87615073e-01 -5.70094407e-01 -1.38238573e+00 4.09965694e-01 -5.67383289e-01 3.04286480e-01 2.91180581e-01 9.23802555e-01 2.47898862e-01 4.04272258e-01 7.49016583e-01 -5.03219306e-01 -5.29988647e-01 -9.68663156e-01 -8.61650407e-01 7.74660408e-01 4.90097791e-01 -8.42780471e-01 -1.27524778e-01 6.07298017e-01]
[9.2923002243042, 1.3538254499435425]
3c881f3f-b8dd-4c16-9884-16cfbdfb30a3
automatic-generation-of-socratic-subquestions
2211.12835
null
https://arxiv.org/abs/2211.12835v1
https://arxiv.org/pdf/2211.12835v1.pdf
Automatic Generation of Socratic Subquestions for Teaching Math Word Problems
Socratic questioning is an educational method that allows students to discover answers to complex problems by asking them a series of thoughtful questions. Generation of didactically sound questions is challenging, requiring understanding of the reasoning process involved in the problem. We hypothesize that such questioning strategy can not only enhance the human performance, but also assist the math word problem (MWP) solvers. In this work, we explore the ability of large language models (LMs) in generating sequential questions for guiding math word problem-solving. We propose various guided question generation schemes based on input conditioning and reinforcement learning. On both automatic and human quality evaluations, we find that LMs constrained with desirable question properties generate superior questions and improve the overall performance of a math word problem solver. We conduct a preliminary user study to examine the potential value of such question generation models in the education domain. Results suggest that the difficulty level of problems plays an important role in determining whether questioning improves or hinders human performance. We discuss the future of using such questioning strategies in education.
['Mrinmaya Sachan', 'Manu Kapur', 'Tanmay Sinha', 'Mennatallah El-Assady', 'Jakub Macina', 'Kumar Shridhar']
2022-11-23
null
null
null
null
['math-word-problem-solving', 'question-generation', 'math-word-problem-solving', 'math-word-problem-solving']
['knowledge-base', 'natural-language-processing', 'reasoning', 'time-series']
[ 1.55257151e-01 3.30539674e-01 4.24176693e-01 -3.27597678e-01 -5.63126683e-01 -9.59695876e-01 2.17699662e-01 5.95418096e-01 -2.23734871e-01 5.10986686e-01 1.55761734e-01 -1.04264116e+00 -5.04304588e-01 -1.32899761e+00 -4.98618364e-01 1.40898392e-01 2.97028959e-01 2.86202371e-01 4.22274172e-01 -7.72247314e-01 6.77144527e-01 4.66716975e-01 -1.72445691e+00 5.81945062e-01 1.61579657e+00 3.65708679e-01 6.64790571e-01 1.06410575e+00 -7.62052715e-01 1.57620180e+00 -9.89199340e-01 -2.51144528e-01 -6.54332265e-02 -9.75355029e-01 -1.43388641e+00 -1.66975364e-01 5.78280032e-01 -2.97512352e-01 1.84024677e-01 8.58150780e-01 3.46748292e-01 6.67542458e-01 3.23314637e-01 -1.00375581e+00 -9.42383111e-01 5.70894003e-01 1.86419547e-01 4.63693887e-01 1.09952378e+00 3.22618455e-01 9.73854005e-01 -5.45433700e-01 3.14061075e-01 1.13618851e+00 2.99892724e-01 6.77359283e-01 -9.44081187e-01 -4.59595084e-01 8.40611383e-02 4.73664939e-01 -1.00356174e+00 1.04206324e-01 5.99518955e-01 -6.06751561e-01 6.65707409e-01 4.83200729e-01 1.08996999e+00 2.93555439e-01 4.71755832e-01 6.41732633e-01 1.27189481e+00 -6.86293244e-01 2.83126444e-01 5.20664275e-01 4.07387286e-01 9.10439193e-01 1.34135023e-01 -7.24083856e-02 -6.23385012e-01 -6.46058321e-02 8.07305753e-01 -4.42344040e-01 -1.26826182e-01 1.19029753e-01 -7.33957171e-01 1.04918218e+00 6.91884458e-02 3.99554551e-01 -1.05274573e-01 -6.65746108e-02 -1.78581834e-01 7.48361647e-01 -2.70849645e-01 1.64497817e+00 -5.75433314e-01 -3.51365328e-01 -8.96075130e-01 8.11315775e-01 1.10542822e+00 8.56570125e-01 4.54631627e-01 -3.32561359e-02 -6.26563251e-01 5.89330792e-01 4.43015069e-01 4.00050521e-01 5.28566837e-01 -1.19884038e+00 3.25802028e-01 1.16710007e+00 4.01749760e-02 -9.83670771e-01 -3.56840901e-02 -2.50349402e-01 1.54872015e-01 2.37861276e-01 6.96762979e-01 -5.47278404e-01 -7.10437298e-01 1.54875422e+00 3.30154389e-01 1.60973556e-02 -7.44723752e-02 6.59772396e-01 1.33162665e+00 9.00949359e-01 4.11626458e-01 5.61163910e-02 1.66168296e+00 -8.42457771e-01 -8.42716157e-01 -1.36034235e-01 7.72516727e-01 -1.04077709e+00 1.51902223e+00 5.48554778e-01 -1.51847386e+00 -9.84885395e-01 -7.08843172e-01 -2.42264196e-01 -2.70058334e-01 -1.07449442e-01 4.39074159e-01 9.58051145e-01 -9.58838165e-01 4.48852241e-01 -2.72556752e-01 -1.61005586e-01 -5.15748700e-03 2.72174120e-01 3.47239971e-01 -1.09041601e-01 -1.35883880e+00 8.56661141e-01 5.68517148e-02 -3.62120271e-01 -6.99561954e-01 -1.10955238e+00 -9.07060444e-01 3.75010729e-01 3.67990553e-01 -5.83133876e-01 1.53873610e+00 -5.84105015e-01 -1.81355536e+00 5.42827070e-01 1.51878148e-01 1.34964257e-01 8.77381340e-02 -4.30837154e-01 1.22757375e-01 1.61724672e-01 8.01257938e-02 8.45104754e-01 2.89473444e-01 -9.45850968e-01 -5.57036459e-01 4.82406691e-02 8.16433728e-01 5.33283055e-01 -3.49432528e-01 -3.72648872e-02 1.86813384e-01 -4.90482777e-01 9.52558815e-02 -6.00732803e-01 -4.88017410e-01 -9.54356045e-02 2.00740486e-01 -7.01357901e-01 3.58010113e-01 -6.99491680e-01 1.40744007e+00 -1.60782611e+00 -2.67535120e-01 3.73799205e-01 1.77178413e-01 3.70849669e-01 -4.15735871e-01 6.70477271e-01 -4.06221896e-02 2.64942259e-01 4.10953164e-01 6.60995483e-01 1.64288327e-01 1.77663285e-02 -2.38143146e-01 -4.61463243e-01 1.29174218e-01 1.06518579e+00 -1.13398933e+00 -5.27346313e-01 2.42209136e-01 -1.28706470e-01 -1.04947424e+00 1.09253287e+00 -6.61811233e-01 3.56147230e-01 -6.73236012e-01 2.67338604e-01 3.00703436e-01 -2.79702872e-01 -6.28896132e-02 4.65063632e-01 8.71349797e-02 6.76824927e-01 -1.47214460e+00 1.45612240e+00 -6.89092934e-01 6.07025027e-01 -3.57575685e-01 -5.12046814e-01 1.12590182e+00 2.44528666e-01 -1.24940112e-01 -8.79144073e-01 2.71010809e-02 -1.74636945e-01 4.28544253e-01 -1.12275529e+00 8.22146535e-01 -3.68067145e-01 2.74567097e-01 6.76588535e-01 9.59228352e-02 -9.37035859e-01 5.48713744e-01 3.90792489e-01 9.31573510e-01 2.98969567e-01 1.60303622e-01 -4.88413155e-01 7.65037775e-01 6.63153008e-02 -9.68532711e-02 1.05337977e+00 -1.07987404e-01 1.56586528e-01 3.33586216e-01 -1.32624120e-01 -3.82880688e-01 -8.93524826e-01 3.07450682e-01 1.48173881e+00 4.29568696e-04 -6.15566134e-01 -9.81489360e-01 -2.71421194e-01 -3.49902004e-01 1.24383557e+00 -2.19243154e-01 -1.87051356e-01 -5.40690005e-01 -2.20200598e-01 3.19071174e-01 4.51781958e-01 3.56160343e-01 -1.41548777e+00 -9.05551732e-01 3.48577261e-01 -3.70506883e-01 -9.10205960e-01 -3.11776072e-01 -1.08394474e-01 -1.03307927e+00 -9.77407277e-01 -4.49789733e-01 -1.13812733e+00 9.34571028e-01 3.39688510e-01 1.27047908e+00 8.70189250e-01 -9.64005068e-02 9.85425830e-01 -5.62120080e-01 -4.04837668e-01 -5.80427349e-01 -1.12220431e-02 -4.11561519e-01 -8.27724576e-01 4.29239541e-01 -3.40375602e-01 -6.88205421e-01 1.40947163e-01 -1.06842327e+00 3.21902394e-01 1.93815798e-01 3.17402154e-01 2.80383602e-03 3.04902554e-01 5.30717313e-01 -8.15473139e-01 1.62253737e+00 -3.11089367e-01 -6.29023373e-01 4.30378973e-01 -5.43363690e-01 3.94549161e-01 6.94252968e-01 -4.56296802e-01 -1.15703285e+00 -5.39390862e-01 -5.33474684e-01 3.05477411e-01 -3.74676585e-01 7.51461744e-01 1.28128394e-01 -4.31965023e-01 9.98425901e-01 3.01990777e-01 -1.68854415e-01 -9.40112621e-02 2.94761539e-01 9.23946947e-02 -4.99933548e-02 -1.39808583e+00 1.01599693e+00 -5.23950458e-01 -2.21228793e-01 -7.67308235e-01 -9.14448202e-01 -3.24311256e-01 -7.05565736e-02 -5.59577167e-01 7.89848089e-01 -5.74460983e-01 -1.27514291e+00 -1.09483480e-01 -1.08638620e+00 -7.80943632e-01 -4.93513405e-01 2.83347487e-01 -2.32415244e-01 1.35131896e-01 -6.86187685e-01 -8.21922839e-01 -4.89474833e-03 -1.34144902e+00 3.67192656e-01 8.84151638e-01 -8.21487546e-01 -1.12893176e+00 1.63596496e-01 1.15217793e+00 4.80656058e-01 -1.65973812e-01 1.48361790e+00 -7.53944159e-01 -7.67029524e-01 1.98242217e-01 1.94268152e-01 2.55431861e-01 -3.56015489e-02 -2.02097535e-01 -5.66238105e-01 1.18165731e-01 3.20579410e-01 -5.72893858e-01 -5.68702072e-02 -7.32282130e-03 1.05274177e+00 -3.78165543e-01 4.04553354e-01 -4.94805761e-02 1.29832649e+00 1.85860917e-01 6.59327507e-01 2.99287200e-01 3.42870355e-01 1.09317994e+00 6.24045134e-01 9.29721445e-02 5.08796155e-01 1.68159813e-01 -1.56741038e-01 3.32857400e-01 -1.48836225e-01 -4.77408320e-01 4.53796417e-01 9.83585775e-01 -1.13806240e-02 -5.77470995e-02 -1.18441153e+00 6.78604007e-01 -1.26356232e+00 -8.41755331e-01 -4.51215178e-01 1.80520773e+00 1.22261500e+00 3.96822020e-02 -1.19267002e-01 6.26449212e-02 1.33601248e-01 -2.57401943e-01 1.16230704e-01 -8.10561895e-01 5.97181439e-01 1.01990473e+00 -2.70631671e-01 1.03742671e+00 4.30489406e-02 1.00957739e+00 6.39668369e+00 6.97395563e-01 -6.08989298e-01 -3.26270401e-01 6.03545964e-01 3.15608621e-01 -7.86881268e-01 2.22516712e-02 -7.70741343e-01 -4.03352976e-02 1.11021090e+00 -3.67371708e-01 4.13980097e-01 6.21647120e-01 4.80797648e-01 -4.07107860e-01 -1.08175528e+00 3.53021979e-01 3.64135057e-02 -1.28934300e+00 4.01578039e-01 -2.35370532e-01 9.73996997e-01 -1.05308020e+00 -9.67614129e-02 5.29689193e-01 5.70034564e-01 -1.31918252e+00 4.59478408e-01 4.78487879e-01 -1.59008130e-02 -8.76395226e-01 5.32094479e-01 5.11327565e-01 -9.98741210e-01 -7.69897252e-02 -2.49644682e-01 -6.86794519e-01 -1.86084732e-01 -4.55283895e-02 -1.11262524e+00 1.50781810e-01 3.02024543e-01 -3.48461986e-01 -9.85877573e-01 1.10733712e+00 -7.20027328e-01 9.55793083e-01 5.92242293e-02 -7.71977425e-01 1.99983254e-01 -2.00822920e-01 1.18700951e-01 7.99581170e-01 4.17086661e-01 1.23038459e+00 3.74350786e-01 1.15464365e+00 3.43332857e-01 6.80606246e-01 -1.36613771e-01 -1.98312625e-01 3.43051434e-01 9.49009180e-01 -7.88030982e-01 -9.34298933e-02 -2.82397568e-01 5.96243203e-01 1.64228514e-01 3.83295357e-01 -4.96553332e-01 -4.72774237e-01 3.67883384e-01 5.88026583e-01 -1.08838595e-01 -4.25109565e-01 -5.32937706e-01 -8.64547610e-01 -1.42119050e-01 -1.57058096e+00 1.29849792e-01 -1.13114095e+00 -9.26349342e-01 2.51320064e-01 5.95219480e-03 -7.54077017e-01 -1.19398713e-01 -8.27176511e-01 -1.08981562e+00 1.07405996e+00 -1.25628674e+00 -4.64218497e-01 -6.07132614e-01 5.08248031e-01 7.22853839e-01 -6.47012964e-02 7.50177383e-01 -7.92035982e-02 -3.00479531e-02 4.69565362e-01 -7.28854537e-01 -8.10807720e-02 5.14319837e-01 -1.49035406e+00 1.19513772e-01 7.05868900e-01 1.54308140e-01 9.01428223e-01 9.79342818e-01 -6.57996356e-01 -1.51682794e+00 -5.95032692e-01 1.13817787e+00 -5.44900179e-01 4.58684564e-01 3.90168130e-02 -1.11341608e+00 1.57631591e-01 5.78980446e-01 -7.14784086e-01 1.41641891e+00 -4.06047180e-02 2.26958215e-01 3.08259279e-01 -1.09583306e+00 8.38759661e-01 5.66795588e-01 -3.23336422e-01 -1.38234890e+00 3.40342432e-01 8.09365332e-01 -5.42977929e-01 -8.86425436e-01 -9.16104838e-02 8.76412913e-02 -7.96133339e-01 9.36206639e-01 -9.57939684e-01 1.02491331e+00 -1.94740608e-01 1.60723373e-01 -1.07506204e+00 -3.41811270e-01 -6.29218638e-01 4.37860861e-02 1.10184693e+00 4.91550982e-01 -1.34376541e-01 9.94726717e-01 1.13671017e+00 1.32980824e-01 -7.28897810e-01 -6.26454577e-02 -4.10066009e-01 5.68186462e-01 -5.00095665e-01 6.06741250e-01 1.02005446e+00 3.76831234e-01 4.30600762e-01 2.76151478e-01 2.14370176e-01 1.45660460e-01 8.35908651e-02 6.86013818e-01 -1.18238866e+00 -4.79952604e-01 -3.51942599e-01 4.88257147e-02 -1.43475950e+00 -7.84787685e-02 -7.33822405e-01 2.36575991e-01 -1.74412405e+00 -2.32341990e-01 -2.15796486e-01 9.67924744e-02 1.91116944e-01 -8.59427154e-01 -3.72973859e-01 3.88076484e-01 -5.28305709e-01 -5.98408639e-01 2.19780192e-01 1.86324131e+00 3.13839763e-01 -3.94864917e-01 7.44521841e-02 -1.12682629e+00 7.43308663e-01 8.62480402e-01 -3.78139466e-01 -1.02383983e+00 -3.88181657e-01 7.91624606e-01 4.77363706e-01 1.39249533e-01 -1.11967754e+00 5.42202592e-01 -9.58704650e-01 4.18748647e-01 -1.23030163e-01 -3.10256898e-01 -5.84126890e-01 -5.30068815e-01 5.58962464e-01 -8.44685733e-01 4.26243097e-01 6.88417852e-01 -1.16085067e-01 -2.15774700e-01 -1.11311352e+00 5.36790431e-01 -4.68558788e-01 -4.27668333e-01 -2.54605651e-01 -9.87409890e-01 5.12776196e-01 8.88064742e-01 -2.26814836e-01 3.83632705e-02 -7.76484609e-01 -5.79078257e-01 6.15844846e-01 -2.52551258e-01 3.99572849e-01 8.77146840e-01 -1.20142984e+00 -7.37779796e-01 6.95283059e-04 -7.33681172e-02 -6.98475689e-02 6.12341277e-02 9.63086355e-03 -1.18586648e+00 6.13270223e-01 -2.15040803e-01 -1.16327502e-01 -1.46296155e+00 1.86167508e-01 3.89780492e-01 -5.71768045e-01 -6.84187263e-02 1.18635201e+00 -5.81535175e-02 -7.96015263e-01 3.11366439e-01 -6.33006871e-01 -7.13499248e-01 -7.66078383e-02 9.76337373e-01 3.97410840e-01 -1.20196633e-01 2.96630561e-01 3.95281911e-01 2.54660279e-01 -1.86124355e-01 -2.63702005e-01 1.02345836e+00 6.20266907e-02 4.53581028e-02 4.42297533e-02 4.99275178e-01 2.02724725e-01 -4.66771513e-01 4.35217135e-02 9.85780433e-02 -3.85696709e-01 -2.62332559e-01 -1.16903269e+00 -4.94584799e-01 7.87461340e-01 4.85058337e-01 3.84378284e-01 8.54265511e-01 -2.40346730e-01 6.41323268e-01 8.43452096e-01 1.74226046e-01 -9.39494133e-01 7.48604357e-01 7.46153116e-01 8.99810374e-01 -1.03558648e+00 -1.29793733e-01 -5.62228441e-01 -3.10358763e-01 1.39201713e+00 1.48080242e+00 -1.23677012e-02 2.64881819e-01 -1.80533156e-02 2.05589071e-01 -4.39537853e-01 -7.98720002e-01 -2.15459242e-02 4.68708336e-01 2.41124392e-01 1.20263505e+00 -2.57293493e-01 -6.49352074e-01 7.03826010e-01 -8.19974124e-01 3.20853084e-01 9.22561049e-01 1.26823211e+00 -1.03475559e+00 -1.42393947e+00 -8.92829359e-01 3.41569930e-01 -2.67050445e-01 -6.02305174e-01 -6.16801620e-01 3.87927562e-01 2.18964934e-01 1.38344920e+00 -3.91793787e-01 -5.02084009e-02 6.14209950e-01 3.80492955e-01 8.51036370e-01 -1.19921768e+00 -1.34246552e+00 -7.16363549e-01 -1.25066251e-01 -2.31759086e-01 6.97757602e-02 -2.57264704e-01 -1.39927387e+00 -4.50041860e-01 -3.82309228e-01 7.88928211e-01 2.24522069e-01 1.24854064e+00 3.00435559e-03 8.37733805e-01 1.39814615e-01 1.48062229e-01 -8.53121102e-01 -8.25503528e-01 -1.28879091e-02 2.62867570e-01 -4.25699316e-02 -1.26899585e-01 -1.32104903e-01 1.66322231e-01]
[10.838167190551758, 7.698014259338379]
c33f51b4-a2e0-4606-b0fd-a363a042e940
neural-attentive-circuits
2210.08031
null
https://arxiv.org/abs/2210.08031v2
https://arxiv.org/pdf/2210.08031v2.pdf
Neural Attentive Circuits
Recent work has seen the development of general purpose neural architectures that can be trained to perform tasks across diverse data modalities. General purpose models typically make few assumptions about the underlying data-structure and are known to perform well in the large-data regime. At the same time, there has been growing interest in modular neural architectures that represent the data using sparsely interacting modules. These models can be more robust out-of-distribution, computationally efficient, and capable of sample-efficient adaptation to new data. However, they tend to make domain-specific assumptions about the data, and present challenges in how module behavior (i.e., parameterization) and connectivity (i.e., their layout) can be jointly learned. In this work, we introduce a general purpose, yet modular neural architecture called Neural Attentive Circuits (NACs) that jointly learns the parameterization and a sparse connectivity of neural modules without using domain knowledge. NACs are best understood as the combination of two systems that are jointly trained end-to-end: one that determines the module configuration and the other that executes it on an input. We demonstrate qualitatively that NACs learn diverse and meaningful module configurations on the NLVR2 dataset without additional supervision. Quantitatively, we show that by incorporating modularity in this way, NACs improve upon a strong non-modular baseline in terms of low-shot adaptation on CIFAR and CUBs dataset by about 10%, and OOD robustness on Tiny ImageNet-R by about 2.5%. Further, we find that NACs can achieve an 8x speedup at inference time while losing less than 3% performance. Finally, we find NACs to yield competitive results on diverse data modalities spanning point-cloud classification, symbolic processing and text-classification from ASCII bytes, thereby confirming its general purpose nature.
['Li Erran Li', 'Nicolas Ballas', 'Bernhard Schölkopf', 'Yoshua Bengio', 'Chris Pal', 'Francesco Locatello', 'Martin Weiss', 'Nasim Rahaman']
2022-10-14
null
null
null
null
['point-cloud-classification']
['computer-vision']
[ 8.87052715e-03 1.25944570e-01 -1.59413666e-01 -4.19364750e-01 -5.41757464e-01 -6.65027738e-01 4.96351063e-01 -5.02981097e-02 -2.83413947e-01 3.84838462e-01 1.05570674e-01 -2.44761929e-01 -5.57130240e-02 -6.94255471e-01 -1.26390707e+00 -4.95815098e-01 -8.92936215e-02 4.69639510e-01 2.07292214e-01 -2.47956008e-01 1.09665636e-02 4.18504834e-01 -1.58143222e+00 5.36443293e-01 6.48725271e-01 9.87985909e-01 2.39168853e-01 6.62023246e-01 9.08616781e-02 7.78364956e-01 -3.44142765e-01 -2.12609336e-01 3.28981936e-01 2.83531584e-02 -8.69601488e-01 1.21461749e-01 6.66533172e-01 -6.80401698e-02 -5.37095547e-01 6.83109820e-01 4.31905925e-01 -1.96088036e-03 8.50475848e-01 -1.07138038e+00 -4.40926701e-01 6.34761631e-01 -6.90119386e-01 2.42315143e-01 -1.84833616e-01 4.72210467e-01 1.05506885e+00 -9.28249776e-01 6.32427096e-01 1.17534304e+00 7.22006679e-01 5.45749366e-01 -1.78749919e+00 -5.40461957e-01 3.43563110e-01 -1.15767360e-01 -1.39461923e+00 -6.62866831e-01 4.41293955e-01 -6.26176119e-01 1.36123514e+00 -5.11413030e-02 2.44770721e-01 9.87990379e-01 1.28124192e-01 7.58490562e-01 5.58866680e-01 -6.55620638e-03 3.59700888e-01 -1.55576572e-01 2.41310634e-02 6.89539611e-01 3.12066793e-01 -3.07884395e-01 -5.66983879e-01 -1.01634778e-01 9.77842569e-01 3.26269120e-02 2.69336198e-02 -6.91221952e-01 -1.21755636e+00 7.31622934e-01 7.65868664e-01 2.56309241e-01 -1.47098899e-01 5.33280194e-01 5.79848170e-01 3.86859477e-01 8.30555409e-02 4.43652540e-01 -7.17702448e-01 1.89454868e-01 -8.28627586e-01 6.34442791e-02 8.60409677e-01 9.90161777e-01 8.48531783e-01 1.90028921e-01 1.92184135e-01 8.97412777e-01 2.14489430e-01 3.84353518e-01 4.55627143e-01 -9.41592157e-01 7.69485176e-01 8.22873712e-01 -4.75704312e-01 -7.47713089e-01 -6.77600443e-01 -5.69600880e-01 -1.10551643e+00 1.96342841e-01 1.54135063e-01 -2.91434646e-01 -1.11065841e+00 2.13619399e+00 -2.03183983e-02 -1.09755360e-02 3.68605889e-02 6.77335501e-01 7.07119703e-01 6.03225231e-01 1.98753789e-01 5.53958535e-01 1.33132482e+00 -7.37454891e-01 1.59195289e-01 -5.75456560e-01 5.23524880e-01 -3.89188915e-01 8.72185707e-01 3.33258092e-01 -1.38174582e+00 -7.08005846e-01 -1.22711575e+00 -9.99535993e-02 -2.97285527e-01 9.76853296e-02 7.58675039e-01 2.87002355e-01 -1.31907237e+00 6.67883277e-01 -1.06662667e+00 -6.03716373e-01 7.50039220e-01 7.44973719e-01 -6.73047900e-01 -8.36407095e-02 -7.08757460e-01 8.11750531e-01 4.75568891e-01 -3.15031916e-01 -1.13451827e+00 -8.93239200e-01 -9.17588055e-01 3.72550070e-01 8.62748772e-02 -1.27524149e+00 1.22988963e+00 -1.42105210e+00 -1.16857791e+00 8.26326728e-01 -1.76133364e-01 -6.65610194e-01 -5.42078875e-02 1.24475963e-01 -1.25471398e-01 3.79393697e-01 -1.51045993e-01 1.14700425e+00 1.00233960e+00 -1.21462965e+00 -4.67745751e-01 -2.14214310e-01 3.11469644e-01 3.45640481e-01 -4.62771803e-01 -2.71073461e-01 -5.86770952e-01 -5.37093401e-01 2.45331973e-01 -7.84528196e-01 -2.64667422e-01 2.82360107e-01 -2.87005931e-01 1.68542191e-02 4.61012453e-01 -4.54770625e-01 7.88870811e-01 -2.30652118e+00 6.29119098e-01 1.88402966e-01 3.14811110e-01 5.61766140e-02 -3.09701025e-01 4.28169191e-01 -3.35668951e-01 1.19210973e-01 -4.51485366e-01 -5.03750741e-01 -2.04086471e-02 2.68004507e-01 -2.88858652e-01 4.00626212e-01 7.08942711e-01 1.10080612e+00 -5.10270596e-01 -2.06822366e-01 5.87777793e-02 3.89436632e-01 -9.33816671e-01 1.20898277e-01 -2.50344664e-01 1.00293964e-01 7.38719627e-02 7.56086290e-01 4.41245526e-01 -6.26971781e-01 3.49562645e-01 -3.95182252e-01 3.06858897e-01 2.92277902e-01 -1.14575911e+00 2.13164043e+00 -6.02773547e-01 7.36620307e-01 4.14844990e-01 -1.27608132e+00 5.28033018e-01 3.07922602e-01 4.35055256e-01 -5.38273156e-01 1.27846241e-01 4.46291789e-02 3.02198082e-01 -2.98617065e-01 1.86020777e-01 -1.26971379e-01 -1.35691658e-01 3.91433895e-01 7.66327679e-01 1.72723502e-01 2.48010635e-01 4.23342019e-01 1.50285053e+00 -1.40232831e-01 1.71786025e-01 -3.48089933e-01 1.51439816e-01 -7.76696876e-02 4.02236015e-01 7.22989559e-01 2.33032018e-01 8.09240341e-01 5.05459487e-01 -2.56468594e-01 -1.20079231e+00 -1.35910213e+00 -2.82479197e-01 1.28903782e+00 3.95128205e-02 -2.27595896e-01 -6.20907426e-01 -3.92440170e-01 2.12170795e-01 2.66373336e-01 -6.00320876e-01 -3.01579684e-01 -4.41753596e-01 -8.50060403e-01 6.75901055e-01 9.34410214e-01 5.98971009e-01 -7.29587913e-01 -4.18149441e-01 2.90556341e-01 -1.38903465e-02 -1.04399085e+00 -2.52307743e-01 6.79602087e-01 -1.32880867e+00 -1.03198767e+00 -5.26967406e-01 -9.52875853e-01 1.01807034e+00 3.93416971e-01 1.50215006e+00 8.41758847e-02 -3.03650826e-01 4.92188841e-01 2.05721818e-02 -3.44562791e-02 -7.70235807e-02 6.15762591e-01 3.89503576e-02 -1.15831181e-01 1.21542677e-01 -1.09814966e+00 -7.10519254e-01 4.08257872e-01 -1.14937067e+00 1.97859153e-01 9.91160810e-01 8.51054847e-01 2.58341283e-01 -2.16634512e-01 6.16691589e-01 -8.82551134e-01 1.60600960e-01 -9.62822378e-01 -2.90645033e-01 5.17316647e-02 -2.09734321e-01 3.34617853e-01 7.71614611e-01 -4.83490825e-01 -7.72635698e-01 3.38578135e-01 -5.43627217e-02 -4.11496043e-01 -4.05083507e-01 5.26652694e-01 -3.31176728e-01 -1.98280334e-01 9.98661280e-01 1.10245727e-01 -1.00508988e-01 -5.27390778e-01 3.92393559e-01 2.50605673e-01 6.69345975e-01 -8.35802734e-01 1.05912077e+00 7.63101280e-01 1.48086809e-02 -6.82872951e-01 -5.37115633e-01 -4.08458412e-01 -7.29912102e-01 3.46671045e-01 6.86108649e-01 -1.41981626e+00 -4.87019390e-01 4.02597338e-01 -9.59996521e-01 -5.06153643e-01 -2.57282048e-01 -2.44489517e-02 -4.52505708e-01 1.85800895e-01 -7.14383006e-01 -3.05292934e-01 -3.92164350e-01 -8.45341265e-01 9.90394175e-01 2.14607984e-01 -1.78763777e-01 -1.06317699e+00 -6.26299949e-03 2.30318844e-01 5.38244247e-01 -1.49305156e-02 1.05772972e+00 -4.00836825e-01 -7.91433632e-01 1.12332860e-02 -3.51370841e-01 3.16264212e-01 -2.40781605e-01 -1.42383561e-01 -1.26892352e+00 -5.14359832e-01 -4.00483668e-01 -4.83173668e-01 1.22780836e+00 2.25906104e-01 1.26821291e+00 -2.19465166e-01 -3.24281305e-01 9.85597670e-01 1.57775760e+00 -3.61864954e-01 6.21836662e-01 3.05788983e-02 6.39345586e-01 3.78116757e-01 -1.39254928e-01 1.34029433e-01 4.93342251e-01 5.11651695e-01 7.31442630e-01 -2.75115728e-01 -1.81036100e-01 -1.74388587e-01 5.31717122e-01 7.10881114e-01 6.12570755e-02 8.40010867e-02 -8.01284611e-01 7.18176961e-01 -2.04801989e+00 -9.49527919e-01 3.56745631e-01 1.91568732e+00 8.78075778e-01 2.03744754e-01 1.59303412e-01 -1.27873436e-01 4.47006017e-01 -8.62264484e-02 -9.96152937e-01 -3.14290524e-01 -3.33595037e-01 3.59161824e-01 5.57665646e-01 1.09670155e-01 -1.27072966e+00 7.76367605e-01 6.21624994e+00 6.20841980e-01 -1.19868994e+00 1.83624759e-01 7.06310332e-01 -3.68694991e-01 -1.93685442e-01 -1.43707134e-02 -6.85112715e-01 2.55276710e-01 1.00323606e+00 1.81091264e-01 6.85316026e-01 8.84989500e-01 -2.76727259e-01 -1.03258066e-01 -1.46021807e+00 1.06804371e+00 1.25280723e-01 -1.57613146e+00 -3.52539904e-02 1.09953985e-01 8.94326866e-01 6.08485937e-01 1.44452855e-01 5.03314257e-01 3.35554689e-01 -1.32512665e+00 7.31977940e-01 4.85611290e-01 6.81686521e-01 -5.85904598e-01 5.48737645e-01 4.95387942e-01 -1.29251051e+00 -2.73402989e-01 -6.77319646e-01 -6.59343600e-02 -2.55633324e-01 3.56425434e-01 -6.32115006e-01 3.26243639e-01 6.90738440e-01 6.60032809e-01 -9.17410433e-01 1.28764999e+00 -9.24114883e-02 6.14070117e-01 -4.97003794e-01 3.34320992e-01 2.36925893e-02 3.34444672e-01 2.39190936e-01 1.26381576e+00 1.66789979e-01 -1.68891758e-01 -6.85264841e-02 9.96046126e-01 -5.53924561e-01 -2.39854857e-01 -6.61364913e-01 1.17332123e-01 3.29655588e-01 1.45686591e+00 -5.25881469e-01 -1.79780096e-01 -4.89093751e-01 9.43076134e-01 7.42556095e-01 4.07250851e-01 -6.67568088e-01 -1.85495153e-01 9.15360093e-01 1.85036227e-01 7.01175094e-01 -2.87280768e-01 -5.83116055e-01 -1.43006468e+00 2.81710979e-02 -8.03636372e-01 3.52516085e-01 -6.83325350e-01 -1.46525574e+00 3.61647010e-01 -9.57003310e-02 -1.10634649e+00 -1.97439909e-01 -9.13703501e-01 -8.81068528e-01 7.92445540e-01 -1.31112731e+00 -1.16313183e+00 -1.05531015e-01 6.45832121e-01 3.75103325e-01 -3.00360650e-01 8.72789919e-01 3.92610401e-01 -6.80096686e-01 6.62814200e-01 1.50240421e-01 3.78272653e-01 5.45181990e-01 -1.24905086e+00 6.26583517e-01 8.24223220e-01 2.68388361e-01 9.95586991e-01 3.24280530e-01 -2.34858617e-01 -1.68427229e+00 -1.23009944e+00 3.34133983e-01 -5.93319178e-01 6.16383851e-01 -6.83897793e-01 -9.28834677e-01 7.60634065e-01 2.70184070e-01 1.35808870e-01 6.28383696e-01 3.81008834e-01 -8.54796708e-01 -3.46715271e-01 -1.02395654e+00 5.76226115e-01 1.19916356e+00 -6.00224018e-01 -6.26440823e-01 1.43010423e-01 5.22259891e-01 -2.90838450e-01 -9.05117035e-01 3.73334229e-01 3.92515838e-01 -8.52157235e-01 1.11049402e+00 -7.75916636e-01 5.43718159e-01 -3.25946867e-01 -2.98824042e-01 -1.15822089e+00 -4.69133377e-01 -3.51252198e-01 -4.36144054e-01 1.08535182e+00 5.95547438e-01 -4.87031013e-01 7.25766361e-01 5.33695042e-01 -1.52224898e-01 -9.09368038e-01 -9.09260094e-01 -6.92392230e-01 3.96321714e-01 -4.46779341e-01 2.53130704e-01 7.13949621e-01 1.31629538e-02 7.56684601e-01 2.20301864e-03 2.48986885e-01 5.03632605e-01 1.01788625e-01 8.53062868e-01 -1.19991767e+00 -6.89739525e-01 -5.75105429e-01 -6.51855648e-01 -1.20003569e+00 9.97659937e-02 -1.16972411e+00 -1.53439745e-01 -1.46974647e+00 4.68993843e-01 -2.82463282e-01 -3.06187481e-01 7.03115582e-01 -5.05817421e-02 2.95370638e-01 2.55373985e-01 3.28119397e-01 -7.71237314e-01 3.53139937e-01 7.11137593e-01 -3.50533009e-01 4.80152220e-02 -2.04189256e-01 -1.05854547e+00 8.01497281e-01 8.09031367e-01 -2.37481475e-01 -4.15008307e-01 -8.79745960e-01 4.50391382e-01 -4.28314447e-01 6.57117069e-01 -1.34838545e+00 3.87640148e-01 1.35330364e-01 7.32178092e-01 -3.26623380e-01 4.77082044e-01 -8.98932278e-01 -5.43953888e-02 3.18393707e-01 -2.00369567e-01 6.29550740e-02 5.43990254e-01 5.28090656e-01 -1.96531564e-02 -1.79555804e-01 8.64887953e-01 -2.16439441e-01 -6.75101519e-01 4.04436857e-01 -3.90090555e-01 2.56263614e-01 7.90506244e-01 -1.37060359e-01 -5.04058421e-01 -3.42924267e-01 -5.20007670e-01 3.13843459e-01 6.08094275e-01 2.90855587e-01 4.51409370e-01 -1.20089769e+00 -5.29327333e-01 2.02621147e-01 1.36842385e-01 3.17051321e-01 2.63531208e-01 7.72056580e-01 -4.09543574e-01 3.13885510e-01 -1.70140088e-01 -8.86528909e-01 -8.34919751e-01 3.66233379e-01 5.38957179e-01 -1.67430863e-01 -4.19758976e-01 8.83008718e-01 5.76811492e-01 -6.58558965e-01 2.62653947e-01 -3.95402372e-01 1.73655421e-01 -1.40794516e-01 3.26952219e-01 -7.27663934e-02 1.07388467e-01 -4.37369853e-01 -3.53340834e-01 5.11101723e-01 4.19197716e-02 -5.40522039e-02 1.46002078e+00 1.59990773e-01 -6.11037686e-02 4.71047580e-01 1.17910326e+00 -3.51423502e-01 -1.54772186e+00 -3.75248969e-01 -1.71092022e-02 1.08501755e-01 -1.75577790e-01 -7.88636565e-01 -1.28832805e+00 1.20563221e+00 3.98262382e-01 -1.13364393e-02 1.10964370e+00 1.85034275e-01 6.39747381e-01 6.79061294e-01 2.20891118e-01 -1.09003878e+00 1.62963450e-01 7.41047859e-01 7.96640694e-01 -1.18357372e+00 -5.33352792e-02 -2.59517133e-02 -5.60654283e-01 1.09583247e+00 7.63979971e-01 -3.91059160e-01 6.58778191e-01 3.92448127e-01 -2.60884166e-01 -1.39875740e-01 -1.22752953e+00 -2.58938581e-01 5.33772528e-01 6.77074850e-01 2.86959112e-01 -1.08241461e-01 5.65409541e-01 7.09675968e-01 -6.23625480e-02 -3.33097875e-01 -2.72004437e-02 7.83671260e-01 -5.04381359e-01 -8.38811457e-01 -1.95042700e-01 6.34777129e-01 -2.14311972e-01 -2.29835078e-01 -3.06322157e-01 7.59257197e-01 2.24899545e-01 6.26597106e-01 2.38508180e-01 -3.89599949e-01 2.94789881e-01 2.42201492e-01 6.29567564e-01 -1.00050855e+00 -8.61477375e-01 -1.12380691e-01 -5.56348078e-02 -5.58630645e-01 -1.06210120e-01 -6.42063141e-01 -1.30351341e+00 -4.08663303e-01 2.16481630e-02 -3.59525800e-01 6.88809097e-01 1.00573897e+00 7.59952307e-01 6.09596372e-01 4.13617551e-01 -1.18140149e+00 -6.19815946e-01 -8.93793166e-01 -2.36984342e-01 3.67610335e-01 1.95319191e-01 -4.26262379e-01 -2.76334673e-01 4.12763394e-02]
[9.517500877380371, 1.6928352117538452]
99e14206-0628-4b68-84a9-3865d7bde9d4
audio-retrieval-with-natural-language-queries-1
2112.09418
null
https://arxiv.org/abs/2112.09418v2
https://arxiv.org/pdf/2112.09418v2.pdf
Audio Retrieval with Natural Language Queries: A Benchmark Study
The objectives of this work are cross-modal text-audio and audio-text retrieval, in which the goal is to retrieve the audio content from a pool of candidates that best matches a given written description and vice versa. Text-audio retrieval enables users to search large databases through an intuitive interface: they simply issue free-form natural language descriptions of the sound they would like to hear. To study the tasks of text-audio and audio-text retrieval, which have received limited attention in the existing literature, we introduce three challenging new benchmarks. We first construct text-audio and audio-text retrieval benchmarks from the AudioCaps and Clotho audio captioning datasets. Additionally, we introduce the SoundDescs benchmark, which consists of paired audio and natural language descriptions for a diverse collection of sounds that are complementary to those found in AudioCaps and Clotho. We employ these three benchmarks to establish baselines for cross-modal text-audio and audio-text retrieval, where we demonstrate the benefits of pre-training on diverse audio tasks. We hope that our benchmarks will inspire further research into audio retrieval with free-form text queries. Code, audio features for all datasets used, and the SoundDescs dataset are publicly available at https://github.com/akoepke/audio-retrieval-benchmark.
['Samuel Albanie', 'Zeynep Akata', 'João F. Henriques', 'Andreea-Maria Oncescu', 'A. Sophia Koepke']
2021-12-17
null
null
null
null
['audio-captioning']
['audio']
[ 1.44532472e-01 -7.07108617e-01 -2.01566685e-02 -1.48963362e-01 -1.86993933e+00 -9.10948575e-01 4.38603461e-01 2.63240188e-01 -2.90909857e-01 2.22596630e-01 6.22552156e-01 3.44961911e-01 -3.53900820e-01 -3.02299589e-01 -4.79145020e-01 -3.33242118e-01 -3.50322753e-01 4.57279950e-01 2.14740142e-01 -2.71369606e-01 1.38219908e-01 -1.07538491e-01 -2.20882082e+00 6.66750550e-01 9.99680012e-02 1.42240953e+00 2.14806959e-01 1.02651513e+00 9.62609276e-02 5.43397903e-01 -6.61696851e-01 -9.61767659e-02 3.72849926e-02 -2.66435385e-01 -9.87552404e-01 -5.46843588e-01 7.28997707e-01 -2.01322258e-01 -5.54492056e-01 6.89800858e-01 1.26616549e+00 4.78832185e-01 6.78370774e-01 -1.41847456e+00 -5.97698390e-01 7.46378064e-01 -1.15026578e-01 4.21901703e-01 1.20052636e+00 1.74393862e-01 1.74882293e+00 -1.00903141e+00 4.17919666e-01 1.29833615e+00 6.01360381e-01 5.91838658e-01 -1.07481945e+00 -1.05387962e+00 -2.90195465e-01 3.64001215e-01 -1.88061416e+00 -9.63191628e-01 6.54630423e-01 -3.96954566e-01 7.92561650e-01 6.26362503e-01 5.84948421e-01 1.21094525e+00 -1.98056966e-01 1.01114428e+00 4.36027050e-01 -3.24843436e-01 -6.09373115e-02 -1.22052930e-01 -5.62977418e-02 8.64764005e-02 -5.53635001e-01 1.55526027e-01 -1.36136591e+00 -4.31532085e-01 1.86077610e-01 -3.83468837e-01 -4.73054439e-01 1.26570359e-01 -1.21472931e+00 4.75875229e-01 3.57812410e-03 1.73296988e-01 -1.24621652e-01 4.81052697e-01 9.72212136e-01 7.41839767e-01 2.99635649e-01 4.91899163e-01 -2.94472635e-01 -5.77668607e-01 -8.95800054e-01 7.09870398e-01 7.95624673e-01 1.07670784e+00 4.91661608e-01 -3.02712284e-02 -2.48068124e-01 1.43331766e+00 2.77573854e-01 8.62662256e-01 7.19852686e-01 -1.13474500e+00 5.51332235e-01 -1.59608290e-01 -1.01968668e-01 -8.84129286e-01 -5.52082397e-02 1.18602939e-01 -3.46513629e-01 -5.35803676e-01 9.93816629e-02 2.18166456e-01 -5.29870629e-01 1.70303583e+00 2.47664866e-03 1.14546292e-01 -1.38449490e-01 1.13308680e+00 1.43685377e+00 1.04737842e+00 5.25027141e-02 1.09902889e-01 1.50935364e+00 -8.89396608e-01 -6.14562631e-01 1.46373350e-03 4.35346484e-01 -1.26542962e+00 1.64502716e+00 3.05729598e-01 -1.21242511e+00 -5.67047656e-01 -5.29790759e-01 -2.20428988e-01 -2.38143295e-01 -1.08080804e-01 2.83710271e-01 1.44017816e-01 -1.11258829e+00 2.76087463e-01 -5.00759423e-01 -3.39633524e-01 -2.18256526e-02 4.23450470e-02 -2.64769882e-01 2.70726770e-01 -1.55543518e+00 -1.88251734e-02 2.83235401e-01 -2.74715275e-01 -1.44089293e+00 -1.18789005e+00 -7.55455971e-01 2.52981365e-01 4.41081971e-01 -5.35855591e-01 1.89650059e+00 -6.39518738e-01 -1.28950334e+00 9.44985688e-01 7.93191884e-03 -1.49143800e-01 7.91255087e-02 -4.06252533e-01 -6.14192188e-01 6.81346834e-01 4.28336710e-01 1.02008641e+00 8.36283565e-01 -7.80704200e-01 -5.98600924e-01 8.98775607e-02 -5.42279966e-02 2.41345167e-01 -5.09029686e-01 4.67207283e-01 -8.51575673e-01 -9.92346942e-01 -2.17178926e-01 -9.91849244e-01 4.78450090e-01 1.31010085e-01 -5.61373949e-01 -3.70597988e-01 5.28928280e-01 -3.66058528e-01 1.53811967e+00 -2.56516433e+00 -1.47481963e-01 -1.74246002e-02 7.33302208e-03 -1.96176142e-01 -6.10839605e-01 9.57433522e-01 -1.70695066e-01 2.71492451e-01 2.35768586e-01 -2.88223147e-01 4.37068850e-01 -3.21163625e-01 -7.71241009e-01 -4.74345237e-02 -5.14104925e-02 6.74702168e-01 -9.43681896e-01 -8.62179577e-01 -2.04712838e-01 4.90886182e-01 -7.16288686e-01 3.26502502e-01 -3.78458589e-01 2.16244638e-01 -4.72122312e-01 8.34786594e-01 -7.78501928e-02 -3.95106897e-02 -5.58785796e-01 -4.40926515e-02 7.87077323e-02 8.68379056e-01 -1.19658005e+00 2.04047418e+00 -6.39220834e-01 9.37778592e-01 9.85088274e-02 -4.45272356e-01 5.65423071e-01 8.27236176e-01 7.17399478e-01 -7.50210762e-01 -9.69394520e-02 3.89974624e-01 -4.55697954e-01 -6.40623987e-01 6.96639478e-01 -1.63563028e-01 -4.53313351e-01 7.70598829e-01 2.40659371e-01 -6.01796508e-01 3.64684165e-01 4.61037874e-01 1.24342132e+00 -1.49394110e-01 -1.34659767e-01 2.50408836e-02 3.58802766e-01 -2.19735235e-01 -7.18070641e-02 9.50776160e-01 -4.78327908e-02 1.10182571e+00 4.84159961e-02 1.11292735e-01 -7.34028041e-01 -1.29580224e+00 -1.54678851e-01 1.97778451e+00 -1.39629915e-01 -1.14944291e+00 -1.97389051e-01 7.86430687e-02 2.64753923e-02 1.20427124e-01 -2.62443453e-01 -9.12879333e-02 -1.59955472e-01 8.58260021e-02 1.22633052e+00 3.21834624e-01 -3.91683318e-02 -1.39784741e+00 -3.30515921e-01 8.30234662e-02 -7.03850091e-01 -9.72093225e-01 -1.27584922e+00 3.45506556e-02 -4.00324464e-01 -7.93420434e-01 -9.08317626e-01 -9.44403946e-01 -2.51019984e-01 2.74921894e-01 1.45882916e+00 2.64127348e-02 -5.61240137e-01 1.12399173e+00 -7.49426484e-01 -4.71977025e-01 -3.79488409e-01 3.80391479e-01 2.10175768e-01 -2.11689800e-01 8.78570080e-02 -8.22263122e-01 -5.53512156e-01 4.22375768e-01 -1.13646555e+00 -3.44068944e-01 8.24155211e-02 6.56663299e-01 8.11906517e-01 -2.21007302e-01 7.86411166e-01 -2.34275505e-01 1.01008093e+00 -5.92156410e-01 -2.22239032e-01 -1.55130643e-02 -1.14644051e-01 -2.08155915e-01 4.82147783e-01 -7.70190120e-01 -3.57223630e-01 -1.56372711e-02 -3.70390713e-01 -7.22065032e-01 -7.57385045e-02 7.69560754e-01 8.69604424e-02 3.68324071e-01 7.92517066e-01 4.01747942e-01 -4.77859110e-01 -7.83592165e-01 1.88586652e-01 1.10330665e+00 7.94095457e-01 -1.08639157e+00 7.11279929e-01 1.10747591e-01 -4.86724615e-01 -1.04570186e+00 -8.21936786e-01 -1.09718704e+00 -3.36726569e-02 -3.85542125e-01 5.55627108e-01 -1.14321756e+00 -8.08990479e-01 9.01617631e-02 -1.03444111e+00 -2.34480098e-01 -5.03532708e-01 4.80335027e-01 -8.77850056e-01 2.26863459e-01 -5.72278261e-01 -8.60474885e-01 -6.56007171e-01 -8.78322840e-01 1.78792775e+00 -1.55398548e-01 -6.17505193e-01 -5.77328384e-01 3.24437082e-01 2.78695583e-01 3.13404590e-01 -2.55499512e-01 6.90486312e-01 -9.28421617e-01 -4.58871812e-01 -2.76378810e-01 1.30291894e-01 1.41052634e-01 -3.63137014e-02 -3.79711539e-02 -1.30734921e+00 -3.08841377e-01 -5.80381632e-01 -1.17751729e+00 9.81597126e-01 7.20453635e-03 1.50483572e+00 -2.55092680e-01 2.45806351e-02 3.71018797e-01 8.85385931e-01 6.18138276e-02 2.55495965e-01 1.65045723e-01 2.12247759e-01 4.87052619e-01 7.63611853e-01 6.73484206e-01 3.36576164e-01 1.02923787e+00 9.26552787e-02 2.95206696e-01 -2.82510966e-01 -5.85403979e-01 4.96794969e-01 1.47562242e+00 4.22102362e-01 -3.73521000e-01 -9.78793323e-01 9.09479618e-01 -1.54623771e+00 -1.09499311e+00 3.93416405e-01 2.04398298e+00 1.44315588e+00 -1.43975511e-01 3.73169005e-01 4.55378860e-01 5.15624702e-01 2.17799619e-01 -2.67526120e-01 1.58890523e-02 1.76081896e-01 4.15486395e-01 -2.73375541e-01 3.70611995e-01 -1.13695526e+00 5.94433129e-01 6.35051775e+00 1.23244929e+00 -1.07269776e+00 1.14384398e-01 -3.10283666e-03 -7.16667473e-01 -3.29071283e-01 -2.73566842e-01 -7.00893819e-01 4.10962313e-01 1.26344585e+00 -5.57755947e-01 7.06438661e-01 5.58122814e-01 2.62606382e-01 9.52671692e-02 -1.70049536e+00 1.42077279e+00 1.83677149e-03 -1.00203645e+00 3.48371416e-01 -4.80534971e-01 2.67628133e-01 1.75780997e-01 2.68976241e-01 5.79297185e-01 -1.62408695e-01 -8.46988976e-01 1.07953417e+00 4.79431838e-01 1.13767791e+00 -4.05269325e-01 1.48991913e-01 -1.16364814e-01 -1.73547149e+00 5.77447787e-02 -7.81096816e-02 2.35361919e-01 1.51602045e-01 2.17303619e-01 -6.89340413e-01 5.37963390e-01 1.24515247e+00 6.61750853e-01 -4.61548865e-01 1.42932808e+00 1.96551651e-01 8.25802982e-01 -6.73636258e-01 -1.20006770e-01 2.84835175e-02 5.27995765e-01 9.39870298e-01 1.54764700e+00 4.51523006e-01 -1.95429042e-01 3.00671667e-01 7.80624270e-01 -3.54936987e-01 5.61951697e-01 -5.71732402e-01 -6.12850308e-01 8.68836224e-01 8.78095150e-01 -2.08020136e-01 -1.05591975e-01 -1.30504265e-01 5.31776428e-01 -2.86488414e-01 2.68744379e-01 -6.50915623e-01 -8.63556564e-01 7.20322311e-01 2.55566031e-01 -1.77708897e-03 1.23073451e-01 5.44264138e-01 -9.85194385e-01 1.23072632e-01 -1.12343121e+00 7.00132608e-01 -1.29933619e+00 -1.51636493e+00 5.34400940e-01 3.30735952e-01 -1.54134607e+00 -6.39835835e-01 -1.67549923e-02 -3.25212240e-01 6.41147316e-01 -1.34134245e+00 -7.58782566e-01 -2.18856871e-01 7.53958702e-01 8.90682459e-01 -2.66912580e-01 9.74735498e-01 9.69340980e-01 7.60751814e-02 7.34874427e-01 -1.27891684e-02 1.48849770e-01 1.26743770e+00 -8.82940292e-01 2.73575336e-01 -5.96646518e-02 6.13139033e-01 5.69711864e-01 7.34400451e-01 -3.01417053e-01 -1.49265599e+00 -9.90360975e-01 8.97790790e-01 -5.04515231e-01 1.20096624e+00 -4.51446891e-01 -8.20755243e-01 4.43960607e-01 6.43548146e-02 -1.82647649e-02 9.34072435e-01 1.02782033e-01 -4.92230415e-01 -2.73297250e-01 -3.84180278e-01 3.81357253e-01 9.87156272e-01 -1.29394519e+00 -5.61033905e-01 4.64817017e-01 1.17173052e+00 -3.45320642e-01 -1.00467360e+00 1.65103838e-01 9.38604116e-01 -3.66700441e-01 1.18364763e+00 -6.10334575e-01 3.81295800e-01 -2.20493212e-01 -5.25760114e-01 -1.17417026e+00 2.92357832e-01 -1.02823973e+00 2.57093161e-01 1.59903800e+00 4.76113915e-01 -4.32137251e-02 3.17535430e-01 -1.42428637e-01 -2.56051302e-01 -3.09926897e-01 -1.03979158e+00 -1.05054545e+00 -1.18899584e-01 -1.09190547e+00 5.14844298e-01 7.31575847e-01 1.77289426e-01 4.70272720e-01 -3.79219711e-01 -1.13809295e-01 1.84830710e-01 2.84342635e-02 8.82025659e-01 -1.30758905e+00 -5.43816507e-01 -4.20766562e-01 -3.17167938e-01 -1.24395847e+00 3.02756488e-01 -1.28112817e+00 4.48944300e-01 -9.81472433e-01 2.17070684e-01 -3.34693491e-01 -2.51805037e-01 6.88183010e-01 2.41298273e-01 6.88508689e-01 4.96347189e-01 3.89543533e-01 -1.14124656e+00 7.54178643e-01 1.09023118e+00 -5.27905464e-01 -3.11461896e-01 9.44437459e-02 -5.74325204e-01 2.78360844e-01 5.35271347e-01 -8.73603821e-01 -5.27737677e-01 -4.45984781e-01 5.30936718e-01 5.04670978e-01 4.66872931e-01 -1.10194528e+00 4.45936292e-01 1.38926968e-01 -3.31702262e-01 -6.71886563e-01 8.32921982e-01 -6.69452190e-01 7.52454847e-02 -2.15149716e-01 -9.81508613e-01 2.16369569e-01 4.98379201e-01 5.28724492e-01 -7.60617077e-01 -4.12305087e-01 3.47589910e-01 4.15020064e-02 -3.20650607e-01 3.51226866e-01 -6.11666858e-01 7.29528308e-01 2.59042561e-01 1.82088077e-01 -1.70900524e-01 -1.16284049e+00 -9.42435861e-01 3.75309229e-01 -8.21138266e-03 7.83341765e-01 7.77962804e-01 -1.71682239e+00 -8.00090373e-01 -1.61317244e-01 6.70708716e-01 -2.36984640e-01 1.00409001e-01 6.09715283e-01 -2.28754386e-01 7.28970706e-01 7.93792084e-02 -7.12330639e-01 -1.46595669e+00 1.94318652e-01 1.96411461e-01 2.39357412e-01 -4.64776367e-01 7.94465244e-01 1.66843459e-01 -3.40760887e-01 9.43433106e-01 -2.73688227e-01 9.72131267e-02 1.70010537e-01 7.58595705e-01 2.84838200e-01 3.09658535e-02 -5.95685005e-01 -3.08639199e-01 5.05224764e-01 2.59927422e-01 -6.51553690e-01 1.20999789e+00 -1.82548910e-01 4.48719300e-02 9.53150094e-01 1.50229704e+00 1.89176753e-01 -3.96674693e-01 -3.30739886e-01 3.59656215e-02 -4.10591722e-01 -7.97030479e-02 -7.32974768e-01 -8.20441842e-01 9.87480342e-01 6.51823580e-01 4.57389563e-01 1.28893232e+00 4.83692855e-01 9.33561563e-01 7.86896408e-01 6.77830055e-02 -1.05716741e+00 6.43315077e-01 6.70370162e-01 1.25956070e+00 -9.71478343e-01 -2.87584364e-01 -1.30767822e-01 -4.24224019e-01 1.01757336e+00 3.61603349e-01 1.90529972e-01 7.96020210e-01 2.51673073e-01 1.26591608e-01 -3.22576433e-01 -1.28986323e+00 -4.87972796e-01 6.70609236e-01 4.81688619e-01 6.46994710e-01 -2.69742370e-01 1.89888760e-01 7.90905356e-01 -6.80214703e-01 -1.18267469e-01 1.70904413e-01 8.07426870e-01 -3.44094187e-01 -1.01972020e+00 -5.49251139e-01 2.81229615e-01 -7.56509900e-01 -3.48458529e-01 -7.75867760e-01 4.27333742e-01 -5.93621612e-01 1.16993916e+00 -2.85006972e-05 -5.14675200e-01 5.12767732e-01 3.58053476e-01 1.56567395e-01 -8.05202723e-01 -7.19164312e-01 4.00860786e-01 1.69582576e-01 -5.90283215e-01 -2.93477088e-01 -6.05839968e-01 -1.24141788e+00 1.33656068e-02 -3.38302225e-01 5.92182994e-01 5.23627460e-01 4.62336123e-01 2.61622906e-01 5.29212892e-01 5.18040121e-01 -1.04526734e+00 -5.23904026e-01 -1.06478488e+00 -6.06405318e-01 5.03601432e-01 5.82241774e-01 -4.57280457e-01 -5.03228784e-01 2.38452390e-01]
[15.316697120666504, 5.010491371154785]
076dd0e3-20cf-4aaf-8bb3-bb907f7f1add
learning-from-weights-a-cost-sensitive
1811.12776
null
http://arxiv.org/abs/1811.12776v2
http://arxiv.org/pdf/1811.12776v2.pdf
Learning From Weights: A Cost-Sensitive Approach For Ad Retrieval
Retrieval models such as CLSM is trained on click-through data which treats each clicked query-document pair as equivalent. While training on click-through data is reasonable, this paper argues that it is sub-optimal because of its noisy and long-tail nature (especially for sponsored search). In this paper, we discuss the impact of incorporating or disregarding the long tail pairs in the training set. Also, we propose a weighing based strategy using which we can learn semantic representations for tail pairs without compromising the quality of retrieval. We conducted our experiments on Bing sponsored search and also on Amazon product recommendation to demonstrate that the methodology is domain agnostic. Online A/B testing on live search engine traffic showed improvements in clicks (11.8\% higher CTR) and as well as improvement in quality (8.2\% lower bounce rate) when compared to the unweighted model. We also conduct the experiment on Amazon Product Recommendation data where we see slight improvements in NDCG Scores calculated by retrieving among co-purchased product.
['Rahul Agrawal', 'Nikit Begwani', 'Shrutendra Harsola']
2018-11-30
null
null
null
null
['product-recommendation']
['miscellaneous']
[-1.03631772e-01 -2.75595278e-01 -4.98541474e-01 -4.76162076e-01 -1.19304109e+00 -7.66474426e-01 6.27042532e-01 1.72540143e-01 -6.27828538e-01 3.98127139e-01 3.34539860e-01 -5.95473528e-01 -5.79735577e-01 -8.20352018e-01 -6.95376635e-01 -2.63976365e-01 -1.68140471e-01 5.90143740e-01 3.48995537e-01 -4.43407595e-01 6.57562673e-01 1.46005144e-02 -1.61206412e+00 4.66037720e-01 7.44116187e-01 1.23155570e+00 2.73594975e-01 5.92182577e-01 -2.77512848e-01 2.42847174e-01 -5.26129246e-01 -6.18008196e-01 5.81543446e-01 -4.26693335e-02 -5.63683271e-01 -1.35115921e-01 7.91979313e-01 -3.41727793e-01 -3.16080987e-01 7.81217575e-01 4.25281674e-01 2.96283484e-01 7.18152285e-01 -1.08525801e+00 -8.53028715e-01 5.16620219e-01 -8.60740364e-01 4.01329368e-01 2.42031261e-01 -2.24181548e-01 1.88191092e+00 -6.84146821e-01 5.30518472e-01 1.00235057e+00 3.36910129e-01 1.34359330e-01 -1.25610781e+00 -7.30476439e-01 1.14681236e-01 1.10959038e-01 -1.26813293e+00 -1.49277970e-01 2.47441441e-01 -2.38353744e-01 1.05446434e+00 4.00074691e-01 2.18859479e-01 7.50247538e-01 1.85892984e-01 8.96050036e-01 7.69329429e-01 -5.54352045e-01 1.57855928e-01 2.85969436e-01 4.90089744e-01 3.80750626e-01 5.97958326e-01 1.61490470e-01 -4.77069318e-01 -4.86170501e-01 3.47542852e-01 5.21427020e-02 8.38573743e-03 -2.24292696e-01 -6.89872622e-01 1.37255740e+00 4.66168135e-01 1.90612540e-01 -4.29996878e-01 3.63677204e-01 2.71350950e-01 5.03479958e-01 6.37225628e-01 5.44143558e-01 -4.25308585e-01 1.40189454e-01 -1.28643572e+00 7.62339294e-01 9.87759590e-01 1.01616585e+00 5.38615763e-01 -4.06186670e-01 -4.18413877e-01 1.03662670e+00 4.65172708e-01 5.52719355e-01 6.81788504e-01 -9.97990727e-01 4.45951551e-01 1.91860840e-01 5.12097538e-01 -1.14102042e+00 -9.35477316e-02 -7.01279461e-01 4.24595773e-02 -1.77410394e-01 3.38849932e-01 6.16484620e-02 -9.87558603e-01 1.63274109e+00 -2.45099202e-01 -4.45773870e-01 -3.79273623e-01 8.58848035e-01 4.58769113e-01 4.41632748e-01 3.29953313e-01 2.28417683e-02 1.32692087e+00 -1.17723763e+00 -8.01888347e-01 -2.35768542e-01 9.36399400e-01 -1.11742032e+00 1.23735499e+00 3.62636715e-01 -1.14737225e+00 -5.25318503e-01 -9.78702426e-01 -1.38499767e-01 -6.26186788e-01 -5.06454185e-02 7.21732020e-01 8.51599693e-01 -8.11868906e-01 5.72163939e-01 -3.76616389e-01 -4.48475957e-01 -6.44767955e-02 1.70237213e-01 2.51645595e-01 -2.47904778e-01 -1.43197167e+00 7.68127739e-01 -4.07029688e-02 -4.91726339e-01 -6.06829524e-01 -6.75001562e-01 -4.37718987e-01 4.59598273e-01 4.16377991e-01 -5.04826725e-01 1.46844566e+00 -5.07898510e-01 -8.37095320e-01 4.78429824e-01 -1.40932664e-01 -6.04144156e-01 4.28487003e-01 -4.62995529e-01 -7.20767081e-01 -6.74786940e-02 3.31984699e-01 5.58374465e-01 5.81480742e-01 -9.75627661e-01 -1.01698065e+00 -4.22709405e-01 3.01514089e-01 2.36623660e-01 -4.40028071e-01 2.31157821e-02 -5.90689003e-01 -7.11987615e-01 3.46978791e-02 -1.02498722e+00 -1.23831462e-02 -2.74532557e-01 -3.37114185e-02 -5.30330420e-01 4.48309392e-01 -6.53430760e-01 1.77664506e+00 -1.99302828e+00 -6.54259026e-01 5.66879928e-01 -9.76322293e-02 -3.95261421e-04 -5.16069293e-01 7.42536664e-01 1.22484997e-01 4.47534531e-01 5.31621277e-01 1.93827525e-02 4.14953709e-01 -6.04167134e-02 -4.42363948e-01 1.61450669e-01 -3.55400473e-01 7.97621846e-01 -6.32966936e-01 -3.08133155e-01 -2.74434865e-01 1.41089529e-01 -8.39208007e-01 -3.31810005e-02 -3.55356187e-01 -5.83019614e-01 -5.86054385e-01 6.59504831e-01 6.17833436e-01 -5.56681037e-01 1.58821903e-02 -2.19222773e-02 1.31985456e-01 5.06062150e-01 -8.72415364e-01 1.61647785e+00 -8.81396890e-01 3.83845985e-01 -1.05496235e-01 -4.86788243e-01 7.00993180e-01 9.50926840e-02 2.78341383e-01 -1.36584306e+00 -1.45809382e-01 3.84681463e-01 -1.37736937e-02 -4.46982443e-01 9.96248186e-01 8.79898369e-02 1.24055743e-02 9.54373300e-01 -2.58399844e-01 6.90147504e-02 4.59676772e-01 5.36954641e-01 1.00486112e+00 -1.25117168e-01 -2.74021298e-01 -5.08014977e-01 1.16094850e-01 1.03856057e-01 -1.38743207e-01 1.23023057e+00 1.32975489e-01 3.95446926e-01 1.60579875e-01 -4.75934669e-02 -1.03642976e+00 -8.94786894e-01 -1.25812083e-01 1.79065239e+00 1.82948440e-01 -6.58214629e-01 -4.17926520e-01 -7.21814096e-01 4.05437768e-01 1.17272973e+00 -5.38527310e-01 -2.15959802e-01 -2.52704650e-01 -7.37647712e-01 2.22388506e-01 5.09695172e-01 2.94940650e-01 -5.35964072e-01 -2.90604889e-01 1.52716920e-01 -1.92137450e-01 -7.56397486e-01 -8.09925079e-01 9.39406604e-02 -8.17838192e-01 -8.98823857e-01 -9.64641333e-01 -5.47912121e-01 1.50483146e-01 7.70339489e-01 1.22343719e+00 5.41487038e-02 -1.96936414e-01 5.30081928e-01 -4.80152160e-01 -3.84447664e-01 1.98199183e-01 2.14692265e-01 -2.37644583e-01 -4.39538836e-01 8.96679759e-01 8.19390863e-02 -1.10457540e+00 6.62973821e-01 -1.02827632e+00 -7.31672585e-01 4.44211781e-01 8.80731761e-01 2.93202102e-01 1.55802563e-01 7.18648791e-01 -9.56571519e-01 1.20976567e+00 -7.82262146e-01 -5.09876490e-01 4.20431733e-01 -1.45743811e+00 3.73158604e-02 2.61402130e-02 -2.17909753e-01 -1.08341599e+00 -6.52579367e-01 2.26084232e-01 1.14153408e-01 3.84945393e-01 6.24050558e-01 5.62298775e-01 1.93917975e-01 7.85002768e-01 -5.29856049e-02 -9.20314863e-02 -8.31025720e-01 5.39749563e-01 9.29294586e-01 -1.73648715e-01 -3.62879455e-01 5.11100292e-01 2.76482999e-01 -4.09682512e-01 -5.17937124e-01 -9.04252172e-01 -9.85134065e-01 2.10438058e-01 1.90888047e-01 5.44257343e-01 -8.43176007e-01 -6.72501683e-01 -2.92213947e-01 -6.72760487e-01 6.59541711e-02 -8.46690685e-02 5.71605265e-01 -4.02101696e-01 3.38115454e-01 -6.75094545e-01 -7.63598800e-01 -3.48440796e-01 -9.29514050e-01 1.04854989e+00 4.93570715e-02 -2.84157336e-01 -9.10702944e-01 1.29223049e-01 7.06286609e-01 8.69413674e-01 -6.61270201e-01 1.09911442e+00 -1.20459282e+00 -3.93447667e-01 -6.98169947e-01 -5.15565753e-01 3.46940249e-01 1.00741223e-01 -1.94460168e-01 -7.20621109e-01 -4.29646194e-01 -2.63489664e-01 -3.24201435e-01 1.09151840e+00 4.11650479e-01 1.10864961e+00 -2.09805727e-01 -3.61305207e-01 -1.67784140e-01 1.50479603e+00 2.02538952e-01 4.54575181e-01 5.95660627e-01 -6.18454814e-02 7.14726627e-01 1.02281725e+00 4.93282288e-01 6.01033047e-02 8.87580752e-01 3.32033038e-01 2.34048799e-01 -1.79631412e-01 -5.51719189e-01 1.03891037e-01 3.96885902e-01 1.28056586e-01 -6.59117818e-01 -4.52235460e-01 5.03235459e-01 -1.68497145e+00 -7.80527353e-01 -1.47431135e-01 2.39562845e+00 6.39783859e-01 2.29310006e-01 2.63602465e-01 -8.94774869e-02 5.93410134e-01 9.23104733e-02 -2.96082526e-01 -5.32643557e-01 1.56900510e-01 4.67073649e-01 9.03264821e-01 6.14391387e-01 -7.52083778e-01 6.78360105e-01 7.10911942e+00 1.19393027e+00 -6.65316939e-01 3.94427078e-03 4.36354399e-01 -3.39904845e-01 -6.49129510e-01 -1.45846575e-01 -9.29278553e-01 6.63573086e-01 1.13247633e+00 -3.63745153e-01 2.37089023e-01 9.59062457e-01 2.01678574e-01 -4.28136498e-01 -9.62135494e-01 6.18657708e-01 9.24906507e-02 -1.13649523e+00 1.72243342e-01 3.93785119e-01 7.30703354e-01 1.11794688e-01 3.13889742e-01 5.29223859e-01 4.73081708e-01 -7.71687806e-01 4.60623533e-01 3.01810265e-01 2.63316482e-01 -6.53980672e-01 9.23748910e-01 2.38211945e-01 -6.77801430e-01 -1.88206565e-02 -4.14404333e-01 2.05703646e-01 5.72102191e-03 4.69340682e-01 -7.89421320e-01 4.68570739e-01 6.99104905e-01 2.52884865e-01 -7.38963425e-01 1.37740397e+00 2.40074873e-01 6.13630652e-01 -3.58836472e-01 -3.85244459e-01 5.80300391e-01 -1.72290370e-01 1.21062897e-01 1.27390623e+00 5.33727527e-01 -3.97585571e-01 -8.08229670e-02 5.95596552e-01 -6.07971065e-02 3.33699405e-01 -4.31438148e-01 -2.98954725e-01 3.61469686e-01 1.08424687e+00 -5.17449021e-01 -3.64714444e-01 -6.91456378e-01 8.39436352e-01 -1.64994970e-01 6.11037970e-01 -8.77439201e-01 -6.10854626e-01 3.12570035e-01 4.67090935e-01 6.34874940e-01 -8.47307891e-02 -1.25296772e-01 -7.87698448e-01 -1.24080367e-01 -5.89277446e-01 6.64029598e-01 -7.46463776e-01 -1.48068762e+00 3.98509741e-01 1.15490228e-01 -9.28424358e-01 -3.05504531e-01 -5.22252977e-01 -3.03276211e-01 1.09284484e+00 -1.62158346e+00 -6.29036427e-01 -1.02608897e-01 3.92817736e-01 6.13200366e-01 -1.10796481e-01 5.96601248e-01 8.32066178e-01 1.71593614e-02 7.68280447e-01 4.83325094e-01 -3.73870522e-01 9.08526659e-01 -1.10387301e+00 1.62629589e-01 6.91557899e-02 3.89366835e-01 1.12493277e+00 8.36756527e-01 -5.95659256e-01 -1.13450205e+00 -7.24676251e-01 1.23214793e+00 -3.46164793e-01 9.59407926e-01 4.23978418e-02 -6.63311064e-01 6.28271401e-01 3.84380400e-01 -4.82039690e-01 9.57301140e-01 5.67819297e-01 -6.66645408e-01 -3.10058817e-02 -1.07840073e+00 6.07283950e-01 7.30692863e-01 -4.34537232e-01 -6.15303159e-01 7.72826016e-01 7.44363606e-01 3.32887203e-01 -7.87889361e-01 2.11957440e-01 8.78327429e-01 -8.94751668e-01 1.02307999e+00 -9.29642558e-01 3.05492282e-01 5.80890141e-02 -5.79484582e-01 -1.08561361e+00 -4.88790244e-01 -7.26730004e-02 3.11068650e-02 8.56138349e-01 8.19645643e-01 -4.66387808e-01 1.02926159e+00 7.97742426e-01 2.92110354e-01 -4.82340038e-01 -3.91666502e-01 -1.17027760e+00 1.81817904e-01 -2.29750097e-01 2.43953452e-01 3.82706970e-01 -1.28494695e-01 3.86868298e-01 -2.30990604e-01 -3.60096067e-01 4.99799997e-01 3.00793350e-01 4.89250004e-01 -9.80914593e-01 -5.15377104e-01 -4.73237723e-01 8.99045542e-02 -1.39612246e+00 -2.76272327e-01 -1.00221610e+00 -3.09999079e-01 -1.33393729e+00 3.76835555e-01 -5.67933857e-01 -6.62406743e-01 1.10382982e-01 1.07289888e-01 4.14926410e-01 3.76498699e-02 2.96249837e-01 -7.02643454e-01 1.44748792e-01 1.00842273e+00 -2.06092626e-01 1.44616663e-01 3.38053495e-01 -9.08019900e-01 1.98367387e-01 6.97933435e-01 -5.01234233e-01 -6.71488941e-01 -4.48968500e-01 4.92612481e-01 -1.33347496e-01 4.07591984e-02 -4.77744728e-01 1.56207919e-01 8.64758044e-02 1.12156890e-01 -6.54826939e-01 3.70628595e-01 -8.01501811e-01 -1.31073952e-01 3.89618337e-01 -1.06673920e+00 1.60039842e-01 -1.19174995e-01 1.03628325e+00 -2.22163498e-01 -6.52040839e-01 1.87110543e-01 -2.59200245e-01 -4.24223989e-01 -1.20372865e-02 -1.07106037e-01 6.72722980e-02 6.53104961e-01 -2.84292847e-02 -4.21806455e-01 -8.49791229e-01 -6.44658685e-01 3.60080630e-01 8.34453329e-02 5.70016384e-01 1.74691901e-01 -1.21836424e+00 -4.58097935e-01 7.47630447e-02 5.06074905e-01 -9.11960065e-01 7.89963454e-02 5.21392167e-01 -3.36684108e-01 1.12013865e+00 1.78565606e-01 -9.84902307e-02 -1.16583240e+00 4.80659366e-01 -1.81290209e-01 -3.20966363e-01 -2.92242821e-02 8.49278808e-01 2.35614590e-02 -7.33693689e-02 6.43572330e-01 -5.64237908e-02 -9.82640460e-02 2.87156701e-01 4.91291493e-01 6.54686034e-01 4.31289345e-01 -2.99386885e-02 -1.44188806e-01 3.44264776e-01 -6.58864975e-01 -3.72925967e-01 1.02659476e+00 -3.39295506e-01 2.05086127e-01 8.42772052e-02 1.50115597e+00 1.36939451e-01 -5.83321750e-01 -3.56830925e-01 4.12735403e-01 -8.62109959e-01 4.02641624e-01 -1.13572288e+00 -1.11245298e+00 6.29016817e-01 7.27430761e-01 7.50223637e-01 6.82654858e-01 1.03987351e-01 9.40777183e-01 6.17136359e-01 4.30560410e-01 -1.49693489e+00 1.17965229e-01 6.55370951e-02 7.40554154e-01 -1.27419209e+00 1.98791295e-01 -1.28671661e-01 -4.93554473e-01 8.52590322e-01 3.26441854e-01 -2.86191434e-01 9.40787792e-01 -3.25633228e-01 -1.42489612e-01 -3.89708549e-01 -9.14671004e-01 -1.94752961e-01 2.50583768e-01 8.50673243e-02 8.15789580e-01 1.07424557e-02 -1.11020494e+00 3.86904687e-01 -7.32243732e-02 7.95724168e-02 2.05621362e-01 9.77606475e-01 -7.15553761e-01 -1.23062193e+00 5.07978275e-02 9.11586583e-01 -8.59769046e-01 -4.76913959e-01 -1.59334078e-01 8.95216763e-01 -4.75879461e-01 1.29666615e+00 1.38322979e-01 -4.64085579e-01 2.55119383e-01 1.55135676e-01 3.22151244e-01 -5.27756929e-01 -5.52220941e-01 4.61151063e-01 4.32387531e-01 -7.28964746e-01 -2.81715780e-01 -3.40335220e-01 -7.08214223e-01 -3.61672252e-01 -8.07683587e-01 7.01554000e-01 1.00580132e+00 4.01436180e-01 4.60876435e-01 3.68499458e-01 6.07396722e-01 -1.38147116e-01 -1.25936759e+00 -1.06793535e+00 -9.91731167e-01 6.30493164e-01 3.22331414e-02 -8.04223895e-01 -7.76430368e-01 -3.83400351e-01]
[11.329263687133789, 7.439615726470947]
9eed1c24-3b58-4bc3-907d-c866539146bd
refrec-pseudo-labels-refinement-via-shape
2110.11036
null
https://arxiv.org/abs/2110.11036v1
https://arxiv.org/pdf/2110.11036v1.pdf
RefRec: Pseudo-labels Refinement via Shape Reconstruction for Unsupervised 3D Domain Adaptation
Unsupervised Domain Adaptation (UDA) for point cloud classification is an emerging research problem with relevant practical motivations. Reliance on multi-task learning to align features across domains has been the standard way to tackle it. In this paper, we take a different path and propose RefRec, the first approach to investigate pseudo-labels and self-training in UDA for point clouds. We present two main innovations to make self-training effective on 3D data: i) refinement of noisy pseudo-labels by matching shape descriptors that are learned by the unsupervised task of shape reconstruction on both domains; ii) a novel self-training protocol that learns domain-specific decision boundaries and reduces the negative impact of mislabelled target samples and in-domain intra-class variability. RefRec sets the new state of the art in both standard benchmarks used to test UDA for point cloud classification, showcasing the effectiveness of self-training for this important problem.
['Luigi Di Stefano', 'Samuele Salti', 'Pierluigi Zama Ramirez', 'Riccardo Spezialetti', 'Adriano Cardace']
2021-10-21
null
null
null
null
['point-cloud-classification']
['computer-vision']
[ 2.67231077e-01 -1.21508129e-01 -1.37215868e-01 -5.15606165e-01 -1.15458429e+00 -8.43391299e-01 7.65253127e-01 2.32957244e-01 -2.38117158e-01 4.77509439e-01 -2.59148031e-01 -9.69217718e-02 -4.66902465e-01 -6.42078996e-01 -8.49319398e-01 -7.82084286e-01 5.05936034e-02 1.17277145e+00 3.82326990e-01 -9.34895501e-02 2.37405762e-01 1.02319026e+00 -1.89560199e+00 3.41120481e-01 6.87389493e-01 9.74135458e-01 5.18858507e-02 2.88723081e-01 -4.09574568e-01 2.54678994e-01 -4.34385121e-01 -7.34312907e-02 5.23648143e-01 2.45290715e-03 -9.07516181e-01 2.98566490e-01 8.12735081e-01 2.22941622e-01 3.95486057e-01 7.64532804e-01 5.61585188e-01 -1.14875793e-01 1.29306567e+00 -1.41292965e+00 -3.56694430e-01 1.30906492e-03 -4.01237130e-01 3.59892219e-05 -5.48656657e-02 -2.99061425e-02 8.34586322e-01 -9.99201834e-01 8.90970230e-01 1.02628481e+00 1.04853833e+00 6.37986839e-01 -1.62173069e+00 -7.07239807e-01 -2.03706846e-01 -6.49112538e-02 -1.46817374e+00 -2.64913440e-01 9.11271811e-01 -8.89999628e-01 8.65179300e-01 1.77268311e-01 1.73647568e-01 1.25845301e+00 -2.33540758e-01 5.21223605e-01 1.38583541e+00 -4.67346519e-01 6.14711404e-01 4.11545008e-01 9.86812711e-02 2.96041593e-02 2.36047998e-01 5.28377712e-01 -1.20120838e-01 -4.35106754e-01 6.59710288e-01 -2.19756350e-01 2.71440804e-01 -1.17973089e+00 -9.58487034e-01 8.00743937e-01 7.41971135e-02 3.51169556e-01 -1.81903332e-01 -4.54712063e-01 5.24390996e-01 7.12306201e-01 7.05831707e-01 5.24438381e-01 -7.93232679e-01 2.05637574e-01 -7.67227471e-01 5.26960492e-01 6.07353628e-01 1.20383608e+00 1.09952939e+00 -2.26607807e-02 3.13460827e-02 1.06301928e+00 2.12959066e-01 7.48605549e-01 3.61615390e-01 -6.45355284e-01 1.63916782e-01 5.82994819e-01 3.37791033e-02 -4.20568973e-01 -3.63230467e-01 -4.30147052e-01 -5.31653583e-01 7.42668331e-01 5.03824234e-01 4.68645291e-03 -9.93751585e-01 1.26872671e+00 5.22339940e-01 4.48760569e-01 2.88501561e-01 5.75003564e-01 9.36763883e-01 1.76375270e-01 6.60553202e-02 5.65395877e-02 9.39622998e-01 -6.12865567e-01 -1.61335878e-02 -9.08202827e-02 6.69424295e-01 -7.78337002e-01 8.96108806e-01 2.78890461e-01 -5.02534032e-01 -9.86772001e-01 -9.68938768e-01 4.09471780e-01 -6.48170412e-01 -7.27647990e-02 1.96728170e-01 6.54952526e-01 -8.72416675e-01 8.81006658e-01 -4.56116617e-01 -4.88835901e-01 7.55789459e-01 5.83684146e-01 -5.57357371e-01 -2.82744430e-02 -6.20663524e-01 9.07885671e-01 2.84855872e-01 -7.10985959e-01 -8.42057228e-01 -9.48786795e-01 -6.57136440e-01 -5.18020391e-01 7.96581954e-02 -4.82281923e-01 1.20791519e+00 -9.54809427e-01 -1.44260108e+00 1.65825856e+00 5.05811535e-02 -2.69499898e-01 5.99687636e-01 -4.23496254e-02 -4.66369897e-01 -1.94467932e-01 4.67427701e-01 7.67069876e-01 1.39377105e+00 -1.85095859e+00 -7.39132047e-01 -6.32143319e-01 -6.03894234e-01 5.62188141e-02 1.75038278e-02 -1.64262846e-01 -4.11015674e-02 -5.90826750e-01 2.05411464e-01 -1.07069123e+00 -1.07009485e-01 -1.22440539e-01 1.63835660e-01 -6.58224404e-01 1.09606540e+00 1.12164803e-01 3.64225447e-01 -2.22533393e+00 1.17503174e-01 4.02952731e-01 1.86599508e-01 4.44821388e-01 -3.74268174e-01 1.89412549e-01 -5.27046502e-01 -2.36143440e-01 -6.27584219e-01 -6.07149661e-01 -8.75222757e-02 5.10462105e-01 -5.72831631e-01 5.62212765e-01 4.75334167e-01 6.20124340e-01 -7.97227323e-01 -5.07952988e-01 4.38443750e-01 1.99122921e-01 -1.39426425e-01 3.69368047e-01 -4.19062823e-01 9.29729700e-01 -3.45872134e-01 7.29967058e-01 1.26517451e+00 -1.55191928e-01 -2.11043701e-01 3.96222668e-03 -2.24090993e-01 -1.98043749e-01 -1.44023371e+00 1.58350420e+00 -3.96784276e-01 2.28224576e-01 -3.95829715e-02 -1.11484730e+00 1.44620419e+00 1.93875834e-01 9.25218999e-01 -6.96247399e-01 1.21508844e-01 5.19502640e-01 -4.56444710e-01 -3.75666857e-01 1.37179688e-01 -5.59589148e-01 -2.79560667e-02 1.71797693e-01 5.66658199e-01 -7.02954173e-01 -3.86584878e-01 -5.21639645e-01 8.68563294e-01 5.46798527e-01 3.09556931e-01 -3.95664752e-01 6.23976231e-01 3.73587638e-01 4.07493144e-01 6.97571099e-01 -2.98906028e-01 1.09078777e+00 -9.00022220e-03 -6.15598023e-01 -1.32139337e+00 -1.22427905e+00 -6.29980564e-01 1.16194344e+00 5.58775701e-02 7.94230178e-02 -4.05550271e-01 -1.00977015e+00 6.41765416e-01 6.73158526e-01 -6.83318138e-01 5.88028431e-02 -3.35851282e-01 -5.32560170e-01 4.41174597e-01 2.94800133e-01 7.20072165e-02 -1.02503657e+00 -2.18024313e-01 9.68601480e-02 3.54702890e-01 -1.39492881e+00 2.19175071e-01 6.46866918e-01 -1.21527779e+00 -1.13563704e+00 -7.75282621e-01 -9.01050448e-01 5.84784031e-01 2.15615287e-01 1.48943102e+00 -3.78714502e-01 -4.02311198e-02 7.84725606e-01 -5.35844564e-01 -7.49459088e-01 -6.89145625e-01 2.16493919e-01 8.37424621e-02 3.89439873e-02 7.16405451e-01 -8.59309256e-01 1.16257504e-01 5.76546490e-01 -8.51444006e-01 -4.30009156e-01 5.37860692e-01 5.91407895e-01 8.34577143e-01 -2.57226169e-01 5.44378042e-01 -1.23473501e+00 3.56347024e-01 -5.15387952e-01 -8.45668972e-01 1.22247927e-01 -6.86678290e-01 1.89957067e-01 4.09703165e-01 -2.12877259e-01 -8.22575152e-01 5.39822638e-01 -3.04548144e-01 -9.14503813e-01 -9.29216504e-01 -2.21579865e-01 -3.73751409e-02 -6.94464326e-01 1.22293627e+00 1.40451133e-01 8.37918818e-02 -5.92543662e-01 3.07850271e-01 7.04711258e-01 5.57850778e-01 -1.01492918e+00 1.08371234e+00 7.39078879e-01 3.98595810e-01 -9.64757204e-01 -7.35528290e-01 -1.14433908e+00 -1.31632257e+00 -8.17107931e-02 7.51961350e-01 -1.03409147e+00 -3.30128580e-01 4.43150431e-01 -1.13508248e+00 -1.95095524e-01 -8.69654655e-01 8.95279571e-02 -9.13675964e-01 4.41305339e-01 3.31408173e-01 -6.49453044e-01 -1.79097086e-01 -9.81088400e-01 1.48425245e+00 -1.58354521e-01 4.13242169e-02 -1.06892204e+00 6.94553137e-01 7.27231652e-02 2.61444181e-01 6.10209346e-01 6.01404071e-01 -1.07406759e+00 -2.28865534e-01 1.02501407e-01 -2.88395286e-01 7.25016356e-01 2.17233121e-01 -1.67217597e-01 -1.43343675e+00 -3.17666322e-01 9.07825157e-02 -5.65328896e-01 6.85992777e-01 1.77404612e-01 9.11517918e-01 4.71835434e-01 -4.87406343e-01 6.38465703e-01 1.63250136e+00 -1.67601198e-01 3.94432664e-01 5.61679840e-01 5.74239075e-01 6.91782773e-01 7.14630604e-01 4.40173268e-01 1.91862926e-01 7.49546409e-01 5.05995035e-01 -1.94009200e-01 -2.93679297e-01 -3.19861695e-02 -8.39555562e-02 5.15295446e-01 -1.14481702e-01 2.68739760e-01 -1.45946300e+00 7.35541523e-01 -1.73410451e+00 -5.03206670e-01 -4.59111094e-01 2.25491571e+00 5.64991057e-01 1.50250033e-01 3.78396332e-01 -7.51697458e-03 5.46680689e-01 -1.90133393e-01 -5.22456586e-01 -6.17175363e-03 -4.58131492e-01 7.72126794e-01 7.74045289e-01 2.08808541e-01 -1.67498362e+00 9.84585404e-01 6.30024433e+00 9.53165233e-01 -1.10540903e+00 3.55257720e-01 8.12448338e-02 6.11778736e-01 -4.44413461e-02 -7.07887933e-02 -7.89733708e-01 2.90232927e-01 7.47211277e-01 2.53185451e-01 -1.21944912e-01 1.32162273e+00 -3.99158001e-01 1.96845621e-01 -1.19623411e+00 1.28802311e+00 9.80920345e-02 -1.39620304e+00 -1.64555132e-01 2.61041988e-02 1.06894183e+00 5.80221295e-01 -3.60247828e-02 3.61440599e-01 4.68386203e-01 -6.57341778e-01 6.67802870e-01 2.97444493e-01 9.27777708e-01 -5.59410870e-01 7.29893923e-01 2.58693218e-01 -1.06422412e+00 9.64037552e-02 -5.15004396e-01 2.64697690e-02 -2.79597640e-01 5.20228446e-01 -1.07048225e+00 8.85950267e-01 6.82053089e-01 9.21449125e-01 -6.92145944e-01 1.14777136e+00 6.32327721e-02 4.20064211e-01 -4.60271269e-01 4.86637384e-01 1.17713876e-01 -2.47406408e-01 7.95501292e-01 1.34945321e+00 1.96276248e-01 -3.87033373e-01 2.11165249e-01 6.20420456e-01 2.55634338e-01 1.24912299e-01 -1.11372519e+00 6.20068669e-01 2.08065704e-01 9.67074931e-01 -6.12071991e-01 -2.21970215e-01 -3.88117403e-01 8.58303308e-01 3.53702188e-01 -3.28535400e-02 -5.60499012e-01 9.42709148e-02 9.38635588e-01 3.55102390e-01 5.66425800e-01 -1.46163985e-01 -7.66606271e-01 -9.21463251e-01 -1.38406217e-01 -7.27163076e-01 4.41055030e-01 -4.79226321e-01 -1.89977968e+00 7.07091808e-01 1.98402449e-01 -1.97833812e+00 -1.14690185e-01 -8.77563953e-01 -4.39929307e-01 6.46886289e-01 -1.86575055e+00 -1.52163589e+00 -5.11632144e-01 5.52578568e-01 3.62073183e-01 -5.72779834e-01 1.17064333e+00 3.53278846e-01 6.16973788e-02 4.23196852e-01 6.03664815e-01 -1.47466511e-01 1.04302478e+00 -1.41156924e+00 5.52385211e-01 3.26903045e-01 1.77198783e-01 -2.43086535e-02 5.42704999e-01 -4.94905949e-01 -8.73470426e-01 -1.32872844e+00 4.81571138e-01 -1.02871883e+00 4.57407355e-01 -4.66615051e-01 -1.15619266e+00 4.31657344e-01 -2.19023928e-01 4.63951141e-01 7.18853891e-01 1.64984420e-01 -6.09137833e-01 -2.21652284e-01 -1.36796403e+00 -1.08829319e-01 1.12008107e+00 -5.48569739e-01 -9.34087038e-01 4.81027961e-01 3.57229739e-01 -4.94266450e-01 -9.28852201e-01 8.40099216e-01 1.26080528e-01 -1.02810681e+00 1.19345641e+00 -6.23561263e-01 2.41237864e-01 -4.35011357e-01 -2.16804788e-01 -1.31440473e+00 -3.10944974e-01 -2.60753661e-01 4.03630733e-01 1.31270576e+00 1.85774297e-01 -5.25346398e-01 1.13863230e+00 2.78596580e-01 -4.07677054e-01 -1.57976881e-01 -1.21679270e+00 -1.17501783e+00 4.99671906e-01 -6.62693381e-01 6.55334711e-01 1.40618348e+00 -6.54788733e-01 2.90880620e-01 8.68593529e-02 2.26928949e-01 9.08655703e-01 2.89870679e-01 1.25893104e+00 -2.02676511e+00 1.50685772e-01 -2.41729125e-01 -5.24614096e-01 -6.93269908e-01 5.05856097e-01 -1.18602014e+00 -2.46421676e-02 -1.07684600e+00 -1.49451941e-01 -9.73603129e-01 -5.88051379e-02 3.69988978e-01 3.72143626e-01 2.75831912e-02 2.64907978e-03 4.61626232e-01 -6.42667174e-01 5.53692281e-01 1.08990109e+00 -1.40712112e-01 -1.81909263e-01 3.38027686e-01 -2.25870281e-01 5.71043313e-01 5.72400928e-01 -8.82814765e-01 -2.37976372e-01 -2.04465032e-01 -1.62795961e-01 -3.30766797e-01 5.25417089e-01 -1.42291617e+00 7.66352490e-02 -1.12219594e-01 2.53233075e-01 -7.43709683e-01 2.28029236e-01 -1.40134108e+00 1.48186743e-01 2.55508516e-02 1.57377616e-01 -1.90389082e-01 5.18143892e-01 5.22393823e-01 -3.12151909e-01 -3.29317510e-01 1.25723684e+00 -1.44954875e-01 -1.01488483e+00 1.87865421e-01 1.26657814e-01 1.25948146e-01 1.22023201e+00 -6.46978080e-01 1.06028691e-01 2.29398072e-01 -8.62303793e-01 -1.23944823e-02 6.52030349e-01 5.14827669e-01 4.95045006e-01 -1.27535093e+00 -9.35506523e-01 6.10973537e-01 7.68024027e-01 4.53618944e-01 1.93477109e-01 2.26555079e-01 -2.32438996e-01 2.31389806e-01 -4.93386894e-01 -1.32359838e+00 -1.32062650e+00 5.59662044e-01 5.17194390e-01 -2.21667677e-01 -5.42065263e-01 8.66494179e-01 1.37013653e-02 -1.15926063e+00 1.38258282e-02 -1.79174006e-01 -2.36290261e-01 2.04366490e-01 -3.77643108e-02 2.37592563e-01 6.62448466e-01 -8.16229522e-01 -4.81982291e-01 1.03544807e+00 1.15954548e-01 2.02276036e-01 1.70763218e+00 2.38277733e-01 5.61325923e-02 4.34582263e-01 1.21419001e+00 -1.86757833e-01 -1.31245279e+00 -6.14112258e-01 4.85039413e-01 -5.46674311e-01 -1.32209390e-01 -7.54080296e-01 -5.43741703e-01 6.40654385e-01 1.07879984e+00 1.39528885e-01 8.30624998e-01 3.11019182e-01 1.37200996e-01 2.34220713e-01 4.76186275e-01 -1.10928822e+00 4.16057371e-02 7.70868361e-01 9.66204107e-01 -1.72679520e+00 -5.88936880e-02 -3.93578976e-01 -6.80780649e-01 1.18475127e+00 6.14035308e-01 -4.32207346e-01 9.83246505e-01 2.10652262e-01 3.19656461e-01 -4.21310335e-01 -2.86961198e-01 -6.23898923e-01 3.97658825e-01 1.27998757e+00 5.41923717e-02 1.66742038e-02 2.97020435e-01 3.52038026e-01 4.05987762e-02 -6.04832657e-02 -1.14104431e-02 9.05885875e-01 -1.93572313e-01 -1.50373852e+00 -7.38416553e-01 2.82950997e-01 2.93622706e-02 4.56089228e-01 -5.78880727e-01 8.42142284e-01 5.82398772e-01 5.42933643e-01 1.68487862e-01 -5.16952991e-01 9.00389791e-01 2.41034776e-01 4.43159640e-01 -8.24830770e-01 -6.50355875e-01 -1.72439411e-01 -1.88360870e-01 -1.62288278e-01 -7.62642980e-01 -1.07348442e+00 -8.92328441e-01 1.25790313e-01 -1.80687964e-01 5.81511036e-02 7.85739303e-01 9.09420133e-01 6.10432148e-01 3.43392521e-01 8.68283629e-01 -1.08843493e+00 -6.78375244e-01 -8.83175910e-01 -5.10992169e-01 8.88858676e-01 5.58787227e-01 -1.05849612e+00 -3.97067845e-01 6.47838116e-02]
[8.076512336730957, -2.7975263595581055]
4f55e778-3e4f-4077-8894-159c987dc398
dense-prediction-with-attentive-feature
2111.00770
null
https://arxiv.org/abs/2111.00770v3
https://arxiv.org/pdf/2111.00770v3.pdf
Dense Prediction with Attentive Feature Aggregation
Aggregating information from features across different layers is an essential operation for dense prediction models. Despite its limited expressiveness, feature concatenation dominates the choice of aggregation operations. In this paper, we introduce Attentive Feature Aggregation (AFA) to fuse different network layers with more expressive non-linear operations. AFA exploits both spatial and channel attention to compute weighted average of the layer activations. Inspired by neural volume rendering, we extend AFA with Scale-Space Rendering (SSR) to perform late fusion of multi-scale predictions. AFA is applicable to a wide range of existing network designs. Our experiments show consistent and significant improvements on challenging semantic segmentation benchmarks, including Cityscapes, BDD100K, and Mapillary Vistas, at negligible computational and parameter overhead. In particular, AFA improves the performance of the Deep Layer Aggregation (DLA) model by nearly 6% mIoU on Cityscapes. Our experimental analyses show that AFA learns to progressively refine segmentation maps and to improve boundary details, leading to new state-of-the-art results on boundary detection benchmarks on BSDS500 and NYUDv2. Code and video resources are available at http://vis.xyz/pub/dla-afa.
['Peter Kontschieder', 'Samuel Rota Bulò', 'Min Sun', 'Fisher Yu', 'Thomas E. Huang', 'Yung-Hsu Yang']
2021-11-01
null
null
null
null
['boundary-detection']
['computer-vision']
[ 1.46027267e-01 6.42092451e-02 -5.03619201e-02 -5.59688687e-01 -8.75819862e-01 -3.38322073e-01 5.06147623e-01 -4.53552380e-02 -2.61790037e-01 5.57314694e-01 4.41058502e-02 -2.94728518e-01 -3.30643426e-03 -9.03916001e-01 -8.19912374e-01 -4.16149795e-01 -1.71022996e-01 2.93526739e-01 6.20934844e-01 -1.87341839e-01 -2.26982996e-01 6.69186413e-01 -1.49938989e+00 6.70848191e-01 9.93247390e-01 1.65610075e+00 2.32325122e-01 7.06983268e-01 -2.29781792e-01 6.41814947e-01 -3.89310211e-01 -4.30532336e-01 5.42633533e-01 -3.20243612e-02 -8.86688471e-01 -9.46364850e-02 9.63890612e-01 -5.59497535e-01 -3.30796868e-01 7.33074605e-01 4.37133968e-01 1.21634133e-01 4.08859462e-01 -1.14290607e+00 -3.68280798e-01 7.08265424e-01 -7.27522612e-01 5.24450243e-01 -2.40602374e-01 1.39348060e-01 1.18408453e+00 -9.38325703e-01 4.58168268e-01 1.25401795e+00 9.02404904e-01 3.37194055e-01 -1.26230216e+00 -8.27319920e-01 5.97155035e-01 1.15023330e-02 -1.46755064e+00 -4.20263052e-01 4.54286486e-01 -3.17781180e-01 1.02508545e+00 3.98906261e-01 8.08339179e-01 6.15888298e-01 -1.76057547e-01 1.07567334e+00 8.89223039e-01 5.44483103e-02 4.86800596e-02 -3.25059325e-01 1.67492986e-01 8.74958336e-01 1.78004920e-01 -1.08047441e-01 -6.10542893e-01 9.39749777e-02 1.07645667e+00 -6.38984814e-02 -3.22243333e-01 -1.24398366e-01 -1.00682425e+00 6.55456722e-01 8.30974162e-01 -2.00699925e-01 -3.84205431e-01 4.95163381e-01 3.58356953e-01 1.34115130e-01 7.89231539e-01 2.00940371e-01 -5.51453948e-01 -5.22228181e-02 -1.07196188e+00 3.79726410e-01 5.74749410e-01 9.29679036e-01 8.23641598e-01 1.58869073e-01 -3.83410513e-01 7.99318194e-01 1.58839226e-01 5.08218944e-01 7.13553419e-03 -1.38606727e+00 5.95681906e-01 5.56417108e-01 -1.57504484e-01 -6.71483219e-01 -6.85247540e-01 -5.95084488e-01 -8.58983934e-01 4.79465604e-01 3.62827182e-01 -2.26049483e-01 -1.29700625e+00 1.55472910e+00 3.94778848e-01 6.68438137e-01 -1.72792032e-01 1.00090218e+00 1.14097738e+00 5.53956032e-01 1.25646070e-01 2.97504902e-01 1.06837571e+00 -1.28330398e+00 -1.76499829e-01 -2.76000857e-01 5.44867158e-01 -5.71677208e-01 1.14719033e+00 2.96454459e-01 -1.34244812e+00 -6.07793927e-01 -9.01829243e-01 -3.93233836e-01 -2.11729199e-01 2.83692367e-02 9.60295558e-01 4.09995258e-01 -1.44531131e+00 7.56103277e-01 -1.08452308e+00 -9.07664523e-02 1.10658050e+00 4.72007811e-01 -1.62114799e-02 1.50659651e-01 -9.58390236e-01 2.55297154e-01 1.53752193e-01 1.84409767e-01 -6.43717349e-01 -1.21576941e+00 -8.81493688e-01 2.27968544e-01 2.93946207e-01 -7.67210841e-01 1.36382782e+00 -8.02638948e-01 -1.30587959e+00 6.54319286e-01 -1.82313606e-01 -7.96863854e-01 7.01592803e-01 -5.00394940e-01 -1.87825948e-01 1.78060085e-01 2.96699964e-02 1.17753994e+00 5.84567606e-01 -1.14185858e+00 -1.02924871e+00 -1.31811678e-01 2.24643454e-01 2.83038378e-01 -2.15260815e-02 -3.54582846e-01 -7.42746770e-01 -7.69573629e-01 2.09120184e-01 -6.83397114e-01 -2.86963761e-01 4.09703285e-01 -4.07014191e-01 -8.82773772e-02 9.87346113e-01 -5.91979980e-01 1.05831254e+00 -2.14200735e+00 -1.10667281e-01 2.70473957e-01 6.50175512e-01 4.44513232e-01 -2.46351093e-01 -2.70927072e-01 2.18534440e-01 2.49902040e-01 -4.47426558e-01 -7.93289065e-01 2.67754984e-03 1.33397907e-01 -3.55187446e-01 2.53461570e-01 2.15564966e-01 1.23765838e+00 -3.97946447e-01 -4.39546883e-01 3.39898795e-01 7.99969375e-01 -8.21514904e-01 -1.10655181e-01 -3.34614515e-01 2.61544347e-01 -3.57598037e-01 8.47886324e-01 7.69419312e-01 -5.57150126e-01 -3.12923193e-01 -3.78118455e-01 -1.56766742e-01 3.64147127e-01 -9.91734862e-01 1.85374796e+00 -4.75721627e-01 8.97975445e-01 2.24212319e-01 -6.31056428e-01 7.16406107e-01 -7.99906179e-02 4.11437154e-01 -1.01974177e+00 1.84640825e-01 2.35113114e-01 -2.08235398e-01 -4.71005449e-03 5.83534718e-01 4.03277755e-01 8.23555440e-02 1.24070376e-01 -2.68455204e-02 2.61737555e-02 2.99901068e-01 2.44278252e-01 1.04325402e+00 2.13910863e-01 -1.54226020e-01 -3.43741387e-01 2.81447768e-01 -1.28477961e-01 7.52186894e-01 7.78304219e-01 -1.92481995e-01 8.26753497e-01 4.81264979e-01 -6.10946894e-01 -9.22311008e-01 -1.37641871e+00 -3.35447520e-01 1.16313434e+00 3.72191638e-01 -4.56801564e-01 -7.74131954e-01 -5.86552143e-01 1.71220869e-01 4.16436434e-01 -6.94753408e-01 2.51425564e-01 -8.24080765e-01 -6.32124424e-01 5.60414255e-01 9.84725595e-01 1.03191209e+00 -9.90453601e-01 -5.50735116e-01 1.16577335e-01 -3.86832021e-02 -1.28262329e+00 -4.94549155e-01 5.10031618e-02 -9.31079686e-01 -8.68309200e-01 -7.89060771e-01 -6.53445065e-01 3.01400721e-01 3.32099378e-01 1.13562465e+00 6.26093671e-02 -2.97037423e-01 5.62476851e-02 -2.02440992e-01 -1.95774317e-01 -8.65158625e-03 4.18474942e-01 -4.81647491e-01 -1.17623605e-01 -3.45801413e-02 -7.07225621e-01 -8.31197321e-01 2.02843174e-01 -6.32846057e-01 4.71324235e-01 5.68514049e-01 4.56452072e-01 9.93561149e-01 -3.90542835e-01 3.33210558e-01 -9.67120945e-01 2.35101521e-01 -2.78514653e-01 -7.54012764e-01 1.08662248e-01 -4.22612101e-01 -9.60472375e-02 4.96358663e-01 -8.53286833e-02 -1.14048052e+00 -5.72209209e-02 -4.62906510e-01 -5.48605382e-01 -2.49224246e-01 2.05382511e-01 -9.51835215e-02 -1.40425697e-01 3.73345524e-01 -7.55952746e-02 -2.16507480e-01 -5.89665234e-01 4.32186067e-01 3.52485001e-01 7.20678926e-01 -4.59723771e-01 4.62485492e-01 7.90524423e-01 -6.85651526e-02 -6.88458383e-01 -9.42677438e-01 -4.04672056e-01 -5.40622890e-01 -1.63687870e-01 9.81177449e-01 -1.17566597e+00 -6.58711374e-01 6.63774073e-01 -9.12390113e-01 -1.03359914e+00 -4.96461511e-01 1.81079611e-01 -5.76319396e-01 -4.69799526e-02 -8.40339661e-01 -2.82479107e-01 -6.21367693e-01 -1.21273613e+00 1.24061847e+00 4.43596512e-01 3.51431035e-02 -8.05378377e-01 -3.07729155e-01 3.92146885e-01 6.14684343e-01 3.36105138e-01 6.00400984e-01 -4.26526159e-01 -8.89393210e-01 2.82508671e-01 -7.81346381e-01 2.93658435e-01 -9.31307152e-02 1.93188004e-02 -1.12502074e+00 -8.91003981e-02 -6.45789504e-01 -1.41685188e-01 1.40978026e+00 7.34427452e-01 1.50571299e+00 -5.17459437e-02 -3.12676370e-01 1.27122104e+00 1.20117795e+00 -3.14196087e-02 5.26318014e-01 2.89504707e-01 1.04103804e+00 1.33471474e-01 3.66367638e-01 4.81593281e-01 5.79458952e-01 6.03085756e-01 4.51384127e-01 -5.84946513e-01 -7.56203771e-01 1.13100275e-01 2.22465154e-02 3.70844513e-01 -2.11670443e-01 -2.68476427e-01 -9.55649734e-01 4.64087725e-01 -1.77391660e+00 -6.59286976e-01 -2.59874582e-01 1.82949018e+00 5.49467444e-01 3.30563068e-01 1.62599146e-01 -1.28935367e-01 5.86509109e-01 4.25355852e-01 -8.31807315e-01 -3.60958636e-01 -2.77029067e-01 4.66023803e-01 7.11953044e-01 7.14522243e-01 -1.46141934e+00 1.23187470e+00 5.53326416e+00 1.03720117e+00 -1.25574660e+00 2.88957447e-01 1.09350586e+00 -4.49935794e-01 -3.31767291e-01 -3.04352969e-01 -8.72541904e-01 2.04268649e-01 7.12936521e-01 2.53959775e-01 3.75143737e-01 7.34355271e-01 9.64474976e-02 -2.45331060e-02 -7.34906256e-01 9.10190463e-01 -2.50585705e-01 -1.93499589e+00 1.13091469e-01 3.84560451e-02 8.56974602e-01 8.10797930e-01 1.72464713e-01 3.09736013e-01 3.63094240e-01 -1.03101277e+00 9.45971847e-01 4.08688307e-01 8.67712736e-01 -7.51620352e-01 6.24584258e-01 -1.32681936e-01 -1.59287441e+00 -2.53938604e-02 -2.20010445e-01 -6.83301836e-02 5.54736733e-01 6.17141128e-01 -3.81568730e-01 3.26264977e-01 1.08285642e+00 8.48354995e-01 -5.99156320e-01 1.11687064e+00 -8.88976008e-02 6.61710143e-01 -6.89585745e-01 4.75300193e-01 4.84639108e-01 -2.21396820e-03 2.52093673e-01 1.21819580e+00 2.18803644e-01 1.66723385e-01 2.09062025e-01 9.24223304e-01 -2.94476032e-01 4.58315760e-03 -4.39426713e-02 3.87413830e-01 4.46717411e-01 1.18305945e+00 -1.02501202e+00 -5.23001432e-01 -5.47057509e-01 9.14597511e-01 4.53832179e-01 4.77961063e-01 -1.17034829e+00 -1.50861561e-01 1.19684660e+00 3.35029423e-01 6.15880787e-01 -1.20682724e-01 -6.24033034e-01 -9.11016822e-01 -8.01942572e-02 -3.73735070e-01 3.04176331e-01 -5.91791093e-01 -9.72938061e-01 7.98278809e-01 -9.73206908e-02 -8.47236812e-01 3.27639818e-01 -3.54455054e-01 -5.84120750e-01 7.79195130e-01 -1.78733432e+00 -1.29990959e+00 -7.09660172e-01 5.81605673e-01 7.83758819e-01 1.03972539e-01 4.31170315e-01 5.22966564e-01 -6.40357494e-01 6.56069398e-01 -2.35343874e-01 1.76513731e-01 2.68071920e-01 -1.20119631e+00 1.17111886e+00 8.05174947e-01 2.15772301e-01 -5.10182679e-02 2.02182129e-01 -6.41818643e-01 -6.42087519e-01 -1.50526786e+00 5.10626554e-01 -2.78926268e-03 5.35972059e-01 -3.77395660e-01 -1.03296840e+00 7.63524592e-01 3.30992751e-02 5.85106671e-01 4.26240325e-01 -3.29275131e-02 -3.22430760e-01 -2.03199044e-01 -9.50525761e-01 5.45422137e-01 1.42922628e+00 -2.58112133e-01 -2.87573263e-02 9.86720696e-02 1.14961505e+00 -7.46755660e-01 -9.25409615e-01 6.22512102e-01 5.11594713e-01 -1.23463511e+00 1.12824988e+00 -2.85719186e-01 3.20488721e-01 -2.06620201e-01 -2.78842032e-01 -8.64135504e-01 -4.50180888e-01 -5.65164804e-01 -2.87712574e-01 1.07929647e+00 6.12046599e-01 -6.30265534e-01 1.02219820e+00 5.95842898e-01 -6.00307763e-01 -1.09349275e+00 -9.48525190e-01 -6.48322225e-01 1.36922941e-01 -7.21794128e-01 7.98587918e-01 6.30382180e-01 -6.03449047e-01 3.29753151e-03 -2.36314349e-03 2.96352386e-01 5.57805002e-01 3.03594977e-01 6.47228777e-01 -1.26704192e+00 -2.48180211e-01 -8.53713751e-01 -3.56268495e-01 -1.46637905e+00 1.09465159e-02 -9.84693885e-01 -1.70124426e-01 -1.78905666e+00 -2.37921432e-01 -8.45947742e-01 -2.30300412e-01 5.99975765e-01 -1.51237369e-01 7.16213286e-01 5.10740876e-01 5.91060668e-02 -8.15376520e-01 4.90999311e-01 1.37786841e+00 8.55639484e-03 -4.22305346e-01 1.03549194e-02 -5.25669515e-01 9.78420079e-01 1.07373965e+00 -1.97187766e-01 -4.62841779e-01 -7.66114712e-01 5.79122417e-02 -2.40762681e-01 5.42751610e-01 -1.22857642e+00 3.81147000e-03 7.03092441e-02 4.08019334e-01 -6.89605176e-01 5.23318410e-01 -4.83851790e-01 -3.20988409e-02 3.50138903e-01 -3.03841233e-01 -5.11023477e-02 5.68003953e-01 4.65123355e-01 -1.05242684e-01 3.94724518e-01 7.27834225e-01 8.00812915e-02 -1.17325139e+00 6.61572635e-01 -1.33221960e-02 2.70335495e-01 9.56177831e-01 -3.11504483e-01 -5.65973938e-01 -1.26574129e-01 -8.65775585e-01 4.21552092e-01 3.41110945e-01 2.77329654e-01 6.11984849e-01 -1.05896425e+00 -6.76591039e-01 3.83217156e-01 -2.10041195e-01 6.68703794e-01 5.60137689e-01 9.69647586e-01 -9.08961117e-01 3.12686890e-01 -3.86183523e-02 -7.76984930e-01 -1.07784748e+00 -7.87932891e-03 3.94398719e-01 -2.51410425e-01 -1.12487042e+00 1.38833904e+00 6.49193048e-01 -1.31380782e-01 3.87875378e-01 -8.54426980e-01 2.24557109e-02 1.23071764e-02 4.67967212e-01 4.15619373e-01 1.41119823e-01 -6.67129040e-01 -2.60591835e-01 4.70268726e-01 -1.82733685e-01 1.69384882e-01 1.36600804e+00 -2.58675784e-01 2.76153624e-01 1.39902681e-01 1.12910390e+00 -2.40589693e-01 -2.00746846e+00 -3.01169097e-01 -3.96836907e-01 -2.94845343e-01 3.43625396e-01 -8.77776027e-01 -1.75674939e+00 9.12725270e-01 5.95618963e-01 -7.01068342e-02 1.22576833e+00 2.43327707e-01 1.25870645e+00 9.09700319e-02 2.46164560e-01 -9.15735543e-01 -1.85582355e-01 4.53702837e-01 8.66304576e-01 -1.24093509e+00 -7.60383084e-02 -7.24647522e-01 -6.16994441e-01 7.95802176e-01 8.05367649e-01 -2.93664634e-01 8.78516734e-01 6.67986989e-01 1.35989994e-01 -1.84456781e-01 -6.01817667e-01 -4.92028445e-01 3.17561388e-01 4.34631467e-01 1.42978355e-01 2.25621998e-01 7.50828385e-02 6.90852642e-01 -2.24266320e-01 -1.23670444e-01 2.62572110e-01 6.44871414e-01 -4.05538410e-01 -6.96669340e-01 -6.27438575e-02 7.37946272e-01 -3.28416586e-01 -3.60332906e-01 4.39953897e-03 9.69135344e-01 2.76606590e-01 4.93517637e-01 6.70332193e-01 -3.31686437e-01 2.87627012e-01 -1.65268689e-01 1.93666056e-01 -3.92268479e-01 -6.02711141e-01 1.88144833e-01 6.36765361e-02 -1.07667029e+00 -3.69571805e-01 -6.48194909e-01 -1.64819789e+00 -4.95952934e-01 -1.49733648e-01 -3.38024259e-01 3.78660977e-01 8.64801288e-01 6.23781621e-01 8.59336019e-01 4.74661142e-02 -1.13412619e+00 9.15348083e-02 -6.46634877e-01 -4.62930441e-01 1.01269946e-01 4.57640707e-01 -7.18712032e-01 -1.35374814e-01 -3.61676142e-02]
[9.513556480407715, 0.057499468326568604]
315a4388-3a35-41f5-9790-acce8dee2c49
cuhk-ee-voice-cloning-system-for-icassp-2021
2103.04699
null
https://arxiv.org/abs/2103.04699v5
https://arxiv.org/pdf/2103.04699v5.pdf
CUHK-EE Voice Cloning System for ICASSP 2021 M2VoC Challenge
This paper presents the CUHK-EE voice cloning system for ICASSP 2021 M2VoC challenge. The challenge provides two Mandarin speech corpora: the AIShell-3 corpus of 218 speakers with noise and reverberation and the MST corpus including high-quality speech of one male and one female speakers. 100 and 5 utterances of 3 target speakers in different voice and style are provided in track 1 and 2 respectively, and the participants are required to synthesize speech in target speaker's voice and style. We take part in the track 1 and carry out voice cloning based on 100 utterances of target speakers. An end-to-end voicing cloning system is developed to accomplish the task, which includes: 1. a text and speech front-end module with the help of forced alignment, 2. an acoustic model combining Tacotron2 and DurIAN to predict melspectrogram, 3. a Hifigan vocoder for waveform generation. Our system comprises three stages: multi-speaker training stage, target speaker adaption stage and target speaker synthesis stage. Our team is identified as T17. The subjective evaluation results provided by the challenge organizer demonstrate the effectiveness of our system. Audio samples are available at our demo page: https://daxintan-cuhk.github.io/CUHK-EE-system-M2VoC-challenge/ .
