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
d46c5ac7-7c41-4d1e-b7bc-c1a2ac3394a4
unsupervised-latent-tree-induction-with-deep-1
null
null
https://aclanthology.org/N19-1116
https://aclanthology.org/N19-1116.pdf
Unsupervised Latent Tree Induction with Deep Inside-Outside Recursive Auto-Encoders
We introduce the deep inside-outside recursive autoencoder (DIORA), a fully-unsupervised method for discovering syntax that simultaneously learns representations for constituents within the induced tree. Our approach predicts each word in an input sentence conditioned on the rest of the sentence. During training we use dynamic programming to consider all possible binary trees over the sentence, and for inference we use the CKY algorithm to extract the highest scoring parse. DIORA outperforms previously reported results for unsupervised binary constituency parsing on the benchmark WSJ dataset.
['Andrew McCallum', 'Patrick Verga', 'Mohit Yadav', 'Andrew Drozdov', 'Mohit Iyyer']
2019-06-01
null
null
null
naacl-2019-6
['constituency-grammar-induction']
['natural-language-processing']
[ 3.91142040e-01 7.88369596e-01 -2.38082156e-01 -8.31737339e-01 -5.49804449e-01 -7.46914983e-01 9.92798731e-02 3.71126294e-01 -4.58384037e-01 7.20191300e-01 5.88672519e-01 -7.90671408e-01 1.45341977e-01 -1.15981948e+00 -9.45218086e-01 -4.99930292e-01 -2.23957896e-01 6.89368963e-01 4.77918833e-02 -1.13277197e-01 -3.06580096e-01 2.55700052e-01 -1.08478236e+00 6.38200462e-01 2.98163354e-01 6.48835242e-01 4.56432730e-01 1.17769074e+00 -5.76324344e-01 1.12231505e+00 -7.99846709e-01 -7.10348666e-01 8.62270147e-02 -3.68584901e-01 -1.08122325e+00 -2.18471095e-01 3.93953741e-01 -2.98851222e-01 -2.56990820e-01 9.48308468e-01 -6.18158430e-02 2.71430463e-02 3.44073355e-01 -6.34449244e-01 -3.56166542e-01 1.50772500e+00 1.49710357e-01 4.68999475e-01 1.41511083e-01 -3.70445102e-01 2.00269437e+00 -4.05268192e-01 8.98918390e-01 1.13273370e+00 2.72176534e-01 5.11744082e-01 -1.47755420e+00 -3.35351735e-01 2.53961414e-01 -1.81071058e-01 -7.30067372e-01 -1.60788432e-01 8.96715939e-01 -4.31384742e-01 1.75144827e+00 2.78075635e-01 5.03173172e-01 1.04687512e+00 3.36612493e-01 9.45022285e-01 1.11874008e+00 -9.20827985e-01 3.49112839e-01 -4.28649187e-01 9.72611845e-01 1.01270199e+00 -4.67406996e-02 1.08371899e-01 -4.00667459e-01 -3.27460021e-01 4.25684750e-01 -4.15285081e-01 4.18720365e-01 2.44724721e-01 -7.55788147e-01 1.29185331e+00 1.61925271e-01 3.97548795e-01 -4.76320952e-01 1.50814116e-01 3.41168076e-01 3.41821104e-01 2.05035850e-01 4.26730275e-01 -1.15246201e+00 -1.44949675e-01 -7.24529862e-01 7.22927228e-02 1.24322450e+00 4.31419373e-01 9.02679443e-01 -3.09142526e-02 -1.20774932e-01 7.67707288e-01 1.57474428e-01 1.21091753e-02 5.19763887e-01 -9.52736795e-01 4.67534810e-01 6.38493359e-01 -6.18060589e-01 -5.09935021e-01 -4.23845053e-01 -2.60688931e-01 -3.51659507e-01 7.01466575e-02 2.45134771e-01 -5.61900735e-01 -9.46146011e-01 1.88012409e+00 4.16286319e-01 -1.72007740e-01 5.97582042e-01 3.27987462e-01 1.07157373e+00 9.51248527e-01 5.63978672e-01 -9.95715559e-02 1.30859959e+00 -7.99908578e-01 -6.44432962e-01 -6.01863623e-01 6.96104944e-01 -3.61980796e-01 5.63396394e-01 2.65724063e-01 -1.05348098e+00 -3.83633137e-01 -1.15362751e+00 -3.41000736e-01 -4.02906299e-01 3.07769418e-01 1.03295887e+00 4.19657201e-01 -1.07344973e+00 7.40334332e-01 -1.08175933e+00 2.09030509e-01 1.36172771e-01 7.00806260e-01 -4.69420195e-01 3.53139609e-01 -1.19255054e+00 6.59536600e-01 9.41891015e-01 -2.00428233e-01 -7.31286824e-01 -1.77163333e-01 -1.21866465e+00 2.10119665e-01 8.34146589e-02 -5.28475761e-01 1.49131238e+00 -1.12286067e+00 -1.70812929e+00 1.10059142e+00 -5.24889827e-01 -9.29396987e-01 -5.43833733e-01 -3.73413980e-01 -2.25112006e-01 7.08814412e-02 1.12074278e-01 6.77182734e-01 5.22232175e-01 -1.02748537e+00 -4.92157489e-01 -4.60589021e-01 3.77886742e-01 -4.40310612e-02 1.39517516e-01 2.99864441e-01 -1.48904353e-01 -6.01526797e-01 2.14641750e-01 -6.66914463e-01 -4.34181839e-01 -9.23504651e-01 -9.71301436e-01 -6.40201092e-01 3.15997750e-01 -1.14694488e+00 1.32844400e+00 -2.10406470e+00 4.25655901e-01 -3.51217873e-02 2.96737820e-01 -1.26521170e-01 -5.02036810e-02 3.02346826e-01 -2.89967626e-01 1.94373429e-02 -5.69920480e-01 -5.83619118e-01 -4.58478481e-02 9.86914814e-01 -4.83229876e-01 -1.70404892e-02 5.68618178e-01 8.18731427e-01 -5.85343063e-01 -6.01889670e-01 -8.14289693e-03 9.65164155e-02 -6.81129515e-01 4.81939018e-01 -7.72701859e-01 -6.50452450e-02 -4.13284034e-01 1.83692962e-01 3.60023797e-01 -2.13737711e-01 7.02947319e-01 2.48046800e-01 -1.66814134e-01 1.15838373e+00 -7.87291229e-01 1.61040938e+00 -4.85246778e-01 5.74974358e-01 -7.25142518e-03 -1.14782214e+00 9.82546031e-01 3.21152389e-01 -3.08699198e-02 -2.07231283e-01 2.98850626e-01 -3.05248145e-02 2.49555901e-01 -3.59623134e-01 3.40260237e-01 -2.63869405e-01 -6.20423675e-01 2.89786726e-01 6.42985225e-01 1.61252394e-01 3.49362075e-01 4.15048838e-01 1.30866551e+00 9.16285664e-02 6.05126679e-01 -2.17565611e-01 2.71129966e-01 2.97240652e-02 7.44434416e-01 7.99387634e-01 1.79539531e-01 1.74107730e-01 8.27640057e-01 -7.97171593e-01 -8.78525496e-01 -1.33180976e+00 -2.22076237e-01 1.58756638e+00 -4.72124219e-01 -4.62373793e-01 -9.71236467e-01 -9.11779046e-01 -2.70639807e-01 1.10713565e+00 -7.05020607e-01 3.65122616e-01 -1.12117124e+00 -5.80346882e-01 5.42740464e-01 6.52442575e-01 2.92938817e-02 -1.61327946e+00 -8.00503492e-01 6.37112141e-01 -1.12535492e-01 -1.33566344e+00 2.85599411e-01 1.21984279e+00 -9.68150198e-01 -9.62209046e-01 2.29054242e-01 -1.34920883e+00 4.89757001e-01 -8.80109489e-01 1.42506135e+00 9.98018607e-02 -3.35841656e-01 -1.24396615e-01 -2.72875190e-01 -1.27181914e-02 -9.45948184e-01 1.51263207e-01 -3.57081085e-01 -3.11317384e-01 7.09116280e-01 -5.69939911e-01 -7.85330459e-02 -7.71323383e-01 -6.25629365e-01 5.66730872e-02 4.18690771e-01 7.85124421e-01 7.19343185e-01 2.91685294e-02 -7.45274648e-02 -1.22553992e+00 4.47756886e-01 -4.38796520e-01 -7.22726703e-01 2.69656479e-01 -3.59077975e-02 5.96831083e-01 8.43039095e-01 -6.56288043e-02 -9.61906433e-01 5.26832998e-01 -6.35071576e-01 3.33758593e-01 -7.44299591e-01 6.01004839e-01 -2.38681212e-01 7.85940111e-01 4.29725051e-01 3.59402984e-01 -5.10378957e-01 -5.56247413e-01 4.78215903e-01 4.57644820e-01 7.54311204e-01 -7.80788600e-01 3.39392960e-01 1.62828013e-01 -1.30238593e-01 -7.82590568e-01 -1.14723861e+00 -2.58424170e-02 -9.78660703e-01 3.65185767e-01 1.55805016e+00 -7.49091029e-01 -4.33885008e-01 -1.32197902e-01 -1.74611616e+00 -4.99835789e-01 -3.90137941e-01 3.59677315e-01 -3.23461562e-01 2.00654715e-01 -1.01544380e+00 -7.74677753e-01 -6.02087736e-01 -1.05825603e+00 1.09282875e+00 2.11017206e-01 -5.95899761e-01 -9.43720937e-01 5.11132717e-01 2.17078105e-01 -1.42987177e-01 4.11559865e-02 1.42044663e+00 -1.28683853e+00 -3.80915731e-01 -1.01904422e-01 2.63104677e-01 4.52717006e-01 -1.35072067e-01 1.06936276e-01 -9.78406489e-01 2.78958857e-01 -4.07410227e-02 -4.68278497e-01 9.47674751e-01 5.30544519e-01 1.10605073e+00 -5.83139837e-01 -1.64020240e-01 3.95199984e-01 1.28767049e+00 2.21714839e-01 2.77427465e-01 4.73016769e-01 2.87734181e-01 6.75586820e-01 -9.71780196e-02 1.38068393e-01 4.24489260e-01 3.05869617e-02 3.06872874e-01 1.54382199e-01 6.45334944e-02 -3.51792753e-01 5.27725279e-01 1.02007806e+00 3.19222003e-01 -3.67613941e-01 -7.54915535e-01 5.71180880e-01 -1.38054514e+00 -6.34973705e-01 9.09965411e-02 1.48950684e+00 1.05173612e+00 4.59485590e-01 -3.47242326e-01 2.62602150e-01 5.96919179e-01 3.56746703e-01 -1.93216920e-01 -1.11116445e+00 -2.96474665e-01 1.16201007e+00 3.16386223e-01 9.36702311e-01 -1.41280532e+00 1.48243868e+00 7.15926933e+00 2.45986387e-01 -7.45536566e-01 9.55327004e-02 8.82456481e-01 1.99723527e-01 -4.04262692e-01 3.13635916e-01 -1.19005835e+00 -5.45300171e-03 1.37199044e+00 4.67358530e-01 4.38399762e-01 1.18292820e+00 -4.27154601e-01 -5.71933202e-02 -9.91395473e-01 1.18535951e-01 -1.68136701e-01 -1.40350199e+00 -2.62545824e-01 -2.19336003e-01 3.54818285e-01 4.02161419e-01 -3.89833570e-01 5.42236507e-01 1.20688331e+00 -9.33505118e-01 3.84236902e-01 -1.21659316e-01 4.08980876e-01 -5.78019261e-01 3.43479514e-01 4.75188047e-01 -9.49829698e-01 -1.24082200e-01 -4.39117551e-01 -5.19819975e-01 2.87252069e-01 7.24563718e-01 -9.34784234e-01 -8.90998989e-02 6.03163302e-01 2.49269292e-01 -3.52434963e-01 1.77937120e-01 -1.06802607e+00 1.25291717e+00 -4.83937949e-01 -3.58372957e-01 4.04419929e-01 -2.12361395e-01 6.13752902e-01 1.61828995e+00 -2.84907401e-01 3.21447313e-01 3.77353311e-01 8.57486665e-01 -1.72756955e-01 1.17043212e-01 -4.58680719e-01 -2.20862612e-01 4.16007549e-01 1.04212773e+00 -7.22526014e-01 -7.44866371e-01 -2.43306771e-01 1.11822104e+00 9.97474968e-01 1.53183162e-01 -2.24043190e-01 -2.54450709e-01 7.05182195e-01 -5.96630335e-01 9.06140447e-01 -4.33998615e-01 -4.26583052e-01 -9.65704918e-01 -8.51115659e-02 -6.54153585e-01 9.37544405e-01 -3.80672902e-01 -9.85589147e-01 9.01466727e-01 -6.83315843e-02 -1.85751542e-01 -5.21390140e-01 -1.11706758e+00 -9.84060287e-01 6.45086944e-01 -1.31331336e+00 -7.61309206e-01 6.21600032e-01 2.10871652e-01 8.54409039e-01 -1.11700818e-01 1.54891300e+00 -4.75751847e-01 -8.10307920e-01 2.34855041e-01 -1.48317143e-01 9.45014000e-01 -3.02367806e-01 -1.63994038e+00 9.22784686e-01 8.25614870e-01 8.98331344e-01 9.13650751e-01 4.84762073e-01 -6.98142946e-01 -1.20959985e+00 -7.54468441e-01 1.66598833e+00 -3.89014542e-01 8.79601479e-01 -6.90233767e-01 -6.70492828e-01 1.20080388e+00 6.29917085e-01 -1.34108141e-01 1.18324363e+00 5.70460498e-01 -5.98524213e-01 3.50495547e-01 -6.59160018e-01 3.10968965e-01 7.53657162e-01 -6.16654456e-01 -1.31910110e+00 1.65316299e-01 1.47604978e+00 -4.37125474e-01 -7.89185584e-01 3.17606091e-01 2.60084629e-01 -7.66010702e-01 7.21588671e-01 -1.09227467e+00 1.02471137e+00 1.27393335e-01 -5.31398952e-01 -1.00410903e+00 -4.06070381e-01 -4.35998768e-01 -4.74462241e-01 1.07912171e+00 1.01631320e+00 -2.66426086e-01 9.96540189e-01 5.05927801e-01 -2.20784888e-01 -7.35814393e-01 -1.19197690e+00 -1.82541490e-01 2.46920198e-01 -6.70362473e-01 5.11728346e-01 6.93708122e-01 1.58387005e-01 8.07279050e-01 1.47165045e-01 4.64325815e-01 5.43486953e-01 6.90720201e-01 2.81502038e-01 -1.01216209e+00 -1.21061218e+00 -1.94575503e-01 -3.33744645e-01 -1.18124568e+00 7.72264242e-01 -1.18965161e+00 3.21397543e-01 -1.39635372e+00 1.06545456e-01 -6.20178273e-03 -2.30457366e-01 6.63525999e-01 -1.56701028e-01 -2.95873940e-01 -5.07087223e-02 -3.01553458e-01 -2.80336320e-01 2.06447706e-01 5.50351143e-01 -1.92034453e-01 2.99538840e-02 -2.09636211e-01 -5.60194314e-01 8.28499198e-01 9.81244087e-01 -8.01693380e-01 2.46507153e-01 -7.04789877e-01 3.56585562e-01 3.16372901e-01 -6.19504191e-02 -5.25563180e-01 -8.16351920e-02 1.45372957e-01 3.87826234e-01 -7.86033332e-01 1.75792143e-01 -5.57179391e-01 -5.37084103e-01 3.66897941e-01 -6.53990805e-01 1.40372127e-01 1.95525646e-01 3.41089696e-01 -2.54967034e-01 -5.23201644e-01 4.14580017e-01 -3.60695601e-01 -5.84126949e-01 -1.47205427e-01 -6.59685791e-01 4.64544035e-02 4.67473209e-01 3.85156035e-01 -7.21331835e-02 9.03172418e-02 -1.28172410e+00 -8.68830308e-02 -1.20933525e-01 1.93892166e-01 3.91520023e-01 -6.60649538e-01 -4.91338313e-01 3.52555126e-01 -4.13860291e-01 1.72965124e-01 -3.40068191e-01 -1.38831466e-01 -5.13571978e-01 3.87875617e-01 1.21354945e-01 -2.20134065e-01 -1.19175673e+00 3.06687593e-01 5.25963753e-02 -7.56620705e-01 -7.68075705e-01 1.30084944e+00 1.79603606e-01 -8.24281454e-01 2.80492395e-01 -4.62075233e-01 -5.46901941e-01 -3.45739663e-01 3.19829077e-01 -2.54743755e-01 -4.13716435e-02 -6.75091386e-01 -2.20868826e-01 2.07240462e-01 -2.75202900e-01 -2.08328962e-01 1.63788688e+00 4.10590172e-01 -5.17895579e-01 5.97827852e-01 1.32549977e+00 -8.91219974e-02 -1.00170505e+00 -2.12509632e-01 5.36672652e-01 3.42474908e-01 9.70929042e-02 -7.16017962e-01 -6.60747409e-01 7.90334046e-01 8.23189989e-02 2.90525168e-01 8.00415277e-01 5.43225408e-01 1.01944554e+00 6.41413808e-01 -1.35172710e-01 -1.00024760e+00 -3.23270053e-01 9.40611899e-01 4.70342010e-01 -9.70748186e-01 -2.66185969e-01 -4.82617795e-01 -6.40142143e-01 1.40141952e+00 3.29829335e-01 -7.11073220e-01 7.54222035e-01 8.87531102e-01 9.12793875e-02 -3.39160919e-01 -9.61855233e-01 -1.44268781e-01 -2.78662890e-02 -1.94301605e-02 7.70314634e-01 4.82157320e-01 -3.63594860e-01 1.08647513e+00 -9.64317679e-01 -5.28963685e-01 2.16369823e-01 1.05148494e+00 -6.01060271e-01 -1.49296582e+00 -1.20782018e-01 4.63151038e-01 -7.73610890e-01 -5.38826466e-01 -6.54825628e-01 3.97543699e-01 1.05456650e-01 9.59445775e-01 3.58677804e-01 -2.60930747e-01 8.64176527e-02 7.22510755e-01 4.72262889e-01 -9.56422508e-01 -8.28938007e-01 1.26548931e-01 6.01557136e-01 -4.72597659e-01 -2.91997135e-01 -7.23670959e-01 -1.85486543e+00 4.30653811e-01 -1.36327252e-01 2.57784963e-01 8.58501196e-01 1.21968126e+00 1.22711957e-01 5.04228234e-01 5.42395651e-01 -3.95437449e-01 -4.19217497e-01 -9.41788137e-01 -2.04129830e-01 2.40298152e-01 4.03594315e-01 -9.46591869e-02 -3.93810451e-01 2.05313951e-01]
[10.395528793334961, 9.591530799865723]
3a64bc58-e309-4ef0-98d0-4f5ae6f7fe74
exploiting-reducibility-in-unsupervised
null
null
https://aclanthology.org/D12-1028
https://aclanthology.org/D12-1028.pdf
Exploiting Reducibility in Unsupervised Dependency Parsing
null
["Zden{\\v{e}}k {\\v{Z}}abokrtsk{\\'y}", 'David Mare{\\v{c}}ek']
2012-07-01
null
null
null
emnlp-2012-7
['unsupervised-dependency-parsing']
['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]
[-7.178629398345947, 3.8342108726501465]
01ee91da-b0bd-4051-846e-6873554b4087
curvature-filtrations-for-graph-generative
2301.12906
null
https://arxiv.org/abs/2301.12906v2
https://arxiv.org/pdf/2301.12906v2.pdf
Curvature Filtrations for Graph Generative Model Evaluation
Graph generative model evaluation necessitates understanding differences between graphs on the distributional level. This entails being able to harness salient attributes of graphs in an efficient manner. Curvature constitutes one such property of graphs, and has recently started to prove useful in characterising graphs. Its expressive properties, stability, and practical utility in model evaluation remain largely unexplored, however. We combine graph curvature descriptors with emerging methods from topological data analysis to obtain robust, expressive descriptors for evaluating graph generative models.