['Tan Lee', 'Guangyan Zhang', 'Hingpang Huang', 'Daxin Tan']
2021-03-08
null
null
null
null
['voice-cloning']
['speech']
[-3.07614416e-01 3.14201154e-02 3.29106838e-01 -2.48810381e-01 -1.43838668e+00 -5.62296867e-01 1.58877149e-01 -5.34528196e-01 -1.58632502e-01 2.85698891e-01 4.09393489e-01 -4.95463014e-01 4.29702908e-01 8.28174502e-02 -4.08058614e-01 -6.92284763e-01 1.99068729e-02 3.30555201e-01 -1.59106448e-01 -2.12368235e-01 -3.20715845e-01 1.31639257e-01 -1.33087587e+00 3.26864183e-01 5.74041128e-01 9.45950031e-01 4.94498640e-01 1.39316571e+00 2.34706968e-01 4.34970498e-01 -7.83426642e-01 -8.19170699e-02 3.30232233e-01 -8.19003284e-01 -6.42780662e-01 7.12608919e-02 2.14512736e-01 -4.13896404e-02 -2.96642572e-01 9.46743071e-01 1.29934680e+00 2.69134164e-01 4.45210755e-01 -9.73627865e-01 -3.26887041e-01 1.22768056e+00 1.14353247e-01 3.97831976e-01 2.11149216e-01 2.83131450e-01 9.41290975e-01 -1.13827300e+00 3.30542356e-01 1.24121737e+00 4.89829779e-01 8.69280636e-01 -9.10324574e-01 -1.00707400e+00 -3.60118538e-01 3.22253853e-01 -1.54440784e+00 -1.26160777e+00 9.42754388e-01 -3.18928361e-01 8.98011923e-01 6.68428898e-01 4.74057466e-01 1.36786234e+00 -2.37010732e-01 7.82127678e-01 7.74811089e-01 -5.75977564e-01 3.35550159e-01 3.91313255e-01 -3.72486003e-02 -9.00258427e-04 -8.17249835e-01 3.08469504e-01 -6.88606381e-01 -5.39515316e-02 3.58310431e-01 -9.03626561e-01 -6.19333088e-01 4.93668258e-01 -1.01097691e+00 6.93434834e-01 -1.62677556e-01 4.88633454e-01 -3.81074518e-01 -1.38732474e-02 5.53228021e-01 7.20603287e-01 4.71279562e-01 1.39194608e-01 -5.92843413e-01 -2.81471670e-01 -9.58555818e-01 2.66933978e-01 7.52749681e-01 1.37374330e+00 -1.01728188e-02 8.45511854e-01 -9.05344337e-02 1.42497706e+00 4.25202966e-01 8.68283212e-01 8.85529220e-01 -8.40932965e-01 4.95432884e-01 -4.76453811e-01 -1.17657177e-01 -1.83276176e-01 -2.71777362e-01 -5.75094998e-01 -5.29872656e-01 -5.89167327e-02 -2.36062538e-02 -5.38844466e-01 -8.20236802e-01 1.60917687e+00 4.43791509e-01 5.85142337e-02 3.59810561e-01 9.39500034e-01 1.10211587e+00 1.06464875e+00 -4.32653457e-01 -4.57157731e-01 1.45358992e+00 -1.26217961e+00 -1.10346985e+00 6.65674582e-02 3.61591071e-01 -1.47300386e+00 1.07972240e+00 5.76779187e-01 -1.39339268e+00 -9.59962428e-01 -8.58156145e-01 1.07493870e-01 2.45749671e-02 4.13895965e-01 -2.38705665e-01 1.04908001e+00 -1.03186750e+00 6.31728321e-02 -4.03460830e-01 -1.16992280e-01 -3.04640740e-01 1.36107773e-01 -7.85761401e-02 5.11039317e-01 -1.26784670e+00 4.58789289e-01 2.13401034e-01 1.81781560e-01 -1.15381360e+00 -6.92307889e-01 -8.52091730e-01 -1.23416081e-01 -1.91326048e-02 -2.30963930e-01 1.95406401e+00 -8.54131103e-01 -2.20525789e+00 5.02209425e-01 -3.14619422e-01 -5.29293120e-01 5.31969786e-01 -1.99397758e-01 -1.14153969e+00 5.77345937e-02 -1.11599518e-02 4.17340159e-01 1.05271721e+00 -9.04696286e-01 -8.54224741e-01 3.78788984e-03 -8.34474444e-01 3.91572595e-01 7.06506297e-02 3.75247151e-01 -3.65251303e-01 -8.69892240e-01 5.09715676e-02 -9.90338624e-01 1.14629202e-01 -1.03487563e+00 -9.20970261e-01 -7.28559494e-02 8.69524419e-01 -1.17007458e+00 1.27779222e+00 -2.38225031e+00 -2.62116035e-03 1.66362152e-02 -3.52836728e-01 3.90298337e-01 -9.23060924e-02 2.88019836e-01 -1.77635968e-01 -4.86970432e-02 1.10528588e-01 -8.01970005e-01 1.21417619e-01 -4.25371259e-01 -4.81674910e-01 3.12542111e-01 -3.93170603e-02 4.61223334e-01 -4.08804148e-01 -1.67571098e-01 2.48131856e-01 4.63634104e-01 -4.83383298e-01 5.37942648e-01 9.76787508e-02 8.25963080e-01 9.09400657e-02 5.48683167e-01 5.91457605e-01 7.65624166e-01 -1.65061593e-01 1.26693742e-02 -4.63690549e-01 9.41621900e-01 -1.46398282e+00 1.41411209e+00 -5.70050955e-01 6.74935043e-01 7.21934795e-01 -2.90165365e-01 1.00067079e+00 1.17550910e+00 2.00946834e-02 -2.87535369e-01 3.74142289e-01 6.40483558e-01 3.09480309e-01 -5.47824383e-01 5.04681885e-01 -1.88836291e-01 2.88201533e-02 8.15612748e-02 3.95765215e-01 -7.95861363e-01 -1.32079035e-01 -1.40755370e-01 6.51564717e-01 -3.11993539e-01 -2.49453401e-03 -3.16941202e-01 6.74662173e-01 -4.20831293e-01 6.02935314e-01 4.04836178e-01 -4.27294731e-01 9.55265343e-01 -1.12756789e-01 1.69956133e-01 -1.08365524e+00 -1.11356688e+00 -1.19367339e-01 9.61087942e-01 -4.02939528e-01 -5.25935590e-01 -9.86953795e-01 7.44425356e-02 -4.49744672e-01 1.12236166e+00 -2.54495621e-01 8.40414912e-02 -7.23373234e-01 4.60821763e-02 1.00368750e+00 1.00875720e-01 2.62618750e-01 -1.28355885e+00 6.82869181e-02 3.94680649e-01 -4.59117711e-01 -1.15609193e+00 -1.28053677e+00 2.55022138e-01 -3.41366827e-01 -4.75870937e-01 -7.20242143e-01 -1.10574591e+00 -1.34954572e-01 5.47853392e-03 5.16296625e-01 -5.39323330e-01 3.26738991e-02 1.82308272e-01 -4.19827104e-01 -7.26583898e-01 -1.03609490e+00 2.07018763e-01 5.69135249e-01 1.95917487e-01 7.78151974e-02 -5.72216690e-01 -2.48624280e-01 4.44001526e-01 -4.51847345e-01 -6.61656260e-02 1.51159987e-01 7.77367532e-01 5.19700587e-01 2.72737909e-02 7.27406979e-01 -4.84039724e-01 6.30246460e-01 -3.58537436e-01 -6.13048434e-01 -2.29352787e-01 -4.78244349e-02 -5.43748856e-01 9.20987844e-01 -6.66794181e-01 -1.04893827e+00 1.30288631e-01 -9.24595833e-01 -5.98256469e-01 -2.20757410e-01 1.23837799e-01 -7.92913556e-01 5.27165473e-01 6.84609354e-01 3.83038968e-01 -2.50671923e-01 -8.40536833e-01 3.25155854e-01 1.63762975e+00 9.71884668e-01 -2.48851135e-01 9.00721133e-01 -3.18383038e-01 -8.72562468e-01 -1.49933922e+00 -4.14366901e-01 -6.14149213e-01 -3.71834099e-01 -1.96446076e-01 7.50654459e-01 -1.13451374e+00 -4.89915073e-01 8.87307525e-01 -1.33711398e+00 -3.89661968e-01 -3.34620416e-01 1.01733899e+00 -6.15987480e-01 -6.03010431e-02 -8.31612945e-01 -9.31050003e-01 -7.49489546e-01 -1.51397347e+00 7.45447576e-01 2.62415528e-01 -2.70262808e-01 -6.63783014e-01 1.67852819e-01 7.23092198e-01 6.17510021e-01 -4.61561114e-01 3.35317343e-01 -8.58241856e-01 -5.11051975e-02 1.16269320e-01 8.88076186e-01 8.65947306e-01 2.38006696e-01 -1.15318283e-01 -1.75223804e+00 -2.14035839e-01 3.51593941e-01 -9.76184011e-02 3.09990644e-01 6.60849094e-01 5.47794819e-01 -3.19203526e-01 2.39925399e-01 6.77677035e-01 5.36820114e-01 5.87842107e-01 4.41323549e-01 -2.88303167e-01 5.39215922e-01 6.55658424e-01 3.06261152e-01 2.47402936e-01 3.16657603e-01 8.55748773e-01 -9.14566442e-02 8.32675099e-02 -6.13680601e-01 -2.12405041e-01 8.65428388e-01 1.98909450e+00 3.20227861e-01 -3.29833239e-01 -6.12500429e-01 6.76281869e-01 -1.12235355e+00 -8.37943196e-01 -4.34528917e-01 2.10035324e+00 1.05298817e+00 9.43493545e-02 3.55492771e-01 3.92487347e-01 9.68890190e-01 9.20396373e-02 -3.53604972e-01 -7.83204615e-01 -1.63384259e-01 2.24479169e-01 1.54876426e-01 1.06612766e+00 -8.62779796e-01 1.08604074e+00 5.36066914e+00 1.11465514e+00 -1.35186541e+00 4.47679073e-01 4.17303175e-01 -4.61305559e-01 -2.13577986e-01 -2.41828457e-01 -1.08060455e+00 4.94599789e-01 1.65564704e+00 -2.77003556e-01 6.65778339e-01 7.81471729e-01 6.10218823e-01 3.95871907e-01 -9.58045542e-01 1.11505187e+00 6.62853345e-02 -9.43172395e-01 -3.96188974e-01 -1.36938378e-01 4.31484520e-01 2.75335968e-01 1.16426557e-01 6.52026951e-01 -1.27880216e-01 -8.48274171e-01 1.40624022e+00 3.69138196e-02 9.36107635e-01 -8.07709634e-01 4.18147177e-01 2.71861583e-01 -1.43366897e+00 2.11831972e-01 -1.08252302e-01 3.42990726e-01 3.85515988e-01 4.22193885e-01 -1.25416899e+00 5.10994196e-01 5.72388768e-01 6.16193525e-02 1.20272869e-02 9.89501536e-01 -3.78825337e-01 1.44763660e+00 -3.42747360e-01 1.18164979e-01 -1.23667203e-01 1.88148573e-01 1.18594325e+00 1.35688877e+00 4.29548711e-01 -7.83895329e-02 -5.10127433e-02 6.20765269e-01 -1.51593417e-01 5.15455604e-01 -2.33387068e-01 -6.65489733e-02 9.30824637e-01 1.11896896e+00 -4.94950153e-02 -1.55640811e-01 -1.12700853e-02 9.87175584e-01 -2.46748880e-01 5.03263235e-01 -7.55204678e-01 -6.89372003e-01 7.79963136e-01 -2.77107894e-01 4.61935967e-01 -2.02142075e-01 -1.51807651e-01 -8.48738134e-01 -5.68550788e-02 -1.12451017e+00 -9.90745500e-02 -7.20780313e-01 -9.36624885e-01 1.20033872e+00 -4.48372781e-01 -1.16405869e+00 -5.97701669e-01 -3.41587543e-01 -7.77354598e-01 1.39901257e+00 -1.13142705e+00 -9.60706949e-01 2.31349334e-01 5.67893445e-01 1.22892570e+00 -5.31401277e-01 9.43978906e-01 6.08320355e-01 -6.54396474e-01 8.32502842e-01 1.52568042e-01 3.54510844e-02 8.05410445e-01 -1.14193559e+00 7.38077998e-01 8.75570536e-01 1.28878891e-01 2.27144271e-01 9.35875535e-01 -4.02775884e-01 -1.24110401e+00 -1.28642142e+00 1.32919407e+00 -1.82308123e-01 4.97690320e-01 -7.92679012e-01 -6.64796293e-01 6.15313649e-01 4.38378274e-01 -1.94233999e-01 7.84330964e-01 -3.14519972e-01 9.59401652e-02 -1.23682730e-01 -7.35482275e-01 5.93296528e-01 6.42544627e-01 -6.90119088e-01 -6.45571232e-01 1.62390605e-01 1.12200773e+00 -8.01743805e-01 -7.12768137e-01 -8.33929181e-02 2.94936299e-01 -6.05571628e-01 3.50612909e-01 -1.45905375e-01 -1.85433552e-01 -5.27697027e-01 -5.39335728e-01 -1.76907384e+00 1.15090264e-02 -1.47053814e+00 3.69165599e-01 1.75724852e+00 7.24023402e-01 -3.79896671e-01 -2.19355873e-03 4.27684225e-02 -7.07409382e-01 -8.76511112e-02 -1.06972194e+00 -1.01431656e+00 8.65266323e-02 -1.05764627e+00 4.54484493e-01 7.27476239e-01 6.76016090e-04 7.28928506e-01 -5.88299394e-01 3.96654278e-01 3.39694351e-01 -3.13520819e-01 8.13491642e-01 -6.85763478e-01 -5.99374235e-01 -3.95860046e-01 1.54717326e-01 -9.32918489e-01 -1.12260982e-01 -8.43908429e-01 3.36575717e-01 -9.11781549e-01 -6.78698123e-01 -1.04042105e-01 8.43621641e-02 1.58705533e-01 -7.54496306e-02 1.58181489e-02 3.84362906e-01 6.51717335e-02 2.91989632e-02 8.93192291e-01 8.83723199e-01 -9.96231809e-02 -7.07960129e-01 6.22351229e-01 -3.54514629e-01 5.32418430e-01 9.90343153e-01 -5.58905482e-01 -2.81662196e-01 -1.63621888e-01 -6.07527018e-01 2.62806326e-01 -9.35305953e-02 -1.11224258e+00 2.55261779e-01 3.76762807e-01 -9.75258350e-02 -7.12880611e-01 6.66228950e-01 -5.37939489e-01 4.34335291e-01 3.74171972e-01 -3.30999792e-01 -2.06078008e-01 4.16219860e-01 -1.91816047e-01 -3.80911589e-01 -2.53313094e-01 1.02891088e+00 1.83304161e-01 -1.92998186e-01 -1.12946248e-02 -8.13509464e-01 1.60065591e-01 5.21048367e-01 1.76729262e-01 -1.37527794e-01 -6.13241673e-01 -9.84325826e-01 1.04747504e-01 -1.34209812e-01 6.08540595e-01 3.66117656e-01 -1.48860502e+00 -1.23672497e+00 4.62894708e-01 -1.27616152e-01 -2.79097944e-01 4.86884505e-01 8.69912624e-01 -3.29813212e-01 5.01216412e-01 1.63918719e-01 -4.32704866e-01 -1.40644646e+00 1.48534775e-03 6.32767022e-01 3.14624220e-01 -4.61420059e-01 1.16532505e+00 4.57977392e-02 -7.01685965e-01 5.70091248e-01 -3.31670314e-01 -4.06833813e-02 -5.36291078e-02 6.98711872e-01 4.43719476e-01 2.57184088e-01 -1.03298104e+00 -3.91634285e-01 1.78662956e-01 3.47081982e-02 -7.90323794e-01 9.88610685e-01 -4.60234731e-01 3.34448218e-01 8.71096671e-01 1.26765919e+00 7.24347770e-01 -7.80842721e-01 -1.25044495e-01 -2.20489830e-01 1.18820881e-02 2.08913550e-01 -8.09371114e-01 -8.32745492e-01 1.00215840e+00 6.14712656e-01 2.09644064e-01 1.05026829e+00 -1.30511180e-01 1.15034223e+00 -1.87858138e-02 -5.61553761e-02 -1.31494904e+00 -3.49269152e-01 9.73446012e-01 1.28189421e+00 -9.22535956e-01 -8.53828132e-01 -1.82948858e-01 -8.92194211e-01 1.01916718e+00 3.88746232e-01 4.00835931e-01 7.81132698e-01 3.33894730e-01 8.62801433e-01 3.66488278e-01 -8.63309383e-01 -1.15857869e-01 4.15772319e-01 5.31801522e-01 6.24737978e-01 3.87727559e-01 -2.97410376e-02 1.03376758e+00 -1.29009402e+00 -5.50031066e-01 5.25125980e-01 3.58819425e-01 -3.33043844e-01 -1.09982288e+00 -7.68408477e-01 -2.76707113e-01 -6.39014781e-01 -3.25989842e-01 -3.68165135e-01 3.18099260e-01 -5.45057878e-02 1.50258422e+00 -1.56309515e-01 -7.19270825e-01 5.78007460e-01 4.21580762e-01 -2.70291301e-03 -5.99665105e-01 -1.04410231e+00 9.77965534e-01 3.81825209e-01 -1.35480016e-01 3.36240143e-01 -7.82138884e-01 -1.28528106e+00 -5.13801686e-02 -3.47845823e-01 6.00002229e-01 1.11436117e+00 7.59606600e-01 1.76809177e-01 6.88162446e-01 1.12325203e+00 -8.71969402e-01 -6.40299618e-01 -1.65540695e+00 -8.83910835e-01 -7.59517252e-02 6.86999023e-01 1.30828723e-01 -6.54134095e-01 3.83691192e-01]
[14.879973411560059, 6.49099063873291]
9d8e712f-631f-43ed-9760-f6cb52761b93
deep-implicit-coordination-graphs-for-multi
2006.11438
null
https://arxiv.org/abs/2006.11438v2
https://arxiv.org/pdf/2006.11438v2.pdf
Deep Implicit Coordination Graphs for Multi-agent Reinforcement Learning
Multi-agent reinforcement learning (MARL) requires coordination to efficiently solve certain tasks. Fully centralized control is often infeasible in such domains due to the size of joint action spaces. Coordination graph based formalization allows reasoning about the joint action based on the structure of interactions. However, they often require domain expertise in their design. This paper introduces the deep implicit coordination graph (DICG) architecture for such scenarios. DICG consists of a module for inferring the dynamic coordination graph structure which is then used by a graph neural network based module to learn to implicitly reason about the joint actions or values. DICG allows learning the tradeoff between full centralization and decentralization via standard actor-critic methods to significantly improve coordination for domains with large number of agents. We apply DICG to both centralized-training-centralized-execution and centralized-training-decentralized-execution regimes. We demonstrate that DICG solves the relative overgeneralization pathology in predatory-prey tasks as well as outperforms various MARL baselines on the challenging StarCraft II Multi-agent Challenge (SMAC) and traffic junction environments.
['Mykel J. Kochenderfer', 'Jayesh K. Gupta', 'Peter Morales', 'Sheng Li', 'Ross Allen']
2020-06-19
null
null
null
null
['smac-1', 'smac']
['playing-games', 'playing-games']
[-5.41593790e-01 3.67589861e-01 -1.31267712e-01 7.90777132e-02 -5.70485234e-01 -7.13237345e-01 8.14968109e-01 2.26302236e-01 -5.60064077e-01 9.83907640e-01 1.44196272e-01 -2.54516363e-01 -2.55713820e-01 -6.05743349e-01 -8.36999476e-01 -6.49097323e-01 -5.99728227e-01 1.06250405e+00 2.31126592e-01 -6.97036803e-01 -9.25754532e-02 7.86995441e-02 -9.39208150e-01 6.58779740e-02 6.40313566e-01 5.38896024e-01 6.20928220e-03 1.04697669e+00 3.78327727e-01 1.39222646e+00 -6.66383922e-01 -1.59452349e-01 6.38452113e-01 -1.98383883e-01 -7.14495242e-01 1.32439584e-01 -4.83105555e-02 -5.56837022e-01 -2.84305394e-01 8.51394176e-01 4.76885587e-01 2.84365863e-01 5.85685432e-01 -2.03291750e+00 -1.96005136e-01 9.76166666e-01 -6.28238618e-01 1.83362126e-01 -1.79246012e-02 8.71078253e-01 1.34632933e+00 1.81199998e-01 5.69655240e-01 1.70686865e+00 3.99492860e-01 7.11371601e-01 -1.24886763e+00 -3.81212026e-01 6.74152434e-01 -1.98286176e-02 -6.11881435e-01 -8.54787901e-02 6.40670657e-01 -5.56923032e-01 1.38899899e+00 -4.30824101e-01 7.15152740e-01 1.20329356e+00 2.48320699e-01 7.01839924e-01 8.91555309e-01 -5.98972989e-03 6.04779720e-01 -6.48827970e-01 -2.02952430e-01 1.02838969e+00 5.39559722e-01 3.37711692e-01 -4.15198684e-01 -4.62093413e-01 8.52546215e-01 -1.63304225e-01 2.68507957e-01 -6.03358984e-01 -1.13065505e+00 1.02907658e+00 4.57300186e-01 -3.09438139e-01 -5.81661820e-01 1.01482379e+00 7.30814636e-01 7.86125720e-01 5.42560741e-02 8.76652062e-01 -6.53625250e-01 -1.07892007e-01 -1.15602292e-01 8.69182110e-01 1.14796853e+00 8.43084216e-01 7.12224960e-01 4.06717986e-01 7.90375024e-02 4.12818432e-01 4.94754076e-01 4.29427832e-01 2.06863061e-01 -1.54023302e+00 8.68238807e-01 5.80511868e-01 4.38952923e-01 -7.29738533e-01 -7.29022861e-01 -2.16390669e-01 -4.19080377e-01 8.12651396e-01 5.99005342e-01 -1.01357687e+00 -3.86570543e-01 1.98581457e+00 7.48263121e-01 -5.20696603e-02 4.82807696e-01 1.00097919e+00 2.18988851e-01 4.05923247e-01 1.10493682e-01 6.25767186e-02 1.26113594e+00 -1.34509444e+00 -3.72723758e-01 -5.49727082e-01 9.01570141e-01 1.84644267e-01 8.07952225e-01 2.26913840e-01 -8.80078733e-01 7.14412034e-02 -1.10084581e+00 2.40770623e-01 -2.19252408e-01 -2.49130741e-01 7.58666158e-01 -2.29476970e-02 -1.18762898e+00 5.50178230e-01 -1.24022400e+00 -1.93201914e-01 3.94944131e-01 6.03577912e-01 -1.33091465e-01 4.59460676e-01 -1.06320214e+00 9.97889400e-01 3.64428610e-01 -1.13035068e-01 -1.81485164e+00 -4.68317717e-01 -8.48197460e-01 7.64478371e-02 1.01833272e+00 -8.24187517e-01 1.68471825e+00 -9.08362269e-01 -1.81324279e+00 2.98427463e-01 8.89902771e-01 -9.66542542e-01 6.58162713e-01 -1.62073731e-01 2.99492121e-01 1.54743791e-01 3.26238126e-01 6.36917949e-01 8.83604884e-01 -1.05317533e+00 -6.95630610e-01 -1.57065719e-01 9.01325524e-01 6.54546738e-01 3.14462245e-01 -4.11579818e-01 1.32288456e-01 -2.83513635e-01 -7.42488027e-01 -1.20748174e+00 -6.48312926e-01 -5.34922145e-02 -2.15306669e-01 -5.51968277e-01 8.12980711e-01 -2.46566504e-01 5.78938007e-01 -1.81317365e+00 5.13609946e-01 -1.09043486e-01 6.37157500e-01 2.09223498e-02 -5.46159208e-01 9.43792880e-01 2.95709133e-01 -1.63821623e-01 1.49927735e-01 -4.19534415e-01 5.58253765e-01 7.11423039e-01 -2.33428270e-01 5.55103660e-01 1.86749116e-01 8.31672966e-01 -1.15132987e+00 -3.29462409e-01 -3.15510370e-02 -1.27697006e-01 -1.09242725e+00 3.93277109e-01 -1.10754406e+00 4.30049509e-01 -8.46814275e-01 4.26414937e-01 8.96881893e-02 -3.69116515e-01 8.16663086e-01 3.07454765e-01 -2.82049505e-03 1.10106081e-01 -1.07571959e+00 1.55877876e+00 -3.85662466e-01 4.31470424e-01 7.39750922e-01 -1.15876079e+00 3.88892293e-01 2.66317517e-01 5.68748951e-01 -4.86270636e-01 8.65303278e-02 -1.03262477e-01 4.15295422e-01 -3.87058914e-01 8.43047574e-02 7.11395368e-02 -3.45442265e-01 8.59643519e-01 7.53344595e-02 -2.35792324e-01 4.56191540e-01 4.55368608e-01 1.50631475e+00 2.14804158e-01 4.18044835e-01 -3.92380416e-01 6.16129190e-02 3.62560332e-01 8.61225426e-01 9.84867275e-01 -4.65367228e-01 -3.42334896e-01 1.38657999e+00 -6.78096116e-01 -1.02059269e+00 -6.84378684e-01 8.87773275e-01 1.47733951e+00 5.54321893e-02 -6.26637518e-01 -7.27330148e-01 -9.06278372e-01 3.82328957e-01 4.47011441e-01 -7.86270261e-01 -8.07108283e-02 -7.35524237e-01 -4.78413969e-01 5.73911428e-01 3.69163960e-01 6.73318923e-01 -1.10310757e+00 -1.17647672e+00 4.03496951e-01 1.87297121e-01 -1.07865429e+00 -5.33850133e-01 2.66584188e-01 -4.67494190e-01 -1.60893118e+00 -1.99602842e-02 -3.11621219e-01 4.62546110e-01 1.14018274e-02 1.02937782e+00 1.85446933e-01 -1.85885906e-01 9.60343182e-01 -2.68968880e-01 -5.04384100e-01 -4.90948468e-01 6.45096879e-03 3.00334632e-01 -2.69051224e-01 -2.53922999e-01 -7.82251477e-01 -6.52838528e-01 2.03882754e-01 -6.49038792e-01 1.12425208e-01 4.36368048e-01 9.68475461e-01 3.73994038e-02 6.92474842e-02 6.16681933e-01 -7.36793578e-01 1.05291092e+00 -5.98993301e-01 -1.32986069e+00 2.38833115e-01 -3.96735877e-01 5.92818499e-01 8.69765341e-01 -5.36853969e-01 -7.28927553e-01 -2.28339061e-02 6.16308212e-01 -4.57018077e-01 1.60190970e-01 5.22344708e-01 1.66305110e-01 8.29157308e-02 8.11897814e-01 -1.59102097e-01 5.34472644e-01 -8.01568106e-02 4.43132371e-01 -2.01383349e-03 1.11613266e-01 -1.39408743e+00 6.62365079e-01 2.75694579e-01 2.61187136e-01 -4.62798387e-01 -5.70694327e-01 1.03471331e-01 -3.23606730e-01 -2.66105115e-01 9.55623269e-01 -1.15929615e+00 -1.52602541e+00 3.57801050e-01 -1.12358952e+00 -1.37959409e+00 -2.12601185e-01 2.62339383e-01 -9.24812078e-01 1.16978496e-01 -6.80438101e-01 -7.04294980e-01 -1.22535080e-01 -1.30300725e+00 8.88639569e-01 -2.45989207e-02 1.51744291e-01 -1.26383626e+00 5.97726822e-01 2.25585133e-01 4.79956359e-01 5.82652807e-01 8.74666274e-01 -7.15174019e-01 -5.84399879e-01 2.34557718e-01 2.15364322e-01 -8.51724148e-02 -1.19234428e-01 -4.77743410e-02 -4.09047574e-01 -6.17153168e-01 -4.88268703e-01 -9.14656401e-01 4.16732848e-01 3.08002114e-01 4.25098181e-01 -8.68075848e-01 -8.81727114e-02 1.65649131e-01 1.22184241e+00 6.36396855e-02 -9.23061892e-02 5.38240731e-01 5.63739538e-01 4.68937516e-01 4.75209862e-01 9.63304222e-01 1.04763484e+00 6.40395284e-01 1.01362062e+00 3.18310887e-01 7.74633661e-02 -4.00649756e-01 7.71838903e-01 2.84583896e-01 -3.86318867e-03 -1.03814773e-01 -1.07505429e+00 2.87370890e-01 -2.69698787e+00 -9.77615118e-01 4.40819025e-01 1.47212446e+00 7.73481965e-01 -4.66995575e-02 7.30555832e-01 -5.54315269e-01 5.30926585e-01 2.37704158e-01 -1.16125226e+00 -4.08930451e-01 1.63821593e-01 -4.92136300e-01 5.55165827e-01 8.36457372e-01 -1.01926911e+00 1.16665578e+00 5.87723923e+00 4.34016377e-01 -7.27878928e-01 2.08558455e-01 2.41350740e-01 -1.59297571e-01 2.15889767e-01 3.08520526e-01 -6.59395635e-01 1.78169668e-01 7.94405043e-01 -2.98465252e-01 1.02086067e+00 1.07056904e+00 2.46696740e-01 -2.22564250e-01 -1.33647847e+00 7.28573740e-01 -4.12409395e-01 -1.35390890e+00 -1.41978115e-01 2.79424787e-01 7.38009393e-01 5.73967099e-01 -1.68820292e-01 7.25212634e-01 1.78129232e+00 -7.54413068e-01 8.06487501e-01 1.04066022e-01 2.82102436e-01 -5.48147619e-01 2.52095133e-01 6.93644047e-01 -1.17405498e+00 -5.71139753e-01 -3.46814662e-01 -5.73094010e-01 -1.97987840e-01 -2.59951860e-01 -9.25324380e-01 2.70660549e-01 4.23040390e-01 8.13864946e-01 -2.49825686e-01 4.35332417e-01 -3.76120508e-01 3.73677909e-01 -4.45237190e-01 -2.18803540e-01 7.74915695e-01 -4.28245097e-01 6.92123532e-01 6.77384913e-01 -4.11438614e-01 1.99202225e-01 8.57980311e-01 7.29071379e-01 4.50909846e-02 -5.18884897e-01 -7.94388771e-01 -4.11324620e-01 3.75132918e-01 1.16941357e+00 -4.63172704e-01 -3.41684908e-01 -2.90013075e-01 4.20409411e-01 1.06983757e+00 5.78664720e-01 -8.70506048e-01 8.57421756e-02 1.16685569e+00 -1.82147458e-01 3.40736032e-01 -6.81375027e-01 3.30847591e-01 -1.37759638e+00 -3.92466098e-01 -1.24238980e+00 8.03264737e-01 -4.86483932e-01 -1.20616221e+00 3.18824619e-01 1.79605111e-01 -9.44619775e-01 -6.67459190e-01 -5.57047129e-01 -9.14744735e-01 1.06793992e-01 -1.39078331e+00 -1.04074407e+00 1.52981177e-01 8.48214507e-01 4.91640925e-01 -5.65330148e-01 5.60572743e-01 -1.91907927e-01 -8.92328799e-01 1.46971971e-01 -1.27115816e-01 2.79109687e-01 3.77691060e-01 -1.52648401e+00 1.11302547e-01 6.50315106e-01 -2.54412591e-01 1.39943808e-01 8.00964475e-01 -6.15444958e-01 -1.83745897e+00 -1.09576654e+00 -1.57924041e-01 -4.85418111e-01 1.20531392e+00 -4.31648850e-01 -2.14605495e-01 1.03473330e+00 6.04224682e-01 3.84557098e-02 2.05228314e-01 9.38497037e-02 -6.21119082e-01 -1.83630943e-01 -9.04085696e-01 9.24657941e-01 8.93664062e-01 -3.17528576e-01 -4.85341817e-01 7.11230040e-01 9.22656357e-01 -4.06071395e-01 -6.85417950e-01 -2.33821809e-01 6.14442527e-02 -6.79694653e-01 5.66434801e-01 -1.16703200e+00 4.35173780e-01 -4.81213361e-01 -1.80505946e-01 -1.71470368e+00 -1.10608786e-01 -1.13719165e+00 -3.13772708e-01 5.88807285e-01 3.27776641e-01 -7.86950350e-01 5.21100461e-01 6.74347341e-01 -1.74030185e-01 -4.37490106e-01 -9.73467767e-01 -6.60997212e-01 1.19188175e-01 4.48796861e-02 3.62717330e-01 7.75799751e-01 3.42206001e-01 6.00999832e-01 -4.80046242e-01 3.70273888e-01 9.35723126e-01 5.12334555e-02 1.25563920e+00 -8.61623704e-01 -8.32414746e-01 -3.91709358e-01 -1.50196865e-01 -7.49896586e-01 5.65813422e-01 -8.38284850e-01 -8.08425248e-02 -1.51569724e+00 -6.53827414e-02 -2.74066031e-01 1.81812406e-01 9.42261696e-01 3.16550404e-01 -6.62997425e-01 4.22653198e-01 1.97658271e-01 -1.40843248e+00 8.16900432e-01 1.31265390e+00 -2.99063265e-01 -2.61931121e-01 -2.29571477e-01 -6.19458914e-01 6.44393146e-01 9.67244565e-01 -4.25130725e-01 -6.74326420e-01 -6.79754972e-01 5.38847804e-01 4.92046833e-01 4.76939827e-01 -8.77179384e-01 5.24104416e-01 -7.34189391e-01 -2.83945799e-01 9.67900380e-02 2.01652825e-01 -8.59059572e-01 -9.68230888e-02 8.88808846e-01 -5.58546185e-01 5.97082734e-01 1.09374516e-01 9.89186525e-01 4.01098989e-02 2.86310405e-01 6.92472279e-01 -4.04953271e-01 -4.48326439e-01 3.96694124e-01 -6.61757052e-01 6.36336863e-01 1.34200239e+00 3.70089382e-01 -7.73049533e-01 -6.90837681e-01 -6.82269037e-01 1.17670619e+00 2.35058635e-01 3.28766368e-02 2.81733066e-01 -1.01848900e+00 -6.42578661e-01 -1.35142118e-01 -1.64692611e-01 2.36805156e-01 5.76668903e-02 8.14967752e-01 -7.76633620e-01 -1.59903757e-05 -5.55760682e-01 -2.50126690e-01 -8.89583170e-01 5.26744127e-01 8.48499477e-01 -7.03538716e-01 -7.44416177e-01 3.17414969e-01 3.06103677e-01 -6.08777106e-01 2.58055568e-01 -4.08394247e-01 -3.13132480e-02 -2.16867581e-01 1.81408659e-01 3.51945609e-01 -5.30620575e-01 -6.69479668e-02 -1.78873435e-01 3.13009657e-02 -2.00946063e-01 -4.31385547e-01 1.55445766e+00 4.51305471e-02 -3.43680680e-02 -4.20853868e-02 5.38800240e-01 -6.15652263e-01 -2.10995054e+00 -2.02204183e-01 9.70768705e-02 1.23318814e-01 -2.52937943e-01 -7.56755233e-01 -9.65614855e-01 6.09810233e-01 -1.28623322e-01 2.71686047e-01 4.15563077e-01 -5.26419096e-02 1.33409023e-01 1.21029282e+00 5.47116995e-01 -1.52656817e+00 6.12293959e-01 9.10147727e-01 1.06302118e+00 -1.23689747e+00 3.20740864e-02 3.20710480e-01 -1.14965105e+00 1.10322797e+00 1.12450254e+00 -6.94065750e-01 6.56870902e-01 4.41153675e-01 -2.66248472e-02 -5.84065139e-01 -1.70135665e+00 -1.86956435e-01 -4.76623327e-01 6.95230067e-01 -3.89675677e-01 1.62369505e-01 1.48078158e-01 3.01360130e-01 -6.11222088e-02 -6.13302410e-01 6.85072124e-01 1.02090049e+00 -3.67339820e-01 -9.43113685e-01 -9.63187665e-02 1.54322118e-01 -1.15334846e-01 3.75401646e-01 -4.98875439e-01 1.08904958e+00 -3.38171601e-01 1.04103041e+00 -6.33698842e-03 -5.57711683e-02 1.15256689e-01 -3.09227943e-01 4.99335617e-01 -6.11088574e-01 -7.76879549e-01 5.61950803e-02 3.48345697e-01 -8.74161243e-01 -4.01954472e-01 -4.88576204e-01 -1.44661260e+00 -3.06618094e-01 1.42134830e-01 3.76767188e-01 4.10987079e-01 8.92520308e-01 5.50131559e-01 6.01611257e-01 4.45830971e-01 -9.04534042e-01 -1.22778022e+00 -7.62147129e-01 -4.50993061e-01 2.54980266e-01 6.67963684e-01 -8.87832642e-01 -4.00962859e-01 -3.50816816e-01]
[3.758683919906616, 1.934697151184082]
0c3ad60d-1c5e-42f8-a12d-17d23e4f0497
embedding-knowledge-graphs-attentive-to
null
null
https://2021.ecmlpkdd.org/wp-content/uploads/2021/07/sub_1096.pdf
https://2021.ecmlpkdd.org/wp-content/uploads/2021/07/sub_1096.pdf
Embedding Knowledge Graphs Attentive to Positional and Centrality Qualities
Knowledge graphs embeddings (KGE) are lately at the center of many artificial intelligence studies due to their applicability for solving downstream tasks, including link prediction and node classification. However, most Knowledge Graph embedding models encode, into the vector space, only the local graph structure of an entity, i.e., information of the 1-hop neighborhood. Capturing not only local graph structure but global features of entities are crucial for prediction tasks on Knowledge Graphs. This work proposes a novel KGE method named Graph Feature Attentive Neural Network (GFA-NN) that computes graphical features of entities. As a consequence, the resulting embeddings are attentive to two types of global network features. First, nodes’ relative centrality is based on the observation that some of the entities are more “prominent” than the others. Second, the relative position of entities in the graph. GFA-NN computes several centrality values per entity, generates a random set of reference nodes’ entities, and computes a given entity’s shortest path to each entity in the reference set. It then learns this information through optimization of objectives specified on each of these features. We investigate GFA-NN on several link prediction benchmarks in the inductive and transductive setting and show that GFA-NN achieves on-par or better results than state-of-the-art KGE solutions.
['Jens Lehmann', 'Damien Graux', 'Diego Collarana', 'Afshin Sadeghi']
2021-06-01
null
null
null
ecml-pkdd-2021-6
['link-property-prediction']
['graphs']
[-3.02964449e-01 5.72082341e-01 -5.52585959e-01 -2.29439855e-01 2.42210329e-01 -5.84773421e-01 6.72424197e-01 9.13917482e-01 -2.93977737e-01 5.77032864e-01 4.17881280e-01 -3.19318771e-01 -7.29164243e-01 -1.65036762e+00 -7.62032866e-01 -5.19329429e-01 -6.15031064e-01 6.94219351e-01 1.39061958e-01 -3.32452267e-01 -1.87001348e-01 5.52554071e-01 -1.07141209e+00 -1.48746595e-01 5.92053711e-01 7.91107953e-01 -3.19502890e-01 4.63965416e-01 -1.19588122e-01 6.93118274e-01 -1.10161401e-01 -6.90950155e-01 1.69212800e-02 -6.27918839e-02 -8.96953702e-01 -5.19849896e-01 1.32410213e-01 8.98523182e-02 -9.10103858e-01 8.75215054e-01 1.91860363e-01 3.15973312e-01 9.10757363e-01 -1.52894902e+00 -1.02739310e+00 9.50899363e-01 -1.72107816e-01 2.60019392e-01 3.49086165e-01 -2.72375345e-01 1.90724087e+00 -7.95610189e-01 8.59807432e-01 1.04052806e+00 7.44483173e-01 1.13913789e-01 -1.33353508e+00 -1.07498281e-01 1.32365584e-01 4.02732104e-01 -1.40516901e+00 1.01562835e-01 8.43265712e-01 -5.05862594e-01 1.01878834e+00 1.76646069e-01 7.73341894e-01 7.06732869e-01 6.06277287e-02 3.68314534e-01 4.03911322e-01 -3.64753664e-01 2.25029185e-01 1.91419750e-01 4.84741896e-01 1.08266854e+00 5.86627364e-01 -3.47033925e-02 -5.74031055e-01 -2.19171971e-01 4.36631083e-01 1.12118892e-01 -3.40028137e-01 -6.24430954e-01 -1.08475208e+00 1.10033321e+00 1.26350248e+00 4.84005660e-01 -5.06345928e-01 4.08607811e-01 2.77638108e-01 2.68050641e-01 4.00272161e-01 6.31595969e-01 -4.92360055e-01 1.66796654e-01 -2.21181512e-01 -1.23690888e-01 9.84410465e-01 5.95339060e-01 1.07994390e+00 -3.43283713e-01 -2.95884818e-01 4.74411786e-01 4.44438994e-01 1.60003483e-01 1.05855785e-01 -1.34475365e-01 4.40278083e-01 1.15046835e+00 -2.79194891e-01 -1.83994567e+00 -4.93065774e-01 -5.84377646e-01 -8.27669263e-01 -3.03856224e-01 1.60563797e-01 -1.98022518e-02 -6.37026072e-01 1.95173597e+00 4.71810520e-01 3.87080878e-01 3.57497088e-03 6.38395548e-01 1.13161802e+00 7.11992264e-01 1.04331084e-01 1.31205305e-01 1.13941562e+00 -7.22939312e-01 -2.94995129e-01 1.65432692e-03 9.95571256e-01 4.80354624e-03 4.71520990e-01 -4.06575352e-01 -6.22100592e-01 -1.49522111e-01 -7.63110220e-01 -6.65130839e-03 -1.04948938e+00 -1.66034341e-01 9.35258389e-01 3.67226183e-01 -1.22958517e+00 7.85578489e-01 -4.96426135e-01 -3.90527278e-01 3.95871341e-01 4.00257140e-01 -7.90417492e-01 -6.52970970e-02 -1.62240481e+00 8.21300864e-01 5.96820712e-01 1.04396045e-01 -2.45877489e-01 -8.56014550e-01 -1.18088841e+00 6.87357724e-01 3.92651439e-01 -8.82706940e-01 3.07839692e-01 -4.99166936e-01 -9.56978619e-01 5.66460907e-01 -3.15477736e-02 -4.94775265e-01 2.46813986e-02 2.54821092e-01 -5.13702929e-01 1.86668877e-02 -5.74032664e-02 4.95086342e-01 4.99773711e-01 -1.02268267e+00 -3.06964487e-01 -4.01931256e-01 3.70020658e-01 8.24093446e-02 -7.12873995e-01 -6.64866924e-01 -3.93135130e-01 -3.44003856e-01 -1.18313119e-01 -8.19566727e-01 -8.29215199e-02 -4.11758339e-03 -7.64655709e-01 -7.27909982e-01 6.48864985e-01 -3.42476964e-01 1.49584866e+00 -1.90136945e+00 4.66844380e-01 9.35617208e-01 9.65468884e-01 3.41146559e-01 -3.36817324e-01 7.48591363e-01 -2.13491231e-01 2.10547686e-01 1.89484119e-01 1.21705838e-01 2.01132372e-01 1.49804711e-01 5.52337524e-03 3.41874033e-01 2.36824572e-01 1.46722853e+00 -1.26705527e+00 -1.95539936e-01 1.43199906e-01 5.01013756e-01 -6.11541688e-01 -3.86782177e-02 -1.55696675e-01 -1.93273291e-01 -6.94917738e-01 3.05201292e-01 1.90426096e-01 -6.45517945e-01 5.53911090e-01 -5.16627908e-01 3.01875234e-01 2.02064961e-01 -9.58163083e-01 1.11025143e+00 -3.10405463e-01 6.84914231e-01 -5.26486695e-01 -1.20520723e+00 9.49490130e-01 3.06187347e-02 4.92444426e-01 -4.41863239e-01 7.75978565e-02 -1.21886574e-01 1.18119590e-01 -1.46389380e-01 4.43801165e-01 2.33711615e-01 2.01813430e-02 1.98619843e-01 2.44290158e-01 6.09588742e-01 3.04048210e-01 9.14422333e-01 1.63617253e+00 -4.24525559e-01 4.47298855e-01 -9.83395651e-02 3.78585696e-01 -2.43347034e-01 3.85256410e-01 5.20568371e-01 2.99825594e-02 -4.83571328e-02 9.82558310e-01 -5.20020783e-01 -7.41556525e-01 -1.17815113e+00 2.25606069e-01 1.12900221e+00 9.54979137e-02 -8.44149709e-01 -3.19046527e-01 -8.89635265e-01 6.23548746e-01 4.30567622e-01 -1.17073512e+00 -6.72552049e-01 -1.28010467e-01 -4.00763452e-01 3.30777973e-01 5.78262568e-01 3.80328819e-02 -9.21542287e-01 9.92112607e-02 1.44865453e-01 1.04034409e-01 -9.89491165e-01 -3.55595529e-01 1.52931899e-01 -6.06587112e-01 -1.32617664e+00 -1.24643981e-01 -8.58318090e-01 8.53628159e-01 1.06680632e-01 1.21877527e+00 3.48923802e-01 -2.27647200e-01 5.03629982e-01 -5.12509763e-01 3.79708670e-02 -5.65256774e-02 2.50719875e-01 5.39832711e-02 3.35395515e-01 4.44976538e-01 -7.47494280e-01 -3.79638433e-01 8.63509029e-02 -6.34761214e-01 -1.77487239e-01 5.40629685e-01 7.95789599e-01 5.28574109e-01 2.59692967e-01 7.04571903e-01 -1.18265486e+00 8.13629031e-01 -9.26465213e-01 -3.81556481e-01 4.88021076e-01 -7.58645713e-01 3.52727503e-01 7.25052536e-01 -1.63662940e-01 -3.68345499e-01 -2.84113318e-01 2.66432434e-01 -3.22026253e-01 3.68503392e-01 1.19583118e+00 -1.75816789e-01 -1.20592676e-01 4.71365988e-01 -4.10455205e-02 -2.92633623e-01 -1.45083547e-01 6.84712589e-01 1.25891536e-01 2.10308805e-01 -3.93474013e-01 1.11538136e+00 1.23165749e-01 4.44798321e-01 -7.29506791e-01 -6.07909799e-01 -5.33251166e-01 -5.88112116e-01 -1.72057152e-01 6.70113087e-01 -5.58718085e-01 -1.05574250e+00 1.44550204e-02 -8.57281387e-01 -9.53559205e-02 -3.07446808e-01 4.26815718e-01 1.16153918e-02 8.15006271e-02 -4.69468385e-01 -2.94712126e-01 -2.75027961e-01 -5.71305454e-01 5.41820228e-01 2.18582794e-01 -1.46149889e-01 -1.54136181e+00 2.48275593e-01 -6.71296567e-02 2.92957306e-01 4.49018985e-01 1.49880791e+00 -9.78168964e-01 -6.98908508e-01 -3.57819378e-01 -4.65674043e-01 -1.16352431e-01 1.83029562e-01 6.89542294e-02 -4.30924416e-01 -1.81976408e-01 -1.17310607e+00 2.02025548e-02 1.03774691e+00 2.37243995e-01 7.12442935e-01 -4.72431511e-01 -8.48312140e-01 5.43718934e-01 1.53715420e+00 -4.30023342e-01 4.48609859e-01 1.22104451e-01 1.02502155e+00 3.57939959e-01 2.10985661e-01 2.83965558e-01 7.43624508e-01 5.85118890e-01 6.51049197e-01 4.01588678e-02 5.12595773e-02 -7.47881949e-01 1.16109915e-01 7.13902175e-01 -2.40585461e-01 -4.97550070e-01 -8.62918973e-01 7.39729166e-01 -1.83313847e+00 -8.94444168e-01 -2.72826821e-01 2.04316235e+00 6.16834164e-01 1.30038327e-02 1.04800880e-01 -1.50874536e-02 7.99932003e-01 3.61569405e-01 -5.16175091e-01 -3.57678950e-01 7.89475963e-02 2.14787960e-01 4.78911072e-01 4.64885294e-01 -9.53287244e-01 9.98085797e-01 4.77783775e+00 4.44983989e-01 -7.60778189e-01 -1.37515932e-01 3.29058677e-01 2.93377638e-01 -6.35365009e-01 -2.93510873e-02 -6.18103504e-01 3.58276993e-01 9.26051617e-01 -5.26754141e-01 5.64118803e-01 7.45271802e-01 -2.23364711e-01 1.77222848e-01 -1.34997451e+00 5.79882324e-01 -2.04509079e-01 -1.62911773e+00 2.31665388e-01 3.79296005e-01 7.04763293e-01 1.28550917e-01 -1.06445096e-01 4.59190696e-01 4.95908618e-01 -1.12939274e+00 1.53654888e-01 6.63431287e-01 6.78850532e-01 -8.22505414e-01 8.09940338e-01 2.64036823e-02 -1.58739221e+00 -1.27865136e-01 -4.01228666e-01 6.90557584e-02 -3.85120809e-02 9.37932074e-01 -1.06806684e+00 8.61131549e-01 4.48678911e-01 9.91498590e-01 -6.61715329e-01 7.34504223e-01 -5.55998981e-01 6.28916323e-01 -3.41503292e-01 -2.45773196e-01 3.46253037e-01 -2.89387316e-01 4.68611449e-01 1.10474765e+00 1.57425404e-01 2.41503850e-01 4.77881096e-02 8.77178311e-01 -7.68101871e-01 1.80573449e-01 -8.72684598e-01 -4.53201354e-01 8.54271352e-01 1.39213836e+00 -5.55692732e-01 -1.40006512e-01 -3.21761936e-01 6.29739881e-01 1.02240646e+00 4.20305818e-01 -6.31853342e-01 -6.65226400e-01 8.51544321e-01 1.39956281e-01 4.77795541e-01 -8.47646222e-02 1.87459096e-01 -8.67372036e-01 -3.68898571e-03 -1.45841211e-01 6.64343894e-01 -6.48538470e-01 -1.47940648e+00 4.56588715e-01 -3.73907745e-01 -7.22792327e-01 3.31154168e-02 -6.24167264e-01 -7.79739738e-01 6.71300292e-01 -1.56430912e+00 -1.09966683e+00 -4.01273727e-01 6.02529883e-01 -4.05991733e-01 -1.31718263e-01 9.76310492e-01 1.46076038e-01 -6.09434783e-01 7.69518971e-01 2.34415725e-01 6.96365476e-01 2.01633736e-01 -1.43431497e+00 4.30065989e-01 3.68655473e-01 5.76303780e-01 7.13289917e-01 2.78039634e-01 -6.14298344e-01 -1.59512305e+00 -1.36238122e+00 1.22934568e+00 -4.47207838e-01 9.87906933e-01 -2.05444872e-01 -8.44348133e-01 9.66714084e-01 -1.32279158e-01 5.95326006e-01 7.37517297e-01 7.03548372e-01 -6.57526493e-01 -1.35955691e-01 -8.94160151e-01 4.19805616e-01 1.21852267e+00 -6.54640436e-01 -4.38126922e-01 1.96273521e-01 8.53868544e-01 -7.59615796e-03 -1.32651079e+00 1.78127065e-01 5.14385998e-01 -6.54053926e-01 1.15299690e+00 -1.05175531e+00 6.14550769e-01 -1.16888553e-01 5.10749929e-02 -1.85699010e+00 -8.35254192e-01 -2.27873102e-01 -6.27070427e-01 1.10675144e+00 7.19437897e-01 -9.19487536e-01 8.31267715e-01 3.97850096e-01 2.02843085e-01 -1.03701675e+00 -8.21533680e-01 -5.95699430e-01 -1.76548958e-01 1.38652911e-02 5.96742511e-01 1.37432754e+00 2.96228796e-01 8.66405249e-01 -7.59821609e-02 4.05295312e-01 4.34054255e-01 2.08016872e-01 4.78252739e-01 -1.81427598e+00 -1.49163783e-01 -4.99735206e-01 -1.22846889e+00 -6.24397814e-01 4.39298064e-01 -1.56384814e+00 -5.13105392e-01 -1.96352792e+00 1.57163367e-01 -5.79381406e-01 -5.34252822e-01 7.11572528e-01 -2.42332101e-01 -1.78344697e-02 2.19336580e-02 -3.14643800e-01 -7.11720705e-01 6.29244089e-01 9.06430602e-01 -3.67870897e-01 -2.95312077e-01 -2.42237255e-01 -6.75770581e-01 4.18829292e-01 6.74823403e-01 -3.71149391e-01 -4.61890966e-01 -2.66709179e-01 8.02053332e-01 -1.91139504e-01 4.78468508e-01 -6.31070077e-01 4.99080986e-01 -8.33622273e-03 3.05027455e-01 -2.07885757e-01 1.91133276e-01 -8.68560851e-01 1.06762707e-01 2.61757076e-01 -4.25771624e-01 -1.28509596e-01 -1.65899247e-01 1.08468127e+00 -1.60982162e-01 3.34450901e-02 2.35651970e-01 2.80562222e-01 -8.09887588e-01 5.47490299e-01 1.55666888e-01 1.55009016e-01 9.95329261e-01 -7.62610584e-02 -5.29226780e-01 -4.03886348e-01 -7.82595575e-01 3.68331790e-01 1.32816225e-01 4.63154078e-01 6.46504581e-01 -1.51958036e+00 -5.94228208e-01 -4.79262788e-03 4.43034559e-01 -1.27569705e-01 1.38632320e-02 7.87704587e-01 -1.83307424e-01 5.32439709e-01 1.29171729e-01 -1.94449738e-01 -1.20521140e+00 6.11747384e-01 3.00496817e-01 -6.20819867e-01 -5.25120199e-01 9.70905185e-01 1.01898864e-01 -5.96324801e-01 2.58576069e-02 -1.85897171e-01 -4.26137149e-01 2.08341211e-01 1.45148695e-01 4.55217004e-01 -1.71671081e-02 -6.77465916e-01 -4.86698866e-01 3.96692157e-01 -5.34789078e-02 5.26306629e-01 1.55284524e+00 2.39673138e-01 -3.33470613e-01 1.77072883e-01 1.37536752e+00 1.06625415e-01 -5.30625761e-01 -7.32245088e-01 1.64061248e-01 -4.14886832e-01 2.20210329e-01 -7.07130909e-01 -1.28917253e+00 5.88302195e-01 -2.11275686e-02 4.71710593e-01 6.96605206e-01 4.12504405e-01 5.20316958e-01 6.99948788e-01 3.40810895e-01 -7.98130870e-01 -6.87872767e-02 5.19117236e-01 5.46300411e-01 -9.94500577e-01 4.18460369e-02 -6.23271644e-01 -2.17095032e-01 1.05259740e+00 4.55538183e-01 6.02161475e-02 1.02124119e+00 -3.59299242e-01 -5.83082974e-01 -5.96682310e-01 -7.49711633e-01 -4.28754300e-01 5.90716660e-01 5.44820666e-01 2.19492540e-01 4.97104764e-01 -1.58013582e-01 5.57375491e-01 -9.67255086e-02 -2.20853075e-01 2.68742532e-01 4.43761587e-01 -4.11066294e-01 -8.70935202e-01 2.57669836e-01 8.68461967e-01 1.01581685e-01 -2.53195405e-01 -6.37047827e-01 7.28908360e-01 -1.26294810e-02 6.74200356e-01 4.33500111e-02 -7.24662721e-01 3.19206238e-01 -1.41081557e-01 2.09222063e-01 -5.76449990e-01 -2.73687214e-01 -8.56173337e-01 1.78295225e-01 -3.90752167e-01 -5.81580959e-02 -1.98710799e-01 -1.40148795e+00 -6.30349636e-01 -5.00122845e-01 4.24513102e-01 4.42904711e-01 8.24188590e-01 6.89863741e-01 6.93580091e-01 7.14578509e-01 -4.57911938e-01 -2.02687606e-01 -7.30698109e-01 -8.53438675e-01 5.47013223e-01 5.08801602e-02 -7.57991552e-01 -3.12350720e-01 -4.17224854e-01]
[8.60036849975586, 7.723392486572266]
8f6ae798-c09b-4c11-a43b-b2e561bbead9
predicting-the-effects-of-news-sentiments-on
1812.04199
null
http://arxiv.org/abs/1812.04199v1
http://arxiv.org/pdf/1812.04199v1.pdf
Predicting the Effects of News Sentiments on the Stock Market
Stock market forecasting is very important in the planning of business activities. Stock price prediction has attracted many researchers in multiple disciplines including computer science, statistics, economics, finance, and operations research. Recent studies have shown that the vast amount of online information in the public domain such as Wikipedia usage pattern, news stories from the mainstream media, and social media discussions can have an observable effect on investors opinions towards financial markets. The reliability of the computational models on stock market prediction is important as it is very sensitive to the economy and can directly lead to financial loss. In this paper, we retrieved, extracted, and analyzed the effects of news sentiments on the stock market. Our main contributions include the development of a sentiment analysis dictionary for the financial sector, the development of a dictionary-based sentiment analysis model, and the evaluation of the model for gauging the effects of news sentiments on stocks for the pharmaceutical market. Using only news sentiments, we achieved a directional accuracy of 70.59% in predicting the trends in short-term stock price movement.
['Haruna Isah', 'Farhana Zulkernine', 'Dev Shah']
2018-12-11
null
null
null
null
['stock-market-prediction', 'stock-price-prediction']
['time-series', 'time-series']
[-8.78810048e-01 -3.41544330e-01 -7.70740807e-01 -1.02370963e-01 -1.18782170e-01 -7.04487026e-01 6.82970464e-01 4.61044908e-01 -5.41924417e-01 6.76281214e-01 8.47710729e-01 -7.18989134e-01 2.46955425e-01 -1.16796887e+00 -4.80606228e-01 -2.73188710e-01 4.72609922e-02 -1.15716718e-01 3.80388737e-01 -7.04438448e-01 1.16380477e+00 1.60820767e-01 -1.25441432e+00 1.59848452e-01 4.88894641e-01 1.35208976e+00 5.90701252e-02 -3.95153165e-02 -7.84663439e-01 1.42221415e+00 -5.62335491e-01 -8.21352422e-01 3.57766062e-01 -1.42957076e-01 -1.50437802e-01 -2.13304818e-01 -3.69236171e-01 -3.32496643e-01 2.05302790e-01 1.25974572e+00 3.38763475e-01 -5.00144243e-01 3.08984965e-01 -8.48672748e-01 -7.73593426e-01 9.39810753e-01 -5.75854778e-01 7.07806230e-01 2.20368162e-01 -2.25119129e-01 1.39431238e+00 -1.00978446e+00 9.49326277e-01 6.16615236e-01 4.48952764e-01 -1.59477234e-01 -2.87367165e-01 -1.18730485e+00 2.17983469e-01 -3.51401381e-02 -6.32374048e-01 -5.80653585e-02 8.29876125e-01 -9.08472776e-01 8.28126013e-01 2.41411239e-01 1.26980650e+00 4.79415447e-01 8.92767906e-01 5.08694172e-01 1.41725230e+00 -2.92872429e-01 3.00663769e-01 7.07894027e-01 2.74496824e-01 -9.66935009e-02 6.29797459e-01 5.42339012e-02 -9.43214715e-01 -3.03729147e-01 3.65399450e-01 2.83944547e-01 1.08774804e-01 3.97480130e-01 -9.97321844e-01 1.34877861e+00 1.60258695e-01 5.61122179e-01 -7.32673585e-01 -3.66881043e-01 4.04719144e-01 6.66616440e-01 1.38631332e+00 5.67133009e-01 -8.83908272e-01 -3.35220277e-01 -6.82589293e-01 5.80435514e-01 1.17379093e+00 4.10680681e-01 1.98737517e-01 -2.10828017e-02 2.32141107e-01 4.90703255e-01 6.02877915e-01 8.15400243e-01 1.00734329e+00 -2.43608311e-01 5.62075853e-01 7.19665110e-01 4.39194381e-01 -1.57660115e+00 -3.58424693e-01 -5.36021471e-01 -1.50817931e-01 2.64332946e-02 1.61787137e-01 -5.38687229e-01 -2.83783108e-01 1.11248469e+00 4.67500202e-02 -1.57690644e-01 1.80207297e-01 6.10037386e-01 7.75218070e-01 8.36766005e-01 8.27120245e-02 -4.70001012e-01 1.22364998e+00 -6.03885770e-01 -1.08962190e+00 -2.77108252e-01 5.39438426e-01 -1.14262938e+00 3.81573826e-01 2.41203487e-01 -1.13032722e+00 -2.38458626e-02 -8.54680240e-01 4.88677621e-01 -7.47484267e-01 -5.04519701e-01 6.46335542e-01 4.97019917e-01 -7.27363825e-01 5.38903356e-01 -4.72497702e-01 1.57544240e-01 1.92574486e-01 -8.03651661e-03 3.90803963e-01 5.54813743e-01 -1.55569708e+00 1.19971585e+00 2.34124005e-01 -2.70170301e-01 3.72335076e-01 -6.81725621e-01 -5.61095536e-01 -1.72466248e-01 -5.47385216e-03 -1.30516902e-01 1.14679933e+00 -1.23981905e+00 -1.18683267e+00 5.40082812e-01 9.86836776e-02 -7.21449375e-01 4.07099038e-01 -7.84298554e-02 -8.77329946e-01 -3.65717411e-01 3.87089163e-01 -2.00116947e-01 3.49894792e-01 -3.29151660e-01 -1.14776993e+00 -4.65394080e-01 -1.91065326e-01 5.73459528e-02 -5.11226714e-01 5.70684791e-01 8.10587183e-02 -1.07993340e+00 1.54672146e-01 -6.72484457e-01 -1.82061076e-01 -7.94099092e-01 9.61359665e-02 -2.65996695e-01 1.68542966e-01 -8.75489593e-01 1.45516014e+00 -1.89964807e+00 -5.66881359e-01 4.61301893e-01 -1.07381731e-01 -2.53515601e-01 5.27117252e-01 6.85526967e-01 -4.36179563e-02 5.66001952e-01 4.42729443e-01 4.97017413e-01 2.49326811e-04 -4.00893956e-01 -1.05381799e+00 3.06298584e-01 -5.50739095e-02 1.01881886e+00 -6.31956220e-01 1.34307459e-01 -7.31070712e-02 -4.87550767e-03 -3.98576975e-01 -3.21169376e-01 -1.78674445e-01 -1.15957841e-01 -7.67159462e-01 7.17117071e-01 4.40922618e-01 -3.90572935e-01 9.59945098e-02 1.89081609e-01 -9.32506323e-01 8.91109407e-01 -8.66816223e-01 5.15074134e-01 6.57087043e-02 6.14665747e-01 -6.05209410e-01 -4.30783093e-01 1.07756054e+00 2.07004517e-01 5.56032062e-01 -1.24630797e+00 2.05812797e-01 6.78167522e-01 -1.75136644e-02 -2.28825942e-01 7.97369540e-01 -6.21837735e-01 -8.04597437e-02 6.57363832e-01 -6.75562382e-01 7.17046019e-03 4.26262289e-01 -9.08154994e-02 4.27825421e-01 -4.12888795e-01 4.60074425e-01 -4.40005362e-01 1.96951210e-01 3.58769268e-01 4.73300457e-01 -2.23352313e-02 4.04626764e-02 -9.12664682e-02 7.27604032e-01 -7.37552881e-01 -7.30230808e-01 3.73746492e-02 -2.76889503e-01 8.31023633e-01 -3.59930890e-03 -2.73991585e-01 -3.22196513e-01 -2.17275582e-02 4.89392817e-01 8.58713210e-01 -5.73580801e-01 2.78692096e-01 -1.47234812e-01 -1.34963679e+00 -2.50266969e-01 3.28660160e-01 3.80512625e-01 -1.01679671e+00 -4.91418868e-01 4.46948022e-01 1.58813447e-01 -8.31654310e-01 -2.92168468e-01 -6.50034845e-02 -9.18230116e-01 -1.13310528e+00 -9.17984903e-01 -4.95787293e-01 2.78609753e-01 5.39070480e-02 1.02059853e+00 -1.69779241e-01 6.29107952e-01 -1.85399592e-01 -3.88497621e-01 -1.53099251e+00 -2.81379163e-01 7.71820545e-02 -5.61267659e-02 -6.85609803e-02 7.88171411e-01 7.78164119e-02 -5.35228610e-01 2.43521333e-01 -8.05712283e-01 -3.41098636e-01 3.22255522e-01 3.38450193e-01 2.15645671e-01 3.41613799e-01 1.24693108e+00 -1.23893821e+00 9.90982771e-01 -1.01915061e+00 -1.04152620e+00 -2.29596540e-01 -1.36104059e+00 -2.43605644e-01 -1.23420008e-01 -1.16803907e-01 -1.03700888e+00 -6.74559116e-01 -2.47763321e-01 4.86505210e-01 6.42384529e-01 1.66620886e+00 7.44168162e-01 1.48849681e-01 3.01285625e-01 1.33908197e-01 2.63055176e-01 -3.35453987e-01 -2.78677762e-01 6.52991712e-01 -2.51147926e-01 3.58813703e-01 6.55416012e-01 3.88155043e-01 -4.72222686e-01 -6.28618717e-01 -1.01333582e+00 -7.39658773e-01 -4.18827496e-02 -4.04855698e-01 5.27695656e-01 -1.30770636e+00 -2.93737829e-01 9.42294776e-01 -7.79577374e-01 6.86565191e-02 -1.34915486e-01 6.97697222e-01 1.45002097e-01 -1.79844514e-01 -8.07194352e-01 -9.94472384e-01 -4.70314384e-01 -8.85705471e-01 3.18011902e-02 1.43388331e-01 -4.80549157e-01 -1.43870556e+00 4.81973559e-01 5.06107152e-01 8.09357226e-01 2.56769240e-01 6.60106421e-01 -1.16961825e+00 -4.63618428e-01 -4.96763617e-01 1.58547461e-01 2.21859485e-01 2.49220207e-01 -1.70985367e-02 -4.92612273e-01 8.42113793e-02 5.59613347e-01 -9.21746120e-02 7.65443802e-01 6.85117602e-01 1.26198798e-01 -4.97912169e-01 2.18438823e-02 -7.04185143e-02 1.34778249e+00 6.34866893e-01 5.71496964e-01 1.22252023e+00 7.06492215e-02 7.49776483e-01 7.98185229e-01 8.93600881e-01 3.87220025e-01 6.46046400e-02 9.82093289e-02 2.30156094e-01 8.05433631e-01 -7.69320726e-02 7.85417318e-01 1.25086653e+00 -1.41418383e-01 1.16298713e-01 -7.94986844e-01 4.53868419e-01 -1.44140100e+00 -1.10779583e+00 -1.32603362e-01 1.80406260e+00 7.71565676e-01 6.54009163e-01 3.66586000e-01 1.40059784e-01 4.82986301e-01 4.00640011e-01 -4.76778656e-01 -1.86703578e-01 -3.79774243e-01 -7.88093954e-02 1.16230917e+00 1.72707498e-01 -7.30501175e-01 6.52255893e-01 6.96915960e+00 3.06861937e-01 -1.39679730e+00 -2.65989095e-01 9.12913442e-01 1.85125638e-02 -9.17625666e-01 -3.23049635e-01 -1.14934552e+00 7.64897346e-01 1.02828002e+00 -8.38576853e-01 -3.25005531e-01 9.14520562e-01 5.91313660e-01 -3.87623519e-01 -2.32105285e-01 4.11525756e-01 5.76464608e-02 -1.88330460e+00 -5.08149453e-02 4.57440108e-01 9.16250765e-01 3.64923865e-01 4.14169163e-01 -1.69817116e-02 1.01203129e-01 -4.23899651e-01 1.29566967e+00 5.20321965e-01 2.39002500e-02 -8.65084231e-01 1.41205847e+00 4.00706857e-01 -7.74986207e-01 -1.09575972e-01 -3.04921746e-01 -6.49053574e-01 1.81461334e-01 8.84179533e-01 -6.79873228e-01 8.57819691e-02 6.50538683e-01 1.12079644e+00 -2.76818007e-01 5.71705163e-01 8.54114071e-02 8.17261934e-01 3.43181635e-03 -5.89456558e-01 2.94162244e-01 -5.59162259e-01 2.19360203e-01 7.05364645e-01 5.01547992e-01 3.10089827e-01 -3.19055408e-01 5.22814274e-01 -3.61049436e-02 7.33966947e-01 -6.34844542e-01 -6.64881527e-01 1.88682079e-01 6.37442112e-01 -9.90362406e-01 -1.90498441e-01 -1.19253182e+00 1.92487657e-01 -3.17925602e-01 1.10639401e-01 -2.72666305e-01 -5.70515841e-02 6.24103248e-01 5.96211076e-01 3.29919755e-01 2.02700391e-01 -7.07098365e-01 -1.18258286e+00 7.24280700e-02 -8.03359926e-01 3.13084781e-01 -5.32194674e-01 -1.30294406e+00 2.37987369e-01 -3.75558168e-01 -1.14095795e+00 -1.93816885e-01 -8.51774454e-01 -7.24473774e-01 1.04479110e+00 -2.04352236e+00 -2.86919683e-01 5.85707068e-01 1.68548614e-01 3.00189167e-01 -7.11921096e-01 4.92681801e-01 6.07165918e-02 -4.06408548e-01 -2.92861640e-01 3.29746783e-01 2.91697264e-01 5.33304572e-01 -1.06981957e+00 6.37828767e-01 5.51883399e-01 -1.85904905e-01 5.97968936e-01 7.25163579e-01 -1.31336379e+00 -1.03919804e+00 -6.86104178e-01 1.55593085e+00 -4.34669763e-01 1.55521548e+00 3.39458495e-01 -5.83071053e-01 5.56280375e-01 3.51864249e-01 -8.04932356e-01 1.30669224e+00 7.87209496e-02 -5.32131875e-03 -5.58066107e-02 -8.24494839e-01 5.04505694e-01 -1.19245656e-01 -2.39144683e-01 -7.44670749e-01 1.93098992e-01 6.44302666e-01 -3.74044299e-01 -1.05823481e+00 -1.37463346e-01 7.48305380e-01 -9.98115301e-01 7.23385036e-01 -4.99747425e-01 7.02556372e-01 2.44298950e-01 -7.14853480e-02 -1.25079477e+00 -2.84291536e-01 -3.05713862e-01 3.46512377e-01 7.85133064e-01 1.07261074e+00 -1.25964677e+00 7.49644399e-01 1.25551081e+00 2.34700128e-01 -9.59273577e-01 -4.70335454e-01 -2.91025609e-01 2.50682771e-01 -4.96188253e-01 8.90171468e-01 1.07464218e+00 5.44566989e-01 5.25041819e-02 -1.01164564e-01 -2.01083586e-01 3.92329842e-02 3.57373059e-01 5.19182146e-01 -1.33849716e+00 1.80410463e-02 -7.26819992e-01 -2.08183125e-01 -5.36240160e-01 -7.46423453e-02 -6.59011722e-01 -8.24279487e-01 -1.28162897e+00 6.81447759e-02 -1.08032063e-01 -5.84652245e-01 -1.66050777e-01 5.46462461e-02 -4.48185094e-02 2.64977723e-01 7.34267592e-01 7.27416575e-02 1.06058754e-01 1.29601502e+00 2.55347937e-02 -2.90367663e-01 2.91685998e-01 -1.24664366e+00 1.01006365e+00 9.58159387e-01 -3.96542370e-01 -3.60079706e-02 1.66757524e-01 1.53225422e+00 4.12618443e-02 -2.61283666e-01 -1.12250462e-01 2.10695669e-01 -5.66276789e-01 3.45648915e-01 -9.11201060e-01 -2.58356988e-01 -7.43826747e-01 1.98787525e-01 7.40354240e-01 -3.16031605e-01 6.36688232e-01 2.31249303e-01 3.23871732e-01 -7.24811673e-01 -3.55991244e-01 1.18778631e-01 -2.42189214e-01 -6.37620211e-01 1.15922980e-01 -6.44532442e-01 1.16598502e-01 9.89241838e-01 -1.48803324e-01 -3.16897333e-01 -5.99879920e-01 -2.16712981e-01 1.69224367e-01 2.64533520e-01 4.53276396e-01 3.58575612e-01 -1.20241928e+00 -7.59008706e-01 9.08598974e-02 -9.59724486e-02 -4.73844737e-01 -3.75401169e-01 6.61214054e-01 -7.53679693e-01 7.86147654e-01 -9.60460231e-02 3.56516749e-01 -7.49045551e-01 3.67993087e-01 9.66135506e-03 -3.13107967e-01 -9.06935632e-02 6.39195025e-01 -8.08335841e-02 3.00822198e-01 -1.45289451e-01 -7.15189278e-01 -8.21705401e-01 1.10966802e+00 8.42374027e-01 5.47142029e-01 -2.47535715e-03 -7.77725875e-01 -1.73583016e-01 5.01868963e-01 -2.12073207e-01 -3.05959851e-01 1.83054781e+00 -2.34089389e-01 -4.51919436e-01 1.17521131e+00 8.26981008e-01 4.11227047e-01 -4.94971931e-01 -1.21417269e-01 6.84323430e-01 -3.59610140e-01 3.56864572e-01 -7.77596831e-01 -1.30324495e+00 8.13815685e-04 1.19995438e-02 8.80685985e-01 6.85665190e-01 -1.78641453e-01 9.50508058e-01 2.31688321e-01 1.50950387e-01 -1.54123211e+00 -2.15872809e-01 5.23478329e-01 9.65034842e-01 -1.39347768e+00 1.84250131e-01 -1.76704407e-01 -9.99668956e-01 1.07487345e+00 -1.61720857e-01 -9.10279602e-02 1.63908100e+00 2.75679350e-01 7.08024979e-01 -5.21660984e-01 -7.28113174e-01 1.92179814e-01 3.31889600e-01 -3.21104109e-01 8.49152565e-01 1.12465307e-01 -7.53006995e-01 1.06286478e+00 -6.25508189e-01 2.65340328e-01 7.74287283e-01 9.30885077e-01 -6.69856787e-01 -1.02066004e+00 -3.73806596e-01 8.50503325e-01 -1.46885800e+00 -3.60846192e-01 -5.24669528e-01 7.72753119e-01 -3.17946702e-01 7.72305727e-01 3.48985612e-01 -1.66639999e-01 4.76698846e-01 2.15388820e-01 -6.18679345e-01 -5.10896087e-01 -7.66095400e-01 3.86012137e-01 3.15520257e-01 -2.55908459e-01 -8.93685460e-01 -1.28613424e+00 -9.95932221e-01 -5.80150187e-01 -6.15020037e-01 3.23036194e-01 9.19687092e-01 1.02786970e+00 1.96850568e-01 1.76312432e-01 9.42449510e-01 -3.27948555e-02 -7.04560757e-01 -9.54755962e-01 -1.17923248e+00 1.59479409e-01 2.27362499e-01 -4.60956365e-01 -6.76068544e-01 -6.68823421e-02]
[4.470396995544434, 4.377518653869629]
74f080d2-be01-426a-ac57-752106132835
vsec-transformer-based-model-for-vietnamese
2111.00640
null
https://arxiv.org/abs/2111.00640v2
https://arxiv.org/pdf/2111.00640v2.pdf
VSEC: Transformer-based Model for Vietnamese Spelling Correction
Spelling error correction is one of topics which have a long history in natural language processing. Although previous studies have achieved remarkable results, challenges still exist. In the Vietnamese language, a state-of-the-art method for the task infers a syllable's context from its adjacent syllables. The method's accuracy can be unsatisfactory, however, because the model may lose the context if two (or more) spelling mistakes stand near each other. In this paper, we propose a novel method to correct Vietnamese spelling errors. We tackle the problems of mistyped errors and misspelled errors by using a deep learning model. The embedding layer, in particular, is powered by the byte pair encoding technique. The sequence to sequence model based on the Transformer architecture makes our approach different from the previous works on the same problem. In the experiment, we train the model with a large synthetic dataset, which is randomly introduced spelling errors. We test the performance of the proposed method using a realistic dataset. This dataset contains 11,202 human-made misspellings in 9,341 different Vietnamese sentences. The experimental results show that our method achieves encouraging performance with 86.8% errors detected and 81.5% errors corrected, which improves the state-of-the-art approach 5.6% and 2.2%, respectively.
['Dinh Hieu Vo', 'Thang Ngoc Bui', 'Ha Thanh Nguyen', 'Dinh-Truong Do']
2021-11-01
null
null
null
null
['spelling-correction']
['natural-language-processing']
[ 3.88170868e-01 -5.07055700e-01 2.56243169e-01 -3.46541882e-01 -8.25300336e-01 -2.43742049e-01 2.72494048e-01 2.71836579e-01 -8.40610385e-01 9.40500200e-01 1.57273456e-01 -2.67983645e-01 4.51886296e-01 -6.95792317e-01 -8.86096001e-01 -6.67176306e-01 2.23916411e-01 1.88952237e-01 4.63735461e-01 -3.29544842e-01 7.45465457e-01 -4.58278209e-02 -1.45561337e+00 4.22491819e-01 1.19529366e+00 5.43029249e-01 4.25179482e-01 6.47186279e-01 -2.08614692e-01 4.94358063e-01 -9.82111037e-01 -7.51721919e-01 8.01517069e-02 -4.15000468e-01 -5.11529028e-01 -1.92310482e-01 1.84926346e-01 -1.41739577e-01 -7.18744546e-02 1.46414888e+00 7.53326714e-01 -4.05102968e-02 4.79472756e-01 -8.78103793e-01 -9.70831275e-01 6.92218721e-01 -5.90040088e-01 3.94916713e-01 1.61585316e-01 5.27887903e-02 9.10617828e-01 -1.23682845e+00 2.69637316e-01 9.34940994e-01 7.62570858e-01 6.91072524e-01 -6.65819228e-01 -6.45531356e-01 -1.12670928e-01 6.23957336e-01 -1.69532514e+00 -3.08028519e-01 6.09724164e-01 -3.21597010e-01 1.20868874e+00 2.98329264e-01 4.23241615e-01 9.60314095e-01 5.53000450e-01 7.60522246e-01 1.07760930e+00 -7.54725218e-01 7.97810126e-03 1.42874688e-01 2.48149201e-01 5.52672923e-01 5.39867520e-01 2.06963241e-01 -4.79758710e-01 2.52707243e-01 1.93525925e-01 -1.44638300e-01 -3.51676434e-01 3.16788435e-01 -1.09525251e+00 6.68418169e-01 1.29055306e-01 4.33724672e-01 -1.29383937e-01 -1.99032575e-01 5.23922086e-01 1.46470577e-01 2.44704902e-01 1.10607259e-01 -4.57484245e-01 -2.79994965e-01 -1.07148302e+00 2.92893827e-01 5.47061980e-01 9.93460119e-01 5.49518645e-01 2.54060209e-01 -2.44408339e-01 9.68035281e-01 2.14059740e-01 4.40168500e-01 9.06977713e-01 -1.54177370e-02 8.08545649e-01 2.93144792e-01 1.79386765e-01 -1.15734005e+00 -5.72083965e-02 -5.86109340e-01 -9.33690488e-01 -1.27880648e-01 3.08789015e-01 -7.78430253e-02 -1.01754165e+00 1.41333175e+00 -5.21917939e-02 5.98460920e-02 1.78315625e-01 9.48621333e-01 5.45682132e-01 1.11385751e+00 -6.24479316e-02 -3.74975592e-01 1.15381813e+00 -1.31731129e+00 -1.08952332e+00 -3.68794680e-01 6.16921842e-01 -1.18668401e+00 1.34340906e+00 4.90467012e-01 -9.71988142e-01 -6.64412558e-01 -1.23805666e+00 -5.58917560e-02 -4.41636801e-01 3.63861442e-01 -4.86103259e-02 8.12355340e-01 -6.98195934e-01 5.58331668e-01 -4.10345286e-01 -1.26427814e-01 -1.05921440e-01 1.48660734e-01 -1.33505791e-01 -4.70650010e-02 -1.50604141e+00 9.16950285e-01 8.53341877e-01 3.39978427e-01 -6.33117557e-01 -5.12645543e-01 -7.12635219e-01 2.65102565e-01 2.83786774e-01 -1.31774440e-01 1.29476988e+00 -9.12080824e-01 -1.28474545e+00 7.27332711e-01 -3.85616273e-01 -4.38984573e-01 4.89835054e-01 -3.32036883e-01 -9.32734072e-01 -3.56835485e-01 3.79424840e-02 8.92107263e-02 6.41748071e-01 -9.35799181e-01 -9.61496830e-01 -3.07357401e-01 -5.65575838e-01 6.67246878e-02 -3.85886133e-01 3.60396355e-01 -3.55638176e-01 -1.07637370e+00 -2.44590104e-01 -7.46829093e-01 1.16939150e-01 -6.04689837e-01 -3.45795751e-01 -2.28910014e-01 6.09620631e-01 -1.06503975e+00 1.89770985e+00 -2.19270349e+00 -1.33184597e-01 -1.16666608e-01 -3.44815463e-01 9.00067627e-01 5.84819168e-02 6.61930978e-01 3.85774300e-02 2.79355407e-01 -5.80685318e-01 -4.15105432e-01 1.12573244e-01 5.22503033e-02 -3.86844963e-01 2.50572741e-01 2.36073181e-01 6.49052262e-01 -8.68488610e-01 -4.77049053e-01 -3.94330211e-02 4.36035186e-01 -5.10914624e-01 4.20122534e-01 9.70073938e-02 8.76712427e-02 1.60437539e-01 5.61381280e-01 9.24157202e-01 3.43342125e-01 1.53575048e-01 3.65482084e-02 -2.94313580e-01 6.05423331e-01 -1.26920915e+00 1.45562458e+00 -3.18420291e-01 4.79639143e-01 -5.46471953e-01 -7.46520042e-01 1.14043331e+00 2.36808553e-01 -3.22906345e-01 -8.06992710e-01 1.04655884e-02 6.86440110e-01 2.29634464e-01 -6.96019828e-01 8.45772624e-01 -2.20166504e-01 -1.63162425e-01 1.60710290e-01 -1.34085327e-01 2.38443837e-01 2.43683726e-01 -9.94780064e-02 8.40898156e-01 -8.62706602e-02 4.91355151e-01 -1.66244641e-01 8.52280140e-01 -1.30872354e-01 9.99042988e-01 6.44253910e-01 -3.63107264e-01 7.17544615e-01 8.55118185e-02 -4.60912526e-01 -1.20134914e+00 -9.51773405e-01 -1.42975189e-02 8.27276587e-01 1.00607667e-02 -5.97879529e-01 -9.02673662e-01 -6.65766060e-01 -2.96125799e-01 8.89887393e-01 -3.56674641e-01 -3.37150693e-01 -9.40916121e-01 -9.64789212e-01 8.23980927e-01 5.52019358e-01 8.35444748e-01 -1.22228611e+00 -2.93136597e-01 4.08617556e-01 -2.75688946e-01 -1.03027260e+00 -8.03335726e-01 -9.00732279e-02 -4.72298354e-01 -7.25064099e-01 -7.14627445e-01 -1.13212097e+00 4.87682521e-01 1.65877506e-01 8.83195639e-01 3.89418036e-01 7.50006288e-02 -4.66526777e-01 -7.27817655e-01 -4.76095557e-01 -5.15509129e-01 1.55654937e-01 2.49527216e-01 7.86400661e-02 8.22312832e-01 -1.61237448e-01 -3.45648408e-01 5.07124774e-02 -9.95160818e-01 -1.31383389e-01 6.77649021e-01 1.09644449e+00 5.97609460e-01 2.69840539e-01 4.84397858e-01 -9.57883358e-01 6.35802448e-01 -3.86173934e-01 -6.02751195e-01 2.83390909e-01 -9.03034091e-01 1.33351803e-01 1.06795263e+00 -1.33515224e-01 -1.07242632e+00 -1.58471853e-01 -5.43931544e-01 7.85986856e-02 -1.95415273e-01 5.60100377e-01 -3.18126917e-01 1.39818564e-01 2.67072141e-01 7.61779487e-01 -4.39616174e-01 -7.29865134e-01 -2.26782039e-01 1.14548397e+00 7.05058157e-01 -2.98957646e-01 7.44109094e-01 -2.86856711e-01 -4.73802447e-01 -7.08056331e-01 -6.39858603e-01 -2.88855672e-01 -3.72498214e-01 1.44312203e-01 5.96975982e-01 -6.92234278e-01 -3.81716400e-01 7.85444617e-01 -1.43619120e+00 7.23803416e-02 1.98685989e-01 4.14233923e-01 -1.35489419e-01 7.90788054e-01 -6.38349235e-01 -7.87777483e-01 -5.62980473e-01 -1.28011751e+00 8.75803947e-01 2.96593875e-01 6.42908141e-02 -6.14506125e-01 6.65946901e-02 1.44825906e-01 3.14540714e-01 -2.21938744e-01 9.31655705e-01 -7.47015059e-01 -3.56402308e-01 -4.83936220e-02 -3.36711928e-02 5.99742174e-01 4.63160761e-02 2.16251835e-02 -8.47807407e-01 -3.36787879e-01 4.49876487e-02 1.12497739e-01 9.15458262e-01 -1.72350451e-01 1.22122860e+00 -4.55424935e-01 -4.45217378e-02 5.01480043e-01 1.57004452e+00 5.26965261e-01 9.92045105e-01 4.92884845e-01 6.23727024e-01 3.52301270e-01 9.01624322e-01 4.75834191e-01 3.71927440e-01 6.79535329e-01 3.10301483e-01 2.88284838e-01 -1.52315184e-01 -5.36212146e-01 5.04396260e-01 1.53389561e+00 1.01591730e-02 -6.11226618e-01 -1.04321289e+00 7.26616621e-01 -1.64857340e+00 -1.01967096e+00 -3.60384792e-01 2.46363521e+00 1.13893771e+00 4.34221059e-01 -1.76977888e-01 7.00477242e-01 9.53426838e-01 1.87178612e-01 -1.68200716e-01 -9.13122416e-01 -7.52731189e-02 2.80904055e-01 4.51231658e-01 6.53725982e-01 -1.04037666e+00 1.24132764e+00 5.49030304e+00 1.03755677e+00 -1.15054500e+00 1.04231127e-01 3.75926048e-01 2.85284609e-01 -2.70458907e-01 -2.42892832e-01 -1.33536971e+00 1.06273794e+00 9.61385369e-01 -6.08160347e-02 4.20143157e-01 4.49287981e-01 1.85403049e-01 4.90346774e-02 -7.65624940e-01 1.07095647e+00 4.96979415e-01 -1.13240349e+00 8.27692896e-02 -3.29913646e-01 7.15316296e-01 -3.81275862e-01 -2.99141165e-02 4.84924316e-01 -2.30164498e-01 -1.04118824e+00 8.50692213e-01 3.58973265e-01 7.86203861e-01 -1.04498875e+00 1.09077299e+00 6.78313911e-01 -1.01252091e+00 4.80656475e-02 -5.49657106e-01 -2.72043616e-01 1.05158694e-01 7.43806958e-01 -8.40450406e-01 4.81476694e-01 7.56322145e-01 5.07727146e-01 -6.55794442e-01 1.26066267e+00 -5.96619725e-01 8.72936726e-01 2.06989236e-02 -5.13351262e-01 1.82546824e-01 -5.93579710e-02 4.37691689e-01 1.66892433e+00 5.76622784e-01 -7.62747079e-02 -2.65768170e-01 5.59181511e-01 -1.35463193e-01 3.84480417e-01 -5.11602163e-01 4.23974590e-03 6.92508757e-01 7.24355161e-01 -3.13297629e-01 -3.51804793e-01 -3.36579502e-01 1.37394094e+00 5.02563357e-01 1.37291908e-01 -1.00395143e+00 -1.07822514e+00 5.00995815e-01 -2.14939103e-01 5.39086878e-01 -2.93539107e-01 -2.80009270e-01 -1.33910525e+00 4.46042627e-01 -1.07604587e+00 5.85214272e-02 -5.00256658e-01 -1.12908864e+00 6.18527532e-01 -4.75180745e-01 -1.15543056e+00 -8.42460394e-02 -7.22826302e-01 -6.00730181e-01 9.09886897e-01 -1.75744736e+00 -7.50908613e-01 -2.44729947e-02 1.56245872e-01 1.03679645e+00 -3.12421113e-01 6.68969750e-01 6.77280962e-01 -7.66736746e-01 1.10856795e+00 3.06533188e-01 2.35896692e-01 9.61236596e-01 -1.15084112e+00 6.67944193e-01 1.44239604e+00 1.55591995e-01 5.89272380e-01 7.13824272e-01 -7.68090189e-01 -1.20239508e+00 -1.20808733e+00 1.96948457e+00 -1.17770076e-01 3.47929776e-01 -4.07938004e-01 -1.06990278e+00 5.40974021e-01 3.51221859e-01 -1.78397343e-01 5.47900021e-01 -2.53822118e-01 -1.36073351e-01 -1.86062410e-01 -1.05444622e+00 6.55439794e-01 1.04458463e+00 -2.84842968e-01 -9.81588900e-01 7.69416168e-02 7.86624551e-01 -4.50000852e-01 -3.88183862e-01 4.12319034e-01 4.41638440e-01 -1.08848906e+00 5.74457109e-01 -5.69204628e-01 4.39319432e-01 -5.99403143e-01 -3.86275440e-01 -1.59710979e+00 -4.38667864e-01 -4.73680645e-01 3.25642265e-02 1.56459165e+00 4.11506027e-01 -4.59045291e-01 3.46315950e-01 1.43692628e-01 -4.86883074e-01 -6.34938776e-01 -7.85387099e-01 -8.89866948e-01 1.64064705e-01 -3.40845942e-01 9.67399001e-01 7.99096346e-01 -2.81508476e-01 2.60121077e-01 -7.00393915e-01 2.96708882e-01 2.63080895e-01 -1.53687268e-01 4.24932033e-01 -7.58444667e-01 -3.01358402e-01 -2.64836758e-01 -2.86487609e-01 -9.88649309e-01 9.07407925e-02 -8.18502426e-01 3.65402699e-01 -1.05091727e+00 1.21506169e-01 -2.29758784e-01 -3.54714125e-01 2.60659844e-01 -6.54659092e-01 1.66981563e-01 3.33545238e-01 -1.12577975e-01 -2.60081291e-01 7.00491071e-01 8.46473217e-01 -2.64769107e-01 1.06221907e-01 -1.72031671e-01 -4.24345225e-01 7.02386916e-01 1.09972715e+00 -6.45365477e-01 2.82350206e-03 -1.04623985e+00 2.78747767e-01 -3.29091042e-01 6.74596950e-02 -1.18325984e+00 2.31850997e-01 -3.94619582e-03 2.94140846e-01 -5.62074780e-01 4.44936939e-02 -7.37125933e-01 -9.68073606e-02 6.78943574e-01 -2.97931075e-01 6.54973984e-01 1.75745234e-01 4.04530108e-01 -3.67368311e-01 -6.98163390e-01 6.42690122e-01 -7.58593008e-02 -1.02245677e+00 3.06074470e-02 -4.72572565e-01 2.94633865e-01 9.20515180e-01 -1.02405079e-01 -3.53167176e-01 1.30863577e-01 -1.82226405e-01 -1.50910705e-01 2.97596514e-01 5.25086939e-01 9.31198716e-01 -1.43849981e+00 -9.19193387e-01 6.17098749e-01 1.68577343e-01 -3.89453560e-01 8.91402811e-02 5.42082250e-01 -7.57962942e-01 4.07691360e-01 -9.97361764e-02 -2.36448541e-01 -1.29977298e+00 7.04764724e-01 1.26497641e-01 -3.08306992e-01 -2.38591686e-01 8.20915222e-01 -3.65375608e-01 -4.69036162e-01 2.77214319e-01 -3.19219142e-01 -1.65745214e-01 -3.57084662e-01 7.68879652e-01 4.16837424e-01 3.86969656e-01 -7.40318418e-01 -4.50892001e-01 6.80500031e-01 -3.14566404e-01 -3.30658862e-03 8.42945278e-01 -9.66847837e-02 -3.58645208e-02 5.17137587e-01 1.13899684e+00 4.14941430e-01 -5.26629031e-01 -1.33862123e-01 2.60147117e-02 -7.80627251e-01 -1.66775882e-01 -9.82744217e-01 -7.23971367e-01 1.21232080e+00 6.19745791e-01 -1.77318733e-02 1.09329522e+00 -6.87429070e-01 1.44279814e+00 2.28304416e-01 5.09853899e-01 -1.30259943e+00 -2.84832895e-01 9.97720182e-01 7.45246410e-01 -1.40424979e+00 -4.08586502e-01 -3.38442594e-01 -6.67552054e-01 1.10452247e+00 7.43081391e-01 -1.58559516e-01 2.90879935e-01 1.58914968e-01 1.97170135e-02 5.05999446e-01 -5.81176639e-01 9.12924239e-04 4.34376523e-02 4.97502923e-01 7.06277013e-01 2.11525977e-01 -1.07021189e+00 9.42930162e-01 -4.97292072e-01 2.36724503e-02 7.16460228e-01 8.80615354e-01 -6.28376782e-01 -1.40012479e+00 -4.83213544e-01 1.87412962e-01 -6.39611959e-01 -6.26571476e-01 -1.89638108e-01 5.42554855e-01 4.04801428e-01 9.67947841e-01 -2.01097336e-02 -6.52626872e-01 3.71509641e-01 1.53184414e-01 2.68275470e-01 -5.62112629e-01 -6.73421800e-01 -1.70721427e-01 -1.68505255e-02 -1.61634848e-01 5.76537196e-03 -6.24924004e-01 -1.30837107e+00 -4.33049917e-01 -1.48246199e-01 2.72395015e-01 4.79820758e-01 7.89739966e-01 3.25141460e-01 4.22095746e-01 6.00499332e-01 -2.70031273e-01 -9.30464149e-01 -9.77031469e-01 -5.58313489e-01 4.42374170e-01 3.76221836e-01 -4.30347502e-01 -2.85069466e-01 6.63745925e-02]
[10.954957008361816, 10.781373977661133]
fa746870-a77c-418f-ab49-e546f73f8419
multiple-instance-learning-for-content
null
null
https://aclanthology.org/2020.bea-1.3
https://aclanthology.org/2020.bea-1.3.pdf
Multiple Instance Learning for Content Feedback Localization without Annotation
Automated Essay Scoring (AES) can be used to automatically generate holistic scores with reliability comparable to human scoring. In addition, AES systems can provide formative feedback to learners, typically at the essay level. In contrast, we are interested in providing feedback specialized to the content of the essay, and specifically for the content areas required by the rubric. A key objective is that the feedback should be localized alongside the relevant essay text. An important step in this process is determining where in the essay the rubric designated points and topics are discussed. A natural approach to this task is to train a classifier using manually annotated data; however, collecting such data is extremely resource intensive. Instead, we propose a method to predict these annotation spans without requiring any labeled annotation data. Our approach is to consider AES as a Multiple Instance Learning (MIL) task. We show that such models can both predict content scores and localize content by leveraging their sentence-level score predictions. This capability arises despite never having access to annotation training data. Implications are discussed for improving formative feedback and explainable AES models.
['Marcia Derr', 'Peter Foltz', 'Mark Rosenstein', 'William Murray', 'Lee Becker', 'Scott Hellman', 'Adam Wiemerslage']
2020-07-01
null
null
null
ws-2020-7
['automated-essay-scoring']
['natural-language-processing']
[ 3.29717785e-01 2.19720855e-01 -2.70042956e-01 -5.44690967e-01 -1.28693664e+00 -9.54656184e-01 2.32297286e-01 6.19086564e-01 -3.69222909e-01 7.83030987e-01 4.46605712e-01 -3.81313890e-01 -8.87788683e-02 -7.13352323e-01 -4.40668374e-01 -1.25877216e-01 7.07953870e-01 4.96546328e-01 1.23171657e-01 -3.32782924e-01 8.58999670e-01 1.18953511e-01 -1.45047414e+00 6.06067061e-01 1.11205709e+00 7.40493119e-01 2.99676448e-01 9.00157392e-01 -5.48263490e-01 1.07241631e+00 -1.11353087e+00 -5.93651116e-01 -6.27373457e-02 -8.08828950e-01 -1.08143175e+00 2.21656203e-01 7.11746156e-01 -2.40630373e-01 2.47270554e-01 8.08368623e-01 -6.16810992e-02 3.21075588e-01 8.97727549e-01 -8.70262921e-01 -6.10088825e-01 6.87452376e-01 -3.65081310e-01 5.49568981e-02 7.32403040e-01 -2.19998494e-01 1.44378018e+00 -9.84232366e-01 3.80482137e-01 3.78808051e-01 5.27704000e-01 6.72413707e-01 -9.49416459e-01 -2.23406032e-01 4.98248339e-02 3.21104616e-01 -7.45811522e-01 -2.87701845e-01 7.69149780e-01 -6.72707915e-01 5.00643790e-01 3.38422030e-01 5.19188762e-01 7.08509982e-01 -2.43180953e-02 9.71700132e-01 1.12849450e+00 -9.08826351e-01 3.16036105e-01 4.59002674e-01 4.86702621e-01 6.50424421e-01 -2.30192039e-02 -5.68165183e-01 -8.50717485e-01 1.50849476e-01 2.16241628e-01 1.12473881e-02 -3.13488871e-01 4.51942272e-02 -8.02762926e-01 7.92893708e-01 1.44707650e-01 1.66691005e-01 -2.81863242e-01 -1.44111753e-01 4.39990729e-01 5.95463037e-01 5.11168063e-01 1.05556107e+00 -6.90374255e-01 -4.12824124e-01 -1.47807407e+00 3.74850482e-01 1.15905857e+00 7.19947934e-01 6.78436279e-01 -3.28065842e-01 -4.42734599e-01 8.98978710e-01 1.40112355e-01 -1.65913906e-02 5.96085131e-01 -8.98681700e-01 7.55588055e-01 9.78266597e-01 2.66131312e-01 -5.74914634e-01 -2.67233998e-01 -3.19234967e-01 -1.31939501e-01 2.17136189e-01 6.84132993e-01 -2.37061113e-01 -4.98129517e-01 1.41938221e+00 1.41101807e-01 -2.69022524e-01 -1.35723472e-01 8.05252552e-01 8.18668723e-01 5.85525453e-01 1.18963979e-01 -3.08711469e-01 1.43577731e+00 -9.72189963e-01 -6.80209279e-01 -7.93232843e-02 1.00784874e+00 -9.25778329e-01 1.56363416e+00 4.48240936e-01 -1.27539539e+00 -4.21072513e-01 -1.04403460e+00 -2.48430267e-01 -1.32570952e-01 3.39790821e-01 2.09420338e-01 4.69674796e-01 -8.72691870e-01 4.80904013e-01 -4.00171906e-01 -9.02149305e-02 7.01349676e-02 1.55566782e-01 -1.14399314e-01 6.56122491e-02 -1.02495468e+00 1.11276114e+00 -6.12268550e-03 -3.14371765e-01 -3.18551511e-01 -7.99104631e-01 -8.35629582e-01 3.92039686e-01 3.28972638e-01 -3.05999517e-01 1.81387234e+00 -1.19642794e+00 -1.50676560e+00 8.78400624e-01 -3.46479058e-01 -1.31389514e-01 3.71432185e-01 -1.53823972e-01 1.69240758e-01 1.80446029e-01 1.30742952e-01 3.69788915e-01 4.39944834e-01 -8.14155936e-01 -6.32574022e-01 -4.34587419e-01 2.68469751e-01 4.80273455e-01 -8.14404011e-01 2.52957016e-01 -3.30917500e-02 -4.69441921e-01 -2.87271198e-03 -6.96317077e-01 1.02288768e-01 -2.97372162e-01 -1.66791975e-01 -6.68687582e-01 5.65675199e-01 -9.77374554e-01 1.54750526e+00 -1.56578434e+00 -4.69008042e-03 8.50272477e-02 2.55076259e-01 -1.11477682e-02 7.46558309e-02 5.32175541e-01 3.05401176e-01 3.75593430e-03 -8.30212608e-02 -2.84619063e-01 1.06226437e-01 -2.51537114e-01 -4.22157615e-01 -4.71194200e-02 1.13512285e-01 8.23359609e-01 -8.74276161e-01 -6.09113514e-01 -1.60462648e-01 -9.19700041e-02 -6.27134919e-01 5.34088075e-01 -4.33669120e-01 3.89473647e-01 -3.99018914e-01 4.75110531e-01 -9.40719806e-03 -3.79720986e-01 1.52067924e-02 2.57992208e-01 -1.46545798e-01 8.64504099e-01 -7.94072688e-01 1.47790158e+00 -8.23349297e-01 8.75336230e-01 -1.92481011e-01 -9.41974998e-01 1.11438251e+00 4.16126281e-01 4.36656415e-01 -4.85868186e-01 -8.68231580e-02 3.27306241e-01 -5.81888333e-02 -5.79073548e-01 9.30302143e-01 -2.23152801e-01 -2.18931496e-01 1.18749583e+00 9.48911011e-02 -3.00202042e-01 4.18847919e-01 2.84607410e-01 1.25963247e+00 1.44542351e-01 4.45172995e-01 -8.06850032e-04 5.34711063e-01 3.49218190e-01 1.44601256e-01 5.99844217e-01 -1.65080085e-01 5.43069005e-01 5.29454827e-01 -2.46246606e-01 -1.06204629e+00 -6.32576048e-01 -6.75376207e-02 1.46288717e+00 -2.44095743e-01 -6.28089547e-01 -9.42689121e-01 -1.08923411e+00 -2.53439516e-01 8.78560364e-01 -6.34756923e-01 8.66419971e-02 -5.28001845e-01 -1.89683542e-01 1.75956264e-01 6.68685794e-01 3.88079099e-02 -1.20862079e+00 -8.11339855e-01 3.05132657e-01 -4.50613290e-01 -6.31010771e-01 -6.71919942e-01 4.81260538e-01 -8.01743925e-01 -1.09478009e+00 -5.68311512e-01 -6.35371327e-01 7.03173697e-01 1.47796065e-01 1.35514486e+00 2.98502088e-01 -6.65128082e-02 6.35016084e-01 -5.56445420e-01 -6.55325770e-01 -3.46036851e-01 3.44596505e-01 -3.58854800e-01 -3.64372909e-01 8.38043630e-01 -2.86053538e-01 -5.88778138e-01 2.17856333e-01 -6.84081137e-01 7.59116337e-02 4.54074204e-01 9.04543102e-01 3.32961202e-01 -5.37615240e-01 8.66495967e-01 -1.11112642e+00 1.08590484e+00 -3.50742221e-01 -3.45075101e-01 5.21128297e-01 -7.17096329e-01 9.89225358e-02 7.53708601e-01 -1.76835582e-01 -1.06140649e+00 9.44250897e-02 -1.10162802e-01 1.58038259e-01 -2.09799707e-01 8.23324263e-01 6.13675080e-02 1.86146144e-02 9.82225657e-01 1.59470528e-01 -7.82550499e-02 -2.65073776e-01 1.07562788e-01 9.05264556e-01 3.48533601e-01 -8.11080813e-01 5.08290231e-01 -2.57693499e-01 -9.09559801e-02 -5.71176469e-01 -1.60753810e+00 -7.49555826e-01 -1.01639438e+00 -6.83045805e-01 5.18431365e-01 -5.83700597e-01 -5.36055624e-01 -3.39531004e-01 -1.10745478e+00 -3.38012010e-01 -4.31495816e-01 3.35336655e-01 -6.76124215e-01 1.92710206e-01 -6.11550152e-01 -7.48361349e-01 -2.83807397e-01 -9.66539502e-01 8.82033050e-01 3.75158042e-01 -8.97737563e-01 -1.23519182e+00 3.47984105e-01 9.88273740e-01 1.66630313e-01 -1.62631720e-01 8.13664496e-01 -1.09061754e+00 -2.71920890e-01 -4.58404958e-01 6.01117760e-02 2.10136876e-01 7.25065218e-03 1.13188149e-02 -1.01599026e+00 2.16096640e-01 2.87388533e-01 -9.10311222e-01 7.26796567e-01 2.24430963e-01 1.42909396e+00 -5.66318989e-01 2.43655249e-01 7.96122402e-02 1.09789145e+00 -1.20897450e-01 1.92374840e-01 4.80713159e-01 4.92929518e-01 1.07297051e+00 7.42539167e-01 3.84746373e-01 5.52775145e-01 7.95309126e-01 1.02337347e-02 3.10106456e-01 -9.10640880e-02 -3.23741883e-01 5.13473868e-01 9.74067688e-01 1.62419662e-01 -3.27674411e-02 -9.20861602e-01 5.58466792e-01 -1.83981717e+00 -1.13839507e+00 -4.55713332e-01 2.10440660e+00 1.29129350e+00 -1.35438684e-02 1.11195065e-01 4.59878683e-01 3.57479960e-01 -1.15477324e-01 -2.93782115e-01 -7.14464545e-01 3.82155240e-01 3.60589266e-01 3.98278832e-02 6.18225634e-01 -7.85387456e-01 5.86350024e-01 6.29253960e+00 4.84991729e-01 -8.33369076e-01 1.15062177e-01 5.83890855e-01 -1.87889621e-01 -5.05488157e-01 -1.05644442e-01 -9.69629109e-01 3.93623084e-01 1.18872941e+00 -1.97372824e-01 2.08646372e-01 9.13982987e-01 2.69024581e-01 -2.29696050e-01 -1.22725511e+00 4.91182774e-01 3.78709555e-01 -1.13102937e+00 -1.71345383e-01 -1.27007812e-01 9.09399390e-01 -6.51190221e-01 -3.50331105e-02 5.10570586e-01 1.47967085e-01 -1.00749099e+00 6.97370470e-01 4.10007477e-01 7.77330399e-01 -6.69749677e-01 7.23436832e-01 6.06456399e-01 -8.10923040e-01 -3.97825949e-02 -4.48627919e-01 -5.24483025e-01 -1.71982914e-01 3.72657567e-01 -1.24367619e+00 7.80702829e-02 1.34162843e-01 4.06180918e-01 -6.74555361e-01 1.22576761e+00 -8.08342338e-01 9.65656936e-01 2.44452611e-01 -5.21953821e-01 1.38353065e-01 2.10596099e-02 8.70814994e-02 1.16876054e+00 4.46183026e-01 2.78708071e-01 2.61941642e-01 7.46597230e-01 -3.94692034e-01 4.25045013e-01 -3.42049003e-01 2.15978473e-02 5.82854927e-01 1.45257306e+00 -6.07183397e-01 -1.98352695e-01 -5.63124955e-01 8.89630914e-01 6.65223002e-01 6.86304569e-02 -3.62619311e-01 -5.30712068e-01 6.76117539e-02 2.25120470e-01 -1.00692295e-01 9.20431614e-02 -9.95842040e-01 -1.04022241e+00 8.49076211e-02 -8.78449202e-01 4.13085669e-01 -8.04008961e-01 -1.20928323e+00 1.68724701e-01 -4.28659558e-01 -1.24611807e+00 -3.67323160e-01 -6.85307980e-01 -8.66024315e-01 9.98698354e-01 -1.43076754e+00 -8.76083791e-01 -5.61529398e-01 3.75017494e-01 9.52328742e-01 -1.00050740e-01 8.12156022e-01 -9.64887068e-02 -2.53701240e-01 7.06722140e-01 -7.38807842e-02 1.46937102e-01 1.13125396e+00 -1.88507533e+00 -1.72520906e-01 6.41755521e-01 5.43837726e-01 5.73192894e-01 8.76711786e-01 -5.49266100e-01 -9.57113028e-01 -8.69165242e-01 1.42327404e+00 -8.54316056e-01 7.42708206e-01 -4.98431064e-02 -1.15913355e+00 4.91105318e-01 1.59856990e-01 -4.15937006e-01 1.26665282e+00 6.53781414e-01 -2.49280423e-01 1.11377858e-01 -7.45088995e-01 3.48416150e-01 4.41921562e-01 -6.99168086e-01 -1.03277743e+00 4.49312747e-01 3.77885729e-01 -5.88544250e-01 -7.38713145e-01 -1.51052445e-01 4.22919869e-01 -6.96571290e-01 2.81140655e-01 -6.45586133e-01 1.15496325e+00 -1.60624340e-01 1.31135046e-01 -1.43359864e+00 -1.01951212e-01 -2.32923806e-01 -2.83168793e-01 1.12685919e+00 6.99119925e-01 1.60264462e-01 1.16500866e+00 1.06692243e+00 -4.67463374e-01 -7.85050809e-01 -4.10350174e-01 -3.96184146e-01 1.32619321e-01 -4.08558846e-01 4.01216775e-01 9.29982781e-01 7.90854633e-01 6.54963076e-01 -4.05256189e-02 -2.86126971e-01 4.00648683e-01 2.70095676e-01 7.33306170e-01 -1.55358458e+00 -1.46856040e-01 -6.33488357e-01 1.34499529e-02 -1.02478766e+00 3.23904961e-01 -1.01180959e+00 3.91630203e-01 -1.66229415e+00 4.79057580e-01 -4.26054329e-01 -3.11608344e-01 5.38852215e-01 -5.99488020e-01 1.83945298e-01 1.69627771e-01 4.06303763e-01 -8.74318719e-01 2.61758476e-01 1.20113659e+00 -1.50246359e-03 -2.36660242e-01 4.09387946e-01 -9.79045630e-01 6.19368374e-01 8.36497426e-01 -3.99076700e-01 -4.86738086e-01 -2.50896424e-01 5.24872839e-01 3.31766665e-01 1.01993456e-01 -8.60210299e-01 7.04365730e-01 -4.76118594e-01 4.68238771e-01 -5.60093999e-01 2.77303867e-02 -4.60144013e-01 -5.91550052e-01 2.62897033e-02 -1.03395915e+00 7.76664689e-02 -1.46472216e-01 1.34279147e-01 -5.48914790e-01 -1.12220919e+00 5.12017548e-01 -2.85616696e-01 -3.84457499e-01 -5.25776409e-02 -3.32083821e-01 8.83516073e-02 9.21260893e-01 -2.86622584e-01 -3.91888291e-01 -5.82397819e-01 -6.22866571e-01 2.20295638e-01 4.46713746e-01 1.10365301e-01 6.28336906e-01 -1.10590613e+00 -7.44308233e-01 -1.68405905e-01 3.33645254e-01 -1.32825330e-01 2.45677447e-03 7.08526909e-01 -2.77084589e-01 7.46891558e-01 -1.31529570e-01 -3.05133849e-01 -1.38695776e+00 1.08892031e-01 5.93725070e-02 -5.94229877e-01 -3.04635167e-01 8.74158382e-01 -8.99041370e-02 -4.66003180e-01 3.40197027e-01 5.65791689e-02 -6.87270284e-01 3.87131602e-01 9.60286915e-01 2.51335502e-01 4.39673841e-01 -1.97958738e-01 2.46797547e-01 9.50957909e-02 -2.54303545e-01 -3.56062651e-01 1.35549200e+00 -3.13749909e-02 -5.61390929e-02 6.90366149e-01 7.21414387e-01 2.97719479e-01 -1.21191823e+00 -5.44588789e-02 5.78985214e-01 -4.02100086e-01 -8.94702598e-03 -1.21559894e+00 -4.11542386e-01 1.07458639e+00 -9.12707299e-02 4.91665721e-01 9.22712684e-01 -1.62610844e-01 6.37920141e-01 4.28304821e-01 2.71666646e-01 -1.30829501e+00 5.79204440e-01 6.95280015e-01 8.24164629e-01 -1.31108499e+00 -5.07520549e-02 3.25985290e-02 -7.20140398e-01 1.57223678e+00 9.57419455e-01 1.47871211e-01 4.24057655e-02 2.26540677e-02 1.61104888e-01 -3.17557871e-01 -9.86171901e-01 5.77162243e-02 8.28517795e-01 3.23556662e-01 1.16906965e+00 3.65168671e-04 -6.24728799e-01 8.88329804e-01 -5.45633495e-01 -1.49080858e-01 1.03696394e+00 8.92411411e-01 -1.02008915e+00 -1.25242400e+00 -4.76896584e-01 9.00025487e-01 -7.73498952e-01 -1.56495959e-01 -8.66087914e-01 2.10976675e-01 -3.81311059e-01 1.07886374e+00 -1.49806872e-01 -1.03717662e-01 1.85289100e-01 6.50082529e-01 5.26150346e-01 -1.21098971e+00 -1.00300205e+00 -3.33445758e-01 6.95700943e-02 -1.28537118e-01 -1.47961825e-01 -8.25709939e-01 -1.14405441e+00 -1.44348487e-01 -4.81570363e-01 6.43633783e-01 8.74222934e-01 1.20353270e+00 -2.23793536e-01 6.42243385e-01 6.75339401e-01 -4.27909881e-01 -8.91515076e-01 -1.05256796e+00 -3.32174987e-01 3.34326297e-01 1.97398812e-01 -3.47225159e-01 -3.00212830e-01 2.23496318e-01]
[11.330339431762695, 9.232720375061035]
aa382d32-fe7e-4cba-b062-728065fab11b
robust-low-rank-training-via-approximate
2306.01485
null
https://arxiv.org/abs/2306.01485v1
https://arxiv.org/pdf/2306.01485v1.pdf
Robust low-rank training via approximate orthonormal constraints
With the growth of model and data sizes, a broad effort has been made to design pruning techniques that reduce the resource demand of deep learning pipelines, while retaining model performance. In order to reduce both inference and training costs, a prominent line of work uses low-rank matrix factorizations to represent the network weights. Although able to retain accuracy, we observe that low-rank methods tend to compromise model robustness against adversarial perturbations. By modeling robustness in terms of the condition number of the neural network, we argue that this loss of robustness is due to the exploding singular values of the low-rank weight matrices. Thus, we introduce a robust low-rank training algorithm that maintains the network's weights on the low-rank matrix manifold while simultaneously enforcing approximate orthonormal constraints. The resulting model reduces both training and inference costs while ensuring well-conditioning and thus better adversarial robustness, without compromising model accuracy. This is shown by extensive numerical evidence and by our main approximation theorem that shows the computed robust low-rank network well-approximates the ideal full model, provided a highly performing low-rank sub-network exists.
['Francesco Tudisco', 'Gianluca Ceruti', 'Emanuele Zangrando', 'Dayana Savostianova']
2023-06-02
null
null
null
null
['adversarial-robustness']
['adversarial']
[-6.68253824e-02 3.72339338e-01 1.20070940e-02 3.87233198e-02 -3.98983538e-01 -7.76020944e-01 4.75138366e-01 -3.03515732e-01 -5.04938304e-01 4.73647118e-01 1.67829506e-02 -4.37082201e-01 -3.42257291e-01 -4.40290153e-01 -9.54136968e-01 -5.52268505e-01 -9.88771468e-02 2.76889771e-01 1.53688505e-01 -4.99533862e-02 1.30595267e-01 8.94293070e-01 -1.32824790e+00 5.83851486e-02 5.80075562e-01 9.33318675e-01 -4.58623081e-01 7.66729116e-01 6.86361372e-01 9.80739892e-01 -3.94595802e-01 -6.64429665e-01 8.32925320e-01 -9.61974859e-02 -6.89525485e-01 -4.61781263e-01 1.01241100e+00 -5.32992721e-01 -1.05706203e+00 1.28536797e+00 3.47425520e-01 3.57104808e-01 4.05401647e-01 -1.32723916e+00 -1.98724106e-01 7.05479741e-01 -2.43411183e-01 1.59491837e-01 -4.10102993e-01 1.60465594e-02 1.05858815e+00 -7.87815511e-01 6.36550784e-01 1.22611308e+00 1.10980499e+00 6.85102403e-01 -1.45091558e+00 -9.26027536e-01 1.76084027e-01 -5.70010282e-02 -1.52105367e+00 -8.18873465e-01 6.89194024e-01 -4.34174448e-01 9.21533406e-01 4.05259699e-01 5.71641326e-01 9.76088405e-01 1.91785425e-01 4.45760310e-01 4.95962381e-01 -2.33024478e-01 2.75011599e-01 5.61826900e-02 2.23750383e-01 9.12287891e-01 6.41277730e-01 2.59794265e-01 -6.90504372e-01 -2.74648875e-01 9.56331253e-01 -3.73146236e-02 -2.75877804e-01 -6.04473829e-01 -8.07602108e-01 6.63975418e-01 4.59135801e-01 2.06086516e-01 -8.54245722e-02 6.52174532e-01 4.50593263e-01 2.26238549e-01 2.63930976e-01 6.75632596e-01 -5.06767094e-01 -2.36567818e-02 -1.25610626e+00 4.68433559e-01 8.21856499e-01 7.34515190e-01 6.11213744e-01 4.08218235e-01 2.21679851e-01 5.10163963e-01 3.30145210e-01 2.86732912e-01 2.25244284e-01 -1.44797540e+00 3.97990674e-01 4.16327626e-01 -1.32331252e-01 -1.38289702e+00 -5.04612744e-01 -8.72605264e-01 -1.09169805e+00 3.69625896e-01 6.57138586e-01 -1.25177398e-01 -8.20621133e-01 2.12459087e+00 7.39287436e-02 3.25682640e-01 -1.92675442e-02 6.61050558e-01 4.93591428e-02 3.42337281e-01 -2.27606699e-01 -2.70365346e-02 8.08103561e-01 -5.17652094e-01 -3.94785821e-01 -2.75906354e-01 6.36316955e-01 -4.38896775e-01 8.60516906e-01 5.15556633e-01 -1.24010432e+00 -1.41178533e-01 -1.40650225e+00 -1.75542414e-01 -3.31051201e-02 -3.12086325e-02 7.51559615e-01 6.84294999e-01 -9.30813253e-01 1.18445110e+00 -1.20237529e+00 1.53836146e-01 6.62318110e-01 5.00801921e-01 -6.50154173e-01 2.56964914e-03 -1.17805707e+00 9.69770312e-01 2.47150138e-01 3.11356664e-01 -9.78320360e-01 -1.24024487e+00 -6.94410205e-01 2.36948892e-01 -7.13897683e-03 -6.91460133e-01 9.67205763e-01 -7.58212924e-01 -1.23311245e+00 4.97196019e-01 3.63125429e-02 -8.76498878e-01 8.62526596e-01 -4.53760356e-01 -5.49798906e-02 4.40779775e-01 -4.16560858e-01 5.47737002e-01 1.17521608e+00 -1.27495742e+00 -2.55108327e-01 -3.74250233e-01 2.24594042e-01 -6.63128346e-02 -8.80961597e-01 -1.66324988e-01 -3.43432069e-01 -6.90285861e-01 3.31707716e-01 -1.13831174e+00 -4.64241445e-01 2.39354625e-01 -3.26610088e-01 3.77158284e-01 6.94216549e-01 -6.75288379e-01 1.41348171e+00 -2.23816419e+00 6.58847094e-02 6.00193977e-01 5.88173568e-01 4.47410762e-01 -1.39205158e-02 1.26532197e-01 -4.04417664e-01 3.37988764e-01 -1.73244059e-01 -5.75627983e-01 6.35684878e-02 3.48251760e-01 -6.43911064e-01 7.84595668e-01 2.77487248e-01 5.80804110e-01 -6.26417220e-01 -2.84295946e-01 -7.15023634e-05 6.82931364e-01 -1.09474683e+00 -1.37894467e-01 1.68585658e-01 -9.70249847e-02 -6.11485355e-02 6.88390732e-02 5.20725429e-01 2.72681862e-02 9.92007703e-02 -6.09638989e-01 1.80284709e-01 2.52811551e-01 -1.54855752e+00 1.51684844e+00 -4.13935155e-01 7.04751790e-01 5.86515188e-01 -8.55694890e-01 3.56062770e-01 1.75859973e-01 4.74416614e-01 -3.21059227e-01 2.66072989e-01 2.16711879e-01 1.86894193e-01 -1.82246119e-02 3.95816982e-01 -1.83441877e-01 3.36558461e-01 2.30956450e-01 7.72363916e-02 1.49096668e-01 3.90534252e-01 6.15379930e-01 1.30957162e+00 6.48363680e-02 -3.76681089e-01 -6.10718846e-01 2.38567010e-01 -1.12880506e-01 7.94249535e-01 7.35913634e-01 -6.94463477e-02 4.17834133e-01 8.06220949e-01 -3.35869431e-01 -1.21684444e+00 -9.39689696e-01 -2.91122228e-01 1.00112617e+00 -3.87611985e-01 -4.56666857e-01 -8.90237451e-01 -2.31907800e-01 1.02615938e-01 5.65941870e-01 -7.53736377e-01 -6.31262839e-01 -6.90311790e-01 -6.35490716e-01 1.10925627e+00 5.76969564e-01 2.02131510e-01 -2.67744809e-01 -5.29870510e-01 -5.63406572e-02 3.81874591e-01 -9.48937953e-01 -2.28227437e-01 5.75859904e-01 -1.02854538e+00 -1.09571517e+00 -2.40695342e-01 -4.61796671e-01 1.01697290e+00 -2.18219385e-02 1.03237891e+00 3.66588354e-01 -1.97938561e-01 -4.70046746e-03 1.20540693e-01 -1.37893200e-01 -3.76580358e-01 7.91475102e-02 6.69096828e-01 -1.44752368e-01 -1.84190333e-01 -1.00268292e+00 -3.46263319e-01 1.12645902e-01 -1.22889936e+00 -4.97254655e-02 4.08993483e-01 9.65827346e-01 4.07734632e-01 3.13062280e-01 1.56527758e-01 -7.11929023e-01 3.34423900e-01 -8.21594447e-02 -9.73577738e-01 -4.39252555e-02 -6.62840068e-01 4.73434001e-01 9.79251683e-01 -6.20942652e-01 -5.41597128e-01 1.37415364e-01 -3.97985056e-02 -1.11486912e+00 4.24264878e-01 3.66654903e-01 -1.88312575e-01 -3.87164474e-01 9.55259085e-01 -3.39370668e-01 1.14979431e-01 -5.10804534e-01 5.67318439e-01 -7.96438903e-02 6.82971179e-01 -6.81084871e-01 1.35681701e+00 4.52376902e-01 5.28748155e-01 -7.02672422e-01 -1.12141740e+00 -4.42580367e-03 -6.83950484e-01 -3.34441029e-02 3.61146092e-01 -9.02259111e-01 -8.37044895e-01 2.36360431e-01 -8.79590333e-01 -3.44151378e-01 -4.40827608e-01 4.00316268e-01 -3.52046669e-01 3.30693126e-01 -9.07224417e-01 -7.86333084e-01 -2.84374714e-01 -1.15445924e+00 5.55865288e-01 -2.95228332e-01 -1.93236873e-01 -9.29045916e-01 -1.53316021e-01 4.70043123e-02 2.84482718e-01 2.79248297e-01 1.03288484e+00 -6.18964255e-01 -3.20518345e-01 -5.82502723e-01 -1.84254274e-01 7.51374304e-01 -5.46164274e-01 2.05564618e-01 -1.12715530e+00 -4.20472682e-01 1.98609203e-01 -2.67075896e-01 8.84264648e-01 1.69960171e-01 9.91411984e-01 -6.37904167e-01 -2.52397843e-02 9.42246974e-01 1.32999480e+00 -5.65044582e-01 3.44168007e-01 2.53562003e-01 1.15199244e+00 4.98539060e-01 7.16916174e-02 2.24977374e-01 -2.47002289e-01 1.89838752e-01 5.15550792e-01 -6.75954670e-03 -2.75479667e-02 -3.84393305e-01 4.20358896e-01 7.32845366e-01 1.92845166e-01 3.42750281e-01 -1.06226790e+00 3.18886518e-01 -2.00361395e+00 -1.07592702e+00 -4.84881736e-02 2.33250666e+00 9.45200026e-01 7.12699294e-01 3.12908888e-02 5.97812176e-01 2.37780303e-01 -1.29338512e-02 -8.56366336e-01 -1.77234128e-01 -1.16524369e-01 1.09298095e-01 1.01700914e+00 8.36078584e-01 -9.07693505e-01 8.21754634e-01 7.41750145e+00 7.85776079e-01 -9.21625078e-01 -6.86986521e-02 5.21808922e-01 -6.97042048e-01 -2.28259966e-01 4.32359576e-02 -7.27496028e-01 6.57159537e-02 1.17514038e+00 -2.11742476e-01 7.51669824e-01 1.08655167e+00 1.31658986e-01 3.11392874e-01 -1.37485862e+00 8.56879771e-01 -5.32723367e-02 -1.57724547e+00 9.64952037e-02 1.52298853e-01 5.96050680e-01 1.30078256e-01 2.35656530e-01 2.64475822e-01 5.16341269e-01 -1.03798890e+00 9.14250076e-01 5.47197342e-01 6.26849771e-01 -1.18492222e+00 3.38240057e-01 3.50710392e-01 -8.47144783e-01 -2.58455485e-01 -6.72813475e-01 -1.89016730e-01 -1.92117333e-01 8.14955652e-01 -4.01542217e-01 1.50873482e-01 6.07408702e-01 5.68876624e-01 -6.53604567e-01 7.50026762e-01 -1.01473980e-01 7.14049876e-01 -8.11129391e-01 5.95202088e-01 1.83892339e-01 -1.83834150e-01 4.67315823e-01 1.01165390e+00 -1.48692921e-01 -1.55421227e-01 5.05082868e-02 7.96388209e-01 -3.43106896e-01 -4.06474769e-01 -4.92977560e-01 -1.72209442e-01 4.78741735e-01 1.30484986e+00 -4.77384925e-01 -1.68788299e-01 9.84501746e-03 6.13238037e-01 5.64620197e-01 3.10366511e-01 -7.77680397e-01 -2.62223154e-01 1.11310256e+00 1.11297354e-01 4.54028957e-02 -5.94224751e-01 -6.70388162e-01 -1.21201050e+00 1.23851120e-01 -1.14890051e+00 1.15495980e-01 -3.84340137e-01 -9.98056352e-01 4.06088710e-01 -1.76519528e-01 -9.03659701e-01 -1.80242419e-01 -8.62903893e-01 -2.67119944e-01 7.25500286e-01 -1.06131577e+00 -9.88297820e-01 2.35428259e-01 5.98868370e-01 -1.47669241e-01 -6.49966449e-02 6.67832613e-01 6.86469972e-01 -1.06341946e+00 1.08054066e+00 2.37264410e-01 2.98522949e-01 4.02176201e-01 -1.12011099e+00 3.01842034e-01 1.35094869e+00 9.76120010e-02 1.29031992e+00 9.82496142e-01 -4.84086007e-01 -1.75144649e+00 -1.01041842e+00 3.68885010e-01 -6.55783296e-01 1.17981803e+00 -5.41588902e-01 -8.06260049e-01 7.53192723e-01 -3.86206031e-01 1.73574537e-01 6.05552197e-01 2.33747900e-01 -8.82901490e-01 -1.96908459e-01 -1.12025678e+00 9.72089946e-01 1.02493024e+00 -8.22673023e-01 -3.66804808e-01 2.69606292e-01 6.93537474e-01 -3.62098187e-01 -1.12734640e+00 3.56277138e-01 6.64119244e-01 -7.48543680e-01 1.00701845e+00 -1.01547670e+00 3.06588024e-01 -2.43813202e-01 -2.05858111e-01 -8.82338762e-01 -4.23032701e-01 -1.04619765e+00 -7.23162532e-01 1.02705872e+00 4.54447895e-01 -3.48181874e-01 1.16862047e+00 1.25629580e+00 -1.55837789e-01 -8.60881031e-01 -1.02097261e+00 -9.21690762e-01 4.21252340e-01 -7.39337623e-01 3.95084098e-02 8.68392825e-01 -8.10498148e-02 2.78957695e-01 -4.32892948e-01 3.57138604e-01 6.50537789e-01 -5.22125661e-01 6.71883941e-01 -1.21980417e+00 -4.01863426e-01 -6.93216383e-01 -4.49654698e-01 -8.23615313e-01 3.40141863e-01 -8.45312357e-01 -6.03531152e-02 -1.02963042e+00 -1.44442007e-01 -4.63863850e-01 -3.82468581e-01 6.81200147e-01 4.22796682e-02 4.95088875e-01 2.48295963e-01 2.12588549e-01 -3.66299838e-01 2.46314257e-01 7.83584535e-01 1.33710697e-01 -9.54289734e-03 -2.75858760e-01 -6.58514261e-01 1.02039945e+00 5.84087789e-01 -5.87080538e-01 -4.52709168e-01 -5.84346592e-01 8.49161267e-01 -3.98533970e-01 3.82287294e-01 -1.29453444e+00 3.72876674e-01 1.15260236e-01 5.48546016e-01 -1.68540210e-01 4.47970062e-01 -9.61259425e-01 4.79825288e-02 7.47453928e-01 -5.92382729e-01 5.54562360e-02 3.85071337e-01 4.59496409e-01 1.35128349e-01 -3.45557690e-01 1.18374872e+00 1.61770642e-01 -5.55164367e-02 4.46729332e-01 -2.64375657e-01 1.25523344e-01 5.37859559e-01 3.09567619e-02 -1.73236325e-01 -8.35672542e-02 -7.67845154e-01 1.21613458e-01 6.25529885e-01 1.74679056e-01 3.19914818e-01 -1.30614340e+00 -3.99030149e-01 2.49309912e-01 -4.70382690e-01 1.18620895e-01 1.67986602e-01 6.43809617e-01 -7.23472416e-01 1.34436384e-01 -1.73188910e-01 -1.15802437e-01 -1.10581946e+00 5.73807597e-01 6.21250212e-01 -3.26644897e-01 -6.22459829e-01 1.15640879e+00 -1.78601354e-01 -1.41472831e-01 7.73166776e-01 -1.30709887e-01 3.82141024e-01 8.96075070e-02 6.72036946e-01 6.65462792e-01 1.87842563e-01 -5.06320179e-01 -2.70020694e-01 2.40313306e-01 -2.20299825e-01 -1.81037530e-01 1.36966288e+00 3.08957398e-01 -2.06403136e-01 1.27501249e-01 1.33593917e+00 9.41853374e-02 -1.56412005e+00 -5.19684479e-02 -1.28698975e-01 -3.10889512e-01 4.12117690e-01 -3.04800391e-01 -1.27391577e+00 9.82722461e-01 3.70876580e-01 1.01187788e-01 8.76409292e-01 -6.93266749e-01 5.90655982e-01 9.27529514e-01 1.90078437e-01 -1.23614240e+00 -1.05318621e-01 7.71556973e-01 7.76976883e-01 -5.87988257e-01 3.13725770e-01 -1.30337924e-01 -2.34858453e-01 9.50441837e-01 6.72431767e-01 -6.91927493e-01 8.28383565e-01 4.76619571e-01 -1.52645782e-01 7.30219483e-02 -6.16231501e-01 4.11627859e-01 1.37955606e-01 4.62602407e-01 1.00765929e-01 -3.83553088e-01 2.13783979e-01 4.95684206e-01 -7.04100192e-01 -4.47381556e-01 4.20124739e-01 5.44178128e-01 -4.17625606e-01 -9.34340894e-01 -3.22437882e-01 6.44720018e-01 -6.70702934e-01 -3.00387532e-01 -3.21069479e-01 6.92376673e-01 -1.00475833e-01 8.31956685e-01 -1.54248089e-01 -4.60132420e-01 3.47725660e-01 8.79312009e-02 5.62766492e-01 -3.41324151e-01 -7.30780542e-01 -2.10193396e-01 1.32058384e-02 -8.10862243e-01 2.23752007e-01 -4.54874963e-01 -1.19802630e+00 -6.46048844e-01 -2.56730855e-01 4.23153825e-02 6.55712426e-01 7.95188248e-01 3.67685854e-01 5.48502147e-01 3.44373345e-01 -1.03957343e+00 -1.14587116e+00 -7.54602075e-01 -4.75535899e-01 4.53764379e-01 4.25992370e-01 -5.05440712e-01 -6.95671678e-01 -1.31892949e-01]
[8.371453285217285, 3.4499330520629883]
0e92a9a5-54af-46e0-85fd-a2494db088c1
orthonormal-convolutions-for-the-rotation
2206.03860
null
https://arxiv.org/abs/2206.03860v1
https://arxiv.org/pdf/2206.03860v1.pdf
Orthonormal Convolutions for the Rotation Based Iterative Gaussianization
In this paper we elaborate an extension of rotation-based iterative Gaussianization, RBIG, which makes image Gaussianization possible. Although RBIG has been successfully applied to many tasks, it is limited to medium dimensionality data (on the order of a thousand dimensions). In images its application has been restricted to small image patches or isolated pixels, because rotation in RBIG is based on principal or independent component analysis and these transformations are difficult to learn and scale. Here we present the \emph{Convolutional RBIG}: an extension that alleviates this issue by imposing that the rotation in RBIG is a convolution. We propose to learn convolutional rotations (i.e. orthonormal convolutions) by optimising for the reconstruction loss between the input and an approximate inverse of the transformation using the transposed convolution operation. Additionally, we suggest different regularizers in learning these orthonormal convolutions. For example, imposing sparsity in the activations leads to a transformation that extends convolutional independent component analysis to multilayer architectures. We also highlight how statistical properties of the data, such as multivariate mutual information, can be obtained from \emph{Convolutional RBIG}. We illustrate the behavior of the transform with a simple example of texture synthesis, and analyze its properties by visualizing the stimuli that maximize the response in certain feature and layer.
['Jesús Malo', 'J. Emmanuel Johnson', 'Alexander Hepburn', 'Valero Laparra']
2022-06-08
null
null
null
null
['texture-synthesis']
['computer-vision']
[ 2.87114590e-01 3.17665815e-01 2.39188090e-01 -3.77848178e-01 -1.63501844e-01 -4.10770595e-01 6.13213181e-01 -4.22489822e-01 -6.70055211e-01 5.08302748e-01 2.34165221e-01 -1.91392407e-01 -3.49558204e-01 -8.13849568e-01 -7.92344689e-01 -1.12828159e+00 -3.09492624e-03 1.82758212e-01 -2.10689068e-01 1.42531693e-02 1.16124161e-01 8.89842689e-01 -1.22078359e+00 2.42248997e-01 5.26652873e-01 8.21068943e-01 1.33647442e-01 6.47720456e-01 1.22680724e-01 5.11968553e-01 -5.69242835e-01 -7.95577187e-03 1.99743643e-01 -3.64780128e-01 -8.63751590e-01 3.34056079e-01 1.09599136e-01 -8.14942122e-02 -2.25638136e-01 9.59582865e-01 3.84103984e-01 3.94565046e-01 1.02322042e+00 -8.54039133e-01 -9.57363248e-01 6.11672819e-01 -6.13661110e-01 -1.04117669e-01 -2.81176388e-01 -3.54872584e-01 7.89443791e-01 -8.79715443e-01 6.31350696e-01 1.34040296e+00 6.12815678e-01 4.67210084e-01 -1.87077868e+00 -2.29251102e-01 1.40017614e-01 -5.62071800e-02 -1.41286516e+00 -2.68793106e-01 6.89682901e-01 -4.44399357e-01 1.06975496e+00 3.75734121e-01 4.18011427e-01 9.03688967e-01 1.13381870e-01 7.07999170e-01 1.04897451e+00 -5.63913107e-01 2.25883484e-01 2.50178397e-01 8.01747590e-02 4.92952317e-01 1.31629661e-01 -1.79488808e-02 -3.09865087e-01 6.43056408e-02 1.24708200e+00 -1.58735752e-01 -3.42840940e-01 -5.38629293e-01 -1.37844753e+00 9.52642679e-01 6.53502822e-01 3.94379616e-01 -5.48357308e-01 4.20993030e-01 2.37583145e-02 4.20812577e-01 5.95223784e-01 6.27476633e-01 -3.50378990e-01 4.10475105e-01 -9.16851819e-01 -2.01560501e-02 7.10816145e-01 7.71633804e-01 1.06833339e+00 2.29474977e-01 2.85340194e-02 7.86592007e-01 2.07305625e-01 5.28548062e-01 5.31469762e-01 -9.08553362e-01 -1.02863550e-01 1.46548629e-01 -1.98358104e-01 -7.12278068e-01 -7.15879917e-01 -5.30476570e-01 -1.14967513e+00 4.02154773e-01 5.21774292e-01 -3.16855460e-01 -1.01528335e+00 1.76743746e+00 -1.16032019e-01 -3.69553976e-02 7.71290660e-02 8.11682284e-01 6.05309308e-01 3.31613481e-01 -9.82727334e-02 5.88399805e-02 1.26819229e+00 -6.28974319e-01 -5.52878857e-01 -9.89041254e-02 5.35751462e-01 -9.13586199e-01 8.84659111e-01 4.84545439e-01 -9.19687450e-01 -6.32453144e-01 -9.50882137e-01 1.45429159e-02 -4.17064011e-01 5.72005212e-01 7.47675836e-01 5.87360442e-01 -1.33179355e+00 7.50261188e-01 -1.01767588e+00 -3.96552533e-01 3.82128268e-01 4.84188378e-01 -7.06876218e-01 2.96161830e-01 -8.51905227e-01 9.47075009e-01 2.69526154e-01 1.52630106e-01 -5.26833415e-01 -5.44009864e-01 -7.96558321e-01 5.22524491e-02 -3.76558527e-02 -8.33220243e-01 7.46824980e-01 -1.26677036e+00 -1.65333104e+00 8.31369877e-01 -2.36203019e-02 -4.41551179e-01 3.56416106e-01 -1.83189571e-01 1.96544975e-02 -9.85191017e-02 -2.77254552e-01 8.86862040e-01 1.25252676e+00 -9.69927967e-01 -1.76513001e-01 -3.59759063e-01 -6.94935471e-02 1.39434263e-01 -5.57826936e-01 8.46057683e-02 -5.26065648e-01 -9.40916955e-01 5.12557507e-01 -1.11159146e+00 -5.88882148e-01 -2.25467414e-01 -3.56695861e-01 1.80058286e-01 6.67867064e-01 -5.88547885e-01 8.41636896e-01 -2.39407587e+00 5.62160969e-01 5.71100652e-01 2.02615470e-01 -2.04934433e-01 -3.09504509e-01 3.07824552e-01 -5.94939411e-01 -1.22835962e-02 -2.54075140e-01 -4.92726773e-01 -2.28734881e-01 2.87449270e-01 -1.60092831e-01 8.18807006e-01 4.64278489e-01 9.41944122e-01 -4.52250451e-01 1.12484738e-01 2.47939363e-01 9.89284813e-01 -7.96081424e-01 -1.12820501e-02 4.94239666e-02 6.69035673e-01 -1.15741782e-01 5.55735826e-02 7.39204168e-01 -3.24175566e-01 -1.08827546e-01 -5.39817095e-01 -2.65914917e-01 -1.01426683e-01 -1.12420750e+00 1.67670918e+00 -6.08816147e-01 1.02015567e+00 -1.41767591e-01 -1.37627423e+00 1.03721893e+00 2.37307072e-01 5.08576274e-01 -1.68628126e-01 9.48940441e-02 1.21794462e-01 2.09472075e-01 -1.98245317e-01 4.07619506e-01 -7.69906119e-02 2.96927065e-01 4.29109603e-01 4.11139339e-01 -3.39571595e-01 2.79356143e-03 -6.14509508e-02 7.58192778e-01 1.35854408e-01 2.08386987e-01 -4.22870666e-01 3.09899956e-01 -4.07291740e-01 -5.77980131e-02 6.50529325e-01 5.12437463e-01 1.04855537e+00 6.24498785e-01 -3.44924122e-01 -1.14997804e+00 -8.71653199e-01 -2.55107999e-01 8.25519383e-01 -3.25712591e-01 -3.72478701e-02 -8.50980043e-01 -3.29304218e-01 -1.33396074e-01 6.38764858e-01 -9.30637002e-01 -2.06589296e-01 -5.66266775e-01 -1.33166826e+00 4.92030859e-01 6.09245181e-01 5.50083458e-01 -9.82900083e-01 -5.28547227e-01 1.11419097e-01 2.46858388e-01 -8.92093480e-01 -1.49919361e-01 7.02504039e-01 -1.00849330e+00 -7.29496121e-01 -1.05270231e+00 -6.08937800e-01 9.88932908e-01 2.38760516e-01 8.69060338e-01 -4.56870079e-01 -3.45197976e-01 3.68416488e-01 -1.10335499e-01 -3.76683712e-01 -1.65304810e-01 2.57054400e-02 -4.95309569e-02 2.24149168e-01 1.85151771e-01 -7.88416803e-01 -3.89407456e-01 2.33910397e-01 -1.07570338e+00 1.77211791e-01 8.64235997e-01 8.98775041e-01 6.77180350e-01 6.27767146e-02 1.78687468e-01 -9.18496192e-01 7.89983928e-01 -1.40484005e-01 -5.10365665e-01 -8.75484049e-02 -3.87294382e-01 3.68146211e-01 3.79171163e-01 -7.13068426e-01 -9.15935934e-01 1.94757998e-01 -8.09547305e-02 -5.50291717e-01 -2.02516347e-01 6.52737737e-01 4.30250466e-02 -2.99728811e-01 9.46165919e-01 1.34104893e-01 2.26932973e-01 -5.18839717e-01 8.39983046e-01 3.17803770e-01 4.61355507e-01 -3.07614356e-01 6.42447293e-01 7.92435884e-01 1.35191694e-01 -1.33036613e+00 -3.21899533e-01 -3.99552435e-01 -7.48024583e-01 -4.79705371e-02 1.02912569e+00 -6.44787610e-01 -7.19666362e-01 4.26136792e-01 -1.17096889e+00 -3.34356129e-01 -6.52693272e-01 1.01505816e+00 -7.87042022e-01 1.36969551e-01 -4.07143921e-01 -4.93275106e-01 -1.38662636e-01 -1.27945459e+00 8.45270932e-01 1.65065244e-01 -1.82348236e-01 -1.04966450e+00 -8.40849951e-02 -4.10590529e-01 7.61445522e-01 -1.02001041e-01 9.77895558e-01 -7.38599956e-01 -2.86494672e-01 -2.25053445e-01 -3.02907467e-01 8.42867434e-01 1.76845521e-01 6.53688163e-02 -1.05958915e+00 -3.27173144e-01 2.70426631e-01 -2.73160413e-02 1.21521485e+00 7.56454229e-01 1.17813933e+00 -2.00299218e-01 -1.32466972e-01 9.48217154e-01 1.29183853e+00 -3.16242650e-02 9.04816091e-01 3.54918033e-01 8.17584932e-01 7.47745931e-01 -7.96664804e-02 2.93227166e-01 -2.68283546e-01 7.12971866e-01 2.34898835e-01 -5.83492815e-01 -2.23857552e-01 3.16343755e-01 1.92180246e-01 6.57567441e-01 -6.84215844e-01 1.58855826e-01 -5.63898861e-01 3.68490852e-02 -1.76702547e+00 -8.05717945e-01 -3.49903554e-02 2.36200237e+00 5.66107273e-01 -1.03103518e-01 -1.96518630e-01 4.19445522e-02 5.23102939e-01 4.39081565e-02 -2.62735486e-01 -3.19851160e-01 -4.81643885e-01 6.30647004e-01 9.35650289e-01 6.08481884e-01 -1.36209226e+00 9.28224444e-01 6.76849556e+00 7.08375812e-01 -1.49160111e+00 -7.96115547e-02 6.17460847e-01 2.50200361e-01 -2.63812304e-01 8.37148353e-02 -5.54566860e-01 -1.05061837e-01 4.72172439e-01 2.85101533e-01 5.79589903e-01 8.05644095e-01 2.47650325e-01 1.67336203e-02 -9.93343413e-01 9.29114103e-01 3.05239167e-02 -1.21514738e+00 6.88849092e-02 2.29224429e-01 7.54794300e-01 8.28356743e-02 4.94433403e-01 -1.38554424e-01 2.08163038e-01 -1.35856664e+00 6.28845215e-01 6.93889797e-01 6.72719300e-01 -7.59943366e-01 4.61821228e-01 1.18731037e-01 -7.36004889e-01 1.84462607e-01 -6.67964160e-01 6.79887012e-02 -1.71720996e-01 4.27662730e-01 -8.88562799e-01 2.79882371e-01 5.30866683e-01 5.51789582e-01 -4.46995735e-01 8.28660250e-01 -3.64499629e-01 4.28399175e-01 -3.48622233e-01 5.14684692e-02 2.99450457e-01 -7.18755364e-01 6.52266681e-01 1.38590002e+00 3.39753270e-01 -2.91983545e-01 -2.54525065e-01 1.04775882e+00 1.55159414e-01 1.91882834e-01 -6.09489977e-01 5.68442643e-02 -3.50574441e-02 1.38895082e+00 -1.00421917e+00 -8.69141072e-02 -4.94961500e-01 1.16782475e+00 3.85327399e-01 8.33447397e-01 -5.43013811e-01 -4.87157792e-01 6.69616222e-01 1.15847634e-02 6.28331363e-01 -3.05162042e-01 -2.53489733e-01 -1.02194476e+00 -2.70466894e-01 -8.09600949e-01 -5.43966889e-02 -6.52949512e-01 -8.48497391e-01 6.96557701e-01 1.07045107e-01 -1.00429904e+00 -3.28719854e-01 -9.04703438e-01 -4.26071703e-01 1.19402826e+00 -1.28931403e+00 -1.17820764e+00 -1.22322023e-01 6.53775871e-01 1.19245857e-01 -2.42612332e-01 1.05947232e+00 1.32805496e-01 -4.21016395e-01 5.01012921e-01 3.27571332e-01 9.57978889e-02 6.22964382e-01 -1.21114194e+00 5.19612968e-01 6.54017150e-01 4.00160968e-01 9.38263535e-01 8.70276451e-01 -2.39424810e-01 -1.23014629e+00 -8.64930034e-01 4.51734424e-01 6.31538332e-02 6.68364823e-01 -3.13540936e-01 -7.70112514e-01 9.41370308e-01 7.83305168e-02 1.71705738e-01 5.59242904e-01 8.71688128e-02 -3.78165781e-01 1.53432846e-01 -7.79894650e-01 7.01161563e-01 6.42338693e-01 -4.79523838e-01 -9.01286602e-02 2.09335193e-01 3.60965848e-01 -1.89129055e-01 -8.32417548e-01 3.29554111e-01 5.20907462e-01 -9.93982613e-01 1.16040158e+00 -8.33299041e-01 1.38244331e-01 -3.63260180e-01 -5.38524576e-02 -1.51043677e+00 -5.88392913e-01 -4.90015030e-01 2.30480537e-01 7.22046971e-01 5.75853050e-01 -7.75392532e-01 8.70426714e-01 3.63577455e-01 -7.55625665e-02 -5.53567588e-01 -8.50732982e-01 -5.62217653e-01 1.49797350e-01 -4.27124202e-01 1.13949291e-01 8.53696167e-01 -2.64633238e-01 1.74561158e-01 -4.29729789e-01 1.50942966e-01 3.51004601e-01 -1.13836609e-01 8.15726340e-01 -9.78192031e-01 -5.58361113e-01 -5.78068435e-01 -6.31720424e-01 -1.21000111e+00 2.38239225e-02 -8.13786209e-01 -8.78404453e-02 -1.25674868e+00 7.44273961e-02 -4.86759067e-01 -4.15515713e-02 4.48991865e-01 2.65398482e-03 3.55307817e-01 2.71219518e-02 2.88957745e-01 1.37812242e-01 2.97362357e-01 1.28815186e+00 -7.00212866e-02 -2.68536329e-01 3.01665127e-01 -4.79320168e-01 8.71409714e-01 7.46225357e-01 -2.78542280e-01 -4.16218847e-01 -4.33594704e-01 3.08017522e-01 -4.14486408e-01 3.72888237e-01 -9.01976466e-01 1.45343706e-01 1.45482644e-01 6.29136264e-01 -1.63641647e-01 4.71770972e-01 -5.94302416e-01 5.25399745e-01 2.48228922e-01 -4.10043746e-01 -1.80592939e-01 1.05495922e-01 2.97778636e-01 -3.25463384e-01 -3.08558017e-01 8.12795758e-01 -2.38960907e-01 -3.64362419e-01 -7.64427008e-03 -5.05956709e-01 -4.74864990e-01 6.08326495e-01 -1.89970195e-01 3.83771956e-02 -4.44884211e-01 -9.87996995e-01 -4.77124244e-01 4.41134602e-01 9.47141368e-03 6.32061064e-01 -1.16836417e+00 -8.00504446e-01 4.99857277e-01 -1.70405671e-01 2.22403966e-02 2.99671423e-02 1.16824055e+00 -6.44664586e-01 3.52286398e-01 -3.60415190e-01 -5.28880477e-01 -1.20839453e+00 6.37500226e-01 3.78108501e-01 -2.08835557e-01 -6.31641507e-01 9.90554571e-01 6.89404905e-01 -1.72170177e-01 1.55220345e-01 -3.74554366e-01 -5.28161407e-01 1.90664291e-01 3.54653835e-01 1.45684510e-01 2.79857576e-01 -6.90071940e-01 -7.88515657e-02 5.42646170e-01 -3.03409584e-02 -3.86264026e-01 1.45403028e+00 -3.54457013e-02 -4.91335928e-01 2.36923397e-01 1.45582199e+00 6.55628070e-02 -1.20217431e+00 -1.26531124e-01 -1.91808254e-01 -1.68738633e-01 2.24177226e-01 -1.36071980e-01 -1.28082287e+00 8.85100186e-01 5.10982931e-01 8.52379724e-02 1.14110875e+00 -8.05170238e-02 -5.17336987e-02 6.29988253e-01 -5.01761176e-02 -8.71414304e-01 1.44975141e-01 6.15477979e-01 1.00973558e+00 -7.73690045e-01 1.51092678e-01 -3.55194509e-01 -4.38744754e-01 1.53605914e+00 9.76643711e-02 -5.07892132e-01 9.07769084e-01 3.91955554e-01 1.40854210e-01 -2.03633413e-01 -1.97437361e-01 -2.91028619e-01 5.28822660e-01 6.29667282e-01 7.17126489e-01 1.84139043e-01 -3.96869242e-01 3.11366141e-01 -2.77578562e-01 -3.81334573e-01 3.29174042e-01 7.22602725e-01 -2.46289060e-01 -1.07085311e+00 -5.56856215e-01 2.56998569e-01 -4.59709406e-01 -3.64230573e-01 -1.77271962e-01 8.16559613e-01 -1.61512017e-01 3.36598158e-01 2.53808647e-01 -4.02152002e-01 1.69801295e-01 -1.70730039e-01 7.04874933e-01 -5.35417736e-01 -2.96781152e-01 5.57254136e-01 -1.42424315e-01 -3.77472639e-01 -5.48059106e-01 -4.53529686e-01 -1.09135425e+00 -7.82016199e-03 -4.42187428e-01 -1.24068156e-01 1.00568521e+00 7.29337633e-01 2.11388152e-03 5.83022475e-01 3.33729267e-01 -1.26087153e+00 -4.32092518e-01 -1.07660902e+00 -7.29681313e-01 3.80029500e-01 9.07380059e-02 -5.54248035e-01 -3.46298754e-01 1.25286043e-01]
[9.120240211486816, 2.3254644870758057]
8c776ef9-61bf-4550-9c46-5a2d30aca8fa
semantic-aware-representation-learning-via-1
2111.06021
null
https://arxiv.org/abs/2111.06021v4
https://arxiv.org/pdf/2111.06021v4.pdf
Probabilistic Contrastive Learning for Domain Adaptation
The standard contrastive learning acts on the extracted features with $\ell_{2}$ normalization. For domain adaptation tasks, however, we find that contrastive learning with the standard paradigm does not perform well. The reason is mainly that the class weights (weights of the final fully connected layer) are not involved during optimization, which does not guarantee the produced features to be clustered around the class weights learned from source data. To tackle this issue, we propose a simple yet powerful probabilistic contrastive learning (PCL) in this paper, which not only produces compact features but also enforces them to be distributed around the class weights. Specifically, we break the traditional contrastive learning paradigm (feature+$\ell_{2}$ normalization) by replacing the features with probabilities and removing $\ell_{2}$ normalization. In this way, we can enforce the probability to approximate the one-hot form, thereby narrowing the distance between the features and the class weights. PCL is generic due to conciseness, which can be used for different tasks. In this paper, we conduct extensive experiments on five tasks, \textit{i.e.}, unsupervised domain adaptation (UDA), semi-supervised domain adaptation (SSDA), semi-supervised learning (SSL), UDA detection, and UDA semantic segmentation. The results demonstrate that our PCL can bring significant gains for these tasks. In particular, for segmentation tasks, with the blessing of PCL, our method achieves or even surpasses CPSL-D with a smaller training cost (1*3090, 5 days vs 4*V100, 11 days). Code is available at https://github.com/ljjcoder/Probabilistic-Contrastive-Learning.
['Keyu Tu', 'Zilei Wang', 'Yixin Zhang', 'Junjie Li']
2021-11-11
semantic-aware-representation-learning-via
https://openreview.net/forum?id=XizHAfgfd3J
https://openreview.net/pdf?id=XizHAfgfd3J
null
['semi-supervised-image-classification']
['computer-vision']
[ 9.35273170e-02 -7.39438683e-02 -3.21426719e-01 -4.74971294e-01 -6.93274319e-01 -4.93787348e-01 5.71822464e-01 2.67563403e-01 -7.46842027e-01 6.08839691e-01 -1.40847161e-01 -2.10615128e-01 -2.65209377e-01 -7.34104216e-01 -6.70487702e-01 -8.88172626e-01 1.88694462e-01 3.54959309e-01 5.29075682e-01 -6.34812191e-02 4.17141430e-02 4.17488992e-01 -1.44058907e+00 5.32708764e-02 1.07250690e+00 1.16995788e+00 4.47153717e-01 2.00595483e-01 -3.87711287e-01 3.63102794e-01 -6.75378680e-01 -4.75978374e-01 1.70343786e-01 -3.62870663e-01 -7.37625718e-01 -1.27215415e-01 1.91801086e-01 -1.61286637e-01 -2.17144266e-01 1.15625739e+00 3.77750635e-01 3.12607318e-01 9.22613502e-01 -1.24722219e+00 -3.36742371e-01 5.29909194e-01 -8.70472610e-01 1.51616573e-01 -9.45481807e-02 6.78802729e-02 1.09614635e+00 -9.63990927e-01 3.57953310e-01 1.17300332e+00 5.21123409e-01 6.10067785e-01 -1.16979468e+00 -8.66573691e-01 4.14456695e-01 1.81342930e-01 -1.52083695e+00 -2.80937880e-01 9.13722575e-01 -3.06550056e-01 7.67879069e-01 2.43335888e-01 3.57178986e-01 9.87645924e-01 -8.80453438e-02 1.07075715e+00 1.12113690e+00 -6.14789367e-01 2.98519313e-01 1.47067174e-01 3.29457164e-01 6.83722675e-01 1.75784245e-01 5.03556915e-02 -4.27615166e-01 -1.49344876e-01 4.68626320e-01 -2.55709961e-02 -1.88016921e-01 -3.45564276e-01 -8.87967348e-01 8.89139771e-01 3.27334791e-01 3.18704337e-01 -1.28486276e-01 -1.34674430e-01 3.33650649e-01 5.14952056e-02 4.63362038e-01 2.54633665e-01 -7.38658071e-01 3.11210472e-03 -8.74783278e-01 2.11243734e-01 4.50527728e-01 9.43458259e-01 9.24800813e-01 -4.24299985e-02 -2.58864552e-01 1.28621018e+00 4.22134817e-01 4.87617701e-01 5.74651718e-01 -7.99909651e-01 3.46749455e-01 6.01453245e-01 -1.79746106e-01 -6.74505830e-01 -4.90830868e-01 -4.91790950e-01 -9.64517474e-01 1.81027412e-01 6.63762152e-01 -2.65072495e-01 -1.06903780e+00 1.98624957e+00 1.90070182e-01 1.20554104e-01 -3.75568937e-03 6.84982538e-01 6.82579160e-01 8.33136141e-01 3.22629124e-01 -2.39081413e-01 1.34782410e+00 -8.57219875e-01 -4.29387718e-01 -5.31001687e-01 4.78775352e-01 -9.20660198e-01 1.43302691e+00 3.32048535e-01 -8.28991473e-01 -6.17317855e-01 -1.16201150e+00 -1.95336845e-02 -3.53531957e-01 2.40485728e-01 5.21281958e-01 5.42229652e-01 -5.62353671e-01 6.02728009e-01 -8.01675558e-01 -1.93533421e-01 6.35131240e-01 3.31817925e-01 -1.19570866e-01 -1.03321478e-01 -1.15530348e+00 6.23841405e-01 6.71308160e-01 -2.64431685e-01 -4.57019061e-01 -5.94293773e-01 -8.68640125e-01 1.43697828e-01 4.28626388e-01 -4.13701981e-01 1.14610755e+00 -9.48189557e-01 -1.56481743e+00 8.36073399e-01 -1.77672848e-01 -3.38108808e-01 3.12643617e-01 -2.04739690e-01 -2.40447029e-01 -1.43134698e-01 1.18568413e-01 8.44598114e-01 8.66240799e-01 -1.22478282e+00 -8.18293214e-01 -2.49331862e-01 -7.88221434e-02 1.37710378e-01 -6.41775072e-01 2.49382155e-03 -7.38084137e-01 -9.29575562e-01 1.98360085e-01 -9.25334632e-01 -1.02637485e-01 -1.37376208e-02 -3.24869573e-01 -5.47275126e-01 6.95722103e-01 -4.44305390e-01 1.36761928e+00 -2.54666972e+00 -1.07082218e-01 4.05882567e-01 9.16386619e-02 4.81033832e-01 -7.17029572e-02 -8.22882876e-02 -4.50602882e-02 -5.83234020e-02 -7.07139313e-01 -4.27463770e-01 7.89459348e-02 3.63714397e-01 -1.20197684e-01 2.76985288e-01 2.74847269e-01 6.95347726e-01 -8.36149275e-01 -5.38081646e-01 1.62172720e-01 2.43758440e-01 -6.62564754e-01 1.15206681e-01 -2.55186230e-01 2.55508661e-01 -3.85230631e-01 5.09036183e-01 8.56835961e-01 -1.36893466e-01 1.61260515e-01 -1.12888001e-01 -2.08635093e-03 1.96698979e-01 -1.47959137e+00 1.58653736e+00 -4.72824931e-01 3.40741128e-01 -4.47245501e-02 -1.11637783e+00 1.01134694e+00 -6.09043166e-02 4.49931026e-01 -7.49535143e-01 1.52422756e-01 3.43986064e-01 4.72020730e-03 -9.91244465e-02 2.09509417e-01 -1.46091104e-01 -3.47792745e-01 2.32830659e-01 3.72287422e-01 -1.13287196e-01 2.52360553e-01 -2.06662510e-02 7.45560646e-01 2.21546337e-01 3.99127930e-01 -3.86383742e-01 5.52018881e-01 1.22739635e-02 9.03650761e-01 6.98818088e-01 -2.06586868e-01 5.92797816e-01 5.64970672e-01 -1.73071418e-02 -8.00605476e-01 -1.28292143e+00 -2.47365355e-01 1.26408064e+00 2.56191522e-01 -3.36725175e-01 -7.61176407e-01 -8.56266916e-01 -1.14623263e-01 7.83518851e-01 -2.91713744e-01 -4.56813186e-01 -6.23491466e-01 -1.02873075e+00 4.66755569e-01 6.65016532e-01 6.79216206e-01 -1.07196772e+00 -1.71873897e-01 2.27896184e-01 -1.36495307e-01 -9.96982217e-01 -5.88388681e-01 5.02894282e-01 -7.80704916e-01 -7.67848670e-01 -8.16716373e-01 -1.03701997e+00 6.33095086e-01 6.42820224e-02 8.63795578e-01 -2.22833291e-01 -5.43219596e-02 -5.91028742e-02 -3.76048386e-01 -4.53024685e-01 -1.58881113e-01 1.99801579e-01 -1.92744061e-02 2.24984232e-02 4.79386002e-01 -5.77369213e-01 -5.07953703e-01 3.01338851e-01 -9.42371368e-01 -1.18586250e-01 6.78392112e-01 1.01623321e+00 8.06707799e-01 1.87382057e-01 5.87845802e-01 -9.73426461e-01 5.23678124e-01 -3.80726904e-01 -5.38411796e-01 3.18364948e-02 -7.17413545e-01 9.70821306e-02 8.04156125e-01 -5.93095720e-01 -1.13500726e+00 2.42132068e-01 -3.65628958e-01 -4.06751722e-01 -3.76516432e-01 3.99720848e-01 -4.35581863e-01 2.80592710e-01 8.01856339e-01 3.51069450e-01 -9.80530605e-02 -5.33198237e-01 3.65788162e-01 7.29682863e-01 4.69416887e-01 -6.83034003e-01 8.70020747e-01 2.93420285e-01 -2.71325469e-01 -5.76925993e-01 -9.07631576e-01 -5.13373971e-01 -4.84649628e-01 1.47502184e-01 5.92180192e-01 -8.91233146e-01 -4.02182102e-01 5.46721756e-01 -8.49094689e-01 -4.59444463e-01 -3.88169974e-01 6.75417244e-01 -4.38503951e-01 4.48354363e-01 -5.11446238e-01 -5.33270121e-01 -3.57508242e-01 -1.01428425e+00 6.25980020e-01 4.70688552e-01 -2.48555139e-01 -7.28292882e-01 -2.55210191e-01 1.46403715e-01 2.91746616e-01 -2.12473366e-02 1.21567106e+00 -7.57847726e-01 -1.07162297e-01 -1.39975231e-02 -3.29368412e-01 7.13746905e-01 2.46155500e-01 -2.31126789e-03 -1.00332177e+00 -2.41542265e-01 -1.75146133e-01 -1.63701817e-01 1.08166814e+00 4.70401496e-01 1.36807394e+00 -1.67875618e-01 -2.31003508e-01 5.37235260e-01 1.22922897e+00 3.23153734e-01 4.44516599e-01 2.93679893e-01 5.53839684e-01 4.98000503e-01 7.23297179e-01 4.89778161e-01 3.11451852e-01 6.76882565e-01 2.72294194e-01 -1.45929322e-01 -2.56346583e-01 -1.18206069e-01 3.54587376e-01 8.15321684e-01 7.97484890e-02 -1.30889699e-01 -9.49095726e-01 5.23593009e-01 -1.76141691e+00 -5.99246562e-01 -4.15629484e-02 2.30317569e+00 1.10494578e+00 5.73981047e-01 8.90677944e-02 2.25052789e-01 7.11708188e-01 1.49982497e-01 -5.29118896e-01 -3.87194514e-01 -1.60597444e-01 6.12949491e-01 3.95629466e-01 4.62908745e-01 -1.35629237e+00 1.03653240e+00 4.69966459e+00 1.43690300e+00 -1.15279365e+00 1.71221197e-01 5.67253411e-01 -2.18732771e-03 -1.48525223e-01 -9.13930833e-02 -7.87429571e-01 7.92002618e-01 6.24566853e-01 -1.12566225e-01 2.82188088e-01 1.00526154e+00 -6.17388412e-02 -1.06576923e-03 -8.83849144e-01 9.72290993e-01 -1.73091829e-01 -9.75014091e-01 2.19335649e-02 -1.67946234e-01 6.14328682e-01 -1.63166746e-02 1.21579461e-01 7.32258439e-01 3.08430612e-01 -6.37110472e-01 7.49840140e-01 1.40191913e-01 6.79385543e-01 -9.96349871e-01 6.95195138e-01 5.30470729e-01 -1.19179201e+00 -1.32758781e-01 -4.35870349e-01 2.42353663e-01 -1.64766937e-01 8.73017848e-01 -5.31214893e-01 5.28831184e-01 8.63468885e-01 4.68296289e-01 -3.82464290e-01 1.04928136e+00 -3.69801641e-01 7.13837266e-01 -5.20377278e-01 -9.08823404e-03 2.88346946e-01 -5.48935346e-02 4.80695873e-01 1.41882086e+00 2.76689470e-01 -6.08684272e-02 2.80137271e-01 6.89603746e-01 -2.85260767e-01 3.84016097e-01 -1.83325797e-01 2.48671770e-01 5.34562230e-01 1.09095943e+00 -7.88161159e-01 -3.02956551e-01 -4.15778011e-01 9.41028237e-01 3.11819196e-01 2.61091024e-01 -9.45443153e-01 -6.90769911e-01 6.14325523e-01 6.84207976e-02 3.30199897e-01 -1.68332115e-01 -4.57859576e-01 -9.89244759e-01 1.92093015e-01 -6.42827988e-01 5.63001156e-01 -2.29253605e-01 -1.38525987e+00 5.07188499e-01 1.83096111e-01 -1.28242874e+00 -6.21127337e-02 -5.44990420e-01 -5.53628623e-01 8.10728371e-01 -1.74645793e+00 -8.82500410e-01 -1.76698551e-01 7.11962938e-01 5.24721205e-01 -6.59341812e-02 7.19918787e-01 5.40146112e-01 -6.85809374e-01 8.38272691e-01 2.43919253e-01 2.21976578e-01 9.84350562e-01 -1.31589353e+00 1.43165067e-01 7.60995328e-01 7.88919777e-02 4.77192193e-01 3.90928864e-01 -2.89438546e-01 -6.95477068e-01 -1.10152173e+00 8.87134552e-01 1.10837584e-02 4.45100814e-01 -3.89675587e-01 -1.04956877e+00 4.01429236e-01 -3.31834964e-02 1.82231784e-01 6.93020403e-01 3.60085927e-02 -3.84501755e-01 -3.74703169e-01 -1.21248400e+00 6.72104239e-01 9.72351730e-01 -2.53445774e-01 -5.31693757e-01 1.65188611e-01 6.11704707e-01 -2.31837526e-01 -7.70307839e-01 6.98845625e-01 2.90736973e-01 -6.86600506e-01 9.48053539e-01 -2.62062013e-01 2.79127598e-01 -3.30032229e-01 -1.94096178e-01 -1.29673338e+00 -3.89585167e-01 -3.81709903e-01 -1.67800352e-01 1.46903849e+00 5.93372226e-01 -7.94538617e-01 7.28413463e-01 3.46034020e-01 -2.44566023e-01 -7.85250008e-01 -1.07850730e+00 -9.39640105e-01 4.33373988e-01 -4.43915427e-01 4.34142470e-01 1.06426978e+00 -3.43711615e-01 3.38514060e-01 -5.72786070e-02 8.62164125e-02 5.51940262e-01 1.73232183e-01 4.35002446e-01 -1.33816242e+00 -4.57860202e-01 -7.44721830e-01 -3.17877159e-02 -1.21520054e+00 1.36184573e-01 -9.25056517e-01 1.35373741e-01 -1.25333345e+00 1.57421738e-01 -8.74521017e-01 -6.31237745e-01 8.04678798e-01 -4.03063744e-01 1.24978058e-01 2.39367738e-01 2.53936350e-01 -4.08234328e-01 6.24591947e-01 1.08197057e+00 -1.67687461e-01 -4.34121400e-01 1.59867689e-01 -7.39761949e-01 1.01733434e+00 1.09706068e+00 -7.16541946e-01 -3.33242118e-01 -3.16376865e-01 -2.38675743e-01 -4.08280164e-01 1.00993223e-01 -9.11092460e-01 1.06703795e-01 -1.50039047e-01 4.30329978e-01 -3.43417495e-01 3.17491204e-01 -6.29893363e-01 -3.40307474e-01 3.82702529e-01 -8.43483284e-02 -3.60417694e-01 3.54628682e-01 3.43470514e-01 -2.53610671e-01 -4.54361469e-01 1.10080802e+00 6.29440323e-02 -8.78986359e-01 2.43997663e-01 -3.41734916e-01 2.25269720e-01 9.48303580e-01 -1.35860428e-01 -7.87458494e-02 6.66252151e-02 -8.31551254e-01 2.34960362e-01 1.76063225e-01 2.40786299e-01 4.24543798e-01 -1.24947906e+00 -7.33757973e-01 2.28169233e-01 2.64205456e-01 3.87598544e-01 1.73515603e-01 6.36642933e-01 -2.61004359e-01 1.72289610e-01 -1.35454565e-01 -5.63670278e-01 -1.00149107e+00 3.62403542e-01 1.20505758e-01 -3.84799093e-01 -3.75455916e-01 9.69414115e-01 2.45768517e-01 -6.25781417e-01 3.12405497e-01 -1.92624852e-01 -2.78330922e-01 2.33096510e-01 1.81486920e-01 1.72691330e-01 2.63404921e-02 -3.58020931e-01 -4.77758825e-01 5.33254743e-01 -3.75009596e-01 1.21262826e-01 1.26392365e+00 1.56780444e-02 1.70807406e-01 1.73301533e-01 1.22901046e+00 6.60770237e-02 -1.42648590e+00 -6.15041912e-01 8.63368437e-03 -1.77504390e-01 5.50523866e-03 -7.38761008e-01 -1.08390164e+00 8.04918408e-01 7.19714701e-01 -6.57816827e-02 1.27469003e+00 1.94192857e-01 8.74261856e-01 2.23323330e-01 2.42853850e-01 -1.31729925e+00 6.23656213e-02 5.46075583e-01 5.78724384e-01 -1.19682062e+00 -5.62470481e-02 -4.90869135e-01 -6.36598170e-01 9.52341497e-01 7.19316006e-01 -2.62179136e-01 7.91056097e-01 2.71275163e-01 -9.98348813e-04 1.69925436e-01 -3.48054469e-01 -4.30054545e-01 2.70893544e-01 4.28809822e-01 2.14229420e-01 1.78351432e-01 -4.68645126e-01 9.27041173e-01 -4.31830809e-02 -3.63634735e-01 1.01905093e-01 9.64439929e-01 -4.39291418e-01 -1.36349332e+00 -1.57768160e-01 4.35802132e-01 -5.13633490e-01 -1.56989351e-01 -1.52875558e-01 8.11172664e-01 4.08851326e-01 8.35022211e-01 9.24923196e-02 -2.72116870e-01 5.01881361e-01 2.10962102e-01 3.18625301e-01 -6.71150982e-01 -3.20813775e-01 2.55996019e-01 -1.74503848e-01 -2.73593724e-01 -2.76663750e-01 -6.58587813e-01 -1.56943369e+00 -1.41225770e-01 -3.23272556e-01 3.17250527e-02 4.37171131e-01 9.22629416e-01 1.73765272e-01 4.75397140e-01 7.35413611e-01 -4.55388188e-01 -5.07167876e-01 -9.50166464e-01 -5.70561111e-01 3.84543985e-01 -1.50344279e-02 -8.14869940e-01 -3.51064861e-01 2.28818916e-02]
[9.588699340820312, 1.5976455211639404]
40ec4134-a6b8-4e7b-92dc-54d3951f5a48
objective-probabilistic-and-generalized-noise
2009.08539
null
https://arxiv.org/abs/2009.08539v2
https://arxiv.org/pdf/2009.08539v2.pdf
Objective, Probabilistic, and Generalized Noise Level Dependent Classifications of sets of more or less 2D Periodic Images into Plane Symmetry Groups
Crystallographic symmetry classifications from real-world images with periodicities in two dimensions (2D) are of interest to crystallographers and practitioners of computer vision studies alike. Currently, these classifications are typically made by both communities in a subjective manner that relies on arbitrary thresholds for judgments, and are reported under the pretense of being definitive, which is impossible. Moreover, the computer vision community tends to use direct space methods to make such classifications instead of more powerful and computationally efficient Fourier space methods. This is because the proper functioning of those methods requires more periodic repeats of a unit cell motif than are commonly present in images analyzed by the computer vision community. We demonstrate a novel approach to plane symmetry group classifications that is enabled by Kenichi Kanatani's Geometric Akaike Information Criterion and associated Geometric Akaike weights. Our approach leverages the advantages of working in Fourier space, is well suited for handling the hierarchic nature of crystallographic symmetries, and yields probabilistic results that are generalized noise level dependent. The latter feature means crystallographic symmetry classifications can be updated when less noisy image data and more accurate processing algorithms become available. We demonstrate the ability of our approach to objectively estimate the plane symmetry and pseudosymmetries of sets of synthetic 2D-periodic images with varying amounts of red-green-blue and spread noise. Additionally, we suggest a simple solution to the problem of too few periodic repeats in an input image for practical application of Fourier space methods. In doing so, we effectively solve the decades-old and heretofore intractable problem from computer vision of symmetry detection and classification from images in the presence of noise.
['Peter Moeck', 'Andrew Dempsey']
2020-09-17
null
null
null
null
['symmetry-detection']
['computer-vision']
[ 7.06124723e-01 -2.59983391e-01 6.26893863e-02 -3.10371280e-01 -8.21253419e-01 -5.97565711e-01 7.84626901e-01 -2.16645658e-01 -4.54790741e-01 6.04407251e-01 1.56628668e-01 -3.34123105e-01 -3.95034403e-01 -6.58797443e-01 -2.87670344e-01 -1.08392167e+00 1.71818256e-01 7.30532050e-01 1.50880277e-01 -1.31445583e-02 7.61220157e-01 6.14843607e-01 -1.75307977e+00 1.19586609e-01 5.85349917e-01 9.06674147e-01 2.71564722e-01 5.03287017e-01 -7.38768131e-02 3.58713180e-01 -3.13059479e-01 -1.33686094e-02 3.12775493e-01 -4.36280042e-01 -8.48452330e-01 4.99858439e-01 5.19562662e-01 1.03825055e-01 -3.11407875e-02 1.19421196e+00 3.55935603e-01 1.40910864e-01 1.14377749e+00 -8.44462097e-01 -6.02065921e-01 -3.82175148e-02 -7.11727619e-01 3.03084314e-01 3.78605098e-01 4.10048455e-01 1.15413320e+00 -7.44761825e-01 9.43279088e-01 1.07645702e+00 7.08303690e-01 1.65711969e-01 -1.59488058e+00 -4.34263825e-01 1.29325977e-02 2.82203287e-01 -1.33453703e+00 -5.99264741e-01 7.07884371e-01 -7.99312234e-01 9.36849415e-01 2.67630666e-01 5.09512126e-01 7.75554955e-01 1.37751564e-01 7.75233954e-02 1.24844682e+00 -6.26627564e-01 4.32510525e-01 -4.76511270e-01 -6.48466051e-02 4.51618344e-01 2.42465138e-01 -7.90073648e-02 -4.68935966e-01 -2.28769198e-01 8.26354980e-01 7.00608119e-02 -3.74524027e-01 -6.77339494e-01 -1.49921131e+00 7.36789167e-01 2.13372558e-01 2.21387342e-01 -4.31405991e-01 -1.30441830e-01 1.97886422e-01 2.20430374e-01 4.27640587e-01 7.63110042e-01 -4.81671169e-02 -1.49445042e-01 -1.01506364e+00 2.77716428e-01 5.06703854e-01 6.71954513e-01 5.55111468e-01 -2.07987741e-01 4.56930935e-01 6.94898069e-01 2.54698098e-01 4.39092040e-01 3.39991152e-01 -1.25397956e+00 -3.90481874e-02 3.65046352e-01 1.25708878e-01 -1.02979565e+00 -3.77963364e-01 -6.36302158e-02 -8.88805866e-01 7.19653785e-01 6.15979493e-01 6.22197628e-01 -8.58072460e-01 1.49839985e+00 2.48336747e-01 -2.40327671e-01 -1.28548130e-01 7.99454927e-01 2.95089990e-01 3.50799024e-01 -3.20041537e-01 -4.49986160e-01 1.32558274e+00 -1.92215249e-01 -4.98087764e-01 2.38183498e-01 3.12867016e-01 -1.08211458e+00 1.13616228e+00 6.45886421e-01 -9.32614565e-01 -2.22493216e-01 -1.17727041e+00 -3.73749323e-02 -6.27989769e-02 -2.51796484e-01 5.87245166e-01 6.46123409e-01 -7.62755156e-01 8.20708930e-01 -9.62943077e-01 -3.77293885e-01 5.05344152e-01 3.42783928e-01 -6.02968931e-01 1.39573179e-02 -3.59924912e-01 7.80016780e-01 1.88460961e-01 -1.64101720e-01 7.20363483e-02 -3.82578105e-01 -7.05817282e-01 -2.47786641e-01 4.20494765e-01 -5.49380898e-01 1.19559109e+00 -8.28824580e-01 -1.13217616e+00 1.08281827e+00 -4.85428274e-01 -8.70666206e-02 5.61669886e-01 4.44109797e-01 -1.28404945e-01 3.54141980e-01 3.02990675e-01 3.09410721e-01 9.05138493e-01 -1.23431110e+00 -5.17081797e-01 -6.86992943e-01 -3.24773818e-01 2.05484182e-01 2.13165328e-01 -1.35200039e-01 -1.63883135e-01 -5.81442654e-01 8.88696313e-01 -7.77465761e-01 -4.58956137e-02 2.67245501e-01 -2.05588803e-01 -1.92509145e-01 7.36544430e-01 -3.91437054e-01 7.66709864e-01 -2.08133888e+00 -6.33891597e-02 5.06095886e-01 4.56916362e-01 -1.37437984e-01 1.68357477e-01 3.96067500e-01 -2.12997526e-01 1.08919404e-01 -3.21312875e-01 7.99678713e-02 -1.02617070e-01 1.05324887e-01 -2.63847500e-01 8.23494256e-01 1.58161446e-01 4.19040263e-01 -8.78353953e-01 -3.10342103e-01 3.68429512e-01 2.46825188e-01 -5.69463789e-01 -2.96719760e-01 7.59596899e-02 4.92521793e-01 -3.50547373e-01 6.74829483e-01 6.66416347e-01 -6.16343796e-01 2.78563112e-01 -3.64095718e-01 -3.90023500e-01 5.32141805e-01 -1.34756994e+00 1.24715352e+00 2.30642837e-02 4.71584857e-01 -7.59138316e-02 -9.75780547e-01 7.87380576e-01 8.34533349e-02 5.69767952e-01 -6.34832740e-01 -1.37438178e-01 1.63476765e-01 2.00065523e-01 -3.52632880e-01 3.73050094e-01 -5.23498356e-01 3.30036730e-01 7.87947834e-01 -8.24655816e-02 -6.99466705e-01 2.51177341e-01 -1.91333130e-01 1.16038036e+00 8.86661038e-02 5.32738566e-01 -4.37665761e-01 3.28447729e-01 2.57188857e-01 4.93332803e-01 8.88220072e-01 -1.87184572e-01 1.04457140e+00 4.59622681e-01 -7.31898367e-01 -1.72292709e+00 -1.18509960e+00 -5.13864338e-01 5.77422142e-01 -4.14374918e-02 -3.31989527e-01 -5.13742924e-01 -1.46802902e-01 -9.40681919e-02 2.38201007e-01 -3.44201982e-01 1.03700921e-01 -4.01568204e-01 -9.24097657e-01 4.28724624e-02 1.92216560e-01 2.65446097e-01 -7.70607948e-01 -5.50555706e-01 7.74848089e-02 -1.10683188e-01 -9.48043227e-01 -3.11631173e-01 4.68146987e-02 -6.35716379e-01 -1.40170991e+00 -7.13465154e-01 -6.03384495e-01 8.09128404e-01 5.70841730e-01 8.71348441e-01 -1.32148312e-02 -6.45850718e-01 3.84459198e-01 -2.05729663e-01 -9.64694992e-02 -4.30369794e-01 -2.78671950e-01 3.46800715e-01 -2.21883953e-01 7.07062721e-01 -8.02127004e-01 -4.38638866e-01 5.57767808e-01 -6.92265570e-01 1.41285151e-01 3.36196512e-01 1.08270586e+00 8.57685030e-01 5.13988256e-01 2.59951264e-01 -7.20080078e-01 4.67775106e-01 -7.30974302e-02 -7.61000812e-01 4.06484678e-02 -6.74413681e-01 1.68614700e-01 4.40112382e-01 -1.62794545e-01 -7.69757509e-01 1.28598347e-01 1.35405868e-01 -2.22660303e-01 -2.13488147e-01 2.79359907e-01 7.95274526e-02 -3.19304109e-01 9.53271449e-01 3.68609875e-01 3.35411102e-01 -2.66904116e-01 5.59997782e-02 8.77503753e-01 6.98839724e-01 -6.40016615e-01 9.16994989e-01 9.17658091e-01 2.52579600e-01 -1.27715993e+00 -4.83841836e-01 -6.95017159e-01 -6.79037571e-01 -8.66239220e-02 7.89532781e-01 -7.06893802e-01 -9.16607678e-01 5.24830222e-01 -8.78981292e-01 2.19955027e-01 -1.59717321e-01 5.27154863e-01 -9.13773298e-01 9.07502413e-01 -2.35133201e-01 -6.93557203e-01 1.81945816e-01 -1.36427450e+00 1.07778025e+00 -1.03581339e-01 -5.67911386e-01 -6.08531296e-01 -1.52394012e-01 5.03109694e-01 2.86341012e-01 1.56658411e-01 1.30763209e+00 -2.63499767e-01 -7.07013428e-01 -8.35106671e-02 -2.73318708e-01 2.32894614e-01 5.43273389e-01 1.42502859e-01 -9.53162074e-01 -2.85673402e-02 4.09587890e-01 -2.31246293e-01 7.90150762e-01 5.98679602e-01 1.11869037e+00 -9.78196487e-02 -9.75179020e-03 5.18240094e-01 1.21231759e+00 2.51351327e-01 4.89321560e-01 4.93102282e-01 5.02965868e-01 8.26774597e-01 2.25836799e-01 3.83031756e-01 8.14825371e-02 7.62775362e-01 8.03769231e-02 1.12264134e-01 1.79180458e-01 1.28527999e-01 -2.15351090e-01 5.70879817e-01 -6.07251346e-01 -4.93892431e-02 -1.06146824e+00 1.38472512e-01 -1.68369901e+00 -1.44926429e+00 7.85716623e-02 2.45061111e+00 7.00978756e-01 2.44273528e-01 4.18397412e-02 4.51175809e-01 7.98504949e-01 1.03934623e-01 -6.06893718e-01 -1.04023926e-02 -2.79079705e-01 5.91082424e-02 5.56633115e-01 3.22265029e-01 -1.15207326e+00 5.03684759e-01 6.70472860e+00 8.18992198e-01 -1.01799393e+00 -4.46879447e-01 6.34052336e-01 7.20016658e-02 -3.42427194e-01 1.02040797e-01 -4.92077082e-01 4.47207659e-01 2.71951109e-01 -1.94114253e-01 5.86902142e-01 7.50440359e-01 3.42363030e-01 -4.48667198e-01 -1.11811936e+00 1.45701706e+00 -3.40601131e-02 -1.60468554e+00 -6.18735626e-02 4.21737999e-01 5.48925698e-01 -1.92821398e-02 2.05851287e-01 -6.39228463e-01 2.08854914e-01 -1.12076926e+00 5.13150156e-01 3.69553059e-01 7.67448008e-01 -5.64501405e-01 5.00103712e-01 7.41373599e-02 -8.49547386e-01 1.56215057e-01 -4.44451153e-01 -1.67349607e-01 1.51830643e-01 6.52347505e-01 -7.76878476e-01 2.29571790e-01 6.80822730e-01 6.58771873e-01 -1.14896312e-01 9.70988691e-01 1.01284899e-01 2.25199714e-01 -5.99484682e-01 1.94904462e-01 2.01813772e-01 -6.56913996e-01 5.29134512e-01 6.32361054e-01 3.71952772e-01 2.27078170e-01 -1.34630147e-02 8.36979568e-01 2.11023033e-01 -2.06225514e-01 -7.66489148e-01 -2.08847910e-01 4.10645932e-01 1.05280077e+00 -1.21460652e+00 -4.32414934e-02 -3.87057155e-01 5.88445067e-01 -5.78526482e-02 3.15005302e-01 8.32011737e-03 -1.43851116e-01 6.90256000e-01 4.66682881e-01 4.81964380e-01 -5.99543571e-01 -5.88928163e-01 -1.13287425e+00 2.55850524e-01 -9.18475211e-01 1.32568002e-01 -9.07817960e-01 -1.49120629e+00 3.32034081e-01 -9.80797410e-02 -1.27422476e+00 -3.80663723e-01 -1.06996274e+00 -4.68507260e-01 8.26265931e-01 -1.10937691e+00 -7.63172984e-01 -1.34464115e-01 2.84316003e-01 3.07975054e-01 -1.60090849e-01 7.71175206e-01 -1.12755299e-01 5.56032360e-02 1.21289194e-02 6.23064041e-01 -1.79803342e-01 6.79742813e-01 -1.25961876e+00 4.66974229e-01 6.69625461e-01 8.55442807e-02 6.85948670e-01 9.74909902e-01 -5.86770654e-01 -1.20196557e+00 -4.66816336e-01 8.35987508e-01 -5.17504871e-01 7.23929346e-01 -2.86701441e-01 -8.34098995e-01 2.98821539e-01 -3.73247176e-01 -3.17035973e-01 6.32497132e-01 2.09664389e-01 -4.67166811e-01 3.46003294e-01 -1.13426077e+00 6.99124098e-01 1.15478396e+00 -7.25827157e-01 -6.91871643e-01 5.86700499e-01 -3.14080939e-02 -8.06054771e-02 -7.31858492e-01 2.98808008e-01 5.41298509e-01 -1.08694744e+00 1.03515124e+00 -3.99466932e-01 3.48567814e-01 -6.42109394e-01 -2.98283607e-01 -8.05242062e-01 -3.46783280e-01 -5.47218621e-01 7.96917200e-01 5.19090831e-01 1.84742242e-01 -6.10582948e-01 9.16412234e-01 5.46330631e-01 -2.99743973e-02 -4.71175730e-01 -1.33195114e+00 -8.12302172e-01 -1.96125641e-01 -3.43909860e-01 2.89238662e-01 1.00537562e+00 -1.87179357e-01 8.15478489e-02 1.55539019e-02 -5.84538840e-02 9.06889915e-01 2.82134533e-01 7.55795419e-01 -1.51655710e+00 -1.29653305e-01 -6.11588895e-01 -1.05873239e+00 -7.58955359e-01 1.21842055e-02 -7.22884715e-01 -5.26237767e-03 -1.07233489e+00 3.38909775e-01 -4.72845137e-01 -4.72483858e-02 2.77505994e-01 3.91185880e-01 4.37478721e-01 -9.82359424e-02 7.20218658e-01 -3.12537938e-01 4.70658243e-01 1.14848375e+00 -1.48311958e-01 -2.87149269e-02 3.08976360e-02 -6.56485260e-01 9.79961336e-01 4.92101163e-01 -3.20927352e-01 -3.15183729e-01 1.07798204e-02 3.70661527e-01 -3.87905389e-01 5.28861463e-01 -1.00389302e+00 2.36471474e-01 -1.96561068e-01 4.94378984e-01 -5.48690856e-01 3.02876800e-01 -5.17133772e-01 3.37260664e-01 1.92104116e-01 1.82372387e-02 1.27633184e-01 -3.58706146e-01 5.25939822e-01 -1.22327611e-01 -2.05228269e-01 1.01855564e+00 -4.32374448e-01 -4.79882836e-01 1.04150355e-01 -6.52106762e-01 -5.89200743e-02 7.06474960e-01 -7.37385869e-01 -2.34595254e-01 -4.22947049e-01 -7.41448581e-01 -2.00662971e-01 1.26079798e+00 2.48966254e-02 5.88614166e-01 -1.13009059e+00 -3.91109377e-01 3.54403526e-01 2.88794905e-01 3.70566696e-02 2.54047245e-01 8.90906453e-01 -8.36898208e-01 1.39764741e-01 -4.30373639e-01 -9.92714524e-01 -1.31476188e+00 3.67742568e-01 2.26535872e-01 2.32433960e-01 -9.17632222e-01 5.39500713e-01 4.44781065e-01 -4.12749410e-01 -1.37371749e-01 -3.20274025e-01 4.26289812e-02 3.63803410e-04 5.75559914e-01 2.91072547e-01 2.58979827e-01 -8.72868180e-01 -2.04278439e-01 8.83363426e-01 -3.74071598e-01 -2.81499863e-01 1.48628747e+00 -1.35780066e-01 -1.09579265e-01 3.06379437e-01 7.99962759e-01 -1.05557762e-01 -1.29589474e+00 -3.54956150e-01 1.35754719e-02 -5.73801696e-01 -1.02572545e-01 -3.86156976e-01 -3.69976848e-01 6.33845031e-01 4.50038612e-01 2.62148857e-01 8.53824735e-01 2.24357098e-01 2.83022761e-01 6.90268934e-01 3.63888890e-01 -1.26888466e+00 -1.47136301e-01 4.46101516e-01 6.61480069e-01 -1.43764877e+00 3.31122905e-01 -3.83865535e-01 -3.25153917e-01 1.39674580e+00 -8.50030046e-04 -4.00371589e-02 4.66700017e-01 3.97868678e-02 1.82866573e-01 -4.21424598e-01 -4.31091696e-01 -7.54289478e-02 1.38804719e-01 7.63112366e-01 2.81133473e-01 -1.06781192e-01 -1.78253919e-01 -8.54109228e-02 -4.93170321e-01 -1.71033263e-01 6.19887590e-01 1.00222170e+00 -6.67219043e-01 -1.13550782e+00 -6.16210759e-01 6.27520978e-01 -3.19934189e-01 -3.76952477e-02 -2.57165432e-01 5.84976733e-01 -7.79577568e-02 7.46105492e-01 2.40360126e-01 -7.28325322e-02 -4.12550941e-02 7.03372806e-02 6.33591890e-01 -5.03834963e-01 2.00925678e-01 2.14594245e-01 -1.41620725e-01 -4.60176706e-01 -7.79205143e-01 -9.19655502e-01 -1.13668752e+00 -3.59234720e-01 -3.16173434e-01 8.37868378e-02 7.90203393e-01 1.12633181e+00 1.86834976e-01 2.02960353e-02 5.37565053e-01 -1.15260875e+00 -5.89188337e-01 -5.43926656e-01 -7.46874869e-01 5.67368329e-01 2.47389257e-01 -9.31145906e-01 -6.65163398e-01 1.43409431e-01]
[8.505260467529297, -2.3243463039398193]
67cbb04f-b405-412f-808d-a879d5257cee
blind-image-deconvolution-using-deep
1802.04073
null
http://arxiv.org/abs/1802.04073v4
http://arxiv.org/pdf/1802.04073v4.pdf
Blind Image Deconvolution using Deep Generative Priors
This paper proposes a novel approach to regularize the \textit{ill-posed} and \textit{non-linear} blind image deconvolution (blind deblurring) using deep generative networks as priors. We employ two separate generative models --- one trained to produce sharp images while the other trained to generate blur kernels from lower-dimensional parameters. To deblur, we propose an alternating gradient descent scheme operating in the latent lower-dimensional space of each of the pretrained generative models. Our experiments show promising deblurring results on images even under large blurs, and heavy noise. To address the shortcomings of generative models such as mode collapse, we augment our generative priors with classical image priors and report improved performance on complex image datasets. The deblurring performance depends on how well the range of the generator spans the image class. Interestingly, our experiments show that even an untrained structured (convolutional) generative networks acts as an image prior in the image deblurring context allowing us to extend our results to more diverse natural image datasets.
['Muhammad Asim', 'Fahad Shamshad', 'Ali Ahmed']
2018-02-12
null
null
null
null
['image-deconvolution']
['computer-vision']
[ 3.41000974e-01 -5.15066423e-02 4.34632331e-01 -1.01907760e-01 -6.55637383e-01 -7.92303026e-01 7.77064979e-01 -1.01682854e+00 -1.99046358e-01 7.74987876e-01 6.64589703e-01 -1.55903175e-01 -1.76095609e-02 -3.37420195e-01 -8.57080400e-01 -1.06022930e+00 2.93682069e-01 2.41779432e-01 -6.48405179e-02 8.86260718e-02 2.90911317e-01 3.28746974e-01 -1.18549514e+00 1.41378164e-01 9.80730236e-01 5.41594446e-01 4.06093925e-01 1.11360979e+00 3.73057246e-01 8.84160399e-01 -6.54925644e-01 -3.70268047e-01 2.89641231e-01 -6.76303029e-01 -7.03451574e-01 4.61264879e-01 5.71620882e-01 -6.32835746e-01 -8.10704947e-01 1.28865004e+00 3.75479162e-01 -1.31935179e-02 9.03171420e-01 -7.65339136e-01 -1.42668986e+00 4.47109878e-01 -4.50410724e-01 3.29239368e-01 2.21234579e-02 3.63564521e-01 5.01583397e-01 -8.28016937e-01 5.27757466e-01 1.18988848e+00 8.43283951e-01 6.05601907e-01 -1.49604452e+00 -3.22680473e-01 -2.31869519e-01 4.68503088e-02 -1.19051147e+00 -8.34223807e-01 5.49471259e-01 -5.67203283e-01 5.06756961e-01 1.94949940e-01 -1.40461884e-02 1.26379979e+00 5.07889353e-02 5.14010847e-01 1.25907683e+00 -3.01793218e-01 1.99922428e-01 -1.01889692e-01 2.66254451e-02 4.06955332e-01 4.10494179e-01 4.20725137e-01 -2.45662898e-01 -2.39861265e-01 1.20800328e+00 -1.40710190e-01 -8.85564804e-01 -6.83972687e-02 -1.15439665e+00 7.30698526e-01 4.11714673e-01 2.61218399e-01 -4.96973783e-01 5.45113444e-01 -2.47174174e-01 7.74608776e-02 6.67511523e-01 5.92999518e-01 -7.53365234e-02 4.51823063e-02 -1.35710943e+00 9.05545056e-03 6.39434099e-01 7.17830122e-01 8.08076501e-01 2.01836854e-01 -3.83150339e-01 9.36764300e-01 2.94411331e-01 5.08740544e-01 6.53558135e-01 -1.15372896e+00 7.92255551e-02 -2.71048397e-01 5.31125963e-01 -7.84569681e-01 6.60289004e-02 -5.83619356e-01 -1.04444885e+00 7.14830533e-02 4.57902551e-01 -4.08374965e-01 -1.44572222e+00 1.65246975e+00 -2.90996522e-01 5.30446589e-01 2.08090693e-01 1.30346990e+00 5.37095487e-01 6.26962900e-01 -2.33327255e-01 -4.69879359e-02 1.19138515e+00 -9.87289190e-01 -7.33540535e-01 -5.80200970e-01 2.01051868e-02 -1.04837275e+00 7.94156134e-01 3.31052542e-01 -1.27091563e+00 -4.48615104e-01 -9.55954671e-01 -3.94525975e-02 9.56913978e-02 4.31418419e-01 3.34049463e-01 7.16932356e-01 -1.56793153e+00 4.23102200e-01 -7.06484199e-01 -2.06671730e-01 3.59207958e-01 8.37351009e-02 -1.37940496e-01 -3.88485968e-01 -1.04914868e+00 9.42832053e-01 1.70555681e-01 1.70982361e-01 -1.34787464e+00 -4.63554621e-01 -7.67795384e-01 3.23595144e-02 -2.03684777e-01 -1.00181353e+00 9.38401461e-01 -8.76846433e-01 -1.48102677e+00 8.21843565e-01 -2.72602707e-01 -4.51937139e-01 7.60695338e-01 -6.18519783e-01 -7.24493638e-02 1.89789057e-01 -1.25818024e-03 6.00084245e-01 1.54638529e+00 -1.72681165e+00 1.29142866e-01 -1.13870487e-01 -2.74144232e-01 1.29650861e-01 -1.24893486e-01 -1.12965979e-01 -4.85391289e-01 -1.10834920e+00 1.05888166e-01 -9.86149728e-01 -2.02490687e-01 -3.55908930e-01 -5.69261432e-01 4.62748915e-01 7.93831408e-01 -1.00142670e+00 9.97941732e-01 -2.23025703e+00 5.42510033e-01 -2.45339513e-01 3.21868867e-01 2.66068071e-01 -2.36424476e-01 9.98022780e-02 -3.19244385e-01 1.08802162e-01 -5.45232415e-01 -6.00469470e-01 -9.15109813e-02 2.69308448e-01 -6.07205689e-01 7.40579605e-01 1.50253683e-01 9.52544212e-01 -7.34942257e-01 1.70053989e-01 4.30332683e-02 8.01971138e-01 -5.73515713e-01 5.34339249e-01 -4.03398424e-02 6.28884912e-01 -1.85008887e-02 9.77431238e-02 1.05955184e+00 -6.08069479e-01 -1.55193567e-01 -3.37452173e-01 1.02000035e-01 -1.46259025e-01 -9.18593943e-01 1.68270791e+00 -2.91832507e-01 9.38214898e-01 3.51968080e-01 -7.18188584e-01 5.52645087e-01 3.73566031e-01 3.21851373e-02 -8.74539837e-02 4.00104709e-02 1.88781112e-01 -1.84996367e-01 -6.64037585e-01 6.35108829e-01 -2.96023458e-01 3.89665633e-01 6.29431784e-01 2.34844714e-01 -3.63476217e-01 -7.42212012e-02 2.52225548e-01 9.65807438e-01 9.20577198e-02 -3.22751522e-01 -3.84079039e-01 2.35620007e-01 -3.80016983e-01 -1.61620155e-01 1.10219455e+00 3.96280363e-02 1.41231894e+00 2.36126676e-01 -1.70102805e-01 -1.28443360e+00 -1.17285407e+00 -1.11202329e-01 9.10546422e-01 2.99630404e-01 1.31087080e-01 -1.09011054e+00 -3.97182912e-01 -3.02146733e-01 8.97301614e-01 -6.41295254e-01 -2.37706631e-01 -4.13797796e-01 -1.43874168e+00 8.09120715e-01 2.83016056e-01 6.34895802e-01 -6.00046575e-01 -1.09872416e-01 4.28094268e-02 -4.15978551e-01 -1.12135684e+00 -8.73353720e-01 -1.84470713e-02 -5.98740578e-01 -8.18930686e-01 -1.40946186e+00 -7.09204793e-01 9.09396231e-01 3.89864445e-01 9.90258873e-01 -1.67265132e-01 -1.37744173e-01 3.63117099e-01 1.22757154e-02 1.98548481e-01 -7.90512919e-01 -3.75884265e-01 -5.55722527e-02 1.39966443e-01 -8.01918283e-02 -7.13790596e-01 -8.07594955e-01 4.39389497e-01 -1.19613659e+00 3.14127021e-02 5.62299609e-01 9.68848646e-01 -3.63509804e-02 5.68603948e-02 2.45222986e-01 -4.85923499e-01 9.21734869e-01 -5.42523742e-01 -4.20420229e-01 1.17947876e-01 -5.63801527e-01 5.05596161e-01 2.11997747e-01 -8.37164104e-01 -1.25433326e+00 -2.64548451e-01 1.50521860e-01 -7.17339396e-01 -3.02434742e-01 2.61929810e-01 7.32992142e-02 -2.52747890e-02 1.00008142e+00 6.62420571e-01 -4.13729362e-02 -7.19162762e-01 7.29654729e-01 5.68544745e-01 1.04408431e+00 -5.40588498e-01 9.39817071e-01 4.68896300e-01 -3.07162225e-01 -7.79263854e-01 -7.19915748e-01 -5.34740239e-02 -3.20407689e-01 1.27480458e-02 1.08963859e+00 -1.11167514e+00 -1.92890063e-01 8.92630696e-01 -1.43171203e+00 -3.57337177e-01 -1.27398977e-02 4.80505496e-01 -5.35439909e-01 7.06760705e-01 -9.99238431e-01 -6.42205238e-01 -2.46207520e-01 -1.41547918e+00 9.67403233e-01 2.37088442e-01 5.26829921e-02 -1.06358349e+00 1.83207035e-01 4.37966436e-01 8.19006503e-01 -1.59690335e-01 5.87230444e-01 -2.27740422e-01 -6.97009504e-01 1.77600998e-02 -6.58176243e-01 5.61204076e-01 3.53415906e-01 -4.03799772e-01 -1.33842468e+00 -3.99897933e-01 2.80315131e-01 -1.43158227e-01 1.38280320e+00 7.46489406e-01 9.99550819e-01 -6.16567492e-01 -6.93352371e-02 9.21318293e-01 1.19878364e+00 -2.25783318e-01 1.04282236e+00 1.70810983e-01 8.46836627e-01 1.33541167e-01 -3.70071620e-01 2.13587135e-01 9.73079652e-02 4.02110517e-01 4.56130058e-01 -2.10044965e-01 -7.30164111e-01 -1.58582535e-02 5.18034637e-01 4.43973213e-01 -2.05882087e-01 -5.35199285e-01 -6.06545806e-01 6.64044738e-01 -1.62122381e+00 -1.01838994e+00 -1.60370991e-01 2.00892019e+00 1.14663434e+00 -2.16623008e-01 -2.44863138e-01 -4.18258905e-01 1.03865004e+00 3.01414281e-01 -5.09451747e-01 1.62293255e-01 -4.62070584e-01 1.13822795e-01 5.76888978e-01 8.73086751e-01 -1.14913094e+00 9.48291898e-01 7.12533045e+00 5.35594046e-01 -1.23722124e+00 2.37224549e-01 6.78279042e-01 5.17447591e-02 -4.42885607e-01 -3.50678563e-02 -2.67439336e-01 6.75469518e-01 6.63974226e-01 1.72067761e-01 8.68635118e-01 3.61794710e-01 2.77255654e-01 -7.70984441e-02 -9.01298106e-01 1.14390182e+00 1.79691494e-01 -1.37051535e+00 4.39703045e-03 1.70687899e-01 1.08482492e+00 2.30519354e-01 4.33623344e-01 -2.55732566e-01 6.81682229e-01 -1.32295656e+00 8.40043604e-01 8.37243617e-01 9.39131737e-01 -1.47393987e-01 4.98723716e-01 2.50976324e-01 -3.72512043e-01 2.25065500e-01 -2.92315036e-01 2.19696790e-01 2.14734510e-01 7.94506252e-01 -7.28352308e-01 3.14910620e-01 5.73257744e-01 7.06103802e-01 -4.98205483e-01 1.04076922e+00 -2.88691044e-01 7.68441916e-01 -6.55727684e-02 7.33498514e-01 -1.34138390e-02 -2.78060228e-01 8.67526889e-01 1.35340142e+00 5.86930811e-01 8.39036852e-02 -3.65057170e-01 1.36293650e+00 -2.25322589e-01 -7.87423432e-01 -2.17681110e-01 -1.57897875e-01 5.42838350e-02 9.42082524e-01 -4.34224337e-01 -3.09684753e-01 1.46377981e-02 1.71968901e+00 -4.72829267e-02 1.07319391e+00 -8.76643479e-01 -1.23346396e-01 8.01970899e-01 -1.49197560e-02 5.45249403e-01 -3.81145805e-01 -4.48348612e-01 -1.74065626e+00 -2.67827898e-01 -6.46729767e-01 -2.92501356e-02 -1.34484363e+00 -1.45379531e+00 7.13009059e-01 -3.40774320e-02 -8.35092485e-01 -3.13287616e-01 -5.06113827e-01 -6.35073066e-01 1.37974644e+00 -1.58111167e+00 -1.05201793e+00 -4.38627481e-01 5.01584113e-01 5.01308203e-01 2.99963858e-02 5.14938712e-01 1.41364381e-01 -4.75188434e-01 2.95599908e-01 4.01580185e-01 1.29621953e-03 9.79436219e-01 -1.20422733e+00 4.51248229e-01 1.44138324e+00 -1.04553789e-01 7.88585722e-01 1.18873549e+00 -6.40633106e-01 -1.13780725e+00 -1.02275014e+00 3.46547753e-01 -4.55628216e-01 6.24252021e-01 -2.56709871e-03 -9.75453019e-01 7.68433094e-01 6.82087183e-01 1.03647932e-01 2.32847199e-01 -4.68633711e-01 -4.76755708e-01 3.78474206e-01 -9.97919202e-01 4.60306227e-01 7.81481624e-01 -6.43676460e-01 -6.75697684e-01 1.98114395e-01 6.16939008e-01 -4.55450207e-01 -4.85487282e-01 1.26496658e-01 2.52563775e-01 -1.08610249e+00 1.17523420e+00 -4.57113564e-01 7.44739771e-01 -4.07708675e-01 -3.89923304e-02 -1.54505348e+00 -5.69641054e-01 -1.05255711e+00 -2.85087407e-01 9.88346159e-01 2.16004848e-01 -5.66803575e-01 4.97592658e-01 5.93217492e-01 -2.01678991e-01 -6.31392095e-03 -4.38222855e-01 -5.00745237e-01 8.00494179e-02 -2.20504344e-01 2.53542364e-01 9.16657031e-01 -5.96681952e-01 1.10226825e-01 -8.28683496e-01 5.52557290e-01 9.02444601e-01 -1.16675660e-01 4.39237416e-01 -7.59543002e-01 -5.77560127e-01 -4.70744461e-01 -2.90255714e-02 -1.54700661e+00 -3.27010639e-02 -5.64428985e-01 3.48799050e-01 -1.45329773e+00 3.63077462e-01 -1.77533239e-01 8.77078250e-02 2.79400229e-01 -4.65489835e-01 5.73580861e-01 6.32221671e-03 6.94185197e-01 -2.05043986e-01 4.94232327e-01 1.39218783e+00 -6.39846995e-02 4.16151976e-04 -9.40137953e-02 -8.54191363e-01 5.75383961e-01 3.60852897e-01 -2.30852798e-01 -3.98532599e-01 -9.35538590e-01 3.26658227e-02 -8.63222703e-02 8.52079213e-01 -7.27770805e-01 2.22442821e-01 -9.01831463e-02 5.99737048e-01 -1.18827283e-01 2.54081547e-01 -3.91627729e-01 4.07221794e-01 1.90539241e-01 -3.03095013e-01 -3.30322117e-01 1.27562478e-01 6.05573595e-01 4.28790189e-02 -1.98671579e-01 1.08886576e+00 -2.33394966e-01 -3.11117649e-01 3.54604945e-02 -4.50771868e-01 7.75434300e-02 3.92959237e-01 -1.01754151e-01 -7.29075551e-01 -7.31775343e-01 -9.54894066e-01 -3.87169659e-01 7.99019337e-01 2.01435551e-01 5.30246079e-01 -1.07608426e+00 -8.94369185e-01 2.92602450e-01 -3.67813468e-01 -8.91762301e-02 2.36642241e-01 7.57432163e-01 -5.66695452e-01 1.60742804e-01 -1.52289737e-02 -5.43095112e-01 -7.50409901e-01 4.76700515e-01 6.51517689e-01 9.02931988e-02 -4.06815946e-01 1.08291519e+00 6.43956244e-01 2.17438310e-01 9.00907442e-03 -1.59883410e-01 1.81575194e-01 -2.63782620e-01 5.16106367e-01 3.46577018e-01 -1.28576774e-02 -7.52294481e-01 -2.89887134e-02 3.62836957e-01 -3.59006710e-02 -3.43927026e-01 1.33542204e+00 -5.45423448e-01 -4.00323004e-01 -1.83187395e-01 1.20409322e+00 1.66353494e-01 -1.88187265e+00 -1.20390639e-01 -3.48540902e-01 -4.03530300e-01 4.25043851e-01 -8.88153911e-01 -9.90558267e-01 7.17061341e-01 6.76032305e-01 3.06528568e-01 1.16180217e+00 1.37932941e-01 5.26136696e-01 -1.41187951e-01 -1.20309196e-01 -3.84184957e-01 2.24386409e-01 5.47324300e-01 1.09593368e+00 -9.87058103e-01 -1.83204755e-01 -1.07401602e-01 -4.63885248e-01 1.10758138e+00 1.40478745e-01 -3.21765870e-01 5.25731444e-01 3.27236414e-01 6.85259625e-02 -1.28052965e-01 -2.94444025e-01 -1.39507726e-01 5.55113912e-01 5.80103397e-01 2.55031407e-01 -2.52534270e-01 1.11014903e-01 4.27250654e-01 -2.37407550e-01 7.17922077e-02 7.62324572e-01 4.19294506e-01 -3.43033820e-01 -7.68143237e-01 -8.37066948e-01 1.78116597e-02 -4.66903418e-01 -4.74715769e-01 -2.79569238e-01 8.88240039e-02 -1.11928629e-02 1.01938355e+00 -4.88086790e-03 -1.38518810e-01 -2.73790330e-01 -1.22666163e-02 6.19502306e-01 -3.15494865e-01 -9.59798992e-02 5.21042943e-01 -2.29596585e-01 -2.93837875e-01 -4.06133771e-01 -5.32145619e-01 -7.00428247e-01 -1.61108121e-01 -3.81668448e-01 2.39413162e-03 5.24011075e-01 9.98834372e-01 4.53537822e-01 3.39625448e-01 4.15208310e-01 -1.21571493e+00 -6.38992190e-01 -1.24526715e+00 -4.30401802e-01 4.41651791e-01 9.16262686e-01 -4.00108337e-01 -7.91601777e-01 7.75395334e-01]
[11.675841331481934, -2.6393215656280518]
c5d2a6b8-e7a5-4ffb-8f62-2d2a03bd7475
a-survey-on-dragonfly-algorithm-and-its
2002.12126
null
https://arxiv.org/abs/2002.12126v3
https://arxiv.org/pdf/2002.12126v3.pdf
A survey on dragonfly algorithm and its applications in engineering
The dragonfly algorithm was developed in 2016. It is one of the algorithms used by researchers to optimize an extensive series of uses and applications in various areas. At times, it offers superior performance compared to the most well-known optimization techniques. However, this algorithm faces several difficulties when it is utilized to enhance complex optimization problems. This work addressed the robustness of the method to solve real-world optimization issues, and its deficiency to improve complex optimization problems. This review paper shows a comprehensive investigation of the dragonfly algorithm in the engineering area. First, an overview of the algorithm is discussed. Besides, we also examined the modifications of the algorithm. The merged forms of this algorithm with different techniques and the modifications that have been done to make the algorithm perform better are addressed. Additionally, a survey on applications in the engineering area that used the dragonfly algorithm is offered. The utilized engineering applications are the applications in the field of mechanical engineering problems, electrical engineering problems, optimal parameters, economic load dispatch, and loss reduction. The algorithm is tested and evaluated against particle swarm optimization algorithm and firefly algorithm. To evaluate the ability of the dragonfly algorithm and other participated algorithms a set of traditional benchmarks (TF1-TF23) were utilized. Moreover, to examine the ability of the algorithm to optimize large-scale optimization problems CEC-C2019 benchmarks were utilized. A comparison is made between the algorithm and other metaheuristic techniques to show its ability to enhance various problems.