['Bastian Rieck', 'Michael Bronstein', 'Jeremy Wayland', 'Joshua Southern']
2023-01-30
null
null
null
null
['topological-data-analysis']
['graphs']
[ 1.23973042e-01 2.27755919e-01 -1.51496142e-01 -1.66478708e-01 -3.39675814e-01 -9.46546197e-01 1.02961576e+00 6.56582475e-01 4.42057848e-02 2.98896343e-01 3.59989524e-01 -3.33481878e-01 -5.33767104e-01 -1.02688813e+00 -1.20078050e-01 -6.30683899e-01 -6.86136842e-01 3.59296530e-01 8.19675177e-02 -2.78371602e-01 2.85346985e-01 8.97627831e-01 -1.28634655e+00 -2.86572307e-01 8.62859309e-01 7.99475551e-01 -2.05555826e-01 8.18958938e-01 7.59469122e-02 3.42360646e-01 -9.31864679e-02 -6.46583259e-01 2.69612838e-02 -3.61199081e-01 -7.48212218e-01 6.77320585e-02 3.89046282e-01 2.21536174e-01 -2.65368283e-01 1.08502519e+00 2.14696303e-01 2.18136944e-02 1.19069147e+00 -1.28508997e+00 -6.29677236e-01 4.00635898e-01 -2.20988646e-01 2.50945896e-01 5.88970959e-01 5.63878529e-02 1.30666637e+00 -5.23682356e-01 8.97916317e-01 1.13757598e+00 6.87118113e-01 1.47230169e-02 -1.71361649e+00 -3.52873206e-02 -1.15679950e-01 -1.27298430e-01 -1.37259495e+00 -2.62748629e-01 1.22833931e+00 -6.83808744e-01 6.49869084e-01 4.92381722e-01 8.13684225e-01 9.04856861e-01 1.26785934e-01 6.22080505e-01 1.02534735e+00 -4.61800009e-01 1.08679056e-01 -4.27172408e-02 1.07914269e-01 8.69723201e-01 4.19605285e-01 8.64499509e-02 -1.92997828e-01 -3.02543253e-01 6.89864457e-01 -2.58078992e-01 -2.53738314e-01 -8.27385664e-01 -9.84156489e-01 8.83090854e-01 5.73892534e-01 4.76515591e-01 -2.24125013e-01 4.74929959e-01 4.05531436e-01 4.21815902e-01 4.50054735e-01 6.27192855e-01 9.90909785e-02 -4.11111206e-01 -7.00853527e-01 4.65452254e-01 7.77969658e-01 9.70366776e-01 8.33456337e-01 -3.57726440e-02 9.33400393e-02 5.81702411e-01 2.97911435e-01 3.74595106e-01 -1.96729273e-01 -5.89059293e-01 9.24967080e-02 1.11971772e+00 -4.41579163e-01 -1.39620686e+00 -4.64287996e-01 -3.69703472e-01 -9.54094350e-01 -3.74468789e-02 4.05925483e-01 3.51710677e-01 -4.79250550e-01 1.61537445e+00 3.26398879e-01 -2.77789742e-01 -3.17777485e-01 4.45165783e-01 5.68295717e-01 1.76366538e-01 2.21950129e-01 9.55008045e-02 9.08379018e-01 3.60460654e-02 -2.41728127e-01 2.26040736e-01 8.36546004e-01 -6.22962654e-01 1.04977965e+00 2.26522982e-01 -1.03950047e+00 -1.90156236e-01 -8.87931883e-01 4.13071364e-02 -6.96501136e-01 -1.52858466e-01 1.05933404e+00 8.86744261e-01 -1.19913030e+00 7.28120327e-01 -7.24912286e-01 -4.19752657e-01 4.23960805e-01 4.39847767e-01 -5.30132115e-01 1.81418043e-02 -9.08927381e-01 9.03450251e-01 3.13724071e-01 -2.45001651e-02 -4.38174069e-01 -5.20891428e-01 -1.10685146e+00 3.72964367e-02 2.24849224e-01 -6.72288895e-01 6.10471487e-01 -4.52104628e-01 -1.23310900e+00 8.52967441e-01 1.82298660e-01 -2.69796401e-01 4.83049452e-01 4.43975270e-01 -4.21633422e-01 2.29336634e-01 -3.22854608e-01 3.12071860e-01 6.71177447e-01 -1.16883349e+00 3.10741700e-02 -4.53505397e-01 3.86755839e-02 -1.08008578e-01 -3.45817566e-01 -2.69350827e-01 -1.32929027e-01 -5.28913319e-01 1.82630360e-01 -9.23735917e-01 -1.54665232e-01 -2.03473285e-01 -6.38209820e-01 -1.31917968e-01 6.75800025e-01 -3.28199774e-01 1.38385558e+00 -1.83989918e+00 2.63674200e-01 7.93018878e-01 9.05506492e-01 7.92981535e-02 -3.00861779e-03 9.79167759e-01 -2.70206351e-02 4.54045027e-01 -2.91958988e-01 8.22475106e-02 3.59448731e-01 -9.90797132e-02 -2.03289270e-01 6.35832131e-01 4.51111972e-01 1.32310665e+00 -1.03206921e+00 -5.49012780e-01 5.44163287e-01 5.50651073e-01 -5.44910550e-01 -6.25181049e-02 -8.86392891e-02 -1.43870339e-02 -6.40516639e-01 7.53406107e-01 4.91348118e-01 -2.75866568e-01 2.85627425e-01 -5.06066764e-03 7.45738000e-02 1.23227835e-01 -1.06582379e+00 1.05752993e+00 -1.71948209e-01 7.23305047e-01 -2.11114302e-01 -1.04416144e+00 1.05455768e+00 -2.31020451e-01 4.61216807e-01 -5.34053147e-01 1.26402184e-01 2.60613188e-02 2.76939347e-02 -3.80837947e-01 4.60582197e-01 -1.42335042e-01 -2.07936898e-01 3.88147444e-01 -4.27243300e-02 -4.77020413e-01 4.87740874e-01 3.71516258e-01 1.31465626e+00 -4.17983048e-02 5.28758347e-01 -6.70298338e-01 3.57029974e-01 -2.99658645e-02 -6.43511042e-02 2.63051301e-01 -9.09764171e-02 3.64686668e-01 8.62311184e-01 -4.44845736e-01 -9.22850788e-01 -1.59275484e+00 -1.66827276e-01 8.00571144e-01 7.32243508e-02 -7.36177206e-01 -6.60637736e-01 -6.07303858e-01 2.94445455e-01 2.92365640e-01 -8.73339832e-01 -4.57592130e-01 -5.10677993e-01 -3.91715467e-01 5.95215201e-01 5.43711305e-01 -1.21146813e-01 -7.02965856e-01 -2.05216706e-01 -5.11177443e-03 3.66148531e-01 -9.71665561e-01 -3.56829286e-01 -1.03798211e-02 -9.61543679e-01 -1.41513693e+00 -4.04673696e-01 -5.48651755e-01 8.74527216e-01 1.12101726e-01 1.43996096e+00 2.32468069e-01 -4.56071645e-01 7.66517818e-01 -1.75413087e-01 -1.50305510e-01 -6.41961157e-01 3.56435835e-01 -1.11122787e-01 -1.74630195e-01 9.00740698e-02 -8.81598890e-01 -3.49121630e-01 3.42866361e-01 -1.14913917e+00 -9.77877080e-02 6.12224698e-01 4.14549470e-01 5.52598715e-01 -6.18960299e-02 3.45788002e-01 -1.03566062e+00 1.16626382e+00 -2.95817375e-01 -4.45307523e-01 2.37327218e-01 -7.89177001e-01 3.38896930e-01 7.07096457e-01 -2.40841836e-01 -2.91613549e-01 -3.13866794e-01 -4.05516895e-03 2.15973835e-02 -1.65588260e-01 7.24771440e-01 -2.76270509e-01 -4.17645425e-01 5.26138544e-01 1.41727716e-01 8.20576400e-02 -1.51482791e-01 6.98696017e-01 5.40033579e-02 3.86034757e-01 -7.09330261e-01 1.08873296e+00 3.90844405e-01 6.74629748e-01 -1.34003532e+00 -3.10895056e-03 -4.47606832e-01 -7.37609208e-01 -3.72066617e-01 5.20501077e-01 -1.87591732e-01 -9.93018150e-01 7.19076954e-03 -7.31055140e-01 -7.31853694e-02 -3.27688545e-01 1.71009257e-01 -6.17658138e-01 4.78403181e-01 -2.72878438e-01 -8.41738999e-01 2.15067854e-03 -8.56965244e-01 1.24281585e+00 -1.92168772e-01 -4.48187023e-01 -1.64980602e+00 3.68878126e-01 -8.01515356e-02 5.80030203e-01 9.14901435e-01 1.26090443e+00 -5.68844199e-01 -4.88426447e-01 -5.87331355e-01 -2.59108990e-01 1.14702202e-01 8.18519816e-02 4.49439049e-01 -4.84005809e-01 -1.29983231e-01 -4.22138810e-01 -6.35119900e-02 8.15943003e-01 4.20531839e-01 8.91916871e-01 -5.80286942e-02 -4.14704651e-01 4.44438010e-01 1.45198810e+00 -4.92391258e-01 4.14133966e-01 -2.04078451e-01 7.39703357e-01 7.01672375e-01 1.75627265e-02 1.71413526e-01 4.93985981e-01 6.13250673e-01 4.45057184e-01 1.13187423e-02 -3.00720315e-02 -4.93133634e-01 3.84917557e-02 9.41384315e-01 -3.52856278e-01 -1.16266198e-01 -1.27292800e+00 4.25856888e-01 -1.50568104e+00 -9.85019267e-01 -4.10153329e-01 2.14129376e+00 2.80606925e-01 3.98930520e-01 3.41761887e-01 2.14594677e-01 6.92107260e-01 3.69727284e-01 -2.41910025e-01 -4.46880549e-01 -2.13036701e-01 3.38550769e-02 3.45918030e-01 4.00972873e-01 -8.73374701e-01 7.32492089e-01 7.72094870e+00 5.71541607e-01 -8.48628163e-01 -4.03662264e-01 5.82033098e-01 4.15276051e-01 -8.83582354e-01 3.82161677e-01 -2.90252894e-01 2.27031261e-01 7.99413741e-01 -4.65862632e-01 4.34625804e-01 8.37220550e-01 -6.51724339e-02 6.29282221e-02 -1.17864847e+00 9.02009964e-01 -7.47484341e-02 -1.40530813e+00 2.01315552e-01 5.53808570e-01 5.71659684e-01 -1.00609474e-02 3.12435627e-01 1.01964347e-01 2.67997593e-01 -1.31172049e+00 2.94757277e-01 7.76207149e-01 6.52609706e-01 -8.18207324e-01 3.82252514e-01 5.45878075e-02 -1.69941676e+00 3.15168887e-01 -3.36531818e-01 -5.89532740e-02 2.54475847e-02 7.68398702e-01 -9.66470361e-01 7.56419480e-01 -4.23934050e-02 6.32747829e-01 -1.03420734e+00 8.14465344e-01 1.03590153e-01 4.49794441e-01 -4.45284426e-01 -2.24343717e-01 2.40126878e-01 -5.74915349e-01 6.17085278e-01 1.39552867e+00 5.64072765e-02 -4.71113533e-01 1.85138538e-01 1.14841628e+00 -3.10494825e-02 3.83135825e-01 -1.26234746e+00 -6.79208040e-01 1.99315146e-01 1.39559841e+00 -1.11984456e+00 -4.71188463e-02 -1.66524991e-01 3.92438948e-01 5.67230403e-01 1.71462327e-01 -6.60052359e-01 -2.54727066e-01 6.24144614e-01 5.39559841e-01 -1.33248493e-01 -8.70315135e-01 -1.42650306e-01 -9.99295056e-01 -1.00137450e-01 -6.03690803e-01 2.94177175e-01 -3.95481139e-01 -1.35543346e+00 5.77173233e-01 2.31593400e-01 -1.08485365e+00 -4.25412238e-01 -7.71786571e-01 -7.84674525e-01 6.39609635e-01 -1.03142178e+00 -1.50195134e+00 -5.50522149e-01 4.99010235e-01 -1.14431843e-01 2.55190432e-01 7.59577394e-01 -1.68935433e-01 -2.78606594e-01 2.81758815e-01 5.27252965e-02 7.81585574e-02 2.38680765e-01 -1.73829341e+00 7.28793979e-01 6.48051202e-01 4.91710782e-01 8.41343582e-01 7.75101364e-01 -5.11866629e-01 -1.88914347e+00 -8.14745545e-01 7.29592085e-01 -6.36959910e-01 8.69948566e-01 -5.39285302e-01 -7.17269421e-01 3.37761164e-01 -3.71949941e-01 2.52074748e-01 6.39595211e-01 4.20838118e-01 -5.25809765e-01 1.49923220e-01 -7.78677225e-01 8.41591179e-01 1.36262989e+00 -9.61825728e-01 -2.08494812e-01 1.34517595e-01 2.01661333e-01 2.14406159e-02 -1.35516465e+00 4.52089906e-01 5.61820805e-01 -1.05545747e+00 1.01434350e+00 -5.69839776e-01 7.84392804e-02 -4.62808050e-02 -2.23425791e-01 -1.18082917e+00 -3.39103460e-01 -1.11126888e+00 -2.95961678e-01 1.18926847e+00 4.44150418e-01 -6.89095855e-01 8.39372754e-01 7.29655445e-01 8.93139094e-02 -8.36012363e-01 -5.77333331e-01 -1.00939417e+00 2.09089458e-01 -5.49239516e-01 6.80880725e-01 8.54139447e-01 3.96061212e-01 2.62034833e-01 1.02998734e-01 -2.84085631e-01 5.12780964e-01 1.73068732e-01 9.26469922e-01 -1.66451991e+00 -6.16351627e-02 -1.24430346e+00 -1.25686252e+00 -5.68476737e-01 2.98882753e-01 -1.33997893e+00 -4.92480308e-01 -1.63752651e+00 -8.09681714e-02 -2.47397527e-01 5.44124357e-02 -1.45153314e-01 -4.69589271e-02 1.93206906e-01 1.02319159e-01 -1.51915535e-01 -5.79100251e-01 6.02338314e-01 1.15211821e+00 1.72053296e-02 5.30599393e-02 -2.24543344e-02 -6.80727839e-01 7.38181651e-01 7.53236651e-01 5.58788441e-02 -5.30879378e-01 2.30377838e-01 4.74395186e-01 -2.27974012e-01 3.76513600e-01 -9.37683582e-01 1.58018619e-02 -9.48932841e-02 3.02451551e-01 -1.78997487e-01 1.18260592e-01 -7.36300528e-01 4.01083261e-01 3.21124882e-01 -3.76822174e-01 3.42454523e-01 -7.96039626e-02 6.44064903e-01 -2.47059271e-01 -3.76951881e-02 7.51780450e-01 1.10018238e-01 -4.81671870e-01 5.10240495e-01 8.06424171e-02 3.18819433e-01 1.06856143e+00 -6.06923044e-01 -3.37912142e-01 -4.52556252e-01 -7.09511518e-01 -1.45801470e-01 9.17757630e-01 2.35855177e-01 5.89048684e-01 -1.43777084e+00 -5.23698926e-01 2.18479276e-01 3.77913058e-01 -5.63303828e-01 6.64002597e-02 9.65455234e-01 -5.17169535e-01 2.96968877e-01 -3.63813192e-02 -6.72669172e-01 -1.14423954e+00 6.09975338e-01 1.01977184e-01 -3.79229337e-01 -6.08433306e-01 3.53147537e-01 1.89339191e-01 -3.00538212e-01 -2.09145352e-01 -4.88258839e-01 3.11088413e-02 7.61153698e-02 7.36256912e-02 6.52486801e-01 1.15501404e-01 -6.88823044e-01 -4.67597157e-01 7.68730223e-01 1.92691475e-01 1.23500876e-01 1.26852703e+00 4.49041352e-02 -4.01614085e-02 4.15330738e-01 1.45200288e+00 1.20135851e-01 -8.53597403e-01 1.04949787e-01 4.18808430e-01 -3.50032240e-01 -1.12656936e-01 -3.05058211e-01 -6.36470020e-01 8.03421974e-01 4.80747260e-02 1.27453709e+00 8.64848733e-01 2.13029906e-01 4.75486606e-01 3.99056934e-02 2.56967992e-01 -9.50932086e-01 1.77421886e-03 3.43156070e-01 8.98909569e-01 -1.06418765e+00 1.74460948e-01 -6.64898574e-01 -4.53817517e-01 1.21033144e+00 1.91355750e-01 -3.54266167e-01 8.76859665e-01 2.78455406e-01 -5.66836298e-01 -6.34164810e-01 -5.68415105e-01 -6.17607415e-01 9.12444293e-01 8.81020069e-01 6.14030659e-01 3.00311595e-01 -1.00734882e-01 -5.77651784e-02 -4.58045006e-01 -8.21684182e-01 2.71323591e-01 7.11948931e-01 -4.84646559e-01 -1.19062257e+00 1.82635739e-01 7.36555219e-01 -2.65163869e-01 8.11344236e-02 -9.07680333e-01 1.10905421e+00 -5.45066655e-01 6.09149158e-01 -2.38132671e-01 -5.38932741e-01 2.66074687e-01 -8.93337429e-02 6.87391400e-01 -4.54786628e-01 -2.68233359e-01 -2.18703344e-01 2.00684160e-01 -6.38441265e-01 -2.57050812e-01 -6.41292036e-01 -7.75342107e-01 -6.49888456e-01 -3.07024598e-01 1.38788551e-01 6.96903765e-01 7.01591551e-01 5.18067062e-01 2.34321818e-01 5.16699791e-01 -8.53034794e-01 -4.58991110e-01 -8.05182874e-01 -7.44462430e-01 5.52730441e-01 1.31762072e-01 -7.82557428e-01 -3.56059194e-01 -1.21084794e-01]
[7.01699161529541, 5.880551815032959]
0fc9b517-511a-451e-82ed-fbd8372e5a69
end-to-end-dereverberation-beamforming-and
2102.11525
null
https://arxiv.org/abs/2102.11525v1
https://arxiv.org/pdf/2102.11525v1.pdf
End-to-End Dereverberation, Beamforming, and Speech Recognition with Improved Numerical Stability and Advanced Frontend
Recently, the end-to-end approach has been successfully applied to multi-speaker speech separation and recognition in both single-channel and multichannel conditions. However, severe performance degradation is still observed in the reverberant and noisy scenarios, and there is still a large performance gap between anechoic and reverberant conditions. In this work, we focus on the multichannel multi-speaker reverberant condition, and propose to extend our previous framework for end-to-end dereverberation, beamforming, and speech recognition with improved numerical stability and advanced frontend subnetworks including voice activity detection like masks. The techniques significantly stabilize the end-to-end training process. The experiments on the spatialized wsj1-2mix corpus show that the proposed system achieves about 35% WER relative reduction compared to our conventional multi-channel E2E ASR system, and also obtains decent speech dereverberation and separation performance (SDR=12.5 dB) in the reverberant multi-speaker condition while trained only with the ASR criterion.