['Seyedali Mirjalili', 'Polla Fattah', 'Tarik A. Rashid', 'Nebojsa Bacanin', 'Abeer Alsadoon', 'Chnoor M. Rahman']
2020-02-19
null
null
null
null
['electrical-engineering']
['miscellaneous']
[-7.58302957e-02 -5.49951732e-01 1.03203788e-01 9.61429551e-02 4.19416070e-01 -4.82106626e-01 6.11593649e-02 2.11666673e-01 -5.32601476e-01 1.20984805e+00 -5.22045612e-01 -7.70646930e-02 -8.77964914e-01 -1.00685811e+00 -2.42325872e-01 -1.00825226e+00 -2.79018849e-01 5.67201316e-01 5.58415279e-02 -8.39495897e-01 8.85672152e-01 7.53403604e-01 -2.13248992e+00 -3.86494368e-01 1.18672621e+00 8.04704309e-01 6.63601458e-01 5.45721471e-01 -2.80358493e-02 1.49183407e-01 -1.17930365e+00 3.74195166e-02 2.15694353e-01 -2.90217400e-01 -4.59702462e-01 -9.71667320e-02 -7.38704562e-01 4.92032200e-01 4.05833304e-01 8.49266410e-01 1.01503873e+00 7.07062662e-01 3.34102958e-01 -1.54480720e+00 -3.94101381e-01 2.67166048e-01 -6.24239504e-01 6.14176333e-01 5.14414847e-01 7.61869270e-03 5.54337084e-01 -5.42532265e-01 3.59501839e-01 1.08561766e+00 4.93345022e-01 -9.26673412e-03 -6.63570940e-01 -4.98221934e-01 -1.81609467e-01 6.79067612e-01 -1.25792122e+00 3.75777513e-01 7.25599647e-01 2.38313209e-02 1.39497876e+00 9.37297046e-01 9.64914024e-01 -8.89122337e-02 7.45058894e-01 4.39726979e-01 9.73694265e-01 -6.52492940e-01 4.30323869e-01 1.70066893e-01 1.03580311e-01 4.82661009e-01 7.92333484e-01 2.99160689e-01 -8.75763968e-02 -1.48752436e-01 -9.77504998e-02 -2.91362405e-01 -1.86033234e-01 -5.31419814e-02 -6.91292346e-01 9.86112237e-01 2.45644927e-01 7.68958688e-01 -4.19741333e-01 -4.06076849e-01 3.54577750e-01 1.21905103e-01 1.87171221e-01 1.19945884e+00 -4.49281335e-01 -2.23377913e-01 -7.17286766e-01 4.95254725e-01 9.62018311e-01 3.86148989e-01 1.54675558e-01 5.36805809e-01 -1.57997742e-01 7.44295478e-01 2.80264795e-01 5.21705091e-01 5.92561483e-01 -5.28500676e-01 2.44457781e-01 8.76868963e-01 1.73252761e-01 -1.31276238e+00 -6.95911169e-01 -7.08293557e-01 -4.52746183e-01 6.81356907e-01 -2.21256107e-01 -2.90230274e-01 -4.66931492e-01 8.59512448e-01 4.34192389e-01 -2.80293196e-01 -3.48308235e-02 7.01234400e-01 6.15557492e-01 1.10159469e+00 -4.37098056e-01 -6.61946416e-01 1.11434877e+00 -1.24065709e+00 -1.10455990e+00 2.37566695e-01 2.73306161e-01 -1.38199759e+00 7.06465900e-01 5.24320126e-01 -1.20543277e+00 -4.16790187e-01 -1.30368352e+00 8.12603116e-01 -8.41481268e-01 7.56522790e-02 5.70493639e-01 1.25939298e+00 -7.73427486e-01 7.51240253e-01 -5.43020546e-01 -4.03092980e-01 -1.01102971e-01 7.37554908e-01 2.24615410e-01 2.51165211e-01 -1.06504607e+00 1.37328434e+00 7.10905135e-01 2.25921065e-01 4.94141094e-02 -5.31462729e-01 -5.90806305e-01 1.45065993e-01 4.89509732e-01 -4.21680450e-01 6.18636191e-01 -5.90304494e-01 -1.72848618e+00 2.64183462e-01 -7.59273618e-02 -2.31839374e-01 3.59861404e-01 1.58451751e-01 -6.55623496e-01 -1.17244042e-01 -1.03151657e-01 -3.22970718e-01 2.33904466e-01 -9.53388810e-01 -1.04893267e+00 -5.58061153e-02 7.00165182e-02 3.13709497e-01 -1.21903658e-01 1.05117448e-01 7.46254772e-02 -8.20420206e-01 -3.55832934e-01 -9.60240126e-01 -2.37268046e-01 -7.90773451e-01 -2.44650304e-01 -1.35659665e-01 1.21112776e+00 -2.67024755e-01 1.76505768e+00 -1.59796154e+00 1.69825912e-01 7.87266970e-01 -4.95459378e-01 6.77364647e-01 6.65987432e-02 9.80660677e-01 -2.93700173e-02 2.47548252e-01 -2.52661437e-01 1.85322165e-01 7.49192014e-02 4.66404051e-01 3.74059945e-01 5.72681010e-01 -9.59301516e-02 4.35367614e-01 -5.40427685e-01 -2.27821574e-01 5.72489917e-01 3.66909802e-01 -6.69602334e-01 -5.30680688e-03 2.16885924e-01 3.92781645e-02 -6.44719839e-01 9.92026925e-01 9.35538709e-01 2.68943846e-01 -1.60896093e-01 -3.29173207e-01 -6.37510955e-01 -6.49593651e-01 -1.63218570e+00 7.03348517e-01 -4.73545253e-01 4.34449613e-01 2.82414109e-01 -1.47273767e+00 8.16737294e-01 2.06804752e-01 1.08878839e+00 -5.21719217e-01 5.93621194e-01 4.59674388e-01 3.36699039e-01 -7.48340964e-01 7.25763500e-01 1.98678240e-01 3.86359304e-01 3.66365433e-01 -2.19109222e-01 -4.25528139e-01 1.31478882e+00 -4.11219716e-01 6.58920050e-01 -1.62879542e-01 3.51841956e-01 -7.41931617e-01 1.25512600e+00 3.39421690e-01 6.34443998e-01 2.93784201e-01 -1.24166422e-01 -2.31135190e-02 -3.42286266e-02 -6.19981587e-01 -6.92444146e-01 -4.54589128e-01 -3.10703754e-01 9.27930176e-01 5.60070336e-01 -3.81517529e-01 -3.57306153e-01 -9.05559435e-02 2.31459901e-01 1.08078086e+00 -4.92692679e-01 -1.82205975e-01 -7.31032908e-01 -1.71176183e+00 1.22638449e-01 -1.89944282e-01 7.07425296e-01 -1.08761942e+00 -1.23299456e+00 4.52977240e-01 9.37880725e-02 -6.04348183e-01 -7.70506188e-02 5.16336039e-02 -7.68402815e-01 -1.44898248e+00 -5.49982369e-01 -8.49038303e-01 6.75439537e-01 1.91342801e-01 1.15737116e+00 6.36625826e-01 -8.41408670e-01 2.54107952e-01 -6.79345727e-01 -8.79875422e-01 -2.88644165e-01 -4.10209261e-02 2.29643602e-02 -3.32579762e-01 1.18451931e-01 -1.71258971e-01 -2.78388172e-01 8.04188132e-01 -9.12151277e-01 -7.87550986e-01 4.52521831e-01 9.89704907e-01 4.67930883e-01 1.23125970e+00 3.77324522e-01 -6.36312008e-01 1.24819779e+00 -2.84486890e-01 -9.73775983e-01 5.01685917e-01 -9.24359441e-01 -1.11084953e-02 7.46194601e-01 -8.63166302e-02 -9.70578611e-01 -3.93710285e-01 -9.45171863e-02 2.12718174e-01 4.13216352e-01 5.62705934e-01 4.61217202e-02 -8.15344274e-01 3.43753636e-01 1.27218753e-01 2.81565785e-01 -3.54039609e-01 -3.79978299e-01 4.08698410e-01 -1.23939388e-01 -6.80525601e-01 8.12637210e-01 6.60566166e-02 5.98695159e-01 -8.67814243e-01 -8.81477743e-02 -4.05306131e-01 1.94367301e-02 -4.03648108e-01 5.80636203e-01 1.79488793e-01 -1.16771805e+00 4.81520742e-01 -7.82119036e-01 3.90466660e-01 -3.30431670e-01 5.00566661e-01 -3.59489709e-01 8.24034661e-02 -1.51235946e-02 -1.02801096e+00 -5.31014979e-01 -1.60681832e+00 2.49192253e-01 7.63665140e-01 1.90605283e-01 -1.06574070e+00 -1.98254399e-02 1.45671174e-01 9.32463586e-01 6.90504193e-01 7.62484729e-01 -4.93492126e-01 1.03736773e-01 -3.70987743e-01 6.15960538e-01 1.48904353e-01 2.98989892e-01 6.68515086e-01 -2.13066235e-01 -6.09652817e-01 3.66394252e-01 3.26711059e-01 7.66591132e-02 6.50571346e-01 9.85264182e-01 -2.31083199e-01 -6.21065736e-01 5.86211979e-01 1.87736189e+00 1.19743991e+00 6.11339569e-01 1.21837699e+00 -1.94292367e-01 5.54540157e-01 1.10202157e+00 6.85029328e-01 -9.84327719e-02 6.89607739e-01 7.27295816e-01 1.78104099e-02 4.01314229e-01 5.71215212e-01 -1.12437308e-01 8.60906005e-01 -7.20049262e-01 -7.17228293e-01 -6.45307064e-01 4.45224531e-02 -1.48254025e+00 -1.07212937e+00 -1.67919695e-01 1.81309211e+00 1.46380022e-01 1.34619638e-01 1.45408764e-01 8.53709817e-01 1.07543409e+00 -8.38665292e-02 -3.25509042e-01 -1.13920784e+00 -4.61479694e-01 4.66799140e-01 6.83625817e-01 2.85497308e-01 -8.67457986e-01 2.01613873e-01 6.10360479e+00 9.54545915e-01 -1.27128792e+00 -2.28644460e-01 1.87612236e-01 -3.34607065e-01 1.12703003e-01 -3.42459500e-01 -7.03996718e-01 8.81684661e-01 4.90160167e-01 -9.36089456e-01 6.97206378e-01 5.76115429e-01 4.12817359e-01 -5.68912268e-01 -3.08460623e-01 9.39103723e-01 1.66316763e-01 -1.19452047e+00 -1.58321559e-01 -1.25136236e-02 1.05561543e+00 -4.16968048e-01 -1.69731975e-02 2.43826106e-01 -1.85564697e-01 -1.03039193e+00 3.20924222e-01 2.34714121e-01 -1.72671989e-01 -1.30926192e+00 1.41673350e+00 6.18579909e-02 -1.30867302e+00 -7.03016937e-01 -4.41615045e-01 -2.13138416e-01 4.66581404e-01 4.91203368e-01 -4.90756452e-01 1.17429411e+00 9.83245432e-01 2.45820314e-01 -4.27588761e-01 1.77946377e+00 2.69635260e-01 1.68270558e-01 -6.50888085e-01 -8.43113899e-01 3.30920368e-01 -6.79055214e-01 1.13891089e+00 7.09508419e-01 7.96833932e-01 -3.11590452e-02 -3.51456180e-03 5.13430119e-01 7.25200295e-01 6.65960968e-01 -2.72917151e-01 -1.42525688e-01 6.40167594e-01 1.19820380e+00 -8.06012273e-01 -6.95811808e-02 -4.28686254e-02 2.55384237e-01 -5.83611369e-01 1.66346177e-01 -1.12106502e+00 -1.03001535e+00 5.60911775e-01 1.01846106e-01 1.50845915e-01 1.66627187e-02 -3.36596072e-01 -4.25517350e-01 -1.70745194e-01 -8.08782816e-01 5.93598485e-01 -6.49690449e-01 -7.98579395e-01 7.27651060e-01 2.83959359e-01 -1.36789858e+00 -3.62389907e-02 -8.73339891e-01 -1.02999842e+00 8.87907207e-01 -1.48275566e+00 -4.33649570e-01 -2.24353507e-01 2.80254781e-01 7.33219087e-01 -8.65501881e-01 5.13002515e-01 4.63846505e-01 -9.21336234e-01 3.34308654e-01 7.42992043e-01 -8.78297925e-01 1.56703845e-01 -9.73287106e-01 -3.88759285e-01 8.38998258e-01 -7.52937019e-01 4.58031774e-01 1.37119222e+00 -6.12463951e-01 -1.58581853e+00 -5.07564902e-01 4.53404665e-01 3.99035603e-01 3.15519005e-01 5.04729748e-01 -3.46830547e-01 -2.26838872e-01 6.42423987e-01 -4.42389160e-01 6.06042862e-01 -4.91244227e-01 1.10173416e+00 -1.63083628e-01 -1.63329303e+00 2.77850151e-01 3.45970243e-01 6.88747823e-01 -4.40929502e-01 3.37805152e-01 4.60904129e-02 -6.01172447e-01 -8.97343874e-01 6.45027339e-01 3.36093277e-01 -9.97997522e-01 1.16605842e+00 -2.63970226e-01 6.35700207e-03 -5.54300487e-01 -3.09908595e-02 -1.63573754e+00 -3.58559102e-01 -7.60559261e-01 5.38737029e-02 1.12920547e+00 3.29187781e-01 -1.21342111e+00 5.72223663e-01 3.48601937e-01 -2.09747016e-01 -1.14081895e+00 -8.66702914e-01 -1.04842520e+00 -1.87706083e-01 3.36612135e-01 1.01804245e+00 6.47692919e-01 -2.03171805e-01 -1.14472575e-01 -5.67947552e-02 -1.82358827e-02 5.08438289e-01 2.19965383e-01 4.94736701e-01 -1.19060743e+00 -1.49735019e-01 -7.55966842e-01 -4.72390294e-01 4.77151237e-02 -2.50088632e-01 -5.30361176e-01 -3.76445532e-01 -1.76792729e+00 -6.22309923e-01 -3.28819454e-01 -2.75123954e-01 -9.86495018e-02 -2.84521878e-01 2.58700028e-02 4.93014455e-01 3.01206354e-02 1.03304990e-01 4.80117381e-01 1.43446386e+00 -1.70512721e-01 -3.74867767e-01 4.81908798e-01 -4.16584641e-01 5.24978280e-01 1.14339077e+00 -5.15815675e-01 -4.67297852e-01 -1.63493335e-01 3.81538332e-01 -3.90007496e-01 -2.89552063e-01 -1.17563212e+00 2.14630857e-01 -3.34818333e-01 1.73029363e-01 -7.83410311e-01 1.53159231e-01 -1.22594845e+00 4.63304162e-01 1.05890262e+00 3.74989599e-01 1.02564478e+00 4.16113645e-01 1.25792399e-01 -4.57175940e-01 -8.44288170e-01 8.31159472e-01 -7.20771477e-02 -8.03993881e-01 -2.89176852e-01 -4.89568114e-01 -6.42055348e-02 1.87833893e+00 -6.99223638e-01 -2.61307180e-01 7.80342892e-03 -6.11046016e-01 5.50070524e-01 3.03984910e-01 3.57970566e-01 2.39593759e-01 -1.05631697e+00 -5.15690744e-01 5.80029078e-02 -5.61367929e-01 -5.13810098e-01 1.42165735e-01 9.51750100e-01 -1.44841909e+00 5.77420831e-01 -9.01458740e-01 -2.43517935e-01 -1.56509113e+00 5.80620766e-01 4.56789762e-01 -3.61601859e-01 4.82915342e-02 6.30489349e-01 -6.70294762e-01 -3.54081213e-01 -6.17150292e-02 -1.12838298e-01 -7.73408413e-01 1.45715000e-02 2.23616704e-01 1.29988754e+00 3.57703239e-01 -6.94218397e-01 -6.64060235e-01 9.72177625e-01 6.13221288e-01 3.63729119e-01 1.64350855e+00 -1.78261608e-01 -4.81165737e-01 -1.61523387e-01 9.21527743e-01 2.54927248e-01 -2.07479358e-01 6.65046215e-01 -1.82682544e-01 -7.79614568e-01 1.35352276e-02 -1.04611373e+00 -1.35343671e+00 3.40294510e-01 8.23666275e-01 6.63038433e-01 1.57952750e+00 -1.04575408e+00 7.17141926e-01 3.52310658e-01 3.44480067e-01 -1.56984830e+00 -3.73353630e-01 6.01956129e-01 9.71918285e-01 -9.84246790e-01 4.95336741e-01 -5.50693750e-01 -4.48778898e-01 1.47058988e+00 5.57593286e-01 -2.11084604e-01 8.88053715e-01 5.29839575e-01 -3.10471743e-01 3.10570560e-02 -3.36498499e-01 4.51276042e-02 2.76829928e-01 4.62438196e-01 3.87723953e-01 -1.73375547e-01 -1.38810062e+00 4.24029678e-01 -5.05982876e-01 9.63439513e-03 5.45614123e-01 1.49380553e+00 -5.47964096e-01 -1.31990910e+00 -1.00076151e+00 5.36189854e-01 -7.90941358e-01 3.92818958e-01 -3.39423567e-02 1.17630458e+00 3.26994926e-01 1.39282906e+00 -3.43392193e-01 -2.79177964e-01 7.64627457e-01 -5.87138951e-01 2.51247287e-01 -3.26904505e-02 -1.15744734e+00 -3.01398486e-01 6.41852841e-02 -3.69291604e-01 -7.11246490e-01 -4.04405683e-01 -1.15938628e+00 -6.99496627e-01 -6.60807490e-01 1.12065017e+00 1.04511499e+00 8.49283159e-01 1.10343151e-01 1.05007672e+00 6.69320822e-01 -8.44550729e-01 -2.41612941e-01 -5.07017016e-01 -6.75850451e-01 6.43148879e-03 -2.44047895e-01 -1.31667173e+00 -4.58711207e-01 -4.86529201e-01]
[5.646997451782227, 3.496082305908203]
b77228d6-9008-4f59-a347-4e29f29a08a4
vimq-a-vietnamese-medical-question-dataset
2304.14405
null
https://arxiv.org/abs/2304.14405v1
https://arxiv.org/pdf/2304.14405v1.pdf
ViMQ: A Vietnamese Medical Question Dataset for Healthcare Dialogue System Development
Existing medical text datasets usually take the form of ques- tion and answer pairs that support the task of natural language gener- ation, but lacking the composite annotations of the medical terms. In this study, we publish a Vietnamese dataset of medical questions from patients with sentence-level and entity-level annotations for the Intent Classification and Named Entity Recognition tasks. The tag sets for two tasks are in medical domain and can facilitate the development of task- oriented healthcare chatbots with better comprehension of queries from patients. We train baseline models for the two tasks and propose a simple self-supervised training strategy with span-noise modelling that substan- tially improves the performance. Dataset and code will be published at https://github.com/tadeephuy/ViMQ
['Steven Q. H. Truong', 'Trung H. Bui', 'Nguyen Phan', 'Nguyen Phuc Minh', 'Tran Hoang Vu', 'Nguyen Anh Tu', 'Ta Duc Huy']
2023-04-27
null
null
null
null
['named-entity-recognition-ner', 'intent-classification']
['natural-language-processing', 'natural-language-processing']
[ 5.67324497e-02 5.22990227e-01 -2.32313976e-01 -6.90781474e-01 -1.39625669e+00 -4.00524557e-01 9.09785032e-02 6.90792382e-01 -7.56057918e-01 8.19923341e-01 7.14316607e-01 -6.35679841e-01 -4.81992848e-02 -5.63644767e-01 -2.68643945e-01 -1.62011892e-01 6.11438751e-02 7.36322165e-01 6.67442977e-02 -4.54532266e-01 -8.54698122e-02 -4.81479764e-01 -5.65303802e-01 9.77664828e-01 1.23761046e+00 7.42826104e-01 2.07455933e-01 1.02891862e+00 -3.14194471e-01 1.53628254e+00 -6.11538649e-01 -8.22472632e-01 -2.78101861e-01 -5.62022328e-01 -1.40561998e+00 -5.00725508e-01 -5.32748690e-03 -2.05233634e-01 -2.60949492e-01 7.65936136e-01 7.28171349e-01 -1.78077385e-01 3.90870124e-01 -9.94574130e-01 -5.91730058e-01 7.33440042e-01 3.58771890e-01 1.90813884e-01 7.68416405e-01 -8.56753439e-04 1.09166622e+00 -4.68767047e-01 9.92927134e-01 8.48449528e-01 7.76181996e-01 1.03779054e+00 -7.80740321e-01 -5.09263873e-01 -5.34829378e-01 1.41465768e-01 -1.10381365e+00 -4.09069091e-01 4.94297773e-01 -3.68150145e-01 1.12186003e+00 6.48201466e-01 5.59072383e-02 1.41029835e+00 3.76624733e-01 9.33554888e-01 8.74753296e-01 -3.04014027e-01 -4.32323851e-02 4.18973446e-01 4.93333012e-01 9.20235038e-01 -2.58663222e-02 -2.83602446e-01 -2.23584980e-01 -6.45049214e-01 1.33381024e-01 9.01474580e-02 -3.02622229e-01 1.85350493e-01 -1.28311872e+00 1.13658178e+00 4.57286388e-01 6.02483273e-01 -4.33479309e-01 -1.56334639e-01 8.72768819e-01 4.44154441e-01 5.32047749e-01 9.44171965e-01 -9.02094424e-01 -2.43216619e-01 -5.64785302e-01 4.70384434e-02 1.41471457e+00 1.19180238e+00 2.14710638e-01 -7.96070933e-01 -7.64980733e-01 9.39786553e-01 5.55963293e-02 2.67614186e-01 6.60860896e-01 -8.66721988e-01 6.68757737e-01 7.36844659e-01 1.34603932e-01 -8.49933863e-01 -7.50667691e-01 -2.54448831e-01 -7.22603559e-01 -9.05454159e-01 3.50306153e-01 -5.82876563e-01 -6.60004735e-01 1.65041161e+00 4.62845415e-01 -2.76202887e-01 5.30796528e-01 4.96132940e-01 1.73671067e+00 3.91646385e-01 5.47565699e-01 5.87909184e-02 1.98126054e+00 -1.12256718e+00 -1.23930013e+00 -2.31254905e-01 1.58464634e+00 -7.46375203e-01 9.03441787e-01 -3.01417917e-01 -8.73110354e-01 -1.04081966e-01 -3.26474875e-01 -5.40608585e-01 -3.87120903e-01 7.08971098e-02 3.38115245e-01 3.95321816e-01 -8.83212805e-01 1.47927463e-01 -7.87384450e-01 -5.95776618e-01 2.38980472e-01 2.10809857e-02 -2.78969795e-01 -5.82254082e-02 -1.56081510e+00 9.83502328e-01 3.92707527e-01 -1.65452883e-01 -6.12337887e-01 -8.22974920e-01 -1.03911901e+00 -1.99473664e-01 3.97786975e-01 -9.86997545e-01 1.75060201e+00 -7.14321256e-01 -1.12970936e+00 1.22800398e+00 -5.85850067e-02 -5.05206287e-01 4.68471080e-01 -1.04060039e-01 -3.71473730e-01 3.64934236e-01 3.44190419e-01 7.83378005e-01 9.52071697e-02 -6.11036777e-01 -5.69192946e-01 -2.74475396e-01 1.77146271e-01 1.31206512e-01 -2.61113703e-01 1.79986909e-01 -1.36901334e-01 -5.99920928e-01 -4.67978776e-01 -5.96266150e-01 -4.58830416e-01 -2.21849427e-01 -8.18219662e-01 -4.22846526e-01 1.82825983e-01 -1.20735919e+00 1.31771386e+00 -2.05979776e+00 -3.00245941e-01 -4.62086171e-01 4.04204428e-01 1.31298617e-01 -3.67660165e-01 8.62519085e-01 3.44635844e-02 2.92480886e-01 -2.91977078e-01 -3.17965835e-01 -1.72402278e-01 3.32552135e-01 7.04455674e-02 -5.62973954e-02 3.18163902e-01 1.31044281e+00 -1.12616682e+00 -9.09891725e-01 -2.61888713e-01 5.27611375e-02 -6.64324701e-01 8.33421588e-01 -4.53435302e-01 6.36317134e-01 -6.38942003e-01 5.15565515e-01 5.87556064e-02 -5.56296766e-01 1.14263922e-01 -1.50410831e-01 2.54639238e-01 9.48077440e-01 -2.68922329e-01 1.76205122e+00 -4.01967734e-01 1.00874171e-01 3.23405892e-01 -7.35583127e-01 5.56762695e-01 9.00284827e-01 7.22057760e-01 -6.67283595e-01 2.96541631e-01 2.80584425e-01 -1.35807842e-02 -1.34968877e+00 2.96594858e-01 -2.14467451e-01 -4.43491161e-01 2.42622003e-01 1.92821771e-01 3.69811580e-02 3.47470045e-02 4.01073426e-01 1.54007220e+00 -4.59756047e-01 6.33139014e-01 -1.27691805e-01 4.71622974e-01 5.68770170e-01 4.51108515e-01 9.70028818e-01 -2.98778713e-01 1.88972086e-01 3.66212040e-01 -1.74197227e-01 -6.45557344e-01 -4.26200300e-01 -3.16240162e-01 1.23213017e+00 -4.25906539e-01 -6.52095735e-01 -8.10927093e-01 -1.12693179e+00 -2.53463924e-01 1.00222194e+00 -6.30263448e-01 -7.88166225e-02 -3.50155592e-01 -4.47429866e-01 1.00030637e+00 3.57246459e-01 3.33485395e-01 -1.27701128e+00 -7.42966354e-01 4.62204725e-01 -9.24530029e-01 -1.27256668e+00 -6.81126595e-01 3.18070978e-01 -7.26635456e-01 -1.25316584e+00 -7.44399250e-01 -1.01571465e+00 4.49674100e-01 -3.35476190e-01 1.35579824e+00 3.46362740e-01 -3.42528790e-01 5.91286123e-01 -7.85536885e-01 -6.37767553e-01 -8.73374045e-01 4.60682929e-01 -6.66622758e-01 -4.79961336e-01 7.33612835e-01 -4.29254882e-02 -7.52007067e-01 1.54120266e-01 -9.45481002e-01 2.52378911e-01 5.22361517e-01 1.05276775e+00 1.86535358e-01 -8.54729414e-01 6.41741157e-01 -1.33141696e+00 9.67923462e-01 -8.36943805e-01 8.59384835e-02 4.11365956e-01 -1.54579133e-01 2.77080182e-02 3.70147705e-01 -2.36042053e-01 -8.59555900e-01 8.34760163e-03 -9.87029731e-01 2.17844665e-01 -5.34519196e-01 9.28111553e-01 2.55088329e-01 3.27183068e-01 8.98575306e-01 2.06543505e-02 -4.75065298e-02 -6.06431901e-01 4.15706575e-01 1.11910665e+00 3.04950774e-01 -1.49561360e-01 9.64549556e-02 2.22292766e-02 -7.60696888e-01 -6.01659715e-01 -1.16943562e+00 -9.07787442e-01 -1.69688255e-01 1.62186593e-01 1.42750239e+00 -7.10467696e-01 -7.45792866e-01 -1.71579812e-02 -1.33346391e+00 -4.49903399e-01 -3.87370735e-01 3.60323608e-01 -3.83501172e-01 2.34156489e-01 -1.23902607e+00 -6.28518045e-01 -8.05637836e-01 -8.09768796e-01 1.09719849e+00 -1.25710264e-01 -6.75598264e-01 -1.28176284e+00 3.81424665e-01 8.31956387e-01 4.37494874e-01 -1.70055404e-02 1.17013454e+00 -1.58863854e+00 -6.68044314e-02 -4.21346664e-01 1.59506090e-02 2.24521812e-02 2.44206771e-01 -7.47628391e-01 -7.11259067e-01 2.23809313e-02 4.16346580e-01 -6.17046893e-01 5.75995803e-01 1.77444726e-01 1.06115377e+00 -9.12330210e-01 -4.32290316e-01 2.18091786e-01 1.02301788e+00 1.63434491e-01 4.19883102e-01 -9.85761918e-03 4.08777833e-01 9.42380548e-01 4.99754250e-01 3.45998079e-01 1.06784511e+00 4.70617652e-01 4.20480706e-02 -2.18830585e-01 1.23554513e-01 -3.09601545e-01 -2.61121150e-02 1.29240489e+00 4.25276726e-01 -3.94732326e-01 -1.18311858e+00 7.50618577e-01 -1.80402470e+00 -6.95251584e-01 -2.06638083e-01 1.58450067e+00 1.49109471e+00 -3.54029447e-01 -2.73310300e-02 -5.46781957e-01 3.95808637e-01 -1.66262865e-01 -2.61842728e-01 -2.40518942e-01 3.18410695e-01 3.02896183e-02 4.64775383e-01 6.52183771e-01 -1.17617023e+00 7.10821688e-01 5.85732460e+00 6.63737237e-01 -7.54277766e-01 6.98667347e-01 6.54924035e-01 3.35451692e-01 -2.73354203e-01 -3.87689829e-01 -6.45999491e-01 3.82110029e-01 1.36457872e+00 3.68316881e-02 -1.60031900e-01 7.13373184e-01 1.30750060e-01 9.74138975e-02 -1.17246664e+00 7.05083489e-01 1.04002200e-01 -1.27542174e+00 -2.86821872e-01 -3.27209562e-01 2.38225624e-01 3.02875549e-01 -3.26125920e-01 4.00728106e-01 3.91559243e-01 -9.64319348e-01 1.08932801e-01 5.96174538e-01 6.60621345e-01 1.37097895e-01 1.19202888e+00 7.39702761e-01 -6.65083528e-01 -8.81310254e-02 -1.30213716e-03 2.22315326e-01 3.43761265e-01 3.01827133e-01 -1.45429718e+00 7.92363346e-01 5.88044047e-01 5.66492379e-01 -4.78505284e-01 1.01446605e+00 -1.61883980e-01 8.68500710e-01 -1.07593201e-01 -4.76190329e-01 1.98169217e-01 1.79947525e-01 4.93014127e-01 1.41702890e+00 4.52549644e-02 4.93355751e-01 2.36162975e-01 7.26276338e-01 -3.26445431e-01 5.49728096e-01 -5.94382524e-01 -4.19938028e-01 2.92247981e-01 1.28054178e+00 -4.26633984e-01 -5.00249803e-01 -4.99465078e-01 1.01142633e+00 4.64588135e-01 1.27348036e-01 -6.07005596e-01 -5.30786991e-01 2.56721824e-01 -1.33999391e-04 -4.33560610e-02 1.25993356e-01 -2.62939751e-01 -1.22308576e+00 -1.29985780e-01 -1.11431086e+00 1.02622366e+00 -4.88259345e-01 -1.59875977e+00 7.98600256e-01 -2.79981673e-01 -9.18715775e-01 -4.66012180e-01 -5.06472647e-01 -3.46239924e-01 4.78379995e-01 -1.32478833e+00 -1.12096298e+00 -2.19379783e-01 7.62711287e-01 5.06364703e-01 2.03805462e-01 1.38856137e+00 5.47976136e-01 -3.80824506e-01 7.65002251e-01 -1.03809357e-01 7.61202633e-01 1.00047159e+00 -1.08125591e+00 9.92059410e-02 9.37960818e-02 -7.16835782e-02 6.74156725e-01 6.29904449e-01 -9.67351794e-01 -1.03493643e+00 -1.26486576e+00 1.68667352e+00 -9.30571258e-01 6.29776597e-01 -4.85998631e-01 -1.03546631e+00 7.20679998e-01 5.51831663e-01 -5.09225905e-01 1.32001078e+00 1.34002000e-01 -1.09380297e-01 4.00475830e-01 -1.36413407e+00 3.57062340e-01 7.49907613e-01 -8.31445754e-01 -9.72971082e-01 1.05115259e+00 1.14086843e+00 -5.68066120e-01 -1.19496274e+00 3.35865676e-01 5.47432862e-02 -3.00026923e-01 7.62080431e-01 -1.28741705e+00 4.68079180e-01 2.49090329e-01 2.83987205e-02 -9.57159758e-01 2.07763631e-02 -5.14584005e-01 1.97041139e-01 8.39286268e-01 9.20105577e-01 -7.38675237e-01 3.98006260e-01 9.02060449e-01 -2.33873278e-01 -8.80091906e-01 -1.10359848e+00 -3.64071518e-01 1.46238774e-01 -1.52853519e-01 4.17188667e-02 1.21733224e+00 6.32782280e-01 6.48308635e-01 -1.47762075e-01 -9.14008245e-02 6.14859425e-02 -1.20896444e-01 2.47974277e-01 -7.94847667e-01 -2.23765075e-01 6.57033408e-03 -7.14098066e-02 -8.26492190e-01 5.00572287e-02 -1.11966228e+00 2.80172259e-01 -1.87080300e+00 3.75906378e-01 -3.69763285e-01 4.68611233e-02 8.98251057e-01 -4.67739940e-01 -1.89014539e-01 -1.96802244e-01 1.33480400e-01 -9.76470947e-01 4.35741842e-01 9.52170789e-01 -1.33700892e-01 -1.14859350e-01 1.58641577e-01 -7.01064289e-01 4.32156533e-01 7.68986464e-01 -9.90342915e-01 -1.07530698e-01 -5.70266843e-01 2.30141774e-01 5.47035098e-01 2.32844785e-01 -4.27242488e-01 3.97962689e-01 -9.99726541e-03 -2.11154088e-01 -2.81314224e-01 1.59078389e-01 -7.36839652e-01 -1.56513304e-01 9.45080698e-01 -1.10336185e+00 2.04053178e-01 9.45606008e-02 2.63696223e-01 -2.79847622e-01 -4.38647687e-01 4.60615844e-01 -4.65976924e-01 -1.55954704e-01 2.01687694e-01 -5.58515489e-01 6.38120472e-01 8.15990388e-01 4.29083109e-01 -4.80748862e-01 -7.75704682e-01 -8.61859977e-01 6.65251255e-01 -3.58743668e-02 3.74678940e-01 4.87081289e-01 -8.92075837e-01 -1.01813293e+00 -3.78127754e-01 4.02519971e-01 -2.76266783e-01 4.77129340e-01 1.18082333e+00 -5.43602288e-01 8.04464281e-01 9.03804228e-02 -2.01529488e-01 -1.39308012e+00 6.14853501e-01 4.96058822e-01 -7.46450663e-01 -3.91386360e-01 1.02533174e+00 1.07400618e-01 -1.04324460e+00 2.80314475e-01 -4.54645485e-01 -4.03605640e-01 -1.20969534e-01 4.99927014e-01 3.14542539e-02 7.62377232e-02 -2.94391274e-01 -4.02083695e-01 -3.20410043e-01 -1.75022021e-01 5.46735562e-02 1.23723006e+00 -7.86943957e-02 -3.07981789e-01 2.92647570e-01 1.44129157e+00 -2.79112339e-01 -1.11448891e-01 -3.44597042e-01 4.84862149e-01 1.08276151e-01 -3.73928189e-01 -1.26557386e+00 -4.75334167e-01 9.31351364e-01 5.38110077e-01 2.59337395e-01 9.40798998e-01 5.54364324e-01 1.09764123e+00 6.83162451e-01 -3.72446068e-02 -8.93507659e-01 1.13652281e-01 6.80928409e-01 8.53658617e-01 -1.33850229e+00 -6.88248456e-01 -5.04380763e-01 -8.55247140e-01 6.86515868e-01 5.90400755e-01 4.58789617e-01 6.15818143e-01 3.16723287e-01 5.86345971e-01 -6.03622913e-01 -8.78356934e-01 -2.40060270e-01 3.61579001e-01 4.01094615e-01 9.17380929e-01 1.20352633e-01 -7.27912664e-01 1.06478512e+00 -1.82297200e-01 6.15728199e-02 1.52002171e-01 1.16882527e+00 -1.09040305e-01 -1.11879492e+00 1.16697513e-01 7.42995799e-01 -9.35883701e-01 -5.70928991e-01 -5.29019237e-01 3.28524500e-01 -2.77238637e-01 1.22689199e+00 -1.72591344e-01 -2.96376973e-01 4.78993773e-01 3.60923350e-01 -1.86283946e-01 -1.01799095e+00 -1.15898120e+00 -2.32942447e-01 9.13701236e-01 -6.32279575e-01 -3.91065031e-01 -1.96357980e-01 -1.19616246e+00 1.80931464e-01 -3.37871641e-01 7.43805945e-01 2.64633626e-01 9.61055398e-01 7.02841878e-01 5.97919464e-01 2.18724936e-01 2.40307689e-01 -5.61338365e-01 -1.25330544e+00 1.14789426e-01 5.90237141e-01 3.66122037e-01 7.32669979e-02 5.22155352e-02 -8.42160061e-02]
[8.736634254455566, 8.733993530273438]
7858e52a-587e-44f4-a79b-2ca308499888
max-min-diversification-with-fairness
2301.02053
null
https://arxiv.org/abs/2301.02053v1
https://arxiv.org/pdf/2301.02053v1.pdf
Max-Min Diversification with Fairness Constraints: Exact and Approximation Algorithms
Diversity maximization aims to select a diverse and representative subset of items from a large dataset. It is a fundamental optimization task that finds applications in data summarization, feature selection, web search, recommender systems, and elsewhere. However, in a setting where data items are associated with different groups according to sensitive attributes like sex or race, it is possible that algorithmic solutions for this task, if left unchecked, will under- or over-represent some of the groups. Therefore, we are motivated to address the problem of \emph{max-min diversification with fairness constraints}, aiming to select $k$ items to maximize the minimum distance between any pair of selected items while ensuring that the number of items selected from each group falls within predefined lower and upper bounds. In this work, we propose an exact algorithm based on integer linear programming that is suitable for small datasets as well as a $\frac{1-\varepsilon}{5}$-approximation algorithm for any $\varepsilon \in (0, 1)$ that scales to large datasets. Extensive experiments on real-world datasets demonstrate the superior performance of our proposed algorithms over existing ones.
['Francesco Fabbri', 'Jia Li', 'Michael Mathioudakis', 'Yanhao Wang']
2023-01-05
null
null
null
null
['data-summarization']
['miscellaneous']
[ 2.59995222e-01 -3.76170548e-03 -5.60600996e-01 -5.10752201e-01 -4.31841195e-01 -3.87691230e-01 -2.15260684e-01 7.85773098e-01 -5.19960463e-01 9.85821068e-01 2.53214631e-02 -5.24039306e-02 -7.63937473e-01 -9.91710722e-01 -2.19056711e-01 -7.98268735e-01 -3.11123461e-01 6.08894646e-01 -3.10081095e-02 -1.20253958e-01 5.57748318e-01 2.46998802e-01 -1.69313037e+00 -7.82773942e-02 1.42517948e+00 1.14628828e+00 -1.81196444e-02 4.10930738e-02 -1.35860726e-01 2.30011582e-01 -7.78508544e-01 -6.69540584e-01 4.89663124e-01 -5.34379959e-01 -5.87934315e-01 5.12941591e-02 3.27247918e-01 4.95966338e-02 3.59396935e-01 1.12575185e+00 5.37493229e-01 4.56977099e-01 5.73000491e-01 -1.31464779e+00 -2.87301362e-01 7.85808980e-01 -9.03022110e-01 3.16834509e-01 1.98648646e-01 -3.08843195e-01 1.22774184e+00 -4.14320558e-01 3.46167266e-01 8.46346676e-01 2.85315037e-01 3.50012481e-01 -1.26361060e+00 -7.23542929e-01 5.40181756e-01 -8.88455138e-02 -1.56035495e+00 -2.91879356e-01 4.51080739e-01 -1.23590454e-01 4.76543725e-01 8.98530900e-01 5.44059277e-01 2.06575558e-01 4.99158502e-02 5.10026336e-01 6.02339387e-01 -3.06486160e-01 4.73903328e-01 3.81536037e-01 3.71624231e-01 2.09383398e-01 8.97170961e-01 -5.05953014e-01 -6.63798451e-01 -6.65834308e-01 3.13642062e-02 2.80212522e-01 -3.29423487e-01 -3.45072269e-01 -1.00304604e+00 1.04583371e+00 1.39337584e-01 3.17198457e-03 -6.83759987e-01 -2.21909955e-01 2.03964502e-01 2.92028040e-01 4.46556687e-01 5.69448233e-01 -3.41410100e-01 1.68882206e-01 -1.07778096e+00 4.62994635e-01 7.95307934e-01 1.00277686e+00 6.77648067e-01 -1.35597646e-01 -2.60427445e-01 8.99402022e-01 -1.71492957e-02 1.53128505e-01 4.06820655e-01 -7.76637912e-01 6.11283064e-01 1.00722802e+00 2.29497552e-01 -1.20195150e+00 -3.25377107e-01 -6.03752255e-01 -1.06850505e+00 -2.45711640e-01 2.95436263e-01 -3.04275542e-01 -3.04521918e-01 1.85361290e+00 5.41914403e-01 -1.52755424e-01 -1.49336234e-01 8.51213634e-01 4.56356406e-01 5.54315567e-01 1.58060983e-01 -1.11921299e+00 1.08017468e+00 -3.31097394e-01 -4.46980357e-01 1.10099446e-02 1.90207899e-01 -5.77035666e-01 6.72913551e-01 5.56496620e-01 -1.27331948e+00 -1.35152042e-01 -8.35766912e-01 4.78508949e-01 -2.83294194e-03 -1.03484921e-01 6.91965342e-01 9.11041141e-01 -5.96739471e-01 4.35173362e-01 -2.29103699e-01 -2.41731212e-01 5.36772430e-01 7.22650290e-01 5.69554570e-04 9.75607783e-02 -9.19144809e-01 3.26532006e-01 6.58700287e-01 -1.13534488e-01 -9.05533805e-02 -5.97401381e-01 -4.18123543e-01 1.11920640e-01 8.60071421e-01 -7.31234968e-01 5.95840514e-01 -1.01653838e+00 -9.96824861e-01 5.18539667e-01 -1.26052201e-01 -5.35297692e-01 3.69250000e-01 2.16516614e-01 -3.29101294e-01 -1.57524750e-01 6.15170226e-02 2.21565425e-01 5.80666006e-01 -1.04983580e+00 -1.07095325e+00 -7.11844981e-01 1.34636596e-01 3.49324256e-01 -5.58320224e-01 -1.36040784e-02 -2.80168146e-01 -5.98298132e-01 1.15985319e-01 -7.91089416e-01 -6.28062248e-01 -4.50046957e-01 -5.22452772e-01 -2.11981386e-01 2.10671663e-01 -3.63146991e-01 1.74906015e+00 -1.85509288e+00 2.66388267e-01 7.16641068e-01 2.11515892e-02 2.87953854e-01 3.23361009e-01 4.94082004e-01 3.42116982e-01 2.79951543e-01 -2.65799254e-01 -7.98044354e-02 -4.23877835e-02 5.71223311e-02 -4.72264625e-02 4.74965543e-01 -2.79925108e-01 1.72805130e-01 -6.49276435e-01 -3.73369545e-01 -6.15089647e-02 -1.44620836e-02 -6.02746725e-01 8.33788700e-03 -1.68543026e-01 -1.96137987e-02 -6.26706839e-01 7.14055657e-01 9.41749454e-01 2.34225318e-02 3.16816121e-01 5.00740558e-02 -2.51910210e-01 -2.09193587e-01 -1.82105160e+00 1.06958747e+00 -7.97758847e-02 -2.36686692e-01 4.87449169e-02 -1.29630089e+00 9.09527302e-01 -1.09844856e-01 8.89538646e-01 -5.22631288e-01 3.21609557e-01 3.52371037e-01 -1.78611465e-02 -1.56935677e-01 8.49089801e-01 -1.53060332e-01 -4.20510650e-01 5.11962652e-01 -3.89095455e-01 4.15392697e-01 6.51499748e-01 1.63826272e-01 5.97711384e-01 -6.46805406e-01 7.47843742e-01 -4.01686281e-01 6.55024230e-01 -1.00255817e-01 8.98478448e-01 8.81833494e-01 -3.89623456e-02 5.07505298e-01 6.46497011e-01 -3.03978503e-01 -7.41973996e-01 -6.57640100e-01 -1.11015096e-01 1.21395755e+00 4.95177507e-01 -1.98677793e-01 -4.76121962e-01 -6.54461980e-01 2.12889686e-01 9.25372899e-01 -5.22811353e-01 -2.02764243e-01 -2.30364040e-01 -1.17920816e+00 1.45322960e-02 -2.51912680e-02 1.30064562e-01 -7.45968759e-01 -9.30740476e-01 2.81074345e-01 5.59086911e-03 -4.44792151e-01 -6.85707688e-01 2.72714168e-01 -7.43880272e-01 -8.98104846e-01 -7.28296697e-01 -5.28768003e-01 8.85413289e-01 3.83902401e-01 9.13462639e-01 1.34515343e-02 -9.33150873e-02 -1.67788357e-01 -4.92030352e-01 -6.00499392e-01 5.51743358e-02 2.83779472e-01 1.67010158e-01 3.59310687e-01 3.31585050e-01 -3.78195852e-01 -7.16269672e-01 4.84962136e-01 -1.05901372e+00 -2.84639299e-01 2.66807437e-01 6.27767384e-01 9.02991176e-01 4.98032272e-01 1.04353333e+00 -1.16136646e+00 7.88331509e-01 -8.55957568e-01 -5.18774271e-01 5.30188799e-01 -8.92800450e-01 5.89292161e-02 8.47331166e-01 -4.06501740e-01 -7.69271791e-01 -3.29393968e-02 2.01696932e-01 6.27408698e-02 2.70740271e-01 5.94575405e-01 -5.05490243e-01 2.66170651e-01 3.08018059e-01 3.17999750e-01 -1.40512856e-02 -3.58954072e-01 2.18359619e-01 7.75963962e-01 3.16032052e-01 -3.97069514e-01 3.16838443e-01 3.90128613e-01 -7.62705412e-03 -4.98543411e-01 -5.57628036e-01 -5.26379168e-01 -2.27641404e-01 4.20538411e-02 1.20629139e-01 -6.59701347e-01 -1.01926327e+00 -6.65997118e-02 -3.78929496e-01 4.71586287e-01 -4.94552106e-01 3.09365958e-01 -3.17033768e-01 3.59413981e-01 3.66883069e-01 -1.04647827e+00 -7.74310112e-01 -8.64769518e-01 4.40568626e-01 7.79919267e-01 -3.21663648e-01 -4.42892969e-01 -1.80136710e-01 3.31354618e-01 3.48903328e-01 4.36843514e-01 9.57814038e-01 -1.06093562e+00 -3.43155086e-01 -3.83064806e-01 1.64132312e-01 8.61033201e-02 3.37135464e-01 -9.27627459e-02 4.44974825e-02 -4.99062240e-01 -3.18491787e-01 -3.88700888e-02 7.56470978e-01 5.25441170e-01 1.26468897e+00 -8.38282347e-01 -3.74069154e-01 5.02388656e-01 1.38803923e+00 4.36330497e-01 2.89767534e-01 1.95298001e-01 1.66915879e-01 7.08458960e-01 9.95145500e-01 1.27992547e+00 2.76583254e-01 6.28560364e-01 4.86382574e-01 3.14527422e-01 6.26689851e-01 1.12100236e-01 -1.02540292e-01 2.88590819e-01 1.27648503e-01 -7.47016311e-01 -3.91764522e-01 7.25464165e-01 -1.95181870e+00 -1.03999317e+00 -1.43562499e-02 3.02100992e+00 8.09474051e-01 3.18495445e-02 6.73408151e-01 1.97899222e-01 9.45401669e-01 -8.94914344e-02 -8.72524500e-01 -7.21755922e-01 -3.32828462e-01 2.02615902e-01 6.29352748e-01 6.90950379e-02 -8.32469761e-01 3.84653449e-01 5.07536793e+00 9.61805701e-01 -1.00413001e+00 -4.01804000e-01 9.38423753e-01 -7.41703808e-01 -5.38575470e-01 -6.79368377e-02 -8.91387224e-01 8.94003332e-01 7.69760847e-01 -9.93696272e-01 3.61567378e-01 5.27917981e-01 3.28582853e-01 -5.47553003e-01 -7.25900352e-01 8.16102147e-01 1.49808392e-01 -1.04007125e+00 1.75595820e-01 1.34917989e-01 9.75359678e-01 -6.64877236e-01 2.59811550e-01 -6.36314377e-02 1.23136371e-01 -7.13213146e-01 5.99200368e-01 3.48063588e-01 5.28018951e-01 -1.46626604e+00 5.62847316e-01 5.97185314e-01 -9.04585958e-01 -3.71161252e-01 -4.72996354e-01 1.67022839e-01 2.04929620e-01 9.25112009e-01 -5.13192773e-01 9.16703343e-01 6.41721189e-01 1.44295767e-01 -2.19205201e-01 1.36814928e+00 4.70631242e-01 2.12524503e-01 -4.60620463e-01 -2.61138469e-01 -6.17371313e-02 -3.10220540e-01 4.56301332e-01 7.12673068e-01 6.01536572e-01 4.08592314e-01 3.02039981e-01 3.80919337e-01 -2.04396233e-01 8.54415774e-01 -1.48552209e-01 5.31720966e-02 8.40031564e-01 1.01736355e+00 -7.58430660e-01 -1.48727506e-01 -4.56279889e-02 5.69908261e-01 1.47639602e-01 -9.68488585e-03 -7.68150866e-01 -7.21569419e-01 6.92086458e-01 2.66588479e-01 3.38922143e-01 1.69867516e-01 -4.99609172e-01 -8.69187653e-01 1.23515263e-01 -8.23111713e-01 1.04590821e+00 1.31320998e-01 -1.21473217e+00 3.66922408e-01 1.20887809e-01 -1.29163957e+00 -6.01453222e-02 1.48078054e-01 -5.39072931e-01 8.05887878e-01 -1.14260900e+00 -4.37944680e-01 -1.28512651e-01 4.77856487e-01 3.66540849e-01 -2.23040164e-01 4.35952306e-01 2.98204124e-01 -8.29917610e-01 9.85904455e-01 5.53317547e-01 -6.25371933e-01 3.91148329e-01 -8.98680151e-01 -3.77136946e-01 6.87163353e-01 -2.77532786e-01 7.07353413e-01 8.58969927e-01 -5.54880798e-01 -1.18856692e+00 -1.03019202e+00 9.04194415e-01 3.24902385e-01 -7.47688040e-02 2.65572798e-02 -7.16627896e-01 1.78915605e-01 1.03532530e-01 -1.93278268e-01 1.12859046e+00 -2.94555873e-02 1.59228459e-01 -4.53698844e-01 -1.70791221e+00 5.54311812e-01 1.00303531e+00 4.77391511e-01 -4.70830277e-02 6.74238801e-02 3.72451305e-01 -2.13050380e-01 -1.00008476e+00 4.50754374e-01 5.69521189e-01 -1.02150202e+00 8.68680537e-01 -6.33568943e-01 4.93519194e-02 -1.88670427e-01 -2.64897913e-01 -1.26248574e+00 -2.24986091e-01 -8.89273167e-01 -1.05570614e-01 1.38610840e+00 5.16965508e-01 -5.66657662e-01 9.04459178e-01 8.90243828e-01 2.86900491e-01 -9.90484834e-01 -1.08393264e+00 -5.29729903e-01 -6.50919601e-02 1.73169270e-01 8.92458677e-01 7.26277232e-01 4.57469039e-02 1.20339401e-01 -6.47626281e-01 4.26604226e-02 7.81925857e-01 6.27851248e-01 6.21822357e-01 -1.32446980e+00 -1.18618615e-01 -4.72625107e-01 -1.31189808e-01 -5.60474932e-01 -1.84208155e-01 -7.41293490e-01 -2.15047419e-01 -1.37928319e+00 4.48017597e-01 -8.90642822e-01 -5.52563131e-01 2.72582710e-01 -4.25733268e-01 -9.54619721e-02 1.19806491e-01 7.41772279e-02 -8.39861453e-01 4.51275736e-01 6.47923946e-01 -5.27402991e-03 -6.82174265e-01 5.49197197e-01 -1.42294610e+00 3.50180835e-01 7.89588213e-01 -5.46123564e-01 -5.19483924e-01 -6.32534921e-02 2.72541314e-01 2.48097911e-01 -4.31325942e-01 -6.96718931e-01 2.11752191e-01 -8.00045848e-01 1.39608189e-01 -7.38460064e-01 1.50259631e-02 -8.26191664e-01 6.49713516e-01 4.31402147e-01 -4.92657721e-01 -1.19229052e-02 -9.90663171e-02 6.02269769e-01 -2.37966124e-02 -5.02144098e-01 7.84583569e-01 -5.84088825e-02 -3.85559916e-01 3.90771002e-01 -1.26645043e-01 1.73185691e-01 1.45794070e+00 -4.28173929e-01 -3.76316011e-01 -2.65634298e-01 -3.70805800e-01 6.60980821e-01 4.19112802e-01 2.51866132e-01 3.71539056e-01 -1.09591711e+00 -8.90881658e-01 -2.71958169e-02 3.03504586e-01 1.31136477e-01 4.96860206e-01 6.59072638e-01 -2.66447753e-01 3.46487790e-01 -2.67288208e-01 -2.37813577e-01 -1.42996001e+00 5.13173759e-01 -2.77989153e-02 -3.26997966e-01 -7.36610172e-03 9.33062673e-01 -2.54794747e-01 -5.63960858e-02 2.22018674e-01 5.88492155e-02 -3.04157823e-01 5.24111867e-01 5.28646290e-01 8.37921083e-01 5.13654500e-02 -6.29422247e-01 -4.92075890e-01 2.83670247e-01 -3.01788002e-01 1.52110696e-01 1.33351433e+00 -3.52680355e-01 -1.53789788e-01 5.53057045e-02 9.56210077e-01 2.62273133e-01 -8.33826780e-01 -3.58791292e-01 -1.70411412e-02 -8.51001203e-01 -2.25416213e-01 -5.82955360e-01 -1.16622484e+00 5.73101938e-02 3.78436178e-01 5.31768560e-01 1.57889462e+00 -1.79659545e-01 7.26134896e-01 2.52559036e-02 5.46414375e-01 -1.23475420e+00 -2.67548621e-01 2.61868238e-02 5.41403532e-01 -1.01743221e+00 3.89320731e-01 -2.46505961e-01 -9.36873257e-01 7.34656274e-01 6.59108162e-01 6.20551221e-03 4.95947033e-01 -1.76921442e-01 -4.81836796e-01 1.25186563e-01 -7.91651011e-01 -1.69089392e-01 1.93669081e-01 1.58554360e-01 2.12358758e-01 3.35343271e-01 -1.30942523e+00 8.72491062e-01 -1.56351626e-01 -7.59961903e-02 5.54224551e-01 1.08585751e+00 -7.59410024e-01 -1.18240428e+00 -3.98151278e-01 1.10323668e+00 -6.86366379e-01 1.13061249e-01 -3.76493603e-01 3.65946680e-01 5.64065993e-01 1.08405054e+00 9.40777436e-02 -1.75840437e-01 4.95039880e-01 -1.64894015e-01 2.16001615e-01 -4.14364547e-01 -8.00129294e-01 5.33565739e-03 -1.84877012e-02 -5.88411130e-02 -3.20497513e-01 -9.13309216e-01 -1.26519620e+00 -4.32824045e-01 -5.12785316e-01 5.43286979e-01 4.65534031e-01 5.50124824e-01 4.68306661e-01 1.55724883e-01 1.05278933e+00 -2.43655384e-01 -6.46724820e-01 -5.53615153e-01 -9.51809227e-01 2.91388422e-01 7.95893464e-03 -5.22028685e-01 -1.68867454e-01 -3.89098912e-01]
[6.591451168060303, 4.8863606452941895]
590dd3ea-6d30-496e-aea9-a6d64596d6a5
displacing-objects-improving-dynamic-vehicle
2306.17536
null
https://arxiv.org/abs/2306.17536v1
https://arxiv.org/pdf/2306.17536v1.pdf
DisPlacing Objects: Improving Dynamic Vehicle Detection via Visual Place Recognition under Adverse Conditions
Can knowing where you are assist in perceiving objects in your surroundings, especially under adverse weather and lighting conditions? In this work we investigate whether a prior map can be leveraged to aid in the detection of dynamic objects in a scene without the need for a 3D map or pixel-level map-query correspondences. We contribute an algorithm which refines an initial set of candidate object detections and produces a refined subset of highly accurate detections using a prior map. We begin by using visual place recognition (VPR) to retrieve a reference map image for a given query image, then use a binary classification neural network that compares the query and mapping image regions to validate the query detection. Once our classification network is trained, on approximately 1000 query-map image pairs, it is able to improve the performance of vehicle detection when combined with an existing off-the-shelf vehicle detector. We demonstrate our approach using standard datasets across two cities (Oxford and Zurich) under different settings of train-test separation of map-query traverse pairs. We further emphasize the performance gains of our approach against alternative design choices and show that VPR suffices for the task, eliminating the need for precise ground truth localization.
['Michael Milford', 'Ankit Vora', 'Shubham Shrivastava', 'Punarjay Chakravarty', 'Sourav Garg', 'Stephen Hausler']
2023-06-30
null
null
null
null
['visual-place-recognition']
['computer-vision']
[ 1.39787853e-01 -1.55607179e-01 4.89548482e-02 -5.53017616e-01 -1.22131574e+00 -9.16345954e-01 7.84184396e-01 5.14233172e-01 -6.39639318e-01 3.99862319e-01 -2.68483549e-01 -5.40586054e-01 3.42657343e-02 -7.21896231e-01 -9.59835529e-01 -2.57928789e-01 -1.84760720e-01 5.70297062e-01 1.00368750e+00 -2.45202899e-01 2.30356798e-01 9.20359671e-01 -1.82818091e+00 2.01022085e-02 5.84326565e-01 8.21427464e-01 5.54242730e-01 9.27912652e-01 3.01110417e-01 4.06099111e-01 -5.14802277e-01 -5.34996614e-02 5.54121852e-01 2.36840710e-01 -5.63154817e-01 1.43104196e-01 8.87188435e-01 -4.75456238e-01 -2.44589195e-01 8.49957466e-01 2.86105931e-01 4.14300114e-01 6.25993550e-01 -1.38082314e+00 -1.72437012e-01 -2.57821023e-01 -5.57545125e-01 7.02985823e-01 3.92267108e-01 3.28108579e-01 7.98367560e-01 -1.02124608e+00 4.73108232e-01 9.95026946e-01 6.62149847e-01 -1.41661540e-01 -1.40872216e+00 -6.27018929e-01 2.21939892e-01 2.46979043e-01 -1.97931015e+00 -7.98744440e-01 4.35918450e-01 -5.78651369e-01 1.17446125e+00 3.17325741e-01 2.73756176e-01 5.67483842e-01 -1.64770141e-01 4.59435701e-01 7.43855298e-01 -2.65172273e-01 2.36011744e-01 4.81631905e-01 -4.80179191e-02 7.46050537e-01 2.40469545e-01 3.33065301e-01 -3.82809550e-01 -1.11177929e-01 5.09501934e-01 -5.66095337e-02 -6.95276186e-02 -7.38995314e-01 -1.21699488e+00 6.85398638e-01 8.76907885e-01 4.09761481e-02 -4.33679521e-01 1.86378255e-01 -7.89851323e-02 3.53276655e-02 3.63047570e-01 3.19126815e-01 -8.50042477e-02 2.55686879e-01 -1.20020437e+00 2.29425147e-01 3.63762230e-01 9.62665319e-01 1.27886939e+00 -2.15696290e-01 4.81007025e-02 5.46646118e-01 2.87325561e-01 8.36789191e-01 -1.05111144e-01 -9.95348752e-01 5.60989916e-01 4.60305274e-01 6.33405268e-01 -1.27056813e+00 -2.39312306e-01 -2.22338319e-01 -1.59709200e-01 5.16848266e-01 3.78654897e-01 1.61550403e-01 -1.09644699e+00 1.46923888e+00 5.28672814e-01 4.23488319e-01 -2.59194244e-03 8.49293470e-01 5.08970916e-01 6.63816750e-01 2.98475623e-02 4.52088982e-01 1.25373781e+00 -5.96201658e-01 6.47925362e-02 -8.73395622e-01 5.59903264e-01 -5.84575236e-01 7.29626596e-01 -9.60975513e-02 -5.65942228e-01 -6.82400644e-01 -1.33453012e+00 -1.40075022e-02 -8.16672385e-01 2.17852846e-01 9.26370844e-02 6.06248260e-01 -1.34542513e+00 6.11150973e-02 -7.50687420e-01 -7.63282716e-01 2.04487994e-01 3.94671798e-01 -5.67805290e-01 -1.11669652e-01 -8.90656948e-01 1.26088274e+00 3.38954777e-01 3.96490432e-02 -1.12800968e+00 -5.50657630e-01 -9.63561177e-01 -8.45173076e-02 1.56816632e-01 -3.73190373e-01 1.09937775e+00 -6.10903203e-01 -6.00191891e-01 1.05446613e+00 -4.50074941e-01 -5.16490638e-01 4.58098948e-01 1.08350709e-01 -4.68420297e-01 2.31427968e-01 7.03298032e-01 1.15224040e+00 7.40778923e-01 -1.41637099e+00 -1.29801416e+00 -2.54228890e-01 1.40211269e-01 2.45582357e-01 3.93349648e-01 -9.52914730e-02 -6.48796618e-01 -1.45935472e-02 3.02887738e-01 -9.20055866e-01 -2.82427430e-01 1.36488289e-01 -2.44428933e-01 -2.03198027e-02 7.14906871e-01 -4.62469220e-01 6.52774572e-01 -2.41354799e+00 -6.94776893e-01 7.27847457e-01 1.12521641e-01 -4.84226756e-02 -2.30959758e-01 2.87769407e-01 4.82159704e-02 -9.42616835e-02 3.15691344e-02 -2.31713027e-01 -2.52996273e-02 1.65942997e-01 -5.42060077e-01 8.61558139e-01 3.71104538e-01 7.89296091e-01 -9.46736693e-01 -4.36624169e-01 5.65250933e-01 3.58688235e-01 -2.08561450e-01 1.86714006e-03 8.36455747e-02 1.94395363e-01 -3.86855692e-01 6.66082084e-01 6.70264900e-01 -1.07642509e-01 -2.84865230e-01 1.10750534e-01 -4.23669785e-01 2.74603635e-01 -1.41783559e+00 1.18779540e+00 -3.10952246e-01 1.13285482e+00 1.67463556e-01 -7.78470337e-01 1.13314664e+00 -1.21272624e-01 9.81467143e-02 -1.00747883e+00 -2.31384635e-01 3.28781962e-01 -3.63172561e-01 -2.20083266e-01 7.69057333e-01 2.41518855e-01 -1.05318703e-01 1.37427986e-01 -2.78688699e-01 -1.07406070e-02 1.03999227e-01 1.44023389e-01 1.22167110e+00 -2.74611503e-01 8.43947753e-02 -2.97139674e-01 3.42444032e-01 5.19261420e-01 5.72951473e-02 1.04464388e+00 -3.88772011e-01 7.27881193e-01 -6.84995623e-03 -5.53658485e-01 -1.13377404e+00 -1.21687734e+00 -2.43391022e-01 1.24156630e+00 6.22731149e-01 1.18992254e-01 -1.54544830e-01 -5.85375130e-01 2.91175485e-01 7.71693468e-01 -6.44951046e-01 2.00574666e-01 -5.12588561e-01 -5.78583419e-01 4.82969731e-01 5.82806528e-01 5.41683018e-01 -5.56788623e-01 -1.08773983e+00 4.16587256e-02 -1.00385472e-01 -1.12750351e+00 -2.68624216e-01 3.46859008e-01 -2.93637455e-01 -9.93842006e-01 -3.74230176e-01 -8.40543807e-01 8.69976759e-01 9.17652428e-01 1.09736991e+00 -3.41859721e-02 -2.08507806e-01 5.18776715e-01 -1.08184479e-01 -3.62126380e-01 -2.61969626e-01 -4.30098735e-02 -1.57690402e-02 -4.03015874e-02 4.04451013e-01 -3.30728024e-01 -6.89454675e-01 5.56071639e-01 -3.10860842e-01 -7.53635615e-02 6.42290890e-01 3.91012043e-01 6.32118165e-01 2.07102802e-02 1.25503734e-01 -2.49345183e-01 1.13656603e-01 -5.57314575e-01 -1.11451149e+00 2.04422504e-01 -4.20717418e-01 8.85885209e-02 -1.50316790e-01 -1.44519582e-01 -5.66180706e-01 6.05579138e-01 1.38703421e-01 -3.26829404e-01 -3.36086541e-01 1.78730354e-01 1.92948505e-01 -3.06545854e-01 1.07038128e+00 1.87773049e-01 -2.68839419e-01 -2.22419962e-01 4.08069432e-01 5.61130941e-01 9.90189314e-01 -2.53339022e-01 1.14999187e+00 8.12321007e-01 -5.67818582e-02 -6.86530948e-01 -4.76874977e-01 -1.06578004e+00 -7.82558560e-01 -2.31620058e-01 9.00629878e-01 -1.21158791e+00 -4.50050473e-01 -2.96204835e-01 -1.10922801e+00 -2.70013511e-01 5.22314347e-02 4.56754923e-01 -3.87961835e-01 -1.38208002e-01 1.37101442e-01 -8.82302701e-01 1.34611592e-01 -1.12464583e+00 1.48601794e+00 1.20459437e-01 2.64037214e-02 -6.49342000e-01 1.40940815e-01 4.60439026e-02 3.22786540e-01 1.95801198e-01 2.86765158e-01 -7.18907833e-01 -1.07761443e+00 -5.33912122e-01 -5.69410443e-01 -1.97312310e-01 -1.13941178e-01 -2.12875575e-01 -1.23193181e+00 -3.31567973e-01 -5.20559013e-01 1.11081779e-01 9.31366324e-01 2.41903797e-01 5.26612461e-01 3.97291929e-02 -8.30420971e-01 4.09369409e-01 1.50532818e+00 1.94403291e-01 4.34648842e-01 7.58756936e-01 4.55551922e-01 5.30480146e-01 8.18944216e-01 9.77413030e-04 6.31588340e-01 8.35842252e-01 5.26534617e-01 -5.01495123e-01 1.56539064e-02 -3.98527503e-01 1.43343687e-01 -3.66682738e-01 2.15796307e-01 -1.73485652e-01 -1.34170473e+00 1.01209748e+00 -1.89788520e+00 -8.62509966e-01 5.30486219e-02 2.43604589e+00 2.08791301e-01 2.61568725e-01 1.52909294e-01 -2.31322601e-01 8.04306805e-01 9.65473801e-02 -4.33004111e-01 4.29931842e-02 2.41440937e-01 -1.60964996e-01 1.31418180e+00 8.06299269e-01 -1.46244431e+00 8.46696377e-01 6.74651909e+00 3.87541443e-01 -1.06138456e+00 5.03729619e-02 5.49811959e-01 4.73602787e-02 4.39962186e-02 -4.87359427e-02 -9.70681310e-01 1.15066752e-01 9.64473665e-01 9.29346010e-02 3.21165144e-01 1.06160033e+00 3.07474673e-01 -6.20442033e-01 -1.14376152e+00 1.05189538e+00 1.59168877e-02 -1.32996297e+00 -5.36746085e-01 1.03032412e-02 5.45283616e-01 6.42660558e-01 8.13136548e-02 3.97427112e-01 5.80347717e-01 -1.00188065e+00 9.60576415e-01 3.33121032e-01 6.27037346e-01 -5.77682018e-01 5.16855896e-01 4.34585989e-01 -1.47146297e+00 -8.30183830e-03 -3.12911272e-01 1.15328014e-01 -2.74353661e-02 -4.86941785e-02 -1.65309834e+00 2.86514282e-01 7.90819049e-01 2.63922900e-01 -1.14098251e+00 1.41951776e+00 -1.02110267e-01 2.47008413e-01 -7.88995922e-01 1.14283383e-01 5.12671232e-01 2.20573708e-01 5.57376266e-01 1.28432810e+00 2.73460478e-01 -1.52099967e-01 5.20417273e-01 8.95001948e-01 4.03285891e-01 -2.63351172e-01 -9.00777340e-01 6.64452374e-01 8.62807453e-01 1.17968023e+00 -1.05901492e+00 -4.36479211e-01 -1.96875706e-01 7.21892178e-01 3.36502016e-01 5.40246189e-01 -8.23127031e-01 -3.32167774e-01 6.12051070e-01 4.57013488e-01 6.82875574e-01 -4.44735557e-01 -6.84468001e-02 -6.55678272e-01 7.76376128e-02 -3.88641685e-01 1.61547914e-01 -1.04304481e+00 -6.05760694e-01 3.33849937e-01 3.35094869e-01 -1.23872685e+00 -3.75348687e-01 -4.53263730e-01 -5.20054996e-01 9.94468629e-01 -1.72849333e+00 -1.09814227e+00 -5.49932778e-01 3.94254148e-01 2.99754053e-01 1.42229497e-01 5.08319676e-01 3.59103262e-01 -2.53459036e-01 2.42462575e-01 1.31073281e-01 1.99438915e-01 5.80254793e-01 -1.27751577e+00 7.09272921e-01 1.20496976e+00 4.98029768e-01 3.87521625e-01 7.89310873e-01 -5.51113665e-01 -1.06624329e+00 -1.37660158e+00 8.39407146e-01 -1.09943867e+00 4.84276205e-01 -7.14644611e-01 -6.45038784e-01 8.22483301e-01 -2.39617258e-01 1.95767745e-01 -4.10106994e-04 2.77999341e-02 -3.93611550e-01 -3.16018194e-01 -1.06363773e+00 4.98236656e-01 7.22879767e-01 -8.44866693e-01 -5.76694250e-01 3.03536296e-01 5.20013690e-01 -5.58947504e-01 -2.03217775e-01 3.42655659e-01 4.20890987e-01 -8.58166635e-01 1.27698112e+00 -1.03279419e-01 -4.69551593e-01 -1.02700901e+00 -3.96443397e-01 -9.45215046e-01 -3.10297966e-01 -7.28432387e-02 5.22376716e-01 9.76779044e-01 6.19171500e-01 -5.52142680e-01 7.35240281e-01 7.72066295e-01 -6.81595877e-02 -4.41047288e-02 -1.14021540e+00 -8.66194129e-01 -5.19094110e-01 -7.83139527e-01 5.30662715e-01 5.00779808e-01 -4.90664184e-01 1.84562266e-01 -4.32524160e-02 1.02887619e+00 4.10728037e-01 1.00451834e-01 1.09472549e+00 -9.98880923e-01 9.76759866e-02 -2.34354481e-01 -9.38877761e-01 -1.06059110e+00 -7.60853291e-02 -6.83746338e-01 5.28240561e-01 -1.67899907e+00 -3.13957594e-02 -5.88165164e-01 -1.86055094e-01 5.96783459e-01 1.62335243e-02 5.71533203e-01 4.24073189e-02 4.15565610e-01 -9.52698827e-01 4.29465175e-02 4.24819976e-01 -2.89994329e-01 -2.78996646e-01 -3.47717032e-02 -4.02355433e-01 3.71577680e-01 5.74365735e-01 -5.31444728e-01 -3.88678253e-01 -4.91634637e-01 1.36726439e-01 -3.08081120e-01 1.03825378e+00 -1.25002515e+00 4.60275739e-01 1.55902188e-02 6.80874527e-01 -1.01937580e+00 4.29724932e-01 -9.62473750e-01 1.84351936e-01 3.24556619e-01 -1.13287650e-01 2.87927836e-01 4.14739579e-01 7.82830834e-01 3.44271213e-02 -6.93282709e-02 6.92236125e-01 3.91214751e-02 -1.38131177e+00 4.47881445e-02 -3.96345168e-01 -2.94997275e-01 1.17054749e+00 -4.60445523e-01 -3.15762341e-01 -2.95320004e-01 -5.95171750e-01 4.12480265e-01 6.11130774e-01 5.28445959e-01 4.83474731e-01 -1.20980060e+00 -6.21819019e-01 3.92086178e-01 4.56744671e-01 -1.14063807e-01 -2.73466617e-01 7.21590221e-01 -6.20632708e-01 6.77828252e-01 6.64673969e-02 -1.02765930e+00 -1.22280228e+00 6.50256336e-01 5.51026762e-01 2.87852228e-01 -5.53938687e-01 6.29507899e-01 3.50378484e-01 -2.36109570e-01 1.93028927e-01 -4.70439941e-01 3.48667800e-02 -9.50930193e-02 5.82318723e-01 2.48894155e-01 1.69765472e-01 -9.63968813e-01 -7.72483110e-01 4.09686446e-01 2.02051014e-01 -4.56393301e-01 9.34606731e-01 -4.57759440e-01 4.39286947e-01 2.88242280e-01 1.12157094e+00 -9.21656471e-03 -1.39595103e+00 -1.01105027e-01 1.68884411e-01 -6.30782843e-01 2.29880288e-01 -4.73639101e-01 -4.62678105e-01 5.19316912e-01 1.13253391e+00 2.76258200e-01 6.75325692e-01 4.03671712e-01 2.36677095e-01 7.20836818e-01 5.32399118e-01 -8.05245042e-01 -1.67344436e-01 3.50945413e-01 5.70053458e-01 -1.63485587e+00 -5.06811142e-02 -1.70710653e-01 -4.21184301e-01 8.11101437e-01 3.94644886e-01 -2.47317687e-01 5.05371571e-01 1.64216876e-01 8.83592665e-02 -4.88880903e-01 -3.64383727e-01 -9.26427841e-01 4.69623983e-01 8.80803704e-01 -3.32324117e-01 3.97196338e-02 6.52952433e-01 -2.97275782e-01 -2.18578339e-01 -4.07720447e-01 9.73890871e-02 9.49238598e-01 -1.00678647e+00 -4.77770120e-01 -6.81471646e-01 1.96572319e-01 1.15465902e-01 -7.29096159e-02 -2.30749339e-01 1.02377200e+00 2.56494194e-01 1.01390576e+00 5.41681230e-01 -4.15618181e-01 4.28642631e-01 -6.15273751e-02 1.33989453e-01 -5.58477759e-01 -7.00151622e-02 -9.91386920e-02 2.54413277e-01 -6.68249130e-01 -3.68572831e-01 -8.25344503e-01 -1.02794278e+00 -9.64903682e-02 -2.09339783e-01 -4.76321615e-02 8.04999709e-01 9.05616403e-01 4.68610436e-01 7.79731423e-02 6.43254697e-01 -1.16696966e+00 -1.21866621e-01 -4.67801273e-01 -2.74493158e-01 7.33317360e-02 8.22682083e-01 -8.86407852e-01 -2.24993825e-01 1.03264600e-01]
[7.633791446685791, -1.9304096698760986]
c9a35cda-a2c3-4ec2-a62e-ad776ebfd218
towards-unified-dialogue-system-evaluation-a
2006.06110
null
https://arxiv.org/abs/2006.06110v1
https://arxiv.org/pdf/2006.06110v1.pdf
Towards Unified Dialogue System Evaluation: A Comprehensive Analysis of Current Evaluation Protocols
As conversational AI-based dialogue management has increasingly become a trending topic, the need for a standardized and reliable evaluation procedure grows even more pressing. The current state of affairs suggests various evaluation protocols to assess chat-oriented dialogue management systems, rendering it difficult to conduct fair comparative studies across different approaches and gain an insightful understanding of their values. To foster this research, a more robust evaluation protocol must be set in place. This paper presents a comprehensive synthesis of both automated and human evaluation methods on dialogue systems, identifying their shortcomings while accumulating evidence towards the most effective evaluation dimensions. A total of 20 papers from the last two years are surveyed to analyze three types of evaluation protocols: automated, static, and interactive. Finally, the evaluation dimensions used in these papers are compared against our expert evaluation on the system-user dialogue data collected from the Alexa Prize 2020.
['Sarah E. Finch', 'Jinho D. Choi']
2020-06-10
null
https://aclanthology.org/2020.sigdial-1.29
https://aclanthology.org/2020.sigdial-1.29.pdf
sigdial-acl-2020-7
['dialogue-management']
['natural-language-processing']
[-7.22772479e-02 3.65794927e-01 -3.27308252e-02 -5.09948552e-01 -6.84203744e-01 -8.88979852e-01 1.13870156e+00 1.71442047e-01 -5.06758213e-01 1.05221868e+00 4.68714952e-01 -4.75597262e-01 -1.80020168e-01 -3.28247994e-01 4.57987458e-01 -2.90035695e-01 5.51292971e-02 8.22684705e-01 2.37291664e-01 -8.30795348e-01 8.48135412e-01 1.91849113e-01 -1.42722678e+00 3.06970149e-01 8.08894515e-01 4.16874796e-01 7.06488341e-02 1.09909058e+00 -5.07575631e-01 1.16516840e+00 -1.36180413e+00 -6.03081763e-01 -1.72459602e-01 -8.67301166e-01 -1.59309864e+00 1.42447278e-01 -4.28327024e-02 -5.48183084e-01 5.26271313e-02 7.40276992e-01 8.63388419e-01 1.74613893e-01 4.97417420e-01 -1.36885262e+00 -2.22449154e-01 6.36834085e-01 3.62508357e-01 1.60289809e-01 1.04274178e+00 4.61652458e-01 1.06138206e+00 -2.97423422e-01 7.97829986e-01 1.28275919e+00 5.06861389e-01 8.44491482e-01 -9.86000240e-01 -1.66787170e-02 -2.13219807e-01 2.10444435e-01 -8.55935276e-01 -6.72566533e-01 5.61575770e-01 -5.39577901e-01 1.12103987e+00 6.93925202e-01 7.57187128e-01 9.81718063e-01 -1.55829817e-01 5.87463796e-01 1.44587362e+00 -8.23976099e-01 2.96403915e-01 8.31170797e-01 3.43403399e-01 3.39741468e-01 -1.59900829e-01 -5.04991055e-01 -4.87583041e-01 -5.00372708e-01 6.43798470e-01 -9.28941727e-01 -1.88464180e-01 -1.33906886e-01 -8.60717535e-01 8.07260454e-01 -4.42177683e-01 8.98087442e-01 -2.24423662e-01 -6.04604185e-01 8.98011804e-01 6.68122053e-01 3.13279182e-01 1.10826528e+00 -2.74512768e-01 -1.22377896e+00 -3.64432007e-01 6.16760492e-01 1.89074004e+00 6.62819505e-01 2.81411499e-01 -1.95737973e-01 -1.77378416e-01 1.14911866e+00 3.49429935e-01 -1.32139668e-01 3.04645538e-01 -1.48425627e+00 3.58229399e-01 7.87719727e-01 4.66591716e-01 -9.08167183e-01 -3.87032300e-01 5.26750505e-01 -3.06692451e-01 3.26924235e-01 6.49699509e-01 -5.15040874e-01 2.01819301e-01 1.29360962e+00 1.12417407e-01 -9.51849401e-01 2.69976377e-01 8.26376200e-01 1.07535803e+00 5.14836788e-01 2.91024353e-02 -4.47340697e-01 1.32197559e+00 -9.92323279e-01 -1.15080428e+00 3.53341281e-01 9.20235991e-01 -9.93335247e-01 1.19639182e+00 4.15187091e-01 -1.34697270e+00 -1.16771325e-01 -9.01014328e-01 1.92399081e-02 -3.41391623e-01 -2.32731417e-01 6.58381343e-01 1.12206018e+00 -1.19110548e+00 2.98936665e-01 -4.14718539e-01 -7.46511877e-01 -6.30536854e-01 1.96296960e-01 -9.64530930e-02 5.12760043e-01 -1.39045048e+00 1.64892054e+00 1.55371591e-01 4.79663648e-02 -4.45021719e-01 1.67111754e-02 -4.48557526e-01 -3.41298163e-01 2.19223037e-01 -2.21196577e-01 1.97260785e+00 -7.98663080e-01 -2.19753838e+00 8.99629414e-01 1.30555466e-01 -2.53293425e-01 7.17337847e-01 -8.71415660e-02 -2.25469217e-01 6.83989227e-02 -1.43120974e-01 1.12484045e-01 -6.64796382e-02 -1.25581181e+00 -8.22073638e-01 -5.43368123e-02 6.44454360e-01 8.74412119e-01 -3.26070577e-01 6.21592522e-01 -2.18511119e-01 5.54201677e-02 -5.84056556e-01 -7.13988304e-01 -1.00038476e-01 -5.83074152e-01 -2.34399334e-01 -6.91353500e-01 6.19876444e-01 -6.42210841e-01 1.65732408e+00 -1.51335251e+00 2.33302325e-01 -1.28635973e-01 4.12270546e-01 4.46365178e-01 1.96497604e-01 1.18313444e+00 5.49111366e-01 2.46535733e-01 -1.21759608e-01 -2.79687494e-01 4.42708939e-01 1.34607553e-01 1.41217753e-01 8.22146982e-02 -9.61473584e-02 4.50908363e-01 -9.14712727e-01 -6.68395817e-01 5.55028558e-01 2.60179192e-01 -3.13152760e-01 5.98177969e-01 -2.23303363e-01 4.58738297e-01 -5.93036890e-01 4.39452499e-01 5.94685320e-03 -8.88963193e-02 5.91843247e-01 3.41270775e-01 -6.31120145e-01 7.31052101e-01 -8.83711696e-01 1.35297704e+00 -3.83724868e-01 9.17682528e-01 3.52960914e-01 -6.53615296e-01 9.24232066e-01 9.59605932e-01 4.46434945e-01 -5.89204431e-01 3.05827469e-01 2.08316743e-01 4.24050868e-01 -8.57543349e-01 9.79548991e-01 9.20796245e-02 -2.50571072e-01 8.20791483e-01 6.39878064e-02 -3.45603377e-01 4.06835526e-01 2.27255151e-01 1.04074776e+00 -2.51475126e-01 4.12946671e-01 -2.91932583e-01 9.29914713e-01 3.72090667e-01 1.24348983e-01 7.49088347e-01 -7.26612806e-01 8.08360055e-03 8.06457162e-01 -3.53030711e-01 -9.36796248e-01 -4.84403908e-01 -1.43227085e-01 1.25541425e+00 -4.93627191e-02 -7.34537601e-01 -1.31558037e+00 -5.85981548e-01 -6.00820720e-01 6.90111220e-01 -2.94780612e-01 4.24244016e-01 -4.60594624e-01 -6.69409811e-01 8.46183360e-01 -1.14070699e-01 7.34953284e-01 -1.36310971e+00 -8.04839492e-01 3.62521261e-01 -6.11276746e-01 -7.57519662e-01 6.34712651e-02 -2.34270364e-01 -5.45504630e-01 -1.16059089e+00 -5.57732284e-01 -5.65074265e-01 9.44536254e-02 2.25229487e-02 1.28650117e+00 3.88564765e-01 8.02100152e-02 8.87978196e-01 -6.31893277e-01 -3.21628749e-01 -1.06843245e+00 4.54458535e-01 -1.37520805e-01 -6.43098652e-01 5.54117680e-01 -2.04722434e-01 -4.00680423e-01 5.89037597e-01 -4.69253451e-01 -1.06697194e-01 2.68422842e-01 8.60934198e-01 -5.77197433e-01 -2.72560328e-01 8.89434218e-01 -9.34086859e-01 1.66888738e+00 -3.12925845e-01 -2.13565782e-01 3.81551802e-01 -8.63295436e-01 -2.94248074e-01 2.98773259e-01 5.56915551e-02 -1.37766504e+00 -4.98182535e-01 -3.64291400e-01 7.05953121e-01 -4.31172431e-01 5.08027792e-01 1.07264435e-02 -1.63326487e-01 8.48684311e-01 -1.41074300e-01 4.16547924e-01 -3.47026259e-01 3.11214894e-01 1.19466019e+00 2.04354629e-01 -8.09276342e-01 2.01729134e-01 -2.41742402e-01 -8.25442493e-01 -1.37240005e+00 -1.07663833e-01 -5.67908466e-01 -5.79159558e-01 -9.02676225e-01 8.01615059e-01 -2.51604766e-01 -9.88330007e-01 4.62387860e-01 -1.34910464e+00 -5.61506569e-01 6.49084672e-02 3.12629223e-01 -7.40702987e-01 6.26828492e-01 -9.69318271e-01 -1.39379406e+00 -4.54690129e-01 -1.32863653e+00 4.65795398e-01 2.64469653e-01 -1.15193546e+00 -1.29260743e+00 5.31266153e-01 7.89650023e-01 6.38461113e-01 -8.45711678e-02 7.13719070e-01 -1.00471556e+00 -6.98693693e-02 -3.06411922e-01 6.64040595e-02 3.72459948e-01 1.54493555e-01 4.20490742e-01 -9.39596057e-01 3.72368544e-02 2.65655607e-01 -6.95071638e-01 -3.60237509e-01 -1.28194287e-01 2.18602329e-01 -4.50530469e-01 1.28073633e-01 -6.44997537e-01 7.99042284e-01 5.91314018e-01 5.80021799e-01 6.42401576e-01 2.96987966e-02 1.26311278e+00 7.86893725e-01 5.14798105e-01 6.35993481e-01 8.00082684e-01 -1.37100205e-01 3.45241427e-01 2.65205085e-01 2.56121933e-01 3.30978423e-01 1.32037413e+00 -4.65638697e-01 -2.56818384e-01 -1.06391263e+00 2.48816386e-01 -1.83903515e+00 -1.11316562e+00 -2.63411719e-02 2.11991787e+00 1.02501619e+00 3.40714343e-02 5.65528035e-01 2.49655232e-01 7.45459199e-01 1.93668306e-01 5.28755970e-02 -9.11790490e-01 2.64890213e-02 -1.40158355e-01 -3.25126559e-01 8.23455989e-01 -7.03491032e-01 6.92829549e-01 7.38139725e+00 2.83393294e-01 -7.37193167e-01 2.14072131e-02 5.28375626e-01 3.17677736e-01 -5.50157055e-02 5.03222272e-02 -3.29155266e-01 2.30652913e-01 1.25748730e+00 -6.43329322e-01 3.23302448e-01 6.88005924e-01 3.48804146e-01 -4.31319267e-01 -9.14349258e-01 6.15379751e-01 -3.70579921e-02 -1.18349278e+00 -3.69307935e-01 6.22863322e-02 1.99099839e-01 -2.65706599e-01 -3.95783186e-01 5.55045068e-01 5.43169260e-01 -8.97297680e-01 2.53785878e-01 4.20758396e-01 2.68663377e-01 -4.57786113e-01 9.96576250e-01 4.39661473e-01 -5.66720724e-01 2.66511530e-01 6.78012893e-02 -5.98225772e-01 3.88621807e-01 -2.99671888e-01 -1.12358248e+00 3.99546057e-01 3.84289354e-01 -1.87833998e-02 -2.44046375e-01 7.99746633e-01 7.35294670e-02 6.17980003e-01 -4.26093489e-02 -7.25117087e-01 2.74923623e-01 -5.00900805e-01 7.57365942e-01 1.34674704e+00 -4.50974733e-01 3.25275719e-01 1.59264207e-01 3.74481678e-01 3.75767767e-01 5.36711097e-01 -7.17901468e-01 -3.11964661e-01 6.66628540e-01 1.39075410e+00 -6.09401762e-01 -3.29463303e-01 -3.70784193e-01 7.08195508e-01 7.47332051e-02 -3.35427336e-02 -4.10915762e-01 -3.80225509e-01 5.10674834e-01 -2.89397359e-01 -7.03090012e-01 -2.55813330e-01 -4.27199692e-01 -7.56193995e-01 -8.96568894e-02 -1.42472577e+00 3.10911387e-01 -2.90944099e-01 -1.17119575e+00 8.81257772e-01 2.45564669e-01 -8.49759340e-01 -8.30793619e-01 -3.52363706e-01 -6.45872593e-01 9.84023690e-01 -6.69739246e-01 -4.20674413e-01 -2.53689080e-01 7.10561201e-02 6.61646605e-01 -2.53051609e-01 1.46341312e+00 2.52047718e-01 -6.66725934e-01 4.09427106e-01 -1.66598797e-01 9.78506655e-02 8.92897010e-01 -1.39326429e+00 7.44839236e-02 2.16464829e-02 -4.93031681e-01 7.29154587e-01 9.79437113e-01 -3.77794027e-01 -1.28947926e+00 5.63834272e-02 9.88111556e-01 -8.36405337e-01 7.37509370e-01 -7.02252463e-02 -8.72261226e-01 2.06106976e-01 1.04310679e+00 -1.34072125e+00 8.62293482e-01 3.57442081e-01 1.50882915e-01 4.06384617e-01 -1.20731401e+00 9.08036709e-01 5.22968888e-01 -6.64771438e-01 -8.73315632e-01 3.50108325e-01 2.33669937e-01 -3.82723361e-01 -1.24939418e+00 1.55059859e-01 7.03187585e-01 -1.28699887e+00 4.10624415e-01 -4.83745962e-01 3.40115905e-01 9.74249989e-02 5.34316264e-02 -1.11762059e+00 2.84742117e-01 -1.17767203e+00 2.22755894e-01 1.40565038e+00 4.89663661e-01 -7.74953842e-01 7.44535446e-01 1.32364368e+00 -1.41081497e-01 -7.04498231e-01 -6.59440339e-01 -3.51262093e-01 1.14722185e-01 -1.50714487e-01 2.16724217e-01 1.16564810e+00 1.00966895e+00 8.78150880e-01 -2.04507232e-01 -5.12589633e-01 1.12520397e-01 -3.95882100e-01 1.21658885e+00 -1.25575757e+00 9.14515108e-02 -9.98662829e-01 -3.85488451e-01 -8.50247204e-01 1.68981086e-02 -2.87356347e-01 1.11641601e-01 -1.76841176e+00 -7.51480758e-02 -1.82764858e-01 3.72529685e-01 5.86286522e-02 5.54777458e-02 -2.48913988e-01 2.51554877e-01 2.76387304e-01 -7.27285028e-01 5.17093897e-01 1.06887519e+00 1.99420884e-01 -4.16191280e-01 1.42971545e-01 -6.23406708e-01 8.00518274e-01 9.84017372e-01 8.87909010e-02 -5.51519573e-01 3.45523842e-02 -2.12695785e-02 4.02914464e-01 -2.45590642e-01 -7.54967332e-01 4.99251276e-01 -8.16358626e-02 -4.73431826e-01 -2.29455397e-01 4.24273252e-01 -4.68808442e-01 -1.86884567e-01 1.65850863e-01 -7.12087572e-01 2.99201757e-01 2.61954051e-02 1.53965084e-02 -4.42813575e-01 -5.95920801e-01 5.06356061e-01 -1.89770162e-01 -6.78576648e-01 -4.85859752e-01 -1.05033922e+00 1.69918910e-01 9.32086825e-01 -3.36263508e-01 -4.37207282e-01 -9.46940958e-01 -5.57325125e-01 4.75282550e-01 4.12304580e-01 3.63358915e-01 2.21082598e-01 -7.87573695e-01 -6.22119606e-01 -5.51486194e-01 2.52513662e-02 -5.07161975e-01 2.06275567e-01 7.56778240e-01 -7.96855271e-01 6.65154874e-01 -4.34871197e-01 -3.65527749e-01 -1.53552556e+00 -1.97048172e-01 3.72212559e-01 -4.19450045e-01 -4.25678790e-01 3.95748258e-01 -3.99907291e-01 -7.92419612e-01 5.84229290e-01 4.53348309e-01 -6.14644527e-01 2.07214266e-01 6.45072043e-01 7.74648905e-01 -5.03400974e-02 -5.89698434e-01 -3.17279138e-02 -2.56472349e-01 -2.10446239e-01 -9.52556729e-01 8.91279459e-01 -4.21950758e-01 -4.27862853e-01 9.90076005e-01 6.57533467e-01 -2.09651589e-01 -5.19919932e-01 -7.20145926e-03 5.45923471e-01 -2.89446533e-01 -4.74249303e-01 -1.02814686e+00 -3.31951231e-02 5.55386722e-01 2.06608430e-01 1.22588992e+00 7.24724352e-01 -3.51813704e-01 2.67196268e-01 8.20472658e-01 4.71942514e-01 -1.53206313e+00 -3.69122177e-02 8.91250789e-01 1.26066375e+00 -1.16873980e+00 -1.84563071e-01 -3.85398149e-01 -1.17394304e+00 1.20214927e+00 8.05992067e-01 3.71067315e-01 3.57200384e-01 1.84902623e-01 5.90505660e-01 -4.91663843e-01 -9.40042138e-01 1.06579870e-01 -2.01277897e-01 4.98209804e-01 1.21076000e+00 6.75035343e-02 -9.91904259e-01 2.46075287e-01 -4.35885102e-01 -1.92654669e-01 7.47227371e-01 1.31518602e+00 -4.10016239e-01 -1.47935963e+00 -3.49391729e-01 4.15587932e-01 -4.00185078e-01 2.94362277e-01 -1.29914010e+00 1.22344494e+00 -7.96133459e-01 1.47392392e+00 -2.85442978e-01 -3.87473971e-01 7.49408245e-01 4.98603106e-01 2.90585399e-01 -6.15514398e-01 -1.31245625e+00 -1.90913320e-01 1.14363182e+00 -2.35381857e-01 -7.77428091e-01 -7.78607965e-01 -8.15087318e-01 -7.52101004e-01 -4.70973581e-01 8.56261671e-01 6.94694996e-01 8.79501581e-01 2.38802712e-02 1.70245230e-01 6.34946108e-01 -7.35938668e-01 -8.41811419e-01 -1.46239233e+00 -3.97125781e-01 2.01672703e-01 -1.74098596e-01 -4.38032746e-01 -3.29570413e-01 -2.22089812e-01]
[12.935574531555176, 7.969776630401611]
139fd8dd-308c-4fcc-8da4-7100e12f82e9
patient-specific-seizure-prediction-using
2011.08982
null
https://arxiv.org/abs/2011.08982v1
https://arxiv.org/pdf/2011.08982v1.pdf
Patient-Specific Seizure Prediction Using Single Seizure Electroencephalography Recording
Electroencephalogram (EEG) is a prominent way to measure the brain activity for studying epilepsy, thereby helping in predicting seizures. Seizure prediction is an active research area with many deep learning based approaches dominating the recent literature for solving this problem. But these models require a considerable number of patient-specific seizures to be recorded for extracting the preictal and interictal EEG data for training a classifier. The increase in sensitivity and specificity for seizure prediction using the machine learning models is noteworthy. However, the need for a significant number of patient-specific seizures and periodic retraining of the model because of non-stationary EEG creates difficulties for designing practical device for a patient. To mitigate this process, we propose a Siamese neural network based seizure prediction method that takes a wavelet transformed EEG tensor as an input with convolutional neural network (CNN) as the base network for detecting change-points in EEG. Compared to the solutions in the literature, which utilize days of EEG recordings, our method only needs one seizure for training which translates to less than ten minutes of preictal and interictal data while still getting comparable results to models which utilize multiple seizures for seizure prediction.
['Bülent Yener', 'Hui Su', 'Lara Marcuse', 'Arun Iyengar', 'Zaid Bin Tariq']
2020-11-14
null
null
null
null
['seizure-prediction']
['medical']
[ 1.89511888e-02 -3.19015115e-01 4.69578564e-01 -3.46521914e-01 -5.84446728e-01 -3.34510267e-01 2.60557860e-01 4.50216308e-02 -3.94710779e-01 6.56897366e-01 6.13142736e-02 -1.66911229e-01 -3.48591506e-01 -5.28209627e-01 -2.73813009e-01 -7.62070537e-01 -6.55233264e-01 3.43874097e-02 5.88579029e-02 -1.55179977e-01 4.07931834e-01 6.39898062e-01 -1.23407447e+00 4.29074466e-01 6.91333234e-01 1.28455353e+00 2.92356551e-01 5.24654746e-01 1.12025455e-01 4.89837676e-01 -8.83728683e-01 3.04544568e-01 3.95337522e-01 -3.96583617e-01 -5.16029716e-01 1.13858730e-01 -1.47722840e-01 -1.18028380e-01 -3.89415950e-01 1.01921809e+00 8.48825216e-01 -2.54441053e-01 4.30719942e-01 -1.25946009e+00 -3.36689651e-01 2.62492001e-01 -3.94597411e-01 7.84528494e-01 3.96577179e-01 1.53990522e-01 2.04852909e-01 -6.77318990e-01 2.91049536e-02 2.24555448e-01 6.64505541e-01 4.12149072e-01 -8.07450235e-01 -1.09428430e+00 -4.16380942e-01 6.14582300e-01 -1.36737013e+00 -2.02150568e-01 1.13469696e+00 -5.66058338e-01 1.27671885e+00 2.70987570e-01 1.49735212e+00 1.16571677e+00 8.62319827e-01 4.31091309e-01 1.29532850e+00 -5.10095470e-02 5.44696689e-01 -9.90201384e-02 1.98028728e-01 9.65057090e-02 6.27948493e-02 1.50942072e-01 -7.69004107e-01 -1.49398208e-01 6.52085662e-01 4.15167928e-01 -6.27338409e-01 2.93551032e-02 -1.07227659e+00 4.60507214e-01 2.95917839e-01 7.89829373e-01 -9.20512080e-01 -1.90496743e-01 5.89715600e-01 6.41857922e-01 5.15763402e-01 7.32339084e-01 -7.07399428e-01 -4.97455865e-01 -1.60406959e+00 5.53357117e-02 6.61964238e-01 5.62509894e-01 5.07089555e-01 2.70399719e-01 2.75894612e-01 5.27108610e-01 -3.42281014e-01 1.74461715e-02 1.28406608e+00 -1.09810069e-01 2.73073554e-01 9.33597803e-01 -2.93482423e-01 -9.37180996e-01 -6.89998090e-01 -8.39478850e-01 -1.13875556e+00 8.12469572e-02 3.15761268e-02 -3.01036894e-01 -8.95037830e-01 1.11999202e+00 -2.89051116e-01 8.15813959e-01 9.32538956e-02 6.45263672e-01 4.24542636e-01 5.08022666e-01 -3.26189011e-01 -4.01251614e-01 1.26843739e+00 -5.39368987e-01 -5.93881071e-01 -1.83642983e-01 6.45740151e-01 -6.50149643e-01 8.43717694e-01 9.93684411e-01 -7.96279073e-01 -1.09789133e-01 -1.21998334e+00 4.53422725e-01 -5.83489776e-01 4.01730463e-02 5.49696326e-01 2.67703682e-01 -1.30345762e+00 6.01691782e-01 -1.13141596e+00 -3.32611680e-01 4.64716077e-01 1.03099525e+00 -6.15163922e-01 2.95498818e-01 -9.25332665e-01 1.06873012e+00 3.30332875e-01 1.18352018e-01 -7.99863338e-01 -8.06563020e-01 -4.93159384e-01 2.07958505e-01 -2.88100928e-01 1.04464553e-01 7.42799878e-01 -9.86418784e-01 -1.42013729e+00 3.00312281e-01 3.35116610e-02 -5.99511504e-01 3.12647074e-01 -1.20849516e-02 -6.61447704e-01 1.96777120e-01 -2.00995877e-01 4.09521818e-01 8.90214682e-01 -4.43057001e-01 -5.56037009e-01 -5.76698482e-01 -3.56616914e-01 -1.15603514e-01 -5.60988843e-01 1.48682594e-01 8.81525874e-02 -6.53261125e-01 2.13568389e-01 -7.84634531e-01 -4.76129763e-02 -7.17516541e-01 -1.53727397e-01 -3.50418210e-01 1.25670314e+00 -9.94567633e-01 1.18817246e+00 -1.91170001e+00 -9.72591266e-02 1.64160103e-01 3.58480692e-01 2.06257299e-01 -1.05532028e-01 3.06673735e-01 -5.79126358e-01 -5.17007291e-01 -3.12650561e-01 6.51102141e-02 -5.45736134e-01 -1.67074293e-01 -1.99430987e-01 5.13931215e-01 4.27483678e-01 7.37724125e-01 -6.88594878e-01 1.23819001e-01 3.24867785e-01 8.17706287e-01 -3.36931348e-01 3.32834780e-01 4.79684919e-01 9.03172731e-01 -3.34494770e-01 5.01485407e-01 5.90605140e-01 -2.29676422e-02 -2.49672398e-01 -1.39383068e-02 -1.63036406e-01 3.04437727e-01 -8.30234289e-01 1.61842895e+00 -3.99374068e-01 9.41519141e-01 -3.67227167e-01 -1.53559577e+00 9.95429933e-01 7.38429487e-01 9.99062777e-01 -6.46455586e-01 2.72425264e-01 4.00102496e-01 3.52422774e-01 -9.23558712e-01 -2.83594489e-01 4.22557183e-02 1.57976836e-01 5.83956420e-01 1.84847862e-01 -1.43902585e-01 3.52609791e-02 -3.77819568e-01 1.65114093e+00 -1.37840122e-01 3.67381513e-01 -4.85032022e-01 5.75373113e-01 -1.73450857e-01 6.66003823e-01 -1.25627369e-01 -2.01725468e-01 3.30039263e-01 1.92920953e-01 -1.07188332e+00 -8.27108264e-01 -4.19254988e-01 -3.06843400e-01 4.10828233e-01 -1.75722301e-01 -2.11583108e-01 -1.07203913e+00 -2.98282743e-01 -4.31353748e-01 3.11161220e-01 -6.21551514e-01 -2.73779556e-02 -6.54109418e-01 -1.16119719e+00 4.17212069e-01 5.62271893e-01 6.20199084e-01 -1.31168604e+00 -1.18490183e+00 5.08571148e-01 -5.01584969e-02 -6.54444754e-01 -1.89804211e-01 6.34750426e-01 -1.01314330e+00 -1.30045605e+00 -8.92223895e-01 -9.41401362e-01 6.97333753e-01 -1.59066036e-01 7.16413319e-01 -1.49730383e-03 -6.23339057e-01 -4.53157350e-02 -3.14873070e-01 -7.82539368e-01 2.45720088e-01 8.31769407e-02 2.66502351e-01 2.41562128e-02 1.08390427e+00 -1.26411808e+00 -8.88972640e-01 -2.13269562e-01 -9.57623422e-01 -2.73681935e-02 7.71463454e-01 7.30571330e-01 3.13851386e-01 2.77756184e-01 9.27585721e-01 -2.38332003e-01 9.99414325e-01 -5.82832336e-01 -6.60176098e-01 4.97029722e-02 -6.07430935e-01 -1.43354729e-01 9.14933681e-01 -6.80362284e-01 -2.63302475e-01 2.87745655e-01 8.77336785e-02 -1.07733153e-01 -2.77635813e-01 4.04841840e-01 2.00740904e-01 -3.55762035e-01 4.43369359e-01 7.32457995e-01 -3.19145411e-01 -3.89370501e-01 -7.00093448e-01 9.12934482e-01 3.05814207e-01 2.19638869e-01 4.76914227e-01 2.00651169e-01 -2.74302419e-02 -8.38478923e-01 -2.83947974e-01 -5.17081678e-01 -8.66962969e-01 -1.33827373e-01 8.54013085e-01 -7.58118272e-01 -6.37758732e-01 5.44631183e-01 -1.16970718e+00 -1.57929733e-02 1.07642069e-01 9.34372485e-01 -4.08129305e-01 8.66808638e-04 -4.64168459e-01 -6.45327866e-01 -7.27243602e-01 -1.43839312e+00 7.32958317e-01 -9.49109420e-02 -2.65900075e-01 -9.13646579e-01 2.36236081e-01 -1.05086476e-01 6.53564990e-01 4.79164332e-01 7.75968730e-01 -7.88554490e-01 -4.09631461e-01 -5.99420190e-01 9.00485292e-02 4.69261914e-01 4.50074106e-01 -6.41283512e-01 -1.02044153e+00 -5.50329745e-01 7.86545932e-01 -9.29239020e-02 3.51806641e-01 4.38063115e-01 1.36768973e+00 -2.44303539e-01 -1.64459169e-01 7.13334322e-01 1.53162336e+00 8.69949758e-01 5.87482572e-01 4.24996346e-01 2.19972804e-01 3.10275257e-01 -1.27679572e-01 4.36481357e-01 1.80865943e-01 2.55154014e-01 3.01341593e-01 -5.62261604e-02 2.02400655e-01 4.20970380e-01 3.24479967e-01 1.42987192e+00 -2.09169582e-01 3.45126539e-01 -1.20025349e+00 7.77825236e-01 -1.30342138e+00 -8.18316579e-01 9.80493426e-02 2.05386305e+00 7.62144148e-01 -1.77736133e-01 -4.19031419e-02 7.46911347e-01 3.64554286e-01 -3.34074080e-01 -6.32845402e-01 -1.40107051e-01 4.23990265e-02 9.20002520e-01 3.26046199e-01 9.56757963e-02 -7.15579093e-01 4.46366757e-01 6.56572771e+00 2.56450593e-01 -2.07631993e+00 1.56441063e-01 2.80846387e-01 -4.17826146e-01 3.00547242e-01 -2.39254698e-01 -9.93552506e-02 8.17846179e-01 1.20753348e+00 -3.02743852e-01 7.52487838e-01 5.35143733e-01 5.45673966e-01 -1.43979415e-01 -1.26831412e+00 1.55995858e+00 1.02949858e-01 -1.08285415e+00 -1.68819264e-01 1.49128094e-01 6.10279858e-01 3.79064918e-01 -1.92721829e-01 -5.96749559e-02 -4.66968507e-01 -1.16848493e+00 1.42850280e-01 6.75972164e-01 5.91503024e-01 -9.30550754e-01 9.86825347e-01 5.03063977e-01 -1.06673694e+00 -2.78328389e-01 -3.68074834e-01 -4.34760273e-01 -2.74625897e-01 5.84142923e-01 -9.08081472e-01 2.99388915e-01 1.04396129e+00 8.42303991e-01 -7.15798020e-01 1.30213475e+00 1.24534637e-01 7.78011858e-01 -3.07904065e-01 -2.49537155e-02 3.59009326e-01 -2.13186890e-01 3.96335334e-01 1.02666903e+00 9.48892653e-01 1.44014731e-01 -1.12459660e-01 4.04436231e-01 1.75221100e-01 9.55958143e-02 -7.10089862e-01 3.73645290e-03 1.95709735e-01 1.14175570e+00 -7.61146188e-01 -2.15690747e-01 -4.83302474e-01 9.99221563e-01 4.73208316e-02 2.82762200e-01 -3.94048810e-01 -8.19893003e-01 3.22265387e-01 1.51355624e-01 -9.12298039e-02 -1.59668341e-01 -5.67592382e-01 -1.24747264e+00 2.66387194e-01 -7.83808887e-01 -1.08893298e-01 -7.05913961e-01 -9.36747432e-01 1.16560698e+00 -2.67021835e-01 -1.69244564e+00 -4.72530693e-01 -6.16075575e-01 -1.01546562e+00 8.73684585e-01 -1.52384675e+00 -7.38016188e-01 -5.10883510e-01 9.16645944e-01 7.45882869e-01 -5.34368336e-01 1.07764614e+00 4.15609986e-01 -3.78850490e-01 1.19207114e-01 -1.70304239e-01 1.82412490e-01 5.09512126e-01 -1.08894813e+00 9.14220959e-02 8.42073083e-01 -2.66782753e-02 4.93609726e-01 6.69564724e-01 -4.80453491e-01 -1.44243550e+00 -9.96505857e-01 9.94203389e-01 -3.39071929e-01 8.07679653e-01 -2.85123765e-01 -9.95147109e-01 6.38086200e-01 5.94094634e-01 1.42499397e-03 9.23262656e-01 -3.71312737e-01 3.59161824e-01 -3.84467155e-01 -1.13088703e+00 2.86360323e-01 6.29940689e-01 -4.13542449e-01 -9.42782998e-01 3.79637390e-01 6.82501961e-03 -3.34301288e-03 -9.69749331e-01 2.95122027e-01 3.20836157e-01 -9.32039142e-01 2.60477811e-01 -2.36775920e-01 1.93899766e-01 -2.07385004e-01 3.33814859e-01 -1.73934674e+00 -1.30996615e-01 -7.05578923e-01 6.34256303e-02 3.50636691e-01 3.42384994e-01 -7.94585288e-01 9.49538410e-01 3.49173248e-01 -2.21443877e-01 -1.20791006e+00 -1.02307487e+00 -7.69983113e-01 2.02440936e-02 -6.87495351e-01 9.25384641e-01 1.09703434e+00 5.27550042e-01 1.20101623e-01 -1.49378181e-01 2.73840874e-01 3.82626742e-01 -5.82298450e-02 2.95675039e-01 -1.53521717e+00 1.60584271e-01 -1.94882363e-01 -1.01108146e+00 -2.90571481e-01 -3.16820592e-02 -9.52161193e-01 -1.63129687e-01 -1.42181563e+00 5.60461581e-02 -9.27073509e-02 -8.05512726e-01 6.96461201e-01 3.55782747e-01 4.46248651e-01 -2.93498784e-01 2.17912376e-01 -8.08017794e-03 2.62488127e-01 1.03366888e+00 -1.32196322e-01 -5.78706145e-01 -4.14608940e-02 -3.95844549e-01 8.00812185e-01 1.00710726e+00 -5.70633829e-01 -6.35674417e-01 -5.62536061e-01 6.39937669e-02 7.95936286e-02 1.61397159e-01 -1.51872230e+00 6.35584235e-01 1.66230425e-01 8.58059347e-01 -6.23560131e-01 3.18010449e-01 -9.50938523e-01 2.47849271e-01 6.89747095e-01 -2.78154612e-02 6.61306322e-01 1.67496905e-01 1.67259857e-01 -6.43019855e-01 1.20721288e-01 6.23706818e-01 -1.30727708e-01 -5.16500354e-01 5.44744492e-01 -6.40733898e-01 -3.34421277e-01 1.29144824e+00 -6.55220747e-01 2.53155440e-01 -1.46712765e-01 -8.63654256e-01 -1.83061827e-02 -1.32137146e-02 4.04558182e-01 9.24377620e-01 -1.17786407e+00 -6.30000591e-01 7.86542177e-01 8.81683175e-03 -1.90500379e-01 4.16026823e-02 1.18673658e+00 -6.23278022e-01 7.17911661e-01 -8.13472092e-01 -5.73006809e-01 -1.07786286e+00 2.36463487e-01 3.82174432e-01 -1.17882922e-01 -1.02382314e+00 6.52133405e-01 -1.60165489e-01 1.01267122e-01 2.36367419e-01 -6.96406782e-01 -3.89353782e-01 -8.85268003e-02 7.89046109e-01 2.84126371e-01 7.07813084e-01 -2.39155084e-01 -4.78871733e-01 3.88984233e-01 5.06737567e-02 2.65261214e-02 2.00617075e+00 4.16189432e-01 -4.99247998e-01 3.32365304e-01 1.23143756e+00 -3.11170846e-01 -1.24128938e+00 2.97744781e-01 3.30225438e-01 -2.24343948e-02 4.16486353e-01 -8.54438245e-01 -1.21024585e+00 1.07200098e+00 1.02053440e+00 3.68146449e-01 1.69034541e+00 -4.70296651e-01 1.04299092e+00 6.01784647e-01 7.90932536e-01 -1.13582480e+00 -1.46096125e-01 3.58207762e-01 1.13874817e+00 -9.78766620e-01 -2.74882227e-01 1.74965099e-01 -3.05786669e-01 1.47242236e+00 2.65098035e-01 -7.37453997e-01 1.10505044e+00 6.21229768e-01 1.06694646e-01 -3.47052336e-01 -6.76625073e-01 4.14077967e-01 3.07939917e-01 6.19471610e-01 3.82340670e-01 1.20879844e-01 -2.93710262e-01 6.76667571e-01 -8.09111476e-01 4.06605035e-01 5.15169322e-01 1.03604019e+00 -2.56354541e-01 -1.04764462e+00 -2.61778533e-01 9.09875691e-01 -6.75467312e-01 -2.83764720e-01 -2.94982463e-01 5.70547640e-01 3.71479422e-01 9.14759815e-01 1.46966428e-01 -5.18181860e-01 1.39658034e-01 3.11393678e-01 5.72499752e-01 -7.21699655e-01 -8.02073658e-01 3.22059169e-02 -6.56577647e-01 -6.11720562e-01 -3.50610405e-01 -4.82297361e-01 -1.11885071e+00 8.76870826e-02 -2.20421225e-01 2.37873465e-01 9.51176584e-01 1.22516167e+00 4.48526502e-01 6.18593574e-01 7.11369991e-01 -8.42213452e-01 -1.37461960e-01 -1.48438525e+00 -8.69168758e-01 1.60323709e-01 7.01998651e-01 -5.38276911e-01 -4.34261322e-01 2.80105084e-01]
[13.208760261535645, 3.513373613357544]
828a3b3e-7c78-4ad6-a6c1-694ca3c63580
conceptnet-at-semeval-2017-task-2-extending
1704.03560
null
http://arxiv.org/abs/1704.03560v2
http://arxiv.org/pdf/1704.03560v2.pdf
ConceptNet at SemEval-2017 Task 2: Extending Word Embeddings with Multilingual Relational Knowledge
This paper describes Luminoso's participation in SemEval 2017 Task 2, "Multilingual and Cross-lingual Semantic Word Similarity", with a system based on ConceptNet. ConceptNet is an open, multilingual knowledge graph that focuses on general knowledge that relates the meanings of words and phrases. Our submission to SemEval was an update of previous work that builds high-quality, multilingual word embeddings from a combination of ConceptNet and distributional semantics. Our system took first place in both subtasks. It ranked first in 4 out of 5 of the separate languages, and also ranked first in all 10 of the cross-lingual language pairs.
['Joanna Lowry-Duda', 'Robyn Speer']
2017-04-11
conceptnet-at-semeval-2017-task-2-extending-1
https://aclanthology.org/S17-2008
https://aclanthology.org/S17-2008.pdf
semeval-2017-8
['multilingual-word-embeddings']
['methodology']
[-6.67066753e-01 -9.93774012e-02 -3.97478938e-01 -2.31173441e-01 -6.27636194e-01 -9.20375586e-01 9.21707809e-01 9.04029608e-01 -9.74166870e-01 6.52999103e-01 8.58855307e-01 -3.66688162e-01 -1.89100042e-01 -5.74716866e-01 -2.66527742e-01 1.15653398e-02 1.12599850e-01 8.12999845e-01 1.08690798e-01 -8.86835456e-01 2.12406099e-01 -2.00465605e-01 -1.28345549e+00 1.89961866e-01 9.35066223e-01 5.42457342e-01 3.02881390e-01 2.22289577e-01 -7.83393204e-01 3.46772730e-01 -8.37934539e-02 -9.51823950e-01 -1.51041821e-01 -1.16765723e-01 -1.21593571e+00 -9.12344277e-01 8.98184359e-01 7.50491023e-01 -5.58960438e-02 1.38025296e+00 5.99461019e-01 3.63931745e-01 6.89573169e-01 -1.12957144e+00 -1.50460982e+00 1.16191018e+00 4.11805473e-02 5.61188340e-01 7.83486068e-01 -3.94934386e-01 1.64487982e+00 -1.27095830e+00 9.44601297e-01 1.52376962e+00 1.07519960e+00 2.20841169e-01 -1.11086416e+00 -5.85066199e-01 -4.32052054e-02 5.27718008e-01 -1.58427286e+00 -1.09390438e-01 1.53250903e-01 -4.88448769e-01 1.70008075e+00 -3.26841235e-01 8.38855147e-01 1.11880124e+00 -2.89718080e-02 3.76799583e-01 1.02862525e+00 -7.87872612e-01 -2.69011054e-02 3.48846108e-01 3.71834725e-01 3.91966462e-01 5.92767656e-01 -5.97735308e-02 -6.65410101e-01 -3.35486621e-01 1.47973001e-01 -4.87811297e-01 -1.99104249e-01 -2.90976405e-01 -1.44264901e+00 1.09179521e+00 2.06861719e-01 1.05342674e+00 -9.55826044e-02 -6.25716671e-02 6.84808731e-01 4.34148312e-01 7.16315210e-01 1.03274906e+00 -7.45805681e-01 -2.00950593e-01 -5.21942794e-01 3.57178509e-01 9.40960109e-01 1.07324839e+00 8.86464715e-01 -2.28780776e-01 1.45670250e-01 1.16948986e+00 1.17349565e-01 5.67722440e-01 9.51262176e-01 -4.85321373e-01 2.02808499e-01 5.12142539e-01 -2.75689602e-01 -8.88962686e-01 -2.12175533e-01 -9.55363810e-02 7.29597453e-03 -5.84213495e-01 -5.60665838e-02 1.99645814e-02 -4.38194185e-01 1.76644647e+00 1.30288735e-01 3.44167799e-01 5.11202812e-01 5.99020302e-01 1.35242510e+00 1.64028570e-01 5.99244833e-01 2.51972795e-01 1.69147527e+00 -6.32231116e-01 -7.72091925e-01 -1.23412617e-01 9.94603157e-01 -1.19882083e+00 1.24134660e+00 -2.54453987e-01 -6.82150543e-01 -4.41752702e-01 -1.06757140e+00 -3.48033339e-01 -1.59213305e+00 -4.71439928e-01 7.36344516e-01 5.00338435e-01 -1.46910226e+00 4.02410418e-01 -7.42392614e-02 -1.28528154e+00 -1.09808080e-01 -2.24033535e-01 -5.78964114e-01 -4.06502545e-01 -2.00409842e+00 1.77783120e+00 1.05508590e+00 -8.87672782e-01 -2.10133910e-01 -1.30531704e+00 -1.36646521e+00 -3.24201256e-01 7.14168549e-02 -6.05229497e-01 9.64722812e-01 -7.42823839e-01 -7.02437699e-01 1.55662906e+00 5.92043810e-02 -5.32518208e-01 -3.60743612e-01 -3.78654450e-01 -1.37956619e+00 -2.45607376e-01 7.97669470e-01 8.31947327e-01 -6.53889179e-02 -8.20528626e-01 -7.56725967e-01 -1.75562456e-01 2.51091748e-01 4.62353408e-01 -6.46139383e-01 4.14851636e-01 -1.80120364e-01 -7.33647645e-01 -4.46270794e-01 -5.51211417e-01 1.07618794e-01 -5.87137938e-01 1.70033336e-01 -7.08794296e-01 3.75701964e-01 -6.03306353e-01 1.07596219e+00 -1.80095112e+00 -1.46204811e-02 4.78070192e-02 -6.38638511e-02 1.12836994e-01 -3.06185246e-01 8.07765245e-01 -4.78692234e-01 3.63493800e-01 1.65634174e-02 -7.42083043e-02 2.71083325e-01 4.15455729e-01 -1.95308700e-01 1.01657897e-01 -8.17043632e-02 1.15433669e+00 -1.68059480e+00 -6.12280548e-01 2.54563391e-01 4.58218634e-01 -3.03576738e-01 -3.17096382e-01 1.92769334e-01 -1.69101030e-01 1.21169917e-01 4.14224118e-01 3.19032341e-01 3.12945485e-01 5.57764769e-01 -4.21719551e-01 -1.76546976e-01 6.75030828e-01 -6.84878111e-01 2.35460973e+00 -8.10387135e-01 5.96791267e-01 -4.96255457e-01 -8.67242396e-01 7.01748610e-01 5.19861221e-01 4.69286919e-01 -8.92870247e-01 -1.83549318e-02 6.39431715e-01 -1.85288325e-01 -3.51854295e-01 1.02906632e+00 -4.11436141e-01 -5.55211306e-01 3.33454460e-01 7.49798656e-01 -5.98655462e-01 5.00554025e-01 3.31856161e-01 6.44384801e-01 2.55155355e-01 9.53883171e-01 -1.17806709e+00 5.62331617e-01 8.78987014e-02 1.98232934e-01 2.06515208e-01 -1.90831974e-01 1.18873335e-01 -9.21886191e-02 -4.08954561e-01 -8.97367239e-01 -1.52072513e+00 -3.92392308e-01 1.27360213e+00 1.34081677e-01 -1.03364182e+00 -3.61790121e-01 -6.41863644e-01 3.70219439e-01 1.15140367e+00 -5.18153846e-01 -7.86336064e-02 -2.89095163e-01 -2.66740650e-01 9.74430799e-01 4.66541529e-01 -1.91621095e-01 -9.64887679e-01 1.19503260e-01 2.22157314e-01 -3.26130539e-01 -1.33387220e+00 -5.85879922e-01 3.97365838e-02 -2.85889983e-01 -1.35134292e+00 -2.77921349e-01 -1.31300306e+00 2.44838167e-02 2.93686390e-02 1.92795980e+00 -2.84464836e-01 -4.12131935e-01 6.03722334e-01 -7.28839815e-01 -6.71122074e-01 -1.68565974e-01 1.01359166e-01 6.56080842e-01 -5.80017507e-01 1.42731118e+00 -5.38950086e-01 3.24083082e-02 -2.93435335e-01 -6.05156064e-01 -7.01887906e-01 -1.59523278e-01 7.27974415e-01 5.11491001e-01 -2.97283143e-01 6.59753263e-01 -6.79970741e-01 1.09349298e+00 -8.36561084e-01 -3.32531214e-01 5.31074405e-01 -5.80141783e-01 1.84888959e-01 4.93871659e-01 -8.93942490e-02 -7.86325991e-01 -5.14379680e-01 -3.15967292e-01 -1.20747268e-01 -1.46716563e-02 6.89509273e-01 9.48883444e-02 1.04976837e-02 6.34725869e-01 -6.24476746e-02 -5.26621103e-01 -5.83859146e-01 1.25708306e+00 5.85984945e-01 6.77349806e-01 -9.13352430e-01 3.54320198e-01 -3.05251256e-02 -4.46098387e-01 -9.34170187e-01 -8.90739977e-01 -1.06503797e+00 -8.80593836e-01 1.78511098e-01 1.14137089e+00 -1.06610107e+00 -3.20353717e-01 -2.42411103e-02 -1.30056286e+00 2.72748858e-01 -5.00280797e-01 8.76550376e-01 -2.35964507e-01 2.68509358e-01 -2.34348476e-01 1.17445871e-01 -4.76234168e-01 -4.46132660e-01 8.02630723e-01 -1.64003372e-01 -8.21748734e-01 -2.08340406e+00 7.25884914e-01 2.72761984e-03 5.70647001e-01 -1.25112966e-01 1.17383409e+00 -1.24131584e+00 4.90445137e-01 3.84245603e-03 -8.13565776e-02 4.93883044e-01 3.16789359e-01 -2.92307258e-01 -5.41075349e-01 -3.09863776e-01 -4.26172882e-01 -7.71084726e-01 7.54135728e-01 -8.57163444e-02 3.93215805e-01 6.93581626e-02 -3.19227427e-01 3.58656347e-01 1.84062850e+00 -2.14112639e-01 2.18997613e-01 3.34081888e-01 8.33010018e-01 6.80076718e-01 4.03793514e-01 -4.68766652e-02 1.06609738e+00 7.36817002e-01 -2.15432450e-01 1.85066193e-01 -4.26484942e-01 -5.49264610e-01 4.79512125e-01 1.83154666e+00 1.45001680e-01 4.39570397e-02 -1.14979684e+00 1.38968909e+00 -1.56206155e+00 -7.16483176e-01 -3.25484991e-01 2.11336040e+00 1.07062614e+00 -4.47309673e-01 -1.88973770e-01 -6.09248281e-01 7.03922927e-01 1.40741512e-01 3.70618887e-02 -8.39425623e-01 -4.97709602e-01 1.08951175e+00 4.49023604e-01 8.52954149e-01 -1.01583421e+00 1.73755515e+00 6.96064615e+00 1.04835558e+00 -7.11422265e-01 7.17946887e-01 -4.24546599e-01 2.57413059e-01 -8.06756556e-01 3.78052145e-01 -7.93773174e-01 2.62120157e-01 1.03292251e+00 -7.63140440e-01 2.27610111e-01 6.64566338e-01 -5.71622610e-01 1.04600325e-01 -1.31420720e+00 1.05426228e+00 7.20483363e-01 -1.34235775e+00 3.68994385e-01 -4.67142433e-01 9.68119621e-01 6.63626492e-01 -4.03215498e-01 5.48329592e-01 9.78270888e-01 -1.17594731e+00 4.87038583e-01 3.09166074e-01 1.06929052e+00 -1.10294068e+00 9.70712006e-01 -3.58145714e-01 -1.74327219e+00 4.53888744e-01 -4.97485310e-01 6.92170039e-02 3.02551717e-01 4.39204961e-01 -5.40977061e-01 9.77524042e-01 8.65422487e-01 1.33175004e+00 -5.39705932e-01 6.62551343e-01 -6.62651241e-01 2.73878574e-01 1.49642274e-01 1.56636350e-02 4.50398445e-01 -6.47683516e-02 6.25877857e-01 1.79503095e+00 3.54355782e-01 -3.54161412e-01 4.98422652e-01 6.07350588e-01 -4.82637584e-02 7.91929007e-01 -1.00482011e+00 -3.36699694e-01 7.87053823e-01 1.08514917e+00 -2.12223202e-01 -5.36737680e-01 -6.12012148e-01 9.68261063e-01 8.02897751e-01 -5.47858775e-02 -4.24737602e-01 -7.85531104e-01 1.20241654e+00 -2.75546700e-01 7.31617883e-02 -2.39318788e-01 -6.73829541e-02 -1.29257810e+00 -2.96481758e-01 -4.57737207e-01 7.56908357e-01 -7.29502320e-01 -2.12576985e+00 7.67745376e-01 4.25108448e-02 -7.20111907e-01 -3.28277439e-01 -8.31701696e-01 -2.81324416e-01 1.09498179e+00 -1.57409668e+00 -1.28706443e+00 2.18672588e-01 9.34147120e-01 2.53949910e-01 -4.65334535e-01 1.70655477e+00 5.73322356e-01 2.43992597e-01 6.91021919e-01 4.48273048e-02 3.56139019e-02 1.24735606e+00 -1.42335939e+00 8.47841144e-01 5.72125196e-01 4.94194210e-01 8.98091614e-01 4.82198685e-01 -6.19620800e-01 -7.23668218e-01 -1.02558088e+00 2.09367633e+00 -8.55985343e-01 1.51734853e+00 -3.61118466e-01 -8.23056221e-01 7.34509468e-01 9.58677471e-01 -1.52333453e-01 1.08147299e+00 5.63172400e-01 -9.69854772e-01 2.04227969e-01 -8.63925338e-01 5.02957165e-01 1.21915817e+00 -1.33085108e+00 -1.58453095e+00 4.53572303e-01 1.24430048e+00 7.98042640e-02 -1.38271117e+00 1.59276307e-01 4.01887029e-01 -2.89523244e-01 1.04493189e+00 -1.03026819e+00 2.77551800e-01 -1.90282822e-01 -6.94734931e-01 -2.02074552e+00 -3.05395424e-01 -2.80122548e-01 4.54790145e-01 1.33341563e+00 5.50440073e-01 -7.07598746e-01 -8.22559446e-02 -3.50346893e-01 -3.98776948e-01 -1.71038613e-01 -1.01899612e+00 -1.29537868e+00 7.36322880e-01 -8.22721660e-01 6.02418780e-01 1.74713516e+00 5.89708805e-01 7.20262527e-01 1.97771847e-01 -4.31906909e-01 4.13006067e-01 -3.38959306e-01 3.97485122e-02 -1.35382700e+00 1.10307597e-01 -5.96673906e-01 -8.80408466e-01 -2.77980834e-01 8.98085475e-01 -1.86753249e+00 -2.33520359e-01 -1.67611265e+00 1.72041446e-01 -2.79723793e-01 -5.90786576e-01 4.83949125e-01 -2.05764517e-01 2.91119814e-01 5.98967075e-03 -2.02249497e-01 -7.29108274e-01 5.38015068e-01 5.82581341e-01 -1.05251141e-01 3.66770148e-01 -9.09265101e-01 -8.56343150e-01 7.81902611e-01 5.35661161e-01 -4.45592225e-01 -3.98164928e-01 -7.12634861e-01 5.36079884e-01 -8.42204988e-01 4.16588224e-02 -6.19143784e-01 7.00529441e-02 5.73597290e-03 -2.24218681e-01 -2.38318816e-01 1.25721604e-01 -6.21481359e-01 -2.19563812e-01 2.09386915e-01 -1.87197641e-01 6.28599346e-01 3.44670564e-01 1.39159247e-01 -6.06758475e-01 -1.62909701e-01 5.51736653e-01 -3.52062047e-01 -1.41938877e+00 -1.99321732e-02 -3.23019981e-01 1.11871266e+00 9.02385294e-01 -5.72334826e-02 -3.84879440e-01 6.87488243e-02 -4.56256628e-01 3.26558441e-01 4.77005810e-01 1.29597557e+00 3.55019927e-01 -2.00325871e+00 -1.04314804e+00 -1.51100621e-01 1.04528546e+00 -9.34513748e-01 -1.85583457e-01 5.84952652e-01 -3.02571207e-01 8.61809433e-01 -3.93468827e-01 -3.68923694e-03 -9.27070856e-01 5.38496196e-01 1.91053569e-01 -1.97678149e-01 -2.94250339e-01 1.00552166e+00 -2.86701202e-01 -1.25510705e+00 -1.94174901e-01 -3.71703133e-02 -5.67808211e-01 4.32835013e-01 5.00262618e-01 3.22256237e-01 1.14501454e-01 -1.21010077e+00 -8.98039401e-01 8.87204409e-01 7.01635629e-02 -3.91628146e-01 1.06380224e+00 -8.80776066e-03 -8.44086111e-01 8.24452341e-01 1.59973395e+00 1.88892528e-01 3.10885698e-01 -4.96190429e-01 7.08441257e-01 -1.80324569e-01 -8.69552270e-02 -8.65226507e-01 -5.11348605e-01 6.61479414e-01 4.52929527e-01 -2.83894002e-01 4.80300158e-01 2.31891289e-01 1.04526186e+00 3.91731054e-01 6.28400683e-01 -1.54187012e+00 -1.87763751e-01 1.48470140e+00 7.58844793e-01 -1.06130230e+00 -2.06789792e-01 -2.10428327e-01 -6.96124256e-01 8.49001288e-01 5.87648988e-01 -7.16125965e-02 9.59251165e-01 -5.16426452e-02 1.81574181e-01 -6.39273047e-01 -5.09733975e-01 -6.59959733e-01 7.16374636e-01 1.02355635e+00 1.02176523e+00 5.20327687e-01 -9.68577921e-01 7.11514592e-01 -6.98138893e-01 -3.49031121e-01 5.41706681e-02 6.41597688e-01 -2.06580848e-01 -1.45924616e+00 1.31260663e-01 -5.00561222e-02 -4.50284988e-01 -1.19246066e+00 -7.07965076e-01 8.16291451e-01 6.99142873e-01 7.81002700e-01 3.15307885e-01 -4.28397626e-01 3.33057731e-01 3.36468279e-01 4.20835018e-01 -1.01472092e+00 -7.51733005e-01 -8.82388115e-01 5.61729848e-01 -6.96766913e-01 -5.74317455e-01 -5.24031818e-01 -1.33000731e+00 -3.41597885e-01 1.74471140e-02 4.04767722e-01 7.29811132e-01 1.09633338e+00 2.94588804e-01 1.27334908e-01 -1.28326518e-02 -2.64769703e-01 -7.68868551e-02 -1.02753091e+00 -7.37731040e-01 7.57507205e-01 -3.85699153e-01 -6.33319795e-01 -1.92538187e-01 -2.17386246e-01]
[10.760143280029297, 9.69489860534668]
b707e5b7-9e67-438e-a57e-d840b43d5fe1
wenlan-2-0-make-ai-imagine-via-a-multimodal
2110.14378
null
https://arxiv.org/abs/2110.14378v2
https://arxiv.org/pdf/2110.14378v2.pdf
Towards artificial general intelligence via a multimodal foundation model
The fundamental goal of artificial intelligence (AI) is to mimic the core cognitive activities of human. Despite tremendous success in the AI research, most of existing methods have only single-cognitive ability. To overcome this limitation and take a solid step towards artificial general intelligence (AGI), we develop a foundation model pre-trained with huge multimodal data, which can be quickly adapted for various downstream cognitive tasks. To achieve this goal, we propose to pre-train our foundation model by self-supervised learning with weak semantic correlation data crawled from the Internet and show that promising results can be obtained on a wide range of downstream tasks. Particularly, with the developed model-interpretability tools, we demonstrate that strong imagination ability is now possessed by our foundation model. We believe that our work makes a transformative stride towards AGI, from our common practice of "weak or narrow AI" to that of "strong or generalized AI".
['Ji-Rong Wen', 'Hao Sun', 'Tao Xiang', 'Xin Gao', 'Ruihua Song', 'Haoyu Lu', 'Jingyuan Wen', 'Yuqi Huo', 'Guoxing Yang', 'Yizhao Gao', 'Zhiwu Lu', 'Nanyi Fei']
2021-10-27
null
null
null
null
['visual-commonsense-reasoning']
['reasoning']
[ 3.11603874e-01 4.80048597e-01 1.41378984e-01 -4.04006064e-01 4.98929136e-02 -4.57328916e-01 8.32569122e-01 -2.15685606e-01 -2.31456712e-01 6.23727202e-01 1.55356348e-01 -1.70355707e-01 -6.08200371e-01 -9.06792223e-01 -4.59874660e-01 -2.89731175e-01 1.45427123e-01 9.17105377e-01 3.49377990e-02 -7.68689334e-01 3.46811414e-01 2.03149900e-01 -1.58729351e+00 4.49843228e-01 1.11390662e+00 7.35204160e-01 3.42995107e-01 3.50294083e-01 -5.53118475e-02 9.88411069e-01 -4.82813746e-01 -6.77045643e-01 -1.29716992e-02 -5.13111234e-01 -1.32192409e+00 1.56952828e-01 -2.41988063e-01 1.17863521e-01 -1.56070128e-01 1.09000826e+00 2.93697208e-01 -2.46040951e-02 5.49207926e-01 -1.54982686e+00 -1.06754863e+00 7.38295794e-01 -3.34029645e-02 -8.55101049e-02 4.76538509e-01 1.66724190e-01 9.37308669e-01 -4.56434190e-01 5.81478894e-01 1.31813860e+00 5.64596891e-01 7.58163452e-01 -1.05882251e+00 -4.62309897e-01 7.46013150e-02 4.64419425e-01 -1.01884282e+00 -1.32718548e-01 9.76522624e-01 -2.86060214e-01 7.66996980e-01 1.42645285e-01 7.78107166e-01 1.28154981e+00 -8.88644159e-02 9.36362505e-01 1.44702923e+00 -6.62463784e-01 2.03013793e-01 2.39359781e-01 3.29688370e-01 8.92855525e-01 2.20577016e-01 3.41133438e-02 -9.72946107e-01 3.73131067e-01 6.13666177e-01 -1.99601706e-02 6.57744333e-02 -1.90895930e-01 -1.50167811e+00 6.98847950e-01 4.30338204e-01 6.24256372e-01 -3.14408362e-01 -7.36678764e-02 1.21121928e-01 5.84302306e-01 9.39182714e-02 9.30906534e-01 -2.70269632e-01 -2.00012654e-01 -5.28555632e-01 1.12769708e-01 8.92459393e-01 8.96168530e-01 7.24655807e-01 -2.19467089e-01 4.53107916e-02 7.71746576e-01 2.15780899e-01 5.43581426e-01 8.63551855e-01 -1.35749674e+00 1.45577535e-01 1.15938962e+00 -2.02897489e-02 -9.27281320e-01 -6.54172122e-01 -7.57159173e-01 -7.52766907e-01 2.79616505e-01 4.84230876e-01 -1.08728915e-01 -6.89745843e-01 1.79151976e+00 1.90845563e-03 -2.09193677e-01 4.49987024e-01 8.08137596e-01 6.09007776e-01 3.56879294e-01 8.69561881e-02 7.40682036e-02 1.40785456e+00 -1.07504857e+00 -6.22992814e-01 -6.14070714e-01 8.32026243e-01 -3.75041485e-01 1.49024975e+00 6.35119796e-01 -1.16279209e+00 -7.73895085e-01 -1.10418129e+00 4.42265905e-02 -8.21423709e-01 -6.26990572e-02 1.13767338e+00 6.19918406e-01 -1.19816637e+00 2.78008491e-01 -3.74445587e-01 -6.82100594e-01 2.29355752e-01 3.44008416e-01 -5.00114620e-01 -2.61303037e-01 -1.26975131e+00 1.10121584e+00 8.09555709e-01 -8.79249871e-02 -7.13524222e-01 -2.18108058e-01 -4.47809905e-01 -2.28856108e-03 6.40663564e-01 -1.17135954e+00 9.49551702e-01 -1.24158978e+00 -1.49757946e+00 9.68467534e-01 -1.74121894e-02 -4.68976140e-01 2.86850214e-01 -1.59630328e-01 -4.14364398e-01 7.78623894e-02 -2.44081989e-02 7.09384084e-01 5.45907855e-01 -1.63630438e+00 -5.93033135e-01 -6.15546107e-01 4.25612181e-01 2.93347061e-01 -8.70108724e-01 -3.27656381e-02 -3.50827813e-01 -5.14227033e-01 1.25998491e-02 -1.11433017e+00 -4.00870889e-02 -4.37107950e-01 -2.15010241e-01 -4.84654188e-01 5.50161719e-01 -3.39020014e-01 1.17255485e+00 -1.90932500e+00 4.17146236e-01 1.70685023e-01 2.92107910e-01 2.93281317e-01 -3.07562351e-01 2.99508363e-01 2.49718443e-01 -5.22017553e-02 -1.02222621e-01 -1.26094326e-01 2.47725353e-01 5.74150681e-01 -1.28218709e-02 -4.54354346e-01 4.80912663e-02 1.23683274e+00 -1.13345945e+00 -4.10735369e-01 -3.52137871e-02 -3.89890336e-02 -6.90906286e-01 4.02565688e-01 -2.58811384e-01 4.67634171e-01 -5.75241148e-01 5.28608441e-01 1.89398304e-01 -6.64507926e-01 3.14386189e-01 -1.02473080e-01 1.58423632e-01 -3.93051878e-02 -6.63286626e-01 2.05472589e+00 -4.54154164e-01 3.31815720e-01 -6.46095276e-02 -1.44350469e+00 1.02999699e+00 1.55908659e-01 3.49622548e-01 -9.43852842e-01 1.44647360e-01 -1.22266291e-02 4.81767833e-01 -7.15748429e-01 9.32312906e-02 -2.25466982e-01 -1.52381837e-01 5.10838747e-01 1.09064236e-01 -3.99840415e-01 4.52132612e-01 2.16173843e-01 1.03988791e+00 3.12348127e-01 5.76477237e-02 -3.58542711e-01 7.66735077e-01 5.02850831e-01 2.87816018e-01 8.47323060e-01 -1.60070896e-01 1.88598007e-01 3.92305672e-01 -6.37726665e-01 -8.47548723e-01 -9.17979062e-01 3.12117636e-01 1.51125467e+00 5.89694567e-02 -4.66602951e-01 -1.04053080e+00 -5.79237342e-01 -4.21074510e-01 9.19081569e-01 -4.97975022e-01 -4.45121497e-01 -2.36436516e-01 -8.82617354e-01 4.95168775e-01 4.78199482e-01 1.08378863e+00 -1.31254923e+00 -5.25051594e-01 1.20834857e-01 -3.42118770e-01 -1.09261692e+00 4.10200328e-01 4.97150943e-02 -7.19545305e-01 -6.99094713e-01 -2.88345426e-01 -7.99416304e-01 4.65067148e-01 3.80628884e-01 1.25858057e+00 3.53047937e-01 -4.80245054e-02 5.34444273e-01 -5.76312482e-01 -8.06135237e-01 -5.40739596e-01 7.67454803e-02 3.88344117e-02 -1.58360735e-01 6.46525562e-01 -6.47739649e-01 -4.97352242e-01 4.38158482e-01 -9.55482900e-01 5.65442383e-01 8.66560757e-01 6.67782068e-01 -2.06230525e-02 4.16172504e-01 8.92756224e-01 -6.95193827e-01 9.70739484e-01 -4.06493336e-01 4.54239734e-02 5.89381456e-01 -8.55141103e-01 1.74594939e-01 7.08215058e-01 -2.55766869e-01 -1.43414664e+00 -1.76325560e-01 2.90676504e-02 1.08326904e-01 -4.07067239e-01 8.49595726e-01 -2.54959404e-01 9.06477869e-02 8.08119178e-01 3.36562693e-01 2.27094218e-01 -3.33989978e-01 3.78359288e-01 6.59011006e-01 7.05125690e-01 -9.49340522e-01 7.34895408e-01 5.06033063e-01 1.32613748e-01 -5.47606826e-01 -1.23754847e+00 3.51698734e-02 -6.87073171e-01 -4.37983155e-01 8.73669684e-01 -6.50534987e-01 -9.07673359e-01 2.24853933e-01 -8.55704844e-01 -5.15189290e-01 1.26305474e-02 3.77716303e-01 -6.08792961e-01 3.82916331e-01 -2.89337814e-01 -6.25022292e-01 -2.07295895e-01 -8.14066470e-01 6.70066059e-01 2.93794274e-01 -4.11700904e-01 -1.30726886e+00 2.54444629e-02 1.07795382e+00 5.26571751e-01 -1.88246220e-01 1.17942786e+00 -7.96173096e-01 -2.82034129e-01 -9.27411243e-02 -3.34239811e-01 4.30183619e-01 1.98599533e-03 -3.80139619e-01 -8.49966764e-01 -2.30261125e-02 4.36647646e-02 -6.65521622e-01 4.77959663e-01 -1.76259473e-01 1.07617831e+00 -9.89696942e-03 -1.80685297e-01 2.91570604e-01 1.13942254e+00 3.95756185e-01 6.66796029e-01 6.64642513e-01 4.59284693e-01 8.76140594e-01 7.21864402e-01 2.62954384e-01 7.87946403e-01 4.71256793e-01 3.39108974e-01 -9.36721042e-02 -2.36536011e-01 -2.33623236e-01 2.75456578e-01 8.34035397e-01 -7.10180938e-01 -1.50030464e-01 -1.33115733e+00 3.78940314e-01 -2.10705018e+00 -8.94910872e-01 1.00584537e-01 1.79553676e+00 7.80348778e-01 2.83198953e-01 4.50813808e-02 1.20512702e-01 2.70210534e-01 -4.14394468e-01 -2.85836726e-01 -3.02368879e-01 -1.31036326e-01 1.88252963e-02 -2.61645228e-01 1.75673574e-01 -7.78004825e-01 1.31501830e+00 6.89353037e+00 5.99824667e-01 -8.70685875e-01 2.45496854e-02 3.87287199e-01 2.10514992e-01 -2.95524597e-01 1.32893808e-02 -4.74478781e-01 3.82294059e-01 1.01431727e+00 -3.07900369e-01 8.49976003e-01 7.05362320e-01 1.54092416e-01 1.42127782e-01 -1.14139032e+00 7.63090909e-01 2.27870926e-01 -1.10876024e+00 3.14674377e-01 1.60496607e-02 5.29428780e-01 -3.03317964e-01 1.53680053e-02 6.05263293e-01 5.48093021e-01 -1.16322184e+00 4.08191025e-01 8.17883193e-01 2.25898579e-01 -6.14261448e-01 7.54403234e-01 7.13786721e-01 -6.14887834e-01 -3.24175924e-01 -5.32102466e-01 -5.91277719e-01 -1.19630732e-01 2.99301267e-01 -7.45012343e-01 7.02321470e-01 6.19600892e-01 5.74211419e-01 -8.81431401e-01 5.32515168e-01 -3.25923324e-01 4.75327432e-01 -9.28686112e-02 -9.94397998e-02 1.73336387e-01 -3.22746217e-01 2.27162480e-01 1.05050099e+00 3.26471061e-01 1.83117777e-01 2.15451837e-01 7.85452843e-01 1.44034415e-01 1.54527659e-02 -8.37051392e-01 -2.93841004e-01 1.96214750e-01 1.23249459e+00 -7.92673886e-01 -5.35289526e-01 -5.78097939e-01 9.93547201e-01 4.49581116e-01 2.73877591e-01 -7.43792713e-01 -1.78578272e-01 2.42565960e-01 -2.57162839e-01 -2.49185309e-01 -3.08156967e-01 -6.51923537e-01 -1.29358602e+00 -3.77580762e-01 -1.04777169e+00 5.44675469e-01 -1.39028716e+00 -1.32192087e+00 5.99336028e-01 1.28556088e-01 -7.72442937e-01 -4.94119763e-01 -1.00748861e+00 -5.22287846e-01 4.94990319e-01 -1.16712987e+00 -1.69654655e+00 -7.12563097e-01 7.88710833e-01 4.41277146e-01 -5.03545761e-01 9.26368773e-01 -7.53912553e-02 -4.32376951e-01 2.87297100e-01 -3.36066991e-01 2.84497440e-02 4.78372574e-01 -9.95631278e-01 1.91296805e-02 6.67924941e-01 2.21332267e-01 7.51460016e-01 8.29463422e-01 -5.00900924e-01 -1.55296266e+00 -5.73452234e-01 7.39672840e-01 -6.53544486e-01 9.38821793e-01 -1.36963904e-01 -7.64577866e-01 8.00242245e-01 4.12598312e-01 -5.90599775e-01 7.84677982e-01 3.01684499e-01 -1.34965286e-01 -1.90528795e-01 -8.48492682e-01 7.06886590e-01 1.62503684e+00 -2.25907862e-01 -1.22369146e+00 5.31961262e-01 5.90266407e-01 3.20195317e-01 -9.91014183e-01 4.86804754e-01 3.82112294e-01 -1.10471737e+00 1.17498744e+00 -8.90565634e-01 6.92551196e-01 -1.62198424e-01 1.02240779e-02 -1.47396886e+00 -6.03650749e-01 -5.22350311e-01 -7.61926733e-03 1.12265277e+00 4.22074646e-01 -7.08083808e-01 8.54619920e-01 8.86121452e-01 -2.49124974e-01 -3.42960745e-01 -5.60164630e-01 -9.20709014e-01 7.88543001e-02 -6.21921599e-01 5.15306234e-01 9.93065894e-01 6.51942015e-01 8.08535397e-01 -2.62136847e-01 -1.21175773e-01 5.48104048e-01 2.71546781e-01 9.18784976e-01 -1.70006037e+00 -2.81827956e-01 -4.45766538e-01 -3.08973432e-01 -8.13876867e-01 3.05373073e-01 -9.71505761e-01 -3.78519773e-01 -1.60797441e+00 4.64904636e-01 -2.23497763e-01 -5.29576004e-01 8.58469009e-01 -9.68881249e-02 3.57130945e-01 1.14299349e-01 3.09126496e-01 -7.79128790e-01 2.67389357e-01 1.50922859e+00 -7.36674760e-03 4.15399559e-02 -2.02532411e-01 -1.22162104e+00 1.02971053e+00 8.54157209e-01 -1.60750002e-02 -8.60903621e-01 -4.95527834e-01 6.07229888e-01 -1.93264708e-01 4.69419032e-01 -1.27365327e+00 3.95348966e-01 -3.20404261e-01 1.79723918e-01 -8.43381062e-02 3.59147638e-01 -8.68739307e-01 4.10605185e-02 5.07580638e-01 -3.81509781e-01 1.65079013e-02 -6.80965930e-02 2.49448761e-01 -2.95068622e-01 -3.33962679e-01 4.36507195e-01 -4.39104438e-01 -1.13627160e+00 -2.38324732e-01 -3.41753185e-01 -2.76924111e-02 9.43570733e-01 -2.08183408e-01 -4.58135217e-01 -4.05690312e-01 -8.23446631e-01 1.86845422e-01 4.63187963e-01 4.89544570e-01 5.23857176e-01 -1.00485539e+00 -6.69162571e-01 1.08193874e-01 3.02012175e-01 -3.14420640e-01 1.69191156e-02 7.34725416e-01 -3.59507233e-01 7.91600168e-01 -6.46721065e-01 -2.73622215e-01 -1.01760530e+00 8.88365507e-01 2.16704700e-03 -1.80617958e-01 -5.26366115e-01 5.10470808e-01 2.78741151e-01 -3.54840398e-01 1.25414833e-01 8.95245075e-02 -3.65974188e-01 -2.50090510e-01 4.37373251e-01 2.18985528e-01 -2.38260418e-01 -3.23586106e-01 -1.09164909e-01 3.48383784e-01 9.75469574e-02 -2.49445572e-01 1.46376836e+00 -2.69295543e-01 -3.46118182e-01 2.40841985e-01 5.97722173e-01 -3.13219726e-01 -9.70433712e-01 -1.29316375e-01 2.48510659e-01 -1.19673520e-01 -8.80944878e-02 -1.31932831e+00 -5.87234139e-01 9.23247457e-01 3.06117088e-01 5.34704983e-01 1.30476320e+00 2.52172172e-01 5.05795896e-01 1.02704084e+00 6.48010671e-01 -1.34024060e+00 4.81088638e-01 6.27715766e-01 7.99287021e-01 -1.36628807e+00 -2.38186300e-01 -2.28528708e-01 -9.96394575e-01 1.11785007e+00 7.30619252e-01 1.16481356e-01 2.54155129e-01 1.07417233e-01 1.60669461e-01 -5.20771205e-01 -8.43087673e-01 -4.93390292e-01 3.32155406e-01 8.99252534e-01 5.43980241e-01 -3.43656763e-02 -4.12668377e-01 7.31107533e-01 -5.21143734e-01 3.88569772e-01 2.51541048e-01 7.67516613e-01 -6.10955954e-01 -1.19178748e+00 -3.14972699e-01 2.90010888e-02 -3.74513119e-02 1.72498003e-02 -8.12666297e-01 1.01246786e+00 2.46896774e-01 1.10364592e+00 -2.08742499e-01 -5.30355811e-01 2.38013476e-01 5.09396493e-01 5.18358111e-01 -4.09509003e-01 -3.88742000e-01 -3.76848161e-01 2.77348340e-01 -4.79301304e-01 -7.51448035e-01 -2.85130590e-01 -1.19766092e+00 -3.88356477e-01 3.03812027e-01 6.70740455e-02 6.30999386e-01 1.31523073e+00 3.17137837e-01 5.05797744e-01 6.87615797e-02 -5.90416849e-01 -3.09736282e-01 -9.03237939e-01 -2.23063752e-01 8.65606129e-01 -5.60166299e-01 -7.27027595e-01 -1.76414028e-01 2.43950620e-01]
[9.163100242614746, 6.494417667388916]
ced402b5-8232-4e44-b3d6-dc918feb8a9b
decoding-anagrammed-texts-written-in-an
null
null
https://aclanthology.org/Q16-1006
https://aclanthology.org/Q16-1006.pdf
Decoding Anagrammed Texts Written in an Unknown Language and Script
Algorithmic decipherment is a prime example of a truly unsupervised problem. The first step in the decipherment process is the identification of the encrypted language. We propose three methods for determining the source language of a document enciphered with a monoalphabetic substitution cipher. The best method achieves 97{\%} accuracy on 380 languages. We then present an approach to decoding anagrammed substitution ciphers, in which the letters within words have been arbitrarily transposed. It obtains the average decryption word accuracy of 93{\%} on a set of 50 ciphertexts in 5 languages. Finally, we report the results on the Voynich manuscript, an unsolved fifteenth century cipher, which suggest Hebrew as the language of the document.
['Grzegorz Kondrak', 'Bradley Hauer']
2016-01-01
null
null
null
tacl-2016-1
['decipherment']
['natural-language-processing']
[ 4.98503506e-01 -3.70870620e-01 5.88833466e-02 -7.11555928e-02 -6.50357664e-01 -1.11941671e+00 1.03635859e+00 -7.37796500e-02 -9.95816588e-01 1.02776015e+00 1.72794238e-01 -9.88758445e-01 2.40635619e-01 -6.36231244e-01 -6.19070590e-01 -8.59871030e-01 2.89423317e-02 4.74185795e-01 -4.83011276e-01 -4.89633352e-01 5.34282148e-01 3.75191927e-01 -1.14281940e+00 6.45741701e-01 6.72859669e-01 6.22749269e-01 1.24946766e-01 1.38397992e+00 -1.60769626e-01 6.47247195e-01 -9.57992136e-01 -1.29720199e+00 3.11070204e-01 -5.09129584e-01 -1.11382031e+00 -3.93467456e-01 -4.96976525e-02 -2.82366604e-01 -3.47949088e-01 1.27143574e+00 1.84756190e-01 -8.31316054e-01 8.78189206e-01 -9.28227484e-01 -2.46067196e-01 1.00340521e+00 1.33388445e-01 -1.41148055e-02 6.35189950e-01 -1.36789292e-01 1.06193292e+00 -7.04125941e-01 5.62198579e-01 9.47616994e-01 5.52876651e-01 1.80406481e-01 -8.38528514e-01 -9.95811582e-01 -8.76437128e-01 2.93129563e-01 -1.83976007e+00 -5.34655035e-01 3.87809873e-01 -3.24255168e-01 8.76624346e-01 7.08402872e-01 6.91224217e-01 9.19151187e-01 5.79149544e-01 5.53801894e-01 1.46745396e+00 -8.28432083e-01 -7.89877623e-02 3.03379655e-01 -5.36677577e-02 7.15852737e-01 5.50676942e-01 1.74472854e-01 -5.05195022e-01 -5.99587023e-01 -1.77341662e-02 -5.12210906e-01 -2.40418121e-01 4.38997477e-01 -1.35952616e+00 8.11312616e-01 -3.64546627e-01 2.83617646e-01 3.84084322e-03 2.22866178e-01 5.22109568e-01 7.55468249e-01 2.56063551e-01 6.64569736e-01 -3.97589058e-01 -6.71310008e-01 -9.90557611e-01 -9.29919779e-02 1.23362112e+00 8.81051004e-01 5.54706454e-01 2.06109229e-03 5.97821534e-01 3.14101040e-01 3.60002331e-02 7.11111307e-01 4.60107327e-01 -3.88969690e-01 6.69917405e-01 1.12214103e-01 -8.87382701e-02 -6.22618496e-01 1.61911994e-02 -2.28621319e-01 -9.00305986e-01 -9.68410894e-02 4.52417940e-01 -2.53895938e-01 -5.08776963e-01 1.14037311e+00 -2.36203820e-01 -3.16295773e-01 7.11085439e-01 4.57657993e-01 5.72752118e-01 7.23952472e-01 -2.58569241e-01 -3.22577983e-01 1.59158063e+00 -2.60735363e-01 -7.04739928e-01 3.65759015e-01 9.15506124e-01 -1.37690055e+00 3.43703181e-01 9.23484981e-01 -8.51810336e-01 -2.27714360e-01 -1.28111708e+00 8.98739137e-03 -5.31211913e-01 1.62390009e-01 5.54527998e-01 1.32880723e+00 -8.43801796e-01 4.61139321e-01 -5.29567376e-02 7.99696892e-02 -3.30911689e-02 5.87434113e-01 -6.12481117e-01 2.47379497e-01 -1.33972669e+00 8.98651540e-01 9.35550034e-01 -4.69194874e-02 -6.66438699e-01 -1.53561190e-01 -4.16181594e-01 -2.28476390e-01 -1.54278949e-01 -4.29652035e-01 9.26436782e-01 -1.19551802e+00 -1.43930650e+00 1.27111793e+00 -3.43973488e-01 -7.06158102e-01 5.82683921e-01 1.32041976e-01 -8.77820909e-01 2.88900733e-01 -2.66507983e-01 6.15494512e-02 1.19826818e+00 -1.17614341e+00 -7.61315703e-01 5.48317656e-03 -3.93867284e-01 -7.54000023e-02 -3.33179832e-01 3.02360237e-01 -1.07041471e-01 -9.73091424e-01 -1.57889113e-01 -9.82522547e-01 2.28376210e-01 -7.19592214e-01 -6.74535215e-01 -6.71056807e-02 5.47765493e-01 -1.09376585e+00 1.53978586e+00 -2.17702270e+00 -5.38310707e-02 6.67795718e-01 3.25710684e-01 3.99375975e-01 8.46808702e-02 9.65492427e-01 -3.58641684e-01 3.22613746e-01 -4.21225250e-01 -2.54347950e-01 4.32817414e-02 3.02247196e-01 -8.28692079e-01 7.54183233e-01 -3.70822012e-01 9.16561186e-01 -8.73824298e-01 -2.73825616e-01 -2.38350213e-01 1.44796401e-01 -1.52041614e-01 -7.18584750e-03 1.54202819e-01 8.26795306e-03 -2.12900955e-02 5.89264631e-01 8.57229352e-01 1.85709968e-01 5.62651932e-01 1.22488491e-01 -1.44010291e-01 3.95828366e-01 -9.14194405e-01 1.17742836e+00 -2.07807884e-01 1.07326376e+00 -1.31261334e-01 -5.27296364e-01 9.74828839e-01 4.38421220e-01 2.92191297e-01 -3.36461395e-01 2.41543308e-01 6.17108524e-01 3.10669035e-01 -1.35326415e-01 1.15041840e+00 -1.42849877e-01 -4.50727791e-01 9.53645587e-01 -1.24369271e-01 -5.66021919e-01 1.58727333e-01 5.17861724e-01 8.61458123e-01 -5.84680378e-01 6.56882524e-01 -5.35299718e-01 7.89538801e-01 -9.21623185e-02 1.02003008e-01 7.34052956e-01 2.60228753e-01 4.11546052e-01 5.49479485e-01 -6.11558080e-01 -1.50873077e+00 -9.09571052e-01 -6.40768483e-02 6.24962270e-01 -1.87801287e-01 -1.25624108e+00 -8.40149701e-01 -3.17879081e-01 -7.35993907e-02 7.26099133e-01 -6.92619607e-02 -4.22588810e-02 -5.42782784e-01 -8.15451503e-01 1.57589877e+00 -3.44943196e-01 6.20335698e-01 -6.66846752e-01 -2.11221248e-01 -2.00403128e-02 -3.55403095e-01 -1.28408480e+00 -2.14265749e-01 4.02306557e-01 -5.14394879e-01 -1.22215056e+00 -5.17852306e-01 -8.90521169e-01 8.13164175e-01 -2.28778139e-01 8.87871683e-01 3.29027563e-01 -5.99694550e-02 1.35719195e-01 -4.05137837e-01 -4.60977226e-01 -1.28767467e+00 2.29913935e-01 3.14317286e-01 -4.02745269e-02 4.54799712e-01 -2.42291093e-01 -2.97964662e-01 -1.29314512e-01 -8.55790377e-01 -4.28085774e-02 3.90196890e-01 6.96361661e-01 1.03213109e-01 3.87883067e-01 -3.35643917e-01 -1.01133347e+00 9.17902768e-01 -2.22548455e-01 -7.52342463e-01 3.54467541e-01 -5.63711166e-01 3.33091646e-01 1.22176528e+00 -9.75836292e-02 -3.62050235e-01 6.73643053e-02 -3.14313620e-01 3.76888603e-01 2.27582231e-01 1.04463726e-01 1.31300129e-02 -4.50419545e-01 5.32407165e-01 1.09298384e+00 -1.08774297e-03 -2.86495715e-01 2.89225578e-01 1.27430785e+00 8.97281408e-01 -5.86475253e-01 1.07642078e+00 3.96034628e-01 -5.80421723e-02 -1.25167584e+00 1.01453394e-01 -2.57811248e-01 -5.69554567e-01 7.25921765e-02 4.55527961e-01 -9.81508553e-01 -9.78018939e-01 7.46347129e-01 -1.44925714e+00 2.33714003e-03 1.00646302e-01 5.60362697e-01 -5.30841947e-01 1.15962994e+00 -4.12196904e-01 -8.06823730e-01 -6.55195832e-01 -1.03560364e+00 9.54106331e-01 -5.40590227e-01 -7.47824669e-01 -8.93372953e-01 -3.92318098e-03 2.60118127e-01 -2.77684182e-01 -1.96980566e-01 1.24641657e+00 -7.04227865e-01 -2.38964379e-01 -2.67306834e-01 1.33116886e-01 5.18872440e-01 2.73156703e-01 7.54430220e-02 -6.88100338e-01 -4.75163341e-01 7.73331076e-02 -5.65881170e-02 8.55831027e-01 -5.95483840e-01 1.32640624e+00 -6.36548758e-01 6.60486221e-02 1.13500524e+00 1.33285367e+00 3.21916789e-01 1.03639662e+00 3.60900432e-01 3.73934507e-01 2.72278935e-01 1.07107162e-01 6.42512083e-01 3.04693282e-01 1.09421320e-01 4.37300242e-02 3.55745971e-01 3.04702103e-01 -4.71940279e-01 6.44918740e-01 1.52261555e+00 -2.50428289e-01 -4.22537059e-01 -1.02522278e+00 1.20956331e-01 -9.20278549e-01 -1.04245365e+00 -3.54992568e-01 2.16814780e+00 9.56080854e-01 6.22276403e-02 -1.29851237e-01 7.87847042e-01 4.85171229e-01 9.63432863e-02 2.57232338e-01 -1.10934246e+00 -6.46971166e-01 7.61856496e-01 1.13974500e+00 9.65445876e-01 -9.85864103e-01 1.13023877e+00 7.89951706e+00 1.10764980e+00 -9.66899455e-01 -3.82027209e-01 5.03991067e-01 5.02121866e-01 -7.00629830e-01 3.17621768e-01 -6.57233000e-01 9.77099359e-01 1.14864373e+00 -2.94106483e-01 8.50463748e-01 1.99881151e-01 -7.65025690e-02 -3.34986418e-01 -9.25025344e-01 1.35579622e+00 3.17723125e-01 -1.23816454e+00 2.48504445e-01 2.11735696e-01 5.55347562e-01 -2.55661517e-01 1.50837272e-01 -2.91946858e-01 3.32748115e-01 -1.31219959e+00 7.78232336e-01 5.31794786e-01 1.31894636e+00 -1.03292358e+00 5.42828619e-01 6.06355071e-01 -8.74983966e-01 1.17247820e-01 -3.13899100e-01 -1.21878810e-01 -3.34944099e-01 2.79050797e-01 -1.17671239e+00 4.89742637e-01 1.58636808e-01 2.19408378e-01 -5.28514206e-01 5.50059199e-01 -5.32709420e-01 7.87693858e-01 -3.17997426e-01 -6.42518401e-01 3.70648026e-01 -6.15389168e-01 6.07186377e-01 1.68275118e+00 4.10660595e-01 3.52003455e-01 -4.45368320e-01 -1.01210857e-02 -3.53484452e-01 1.20969243e-01 -7.32739151e-01 -3.06223571e-01 5.10617256e-01 6.86878264e-01 -6.00199461e-01 -6.09303713e-01 2.41234615e-01 1.43941987e+00 -1.97166264e-01 2.26896212e-01 -3.84702444e-01 -7.08757579e-01 6.47622228e-01 -2.99135566e-01 4.00149167e-01 -6.01087809e-01 -4.00329143e-01 -1.20365393e+00 -1.09099075e-01 -1.58148396e+00 2.15902507e-01 -5.30928254e-01 -7.56964505e-01 5.10533035e-01 -2.48609200e-01 -1.17270637e+00 -4.08737391e-01 -9.30495620e-01 -3.38664621e-01 8.49324584e-01 -1.11433744e+00 -7.75639892e-01 3.13339859e-01 6.60276115e-01 -1.99951101e-02 -9.54086304e-01 1.24172592e+00 8.05493519e-02 7.31433406e-02 8.36321533e-01 8.86760771e-01 4.21370357e-01 5.03398180e-01 -1.12315488e+00 6.81557238e-01 8.53034854e-01 5.73018372e-01 8.89141500e-01 7.33496487e-01 -5.59499860e-01 -1.88217199e+00 -6.46067321e-01 1.68773675e+00 -4.55301553e-01 7.14739621e-01 -8.47060740e-01 -1.67384878e-01 4.03948635e-01 3.56521547e-01 -8.36961448e-01 8.95289063e-01 -4.76738274e-01 -5.21247447e-01 -1.28375692e-02 -8.94535005e-01 7.87815988e-01 7.98067391e-01 -1.29832184e+00 -5.53764820e-01 5.53789794e-01 2.68674433e-01 -1.63335681e-01 -6.87484562e-01 -2.24442989e-01 8.76996636e-01 -7.33043313e-01 7.81836867e-01 -5.61740160e-01 4.22100306e-01 -2.16128334e-01 -3.32096040e-01 -1.06298840e+00 1.13752551e-01 -1.51125169e+00 -5.64957894e-02 5.94007194e-01 3.27909082e-01 -9.58396137e-01 8.02423060e-01 -1.62571326e-01 3.86649489e-01 -1.94835886e-01 -1.14534628e+00 -7.25627780e-01 2.94982791e-01 -8.43860269e-01 8.70558321e-01 9.44082797e-01 -7.60884807e-02 2.95402221e-02 -7.76084483e-01 -1.74433157e-01 7.03757107e-01 -6.40895069e-02 1.02314007e+00 -6.37172103e-01 -6.67942837e-02 -2.53258437e-01 -6.58907533e-01 -1.12004745e+00 4.04638618e-01 -1.36110580e+00 -4.35523003e-01 -5.45393527e-01 -3.18668671e-02 -2.86586404e-01 1.35590419e-01 -9.40006524e-02 2.53071338e-01 5.23181200e-01 1.29688427e-01 2.25710034e-01 -6.41692579e-02 2.20039442e-01 7.68854022e-01 -4.04869497e-01 4.58342224e-01 1.19451573e-03 -7.25272417e-01 6.34773791e-01 9.47781444e-01 -5.78603446e-01 -6.84473366e-02 -4.53267574e-01 7.26606905e-01 -3.54416937e-01 1.85441419e-01 -6.52404964e-01 5.35746217e-01 2.77319223e-01 3.49538118e-01 -8.39218438e-01 3.80229443e-01 -9.03879821e-01 4.08321857e-01 1.00554788e+00 -9.39229652e-02 7.23226190e-01 -3.12857926e-01 1.37500271e-01 -2.51970559e-01 -3.43223423e-01 5.26972830e-01 -1.40893072e-01 -8.01090658e-01 3.12777124e-02 -8.32552016e-01 -1.63928848e-02 8.06576192e-01 -4.47439224e-01 -1.49017975e-01 -4.27280515e-01 -1.94149867e-01 -3.04156661e-01 5.31917155e-01 9.59894955e-02 8.62811327e-01 -9.25097704e-01 -1.02374780e+00 5.03203750e-01 2.65437718e-02 -9.48534071e-01 -5.72012067e-01 3.59925210e-01 -1.45753157e+00 6.64257884e-01 -1.29765430e-02 -2.97037989e-01 -1.81662941e+00 1.78375691e-01 4.72614877e-02 -8.80156457e-02 -2.35777318e-01 6.79302275e-01 -5.23929298e-01 -1.73642620e-01 -1.20688796e-01 1.14693148e-02 1.72247261e-01 7.68559277e-02 6.02691770e-01 3.65556538e-01 2.45143741e-01 -1.06240058e+00 -3.94324422e-01 5.75818419e-01 -5.07593229e-02 -3.93315166e-01 8.25726211e-01 2.01865241e-01 -9.10256743e-01 1.26642123e-01 1.79413867e+00 4.42044884e-01 5.77229634e-02 -2.97237672e-02 7.06971288e-02 -4.91692185e-01 -4.72859412e-01 -4.52499181e-01 -3.43693644e-01 4.90657449e-01 2.28446960e-01 6.03946634e-02 1.04348624e+00 -4.48969901e-01 1.00048375e+00 1.03026724e+00 4.64666784e-01 -1.25716734e+00 -5.55781007e-01 9.87393677e-01 3.52114230e-01 -8.22552443e-01 4.86976385e-01 -2.33148843e-01 -5.02570391e-01 1.62156284e+00 -5.53270936e-01 -1.12959675e-01 5.53847492e-01 4.65281546e-01 1.12038488e-02 1.71707749e-01 -6.36332870e-01 1.79832235e-01 1.31199481e-02 4.39327151e-01 1.51437864e-01 4.00492340e-01 -8.27506304e-01 4.02666807e-01 -1.33334196e+00 -4.29444015e-01 5.53708971e-01 8.47410738e-01 -7.78337196e-02 -1.45593357e+00 -7.18706846e-01 2.79631555e-01 -8.41571748e-01 -7.19052851e-01 -1.06647742e+00 7.28340864e-01 2.51884282e-01 8.28280389e-01 -8.18719640e-02 -8.54177713e-01 -3.02123815e-01 1.07579663e-01 5.49417317e-01 -5.23166098e-02 -1.06464469e+00 -4.46026593e-01 4.42084968e-01 9.14881304e-02 -1.70891628e-01 -4.46634233e-01 -9.11756814e-01 -1.18463004e+00 -1.43979033e-02 8.05058837e-01 7.59815931e-01 8.49416256e-01 -1.62197873e-01 -2.83920944e-01 1.16195297e+00 5.58212325e-02 -5.90459108e-01 -4.39758599e-01 -6.92430437e-01 1.87968433e-01 4.32436109e-01 4.44139004e-01 -3.51886779e-01 2.18286842e-01]
[10.23220443725586, 10.515986442565918]
ccf42a9b-6df1-4f32-8c0c-573ae06ed0e5
deep-iterative-frame-interpolation-for-full
1909.02641
null
https://arxiv.org/abs/1909.02641v1
https://arxiv.org/pdf/1909.02641v1.pdf
Deep Iterative Frame Interpolation for Full-frame Video Stabilization
Video stabilization is a fundamental and important technique for higher quality videos. Prior works have extensively explored video stabilization, but most of them involve cropping of the frame boundaries and introduce moderate levels of distortion. We present a novel deep approach to video stabilization which can generate video frames without cropping and low distortion. The proposed framework utilizes frame interpolation techniques to generate in between frames, leading to reduced inter-frame jitter. Once applied in an iterative fashion, the stabilization effect becomes stronger. A major advantage is that our framework is end-to-end trainable in an unsupervised manner. In addition, our method is able to run in near real-time (15 fps). To the best of our knowledge, this is the first work to propose an unsupervised deep approach to full-frame video stabilization. We show the advantages of our method through quantitative and qualitative evaluations comparing to the state-of-the-art methods.
['Jinsoo Choi', 'In So Kweon']
2019-09-05
null
null
null
null
['video-stabilization']
['computer-vision']
[ 1.18202396e-01 8.59299153e-02 -7.15099722e-02 -1.18855521e-01 -5.61283946e-01 -4.86801535e-01 4.21010464e-01 9.07292962e-03 -3.19276929e-01 7.15165675e-01 2.17444211e-01 -9.96381491e-02 4.10165966e-01 -4.97742474e-01 -1.04593956e+00 -4.84953582e-01 -6.58617318e-02 -2.62641162e-01 5.55233121e-01 -2.84210920e-01 2.51879603e-01 4.08299029e-01 -1.68669593e+00 2.30338633e-01 7.66263366e-01 7.82159388e-01 8.16484243e-02 7.05968082e-01 4.29449528e-01 1.07165265e+00 -3.62395495e-01 -2.50675857e-01 3.84725571e-01 -5.42879701e-01 -7.38503277e-01 4.66391057e-01 6.45540237e-01 -7.84896016e-01 -4.17100400e-01 8.26842129e-01 5.27348697e-01 2.55330443e-01 1.71943277e-01 -8.85315597e-01 -5.02853215e-01 4.98750687e-01 -6.31996512e-01 2.14098543e-01 5.73205888e-01 3.70061010e-01 5.52335560e-01 -7.02037871e-01 7.12034881e-01 1.11463356e+00 6.07425451e-01 7.04804480e-01 -1.39653647e+00 -4.23563242e-01 1.56994894e-01 1.53901735e-02 -1.08392537e+00 -9.63231206e-01 6.83676422e-01 -3.57404619e-01 8.11225832e-01 -2.87074968e-02 5.63088715e-01 9.64679539e-01 1.86602682e-01 6.34807348e-01 8.32887053e-01 -5.65705359e-01 2.46564120e-01 -3.68639499e-01 -2.99547642e-01 6.06145084e-01 1.03101201e-01 6.83213919e-02 -5.95732331e-01 3.19992274e-01 1.10082424e+00 -1.54825404e-01 -5.50233364e-01 -3.47808391e-01 -1.33449864e+00 4.12715346e-01 2.55468905e-01 1.56245187e-01 -4.10754919e-01 6.27995670e-01 5.03569007e-01 8.41875896e-02 4.88005936e-01 7.95485899e-02 -1.79376692e-01 -3.64695430e-01 -1.45876944e+00 2.87456602e-01 4.81069744e-01 9.27030325e-01 6.69425607e-01 3.24841410e-01 -2.26905093e-01 4.07139391e-01 2.84085244e-01 1.06946312e-01 4.21507806e-01 -1.58258069e+00 2.88046092e-01 5.90542816e-02 3.41171324e-01 -8.76521051e-01 -7.94788525e-02 -3.84690464e-02 -5.70133448e-01 7.25567639e-01 4.65684325e-01 -3.77740681e-01 -7.61804521e-01 1.83783686e+00 2.37221137e-01 5.84925115e-01 -1.14605859e-01 1.01539683e+00 5.91719627e-01 6.03976607e-01 -1.22513086e-01 -4.68053907e-01 1.00995898e+00 -1.16215456e+00 -1.01028788e+00 1.12286747e-01 2.15588242e-01 -1.00297058e+00 9.38490510e-01 4.67810184e-01 -1.55703008e+00 -6.21374905e-01 -1.04591382e+00 -2.81244189e-01 3.08700114e-01 5.79343885e-02 3.10291737e-01 6.46724463e-01 -1.51440930e+00 8.58221471e-01 -1.19564712e+00 -3.33890468e-01 4.22621101e-01 5.12497723e-01 -4.53525543e-01 3.30419928e-01 -9.86975729e-01 6.28761113e-01 1.90226018e-01 -2.04401202e-02 -1.07490230e+00 -7.73762882e-01 -9.10117626e-01 6.93445951e-02 4.09087151e-01 -7.40564585e-01 1.55713058e+00 -1.53570604e+00 -2.19512916e+00 6.36717439e-01 -4.82717961e-01 -7.37059891e-01 7.01633930e-01 -7.07322419e-01 1.20074954e-02 4.30613130e-01 -2.38314733e-01 9.20083344e-01 1.08082592e+00 -1.30231905e+00 -4.33841377e-01 2.51778930e-01 3.58817309e-01 1.98227167e-02 -3.53285760e-01 2.63249367e-01 -4.99473661e-01 -8.71524811e-01 -2.92876869e-01 -9.43759739e-01 -1.09640732e-01 2.28391364e-01 -3.48437838e-02 2.48581395e-01 9.04053211e-01 -6.16794109e-01 1.24192309e+00 -2.07022786e+00 2.66006023e-01 -4.34989095e-01 2.39132598e-01 5.37212312e-01 9.60218683e-02 4.36780274e-01 -2.13831097e-01 1.32283002e-01 -1.40888885e-01 -5.94531178e-01 -2.47648016e-01 -3.90458182e-02 -3.40195239e-01 6.05213225e-01 2.51284301e-01 6.57147586e-01 -9.69846964e-01 -3.81790161e-01 5.51593900e-01 8.15383315e-01 -8.29875112e-01 2.45526150e-01 -1.49592519e-01 6.24321878e-01 4.60443608e-02 4.57203835e-01 6.58349752e-01 -4.25562225e-02 2.07997918e-01 -3.90914530e-01 -5.61962724e-01 1.51314378e-01 -1.14949894e+00 1.94027531e+00 -2.49990344e-01 9.98569906e-01 7.49286786e-02 -7.23540127e-01 5.27612209e-01 6.75495803e-01 6.26176655e-01 -3.77064675e-01 4.12018329e-01 5.24472259e-02 -2.30662808e-01 -4.38058764e-01 7.12717831e-01 6.94393143e-02 5.15426576e-01 2.07137778e-01 4.06088214e-03 -4.11193594e-02 5.02947688e-01 2.28922486e-01 8.10063303e-01 6.73342526e-01 2.11136341e-01 -4.31183189e-01 7.02314317e-01 -4.77420568e-01 5.75314581e-01 3.30632716e-01 -4.37217623e-01 1.08710277e+00 5.65019965e-01 -5.14840603e-01 -1.26227343e+00 -8.93738270e-01 2.61885762e-01 7.27448642e-01 1.96167052e-01 -6.67978108e-01 -1.27210271e+00 -2.11727574e-01 -5.05901277e-01 2.78381735e-01 -5.21048665e-01 2.19407171e-01 -9.34126794e-01 -8.25784802e-02 3.84291738e-01 4.24231112e-01 3.98852378e-01 -9.41231608e-01 -8.42184007e-01 3.79258454e-01 -3.02803546e-01 -1.30421054e+00 -7.17756689e-01 -3.17231834e-01 -1.16915631e+00 -9.70202088e-01 -8.53250980e-01 -8.84531915e-01 6.90857947e-01 3.81049395e-01 1.11749136e+00 2.84598261e-01 4.76445109e-02 1.46393731e-01 -3.20413262e-01 -1.67627949e-02 -3.39634508e-01 -1.25467241e-01 1.87583834e-01 1.17658183e-01 -1.67625532e-01 -5.41544437e-01 -1.03637362e+00 4.28821474e-01 -1.13585758e+00 2.18442917e-01 1.40622526e-01 5.39722919e-01 7.21408486e-01 -1.42140493e-01 1.02689825e-01 -6.11368001e-01 3.36940587e-01 -6.47533163e-02 -7.31331944e-01 -1.62430674e-01 -1.71548560e-01 4.24863324e-02 8.63127172e-01 -3.50930095e-01 -1.02460873e+00 2.82450438e-01 -4.85350452e-02 -6.53401852e-01 -2.32728586e-01 7.01524913e-02 1.64397880e-01 -2.88487345e-01 7.05700397e-01 -1.46559387e-01 2.05000892e-01 -2.29807973e-01 3.52190554e-01 1.26953870e-01 6.86131179e-01 -4.59755123e-01 7.62889504e-01 7.61411250e-01 6.74647465e-02 -8.80170047e-01 -4.88798708e-01 -1.76597342e-01 -7.40774393e-01 -4.96277511e-01 7.98222601e-01 -1.09621894e+00 -8.07932794e-01 7.39739597e-01 -1.07562685e+00 -6.25256479e-01 -7.71036595e-02 4.50082690e-01 -7.74050653e-01 8.70946050e-01 -7.96871662e-01 -5.80774128e-01 -2.98558146e-01 -1.46284604e+00 1.05803120e+00 3.25369447e-01 -1.98310152e-01 -8.13920617e-01 4.50135767e-02 1.09836556e-01 5.19000411e-01 5.61379731e-01 -3.64084132e-02 2.92608082e-01 -7.04689801e-01 1.84881747e-01 4.45469767e-02 3.52760881e-01 2.84355164e-01 6.32894516e-01 -7.64252365e-01 -5.34091115e-01 -1.41422683e-02 -3.24937284e-01 6.12776399e-01 7.08588958e-01 1.13277781e+00 -1.83145985e-01 3.51330265e-02 6.29399836e-01 1.44087756e+00 1.98104680e-02 1.12932491e+00 5.43831885e-01 5.40445387e-01 1.94336936e-01 6.69719517e-01 4.55288172e-01 1.65286303e-01 7.80691743e-01 5.45738161e-01 -3.05513144e-01 -3.20772737e-01 1.15642473e-01 6.62088454e-01 5.37045002e-01 -5.26072681e-01 -3.73826891e-01 -6.18887782e-01 5.50018907e-01 -1.95939898e+00 -1.37811410e+00 -2.77638555e-01 2.18844748e+00 8.62994552e-01 2.02645570e-01 2.68750012e-01 2.24824697e-01 9.35569942e-01 2.64140099e-01 -1.43688083e-01 -4.05898809e-01 1.20771438e-01 2.68653572e-01 4.46262121e-01 6.82582259e-01 -1.24008465e+00 1.04241633e+00 6.71997023e+00 3.82713407e-01 -1.34299982e+00 -8.06935132e-02 6.17971599e-01 -3.12455475e-01 1.75600890e-02 1.53388560e-01 -5.18647790e-01 6.32765234e-01 1.02059162e+00 1.04679987e-01 4.72387552e-01 5.89262962e-01 8.88710380e-01 -1.50500357e-01 -1.08968806e+00 9.63257968e-01 1.00558124e-01 -1.50124991e+00 -1.45450220e-01 -5.78957004e-03 9.26955760e-01 -1.81477338e-01 -1.02317734e-02 -2.37938568e-01 -8.86256248e-02 -8.11469376e-01 1.24440110e+00 5.12384355e-01 7.21780002e-01 -8.42213333e-01 5.23573101e-01 -3.95771824e-02 -1.24013221e+00 2.10641608e-01 -1.50204271e-01 -2.64217407e-01 6.90939069e-01 5.81577599e-01 -2.45327622e-01 4.55093473e-01 6.92739904e-01 8.42874587e-01 -4.59621042e-01 1.06319308e+00 -3.11294407e-01 6.79168761e-01 -1.91427231e-01 4.56672430e-01 1.50415882e-01 -1.27494887e-01 4.51867133e-01 1.14577794e+00 4.73825067e-01 1.87833309e-01 8.85087326e-02 4.46251899e-01 -1.96130976e-01 -1.12528197e-01 -3.83829236e-01 1.60515100e-01 2.13167861e-01 1.01945889e+00 -7.50411093e-01 -4.17528033e-01 -5.09571671e-01 1.06385398e+00 1.42038897e-01 3.13747704e-01 -1.25908923e+00 -2.55872488e-01 6.99061573e-01 4.77555245e-01 5.18498242e-01 -5.34671545e-01 -2.06606075e-01 -1.34834313e+00 1.59016907e-01 -1.03075314e+00 -2.92784553e-02 -7.57561564e-01 -7.65392125e-01 6.71274245e-01 -9.82713401e-02 -1.43717861e+00 -2.68513650e-01 -2.40981579e-01 -5.29846013e-01 5.34058690e-01 -1.65260220e+00 -9.77091193e-01 -4.45666850e-01 7.02203989e-01 7.65502572e-01 9.24367979e-02 4.65357065e-01 4.46933746e-01 -4.48146999e-01 5.82013726e-01 1.60019591e-01 -6.34490848e-02 1.06475723e+00 -1.02531016e+00 5.57562828e-01 1.48609173e+00 -2.60995984e-01 7.55642772e-01 1.09089994e+00 -4.11326110e-01 -1.39811122e+00 -9.32789326e-01 6.22080624e-01 -1.94773138e-01 5.04049361e-01 -2.12486520e-01 -9.24056768e-01 5.10189772e-01 7.62564123e-01 1.46682546e-01 4.85116303e-01 -4.53542262e-01 -6.56708470e-03 -2.41947934e-01 -9.53442097e-01 7.70901978e-01 8.14308822e-01 -2.12881073e-01 -3.36471856e-01 9.16682556e-02 7.16380596e-01 -8.60803187e-01 -6.95223212e-01 3.06426674e-01 5.68217158e-01 -1.56412411e+00 9.04870093e-01 1.34229779e-01 9.55457926e-01 -7.87992895e-01 1.75215706e-01 -9.68676865e-01 -3.87418985e-01 -1.23699582e+00 -4.36564952e-01 1.35232043e+00 -1.31684719e-02 -2.46355340e-01 7.56953716e-01 6.01753056e-01 -2.94071853e-01 -5.14459729e-01 -6.07475758e-01 -5.67022502e-01 -1.97879329e-01 -1.18815787e-01 2.74398774e-01 6.45202339e-01 -1.36271492e-01 -1.64808244e-01 -7.96206415e-01 2.09694192e-01 5.94410300e-01 -1.49800390e-01 9.50998902e-01 -7.07119167e-01 -2.47016057e-01 -2.65228599e-01 -4.67857599e-01 -1.05176699e+00 1.80635732e-02 -1.70077160e-01 1.25294417e-01 -1.37094688e+00 -1.07913971e-01 1.86973572e-01 -1.50158152e-01 2.56051987e-01 -2.06665874e-01 5.54278374e-01 3.50385636e-01 1.55275837e-01 -5.45781076e-01 4.67076272e-01 1.08572173e+00 2.72155434e-01 -2.66713798e-01 -4.63249028e-01 -4.89629447e-01 6.59164310e-01 9.69502330e-01 -2.68968672e-01 -4.62929815e-01 -7.86784649e-01 -1.63970068e-01 -1.61505982e-01 2.72826970e-01 -1.21239936e+00 1.11479327e-01 -2.00180173e-01 2.69209892e-01 -3.42877924e-01 1.79120600e-01 -6.41673625e-01 3.72947276e-01 5.03544033e-01 -1.87574431e-01 2.54625440e-01 3.63919556e-01 3.65299523e-01 -3.10847461e-01 -4.61986698e-02 1.20797610e+00 -6.82202429e-02 -6.00849509e-01 2.18421072e-01 -7.96168804e-01 -1.76114365e-01 1.18999839e+00 -3.03229272e-01 -1.50089517e-01 -4.83465225e-01 -3.55927289e-01 -1.37490049e-01 1.04897189e+00 3.25682491e-01 4.81985331e-01 -1.23421347e+00 -6.15611434e-01 -8.74601752e-02 -4.36748832e-01 -2.43350659e-02 2.51235485e-01 7.17874289e-01 -1.25012922e+00 2.34754458e-01 -3.40176970e-01 -7.71469116e-01 -1.33121967e+00 7.79603720e-01 2.04189256e-01 2.28974242e-02 -6.76306367e-01 5.73134780e-01 1.27512202e-01 3.30703646e-01 1.42621949e-01 -4.22611773e-01 -8.08219835e-02 -2.18156561e-01 7.55613148e-01 3.48163694e-01 6.02828637e-02 -7.66760886e-01 -2.91754723e-01 7.26267934e-01 -2.88956892e-02 -2.61465460e-01 1.28278005e+00 -3.49244744e-01 -1.65511727e-01 4.10566032e-02 9.09186482e-01 1.93464369e-01 -1.80735981e+00 3.24677497e-01 -4.73447978e-01 -8.88149619e-01 2.21808758e-02 -2.78538525e-01 -1.25530493e+00 6.63262427e-01 6.66934252e-01 1.79527290e-02 1.40776312e+00 -5.73659718e-01 9.76148129e-01 -1.22361239e-02 1.68674007e-01 -1.05369270e+00 1.66629225e-01 4.02178884e-01 6.78486764e-01 -1.11775446e+00 1.02950551e-01 -4.87568974e-01 -3.66787463e-01 1.30450857e+00 6.82154417e-01 -4.12463933e-01 3.26435149e-01 4.77831542e-01 1.51850909e-01 2.71050066e-01 -7.11595118e-01 7.33440323e-03 6.39907643e-02 5.31359136e-01 9.96513188e-01 -4.95110393e-01 -5.76084971e-01 -1.77392438e-01 1.27738819e-01 5.40773451e-01 9.21549737e-01 1.10732138e+00 -3.52751523e-01 -1.39278150e+00 -4.04575676e-01 -2.84759700e-01 -9.03201163e-01 -5.89424260e-02 6.12516236e-03 5.10615528e-01 -1.17250778e-01 1.11085117e+00 -9.97451916e-02 -9.27556586e-03 2.43014500e-01 -2.27380335e-01 5.77948093e-01 -2.48622105e-01 -6.54302597e-01 3.38980436e-01 -1.40167505e-01 -9.04728413e-01 -9.89199042e-01 -5.15529633e-01 -1.35066855e+00 -6.31540656e-01 -1.68381765e-01 -8.13176408e-02 3.41838390e-01 6.65951788e-01 5.99170327e-01 6.76835239e-01 6.17396295e-01 -1.50685298e+00 7.34856203e-02 -5.13801932e-01 -1.94610842e-02 6.11699998e-01 6.46334350e-01 -4.51614708e-01 -1.22874863e-01 8.20455611e-01]
[10.660002708435059, -1.3691519498825073]
e898b0a3-8b31-486d-80e2-1536035df3d5
cnn-aided-factor-graphs-with-estimated-mutual
2203.05950
null
https://arxiv.org/abs/2203.05950v1
https://arxiv.org/pdf/2203.05950v1.pdf
CNN-Aided Factor Graphs with Estimated Mutual Information Features for Seizure Detection
We propose a convolutional neural network (CNN) aided factor graphs assisted by mutual information features estimated by a neural network for seizure detection. Specifically, we use neural mutual information estimation to evaluate the correlation between different electroencephalogram (EEG) channels as features. We then use a 1D-CNN to extract extra features from the EEG signals and use both features to estimate the probability of a seizure event.~Finally, learned factor graphs are employed to capture the temporal correlation in the signal. Both sets of features from the neural mutual estimation and the 1D-CNN are used to learn the factor nodes. We show that the proposed method achieves state-of-the-art performance using 6-fold leave-four-patients-out cross-validation.
['Nariman Farsad', 'Sandrine de Ribaupierre', 'Nir Shlezinger', 'Eyal Fishel Ben-Knaan', 'Bahareh Salafian']
2022-03-11
null
null
null
null
['mutual-information-estimation']
['methodology']
[-1.75667211e-01 -1.93024203e-01 -1.85741559e-02 -6.84245527e-01 -6.38455987e-01 -7.85388350e-02 1.57390639e-01 -2.16266923e-02 -6.14471138e-01 6.81455612e-01 2.28189066e-01 -6.14116602e-02 -5.49368680e-01 -4.41298574e-01 -4.36733514e-01 -5.62789500e-01 -9.86726046e-01 -2.02244073e-01 -2.28389889e-01 1.18565574e-01 -3.66193950e-02 6.42393529e-01 -9.57222521e-01 3.47126067e-01 6.79102361e-01 1.37001014e+00 -1.32465050e-01 4.89838213e-01 3.22692513e-01 6.97037637e-01 -7.03125179e-01 2.37847283e-01 4.91667762e-02 -3.80656809e-01 -3.83467138e-01 -1.17561452e-01 -7.36272871e-01 -2.84623355e-01 -1.11016262e+00 9.89061594e-01 6.10355198e-01 -2.30128333e-01 7.90866256e-01 -1.37191105e+00 -7.45486468e-02 8.61441731e-01 -3.86449218e-01 8.01753223e-01 4.60114241e-01 1.20614663e-01 7.46882915e-01 -6.89870834e-01 -5.72920740e-02 7.14966714e-01 6.16098762e-01 3.02745879e-01 -1.09284735e+00 -1.38911939e+00 -3.68944794e-01 2.31250033e-01 -1.58700216e+00 -1.04049735e-01 9.63763237e-01 -3.71941477e-01 1.20536292e+00 -6.45428821e-02 1.25425613e+00 9.00499284e-01 9.56943750e-01 9.17890549e-01 1.27724183e+00 -6.97046220e-02 2.17593946e-02 -3.06378722e-01 2.80210227e-01 7.13271379e-01 -7.50897378e-02 3.42365921e-01 -1.04381716e+00 -3.29585612e-01 8.41547728e-01 3.89232427e-01 -7.22548902e-01 1.78368449e-01 -1.21295691e+00 5.76325476e-01 7.10178733e-01 6.61324620e-01 -6.02947056e-01 2.21439511e-01 1.89071000e-01 4.83286530e-01 5.15714943e-01 3.65268558e-01 -7.32473016e-01 -1.81617007e-01 -9.53223050e-01 4.78964411e-02 7.26081014e-01 4.13923860e-01 7.71549702e-01 -1.14724068e-02 -8.29735547e-02 4.04115409e-01 4.05535132e-01 2.99185485e-01 7.21010029e-01 -2.04679534e-01 7.74928257e-02 7.88100660e-01 -5.57754695e-01 -1.00832677e+00 -1.05974197e+00 -8.35207105e-01 -1.25360072e+00 -1.11430079e-01 -1.35055244e-01 -5.31299353e-01 -7.95046508e-01 1.77365029e+00 -4.24673498e-01 6.35986805e-01 -7.18311444e-02 6.56459332e-01 6.94758713e-01 1.00873359e-01 -3.19337636e-01 -4.30714428e-01 1.35540736e+00 -1.20049149e-01 -9.76100147e-01 -5.06752394e-02 6.41259193e-01 -2.74456412e-01 9.41971689e-02 4.63698745e-01 -7.08566904e-01 -1.63033292e-01 -1.36286569e+00 6.92004025e-01 -1.84161976e-01 2.10712552e-02 6.16906047e-01 1.22035533e-01 -9.43058968e-01 9.26425874e-01 -1.27103186e+00 2.96149850e-01 7.93922961e-01 9.09596384e-01 -8.13548982e-01 1.78849816e-01 -1.53156555e+00 6.06191814e-01 3.67363334e-01 2.52142306e-02 -9.28908467e-01 -7.71958709e-01 -8.49797249e-01 3.28533500e-02 -5.34152806e-01 -5.87899446e-01 7.48751402e-01 -5.87671995e-01 -1.08530533e+00 3.18848312e-01 -5.62283434e-02 -5.77261686e-01 2.01272443e-01 -1.81707114e-01 -5.34719110e-01 1.73399448e-01 1.24620184e-01 4.16080773e-01 7.34030962e-01 -5.69134831e-01 -5.46326220e-01 -6.16793931e-01 -3.95990223e-01 -2.06890460e-02 -3.63708943e-01 2.09848538e-01 -3.73720229e-01 -6.94132507e-01 3.47561389e-01 -5.83689570e-01 -3.32315415e-01 -3.37631941e-01 -5.49045384e-01 -2.21114352e-01 5.57159126e-01 -6.30851626e-01 1.26445675e+00 -2.33050132e+00 3.53384800e-02 3.87134105e-01 8.75027895e-01 -1.45710394e-01 5.59918024e-02 -9.33293486e-04 -6.43736660e-01 -1.88524112e-01 -2.01514766e-01 -2.14017898e-01 -3.26805174e-01 -1.61624089e-01 1.61245257e-01 5.45328379e-01 4.76163685e-01 9.36267674e-01 -8.27392519e-01 -7.62031451e-02 6.37579933e-02 8.18090439e-01 -2.00677440e-01 5.51004708e-01 5.83453715e-01 6.34205878e-01 -4.34549421e-01 1.60527099e-02 5.02554715e-01 -2.02412143e-01 1.20424576e-01 -3.73120427e-01 3.11435401e-01 2.12529182e-01 -8.15143228e-01 1.84697509e+00 -4.69395965e-02 8.57234776e-01 -6.78440809e-01 -8.85702014e-01 8.91926289e-01 5.22207618e-01 7.44129717e-01 -4.59056288e-01 6.34856045e-01 6.38338253e-02 3.01098317e-01 -4.06665921e-01 -4.63014632e-01 4.42813300e-02 1.25342272e-02 4.54531670e-01 7.18740463e-01 1.39324605e-01 -2.60915309e-01 1.72491431e-01 1.65078473e+00 -4.39099789e-01 5.58341980e-01 -4.62955117e-01 4.68408093e-02 -7.14849055e-01 8.02316189e-01 3.96966815e-01 -1.45505503e-01 1.71806782e-01 7.82648921e-01 -7.03220904e-01 -2.63097852e-01 -8.33585918e-01 -3.34129542e-01 4.14235927e-02 -2.97444612e-01 -5.70044339e-01 -6.72704816e-01 -6.19594276e-01 5.98927550e-02 2.55218416e-01 -9.89797771e-01 -4.59756762e-01 -7.78249130e-02 -1.02329004e+00 6.86591804e-01 7.97565877e-01 4.06745315e-01 -8.03117037e-01 -7.56080627e-01 3.26212317e-01 1.20990016e-01 -9.36562717e-01 -1.43781260e-01 1.08517742e+00 -7.65375257e-01 -1.29727554e+00 -4.62813318e-01 -6.47143543e-01 5.13850212e-01 -2.08001256e-01 7.53062129e-01 -1.46779761e-01 -4.68268126e-01 5.83654419e-02 -1.84064493e-01 -4.55033273e-01 3.23338658e-01 -2.91781604e-01 5.59725821e-01 1.72206298e-01 7.09232032e-01 -1.29275334e+00 -9.11964417e-01 1.78315729e-01 -7.07048655e-01 -1.50345117e-01 5.67624390e-01 8.23641539e-01 4.49245095e-01 2.72359550e-01 6.63235426e-01 -2.31999353e-01 1.01337826e+00 -4.46614385e-01 -3.26287359e-01 -1.03626564e-01 -3.18953186e-01 3.16702276e-01 3.33845735e-01 -4.03651297e-01 -1.61606818e-01 2.55965114e-01 1.18095525e-01 -8.39666963e-01 -9.90108773e-02 5.93733609e-01 -1.05065025e-01 3.80191170e-02 3.56470883e-01 2.88917691e-01 -3.21442097e-01 -1.43858030e-01 -7.48768970e-02 7.47070849e-01 4.76037025e-01 2.12665379e-01 3.96725327e-01 1.08796857e-01 3.73390801e-02 -3.59564990e-01 -7.00812995e-01 -4.58080024e-01 -7.23201394e-01 -2.12041631e-01 1.27857137e+00 -1.14918339e+00 -9.30728853e-01 6.74129307e-01 -1.27615702e+00 -9.75792855e-02 1.54388711e-01 1.20743084e+00 -4.08614248e-01 -1.57099754e-01 -7.73111284e-01 -7.57511079e-01 -6.21569932e-01 -1.36335647e+00 8.98728430e-01 2.30752796e-01 -2.27179229e-01 -8.55436504e-01 2.37167537e-01 -4.78727311e-01 2.00284123e-01 2.67957866e-01 8.51041615e-01 -1.10275447e+00 -2.52812237e-01 -5.37970901e-01 -3.28870147e-01 3.07890683e-01 4.95362908e-01 -5.11957824e-01 -1.14056396e+00 -2.61721402e-01 2.16157004e-01 4.82050814e-02 1.00749099e+00 6.61001980e-01 1.17213678e+00 -1.00865439e-01 -4.51295763e-01 9.00319397e-01 1.16752183e+00 4.70524788e-01 5.06809413e-01 -2.33512372e-01 5.29964924e-01 1.34045510e-02 -1.42395243e-01 8.66549492e-01 2.46217117e-01 1.94808915e-01 2.01348275e-01 -5.78488829e-03 3.56660545e-01 8.57713223e-02 5.77981323e-02 1.47042978e+00 1.63205322e-02 1.26258373e-01 -9.59221065e-01 4.50638533e-01 -1.78987157e+00 -3.75036210e-01 1.45320848e-01 1.95108640e+00 6.14864051e-01 2.03515798e-01 -1.38386443e-01 4.44550365e-01 4.43084925e-01 -1.33750096e-01 -4.76395667e-01 3.57656747e-01 -1.33488283e-01 5.73409915e-01 3.45857650e-01 1.72340363e-01 -1.02795792e+00 5.19308567e-01 6.89469719e+00 5.31590700e-01 -1.19943845e+00 -5.14362343e-02 7.19438076e-01 8.49528909e-02 2.72900295e-02 -2.06639230e-01 -1.39803469e-01 3.49981606e-01 1.03982830e+00 -2.09681258e-01 7.40912855e-01 3.02757919e-01 -1.21415136e-02 -1.50442030e-02 -1.29664314e+00 1.54467535e+00 5.02622984e-02 -1.11390805e+00 -2.36807451e-01 2.78778654e-02 3.55731875e-01 4.86935943e-01 -1.33847684e-01 -1.45631626e-01 1.87848464e-01 -1.16812384e+00 4.38919924e-02 1.16218936e+00 7.91003764e-01 -1.17340446e+00 9.59476233e-01 3.44939739e-01 -1.32977605e+00 -4.52135969e-03 -5.95455915e-02 -1.18468881e-01 -2.21603792e-02 9.95395124e-01 -8.48129511e-01 6.22506559e-01 1.05729866e+00 1.11871374e+00 -3.41610223e-01 1.10015833e+00 -2.28207812e-01 5.25921345e-01 -1.55676976e-01 -6.37122840e-02 -6.52373303e-03 2.97703952e-01 4.68510091e-01 9.92158592e-01 2.50828743e-01 3.32410902e-01 -1.38055414e-01 7.96530843e-01 -1.75736234e-01 -2.02966467e-01 -7.25481749e-01 -1.69118404e-01 1.14740208e-01 1.13600194e+00 -7.76908159e-01 -1.80626065e-01 -2.70526439e-01 9.84972358e-01 4.04652417e-01 1.13101639e-01 -4.19529051e-01 -9.22401726e-01 7.42681086e-01 -5.15330195e-01 3.84301022e-02 -9.95578021e-02 5.06302007e-02 -1.44146299e+00 -1.70936882e-01 -5.60711443e-01 1.28747553e-01 -7.20551014e-01 -1.49765003e+00 1.07895136e+00 -1.08573623e-01 -1.27889359e+00 -3.65209103e-01 -6.03736341e-01 -7.42347181e-01 9.05704558e-01 -1.33762491e+00 -4.70027655e-01 -1.68315634e-01 9.46081340e-01 2.27291837e-01 -3.96316469e-01 1.02220654e+00 2.24327654e-01 -5.68442702e-01 5.83885014e-01 -2.74376780e-01 6.74327075e-01 4.84769553e-01 -1.17241085e+00 2.51322389e-01 5.35493433e-01 1.20046347e-01 6.47797048e-01 2.13079333e-01 -6.59763992e-01 -1.32741356e+00 -9.34000373e-01 6.97158813e-01 -1.42402694e-01 9.68643785e-01 -4.47295427e-01 -6.53692544e-01 6.10580087e-01 1.93169042e-01 5.11464655e-01 1.02408218e+00 -1.11826614e-01 -2.34683335e-01 -1.31386280e-01 -9.66995239e-01 5.79786114e-02 8.32456946e-01 -8.44680965e-01 -6.83795571e-01 1.68039650e-01 7.98886061e-01 -1.72380120e-01 -1.25963068e+00 5.83944619e-01 7.68932700e-01 -7.64855444e-01 4.10532951e-01 -6.49529636e-01 8.83531943e-02 -1.65126321e-03 -2.25383684e-01 -1.75970113e+00 -4.12145376e-01 -6.90894186e-01 1.20569251e-01 4.86325443e-01 6.02163076e-01 -7.45137751e-01 3.90533626e-01 3.17233026e-01 -1.11230746e-01 -1.10300434e+00 -1.09302247e+00 -3.60524148e-01 -2.46121928e-01 -9.38441455e-01 8.31230521e-01 1.00331044e+00 7.87212193e-01 4.61905062e-01 -4.07621980e-01 2.78023839e-01 5.41903555e-01 -4.40456986e-01 1.56169027e-01 -1.30619502e+00 -2.88683891e-01 -9.81819704e-02 -1.14754546e+00 -7.08888113e-01 3.28064382e-01 -1.03223908e+00 3.66019085e-03 -1.17973328e+00 5.47874808e-01 1.04360558e-01 -1.33208120e+00 5.34231424e-01 -1.75011784e-01 2.54386902e-01 7.39781931e-03 -1.14750668e-01 -5.18531621e-01 5.98911107e-01 9.10632908e-01 -1.79263219e-01 -3.72678280e-01 2.22505629e-02 -4.52249259e-01 6.70754731e-01 8.03635359e-01 -7.72918284e-01 -1.13822073e-01 -2.39776149e-01 9.58271548e-02 6.29314959e-01 1.05257571e-01 -1.45800614e+00 3.62975210e-01 4.16408449e-01 1.04040051e+00 -6.17217720e-01 2.11161718e-01 -7.84234703e-01 -4.58738953e-02 7.44010270e-01 -2.53465891e-01 3.89719397e-01 2.12894261e-01 7.28582680e-01 -3.92775774e-01 4.25740540e-01 3.39501500e-01 2.33142581e-02 8.45085382e-02 6.52885377e-01 -5.23651540e-01 -5.15633821e-01 7.37813830e-01 3.25798392e-01 1.95794538e-01 -4.04337227e-01 -9.59941626e-01 8.63188207e-02 -4.11026984e-01 1.85288548e-01 1.08562231e+00 -1.48813844e+00 -5.47619164e-01 7.13662088e-01 2.49649763e-01 -3.48173112e-01 1.92355901e-01 1.14097273e+00 -3.69771942e-02 3.93228650e-01 -5.25202036e-01 -7.43989348e-01 -8.97532463e-01 1.44033819e-01 4.55270737e-01 -4.85307425e-01 -4.82464284e-01 9.89910483e-01 2.77066410e-01 -1.11360222e-01 1.88015625e-01 -5.28242290e-01 -3.24166417e-01 -9.23828632e-02 8.88423443e-01 -4.50437479e-02 2.65323430e-01 -3.37798208e-01 -7.11711109e-01 2.77046293e-01 5.28492890e-02 -2.59302914e-01 1.76725662e+00 5.30364275e-01 -3.44594181e-01 5.11566579e-01 1.63322937e+00 -4.94426370e-01 -1.09331977e+00 -2.27539539e-01 -1.73219457e-01 -1.31718442e-01 5.16198933e-01 -7.12185085e-01 -1.43376005e+00 9.30877805e-01 1.19643497e+00 -1.98225915e-01 1.40256786e+00 -1.08953781e-01 5.38764119e-01 2.96517700e-01 4.48812455e-01 -4.83325750e-01 3.52253877e-02 2.65328705e-01 5.98835349e-01 -8.27296317e-01 -1.18865982e-01 9.72178653e-02 -4.48470354e-01 1.35942030e+00 1.47763938e-01 -5.56606233e-01 1.54153132e+00 6.70360088e-01 -3.40873525e-02 -5.48228741e-01 -8.71372283e-01 -3.52288000e-02 6.58528030e-01 4.39432830e-01 3.00166816e-01 4.70738262e-01 -4.45738249e-02 1.12076616e+00 -2.68998057e-01 3.24865937e-01 -6.94901273e-02 8.19930136e-01 -1.46719977e-01 -7.02521086e-01 1.07054800e-01 1.01992428e+00 -5.89321494e-01 -2.50171244e-01 -3.43103588e-01 4.46348578e-01 4.69263978e-02 9.35767412e-01 3.47426772e-01 -1.28038001e+00 1.64256722e-01 1.24088846e-01 5.74004710e-01 -2.69480348e-01 -6.73638284e-01 2.88766295e-01 -3.49169374e-01 -7.72783637e-01 -3.64144385e-01 -3.73536825e-01 -1.11574793e+00 3.44826877e-01 -6.48155272e-01 1.45531341e-01 5.71286142e-01 1.25219607e+00 4.45951790e-01 9.70116556e-01 9.74003911e-01 -7.35580742e-01 1.11544877e-01 -1.25136626e+00 -9.01383162e-01 6.95579648e-02 6.49860740e-01 -6.78827703e-01 -4.66933399e-01 -4.00623947e-01]
[13.1997652053833, 3.5113449096679688]
5f5652b4-3e93-4a18-be0b-7b4b46a26d7b
korean-drama-scene-transcript-dataset-for
null
null
https://ieeexplore.ieee.org/document/9946853
https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=9946853
Korean Drama Scene Transcript Dataset for Emotion Recognition in Conversations
Understanding emotions in conversation is a challenging task, as the sentences often have an implied meaning that is not generally understood in isolation. Efficient use of contextual information is essential for emotion recognition in conversations. Many published datasets provide contextual information for situations such as text-based online messaging, chatbots, and movie dialogues. However, such dialogue-based datasets are collected by selecting ideal conversational situations and thus do not include many variations in dialogue length and number of participants. Therefore, such datasets may not be applicable for emotion recognition in text-based movie transcripts, where scenes contain variations in the number of speakers and length of spoken sentences. We present a conversation dataset based on the Korean television show transcripts to analyze the emotions in presence of scene context. The Korean Drama Scene Transcript dataset for Emotion Recognition (KD-EmoR) is a text-based conversation dataset. We analyze three classes of complex emotions: euphoria, dysphoria, and neutral, in the scenes of a television drama to build a publicly available dataset for further research. We developed a context-aware deep learning model to classify emotions using the speaker-level context and scene context and achieved an F1-score of 0.63 on the proposed dataset.
['Hyerim Jang', 'Young-Shin Kang', 'Soo-Hyung Kim', 'Guee-Sang Lee', 'Hyung-Jeong Yang', 'Eunchae Lim', 'Sudarshan Pant']
2022-11-11
null
null
null
ieee-access-2022-11
['emotion-recognition-in-conversation']
['natural-language-processing']
[ 1.44780648e-03 -5.15320241e-01 2.81207263e-01 -9.63203728e-01 -5.32440543e-01 -5.25107920e-01 6.29498541e-01 7.79720470e-02 -2.43113101e-01 5.60283542e-01 6.56289339e-01 1.90342575e-01 3.62774193e-01 -5.00341892e-01 -1.92386210e-01 -7.51328230e-01 1.36719674e-01 -4.61315028e-02 -3.69312435e-01 -6.07869983e-01 2.90188760e-01 3.37041199e-01 -1.60404992e+00 1.12352467e+00 4.71440643e-01 1.32783723e+00 1.14153489e-01 8.68772328e-01 -5.96063793e-01 1.13896096e+00 -1.14143455e+00 -2.78343588e-01 -2.60269850e-01 -1.00564516e+00 -8.16715002e-01 6.05758369e-01 -7.11746663e-02 -1.78746983e-01 -2.56012410e-01 5.31561315e-01 7.66516805e-01 4.42800760e-01 4.64450538e-01 -1.22790122e+00 -3.04260552e-01 4.63007838e-01 -1.00946710e-01 1.01347409e-01 6.40007973e-01 -5.30695654e-02 8.22402179e-01 -7.25755334e-01 6.37154818e-01 1.41248393e+00 3.23199302e-01 6.24813557e-01 -6.47877336e-01 -5.55400610e-01 2.04111040e-01 3.97267163e-01 -1.08528686e+00 -5.88533044e-01 1.21423364e+00 -2.59718865e-01 9.38757002e-01 5.66773117e-01 7.80870497e-01 1.76192105e+00 8.95026792e-03 8.34834754e-01 1.38711190e+00 -3.27866465e-01 3.23799044e-01 4.62427378e-01 2.18601242e-01 2.07690254e-01 -1.00903535e+00 -5.64211607e-01 -7.46196747e-01 -3.34740989e-02 2.79824495e-01 5.46675622e-02 -3.37109476e-01 3.46076131e-01 -1.02664673e+00 9.67834532e-01 1.79104269e-01 4.90124941e-01 -3.38436425e-01 -5.04293144e-01 1.08168697e+00 7.10430920e-01 6.37069821e-01 2.31112763e-01 -4.15791869e-01 -7.23755300e-01 -2.52927423e-01 4.04053442e-02 1.09551764e+00 6.93634391e-01 5.58442771e-01 -7.50382151e-03 -2.28481695e-01 1.55046880e+00 -1.91535398e-01 2.20322564e-01 5.99033713e-01 -8.41369748e-01 2.59676367e-01 6.75546885e-01 -3.97683084e-02 -1.46074450e+00 -5.93415618e-01 2.06159115e-01 -8.50917101e-01 -3.91995817e-01 7.81023651e-02 -5.86766064e-01 -2.06275895e-01 1.73806083e+00 4.35975611e-01 -6.66339844e-02 4.13925052e-01 1.06333995e+00 1.53850079e+00 1.04815876e+00 -1.24130607e-01 -7.10012496e-01 1.44054377e+00 -1.00512338e+00 -1.40051067e+00 -4.21996713e-02 6.44350290e-01 -8.59323204e-01 1.67461288e+00 5.51451921e-01 -5.56499958e-01 -5.05002439e-01 -7.76710808e-01 -8.68223459e-02 -5.54409146e-01 1.65425707e-02 6.32682621e-01 6.40569270e-01 -6.04153633e-01 4.75758687e-02 -3.75296883e-02 -4.84888583e-01 -1.49236113e-01 -1.27345905e-01 -4.62694436e-01 1.95924103e-01 -1.41418409e+00 7.05288410e-01 -9.73616093e-02 5.46393633e-01 -7.03867674e-01 -9.22059342e-02 -8.93047333e-01 -1.30180076e-01 2.68283248e-01 1.18136257e-01 1.21819699e+00 -1.56577563e+00 -2.01120400e+00 1.01445317e+00 -3.30498397e-01 1.40567183e-01 2.14865088e-01 -1.20715931e-01 -7.92988718e-01 3.21739197e-01 -3.92141193e-01 3.20602030e-01 5.87815881e-01 -1.14546609e+00 -2.00367719e-01 -3.52727920e-01 2.80677348e-01 5.52968979e-01 -4.25291270e-01 5.09800375e-01 -3.86622012e-01 -3.09330523e-01 -1.19198762e-01 -8.42869580e-01 1.09026946e-01 -4.64599729e-01 -3.46798927e-01 -2.95093767e-02 1.13454664e+00 -6.33137703e-01 1.07022870e+00 -2.32145786e+00 -9.39723179e-02 -2.42353514e-01 -2.42648765e-01 -1.33739486e-01 -6.33425489e-02 8.04805458e-01 6.91092685e-02 -1.19519355e-02 -2.40032580e-02 -2.59581387e-01 -3.42024341e-02 3.08995247e-01 -2.56980926e-01 2.10923418e-01 -4.61684242e-02 4.42888111e-01 -7.58662820e-01 -5.73495328e-01 2.31583193e-01 3.75313163e-01 -3.09618860e-01 6.08408689e-01 -1.83386430e-01 7.47449160e-01 -3.99404198e-01 4.97257680e-01 4.00307685e-01 2.74695396e-01 1.59170434e-01 -3.17060620e-01 -1.66463971e-01 2.48105869e-01 -9.03432906e-01 1.66277814e+00 -9.92769599e-01 1.06791282e+00 4.11739677e-01 -8.44758987e-01 1.21951294e+00 7.11462379e-01 4.38654959e-01 -6.74020767e-01 5.35355151e-01 -3.04246604e-01 -2.77283601e-03 -1.25312555e+00 7.24003911e-01 -3.36962372e-01 -5.69275618e-01 3.30362469e-01 -1.72890887e-01 -4.09629017e-01 -2.11894270e-02 1.42285287e-01 8.61114502e-01 -6.02882147e-01 1.48155034e-01 4.97331023e-02 5.00167906e-01 -2.11582109e-01 6.73533082e-01 4.51216042e-01 -5.97401738e-01 6.21058047e-01 8.69098186e-01 -5.29805362e-01 -5.99011123e-01 -4.77469057e-01 -1.72770441e-01 1.33133686e+00 8.71427357e-02 -5.17719448e-01 -6.51785314e-01 -4.68150139e-01 -8.18938851e-01 5.99792123e-01 -6.51527941e-01 -1.19647630e-01 -1.40088379e-01 -7.19191492e-01 4.42871720e-01 4.60176244e-02 8.30823660e-01 -1.66861212e+00 -3.50758284e-01 1.74501508e-01 -8.35746169e-01 -1.53810477e+00 -3.17624986e-01 2.00923737e-02 -2.95242876e-01 -9.43234444e-01 -1.90693691e-01 -7.40648389e-01 3.84546071e-01 1.24011964e-01 1.04613960e+00 -2.20084235e-01 -2.45716661e-01 4.94890004e-01 -8.89791369e-01 -4.50948864e-01 -5.17248213e-01 -2.99976110e-01 -6.95560351e-02 4.62193787e-01 4.35316771e-01 -3.60532135e-01 -4.74581152e-01 4.92342740e-01 -8.17472398e-01 1.79507136e-01 -7.17089325e-02 9.89277005e-01 1.26735270e-01 7.63696134e-02 8.69123459e-01 -7.45531797e-01 1.15640819e+00 -4.91915494e-01 3.61487806e-01 1.06103167e-01 5.88954926e-01 -7.91566968e-01 7.16487110e-01 -7.12952256e-01 -1.52513433e+00 -2.77990192e-01 -3.85345459e-01 -9.20875147e-02 -6.18684471e-01 4.44230676e-01 -6.04467571e-01 4.00861114e-01 4.85041827e-01 1.50694191e-01 -2.14844152e-01 -7.26403520e-02 1.16383821e-01 1.48874784e+00 2.56231606e-01 -7.15587497e-01 -4.05166447e-01 2.09151998e-01 -6.24417007e-01 -1.47171509e+00 -8.08457911e-01 -4.36760247e-01 -2.85268873e-01 -8.50357115e-01 9.86302316e-01 -8.02432358e-01 -8.50732446e-01 7.89226830e-01 -1.16837585e+00 -3.94409537e-01 1.89064890e-01 4.52543110e-01 -5.13462603e-01 2.80881703e-01 -9.14589107e-01 -1.09543431e+00 -2.44168624e-01 -1.19631183e+00 9.16952252e-01 2.46383086e-01 -5.58234692e-01 -8.16311896e-01 -2.26388782e-01 6.82423949e-01 3.10902089e-01 5.74238122e-01 7.81710923e-01 -7.76749015e-01 3.16840917e-01 -1.40843123e-01 1.94088206e-01 4.70693141e-01 5.11755228e-01 1.23254523e-01 -1.31448150e+00 1.83246940e-01 5.45949459e-01 -1.05781603e+00 2.19627291e-01 5.46848327e-02 1.33511138e+00 -4.01155323e-01 3.48504603e-01 8.73627141e-02 7.51676798e-01 5.75319827e-01 6.46694899e-01 5.83982468e-02 4.21689630e-01 1.13512814e+00 8.29836607e-01 8.38772655e-01 2.82019615e-01 5.82148671e-01 2.00717762e-01 1.79909188e-02 5.41456699e-01 2.14774191e-01 4.91928428e-01 1.25023842e+00 2.76191980e-01 -3.48480970e-01 -7.12718129e-01 4.86034155e-01 -1.79547012e+00 -1.13179100e+00 -1.52835503e-01 1.60307622e+00 1.04093015e+00 -2.11913168e-01 1.41737582e-02 8.58534202e-02 8.43473613e-01 5.03627419e-01 -4.78353918e-01 -1.23436487e+00 -3.10559422e-01 -4.46777523e-01 -6.07171297e-01 2.53493488e-01 -1.07752836e+00 9.50859785e-01 5.86805820e+00 6.87850654e-01 -1.44549847e+00 5.09375744e-02 9.75806355e-01 -3.57361406e-01 3.49322669e-02 -4.51843798e-01 -3.04209560e-01 4.98427182e-01 1.09231579e+00 1.95580602e-01 3.35521638e-01 9.37739074e-01 7.31878757e-01 -1.06120773e-01 -1.00733876e+00 1.35434711e+00 3.77977639e-01 -8.64864588e-01 -1.85609788e-01 -5.31999826e-01 5.13394535e-01 -3.90931189e-01 -1.12625510e-01 7.16163278e-01 -3.03246558e-01 -8.86560440e-01 2.07331493e-01 2.55594313e-01 4.42793638e-01 -8.15025568e-01 7.93112874e-01 3.70985717e-01 -7.06343412e-01 5.60804605e-02 -2.72378087e-01 -4.85927522e-01 2.53727973e-01 4.41425085e-01 -7.14103580e-01 7.62394816e-02 9.57542181e-01 5.94185293e-01 1.02033559e-02 3.03062260e-01 4.53685708e-02 7.16494024e-01 -1.83025658e-01 -4.85986620e-01 3.05645168e-01 -3.35446328e-01 3.92886013e-01 1.65568268e+00 6.33185804e-02 5.39359152e-01 1.04358405e-01 3.02158445e-01 -2.45761603e-01 6.72254980e-01 -7.17234075e-01 -1.47927538e-01 4.21203792e-01 1.54022682e+00 -6.21333301e-01 -2.82045394e-01 -4.96958941e-01 1.09636152e+00 6.72642514e-02 3.27896059e-01 -8.68468463e-01 -3.66830856e-01 8.22116852e-01 -5.62853277e-01 -9.62058902e-02 -7.83566460e-02 -3.09795397e-03 -1.19543779e+00 6.72565773e-02 -1.42019856e+00 2.80132204e-01 -9.17979121e-01 -1.44372535e+00 8.03431511e-01 -8.97383764e-02 -1.02741778e+00 -9.09767225e-02 -4.37460899e-01 -1.01318574e+00 5.14294565e-01 -9.04268086e-01 -8.41903627e-01 -7.16455877e-01 7.57541835e-01 1.18707395e+00 -1.01087295e-01 1.09849370e+00 1.82534471e-01 -8.35382879e-01 4.34444010e-01 -4.85714301e-02 2.76170254e-01 1.04842341e+00 -8.90322268e-01 -4.07576025e-01 1.50758207e-01 -1.59944549e-01 4.41224575e-01 8.02828908e-01 -1.32763848e-01 -1.41422474e+00 -7.29272246e-01 6.48021221e-01 -9.61880088e-02 5.57592750e-01 -6.60942733e-01 -8.72836709e-01 3.69473279e-01 6.62464023e-01 -3.34700346e-01 1.23919499e+00 4.78659511e-01 -7.79896751e-02 -2.13147551e-01 -1.31765461e+00 7.91168988e-01 7.24606931e-01 -7.19819963e-01 -6.42392755e-01 4.21946108e-01 6.81242704e-01 -4.83636349e-01 -8.56071949e-01 2.52967507e-01 4.64249015e-01 -1.23507190e+00 4.93377119e-01 -5.13284922e-01 7.94673324e-01 2.23414704e-01 -4.60702360e-01 -1.42362404e+00 4.71470475e-01 -5.69588482e-01 2.71682173e-01 1.64985573e+00 9.48279575e-02 -3.18734020e-01 3.69991839e-01 8.50463808e-01 -2.23930538e-01 -7.38285840e-01 -8.00032198e-01 -1.46831349e-01 -2.11128414e-01 -6.52923465e-01 5.67321658e-01 1.36252606e+00 5.47299087e-01 7.82871842e-01 -6.53922260e-01 -2.49652922e-01 -2.38081247e-01 3.15274328e-01 1.07264948e+00 -6.50242865e-01 6.11781813e-02 -3.13951135e-01 -3.19134921e-01 -8.23900342e-01 4.30184007e-01 -2.95119911e-01 1.83185235e-01 -1.33270907e+00 1.70854136e-01 -1.20902501e-01 -8.09851009e-03 1.57348022e-01 -3.07670124e-02 -8.46786126e-02 4.46490571e-02 -1.32150918e-01 -6.82375908e-01 9.74632382e-01 1.20572901e+00 -2.52187133e-01 -4.02708292e-01 -2.16642693e-01 -5.45740604e-01 8.90376031e-01 9.42282259e-01 7.45936809e-03 -5.74078143e-01 -4.76551428e-02 1.07548781e-01 4.06583905e-01 -1.06523775e-01 -6.78628981e-01 -7.26104379e-02 -4.40550864e-01 2.48964533e-01 -5.62110484e-01 9.82988179e-01 -7.67619967e-01 -1.54628411e-01 -2.53695041e-01 -8.38268936e-01 -9.90841985e-02 2.06957579e-01 3.65244299e-01 -7.02981770e-01 -1.19551681e-01 5.77237487e-01 -2.52051145e-01 -8.26464772e-01 -1.81895886e-02 -8.22871387e-01 1.22872733e-01 9.89454985e-01 -1.80529267e-01 -3.67205858e-01 -1.15668190e+00 -7.81969011e-01 2.10559890e-01 1.98518693e-01 7.11211860e-01 7.09974289e-01 -1.22461390e+00 -6.31355166e-01 -2.24669427e-01 2.78459579e-01 -1.32682085e-01 9.39400971e-01 6.48612678e-01 -3.21338981e-01 -5.44065945e-02 -3.17285687e-01 -5.39050043e-01 -1.62515259e+00 2.19313025e-01 3.40228498e-01 1.26771644e-01 -4.45368290e-01 5.95029414e-01 3.32373232e-01 -6.65370405e-01 3.46909374e-01 -1.85173348e-01 -6.16253912e-01 4.58471060e-01 7.84942031e-01 2.92376876e-02 -7.61161819e-02 -8.96887898e-01 -1.65722594e-01 1.38770193e-01 1.00738987e-01 -1.98748007e-01 1.19173241e+00 -5.25764346e-01 -1.54085636e-01 1.18279839e+00 1.52453899e+00 6.99917972e-02 -8.60504031e-01 -9.37261656e-02 -4.01430637e-01 -4.09362108e-01 -1.27104923e-01 -8.30943346e-01 -9.44320738e-01 1.08587432e+00 3.44327241e-01 4.15374368e-01 1.30813587e+00 -1.58125728e-01 8.73715639e-01 8.40610027e-01 8.89510363e-02 -1.47564650e+00 4.99981791e-01 8.34977210e-01 1.41399038e+00 -1.52009332e+00 -4.29497063e-01 -1.89069927e-01 -1.47118592e+00 1.12959361e+00 9.45824444e-01 4.32536811e-01 4.60023701e-01 2.05003396e-01 7.62343943e-01 -2.47093633e-01 -9.93416011e-01 5.03414012e-02 -2.03110471e-01 3.61229807e-01 6.67071760e-01 4.25760522e-02 -3.13041359e-01 1.04230857e+00 -4.46077466e-01 -4.86685574e-01 6.34704351e-01 8.94705296e-01 -3.04790467e-01 -7.01921880e-01 -4.29988116e-01 2.03546911e-01 -5.61550915e-01 1.36383384e-01 -9.57360327e-01 6.60485625e-01 -6.66700006e-02 1.63557684e+00 1.43839523e-01 -5.43513238e-01 3.88916254e-01 3.77038836e-01 5.57331704e-02 -4.49940532e-01 -9.42526221e-01 -7.07349256e-02 7.46521056e-01 -3.11888486e-01 -7.05185115e-01 -3.82547766e-01 -1.16705692e+00 -3.63632053e-01 -1.00583911e-01 3.54084551e-01 8.77284229e-01 1.04553616e+00 3.11332017e-01 4.57372189e-01 1.19164026e+00 -9.23385024e-01 1.60289451e-01 -1.33683276e+00 -6.31759703e-01 8.78601432e-01 1.14766017e-01 -4.23045546e-01 -4.67526764e-01 -2.84167919e-02]
[13.084771156311035, 6.013156890869141]
1c84270b-e62b-4de8-b971-9bee397b68ce
representation-learning-with-fine-grained
2005.09681
null
https://arxiv.org/abs/2005.09681v3
https://arxiv.org/pdf/2005.09681v3.pdf
Weakly Supervised Representation Learning with Coarse Labels
With the development of computational power and techniques for data collection, deep learning demonstrates a superior performance over most existing algorithms on visual benchmark data sets. Many efforts have been devoted to studying the mechanism of deep learning. One important observation is that deep learning can learn the discriminative patterns from raw materials directly in a task-dependent manner. Therefore, the representations obtained by deep learning outperform hand-crafted features significantly. However, for some real-world applications, it is too expensive to collect the task-specific labels, such as visual search in online shopping. Compared to the limited availability of these task-specific labels, their coarse-class labels are much more affordable, but representations learned from them can be suboptimal for the target task. To mitigate this challenge, we propose an algorithm to learn the fine-grained patterns for the target task, when only its coarse-class labels are available. More importantly, we provide a theoretical guarantee for this. Extensive experiments on real-world data sets demonstrate that the proposed method can significantly improve the performance of learned representations on the target task, when only coarse-class information is available for training. Code is available at \url{https://github.com/idstcv/CoIns}.
['Qi Qian', 'Juhua Hu', 'Yuanhong Xu', 'Rong Jin', 'Hao Li']
2020-05-19
null
http://openaccess.thecvf.com//content/ICCV2021/html/Xu_Weakly_Supervised_Representation_Learning_With_Coarse_Labels_ICCV_2021_paper.html
http://openaccess.thecvf.com//content/ICCV2021/papers/Xu_Weakly_Supervised_Representation_Learning_With_Coarse_Labels_ICCV_2021_paper.pdf
iccv-2021-1
['learning-with-coarse-labels']
['computer-vision']
[-1.33627821e-02 -4.08764482e-01 -5.85663557e-01 -5.09950399e-01 -6.73655868e-01 -5.81744015e-01 4.56226170e-01 9.68900099e-02 -2.86528379e-01 5.40966928e-01 -1.56869769e-01 -3.05143893e-01 -1.78814426e-01 -7.66613126e-01 -8.59942496e-01 -8.78918290e-01 9.84199122e-02 3.59534681e-01 2.73855254e-02 3.48124467e-02 2.35161185e-01 4.13579255e-01 -1.55239153e+00 4.14319724e-01 6.59441590e-01 1.37205017e+00 5.93874097e-01 1.88135102e-01 -1.69288337e-01 5.98541379e-01 -4.08119977e-01 -2.06806824e-01 4.34617430e-01 -1.21808022e-01 -6.56726778e-01 2.07453474e-01 4.08840686e-01 -3.63219291e-01 -4.63145047e-01 1.17020476e+00 3.27644676e-01 1.89816177e-01 6.62364423e-01 -1.23601210e+00 -7.67667115e-01 2.95190603e-01 -7.87494481e-01 3.07968557e-01 -6.55843616e-02 1.88404441e-01 1.20309997e+00 -1.02790296e+00 4.88404900e-01 1.02580011e+00 2.27101713e-01 4.11231756e-01 -1.07519770e+00 -9.20887828e-01 3.50424200e-01 5.06110609e-01 -1.43464565e+00 -2.78486401e-01 9.55608308e-01 -4.84203875e-01 3.56700420e-01 1.91387847e-01 3.98604155e-01 1.35322261e+00 -6.17920943e-02 1.22612035e+00 1.25161743e+00 -2.58399993e-01 1.93469509e-01 3.18904400e-01 3.84733886e-01 8.05972755e-01 4.39676285e-01 1.09292075e-01 -4.64469194e-01 1.44193927e-02 9.66715634e-01 5.76519310e-01 -1.94434181e-01 -5.68424761e-01 -1.13985896e+00 8.82471800e-01 7.25047171e-01 4.46578383e-01 -4.06808466e-01 1.29462425e-02 3.94454777e-01 3.27673763e-01 3.61286581e-01 2.42865190e-01 -4.67582345e-01 6.26030192e-03 -6.38713777e-01 -3.23341787e-02 5.34107864e-01 1.23258400e+00 9.62541521e-01 2.75039449e-02 -2.98842788e-01 9.67265069e-01 2.56783348e-02 4.20531034e-01 5.46531200e-01 -6.21848643e-01 5.32047570e-01 6.43690467e-01 1.15006767e-01 -1.23257113e+00 -2.94928432e-01 -6.70240879e-01 -1.21057463e+00 8.17732289e-02 7.39935398e-01 1.32888943e-01 -1.17861545e+00 1.53712308e+00 1.92512497e-01 -1.28129780e-01 -3.69116336e-01 1.23028839e+00 6.66306257e-01 7.28972197e-01 9.37656462e-02 1.05505455e-02 1.36678863e+00 -1.08476114e+00 -5.03936827e-01 -4.12540466e-01 4.69930410e-01 -6.63582504e-01 1.37335014e+00 4.23349112e-01 -5.50410926e-01 -7.46376872e-01 -9.71030593e-01 -1.40693069e-01 -3.96159351e-01 4.26601917e-01 8.87766063e-01 3.92639905e-01 -5.58225572e-01 3.45545828e-01 -6.23840153e-01 -2.02918142e-01 8.49525869e-01 3.01218688e-01 -3.35150570e-01 -4.78063971e-01 -8.25360060e-01 3.23042661e-01 3.40237468e-01 1.28062814e-01 -1.14669573e+00 -3.69637311e-01 -6.83755636e-01 1.84133738e-01 6.17266119e-01 -2.85410196e-01 1.11364865e+00 -1.07714772e+00 -9.79239225e-01 9.01925087e-01 -2.34097764e-01 -1.15617469e-01 4.63452548e-01 -1.87367111e-01 -1.00495897e-01 5.97647354e-02 1.24126226e-01 4.38872576e-01 1.06042886e+00 -1.48811901e+00 -7.44826078e-01 -3.56580466e-01 3.91781837e-01 -8.22132006e-02 -6.66337013e-01 -2.96830595e-01 -6.96079612e-01 -6.94997072e-01 -6.43961504e-02 -9.10344720e-01 -2.56429344e-01 2.40544230e-01 -4.54927087e-01 -5.53241909e-01 8.63805413e-01 -4.46015447e-01 8.70006025e-01 -2.48151755e+00 -9.04051885e-02 2.17499092e-01 3.73336077e-01 4.08499628e-01 -2.31382072e-01 2.15525061e-01 1.63514167e-01 3.41731943e-02 6.29526749e-02 -1.35638148e-01 4.63358574e-02 3.37034881e-01 -3.11789572e-01 3.95919561e-01 4.79116105e-02 9.52700019e-01 -8.34250987e-01 -4.53055292e-01 2.22850382e-01 2.05114365e-01 -1.89409092e-01 4.22890961e-01 -2.40296111e-01 4.22193706e-01 -7.84897149e-01 9.21742797e-01 6.88660026e-01 -7.98326552e-01 2.79334426e-01 -2.37028822e-01 2.52556384e-01 1.45837054e-01 -9.58512783e-01 1.66225421e+00 -6.00980163e-01 5.48728287e-01 -2.51554493e-02 -1.48592103e+00 1.03144848e+00 -2.26684269e-02 2.89174467e-01 -1.16835070e+00 1.34947553e-01 1.77424878e-01 -9.81244743e-02 -4.71377075e-01 2.02994794e-01 4.66381609e-02 -1.90551460e-01 5.11605203e-01 -2.07955074e-02 4.87881869e-01 2.55110383e-01 7.16732144e-02 9.04809713e-01 -5.63483164e-02 3.20878714e-01 -2.03248218e-01 2.34717652e-01 5.33778518e-02 6.94441319e-01 7.85012960e-01 -9.54989567e-02 5.23054659e-01 3.32575947e-01 -8.88182759e-01 -1.05518937e+00 -8.28264952e-01 -1.28321439e-01 1.56008852e+00 3.79977286e-01 -2.73753077e-01 -4.76037711e-01 -1.04311454e+00 1.81144252e-01 3.63961905e-01 -7.97797322e-01 -1.29181296e-01 -5.63800395e-01 -6.42168462e-01 8.08159728e-03 7.23930478e-01 4.54893619e-01 -1.15984809e+00 -5.37814558e-01 3.61014456e-02 -1.25551838e-02 -1.02210057e+00 -4.42024678e-01 2.38487333e-01 -8.40823352e-01 -1.17840052e+00 -8.65418971e-01 -9.27543283e-01 9.31794167e-01 7.68055379e-01 1.05404198e+00 3.71876866e-01 -3.46302509e-01 1.26152605e-01 -4.30283576e-01 -2.92845786e-01 4.43273894e-02 1.90114051e-01 -1.85402170e-01 1.82643935e-01 5.89645088e-01 -3.93319577e-01 -8.52634609e-01 5.43306291e-01 -8.50464821e-01 2.02805609e-01 9.23407376e-01 1.06993520e+00 7.56740570e-01 1.29979163e-01 6.45267308e-01 -9.95426059e-01 3.90474796e-01 -6.55909538e-01 -6.16966784e-01 2.29619831e-01 -5.75566590e-01 1.19289391e-01 8.15441132e-01 -5.90883195e-01 -8.76120925e-01 1.07166521e-01 6.97210878e-02 -5.87005615e-01 -3.26898783e-01 4.18334275e-01 -1.73971176e-01 1.77160412e-01 4.89195883e-01 4.11250979e-01 -2.15794817e-01 -8.81059289e-01 1.51313186e-01 7.67413199e-01 2.18070537e-01 -8.62062037e-01 7.27161348e-01 4.64703709e-01 -1.67513505e-01 -5.79865396e-01 -1.12514579e+00 -5.20935833e-01 -5.11009216e-01 -1.56967621e-02 5.46130121e-01 -8.96465778e-01 -5.32796144e-01 2.11919516e-01 -9.12288427e-01 -4.37459767e-01 -2.37568840e-01 3.26721340e-01 -3.65101039e-01 2.51322895e-01 -4.23239321e-01 -5.53507209e-01 -2.34183058e-01 -1.06636631e+00 1.03879130e+00 3.02812606e-01 3.11189499e-02 -8.75082254e-01 -2.68653691e-01 3.00580591e-01 2.33522832e-01 7.52035007e-02 1.03697979e+00 -6.89402580e-01 -9.20221388e-01 -3.16494763e-01 -6.30743980e-01 3.90186161e-01 3.70942891e-01 -3.56052518e-01 -1.00345767e+00 -5.59880078e-01 -1.21204875e-01 -5.51355422e-01 9.30068135e-01 2.35241964e-01 1.81596935e+00 -2.39754871e-01 -4.64162976e-01 5.59708297e-01 1.45630097e+00 2.46977925e-01 2.14834169e-01 4.58912939e-01 9.13142502e-01 4.12622333e-01 8.69672596e-01 6.09833598e-01 2.88631141e-01 7.48261750e-01 4.05108750e-01 -1.63611516e-01 -3.03522795e-01 -3.91318232e-01 -3.29149328e-02 6.87370002e-01 -1.65722623e-01 -1.55150697e-01 -9.59518611e-01 7.02426255e-01 -1.94525170e+00 -6.08891547e-01 1.46862030e-01 2.15775132e+00 8.03320765e-01 9.32521522e-02 7.40600154e-02 1.91620663e-01 6.93507493e-01 2.08086461e-01 -7.33407080e-01 -1.53497392e-02 1.35560244e-01 8.13194737e-02 3.58839005e-01 1.64512657e-02 -1.19734597e+00 7.42581010e-01 5.64564037e+00 1.13686872e+00 -1.14939797e+00 2.34162554e-01 7.66272306e-01 -8.68857056e-02 -2.87371248e-01 -2.85254568e-01 -8.47805679e-01 6.84472859e-01 4.69240963e-01 -1.89539492e-01 3.55165124e-01 1.26667225e+00 -1.51769603e-02 1.20873287e-01 -1.27194309e+00 1.27114522e+00 -9.50641632e-02 -1.34445167e+00 2.61723120e-02 1.60713226e-01 6.80043817e-01 -1.77970171e-01 2.43913889e-01 4.19145852e-01 2.95392960e-01 -9.85782921e-01 6.21636033e-01 1.45044625e-01 9.43755805e-01 -7.39792764e-01 6.60651863e-01 4.44401354e-01 -1.18930471e+00 -2.48738959e-01 -7.25804627e-01 -1.18203357e-01 -2.88839191e-01 8.17961693e-01 -7.47864664e-01 4.15785283e-01 9.02327061e-01 8.96092117e-01 -5.83911777e-01 9.92797017e-01 -3.00504476e-01 6.45326793e-01 -1.75100654e-01 -1.54963002e-01 3.17893714e-01 -1.09393764e-02 9.63134021e-02 1.16787088e+00 2.26809576e-01 -4.01252583e-02 4.06746149e-01 6.88808799e-01 -3.42880249e-01 7.06800893e-02 -7.38507211e-01 -1.11064896e-01 4.26209927e-01 1.38634086e+00 -6.66908860e-01 -3.25592250e-01 -5.61646461e-01 8.38828087e-01 6.43261075e-01 5.38114488e-01 -6.50853932e-01 -2.19266832e-01 5.72727919e-01 1.71069518e-01 5.16693056e-01 -2.58351266e-01 -3.80429059e-01 -1.12713015e+00 2.35986695e-01 -1.04154420e+00 5.38367391e-01 -3.36902767e-01 -1.58371556e+00 4.73878056e-01 -1.64127931e-01 -1.43050981e+00 -1.22343168e-01 -7.69259632e-01 -3.13613415e-01 6.42104149e-01 -1.49073911e+00 -1.06685245e+00 -5.09015262e-01 7.77025521e-01 7.23226786e-01 -2.11413577e-01 6.80500567e-01 3.95526350e-01 -6.20490074e-01 8.02178502e-01 4.23263550e-01 3.67964417e-01 5.81438959e-01 -1.06295240e+00 5.49097098e-02 5.61827719e-01 4.65923458e-01 4.11186844e-01 3.73581320e-01 -3.03553224e-01 -1.54620898e+00 -1.02755272e+00 4.40026045e-01 -1.79523885e-01 4.83034104e-01 -6.08289361e-01 -9.52529907e-01 7.12977231e-01 7.46131241e-02 3.33065569e-01 7.19377279e-01 3.40883821e-01 -6.04796410e-01 -3.76911134e-01 -9.28305745e-01 2.71883279e-01 1.12250113e+00 -4.81774718e-01 -4.72478032e-01 5.47643483e-01 4.19287533e-01 -2.87477285e-01 -7.28325963e-01 2.97343224e-01 5.02066255e-01 -7.70087719e-01 9.21076179e-01 -6.04526341e-01 5.25418580e-01 -2.42547497e-01 -1.80122793e-01 -1.23904228e+00 -5.82997262e-01 -1.10144727e-01 -1.91007435e-01 1.02992010e+00 2.09024131e-01 -6.00868762e-01 8.85198355e-01 4.47427899e-01 1.19806573e-01 -9.40704405e-01 -7.35777557e-01 -9.56226766e-01 -9.58974063e-02 -1.49011046e-01 5.45156956e-01 1.00229764e+00 -4.31900054e-01 2.43009597e-01 -5.80969214e-01 1.04199544e-01 7.62155116e-01 8.93795192e-01 7.97744870e-01 -1.25262094e+00 -3.37738305e-01 -2.29315445e-01 -3.79429579e-01 -1.34381783e+00 1.14993744e-01 -8.33764791e-01 1.23176503e-03 -1.58242583e+00 5.98168731e-01 -9.46140230e-01 -7.99193740e-01 7.82341123e-01 -2.51172930e-01 2.41654053e-01 3.39678854e-01 5.43605745e-01 -7.54485846e-01 4.80534226e-01 1.54831660e+00 -4.28668141e-01 2.22767368e-01 -4.20650132e-02 -8.58051300e-01 5.79192281e-01 9.02569890e-01 -6.07181251e-01 -4.44921553e-01 -5.59566021e-01 -7.16848392e-03 -2.32144341e-01 2.92407542e-01 -7.58575916e-01 3.13890800e-02 -3.64430815e-01 6.12313330e-01 -4.07198519e-01 3.09014827e-01 -9.42302883e-01 -1.84564859e-01 4.16166872e-01 -3.45960796e-01 -2.75033444e-01 6.03070529e-03 7.24453449e-01 -3.42683375e-01 -1.99278548e-01 6.68830276e-01 -1.14102431e-01 -1.04359627e+00 6.33259475e-01 -2.26212945e-02 -2.74972133e-02 1.05814755e+00 -3.29739670e-03 -4.49580491e-01 -3.26816231e-01 -5.35341442e-01 8.60561654e-02 4.08740878e-01 6.28104031e-01 6.17580712e-01 -1.58004391e+00 -5.20235479e-01 2.46890500e-01 2.70376146e-01 1.82076707e-01 1.80888891e-01 5.63529611e-01 1.28664747e-02 4.99268800e-01 -5.08272052e-01 -7.18721032e-01 -1.12500203e+00 1.03643799e+00 -1.32379144e-01 -3.12442958e-01 -6.16959453e-01 7.65486717e-01 8.22023153e-01 -3.83460037e-02 3.09003592e-01 -1.59134716e-01 -1.35341451e-01 3.01184524e-02 6.65538609e-01 1.59050316e-01 -7.40777776e-02 -3.14930260e-01 -2.76303619e-01 5.58772326e-01 -4.93364513e-01 5.92065334e-01 1.44842374e+00 -7.84653798e-02 3.09013277e-01 3.95828128e-01 1.34294856e+00 -2.18781888e-01 -1.54380214e+00 -6.39216065e-01 -6.16839901e-02 -8.93619120e-01 -1.14252148e-02 -5.26997030e-01 -1.54578424e+00 1.14807892e+00 6.62848830e-01 3.30760330e-01 1.23266315e+00 1.59174725e-01 7.94632554e-01 5.59565723e-01 6.55344307e-01 -1.03928077e+00 4.28252190e-01 1.78407878e-01 8.31900418e-01 -1.48498321e+00 -9.57839787e-02 -4.77357239e-01 -6.37708604e-01 9.46929693e-01 7.07970977e-01 -2.52551734e-01 5.55170953e-01 3.50071900e-02 -8.58563650e-03 -2.33377352e-01 -6.06730282e-01 -1.83208704e-01 3.54062021e-01 5.00506520e-01 2.30407998e-01 3.26771051e-01 -5.19821607e-02 6.52446687e-01 2.71153927e-01 1.64298061e-02 9.91981104e-02 9.44355667e-01 -4.09257233e-01 -1.17360342e+00 -2.24372849e-01 6.83072507e-01 -3.58567655e-01 6.13189824e-02 -1.09058157e-01 7.35778332e-01 6.29936904e-03 7.72696078e-01 4.60991897e-02 -3.92133832e-01 2.98771888e-01 -1.90195888e-01 4.44456577e-01 -5.76886356e-01 -2.69120727e-02 -1.49524331e-01 4.30116343e-04 -6.36733115e-01 -2.48173878e-01 -3.74872535e-01 -9.59032118e-01 -4.04483527e-01 -2.80052871e-02 -1.58897080e-02 4.17039812e-01 7.88448453e-01 3.60793531e-01 2.96347350e-01 8.72929633e-01 -8.33559453e-01 -6.53396904e-01 -7.47192264e-01 -6.48397565e-01 7.60838449e-01 2.53977567e-01 -1.01563489e+00 -2.30638802e-01 2.62481067e-02]
[9.65853214263916, 2.2060134410858154]
4ef05398-d01e-4f15-ab3d-9fd0cbec7877
intelligent-problem-solving-as-integrated
2208.08731
null
https://arxiv.org/abs/2208.08731v1
https://arxiv.org/pdf/2208.08731v1.pdf
Intelligent problem-solving as integrated hierarchical reinforcement learning
According to cognitive psychology and related disciplines, the development of complex problem-solving behaviour in biological agents depends on hierarchical cognitive mechanisms. Hierarchical reinforcement learning is a promising computational approach that may eventually yield comparable problem-solving behaviour in artificial agents and robots. However, to date the problem-solving abilities of many human and non-human animals are clearly superior to those of artificial systems. Here, we propose steps to integrate biologically inspired hierarchical mechanisms to enable advanced problem-solving skills in artificial agents. Therefore, we first review the literature in cognitive psychology to highlight the importance of compositional abstraction and predictive processing. Then we relate the gained insights with contemporary hierarchical reinforcement learning methods. Interestingly, our results suggest that all identified cognitive mechanisms have been implemented individually in isolated computational architectures, raising the question of why there exists no single unifying architecture that integrates them. As our final contribution, we address this question by providing an integrative perspective on the computational challenges to develop such a unifying architecture. We expect our results to guide the development of more sophisticated cognitively inspired hierarchical machine learning architectures.