['Yanmin Qian', 'Reinhold Haeb-Umbach', 'Naoyuki Kamo', 'Tsubasa Ochiai', 'Keisuke Kinoshita', 'Marc Delcroix', 'Tomohiro Nakatani', 'Shinji Watanabe', 'Christoph Boeddeker', 'Wangyou Zhang']
2021-02-23
null
null
null
null
['speech-dereverberation']
['speech']
[ 1.84108235e-03 -5.06191432e-01 6.73000813e-01 3.76940034e-02 -1.34172428e+00 -6.34875476e-01 6.88122287e-02 -4.03997719e-01 -3.04215461e-01 6.08814597e-01 5.17956495e-01 -5.52451551e-01 1.90294404e-02 2.18222931e-01 -1.44281283e-01 -9.62839603e-01 3.16066816e-02 -2.50772178e-01 -2.08521351e-01 -2.27026016e-01 -3.71717453e-01 3.36454481e-01 -1.48252809e+00 3.92350182e-02 9.54990208e-01 9.65986371e-01 4.64449227e-01 1.36095774e+00 3.94357830e-01 2.35814214e-01 -1.15696442e+00 5.00950813e-02 1.02814175e-01 -5.57849348e-01 -7.16040730e-02 -4.50468175e-02 3.34650338e-01 -5.12904823e-02 -5.46391547e-01 1.02980983e+00 1.42143512e+00 2.92479366e-01 2.76930004e-01 -6.74496293e-01 -1.07486628e-01 5.94859719e-01 -3.91738176e-01 4.72375154e-01 3.50418866e-01 -4.84450087e-02 7.31258452e-01 -1.23367834e+00 -3.76488179e-01 1.22265232e+00 6.11473441e-01 3.39945465e-01 -8.31902981e-01 -9.53670800e-01 -1.38396353e-01 1.89865842e-01 -1.68610227e+00 -1.31836796e+00 5.14331698e-01 -3.45325507e-02 9.37129915e-01 8.49820316e-01 1.74820647e-01 9.12672698e-01 -3.67980152e-01 5.69972217e-01 1.07460082e+00 -5.54704607e-01 4.54576574e-02 -1.20248742e-01 1.29620209e-01 2.61885226e-02 -3.03617828e-02 3.50187600e-01 -6.60568655e-01 -8.28151703e-02 4.05732304e-01 -6.12447083e-01 -9.36213672e-01 7.14641511e-01 -1.15188348e+00 1.56334832e-01 4.06907238e-02 6.71905935e-01 -1.62334844e-01 -1.18503511e-01 1.45306990e-01 6.06612861e-01 7.16853619e-01 2.59064257e-01 -3.38469416e-01 -2.34401584e-01 -1.17105329e+00 -5.54132415e-03 9.28687036e-01 8.58030558e-01 -6.54904684e-03 8.65747392e-01 -2.38774151e-01 1.50231099e+00 4.98493820e-01 1.18901312e+00 5.12290120e-01 -4.79297340e-01 7.49063551e-01 -9.27716315e-01 2.29569674e-01 -5.62849402e-01 -2.57125884e-01 -1.24587893e+00 -9.14754272e-01 -8.63278806e-02 1.92260295e-01 -6.76605999e-01 -8.01050663e-01 1.47877109e+00 3.94917220e-01 5.90400338e-01 6.08827472e-01 1.12030303e+00 7.37654328e-01 9.82110858e-01 -5.05695641e-01 -7.68270969e-01 1.13121450e+00 -1.14345300e+00 -1.44970286e+00 -2.75287122e-01 1.78554691e-02 -1.45774484e+00 7.09765673e-01 5.91267228e-01 -1.22021937e+00 -7.80258119e-01 -1.04473352e+00 4.25954223e-01 1.49515107e-01 1.92011192e-01 -3.24246436e-01 1.44699168e+00 -1.21815884e+00 2.29271059e-03 -5.45783699e-01 2.08216026e-01 -3.13283533e-01 9.41868424e-02 -1.00444250e-01 -1.25276044e-01 -1.25720692e+00 5.88147223e-01 -1.59313291e-01 6.83067799e-01 -9.18968260e-01 -6.23645425e-01 -7.21382260e-01 2.40349621e-01 2.03115344e-01 -2.55217791e-01 1.39443552e+00 -5.99532127e-01 -1.91540551e+00 1.14857897e-01 -7.04325616e-01 -5.74763082e-02 4.25496787e-01 -5.53509653e-01 -1.43477058e+00 -1.90332830e-01 -3.49885434e-01 -5.86136043e-01 9.08979058e-01 -1.42903113e+00 -6.16828501e-01 -2.11825073e-01 -5.37744701e-01 6.81099296e-01 -3.13793451e-01 4.68805015e-01 -1.39976159e-01 -9.76794839e-01 4.31792587e-01 -5.82957625e-01 -6.68796450e-02 -7.27990746e-01 -5.48800826e-01 4.03975606e-01 7.06366360e-01 -1.26398349e+00 1.35596311e+00 -2.57907104e+00 -8.66126567e-02 1.27023548e-01 -1.88975617e-01 7.82947183e-01 -4.33501340e-02 2.20653877e-01 -2.00596377e-01 -2.34378397e-01 -9.43660140e-02 -6.49059296e-01 -1.21003464e-01 -3.44240487e-01 -1.79877579e-01 6.51530862e-01 -3.76408488e-01 -1.20023638e-02 -5.89626193e-01 -1.04092635e-01 3.38831961e-01 7.79650211e-01 -4.42700028e-01 6.37210667e-01 7.54560590e-01 6.12811327e-01 1.12081669e-01 6.73748136e-01 9.70698416e-01 5.67571998e-01 -1.48222581e-01 1.29623143e-02 -3.63313943e-01 3.19955319e-01 -1.56816554e+00 1.36074996e+00 -1.00410008e+00 8.05832326e-01 1.27619362e+00 -6.95108593e-01 8.51832032e-01 1.07128477e+00 -1.66236237e-01 -4.43985343e-01 8.28554779e-02 7.44361043e-01 4.33961898e-01 -5.20894825e-01 3.71599764e-01 -5.02864778e-01 2.93664992e-01 -7.98160769e-03 2.79171169e-02 -1.94652900e-01 -3.14379632e-01 -3.15706164e-01 8.31786990e-01 -6.05871797e-01 2.70942330e-01 -3.18666369e-01 9.19435978e-01 -1.03942370e+00 4.43287879e-01 6.33275092e-01 -2.32822970e-01 8.90755057e-01 -3.69571537e-01 7.13821113e-01 -6.22220397e-01 -1.22398150e+00 -1.37800187e-01 1.13675106e+00 1.37675996e-03 -8.96526054e-02 -7.97994852e-01 3.18362355e-01 -3.26880783e-01 9.80521202e-01 3.55712503e-01 1.68524027e-01 -6.84339583e-01 -7.88754046e-01 1.08272052e+00 1.55305639e-01 5.52483797e-01 -2.88503557e-01 4.48385626e-01 4.49379057e-01 -4.88968760e-01 -1.21448600e+00 -8.19380462e-01 1.76960871e-01 -2.31107816e-01 -3.43428493e-01 -1.16755700e+00 -8.31079781e-01 2.78579026e-01 5.34871697e-01 4.14722681e-01 -3.81693274e-01 1.95712727e-02 2.83951402e-01 -4.40265298e-01 -3.48628372e-01 -6.63168967e-01 -3.93139511e-01 4.94446695e-01 4.38307792e-01 -3.07674825e-01 -6.74317241e-01 -7.01116323e-01 8.46940100e-01 -4.72416520e-01 -4.54853058e-01 2.80377090e-01 8.68029237e-01 -1.03514425e-01 4.03049499e-01 9.88952339e-01 1.08575538e-01 9.14799094e-01 -2.80706346e-01 -4.04649526e-01 1.17674172e-01 -3.85659277e-01 -6.19939864e-01 8.72326910e-01 -4.63837981e-01 -1.75647974e+00 -4.55563843e-01 -8.74806166e-01 -1.65101483e-01 -3.35187405e-01 3.16159517e-01 -7.30443776e-01 7.25124478e-02 5.55092394e-01 3.91245753e-01 -2.02551126e-01 -7.60320365e-01 2.33154178e-01 1.70965493e+00 6.39883399e-01 -1.36850595e-01 8.76426697e-01 9.46311373e-03 -5.03306925e-01 -1.55482364e+00 -1.15880705e-01 -9.88331974e-01 2.72748601e-02 -5.27316257e-02 3.38720232e-01 -1.34930813e+00 -5.88328302e-01 9.30328369e-01 -1.18200874e+00 -1.83342263e-01 2.80333489e-01 1.17584753e+00 -8.42217356e-02 4.97146845e-01 -7.18497694e-01 -1.46371353e+00 -5.31668007e-01 -1.25570679e+00 6.71025515e-01 3.56687754e-02 -9.27829370e-02 -6.94003582e-01 -7.79874772e-02 5.53611636e-01 8.85039687e-01 -4.19094145e-01 3.05101633e-01 -5.10951459e-01 9.07226354e-02 -1.25460953e-01 3.52739573e-01 7.81318784e-01 3.89257878e-01 -5.53315461e-01 -1.68429494e+00 -5.75371087e-01 4.94511902e-01 1.44179508e-01 5.35342813e-01 7.10142076e-01 6.20254993e-01 -3.01966012e-01 -9.93678719e-02 5.78679919e-01 1.03390861e+00 6.22826457e-01 5.95515907e-01 -4.71028000e-01 4.52423394e-01 3.43048155e-01 4.16823775e-01 3.19087148e-01 -1.46443084e-01 5.85401475e-01 -6.46507088e-03 -5.76183200e-01 -5.45990050e-01 2.65757591e-01 5.85848451e-01 1.58387530e+00 1.91735014e-01 -9.56942558e-01 -5.97041368e-01 5.88854313e-01 -1.18294406e+00 -8.08327079e-01 -3.06820005e-01 2.46871114e+00 7.90096760e-01 -3.44303489e-01 -1.10671604e-02 6.45698667e-01 1.12959456e+00 3.68335992e-01 -3.58656555e-01 -3.32750082e-01 -4.25549865e-01 2.99941212e-01 5.59804142e-01 1.10449052e+00 -7.58815289e-01 5.99077940e-01 6.23901701e+00 1.17837465e+00 -1.22076952e+00 4.10763621e-01 6.05371356e-01 -3.22529465e-01 1.52092827e-02 -5.64125001e-01 -6.07819319e-01 1.54097870e-01 1.25621974e+00 -1.97367653e-01 6.44811988e-01 3.17304969e-01 6.72049701e-01 8.90997984e-03 -8.26989174e-01 1.40082729e+00 2.92636395e-01 -4.56375957e-01 -6.69428766e-01 -1.93328097e-01 4.15104955e-01 1.65579300e-02 2.71536857e-01 9.40070748e-02 -1.99833333e-01 -1.12880683e+00 8.18423092e-01 1.60685182e-02 1.13902640e+00 -6.75947726e-01 5.26932180e-01 4.06298876e-01 -1.33426571e+00 -2.26394162e-01 -2.14492977e-02 -5.65975495e-02 4.25529987e-01 1.02537727e+00 -9.90866840e-01 8.64792228e-01 3.50033462e-01 -3.68906744e-02 2.43818909e-01 1.30188239e+00 -2.71681666e-01 1.10559356e+00 -3.29805821e-01 1.11313006e-02 -1.68088153e-01 6.95944950e-02 1.05996847e+00 1.57309520e+00 8.10443521e-01 2.24556580e-01 -2.50767261e-01 2.96651930e-01 3.78623344e-02 2.37152795e-04 -1.47505462e-01 2.66737163e-01 6.01016462e-01 9.62150991e-01 -1.91332564e-01 -1.86503053e-01 -2.44212121e-01 1.07114661e+00 -6.28318250e-01 1.15173674e+00 -8.73994172e-01 -9.78155077e-01 1.01497495e+00 -2.29665861e-01 3.40405196e-01 -4.95884031e-01 -2.05975503e-01 -1.08202696e+00 4.93892394e-02 -1.17271245e+00 -1.98247626e-01 -6.08930409e-01 -8.00927818e-01 8.72094572e-01 -5.25518894e-01 -1.14940226e+00 -5.63044548e-02 -5.88955343e-01 -6.10565066e-01 1.46098137e+00 -1.57946527e+00 -5.47820330e-01 2.52841443e-01 6.71655715e-01 7.63704777e-01 -1.91628441e-01 8.70753884e-01 1.00212538e+00 -5.51881492e-01 1.01121628e+00 8.75406563e-01 -1.30499110e-01 7.70719230e-01 -1.05567944e+00 3.10773641e-01 1.29140437e+00 -2.50004292e-01 6.61073327e-01 1.21525574e+00 -1.55315086e-01 -1.33248627e+00 -9.50337350e-01 8.19224834e-01 2.47743115e-01 3.30154359e-01 -7.19823360e-01 -6.51223123e-01 1.46965414e-01 3.97647798e-01 -9.76750478e-02 9.01521742e-01 1.22971758e-01 -9.15019065e-02 -4.10993159e-01 -1.00021100e+00 6.39235198e-01 8.53164375e-01 -6.26960039e-01 -3.70824218e-01 2.19559342e-01 8.29206586e-01 -4.83881921e-01 -4.66876358e-01 1.28207773e-01 4.79410410e-01 -7.65609205e-01 1.01803255e+00 7.06098452e-02 -3.07860136e-01 -5.17641127e-01 -6.14175498e-01 -2.00514388e+00 -1.74561545e-01 -1.59360516e+00 1.46974906e-01 1.69868326e+00 7.64703214e-01 -9.25102770e-01 -3.28821763e-02 8.98868889e-02 -5.79796672e-01 6.88500190e-03 -1.23093045e+00 -1.21090865e+00 -3.53234485e-02 -5.22894025e-01 2.57376194e-01 5.18331647e-01 8.63675550e-02 3.11191291e-01 -7.61422873e-01 8.54096174e-01 6.08976066e-01 -3.28538239e-01 5.04267693e-01 -6.31839693e-01 -6.61484301e-01 -1.03163756e-01 1.29014924e-01 -1.67269838e+00 -1.66676752e-02 -5.04256606e-01 6.62486255e-01 -1.29974699e+00 -6.78553522e-01 -1.22409321e-01 -4.63658720e-01 -3.45780760e-01 -5.09632885e-01 -1.43964872e-01 8.82095546e-02 -2.77212471e-01 -9.64479446e-02 6.73251331e-01 1.09508324e+00 2.87922844e-03 -2.42403969e-01 5.20669043e-01 -6.59209788e-01 4.99558926e-01 4.64277238e-01 -2.21711621e-01 -3.08330059e-01 -4.57515240e-01 -6.35268331e-01 5.92733920e-01 -6.02541864e-02 -1.22702885e+00 2.89093971e-01 3.27513039e-01 2.89485021e-03 -4.00922507e-01 8.20911705e-01 -7.76184082e-01 1.83908343e-01 2.88585812e-01 -2.29682818e-01 -5.15485704e-01 3.19582134e-01 5.27227461e-01 -5.18153191e-01 2.24966824e-01 9.47769284e-01 1.87933266e-01 1.25959674e-02 -1.48439184e-01 -7.69528747e-01 -6.80334345e-02 5.31495452e-01 -1.44200698e-02 -2.20705628e-01 -7.63440192e-01 -8.04487765e-01 5.88501319e-02 -4.42966044e-01 1.25014275e-01 6.42566502e-01 -1.08067966e+00 -1.10619533e+00 2.82620907e-01 -4.86766756e-01 -3.77984911e-01 7.14487553e-01 9.24311578e-01 -1.04222231e-01 6.74851716e-01 3.78172547e-01 -4.77124363e-01 -1.74060023e+00 -1.34973088e-04 6.13188446e-01 2.12329984e-01 -1.48873776e-01 1.23626935e+00 2.53802180e-01 -1.08596735e-01 5.43036938e-01 -2.53503412e-01 -2.31589079e-02 -1.93785623e-01 7.88986504e-01 9.45243418e-01 4.41903442e-01 -8.19232583e-01 -2.93816626e-01 4.52295989e-01 3.61276448e-01 -7.19745040e-01 8.80877972e-01 -7.02791274e-01 1.23151928e-01 5.02243936e-01 1.37413621e+00 8.05929005e-01 -7.58330464e-01 -2.32239395e-01 -5.16676605e-01 -7.11577475e-01 4.26259071e-01 -1.00765753e+00 -7.22521842e-01 9.93222892e-01 7.89019108e-01 3.69491160e-01 1.32105350e+00 -3.95751864e-01 1.00581443e+00 2.25652590e-01 1.09701298e-01 -9.76958334e-01 -6.62863851e-02 6.30421281e-01 1.09115815e+00 -7.62576520e-01 -5.36104083e-01 -4.79158014e-01 -2.82836318e-01 8.35953295e-01 2.16307923e-01 4.75433350e-01 7.60381520e-01 6.55513465e-01 6.94829702e-01 5.38966596e-01 -2.17354596e-01 -2.06248507e-01 8.44400376e-02 6.37526512e-01 7.51338601e-01 2.68306673e-01 -1.37205929e-01 6.00797594e-01 -5.49350798e-01 -7.05559194e-01 3.67034435e-01 4.98004496e-01 -6.53775692e-01 -7.57646620e-01 -1.26516581e+00 1.82221755e-02 -8.15612912e-01 -5.14373064e-01 -2.41613394e-04 6.74922541e-02 -3.53812128e-01 2.08645105e+00 -1.44166321e-01 -3.61893147e-01 7.86425591e-01 2.38057435e-01 -2.04436611e-02 -3.43969285e-01 -7.76967525e-01 1.13185287e+00 5.10771990e-01 -1.25754431e-01 -1.66276172e-01 -5.72612047e-01 -1.00165117e+00 -2.52546221e-01 -8.84492159e-01 4.58569616e-01 8.97449672e-01 8.27369690e-01 1.90259382e-01 1.10485220e+00 1.07312357e+00 -6.94513500e-01 -6.42251432e-01 -1.35651684e+00 -1.06289637e+00 -8.83449242e-02 1.12621629e+00 -1.94545120e-01 -9.39584970e-01 -1.24412745e-01]
[14.962642669677734, 5.874068260192871]
3a1dc508-0bdd-45fa-8ff2-8ce4352c8351
findings-of-the-nlp4if-2019-shared-task-on
1910.09982
null
https://arxiv.org/abs/1910.09982v1
https://arxiv.org/pdf/1910.09982v1.pdf
Findings of the NLP4IF-2019 Shared Task on Fine-Grained Propaganda Detection
We present the shared task on Fine-Grained Propaganda Detection, which was organized as part of the NLP4IF workshop at EMNLP-IJCNLP 2019. There were two subtasks. FLC is a fragment-level task that asks for the identification of propagandist text fragments in a news article and also for the prediction of the specific propaganda technique used in each such fragment (18-way classification task). SLC is a sentence-level binary classification task asking to detect the sentences that contain propaganda. A total of 12 teams submitted systems for the FLC task, 25 teams did so for the SLC task, and 14 teams eventually submitted a system description paper. For both subtasks, most systems managed to beat the baseline by a sizable margin. The leaderboard and the data from the competition are available at http://propaganda.qcri.org/nlp4if-shared-task/.