['Stefan Wermter', 'Martin V. Butz', 'Phuong D. H. Nguyen', 'Matthias Kerzel', 'Christian Gumbsch', 'Manfred Eppe']
2022-08-18
null
null
null
null
['hierarchical-reinforcement-learning']
['methodology']
[ 3.54009897e-01 5.12152135e-01 3.96422833e-01 6.25562966e-02 2.08749235e-01 -4.78408873e-01 6.35287702e-01 4.60376054e-01 -3.82089704e-01 8.52010489e-01 -1.12095363e-01 -7.97841996e-02 -6.36891425e-01 -8.04123223e-01 -3.84659052e-01 -6.33980393e-01 -2.56538332e-01 5.38477838e-01 3.41674149e-01 -6.70799851e-01 7.17388570e-01 4.89369154e-01 -2.03082609e+00 3.56997550e-01 9.42768276e-01 5.88360131e-01 6.50550544e-01 7.50398695e-01 2.39014879e-01 1.14199042e+00 -4.31038797e-01 -2.63796002e-01 -2.44540758e-02 -7.03699112e-01 -1.05357718e+00 5.16222939e-02 -1.91560581e-01 8.27904642e-02 1.18942950e-02 7.04887450e-01 2.21652731e-01 4.53501008e-02 8.66235018e-01 -9.04993415e-01 -8.33583355e-01 5.18112004e-01 1.74762830e-01 2.22221121e-01 3.07132185e-01 2.90857613e-01 9.61187780e-01 -5.06287992e-01 4.93006736e-01 1.26602995e+00 4.31570530e-01 8.47673655e-01 -1.26010263e+00 -1.13430828e-01 7.26927742e-02 5.26562631e-01 -1.00850070e+00 -3.48234266e-01 6.43497646e-01 -6.30495310e-01 1.24129510e+00 -2.33424053e-01 1.19547081e+00 6.76136076e-01 5.81315994e-01 5.08626223e-01 1.29621673e+00 -5.30599177e-01 6.23034775e-01 -2.67023146e-01 -2.61133946e-02 7.72087038e-01 3.16448182e-01 6.43754840e-01 -7.28948176e-01 1.14440978e-01 9.03085172e-01 -3.42034817e-01 1.96383461e-01 -2.33257502e-01 -1.03377163e+00 9.24595118e-01 7.89138496e-01 7.67215848e-01 -6.51919782e-01 1.85655326e-01 2.94061959e-01 1.00365281e-01 -4.36752319e-01 1.13702512e+00 -4.61357415e-01 2.17075065e-01 -4.87500668e-01 2.19230488e-01 5.98952234e-01 6.07416332e-01 7.26633370e-01 4.04576868e-01 2.21418321e-01 7.53936648e-01 2.90079027e-01 -1.82397943e-02 6.16269708e-01 -1.43106747e+00 -5.29066324e-01 7.00718164e-01 -1.87822923e-01 -9.53208327e-01 -8.53857577e-01 -4.58554268e-01 -6.70880556e-01 5.81699491e-01 2.57520705e-01 9.25351605e-02 -3.64073396e-01 1.85644996e+00 1.65332720e-01 -5.04874408e-01 3.28495026e-01 7.73462474e-01 5.49014509e-01 4.32989448e-01 5.33623397e-01 -2.63939500e-01 1.43913460e+00 -9.70386684e-01 -3.05042088e-01 -3.89797151e-01 3.51587802e-01 -2.82514453e-01 8.78290534e-01 5.11712909e-01 -1.27637815e+00 -7.88286507e-01 -1.22500241e+00 -1.31377146e-01 -5.27256906e-01 -3.96083504e-01 1.11229646e+00 7.22595572e-01 -1.32019186e+00 7.82570660e-01 -6.55255020e-01 -8.01776230e-01 2.83519536e-01 3.71065259e-01 -1.09365553e-01 4.25820172e-01 -1.18017781e+00 1.45822668e+00 1.07801497e+00 -1.64753020e-01 -1.09363377e+00 -3.82936001e-01 -5.35133600e-01 1.07313439e-01 1.70715824e-01 -1.25508022e+00 1.34775889e+00 -1.02924252e+00 -1.79063880e+00 1.05181205e+00 -1.46885989e-02 -5.57588816e-01 -2.69418508e-01 2.42248088e-01 2.25756750e-01 4.57154036e-01 -1.05675004e-01 1.03686750e+00 4.48938429e-01 -1.57611954e+00 -7.14159667e-01 -4.89816874e-01 2.66883850e-01 2.46473253e-01 -2.08537996e-01 -1.57473773e-01 3.34013909e-01 -5.20875752e-01 9.47855115e-02 -8.86403382e-01 -3.14399123e-01 -2.19558790e-01 5.06755531e-01 -6.14682257e-01 -6.05384149e-02 -2.43602127e-01 8.95985961e-01 -1.99229896e+00 7.02837169e-01 -2.96709031e-01 3.30292493e-01 -1.08040553e-02 -6.80780336e-02 6.56490445e-01 8.75614360e-02 -5.98167442e-02 -5.13319194e-01 3.48516405e-01 8.80529433e-02 4.89716470e-01 -1.52255043e-01 -2.62873061e-02 2.96798110e-01 1.16342843e+00 -9.69534159e-01 -3.43122691e-01 2.31809318e-01 2.07522810e-01 -8.10428739e-01 3.46946001e-01 -3.55893612e-01 5.65797746e-01 -2.91486055e-01 5.11195183e-01 1.12141460e-01 -2.04093009e-01 4.99071032e-01 3.79653513e-01 -2.58949339e-01 4.39297765e-01 -4.73108023e-01 1.63287580e+00 -2.41863772e-01 4.36846614e-01 -5.86790890e-02 -1.24557877e+00 8.57533336e-01 3.14136833e-01 2.28726760e-01 -9.05143380e-01 2.14483142e-01 1.59675673e-01 8.30656886e-01 -2.85559386e-01 1.61692470e-01 -4.05895412e-01 7.58974347e-03 3.02459508e-01 4.47547615e-01 -6.78164184e-01 4.25389796e-01 -1.53297812e-01 9.48380649e-01 4.01883274e-01 7.80917883e-01 -6.98679626e-01 8.10253084e-01 2.24291325e-01 4.04702276e-01 9.92572248e-01 -4.91565555e-01 6.20381720e-02 3.68393630e-01 -6.75537527e-01 -1.08377647e+00 -1.14112687e+00 -1.77781537e-01 1.65651929e+00 -1.24993794e-01 -2.70866454e-01 -1.09211671e+00 3.12942147e-01 -4.81523752e-01 7.56814718e-01 -7.55776405e-01 -3.51860821e-01 -6.42162144e-01 -9.10206139e-01 4.90118057e-01 2.45692044e-01 4.59747672e-01 -1.94147003e+00 -1.41483915e+00 5.24431050e-01 2.00519130e-01 -8.25835109e-01 8.32022488e-01 5.62522709e-01 -1.06589496e+00 -1.03172874e+00 -2.58477539e-01 -1.05369961e+00 2.85117328e-01 2.79991120e-01 1.14879131e+00 6.86219037e-01 -3.01225334e-01 3.34344566e-01 -3.99777055e-01 -5.36286056e-01 -5.95682919e-01 8.68085846e-02 1.97107345e-01 -9.31318462e-01 1.31173223e-01 -1.02194023e+00 -6.30482316e-01 1.26993498e-02 -9.05699968e-01 2.21798882e-01 6.12833023e-01 7.82426417e-01 1.27111390e-01 2.54931092e-01 9.30842698e-01 -4.02634531e-01 7.52082050e-01 -4.09052193e-01 -3.14925224e-01 2.09995463e-01 -4.64651793e-01 8.69522244e-02 7.88003087e-01 -1.25875339e-01 -9.92161036e-01 -1.25116542e-01 -2.35685140e-01 6.32998586e-01 -7.08032250e-01 5.11427641e-01 8.97951946e-02 -7.64301345e-02 8.24932218e-01 7.03795373e-01 -1.43837184e-01 8.09259117e-02 3.01206917e-01 1.90008804e-01 3.46787930e-01 -1.07226670e+00 3.47685635e-01 2.34256998e-01 1.44649088e-01 -1.22308087e+00 -5.30973732e-01 1.61661759e-01 -8.65302920e-01 -3.12370688e-01 1.07095587e+00 -3.93473119e-01 -1.17311907e+00 3.23934466e-01 -1.15858245e+00 -4.45614159e-01 -2.02296853e-01 2.33068224e-02 -1.36643827e+00 2.48581573e-01 -7.72859037e-01 -8.82861912e-01 -2.73107052e-01 -1.21151853e+00 7.90702045e-01 4.35628951e-01 -5.48655510e-01 -8.84837925e-01 3.49366635e-01 3.46519470e-01 2.91583866e-01 -2.99895350e-02 1.56211114e+00 -4.24463540e-01 -4.18384790e-01 4.10743952e-01 8.67343768e-02 -3.22359592e-01 -1.67573914e-01 2.32983008e-02 -9.68541026e-01 4.38867472e-02 3.40909362e-01 -7.49183297e-01 6.21450961e-01 2.20735624e-01 7.38189578e-01 4.10978012e-02 -4.27329205e-02 9.25317928e-02 1.16017509e+00 7.52245486e-01 7.00780272e-01 7.53566325e-01 -4.26849723e-02 1.36720467e+00 2.13735998e-01 2.38166481e-01 5.78760564e-01 2.52888530e-01 4.16106105e-01 3.92896414e-01 9.64050293e-02 -8.48358124e-03 1.86141402e-01 1.10007393e+00 -4.66070384e-01 1.27500236e-01 -1.00756848e+00 2.66891629e-01 -1.82754230e+00 -1.10792410e+00 1.61187187e-01 1.75473046e+00 8.27183306e-01 2.03853711e-01 2.98860282e-01 2.74092168e-01 4.80511039e-01 -3.73585880e-01 -6.71338439e-01 -7.35465050e-01 -2.18225926e-01 2.91972816e-01 -4.69484448e-01 3.73707712e-01 -7.96735942e-01 1.41815281e+00 7.86113977e+00 4.09566253e-01 -7.69955635e-01 -5.70973866e-02 3.65964860e-01 3.81925613e-01 1.13146275e-01 -8.44698772e-02 -3.98147941e-01 6.11422807e-02 1.31371427e+00 -1.99400470e-01 7.27501690e-01 6.69501603e-01 1.76762253e-01 -2.99701661e-01 -1.21114373e+00 5.23726821e-01 -1.57226592e-01 -1.20189369e+00 1.89670354e-01 5.94322607e-02 3.90969306e-01 -3.39572966e-01 -1.41259842e-02 5.65666020e-01 5.50163329e-01 -1.26121986e+00 7.35979021e-01 4.74035412e-01 -2.50218093e-01 -6.06239438e-01 2.07845688e-01 6.40468001e-01 -8.59463155e-01 -4.01292473e-01 -6.66588485e-01 -8.56566608e-01 -3.45288403e-02 -5.34099936e-02 -4.60480720e-01 2.31130183e-01 6.01783812e-01 3.48199993e-01 -6.87805474e-01 1.00025749e+00 -2.97319204e-01 6.69097081e-02 1.20882891e-01 -4.09276098e-01 2.05028102e-01 -7.31647089e-02 1.39334396e-01 9.62888420e-01 7.48428795e-03 6.61433280e-01 -4.85835932e-02 9.68185782e-01 6.03930652e-01 7.75017962e-02 -5.98835468e-01 -1.95982397e-01 3.45906794e-01 1.16081345e+00 -1.16445827e+00 -2.86032140e-01 -3.65628541e-01 6.63971245e-01 9.07754362e-01 -9.01699215e-02 -5.61909497e-01 1.25577688e-01 4.31601137e-01 -3.24476361e-01 1.50160402e-01 -6.25763774e-01 -5.28089643e-01 -8.68876576e-01 -7.18754709e-01 -1.04436517e+00 2.95042306e-01 -8.58712792e-01 -1.13053882e+00 6.68966711e-01 -1.60579443e-01 -4.95721340e-01 -4.33770657e-01 -9.40524876e-01 -2.12582633e-01 2.92968780e-01 -1.03594732e+00 -1.00384951e+00 -1.37821585e-01 4.08507645e-01 6.18209839e-01 -3.29552978e-01 1.30836725e+00 -4.59230214e-01 -4.45737094e-01 4.21097390e-02 -1.39275521e-01 -2.50481009e-01 2.15652771e-02 -1.27662420e+00 7.74109662e-02 3.80105048e-01 -5.49268685e-02 7.59459436e-01 8.55355084e-01 -3.53200942e-01 -1.17814219e+00 -2.62912154e-01 4.24146503e-01 -4.93350267e-01 6.75314903e-01 -2.74698615e-01 -1.01403666e+00 3.95302117e-01 7.69207418e-01 -7.98109412e-01 1.01595569e+00 2.65762955e-02 -1.87599123e-01 3.81974757e-01 -9.97750700e-01 8.12831283e-01 1.15062559e+00 -3.13647568e-01 -1.48118460e+00 1.12251928e-02 5.63453615e-01 3.41346562e-01 -9.42968726e-01 4.51706499e-01 7.04623997e-01 -1.51548409e+00 9.95389998e-01 -7.86971688e-01 7.61689067e-01 -4.88831311e-01 -1.44412100e-01 -1.10492027e+00 -1.05838656e+00 -2.22150907e-01 7.55653232e-02 7.34461367e-01 1.81354687e-01 -6.28035486e-01 5.88921010e-01 5.32912791e-01 -3.10601979e-01 -4.32896495e-01 -7.67742634e-01 -4.70154732e-01 8.44172060e-01 -2.03972667e-01 8.35941732e-02 7.29313791e-01 6.19606256e-01 5.78683436e-01 2.02809528e-01 -1.56756610e-01 6.47737086e-01 1.68413192e-01 3.73556286e-01 -1.50476170e+00 -2.97058523e-01 -1.00036538e+00 -3.62203091e-01 -4.45303172e-01 4.45303172e-01 -9.02848721e-01 1.38076097e-01 -1.50241208e+00 2.59481907e-01 -4.58138250e-03 -3.24215114e-01 2.89594054e-01 -9.12133157e-02 5.46372347e-02 3.19137514e-01 2.58742601e-01 -5.26899695e-01 6.43852293e-01 1.19038272e+00 3.07018816e-01 -1.21694565e-01 -6.61370099e-01 -1.00987303e+00 1.18129158e+00 1.18057060e+00 -2.32903883e-01 -4.39062178e-01 -3.48009199e-01 4.31591362e-01 -5.06581366e-02 2.71355718e-01 -1.37062609e+00 4.68111336e-01 -4.87403482e-01 5.22290111e-01 -9.47285444e-02 2.66202807e-01 -7.85645902e-01 -2.27732077e-01 1.10120857e+00 -3.41263324e-01 1.57315642e-01 3.30271453e-01 3.07273656e-01 3.21175270e-02 -5.37098169e-01 1.04324460e+00 -7.69732594e-01 -9.46310163e-01 -4.43333924e-01 -1.58482850e+00 -9.34472680e-02 1.20837200e+00 -4.68781173e-01 -3.75458002e-01 -1.28413588e-02 -9.55927670e-01 7.53437951e-02 3.85711968e-01 3.53092372e-01 5.98762751e-01 -9.10864651e-01 -4.61986154e-01 -1.53227329e-01 8.60939994e-02 -4.25520241e-01 2.72696882e-01 3.34691226e-01 -7.77877092e-01 8.71850848e-01 -1.08652329e+00 -2.01023281e-01 -7.23323405e-01 1.06897044e+00 4.91092533e-01 6.34072572e-02 -1.61131337e-01 5.63134193e-01 7.57087231e-01 -4.68477696e-01 -3.40278260e-02 2.95121688e-02 -6.41579509e-01 1.17814988e-02 5.50348938e-01 3.06996554e-01 -2.31079772e-01 -5.72487354e-01 -3.76063257e-01 4.91399258e-01 1.28194943e-01 -2.29752496e-01 1.46521676e+00 -1.03049114e-01 -6.88729525e-01 4.67364430e-01 2.30441898e-01 -5.88630974e-01 -8.75109673e-01 2.11982876e-01 5.04584610e-01 2.70364344e-01 -3.92147183e-01 -6.35831058e-01 -3.44942302e-01 1.06162035e+00 2.52320468e-01 5.67513704e-01 1.24335396e+00 2.01823473e-01 2.40500979e-02 7.66620755e-01 7.87390530e-01 -1.09707844e+00 4.84040439e-01 1.01987648e+00 1.20536733e+00 -9.53331590e-01 -3.28500755e-02 -2.93449104e-01 -4.06270593e-01 1.28692281e+00 9.11569476e-01 -5.27171671e-01 3.34258676e-01 5.17866984e-02 -3.88506025e-01 -2.54826635e-01 -1.20448291e+00 -5.83774626e-01 -1.28473630e-02 9.73146498e-01 9.12858248e-01 -1.17261618e-01 -6.38034940e-01 4.62057263e-01 -7.97497630e-01 -5.22128604e-02 4.27273333e-01 1.01145971e+00 -1.15670562e+00 -1.14727354e+00 -5.53818882e-01 -1.93349853e-01 -1.00505255e-01 -1.78383440e-01 -6.86279774e-01 7.13078022e-01 2.96452135e-01 1.04218268e+00 -1.97325587e-01 -2.08232537e-01 5.73311560e-02 2.75599360e-01 1.10629034e+00 -6.93325341e-01 -8.61713886e-01 -2.81962100e-02 -2.43573129e-01 -1.95868686e-01 -7.21334219e-01 -6.88383281e-01 -1.48328829e+00 -2.63806790e-01 4.55215126e-01 1.05766267e-01 2.74426043e-01 9.72548246e-01 2.11997330e-01 7.52832174e-01 -1.26970053e-01 -1.15541208e+00 -6.46580383e-02 -7.37368286e-01 -3.35466057e-01 1.47470534e-02 -4.91709970e-02 -8.03932786e-01 -5.94602488e-02 3.60251665e-01]
[4.295072078704834, 1.4249848127365112]
aeffdc2d-118b-4534-a5f1-c027bae9dfa1
mvtrans-multi-view-perception-of-transparent
2302.11683
null
https://arxiv.org/abs/2302.11683v1
https://arxiv.org/pdf/2302.11683v1.pdf
MVTrans: Multi-View Perception of Transparent Objects
Transparent object perception is a crucial skill for applications such as robot manipulation in household and laboratory settings. Existing methods utilize RGB-D or stereo inputs to handle a subset of perception tasks including depth and pose estimation. However, transparent object perception remains to be an open problem. In this paper, we forgo the unreliable depth map from RGB-D sensors and extend the stereo based method. Our proposed method, MVTrans, is an end-to-end multi-view architecture with multiple perception capabilities, including depth estimation, segmentation, and pose estimation. Additionally, we establish a novel procedural photo-realistic dataset generation pipeline and create a large-scale transparent object detection dataset, Syn-TODD, which is suitable for training networks with all three modalities, RGB-D, stereo and multi-view RGB. Project Site: https://ac-rad.github.io/MVTrans/
['Animesh Garg', 'Florian Shkurti', 'Alan Aspuru-Guzik', 'Saggi Eppel', 'Haoping Xu', 'Yuchi Zhao', 'Yi Ru Wang']
2023-02-22
null
null
null
null
['transparent-objects', 'transparent-object-detection', 'robot-manipulation']
['computer-vision', 'computer-vision', 'robots']
[ 2.71477163e-01 3.73542905e-02 2.43351325e-01 -5.76270998e-01 -4.62980509e-01 -6.38422906e-01 3.56314108e-02 -2.45409623e-01 -1.84669480e-01 3.28050166e-01 -2.14672923e-01 -2.50244945e-01 3.18751365e-01 -8.46516311e-01 -8.23800743e-01 -3.92090380e-01 5.91875613e-01 4.07702267e-01 6.73767924e-01 6.49872646e-02 2.05605060e-01 5.08113801e-01 -1.45482039e+00 4.14449424e-01 7.96391308e-01 1.23582232e+00 7.02729702e-01 7.70097017e-01 1.42902127e-02 2.89718300e-01 3.28304013e-03 -3.90187949e-01 5.58386683e-01 1.60912514e-01 -5.33942282e-01 2.84004748e-01 6.60373390e-01 -8.25767338e-01 -8.14120173e-02 7.91739762e-01 6.81670368e-01 -1.63100958e-01 3.31434757e-01 -1.25145292e+00 -3.42592210e-01 4.54548048e-03 -7.21867979e-01 -2.70877033e-01 7.74324834e-01 3.67522150e-01 4.23083425e-01 -9.06214714e-01 5.85097075e-01 1.26379716e+00 5.79846621e-01 7.73501575e-01 -8.86236310e-01 -4.33872968e-01 2.64141798e-01 -5.89440838e-02 -8.66913676e-01 -2.75029778e-01 9.11445558e-01 -4.84035462e-01 7.15326309e-01 1.76073477e-01 7.71324933e-01 1.33322632e+00 2.10937500e-01 8.97279680e-01 1.44380522e+00 -4.76148315e-02 1.08107790e-01 2.35426396e-01 -3.03645283e-01 8.24929535e-01 1.48148835e-01 3.02583333e-02 -6.66262090e-01 2.64427871e-01 1.31149971e+00 3.02552730e-01 -3.79548788e-01 -9.31273043e-01 -1.42378020e+00 2.33962178e-01 4.85212684e-01 -3.90851110e-01 -3.57902318e-01 1.25518844e-01 4.78846878e-02 1.88154075e-02 4.31475133e-01 2.80775875e-02 -8.70608509e-01 -1.34025097e-01 -1.91410527e-01 1.88565832e-02 6.45248353e-01 1.42319429e+00 8.43519330e-01 -2.90427297e-01 2.55006194e-01 5.70216060e-01 6.11361682e-01 5.56721270e-01 2.13757992e-01 -1.30861866e+00 7.84761727e-01 6.93976998e-01 2.64823228e-01 -5.44831276e-01 -6.35595977e-01 2.24406332e-01 -4.53963518e-01 4.68906224e-01 4.93908226e-01 -1.94309168e-02 -1.07560515e+00 1.19755828e+00 1.00681829e+00 -3.39626849e-01 1.58927729e-03 1.19154286e+00 1.20469737e+00 2.62891114e-01 -3.70584607e-01 2.22824141e-01 1.25068760e+00 -1.17944920e+00 -3.06425899e-01 -5.40536463e-01 1.95225164e-01 -7.89369285e-01 1.33230686e+00 7.39313483e-01 -1.17061770e+00 -4.81995732e-01 -8.56927395e-01 -5.39094806e-01 -4.69312489e-01 2.47904614e-01 1.00185454e+00 6.08420908e-01 -8.25294971e-01 3.66320938e-01 -9.46204662e-01 -4.81884122e-01 3.85237545e-01 3.94822180e-01 -6.55403316e-01 -3.59827071e-01 -4.83311832e-01 5.97746432e-01 3.21287036e-01 3.59373957e-01 -8.56265664e-01 -5.79469264e-01 -8.80037248e-01 -6.05874538e-01 5.48055232e-01 -1.11102879e+00 1.26060653e+00 -3.96746963e-01 -1.72122657e+00 1.21000576e+00 -8.93169418e-02 1.75271899e-01 7.73016453e-01 -5.83710015e-01 2.68032432e-01 4.45344120e-01 3.86980511e-02 7.51956642e-01 7.71510541e-01 -1.76603448e+00 -3.05484593e-01 -8.69166791e-01 3.36119920e-01 4.01331723e-01 -4.06902209e-02 -2.12494507e-01 -6.25425756e-01 -2.07035407e-01 7.91155636e-01 -8.20563912e-01 -1.35255173e-01 5.99249899e-01 -7.20162213e-01 3.75091285e-02 7.70620823e-01 -6.31814063e-01 8.72188136e-02 -1.90788150e+00 9.21111181e-02 -1.23044059e-01 3.15163106e-01 -1.53722361e-01 1.32868066e-01 2.27709264e-01 2.21808627e-01 -1.70165554e-01 -1.77083284e-01 -7.69205093e-01 5.64437397e-02 -1.60366576e-02 -2.00921725e-02 5.50827026e-01 -1.98459789e-01 6.67469919e-01 -7.74470389e-01 -6.62958503e-01 6.06285512e-01 4.31944877e-01 -4.25079644e-01 4.60588634e-01 -4.74335134e-01 9.64664340e-01 -6.31866634e-01 1.21428168e+00 1.05293345e+00 -5.30206934e-02 -1.02084503e-01 -4.24253494e-01 -1.92284569e-01 1.20642126e-01 -1.10575438e+00 2.41995382e+00 -6.35843694e-01 2.32147306e-01 4.27315652e-01 -2.01946661e-01 7.57341981e-01 2.31912106e-01 6.08197093e-01 -5.24731398e-01 1.87551558e-01 2.43112504e-01 -5.58809698e-01 -8.71817112e-01 4.57118064e-01 9.17591229e-02 5.32439165e-02 2.52904356e-01 -2.10221842e-01 -8.85974765e-01 -2.55668461e-01 -4.83648367e-02 8.38155210e-01 9.79472458e-01 -9.91811231e-02 5.28182626e-01 3.06589361e-02 3.80120650e-02 5.49678802e-01 3.10705304e-01 -1.83629870e-01 1.15474713e+00 8.68820027e-02 -2.66312391e-01 -9.46080983e-01 -1.50710285e+00 -8.03691968e-02 7.96134174e-01 6.74195349e-01 -3.24790627e-02 -5.14753342e-01 -5.40396035e-01 9.93579179e-02 3.03982526e-01 -2.31349602e-01 2.10577354e-01 -4.00481492e-01 -4.63263094e-01 -1.33580640e-01 6.03985071e-01 8.33030522e-01 -9.72294986e-01 -9.82622147e-01 2.66300067e-02 -2.00951502e-01 -1.52355146e+00 -2.86827624e-01 1.80158421e-01 -1.10552418e+00 -1.16823137e+00 -6.24537468e-01 -5.29514372e-01 6.47788525e-01 6.95437908e-01 9.35998261e-01 -5.11147022e-01 -3.07509512e-01 7.27815092e-01 -3.39644641e-01 -5.31359136e-01 2.13486291e-02 -1.91384539e-01 2.26993226e-02 -3.32646251e-01 -1.90073531e-02 -7.60670602e-01 -1.10052502e+00 4.51714396e-01 -6.55448020e-01 4.89093781e-01 6.43658459e-01 7.40670562e-02 7.43465543e-01 -5.44033110e-01 -3.43364291e-02 -6.84913933e-01 1.34228870e-01 -1.61594912e-01 -5.97763658e-01 2.16488503e-02 -3.21602561e-02 -6.17774665e-01 1.96350649e-01 -2.35498115e-01 -1.04974651e+00 6.42357290e-01 -1.99644744e-01 -7.30937600e-01 -5.05714238e-01 -7.16843233e-02 -6.00965202e-01 -1.95258513e-01 3.88790429e-01 1.48504838e-01 -1.73594877e-01 -4.52591121e-01 3.52759212e-01 7.95158327e-01 4.20706540e-01 -4.58978891e-01 6.70980752e-01 9.76517916e-01 -2.07935706e-01 -6.93159223e-01 -7.38449931e-01 -6.41644835e-01 -1.01112592e+00 -3.57834071e-01 9.66037273e-01 -1.16716433e+00 -1.06551123e+00 7.14420557e-01 -1.42296827e+00 -6.91511631e-01 3.37984413e-02 5.75139046e-01 -9.13300991e-01 4.29916710e-01 -5.28870225e-01 -8.09978783e-01 -3.07710320e-01 -1.21420586e+00 1.67524397e+00 2.90233761e-01 3.55353206e-02 -5.53869307e-01 -3.36431533e-01 9.43500996e-01 -9.93275717e-02 4.91042405e-01 3.98337692e-01 2.65913665e-01 -1.19440508e+00 -1.13874875e-01 -3.28698546e-01 1.02881715e-01 1.44566029e-01 -1.73015863e-01 -1.39707255e+00 -2.25269068e-02 6.54085726e-02 -6.92220271e-01 6.85633063e-01 3.26104313e-01 1.43970764e+00 8.39979351e-02 -3.34603280e-01 9.92806077e-01 1.42710781e+00 -5.04128188e-02 4.03964877e-01 3.38354379e-01 1.27108228e+00 7.07332253e-01 8.94617796e-01 4.18346941e-01 8.87587667e-01 7.79177904e-01 9.54308569e-01 -9.07994136e-02 -1.76899098e-02 -1.57353237e-01 3.06568533e-01 6.60068154e-01 -2.31806278e-01 -2.82094747e-01 -9.69145477e-01 2.98321754e-01 -1.57261956e+00 -4.14682567e-01 -5.45948684e-01 2.18582892e+00 7.24589109e-01 1.25453442e-01 6.63403943e-02 -8.30372497e-02 3.14322382e-01 -3.23260754e-01 -8.84347796e-01 -3.30631018e-01 9.88458246e-02 -2.50873417e-01 6.61118984e-01 2.26212263e-01 -9.45193887e-01 9.94425178e-01 5.00568295e+00 2.37235144e-01 -1.07257295e+00 2.01209560e-01 3.32589775e-01 6.97996980e-03 -3.33995402e-01 -3.33540082e-01 -5.47801912e-01 2.75124490e-01 1.87980443e-01 5.98145366e-01 2.01175779e-01 9.59534466e-01 3.14545482e-01 -7.42226720e-01 -1.31226623e+00 1.28274143e+00 1.43135175e-01 -6.84142172e-01 -2.79703110e-01 -1.72386058e-02 6.59953356e-01 3.52345109e-01 4.12901640e-02 -2.40251735e-01 2.90247977e-01 -6.53552055e-01 9.29146171e-01 4.34897155e-01 7.50115514e-01 -1.36438012e-01 4.13936913e-01 4.79033798e-01 -1.11373651e+00 4.23312932e-02 -2.80791432e-01 2.20014513e-01 2.72239000e-01 6.92775071e-01 -8.89302254e-01 5.12667835e-01 9.90450680e-01 6.80656195e-01 -4.75940675e-01 1.23906946e+00 -3.99718553e-01 -6.05175532e-02 -4.55976635e-01 1.99648693e-01 -1.95740372e-01 -2.99709141e-01 3.82644415e-01 7.21992850e-01 2.11313441e-01 -9.82891768e-02 3.02023947e-01 9.92548823e-01 2.62455307e-02 -4.01780933e-01 -7.66778648e-01 4.36249107e-01 3.61656368e-01 1.40624452e+00 -9.06486452e-01 2.76995823e-02 -4.08875138e-01 1.46567011e+00 2.69015342e-01 3.86288196e-01 -8.93945694e-01 -1.99405923e-01 4.38654631e-01 2.28272304e-02 1.23951815e-01 -6.03961468e-01 -5.49574912e-01 -1.20397568e+00 5.75098217e-01 -4.12653297e-01 -1.39487907e-01 -1.48996949e+00 -1.09782112e+00 3.51971567e-01 -3.86507809e-02 -1.58429396e+00 1.21954940e-01 -1.01259911e+00 -3.23787272e-01 8.01593840e-01 -1.69768119e+00 -1.57033098e+00 -8.90374839e-01 5.53543746e-01 6.33938670e-01 3.87572557e-01 6.34678662e-01 1.31321579e-01 -3.76439512e-01 4.19335328e-02 -2.35854879e-01 -1.09446801e-01 7.36874402e-01 -1.36857617e+00 1.34968057e-01 4.88003999e-01 -3.39173526e-01 4.18594688e-01 4.48353529e-01 -5.56818843e-01 -1.99905717e+00 -1.06271124e+00 2.34131347e-02 -9.39912915e-01 2.58465737e-01 -7.66009152e-01 -3.92144799e-01 7.88041353e-01 -1.55750692e-01 1.00426786e-01 2.87913054e-01 -3.46338332e-01 -1.61287159e-01 -2.35882580e-01 -1.52336502e+00 5.31639755e-01 1.41451895e+00 -5.79182804e-01 -2.15222910e-01 5.42238772e-01 9.46413457e-01 -1.10284758e+00 -1.01404285e+00 3.67789119e-01 9.47031736e-01 -1.36815715e+00 1.23661506e+00 2.23984167e-01 7.08909452e-01 -3.46347421e-01 -1.23192616e-01 -1.12939692e+00 3.41958910e-01 -4.20728356e-01 -1.15905739e-01 1.14534271e+00 2.77853251e-01 -6.94435894e-01 1.13253164e+00 7.66412318e-01 -5.48775911e-01 -8.05299342e-01 -9.42982852e-01 -5.33513010e-01 -3.55928779e-01 -6.34748340e-01 3.99970829e-01 5.79997301e-01 -1.80829525e-01 1.12931587e-01 4.14933749e-02 5.16749442e-01 8.04627120e-01 6.82467461e-01 1.13214600e+00 -1.09200275e+00 -1.41919374e-01 -4.13361751e-02 -2.12502241e-01 -1.36111522e+00 -2.74156958e-01 -5.19958794e-01 3.51487666e-01 -2.08874536e+00 8.11442733e-02 -5.13589144e-01 4.73152757e-01 3.63342524e-01 1.23648778e-01 2.86863357e-01 1.87916998e-02 9.33665633e-02 -5.72213054e-01 6.13684595e-01 1.92068565e+00 3.80506143e-02 -2.94549912e-01 1.65417477e-01 -3.09475780e-01 9.82944191e-01 1.07744277e+00 -1.67093098e-01 -3.76988918e-01 -9.02785778e-01 2.09219545e-01 2.16935799e-01 7.77613878e-01 -1.12705898e+00 3.84819731e-02 -3.26722920e-01 5.32545686e-01 -9.33293283e-01 1.06494737e+00 -1.09946668e+00 -2.14300990e-01 2.39166006e-01 2.46475220e-01 -6.57680929e-02 1.63265213e-01 5.08456707e-01 2.97586530e-01 1.37661293e-01 6.22942388e-01 -7.09085166e-01 -8.25927258e-01 5.55177987e-01 -9.60082784e-02 -2.83843964e-01 1.17672431e+00 -9.19381082e-01 -3.37198943e-01 -1.79148689e-02 -6.89812899e-01 4.44679856e-01 7.90949047e-01 4.40529108e-01 1.27169502e+00 -9.27161098e-01 -3.16910386e-01 2.48480827e-01 3.45821798e-01 8.62757564e-01 2.34834567e-01 9.64875281e-01 -9.15108323e-01 1.70201331e-01 -1.98218510e-01 -8.55795562e-01 -1.27380013e+00 2.56514937e-01 3.06379139e-01 3.66797894e-01 -6.74794614e-01 9.76274669e-01 2.60958523e-01 -1.05014420e+00 1.86648279e-01 -8.43624413e-01 2.59606034e-01 -3.77688468e-01 -1.61341857e-02 2.57146567e-01 1.14537314e-01 -3.28316897e-01 -2.72174448e-01 9.05009985e-01 4.20105368e-01 -5.69108501e-02 1.39543819e+00 -5.99248707e-01 -1.60237640e-01 7.01523483e-01 9.29071426e-01 -7.12210685e-02 -1.61119127e+00 2.13627800e-01 -4.84681606e-01 -5.92780769e-01 -8.92228484e-02 -6.63866043e-01 -9.99418914e-01 1.20476758e+00 6.58142328e-01 -1.84453860e-01 9.36111212e-01 1.77124947e-01 8.40791881e-01 4.67630655e-01 9.56501245e-01 -1.10139990e+00 3.95666152e-01 4.54862744e-01 9.83938932e-01 -1.61472750e+00 8.69351998e-02 -1.14911258e+00 -5.68773806e-01 1.05545390e+00 1.29995143e+00 2.76217580e-01 3.26272398e-01 1.90272838e-01 2.09092274e-01 -2.71704346e-01 -3.73757452e-01 -2.96093136e-01 -1.25878558e-01 9.33173060e-01 2.67457515e-01 -5.34738190e-02 4.71294284e-01 1.14062756e-01 -2.50245929e-01 1.25714540e-01 5.73674560e-01 1.18816137e+00 -3.32812250e-01 -7.34614134e-01 -5.40106833e-01 3.02766472e-01 -7.28381351e-02 9.64017361e-02 -4.22413707e-01 6.33555412e-01 4.21955615e-01 8.62522483e-01 -7.13208914e-02 -3.77652049e-01 4.09026861e-01 -2.99591511e-01 8.93564820e-01 -9.55058157e-01 -3.25368553e-01 -1.23562895e-01 5.05808629e-02 -9.28556025e-01 -6.10229254e-01 -5.62032044e-01 -1.28789842e+00 -5.26013374e-02 -1.94765493e-01 -6.35240853e-01 1.07529271e+00 5.97641885e-01 2.11428925e-01 3.87604147e-01 6.72055364e-01 -1.52182305e+00 -4.23948467e-02 -1.01318455e+00 -4.33822185e-01 1.66566908e-01 3.56201768e-01 -5.71788073e-01 -6.79471996e-03 1.37084916e-01]
[7.113190650939941, -1.978243112564087]
24c99074-245f-4c3a-90e8-5382bb5463e7
query-focused-abstractive-summarization
1801.07704
null
http://arxiv.org/abs/1801.07704v2
http://arxiv.org/pdf/1801.07704v2.pdf
Query Focused Abstractive Summarization: Incorporating Query Relevance, Multi-Document Coverage, and Summary Length Constraints into seq2seq Models
Query Focused Summarization (QFS) has been addressed mostly using extractive methods. Such methods, however, produce text which suffers from low coherence. We investigate how abstractive methods can be applied to QFS, to overcome such limitations. Recent developments in neural-attention based sequence-to-sequence models have led to state-of-the-art results on the task of abstractive generic single document summarization. Such models are trained in an end to end method on large amounts of training data. We address three aspects to make abstractive summarization applicable to QFS: (a)since there is no training data, we incorporate query relevance into a pre-trained abstractive model; (b) since existing abstractive models are trained in a single-document setting, we design an iterated method to embed abstractive models within the multi-document requirement of QFS; (c) the abstractive models we adapt are trained to generate text of specific length (about 100 words), while we aim at generating output of a different size (about 250 words); we design a way to adapt the target size of the generated summaries to a given size ratio. We compare our method (Relevance Sensitive Attention for QFS) to extractive baselines and with various ways to combine abstractive models on the DUC QFS datasets and demonstrate solid improvements on ROUGE performance.
['Tal Baumel', 'Matan Eyal', 'Michael Elhadad']
2018-01-23
null
null
null
null
['query-based-extractive-summarization']
['natural-language-processing']
[ 6.26626134e-01 2.17536911e-01 -1.77089378e-01 -3.97584476e-02 -1.41146469e+00 -5.56980908e-01 6.76340461e-01 3.00271034e-01 -5.85398555e-01 8.96425128e-01 9.59045172e-01 -2.18716115e-02 2.58363318e-02 -6.55763865e-01 -7.88631856e-01 -2.33385712e-01 1.89092278e-01 6.84167504e-01 4.51206475e-01 -7.35837042e-01 7.40662813e-01 2.55283803e-01 -1.42875612e+00 4.91004705e-01 1.24659729e+00 4.76431996e-01 3.15556824e-01 1.50063741e+00 -1.07770771e-01 5.70863485e-01 -1.31107092e+00 -3.29613209e-01 3.79996970e-02 -8.21471393e-01 -1.19588912e+00 4.07152213e-02 8.57336998e-01 -4.98787791e-01 -2.79020071e-01 7.83433080e-01 8.40796888e-01 3.70702147e-01 1.01490164e+00 -8.40762675e-01 -9.23796654e-01 7.88015842e-01 -2.28268251e-01 5.11467218e-01 5.34203589e-01 1.39796287e-01 1.30564272e+00 -7.18205452e-01 5.89552701e-01 1.35074413e+00 2.94603318e-01 7.40489602e-01 -1.24115872e+00 -4.50077206e-02 1.38712004e-01 3.67412507e-03 -9.36454713e-01 -7.44269550e-01 3.58000159e-01 2.15820801e-02 1.58338523e+00 6.14816248e-01 5.20151734e-01 1.13301516e+00 2.58679390e-01 1.00581086e+00 2.99288660e-01 -6.65677905e-01 1.96139649e-01 -1.25613108e-01 1.51611775e-01 2.37745464e-01 1.89187542e-01 -3.03852558e-01 -5.53812981e-01 -2.16363132e-01 3.09339941e-01 -3.36151510e-01 -3.66475910e-01 1.70514598e-01 -1.17945027e+00 1.12601531e+00 2.30561718e-01 2.59575456e-01 -4.79763150e-01 3.70506167e-01 8.18396926e-01 4.27454710e-01 7.27969408e-01 1.02317965e+00 -4.22143281e-01 -8.15301389e-02 -1.33529818e+00 8.02110553e-01 9.42886055e-01 1.16184556e+00 4.60123003e-01 2.66256094e-01 -1.16588318e+00 1.06042004e+00 -1.42229736e-01 6.68342352e-01 7.74240851e-01 -9.67257619e-01 8.24558258e-01 2.41009101e-01 3.17807585e-01 -6.01557612e-01 -1.76564336e-01 -4.71302569e-01 -6.74678802e-01 -4.03543234e-01 -2.18542516e-01 -2.14814737e-01 -9.64764655e-01 1.49629629e+00 -2.38293752e-01 -4.99696821e-01 2.98368335e-01 6.72128677e-01 1.01429880e+00 1.08029974e+00 2.51068803e-03 -3.38070482e-01 1.21050894e+00 -1.30695331e+00 -6.76125348e-01 -4.50206876e-01 7.29777634e-01 -7.26373553e-01 1.32151866e+00 2.62162983e-01 -1.45079112e+00 -5.45582473e-01 -1.15863216e+00 -2.48255074e-01 -2.13760614e-01 2.64805317e-01 3.65697034e-02 4.92825955e-01 -1.47918046e+00 7.46563435e-01 -4.27598983e-01 -6.66270673e-01 1.37413740e-01 3.88360530e-01 1.83876546e-03 1.92061216e-01 -1.31938958e+00 9.14355636e-01 8.30480933e-01 -2.79075474e-01 -7.50691891e-01 -3.87234092e-01 -7.43067622e-01 4.55577493e-01 5.90976179e-01 -1.33643925e+00 1.69102085e+00 -9.36205268e-01 -1.63345122e+00 3.72309774e-01 -2.02190161e-01 -6.24842107e-01 2.76184678e-01 -5.45135260e-01 -9.05588642e-02 5.41193783e-01 2.36668006e-01 7.83033550e-01 9.44680989e-01 -1.09608233e+00 -5.63126206e-01 -1.34340916e-02 3.20567861e-02 5.05201876e-01 -5.47136188e-01 1.00910544e-01 -4.56997037e-01 -9.20760751e-01 -4.34065938e-01 -7.94998109e-01 -2.78887421e-01 -6.80049360e-01 -7.71229684e-01 -3.59762341e-01 4.90577847e-01 -7.51321733e-01 1.60690379e+00 -1.46029162e+00 2.84148127e-01 -3.47143561e-01 -8.89732689e-02 6.78129435e-01 -6.68068349e-01 9.98629451e-01 2.83865988e-01 3.98589224e-01 -4.84923393e-01 -5.52592695e-01 3.66308950e-02 -7.45794773e-02 -5.50906062e-01 -1.63427651e-01 4.89610910e-01 1.13358724e+00 -1.08062923e+00 -5.89512229e-01 -2.29187086e-01 -7.50329494e-02 -9.11504507e-01 4.91937697e-01 -5.49613237e-01 -8.97015855e-02 -5.27009845e-01 1.86316997e-01 3.77005011e-01 -5.31285033e-02 -3.12638551e-01 5.02285026e-02 5.28492257e-02 5.63719690e-01 -6.67305648e-01 1.73559785e+00 -5.24934888e-01 4.60203290e-01 -3.25794250e-01 -9.62200701e-01 7.22161472e-01 2.21199915e-01 1.55107275e-01 -6.13794982e-01 -1.42447785e-01 2.00957254e-01 -2.14121953e-01 -5.57057321e-01 1.46305716e+00 -2.28059292e-02 -2.17159748e-01 6.11024976e-01 3.04036319e-01 -5.52094042e-01 8.82460415e-01 6.90418005e-01 1.27441168e+00 1.17634408e-01 3.68518263e-01 -7.48020411e-02 4.59327459e-01 2.07900956e-01 1.24741316e-01 1.25525045e+00 1.45699665e-01 1.10025239e+00 4.89920944e-01 1.31930918e-01 -1.23117828e+00 -7.58805394e-01 4.48758245e-01 1.44829452e+00 -1.59339875e-01 -6.94377720e-01 -8.42644334e-01 -9.68238890e-01 -1.28094554e-01 1.23123038e+00 -3.90893936e-01 -4.87080276e-01 -9.24881458e-01 -6.79511964e-01 6.52571023e-01 6.26759112e-01 2.86941469e-01 -1.37936652e+00 -7.24714100e-01 5.27445793e-01 -3.67402673e-01 -8.77194405e-01 -1.10953426e+00 1.05220713e-01 -1.07132626e+00 -4.77149487e-01 -1.06928694e+00 -5.26032329e-01 2.90260047e-01 3.27559531e-01 1.63777196e+00 -4.96935472e-02 2.43869022e-01 4.64199692e-01 -6.73746943e-01 -5.58408678e-01 -8.93379927e-01 6.93037033e-01 -2.23578602e-01 -4.73066956e-01 3.59180123e-02 -2.42091760e-01 -6.83585942e-01 -1.47733256e-01 -1.47069561e+00 -5.56587242e-02 9.28669155e-01 8.42386425e-01 1.58134297e-01 -6.21542931e-01 1.20050287e+00 -9.36524630e-01 1.38759458e+00 -3.03651780e-01 -7.29663372e-02 3.42132181e-01 -6.87454343e-01 4.09628808e-01 8.86586130e-01 -3.65029901e-01 -9.16360795e-01 -4.73127246e-01 -4.46530402e-01 -1.66029304e-01 1.28291711e-01 6.45893514e-01 1.22778744e-01 6.16480529e-01 1.01638567e+00 4.93725836e-01 2.45426055e-02 -3.24477434e-01 7.20865786e-01 9.41950738e-01 4.01807100e-01 -4.28411096e-01 4.48663503e-01 -9.13999230e-03 -3.31846505e-01 -8.99063289e-01 -7.42663801e-01 -6.36607349e-01 -4.57309425e-01 1.51379809e-01 5.91200888e-01 -6.70246601e-01 -8.31280649e-02 3.20180506e-02 -1.44143069e+00 -2.48712704e-01 -6.44855499e-01 1.59445807e-01 -8.76608551e-01 6.68651581e-01 -6.06608033e-01 -7.01970875e-01 -1.20045245e+00 -8.29693377e-01 1.59355259e+00 1.51745960e-01 -6.04359567e-01 -7.97774434e-01 3.54403287e-01 2.53495514e-01 5.88261306e-01 -2.18673855e-01 8.46545279e-01 -9.64384437e-01 -1.78531274e-01 -1.23335965e-01 -1.49011657e-01 5.48180103e-01 5.74504845e-02 2.62855925e-02 -6.26559377e-01 -4.39523071e-01 -2.91184247e-01 -5.30786753e-01 1.40228200e+00 4.11021858e-01 9.11165118e-01 -8.35570097e-01 -1.89272776e-01 8.30570310e-02 1.17983067e+00 -3.06623429e-02 6.58397377e-01 2.32560083e-01 4.51871186e-01 4.96900052e-01 6.62466228e-01 2.34941721e-01 2.63183415e-01 7.77505457e-01 2.29368389e-01 7.75615349e-02 -3.77713263e-01 -3.06489587e-01 5.73219180e-01 1.05972588e+00 5.25509790e-02 -8.58270824e-01 -4.31620598e-01 8.72627258e-01 -2.04933262e+00 -1.08545566e+00 6.83799312e-02 2.21173072e+00 9.84389782e-01 1.37116656e-01 3.88234943e-01 -1.05603293e-01 4.55926985e-01 3.52440029e-01 -4.28054810e-01 -1.01971090e+00 -7.35590309e-02 2.75377721e-01 3.90808433e-01 3.87238771e-01 -9.31560993e-01 9.80017900e-01 6.56936789e+00 9.98092949e-01 -1.06889331e+00 -1.15804389e-01 3.80439281e-01 -3.87907654e-01 -4.97396082e-01 -1.26927063e-01 -9.53948200e-01 3.15315664e-01 1.37905169e+00 -4.80833411e-01 1.55459689e-02 5.62194824e-01 3.42415094e-01 -6.61392063e-02 -1.13223851e+00 4.81618792e-01 4.72966552e-01 -1.31206632e+00 6.77311778e-01 -1.18826486e-01 7.89475203e-01 -6.30231574e-02 -2.15910345e-01 8.13024223e-01 1.93755105e-01 -9.33447242e-01 6.63682759e-01 5.82138240e-01 7.85198033e-01 -7.19004691e-01 8.13407242e-01 4.68960106e-01 -5.91330945e-01 -5.73558733e-02 -7.13603079e-01 2.04991013e-01 2.78525025e-01 4.31286961e-01 -1.16261613e+00 8.74195218e-01 3.17769676e-01 4.75013167e-01 -7.29207516e-01 1.00099003e+00 -8.76939669e-02 6.49977565e-01 -4.04204521e-03 -4.79647696e-01 4.68395084e-01 -2.80283508e-03 9.85110939e-01 1.59572756e+00 3.95307839e-01 -2.16089353e-01 -1.12827145e-01 8.13557208e-01 -2.41755739e-01 3.98414910e-01 -5.20346403e-01 -2.03437001e-01 2.06615597e-01 8.83222759e-01 -4.98261154e-01 -6.36690259e-01 -2.28782743e-01 1.33951688e+00 3.58931988e-01 4.16365802e-01 -4.43287879e-01 -7.57032514e-01 -2.39590462e-02 7.41940886e-02 5.82725585e-01 6.44542947e-02 -5.97891398e-03 -1.09478867e+00 3.24984118e-02 -9.72150147e-01 3.45954388e-01 -8.39652479e-01 -1.11379302e+00 6.59825861e-01 3.87400448e-01 -1.13438892e+00 -8.90962005e-01 2.99871638e-02 -7.84299672e-01 9.28939462e-01 -1.48388743e+00 -9.04523909e-01 1.29750386e-01 7.42593482e-02 1.17733598e+00 -1.62484184e-01 8.44632626e-01 -1.52666405e-01 -3.51259053e-01 4.68135566e-01 3.71857524e-01 -2.45812103e-01 9.56671238e-01 -1.56973767e+00 9.00314212e-01 8.83693516e-01 1.56906798e-01 7.15227127e-01 1.04161608e+00 -7.04792440e-01 -1.21823788e+00 -1.34639275e+00 1.13007140e+00 -5.90970933e-01 1.62435472e-01 -2.50180572e-01 -8.69383991e-01 4.76847738e-01 6.49400592e-01 -4.27835733e-01 4.42111373e-01 5.29271476e-02 1.28279895e-01 3.57102565e-02 -8.30678761e-01 7.56004870e-01 8.52827609e-01 -2.21155405e-01 -1.01137042e+00 4.66878563e-01 1.15425813e+00 -1.59434736e-01 -4.95864034e-01 2.66742676e-01 3.02242607e-01 -8.27694654e-01 7.80988038e-01 -7.79634178e-01 7.94570863e-01 -1.13200992e-01 1.63750306e-01 -1.89936781e+00 -2.35279560e-01 -6.86469197e-01 -4.58629787e-01 1.25252497e+00 4.94971275e-01 -4.21311438e-01 6.68525279e-01 1.28048450e-01 -5.50366223e-01 -8.05289388e-01 -7.01528251e-01 -7.80032337e-01 3.67371202e-01 2.16157794e-01 4.58588153e-01 4.42819633e-02 3.84878367e-02 1.03992069e+00 -3.37454021e-01 -4.15928304e-01 1.34483829e-01 1.12361386e-01 9.10860240e-01 -8.93830657e-01 -3.00372720e-01 -6.62165046e-01 1.35568038e-01 -1.55098939e+00 8.23455155e-02 -6.86685145e-01 3.18554044e-01 -1.99745524e+00 4.15659577e-01 1.14067756e-01 5.29886894e-02 8.53201225e-02 -8.47936630e-01 5.89375757e-02 3.80003005e-01 2.63456762e-01 -7.31291950e-01 1.06886005e+00 1.21062076e+00 -3.82863909e-01 -4.49670017e-01 1.85781553e-01 -1.09680295e+00 4.64792028e-02 6.47467315e-01 -4.56793457e-01 -6.23692811e-01 -3.62259001e-01 2.20861956e-01 2.73859471e-01 -6.47488534e-02 -7.75653303e-01 1.61048517e-01 8.87660757e-02 5.80117330e-02 -8.23813140e-01 1.26589611e-01 -2.72765830e-02 -5.20287871e-01 3.60578567e-01 -7.25092649e-01 1.86630830e-01 2.89801359e-01 6.29397869e-01 -3.03409517e-01 -7.49797463e-01 4.93601501e-01 -3.51947308e-01 -2.60074973e-01 7.95671716e-02 -5.12418866e-01 3.11759055e-01 3.32764387e-01 -2.32792020e-01 -4.10264611e-01 -8.47739756e-01 -2.53691703e-01 3.99489701e-01 2.64426827e-01 3.72524977e-01 6.01633191e-01 -1.03119969e+00 -1.35896719e+00 -2.55831748e-01 2.90405095e-01 2.37582829e-02 7.77859315e-02 5.04030526e-01 -5.17798841e-01 9.16522205e-01 1.86110884e-02 -3.92704219e-01 -1.23603415e+00 6.00370646e-01 -2.54727378e-02 -6.70881450e-01 -4.98009354e-01 6.49303138e-01 9.53644440e-02 -3.51330519e-01 -5.63368984e-02 -3.70423079e-01 -6.03540301e-01 3.18738431e-01 5.24756610e-01 4.68027890e-01 4.59584355e-01 -4.83880758e-01 7.35081285e-02 3.04764211e-01 -5.19318163e-01 -3.14222962e-01 1.28079569e+00 -1.50828049e-01 1.92747369e-01 1.16666444e-01 1.19542336e+00 1.87026158e-01 -9.38280463e-01 -1.03769556e-01 2.13773437e-02 -6.53093532e-02 -3.01129036e-02 -7.14851916e-01 -4.61936504e-01 5.84777296e-01 -3.59373130e-02 5.03564000e-01 1.12125278e+00 -1.44873932e-01 9.94966686e-01 6.43800259e-01 -2.17357919e-01 -1.18717933e+00 4.08909440e-01 7.83642411e-01 1.35649359e+00 -1.05988061e+00 2.82674134e-01 3.26812156e-02 -8.00713539e-01 1.28035355e+00 3.11367154e-01 -1.46619305e-01 -1.87145323e-01 -2.44776979e-01 -1.80225804e-01 -1.32778868e-01 -1.01164317e+00 -2.94949800e-01 5.24545610e-01 5.47805369e-01 4.58622992e-01 -2.49965027e-01 -6.51441693e-01 4.45694804e-01 -4.97150183e-01 -4.28417027e-02 9.77871537e-01 9.12988067e-01 -6.80807769e-01 -9.33404922e-01 -3.30259532e-01 8.68317664e-01 -6.24434054e-01 -4.00998294e-01 -5.86313963e-01 5.64288199e-01 -6.74482942e-01 1.22007918e+00 -1.27238274e-01 2.36616712e-02 5.78336596e-01 9.21054408e-02 3.92691165e-01 -1.08441091e+00 -9.52409863e-01 3.95637564e-02 4.89066124e-01 -1.94989115e-01 -2.73788631e-01 -5.52257895e-01 -8.81447673e-01 6.91717416e-02 -5.13323486e-01 5.13323605e-01 4.85892177e-01 8.22729111e-01 5.04091799e-01 7.65628874e-01 5.49030483e-01 -1.05831945e+00 -1.09367681e+00 -1.53588867e+00 -3.51294219e-01 3.20153922e-01 5.64582467e-01 5.34986630e-02 -2.90030181e-01 -2.24143546e-02]
[12.32159423828125, 9.331268310546875]
a50a2ad7-e20f-42a5-b4b8-4b10694ba0f9
cross-corpora-spoken-language-identification
2302.05110
null
https://arxiv.org/abs/2302.05110v1
https://arxiv.org/pdf/2302.05110v1.pdf
Cross-Corpora Spoken Language Identification with Domain Diversification and Generalization
This work addresses the cross-corpora generalization issue for the low-resourced spoken language identification (LID) problem. We have conducted the experiments in the context of Indian LID and identified strikingly poor cross-corpora generalization due to corpora-dependent non-lingual biases. Our contribution to this work is twofold. First, we propose domain diversification, which diversifies the limited training data using different audio data augmentation methods. We then propose the concept of maximally diversity-aware cascaded augmentations and optimize the augmentation fold-factor for effective diversification of the training data. Second, we introduce the idea of domain generalization considering the augmentation methods as pseudo-domains. Towards this, we investigate both domain-invariant and domain-aware approaches. Our LID system is based on the state-of-the-art emphasized channel attention, propagation, and aggregation based time delay neural network (ECAPA-TDNN) architecture. We have conducted extensive experiments with three widely used corpora for Indian LID research. In addition, we conduct a final blind evaluation of our proposed methods on the Indian subset of VoxLingua107 corpus collected in the wild. Our experiments demonstrate that the proposed domain diversification is more promising over commonly used simple augmentation methods. The study also reveals that domain generalization is a more effective solution than domain diversification. We also notice that domain-aware learning performs better for same-corpora LID, whereas domain-invariant learning is more suitable for cross-corpora generalization. Compared to basic ECAPA-TDNN, its proposed domain-invariant extensions improve the cross-corpora EER up to 5.23%. In contrast, the proposed domain-aware extensions also improve performance for same-corpora test scenarios.
['Goutam Saha', 'Md Sahidullah', 'Spandan Dey']
2023-02-10
null
null
null
null
['spoken-language-identification']
['speech']
[-8.34475011e-02 -2.39929691e-01 -2.24472642e-01 -2.87093759e-01 -1.22563696e+00 -6.88322961e-01 5.66919446e-01 -3.58146012e-01 -6.19737864e-01 8.54693592e-01 4.99344230e-01 -2.67358035e-01 2.31864937e-02 -6.74619228e-02 -4.19797212e-01 -8.23682666e-01 3.45055461e-02 5.38934886e-01 -4.00552228e-02 -5.24934590e-01 -3.34916562e-02 4.58463937e-01 -1.74436378e+00 2.00945795e-01 1.08921695e+00 9.68184590e-01 9.10137221e-02 4.57358122e-01 -1.41662657e-01 1.19297773e-01 -7.34325528e-01 -3.15835893e-01 4.78290647e-01 -5.74116349e-01 -8.81933928e-01 2.24747196e-01 4.68732536e-01 1.80961844e-02 -1.38003156e-01 8.45944345e-01 1.12612855e+00 2.80594468e-01 5.26846468e-01 -1.11186230e+00 -8.26828480e-01 7.40785778e-01 -5.13826251e-01 2.74926335e-01 1.85818281e-02 2.38912404e-01 8.86128783e-01 -8.74721348e-01 4.94518697e-01 1.23258841e+00 7.24872053e-01 1.21620822e+00 -1.20659781e+00 -1.04822147e+00 3.52521300e-01 2.97627687e-01 -1.28981173e+00 -5.18076301e-01 9.69102979e-01 -1.32580772e-01 8.93648863e-01 7.73054641e-03 1.61558792e-01 1.44545996e+00 -4.56744313e-01 9.54484999e-01 1.30533493e+00 -7.08310068e-01 4.19215798e-01 5.60700536e-01 1.50356710e-01 -3.39368768e-02 -8.62724259e-02 1.62299335e-01 -6.41010404e-01 4.33434211e-02 2.93855608e-01 -7.87125111e-01 -3.15110117e-01 -2.14828938e-01 -7.46577501e-01 8.88694882e-01 2.12608427e-01 5.64755023e-01 -2.86380023e-01 -4.89811480e-01 7.23124444e-01 4.04617816e-01 7.99685001e-01 5.93920708e-01 -7.90563941e-01 -4.84646745e-02 -1.06867874e+00 3.12001519e-02 6.16403759e-01 9.21007633e-01 3.64380926e-01 3.09381008e-01 -2.55962968e-01 1.26861227e+00 3.67890522e-02 2.78217494e-01 9.27271724e-01 -6.09096825e-01 4.85031277e-01 1.95304513e-01 -1.93929091e-01 -5.12685813e-02 -5.13607800e-01 -7.70754576e-01 -8.23309004e-01 1.02676995e-01 3.07239830e-01 -5.32030463e-01 -9.66597676e-01 2.11149359e+00 7.90774003e-02 7.10869879e-02 6.27028525e-01 8.12278509e-01 9.12363648e-01 5.74160159e-01 4.24684078e-01 -4.05294269e-01 1.31113112e+00 -1.02497649e+00 -7.09453225e-01 -1.47430718e-01 4.67993230e-01 -7.63819635e-01 1.35753083e+00 4.77310121e-01 -9.70853090e-01 -6.67728007e-01 -9.87104535e-01 2.70583600e-01 -3.79451483e-01 3.53275031e-01 3.63748431e-01 9.12634254e-01 -1.23925579e+00 3.37442644e-02 -2.09736452e-01 -8.45693529e-01 1.74229473e-01 6.01752579e-01 -2.94951558e-01 1.29376784e-01 -1.36294997e+00 7.09736109e-01 3.32537502e-01 -3.19063634e-01 -7.04456627e-01 -7.05450177e-01 -7.21547186e-01 -2.38554686e-01 1.53459953e-02 -3.10333103e-01 1.37158263e+00 -1.12670767e+00 -1.77536023e+00 1.07084680e+00 -7.12705925e-02 -8.31948221e-01 2.33236134e-01 -1.64845526e-01 -8.15978646e-01 -2.56263502e-02 3.51837538e-02 8.66072953e-01 7.78607368e-01 -1.41483331e+00 -6.10042572e-01 -2.77054161e-01 -3.90718818e-01 4.37572360e-01 -8.15024793e-01 -4.95269382e-03 -4.72426385e-01 -9.38522995e-01 -1.72719866e-01 -1.02854991e+00 6.30265847e-02 -4.86077458e-01 -1.75608754e-01 -2.64264852e-01 1.00916684e+00 -7.74642348e-01 1.37179410e+00 -2.40529799e+00 9.10061896e-02 -2.12656215e-01 -3.78337145e-01 6.67358279e-01 -3.17002982e-01 3.52521449e-01 -2.37222552e-01 -3.65531771e-04 -4.60539013e-01 -7.07079411e-01 -2.44330708e-02 1.35718152e-01 -2.85872638e-01 2.97646731e-01 2.20857203e-01 3.20825994e-01 -3.57404888e-01 -3.25862229e-01 -5.17653897e-02 5.56869626e-01 -6.41664684e-01 2.13017672e-01 -3.68546486e-01 7.24030912e-01 -4.21905704e-02 6.70488000e-01 8.46498847e-01 3.59566569e-01 6.69976845e-02 -1.59281865e-01 -3.40752333e-01 3.04554641e-01 -1.08244371e+00 2.00233078e+00 -7.77502596e-01 5.82354367e-01 6.93788305e-02 -1.04281414e+00 1.23749208e+00 4.48387265e-01 2.69746691e-01 -9.79570329e-01 1.07862413e-01 4.70801383e-01 2.09367368e-02 -2.55307525e-01 4.88136590e-01 -2.09058806e-01 -2.85809129e-01 3.58176053e-01 5.05061328e-01 9.50893573e-03 7.65382349e-02 -4.96393442e-02 5.84435344e-01 9.33521613e-02 1.96254656e-01 -4.02874053e-01 8.72177720e-01 -1.08583570e-01 5.69797933e-01 5.36477745e-01 -4.53681737e-01 5.93459010e-01 1.25611231e-01 -4.41234373e-02 -9.95733440e-01 -9.59580779e-01 -4.07683611e-01 1.34937537e+00 4.16233689e-02 -1.62125938e-02 -8.36870611e-01 -8.16273749e-01 -2.16221720e-01 8.63818228e-01 -3.78242582e-01 -1.36495352e-01 -5.39351583e-01 -9.56750333e-01 9.53835309e-01 5.99601924e-01 1.01689935e+00 -1.03787887e+00 -4.91274595e-02 2.74173409e-01 -1.59888953e-01 -1.18089104e+00 -6.81406140e-01 5.18195808e-01 -6.85233891e-01 -4.61392194e-01 -1.08672845e+00 -1.17748129e+00 8.56792405e-02 -5.50798662e-02 9.36156690e-01 -6.28351152e-01 -5.99616542e-02 6.53753459e-01 -4.91845876e-01 -4.46411669e-01 -4.84403312e-01 3.86743844e-01 4.88600463e-01 6.74146935e-02 7.52324522e-01 -6.53026998e-01 -4.12924111e-01 3.60334128e-01 -5.82412660e-01 -6.62537873e-01 7.04988182e-01 1.17779875e+00 4.96641219e-01 9.39024836e-02 1.12489319e+00 -3.72914165e-01 1.06489885e+00 -3.41519922e-01 -5.28844893e-01 1.44371793e-01 -6.91755533e-01 3.68599862e-01 7.42279589e-01 -7.75831997e-01 -1.51140451e+00 1.46084607e-01 -3.46432507e-01 -2.47396901e-01 -4.10230249e-01 1.74077880e-02 -5.09953916e-01 -1.13946386e-01 8.11082602e-01 3.81913990e-01 -2.84986347e-02 -8.14538896e-01 4.91308421e-01 1.13522768e+00 6.99668467e-01 -7.27954507e-01 5.16390860e-01 2.08507478e-01 -5.37508547e-01 -1.11374331e+00 -3.85068804e-01 -6.99076176e-01 -6.57958150e-01 1.33938253e-01 7.99379051e-01 -1.10490799e+00 -4.86194551e-01 6.15114272e-01 -9.97234464e-01 -3.49682808e-01 -4.20685172e-01 5.75215578e-01 -5.29811502e-01 3.42377752e-01 -2.69604594e-01 -9.30069447e-01 -4.77413416e-01 -1.01704371e+00 9.73554552e-01 3.16228807e-01 -2.17514873e-01 -9.96872187e-01 3.17458957e-01 1.98968992e-01 4.91544008e-01 -3.62971812e-01 7.76154935e-01 -1.27192414e+00 -3.77912372e-02 1.92621097e-01 3.78216691e-02 6.95147276e-01 1.01018324e-01 -4.42050934e-01 -1.50678146e+00 -3.32391471e-01 -8.21344405e-02 -3.67371559e-01 8.66409242e-01 4.80683774e-01 8.68547320e-01 -1.56180814e-01 -1.45374984e-01 7.49454737e-01 1.31451702e+00 5.27306736e-01 4.32414263e-01 6.65472031e-01 7.55489916e-02 6.33698285e-01 6.25117898e-01 4.92328495e-01 3.17488760e-01 9.64815676e-01 4.51075509e-02 -3.78509015e-01 -5.30961812e-01 -8.64854455e-02 5.01678526e-01 6.18420243e-01 2.12432787e-01 -4.07129943e-01 -8.39446664e-01 1.01701963e+00 -1.38410950e+00 -8.48102808e-01 2.97362745e-01 2.24505854e+00 1.07673919e+00 -5.42768724e-02 7.54876733e-01 3.74490768e-01 8.54180872e-01 -1.54092029e-01 -6.21648848e-01 -8.21423292e-01 -4.92837191e-01 4.87576365e-01 3.55274290e-01 4.65071350e-01 -1.29736924e+00 1.15422010e+00 5.55940008e+00 9.46736693e-01 -1.34352136e+00 3.34205687e-01 4.21491951e-01 -1.02756940e-01 -5.33283874e-02 -5.25523901e-01 -1.20404506e+00 3.05218786e-01 9.84487295e-01 -1.94197595e-01 2.13882938e-01 8.20969403e-01 -4.93813232e-02 1.76588491e-01 -8.42054605e-01 9.72023785e-01 2.70555049e-01 -8.21518540e-01 1.27302501e-02 -1.19506396e-01 7.97648787e-01 -3.47657539e-02 6.67121172e-01 5.05505562e-01 1.29722089e-01 -7.63885140e-01 4.58219856e-01 -1.61689773e-01 1.04870820e+00 -8.65372181e-01 7.38912821e-01 -5.74207976e-02 -1.01082766e+00 -3.71500224e-01 -1.83459416e-01 3.47943574e-01 5.45980828e-03 5.43915853e-02 -1.08474088e+00 4.53777373e-01 8.44395399e-01 4.01105016e-01 -4.31525260e-01 1.04854000e+00 -3.08995750e-02 7.82245576e-01 -2.99301088e-01 7.96075165e-02 4.40161735e-01 1.72460809e-01 7.65416920e-01 1.55952013e+00 2.12895826e-01 -8.13894495e-02 -2.69866407e-01 5.94062150e-01 -2.83901971e-02 2.82130152e-01 -5.39576948e-01 2.08659053e-01 4.89890993e-01 8.58381510e-01 -2.38048106e-01 1.58345290e-02 -4.88907635e-01 1.08263838e+00 4.75088358e-02 5.50942123e-01 -7.27289319e-01 -3.34043831e-01 1.00919545e+00 -2.21659675e-01 5.21085560e-01 -5.65287992e-02 -3.27520162e-01 -9.14021194e-01 -4.55408506e-02 -1.11729503e+00 5.70478559e-01 -3.87332916e-01 -1.53397393e+00 1.29588544e+00 3.15904096e-02 -1.52584088e+00 -3.66209209e-01 -4.09859955e-01 -3.75833929e-01 1.04085374e+00 -1.77090430e+00 -1.34628844e+00 -1.92259587e-02 1.02436996e+00 9.37367082e-01 -7.40415037e-01 9.56364036e-01 5.67913294e-01 -4.93281603e-01 1.29708695e+00 4.45905149e-01 -1.09743483e-01 1.09649348e+00 -1.03668666e+00 2.07987487e-01 9.30003881e-01 3.35146934e-02 3.60705197e-01 6.83051825e-01 -3.65501195e-01 -8.14628959e-01 -1.14829671e+00 7.74320424e-01 -1.30298287e-01 4.58945811e-01 -5.27198434e-01 -9.36389863e-01 4.04439002e-01 5.04264534e-01 -4.08159018e-01 7.82960415e-01 1.63514853e-01 -4.48440373e-01 -3.66203338e-01 -1.40188968e+00 5.53589344e-01 1.02637672e+00 -5.03091991e-01 -6.69595242e-01 1.23340972e-01 9.63074088e-01 -6.63313121e-02 -6.12306416e-01 3.31216812e-01 3.79820913e-01 -8.94621074e-01 8.83041441e-01 -4.66052353e-01 -1.23836800e-01 -7.10219517e-02 -4.61702615e-01 -1.51412070e+00 -1.08253278e-01 -7.27569878e-01 4.06142294e-01 1.86161864e+00 5.78533530e-01 -7.00442076e-01 5.83058476e-01 1.67728394e-01 -2.07689732e-01 -3.13530177e-01 -1.07470119e+00 -1.20909643e+00 5.22418261e-01 -4.62987214e-01 5.87130249e-01 8.24218750e-01 -7.10942447e-02 4.48769629e-01 -5.00976086e-01 2.25504369e-01 6.04279935e-01 -1.70718208e-01 5.28540194e-01 -1.15600908e+00 -2.03586474e-01 -3.64436001e-01 -1.73381403e-01 -9.34668481e-01 4.85976398e-01 -6.59629941e-01 3.55295651e-02 -9.94618297e-01 -2.45622948e-01 -5.59236109e-01 -2.65889943e-01 4.00482804e-01 1.97532564e-01 2.64509797e-01 2.11682603e-01 8.01857859e-02 -3.60689461e-01 7.07363069e-01 9.15404379e-01 -1.93453059e-01 -8.27164352e-01 2.85925537e-01 -8.51032734e-01 3.43922824e-01 1.04915369e+00 -1.88116923e-01 -4.99678195e-01 -3.36010486e-01 -5.13743401e-01 -2.72307068e-01 1.55544886e-02 -1.19777071e+00 2.69243956e-01 3.87561202e-01 -5.54397181e-02 -5.18715858e-01 4.71070737e-01 -7.95609355e-01 -3.72306734e-01 1.48684770e-01 -3.26273441e-01 -2.71717817e-01 9.65277135e-01 3.18959802e-01 -5.11393726e-01 6.04164451e-02 1.07413530e+00 1.49262771e-01 -1.21049511e+00 2.95189712e-02 -2.89979011e-01 3.20514292e-01 8.65896881e-01 -3.61093044e-01 -1.59504518e-01 -5.29503822e-01 -8.92807841e-01 9.26069692e-02 1.63595095e-01 5.42161524e-01 2.32463717e-01 -1.27837133e+00 -8.91137123e-01 5.07825971e-01 2.54963875e-01 -5.08831382e-01 3.44596118e-01 5.83675623e-01 1.88517630e-01 7.05094814e-01 -3.04046005e-01 -6.23957455e-01 -1.46762621e+00 6.31734431e-01 4.47765261e-01 -3.43119428e-02 -2.73281336e-01 1.21437633e+00 3.86027664e-01 -5.76671302e-01 6.89579725e-01 -1.49685830e-01 -3.66162390e-01 2.54810870e-01 3.98738205e-01 2.36741737e-01 2.42924720e-01 -7.17234015e-01 -5.77194631e-01 8.41164768e-01 -1.85143828e-01 -4.12327260e-01 1.32398129e+00 -3.47844481e-01 3.61969650e-01 2.68778265e-01 1.08306754e+00 9.91692469e-02 -1.08828568e+00 -4.32025939e-01 -9.12773758e-02 9.64705795e-02 -4.51464206e-02 -1.36539507e+00 -8.57985854e-01 1.00018704e+00 1.03552985e+00 -5.09905815e-02 1.87446785e+00 9.89033505e-02 6.56790972e-01 4.01713625e-02 2.10799396e-01 -1.19392073e+00 4.38041762e-02 5.51038086e-01 1.04816616e+00 -1.23885846e+00 -4.47017580e-01 -6.47716150e-02 -1.15396559e+00 8.42534184e-01 9.16648090e-01 2.71714628e-01 4.36261505e-01 2.17501253e-01 3.03451866e-01 3.23745638e-01 -6.38032734e-01 -5.60683668e-01 3.47806394e-01 1.14009595e+00 4.79347467e-01 -2.14151561e-01 -4.74639535e-01 9.31357443e-01 -2.38181949e-01 -1.99268758e-01 2.83652991e-01 3.04111153e-01 -6.95142746e-02 -1.38250422e+00 -5.76150775e-01 -3.05714399e-01 -3.83351356e-01 -3.21973085e-01 -5.55510044e-01 1.14036965e+00 2.70820975e-01 1.00813997e+00 -1.32318586e-02 -4.86009687e-01 3.79303873e-01 4.59388256e-01 3.54400128e-01 -4.50304985e-01 -7.97562301e-01 2.13867143e-01 7.95120075e-02 -1.24351956e-01 -5.51113307e-01 -8.07083726e-01 -1.06216943e+00 3.93890738e-02 -2.52016455e-01 3.19989085e-01 7.46499002e-01 6.13041997e-01 4.33177769e-01 4.28058654e-01 7.38896430e-01 -5.87698460e-01 -6.25954807e-01 -1.11268413e+00 -6.59438968e-01 2.58241773e-01 5.88034093e-01 -6.25981390e-01 -5.70912898e-01 -2.82220636e-02]
[14.254858016967773, 6.401511192321777]
c7cf795b-3954-4bdb-8bab-08607c3ba68e
police-text-analysis-topic-modeling-and
2202.04176
null
https://arxiv.org/abs/2202.04176v2
https://arxiv.org/pdf/2202.04176v2.pdf
Police Text Analysis: Topic Modeling and Spatial Relative Density Estimation
We analyze a large corpus of police incident narrative documents in understanding the spatial distribution of the topics. The motivation for doing this is that police narratives in each incident report contains very fine-grained information that is richer than the category that is manually assigned by the police. Our approach is to split the corpus into topics using two different unsupervised machine learning algorithms - Latent Dirichlet Allocation and Non-negative Matrix Factorization. We validate the performance of each learned topic model using model coherence. Then, using a k-nearest neighbors density ratio estimation (kNN-DRE) approach that we propose, we estimate the spatial density ratio per topic and use this for data discovery and analysis of each topic, allowing for insights into the described incidents at scale. We provide a qualitative assessment of each topic and highlight some key benefits for using our kNN-DRE model for estimating spatial trends.
['Yao Xie', 'Xiuyuan Cheng', 'Sarah Huestis-Mitchell']
2022-02-08
null
null
null
null
['density-ratio-estimation']
['methodology']
[-3.08695167e-01 3.63375306e-01 -4.15712565e-01 -3.28128517e-01 -1.07792330e+00 -6.84570074e-01 8.35948288e-01 6.90556884e-01 -3.20532739e-01 5.72433293e-01 1.38019645e+00 -5.27693808e-01 -4.59520161e-01 -9.29967463e-01 -2.89136648e-01 -6.60464525e-01 -2.48318717e-01 8.66836131e-01 1.75117731e-01 2.12526575e-01 4.23939884e-01 2.54704148e-01 -1.03269160e+00 6.44777358e-01 3.97254378e-01 3.63316864e-01 1.87012374e-01 5.64912021e-01 -2.20168322e-01 9.90994513e-01 -5.00124156e-01 -2.18942940e-01 -2.43974812e-02 -5.03930561e-02 -1.00670981e+00 2.72078104e-02 -1.11749053e-01 -4.90342259e-01 -2.46220410e-01 5.73816836e-01 1.13619678e-01 2.82067448e-01 1.39372325e+00 -1.05588806e+00 -4.46549922e-01 8.25715363e-01 -8.82155240e-01 7.33112216e-01 3.35262269e-01 -2.85776436e-01 1.11706734e+00 -7.20686436e-01 9.34552133e-01 1.37582886e+00 9.07968163e-01 -2.19100595e-01 -1.32835817e+00 -6.54109955e-01 1.08741827e-01 1.50664225e-01 -1.72150218e+00 -2.78674483e-01 6.68292642e-01 -1.25844026e+00 9.96627867e-01 -6.08066097e-02 4.14418787e-01 8.31047654e-01 1.11462563e-01 4.26671803e-01 8.43754053e-01 -3.05387944e-01 4.66581941e-01 3.18055779e-01 3.29618454e-01 -1.34246469e-01 2.92805225e-01 -3.18872929e-01 -3.70120585e-01 -7.98199773e-01 5.49388707e-01 3.18785816e-01 1.96542293e-01 7.86767676e-02 -8.68914306e-01 1.20653844e+00 -2.06822879e-03 5.07775307e-01 -7.85988688e-01 -4.19386812e-02 5.44540763e-01 -2.78463364e-01 1.18527770e+00 2.23910406e-01 -2.91422993e-01 -4.88051802e-01 -1.32543135e+00 5.59033334e-01 6.90076888e-01 4.53534693e-01 9.28746939e-01 -5.73793709e-01 -2.01489270e-01 7.93180585e-01 3.32351238e-01 3.24883521e-01 -4.48299497e-02 -8.53417397e-01 6.11776054e-01 5.42714655e-01 5.15917182e-01 -1.68369567e+00 -4.85549003e-01 2.81151384e-01 -5.55408478e-01 -3.64498258e-01 3.35410953e-01 -3.93550366e-01 -5.32617629e-01 1.36671197e+00 3.40027511e-01 1.80016220e-01 -1.83444098e-01 4.63439316e-01 4.22056794e-01 1.10435569e+00 6.11632526e-01 -4.21425492e-01 1.52727389e+00 -1.22647390e-01 -8.72330129e-01 1.01628318e-01 7.21243560e-01 -4.47510332e-01 7.02995479e-01 4.10038382e-02 -7.81246722e-01 -6.94703013e-02 -2.21337751e-01 7.39658028e-02 -2.18656510e-01 -2.27892220e-01 6.45939112e-01 4.98188615e-01 -8.26285899e-01 1.20673187e-01 -9.40063655e-01 -5.77213228e-01 4.65955645e-01 -2.65858412e-01 -2.28461996e-01 -5.51446564e-02 -1.15484357e+00 6.45512640e-01 4.52761382e-01 -6.31605268e-01 -5.03583193e-01 -1.10731065e+00 -9.42594945e-01 8.50922242e-02 6.62734360e-02 -7.82666877e-02 1.11534548e+00 1.24369478e-02 -4.53651309e-01 7.39625812e-01 -7.06200421e-01 -2.70450562e-01 -1.21517871e-02 1.63786516e-01 -2.18342572e-01 4.21766639e-01 9.88198102e-01 4.32084858e-01 2.55978405e-01 -1.43064153e+00 -9.19547617e-01 -4.11853671e-01 -1.27992690e-01 -1.48207784e-01 -5.31495810e-01 5.52297473e-01 6.18789159e-02 -7.52239883e-01 7.86616057e-02 -6.53145492e-01 -2.73771524e-01 -9.19719517e-01 -2.85172433e-01 -6.19789660e-01 7.26207256e-01 -1.05285215e+00 1.82802463e+00 -2.26921105e+00 -3.56855690e-01 3.79333228e-01 4.43916678e-01 -5.66437364e-01 3.50348592e-01 9.02157128e-01 -7.37007409e-02 4.15874600e-01 -1.49310797e-01 -3.85074407e-01 -5.90760708e-02 7.96219483e-02 -9.34508920e-01 5.14986396e-01 9.72936079e-02 5.99114299e-01 -9.77707148e-01 -5.34544587e-01 -1.75129637e-01 2.92766720e-01 -6.77189529e-01 -1.92067489e-01 1.36099681e-01 3.73861402e-01 -3.85234982e-01 2.89998412e-01 6.78569853e-01 -1.08497590e-01 3.24331790e-01 -2.55115796e-02 -5.42422056e-01 5.83104491e-01 -9.58549023e-01 1.09104979e+00 -1.45724386e-01 9.71987307e-01 -1.70713022e-01 -8.22835028e-01 8.16552281e-01 4.61462080e-01 8.42920363e-01 -2.71913022e-01 -2.04184875e-01 -3.27094734e-01 -5.16156197e-01 -4.52455282e-01 9.14870799e-01 -5.38762152e-01 -5.44822872e-01 1.23424268e+00 -9.55523774e-02 3.72085199e-02 1.72097966e-01 5.42772651e-01 1.17464352e+00 -6.98761880e-01 3.92266691e-01 -6.49987936e-01 -5.02063215e-01 6.05100334e-01 5.66959798e-01 8.11064363e-01 1.12083897e-01 5.14396012e-01 9.28855240e-01 -7.07485020e-01 -1.49203348e+00 -1.05710018e+00 -6.41866148e-01 1.04346812e+00 -3.82504761e-01 -7.67329395e-01 -6.12206936e-01 -3.11780274e-01 -1.10165864e-01 1.05372500e+00 -8.46696258e-01 3.44116509e-01 -3.59303147e-01 -1.16861391e+00 4.25610930e-01 6.08272016e-01 -1.48803815e-01 -7.06412673e-01 -5.79702914e-01 2.10536048e-01 -5.22450745e-01 -1.04771781e+00 1.71449576e-02 -1.32712319e-01 -5.59112847e-01 -8.66955459e-01 -4.38643098e-01 -2.32978538e-01 5.95881820e-01 3.03495020e-01 9.93891895e-01 -4.72926408e-01 -6.42175600e-02 5.88509619e-01 -6.03734076e-01 -5.68698943e-01 -4.05453593e-01 2.03039929e-01 1.79268703e-01 -1.75965920e-01 1.01272511e+00 -7.97859073e-01 -3.96851957e-01 -2.74200551e-02 -8.68451536e-01 -2.50608116e-01 1.31702274e-01 2.37362653e-01 3.93295251e-02 6.75587237e-01 5.37023425e-01 -9.25665975e-01 1.06547928e+00 -1.33666027e+00 -1.17462248e-01 -2.23901778e-01 -2.01799959e-01 -4.00880277e-01 -2.12818801e-01 -1.20286360e-01 -1.12501514e+00 -5.08035481e-01 2.32430235e-01 4.47219238e-02 -5.99820852e-01 7.88365901e-01 2.66833872e-01 1.05286729e+00 8.41477036e-01 -2.26025194e-01 -4.40662354e-01 -4.31274116e-01 2.58837521e-01 8.99617612e-01 3.03987175e-01 -5.45128882e-01 6.32877648e-01 8.37633610e-01 -6.13815665e-01 -1.07145476e+00 -9.25748110e-01 -1.12893069e+00 -9.61266816e-01 -2.23675981e-01 1.33471859e+00 -1.07292426e+00 -4.88974154e-01 5.27800620e-02 -1.57268584e+00 -4.55574781e-01 -3.85003030e-01 7.78606236e-01 -3.09177130e-01 8.75693932e-02 -5.68213999e-01 -1.19633412e+00 3.61146390e-01 -6.90447152e-01 1.12066054e+00 -3.15881044e-01 -9.71084833e-01 -1.49348235e+00 8.27494740e-01 1.80240303e-01 1.58867016e-01 3.71721089e-01 1.22833574e+00 -7.92987704e-01 -8.29840675e-02 -1.22072399e-01 -5.30840158e-01 -5.18255353e-01 3.08674395e-01 -1.21379092e-01 -1.01055670e+00 3.42966318e-02 -7.05561414e-02 1.11171968e-01 1.00844407e+00 8.17655206e-01 5.30737996e-01 -6.97479308e-01 -8.29629004e-01 -9.60610434e-02 1.23150492e+00 8.43075365e-02 5.92724681e-01 5.70592463e-01 5.31968296e-01 9.17771518e-01 3.62394214e-01 9.06898797e-01 1.01347375e+00 4.60589975e-01 -1.60294071e-01 1.92785505e-02 5.11608720e-01 -2.28670597e-01 1.16484053e-01 5.49309254e-01 -9.99710113e-02 -4.93650585e-02 -1.80371976e+00 1.24815786e+00 -1.85804105e+00 -1.37631583e+00 -2.11589262e-01 1.58878398e+00 8.05557191e-01 4.37891744e-02 4.77499783e-01 9.84606221e-02 7.25347519e-01 2.78054506e-01 3.14684570e-01 -4.08114046e-01 2.51942307e-01 -2.54375070e-01 4.22267944e-01 5.80995083e-01 -1.20830405e+00 7.64652133e-01 7.59304094e+00 7.81669140e-01 -4.46073055e-01 3.71195108e-01 8.50564480e-01 8.75614807e-02 -5.66382766e-01 1.79140940e-01 -8.51067781e-01 3.16891670e-01 1.16814840e+00 -1.03298031e-01 -1.06834419e-01 8.94511223e-01 8.90972316e-01 -5.99788308e-01 -6.17504716e-01 4.57432002e-01 -1.00316264e-01 -1.76755607e+00 -2.97508538e-02 4.77230012e-01 9.54317033e-01 -1.95843652e-01 -1.94632113e-01 5.52454889e-02 8.06433856e-01 -8.10814679e-01 6.63656950e-01 5.98070204e-01 8.02402198e-01 -8.52358937e-01 7.40267456e-01 4.33560550e-01 -9.80999172e-01 -3.27647954e-01 -4.48132813e-01 -5.78865111e-01 7.10045338e-01 1.15835035e+00 -1.26919818e+00 4.84981984e-02 9.57903206e-01 5.73644936e-01 -4.62051369e-02 6.72370315e-01 4.07901406e-02 1.13308311e+00 -2.80796081e-01 4.02748853e-01 3.88059407e-01 -1.01173401e-03 7.14392424e-01 1.66656458e+00 2.80363172e-01 5.20195425e-01 2.24253371e-01 9.36051130e-01 3.90041769e-01 -1.80056512e-01 -9.05415416e-01 5.60997762e-02 8.44964445e-01 9.92143035e-01 -1.15078855e+00 -2.91250080e-01 -1.66795224e-01 2.41831914e-01 2.86527455e-01 3.98483813e-01 -3.06169987e-01 2.68555642e-03 6.95114672e-01 7.14973867e-01 2.02224448e-01 -6.03922248e-01 -5.35041392e-01 -9.20859456e-01 -4.05527711e-01 -2.47612730e-01 2.73689091e-01 -5.65554559e-01 -1.42314780e+00 3.45062226e-01 9.70611393e-01 -7.11646557e-01 -5.34023821e-01 5.47325946e-02 -8.84486496e-01 7.56496727e-01 -9.90741789e-01 -9.17850912e-01 1.19328760e-01 2.24248677e-01 4.12861943e-01 -4.52026650e-02 7.93491066e-01 1.28665522e-01 -3.34598333e-01 -1.48024827e-01 2.18093142e-01 3.52909982e-01 5.07176399e-01 -1.22654045e+00 5.86998582e-01 4.75491554e-01 -9.74954516e-02 7.37218380e-01 6.95325315e-01 -1.10733247e+00 -4.05599415e-01 -1.06892586e+00 1.39383125e+00 -8.47343504e-01 1.11768544e+00 -4.55951005e-01 -7.20091760e-01 8.90907943e-01 1.26406550e-01 -6.64362192e-01 1.38088155e+00 6.21806979e-01 -1.02495559e-01 4.43140209e-01 -9.10507619e-01 2.58272499e-01 4.65568155e-01 -6.68567538e-01 -8.91558945e-01 3.34264755e-01 7.08047032e-01 1.95910394e-01 -9.44484472e-01 -7.46782199e-02 2.64929444e-01 -7.29840517e-01 6.33281648e-01 -6.81400120e-01 5.36514580e-01 -9.49220136e-02 -3.35018814e-01 -1.30171084e+00 -7.76321769e-01 -2.81147629e-01 5.38579404e-01 1.52061927e+00 5.60042322e-01 -2.61465907e-01 7.34948277e-01 1.11181188e+00 1.79705530e-01 -2.38072276e-01 -8.71095717e-01 -2.43384480e-01 2.81024069e-01 -1.08315551e+00 6.12725258e-01 1.54709554e+00 5.67949355e-01 2.20131010e-01 -2.09968299e-01 2.94019431e-01 4.52001482e-01 -2.35235408e-01 7.66681194e-01 -1.51623178e+00 1.78928107e-01 3.49150114e-02 -4.75204080e-01 -6.64824665e-01 2.36394733e-01 -5.07548749e-01 -3.21305424e-01 -1.71973825e+00 6.96710289e-01 -7.82095492e-01 4.05206949e-01 4.33966368e-01 2.66830791e-02 2.52123326e-01 4.24848543e-03 5.79084516e-01 -5.27083039e-01 6.60437867e-02 4.35150057e-01 -1.82108376e-02 -4.44839090e-01 -2.40704969e-01 -8.56542945e-01 9.56460416e-01 6.92495644e-01 -8.80363345e-01 -2.19494954e-01 -2.92814434e-01 7.02710509e-01 -2.57120263e-02 4.89624649e-01 -6.33867085e-01 3.03426057e-01 -5.16593874e-01 3.18863541e-01 -9.38034236e-01 1.95428152e-02 -6.33749843e-01 8.37954283e-02 -2.13776290e-01 -4.48138654e-01 6.52595311e-02 3.56081486e-01 5.81613362e-01 -2.50788599e-01 -1.81143403e-01 8.97660702e-02 -1.31845877e-01 -4.77699459e-01 -5.80523945e-02 -1.21395719e+00 1.94973368e-02 9.34132576e-01 -2.37455606e-01 -2.04615638e-01 -7.35213339e-01 -9.52109635e-01 6.19183145e-02 2.47362882e-01 1.26640171e-01 3.01865429e-01 -1.45176983e+00 -9.69718456e-01 -1.97497636e-01 5.95956817e-02 -4.29989286e-02 4.80328649e-01 8.31245303e-01 -1.50847390e-01 6.77177548e-01 2.67580718e-01 -5.10388196e-01 -6.04741573e-01 4.18981135e-01 -3.71245325e-01 -5.39591432e-01 -7.00283587e-01 2.54464477e-01 5.32864094e-01 -2.29792684e-01 -3.20844531e-01 -9.54133198e-02 -6.12952948e-01 8.13833356e-01 9.13110256e-01 7.29543984e-01 -3.83926392e-01 -9.15235460e-01 -2.55024284e-01 3.18689615e-01 4.83663604e-02 -5.58702528e-01 1.93915200e+00 -4.19190526e-01 -3.71221900e-01 7.81413674e-01 1.11692965e+00 1.69393152e-01 -1.11587048e+00 -1.13471963e-01 3.43787849e-01 -3.74072403e-01 1.06116496e-01 -1.76819846e-01 -2.93220639e-01 5.93288600e-01 -5.32262772e-03 8.81801069e-01 5.79135060e-01 5.19850314e-01 4.26852375e-01 1.21020585e-01 -2.77254879e-02 -1.02463281e+00 -2.82535106e-01 5.21470428e-01 6.83233082e-01 -9.11257148e-01 2.15487897e-01 -3.15844089e-01 -8.78065884e-01 9.87557054e-01 -1.38312995e-01 -2.82752097e-01 1.22036695e+00 5.48679590e-01 -1.46794409e-01 -7.03533471e-01 -6.58977985e-01 -9.00808349e-03 1.84693888e-01 7.32037127e-01 4.16851342e-01 2.51371205e-01 -1.01131074e-01 8.28182697e-01 -5.21719575e-01 -3.57892662e-01 5.84571600e-01 5.78618765e-01 -7.13583648e-01 -6.44223869e-01 -6.82589710e-01 6.35490358e-01 -6.80107057e-01 -1.92736700e-01 -2.65740097e-01 7.17522323e-01 -1.06620719e-03 1.15969527e+00 7.75320411e-01 -2.90527165e-01 -7.16305971e-02 5.59815988e-02 -3.30073833e-01 -8.63335669e-01 -1.94812000e-01 2.67149478e-01 1.45908460e-01 -1.26796544e-01 -4.76024747e-01 -1.20959675e+00 -1.03302145e+00 -3.67380649e-01 -1.15890890e-01 7.24492788e-01 5.55657864e-01 1.21568942e+00 3.06871146e-01 1.94588244e-01 6.05873406e-01 -1.04405940e+00 2.89610535e-01 -1.24071729e+00 -7.69261897e-01 2.72551686e-01 2.11514518e-01 -7.42693782e-01 -3.23012590e-01 3.32377851e-01]
[10.347222328186035, 7.0734686851501465]
4f399990-c658-4c5c-b828-092fae0e1cba
learning-deep-multi-level-similarity-for
1906.03568
null
https://arxiv.org/abs/1906.03568v1
https://arxiv.org/pdf/1906.03568v1.pdf
Learning Deep Multi-Level Similarity for Thermal Infrared Object Tracking
Existing deep Thermal InfraRed (TIR) trackers only use semantic features to describe the TIR object, which lack the sufficient discriminative capacity for handling distractors. This becomes worse when the feature extraction network is only trained on RGB images.To address this issue, we propose a multi-level similarity model under a Siamese framework for robust TIR object tracking. Specifically, we compute different pattern similarities on two convolutional layers using the proposed multi-level similarity network. One of them focuses on the global semantic similarity and the other computes the local structural similarity of the TIR object. These two similarities complement each other and hence enhance the discriminative capacity of the network for handling distractors. In addition, we design a simple while effective relative entropy based ensemble subnetwork to integrate the semantic and structural similarities. This subnetwork can adaptive learn the weights of the semantic and structural similarities at the training stage. To further enhance the discriminative capacity of the tracker, we construct the first large scale TIR video sequence dataset for training the proposed model. The proposed TIR dataset not only benefits the training for TIR tracking but also can be applied to numerous TIR vision tasks. Extensive experimental results on the VOT-TIR2015 and VOT-TIR2017 benchmarks demonstrate that the proposed algorithm performs favorably against the state-of-the-art methods.
['Hongpeng Wang', 'Nana Fan', 'Qiao Liu', 'Zhenyu He', 'Xin Li', 'Di Yuan']
2019-06-09
null
null
null
null
['thermal-infrared-object-tracking']
['computer-vision']
[ 5.39635867e-02 -6.62862957e-01 -5.34133613e-02 -2.48442769e-01 -3.59250844e-01 -2.47631803e-01 4.58249480e-01 -7.57992208e-01 -5.74150681e-01 2.58908391e-01 -6.85917139e-02 1.33757278e-01 -3.76553349e-02 -1.18446440e-01 -4.16480482e-01 -1.22237170e+00 5.88038981e-01 -1.25318483e-01 5.41152000e-01 -1.94083545e-02 1.90814644e-01 2.46857941e-01 -1.88135779e+00 1.02374814e-02 1.15347791e+00 1.56991708e+00 5.13851881e-01 1.06050074e-01 2.48689130e-02 7.14220107e-01 -3.47498715e-01 -1.99080519e-02 4.13129002e-01 -7.27302954e-02 -1.65639818e-01 -8.57456997e-02 7.58312166e-01 -1.82027414e-01 -6.02982163e-01 1.12738729e+00 5.14456451e-01 4.82668370e-01 5.04805863e-01 -1.34711182e+00 -5.80126345e-01 -1.37382904e-02 -6.69104874e-01 3.71786535e-01 -9.16579813e-02 5.38346887e-01 8.42908561e-01 -1.00729513e+00 2.79359967e-01 1.31147099e+00 6.10812724e-01 5.37213027e-01 -5.87684810e-01 -9.09331918e-01 4.04965430e-01 5.07254004e-01 -1.14002085e+00 -3.44806165e-01 8.84666026e-01 -2.08370417e-01 6.25417650e-01 2.31644008e-02 3.15909564e-01 1.08801019e+00 2.62662888e-01 1.10853875e+00 8.90120864e-01 1.62440002e-01 -3.90894338e-03 -3.41505744e-02 3.20096612e-01 5.12875199e-01 3.49322557e-01 3.62435281e-01 -2.56759167e-01 1.41265601e-01 4.07260150e-01 1.93589285e-01 -6.56474382e-02 -3.94149840e-01 -1.33123660e+00 4.61076766e-01 6.18030548e-01 3.59213680e-01 -4.04979736e-02 4.60177392e-01 6.08768761e-01 -1.42361104e-01 2.71985859e-01 -7.82943368e-02 -3.77903908e-01 9.31614116e-02 -6.91317737e-01 -6.29542544e-02 2.17750911e-02 9.38784838e-01 7.26649642e-01 2.33349040e-01 -4.51998949e-01 6.00216150e-01 6.51249588e-01 1.01084673e+00 4.31010753e-01 -7.70089805e-01 4.62383628e-01 4.47664768e-01 2.05689594e-01 -9.62735772e-01 -2.51042873e-01 -5.09987056e-01 -8.60667169e-01 1.38505101e-01 3.90184432e-01 1.12501040e-01 -1.11436307e+00 1.67949116e+00 4.80308354e-01 5.08497596e-01 7.53854290e-02 1.20971179e+00 1.01376915e+00 3.70443255e-01 2.41227731e-01 -1.36456564e-01 1.41120005e+00 -1.05358124e+00 -9.19709802e-01 -5.59648871e-01 5.66123307e-01 -8.87001216e-01 7.13298917e-01 1.03067048e-01 -6.01801336e-01 -1.03962719e+00 -9.40252066e-01 -2.71293707e-02 -2.82522708e-01 8.01805854e-01 5.67064464e-01 5.31756282e-01 -7.48934567e-01 3.74026984e-01 -6.85378075e-01 -2.29258657e-01 5.36744356e-01 3.57454807e-01 -1.02095544e-01 -2.91760534e-01 -1.34357572e+00 1.01936662e+00 4.68411952e-01 7.83393085e-01 -5.98768234e-01 -3.25325847e-01 -8.69667768e-01 -1.11097150e-01 3.78804952e-01 -6.17918849e-01 7.45893002e-01 -9.69645441e-01 -1.33407807e+00 2.93146700e-01 -2.10600629e-01 -5.07290149e-03 2.22399652e-01 -1.51380107e-01 -5.39279878e-01 2.69462373e-02 -3.46563980e-02 6.05314016e-01 1.24064612e+00 -1.20077419e+00 -7.41714537e-01 -3.37317228e-01 -3.28576565e-01 2.69622833e-01 -2.77875721e-01 2.21168935e-01 -4.90424216e-01 -4.21480507e-01 7.79070854e-02 -1.22802186e+00 -7.00331405e-02 4.47125956e-02 -1.22058883e-01 -6.20461524e-01 1.36598051e+00 -3.40689957e-01 1.02798331e+00 -2.36779666e+00 -2.67842878e-02 -2.61370186e-02 9.17996243e-02 7.32399404e-01 -2.19202057e-01 -6.47418126e-02 -1.31956950e-01 -3.05727601e-01 1.26939073e-01 -4.41776663e-01 1.56688653e-02 7.49208555e-02 -1.39706790e-01 6.92882001e-01 3.18717003e-01 9.61585402e-01 -1.01063800e+00 -7.07123995e-01 6.39153183e-01 5.44123232e-01 -4.87457030e-02 8.85181054e-02 3.25056687e-02 5.27106941e-01 -8.79745603e-01 6.17457330e-01 1.05093908e+00 -1.14010401e-01 -1.36779368e-01 -9.29832458e-01 -3.20368379e-01 -8.19626600e-02 -1.00819254e+00 1.54288006e+00 -2.55248129e-01 5.59098423e-01 -2.77889997e-01 -9.28117514e-01 1.09413767e+00 2.60841735e-02 7.33280003e-01 -7.35365748e-01 5.26308656e-01 1.80757985e-01 3.58444825e-02 -5.95118582e-01 5.78286409e-01 1.18116997e-01 3.37497368e-02 1.26527444e-01 1.19942650e-02 3.61451328e-01 -1.92192737e-02 3.96361016e-02 8.00840259e-01 4.46561545e-01 -2.65581846e-01 -1.00148227e-02 9.75202322e-01 -2.22374320e-01 9.02101994e-01 4.03728664e-01 -6.86581016e-01 4.48990822e-01 -2.16340795e-01 -3.97354245e-01 -7.83344686e-01 -1.01537490e+00 -2.31100582e-02 7.91936696e-01 7.71877646e-01 1.28404841e-01 -3.51920247e-01 -8.60794127e-01 -6.17825165e-02 4.25768495e-01 -5.72388530e-01 -5.58014870e-01 -5.41596532e-01 -4.49415624e-01 4.49803352e-01 8.57647836e-01 9.26012993e-01 -1.06487179e+00 -6.41005099e-01 -2.27179363e-01 -5.63811839e-01 -1.53583133e+00 -7.33831942e-01 -1.41667826e-02 -8.07248235e-01 -9.05378461e-01 -4.45231587e-01 -6.55330360e-01 3.88276249e-01 9.31424379e-01 4.79440778e-01 1.19323328e-01 -3.78154784e-01 3.54539931e-01 -3.60179693e-01 -3.27619553e-01 2.43318319e-01 -1.25940695e-01 3.16983312e-01 4.03173178e-01 6.85988784e-01 1.16377054e-02 -7.21821964e-01 5.81640303e-01 -7.44298697e-01 -3.84486079e-01 7.81168878e-01 7.53366232e-01 4.18631434e-01 3.04279812e-02 4.20045704e-01 1.34225726e-01 4.36911769e-02 -2.80800402e-01 -5.93030155e-01 2.37835899e-01 -4.12391067e-01 1.93319514e-01 6.98491752e-01 -5.35523772e-01 -1.39400184e+00 3.92264694e-01 1.44061550e-01 -1.09691691e+00 -4.48903330e-02 8.73467773e-02 -2.06538796e-01 -4.62306619e-01 1.36265948e-01 4.33241546e-01 -1.06005311e-01 -3.70958328e-01 3.73100877e-01 4.68230784e-01 6.70906961e-01 -3.15664411e-01 1.41300213e+00 6.44806027e-01 1.85029417e-01 -6.96795821e-01 -1.08152080e+00 -1.00502872e+00 -5.69250047e-01 -4.49789971e-01 1.17020440e+00 -1.05619371e+00 -1.16450107e+00 5.78084946e-01 -1.01403284e+00 2.75621414e-01 1.68162473e-02 7.97583103e-01 -4.71150249e-01 8.51663411e-01 -1.22820571e-01 -1.06920159e+00 -5.69956362e-01 -1.13258135e+00 1.34421134e+00 6.37179375e-01 6.03422880e-01 -8.42868686e-01 8.42540562e-02 3.90124172e-01 3.97550464e-01 4.02393788e-02 2.04845250e-01 -2.33678922e-01 -7.92859733e-01 -1.15395047e-01 -5.90403616e-01 5.52542806e-01 1.31437764e-01 1.28733113e-01 -1.19920027e+00 -3.68780345e-01 -5.35635315e-02 -4.80411768e-01 1.27409685e+00 3.59260440e-01 1.13542879e+00 1.83610424e-01 -5.09450972e-01 7.05238104e-01 1.41841233e+00 8.07811841e-02 8.26526105e-01 2.82207727e-01 1.13470626e+00 3.80984187e-01 1.43561351e+00 2.08811626e-01 3.53751361e-01 8.73828650e-01 6.43151939e-01 -9.72540677e-02 5.68798818e-02 -3.11642140e-02 6.84098899e-01 6.89188302e-01 -1.89235032e-01 -7.12318569e-02 -4.21145648e-01 3.90705436e-01 -2.17816138e+00 -1.22209930e+00 -2.34563515e-01 2.13013029e+00 3.14363629e-01 -3.93215865e-02 -1.95740182e-02 -5.69186881e-02 5.65121531e-01 3.82957935e-01 -6.31581187e-01 -9.02442485e-02 -1.77411124e-01 -6.92635849e-02 7.20787585e-01 -1.43344074e-01 -1.24915195e+00 1.13645256e+00 5.19426298e+00 1.11478174e+00 -1.12351334e+00 4.97208834e-02 -5.15021309e-02 9.20571387e-02 3.34778503e-02 -4.59038317e-02 -1.02297413e+00 5.70137322e-01 7.84862638e-01 1.35696977e-01 2.79684514e-01 8.43495965e-01 4.23252016e-01 -2.31484339e-01 -7.97267616e-01 1.16463494e+00 2.74025984e-02 -7.43422985e-01 -3.00262868e-01 -5.50112240e-02 5.44999838e-01 3.82472217e-01 3.94131094e-01 2.74396628e-01 -3.85563867e-03 -7.88672507e-01 6.30120337e-01 9.00103569e-01 6.43138111e-01 -6.88610196e-01 9.06520724e-01 1.26567990e-01 -1.88877809e+00 -3.65957439e-01 -8.62315357e-01 5.90288043e-01 -1.42148793e-01 3.17279547e-01 -2.86708742e-01 9.88744318e-01 9.60600078e-01 1.13414717e+00 -6.33278728e-01 1.13316405e+00 8.40230957e-02 2.34261557e-01 -3.10142875e-01 2.21584532e-02 5.46853244e-01 -2.68059462e-01 4.29809719e-01 9.19018388e-01 3.35287213e-01 -1.33741759e-02 3.15910339e-01 8.20361853e-01 2.77651250e-01 -3.22088093e-01 -6.21993542e-01 3.53207369e-03 3.54185611e-01 1.69128561e+00 -2.71951109e-01 -1.35007069e-01 -5.40643334e-01 7.87125289e-01 9.50096101e-02 4.69254792e-01 -1.25276470e+00 -4.42476571e-01 1.03378129e+00 -5.07203579e-01 7.18868375e-01 -4.49440062e-01 1.01594264e-02 -1.24958849e+00 1.72160361e-02 -6.63843095e-01 2.38764033e-01 -1.09257269e+00 -1.19928217e+00 5.42249858e-01 -1.40440434e-01 -1.76383948e+00 1.22967251e-01 -7.28075743e-01 -6.79821491e-01 9.51029956e-01 -1.91252768e+00 -1.34802794e+00 -7.93644249e-01 8.69543016e-01 3.26382220e-01 9.57668852e-03 9.73652974e-02 2.79073775e-01 -8.49490404e-01 6.92009330e-01 3.11091363e-01 5.56639908e-03 8.80816340e-01 -6.64538562e-01 3.21260840e-01 9.06151593e-01 -3.09223324e-01 6.68403804e-01 5.42994320e-01 -6.43437207e-01 -1.64534473e+00 -1.33886182e+00 5.76496601e-01 -6.30624950e-01 5.85283220e-01 -1.15887202e-01 -9.44969833e-01 3.12049389e-01 -1.02029657e-02 3.62050533e-01 1.40420869e-01 -2.94514507e-01 -4.80539739e-01 -4.88612890e-01 -1.05542302e+00 3.39131266e-01 1.06411290e+00 -4.37800437e-01 -5.73467255e-01 2.33192831e-01 6.93353713e-01 -1.30585924e-01 -7.93989956e-01 4.15668488e-01 6.79022729e-01 -8.61911595e-01 1.21978462e+00 -2.81598747e-01 -8.30043480e-02 -8.72430861e-01 6.25919085e-03 -9.50275302e-01 -4.39640820e-01 -2.82360643e-01 -5.95142245e-02 1.08122671e+00 -2.88424164e-01 -6.29355729e-01 5.56489348e-01 4.00421262e-01 -3.11109513e-01 -5.47387362e-01 -9.38999057e-01 -1.18614054e+00 -2.92947620e-01 -3.65267783e-01 3.01857531e-01 5.52146077e-01 -2.95298666e-01 1.59240186e-01 -6.39079452e-01 1.71040639e-01 1.11412883e+00 9.30178314e-02 5.60720384e-01 -1.30383086e+00 2.62972087e-01 -1.50771469e-01 -5.03986418e-01 -1.13175809e+00 3.61366153e-01 -7.00408220e-01 4.42678094e-01 -1.24284136e+00 3.08751374e-01 -5.35958529e-01 -6.30882382e-01 3.29547614e-01 -4.91336048e-01 1.88448906e-01 3.86413217e-01 3.80610406e-01 -1.04487491e+00 1.02571166e+00 1.32987368e+00 -1.51688457e-01 4.99127544e-02 -2.16410663e-02 -4.40663099e-01 4.87879187e-01 6.47741020e-01 -3.93517554e-01 -4.05210733e-01 -4.96171832e-01 -1.71739921e-01 -5.21758914e-01 5.81466973e-01 -1.26474023e+00 5.17860830e-01 -3.22315335e-01 5.01539886e-01 -9.62279260e-01 3.96074295e-01 -1.13772285e+00 -1.58007458e-01 4.33672041e-01 -2.06853449e-02 -2.50236332e-01 2.64653981e-01 1.00141764e+00 -4.29039970e-02 4.87738028e-02 1.03407943e+00 1.40604585e-01 -9.68074620e-01 7.62649238e-01 -6.76339790e-02 -3.24404519e-03 9.68078732e-01 -4.99122322e-01 -5.86540163e-01 -1.03819231e-02 -7.63242841e-02 4.81672168e-01 3.85589063e-01 8.15220535e-01 7.95315146e-01 -1.55928457e+00 -3.19211960e-01 -6.13804087e-02 4.76866275e-01 -3.35698396e-01 4.48522121e-01 1.31841564e+00 1.75296858e-01 6.34636581e-01 -3.53971243e-01 -8.93210709e-01 -1.30651641e+00 8.43288362e-01 4.83158648e-01 -6.64848387e-02 -4.45777655e-01 6.80394530e-01 5.87034583e-01 -1.36330649e-01 1.08376272e-01 -4.15321916e-01 -4.31034476e-01 -2.33071461e-01 4.52027023e-01 3.54870647e-01 -2.35761538e-01 -1.10160673e+00 -8.70726764e-01 1.13883960e+00 3.11842673e-02 4.58652258e-01 8.74948442e-01 -4.56179619e-01 2.38066643e-01 2.31280565e-01 1.19329643e+00 -4.99306798e-01 -1.37741864e+00 -6.13466859e-01 -1.67361408e-01 -7.13843226e-01 2.99116433e-01 -4.33770269e-01 -1.42586267e+00 1.00197148e+00 9.44016337e-01 -4.28843319e-01 1.36509645e+00 -2.26300314e-01 1.00310850e+00 5.41230917e-01 2.68263698e-01 -1.32933474e+00 3.86093348e-01 6.44552529e-01 3.32100391e-01 -1.47600234e+00 -6.45048916e-02 -2.56790698e-01 -9.11724210e-01 1.06743395e+00 8.37262273e-01 -1.99973851e-01 3.12224299e-01 -6.57251328e-02 4.23479006e-02 -1.34838551e-01 -7.03974545e-01 -7.31207013e-01 7.04100192e-01 7.11631000e-01 2.23837003e-01 -3.60866845e-01 -3.21753025e-01 2.02636749e-01 5.77112794e-01 2.67331414e-02 3.70130055e-02 6.19870782e-01 -6.79968297e-01 -7.04934657e-01 -3.00560266e-01 2.54674911e-01 -1.84824154e-01 -6.91830814e-02 -3.27614039e-01 5.71361125e-01 3.19968581e-01 1.10525763e+00 -7.55879134e-02 -8.16706836e-01 1.87655196e-01 -2.58635730e-01 4.13325995e-01 -6.78344443e-02 -5.70001781e-01 9.42227989e-02 -1.26146168e-01 -9.23683345e-01 -7.85187125e-01 -7.50160635e-01 -1.17479074e+00 -2.28452310e-01 -7.01726854e-01 -1.00301743e-01 7.03391016e-01 1.16365969e+00 2.93036878e-01 4.79406416e-01 9.38646019e-01 -8.82566094e-01 -7.90697753e-01 -8.43434870e-01 -4.19017166e-01 2.39294678e-01 3.61998469e-01 -1.13490093e+00 -4.89414781e-01 -2.72265226e-01]
[6.3203277587890625, -2.213862180709839]
47ce01fa-543a-4ea2-bff5-a61ddfcf7115
dinosr-self-distillation-and-online
2305.10005
null
https://arxiv.org/abs/2305.10005v1
https://arxiv.org/pdf/2305.10005v1.pdf
DinoSR: Self-Distillation and Online Clustering for Self-supervised Speech Representation Learning
In this paper, we introduce self-distillation and online clustering for self-supervised speech representation learning (DinoSR) which combines masked language modeling, self-distillation, and online clustering. We show that these concepts complement each other and result in a strong representation learning model for speech. DinoSR first extracts contextualized embeddings from the input audio with a teacher network, then runs an online clustering system on the embeddings to yield a machine-discovered phone inventory, and finally uses the discretized tokens to guide a student network. We show that DinoSR surpasses previous state-of-the-art performance in several downstream tasks, and provide a detailed analysis of the model and the learned discrete units. The source code will be made available after the anonymity period.