['Alberto Barrón-Cedeño', 'Giovanni Da San Martino', 'Preslav Nakov']
2019-10-20
findings-of-the-nlp4if-2019-shared-task-on-1
https://aclanthology.org/D19-5024
https://aclanthology.org/D19-5024.pdf
ws-2019-11
['propaganda-detection']
['natural-language-processing']
[ 2.53244311e-01 9.39965546e-02 -4.37480539e-01 -1.97710425e-01 -1.15993690e+00 -7.70985961e-01 1.31236887e+00 5.12894809e-01 -4.55886096e-01 7.65845656e-01 8.86227846e-01 -6.02412045e-01 7.43060187e-02 -4.14741486e-01 -8.69124770e-01 -5.63927233e-01 -4.07888703e-02 4.47874993e-01 1.20150208e-01 -2.76941568e-01 9.04618561e-01 1.06485270e-01 -6.53213322e-01 1.32757854e+00 8.97290349e-01 5.64267039e-01 3.43833352e-03 1.04679561e+00 -3.12447790e-02 1.69434464e+00 -1.10867143e+00 -4.22451675e-01 -1.53324381e-01 -6.59457862e-01 -1.34568942e+00 -3.47950906e-01 6.85777247e-01 2.57799011e-02 -1.97429001e-01 9.52989221e-01 3.10447365e-01 -1.68096930e-01 8.69398952e-01 -8.43382061e-01 -6.11828089e-01 1.33411443e+00 -5.70216179e-01 7.17814863e-01 7.42606819e-01 -1.81688070e-01 1.08920240e+00 -8.48395348e-01 1.04483831e+00 1.64414763e+00 5.62417924e-01 3.43030304e-01 -1.20402336e+00 -5.12166142e-01 -8.84363279e-02 3.56236815e-01 -9.29107845e-01 -4.37661380e-01 5.67902088e-01 -9.99356151e-01 1.13364267e+00 3.76965880e-01 3.91181022e-01 1.65646052e+00 7.04601824e-01 1.16175783e+00 1.26974142e+00 -5.89295864e-01 1.10016419e-02 1.86054587e-01 7.54444122e-01 8.27201247e-01 3.91348526e-02 -9.15950239e-02 -7.33903706e-01 -5.40217936e-01 1.62539974e-01 -7.67955482e-01 -2.03348175e-01 3.73712897e-01 -1.28848660e+00 1.32919657e+00 4.31721985e-01 7.62239933e-01 -1.92339048e-01 1.49603471e-01 8.07857454e-01 4.83094156e-01 1.04655242e+00 6.57877684e-01 -3.83330762e-01 -2.02835575e-01 -9.09033775e-01 9.69979048e-01 1.05392897e+00 4.12268549e-01 1.35922745e-01 -4.22102064e-01 -6.12727582e-01 6.75763190e-01 -1.33562520e-01 5.15985191e-01 3.30915958e-01 -5.80685914e-01 1.05705512e+00 2.03002185e-01 1.42509133e-01 -1.09520781e+00 -8.08874786e-01 -3.78743529e-01 -5.55047393e-01 -1.35982305e-01 5.15385330e-01 -3.82198870e-01 -9.67788935e-01 1.37748063e+00 3.27587277e-02 -4.39487755e-01 -3.20917606e-01 7.33093441e-01 9.99889970e-01 9.86067176e-01 -2.95047350e-02 -3.87708008e-01 1.42165518e+00 -1.33463347e+00 -8.58407319e-01 -1.16803035e-01 1.29133260e+00 -1.03657329e+00 6.40678167e-01 3.81680399e-01 -8.49873304e-01 -2.08930120e-01 -8.57102871e-01 -7.90605769e-02 -1.21347487e-01 -8.11180100e-02 3.19350988e-01 2.59561419e-01 -5.50296307e-01 4.19747710e-01 -5.20002663e-01 -3.33953440e-01 3.17493051e-01 -6.04623675e-01 -2.33989716e-01 5.58470115e-02 -1.35859978e+00 1.43267858e+00 5.18812358e-01 -1.13953680e-01 -1.14600682e+00 -6.72385871e-01 -6.92469001e-01 -1.55716315e-01 2.66370118e-01 -1.80780619e-01 1.37464690e+00 -9.12229598e-01 -8.82529616e-01 1.03015280e+00 1.51894957e-01 -8.45261395e-01 5.65675914e-01 -5.59815764e-01 -5.42522252e-01 -5.35418950e-02 7.60217071e-01 1.51199698e-01 8.55588496e-01 -7.71271288e-01 -6.02845430e-01 -1.27880946e-01 -6.41630068e-02 -1.51815057e-01 3.47801805e-01 8.29111278e-01 3.60832423e-01 -8.65151465e-01 -3.99957418e-01 -7.47104704e-01 2.25627348e-01 -9.05015349e-01 -9.58522618e-01 -7.20668733e-01 5.63464284e-01 -1.00307477e+00 1.22535944e+00 -1.98331738e+00 3.16414893e-01 -1.63689628e-01 2.46776074e-01 2.90338099e-01 -3.79781753e-01 7.62718022e-01 -1.63086355e-01 5.19717157e-01 2.63277143e-01 1.19702583e-02 -9.24371481e-02 -5.12404203e-01 -7.56452918e-01 7.86409199e-01 1.99169353e-01 6.38344347e-01 -9.15405273e-01 -4.72110987e-01 -3.37466329e-01 -1.15054548e-01 -2.85904080e-01 -1.16103910e-01 -5.46296358e-01 2.19033688e-01 -5.43395758e-01 1.60492882e-01 3.70642424e-01 -3.15932840e-01 -1.90970674e-01 1.82500884e-01 -4.51899081e-01 1.01436651e+00 -4.61189270e-01 1.41784275e+00 -1.45429209e-01 8.67357373e-01 3.52453142e-01 -7.38753438e-01 3.89671594e-01 3.65858287e-01 1.22987740e-02 -5.22101939e-01 3.18088531e-01 2.11782783e-01 1.57144576e-01 -3.23997051e-01 5.02897561e-01 -1.10919289e-01 -6.32763803e-01 8.77452910e-01 4.79261130e-02 -5.08059561e-02 6.62261784e-01 8.24159801e-01 1.67663121e+00 -1.77897245e-01 6.50968850e-01 -5.60470879e-01 5.03791809e-01 7.13349462e-01 5.32103539e-01 1.26428103e+00 -2.91874170e-01 3.02945197e-01 7.41060674e-01 -9.25367415e-01 -6.70628428e-01 -8.80523145e-01 -3.91661555e-01 1.35633302e+00 -3.07015479e-01 -9.23882723e-01 -6.25728250e-01 -1.06934202e+00 -1.19259007e-01 1.09085655e+00 -1.03845048e+00 3.61815304e-01 -7.67114699e-01 -5.89665532e-01 8.32471132e-01 2.43666116e-02 3.41480672e-01 -1.43235862e+00 -4.22902554e-01 3.19730282e-01 -5.53300679e-01 -7.09997535e-01 -3.87728721e-01 5.61091483e-01 -2.22612619e-01 -1.22690713e+00 -4.50238436e-01 -7.13128865e-01 1.25711560e-01 3.45119201e-02 9.64208305e-01 1.04387634e-01 -4.41540301e-01 -4.97807533e-01 -8.19402874e-01 -5.56526959e-01 -1.11539710e+00 2.59220362e-01 -2.82941461e-01 -5.84380150e-01 8.79256353e-02 1.89314350e-01 2.12521613e-01 -2.21324474e-01 -3.46803576e-01 4.68770236e-01 3.31646711e-01 9.59979951e-01 -9.02277902e-02 -1.20876499e-01 5.88814318e-01 -1.31091607e+00 1.05524755e+00 -3.99997979e-01 -3.23725671e-01 2.51066178e-01 -6.91453665e-02 -6.65998133e-03 6.68905318e-01 1.92906987e-02 -1.03055859e+00 -2.91263819e-01 -1.11324750e-01 4.73885596e-01 -2.55274415e-01 7.20112264e-01 4.50656772e-01 5.38843989e-01 1.47769415e+00 -1.38551667e-01 -3.27869028e-01 -6.23293996e-01 5.47292590e-01 7.33175397e-01 3.30955029e-01 -3.30348849e-01 7.74070561e-01 -8.66582617e-02 -6.51654363e-01 -6.56211793e-01 -1.66234112e+00 -5.77011585e-01 -4.38438654e-01 7.72225065e-03 1.09246528e+00 -8.49572778e-01 -9.61782411e-02 5.02940118e-01 -1.81142545e+00 -4.26748931e-01 7.06225336e-02 2.94177473e-01 -4.29249793e-01 4.62965146e-02 -1.20884550e+00 -7.30153620e-01 -5.67615688e-01 -6.66538239e-01 7.97512472e-01 -3.83003086e-01 -5.71006358e-01 -8.25938165e-01 5.11589229e-01 7.56572187e-01 1.00877911e-01 1.97017044e-01 1.27924263e+00 -1.11305892e+00 1.22102425e-01 -2.95512259e-01 -1.70059100e-01 6.30498379e-02 -1.58596084e-01 -1.31378829e-01 -6.31516695e-01 -7.99829885e-02 2.40434352e-02 -8.67556214e-01 1.34515035e+00 2.95544147e-01 6.10984087e-01 -9.05026317e-01 -3.79500419e-01 -2.08625674e-01 6.40664458e-01 -4.77347188e-02 2.96285391e-01 5.87323725e-01 4.11506414e-01 5.32126069e-01 6.24086499e-01 2.23881841e-01 -6.27513826e-02 7.09704936e-01 1.66436642e-01 2.15596631e-01 -2.93205798e-01 -2.11921692e-01 7.40724206e-01 6.22481346e-01 6.37775809e-02 -5.68958342e-01 -1.21432829e+00 5.03094554e-01 -1.98883176e+00 -1.64892113e+00 -7.17615068e-01 1.39865661e+00 1.05970478e+00 2.32840389e-01 2.17860207e-01 2.94252802e-02 8.15720201e-01 6.40382946e-01 1.97418571e-01 -1.02070475e+00 -1.42470151e-01 -3.43902856e-02 1.17886171e-01 9.54175532e-01 -1.66134179e+00 1.20932376e+00 6.39135933e+00 1.31257105e+00 -6.38220787e-01 5.12528658e-01 4.23882514e-01 -4.08466488e-01 -1.49751902e-01 -1.63782269e-01 -9.88894880e-01 5.17957091e-01 1.21599019e+00 -2.97362983e-01 2.61291325e-01 7.24081874e-01 2.43368804e-01 -2.17088625e-01 -8.76353145e-01 3.00381333e-01 3.97128582e-01 -1.86310494e+00 6.73463196e-02 -4.93327938e-02 8.67522657e-01 5.35023093e-01 -3.29086304e-01 4.97487366e-01 5.15820384e-01 -1.12616420e+00 1.25173938e+00 1.95572544e-02 1.18613377e-01 -4.98639464e-01 7.89843798e-01 1.02633250e+00 -4.17955995e-01 5.91547936e-02 -1.66406557e-01 -3.93678457e-01 6.27693474e-01 8.80791783e-01 -9.90295231e-01 3.45187515e-01 4.81543183e-01 6.87771678e-01 -7.03422844e-01 7.09311008e-01 -8.19546580e-01 1.11069465e+00 6.78621083e-02 -4.62314516e-01 5.59138596e-01 4.35806036e-01 9.68746722e-01 1.82636702e+00 -2.11321503e-01 -8.90301317e-02 3.80209893e-01 7.17697859e-01 -1.93731308e-01 2.27868706e-01 -3.52477521e-01 -3.80692512e-01 2.46571735e-01 1.27127409e+00 -5.63124537e-01 -3.54055434e-01 1.80593818e-01 7.88201332e-01 6.16857946e-01 -3.33822658e-03 -8.99825573e-01 -4.08935070e-01 7.34418491e-03 2.64185548e-01 3.87605242e-02 -8.70570634e-03 -3.30881864e-01 -1.09915912e+00 -2.07546368e-01 -7.83019722e-01 4.65380132e-01 -4.92595196e-01 -1.50574911e+00 7.03208923e-01 -9.59189981e-02 -3.30602050e-01 -5.21179020e-01 -5.37956655e-01 -8.91150653e-01 7.05436289e-01 -6.79724157e-01 -1.07128859e+00 2.33846009e-01 3.35992545e-01 7.91119933e-01 -2.32093483e-01 8.10742438e-01 -1.59140736e-01 -3.71388644e-01 7.84755126e-02 4.57152352e-03 4.38560367e-01 1.00794125e+00 -1.08593225e+00 4.46551293e-01 8.74755025e-01 3.82732183e-01 5.77517927e-01 1.02315056e+00 -1.03621197e+00 -7.81435490e-01 -1.14940345e+00 1.74325657e+00 -7.54395604e-01 1.26328719e+00 -6.69628203e-01 -5.91840565e-01 5.42583406e-01 5.03330946e-01 -5.08168578e-01 4.92091119e-01 6.55470967e-01 -8.36799026e-01 6.54618561e-01 -7.99660802e-01 2.51152575e-01 9.39799547e-01 -4.88915741e-01 -1.13226962e+00 1.38999820e+00 5.52105308e-01 -4.56713736e-01 -3.04034323e-01 -1.00869045e-01 3.17434669e-01 -5.86974502e-01 3.33227247e-01 -9.19953167e-01 1.14951801e+00 8.43931884e-02 -5.46292327e-02 -1.36974752e+00 -8.64139020e-01 -3.63316894e-01 -1.49881635e-02 1.00610554e+00 8.35551918e-01 -1.80305794e-01 2.50158221e-01 -2.93687284e-01 -2.40059391e-01 -3.36904109e-01 -1.09931111e+00 -9.59554970e-01 5.82819700e-01 -3.65713835e-01 -5.38943768e-01 8.98353279e-01 4.92963672e-01 1.13631856e+00 -4.78247702e-01 -2.43262649e-01 5.99077582e-01 1.51138961e-01 5.14321208e-01 -8.73766661e-01 -2.95836210e-01 -6.69073761e-01 -1.47425652e-01 -7.29953349e-01 3.99207205e-01 -1.36281753e+00 4.22894716e-01 -1.82690787e+00 6.26167715e-01 2.15031654e-02 1.41106471e-01 6.80751383e-01 -1.33120134e-01 7.65350237e-02 4.66621876e-01 4.12775159e-01 -8.79198253e-01 1.39416799e-01 7.46844232e-01 -4.99264538e-01 -1.13909900e-01 -1.97635964e-01 -7.51711309e-01 6.91996813e-01 8.15342367e-01 -8.54107320e-01 2.19822288e-01 -6.09746933e-01 2.62595922e-01 -7.04415664e-02 5.65381885e-01 -6.35300517e-01 1.99061185e-01 -2.10030988e-01 3.31451178e-01 -9.07295227e-01 -6.66447654e-02 2.92093188e-01 -2.12179631e-01 8.30183089e-01 -6.77313149e-01 -3.15226853e-01 3.55849676e-02 4.93863404e-01 -8.62954408e-02 -3.77254009e-01 6.38868749e-01 -3.31063531e-02 -1.74013257e-01 -3.82806420e-01 -1.14558411e+00 4.60178375e-01 9.15239930e-01 6.39300704e-01 -1.32174432e+00 -1.15974054e-01 -6.79482579e-01 1.05231106e-01 -6.66225031e-02 5.64493835e-01 7.14250803e-02 -9.88734484e-01 -1.49982691e+00 -3.66562665e-01 -4.35926160e-03 -5.80370128e-01 8.18916503e-03 1.11676323e+00 -6.66973591e-01 9.76700664e-01 1.18001550e-01 -1.65903345e-01 -1.40206873e+00 4.83628720e-01 -1.51049241e-01 -7.27675378e-01 -7.77395964e-01 1.11490417e+00 1.99510023e-01 -1.99837372e-01 -1.67236459e-02 1.22221440e-01 -4.50089611e-02 8.39614868e-02 1.13640821e+00 4.10583973e-01 7.60855004e-02 -6.51798427e-01 -7.11325288e-01 -1.72548547e-01 -7.13846684e-01 -2.43278906e-01 1.42317724e+00 4.47416812e-01 -3.51879328e-01 5.20391762e-01 1.04057014e+00 2.92324930e-01 -4.64830756e-01 -1.77632958e-01 3.25566769e-01 -1.66422632e-02 2.45602116e-01 -1.64045870e+00 -1.85835779e-01 6.56740248e-01 -2.37236142e-01 5.07350385e-01 2.14590386e-01 4.17740583e-01 4.95229930e-01 3.94271582e-01 3.59132171e-01 -1.22170448e+00 1.54720554e-02 1.22031724e+00 1.72953713e+00 -8.14006865e-01 2.74112165e-01 -5.25686741e-01 -5.19687593e-01 9.61228669e-01 1.79276764e-01 -3.32404166e-01 4.13298786e-01 2.44112641e-01 -5.62612265e-02 -5.77294827e-01 -1.10121930e+00 2.77311325e-01 5.04749179e-01 2.20324218e-01 8.07685375e-01 3.05522501e-01 -9.39819336e-01 7.18485177e-01 -3.30861926e-01 -3.83811176e-01 4.92827654e-01 9.44375515e-01 -1.00883019e+00 -6.54278994e-01 -4.10639107e-01 5.39220870e-01 -5.96652567e-01 -3.60347360e-01 -1.23327732e+00 6.06004477e-01 6.48189187e-02 1.35396135e+00 -1.99086368e-01 -4.97710288e-01 3.83072682e-02 -3.31047736e-03 5.69591939e-01 -9.82451200e-01 -1.34397602e+00 1.95198968e-01 9.73449647e-01 -7.08135962e-01 -1.68773919e-01 -7.65518486e-01 -1.10268307e+00 -6.64820910e-01 -3.95075351e-01 6.80430532e-01 5.01668751e-01 1.07188177e+00 -1.38306722e-01 3.60915631e-01 3.66680145e-01 -7.07738161e-01 -8.98211718e-01 -1.51947546e+00 -5.64987004e-01 2.31350958e-01 2.74666309e-01 -3.49672854e-01 -6.82924211e-01 1.32205253e-02]
[8.470013618469238, 10.656709671020508]
59348bec-2539-4c71-b4f6-16f93681c4cf
intersectionality-and-testimonial-injustice
2306.13675
null
https://arxiv.org/abs/2306.13675v1
https://arxiv.org/pdf/2306.13675v1.pdf
Intersectionality and Testimonial Injustice in Medical Records
Detecting testimonial injustice is an essential element of addressing inequities and promoting inclusive healthcare practices, many of which are life-critical. However, using a single demographic factor to detect testimonial injustice does not fully encompass the nuanced identities that contribute to a patient's experience. Further, some injustices may only be evident when examining the nuances that arise through the lens of intersectionality. Ignoring such injustices can result in poor quality of care or life-endangering events. Thus, considering intersectionality could result in more accurate classifications and just decisions. To illustrate this, we use real-world medical data to determine whether medical records exhibit words that could lead to testimonial injustice, employ fairness metrics (e.g. demographic parity, differential intersectional fairness, and subgroup fairness) to assess the severity to which subgroups are experiencing testimonial injustice, and analyze how the intersectionality of demographic features (e.g. gender and race) make a difference in uncovering testimonial injustice. From our analysis, we found that with intersectionality we can better see disparities in how subgroups are treated and there are differences in how someone is treated based on the intersection of their demographic attributes. This has not been previously studied in clinical records, nor has it been proven through empirical study.