['James R. Glass', 'Wei-Ning Hsu', 'Michael Auli', 'Heng-Jui Chang', 'Alexander H. Liu']
2023-05-17
null
null
null
null
['online-clustering']
['computer-vision']
[-6.06307462e-02 5.58223128e-01 -3.05578589e-01 -5.29759765e-01 -1.05210543e+00 -8.31147969e-01 5.25665283e-01 2.63184130e-01 -2.81594992e-01 4.04141277e-01 6.69240892e-01 -6.33608282e-01 1.63362712e-01 -4.27489430e-01 -5.13653278e-01 -4.40361500e-01 -2.00889930e-01 7.14708805e-01 1.27534373e-02 2.68844813e-02 -3.87968302e-01 3.62819970e-01 -1.29102039e+00 4.65783358e-01 6.05591655e-01 7.63073564e-01 8.73857811e-02 8.09710026e-01 -3.50358576e-01 1.04823256e+00 -5.54946303e-01 -2.86378086e-01 -1.90529246e-02 -4.62906659e-01 -8.93276215e-01 6.64967597e-02 -3.89555320e-02 -3.46425086e-01 -8.98074389e-01 8.56525064e-01 6.76058412e-01 8.14618886e-01 6.87376082e-01 -9.72238898e-01 -9.66949463e-01 1.54623139e+00 -7.42888302e-02 1.00492381e-01 8.14102963e-02 -2.31339887e-01 1.19550574e+00 -1.06407940e+00 4.00065064e-01 1.38427472e+00 3.88812512e-01 7.05580950e-01 -1.47613895e+00 -8.38290215e-01 3.58961165e-01 -8.72935280e-02 -1.47166753e+00 -7.90549278e-01 8.60663235e-01 -3.68758112e-01 1.09540534e+00 -5.94803365e-03 3.02377552e-01 1.37872243e+00 -8.88510525e-01 9.81829345e-01 6.70582175e-01 -6.04647875e-01 5.65382659e-01 4.33014005e-01 4.56999779e-01 6.34110272e-01 -2.97693223e-01 8.74705315e-02 -6.03862941e-01 -2.94914722e-01 4.62210804e-01 1.80703439e-02 1.34267673e-01 -5.56004681e-02 -7.28128374e-01 9.79377151e-01 3.42586964e-01 3.64441991e-01 6.32879362e-02 8.09366032e-02 3.96954745e-01 4.33818877e-01 6.91986322e-01 2.45934874e-01 -4.32582200e-01 -4.50913012e-01 -1.03952479e+00 -4.16074216e-01 8.65318894e-01 9.24686253e-01 8.62568319e-01 4.87945467e-01 2.48971850e-01 1.33627236e+00 1.75317883e-01 9.70189646e-02 1.11946094e+00 -1.06503630e+00 3.88410240e-01 2.29878172e-01 -4.52379525e-01 -4.13376898e-01 -1.03278466e-01 -4.43512261e-01 -6.72353923e-01 -1.54313594e-01 5.60076609e-02 -5.79885066e-01 -7.24780560e-01 1.68198454e+00 1.94505788e-02 4.33016509e-01 3.47719193e-01 4.28649485e-01 7.70862579e-01 9.56221581e-01 -1.32818207e-01 3.59991146e-03 8.93930137e-01 -1.15023291e+00 -6.29018903e-01 -2.49485493e-01 7.63889432e-01 -1.93987399e-01 8.14205647e-01 3.85805488e-01 -9.51222479e-01 -5.86896956e-01 -8.93876970e-01 -2.78717190e-01 -5.74857950e-01 1.81079626e-01 4.26323563e-01 6.17723644e-01 -1.30945683e+00 7.86728978e-01 -1.11569941e+00 -3.11424524e-01 2.92850703e-01 2.66857117e-01 -2.76661605e-01 1.06851280e-01 -1.19353628e+00 3.26812685e-01 3.84274036e-01 -3.53089452e-01 -1.22221947e+00 -6.56799257e-01 -9.84209001e-01 7.53232837e-02 2.05720440e-01 -1.85171708e-01 1.50082898e+00 -8.44660342e-01 -1.91955531e+00 7.49937534e-01 -3.59566897e-01 -9.38633859e-01 3.20962548e-01 -2.59017311e-02 -4.31540519e-01 2.10878149e-01 -3.02587837e-01 7.65941381e-01 8.73648405e-01 -1.11748838e+00 -4.91958797e-01 -1.57571420e-01 -2.33446702e-01 1.29308224e-01 -7.14003861e-01 -7.48129860e-02 -3.64230603e-01 -7.29957044e-01 1.16891250e-01 -6.70660257e-01 -2.56001979e-01 -5.75231791e-01 -5.20735741e-01 -5.98109603e-01 5.02311349e-01 -6.27910733e-01 1.31731558e+00 -2.69614625e+00 1.75117292e-02 2.40382880e-01 2.55258918e-01 1.05444781e-01 -1.85089841e-01 5.73732316e-01 -3.50664318e-01 3.83780390e-01 -1.87621027e-01 -1.17448497e+00 2.37090349e-01 2.89651990e-01 -8.17421317e-01 4.13345337e-01 -4.95604500e-02 6.38277233e-01 -9.53809440e-01 -2.62365699e-01 1.96601942e-01 4.93863851e-01 -7.23189056e-01 3.80888879e-01 -2.54194409e-01 2.56119519e-02 1.45621538e-01 2.30437800e-01 2.14329228e-01 7.21412748e-02 3.02574724e-01 2.04285488e-01 4.57988121e-02 1.13947749e+00 -1.13869846e+00 1.67043030e+00 -6.66666269e-01 7.82534838e-01 2.23028600e-01 -1.08785200e+00 1.10090125e+00 4.78345335e-01 3.27339619e-01 -1.34817660e-01 -5.21588400e-02 9.94738489e-02 -2.51862764e-01 1.92503035e-02 5.24470806e-01 -3.39078084e-02 -8.58254209e-02 8.17529380e-01 5.47784030e-01 -5.28938137e-02 -1.98302329e-01 7.09902585e-01 1.07371593e+00 -3.55010271e-01 1.02690883e-01 7.33821914e-02 8.95172432e-02 -3.07551563e-01 3.40674788e-01 7.59199381e-01 -1.49833649e-01 6.39107883e-01 4.03105259e-01 1.54575158e-03 -8.22038352e-01 -1.50676107e+00 9.00034308e-02 1.43621004e+00 -5.74735999e-01 -6.51238322e-01 -6.90784454e-01 -7.16944933e-01 2.54833877e-01 8.87051105e-01 -5.17694175e-01 -2.65310109e-01 -2.41514772e-01 8.99909660e-02 8.43464971e-01 7.50960112e-01 -8.16931576e-02 -8.09407353e-01 4.41363633e-01 3.43378097e-01 1.80670060e-02 -1.06494367e+00 -5.77116370e-01 7.90359318e-01 -7.20441639e-01 -2.87417948e-01 -4.88509834e-01 -1.01971924e+00 5.30890048e-01 2.62636423e-01 7.66311288e-01 -1.01961240e-01 -7.74608552e-02 3.73991400e-01 -4.16841507e-01 -2.15982255e-02 -7.59885013e-01 4.53271925e-01 5.27214170e-01 1.54980505e-02 4.07464355e-01 -1.04746103e+00 -4.87866886e-02 -1.82309926e-01 -6.78535223e-01 -3.85867417e-01 2.13592246e-01 8.16336215e-01 3.88470292e-01 2.51384079e-01 8.23072791e-01 -9.62595224e-01 5.70064664e-01 -7.17433929e-01 -3.62247676e-01 -1.99793816e-01 -3.22216451e-01 1.79614559e-01 9.21404541e-01 -6.20546639e-01 -6.63794160e-01 2.49679610e-01 -2.36655593e-01 -8.66253138e-01 -3.85680348e-01 4.03620541e-01 -1.31866366e-01 5.41007876e-01 6.59966707e-01 3.21265787e-01 6.11577965e-02 -7.28675246e-01 1.01807964e+00 1.37950516e+00 8.18788826e-01 -6.10825062e-01 8.69524002e-01 3.29766870e-01 -9.64959562e-01 -1.25131130e+00 -9.00038064e-01 -4.87749696e-01 -6.99236393e-01 3.10568571e-01 5.46276808e-01 -1.44699383e+00 -5.22813678e-01 -4.20059962e-03 -1.02588165e+00 -7.02882886e-01 -7.87445188e-01 6.29801929e-01 -4.44613099e-01 2.07438722e-01 -7.88646221e-01 -1.35530138e+00 3.46174613e-02 -6.93068683e-01 5.62115967e-01 4.60660532e-02 -4.75937426e-01 -1.15061188e+00 4.41100858e-02 4.40505296e-01 2.08473250e-01 -6.50011599e-01 7.94483900e-01 -1.43092561e+00 -2.53196210e-01 -5.40706329e-03 1.65077448e-01 7.46704698e-01 2.51557082e-01 7.89477304e-02 -1.62336969e+00 -2.50543475e-01 -2.86263198e-01 -5.59731722e-01 1.21921444e+00 2.77870595e-01 1.34530020e+00 -4.87861454e-01 -1.59614533e-01 6.15093708e-01 7.40693748e-01 1.32156566e-01 1.51122138e-01 -1.74320251e-01 6.63763106e-01 7.39527762e-01 -1.70493662e-01 5.28969407e-01 4.40852314e-01 6.35280162e-02 -6.59806356e-02 -4.93603051e-02 -2.00022280e-01 -9.60397422e-01 7.43634284e-01 1.61923492e+00 8.10094476e-01 8.19222722e-03 -6.72552228e-01 9.82192695e-01 -1.67013419e+00 -9.45291758e-01 6.02034509e-01 1.88692296e+00 1.12887442e+00 1.29359841e-01 5.96594930e-01 2.10133612e-01 7.69501746e-01 3.40153933e-01 -5.19693434e-01 -6.02867246e-01 1.54538468e-01 4.77831274e-01 5.25129974e-01 1.00689566e+00 -1.04636002e+00 1.38625014e+00 6.76327944e+00 8.79130244e-01 -7.09409297e-01 3.52289170e-01 6.24056339e-01 -1.78477779e-01 -5.27506232e-01 5.04171252e-02 -8.02173138e-01 3.08675706e-01 1.48662174e+00 -3.55879962e-01 8.70661020e-01 1.16355217e+00 2.08771273e-01 4.19285387e-01 -1.37545931e+00 9.42162097e-01 3.74353165e-03 -1.35622203e+00 -2.56835539e-02 -2.80359946e-02 6.11852050e-01 3.55317622e-01 2.25866064e-01 6.16011024e-01 1.36107194e+00 -1.01915836e+00 4.56030637e-01 4.39394964e-03 8.16920221e-01 -9.76325095e-01 7.66769275e-02 4.47624058e-01 -1.10465384e+00 -2.82657981e-01 -4.18485582e-01 -1.58696294e-01 -9.89217609e-02 6.53980494e-01 -1.15852082e+00 4.10482109e-01 3.73524636e-01 7.62153506e-01 -3.74089271e-01 5.22858322e-01 -3.51443261e-01 1.29845047e+00 -4.28104341e-01 -1.30173445e-01 2.15630591e-01 -1.12665497e-01 2.24993631e-01 1.42235231e+00 -7.60783115e-03 -4.22937162e-02 4.28865671e-01 1.02395928e+00 -5.48067510e-01 -1.06105760e-01 -7.52894640e-01 -6.59068406e-01 1.22106588e+00 1.04998112e+00 -3.77595603e-01 -6.12032592e-01 -1.44468024e-01 1.10847616e+00 7.30526209e-01 5.27024686e-01 -2.42354944e-01 -6.57828689e-01 1.07780480e+00 -2.61268802e-02 4.47547913e-01 -5.26871681e-01 -1.90059826e-01 -1.09336293e+00 -4.65010881e-01 -7.34845281e-01 3.89889628e-01 -5.60227633e-01 -1.44602644e+00 5.78099668e-01 -1.98389649e-01 -7.70355105e-01 -7.84420192e-01 -3.71414363e-01 -6.62552357e-01 7.12158799e-01 -1.20105326e+00 -6.90948069e-01 3.60927850e-01 5.11400163e-01 6.48092568e-01 -6.03132248e-01 1.06143403e+00 1.92442223e-01 -7.56262302e-01 9.22509849e-01 3.93443078e-01 7.89863229e-01 5.50410569e-01 -1.47211981e+00 7.59894967e-01 7.04534709e-01 9.05618727e-01 8.03197443e-01 3.27209413e-01 -4.38623577e-01 -1.05118215e+00 -1.21718109e+00 8.02494943e-01 -4.17773366e-01 1.10723245e+00 -9.35091436e-01 -8.45860720e-01 1.04482484e+00 1.94616616e-01 1.71394542e-01 1.02264678e+00 2.91048288e-01 -7.27110922e-01 -1.33959606e-01 -6.92500472e-01 5.78084588e-01 1.05625165e+00 -1.30732048e+00 -8.36007953e-01 3.93290669e-01 1.50190330e+00 -1.07923940e-01 -6.03658319e-01 -5.30162513e-01 1.05430029e-01 -5.38514912e-01 7.83919275e-01 -7.77651429e-01 -3.33180092e-02 -6.99703023e-02 -4.36105222e-01 -1.54386127e+00 -2.56463140e-01 -1.18637896e+00 -2.22286865e-01 1.70637405e+00 6.95040941e-01 -7.68498719e-01 7.64506876e-01 2.28281423e-01 -3.58022094e-01 -3.81003499e-01 -1.02509904e+00 -9.60048556e-01 3.98059219e-01 -6.68057144e-01 4.68429953e-01 1.10082173e+00 5.03960490e-01 4.42260742e-01 -3.42433080e-02 4.35329139e-01 5.88379264e-01 -1.27919346e-01 6.14228070e-01 -1.13139606e+00 -5.11811733e-01 -2.66980767e-01 -2.62987792e-01 -1.46801627e+00 8.65634203e-01 -1.43664610e+00 -5.79565279e-02 -1.36084509e+00 -4.22710478e-01 -3.34059894e-01 -3.21874022e-01 5.22598624e-01 3.28487933e-01 -1.94177762e-01 5.20403832e-02 3.01869046e-02 -6.39244437e-01 6.05648875e-01 1.78523377e-01 -1.49453774e-01 -6.06386483e-01 -6.62877336e-02 -1.05848181e+00 5.32697618e-01 8.06221247e-01 -5.17127216e-01 -4.48086619e-01 -2.90179908e-01 -4.76909578e-01 -2.54347324e-01 -2.02450659e-02 -9.37042832e-01 5.32793343e-01 3.21416736e-01 2.18054906e-01 -5.99012256e-01 6.57542646e-01 -5.17881572e-01 -5.75460970e-01 1.35484636e-01 -9.89096224e-01 -3.37567180e-01 -2.52162688e-03 6.20328188e-01 -2.60465175e-01 -1.23025402e-01 7.46872723e-01 -5.03706485e-02 -3.06077540e-01 1.47952482e-01 -7.31357336e-01 2.99080253e-01 4.62176383e-01 1.60186499e-01 -1.03368528e-01 -8.81827295e-01 -1.12113154e+00 2.39167318e-01 1.51399121e-01 4.35646057e-01 4.82976466e-01 -1.40431190e+00 -4.95400190e-01 4.51515436e-01 -2.57400542e-01 4.32780236e-02 7.52042681e-02 2.34440595e-01 1.95495084e-01 1.89183697e-01 6.97882712e-01 -3.34552526e-01 -8.92683983e-01 5.23873031e-01 2.87912607e-01 8.56518298e-02 -6.99673057e-01 1.03832459e+00 -7.11576715e-02 -8.75562847e-01 1.01255500e+00 -5.90838015e-01 1.51141346e-01 3.40655386e-01 5.14406860e-01 2.95090646e-01 -4.45289612e-02 -2.96384186e-01 -4.62735534e-01 -1.45550743e-01 -1.46200478e-01 -7.31132627e-01 1.42094004e+00 -3.02495629e-01 3.03723454e-01 1.08537221e+00 1.61119449e+00 1.75999820e-01 -1.20191967e+00 -6.18283331e-01 7.07671419e-03 1.07813194e-01 7.17839673e-02 -4.05173063e-01 -7.48475969e-01 1.24682164e+00 2.59558052e-01 4.86491859e-01 5.07005751e-01 4.43502545e-01 8.25137913e-01 6.13923728e-01 -2.23177493e-01 -1.41487300e+00 2.57704288e-01 7.71391690e-01 4.28700894e-01 -8.32447946e-01 -4.15709406e-01 -4.64898683e-02 -7.38184333e-01 1.00636637e+00 2.69547522e-01 -2.94524759e-01 1.02689099e+00 6.56552315e-01 1.01520024e-01 4.06333596e-01 -1.16010606e+00 -3.60107511e-01 -1.48054034e-01 9.42140400e-01 4.72655088e-01 2.96521544e-01 7.60113835e-01 1.23510647e+00 -7.76112854e-01 -4.13977087e-01 4.95804697e-01 5.11198521e-01 -6.72813296e-01 -9.73338366e-01 -2.66759485e-01 1.87923312e-01 -1.66416809e-01 -2.57429391e-01 -8.75309765e-01 2.61273354e-01 -2.37396196e-01 1.29199255e+00 4.99410570e-01 -8.44724655e-01 1.92620933e-01 6.79668605e-01 -1.36767939e-01 -1.04846573e+00 -5.42762399e-01 2.14483008e-01 1.88979074e-01 -2.80586332e-01 1.66169107e-01 -5.39678752e-01 -1.58750463e+00 -1.70125157e-01 -1.22319706e-01 5.80193162e-01 7.08234072e-01 7.99264312e-01 6.23493493e-01 5.55561543e-01 1.18607473e+00 -7.74989069e-01 -8.25715482e-01 -1.09126842e+00 -7.86188066e-01 -1.54630849e-02 6.34557366e-01 -3.25616419e-01 -1.02645123e+00 -6.48940802e-02]
[14.421330451965332, 6.6013898849487305]
4452e2a4-5dea-4f82-8826-f2bc0b06adde
shift-invariant-waveform-learning-on
2108.03177
null
https://arxiv.org/abs/2108.03177v2
https://arxiv.org/pdf/2108.03177v2.pdf
Shift-invariant waveform learning on epileptic ECoG
Seizure detection algorithms must discriminate abnormal neuronal activity associated with a seizure from normal neural activity in a variety of conditions. Our approach is to seek spatiotemporal waveforms with distinct morphology in electrocorticographic (ECoG) recordings of epileptic patients that are indicative of a subsequent seizure (preictal) versus non-seizure segments (interictal). To find these waveforms we apply a shift-invariant k-means algorithm to segments of spatially filtered signals to learn codebooks of prototypical waveforms. The frequency of the cluster labels from the codebooks is then used to train a binary classifier that predicts the class (preictal or interictal) of a test ECoG segment. We use the Matthews correlation coefficient to evaluate the performance of the classifier and the quality of the codebooks. We found that our method finds recurrent non-sinusoidal waveforms that could be used to build interpretable features for seizure prediction and that are also physiologically meaningful.
['Austin J. Brockmeier', 'Carlos H. Mendoza-Cardenas']
2021-08-06
null
null
null
null
['seizure-prediction']
['medical']
[ 2.95334011e-01 -3.21858585e-01 2.93401450e-01 -3.31721097e-01 -7.87926137e-01 -7.28477418e-01 5.14984846e-01 3.52807373e-01 7.27037564e-02 6.29803300e-01 5.51460207e-01 -1.60860389e-01 -4.32807326e-01 -1.06282435e-01 -1.70595169e-01 -1.01185608e+00 -7.54217386e-01 2.01170027e-01 1.05563447e-01 2.21900180e-01 7.04939008e-01 7.35931098e-01 -1.48212218e+00 5.97725689e-01 6.02552652e-01 1.18797708e+00 1.22942992e-01 6.30139172e-01 2.31823072e-01 1.22365370e-01 -9.37124491e-01 5.81668973e-01 7.98029378e-02 -1.04280245e+00 -5.16191661e-01 -7.22178910e-03 -1.90269649e-01 5.90741873e-01 -8.97511765e-02 9.24246669e-01 6.21894240e-01 -1.97283030e-01 9.81254101e-01 -1.09965706e+00 -5.35937548e-02 3.04401010e-01 -1.19782425e-01 8.33056867e-01 3.80563140e-01 -5.03213890e-02 6.75290942e-01 -7.62988269e-01 6.82483852e-01 3.74671280e-01 6.32288158e-01 3.56211543e-01 -1.20511603e+00 -8.30554307e-01 -4.07625049e-01 3.55775297e-01 -1.61766136e+00 -6.96886897e-01 8.80927086e-01 -7.50567138e-01 8.45619857e-01 3.80050510e-01 1.20653713e+00 7.87505925e-01 9.11409378e-01 1.89392433e-01 1.15516829e+00 -1.69400468e-01 6.40767694e-01 -2.10799843e-01 -1.49086639e-01 1.72433227e-01 1.19934394e-03 -1.88343991e-02 -6.06006205e-01 -5.69321990e-01 3.59555751e-01 -3.00562590e-01 -7.34556496e-01 -1.96017362e-02 -1.11838269e+00 7.20793128e-01 4.70365509e-02 9.61246550e-01 -7.99558163e-01 -1.04517020e-01 2.25489885e-01 3.33386242e-01 2.36982524e-01 8.56951475e-01 -3.12257051e-01 -3.07583809e-01 -1.46039200e+00 1.51999757e-01 5.02274096e-01 9.76722687e-02 5.57761014e-01 -1.51330298e-02 -1.21136736e-02 6.23100996e-01 -1.28554210e-01 1.54394135e-01 1.12359631e+00 -5.53396046e-01 -3.17419320e-01 5.17656684e-01 -2.27946997e-01 -8.37316096e-01 -6.64302707e-01 -2.86364585e-01 -4.03604805e-01 -7.63007104e-02 2.23303944e-01 -1.74255088e-01 -7.01014936e-01 1.34264207e+00 -1.61513597e-01 6.12978995e-01 1.21553995e-01 6.62456512e-01 5.21440566e-01 4.59077328e-01 -1.03623860e-01 -6.51658773e-01 1.02128208e+00 1.14460953e-01 -5.24501860e-01 -5.08092083e-02 8.51446927e-01 -5.74328184e-01 5.78077555e-01 3.52933854e-01 -7.26114511e-01 2.97935661e-02 -9.02143478e-01 8.54348660e-01 -1.07217267e-01 -1.20972835e-01 1.45072311e-01 3.00971717e-01 -1.01951623e+00 5.71208954e-01 -8.72578084e-01 -3.64111722e-01 2.91382492e-01 3.80623549e-01 -3.84295523e-01 6.01526618e-01 -9.22346652e-01 9.07926083e-01 2.67941535e-01 -4.29794878e-01 -6.21749520e-01 -6.19066119e-01 -4.95576233e-01 -9.01039504e-03 -4.68595356e-01 2.07069159e-01 7.35116839e-01 -8.51244330e-01 -7.17167258e-01 8.97400975e-01 -4.50148731e-01 -4.39744532e-01 -2.86204278e-01 7.19905674e-01 -8.31421852e-01 5.32218635e-01 1.67662073e-02 1.93264976e-01 8.74580443e-01 -8.99904966e-01 -6.61992192e-01 -4.73782659e-01 -9.32447970e-01 5.67769669e-02 5.67610227e-02 2.22895250e-01 3.29155982e-01 -7.61870861e-01 6.26389682e-01 -8.76890957e-01 1.78760111e-01 -5.71012080e-01 -2.65989900e-01 -4.40487176e-01 1.05900431e+00 -5.78141451e-01 1.32486284e+00 -2.27802658e+00 -1.17438987e-01 6.66932583e-01 1.59651980e-01 -9.15101916e-02 -1.28433689e-01 4.82519418e-01 -5.02775908e-01 -1.18092254e-01 -2.96534181e-01 6.60096407e-01 -4.75457758e-01 -1.97324887e-01 -4.36927199e-01 7.03677833e-01 2.34979227e-01 7.16779470e-01 -7.08173513e-01 -1.64152130e-01 -1.53801262e-01 4.03227389e-01 -1.67837009e-01 1.49711072e-01 1.91724539e-01 7.97728181e-01 -2.97620386e-01 6.02408171e-01 -2.20856182e-02 9.00276676e-02 -9.57132056e-02 -1.56286016e-01 -1.80123538e-01 5.02414644e-01 -7.81960011e-01 1.15710044e+00 1.38279036e-01 1.11190856e+00 -4.20834422e-01 -1.27899623e+00 1.27722824e+00 5.92746913e-01 7.94785500e-01 -6.19879246e-01 3.01835805e-01 4.87727165e-01 5.97899318e-01 -6.25786960e-01 -3.94150943e-01 -2.12295413e-01 3.68296765e-02 6.43843949e-01 -1.21332919e-02 -1.60012856e-01 4.17891853e-02 -2.96782076e-01 1.46102953e+00 -4.30248529e-01 3.41045141e-01 -7.38163650e-01 2.63947964e-01 -1.41938001e-01 5.41995704e-01 4.05449569e-01 -1.31924674e-01 6.62124693e-01 4.70589638e-01 -5.00078201e-01 -7.44362950e-01 -9.51699257e-01 -6.38448119e-01 4.17188108e-01 8.12458172e-02 -1.90742254e-01 -6.70938492e-01 -1.90631330e-01 -1.86374769e-01 7.17998087e-01 -5.74458241e-01 -3.71283680e-01 -2.93691248e-01 -1.02393568e+00 5.81895351e-01 3.00831556e-01 -2.40088269e-01 -1.31571960e+00 -1.21465158e+00 3.08104992e-01 -1.37889162e-01 -5.65424323e-01 -2.89657325e-01 6.27187848e-01 -7.38921821e-01 -1.09378278e+00 -4.38483030e-01 -9.69295323e-01 6.42123520e-01 -4.33787465e-01 4.97732788e-01 -5.48605472e-02 -6.23614728e-01 1.58078060e-01 -4.86980319e-01 -3.26470822e-01 -1.82669103e-01 -5.91107786e-01 3.47057074e-01 1.78846866e-01 7.17047989e-01 -8.78105879e-01 -6.79555774e-01 1.83659419e-01 -6.29560947e-01 -3.85945082e-01 3.98792595e-01 4.66846853e-01 8.69241655e-01 1.66205332e-01 9.75579441e-01 -1.72225386e-01 9.66082036e-01 -6.55303180e-01 -3.24543685e-01 1.35373935e-01 -4.60794419e-01 -6.14600442e-02 6.50836110e-01 -8.57267976e-01 -1.64612472e-01 1.78753302e-01 1.42073929e-01 -3.58092070e-01 -4.48902667e-01 5.70453823e-01 2.03625649e-01 -2.08688468e-01 9.04689193e-01 8.90223682e-01 -4.11084980e-01 -1.05003431e-03 -3.29912096e-01 9.66574728e-01 7.84685016e-01 -6.49879724e-02 3.05211544e-01 2.68880427e-01 -1.08897217e-01 -1.03592467e+00 -4.28570695e-02 -5.09508967e-01 -4.59159523e-01 -2.79901356e-01 8.53259265e-01 -4.20880675e-01 -3.42627943e-01 2.35628754e-01 -7.89784670e-01 -1.84472457e-01 -1.94718525e-01 9.80965376e-01 -6.53811634e-01 -8.39060619e-02 -1.54870436e-01 -7.95651376e-01 -3.70992631e-01 -1.01572752e+00 7.08159089e-01 2.53588945e-01 -9.10204411e-01 -7.90194929e-01 4.60038424e-01 -4.51060206e-01 2.17386976e-01 4.51760232e-01 1.17827618e+00 -1.40696406e+00 1.11891866e-01 -5.45698285e-01 3.27092171e-01 -1.92104623e-01 3.40781838e-01 -2.48292625e-01 -9.61757720e-01 -2.20485732e-01 4.43129033e-01 -3.20294313e-02 3.69052291e-01 7.95447886e-01 1.05366135e+00 -3.03985924e-01 -6.05925202e-01 7.63555348e-01 1.18080926e+00 1.12191510e+00 6.34985924e-01 -2.76109893e-02 -7.11297020e-02 5.65455616e-01 7.94353485e-02 2.52753496e-01 -1.71003878e-01 5.02098978e-01 -1.68448240e-01 2.60916293e-01 2.10113302e-01 9.06663686e-02 3.50645125e-01 8.50316107e-01 1.57338038e-01 5.14837056e-02 -1.28213084e+00 9.83536839e-01 -1.30441427e+00 -1.04386723e+00 9.48043242e-02 2.10747790e+00 7.61865079e-01 -6.90792277e-02 9.13533494e-02 4.91393030e-01 6.19950533e-01 -2.94268459e-01 -6.20544374e-01 -4.94354218e-01 -2.50851393e-01 5.61017156e-01 7.22145885e-02 5.33050224e-02 -6.74303949e-01 3.93690556e-01 7.02182102e+00 4.00243580e-01 -1.63866484e+00 -4.02961373e-01 5.12628913e-01 -1.66314557e-01 4.54212576e-02 -7.36095756e-02 -3.01011056e-01 6.31146312e-01 1.43948543e+00 -6.69434547e-01 4.53704268e-01 3.46674472e-01 4.34008867e-01 -3.59929889e-01 -1.24827611e+00 1.12609339e+00 2.02754334e-01 -1.49750555e+00 -3.07673126e-01 4.15738560e-02 6.40347302e-01 2.27444500e-01 -2.12647736e-01 -3.89419943e-01 -3.78241360e-01 -1.16559851e+00 4.28946525e-01 7.11443663e-01 7.74184644e-01 -6.47547603e-01 5.75840771e-01 1.87417522e-01 -1.19428408e+00 -4.51823324e-02 -3.39001343e-02 1.19735658e-01 -1.36412308e-01 2.52405107e-01 -1.04833591e+00 -1.30784884e-01 6.06695294e-01 6.98312998e-01 -6.70743585e-01 1.64592862e+00 1.04197554e-01 8.52466643e-01 -3.15270931e-01 -6.66422918e-02 7.60190114e-02 -3.40214968e-02 7.40437388e-01 1.20320976e+00 7.99991310e-01 3.50031227e-01 -1.42990306e-01 8.52899492e-01 4.01844144e-01 1.39096394e-01 -6.86090946e-01 -3.74932855e-01 8.72694850e-01 1.02310073e+00 -1.26640189e+00 -1.75141543e-01 -6.89072162e-02 5.65333486e-01 -3.48384649e-01 2.18703628e-01 -6.76186085e-02 -8.57994616e-01 5.38650453e-01 2.32810289e-01 1.91166118e-01 1.89308569e-01 -4.59690750e-01 -8.85735631e-01 -8.86179656e-02 -6.71892643e-01 3.03220779e-01 -8.02908719e-01 -9.77643132e-01 9.01192904e-01 -1.05863409e-02 -1.42488432e+00 -7.19876766e-01 -1.23705231e-01 -1.32076526e+00 8.99913609e-01 -7.74994195e-01 -4.29987788e-01 -8.52269121e-03 6.54978514e-01 2.85981417e-01 -3.04144174e-01 1.06365132e+00 -2.25487754e-01 6.31567314e-02 2.35116169e-01 1.48305789e-01 2.45476365e-01 3.11378062e-01 -1.02787817e+00 -4.61245254e-02 7.14845538e-01 3.40209514e-01 4.82367337e-01 8.95596981e-01 -8.04968596e-01 -8.67787421e-01 -9.76381004e-01 1.23443604e+00 -1.33036613e-01 8.22584033e-01 -5.93351722e-02 -1.09516203e+00 2.95348793e-01 -1.40369534e-01 -7.70258680e-02 9.82433379e-01 -4.30058211e-01 1.21691503e-01 1.37997463e-01 -1.01092839e+00 3.04137290e-01 4.66130316e-01 -6.59030557e-01 -1.15946782e+00 2.52921224e-01 -1.85977757e-01 1.04606628e-01 -1.00718462e+00 3.15129459e-01 7.11074352e-01 -7.75501609e-01 3.66894901e-01 -3.39670032e-01 -1.01389447e-02 -3.49943489e-01 8.54701996e-02 -1.64932060e+00 -1.78956926e-01 -7.22251356e-01 4.46121752e-01 4.41548347e-01 2.97535717e-01 -6.81337237e-01 5.87688088e-01 1.69393122e-01 -2.67755747e-01 -1.06739020e+00 -9.71222699e-01 -7.57479548e-01 1.09237805e-01 -2.82327473e-01 4.51465249e-01 9.13851917e-01 9.34208572e-01 -7.05791637e-02 3.08016717e-01 1.00778401e-01 3.03858578e-01 3.08542401e-01 -1.47076055e-01 -1.37821043e+00 1.69490620e-01 -5.69777012e-01 -1.10516393e+00 6.48720115e-02 1.42812937e-01 -1.29572725e+00 2.77718604e-01 -1.34237957e+00 -1.23068094e-01 -1.97648868e-01 -3.84121180e-01 6.70871019e-01 3.38806957e-01 6.18354321e-01 -2.99048901e-01 4.65513378e-01 -3.15833427e-02 6.12351904e-03 5.09463847e-01 1.90271154e-01 -7.82476962e-01 -2.65960693e-02 -4.05629218e-01 8.02450955e-01 9.16192174e-01 -7.66887248e-01 -3.41939270e-01 3.93175393e-01 -2.65364081e-01 2.94456214e-01 1.30307108e-01 -1.36699283e+00 3.68288487e-01 4.58374843e-02 7.12528408e-01 -6.26247942e-01 5.06090075e-02 -3.98666590e-01 4.16039556e-01 5.92528224e-01 -5.03591537e-01 9.53740552e-02 1.63873971e-01 2.31001124e-01 -2.01463357e-01 -2.18527630e-01 8.86474073e-01 2.92743593e-01 -3.88480693e-01 1.31457418e-01 -1.13961852e+00 2.52096027e-01 1.10866368e+00 -6.34586275e-01 1.16382770e-01 -3.60599875e-01 -9.97283638e-01 -2.39773721e-01 3.95213157e-01 1.90337911e-01 9.14473832e-01 -1.23113191e+00 -7.13629246e-01 6.47038758e-01 2.38564447e-01 -7.01951027e-01 -2.30926573e-01 1.17657650e+00 -2.89811134e-01 6.48565233e-01 -6.07847989e-01 -8.03719938e-01 -1.23402536e+00 -2.22480208e-01 4.25524205e-01 3.22466075e-01 -8.94995689e-01 7.62052834e-01 -2.22654268e-02 5.50688863e-01 7.59361917e-03 -3.42005610e-01 -5.28062344e-01 2.64316082e-01 8.70450139e-01 3.94170843e-02 4.22441393e-01 -9.42898333e-01 -5.74514151e-01 5.63703895e-01 3.52016985e-01 -1.46581456e-01 1.41324878e+00 3.31825346e-01 -1.12112604e-01 6.88961148e-01 1.31444895e+00 -2.12855652e-01 -1.02281570e+00 2.15626925e-01 4.65994745e-01 -1.64471596e-01 6.67367056e-02 -8.28835249e-01 -8.17430377e-01 5.35288095e-01 9.25269186e-01 4.34038192e-01 1.37427974e+00 3.49723846e-01 6.08486295e-01 9.85493287e-02 4.28330928e-01 -8.12910199e-01 -3.14463735e-01 9.60109979e-02 8.54738891e-01 -2.79192656e-01 -3.52488160e-01 2.87488103e-01 -6.20414078e-01 1.32016230e+00 -1.75040916e-01 -5.26342571e-01 9.04550314e-01 4.85166699e-01 1.66909873e-01 -5.41793227e-01 -8.48136604e-01 -2.02182103e-02 7.04490900e-01 7.92649508e-01 3.54860723e-01 -2.36542430e-02 -6.29844487e-01 6.07313931e-01 -6.19353771e-01 3.48896682e-02 4.68349248e-01 7.68476009e-01 -7.34297097e-01 -5.68178535e-01 -3.26322228e-01 1.14666545e+00 -3.49567205e-01 -3.84873003e-02 -6.68574750e-01 2.68099546e-01 2.31547132e-01 8.44047308e-01 4.74073678e-01 -6.19381726e-01 -6.76400512e-02 5.83656967e-01 3.81654203e-01 -6.53393567e-01 -4.80890065e-01 5.83523810e-01 -2.89090276e-01 -5.47736943e-01 -2.32363746e-01 -1.01637530e+00 -1.67271686e+00 5.66809416e-01 -1.47563547e-01 5.94576001e-01 6.79541588e-01 1.04399240e+00 3.21666151e-01 1.39410153e-01 6.23576820e-01 -7.37439513e-01 9.98819619e-02 -9.05851901e-01 -9.45391059e-01 4.69814897e-01 4.47527379e-01 -5.20115316e-01 -6.89365208e-01 5.05068481e-01]
[13.199695587158203, 3.5357720851898193]
db6619c7-123f-4071-8c99-e96fa8240796
simulation-based-frequentist-inference-with
2306.07769
null
https://arxiv.org/abs/2306.07769v1
https://arxiv.org/pdf/2306.07769v1.pdf
Simulation-Based Frequentist Inference with Tractable and Intractable Likelihoods
High-fidelity simulators that connect theoretical models with observations are indispensable tools in many sciences. When coupled with machine learning, a simulator makes it possible to infer the parameters of a theoretical model directly from real and simulated observations without explicit use of the likelihood function. This is of particular interest when the latter is intractable. We introduce a simple modification of the recently proposed likelihood-free frequentist inference (LF2I) approach that has some computational advantages. The utility of our algorithm is illustrated by applying it to three pedagogically interesting examples: the first is from cosmology, the second from high-energy physics and astronomy, both with tractable likelihoods, while the third, with an intractable likelihood, is from epidemiology.
['Olivia F. Prosper', 'Harrison B. Prosper', 'Ali Al Kadhim']
2023-06-13
null
null
null
null
['epidemiology', 'astronomy']
['medical', 'miscellaneous']
[ 1.57984570e-01 1.31213292e-01 -2.72572875e-01 5.26309423e-02 -5.03138244e-01 -4.93735999e-01 9.96350288e-01 2.35820606e-01 -5.66488981e-01 1.29806638e+00 -3.51669431e-01 -8.97822261e-01 -6.83851361e-01 -7.81195164e-01 -8.80922675e-01 -7.87327945e-01 -1.60032332e-01 5.96281707e-01 3.20180386e-01 -1.81593057e-02 4.03995335e-01 6.71554327e-01 -1.73852158e+00 -6.05764329e-01 1.22191739e+00 4.45559949e-01 3.38416666e-01 6.56754017e-01 4.24042298e-03 5.44553697e-01 -3.82063955e-01 -3.63534808e-01 -1.98346883e-01 -4.25419122e-01 -7.23316729e-01 -1.24305993e-01 -7.12227169e-03 -4.82927375e-02 -2.50725299e-01 9.96318877e-01 1.12105548e-01 1.45965531e-01 9.46463227e-01 -1.15777338e+00 8.67726747e-03 8.58704373e-02 -3.23451102e-01 3.17083865e-01 6.47153437e-01 8.78048502e-03 6.66536808e-01 -3.64604563e-01 5.84227443e-01 1.23546362e+00 5.22939920e-01 -4.74912673e-02 -1.42789567e+00 -5.86058617e-01 -9.85977873e-02 3.54011148e-01 -1.66866112e+00 -6.41336888e-02 3.37858140e-01 -6.09296203e-01 6.86485350e-01 1.37083009e-01 9.49646771e-01 1.03019273e+00 3.58840257e-01 5.14132082e-01 1.41872740e+00 -8.11620116e-01 7.13471055e-01 5.87989271e-01 -1.19540259e-01 7.55139172e-01 6.29310250e-01 7.72160888e-01 -4.06783044e-01 -5.25764465e-01 8.34251165e-01 -2.08708838e-01 -2.49580517e-01 -3.48748803e-01 -1.03046727e+00 1.02650738e+00 3.67884785e-02 1.09555654e-01 -1.77344263e-01 3.72583747e-01 -2.04103529e-01 3.19305658e-01 2.65460610e-01 3.79282594e-01 -3.53687167e-01 -1.82358161e-01 -9.08102393e-01 4.71854031e-01 1.11743808e+00 6.86505437e-01 8.51079583e-01 3.26812174e-03 6.78314865e-01 1.50711879e-01 6.73830926e-01 6.21884465e-01 1.70811579e-01 -8.94777656e-01 -1.40138507e-01 7.33225867e-02 5.72109938e-01 -5.74509919e-01 -1.83390632e-01 -3.43302816e-01 -5.34110963e-01 3.38020474e-01 5.32793820e-01 -1.48000836e-01 -5.70078790e-01 1.56421518e+00 7.26872504e-01 7.36534357e-01 -1.65885258e-02 3.86433572e-01 6.63576186e-01 7.49876738e-01 -4.30824031e-04 -5.26828349e-01 1.27347076e+00 -1.53396234e-01 -6.10683262e-01 -4.81579863e-02 3.95721227e-01 -6.65219069e-01 7.41802812e-01 6.61923468e-01 -8.37768257e-01 -7.71867037e-02 -9.45770681e-01 3.95584792e-01 -3.42424750e-01 -2.88225144e-01 9.59498823e-01 1.02784503e+00 -9.47987139e-01 4.72271532e-01 -8.92068863e-01 -5.00257075e-01 2.11382620e-02 2.97525048e-01 -9.16506648e-02 2.19110027e-01 -1.27191520e+00 1.11021197e+00 1.70178980e-01 -3.47721219e-01 -8.01834404e-01 -8.29760253e-01 -6.96884871e-01 9.25007239e-02 5.23057342e-01 -7.84242272e-01 1.37599325e+00 -4.06082958e-01 -1.99666798e+00 4.64682311e-01 -4.25140619e-01 -2.08174989e-01 5.35182476e-01 2.18139328e-02 -3.91210169e-01 1.59562111e-01 -3.63442242e-01 1.82185918e-01 4.73732144e-01 -1.12731659e+00 -3.00742298e-01 6.24694638e-02 -3.63152660e-02 -1.24273129e-01 2.55196571e-01 -5.07131964e-02 5.04766181e-02 -2.64801979e-01 3.40661816e-02 -8.83873165e-01 -5.06651461e-01 1.31906018e-01 -2.70831823e-01 -3.13914746e-01 3.64910662e-01 -3.35263699e-01 8.00525665e-01 -1.59424949e+00 -1.75435618e-02 5.82692564e-01 2.40904894e-02 1.21111041e-02 2.35086933e-01 8.78581345e-01 4.53193374e-02 4.18763608e-02 -2.19020337e-01 6.18588626e-02 -4.86623533e-02 3.40535045e-01 -2.93247789e-01 7.18360662e-01 1.97173446e-01 7.33180881e-01 -1.00883853e+00 -5.81136048e-01 5.47732174e-01 3.75017285e-01 -5.74250400e-01 1.18271872e-01 -2.90927321e-01 7.09565699e-01 -7.08247483e-01 6.26901612e-02 3.41368020e-01 -5.92288077e-01 3.49721372e-01 3.12101185e-01 -4.63157505e-01 5.49883604e-01 -1.35745633e+00 1.16808629e+00 -5.69957733e-01 4.30821687e-01 -1.54264972e-01 -1.23971283e+00 4.25921559e-01 5.17127633e-01 1.60029918e-01 -3.54978979e-01 6.90454766e-02 1.79694090e-02 -1.83431525e-02 -7.13687181e-01 1.03435144e-01 -5.16013741e-01 9.96363312e-02 6.99774325e-01 2.11096063e-01 -5.88456988e-01 1.18551508e-01 1.16823077e-01 8.08081567e-01 2.11107194e-01 1.00611079e+00 -5.34636676e-01 3.72214139e-01 -1.89398542e-01 3.03723931e-01 1.01786137e+00 2.21910670e-01 2.03404456e-01 5.09164572e-01 -1.81024894e-01 -8.74042273e-01 -1.38611329e+00 -4.48743701e-01 3.86332512e-01 1.11914910e-01 -1.79633781e-01 -4.59593326e-01 -2.71513671e-01 5.98423220e-02 9.07661080e-01 -4.24950331e-01 -1.06681243e-01 -1.16085529e-01 -1.02336121e+00 1.87716931e-01 -2.17377897e-02 -1.34085659e-02 -7.09352314e-01 -5.21001875e-01 -2.50882842e-02 -2.07522046e-02 -1.00267828e+00 3.86831284e-01 2.24534258e-01 -8.96258354e-01 -8.04457664e-01 -3.01869214e-01 -2.75486141e-01 5.20854950e-01 1.50998458e-01 1.13271081e+00 2.35518873e-01 -4.05513793e-01 4.49778318e-01 -7.48594105e-02 -6.52725220e-01 -7.35699952e-01 -2.16784716e-01 4.04242128e-01 -4.09055322e-01 1.41467124e-01 -9.75525022e-01 -1.61918432e-01 1.85988680e-01 -9.95380104e-01 -2.35996276e-01 5.26400149e-01 5.35195589e-01 4.68206346e-01 1.55882671e-01 6.84792995e-01 -8.14819276e-01 1.18139558e-01 -8.63870025e-01 -1.34195197e+00 3.53654355e-01 -6.69762909e-01 1.25571832e-01 5.73789358e-01 -4.66381639e-01 -9.51951265e-01 1.74168907e-02 -2.66122520e-01 1.74821448e-02 -3.57881278e-01 6.90168023e-01 -1.19585931e-01 -3.97750258e-01 6.61619663e-01 3.55063170e-01 -2.00844765e-01 -5.21810710e-01 1.70399427e-01 4.27203417e-01 4.13035482e-01 -6.09094977e-01 9.29429889e-01 3.32290500e-01 3.33509654e-01 -1.41970825e+00 -7.05336750e-01 -2.41513923e-01 -4.94584739e-01 -1.43800035e-01 4.35185820e-01 -7.20285237e-01 -1.19170702e+00 7.87371621e-02 -9.67638433e-01 -1.88513577e-01 -3.65463436e-01 1.16861081e+00 -7.37901390e-01 4.05816942e-01 -1.67157769e-01 -1.18663919e+00 4.15952146e-01 -9.43876565e-01 6.91295266e-01 4.57890928e-01 -2.23438770e-01 -1.37114537e+00 4.65665370e-01 -1.45787224e-01 1.58127137e-02 -6.64937496e-02 1.21571231e+00 -5.32333612e-01 -7.08929658e-01 -2.75559187e-01 1.00319020e-01 -3.07241619e-01 -1.69280991e-01 2.40500450e-01 -1.06616342e+00 -2.09838763e-01 2.16473132e-01 -2.71080017e-01 5.51924109e-01 6.20669425e-01 1.08718979e+00 -1.49359375e-01 -5.74296296e-01 5.42499244e-01 1.50299716e+00 7.34191760e-02 5.94388843e-01 -1.44891376e-02 1.68937445e-01 5.39545298e-01 2.98002630e-01 5.06600499e-01 1.98784173e-01 7.35804498e-01 1.11512549e-01 2.01721698e-01 2.91559428e-01 -4.74889159e-01 2.25279480e-01 9.48141456e-01 2.38791071e-02 -3.91258121e-01 -7.90567398e-01 3.98330420e-01 -1.64023912e+00 -1.07179356e+00 -3.26316416e-01 2.60432863e+00 9.06469822e-01 3.79444212e-02 8.50767344e-02 -5.87895662e-02 3.91419709e-01 -3.35505068e-01 -5.04420042e-01 -3.29986870e-01 1.91448510e-01 6.08845770e-01 3.51314068e-01 8.14296603e-01 -6.37012243e-01 4.54713315e-01 8.31441402e+00 1.06989264e+00 -5.28508544e-01 5.94692156e-02 3.50617379e-01 1.50254279e-01 -6.34165466e-01 4.32443678e-01 -8.50551903e-01 5.29545009e-01 1.57255805e+00 -3.90712023e-01 3.80785584e-01 3.71367872e-01 4.64221299e-01 -8.15280378e-01 -9.94006157e-01 6.58266425e-01 -4.07518089e-01 -1.36602223e+00 -1.37651786e-01 2.66767442e-01 6.29388988e-01 -7.36569390e-02 -1.51321918e-01 1.17713369e-01 8.97428930e-01 -1.18876314e+00 4.47388977e-01 5.41157246e-01 6.85201049e-01 -7.04014480e-01 4.99305665e-01 8.33180487e-01 -8.33525777e-01 3.46826851e-01 -3.21313977e-01 -3.75939876e-01 2.12145671e-01 1.05212879e+00 -1.01760328e+00 5.86845279e-01 2.78287649e-01 2.90730655e-01 -5.75789995e-02 1.54107809e+00 -5.46409965e-01 1.06519330e+00 -9.08798218e-01 -4.39113796e-01 1.55280549e-02 -5.46312869e-01 6.57323778e-01 1.05739450e+00 4.74298000e-01 2.25591749e-01 -7.12674158e-03 1.22033846e+00 3.11159521e-01 -1.76932767e-01 -8.23411524e-01 1.52848989e-01 7.30174363e-01 1.16469598e+00 -7.37253845e-01 -2.04850912e-01 -3.33805531e-01 2.45149553e-01 1.50781319e-01 4.15966511e-01 -7.11060226e-01 -5.05672321e-02 5.68420172e-01 4.49542999e-02 4.68891054e-01 -3.28838140e-01 4.00198027e-02 -1.01407242e+00 -4.79958802e-01 -7.00669587e-01 1.53572723e-01 -6.19139314e-01 -1.14417696e+00 -1.30225811e-02 6.35178030e-01 -1.08710754e+00 -8.36660445e-01 -4.52378482e-01 -6.15246356e-01 1.04780018e+00 -1.47445273e+00 -6.46683991e-01 2.50298023e-01 2.56533891e-01 2.01941635e-02 1.41699970e-01 8.82560432e-01 7.48470277e-02 -5.40153325e-01 3.49943161e-01 4.54318494e-01 -6.18231475e-01 -4.88971919e-03 -1.39388287e+00 4.51757580e-01 5.76125860e-01 1.75324053e-01 6.80737674e-01 1.33051956e+00 -7.49781013e-01 -1.48289907e+00 -7.21089363e-01 9.28843915e-01 -3.56287718e-01 9.75403607e-01 -3.98625761e-01 -8.84618521e-01 6.08790457e-01 -5.41663729e-02 -9.11619514e-02 7.63855875e-01 2.03241408e-01 1.10179661e-02 4.35212523e-01 -1.22519791e+00 4.75077808e-01 6.61778808e-01 -4.53485399e-01 -5.78545749e-01 7.07107782e-01 4.25188035e-01 -4.74577248e-02 -7.89496899e-01 2.80794427e-02 6.85253739e-01 -7.36848295e-01 9.91948247e-01 -5.76067269e-01 1.18275672e-01 -3.18869919e-01 1.11985296e-01 -1.26582813e+00 5.56688011e-02 -1.00002372e+00 -2.19390020e-01 7.56489158e-01 2.84580797e-01 -8.78901958e-01 6.65750086e-01 1.41792372e-01 4.68657970e-01 -5.53276956e-01 -1.02993608e+00 -1.20913875e+00 2.18685091e-01 -5.49157262e-01 4.00462180e-01 9.06567574e-01 1.14962853e-01 1.78010911e-01 -3.76401395e-01 3.98840994e-01 9.62210476e-01 1.98754087e-01 8.23338389e-01 -1.65249264e+00 -6.76483572e-01 -1.09911777e-01 -4.74487245e-01 -1.29295123e+00 1.15329958e-01 -5.66140413e-01 1.22117504e-01 -9.91335571e-01 4.21172172e-01 -3.77823442e-01 7.92815089e-02 -9.39656273e-02 -6.64064363e-02 -6.66851327e-02 -3.51493090e-01 3.22445785e-03 -1.99265510e-01 6.05026782e-01 9.40436721e-01 5.05753517e-01 -6.04165234e-02 4.32737291e-01 -4.15792078e-01 1.15563846e+00 5.50823808e-01 -8.43687415e-01 -4.57243234e-01 2.27649391e-01 5.84646225e-01 3.36757630e-01 7.28491306e-01 -7.45339811e-01 2.44480029e-01 -5.20266294e-01 1.37298316e-01 -4.94560629e-01 3.17329258e-01 -5.83066702e-01 4.73481119e-01 6.62847459e-01 -1.33629516e-01 1.58995986e-02 1.52937710e-01 7.75648296e-01 5.00268638e-02 -8.36277544e-01 8.55315089e-01 -1.49735063e-01 -4.38659459e-01 -1.55439572e-02 -9.23670173e-01 -8.90020840e-03 8.96628022e-01 9.12726372e-02 -2.17792660e-01 -7.69959271e-01 -6.68343842e-01 -1.80750460e-04 2.52306908e-01 -1.51836604e-01 4.77911025e-01 -1.01407456e+00 -5.89522421e-01 2.09058568e-01 -1.38271704e-01 -2.86953688e-01 -2.03770455e-02 1.10260069e+00 -5.03930151e-01 5.84469259e-01 1.99070439e-01 -6.71121478e-01 -8.25086117e-01 4.57235485e-01 3.60084385e-01 -3.69950384e-01 -5.99740922e-01 5.36586642e-01 4.48193222e-01 -5.86605370e-01 -3.05816885e-02 -3.20117623e-01 -4.70823459e-02 -4.02171433e-01 6.69259548e-01 5.31903982e-01 -5.15997000e-02 -2.15243608e-01 -2.38224849e-01 5.53961039e-01 8.56332853e-02 -4.40962642e-01 1.29880869e+00 -2.06477642e-01 4.98177111e-02 8.65644038e-01 9.07495975e-01 6.77585751e-02 -9.94039237e-01 -2.99869448e-01 7.99019337e-02 -4.64347959e-01 3.04294705e-01 -6.34479046e-01 -3.77403229e-01 8.14434230e-01 3.95384818e-01 4.90560025e-01 5.57885528e-01 -4.46696021e-02 4.87015635e-01 6.42003119e-01 5.86782277e-01 -5.39863825e-01 -2.51214832e-01 2.36283153e-01 3.31864566e-01 -1.06861937e+00 5.42052269e-01 -4.50783163e-01 -4.28535342e-02 8.24289322e-01 2.03000214e-02 -1.47697628e-01 8.99011016e-01 3.61538947e-01 -4.53168690e-01 -2.96460599e-01 -9.10104513e-01 -9.26952064e-02 2.31122002e-01 5.14502227e-01 1.33175865e-01 -8.25549737e-02 -2.80605435e-01 1.13262646e-01 -2.39907563e-01 1.99531361e-01 7.92299986e-01 8.55268657e-01 -7.14543819e-01 -1.11865938e+00 -4.68691409e-01 3.70791197e-01 -3.22105408e-01 -1.06437407e-01 -3.12959462e-01 1.11849034e+00 -1.33962378e-01 9.33209836e-01 -8.44677165e-02 -2.00573299e-02 -1.53517619e-01 1.69352405e-02 8.61728609e-01 -4.95411754e-01 -2.05395803e-01 -7.67166689e-02 -1.12907104e-02 -4.43333119e-01 -5.41290939e-01 -1.01666331e+00 -9.05443311e-01 -6.85372114e-01 -5.88747740e-01 4.37442273e-01 7.31488764e-01 1.40057349e+00 -5.57435788e-02 2.40285754e-01 6.57162428e-01 -7.84818649e-01 -6.89969242e-01 -7.25770235e-01 -8.13086867e-01 -2.81287819e-01 4.83266771e-01 -1.03235304e+00 -6.49460971e-01 -9.36831087e-02]
[6.6506547927856445, 3.9624979496002197]
b795c1dd-b39b-4fd0-8321-56733566555a
mathbert-a-pre-trained-model-for-mathematical
2105.00377
null
https://arxiv.org/abs/2105.00377v1
https://arxiv.org/pdf/2105.00377v1.pdf
MathBERT: A Pre-Trained Model for Mathematical Formula Understanding
Large-scale pre-trained models like BERT, have obtained a great success in various Natural Language Processing (NLP) tasks, while it is still a challenge to adapt them to the math-related tasks. Current pre-trained models neglect the structural features and the semantic correspondence between formula and its context. To address these issues, we propose a novel pre-trained model, namely \textbf{MathBERT}, which is jointly trained with mathematical formulas and their corresponding contexts. In addition, in order to further capture the semantic-level structural features of formulas, a new pre-training task is designed to predict the masked formula substructures extracted from the Operator Tree (OPT), which is the semantic structural representation of formulas. We conduct various experiments on three downstream tasks to evaluate the performance of MathBERT, including mathematical information retrieval, formula topic classification and formula headline generation. Experimental results demonstrate that MathBERT significantly outperforms existing methods on all those three tasks. Moreover, we qualitatively show that this pre-trained model effectively captures the semantic-level structural information of formulas. To the best of our knowledge, MathBERT is the first pre-trained model for mathematical formula understanding.
['Zhi Tang', 'Liangcai Gao', 'Ke Yuan', 'Shuai Peng']
2021-05-02
null
null
null
null
['headline-generation']
['natural-language-processing']
[ 4.62756604e-01 1.70657441e-01 -3.20319593e-01 -8.02136719e-01 -7.22743750e-01 -2.81553775e-01 3.65492314e-01 3.90195638e-01 2.49053434e-01 3.80246878e-01 1.87392101e-01 -4.47024435e-01 -1.37401238e-01 -1.28154624e+00 -8.75572979e-01 -1.02617882e-01 3.44825864e-01 6.79023087e-01 2.86939353e-01 -3.76690596e-01 5.21314144e-01 1.09732181e-01 -1.60461867e+00 8.54279280e-01 1.15682638e+00 1.22500479e+00 3.11334491e-01 3.21878701e-01 -8.94293010e-01 1.19577098e+00 -4.02149796e-01 -8.43617916e-01 -2.79250503e-01 -6.05540514e-01 -9.48750556e-01 -6.72859922e-02 6.17649734e-01 -2.87085980e-01 -2.63680398e-01 1.08664525e+00 -1.21437557e-01 -1.00028343e-01 5.69561481e-01 -1.04142296e+00 -7.23227978e-01 1.04058278e+00 -1.67751059e-01 1.57849476e-01 5.81754267e-01 -2.81498492e-01 1.38590097e+00 -8.09118986e-01 6.98796213e-01 1.49903297e+00 6.21474862e-01 2.44539410e-01 -8.94393623e-01 -6.67485297e-01 3.48229796e-01 4.79426622e-01 -1.37443578e+00 -3.25602680e-01 1.03981543e+00 -3.39417607e-01 9.46598470e-01 2.91474283e-01 5.39808273e-01 4.96680886e-01 1.18100591e-01 1.20791411e+00 8.50980997e-01 -5.12443006e-01 -1.71479836e-01 1.05503187e-01 4.60667938e-01 1.43677866e+00 -7.39667043e-02 -4.01278883e-01 -6.28994942e-01 -1.28627971e-01 6.15105629e-01 -3.10493000e-02 -2.95597672e-01 -1.43524304e-01 -9.34357166e-01 9.60646570e-01 2.95090973e-01 2.51988798e-01 -1.01583079e-01 1.15506358e-01 2.87268192e-01 4.26161140e-01 6.01306319e-01 5.93341589e-01 -6.13902986e-01 -6.77054822e-02 -8.46517146e-01 4.35619891e-01 9.78933454e-01 1.45090461e+00 9.25947309e-01 -3.19375932e-01 -3.25661838e-01 7.15685189e-01 3.21452171e-01 3.95583332e-01 1.92501739e-01 -7.97039628e-01 8.76876175e-01 1.20760071e+00 -3.39967102e-01 -1.19019270e+00 -2.30682671e-01 -2.92182654e-01 -3.85588706e-01 -8.49579751e-01 2.88671523e-01 3.36511433e-01 -6.68886602e-01 1.40610743e+00 2.37881616e-01 1.86507627e-01 -1.03187270e-01 5.13484240e-01 1.39374316e+00 7.59036779e-01 -6.76845908e-02 -2.54468843e-02 1.19920540e+00 -1.26460564e+00 -7.90285110e-01 8.00354853e-02 9.08048928e-01 -1.03082252e+00 1.27495015e+00 1.91514507e-01 -1.15625918e+00 -4.92628664e-01 -9.16150510e-01 -5.68271458e-01 -5.48871100e-01 4.71470058e-01 1.14232814e+00 3.84548247e-01 -5.75786531e-01 5.33038795e-01 -4.94804323e-01 -6.77201971e-02 5.66447556e-01 4.08692397e-02 1.23891383e-01 -6.31658912e-01 -1.17454886e+00 6.56035602e-01 1.00796357e-01 4.08028141e-02 -5.39108992e-01 -1.08935618e+00 -1.34867835e+00 4.78573382e-01 7.07171261e-01 -7.75089800e-01 1.39468884e+00 -8.98211479e-01 -1.40019345e+00 8.53169560e-01 -8.64049911e-01 -2.71239281e-01 5.19095436e-02 3.58111747e-02 -2.58778512e-01 2.42186800e-01 8.52520466e-02 5.33231854e-01 7.69854128e-01 -9.80633914e-01 -4.99340802e-01 -2.37143785e-01 4.82132375e-01 7.30852857e-02 -1.84119925e-01 2.83218116e-01 -2.84206539e-01 -6.52388096e-01 2.57789969e-01 -3.97765219e-01 1.59681112e-01 -2.95583904e-02 -3.85187089e-01 -7.45735109e-01 2.57685393e-01 -5.31289995e-01 1.57415485e+00 -1.90764165e+00 1.49872974e-01 1.47836521e-01 2.83325791e-01 -1.31888408e-02 -2.42232770e-01 4.28745747e-01 2.39096984e-01 5.11688180e-02 -2.36972168e-01 -3.75189364e-01 3.53706002e-01 1.37792543e-01 -8.18682671e-01 -2.20771357e-01 4.18993682e-01 1.19909036e+00 -9.85852182e-01 -4.75037009e-01 -7.54981674e-03 1.75593436e-01 -7.53390014e-01 1.51747346e-01 -9.36826289e-01 -1.32698804e-01 -6.85665488e-01 8.31998348e-01 7.37325490e-01 -5.49699366e-01 8.89382437e-02 -2.14146972e-01 2.46129274e-01 9.54095721e-01 -7.19781101e-01 1.57635677e+00 -6.24636352e-01 5.35687268e-01 -3.05576563e-01 -1.12936294e+00 8.44449341e-01 -5.09305596e-02 2.55389452e-01 -7.12392449e-01 1.01032801e-01 2.25595430e-01 6.98758382e-03 -5.76361001e-01 4.35997337e-01 -1.54204488e-01 1.36782721e-01 4.84020472e-01 3.50262481e-03 -4.21099752e-01 4.64463502e-01 4.47030842e-01 1.14071393e+00 1.32334560e-01 2.24079266e-01 -2.61463970e-01 7.78837085e-01 1.42162576e-01 2.61068821e-01 5.94637275e-01 2.37270057e-01 3.73663455e-01 6.37237430e-01 -4.62917686e-01 -3.14876825e-01 -9.81573284e-01 -1.30747497e-01 1.11651146e+00 3.10833335e-01 -1.26114917e+00 -5.98269165e-01 -7.94609427e-01 1.17079318e-01 8.99772108e-01 -5.05482912e-01 -1.94554716e-01 -7.56864429e-01 -2.86798924e-01 4.00458872e-01 4.13742244e-01 7.49219120e-01 -9.51235175e-01 -1.83078378e-01 1.98985919e-01 -2.87185997e-01 -1.24623215e+00 -4.12867367e-01 -2.07214385e-01 -8.48373055e-01 -1.48885846e+00 -1.49272189e-01 -9.88517165e-01 8.79518867e-01 2.61630744e-01 1.49313045e+00 6.13075972e-01 -1.90339804e-01 4.08166170e-01 -3.41980428e-01 -5.21234512e-01 -2.40644038e-01 1.32212723e-02 -8.13344479e-01 -9.42072421e-02 6.90763295e-01 -3.90800387e-01 -3.00434288e-02 2.11748239e-02 -7.81822383e-01 3.92259717e-01 4.61459517e-01 6.95714653e-01 6.31550789e-01 4.64723349e-01 2.83146769e-01 -1.29791212e+00 4.58138525e-01 -3.22659492e-01 -5.86936295e-01 7.50883400e-01 -5.38561702e-01 1.57323152e-01 6.46493495e-01 -2.24514157e-01 -1.20325863e+00 -1.69470057e-01 -1.56414926e-01 -7.05837086e-02 8.69017392e-02 1.02076054e+00 -3.97121757e-01 1.96321923e-02 1.17833950e-01 4.68808740e-01 -2.63216615e-01 -5.59838712e-01 2.54259557e-01 3.33529115e-01 9.27036852e-02 -9.09850597e-01 6.88167155e-01 2.33353779e-01 1.72140673e-01 -6.66718066e-01 -1.57888591e+00 -2.15451896e-01 -4.79349673e-01 2.08682194e-01 5.75767815e-01 -7.55856633e-01 -3.84146988e-01 1.90741718e-01 -1.49510968e+00 -3.17821026e-01 4.95894551e-02 -1.21102910e-02 -4.90378320e-01 3.52300704e-01 -7.06129789e-01 -5.41208744e-01 -1.96137458e-01 -1.15829182e+00 1.25184298e+00 -6.65192902e-02 -1.53048128e-01 -1.15468526e+00 -3.74833167e-01 6.95491612e-01 4.11116451e-01 -9.78770629e-02 1.76023161e+00 -4.85321432e-01 -1.10658896e+00 1.23501293e-01 -5.67837417e-01 1.34757191e-01 2.73678869e-01 -3.18594426e-02 -6.79452837e-01 4.75447297e-01 1.26145065e-01 -2.95312375e-01 1.12327027e+00 1.16334423e-01 1.49198246e+00 -3.87922138e-01 -3.34349602e-01 4.24535394e-01 9.43804204e-01 -5.14938384e-02 5.10009885e-01 -7.93389603e-02 6.26315475e-01 5.80943108e-01 6.60209894e-01 3.86902615e-02 8.59166145e-01 4.78234082e-01 -4.72967774e-02 1.95629254e-01 -3.42070729e-01 -6.69527948e-01 2.16789469e-01 8.81707251e-01 2.39767745e-01 -3.92207466e-02 -7.46262372e-01 1.81840733e-01 -1.76996756e+00 -6.92612708e-01 -1.05618097e-01 1.74685431e+00 1.16224265e+00 1.86513200e-01 -3.24493974e-01 -1.59197289e-03 3.37159008e-01 7.24887475e-03 -1.52089134e-01 -1.96140915e-01 -5.59895439e-03 6.15866482e-01 -3.32398355e-01 5.89441061e-01 -1.20247209e+00 1.19223690e+00 6.00533295e+00 1.11584091e+00 -8.80790353e-01 -1.06582992e-01 5.25116146e-01 2.71260291e-01 -5.72907150e-01 8.35399628e-02 -9.29990351e-01 3.22292566e-01 5.09449482e-01 -2.76354432e-01 6.24989927e-01 8.75912428e-01 -2.23739982e-01 -3.71111035e-02 -1.49095035e+00 9.42466140e-01 4.93915081e-01 -1.52646244e+00 5.12530744e-01 -4.53387290e-01 7.39331067e-01 -5.65973818e-01 -8.53950530e-02 7.55418301e-01 -1.27219602e-01 -1.10565329e+00 5.90987563e-01 5.97546697e-01 4.79345024e-01 -6.67982280e-01 6.88174665e-01 3.55976582e-01 -1.47435653e+00 2.00947449e-01 -5.38362801e-01 -3.93239111e-01 2.12589949e-02 6.87333286e-01 -5.77710509e-01 5.35038352e-01 9.54548344e-02 1.13629079e+00 -8.22018564e-01 8.57486188e-01 -7.25033998e-01 6.73849165e-01 -1.82795435e-01 -3.00996542e-01 1.35618195e-01 -1.56034619e-01 2.13102967e-01 1.05831718e+00 2.49058634e-01 3.49009722e-01 2.97976017e-01 1.38852322e+00 -3.33579898e-01 2.82442331e-01 -5.66356003e-01 -4.00361061e-01 3.05640250e-01 9.79861081e-01 -6.71972275e-01 -7.53112257e-01 -6.49103105e-01 7.10797369e-01 4.52602953e-01 3.56983989e-01 -9.21280682e-01 -5.76528966e-01 3.88568401e-01 8.24861676e-02 2.63841003e-01 4.30292003e-02 -4.42410946e-01 -1.46460629e+00 2.48436749e-01 -7.50941992e-01 4.92327273e-01 -7.50770569e-01 -1.32035124e+00 4.24005985e-01 8.79380777e-02 -6.83084309e-01 -6.45066053e-02 -7.01813757e-01 -5.91747105e-01 5.52629828e-01 -1.85132623e+00 -1.40139723e+00 -1.71545491e-01 3.70901853e-01 7.46407151e-01 -2.75099009e-01 7.50299513e-01 2.62094259e-01 -3.56913209e-01 7.56800652e-01 -4.94620889e-01 2.08078563e-01 4.61797029e-01 -1.19353294e+00 -6.51656091e-02 5.99208117e-01 4.30237085e-01 1.20252681e+00 3.79165560e-01 -7.50718474e-01 -1.65237737e+00 -1.15922272e+00 1.50404632e+00 -5.15803337e-01 7.94874549e-01 -4.03098941e-01 -8.05322170e-01 8.44079614e-01 1.35759963e-02 -3.05504680e-01 7.28817403e-01 3.45486671e-01 -6.07519925e-01 -1.31273836e-01 -9.60657597e-01 5.89727759e-01 1.20654655e+00 -6.86167240e-01 -8.49112511e-01 7.77744532e-01 1.07223284e+00 -6.94276571e-01 -6.57108843e-01 5.21750629e-01 2.44248316e-01 -7.84387171e-01 9.08612669e-01 -7.27631152e-01 1.12701952e+00 -6.19346164e-02 -2.84745067e-01 -9.60668206e-01 5.75748794e-02 -5.42217791e-01 -4.89326090e-01 1.31419933e+00 5.70434093e-01 -4.17369813e-01 6.42716885e-01 6.97446704e-01 -3.46922815e-01 -1.19949055e+00 -6.50381923e-01 -5.64685106e-01 3.68780792e-02 -5.36680102e-01 9.53261435e-01 8.23044181e-01 1.64517179e-01 7.73268461e-01 1.86919749e-01 -1.51797548e-01 2.52426445e-01 1.03387129e+00 6.80634081e-01 -9.88912404e-01 -3.19171101e-01 -6.45804524e-01 -2.58109003e-01 -1.60109270e+00 6.66383862e-01 -1.43011498e+00 -1.08410843e-01 -1.49169075e+00 4.43274736e-01 -4.33082908e-01 -6.88955933e-02 4.15214658e-01 -3.33774030e-01 -1.90405384e-01 1.52087748e-01 -2.63433307e-01 -9.38090980e-01 7.45052457e-01 1.52908146e+00 -5.09346008e-01 -3.36828679e-02 1.89524680e-01 -8.12151670e-01 8.35376263e-01 5.05433559e-01 -2.75345176e-01 -6.68106139e-01 -6.36213899e-01 4.10709292e-01 -2.44305372e-01 2.31980592e-01 -4.24356073e-01 1.67694613e-01 -4.25964803e-01 7.54930824e-02 -7.07574427e-01 1.66035190e-01 -6.61996901e-01 -3.88249099e-01 7.91167468e-02 -5.23149371e-01 -8.49203393e-02 1.54909402e-01 2.97336191e-01 -3.98911297e-01 -5.14869988e-01 2.98514217e-01 -2.94409245e-01 -3.76043648e-01 1.44623086e-01 -1.70459971e-01 1.60841703e-01 4.39912468e-01 3.54527056e-01 -5.02583444e-01 -2.67214090e-01 -2.92502254e-01 2.92046547e-01 1.00637503e-01 2.00066611e-01 9.41301227e-01 -1.31353605e+00 -3.28694731e-01 2.63937503e-01 2.11557359e-01 3.38277698e-01 9.51553229e-03 8.66889894e-01 -4.65266436e-01 6.92848682e-01 4.81981128e-01 -3.08584899e-01 -1.33361614e+00 5.98228574e-01 1.88838899e-01 -5.34019947e-01 -6.43585503e-01 1.03805971e+00 3.04080784e-01 -6.90183461e-01 3.48740906e-01 -1.19125676e+00 -2.73287995e-03 -3.07297975e-01 7.18261182e-01 1.47746906e-01 2.98260129e-03 -1.68084189e-01 -1.26624569e-01 6.07360542e-01 -1.20472871e-01 4.82548922e-01 1.34891248e+00 9.78618860e-02 -7.93583095e-01 2.93594301e-01 1.06889248e+00 1.04041904e-01 -4.77054685e-01 -5.29386103e-01 3.07639420e-01 -5.24653077e-01 -7.81172588e-02 -8.95428419e-01 -9.94004369e-01 1.21286571e+00 -4.13836718e-01 -7.74886981e-02 1.02445316e+00 1.70838788e-01 1.18261766e+00 9.40631092e-01 4.53847110e-01 -6.71168327e-01 2.29423046e-01 9.57971692e-01 9.90761995e-01 -1.00517952e+00 8.12726244e-02 -1.23118007e+00 -2.05415234e-01 1.47739792e+00 6.48480833e-01 7.40117431e-02 7.68042922e-01 2.29878709e-01 -1.85007498e-01 -4.67999548e-01 -8.87884974e-01 -3.24477702e-01 6.61583722e-01 1.32012919e-01 6.01752043e-01 -4.97450791e-02 -1.81071728e-01 1.16347492e+00 -5.21277785e-01 2.02842712e-01 1.00898325e-01 1.01799130e+00 -5.70692301e-01 -1.16904640e+00 3.98235619e-02 6.82192504e-01 -9.80849043e-02 -5.21860659e-01 -7.46838868e-01 5.74221313e-01 1.40244588e-01 8.45549405e-01 -2.23829150e-01 -1.14209630e-01 1.85688674e-01 2.89497703e-01 1.01711643e+00 -9.40285683e-01 -4.11208272e-01 -1.04015075e-01 2.40436807e-01 -5.53335190e-01 -2.89418548e-01 -3.39238733e-01 -1.44512427e+00 -2.08218709e-01 -2.37127677e-01 1.07296169e-01 1.68008432e-01 1.31834269e+00 3.21282387e-01 6.37661934e-01 2.61045188e-01 -3.22557777e-01 -6.29678428e-01 -8.14179599e-01 -2.82910854e-01 4.54374611e-01 7.51440525e-02 -8.18714440e-01 -1.37661204e-01 4.34190780e-02]
[9.600214004516602, 7.522762298583984]
de82a6e0-b589-4438-adeb-c9d00baa1782
conjr-conjunctive-sentence-splitter-without
null
null
https://openreview.net/forum?id=nUcR4c0hg5q
https://openreview.net/pdf?id=nUcR4c0hg5q
CONJR: Conjunctive Sentence Splitter without Parsing
In this paper, we observe and address the challenges of splitting conjunctive sentences around each group of conjuncts. Most existing methods rely on parsers to identify the conjuncts in a sentence and detect the coordination boundaries. However, state-of-the-art syntactic parsers are slow and suffer from errors, especially for long and complicated sentences. In order to better solve the problems, we formulate coordination boundary detection as a sequence tagging task and propose a specialized model CONJR without using syntactic parsers. We introduce both semantic and syntactic features and a specially designed attention mechanism to capture the symmetry among the potential conjuncts. The experimental results on datasets from various domains demonstrate the effectiveness of our proposed methods.
['Anonymous']
2021-11-16
null
null
null
acl-arr-november-2021-11
['boundary-detection']
['computer-vision']
[-6.72782362e-02 6.36038482e-02 -2.31069386e-01 -4.95166630e-01 -7.84413457e-01 -6.13842130e-01 5.33275492e-02 3.69386077e-01 -2.06333786e-01 5.38773298e-01 2.91901737e-01 -3.91994834e-01 1.53480768e-01 -5.77093542e-01 -5.55569589e-01 -2.92036176e-01 -5.31441253e-03 2.29212254e-01 5.90612710e-01 -3.23942870e-01 3.70930672e-01 -7.03634545e-02 -1.13277435e+00 4.84229654e-01 1.12805903e+00 7.69102812e-01 5.86568832e-01 3.83275986e-01 -6.50554478e-01 8.20808768e-01 -8.44113529e-01 -6.23192668e-01 5.69681935e-02 -4.51222539e-01 -1.07930160e+00 -3.00862156e-02 8.26082453e-02 6.02193400e-02 -3.49879009e-03 1.22182095e+00 2.05766961e-01 -2.72668540e-01 2.99393777e-02 -9.82676625e-01 -6.25964582e-01 1.12169266e+00 -5.92771769e-01 6.35443687e-01 6.10410094e-01 -2.69715846e-01 1.60230160e+00 -7.65025854e-01 5.37899017e-01 1.29963386e+00 6.21310294e-01 5.27159095e-01 -6.60933673e-01 -4.70136285e-01 6.67213202e-01 3.30505580e-01 -1.28320813e+00 -3.56057525e-01 8.62672746e-01 -1.29313871e-01 1.16459858e+00 1.11824051e-01 1.12093948e-01 5.83275080e-01 2.07884789e-01 9.72777247e-01 8.13604593e-01 -5.50789058e-01 4.57704812e-02 -3.70859832e-01 7.15432286e-01 6.95745289e-01 1.95899591e-01 -6.22336030e-01 -2.85427302e-01 -5.71347773e-02 3.29646617e-01 -1.90980345e-01 -1.56850532e-01 2.72835195e-01 -8.85500848e-01 7.56883740e-01 1.53731197e-01 6.93566501e-01 -4.00864035e-01 -1.02307603e-01 6.06881082e-01 6.23732172e-02 3.92591923e-01 3.17015678e-01 -7.46280670e-01 -1.36831939e-01 -5.77117860e-01 7.41397813e-02 6.59285605e-01 1.24963546e+00 6.28292263e-01 -4.67896849e-01 -2.75877565e-01 8.38466048e-01 3.31946582e-01 -5.97680546e-02 4.69377339e-01 -6.58817410e-01 1.12309706e+00 9.64590430e-01 -3.25107127e-02 -8.15459967e-01 -5.89935243e-01 -3.14848393e-01 -4.43727583e-01 -6.92731619e-01 2.95124650e-01 -4.10117507e-01 -6.29921257e-01 1.63141811e+00 4.46944535e-01 1.03819005e-01 1.07934147e-01 8.88908923e-01 9.68747318e-01 8.31101954e-01 1.42028943e-01 -4.29061472e-01 1.61313057e+00 -1.10100532e+00 -9.28646445e-01 -4.20599997e-01 8.85934591e-01 -8.69421422e-01 8.71886313e-01 -3.86182889e-02 -1.30296028e+00 -4.20531482e-01 -8.62836063e-01 -1.75407946e-01 1.22675337e-01 1.25293806e-01 7.26567268e-01 3.16837996e-01 -7.73939729e-01 3.63513321e-01 -8.73454392e-01 -1.68275714e-01 1.33764043e-01 4.28755552e-01 -2.35401645e-01 8.43586475e-02 -1.30551624e+00 5.05207598e-01 4.28499699e-01 4.62249011e-01 -1.81228027e-01 -4.20152962e-01 -1.04290593e+00 2.87498951e-01 5.21013975e-01 -3.67333412e-01 1.48695230e+00 -1.01321328e+00 -1.45693576e+00 7.15395808e-01 -6.51347816e-01 -1.91011950e-01 -7.98830986e-02 -3.57304573e-01 -2.70727515e-01 1.53366044e-01 4.46033955e-01 9.36814621e-02 2.76658654e-01 -9.45427954e-01 -9.32852864e-01 -3.28155518e-01 3.34561825e-01 3.48842651e-01 -1.47205859e-01 7.97494233e-01 -8.05204034e-01 -5.13903797e-01 4.19660330e-01 -5.63836932e-01 -2.15984315e-01 -8.46271515e-01 -7.26092100e-01 -8.13705742e-01 6.33277535e-01 -8.09913933e-01 1.62432039e+00 -2.11767888e+00 2.49186084e-01 -1.49191573e-01 2.53422886e-01 4.15684491e-01 -1.44305617e-01 4.78626013e-01 -6.30651265e-02 3.69917512e-01 -1.31484270e-01 -5.41420281e-01 -1.80366188e-01 3.47606450e-01 -2.31548116e-01 1.52858719e-01 4.87484992e-01 6.96444392e-01 -8.54060948e-01 -6.85726643e-01 -2.60052174e-01 -1.31167114e-01 -4.69022393e-01 4.12277699e-01 -2.43918836e-01 2.91644275e-01 -7.77668357e-01 6.36819601e-01 7.59804606e-01 -2.26478338e-01 6.63840950e-01 8.02117586e-02 -3.22427392e-01 1.10065675e+00 -1.14900887e+00 1.45705259e+00 -3.52782995e-01 1.59952007e-02 2.51429766e-01 -1.14422190e+00 8.25571299e-01 3.30342621e-01 4.96938303e-02 -5.78059375e-01 2.41496846e-01 1.95000917e-01 3.88505906e-01 -7.04907656e-01 3.77334535e-01 -1.74696177e-01 -4.63930815e-01 3.10233295e-01 -4.33071293e-02 5.23194194e-01 5.54936290e-01 2.85788774e-01 1.30651522e+00 -1.01828121e-01 3.55662793e-01 -2.24681050e-01 8.17513347e-01 -5.34784757e-02 1.32608426e+00 6.04662299e-01 -3.16669285e-01 5.67421079e-01 1.02568233e+00 -4.26200181e-01 -6.61739767e-01 -5.78504324e-01 1.67043045e-01 1.19330013e+00 3.09456438e-01 -8.58803689e-01 -1.02768135e+00 -1.01298106e+00 -3.62594634e-01 3.76744926e-01 -3.76394838e-01 2.31362283e-01 -1.14968336e+00 -7.16964364e-01 4.04109091e-01 8.23417962e-01 3.87559891e-01 -1.07590270e+00 -3.64379734e-01 3.65898013e-01 -5.17979622e-01 -1.45630908e+00 -5.84618866e-01 1.05760239e-01 -6.23598278e-01 -1.11879158e+00 -5.94199486e-02 -1.37356973e+00 6.64531529e-01 3.34880680e-01 9.61856365e-01 4.73751754e-01 6.66368902e-02 -3.03682864e-01 -7.82996833e-01 -9.49738696e-02 -2.91758478e-01 1.88983113e-01 -1.48398057e-01 -1.33898675e-01 4.47060734e-01 -2.94963270e-01 -1.29648328e-01 1.95647955e-01 -5.55217683e-01 -3.43904309e-02 5.72753131e-01 7.79047191e-01 4.93757099e-01 2.04608053e-01 5.57259381e-01 -1.05531108e+00 6.82014883e-01 -4.82694149e-01 -5.35013914e-01 5.12694120e-01 7.66266976e-03 2.55864799e-01 1.07534742e+00 -7.08191395e-02 -1.33756185e+00 -1.04389712e-02 -2.59845436e-01 7.77877197e-02 -1.55781329e-01 8.95302474e-01 -4.81540650e-01 4.42305207e-01 -7.22258240e-02 1.49363846e-01 -4.45253015e-01 -7.02370346e-01 1.55044654e-02 6.63274825e-01 4.10650045e-01 -7.68557310e-01 5.50883293e-01 8.09732378e-02 -3.16426873e-01 -6.58501029e-01 -1.07405972e+00 -5.75277865e-01 -6.47130132e-01 3.03530931e-01 8.26745272e-01 -8.70584488e-01 -6.23239636e-01 3.97770822e-01 -1.74814558e+00 -2.44731735e-02 3.80542845e-01 1.17448114e-01 -9.15375054e-02 9.07764733e-01 -9.47366297e-01 -6.87687635e-01 -5.05463004e-01 -1.10017705e+00 9.24612045e-01 3.92314821e-01 -1.19916603e-01 -7.85888135e-01 -2.11970862e-02 4.42758262e-01 -4.45514694e-02 -3.30129415e-01 1.07486498e+00 -7.81205297e-01 -4.82439846e-01 -1.27092481e-01 -2.44958505e-01 9.70208719e-02 1.81976095e-01 1.48606330e-01 -3.06154728e-01 2.52177149e-01 6.31926656e-02 -8.42313990e-02 6.40746951e-01 1.81743786e-01 9.17642772e-01 -2.84094363e-01 -4.50126976e-01 3.10781538e-01 1.10296440e+00 3.08527350e-01 5.13045311e-01 2.81443119e-01 7.09342360e-01 6.95788801e-01 7.86765218e-01 3.87086600e-01 7.75555789e-01 5.71361482e-01 2.65986115e-01 2.15308025e-01 1.17215529e-01 -2.88828671e-01 3.56787294e-01 1.24836981e+00 1.65716723e-01 -3.39588255e-01 -1.05666459e+00 6.13601029e-01 -2.02589846e+00 -7.85907805e-01 -5.37131369e-01 1.63440490e+00 1.07870984e+00 3.67162049e-01 -5.94111644e-02 2.03468919e-01 1.14631474e+00 6.63152412e-02 1.00397684e-01 -5.87023914e-01 -1.74302712e-01 2.48642460e-01 1.43309981e-01 6.10101700e-01 -1.24504018e+00 1.39808357e+00 6.57778454e+00 5.40470421e-01 -9.72193301e-01 6.29936084e-02 3.00241739e-01 2.62035459e-01 -2.47499347e-01 3.02318126e-01 -1.30970418e+00 5.66771805e-01 6.05814278e-01 1.10543624e-01 1.35406956e-01 6.08879805e-01 2.57137239e-01 -6.64997101e-02 -8.76690805e-01 5.17117500e-01 -5.56872226e-03 -1.28523552e+00 -2.86975414e-01 -4.79690552e-01 3.79371226e-01 -1.22797703e-02 -3.73599321e-01 2.92380631e-01 2.68382460e-01 -5.38952291e-01 8.87134612e-01 2.48071668e-03 1.05203845e-01 -5.14218569e-01 1.02093005e+00 4.04214084e-01 -1.40921545e+00 -1.83155075e-01 -4.33581620e-01 -5.98596454e-01 3.52115452e-01 5.67985535e-01 -4.76637810e-01 7.47072339e-01 6.31823182e-01 6.73110425e-01 -5.20449638e-01 9.49166179e-01 -9.48750257e-01 7.67062545e-01 -3.72037172e-01 -3.15608352e-01 3.17479938e-01 -1.08160056e-01 4.25227284e-01 1.23652673e+00 -1.05891516e-02 2.96552628e-01 2.72357434e-01 7.59739637e-01 -1.57152265e-01 4.06483352e-01 -9.30564255e-02 5.39889466e-03 7.85782099e-01 1.23259246e+00 -8.25314760e-01 -2.38527685e-01 -8.08944166e-01 7.62942672e-01 9.26238120e-01 1.68200478e-01 -8.95702600e-01 -5.95493197e-01 5.69326520e-01 -1.45918652e-01 6.52982473e-01 -3.28526348e-01 -4.47547317e-01 -1.37798774e+00 6.50592566e-01 -7.22176492e-01 5.28367698e-01 -3.25379133e-01 -1.25727987e+00 5.58762193e-01 -2.59710819e-01 -7.51830876e-01 2.13735700e-01 -3.56849343e-01 -1.21120942e+00 7.82732844e-01 -1.52333975e+00 -8.63569558e-01 1.62613064e-01 2.03842223e-01 6.51682854e-01 3.32945645e-01 5.86000025e-01 2.40021944e-01 -1.27539587e+00 6.02222323e-01 -2.08067149e-01 5.63057065e-01 3.95883799e-01 -1.16049564e+00 5.98996937e-01 1.15640318e+00 -5.45063689e-02 9.92235780e-01 5.20536661e-01 -6.42280281e-01 -1.38483441e+00 -8.94973576e-01 1.78055608e+00 -3.55363414e-02 8.05372238e-01 -5.71972549e-01 -9.69219863e-01 8.33906472e-01 4.36252266e-01 -1.09274248e-02 6.85493171e-01 5.53892016e-01 -3.42509210e-01 -1.05943553e-01 -6.83701873e-01 3.26419175e-01 1.10673296e+00 -1.97254628e-01 -1.17591631e+00 2.71647632e-01 8.42896998e-01 -4.93822575e-01 -6.18104339e-01 3.51012737e-01 3.78696024e-02 -8.90167475e-01 3.74125153e-01 -6.46613896e-01 5.98072946e-01 -4.30055320e-01 6.73065335e-02 -9.54872072e-01 -4.55845386e-01 -6.71426952e-01 1.57119513e-01 1.58118761e+00 6.67846799e-01 -4.17969555e-01 6.49335086e-01 6.11262560e-01 -7.18227029e-01 -6.40619636e-01 -9.84343171e-01 -6.58221543e-01 -4.17061448e-02 -4.32674706e-01 4.95458513e-01 7.46810555e-01 6.74082279e-01 8.65824282e-01 -1.08272590e-01 4.57402557e-01 2.03458279e-01 4.82079864e-01 3.75696599e-01 -9.91523445e-01 -3.84875089e-01 -3.18062842e-01 -2.24441364e-01 -1.25063705e+00 5.35734057e-01 -7.19598532e-01 3.47460032e-01 -1.60791051e+00 3.58322263e-01 -6.15061879e-01 -1.20367579e-01 6.49460733e-01 -6.75660074e-01 -3.27705860e-01 2.16112047e-01 1.60725564e-01 -1.13455570e+00 4.26030874e-01 8.60309362e-01 2.54159793e-03 -2.74028212e-01 -2.64324456e-01 -1.03020275e+00 8.16339791e-01 8.90886843e-01 -6.17203653e-01 9.52567626e-03 -8.33604336e-01 2.57292032e-01 7.18000233e-02 -2.58721143e-01 -6.41980648e-01 2.23815113e-01 -2.94133663e-01 -2.21955299e-01 -6.94403648e-01 -1.72534138e-01 -2.96487242e-01 -3.49757642e-01 3.12749505e-01 -1.96511775e-01 2.41012007e-01 1.18318059e-01 3.37620914e-01 -4.40286458e-01 -4.91328180e-01 4.49832022e-01 -3.84975784e-02 -6.17449284e-01 4.93218452e-02 -4.37491059e-01 3.44084352e-01 1.08316410e+00 3.27792764e-01 -4.68634248e-01 7.44436830e-02 -5.18701613e-01 5.32897651e-01 2.33512014e-01 4.16515380e-01 4.52394158e-01 -9.15381134e-01 -5.45266747e-01 1.01756833e-01 3.24316472e-02 3.15567166e-01 1.57693848e-01 8.23429644e-01 -6.21618629e-01 5.37138999e-01 1.70395166e-01 -2.28823885e-01 -1.45368075e+00 6.52784765e-01 1.75220266e-01 -4.97789264e-01 -5.13572276e-01 9.91461694e-01 2.87912399e-01 -2.33429626e-01 1.34082228e-01 -5.49275815e-01 -4.26203460e-01 -2.12262064e-01 5.51107287e-01 -1.34550944e-01 1.58913154e-02 -7.53663123e-01 -8.23288500e-01 5.10908961e-01 -3.72449815e-01 4.09279823e-01 1.20549035e+00 -2.57341146e-01 -6.18111789e-01 2.89617162e-02 9.23254907e-01 1.01830773e-01 -7.88709700e-01 -2.99177557e-01 3.90462875e-01 -8.75992551e-02 -1.70062959e-01 -3.64308685e-01 -8.83777022e-01 7.90703118e-01 -3.60904068e-01 4.15371120e-01 9.96514261e-01 3.42475772e-01 1.13455343e+00 1.96464688e-01 1.93027958e-01 -1.14952517e+00 -1.72985867e-01 9.80983973e-01 4.94563371e-01 -9.20842946e-01 -3.50654185e-01 -1.16448355e+00 -5.36484718e-01 1.11812496e+00 7.80896664e-01 -1.18709579e-01 5.33978224e-01 4.21706527e-01 3.34688239e-02 -1.79393813e-02 -9.36384797e-01 -3.74396086e-01 -1.53143793e-01 2.80400991e-01 7.75780737e-01 5.52652637e-03 -1.15056562e+00 1.25836706e+00 1.07910872e-01 -2.87699610e-01 5.67273140e-01 1.21738374e+00 -5.64654112e-01 -1.58581686e+00 -2.81072348e-01 4.80407327e-02 -9.24061239e-01 -3.65151912e-01 -5.13169825e-01 4.63366449e-01 2.85867840e-01 1.20777094e+00 1.64097294e-01 -3.44231904e-01 3.94250274e-01 4.68895249e-02 3.85879695e-01 -1.02763510e+00 -8.34133506e-01 1.65475041e-01 4.03208137e-01 -3.28939617e-01 -5.10865211e-01 -9.29339051e-01 -1.73278224e+00 2.34398618e-02 -4.90861297e-01 3.76618505e-01 2.21107662e-01 1.21716487e+00 5.23703814e-01 4.88743335e-01 7.33471155e-01 -3.26928765e-01 -5.66484571e-01 -1.00962186e+00 -3.64868283e-01 4.58733201e-01 -1.51625527e-02 -5.82200527e-01 -1.03006765e-01 -1.25016704e-01]
[10.341859817504883, 9.51630973815918]
b15b3d19-537d-4087-b921-72c467b4bd8b
incomplete-utterance-rewriting-as-semantic
2009.13166
null
https://arxiv.org/abs/2009.13166v1
https://arxiv.org/pdf/2009.13166v1.pdf
Incomplete Utterance Rewriting as Semantic Segmentation
Recent years the task of incomplete utterance rewriting has raised a large attention. Previous works usually shape it as a machine translation task and employ sequence to sequence based architecture with copy mechanism. In this paper, we present a novel and extensive approach, which formulates it as a semantic segmentation task. Instead of generating from scratch, such a formulation introduces edit operations and shapes the problem as prediction of a word-level edit matrix. Benefiting from being able to capture both local and global information, our approach achieves state-of-the-art performance on several public datasets. Furthermore, our approach is four times faster than the standard approach in inference.
['Jian-Guang Lou', 'Bei Chen', 'Qian Liu', 'Dongmei Zhang', 'Bin Zhou']
2020-09-28
null
https://aclanthology.org/2020.emnlp-main.227
https://aclanthology.org/2020.emnlp-main.227.pdf
emnlp-2020-11
['context-query-reformulation', 'dialogue-rewriting']
['natural-language-processing', 'natural-language-processing']
[ 7.63281286e-01 4.47589666e-01 -2.70806402e-02 -6.52003706e-01 -1.10669446e+00 -6.60216928e-01 8.99470389e-01 -4.38794121e-02 -4.14819628e-01 8.66120219e-01 2.60850877e-01 -5.03130674e-01 4.79452670e-01 -6.44062817e-01 -1.01380646e+00 -1.96311787e-01 4.21061486e-01 7.22342551e-01 2.05207765e-01 -6.21664941e-01 3.54164004e-01 -9.10076648e-02 -1.33458352e+00 3.97020549e-01 1.06953001e+00 7.20795453e-01 2.23239303e-01 8.09680939e-01 -3.25313002e-01 5.82756579e-01 -7.00615883e-01 -9.24730659e-01 3.12379122e-01 -8.21417570e-01 -1.26088858e+00 6.14487845e-03 4.53166991e-01 -5.35951443e-02 1.01495922e-01 1.14505076e+00 4.25843477e-01 3.25414211e-01 4.83036339e-01 -8.77039492e-01 -9.31564808e-01 8.08576763e-01 -9.94687229e-02 -9.10785422e-02 6.88522935e-01 -6.91030622e-02 1.34454155e+00 -9.59904373e-01 9.14429069e-01 1.24169314e+00 5.40222883e-01 6.73999786e-01 -1.19336271e+00 -3.71679477e-02 2.27434173e-01 1.93545759e-01 -1.14241338e+00 -5.06694794e-01 5.17802835e-01 -1.27276957e-01 1.50571442e+00 4.20513481e-01 2.74365574e-01 1.14615393e+00 2.14332758e-04 8.34693015e-01 1.07901025e+00 -7.40925252e-01 2.61910319e-01 -1.46768704e-01 1.10179633e-01 7.49687254e-01 -2.91880518e-01 -3.54172252e-02 -2.69707263e-01 1.33528367e-01 2.61084378e-01 -2.72893846e-01 -5.17179407e-02 -3.23378853e-02 -9.94831145e-01 8.02897036e-01 4.08070274e-02 1.61361545e-01 -1.33407220e-01 1.45360634e-01 6.15938485e-01 5.67515910e-01 6.49300218e-01 3.65271568e-01 -5.25436401e-01 -5.00586331e-01 -8.24454844e-01 4.14511591e-01 1.13936484e+00 1.31879759e+00 6.73053265e-01 -1.38565302e-01 -2.47647136e-01 8.77357423e-01 1.72535330e-01 1.23887241e-01 4.32113111e-01 -9.72052097e-01 6.45537019e-01 4.69679534e-01 1.50270723e-02 -4.85339016e-01 -2.40960997e-02 -1.04940720e-01 -4.44812983e-01 -7.85533413e-02 1.71990976e-01 -1.10087417e-01 -9.24060762e-01 1.83091819e+00 3.22297305e-01 2.85619557e-01 5.51293977e-02 8.23280156e-01 4.70133126e-01 6.74249589e-01 -3.86150256e-02 -2.35269144e-01 1.35851705e+00 -1.52182543e+00 -8.62078249e-01 -3.28496575e-01 5.80284715e-01 -8.88230205e-01 1.10948312e+00 4.01961923e-01 -1.51042402e+00 -5.53954899e-01 -1.17972660e+00 -4.62867916e-01 -5.48788249e-01 -1.63413420e-01 3.55981588e-01 6.09007478e-01 -1.30472672e+00 7.81795502e-01 -7.26094425e-01 -4.58561689e-01 -4.09115367e-02 1.76025271e-01 -2.00239763e-01 -1.21419281e-02 -1.43794620e+00 1.38888574e+00 4.67405140e-01 1.24650169e-02 -4.82919455e-01 -6.06621444e-01 -1.09151530e+00 -4.52733301e-02 6.19279921e-01 -1.01303840e+00 1.74330115e+00 -1.03485465e+00 -2.08133864e+00 8.35576057e-01 -5.43719828e-01 -7.31860399e-01 5.28839231e-01 -2.75264233e-01 -2.47726321e-01 -9.74666327e-02 -8.74690861e-02 4.78915930e-01 8.11653912e-01 -8.79705012e-01 -5.89515150e-01 -3.71658683e-01 3.53737384e-01 2.75805503e-01 2.03500673e-01 4.06727135e-01 -5.40574551e-01 -8.26721489e-01 -2.07861468e-01 -1.08843493e+00 -2.76921958e-01 -5.07521570e-01 -2.16327727e-01 -3.94313425e-01 4.97984827e-01 -9.76705849e-01 1.41239965e+00 -1.66862631e+00 6.98188126e-01 -2.01365426e-01 -2.29090657e-02 3.67684245e-01 -1.17372334e-01 8.18243921e-01 1.01870827e-01 3.46668571e-01 -8.84328961e-01 -7.90557504e-01 3.73798907e-01 4.84038800e-01 -2.37303123e-01 1.31375283e-01 4.94193971e-01 1.30236328e+00 -9.89348531e-01 -3.20665658e-01 2.43157409e-02 1.84938386e-01 -6.73045576e-01 3.24517846e-01 -6.78200722e-01 3.27089339e-01 -3.64903718e-01 5.31714380e-01 5.42834878e-01 1.30021468e-01 2.74551213e-01 3.03885400e-01 -7.08999112e-04 6.32665694e-01 -7.15123951e-01 2.36878157e+00 -6.26587152e-01 2.97350019e-01 6.92610070e-02 -1.34855938e+00 8.76055062e-01 3.34723800e-01 6.53346581e-03 -6.54068768e-01 6.00126162e-02 2.64409274e-01 -2.76643723e-01 -4.78260845e-01 8.46036613e-01 -3.31858724e-01 -4.03939426e-01 5.18560529e-01 1.83513448e-01 -3.52631330e-01 3.05126101e-01 1.11124612e-01 1.20414889e+00 8.24329972e-01 4.97482419e-01 -1.17580041e-01 7.14108229e-01 -5.61985886e-04 5.05916238e-01 7.34904945e-01 5.06856628e-02 7.33309746e-01 5.29438078e-01 -2.15842858e-01 -1.26009381e+00 -9.13945615e-01 1.87901869e-01 1.29980612e+00 -1.56266376e-01 -4.82379168e-01 -1.20803189e+00 -8.13884735e-01 -2.61969954e-01 9.27222908e-01 -6.30596161e-01 -1.62866246e-02 -9.90855813e-01 -5.79919815e-01 7.75940895e-01 3.84103924e-01 3.57601017e-01 -1.08668196e+00 -3.52210760e-01 3.16577047e-01 -3.88597578e-01 -1.28555238e+00 -7.36618817e-01 -1.73861086e-01 -6.57575846e-01 -6.06678665e-01 -4.18093741e-01 -8.44031751e-01 4.87433553e-01 -1.71106115e-01 1.44562781e+00 1.87083349e-01 -3.39888334e-01 2.21307185e-02 -7.79427648e-01 -4.36397314e-01 -8.79472375e-01 3.97654146e-01 -2.05628932e-01 -5.20531349e-02 2.45179445e-01 -4.51674908e-01 -8.84348676e-02 -8.49861205e-02 -9.79420245e-01 1.42302677e-01 5.22949696e-01 9.09060061e-01 4.39778119e-01 -8.43974888e-01 7.08452821e-01 -1.29195344e+00 9.15078819e-01 -3.53356838e-01 -4.89023358e-01 5.91360152e-01 -6.76240325e-01 3.37665617e-01 7.86162615e-01 1.60358459e-01 -1.36179721e+00 7.17075169e-02 -4.27795142e-01 -9.66528654e-02 -3.55802417e-01 5.26820779e-01 -9.37898979e-02 3.06156009e-01 3.27795893e-01 3.37236732e-01 1.72324300e-01 -5.81367195e-01 7.64153063e-01 7.14657187e-01 5.96887887e-01 -7.32624471e-01 6.85730100e-01 1.40320182e-01 -2.49863967e-01 -5.49890220e-01 -9.37403023e-01 -3.18550169e-01 -9.35136080e-01 1.37378156e-01 1.01798773e+00 -6.07814372e-01 -4.20651972e-01 2.75523961e-01 -1.52588689e+00 -3.02923560e-01 -1.61042646e-01 -5.50688505e-02 -7.53922164e-01 6.04850113e-01 -7.17177570e-01 -6.89866185e-01 -4.98633862e-01 -1.28157091e+00 1.31542850e+00 -1.25785470e-01 -5.01379669e-01 -1.08398807e+00 3.17702264e-01 3.87560248e-01 5.66818595e-01 2.76913673e-01 8.22929919e-01 -8.51602614e-01 -6.00068569e-01 -4.86384556e-02 -7.37451166e-02 4.87067640e-01 4.39831242e-02 2.32400442e-03 -8.09844196e-01 -3.15343365e-02 2.28012919e-01 -2.13656828e-01 7.68297553e-01 -8.94010663e-02 9.89728272e-01 -3.31111640e-01 7.54081607e-02 6.04686737e-01 1.40593565e+00 1.00489736e-01 6.38219178e-01 1.50675207e-01 5.36705554e-01 7.25978792e-01 7.85834908e-01 2.32523814e-01 5.41890800e-01 9.20683324e-01 3.60244513e-01 1.20801426e-01 -7.79047161e-02 -2.60725975e-01 5.86579740e-01 1.22691727e+00 -6.16910011e-02 -2.18063965e-01 -7.11312354e-01 4.65611190e-01 -2.09766912e+00 -1.02395606e+00 -2.78482456e-02 1.94680011e+00 1.08045900e+00 -8.55251402e-03 -1.09786190e-01 -3.57791632e-01 5.59523225e-01 2.02908844e-01 -3.17247301e-01 -1.27592385e+00 -8.56269300e-02 7.03411996e-01 3.18269014e-01 9.65406895e-01 -9.69263136e-01 1.42670691e+00 6.72760201e+00 7.03715622e-01 -7.09469676e-01 2.76372463e-01 4.15659696e-01 6.21067844e-02 -3.82165551e-01 1.37723401e-01 -7.20993102e-01 2.95585454e-01 1.11022794e+00 -1.19345583e-01 7.30768383e-01 5.37541866e-01 9.27081928e-02 -2.73751132e-02 -1.24298072e+00 4.09202635e-01 4.23522800e-01 -1.15036273e+00 4.49919961e-02 -2.67654032e-01 8.52679372e-01 3.17113996e-02 -2.32416078e-01 5.14865875e-01 5.01726270e-01 -1.24060011e+00 8.16744685e-01 4.08767015e-01 6.92681134e-01 -7.29923844e-01 7.36978352e-01 3.94729793e-01 -1.03425300e+00 1.65565118e-01 -2.15514213e-01 -4.13044870e-01 5.76004148e-01 1.82289094e-01 -7.40412533e-01 1.03139353e+00 2.09575295e-01 7.28620946e-01 -3.87385368e-01 5.40307581e-01 -5.22592902e-01 5.59852540e-01 -4.30915840e-02 -6.65415823e-02 5.29003739e-01 -6.00797415e-01 4.51191932e-01 1.56981122e+00 1.46826059e-01 1.58870220e-01 2.70113617e-01 7.08538532e-01 -2.37856179e-01 1.41824916e-01 -6.64631307e-01 3.46379675e-04 3.90016645e-01 9.25825179e-01 -3.00866097e-01 -6.33714199e-01 -5.05998731e-01 1.48785830e+00 6.69614077e-01 1.56440482e-01 -1.10322559e+00 -4.43728536e-01 6.75640702e-01 -4.35296297e-01 5.99206388e-01 -2.03998014e-01 -4.49909568e-01 -1.15816879e+00 2.98793405e-01 -1.13828576e+00 2.12225810e-01 -5.33578098e-01 -1.00721633e+00 7.59134173e-01 -1.81200900e-04 -8.81202757e-01 -5.82436383e-01 -5.47912717e-01 -6.21312320e-01 9.80385423e-01 -1.43020225e+00 -1.11656356e+00 6.21068999e-02 6.27044067e-02 1.14947057e+00 1.17604397e-01 9.91394937e-01 2.46607542e-01 -5.67475915e-01 6.51240110e-01 2.45461017e-02 3.92411686e-02 6.87342823e-01 -1.47705090e+00 1.11689436e+00 1.13164520e+00 2.02418327e-01 7.61639655e-01 8.97243083e-01 -5.24935722e-01 -1.46969521e+00 -1.03277349e+00 1.40679884e+00 -7.58919954e-01 7.41852164e-01 -5.73010564e-01 -8.72931123e-01 7.86869049e-01 6.76126719e-01 -4.71953630e-01 3.72336060e-01 -5.84753379e-02 -2.28738129e-01 3.27819496e-01 -9.71715152e-01 5.77221036e-01 1.51456583e+00 -6.86875463e-01 -1.02464736e+00 2.81201929e-01 1.18792868e+00 -9.32786644e-01 -8.24883044e-01 1.81311667e-01 3.57231915e-01 -7.31342375e-01 7.62994051e-01 -9.14829552e-01 7.83633828e-01 -8.59981328e-02 -1.22436196e-01 -1.48952425e+00 1.10898249e-01 -1.02319801e+00 -8.91518220e-02 1.20262063e+00 7.25118577e-01 -6.10234022e-01 5.16729534e-01 4.76583391e-01 -6.89355075e-01 -9.20850396e-01 -8.99622977e-01 -8.02578092e-01 1.05349950e-01 -4.10791725e-01 7.19321191e-01 7.34098077e-01 1.59517154e-01 6.09959424e-01 -4.85228419e-01 -6.61419258e-02 3.19865137e-01 2.13389397e-01 5.88582456e-01 -8.23085368e-01 -4.33129728e-01 -4.59172636e-01 -1.19895667e-01 -1.30851519e+00 6.29350066e-01 -1.24095583e+00 3.46789807e-01 -1.52243626e+00 -6.05582930e-02 -4.15546484e-02 -3.94756980e-02 4.23881382e-01 -3.38452667e-01 2.41696849e-01 1.61361456e-01 -3.17268789e-01 -6.31236970e-01 6.03936553e-01 9.21942890e-01 -2.88212579e-02 1.86172709e-01 -9.78982747e-02 -6.63014412e-01 4.78906900e-01 8.47235918e-01 -3.90163213e-01 -3.74439031e-01 -9.10583913e-01 1.98234007e-01 1.04692973e-01 8.03916976e-02 -4.40140873e-01 1.11402646e-01 -1.29781529e-01 -4.24247682e-01 -4.20325994e-01 3.82831872e-01 -3.50545228e-01 6.01877086e-02 3.06822568e-01 -4.86285210e-01 4.74519342e-01 -9.30273831e-02 5.55789709e-01 -5.03668845e-01 -5.95484972e-01 4.88897711e-01 -4.94888544e-01 -6.92895114e-01 1.54791400e-02 -2.06838980e-01 2.64902622e-01 9.27604795e-01 -4.79452722e-02 -1.98126182e-01 -1.65431112e-01 -7.79577434e-01 2.29035378e-01 5.91699481e-01 7.28948534e-01 4.22069460e-01 -9.56427872e-01 -9.42726910e-01 -1.59496702e-02 -2.17774697e-02 1.39725581e-03 -6.67393506e-02 8.49229813e-01 -5.74660063e-01 6.15878582e-01 1.20427601e-01 -3.25821519e-01 -1.30125129e+00 5.26883662e-01 9.66103896e-02 -4.61204290e-01 -4.76926982e-01 8.71080160e-01 -1.01456724e-01 -8.09205055e-01 1.88434906e-02 -3.43683869e-01 1.02496631e-01 -6.19858429e-02 3.63350987e-01 2.95044124e-01 3.27608854e-01 -6.06722772e-01 -1.57404900e-01 3.88210118e-01 -1.99526146e-01 -2.76742160e-01 1.23998582e+00 -3.16410482e-01 -4.64894295e-01 3.56312484e-01 1.37425518e+00 -2.16871157e-01 -8.15875411e-01 -3.78573865e-01 4.55097735e-01 -1.85156390e-01 -5.24900198e-01 -8.13991785e-01 -5.19708276e-01 8.49956095e-01 -9.11670774e-02 1.45365134e-01 9.37730789e-01 -4.21967953e-02 1.07139719e+00 4.75539327e-01 4.12586749e-01 -1.39472771e+00 -1.92580223e-01 1.09159482e+00 8.80599320e-01 -1.23609841e+00 -2.67689675e-01 -5.48112392e-01 -7.25427449e-01 9.54690576e-01 3.82831603e-01 -2.09146276e-01 8.18375424e-02 2.38845810e-01 -2.87539773e-02 1.21988773e-01 -1.04758358e+00 -2.28475794e-01 2.23568887e-01 4.21562523e-01 5.72250128e-01 1.81956276e-01 -7.36562312e-01 5.31609774e-01 -4.26400065e-01 -1.08927175e-01 4.90795910e-01 1.01594210e+00 -4.79652047e-01 -1.76165271e+00 8.87325108e-02 2.63621718e-01 -6.05223656e-01 -4.06829476e-01 -6.78267598e-01 6.31875217e-01 -1.83832362e-01 8.57303619e-01 -7.94212595e-02 -1.89820573e-01 4.88795996e-01 5.55805087e-01 5.30611157e-01 -9.86217856e-01 -8.22861314e-01 -3.51904422e-01 6.44112527e-01 -6.51364028e-01 -3.24833483e-01 -7.35977590e-01 -1.34578502e+00 -1.20276101e-01 -1.74434185e-01 8.05065632e-02 7.23907650e-01 1.33162773e+00 3.81600678e-01 5.32194972e-01 5.15331745e-01 -5.44501543e-01 -1.10795867e+00 -1.06780112e+00 2.28557605e-02 5.74924409e-01 1.62670940e-01 -2.61256009e-01 1.27054811e-01 2.39354745e-01]
[11.331196784973145, 9.37177848815918]
b937e453-2da1-4086-9484-02db218ace44
unsupervised-wasserstein-distance-guided
2009.02831
null
https://arxiv.org/abs/2009.02831v1
https://arxiv.org/pdf/2009.02831v1.pdf
Unsupervised Wasserstein Distance Guided Domain Adaptation for 3D Multi-Domain Liver Segmentation
Deep neural networks have shown exceptional learning capability and generalizability in the source domain when massive labeled data is provided. However, the well-trained models often fail in the target domain due to the domain shift. Unsupervised domain adaptation aims to improve network performance when applying robust models trained on medical images from source domains to a new target domain. In this work, we present an approach based on the Wasserstein distance guided disentangled representation to achieve 3D multi-domain liver segmentation. Concretely, we embed images onto a shared content space capturing shared feature-level information across domains and domain-specific appearance spaces. The existing mutual information-based representation learning approaches often fail to capture complete representations in multi-domain medical imaging tasks. To mitigate these issues, we utilize Wasserstein distance to learn more complete representation, and introduces a content discriminator to further facilitate the representation disentanglement. Experiments demonstrate that our method outperforms the state-of-the-art on the multi-modality liver segmentation task.