['Lu Cheng', 'Bhuvani Shah', 'Kenya S. Andrews']
2023-06-20
null
null
null
null
['fairness', 'fairness']
['computer-vision', 'miscellaneous']
[ 1.85192198e-01 3.24868441e-01 -5.75590730e-01 -3.33379358e-01 -4.49946463e-01 -5.94938874e-01 4.65763837e-01 1.01587570e+00 -5.05110443e-01 6.01101637e-01 1.18319428e+00 -9.49154496e-01 -6.41267955e-01 -6.24333858e-01 -4.17766184e-01 -4.79066283e-01 3.34766686e-01 -4.41166200e-02 -7.47267962e-01 5.20572551e-02 6.84722364e-01 2.66819358e-01 -8.70955825e-01 3.97952825e-01 9.68593955e-01 4.66214418e-01 -6.59972489e-01 1.55831575e-01 6.68269098e-02 1.24744129e+00 -5.74838400e-01 -7.30563641e-01 1.86680883e-01 -6.64200962e-01 -5.01858354e-01 -4.57621247e-01 5.99519551e-01 -7.94664443e-01 -2.43894711e-01 1.01989412e+00 7.01371491e-01 -3.14836979e-01 9.14001644e-01 -9.89746749e-01 -7.38869309e-01 9.21767592e-01 -3.31940800e-01 3.49700212e-01 6.59168065e-01 3.81661326e-01 8.60139251e-01 -2.32993558e-01 6.84543669e-01 1.43234992e+00 6.84568048e-01 4.77662921e-01 -1.28881788e+00 -1.04114485e+00 -3.12186390e-01 2.91894916e-02 -8.46870780e-01 -8.23380530e-01 5.71893275e-01 -7.64206111e-01 1.00038402e-01 4.61466938e-01 5.74664533e-01 9.40533578e-01 4.99785274e-01 -1.29472658e-01 1.37964714e+00 -7.87656531e-02 1.42271340e-01 1.07692167e-01 3.68766397e-01 3.58802170e-01 1.02921855e+00 4.66488935e-02 -3.17594260e-01 -8.88486505e-01 3.28217566e-01 3.49671990e-01 -4.28809911e-01 1.28873408e-01 -1.28849113e+00 9.46053326e-01 2.49570057e-01 2.83873603e-02 -4.09626901e-01 -1.11730054e-01 6.68689191e-01 1.75902963e-01 1.15908287e-01 8.61483693e-01 -1.30248563e-02 -3.16502273e-01 -8.03639829e-01 1.69822782e-01 8.37383807e-01 -1.03172190e-01 6.80380538e-02 -3.66029918e-01 -3.22750479e-01 5.05871117e-01 -1.65185824e-01 4.07513469e-01 1.00443035e-01 -1.42889178e+00 4.99610811e-01 5.19866705e-01 1.12141810e-01 -1.33205676e+00 -3.45790595e-01 -1.29007697e-01 -4.76189017e-01 3.91643792e-01 7.64457107e-01 -2.87445158e-01 -2.93685466e-01 1.88330162e+00 1.01338122e-02 -3.49865615e-01 2.35191770e-02 9.39309537e-01 7.78972805e-01 -1.52352229e-01 4.96764123e-01 -2.68721044e-01 1.70905554e+00 1.69843405e-01 -8.05365086e-01 -1.41138852e-01 8.90489936e-01 -7.74203718e-01 8.87053490e-01 1.46313220e-01 -1.21420574e+00 1.78186595e-01 -7.67814577e-01 1.33565776e-02 6.31046307e-04 -8.20408940e-01 3.89101565e-01 1.00131071e+00 -2.84216404e-01 7.56453156e-01 -1.39171243e-01 -3.95286262e-01 7.58625150e-01 -2.02395901e-01 -3.11341286e-01 -7.77867511e-02 -1.06739902e+00 1.12496364e+00 -1.16595179e-01 -3.38545382e-01 -4.15289938e-01 -1.10810876e+00 -6.87376320e-01 1.85584500e-01 6.70960665e-01 -8.20108891e-01 6.08209610e-01 -9.51484799e-01 -3.11021119e-01 9.78235960e-01 -6.79882662e-03 -2.29544699e-01 6.29308283e-01 2.09574580e-01 -5.20873249e-01 2.11012095e-01 3.51918161e-01 1.29410774e-01 2.29528889e-01 -1.13457370e+00 -2.38492683e-01 -6.54014230e-01 2.95888990e-01 1.30769148e-01 -3.71586591e-01 5.10424316e-01 1.08656061e+00 -5.38159728e-01 -2.08372310e-01 -4.90760654e-01 -5.98329380e-02 2.38564402e-01 -3.30632865e-01 2.24197641e-01 4.09272462e-01 -7.62378931e-01 1.26621735e+00 -2.29344153e+00 -8.12497199e-01 1.57065421e-01 5.57128847e-01 1.22132525e-01 2.53595591e-01 4.65291232e-01 2.03309208e-01 7.72073209e-01 -8.25785995e-02 3.33738625e-01 1.10828854e-01 4.24601398e-02 -1.42554432e-01 8.43376160e-01 -2.00554550e-01 8.20829093e-01 -1.09843183e+00 -7.22848058e-01 3.97393405e-02 3.91870767e-01 -9.42028165e-01 -4.01341289e-01 5.33841550e-01 1.09300621e-01 -7.46055692e-02 6.66810274e-01 5.63078642e-01 9.32618976e-02 2.93390840e-01 -1.44215718e-01 -1.38519198e-01 7.67126799e-01 -4.80918676e-01 6.68491185e-01 1.46982623e-02 7.56792665e-01 2.70441502e-01 -7.28142321e-01 6.66474700e-01 3.53682876e-01 2.75115669e-01 -5.64793825e-01 3.26554745e-01 3.42500716e-01 7.53212571e-01 -6.50189877e-01 5.10513246e-01 -7.67132580e-01 -1.32320449e-01 7.75493562e-01 -7.49586821e-01 4.30850983e-02 -3.74883890e-01 3.01745355e-01 1.20181727e+00 -8.48972738e-01 3.75745982e-01 -5.20725310e-01 -2.11898863e-01 1.01358965e-01 9.67519641e-01 6.88697696e-01 -7.20546305e-01 6.70735419e-01 9.09642696e-01 -3.86912040e-02 -8.97665620e-01 -1.12100124e+00 -4.98405665e-01 3.89130592e-01 1.59705237e-01 5.16320986e-04 -4.04075146e-01 -6.13664746e-01 4.30385679e-01 1.30312669e+00 -6.61787391e-01 -6.32096112e-01 -3.95646840e-02 -7.16768444e-01 8.83431852e-01 2.32199982e-01 1.73882201e-01 -7.75121391e-01 -1.36700296e+00 -2.65351295e-01 -2.12394625e-01 -9.43532050e-01 -6.14417195e-01 -4.81264204e-01 -5.32439649e-01 -1.60946131e+00 -3.35229188e-01 -1.42055944e-01 6.00478292e-01 2.59918422e-01 9.83499885e-01 4.19697732e-01 -5.47811151e-01 4.03164685e-01 -2.12276921e-01 -6.80957913e-01 -8.13325763e-01 -7.44170070e-01 -1.18619971e-01 -3.36693525e-01 6.58706486e-01 -4.64779019e-01 -9.96268868e-01 -1.23000793e-01 -9.71394598e-01 -2.20353723e-01 3.87261838e-01 5.73961079e-01 -2.18336299e-01 -4.09113944e-01 9.27733600e-01 -1.31522703e+00 1.01162279e+00 -1.06683159e+00 3.49569261e-01 9.36760753e-02 -9.84947383e-01 -3.91500086e-01 4.79408056e-01 -4.74748969e-01 -9.34416056e-01 -1.08188677e+00 2.74947405e-01 1.25913993e-01 -2.19609663e-01 6.08777344e-01 2.55170524e-01 1.49677128e-01 7.28641629e-01 -5.22475421e-01 5.04381597e-01 -8.30167979e-02 -1.09913938e-01 9.01152670e-01 2.49830350e-01 -6.33232296e-01 1.66926250e-01 7.95731366e-01 -2.57278651e-01 -3.49313617e-01 -9.59205210e-01 -2.29899481e-01 1.86855912e-01 -7.66626075e-02 9.15643632e-01 -6.70541286e-01 -1.21181881e+00 -3.34165096e-01 -7.54275560e-01 -2.42703468e-01 -4.02331442e-01 9.50868785e-01 -1.81193486e-01 5.36076784e-01 -5.20874739e-01 -7.48749912e-01 -2.65964359e-01 -8.91376376e-01 1.10990144e-01 -9.10469443e-02 -1.05168033e+00 -8.87337923e-01 1.19721211e-01 7.38023818e-01 3.47132087e-01 7.07978964e-01 1.30677199e+00 -9.06071782e-01 1.12563416e-01 -2.38283351e-01 -2.69274414e-01 5.74701503e-02 5.46582222e-01 -2.30187946e-03 -7.31291294e-01 -2.20827200e-02 2.99666107e-01 -4.28840786e-01 4.50956970e-01 4.04636055e-01 7.36903727e-01 -8.51598263e-01 -1.75464362e-01 3.45641315e-01 1.26286399e+00 3.58204097e-01 5.03644109e-01 1.69747531e-01 5.25710464e-01 1.21079350e+00 4.53606665e-01 6.36003971e-01 5.34762204e-01 3.07175964e-01 3.02558750e-01 3.27866487e-02 -3.68321203e-02 -3.50214481e-01 3.89537960e-01 2.26324443e-02 9.65417475e-02 2.03858316e-01 -1.13011718e+00 6.18468344e-01 -1.19662845e+00 -1.45914245e+00 -2.07255095e-01 2.21381950e+00 7.11438000e-01 1.41574353e-01 1.83292240e-01 1.29375473e-01 7.11762130e-01 1.89464465e-01 -4.79354978e-01 -1.14463389e+00 -7.14571849e-02 -1.46420985e-01 5.95928788e-01 3.72293383e-01 -1.41384050e-01 6.19868971e-02 6.53652811e+00 2.15248391e-01 -9.76132810e-01 4.47386019e-02 1.04582894e+00 -5.86906910e-01 -1.08275902e+00 1.05968259e-01 -5.45654539e-03 6.94101572e-01 6.19560003e-01 -5.71913421e-01 1.35554150e-01 4.12812173e-01 7.80869603e-01 -2.38290817e-01 -1.16424346e+00 5.38564742e-01 2.03892067e-01 -1.02048254e+00 -1.34319916e-01 3.13505113e-01 4.11542773e-01 -5.96733809e-01 3.14362526e-01 -1.80554420e-01 4.86616760e-01 -1.37374687e+00 9.40316498e-01 2.00275704e-01 8.95029724e-01 -6.78451777e-01 8.16026092e-01 8.57246816e-02 -5.66374622e-02 -3.44401836e-01 -3.23518753e-01 -4.86865997e-01 2.04336897e-01 7.84185588e-01 -8.28136325e-01 7.53830895e-02 4.51675624e-01 2.44458839e-01 -3.92529229e-03 8.13242793e-01 9.06473696e-02 7.19481528e-01 1.01855621e-01 2.15523571e-01 8.38781297e-02 1.23910839e-02 6.18325830e-01 1.01807547e+00 2.73437262e-01 4.43920940e-01 -6.12451851e-01 1.05422449e+00 -1.98323146e-01 1.80184320e-01 -7.36779928e-01 -4.07544464e-01 8.67464185e-01 9.91381288e-01 -5.12872934e-01 -3.75543594e-01 -4.83311445e-01 2.31062651e-01 -8.46700668e-02 1.67875022e-01 -3.35285634e-01 -1.99236888e-02 9.98840094e-01 4.77041066e-01 -4.94984567e-01 3.05912018e-01 -7.93659806e-01 -8.39696407e-01 -3.27434242e-01 -1.14977419e+00 6.07393801e-01 -4.85961795e-01 -1.28920293e+00 -3.31691116e-01 -3.32302712e-02 -6.57575250e-01 1.85090676e-01 -1.62965283e-01 -6.63974226e-01 8.78077567e-01 -1.16293037e+00 -6.62666678e-01 -4.75718416e-02 2.54444212e-01 -1.98727876e-01 4.08079058e-01 6.25800967e-01 1.61417708e-01 -1.59777328e-01 7.93222904e-01 -2.78506130e-01 7.38582015e-02 1.25664902e+00 -8.64992678e-01 -2.47304723e-01 4.48755443e-01 -4.11362916e-01 8.36555958e-01 8.27743113e-01 -8.11548889e-01 -8.80446792e-01 -4.43458974e-01 1.07413447e+00 -6.18938506e-01 5.87396264e-01 5.77328913e-02 -9.75129604e-01 6.47418916e-01 -6.81764632e-02 -5.50080180e-01 1.11804402e+00 4.38808322e-01 -8.98865581e-01 1.77088618e-01 -1.81875014e+00 8.92330885e-01 1.28545833e+00 -7.55229294e-01 -7.89589882e-01 -1.62426114e-01 5.12612641e-01 1.79622665e-01 -8.91097903e-01 3.60211849e-01 8.98526192e-01 -1.30442941e+00 8.49431455e-01 -8.75527143e-01 1.01234221e+00 3.82972747e-01 -7.25492835e-02 -9.98714924e-01 -3.76134098e-01 -2.90067732e-01 6.62133098e-01 1.18792951e+00 3.25619161e-01 -1.01891530e+00 4.96076822e-01 1.42963529e+00 -9.99649614e-02 -8.63918126e-01 -6.03871226e-01 -4.02545929e-01 5.53304911e-01 -1.20758750e-01 8.55164766e-01 1.67171896e+00 7.10513890e-01 -2.33135045e-01 7.08255963e-03 -6.06440865e-02 5.06656826e-01 8.61500353e-02 3.42682630e-01 -9.46702659e-01 -3.53191905e-02 -6.99314713e-01 -2.85733908e-01 2.12670952e-01 4.09930684e-02 -7.44595587e-01 -4.94800031e-01 -1.63289475e+00 1.03529167e+00 -6.89809442e-01 -1.26011163e-01 2.74253428e-01 -4.02461588e-01 -1.50812492e-01 5.78802466e-01 2.90819198e-01 -7.19647557e-02 -1.36612549e-01 1.00782514e+00 -7.49717001e-03 -6.79888111e-03 -4.50849831e-01 -1.64198995e+00 6.67392313e-01 8.48190606e-01 -6.28313124e-01 -2.31677383e-01 -2.16604501e-01 5.41062653e-01 3.20990354e-01 8.08034778e-01 -3.48699510e-01 1.38410568e-01 -7.02275217e-01 4.13323417e-02 1.99303031e-01 1.67861015e-01 -6.85740352e-01 2.23520279e-01 9.17476892e-01 -7.25944519e-01 -6.88347965e-02 2.29308326e-02 1.17393091e-01 9.19475704e-02 -3.19603443e-01 5.86879730e-01 -1.56791478e-01 2.99720556e-01 -1.67714044e-01 -4.45864201e-01 5.29920280e-01 9.22585666e-01 -3.79895836e-01 -7.96817362e-01 -6.47805929e-01 -2.19188601e-01 -5.72482869e-02 1.06773853e+00 3.45547423e-02 3.89598310e-01 -1.03196979e+00 -1.03529739e+00 -4.51921403e-01 4.03606668e-02 -6.90728784e-01 4.65668052e-01 1.21862829e+00 -3.22361231e-01 -1.49472896e-02 -1.95592284e-01 2.24716872e-01 -1.46584189e+00 7.04510689e-01 2.81583130e-01 2.02668145e-01 -4.64043677e-01 3.70850176e-01 4.03675914e-01 1.43290251e-01 -1.44577131e-01 9.51997414e-02 2.06281338e-02 3.78394276e-01 4.11708117e-01 8.14588606e-01 -4.88975376e-01 -6.85589194e-01 -5.79429626e-01 2.09189743e-01 4.19562608e-02 -1.95863888e-01 1.03482294e+00 -1.06701046e-01 -3.02984387e-01 3.94820750e-01 1.09593177e+00 5.28429270e-01 -8.37816596e-01 3.72487932e-01 -2.30331421e-01 -9.00024295e-01 -3.20166737e-01 -9.19409513e-01 -4.98541772e-01 8.75902534e-01 4.02037948e-02 3.13465685e-01 8.30179393e-01 -4.42506149e-02 6.48753107e-01 -1.79782927e-01 1.37980118e-01 -8.94769609e-01 -4.13765125e-02 -4.67113972e-01 6.01401389e-01 -1.17268980e+00 1.49194941e-01 -1.89320311e-01 -6.41374350e-01 5.77869475e-01 2.09969163e-01 1.58424824e-01 3.57211173e-01 -8.82693939e-03 2.87169248e-01 -4.48688895e-01 -6.55398726e-01 2.61723369e-01 -1.93110853e-01 3.37644279e-01 6.34233057e-01 6.64520919e-01 -1.42318666e+00 6.14733815e-01 -4.92260903e-01 -1.68191865e-01 1.00364029e+00 6.48761988e-01 -1.66981876e-01 -6.76939726e-01 -6.13285661e-01 1.00981283e+00 -1.21433485e+00 -2.47544006e-01 -6.49061084e-01 6.08012497e-01 4.13299650e-01 1.19580436e+00 2.79617846e-01 -2.18828291e-01 9.50546637e-02 -1.33457080e-01 3.15984190e-01 -6.09153032e-01 -8.27249587e-01 -4.97707754e-01 6.41875505e-01 -4.64517087e-01 -2.66875893e-01 -8.24556887e-01 -1.10717964e+00 -1.05269957e+00 -1.74221843e-02 8.03585127e-02 2.86320120e-01 8.68431091e-01 5.52722752e-01 1.96535215e-01 2.27452338e-01 1.05263218e-01 -7.21197784e-01 -3.09021443e-01 -4.80010420e-01 1.00561285e+00 5.51108181e-01 -3.25090677e-01 -7.13422835e-01 -4.89647061e-01]
[8.832578659057617, 5.604638576507568]
30d9c954-e07b-4601-a7f9-796c47414d34
improving-visual-representation-learning
2212.14504
null
https://arxiv.org/abs/2212.14504v2
https://arxiv.org/pdf/2212.14504v2.pdf
Improving Visual Representation Learning through Perceptual Understanding
We present an extension to masked autoencoders (MAE) which improves on the representations learnt by the model by explicitly encouraging the learning of higher scene-level features. We do this by: (i) the introduction of a perceptual similarity term between generated and real images (ii) incorporating several techniques from the adversarial training literature including multi-scale training and adaptive discriminator augmentation. The combination of these results in not only better pixel reconstruction but also representations which appear to capture better higher-level details within images. More consequentially, we show how our method, Perceptual MAE, leads to better performance when used for downstream tasks outperforming previous methods. We achieve 78.1% top-1 accuracy linear probing on ImageNet-1K and up to 88.1% when fine-tuning, with similar results for other downstream tasks, all without use of additional pre-trained models or data.
['Ken Chatfield', 'Frederick Hoffman', 'Samyakh Tukra']
2022-12-30
null
http://openaccess.thecvf.com//content/CVPR2023/html/Tukra_Improving_Visual_Representation_Learning_Through_Perceptual_Understanding_CVPR_2023_paper.html
http://openaccess.thecvf.com//content/CVPR2023/papers/Tukra_Improving_Visual_Representation_Learning_Through_Perceptual_Understanding_CVPR_2023_paper.pdf
cvpr-2023-1
['self-supervised-image-classification']
['computer-vision']
[ 5.84840178e-01 3.54208320e-01 3.26168239e-01 -2.93337971e-01 -8.75766993e-01 -5.42138040e-01 9.89141464e-01 -2.28334695e-01 -5.11445224e-01 7.84306705e-01 5.11343241e-01 -2.68308848e-01 2.81989634e-01 -7.66633213e-01 -1.10555434e+00 -6.01925373e-01 -1.45116284e-01 -3.38091031e-02 3.97546113e-01 -4.76810992e-01 2.47901883e-02 7.30798781e-01 -1.28215098e+00 6.92478597e-01 3.96275073e-01 1.04716671e+00 1.80131927e-01 8.15084636e-01 3.20193738e-01 1.04473805e+00 -5.39449871e-01 -5.01158118e-01 6.15921676e-01 -3.50606084e-01 -7.80616939e-01 2.16925412e-01 8.16007018e-01 -5.07816255e-01 -5.36571860e-01 8.30940187e-01 4.44545746e-01 2.09622368e-01 6.41786098e-01 -9.12111282e-01 -9.58181322e-01 4.64266807e-01 -5.36105752e-01 2.48811990e-01 -8.82689096e-03 4.32042420e-01 8.61279368e-01 -8.77632141e-01 6.07817948e-01 1.34391248e+00 9.06186640e-01 6.42655015e-01 -1.69616282e+00 -6.67908490e-01 4.99173924e-02 -3.75257252e-04 -1.13599622e+00 -6.28596246e-01 5.82743049e-01 -3.36466789e-01 1.23649311e+00 1.32679287e-02 3.09105664e-01 1.41095996e+00 2.14533925e-01 6.28720999e-01 1.47188830e+00 -4.84012991e-01 -4.29967679e-02 1.92957968e-01 -5.40527701e-01 5.68037868e-01 -7.62123466e-02 6.90591633e-01 -8.39005262e-02 1.87348530e-01 1.05667150e+00 -2.08910093e-01 -2.95734406e-01 -3.48075747e-01 -1.26480532e+00 8.47049713e-01 1.02194881e+00 3.59547824e-01 -4.63571340e-01 5.47574520e-01 3.18115503e-01 3.38826001e-01 3.67030352e-01 7.42423654e-01 -4.54465628e-01 3.91627610e-01 -1.09084988e+00 8.24987814e-02 3.00114870e-01 7.66812682e-01 8.43190372e-01 6.89447761e-01 -2.62817532e-01 6.19078577e-01 -6.70066401e-02 2.32155800e-01 6.44889176e-01 -1.33836520e+00 3.64229620e-01 5.20382151e-02 9.26252007e-02 -7.94787109e-01 -2.32951120e-01 -5.69780529e-01 -8.20260406e-01 8.28509390e-01 4.41657603e-01 -1.41397730e-01 -1.41100848e+00 2.00837636e+00 -2.02755883e-01 4.53467816e-01 3.50228220e-01 6.02526724e-01 5.35121143e-01 7.49132216e-01 3.69871020e-01 2.65801996e-01 1.24489939e+00 -1.11774027e+00 -3.89566004e-01 -3.88226539e-01 3.24218661e-01 -8.53560388e-01 1.02819538e+00 3.99005383e-01 -1.06286323e+00 -1.06856585e+00 -1.24288130e+00 -1.91362709e-01 -6.60960555e-01 -6.81373617e-03 6.10300660e-01 6.78838074e-01 -1.56321180e+00 1.00528491e+00 -6.82683468e-01 -1.41196802e-01 6.11891627e-01 3.00833285e-01 -6.11727059e-01 -5.59784994e-02 -1.12419569e+00 1.03821921e+00 4.21998054e-01 -2.58735806e-01 -1.13038778e+00 -7.96362936e-01 -1.12152088e+00 -1.72112789e-02 5.05932048e-02 -7.64207065e-01 1.16194308e+00 -1.40167272e+00 -1.47811925e+00 9.38185155e-01 1.49624258e-01 -8.63546550e-01 5.55416167e-01 -3.04465860e-01 -5.48716187e-01 3.13780755e-01 7.98200741e-02 1.44144046e+00 1.09369802e+00 -1.56855214e+00 -4.47477996e-01 7.94188753e-02 2.39734501e-01 1.15752928e-01 -1.46451458e-01 -2.35034823e-01 -2.27046385e-01 -1.12928391e+00 -3.68883789e-01 -9.11644638e-01 -4.99959528e-01 1.65905744e-01 -3.12411398e-01 3.59210223e-01 7.88208902e-01 -9.04669404e-01 4.92747009e-01 -2.23313975e+00 -1.48833189e-02 8.21058173e-03 1.96799606e-01 5.21937966e-01 -5.67116797e-01 3.91515702e-01 -5.11617243e-01 3.42126861e-02 -4.09327179e-01 -5.59344113e-01 -1.82018235e-01 3.91036242e-01 -6.33846879e-01 2.96836078e-01 8.00600052e-01 1.19626665e+00 -7.05122352e-01 -1.73963591e-01 5.49681723e-01 5.74837506e-01 -7.14237392e-01 5.20010404e-02 -2.15120260e-02 5.47856152e-01 6.66217133e-02 2.22672403e-01 5.42738795e-01 -1.97775513e-01 -2.58680761e-01 -2.60037810e-01 8.36703554e-02 1.98468372e-01 -1.00250399e+00 1.83318913e+00 -7.09114134e-01 7.97214270e-01 1.22258186e-01 -9.26826656e-01 5.32688916e-01 3.14406604e-01 9.32874680e-02 -7.36328483e-01 8.66525434e-03 1.47901168e-02 5.10644503e-02 -1.52119011e-01 5.27515054e-01 -3.44358146e-01 8.50262865e-02 7.38416612e-03 3.35005879e-01 -2.15047881e-01 -1.74919814e-01 3.32700789e-01 1.14917850e+00 3.30797106e-01 2.83252001e-01 -3.20887566e-01 3.83135408e-01 -2.12859455e-03 2.04400852e-01 9.90272462e-01 -1.73646435e-01 8.86520147e-01 4.03617993e-02 -1.70058131e-01 -1.34301662e+00 -1.16965795e+00 -1.59132466e-01 9.28643286e-01 -6.03794344e-02 -6.71875998e-02 -6.30439579e-01 -5.29407978e-01 3.33530679e-02 8.87145996e-01 -9.84515011e-01 -3.20210099e-01 -6.24061227e-01 -4.21819568e-01 8.29621613e-01 9.30224061e-01 6.70768559e-01 -1.26317608e+00 -3.84668022e-01 2.36385629e-01 1.50205851e-01 -1.34541118e+00 -3.11775059e-01 4.68570948e-01 -8.68343174e-01 -5.22574306e-01 -7.87113369e-01 -7.10494101e-01 3.80061477e-01 1.41189039e-01 1.35220981e+00 -1.46973496e-02 -3.77460480e-01 3.57721448e-01 -3.53081316e-01 -8.23304430e-02 -7.58966386e-01 -1.82885498e-01 -1.09505899e-01 -2.07325682e-01 -1.43038586e-01 -8.57212424e-01 -6.76508188e-01 1.28101513e-01 -1.19735312e+00 1.06091745e-01 1.03251255e+00 1.08390057e+00 5.16866028e-01 -5.08429743e-02 6.00513875e-01 -1.02904248e+00 3.03179443e-01 -3.06774646e-01 -3.17198366e-01 -1.00990728e-01 -5.81949413e-01 3.40382904e-01 9.77661788e-01 -5.77466905e-01 -1.02703679e+00 1.37840182e-01 -4.40576941e-01 -6.88999355e-01 -5.30783594e-01 -8.88654664e-02 -1.98553681e-01 -3.17959428e-01 9.57708538e-01 1.63892627e-01 -6.06734231e-02 -3.61139745e-01 8.84602070e-01 2.69794285e-01 1.01304913e+00 -4.86643791e-01 1.13552141e+00 6.18322611e-01 2.84626875e-02 -5.28281748e-01 -6.58751845e-01 -1.17693186e-01 -5.48334062e-01 2.83959717e-01 1.09290767e+00 -1.14491105e+00 -1.17199048e-01 2.77172536e-01 -8.66431534e-01 -7.37697601e-01 -7.73913980e-01 3.80048633e-01 -6.77185297e-01 2.95638323e-01 -8.84422660e-01 -3.79360199e-01 -6.04120977e-02 -1.13774610e+00 1.14835835e+00 -1.06550485e-01 -3.66418138e-02 -9.99469280e-01 -9.33623388e-02 2.73159832e-01 6.14417136e-01 3.63065153e-01 7.60659397e-01 -5.38983881e-01 -5.49467742e-01 -7.32464045e-02 -4.29827899e-01 7.98400462e-01 -2.62888074e-02 -3.07340711e-01 -1.42107821e+00 -4.02763039e-01 -9.39793512e-02 -5.62208593e-01 1.29600656e+00 3.35839093e-01 1.16801834e+00 -3.61696571e-01 -8.85201767e-02 8.35635245e-01 1.65951669e+00 -4.64411452e-02 9.92401540e-01 4.42088157e-01 5.62443256e-01 4.62214857e-01 1.47463679e-01 7.96311814e-03 -2.68868152e-02 6.18005991e-01 5.73815167e-01 -5.97713947e-01 -8.25726032e-01 -5.39399862e-01 4.08420861e-01 2.87526488e-01 2.82940659e-04 -1.38194844e-01 -4.25644845e-01 6.91047013e-01 -1.47253048e+00 -1.09322274e+00 3.72487828e-02 1.85340846e+00 8.22658122e-01 2.20565170e-01 3.16277631e-02 1.09017260e-01 4.75895762e-01 2.97228426e-01 -6.41179621e-01 -5.23200154e-01 -2.89092928e-01 8.53598416e-01 9.31534708e-01 6.45880878e-01 -1.35135949e+00 1.21988082e+00 7.34217215e+00 8.98736894e-01 -9.46845770e-01 8.44621584e-02 7.20720053e-01 1.35522887e-01 -1.66754231e-01 -5.56247905e-02 -3.96046072e-01 1.71780273e-01 9.55094039e-01 2.99827933e-01 5.49901068e-01 8.33105505e-01 -1.25234246e-01 1.28133386e-01 -1.03892934e+00 5.77086866e-01 -3.51621024e-02 -1.44928110e+00 7.72314966e-02 1.69520110e-01 9.97264028e-01 1.49699122e-01 3.47304016e-01 4.88328427e-01 7.59210944e-01 -1.32618392e+00 7.64368951e-01 2.73385346e-01 9.46837902e-01 -5.15509784e-01 4.84111607e-01 1.30050644e-01 -9.44159448e-01 -1.12646125e-01 -4.05186951e-01 -9.00076181e-02 9.23137181e-03 3.05941671e-01 -6.63978934e-01 4.94819880e-01 6.48141861e-01 6.68037534e-01 -8.09177041e-01 7.23711491e-01 -4.82724756e-01 5.76664031e-01 -1.07243590e-01 5.67466378e-01 4.48314488e-01 2.17853665e-01 3.29655319e-01 1.42270029e+00 3.32499780e-02 -1.67047426e-01 1.99397560e-02 8.42697620e-01 -2.54647970e-01 -2.23109722e-01 -7.75583029e-01 3.52017850e-01 1.43068060e-01 1.23237598e+00 -4.61608231e-01 -4.63582814e-01 -5.60742319e-01 1.52544773e+00 3.26340377e-01 5.69155812e-01 -7.85480082e-01 -2.51252145e-01 8.13651025e-01 1.28368840e-01 7.48039424e-01 -2.56932646e-01 -1.66644484e-01 -9.79751408e-01 -2.49142975e-01 -9.83009756e-01 1.98725864e-01 -1.07367909e+00 -1.46555710e+00 7.97092676e-01 -1.46134704e-01 -1.06780243e+00 -5.48769772e-01 -9.56561685e-01 -5.35205185e-01 9.87049341e-01 -1.66258156e+00 -1.30467641e+00 -1.09903067e-01 7.23442078e-01 4.27232742e-01 -1.52788445e-01 1.07709301e+00 2.13800386e-01 -2.04303697e-01 7.31317937e-01 1.41766936e-01 2.99530685e-01 7.14929938e-01 -1.44545710e+00 7.60396302e-01 1.19308507e+00 5.25836110e-01 3.64544570e-01 7.75847495e-01 -2.89259911e-01 -7.36360610e-01 -1.25268006e+00 3.85116816e-01 -5.81278086e-01 7.95875549e-01 -3.62164080e-01 -8.14892232e-01 9.57888544e-01 6.04264438e-01 2.59832144e-01 4.82739747e-01 -2.07499146e-01 -8.03358316e-01 5.47127314e-02 -1.33113444e+00 5.82901120e-01 1.04506910e+00 -6.61146939e-01 -6.84803665e-01 5.23758456e-02 9.44901645e-01 -3.39963734e-01 -8.29664469e-01 3.70491862e-01 2.20478252e-01 -1.10097277e+00 1.45107698e+00 -8.63886476e-01 8.50058198e-01 -2.12207898e-01 -4.45117623e-01 -1.51288784e+00 -7.15876937e-01 -4.49503481e-01 6.44071475e-02 1.04921412e+00 3.75966400e-01 -5.30981481e-01 6.36252046e-01 3.37396950e-01 -3.21895391e-01 -5.74654460e-01 -7.48394489e-01 -8.97057831e-01 4.88941938e-01 -3.97726148e-01 2.73775995e-01 8.44338357e-01 -6.88365281e-01 3.69529516e-01 -6.32821143e-01 3.07081223e-01 5.37280858e-01 -5.27782589e-02 6.80977345e-01 -6.83201492e-01 -7.16965139e-01 -4.81750906e-01 -7.25416422e-01 -1.22890174e+00 1.86519161e-01 -8.57641280e-01 3.42197530e-02 -1.26537120e+00 -1.58208251e-01 -3.02355558e-01 -4.46431935e-01 6.73750699e-01 -1.91112310e-01 8.80589843e-01 2.30654761e-01 1.43573329e-01 -2.38326818e-01 5.37740648e-01 1.27730179e+00 -6.13614880e-02 1.67002499e-01 -2.56788641e-01 -9.62074518e-01 5.91265738e-01 7.67758548e-01 -3.47709984e-01 -3.58820796e-01 -5.45668006e-01 -4.57953483e-01 -2.66610175e-01 8.12241912e-01 -1.34972644e+00 -2.69201398e-01 1.51367664e-01 9.69846785e-01 -3.70786674e-02 6.64727092e-01 -7.62698650e-01 4.34657261e-02 5.47078431e-01 -4.94871885e-01 -2.00959761e-02 6.55316174e-01 7.55690217e-01 -1.14947565e-01 -1.43905729e-01 1.17823529e+00 -2.26521283e-01 -1.11288655e+00 7.00382143e-02 -4.04892206e-01 -7.58133754e-02 8.48984122e-01 -3.04518640e-01 -3.34050894e-01 -5.54982960e-01 -9.31474686e-01 -2.33930677e-01 6.87547982e-01 4.26752001e-01 3.87036294e-01 -1.32784343e+00 -7.39851058e-01 2.38925308e-01 -3.66865769e-02 -3.06296647e-01 1.08130872e-01 4.50543195e-01 -4.70055699e-01 3.07867557e-01 -4.22197640e-01 -2.74968326e-01 -8.16996217e-01 8.16919684e-01 4.16829824e-01 -3.82508844e-01 -8.68804455e-01 1.01760948e+00 4.64472026e-01 -2.15393558e-01 3.58294547e-02 -1.57417253e-01 1.47826388e-01 -3.85881394e-01 5.02215743e-01 -6.79326281e-02 2.43872553e-02 -8.23931336e-01 -1.45430654e-01 2.54875749e-01 -1.16642535e-01 -3.96850407e-01 1.41360652e+00 5.61973825e-03 3.50144148e-01 8.86433423e-02 1.31172848e+00 1.89150020e-01 -1.79294515e+00 -3.27843964e-01 -2.11754635e-01 -4.71588314e-01 2.67677724e-01 -1.03992164e+00 -1.21955264e+00 9.90039825e-01 8.31646621e-01 2.74506230e-02 1.23325717e+00 -8.60980339e-03 7.13747442e-01 2.49792248e-01 1.78903997e-01 -6.27248585e-01 3.20157617e-01 1.40602931e-01 1.00458062e+00 -1.19620013e+00 -1.02160107e-02 -2.05489442e-01 -7.06590295e-01 8.17420483e-01 4.57356036e-01 -5.64191341e-01 4.55986440e-01 3.87526304e-01 2.42774680e-01 1.43685147e-01 -6.38526559e-01 -4.77123439e-01 2.78267771e-01 9.69558895e-01 2.28353441e-01 -1.26856372e-01 2.62904704e-01 2.75878936e-01 -2.28715301e-01 -1.75319895e-01 3.09780568e-01 6.77053630e-01 -2.47862935e-01 -1.20905983e+00 -2.52739489e-01 3.46467674e-01 -5.68948984e-01 -4.79548693e-01 -1.95348457e-01 1.10545433e+00 2.48813838e-01 8.93061757e-01 1.29988968e-01 -4.93340611e-01 2.54435986e-01 -4.67246138e-02 5.50850272e-01 -6.25696778e-01 -6.95647836e-01 -1.03525370e-01 3.86496224e-02 -6.83744431e-01 -4.13649946e-01 -4.83122170e-01 -9.31024790e-01 -2.41875082e-01 -1.13377139e-01 -3.30586195e-01 4.35372055e-01 6.96733117e-01 4.22759265e-01 8.05046558e-01 5.18969715e-01 -1.29771113e+00 -5.50977230e-01 -1.05192184e+00 -3.97123963e-01 6.93070292e-01 4.17640686e-01 -4.59369630e-01 -3.80365163e-01 2.71443546e-01]
[11.44682502746582, -0.43848925828933716]
b6a4c0d2-2398-4b5d-a5d8-ca085444273b
evolvement-constrained-adversarial-learning
1811.02476
null
http://arxiv.org/abs/1811.02476v1
http://arxiv.org/pdf/1811.02476v1.pdf
Evolvement Constrained Adversarial Learning for Video Style Transfer
Video style transfer is a useful component for applications such as augmented reality, non-photorealistic rendering, and interactive games. Many existing methods use optical flow to preserve the temporal smoothness of the synthesized video. However, the estimation of optical flow is sensitive to occlusions and rapid motions. Thus, in this work, we introduce a novel evolve-sync loss computed by evolvements to replace optical flow. Using this evolve-sync loss, we build an adversarial learning framework, termed as Video Style Transfer Generative Adversarial Network (VST-GAN), which improves upon the MGAN method for image style transfer for more efficient video style transfer. We perform extensive experimental evaluations of our method and show quantitative and qualitative improvements over the state-of-the-art methods.