['Junlin Yang', 'James S. Duncan', 'Chenyu You', 'Julius Chapiro']
2020-09-06
null
null
null
null
['liver-segmentation']
['medical']
[ 3.81088763e-01 5.05793765e-02 -4.36692029e-01 -4.74015862e-01 -1.11564004e+00 -3.59443098e-01 4.02575701e-01 5.29443622e-02 -1.91470593e-01 6.65558934e-01 4.94163662e-01 -8.44296534e-04 -3.84793788e-01 -4.99051839e-01 -3.15878749e-01 -1.02338552e+00 -9.17055905e-02 5.42209744e-01 -3.51454318e-01 9.74884257e-02 -3.04012269e-01 3.26700360e-01 -6.67028069e-01 1.47195533e-01 1.05115950e+00 6.37898743e-01 7.22760707e-02 1.97369203e-01 3.08273314e-03 5.96582592e-01 -3.21062833e-01 3.82747650e-02 3.75007123e-01 -7.01312244e-01 -9.17731464e-01 3.64605427e-01 1.62307784e-01 -4.10977423e-01 -6.05079174e-01 1.21495342e+00 4.85502124e-01 2.60642655e-02 1.01234245e+00 -1.09354925e+00 -1.20374763e+00 4.61295068e-01 -8.58225107e-01 1.42567858e-01 1.36763364e-01 -3.45172472e-02 8.48434150e-01 -7.42654800e-01 7.26145864e-01 9.97832060e-01 4.07536179e-01 7.39871681e-01 -1.54296327e+00 -8.35054815e-01 -8.98020640e-02 -1.89861879e-01 -1.20484018e+00 -3.97258438e-02 1.13587999e+00 -6.68650329e-01 3.06140184e-01 -6.16578832e-02 3.48982334e-01 1.28596449e+00 1.04694754e-01 1.03219318e+00 1.33432353e+00 -2.05448970e-01 -3.04886829e-02 3.68875191e-02 9.14062187e-02 9.48367119e-01 5.04140377e-01 2.24491864e-01 -3.25573832e-01 -3.28085214e-01 1.07850158e+00 3.03293854e-01 -4.58161414e-01 -1.14509964e+00 -1.69381034e+00 1.21004677e+00 7.32255042e-01 5.54462731e-01 -4.02949512e-01 -2.66278744e-01 5.80538034e-01 3.44265521e-01 6.63415194e-01 5.97861588e-01 -4.13485259e-01 4.47107881e-01 -8.77077043e-01 -4.99671176e-02 6.48370564e-01 8.68303299e-01 6.11200809e-01 1.24350555e-01 -2.21591070e-01 8.64358187e-01 4.34838176e-01 2.05008775e-01 8.46271932e-01 -6.90044165e-01 2.81840950e-01 6.33255780e-01 -4.88911808e-01 -8.50164771e-01 -4.42528635e-01 -6.67868555e-01 -1.32072639e+00 7.10016564e-02 5.12603462e-01 -6.71538264e-02 -1.07610488e+00 1.81843042e+00 3.31203163e-01 3.69516820e-01 4.31416750e-01 1.12168264e+00 1.09428394e+00 1.22995935e-01 2.93062925e-01 -1.87306870e-02 1.39862299e+00 -9.76695299e-01 -6.34516418e-01 -1.62416488e-01 7.33557999e-01 -4.93333131e-01 6.70013964e-01 9.04500410e-02 -6.76971912e-01 -2.94844687e-01 -1.19875431e+00 1.07597865e-01 -2.21550316e-02 4.14863741e-03 6.18052781e-01 4.72733915e-01 -7.31372654e-01 3.95961016e-01 -1.01150584e+00 -2.04968408e-01 7.64681458e-01 1.90748364e-01 -8.75138342e-01 -5.52783310e-01 -1.14218867e+00 8.66805136e-01 4.18482721e-01 -3.22005689e-01 -1.20990324e+00 -9.55794632e-01 -1.33160102e+00 -2.57272840e-01 1.50378078e-01 -1.04730976e+00 8.33556652e-01 -8.98627877e-01 -1.37998760e+00 1.32736552e+00 3.77310604e-01 -2.57161468e-01 4.37649161e-01 1.70124978e-01 -2.61503279e-01 5.55532455e-01 2.89277315e-01 5.21094203e-01 8.96235168e-01 -1.31722593e+00 4.26541828e-02 -6.31209671e-01 -2.06720948e-01 2.64760852e-01 -2.86086321e-01 -3.18128020e-01 2.40746401e-02 -1.03227222e+00 4.87865627e-01 -1.05130339e+00 -5.37740171e-01 2.72850275e-01 -4.72320467e-01 1.07425554e-02 6.27004623e-01 -7.46471703e-01 4.88481671e-01 -2.28810668e+00 6.24031961e-01 3.13513689e-02 6.45101726e-01 -6.65279627e-02 -3.84027511e-01 -5.08949980e-02 -4.31811213e-01 2.03818679e-02 -4.73077923e-01 -3.34735394e-01 -1.16951577e-01 5.41952789e-01 1.55020893e-01 9.72129107e-01 2.24600866e-01 7.52373219e-01 -1.17634177e+00 -6.38233483e-01 1.26058221e-01 4.86468732e-01 -4.64167535e-01 3.50229830e-01 2.36029908e-01 1.12602019e+00 -7.00212598e-01 5.39467871e-01 8.14325511e-01 -6.13112152e-01 2.98079491e-01 -4.44495052e-01 6.65554821e-01 -4.14878316e-02 -6.67536795e-01 2.84697962e+00 -4.02403206e-01 4.34110641e-01 1.80451944e-02 -1.56604290e+00 1.01465392e+00 4.68953669e-01 1.01324713e+00 -4.59590822e-01 3.58817652e-02 2.49110818e-01 5.15373722e-02 -5.99933445e-01 -2.03643348e-02 -6.45289242e-01 -2.28952795e-01 4.63396370e-01 4.58934665e-01 -3.07645965e-02 -3.45319450e-01 2.28160813e-01 9.09280598e-01 1.81036964e-02 6.64014399e-01 -4.24246043e-01 4.14813906e-01 -6.15413673e-02 7.79057443e-01 4.58046138e-01 -6.23582304e-01 8.92047942e-01 3.17034304e-01 -3.27506781e-01 -8.27925980e-01 -1.39902198e+00 -3.50388646e-01 8.51972640e-01 2.60083526e-01 1.56117648e-01 -4.03174788e-01 -1.23000216e+00 1.32134274e-01 3.14755976e-01 -8.22091877e-01 -4.48844612e-01 -3.75146449e-01 -9.25382793e-01 6.32425725e-01 5.82552493e-01 5.07096529e-01 -4.38879967e-01 -2.88652420e-01 7.54884109e-02 -2.11951777e-01 -1.00769258e+00 -4.43776518e-01 1.78684592e-01 -1.12668526e+00 -1.18754292e+00 -1.39975178e+00 -9.25971091e-01 9.28478360e-01 2.74311244e-01 1.12689435e+00 -2.82624930e-01 -5.07936060e-01 5.77378750e-01 -3.13827366e-01 8.30148980e-02 -5.42323530e-01 1.56206377e-02 -1.17166862e-02 -1.14020258e-01 4.71385062e-01 -4.91769910e-01 -9.48919952e-01 3.39039922e-01 -1.05272031e+00 1.58183441e-01 7.18303323e-01 1.39154124e+00 6.03453994e-01 -4.88147944e-01 7.24094093e-01 -8.62202525e-01 5.62139511e-01 -8.86542201e-01 1.70684028e-02 2.26479039e-01 -6.51729763e-01 3.09601694e-01 2.35449001e-01 -6.00906193e-01 -1.03329158e+00 1.03825390e-01 3.63815635e-01 -5.87476730e-01 -3.44984293e-01 6.93209469e-01 9.40175056e-02 -1.35821313e-01 8.95842731e-01 3.57686102e-01 5.14632344e-01 -5.40119410e-01 4.65557337e-01 3.24892581e-01 2.60778695e-01 -7.29618490e-01 7.12694824e-01 5.95507622e-01 4.61722240e-02 -3.36035579e-01 -8.88231397e-01 -5.70636272e-01 -7.98827529e-01 2.82156527e-01 1.11946142e+00 -1.27536833e+00 -4.97505581e-03 1.19008809e-01 -8.75518978e-01 -6.48775995e-02 -3.36231291e-01 8.41421008e-01 -6.70311034e-01 6.00498021e-01 -6.89041734e-01 -4.97304685e-02 -1.87418357e-01 -1.43049741e+00 1.05050325e+00 5.80363013e-02 -9.21484903e-02 -1.41715407e+00 3.49143565e-01 3.78694117e-01 3.35663378e-01 7.21290827e-01 9.68054116e-01 -1.10372400e+00 -3.12944561e-01 -5.64866513e-02 -4.94319946e-01 4.31082845e-01 4.71075565e-01 -7.76435971e-01 -6.75761759e-01 -6.85640872e-01 1.81712732e-01 -4.21320081e-01 9.99366701e-01 5.03991663e-01 1.13516498e+00 -1.45403832e-01 -3.47768605e-01 8.85031044e-01 1.47498059e+00 -2.10433185e-01 1.56131700e-01 2.48647884e-01 6.27456129e-01 6.21186554e-01 2.93952823e-01 3.69710505e-01 4.74498421e-01 2.65065998e-01 3.91113043e-01 -5.58913410e-01 -3.30429584e-01 -8.00745115e-02 -6.00128137e-02 9.60188687e-01 2.50986814e-01 1.99132964e-01 -9.68080640e-01 6.85091138e-01 -1.64029813e+00 -5.23955762e-01 2.70413578e-01 1.77374649e+00 9.32853281e-01 -4.01096582e-01 -5.70633747e-02 -3.93121332e-01 5.22334993e-01 3.20440233e-01 -6.60107791e-01 1.71965465e-01 -1.23127140e-02 -4.21127714e-02 4.26780879e-01 4.53350134e-02 -1.23408329e+00 3.84256542e-01 6.05848074e+00 6.29584432e-01 -9.24745262e-01 4.60850388e-01 5.93617857e-01 1.09061822e-01 -4.09561336e-01 -3.23681921e-01 6.01920709e-02 6.77791685e-02 3.84364277e-01 -3.29987228e-01 3.65504958e-02 9.12858665e-01 -2.96374947e-01 3.12912315e-01 -1.45557880e+00 1.34330893e+00 3.57838899e-01 -1.17786574e+00 8.79426077e-02 1.80950537e-01 9.94368613e-01 1.33988366e-01 4.40186679e-01 2.76181042e-01 4.45516407e-01 -1.22792804e+00 -8.68751556e-02 3.61445934e-01 9.26599741e-01 -3.36793244e-01 5.91846466e-01 1.99477300e-01 -9.72788930e-01 2.57785410e-01 -3.23319405e-01 4.09159124e-01 -9.47993845e-02 5.16224205e-01 -1.07611573e+00 9.18680131e-01 2.29874909e-01 1.14978099e+00 -2.64877141e-01 8.69521260e-01 -1.39745250e-01 2.99667388e-01 1.14759229e-01 4.92552131e-01 3.16937268e-01 -2.36657247e-01 6.35918140e-01 1.11674654e+00 2.50402719e-01 6.23741001e-03 4.69825417e-01 1.09932578e+00 -1.25635684e-01 -1.26348203e-02 -8.84826481e-01 6.79966882e-02 -3.45777385e-02 1.27277756e+00 -5.44605911e-01 -3.26354831e-01 -6.04899406e-01 1.17587328e+00 1.81911245e-01 7.00931370e-01 -6.94077969e-01 5.27976416e-02 8.65299940e-01 -3.66070181e-01 -7.36829564e-02 -1.89616799e-01 -3.57315511e-01 -1.66130221e+00 -3.16350818e-01 -8.95481110e-01 7.03381121e-01 -3.33935916e-01 -2.05620933e+00 6.11101806e-01 8.82365480e-02 -1.58254337e+00 -2.30558768e-01 -6.59029961e-01 -2.39448696e-01 9.41812158e-01 -1.83202076e+00 -1.44389153e+00 -2.02198744e-01 9.28230107e-01 4.85762894e-01 -5.50407469e-01 1.16046488e+00 2.54185319e-01 -2.96259254e-01 5.68966746e-01 4.34852093e-01 6.00461185e-01 9.49428737e-01 -1.36232805e+00 -3.60829145e-01 5.11078119e-01 5.20361438e-02 6.39116824e-01 4.85411316e-01 -4.12549227e-01 -1.29713356e+00 -1.00631714e+00 1.27907202e-01 -2.50488877e-01 5.56021810e-01 -6.03271183e-03 -1.09092379e+00 8.28308105e-01 3.41262907e-01 5.58762729e-01 1.52623022e+00 1.25885159e-01 -7.28000343e-01 1.82747945e-01 -1.51516783e+00 4.04218823e-01 8.25903654e-01 -8.05679083e-01 -9.48897779e-01 3.94112498e-01 6.65097356e-01 -3.82127255e-01 -1.44685793e+00 4.74928647e-01 3.86412740e-01 -6.03780448e-01 1.21458817e+00 -9.42629278e-01 5.63503027e-01 -1.76026765e-02 -2.90930748e-01 -1.67840111e+00 -4.78774756e-01 -2.75196195e-01 2.26167850e-02 6.91361785e-01 2.51555949e-01 -6.62172616e-01 7.48793960e-01 6.43835902e-01 2.15473883e-02 -6.81927800e-01 -1.14306223e+00 -6.48947716e-01 7.01544642e-01 4.18823175e-02 5.07433474e-01 1.58679020e+00 3.31739843e-01 1.00808501e-01 -3.32936674e-01 2.12392539e-01 1.15420413e+00 4.25788045e-01 4.28561717e-01 -1.22350061e+00 -2.62509614e-01 -3.83799851e-01 -5.63410640e-01 -9.69067156e-01 4.54537868e-01 -1.36930323e+00 -1.42044887e-01 -1.49259126e+00 7.06372917e-01 -4.38193887e-01 -8.43431771e-01 3.95141453e-01 -1.01274967e-01 2.28536457e-01 1.30236030e-01 3.67601454e-01 -5.94905734e-01 6.16309404e-01 1.83590972e+00 -6.28144860e-01 1.17755584e-01 -3.14100236e-01 -9.94814754e-01 6.02328718e-01 6.57804787e-01 -6.44044042e-01 -4.88618016e-01 -6.08548760e-01 -3.36115479e-01 3.96694839e-01 4.50043261e-01 -5.89369655e-01 1.33496463e-01 -6.59226328e-02 6.26587808e-01 -1.07975669e-01 1.63338110e-01 -9.22811687e-01 -2.31091268e-02 5.75698853e-01 -5.94610989e-01 -4.39886183e-01 -1.46406470e-02 8.89414370e-01 -4.43769723e-01 -8.16331208e-02 9.56444740e-01 -3.52078557e-01 -6.39795721e-01 7.61590302e-01 -1.13106444e-01 2.22182319e-01 1.12169003e+00 -9.77257937e-02 -1.21310845e-01 -7.06932172e-02 -1.25419164e+00 1.46215335e-01 3.40840280e-01 4.48130429e-01 1.00752771e+00 -1.66520762e+00 -1.03279471e+00 4.61336434e-01 3.93819690e-01 9.93443802e-02 4.46951449e-01 1.00492299e+00 -3.10699821e-01 2.25266874e-01 -4.47203904e-01 -9.26164567e-01 -1.03478789e+00 5.75828731e-01 5.38113296e-01 -6.86571896e-01 -7.36575842e-01 6.96910203e-01 7.45181739e-01 -6.61781251e-01 -1.31643802e-01 -2.64045447e-01 -9.86212194e-02 -1.43083692e-01 3.29866350e-01 3.46681662e-02 -2.68132210e-01 -5.92601717e-01 -3.58597189e-01 4.73011911e-01 -3.00105035e-01 7.76930228e-02 1.38178718e+00 -1.48843586e-01 6.71865940e-02 1.57633334e-01 1.76468194e+00 -5.19153416e-01 -1.34445143e+00 -8.36883068e-01 -1.09603532e-01 -5.93825698e-01 1.62966385e-01 -8.25853646e-01 -1.29990900e+00 9.11366880e-01 8.68762255e-01 -1.31042838e-01 1.14449096e+00 2.71184146e-01 6.63112521e-01 -1.87323503e-02 2.37374485e-01 -6.03978455e-01 4.67056662e-01 1.17894739e-01 8.78118694e-01 -1.72116745e+00 3.18943039e-02 -2.38490358e-01 -9.06589806e-01 1.14230001e+00 5.00234604e-01 -3.92681360e-01 8.04119825e-01 -3.50355729e-02 9.62205827e-02 -3.60870361e-01 -2.44203895e-01 -3.75802368e-02 6.02719128e-01 9.49158251e-01 4.58576292e-01 3.61852586e-01 -1.12761520e-01 6.34221017e-01 4.30270553e-01 -2.56948657e-02 3.14157099e-01 6.23447776e-01 1.90435406e-02 -1.03397167e+00 -1.83956429e-01 3.04751635e-01 -5.64003766e-01 1.04514554e-01 6.84317052e-02 8.10324550e-01 -1.65295660e-01 5.33631682e-01 -1.86415315e-01 -2.84359064e-02 6.09474480e-02 -3.67080830e-02 6.77935839e-01 -7.59465516e-01 -1.94248751e-01 9.51763019e-02 -4.04192746e-01 -4.30023462e-01 -7.52235472e-01 -5.59824705e-01 -1.11809242e+00 2.36269102e-01 -1.47275195e-01 -1.56633519e-02 3.65706146e-01 8.17087591e-01 3.03189456e-01 6.53541982e-01 6.42816663e-01 -7.01485634e-01 -9.44491863e-01 -8.45039606e-01 -8.69918585e-01 9.52676654e-01 7.27538288e-01 -8.43450963e-01 -1.93833768e-01 -2.49471348e-02]
[14.61380386352539, -2.0127675533294678]
7a8af637-84d3-468a-b2e7-74a99cc98187
graph-gpa-2-0-a-graphical-model-for-multi
2204.06714
null
https://arxiv.org/abs/2204.06714v1
https://arxiv.org/pdf/2204.06714v1.pdf
graph-GPA 2.0: A Graphical Model for Multi-disease Analysis of GWAS Results with Integration of Functional Annotation Data
Genome-wide association studies (GWAS) have successfully identified a large number of genetic variants associated with traits and diseases. However, it still remains challenging to fully understand functional mechanisms underlying many associated variants. This is especially the case when we are interested in variants shared across multiple phenotypes. To address this challenge, we propose graph-GPA 2.0 (GGPA 2.0), a novel statistical framework to integrate GWAS datasets for multiple phenotypes and incorporate functional annotations within a unified framework. We conducted simulation studies to evaluate GGPA 2.0. The results indicate that incorporating functional annotation data using GGPA 2.0 does not only improve detection of disease-associated variants, but also allows to identify more accurate relationships among diseases. We analyzed five autoimmune diseases and five psychiatric disorders with the functional annotations derived from GenoSkyline and GenoSkyline-Plus and the prior disease graph generated by biomedical literature mining. For autoimmune diseases, GGPA 2.0 identified enrichment for blood, especially B cells and regulatory T cells across multiple diseases. Psychiatric disorders were enriched for brain, especially prefrontal cortex and inferior temporal lobe for bipolar disorder (BIP) and schizophrenia (SCZ), respectively. Finally, GGPA 2.0 successfully identified the pleiotropy between BIP and SCZ. These results demonstrate that GGPA 2.0 can be a powerful tool to identify associated variants associated with each phenotype or those shared across multiple phenotypes, while also promoting understanding of functional mechanisms underlying the associated variants.
['Dongjun Chung', 'Hang J. Kim', 'Lang Li', 'Maciej Pietrzak', 'Won Chang', 'Ayse Selen Yilmaz', 'Jin Hyun Nam', 'Qiaolan Deng']
2022-04-14
null
null
null
null
['literature-mining']
['natural-language-processing']
[ 3.86531875e-02 -2.31777653e-01 9.37752724e-02 -2.29761094e-01 -4.75822031e-01 -4.53466505e-01 -9.67062265e-02 3.92460078e-01 1.81580871e-01 1.03238761e+00 -5.97485714e-03 -1.52776912e-01 -6.16358757e-01 -7.77477086e-01 -4.02636617e-01 -6.63151622e-01 -6.04268551e-01 5.45001805e-01 -1.09511301e-01 2.01422453e-01 -6.48567975e-02 1.08823389e-01 -9.61594582e-01 8.71497616e-02 1.54349220e+00 4.25458610e-01 1.07512683e-01 1.17649682e-01 5.63071743e-02 -7.76280940e-01 -4.85943973e-01 -5.74310541e-01 -1.54438525e-01 -4.75655526e-01 -1.86645314e-01 -4.74547505e-01 2.94611808e-02 3.17033291e-01 1.85611725e-01 1.18733943e+00 8.68587673e-01 -3.78122181e-01 5.54791212e-01 -1.21990633e+00 -7.34535515e-01 5.98308623e-01 -9.70210314e-01 9.62107256e-02 1.12015709e-01 4.19436634e-01 1.21299005e+00 -6.42695367e-01 8.37984681e-01 1.47127748e+00 6.92861795e-01 2.66786098e-01 -1.71499562e+00 -6.51580632e-01 -7.04139322e-02 -1.76270634e-01 -1.62370169e+00 -2.25644726e-02 -1.32784974e-02 -7.43363261e-01 9.55837548e-01 4.21676427e-01 1.04021001e+00 8.22528899e-01 3.82092088e-01 1.16894409e-01 9.51316237e-01 3.05875577e-02 2.29204878e-01 -5.29998899e-01 3.94734263e-01 8.08230460e-01 8.39878857e-01 -2.91427076e-01 -3.74370724e-01 -1.01618755e+00 5.16885519e-01 -1.15355119e-01 -3.90353173e-01 5.54900430e-02 -1.31787527e+00 5.82394958e-01 -7.39210993e-02 1.20307729e-01 -4.40763950e-01 8.64762366e-02 4.80113059e-01 -1.68867648e-01 5.43206632e-01 6.36110783e-01 -8.11391234e-01 9.06652678e-03 -6.28253341e-01 2.59050727e-01 2.64861405e-01 5.42791307e-01 5.13013959e-01 1.34181902e-02 -4.89985913e-01 1.25720179e+00 4.18051183e-01 4.92397934e-01 3.37883949e-01 -4.92990971e-01 -7.17788935e-02 8.22662950e-01 -1.53379515e-01 -1.05016518e+00 -8.71898234e-01 -6.34077489e-01 -9.78147745e-01 -1.98707789e-01 1.47469521e-01 -4.41385746e-01 -5.53851783e-01 2.14851618e+00 6.53544486e-01 4.73510951e-01 -2.89366484e-01 8.07809174e-01 6.68434143e-01 1.00310542e-01 3.21223915e-01 -7.93341622e-02 1.95720160e+00 -3.77560377e-01 -5.81751764e-01 5.03092930e-02 5.44203997e-01 -3.57774496e-01 8.02396417e-01 4.29982513e-01 -9.48427558e-01 -2.70178676e-01 -5.22883058e-01 3.64777058e-01 -2.49909073e-01 2.93274969e-01 7.93984711e-01 5.35574317e-01 -9.34322655e-01 3.13841760e-01 -9.03830647e-01 -5.02463281e-01 5.46726763e-01 3.46273839e-01 -4.80957896e-01 -1.40465915e-01 -1.53818774e+00 6.52994514e-01 1.25084147e-01 1.06462799e-01 -6.55853093e-01 -1.25449014e+00 -5.20060897e-01 1.76088989e-01 2.22760096e-01 -1.11054564e+00 2.55017400e-01 -2.29133710e-01 -6.79612517e-01 6.09266043e-01 -3.21269780e-01 -1.92106776e-02 -2.11487442e-01 -2.52892584e-01 -8.66787255e-01 1.23155661e-01 3.67129415e-01 4.60113317e-01 -2.79253814e-02 -3.12722117e-01 -1.34999394e-01 -8.64326000e-01 -4.44211602e-01 -2.49875307e-01 -1.05292294e-02 1.09388895e-01 -3.06186587e-01 -6.99143052e-01 -1.94502827e-02 -8.90199423e-01 -3.71539801e-01 1.45379752e-01 -8.12723279e-01 -1.76724851e-01 1.08522944e-01 -7.78111160e-01 1.05722260e+00 -1.95163155e+00 2.50317097e-01 2.95538217e-01 4.24929023e-01 1.57043919e-01 -3.55987936e-01 3.85685563e-01 -2.50733852e-01 6.15996420e-01 -1.32844821e-01 5.08872986e-01 2.98680514e-02 -3.33848238e-01 3.66633326e-01 5.99628150e-01 8.11555088e-01 1.17450190e+00 -7.36237824e-01 1.43632710e-01 -3.33168536e-01 6.77506030e-01 -9.12185252e-01 -2.04496562e-01 -4.65426058e-01 7.27712214e-01 -8.02658796e-01 7.86959350e-01 9.93728817e-01 -4.06074524e-01 7.34536350e-01 -1.01637512e-01 -6.13964126e-02 1.62448213e-01 -7.69351542e-01 1.72130024e+00 7.65651688e-02 9.97710600e-02 2.76024848e-01 -8.29374194e-01 1.02704632e+00 2.16196105e-01 4.00964558e-01 -4.30798680e-01 -1.93768084e-01 6.39833584e-02 7.49240696e-01 -7.91059494e-01 -4.33150828e-01 9.16603133e-02 -1.17181495e-01 2.20505998e-01 -1.85640827e-01 5.64661503e-01 1.30773157e-01 7.76517838e-02 1.53420067e+00 -4.74248780e-03 5.21465361e-01 -5.56864858e-01 2.16079384e-01 1.38512582e-01 1.19857728e+00 3.34943771e-01 7.94254243e-02 4.31264341e-01 1.18653655e+00 -4.80560251e-02 -7.92677104e-01 -1.04663765e+00 -4.94081587e-01 6.06369615e-01 -3.84527713e-01 -5.46437323e-01 -2.79390246e-01 -3.89834046e-01 4.39555615e-01 3.74222100e-01 -5.51262081e-01 -2.44030997e-01 2.45150179e-01 -1.70275307e+00 1.09415436e+00 1.60211712e-01 2.85636991e-01 -2.43022099e-01 2.01965822e-03 6.02360405e-02 7.09189195e-03 -7.34168649e-01 -3.88838440e-01 -1.60329625e-01 -6.21671021e-01 -1.40914011e+00 -5.51107347e-01 -1.94525003e-01 8.30708325e-01 -2.17909813e-01 8.56636167e-01 -9.49480757e-02 -9.96002376e-01 -8.23776498e-02 -1.70413271e-01 -3.05850834e-01 -2.28259377e-02 -3.50856930e-01 1.86992779e-01 -1.86239094e-01 3.84877980e-01 -7.07520843e-01 -7.05014825e-01 3.62656325e-01 -8.27725053e-01 -1.11473560e-01 7.59157658e-01 8.57137918e-01 5.43020904e-01 2.21180961e-01 1.24583888e+00 -8.39234710e-01 6.86992288e-01 -1.02120245e+00 -7.62564957e-01 3.04747909e-01 -5.61481118e-01 -1.96835369e-01 1.27821863e-01 -2.21696317e-01 -9.68695998e-01 -1.82955116e-01 -3.54783118e-01 2.23114610e-01 -2.10334733e-01 1.15590465e+00 -7.62926757e-01 4.12250638e-01 1.42914340e-01 -1.52071118e-01 1.87500492e-01 -6.69434845e-01 5.02211647e-03 3.18237245e-01 -1.88851785e-02 -4.67011839e-01 -2.49803662e-02 1.12461023e-01 3.54471356e-01 -7.70564675e-01 -3.62495631e-01 -3.63124549e-01 -4.19857115e-01 3.42895031e-01 1.09987235e+00 -1.01527750e+00 -1.05403674e+00 2.08484516e-01 -7.23322213e-01 -1.18747450e-01 7.69516408e-01 7.45400310e-01 3.03886771e-01 1.92109153e-01 -4.48906869e-01 -4.81027901e-01 -5.52161753e-01 -1.20545256e+00 9.26022887e-01 4.02108580e-02 -6.46587431e-01 -1.12455463e+00 6.21551335e-01 2.96689242e-01 1.13675117e-01 4.38924313e-01 1.63327587e+00 -9.87212718e-01 1.45396432e-02 1.11325197e-01 -3.63976151e-01 -1.28151953e-01 3.80735248e-01 2.16428399e-01 -5.30597866e-01 2.40489771e-03 -8.52359533e-01 5.95880731e-04 5.55501163e-01 5.56252122e-01 1.00595653e+00 -1.37480885e-01 -6.66168332e-01 5.09129524e-01 1.15049148e+00 1.91864446e-01 6.64132476e-01 -2.93401569e-01 4.87475425e-01 6.14295840e-01 6.19405270e-01 4.41410214e-01 7.68228993e-02 9.84913707e-01 1.34572297e-01 -2.31276937e-02 2.57036686e-01 2.30245553e-02 2.34203488e-01 3.19578856e-01 4.33651879e-02 -2.93619335e-01 -1.09261405e+00 5.07364452e-01 -1.66688502e+00 -5.37224352e-01 -8.33939135e-01 2.24458385e+00 9.90649879e-01 -3.48358124e-01 2.09066018e-01 -6.67650580e-01 9.44000781e-01 -6.27708614e-01 -8.76323164e-01 -1.83731332e-01 -3.56766045e-01 4.30931956e-01 -1.16169296e-01 -2.34253742e-02 -5.20357013e-01 6.04679406e-01 6.41889524e+00 3.94165963e-01 -7.68940687e-01 1.15192853e-01 6.21025801e-01 -2.20708489e-01 -4.40716654e-01 5.78150675e-02 -7.52150893e-01 5.99123776e-01 9.86235738e-01 -1.10241078e-01 9.84673649e-02 3.75671297e-01 6.73060954e-01 -1.71564490e-01 -6.85259461e-01 3.78764242e-01 -4.20617849e-01 -1.30306470e+00 -1.47823974e-01 4.53076452e-01 6.93559825e-01 3.34615111e-02 -1.56633049e-01 -7.95489997e-02 3.40872198e-01 -9.82520640e-01 -3.79667759e-01 9.27112281e-01 9.54043031e-01 -5.98463833e-01 7.58324802e-01 -2.41699636e-01 -7.98631728e-01 4.57371682e-01 -6.70543253e-01 2.09708065e-01 4.33731079e-02 1.67672849e+00 -1.18822312e+00 6.86154008e-01 5.28011143e-01 6.12772822e-01 -4.00300682e-01 1.24976611e+00 -2.82572716e-01 7.70912588e-01 1.37085587e-01 1.90540135e-01 -3.25044125e-01 -4.53674406e-01 6.78607523e-01 1.04966903e+00 7.46758938e-01 1.52616918e-01 7.70621970e-02 1.52592921e+00 2.45175496e-01 1.36986017e-01 -3.06975961e-01 -5.33518434e-01 5.90901911e-01 1.70225739e+00 -6.61924303e-01 -2.07015201e-01 -3.63552600e-01 7.04962671e-01 2.53837168e-01 2.47018710e-01 -7.95559466e-01 -3.07288021e-01 1.40572858e+00 -6.78463504e-02 -3.75385173e-02 1.28784049e-02 8.22752044e-02 -1.07339466e+00 -4.61553514e-01 -8.00632656e-01 5.27248800e-01 -8.39074016e-01 -1.61250579e+00 1.09297343e-01 -2.01634943e-01 -6.77676320e-01 1.53905958e-01 -3.54044586e-01 -6.20695174e-01 1.31854904e+00 -1.06838834e+00 -6.97888374e-01 -7.65088499e-02 1.11279048e-01 -1.19607344e-01 1.15237519e-01 1.15645432e+00 3.21804523e-01 -1.34973919e+00 2.34124616e-01 1.87933192e-01 -2.70275205e-01 1.03148532e+00 -9.74671125e-01 3.32532972e-01 3.59531254e-01 -5.07390678e-01 1.27929890e+00 4.46533114e-01 -1.35753155e+00 -1.30383360e+00 -1.39448512e+00 4.30309236e-01 1.14828750e-01 8.43510747e-01 -4.02043045e-01 -1.00810611e+00 5.97036242e-01 -7.40823820e-02 -2.20359832e-01 1.38213575e+00 4.44568425e-01 -3.16162705e-01 8.91659260e-02 -1.22650564e+00 5.00185251e-01 1.12723255e+00 -5.51395416e-02 -2.49995261e-01 2.61989504e-01 6.92463815e-01 2.09620088e-01 -1.42678916e+00 3.56337517e-01 7.53201246e-01 -5.85401893e-01 9.71354306e-01 -8.28509033e-01 2.37809718e-01 -4.68723297e-01 5.85003011e-02 -1.85959518e+00 -5.80994904e-01 -2.54103750e-01 3.62663776e-01 1.42042375e+00 5.43553710e-01 -9.14909363e-01 9.14492607e-02 5.13132393e-01 -3.01346570e-01 -3.85110229e-01 -8.19808245e-01 -5.86239398e-01 -1.58834308e-01 -1.95926562e-01 7.41931617e-01 1.13725841e+00 2.83694923e-01 3.22018385e-01 -9.64414254e-02 5.78031719e-01 4.99630809e-01 1.79630816e-02 4.67567146e-01 -1.52998233e+00 -3.88233900e-01 -4.54964250e-01 -4.51380283e-01 1.24716476e-01 3.11584994e-02 -1.13671887e+00 -3.49079758e-01 -1.63164639e+00 6.46207631e-01 -6.26220524e-01 -3.66291553e-01 7.81372309e-01 -5.46776235e-01 2.52849668e-01 -7.41812825e-01 -4.14020747e-01 -1.51474848e-01 2.90639222e-01 1.10606527e+00 -4.94940914e-02 -1.79229349e-01 -2.46736199e-01 -1.12928343e+00 5.64211130e-01 1.06517982e+00 -2.81880915e-01 -4.81958330e-01 -1.32691234e-01 6.03578269e-01 -7.73851126e-02 6.76187754e-01 -3.61277282e-01 -5.31983018e-01 -2.93909907e-01 5.73877692e-01 -4.25153047e-01 1.72902822e-01 -2.76596785e-01 8.23513508e-01 8.16871822e-01 -1.92689538e-01 1.14939764e-01 5.70062280e-01 5.81268907e-01 1.96459919e-01 5.07333092e-02 4.84152615e-01 3.14124301e-02 -2.39402518e-01 2.08831385e-01 -4.24472153e-01 1.11094620e-02 9.01599705e-01 4.50642675e-01 -8.09850872e-01 9.57041457e-02 -1.18358159e+00 3.23127776e-01 6.78564236e-02 1.60437912e-01 6.37624085e-01 -1.07309914e+00 -1.01682758e+00 2.47688293e-01 2.49670982e-01 -5.77768862e-01 8.23446572e-01 1.49281776e+00 -1.39285684e-01 6.11501634e-01 -3.63756001e-01 -4.61716413e-01 -1.49832189e+00 5.30834317e-01 -4.87520434e-02 -6.30537719e-02 -4.02272671e-01 8.05270672e-01 6.02122724e-01 -4.75735784e-01 -2.33041480e-01 -1.45772368e-01 -1.80940583e-01 1.23634338e-01 6.54266953e-01 4.77470815e-01 -8.09429288e-02 -1.03333108e-01 -6.93344653e-01 3.20012540e-01 1.69150025e-01 2.33159781e-01 1.55202329e+00 -4.77290787e-02 -9.35062230e-01 2.91053414e-01 8.26173365e-01 5.08920372e-01 -3.12840939e-01 2.52648950e-01 7.42937773e-02 -3.90102804e-01 -2.87245929e-01 -1.30456936e+00 -1.17852068e+00 6.19632602e-01 4.18640703e-01 -1.66957989e-01 9.22487617e-01 1.89124346e-02 2.18047842e-01 -2.36946911e-01 5.23825645e-01 -3.82589400e-01 -4.92023766e-01 3.65575813e-02 8.26390624e-01 -6.00384891e-01 1.17325164e-01 -7.42925227e-01 -4.46869969e-01 7.71408737e-01 8.10983419e-01 1.13784194e-01 4.22168821e-01 1.66800216e-01 -5.32995701e-01 -6.19987309e-01 -1.16894948e+00 7.37397149e-02 5.55251777e-01 8.02327752e-01 7.95795977e-01 2.45522425e-01 -8.44186842e-01 1.15932035e+00 2.21993163e-01 -1.38512895e-01 1.90018743e-01 2.64244080e-01 -1.11878268e-01 -1.42909169e+00 -4.67716485e-01 8.28262746e-01 -8.33158314e-01 -5.96337080e-01 -6.63945019e-01 6.95592403e-01 3.99716079e-01 7.21734643e-01 1.18538342e-01 -1.31980672e-01 -4.53509241e-02 3.11858177e-01 1.53655335e-01 -8.34803760e-01 -3.78391236e-01 8.38016331e-01 6.23760521e-01 -5.06431401e-01 -1.64263189e-01 -6.99432194e-01 -1.33403659e+00 7.43912533e-02 -4.41215307e-01 2.82710522e-01 4.03353333e-01 4.77704644e-01 1.23366296e+00 1.12723911e+00 -6.02699444e-03 1.20005965e-01 2.08595410e-01 -9.85766053e-01 -1.05547786e+00 -1.60024002e-01 -2.82211035e-01 -8.21697354e-01 1.91563740e-01 -2.53742307e-01]
[6.492081165313721, 5.546047210693359]
900c7366-74a8-40b4-8e21-29fdc750fa5f
energy-based-sliced-wasserstein-distance
2304.13586
null
https://arxiv.org/abs/2304.13586v1
https://arxiv.org/pdf/2304.13586v1.pdf
Energy-Based Sliced Wasserstein Distance
The sliced Wasserstein (SW) distance has been widely recognized as a statistically effective and computationally efficient metric between two probability measures. A key component of the SW distance is the slicing distribution. There are two existing approaches for choosing this distribution. The first approach is using a fixed prior distribution. The second approach is optimizing for the best distribution which belongs to a parametric family of distributions and can maximize the expected distance. However, both approaches have their limitations. A fixed prior distribution is non-informative in terms of highlighting projecting directions that can discriminate two general probability measures. Doing optimization for the best distribution is often expensive and unstable. Moreover, designing the parametric family of the candidate distribution could be easily misspecified. To address the issues, we propose to design the slicing distribution as an energy-based distribution that is parameter-free and has the density proportional to an energy function of the projected one-dimensional Wasserstein distance. We then derive a novel sliced Wasserstein metric, energy-based sliced Waserstein (EBSW) distance, and investigate its topological, statistical, and computational properties via importance sampling, sampling importance resampling, and Markov Chain methods. Finally, we conduct experiments on point-cloud gradient flow, color transfer, and point-cloud reconstruction to show the favorable performance of the EBSW.
['Nhat Ho', 'Khai Nguyen']
2023-04-26
null
null
null
null
['point-cloud-reconstruction']
['computer-vision']
[-9.67149362e-02 -4.14345354e-01 -7.78750405e-02 -2.23227188e-01 -6.06571376e-01 -3.73993099e-01 3.01595151e-01 -2.47701127e-02 -3.90401393e-01 6.86719000e-01 -1.11139454e-01 -2.81404704e-01 -5.43783665e-01 -9.74558890e-01 -3.37801874e-01 -9.14356768e-01 9.07654539e-02 4.01678383e-01 5.74911118e-01 1.28630072e-01 6.77909970e-01 8.26782107e-01 -1.47991037e+00 -4.45571840e-01 1.41524673e+00 9.53190863e-01 2.84684747e-01 3.68792593e-01 -3.18926096e-01 6.48196042e-02 -3.49827528e-01 -3.50176245e-01 2.72055149e-01 -5.62959552e-01 -6.67144418e-01 -1.23904511e-01 1.01302534e-01 -2.24854052e-01 1.70586199e-01 1.35383010e+00 3.42568576e-01 3.22475702e-01 9.95927274e-01 -1.42677367e+00 -6.02401972e-01 7.29897618e-02 -6.88238025e-01 1.49809420e-01 5.98483160e-02 -1.05822906e-01 7.95019150e-01 -9.21296120e-01 5.51089644e-01 1.23146689e+00 5.26348889e-01 2.87854880e-01 -1.20897698e+00 -4.08424318e-01 -2.33745605e-01 2.45572686e-01 -1.64830315e+00 3.14132124e-01 9.93738174e-01 -6.05335772e-01 1.35811836e-01 3.61670136e-01 7.33325779e-01 3.71054322e-01 2.01592058e-01 5.74360907e-01 1.16915154e+00 -4.84144509e-01 4.42860782e-01 2.44031921e-01 -3.41908708e-02 7.90177763e-01 4.16331619e-01 1.05325826e-01 -2.15391770e-01 -4.17939365e-01 8.82056534e-01 6.06637262e-02 -5.17108381e-01 -8.78083885e-01 -1.06670570e+00 8.68875325e-01 2.79460758e-01 3.38412136e-01 -3.80390659e-02 -1.22548200e-01 -2.22752038e-02 -8.31815228e-02 4.69076604e-01 3.40852231e-01 -4.85525131e-02 -2.18787178e-01 -9.28705335e-01 2.48312533e-01 4.93226349e-01 7.25280166e-01 1.05934930e+00 -3.34100813e-01 -2.38561541e-01 6.22048378e-01 5.89287877e-01 5.97300589e-01 1.62141994e-01 -9.24078226e-01 2.55601078e-01 5.08343101e-01 3.99910033e-01 -1.38421404e+00 -1.34919897e-01 -5.05915172e-02 -6.73396885e-01 3.96499932e-01 6.67846918e-01 1.46407168e-02 -4.83679980e-01 1.53773499e+00 6.18461728e-01 -7.78517127e-02 -3.52978349e-01 1.04538488e+00 2.02160925e-01 6.46481752e-01 -1.78609952e-01 -3.05024356e-01 9.09628987e-01 -6.75683022e-01 -5.47595739e-01 4.96501893e-01 2.58478403e-01 -8.27311814e-01 1.31562400e+00 1.49914041e-01 -7.72107124e-01 -1.61873117e-01 -1.08726311e+00 2.56896585e-01 -5.94095401e-02 7.99995139e-02 2.18858257e-01 6.97341859e-01 -8.54239583e-01 1.06521153e+00 -8.61637592e-01 -2.27868900e-01 1.51104167e-01 -1.77042738e-01 2.49949172e-02 6.27342537e-02 -9.94146585e-01 8.86438131e-01 2.25531563e-01 -5.47478423e-02 -3.38981658e-01 -6.73803568e-01 -4.38041568e-01 1.32597238e-01 -1.31595805e-01 -5.17293215e-01 5.66163719e-01 -5.49048841e-01 -1.70435333e+00 4.11300272e-01 -1.31158950e-03 1.30257839e-02 6.84602678e-01 1.37867689e-01 -1.61773056e-01 4.06881571e-01 6.60748482e-02 3.02982837e-01 8.76552939e-01 -1.16159034e+00 -2.68687636e-01 -2.66098082e-01 -2.45903671e-01 2.40009263e-01 -4.46050078e-01 -1.05957352e-01 -1.22553274e-01 -5.51766694e-01 3.94838303e-01 -7.65127003e-01 -1.29258871e-01 5.57022095e-01 -4.60815817e-01 -2.22307265e-01 7.54088581e-01 -4.83479768e-01 1.39015901e+00 -2.35867929e+00 7.73728341e-02 4.70975965e-01 3.85733359e-02 9.22294855e-02 1.41242996e-01 3.93487155e-01 1.90244034e-01 2.35733405e-01 -5.17847180e-01 2.00717926e-01 -3.96330021e-02 -1.53075859e-01 -2.80288029e-02 6.90521181e-01 4.44982015e-02 1.58956423e-01 -1.17391443e+00 -7.40412772e-01 3.65232348e-01 6.27098262e-01 -4.17533189e-01 1.29358187e-01 1.70684591e-01 3.69573236e-01 -6.56612933e-01 3.78965229e-01 1.10736024e+00 -1.01628199e-01 -1.55312121e-02 -4.52387810e-01 -3.80556256e-01 -1.92843810e-01 -1.41907811e+00 1.05756855e+00 -1.10127844e-01 4.85639632e-01 -1.43969834e-01 -8.25903356e-01 1.28079772e+00 -1.74616218e-01 6.78693771e-01 -1.52282238e-01 1.24156810e-01 4.26574349e-01 -2.30271235e-01 -3.32965165e-01 4.91440207e-01 -5.27436316e-01 3.91094446e-01 3.64259452e-01 -3.08582515e-01 -2.57788002e-01 1.47400334e-01 5.69051579e-02 8.37439299e-01 1.48252591e-01 1.94931060e-01 -8.15642893e-01 5.62715232e-01 -1.97522640e-01 6.14422023e-01 3.23583454e-01 -3.87198031e-01 8.95250380e-01 6.04792416e-01 -3.09612483e-01 -9.21423018e-01 -1.34741247e+00 -5.67552030e-01 3.00114006e-01 7.51186609e-01 -2.71035761e-01 -6.87545717e-01 -7.11737692e-01 1.16248131e-02 8.13941181e-01 -3.18502545e-01 -1.52289942e-01 -3.72667879e-01 -7.03941047e-01 2.06788823e-01 2.51462489e-01 5.92237771e-01 -5.10804832e-01 -8.28115463e-01 -1.76153146e-03 -3.08879644e-01 -4.89725262e-01 -7.08112538e-01 -2.80129552e-01 -9.86824691e-01 -1.30659461e+00 -9.97095466e-01 -4.44431782e-01 9.57180858e-01 3.28512371e-01 7.90280044e-01 -1.29821315e-01 -1.95816293e-01 3.24433178e-01 -3.51209193e-01 6.19827732e-02 -1.86517239e-01 -3.12834203e-01 -7.07666054e-02 2.82709062e-01 1.83996379e-01 -4.20889258e-01 -9.10724163e-01 6.36524975e-01 -8.15242350e-01 -1.33497909e-01 3.36623907e-01 7.15379596e-01 8.91517699e-01 2.67912298e-01 1.83525309e-01 -2.75561810e-01 6.14190698e-01 -3.16215575e-01 -7.20758319e-01 3.74872684e-01 -6.37113512e-01 3.59624475e-01 5.01116395e-01 -3.96090329e-01 -9.65158463e-01 -8.39368403e-02 7.71443471e-02 -5.31684279e-01 1.58156395e-01 2.61744678e-01 -6.85287118e-02 -2.53478646e-01 3.80919307e-01 1.02010250e-01 2.05508456e-01 -4.05106515e-01 2.88886160e-01 4.44900632e-01 1.70134734e-02 -8.29849839e-01 7.67239571e-01 7.30652034e-01 2.76393503e-01 -9.80509460e-01 -5.96073925e-01 -4.65826690e-01 -4.78642523e-01 -4.46106195e-01 7.60675073e-01 -1.20340651e-02 -6.57339811e-01 4.08185601e-01 -1.11206257e+00 1.58995166e-02 -2.92668790e-01 7.07791924e-01 -4.59481567e-01 8.10176969e-01 -2.38863066e-01 -9.52634454e-01 -2.20978245e-01 -1.22854662e+00 8.76591504e-01 2.26103947e-01 -3.78546603e-02 -1.08953452e+00 2.13978231e-01 -2.90784568e-01 5.58845341e-01 5.80726147e-01 1.10440314e+00 -3.59857753e-02 -5.59907615e-01 -1.19870938e-01 -3.74562621e-01 3.67948771e-01 1.27356932e-01 6.02500796e-01 -2.54424065e-01 -2.53771275e-01 1.24451198e-01 1.43677831e-01 7.14408278e-01 5.54423749e-01 1.29309845e+00 -2.22082704e-01 -3.38185579e-01 5.22518158e-01 1.72303176e+00 2.55916595e-01 6.64993882e-01 2.50909358e-01 4.58841383e-01 5.50905406e-01 8.16407025e-01 4.63438839e-01 2.59556234e-01 5.74498832e-01 2.94972122e-01 1.59951925e-01 9.40097719e-02 -3.62638861e-01 2.07126424e-01 8.24581027e-01 -1.71495304e-01 3.14450078e-03 -7.45264828e-01 4.66505289e-01 -1.72761846e+00 -9.84581411e-01 -1.16642945e-01 2.66194987e+00 6.47781193e-01 4.95400950e-02 2.05479950e-01 1.50763899e-01 1.14869285e+00 -4.83112689e-03 -4.44623202e-01 -1.92621469e-01 3.82181965e-02 -7.44961575e-02 5.51409960e-01 6.60299420e-01 -9.00236309e-01 2.59051144e-01 5.97725010e+00 1.12978303e+00 -1.09996200e+00 3.07104718e-02 3.59919846e-01 1.57607317e-01 -7.21927404e-01 2.81647950e-01 -7.22456753e-01 9.19169962e-01 3.57691377e-01 -2.97044933e-01 1.16330504e-01 8.51677775e-01 1.12831444e-01 -3.59691381e-01 -7.23326862e-01 1.01111019e+00 -2.46773735e-01 -1.16642845e+00 3.78765799e-02 1.48242131e-01 6.17708266e-01 -2.72981644e-01 -1.00180216e-01 -2.25962922e-01 -5.51303141e-02 -6.66275561e-01 7.64789879e-01 7.38028646e-01 7.51746953e-01 -8.16054404e-01 5.68904459e-01 2.63994843e-01 -1.28116608e+00 2.11003095e-01 -5.58081746e-01 3.68961185e-01 3.01889062e-01 1.01859963e+00 -4.00249004e-01 3.25901419e-01 7.70135999e-01 4.97150630e-01 -2.14661881e-01 1.51212859e+00 -6.62273690e-02 3.36827666e-01 -5.65991044e-01 -4.13332999e-01 6.79850280e-02 -8.50902438e-01 9.17417467e-01 8.78832042e-01 8.05298507e-01 -1.42019510e-01 1.51641816e-01 1.05295146e+00 3.29347432e-01 2.87984371e-01 -4.26949680e-01 1.73345730e-01 8.30546677e-01 1.14202094e+00 -1.08786678e+00 -7.03879399e-03 -1.31441817e-01 7.01896131e-01 1.56431779e-01 3.00354987e-01 -8.57635260e-01 -9.08114851e-01 4.97792423e-01 3.29364747e-01 3.14400017e-01 -3.53768080e-01 -2.81136811e-01 -1.11772311e+00 2.27691188e-01 -1.36701092e-01 2.69100100e-01 -4.81339186e-01 -1.46780097e+00 4.74255383e-01 3.87198269e-01 -1.58213806e+00 4.17810619e-01 -6.64161742e-01 -7.55515277e-01 8.28371108e-01 -1.47525764e+00 -4.71235245e-01 -3.54279637e-01 4.51229990e-01 4.42977399e-02 2.15553567e-01 4.22629237e-01 4.11852032e-01 -3.80292594e-01 3.30273449e-01 4.30318832e-01 -2.31406212e-01 5.62025547e-01 -1.32515073e+00 -3.20058554e-01 8.07060599e-01 -3.13290387e-01 5.70374429e-01 7.63955712e-01 -5.63252747e-01 -1.11973679e+00 -8.58661175e-01 6.07514799e-01 -5.57751507e-02 6.02188289e-01 5.22444807e-02 -1.02583444e+00 2.39851046e-02 -2.51678288e-01 -3.27726677e-02 5.03106773e-01 -2.92739987e-01 -6.83261305e-02 -2.60561854e-01 -1.53717566e+00 6.30289853e-01 7.12544084e-01 -7.85614997e-02 -3.17955405e-01 2.85203874e-01 3.03166509e-01 -6.44994900e-02 -1.05740428e+00 5.57942569e-01 5.61569393e-01 -1.26083601e+00 8.83742392e-01 7.92230591e-02 2.37693742e-01 -6.69723094e-01 -1.00999765e-01 -1.38256538e+00 -2.95134991e-01 -6.25052810e-01 1.40351743e-01 1.23338187e+00 1.33298025e-01 -8.53509128e-01 6.98305845e-01 4.89377886e-01 1.42451465e-01 -1.22148037e+00 -1.00475931e+00 -1.16365516e+00 1.64685965e-01 -1.33248001e-01 6.98072493e-01 6.91111863e-01 2.25268509e-02 -1.01384051e-01 -7.65372515e-02 7.56113231e-02 1.00714731e+00 2.76111990e-01 4.33103472e-01 -1.36594987e+00 -9.48099270e-02 -6.17375731e-01 -4.46890116e-01 -1.08936417e+00 -1.94294840e-01 -7.67224908e-01 1.22292839e-01 -1.65567994e+00 1.40632913e-01 -7.97158837e-01 -1.90137953e-01 -1.43941075e-01 -1.32145137e-01 -2.20753357e-01 -6.75406456e-02 5.44819593e-01 -2.44116768e-01 1.00020790e+00 1.52121103e+00 1.59057677e-01 -1.35780513e-01 2.90186275e-02 -3.62720996e-01 7.85774887e-01 8.15147579e-01 -4.33464259e-01 -4.84425992e-01 -1.59296796e-01 1.79091603e-01 2.54483111e-02 2.27143645e-01 -8.09224844e-01 2.78919879e-02 -5.33627272e-01 1.42557714e-02 -8.94913912e-01 1.42662928e-01 -7.82910585e-01 -2.98930500e-02 3.42131376e-01 -1.24089858e-02 -5.29508702e-02 -3.89634043e-01 6.79605663e-01 -2.28381485e-01 -5.26904345e-01 1.29062259e+00 9.29813012e-02 -4.09573197e-01 3.64395350e-01 -1.56048715e-01 1.35901809e-01 1.16138470e+00 -4.41187322e-01 -2.03406557e-01 -2.16725960e-01 -2.14976370e-01 -2.11763382e-02 7.00713098e-01 -1.86841823e-02 8.33200276e-01 -1.63572323e+00 -4.72604036e-01 1.39719129e-01 -5.81572242e-02 -1.29270181e-01 4.78593856e-02 1.00594616e+00 -8.94665837e-01 -4.13244823e-03 -1.34433597e-01 -6.98258221e-01 -8.16838682e-01 4.80757743e-01 3.81106704e-01 -9.38671604e-02 -6.02878332e-01 6.85266972e-01 -4.91975844e-02 -4.41903323e-01 -4.76202928e-02 -3.58456135e-01 -5.76326475e-02 -1.51253954e-01 3.82073671e-01 9.92378533e-01 -2.32370630e-01 -4.67535317e-01 -5.80460131e-01 9.48699355e-01 4.84442741e-01 -1.84935600e-01 9.57103193e-01 -2.08663747e-01 -3.08585256e-01 3.01809520e-01 1.40654469e+00 2.83879810e-03 -1.28809047e+00 9.79866832e-02 -3.72071378e-03 -9.69451904e-01 5.96985817e-02 -2.59628415e-01 -1.06775224e+00 9.27476943e-01 5.78824818e-01 6.61419511e-01 9.44420338e-01 -1.39578775e-01 7.79201508e-01 -1.12449728e-01 4.08719659e-01 -1.11553514e+00 1.59055546e-01 2.67365485e-01 9.10630465e-01 -9.17155743e-01 1.42182052e-01 -6.08108759e-01 -4.60352033e-01 1.40288734e+00 5.71412146e-01 -2.86126345e-01 1.11443257e+00 3.48460823e-02 -2.83308119e-01 -1.29386768e-01 1.75292790e-02 -1.91634092e-02 3.64867538e-01 4.52860177e-01 2.99178302e-01 2.15827897e-01 -7.59075761e-01 4.14114930e-02 -1.50274724e-01 -1.93890363e-01 3.76013845e-01 7.13750482e-01 -5.78882754e-01 -9.36000466e-01 -3.58536243e-01 3.69159669e-01 -1.64414734e-01 1.51547745e-01 1.29239306e-01 5.10896504e-01 -3.53098847e-02 7.41796136e-01 6.89586550e-02 -2.12236330e-01 2.59477019e-01 -1.41885713e-01 5.74680209e-01 -2.21312046e-01 3.31007481e-01 -7.13418946e-02 -3.25629354e-01 -3.82065177e-01 -4.59661722e-01 -6.60758197e-01 -1.10325909e+00 -3.17119867e-01 -7.07790315e-01 4.92658794e-01 6.24249160e-01 8.32270205e-01 3.53763998e-01 -3.21337320e-02 9.90755141e-01 -7.76159942e-01 -7.42905140e-01 -5.58878303e-01 -8.21981788e-01 2.26118937e-01 1.10379875e-01 -1.06339455e+00 -7.20021546e-01 -4.12081212e-01]
[7.413590908050537, 3.873350143432617]
d09db522-f1ec-40e9-94f9-41355511d51d
multi-task-cross-attention-network-in-facial
2207.10293
null
https://arxiv.org/abs/2207.10293v2
https://arxiv.org/pdf/2207.10293v2.pdf
Affective Behavior Analysis using Action Unit Relation Graph and Multi-task Cross Attention
Facial behavior analysis is a broad topic with various categories such as facial emotion recognition, age, and gender recognition. Many studies focus on individual tasks while the multi-task learning approach is still an open research issue and requires more research. In this paper, we present our solution and experiment result for the Multi-Task Learning challenge of the Affective Behavior Analysis in-the-wild competition. The challenge is a combination of three tasks: action unit detection, facial expression recognition, and valance-arousal estimation. To address this challenge, we introduce a cross-attentive module to improve multi-task learning performance. Additionally, a facial graph is applied to capture the association among action units. As a result, we achieve the evaluation measure of 128.8 on the validation data provided by the organizers, which outperforms the baseline result of 30.
['Hyung-Jeong Yang', 'Soo-Huyng Kim', 'Guee-Sang Lee', 'Ngoc-Huynh Ho', 'Sudarshan Pant', 'Dang-Khanh Nguyen']
2022-07-21
null
null
null
null
['facial-emotion-recognition', 'facial-expression-recognition', 'action-unit-detection']
['computer-vision', 'computer-vision', 'computer-vision']
[ 2.52305511e-02 -1.64520293e-01 -3.86385769e-01 -7.77494192e-01 -8.40477824e-01 -1.57977477e-01 3.12368959e-01 -1.53002888e-01 -4.03188854e-01 4.25071388e-01 8.72320235e-02 6.05570436e-01 4.07151520e-01 -1.60391107e-01 -3.42411786e-01 -8.17510366e-01 -2.00912744e-01 -9.17368159e-02 -2.00307682e-01 5.27664945e-02 -5.72206862e-02 1.75036177e-01 -1.59086108e+00 6.71480954e-01 4.80191618e-01 1.59817374e+00 -6.79716945e-01 4.46908981e-01 1.61494300e-01 8.44394505e-01 -5.65772653e-01 -6.12692058e-01 -1.25582427e-01 -3.77885163e-01 -5.71559012e-01 9.23460349e-02 7.23028421e-01 -2.11422980e-01 1.71229199e-01 9.18300092e-01 7.90214956e-01 1.37928322e-01 6.85497344e-01 -1.80274284e+00 -1.41614497e-01 1.84240833e-01 -1.17174637e+00 5.45719303e-02 2.37126112e-01 -2.44172409e-01 8.83166313e-01 -8.75783622e-01 2.42888346e-01 1.44898760e+00 3.87553573e-01 8.84930193e-01 -8.04227948e-01 -1.26214314e+00 2.92288512e-01 3.62552047e-01 -1.46546519e+00 -7.09165633e-01 7.67175972e-01 -3.52026671e-01 7.65337527e-01 -7.13452790e-03 3.66846263e-01 1.46878302e+00 2.41969571e-01 9.18568790e-01 1.28658116e+00 -1.58995524e-01 -2.16720086e-02 1.81259379e-01 1.43313944e-01 1.00469768e+00 -3.62892687e-01 -2.67923802e-01 -8.70940447e-01 -1.20669484e-01 4.12453920e-01 -7.91308954e-02 2.87475437e-01 -1.14973962e-01 -3.81022513e-01 7.78751194e-01 3.47492509e-02 2.50256866e-01 -3.21063936e-01 1.80835187e-01 8.28918278e-01 2.69991904e-01 8.58587801e-01 -1.24226384e-01 -4.10136372e-01 -3.21961313e-01 -7.12683499e-01 2.26048231e-01 5.60594320e-01 3.93901348e-01 8.66577685e-01 6.61258772e-02 -3.01278293e-01 1.20529151e+00 3.80415320e-01 3.46230090e-01 5.34731030e-01 -8.27052534e-01 6.51741447e-03 7.51973569e-01 -4.06549275e-01 -1.03217745e+00 -6.32808685e-01 9.37232189e-03 -7.75398731e-01 3.40847701e-01 1.72032759e-01 -6.57160699e-01 -5.88926077e-01 2.08755898e+00 4.16702271e-01 3.50969225e-01 -1.69262350e-01 8.34557891e-01 9.51299012e-01 5.20335972e-01 5.94479501e-01 -4.50494975e-01 1.69942141e+00 -1.27031577e+00 -8.93850207e-01 -3.52109194e-01 7.31373787e-01 -5.09465873e-01 8.06661010e-01 5.84275723e-01 -7.69342303e-01 -5.73423266e-01 -7.47322619e-01 9.87200513e-02 -3.42883140e-01 4.81283426e-01 9.40080643e-01 8.56736302e-01 -8.69093120e-01 1.10789984e-01 -6.71080172e-01 -3.09818387e-01 6.62196398e-01 3.37208956e-01 -6.85862601e-01 2.04609528e-01 -1.06808496e+00 7.82494187e-01 -1.15117915e-02 -1.04548305e-01 -8.51451457e-01 -5.31984568e-01 -8.40434253e-01 -1.58197209e-02 2.86992103e-01 -2.04922304e-01 1.22412300e+00 -1.59523475e+00 -1.61362076e+00 1.32401490e+00 -1.74536407e-01 -2.58894749e-02 1.43121555e-01 -3.91525447e-01 -6.38754547e-01 3.29610938e-03 -1.40602037e-01 6.14373088e-01 1.11885238e+00 -8.00282657e-01 -6.57500327e-01 -9.04894829e-01 -1.25479609e-01 3.05252790e-01 -8.20387244e-01 6.51499629e-01 -6.19040668e-01 -3.90352070e-01 -5.49267948e-01 -9.70812917e-01 -5.04181953e-03 -1.78554520e-01 -1.33039402e-02 -6.38937473e-01 8.15575957e-01 -4.94446039e-01 1.42223907e+00 -2.61781454e+00 9.50437263e-02 1.13620691e-01 2.43980452e-01 3.08551304e-02 -2.28039131e-01 -2.14083880e-01 -2.96726853e-01 3.51417484e-03 4.09271151e-01 -8.86421800e-01 -1.38475075e-02 -3.01446021e-01 2.37601697e-01 4.48377132e-01 1.76692188e-01 6.53038681e-01 -3.78587127e-01 -7.29875565e-01 -2.05284119e-01 3.08929354e-01 -4.29835618e-01 4.01897997e-01 1.60406500e-01 1.76826164e-01 -5.53400457e-01 1.18842995e+00 5.55689573e-01 -7.68606924e-03 2.12067783e-01 -3.83920193e-01 2.38664120e-01 -5.29153228e-01 -9.77859139e-01 1.74935889e+00 -4.32182074e-01 3.35110813e-01 3.45868498e-01 -8.45382214e-01 1.11604631e+00 3.61498982e-01 8.29818904e-01 -8.01143765e-01 3.79492134e-01 -5.02380580e-02 -3.42870913e-02 -6.12984657e-01 1.90655008e-01 -1.37135565e-01 -3.89674067e-01 4.16042209e-01 3.70382488e-01 4.74911898e-01 7.52882883e-02 -1.39624730e-01 1.06759441e+00 3.25083226e-01 4.15252954e-01 -3.47612128e-02 5.81491709e-01 -6.79462075e-01 8.81268382e-01 7.58889019e-02 -8.56087327e-01 1.87859014e-01 8.04607868e-01 -5.21883667e-01 -3.81461024e-01 -6.34735942e-01 3.45190056e-02 1.92500699e+00 -2.27360338e-01 -8.91897619e-01 -9.01387036e-01 -8.90594304e-01 1.63994376e-02 -9.93742608e-03 -1.13576317e+00 -4.17281747e-01 -2.84597993e-01 -1.01966536e+00 6.50295854e-01 6.14887595e-01 4.44925785e-01 -1.18333197e+00 -4.55341309e-01 -2.21284881e-01 -2.40622431e-01 -1.34789026e+00 -3.80127221e-01 -8.61657709e-02 -3.84962052e-01 -9.40398514e-01 -4.02839899e-01 -6.47789657e-01 3.68199259e-01 -1.68324500e-01 9.65694964e-01 -7.76002482e-02 -3.94694805e-01 4.65490520e-01 -2.39731148e-01 -7.22245514e-01 1.63658276e-01 6.82400614e-02 1.20772332e-01 7.52397239e-01 7.94007242e-01 -5.78548789e-01 -5.05884051e-01 4.35415238e-01 -5.69781542e-01 -1.41389072e-01 6.43238842e-01 4.63270098e-01 4.17264640e-01 -5.26563048e-01 7.48155296e-01 -6.04413927e-01 5.99599421e-01 -4.22888041e-01 -2.31582731e-01 3.07980150e-01 -3.32999200e-01 -3.72090220e-01 1.80392429e-01 -5.49735904e-01 -1.07137036e+00 3.88562620e-01 -4.28106599e-02 -6.01374924e-01 -3.81038338e-01 2.93638021e-01 -4.10309911e-01 -1.73628882e-01 3.37519705e-01 -1.26213282e-01 2.29496434e-01 -2.15759009e-01 1.73299611e-02 7.48874485e-01 5.60012124e-02 -5.09903073e-01 1.21578462e-01 1.81076229e-01 1.32553592e-01 -8.17249358e-01 -1.13430357e+00 -3.41116279e-01 -4.35363919e-01 -7.88249195e-01 1.13413036e+00 -1.36886501e+00 -1.10272670e+00 6.45540237e-01 -9.83732045e-01 -2.95563459e-01 4.82139170e-01 4.07302439e-01 -5.08313596e-01 7.57631660e-02 -5.62904298e-01 -9.10843372e-01 -6.01533294e-01 -1.04965889e+00 1.32952189e+00 3.95440102e-01 -5.26149511e-01 -8.75547945e-01 2.46575937e-01 6.51516318e-01 2.92603165e-01 5.30968130e-01 5.77085853e-01 -7.16891289e-01 1.29964247e-01 -1.31244034e-01 -1.29670262e-01 5.21693468e-01 1.10344738e-01 1.98863417e-01 -1.25969517e+00 -1.41185880e-01 -1.49337649e-01 -1.19742513e+00 8.99651706e-01 3.24383587e-01 1.59142125e+00 7.22003505e-02 -2.99007088e-01 6.66118264e-01 1.03227818e+00 6.81253150e-02 6.09986901e-01 2.96033937e-02 7.30621934e-01 7.98332810e-01 7.22672224e-01 8.84122431e-01 5.15486598e-01 8.04050088e-01 4.35343087e-01 -2.70044863e-01 3.39210689e-01 2.36887082e-01 6.64223552e-01 3.56833607e-01 -3.75720680e-01 6.35518879e-02 -7.22635567e-01 1.24836527e-03 -1.95091081e+00 -9.70280409e-01 1.66057274e-01 1.68389404e+00 6.49361312e-01 -2.17844218e-01 6.75412595e-01 -3.31554025e-01 5.96301854e-01 4.77017730e-01 -4.86714929e-01 -7.87124097e-01 9.62208062e-02 1.97232604e-01 -4.33735885e-02 1.67734355e-01 -1.63642454e+00 1.08978581e+00 6.18643332e+00 9.48465049e-01 -1.35604763e+00 2.41352424e-01 1.04191661e+00 -4.57927644e-01 3.81494641e-01 -7.06803918e-01 -1.04881799e+00 4.22709912e-01 9.95815396e-01 -1.50695041e-01 3.06225531e-02 9.75002646e-01 9.11971629e-02 -1.25028029e-01 -1.01263189e+00 1.48548651e+00 7.77696967e-01 -3.92176300e-01 -2.07550868e-01 -1.96656045e-02 3.64908516e-01 -2.96235472e-01 1.12156548e-01 8.24468374e-01 -2.62422740e-01 -1.16093123e+00 2.66879976e-01 4.84172583e-01 7.39809394e-01 -8.79982412e-01 5.13929009e-01 -6.09950349e-02 -1.18797994e+00 -2.70946264e-01 -8.83182287e-02 -1.06143974e-01 4.67838300e-03 3.82849246e-01 -1.51138186e-01 3.26262176e-01 8.51079285e-01 8.41025472e-01 -8.24595988e-01 6.66990280e-01 1.00798607e-02 4.37535495e-01 -4.79019098e-02 -2.30375215e-01 2.52496023e-02 -2.32160643e-01 -1.22680636e-02 1.30530262e+00 1.85601324e-01 1.15867052e-02 3.01041961e-01 2.86085635e-01 -3.57447237e-01 5.92348516e-01 -4.81950849e-01 -6.54036701e-02 -1.82522133e-01 2.18834448e+00 -3.57960731e-01 -1.46734774e-01 -6.58709466e-01 1.02491295e+00 6.04618073e-01 1.24587700e-01 -1.20658851e+00 -3.40800770e-02 1.13699317e+00 -3.19582701e-01 5.55244535e-02 1.83084667e-01 1.89608913e-02 -1.11245298e+00 -9.82627198e-02 -1.01439738e+00 6.52930737e-01 -4.82749879e-01 -1.18089461e+00 6.01176977e-01 -1.37483850e-01 -9.16164517e-01 -3.10512096e-01 -7.43055999e-01 -7.20026076e-01 4.53408837e-01 -1.21666777e+00 -1.36192167e+00 -5.69334149e-01 7.24674761e-01 3.50910574e-01 -4.06554550e-01 1.09967136e+00 7.43341208e-01 -1.09917438e+00 1.02577603e+00 -5.47667146e-01 2.15886608e-01 1.26777887e+00 -9.38511074e-01 -3.10729623e-01 4.56185073e-01 -1.19573109e-01 -5.83468121e-04 2.60754377e-01 -2.01260358e-01 -1.13755631e+00 -1.10574484e+00 6.24767721e-01 -3.55755061e-01 6.69154763e-01 -7.00508475e-01 -7.67476737e-01 6.80259705e-01 4.05529827e-01 1.82987064e-01 1.37044966e+00 6.66757822e-01 -5.42818487e-01 -5.53269625e-01 -9.03582394e-01 3.27753901e-01 7.52199948e-01 -4.03871804e-01 1.11270800e-01 1.82134509e-01 4.43806453e-03 -1.71719313e-01 -9.64839578e-01 5.12297213e-01 9.07834649e-01 -1.00131047e+00 5.77660382e-01 -8.48013461e-01 6.58015847e-01 2.90674239e-01 -8.53485465e-02 -1.31023490e+00 -2.40451425e-01 -5.13055086e-01 -2.01344192e-01 1.55254745e+00 3.31721604e-01 -7.83901215e-02 9.26206052e-01 5.99047959e-01 1.53132573e-01 -1.06535220e+00 -7.65720367e-01 -4.75828826e-01 -1.00221306e-01 -1.87831789e-01 1.91884831e-01 9.63474810e-01 1.92924991e-01 8.82092237e-01 -7.95407355e-01 -3.19104791e-01 5.34817517e-01 8.79094079e-02 8.67912233e-01 -1.16032541e+00 -2.04591379e-02 -5.39719522e-01 -4.56622064e-01 -2.49657854e-01 8.63607526e-01 -7.10254610e-01 -1.95429921e-01 -9.40192759e-01 5.11497915e-01 3.11442316e-01 -5.06762445e-01 9.09733355e-01 -3.25199604e-01 3.67960155e-01 4.87138517e-02 -3.62302482e-01 -1.23110580e+00 6.93772376e-01 8.74298275e-01 6.66670203e-02 1.35460785e-02 -2.02104840e-02 -7.94894457e-01 9.11973953e-01 1.07501543e+00 -2.24350929e-01 -2.70004213e-01 -1.43641114e-01 -2.98277810e-02 -1.36961907e-01 7.23245144e-02 -9.18089569e-01 2.67098099e-02 -1.43847331e-01 5.24159253e-01 -2.01796964e-01 7.52442718e-01 -5.29153705e-01 -2.64855117e-01 3.25166024e-02 -2.37678796e-01 3.60645592e-01 3.03365052e-01 1.47163525e-01 -3.21000814e-01 1.96131885e-01 9.93243158e-01 6.56201094e-02 -9.68254685e-01 7.90040493e-01 -2.58744627e-01 -4.54323292e-02 1.42152035e+00 1.25003934e-01 -2.99982995e-01 -3.48034889e-01 -8.59772921e-01 4.11964834e-01 -1.00649633e-01 6.89773858e-01 5.09513974e-01 -1.61088753e+00 -8.56413364e-01 1.33210942e-01 4.44767505e-01 -7.37161040e-01 5.22909760e-01 1.06831658e+00 2.36656711e-01 -1.35707825e-01 -6.91650689e-01 -4.38695103e-01 -2.03444576e+00 2.33992219e-01 4.63032335e-01 -3.53466511e-01 1.98799103e-01 8.27498734e-01 2.97436148e-01 -1.70974866e-01 4.41350847e-01 3.32370818e-01 -7.58606315e-01 7.86935270e-01 8.38940203e-01 3.78259242e-01 -3.77976634e-02 -7.37604082e-01 -6.14062548e-01 6.04203641e-01 -1.28165990e-01 1.27114579e-01 1.31560123e+00 1.97617654e-02 -3.31242770e-01 4.78091449e-01 1.52216113e+00 -2.21953809e-01 -9.17374671e-01 -9.24759954e-02 -3.29545862e-03 -2.53619909e-01 2.67648268e-02 -8.00514698e-01 -1.33681929e+00 9.68901873e-01 9.71070707e-01 -1.51271448e-01 1.42805398e+00 5.58603518e-02 3.85166585e-01 2.81859100e-01 -1.88782737e-02 -1.45853090e+00 6.23687983e-01 5.00553012e-01 9.00467813e-01 -1.54461849e+00 -1.57669574e-01 -3.18141609e-01 -9.37235653e-01 1.00150728e+00 1.48620582e+00 1.13565154e-01 8.85032892e-01 3.40831161e-01 3.24752182e-01 -2.82222420e-01 -1.05491376e+00 -2.38784760e-01 3.34220409e-01 3.76736283e-01 8.43191564e-01 2.09823642e-02 -2.86322743e-01 9.64272916e-01 2.05799356e-01 8.71537700e-02 2.92760860e-02 6.29977345e-01 -3.01803321e-01 -1.07350290e+00 8.35459027e-03 4.27104831e-01 -9.71894383e-01 2.67102599e-01 -6.89169049e-01 4.67239648e-01 7.36577064e-02 7.53593504e-01 1.73961520e-01 -6.15823984e-01 3.92385036e-01 3.78839850e-01 6.09234571e-01 -5.09918749e-01 -7.27340400e-01 2.44998619e-01 2.87552565e-01 -1.01408184e+00 -7.31174469e-01 -7.02507794e-01 -9.33820844e-01 -1.36648715e-01 1.44962385e-01 -9.49599594e-02 5.81380069e-01 8.48797858e-01 4.05003756e-01 4.12497163e-01 9.63302135e-01 -6.71118081e-01 -2.13997290e-01 -1.16342902e+00 -9.30194736e-01 5.95308304e-01 5.80433160e-02 -8.62291873e-01 -2.35054061e-01 -1.33724794e-01]
[13.601109504699707, 1.8698492050170898]
7af29229-ec55-4491-ae9b-5ceb8d09da01
pay-more-attention-to-relation-exploration
2305.02118
null
https://arxiv.org/abs/2305.02118v2
https://arxiv.org/pdf/2305.02118v2.pdf
Pay More Attention to Relation Exploration for Knowledge Base Question Answering
Knowledge base question answering (KBQA) is a challenging task that aims to retrieve correct answers from large-scale knowledge bases. Existing attempts primarily focus on entity representation and final answer reasoning, which results in limited supervision for this task. Moreover, the relations, which empirically determine the reasoning path selection, are not fully considered in recent advancements. In this study, we propose a novel framework, RE-KBQA, that utilizes relations in the knowledge base to enhance entity representation and introduce additional supervision. We explore guidance from relations in three aspects, including (1) distinguishing similar entities by employing a variational graph auto-encoder to learn relation importance; (2) exploring extra supervision by predicting relation distributions as soft labels with a multi-task scheme; (3) designing a relation-guided re-ranking algorithm for post-processing. Experimental results on two benchmark datasets demonstrate the effectiveness and superiority of our framework, improving the F1 score by 5.7% from 40.5 to 46.3 on CWQ and 5.8% from 62.8 to 68.5 on WebQSP, better or on par with state-of-the-art methods.
['Daniel Hershcovich', 'Min Chen', 'Bin Wang', 'Shuai Chen', 'Wen Dai', 'Huiwen Liu', 'Xianzhi Li', 'Yong Cao']
2023-05-03
null
null
null
null
['knowledge-base-question-answering']
['natural-language-processing']
[ 7.52643347e-02 4.44597304e-01 -4.11330521e-01 -2.42604032e-01 -9.70068038e-01 -4.29871649e-01 4.03917342e-01 4.53491509e-01 -4.25823063e-01 9.46387231e-01 3.85987490e-01 -4.25714105e-01 -4.87932414e-01 -1.13178039e+00 -8.03699911e-01 -3.13206196e-01 1.21095106e-01 7.80609906e-01 7.38789916e-01 -5.79479694e-01 1.32829934e-01 1.11107886e-01 -1.35913038e+00 4.28802431e-01 1.45840180e+00 9.07443225e-01 -1.98132813e-01 3.34854603e-01 -4.36176270e-01 1.29423916e+00 -3.32683772e-01 -1.24208140e+00 -1.91924423e-01 -3.91827255e-01 -1.41338646e+00 -3.66812885e-01 9.59178060e-02 -1.51682571e-01 -3.62753302e-01 1.00865078e+00 2.75973201e-01 3.34173292e-01 5.70589781e-01 -1.01121545e+00 -9.72560823e-01 7.54033566e-01 -4.77739304e-01 3.53762597e-01 5.78995407e-01 -3.85566086e-01 1.71098626e+00 -7.55168140e-01 6.25391245e-01 1.22463822e+00 3.62338334e-01 3.20547491e-01 -1.02365649e+00 -4.40411359e-01 8.84332582e-02 9.70763266e-01 -1.52298033e+00 -2.45820984e-01 8.38309169e-01 -3.23738009e-01 1.15175259e+00 2.67734706e-01 2.65806556e-01 5.91744959e-01 -1.52233496e-01 7.91340709e-01 7.72115350e-01 -4.18426782e-01 2.78981403e-02 9.53474119e-02 4.61580575e-01 9.37532187e-01 2.12390393e-01 -5.11058986e-01 -5.92020512e-01 -2.79316813e-01 3.80196244e-01 -2.13590339e-01 -4.10065860e-01 -3.17334235e-01 -1.03387690e+00 8.77726316e-01 6.86441898e-01 8.50656256e-03 -5.82952142e-01 -1.15642399e-01 1.84997723e-01 4.16931480e-01 4.32724983e-01 4.26408350e-01 -5.53683043e-01 4.93808575e-02 -3.16638649e-01 3.66186798e-01 9.45287168e-01 9.42209959e-01 9.04480517e-01 -5.16619921e-01 -6.18562877e-01 9.52831566e-01 3.46659362e-01 3.85850579e-01 9.60848406e-02 -8.44695926e-01 9.80476677e-01 1.07492685e+00 1.41168758e-01 -1.02170861e+00 -2.64139026e-01 -5.70850790e-01 -6.44076228e-01 -6.27754271e-01 1.80763170e-01 1.09637648e-01 -6.19589865e-01 1.51049912e+00 6.95745885e-01 1.62316471e-01 3.12407732e-01 8.74587297e-01 1.10459733e+00 5.82081497e-01 1.79597795e-01 -9.00630727e-02 1.70063627e+00 -1.31968510e+00 -8.81683767e-01 1.07856086e-02 7.22144127e-01 -6.08583391e-01 1.11227298e+00 1.37470394e-01 -1.00255823e+00 -3.76819402e-01 -8.52513492e-01 -4.08762664e-01 -2.57708997e-01 8.82177502e-02 6.82653069e-01 3.53760093e-01 -7.78655231e-01 3.77962530e-01 -5.74780405e-01 1.58830006e-02 4.80634451e-01 3.07612062e-01 -1.46886393e-01 -3.69936496e-01 -1.78545702e+00 9.13913071e-01 5.22388756e-01 -6.18030839e-02 -3.83727282e-01 -8.23754132e-01 -8.37665796e-01 3.67361128e-01 1.10509551e+00 -1.19487011e+00 1.05089498e+00 -1.06814884e-01 -1.51545954e+00 5.14164388e-01 -4.76106524e-01 -4.39971358e-01 2.41561860e-01 -3.79473090e-01 -6.00776076e-01 3.70520175e-01 2.70605057e-01 1.71628267e-01 3.63062382e-01 -1.17819488e+00 -7.24662602e-01 -4.31451142e-01 5.72683275e-01 4.91932243e-01 -3.04022878e-01 -1.02714211e-01 -9.14286733e-01 -3.28667611e-01 1.22647360e-01 -7.18215883e-01 -6.36233464e-02 -5.52065194e-01 -4.62148011e-01 -7.87713945e-01 2.67327964e-01 -9.76392567e-01 1.49246800e+00 -1.45591807e+00 3.73326123e-01 1.99971512e-01 4.75930512e-01 4.43042964e-01 2.20017359e-02 5.01296580e-01 2.95286387e-01 3.91115732e-02 -1.25378251e-01 -1.34171844e-01 8.82386565e-02 1.01855762e-01 -3.47356737e-01 -4.52427827e-02 4.74917352e-01 1.31951058e+00 -1.07343411e+00 -7.39586413e-01 -2.86777526e-01 4.32339668e-01 -6.77217901e-01 2.44394004e-01 -2.59821624e-01 1.56005383e-01 -8.35737646e-01 7.14805067e-01 3.51117343e-01 -6.09354377e-01 1.74398586e-01 -3.55186075e-01 4.10068065e-01 7.59881735e-01 -1.04660964e+00 1.42371464e+00 -4.59653467e-01 6.50392100e-02 -3.21116179e-01 -1.01270580e+00 9.18067813e-01 1.39486939e-01 2.13095248e-01 -9.33320522e-01 -9.57895294e-02 1.64443210e-01 5.48571385e-02 -8.81375849e-01 5.97749352e-01 3.86827290e-02 1.42254904e-01 2.34134212e-01 1.62438452e-01 1.77307740e-01 4.84327525e-01 5.79563916e-01 1.31341779e+00 8.26936960e-02 2.23690510e-01 3.98582825e-03 9.51787531e-01 9.08405520e-03 6.44208789e-01 5.60130060e-01 8.72846916e-02 1.85194463e-01 7.02469587e-01 3.77847478e-02 -3.57136905e-01 -8.93450856e-01 1.67654827e-01 1.06974769e+00 3.74201864e-01 -6.87173784e-01 -5.04452884e-01 -9.98544037e-01 2.52072811e-01 9.28063691e-01 -6.22735441e-01 -3.97806793e-01 -5.88748693e-01 -6.33490622e-01 3.17621231e-01 5.65483034e-01 5.57657838e-01 -9.36897635e-01 1.88008118e-02 1.69201747e-01 -6.70283020e-01 -1.30136359e+00 -1.83844864e-01 -1.37519553e-01 -7.31451809e-01 -1.24856007e+00 -4.83557075e-01 -4.99683440e-01 5.35400569e-01 3.17231715e-01 1.32890081e+00 2.60205418e-01 1.91865653e-01 3.38682771e-01 -6.07388020e-01 -1.08404107e-01 -6.98323175e-02 4.07481819e-01 -2.41998151e-01 9.04964134e-02 5.34424067e-01 -4.26477760e-01 -6.39641583e-01 1.82540372e-01 -6.77953959e-01 -1.56642795e-01 7.28458941e-01 8.06137025e-01 5.85329413e-01 2.02911735e-01 1.01601064e+00 -1.36357570e+00 7.95373917e-01 -7.07387209e-01 -2.85937697e-01 8.46268952e-01 -1.01493955e+00 2.18233883e-01 5.74399471e-01 9.84502658e-02 -1.44000506e+00 -6.22279882e-01 -1.98046207e-01 -1.02709763e-01 2.84775734e-01 8.71018231e-01 -1.67473689e-01 8.41041952e-02 4.02137011e-01 7.68552423e-02 -3.74791056e-01 -3.28342378e-01 6.55470610e-01 4.88708556e-01 3.67764205e-01 -7.82562494e-01 8.41970623e-01 2.99011230e-01 -4.27465476e-02 -2.51660883e-01 -1.43642437e+00 -7.59735346e-01 -5.18900871e-01 3.37842293e-02 7.73576856e-01 -9.80273187e-01 -9.80672896e-01 -7.69333616e-02 -1.03789139e+00 2.25614328e-02 -1.58965051e-01 3.97895008e-01 -8.30476657e-02 4.75463152e-01 -6.98239565e-01 -7.62162030e-01 -5.70682645e-01 -8.51063550e-01 8.90730679e-01 3.43653709e-01 6.60276189e-02 -1.05467892e+00 1.28311381e-01 1.37675548e+00 2.80836582e-01 -1.22952342e-01 1.23376501e+00 -8.05739760e-01 -8.69967222e-01 -3.66711058e-02 -5.71724594e-01 7.27271438e-02 1.24131307e-01 -3.69908631e-01 -7.25137711e-01 7.57789612e-02 -4.27605033e-01 -4.78993803e-01 1.17368746e+00 -2.24382788e-01 9.52049136e-01 -1.59525186e-01 -4.03465211e-01 1.82753995e-01 1.12461138e+00 8.13877359e-02 6.16577923e-01 2.36784101e-01 9.73986566e-01 8.91440928e-01 8.21646690e-01 1.70055866e-01 1.22347093e+00 6.83111429e-01 3.05351347e-01 2.89899617e-01 -2.73835361e-01 -5.35248220e-01 4.68390696e-02 1.15816057e+00 -2.40633696e-01 -2.17673779e-01 -1.05836165e+00 7.27601409e-01 -2.07985783e+00 -7.03907132e-01 -3.48767996e-01 1.88498080e+00 1.11166584e+00 1.29646242e-01 6.82554347e-03 2.05241330e-02 6.76423132e-01 5.82813658e-02 -5.65567672e-01 1.39025569e-01 6.17607012e-02 3.76181901e-01 9.46371034e-02 6.77845716e-01 -9.77290094e-01 1.11466432e+00 4.30213499e+00 9.15166974e-01 -3.76100183e-01 1.20538816e-01 2.81984121e-01 2.98523635e-01 -7.54524291e-01 2.38220513e-01 -9.27522302e-01 1.87213495e-01 6.93356514e-01 -3.61368328e-01 4.12842005e-01 5.52742422e-01 -4.00287151e-01 1.03355892e-01 -7.76296437e-01 7.56461501e-01 2.41959944e-01 -1.23401308e+00 1.61489204e-01 -1.58894420e-01 8.17169964e-01 -2.95110375e-01 -2.41520599e-01 9.29290652e-01 3.87167096e-01 -7.51469076e-01 2.96160877e-01 7.22231209e-01 3.25114548e-01 -9.18520927e-01 1.00517404e+00 2.58482128e-01 -1.31433415e+00 2.91345902e-02 -4.29994792e-01 1.13742210e-01 3.20066661e-01 8.40852141e-01 -7.66050339e-01 1.31108868e+00 7.24443913e-01 6.38167739e-01 -6.28593504e-01 7.28523552e-01 -7.57094085e-01 9.05993760e-01 -3.42866890e-02 -2.07085282e-01 4.75135706e-02 -2.35506549e-01 2.63693184e-01 8.83013010e-01 -2.34084558e-02 4.62922245e-01 7.56802708e-02 8.22820008e-01 -5.23906887e-01 3.74623924e-01 -4.44107465e-02 -4.03576456e-02 5.60843885e-01 1.24211526e+00 -4.02920634e-01 -2.81976253e-01 -5.02046645e-01 1.00550830e+00 8.55577409e-01 4.38710213e-01 -9.33113277e-01 -6.34580910e-01 2.92236894e-01 -2.78345883e-01 5.14503837e-01 1.63264006e-01 3.83508462e-03 -1.30253124e+00 2.99412161e-01 -6.18162870e-01 8.65316570e-01 -5.52674770e-01 -1.49617374e+00 5.63148379e-01 -9.09898132e-02 -7.18204796e-01 -1.58658713e-01 -2.98813313e-01 -2.95464188e-01 9.11575913e-01 -2.11220932e+00 -1.16422915e+00 -3.29410732e-01 4.61780518e-01 5.32286689e-02 -9.10822675e-02 6.33012235e-01 6.53163671e-01 -7.93556571e-01 6.71424270e-01 -1.37130961e-01 4.82272133e-02 6.15613580e-01 -1.41699445e+00 2.40173459e-01 5.79039514e-01 3.37537646e-01 7.58791924e-01 2.33482882e-01 -6.40249133e-01 -1.50527132e+00 -1.05299294e+00 1.31820011e+00 -6.08197570e-01 7.93508351e-01 9.10819872e-05 -1.41224182e+00 6.09337449e-01 7.53870383e-02 -3.86034288e-02 8.56520832e-01 7.86124885e-01 -5.49270570e-01 -2.10101798e-01 -9.83176291e-01 6.50333881e-01 1.19768393e+00 -5.89055777e-01 -8.57604086e-01 2.50421554e-01 1.07118642e+00 -5.28758585e-01 -1.13251686e+00 6.62387550e-01 2.23683670e-01 -7.86695778e-01 1.12015378e+00 -8.55764985e-01 5.98821878e-01 -4.02641982e-01 -4.89807539e-02 -1.06885898e+00 -4.78334576e-01 -3.95694107e-01 -9.23913062e-01 1.54292667e+00 6.35066867e-01 -7.68748164e-01 8.17255437e-01 5.56003392e-01 4.40357439e-02 -1.29394770e+00 -7.46310830e-01 -4.90195751e-01 -2.63770729e-01 -2.21078172e-01 6.81831419e-01 1.01468301e+00 -2.85417773e-02 9.28840816e-01 -1.03813283e-01 4.67258334e-01 4.55254912e-01 4.91328448e-01 6.04912817e-01 -1.20869422e+00 -3.11631262e-01 -2.66995400e-01 -9.12200660e-02 -1.29148686e+00 4.85015601e-01 -1.12310791e+00 -2.88199633e-01 -2.01398706e+00 3.76965404e-01 -4.40662920e-01 -5.45672655e-01 4.37461108e-01 -9.09842074e-01 -1.12630643e-01 -1.08698532e-01 1.17381155e-01 -1.15355325e+00 1.06429935e+00 1.20603311e+00 -9.56961438e-02 5.04379198e-02 2.56624054e-02 -1.05107856e+00 4.79918212e-01 6.71011031e-01 -5.14978349e-01 -8.22139680e-01 -4.21103030e-01 6.53042018e-01 1.87388644e-01 2.55004883e-01 -5.67065895e-01 3.87446523e-01 -5.82074262e-02 -1.85831890e-01 -5.36776960e-01 2.30279252e-01 -5.76755106e-01 -1.94385678e-01 1.80354089e-01 -3.92332882e-01 -2.24093437e-01 -7.89138302e-02 9.45544899e-01 -5.05760431e-01 -3.82865220e-01 7.47080669e-02 2.91038036e-01 -8.06732893e-01 3.00440222e-01 3.56936336e-01 6.29385591e-01 7.96640277e-01 4.21909600e-01 -5.52039683e-01 -3.62583905e-01 -3.69274318e-01 7.61381865e-01 -1.67112023e-01 4.10569906e-01 6.02443635e-01 -1.39888167e+00 -8.57896626e-01 -3.22413594e-01 3.43017310e-01 2.60343581e-01 5.05572200e-01 8.81634533e-01 -3.62837523e-01 6.77922308e-01 3.86647522e-01 -2.57178158e-01 -1.11316562e+00 4.25725013e-01 1.30043417e-01 -9.10509467e-01 -4.61006403e-01 1.15678608e+00 -1.57892153e-01 -5.40543377e-01 2.76778564e-02 -1.28279150e-01 -7.07291663e-01 2.08649740e-01 3.48190248e-01 5.20455956e-01 2.85109669e-01 -3.56924564e-01 -4.40789193e-01 4.19138223e-01 -3.48668069e-01 2.38089934e-01 1.24421918e+00 -1.20185472e-01 -3.15038055e-01 1.22982204e-01 9.92018044e-01 1.64087579e-01 -7.88426399e-01 -8.07829380e-01 2.94005841e-01 -3.10997933e-01 -6.42404333e-02 -8.74092400e-01 -1.00866747e+00 6.41762614e-01 -1.96072966e-01 2.96178818e-01 9.86483335e-01 3.36104929e-01 1.15817666e+00 8.14733088e-01 4.14336592e-01 -7.99195349e-01 2.11132422e-01 6.46775723e-01 9.30370986e-01 -1.30623543e+00 8.43602046e-02 -8.79432023e-01 -8.49483788e-01 7.39161372e-01 8.00540507e-01 1.78134009e-01 5.17627776e-01 -3.46556246e-01 -1.71428800e-01 -5.52632630e-01 -9.66183126e-01 -6.16108477e-01 8.46169651e-01 1.44877374e-01 4.53157008e-01 -2.81348377e-02 -4.35900152e-01 9.97441709e-01 -1.49783745e-01 -1.24606155e-01 3.75002511e-02 7.50150084e-01 -3.10184658e-01 -1.08379924e+00 2.25462392e-02 6.14934802e-01 -4.69343066e-01 -2.44236469e-01 -3.60959828e-01 4.49326664e-01 -4.00770232e-02 1.17278159e+00 -4.38329756e-01 -3.57388705e-01 6.79695249e-01 2.39118978e-01 4.97339994e-01 -6.94161415e-01 -4.52290565e-01 -5.18476605e-01 4.70655978e-01 -4.26552504e-01 -5.05338609e-01 -4.40435529e-01 -1.35568619e+00 -1.53271094e-01 -6.28585756e-01 5.35647869e-01 1.59432948e-01 1.15284169e+00 6.54102743e-01 9.95747030e-01 3.33493561e-01 1.49018303e-01 -6.67698383e-01 -7.88319230e-01 -4.20107275e-01 4.20226723e-01 1.18004233e-02 -8.82385492e-01 -1.84942082e-01 -1.62551984e-01]
[10.437469482421875, 7.976081371307373]
2b5fb4ce-bbdd-48f1-848e-17b482ca5f91
learning-lie-group-symmetry-transformations
2307.01583
null
https://arxiv.org/abs/2307.01583v1
https://arxiv.org/pdf/2307.01583v1.pdf
Learning Lie Group Symmetry Transformations with Neural Networks
The problem of detecting and quantifying the presence of symmetries in datasets is useful for model selection, generative modeling, and data analysis, amongst others. While existing methods for hard-coding transformations in neural networks require prior knowledge of the symmetries of the task at hand, this work focuses on discovering and characterizing unknown symmetries present in the dataset, namely, Lie group symmetry transformations beyond the traditional ones usually considered in the field (rotation, scaling, and translation). Specifically, we consider a scenario in which a dataset has been transformed by a one-parameter subgroup of transformations with different parameter values for each data point. Our goal is to characterize the transformation group and the distribution of the parameter values. The results showcase the effectiveness of the approach in both these settings.
['Efstratios Gavves', 'Rick Quax', 'Kevin Webster', 'Jeroen S. W. Lamb', 'Riccardo Valperga', 'Victoria Klein', 'Alex Gabel']
2023-07-04
null
null
null
null
['model-selection']
['methodology']
[ 4.85329390e-01 3.57876942e-02 -2.64340490e-01 -9.44386497e-02 -2.62330979e-01 -8.19357753e-01 1.28483605e+00 -1.17922224e-01 -1.58420056e-01 5.59726655e-01 3.56101751e-01 -6.28631935e-02 -4.98683453e-01 -4.53364283e-01 -8.39784443e-01 -8.64414632e-01 7.05969706e-02 6.84458435e-01 -1.41256317e-01 -6.08367659e-02 4.90264088e-01 7.79076934e-01 -1.42518020e+00 -1.11809485e-01 4.17490482e-01 8.08901548e-01 -3.30576710e-02 3.63089949e-01 2.64560610e-01 4.26759928e-01 -4.40899223e-01 -2.46091068e-01 6.80457413e-01 -4.85707879e-01 -7.92196810e-01 4.12087053e-01 5.37639737e-01 1.90365851e-01 -2.99643695e-01 1.26827776e+00 1.66549510e-03 3.19320053e-01 8.85807991e-01 -1.56273580e+00 -2.26156697e-01 5.62865674e-01 -2.34317198e-01 7.50866309e-02 8.85675177e-02 -3.46094698e-01 1.13322878e+00 -6.22773767e-01 9.21568334e-01 1.01749885e+00 2.76742250e-01 3.25348347e-01 -1.50140119e+00 -6.32447720e-01 -3.16773415e-01 2.49189496e-01 -1.44782853e+00 -6.34112060e-01 9.78443682e-01 -8.18511844e-01 4.66158420e-01 2.03278586e-01 3.75140727e-01 1.11633587e+00 1.96120262e-01 2.17192039e-01 7.47877598e-01 -4.35327858e-01 5.68110108e-01 -1.14339434e-01 -1.00269735e-01 2.81912088e-01 2.83882588e-01 -2.84458306e-02 -6.53035700e-01 -2.82653362e-01 8.58166218e-01 -2.42437959e-01 -1.14703022e-01 -9.37122047e-01 -1.38201463e+00 9.02432501e-01 -5.65049946e-02 4.10310090e-01 -4.47803020e-01 8.92448947e-02 1.81938037e-01 3.07132810e-01 3.55987787e-01 6.88103914e-01 -2.30092034e-01 -2.45735452e-01 -9.74712133e-01 2.63872117e-01 6.52151763e-01 1.07857931e+00 7.70723224e-01 -7.32481107e-02 2.70143431e-02 5.78235447e-01 -1.06794320e-01 -2.80265268e-02 4.84239548e-01 -7.50867188e-01 5.66754818e-01 4.60622936e-01 -1.69412017e-01 -9.86294270e-01 -3.82416725e-01 -3.69373232e-01 -9.53267574e-01 -5.90781905e-02 5.92352450e-01 1.61834404e-01 -8.49236190e-01 2.01014757e+00 2.39107832e-01 2.84927696e-01 -1.37666047e-01 4.62798268e-01 1.46721125e-01 2.76215374e-01 -3.00958902e-01 -7.04616085e-02 1.09780574e+00 -3.25524479e-01 -4.46083784e-01 -4.46078442e-02 6.35951340e-01 -8.55462670e-01 6.27538025e-01 3.52413058e-01 -8.48465860e-01 -1.86659768e-01 -8.42206120e-01 -8.02528560e-02 -9.04020816e-02 2.26385713e-01 4.38367993e-01 4.39778805e-01 -8.61404717e-01 5.82261264e-01 -9.06258881e-01 -2.38949403e-01 2.40699917e-01 4.70568925e-01 -6.28920913e-01 2.20726840e-02 -7.96324372e-01 7.87146389e-01 5.08276820e-01 4.71889041e-02 -3.58251154e-01 -7.51844585e-01 -7.69510210e-01 8.31141248e-02 3.43595147e-01 -4.41631854e-01 8.70249569e-01 -7.39570558e-01 -1.12055850e+00 9.45124328e-01 -2.15394109e-01 -4.14411783e-01 6.97545648e-01 1.69192851e-01 -9.04743001e-02 -2.12775245e-01 1.06729730e-03 3.59074265e-01 8.68939221e-01 -7.88261890e-01 -1.06618702e-01 -4.67521101e-01 7.10550742e-03 -1.56416390e-02 2.19030343e-02 1.69854164e-01 -1.15974136e-01 -7.28853822e-01 8.08198571e-01 -1.22287941e+00 1.44749796e-02 -1.24557257e-01 -8.30598772e-01 -1.29051730e-01 9.15865660e-01 -5.67176044e-01 6.88887715e-01 -2.17394137e+00 4.56562877e-01 7.08037198e-01 1.28359035e-01 -2.82474399e-01 6.51871040e-02 6.06274307e-01 -5.71444809e-01 1.37166694e-01 -1.25320002e-01 -2.99390614e-01 9.29367840e-02 2.27609947e-01 -6.01477742e-01 8.39145660e-01 2.87637532e-01 5.74037552e-01 -2.83646762e-01 -1.82511732e-01 6.75110742e-02 4.37549621e-01 -7.04582870e-01 -8.67257267e-02 -1.60299271e-01 6.21596813e-01 -4.03854519e-01 1.04471818e-01 5.02806425e-01 -9.16220769e-02 1.95781305e-01 -2.42011949e-01 -2.77163330e-02 4.04049546e-01 -1.69449580e+00 1.17035782e+00 -2.39331022e-01 7.13166296e-01 -4.73116636e-01 -9.83287573e-01 1.06439853e+00 1.75027654e-01 8.01683068e-01 -2.74175972e-01 3.64522219e-01 -5.17460853e-02 4.02428478e-01 -2.40788594e-01 2.76309550e-01 -9.01387855e-02 9.43959504e-02 6.70921087e-01 2.39865437e-01 -1.10419042e-01 4.96450871e-01 -5.25983684e-02 1.02778256e+00 -1.36937231e-01 5.32817006e-01 -3.89176488e-01 2.61130542e-01 -3.67928326e-01 4.66316789e-01 5.14272749e-01 3.47246587e-01 7.65974939e-01 1.01495588e+00 -3.15158576e-01 -1.36762202e+00 -7.91190922e-01 -2.63413429e-01 6.92787349e-01 -1.94570303e-01 -4.29088563e-01 -7.25300193e-01 -1.70120090e-01 -1.28780410e-01 6.81389213e-01 -8.07279587e-01 -4.77349043e-01 -5.94418585e-01 -7.58514225e-01 3.37288260e-01 1.75185680e-01 4.46088225e-01 -8.35201681e-01 -4.83497620e-01 -1.92965046e-01 -2.65338808e-01 -1.29665112e+00 -4.44876671e-01 1.91826060e-01 -8.71713519e-01 -1.25650167e+00 -1.94510534e-01 -4.36167538e-01 8.26847732e-01 -2.11692944e-01 6.24200881e-01 -4.13038254e-01 -2.34066308e-01 6.85551763e-02 -5.82475588e-02 -1.55037880e-01 -5.30095518e-01 1.57050833e-01 1.18274182e-01 2.70011455e-01 1.24140419e-01 -6.61688149e-01 -1.54268974e-02 3.93556565e-01 -1.12275565e+00 9.10529047e-02 3.52331609e-01 6.62946582e-01 4.38671589e-01 3.26242775e-01 1.05815362e-02 -8.34758997e-01 6.68486714e-01 -4.64535296e-01 -6.39379561e-01 2.41735548e-01 -1.68267414e-01 5.86746097e-01 5.58417618e-01 -7.57990241e-01 -5.93979776e-01 4.18181896e-01 3.10753256e-01 -4.40958381e-01 -2.17372388e-01 5.29499888e-01 -5.23152709e-01 -3.75372976e-01 6.44200802e-01 2.61172444e-01 -1.03050292e-01 -4.97644931e-01 5.69170155e-02 1.18752219e-01 5.76409936e-01 -6.18690789e-01 1.04615045e+00 3.37281346e-01 5.64820647e-01 -1.01044786e+00 -3.65318060e-01 -3.49944502e-01 -1.00476170e+00 1.04999669e-01 5.25118768e-01 -4.99892682e-01 -3.70268881e-01 4.57492262e-01 -9.46833909e-01 -9.62560344e-03 -5.00794828e-01 4.36790019e-01 -8.67999971e-01 3.49571139e-01 1.23604424e-01 -4.94061857e-01 3.47337306e-01 -1.21178222e+00 8.98098469e-01 -1.89245760e-01 -4.52730000e-01 -1.03777754e+00 1.07695803e-01 9.66344625e-02 2.50588268e-01 3.94918233e-01 1.38949263e+00 -9.38630462e-01 -7.25183547e-01 -3.75288904e-01 1.28310829e-01 2.00007007e-01 2.25106925e-01 1.53442666e-01 -7.82683611e-01 -4.75342534e-02 3.29620600e-01 7.07089305e-02 5.99137127e-01 3.28521878e-01 1.36992860e+00 -3.82976443e-01 -2.24034991e-02 7.50881433e-01 9.16914046e-01 1.82447627e-01 4.30628479e-01 3.39944214e-01 6.18624985e-01 6.61573827e-01 7.45106116e-02 5.18021345e-01 1.03093404e-02 1.02802622e+00 4.40461427e-01 1.80079281e-01 2.76578218e-01 -2.69059032e-01 -6.12506494e-02 3.48662347e-01 -1.29960313e-01 -2.14852512e-01 -9.98969674e-01 2.49380529e-01 -1.60316765e+00 -8.59374523e-01 3.22604656e-01 2.32615829e+00 5.72120190e-01 7.07842261e-02 4.95962985e-02 5.79725862e-01 8.17793787e-01 2.20667750e-01 -4.99693215e-01 -1.12995945e-01 -1.41271606e-01 2.27829963e-01 6.07086778e-01 3.07844162e-01 -1.11777961e+00 5.03896058e-01 6.71535826e+00 4.45793301e-01 -1.27672327e+00 -4.58289027e-01 1.79657951e-01 9.52182785e-02 -2.86115378e-01 2.18924105e-01 -6.95469022e-01 3.27895820e-01 4.57868099e-01 -4.14698929e-01 6.93326831e-01 6.01157188e-01 1.05750032e-01 5.30537292e-02 -1.42374611e+00 7.71729290e-01 2.32728973e-01 -1.28425205e+00 2.60680079e-01 5.11162162e-01 7.33478963e-01 -1.59580931e-01 2.23258510e-01 -3.02546859e-01 2.32017413e-02 -8.40625703e-01 8.18801403e-01 5.15124619e-01 6.05074465e-01 -7.31906891e-01 4.26783174e-01 5.35038710e-01 -7.04803467e-01 -3.16893980e-02 -2.50226259e-01 -8.53552148e-02 -1.60974815e-01 4.39920276e-01 -1.19894528e+00 3.50033015e-01 4.13269475e-02 6.51864171e-01 -6.96069658e-01 1.14559710e+00 -1.14099644e-01 6.09794259e-01 -5.14178932e-01 2.07127869e-01 -3.14408615e-02 -6.02279842e-01 5.96397519e-01 5.83782494e-01 5.65168798e-01 -2.17954010e-01 -1.35176376e-01 9.63643551e-01 -1.97405696e-01 -6.62617683e-02 -6.90508604e-01 -1.85396567e-01 3.23381066e-01 8.78807902e-01 -8.92054856e-01 1.69322386e-01 -4.78506200e-02 4.15663511e-01 5.30748069e-02 4.30835336e-01 -5.34506977e-01 -1.25418514e-01 7.93937862e-01 3.80804360e-01 3.23196113e-01 -5.88652432e-01 -1.40913352e-01 -1.22524643e+00 2.84939885e-01 -7.60527432e-01 2.22631142e-01 -4.63971257e-01 -8.97550583e-01 2.27025166e-01 5.75916648e-01 -1.19580078e+00 -7.91538358e-01 -7.62344539e-01 -7.13289320e-01 6.78441405e-01 -7.15646625e-01 -1.09403896e+00 -1.69622004e-01 5.19031703e-01 1.46593302e-01 -3.78981590e-01 5.49469829e-01 3.33444513e-02 -3.68573010e-01 3.83173317e-01 3.81744444e-01 8.92631337e-02 4.00680214e-01 -1.04686344e+00 5.04741967e-01 1.01879442e+00 5.56814730e-01 7.97535956e-01 1.14353311e+00 -4.33672965e-01 -1.27923357e+00 -9.06077981e-01 1.07889163e+00 -1.78007007e-01 8.75400543e-01 -8.54949534e-01 -7.40197062e-01 9.92522597e-01 -4.96580213e-01 2.86983587e-02 3.89821053e-01 -3.33462469e-02 -4.29539561e-01 2.36388400e-01 -1.09856975e+00 6.47847891e-01 1.13733757e+00 -6.55077219e-01 -3.33167732e-01 4.77198809e-01 1.84204683e-01 -5.39648116e-01 -5.75616062e-01 2.52264291e-01 3.64701092e-01 -6.76396132e-01 8.77768397e-01 -1.10138834e+00 4.91031379e-01 -1.68966621e-01 -1.68785065e-01 -1.34645903e+00 -2.57864535e-01 -6.00331903e-01 1.33154169e-01 1.07802320e+00 1.28351450e-01 -6.27427578e-01 1.01678216e+00 7.93097496e-01 1.93367258e-01 -2.90160596e-01 -1.30097687e+00 -7.42446721e-01 9.50358436e-02 -3.93527329e-01 7.27835536e-01 1.12656748e+00 -4.34464961e-01 1.53602362e-01 -3.76603842e-01 1.21758670e-01 5.21904111e-01 2.91705821e-02 8.82441223e-01 -1.49610341e+00 -1.08110338e-01 -4.53466207e-01 -1.13597870e+00 -6.56329930e-01 4.50028867e-01 -1.08088386e+00 -1.94336921e-01 -9.31158602e-01 1.32125258e-01 -2.27507934e-01 1.44633368e-01 4.26129460e-01 4.17233437e-01 1.03418536e-01 1.25828654e-01 2.21003503e-01 -1.50654819e-02 4.87653375e-01 1.01745641e+00 1.01445839e-01 -3.71685103e-02 4.11710829e-01 -5.09628892e-01 6.50668740e-01 6.58554256e-01 -6.17085099e-01 -4.85752940e-01 -6.76469654e-02 5.31376719e-01 -1.96690470e-01 6.71570837e-01 -1.07365453e+00 4.86458063e-01 -4.82695371e-01 2.06813186e-01 -3.25227201e-01 2.89688170e-01 -8.81509542e-01 6.24609828e-01 1.60083339e-01 -6.19457126e-01 2.15890899e-01 -6.16423637e-02 3.33268434e-01 -3.80087942e-01 -5.55212796e-01 6.39198363e-01 2.31389537e-01 -3.06047171e-01 3.09797883e-01 -1.98433354e-01 2.36975715e-01 9.72713232e-01 -1.67578310e-01 -2.30928704e-01 -5.24470568e-01 -7.57959306e-01 -3.67751330e-01 5.30265331e-01 6.21236503e-01 2.47556821e-01 -1.36362720e+00 -4.98953670e-01 7.37018228e-01 1.94466501e-01 -4.02330566e-04 -7.52667040e-02 7.03977525e-01 -2.88804263e-01 3.12327951e-01 -5.49056411e-01 -5.33479929e-01 -1.33089387e+00 2.13754341e-01 4.83333051e-01 -5.36852144e-02 -3.53486627e-01 2.95238703e-01 2.20519796e-01 -4.44151729e-01 1.69092372e-01 -4.96576488e-01 -9.13748518e-02 1.64532870e-01 -1.89723112e-02 3.96100760e-01 3.47238868e-01 -9.66594756e-01 -1.46898553e-01 5.46003759e-01 6.84288889e-02 -6.32202253e-02 1.31488729e+00 3.46236557e-01 -1.38754129e-01 4.49337989e-01 1.28374398e+00 -9.78681594e-02 -9.21879232e-01 -4.79069829e-01 1.81745470e-01 -4.05469745e-01 -1.55911490e-01 -1.74696922e-01 -9.08700168e-01 7.25825727e-01 2.15311006e-01 3.08211029e-01 8.59323144e-01 1.16925761e-01 -7.30772913e-02 6.51603222e-01 1.91581845e-01 -7.98127294e-01 -1.22641861e-01 6.51673794e-01 1.10655391e+00 -8.01636696e-01 6.60308674e-02 -4.42495018e-01 -3.98070574e-01 1.22536278e+00 1.53196290e-01 -2.83104360e-01 7.86996543e-01 1.62925601e-01 -2.70459861e-01 -6.31191581e-02 -6.21723175e-01 1.73509806e-01 7.42980063e-01 4.11487341e-01 1.64751679e-01 -1.77843109e-01 -8.41908753e-02 4.74319886e-03 -1.00168347e+00 -2.31615841e-01 7.17213035e-01 7.24952519e-01 1.12558313e-01 -1.16880703e+00 -2.95113981e-01 5.60314834e-01 -2.25875750e-01 1.80786908e-01 -7.15611517e-01 9.36411977e-01 1.33842468e-01 3.32439691e-01 2.55963326e-01 -1.99987754e-01 3.79603148e-01 2.01316133e-01 5.26793540e-01 -5.02918363e-01 -1.97711185e-01 -1.24561146e-01 -1.43666074e-01 -3.83936167e-01 -3.29515636e-01 -1.16527307e+00 -7.52851546e-01 -1.43364653e-01 -1.74986571e-01 -1.37391523e-01 9.35902655e-01 1.16509938e+00 1.03175998e-01 8.92116353e-02 6.34301186e-01 -7.09646881e-01 -7.82164574e-01 -9.51944709e-01 -7.88247406e-01 6.97345614e-01 4.38569129e-01 -1.03641272e+00 -7.47933745e-01 2.26523891e-01]
[9.052184104919434, 2.4678966999053955]