['Xiao Bian', 'Longyin Wen', 'Siwei Lyu', 'Wenbo Li']
2018-11-06
null
null
null
null
['video-style-transfer']
['computer-vision']
[ 1.85119554e-01 -3.10365707e-01 9.91925374e-02 -1.04760058e-01 -3.23497266e-01 -6.47144496e-01 5.60630620e-01 -7.33605087e-01 -1.31765723e-01 9.67553079e-01 1.51847929e-01 -1.98725283e-01 4.22930837e-01 -8.53296399e-01 -8.01275790e-01 -3.90481055e-01 3.12771231e-01 -1.56890765e-01 3.28334421e-01 -2.82813340e-01 1.00581661e-01 4.75506842e-01 -1.07497823e+00 3.04569006e-01 8.97884488e-01 8.64466906e-01 -2.73275405e-01 7.39363194e-01 -1.13349602e-01 1.18854249e+00 -5.71972728e-01 -6.28780127e-01 4.32339042e-01 -9.44807589e-01 -6.22583508e-01 2.82410607e-02 5.28008699e-01 -6.96426928e-01 -8.95289004e-01 1.04003584e+00 4.11710054e-01 3.93076122e-01 4.83843446e-01 -1.38621283e+00 -8.98736715e-01 -1.30846173e-01 -6.62646890e-01 1.14884183e-01 6.53471947e-01 3.20841700e-01 6.10286772e-01 -8.52465868e-01 8.48769367e-01 1.47514439e+00 5.79086602e-01 7.94384539e-01 -1.16029179e+00 -8.23090374e-01 8.76112133e-02 1.33395478e-01 -1.23552120e+00 -2.26926386e-01 1.03557241e+00 -3.02070677e-01 2.09100753e-01 3.67185712e-01 8.48981440e-01 1.14082992e+00 2.74772584e-01 8.33975017e-01 1.12694979e+00 -2.32155234e-01 9.82848331e-02 -1.86341345e-01 -7.55570829e-01 8.36700916e-01 -2.11041361e-01 4.11516070e-01 -4.66151178e-01 -1.91059504e-02 1.52557218e+00 8.31133351e-02 -5.02803147e-01 -3.44022155e-01 -9.99773145e-01 7.64867961e-01 6.17177188e-01 -2.11980045e-01 -2.67269194e-01 3.95278364e-01 3.12664360e-01 4.30561990e-01 7.81699479e-01 1.78191125e-01 2.40342226e-02 -2.42088705e-01 -7.74499655e-01 3.39878649e-01 6.57937050e-01 9.60136533e-01 7.17345536e-01 5.57568133e-01 -3.20393860e-01 6.90890551e-01 2.43538246e-01 3.12277675e-01 2.82414317e-01 -1.44930565e+00 1.75236270e-01 3.55406702e-01 2.14359034e-02 -1.05955386e+00 2.23081440e-01 -2.02650367e-03 -8.47500682e-01 6.37013316e-01 -9.04087792e-04 -3.17964107e-01 -9.16786671e-01 1.61043906e+00 4.14758623e-01 8.28553021e-01 -1.15391888e-01 9.90449548e-01 7.80290842e-01 8.28954935e-01 8.65462124e-02 -6.43344074e-02 7.22787857e-01 -1.15079093e+00 -8.28616560e-01 -1.01678163e-01 6.31073639e-02 -1.00884104e+00 9.82483864e-01 8.02008286e-02 -1.30163801e+00 -6.35941982e-01 -9.28996027e-01 -2.94359624e-01 2.33955771e-01 -3.68003815e-01 6.06377423e-01 5.63906491e-01 -9.74973917e-01 7.37500131e-01 -8.61612141e-01 1.89551990e-02 5.04811645e-01 1.05734617e-01 -4.04638052e-01 -1.33628100e-01 -1.02709150e+00 5.73893845e-01 1.17575854e-01 3.34259123e-02 -8.45913470e-01 -1.00974548e+00 -1.03366327e+00 -1.45137757e-01 3.02697867e-01 -1.04784238e+00 1.05213296e+00 -1.29961574e+00 -2.23835802e+00 8.02200437e-01 -6.93844631e-02 -1.72515556e-01 1.03467047e+00 -4.17307794e-01 -3.18135172e-01 8.79739001e-02 -1.11426942e-01 7.35518754e-01 1.00217450e+00 -1.30225170e+00 -4.16906029e-01 1.12982087e-01 3.85100752e-01 2.19093561e-01 -4.42372374e-02 1.24823123e-01 -6.79872036e-01 -1.16005027e+00 -3.27707976e-01 -1.05700254e+00 -1.07836001e-01 6.97957456e-01 -1.90879196e-01 2.35230163e-01 1.26630914e+00 -7.55742908e-01 1.03607643e+00 -2.20896602e+00 3.66536647e-01 -1.35069400e-01 1.73087344e-01 5.23991942e-01 -2.41614521e-01 2.50613868e-01 -3.59908827e-02 -1.47463186e-02 -3.50195616e-01 -4.18321103e-01 -3.24333757e-01 3.13330144e-01 -3.84057790e-01 3.63896370e-01 3.45141023e-01 1.06798971e+00 -1.04063666e+00 -3.62274796e-01 4.57585603e-01 8.74440134e-01 -7.50623167e-01 5.75336337e-01 -1.21619806e-01 9.98873472e-01 -4.77851689e-01 4.83334303e-01 7.47961044e-01 9.96734425e-02 -1.61336854e-01 -1.27765417e-01 7.99661651e-02 1.40561648e-02 -9.06774998e-01 1.99425364e+00 -4.78811234e-01 7.68784463e-01 -3.15961055e-02 -5.16511083e-01 8.96537125e-01 2.16111019e-01 5.06724417e-01 -5.38137138e-01 2.58485287e-01 1.88174427e-01 -2.65780836e-01 -2.93313861e-01 4.44459736e-01 -2.14186653e-01 3.56899202e-01 1.66405663e-01 -7.33562410e-02 -4.45293188e-01 -3.56475376e-02 2.17625201e-01 9.73113239e-01 7.38434196e-01 2.44766846e-02 -8.73295870e-03 7.05788791e-01 -4.37435538e-01 8.08787286e-01 1.24349400e-01 -2.29089156e-01 1.12195373e+00 5.22634625e-01 -6.54748380e-01 -1.16073179e+00 -1.15838218e+00 2.91668534e-01 5.47193289e-01 4.10178602e-01 -3.99905384e-01 -7.75527775e-01 -7.88895965e-01 -1.56262994e-01 3.67186666e-01 -5.99111557e-01 -3.12121242e-01 -8.86418581e-01 -1.17526643e-01 3.34480286e-01 7.16702104e-01 7.40357041e-01 -1.14610100e+00 -9.18376222e-02 1.97635248e-01 -1.64933071e-01 -1.32608771e+00 -1.01546085e+00 -8.03464830e-01 -7.08597183e-01 -9.44928348e-01 -1.02107406e+00 -6.77991986e-01 5.31356692e-01 2.38661930e-01 1.10896575e+00 1.77281156e-01 -1.00164995e-01 3.21710587e-01 -3.28744560e-01 -1.79835841e-01 -6.45475090e-01 -2.38044009e-01 -1.77058786e-01 2.68241346e-01 -2.35064209e-01 -7.91989625e-01 -9.72338319e-01 4.61508304e-01 -1.24632192e+00 2.44042605e-01 5.64498045e-02 9.32733655e-01 7.10921884e-01 -3.97949725e-01 1.78507939e-01 -1.10609901e+00 5.84118307e-01 -1.89763516e-01 -6.30106449e-01 1.43966556e-01 -2.19812423e-01 -3.53900678e-02 7.04645872e-01 -5.89624643e-01 -1.20211911e+00 -7.59431049e-02 -2.56662726e-01 -1.12045240e+00 2.80211002e-01 9.95916203e-02 -8.05378333e-02 -6.97904944e-01 4.28522915e-01 1.04963385e-01 1.37059852e-01 -2.06322491e-01 4.25795883e-01 2.40248740e-01 7.50440300e-01 -4.74068224e-01 1.19617736e+00 6.50141120e-01 2.15883344e-01 -6.48742795e-01 -6.17407799e-01 -2.17004158e-02 -3.03067267e-01 -2.93233186e-01 8.54849339e-01 -9.14605856e-01 -6.08484924e-01 7.41078615e-01 -1.33886385e+00 -5.12497425e-01 -3.66070569e-01 3.75120819e-01 -8.05942178e-01 5.31008303e-01 -8.60418081e-01 -4.85950768e-01 -4.12669688e-01 -1.22347999e+00 1.05239844e+00 4.05043453e-01 -6.20580325e-03 -1.17479074e+00 2.93720186e-01 1.26809254e-01 4.90550816e-01 8.74886513e-01 3.26230794e-01 3.76861632e-01 -7.64299154e-01 -6.30642101e-02 -3.06185514e-01 5.61990023e-01 4.31210905e-01 2.10594222e-01 -7.54542589e-01 -4.18293536e-01 -1.68300852e-01 -1.58729330e-01 6.67163014e-01 1.17765568e-01 1.28220475e+00 -2.34115317e-01 -4.31926399e-02 1.28272116e+00 1.35645330e+00 2.82082587e-01 1.23828673e+00 1.51112065e-01 1.19473600e+00 2.19423845e-01 4.05746788e-01 2.63493747e-01 1.72377929e-01 7.96196342e-01 2.74528652e-01 -2.94372201e-01 -4.66677517e-01 -5.02031207e-01 3.78231525e-01 6.42207563e-01 -4.85644758e-01 -2.94452906e-01 -3.39717537e-01 1.78746954e-01 -1.75704062e+00 -8.41038942e-01 1.24624297e-01 1.99532700e+00 8.02530706e-01 -1.93603951e-02 -5.78514785e-02 7.77421892e-03 6.96272612e-01 3.50676686e-01 -5.52705586e-01 -3.85743201e-01 -3.65107730e-02 5.36449432e-01 2.35821828e-01 6.32710516e-01 -9.51522708e-01 1.18758130e+00 6.35419559e+00 7.67176032e-01 -1.38270819e+00 1.47164404e-01 5.63108623e-01 6.71421215e-02 -5.48625886e-01 -5.50055038e-03 -2.04523206e-01 6.28757298e-01 3.88731122e-01 -2.58432120e-01 6.80976152e-01 6.24163806e-01 1.76818296e-01 3.47702205e-01 -8.08116615e-01 1.02605271e+00 1.14886306e-01 -1.37817359e+00 1.96344912e-01 1.09991999e-02 1.14098108e+00 -4.21655864e-01 9.59000438e-02 1.02567337e-01 3.18265885e-01 -7.52121329e-01 6.45747244e-01 4.77409661e-01 1.08694553e+00 -8.10761452e-01 3.75994027e-01 -2.78950959e-01 -1.23117542e+00 3.95143598e-01 -1.83079332e-01 1.15537085e-01 5.50317287e-01 2.89688349e-01 4.91803400e-02 6.94441438e-01 4.50480789e-01 1.03521550e+00 -1.74536839e-01 1.04701340e+00 -5.20018578e-01 4.00517792e-01 8.89184326e-02 4.59664553e-01 2.31770918e-01 -5.03064871e-01 7.05446720e-01 7.40510702e-01 3.76814842e-01 2.14291871e-01 9.07483548e-02 8.94578874e-01 -4.65114355e-01 -1.51473237e-02 -6.88231349e-01 1.32007375e-01 1.42635182e-01 1.07361364e+00 -4.51943129e-01 -2.00264141e-01 -5.17564237e-01 1.61958694e+00 5.96048497e-02 6.14016593e-01 -1.02928960e+00 -3.73958915e-01 1.01340771e+00 7.27117881e-02 2.05811560e-01 -3.64966244e-01 2.50628628e-02 -1.49483883e+00 1.03114814e-01 -7.97640860e-01 -1.12705715e-02 -9.62079227e-01 -1.15048039e+00 8.08722973e-01 -2.76681095e-01 -1.55962718e+00 -3.20813537e-01 -4.52526718e-01 -9.30618107e-01 8.95491421e-01 -1.69862497e+00 -1.25220060e+00 -6.36991203e-01 6.56245530e-01 6.23025835e-01 -1.25186980e-01 4.99719173e-01 3.92437905e-01 -3.21513653e-01 6.07801437e-01 5.53010516e-02 1.83701769e-01 9.56380725e-01 -1.05425847e+00 7.39763319e-01 1.07952690e+00 -8.78948346e-02 2.31969744e-01 5.77938676e-01 -5.69211245e-01 -1.27701890e+00 -1.32684815e+00 2.01000988e-01 -2.49319822e-01 5.13250232e-01 -5.92713840e-02 -9.93837357e-01 7.65354335e-01 4.91017222e-01 6.06368005e-01 3.59166592e-01 -7.26693869e-01 -2.36633390e-01 -1.77686915e-01 -1.28590333e+00 7.68844783e-01 1.24309683e+00 -6.11827195e-01 -9.45580006e-02 -2.69027725e-02 9.74853337e-01 -8.31379950e-01 -8.35544229e-01 3.95068288e-01 6.58710361e-01 -1.05787301e+00 1.04583097e+00 -5.53491831e-01 8.02436233e-01 -4.37318385e-01 1.41316324e-01 -1.56238198e+00 -2.85007805e-01 -1.14250004e+00 -2.42538363e-01 1.23983240e+00 -1.68247208e-01 -5.35808086e-01 8.14812005e-01 5.79970360e-01 -1.68725044e-01 -4.83389467e-01 -6.32734954e-01 -7.47713506e-01 8.27747509e-02 -9.20500327e-03 5.56576312e-01 9.97789025e-01 -6.32071972e-01 9.55592543e-02 -7.63682783e-01 -1.57228112e-01 6.88181520e-01 -1.49047688e-01 8.74707937e-01 -8.48637938e-01 -3.57776821e-01 -2.86656946e-01 -6.93844855e-01 -1.09177852e+00 3.63100886e-01 -3.18585962e-01 -1.43192217e-01 -1.24392664e+00 -2.28680700e-01 -2.32886955e-01 -1.18047871e-01 9.29158852e-02 -4.83518481e-01 7.02266932e-01 4.16807532e-01 1.08806670e-01 -2.96212912e-01 9.71258342e-01 1.96827877e+00 -1.78969994e-01 -2.37689614e-01 -4.78398912e-02 -3.72097880e-01 6.03473663e-01 6.13590658e-01 -2.05451190e-01 -5.37154198e-01 -6.11571133e-01 -3.23064253e-02 1.89118490e-01 3.15536380e-01 -9.98758733e-01 -1.13725647e-01 -3.37851465e-01 3.28854918e-01 -1.00829288e-01 4.41873372e-01 -7.30094671e-01 4.14242625e-01 3.21779668e-01 -1.11633107e-01 2.70341396e-01 2.69608587e-01 5.13836980e-01 -4.51331705e-01 1.60759896e-01 9.02262509e-01 -6.57672882e-02 -6.00296736e-01 7.29512632e-01 1.88169330e-02 1.86954379e-01 1.08858943e+00 -7.00142086e-02 -1.96640491e-01 -6.91158175e-01 -4.29354072e-01 -3.88049856e-02 7.16013730e-01 5.51896989e-01 8.79218578e-01 -1.77154601e+00 -7.46896029e-01 2.57663786e-01 -1.92375451e-01 9.35569331e-02 2.51100659e-01 5.29144406e-01 -1.17144644e+00 -7.92595223e-02 -5.68428874e-01 -2.90532738e-01 -1.06160867e+00 5.18759847e-01 4.11549628e-01 -1.16879065e-02 -6.70625627e-01 6.36578083e-01 6.08379006e-01 -1.77564219e-01 -3.15964110e-02 -6.52655587e-02 5.88757358e-02 -5.96290946e-01 6.38700545e-01 5.14832318e-01 -2.75281191e-01 -5.66917598e-01 -7.10913017e-02 7.12166488e-01 1.45653486e-01 -1.98177263e-01 1.14644384e+00 -1.27236038e-01 -1.32055044e-01 7.40141273e-02 1.19936931e+00 3.30326781e-02 -1.88638592e+00 -1.19510479e-01 -7.51935244e-01 -9.98469472e-01 1.37456805e-02 -2.87088454e-01 -1.59095311e+00 7.61672378e-01 4.91361797e-01 9.41258483e-03 1.26246464e+00 -6.23212516e-01 1.12189686e+00 -2.23313197e-01 3.95594835e-01 -4.55655366e-01 2.70909727e-01 3.61550689e-01 9.87325668e-01 -1.10188949e+00 -1.67115405e-01 -7.02297866e-01 -6.33996189e-01 9.90616441e-01 8.14168990e-01 -5.18837988e-01 6.77516103e-01 2.06767917e-01 2.38690630e-01 2.89669812e-01 -4.20908660e-01 8.34615752e-02 4.36124593e-01 6.34562135e-01 4.78023231e-01 -1.84948057e-01 -4.90992486e-01 -2.63706088e-01 1.47502869e-01 3.66160154e-01 4.60014522e-01 9.07185733e-01 1.94329947e-01 -1.45532131e+00 -1.76008224e-01 -8.05205703e-02 -5.40075004e-01 -3.95953506e-02 -3.08976710e-01 6.15223050e-01 -1.68148205e-01 5.90003073e-01 8.31616297e-02 -4.63865548e-01 3.48687172e-01 -4.22850341e-01 7.23775327e-01 -3.42627794e-01 -4.51576799e-01 1.58270344e-01 -2.37460032e-01 -7.94755936e-01 -7.11824596e-01 -3.06171328e-01 -1.02019978e+00 -6.16651893e-01 1.26768574e-02 -1.30231544e-01 3.62237483e-01 6.77487075e-01 3.58348668e-01 7.06079364e-01 9.87907708e-01 -1.03183877e+00 3.08513995e-02 -7.20721722e-01 -3.71256441e-01 7.35131502e-01 3.41223776e-01 -6.20164275e-01 -1.56511113e-01 3.55964243e-01]
[11.230539321899414, -0.950326144695282]
3c22d30c-deb5-4479-84b5-c3f967bbc420
scalable-computation-of-optimized-queries-for
1612.04791
null
http://arxiv.org/abs/1612.04791v3
http://arxiv.org/pdf/1612.04791v3.pdf
Scalable Computation of Optimized Queries for Sequential Diagnosis
In many model-based diagnosis applications it is impossible to provide such a set of observations and/or measurements that allow to identify the real cause of a fault. Therefore, diagnosis systems often return many possible candidates, leaving the burden of selecting the correct diagnosis to a user. Sequential diagnosis techniques solve this problem by automatically generating a sequence of queries to some oracle. The answers to these queries provide additional information necessary to gradually restrict the search space by removing diagnosis candidates inconsistent with the answers. During query computation, existing sequential diagnosis methods often require the generation of many unnecessary query candidates and strongly rely on expensive logical reasoners. We tackle this issue by devising efficient heuristic query search methods. The proposed methods enable for the first time a completely reasoner-free query generation while at the same time guaranteeing optimality conditions, e.g. minimal cardinality or best understandability, of the returned query that existing methods cannot realize. Hence, the performance of this approach is independent of the (complexity of the) diagnosed system. Experiments conducted using real-world problems show that the new approach is highly scalable and outperforms existing methods by orders of magnitude.
['Wolfgang Schmid', 'Kostyantyn Shchekotykhin', 'Patrick Rodler']
2016-12-14
null
null
null
null
['sequential-diagnosis']
['medical']
[ 2.92872667e-01 3.00720274e-01 -1.72469854e-01 -2.94157594e-01 -9.28654373e-01 -4.83198583e-01 3.22118074e-01 7.40957320e-01 -8.11558589e-03 7.60245383e-01 -5.53978741e-01 -5.80031216e-01 -7.03658104e-01 -1.05072749e+00 -4.68688548e-01 -5.98464429e-01 1.51324823e-01 1.23052895e+00 5.81988215e-01 2.04853952e-01 2.24153236e-01 5.70269465e-01 -1.87269461e+00 6.79553151e-02 1.12843895e+00 1.19347417e+00 2.83403724e-01 5.02204537e-01 -1.23792954e-01 1.08446729e+00 -7.84019172e-01 8.49552304e-02 8.40976313e-02 -7.37223506e-01 -1.08922744e+00 5.01786351e-01 -1.08445913e-01 -4.00305599e-01 1.93252772e-01 1.38353229e+00 1.89567029e-01 6.43185154e-02 2.07009286e-01 -1.40775239e+00 -2.84714978e-02 6.24545872e-01 7.43872896e-02 -7.42329657e-02 8.42498958e-01 -1.35249212e-01 1.00619972e+00 -6.88944161e-01 6.77332401e-01 9.63731766e-01 2.42695510e-01 4.67022449e-01 -1.16275048e+00 -2.02231571e-01 1.03595234e-01 4.51899111e-01 -1.66232026e+00 -4.64069545e-01 6.84556782e-01 -2.05368623e-01 8.93208742e-01 8.86208475e-01 4.31923658e-01 4.13162082e-01 -8.27807039e-02 4.27980810e-01 8.90798509e-01 -3.93260360e-01 8.56546879e-01 3.52251172e-01 1.06125280e-01 8.12919438e-01 4.02913600e-01 -2.79868543e-01 -2.49262750e-01 -7.12745309e-01 5.58831632e-01 2.25841120e-01 -5.53081393e-01 -4.29281324e-01 -1.07327104e+00 5.89381456e-01 4.03556898e-02 2.63749152e-01 -5.64196825e-01 2.31881477e-02 1.17446497e-01 6.52169824e-01 -8.34371243e-03 6.75134480e-01 -6.24243557e-01 1.42315045e-01 -8.46352696e-01 4.50690985e-01 1.13419676e+00 8.23726833e-01 7.48065770e-01 -2.50517696e-01 1.26647815e-01 3.85539502e-01 6.78140894e-02 6.26710117e-01 2.20174298e-01 -8.35723221e-01 3.22151750e-01 1.00392747e+00 5.11818230e-01 -1.01608539e+00 -1.77505657e-01 -4.87332314e-01 -5.22971511e-01 1.03226490e-01 3.94084632e-01 8.43822733e-02 -5.84692240e-01 1.49137604e+00 6.62681162e-01 -1.71009414e-02 8.98756757e-02 9.11132514e-01 8.53103027e-02 5.13111770e-01 -5.13338149e-01 -7.75268853e-01 1.18233502e+00 -5.74203432e-01 -7.27494657e-01 -1.56039208e-01 6.60261571e-01 -5.08303404e-01 8.24453712e-01 7.98115373e-01 -1.15393960e+00 -4.93454598e-02 -8.77398193e-01 5.09150863e-01 1.46934807e-01 4.03943509e-02 5.42703390e-01 2.44141534e-01 -9.52132821e-01 3.77554893e-01 -9.81786132e-01 -2.27737218e-01 -1.81467026e-01 6.52550578e-01 -2.95634627e-01 -5.35212934e-01 -8.63171935e-01 6.67423069e-01 4.30587858e-01 1.41941085e-01 -8.46188426e-01 -3.51857424e-01 -5.73024988e-01 3.25205505e-01 1.15915203e+00 -7.28360057e-01 1.48790503e+00 -8.33177567e-01 -1.09667015e+00 2.93706685e-01 -5.18358648e-01 -4.75605041e-01 6.75105274e-01 4.15597521e-02 -3.20087880e-01 6.59175277e-01 1.27041236e-01 -9.21300203e-02 7.43753195e-01 -1.10059679e+00 -8.36499214e-01 -2.96399921e-01 4.15423155e-01 3.04004177e-03 -1.39651880e-01 -1.51559457e-01 -4.10641015e-01 1.68613568e-01 7.30205059e-01 -7.38451421e-01 -5.01963615e-01 3.26216146e-02 -6.16721034e-01 -2.49427661e-01 6.19475245e-01 -4.19623286e-01 1.27113867e+00 -1.77314270e+00 1.68947041e-01 6.04406297e-01 3.73811156e-01 -6.91230074e-02 2.19811678e-01 7.16455221e-01 -4.50582579e-02 -4.80457731e-02 -3.27252418e-01 -2.33644396e-02 -1.49538899e-02 4.14201051e-01 -3.18228185e-01 4.94159788e-01 2.00369313e-01 4.04542029e-01 -9.67835963e-01 -6.98911130e-01 1.68772176e-01 -2.12700158e-01 -7.71024108e-01 4.15139735e-01 -7.13343263e-01 3.83300453e-01 -7.24449098e-01 7.13770509e-01 4.03662920e-01 -6.22092187e-01 4.82247949e-01 1.46488294e-01 3.58667523e-01 5.33887267e-01 -1.64432740e+00 1.07722807e+00 -4.83928919e-01 -1.28296897e-01 1.58718616e-01 -1.27718627e+00 7.67453849e-01 6.89801574e-01 6.66405082e-01 -3.74064833e-01 -4.83404659e-02 5.97918332e-01 -3.56891863e-02 -7.00666308e-01 -5.89787401e-02 -1.18258350e-01 -7.66427889e-02 5.56854367e-01 -4.06291187e-01 -3.13972980e-02 1.99178010e-01 -3.34420949e-02 1.49884725e+00 -5.81447661e-01 4.29994941e-01 -1.86778486e-01 8.12906861e-01 2.74615318e-01 9.62081015e-01 8.65997493e-01 3.38928193e-01 3.07234496e-01 8.20371568e-01 -3.67341429e-01 -6.14129663e-01 -7.04377174e-01 2.22113758e-01 2.45820671e-01 2.24704504e-01 -4.43720579e-01 -7.04507411e-01 -7.83218086e-01 -8.56471136e-02 7.09260881e-01 -2.71229118e-01 -1.18224718e-01 -3.75566691e-01 -2.63224661e-01 8.96402970e-02 1.18263900e-01 6.56097680e-02 -8.43315840e-01 -1.01403391e+00 3.97593439e-01 -3.52622509e-01 -1.01362574e+00 3.20157222e-02 2.37564385e-01 -1.13090241e+00 -1.51402807e+00 -7.72087947e-02 -4.92780834e-01 1.28723848e+00 3.01426668e-02 1.01694441e+00 7.03885198e-01 -9.87197384e-02 5.16691543e-02 -2.67814428e-01 7.44178817e-02 -6.37255251e-01 -1.49206638e-01 -6.53941780e-02 1.06814401e-02 -2.67692190e-02 -4.14023668e-01 -2.04779595e-01 2.56593615e-01 -1.24928510e+00 -7.82602429e-02 5.73358715e-01 9.93199468e-01 9.38887239e-01 8.93985689e-01 6.70077682e-01 -1.17371702e+00 5.22131681e-01 -4.83119607e-01 -1.19679141e+00 6.40540063e-01 -1.05663872e+00 4.34183359e-01 9.29671228e-01 -4.09646630e-01 -8.10155571e-01 2.30599761e-01 4.04056199e-02 -3.57232213e-01 -1.83304638e-01 9.08853173e-01 -4.83883500e-01 2.61926770e-01 4.73966628e-01 2.87741125e-01 -3.62255543e-01 -5.27832210e-01 -1.74721535e-02 3.40612501e-01 5.24726450e-01 -4.91715908e-01 7.02630639e-01 3.42190623e-01 2.25213170e-01 -4.38571632e-01 -4.70802784e-01 -5.09017289e-01 -8.05728361e-02 6.84320256e-02 2.46834874e-01 -4.02892143e-01 -6.69899821e-01 2.31784601e-02 -1.00499892e+00 2.11863279e-01 -4.02051568e-01 2.42243439e-01 -3.59468520e-01 2.57317424e-01 -3.54733706e-01 -1.14846134e+00 -1.20340116e-01 -1.39592075e+00 8.98851216e-01 -2.00441003e-01 -4.29387122e-01 -6.77801073e-01 -2.63113856e-01 2.42589578e-01 1.49856403e-01 1.38752490e-01 1.25990987e+00 -9.43464994e-01 -9.04351711e-01 -7.53768086e-01 9.27710906e-02 -8.59674960e-02 2.75341809e-01 -2.60213137e-01 -6.77612960e-01 -1.03400141e-01 4.61390138e-01 3.52053642e-02 3.29042464e-01 -8.06299597e-03 1.05243337e+00 -7.87238657e-01 -4.55487669e-01 -1.02009056e-02 1.52783382e+00 2.35844195e-01 3.31669837e-01 -1.40790075e-01 3.61753881e-01 5.17020166e-01 8.96652102e-01 7.18132436e-01 1.48335308e-01 5.93609571e-01 4.90313709e-01 2.46002674e-01 3.40986937e-01 -2.69750245e-02 -5.49863912e-02 5.57823241e-01 3.50046605e-01 -3.82819057e-01 -1.07528973e+00 8.82290721e-01 -1.98120666e+00 -7.60151386e-01 4.53742640e-03 2.55703139e+00 9.70700264e-01 3.01620483e-01 -2.00456098e-01 8.79311800e-01 5.33205152e-01 -6.11252308e-01 -6.66658401e-01 -1.78888574e-01 2.93024302e-01 1.21893190e-01 1.82073176e-01 8.44298422e-01 -5.98378956e-01 3.94473255e-01 5.48341370e+00 4.90204513e-01 -1.03816962e+00 -1.91116259e-01 2.36857042e-01 -4.83957604e-02 -7.02722847e-01 4.02199477e-01 -3.84413421e-01 6.04610369e-02 9.69513357e-01 -3.48285854e-01 4.90981042e-01 1.06589925e+00 1.96308941e-01 -5.35799503e-01 -1.49150050e+00 8.07157993e-01 -2.37176672e-01 -9.94629443e-01 -2.19224706e-01 3.27180810e-02 6.54221952e-01 -4.78240639e-01 -4.90827173e-01 -1.69242606e-01 -4.95675728e-02 -6.48278892e-01 7.00757325e-01 5.79788506e-01 4.65575308e-01 -8.52094650e-01 9.40420628e-01 6.97035551e-01 -8.66396010e-01 -4.07589704e-01 -7.32856914e-02 -1.95448518e-01 7.22462833e-02 1.16395390e+00 -1.23031723e+00 6.82477474e-01 2.66631126e-01 1.09457886e-02 -3.25162590e-01 1.14118433e+00 -3.53534847e-01 4.31588471e-01 -6.12493753e-01 7.00324401e-02 -1.99408680e-01 2.17264563e-01 5.66437602e-01 6.21942043e-01 4.32557017e-01 7.01508895e-02 4.75598663e-01 8.81980300e-01 2.41689831e-01 -1.83010757e-01 -4.31514293e-01 -1.84142783e-01 8.39505196e-01 7.26199925e-01 -1.03310001e+00 -6.37144327e-01 -2.30872899e-01 8.90372574e-01 6.62158579e-02 2.64084429e-01 -6.30100489e-01 -2.86972493e-01 3.74521852e-01 1.68285713e-01 1.88962728e-01 5.38284704e-02 -2.31830016e-01 -1.06011283e+00 5.28225660e-01 -1.25292861e+00 4.52165574e-01 -3.05483311e-01 -1.00981188e+00 6.50877297e-01 9.43790972e-02 -1.06578720e+00 -1.06986046e+00 6.78647608e-02 -1.25920340e-01 7.71056652e-01 -1.22983789e+00 -4.81344253e-01 -1.57952622e-01 5.96461356e-01 3.48960936e-01 4.76086944e-01 1.10442853e+00 3.40409547e-01 -4.31275219e-01 1.72044069e-01 -4.33275461e-01 -4.30258870e-01 2.83350646e-01 -1.38287413e+00 -1.31684840e-01 1.08185303e+00 -1.01576410e-01 6.17400587e-01 1.07442999e+00 -7.35552967e-01 -1.80280721e+00 -8.15602422e-01 1.49031126e+00 -5.64538725e-02 4.71641302e-01 -1.98553987e-02 -1.00322282e+00 3.93415779e-01 -4.09078658e-01 -2.65879501e-02 2.05669135e-01 -8.02288577e-02 -2.93551944e-02 -3.37236971e-01 -1.43299794e+00 2.89102107e-01 6.32666528e-01 -5.97970486e-01 -4.36641604e-01 5.99560201e-01 6.93623781e-01 -2.20970228e-01 -7.01639235e-01 3.29391956e-01 1.11839972e-01 -8.94898891e-01 6.69167995e-01 -3.41783971e-01 2.80589927e-02 -7.13785648e-01 -1.78518202e-02 -7.30616212e-01 1.52558565e-01 -4.61823732e-01 -4.28174734e-01 8.72441709e-01 4.63037074e-01 -8.96269143e-01 5.43941081e-01 1.02603567e+00 4.64178957e-02 -9.99621928e-01 -1.00290346e+00 -7.51572490e-01 -7.83317089e-01 -5.45573354e-01 9.88690376e-01 8.87949467e-01 1.67428106e-01 5.46095744e-02 -1.01893537e-01 8.48281503e-01 5.21171451e-01 6.59231544e-01 4.17487264e-01 -1.49331951e+00 -8.16379488e-01 -1.60115838e-01 -3.01530629e-01 -7.22376943e-01 -1.25474453e-01 -4.68749791e-01 3.44350159e-01 -1.82686174e+00 -5.92109077e-02 -7.69383252e-01 -2.30309606e-01 7.33384252e-01 -1.31557733e-01 -4.98153895e-01 -2.08758593e-01 2.87115216e-01 -7.66934693e-01 3.60260084e-02 8.65878761e-01 -1.65430363e-02 -1.33313030e-01 3.53495568e-01 -4.30520415e-01 6.90934241e-01 6.75363600e-01 -7.56675243e-01 -8.40505958e-01 -3.33068401e-01 6.13895655e-01 8.64325047e-01 4.32571828e-01 -1.01963425e+00 4.99593318e-01 -5.55195570e-01 -2.42193550e-01 -5.44322670e-01 1.82610154e-01 -1.20557690e+00 6.06257081e-01 8.02547872e-01 -2.93137521e-01 2.21035808e-01 -3.37995142e-01 4.90187109e-01 -3.64841491e-01 -5.45098543e-01 3.49738270e-01 -1.46656418e-02 -4.26672161e-01 8.80162790e-02 -3.97336394e-01 -2.15208575e-01 9.15648580e-01 9.76049677e-02 1.07191153e-01 -4.18899268e-01 -6.95589483e-01 3.60509187e-01 5.70113778e-01 -7.28529766e-02 7.61053801e-01 -8.55961680e-01 -3.26499343e-01 1.59237981e-01 2.01115072e-01 3.98549825e-01 -1.47693921e-02 9.34406221e-01 -4.69053745e-01 6.19275630e-01 3.63606483e-01 -4.51474607e-01 -1.24956489e+00 8.79993081e-01 3.19774806e-01 -4.80091035e-01 -6.31890774e-01 5.65064132e-01 -6.67330176e-02 -1.23427249e-01 2.48148754e-01 -8.23944151e-01 1.25166684e-01 -2.93905467e-01 6.44580424e-01 3.63526404e-01 4.17381883e-01 8.81514177e-02 -5.52866101e-01 -1.74010769e-02 2.05507770e-01 -2.52950221e-01 1.22063565e+00 -2.32038677e-01 -4.33241099e-01 3.94060969e-01 6.36312306e-01 4.15934660e-02 -6.81200266e-01 -2.78530329e-01 4.23976749e-01 -3.84187102e-01 -3.98231857e-02 -8.32132101e-01 -1.06996000e+00 5.19960821e-01 1.76542372e-01 6.83234990e-01 1.81204414e+00 6.75718933e-02 7.66212106e-01 9.11640525e-01 7.49477267e-01 -7.78989792e-01 -3.50698650e-01 -2.74813659e-02 6.03321552e-01 -8.20435643e-01 -1.20840445e-01 -6.70502782e-01 -1.82854503e-01 9.83268023e-01 4.54681575e-01 1.30858704e-01 2.79127568e-01 3.48952800e-01 -4.74732406e-02 -3.18655580e-01 -1.17699337e+00 -5.15813641e-02 6.53955415e-02 2.77493417e-01 -1.71541035e-01 1.17401659e-01 -5.06614208e-01 5.71045876e-01 -1.32166687e-02 1.14651404e-01 5.06847203e-01 1.22036684e+00 -5.63682735e-01 -1.48181224e+00 -6.86118662e-01 4.12570119e-01 -4.32150573e-01 1.74446777e-01 -3.97727162e-01 5.19409418e-01 9.51071382e-02 1.34402752e+00 -2.76043028e-01 -1.16219863e-01 2.84329921e-01 8.03258643e-02 4.03317660e-01 -7.40395725e-01 -1.36642337e-01 1.82655722e-01 3.07013273e-01 -6.77531302e-01 -3.75688286e-03 -5.72205067e-01 -1.57686627e+00 -1.87395830e-02 -7.30907083e-01 6.34282291e-01 3.05352241e-01 1.30924785e+00 2.72537440e-01 5.20070791e-01 5.99242270e-01 4.33121957e-02 -1.04371059e+00 -5.90588331e-01 -4.93757218e-01 1.95748776e-01 4.33766872e-01 -5.34972668e-01 -1.17460825e-01 4.66397740e-02]
[5.463347911834717, 2.77248215675354]
e17c523f-c18b-4a05-bec8-aee3416748b7
fast-and-accurate-least-mean-squares-solvers
1906.04705
null
https://arxiv.org/abs/1906.04705v2
https://arxiv.org/pdf/1906.04705v2.pdf
Fast and Accurate Least-Mean-Squares Solvers
Least-mean squares (LMS) solvers such as Linear / Ridge / Lasso-Regression, SVD and Elastic-Net not only solve fundamental machine learning problems, but are also the building blocks in a variety of other methods, such as decision trees and matrix factorizations. We suggest an algorithm that gets a finite set of $n$ $d$-dimensional real vectors and returns a weighted subset of $d+1$ vectors whose sum is \emph{exactly} the same. The proof in Caratheodory's Theorem (1907) computes such a subset in $O(n^2d^2)$ time and thus not used in practice. Our algorithm computes this subset in $O(nd+d^4\log{n})$ time, using $O(\log n)$ calls to Caratheodory's construction on small but "smart" subsets. This is based on a novel paradigm of fusion between different data summarization techniques, known as sketches and coresets. For large values of $d$, we suggest a faster construction that takes $O(nd)$ time (linear in the input's size) and returns a weighted subset of $O(d)$ sparsified input points. Here, sparsified point means that some of its entries were replaced by zeroes. As an example application, we show how it can be used to boost the performance of existing LMS solvers, such as those in scikit-learn library, up to x100. Generalization for streaming and distributed (big) data is trivial. Extensive experimental results and complete open source code are also provided.
['Dan Feldman', 'Ibrahim Jubran', 'Alaa Maalouf']
2019-06-11
fast-and-accurate-least-mean-squares-solvers-1
http://papers.nips.cc/paper/9040-fast-and-accurate-least-mean-squares-solvers
http://papers.nips.cc/paper/9040-fast-and-accurate-least-mean-squares-solvers.pdf
neurips-2019-12
['data-summarization']
['miscellaneous']
[ 8.25528502e-02 4.98272553e-02 -2.80256599e-01 -2.62807548e-01 -9.89653826e-01 -5.87827742e-01 -1.16150275e-01 2.01729029e-01 -6.29446626e-01 7.33123243e-01 1.07307822e-01 -5.09379506e-01 -4.09710139e-01 -8.42014968e-01 -9.69007432e-01 -9.73480582e-01 -6.53483033e-01 6.28618360e-01 -1.10690676e-01 -4.23313975e-01 1.99784279e-01 2.86239296e-01 -1.49715543e+00 1.89249322e-01 6.56883836e-01 1.17695403e+00 -1.39600560e-01 5.65333486e-01 -4.04605865e-01 2.77769208e-01 -2.35366687e-01 -5.27039230e-01 8.14646959e-01 -2.67746896e-01 -8.14515710e-01 -1.29247472e-01 4.84837413e-01 -1.38062268e-01 -2.98153728e-01 9.57096934e-01 3.46699089e-01 2.83613563e-01 3.48066270e-01 -1.09480226e+00 -1.51391834e-01 1.00946069e+00 -9.07841444e-01 -1.68277413e-01 2.11599395e-01 -1.27436398e-02 1.12466681e+00 -1.03027558e+00 5.36272466e-01 1.06606841e+00 8.13935876e-01 3.20519775e-01 -1.51594043e+00 -9.73970950e-01 2.83927858e-01 -3.95400375e-02 -1.43189204e+00 -3.26675653e-01 6.17041767e-01 -2.24180579e-01 7.82044113e-01 7.10158467e-01 6.74608171e-01 4.84170616e-01 -3.78869921e-01 9.41207230e-01 8.78630400e-01 -3.76255214e-01 3.26820910e-01 4.38600369e-02 1.83134422e-01 8.52518380e-01 4.90099370e-01 -2.48121068e-01 -6.23435795e-01 -5.65918505e-01 6.60423994e-01 3.35097164e-01 -1.25973687e-01 -3.07262510e-01 -1.31641960e+00 1.10665429e+00 1.96293935e-01 2.20272735e-01 -1.85667738e-01 3.91965419e-01 5.63958049e-01 5.70386171e-01 5.05473495e-01 2.28256300e-01 -8.43762755e-01 -1.44436017e-01 -1.11759889e+00 4.76945668e-01 8.93761218e-01 8.89156163e-01 1.00182712e+00 1.25491381e-01 4.24556136e-01 8.58127594e-01 -1.90266788e-01 8.27831864e-01 3.59296739e-01 -1.17378211e+00 7.79330313e-01 6.51143193e-01 -2.54675164e-03 -1.11589837e+00 -3.64945829e-01 -3.35924506e-01 -1.40728128e+00 2.47084834e-02 5.22627950e-01 -3.71862471e-01 -5.07789612e-01 1.83799994e+00 4.65193421e-01 1.17632657e-01 -3.16733450e-01 7.38233984e-01 4.68082815e-01 6.19440079e-01 -3.69514227e-01 -5.81309974e-01 1.21605730e+00 -5.63877404e-01 -3.54230881e-01 -2.78705806e-01 9.23694849e-01 -6.53876722e-01 9.85957265e-01 9.09583271e-01 -1.33893239e+00 -1.14365354e-01 -7.18686521e-01 3.65899801e-02 -2.26830423e-01 -1.43506164e-02 1.22326767e+00 5.63950121e-01 -8.69069278e-01 7.83221960e-01 -6.79079175e-01 5.27159050e-02 4.92259920e-01 6.93049550e-01 -5.57754576e-01 -4.08350490e-02 -8.03275406e-01 3.29262674e-01 -5.76919504e-02 8.09487049e-03 -4.58405644e-01 -8.48665059e-01 -6.96861207e-01 3.29088792e-02 6.52326584e-01 -6.40616238e-01 7.31808960e-01 -7.98577964e-01 -1.05082941e+00 8.11496556e-01 -3.61758798e-01 -5.37831843e-01 3.79924119e-01 -1.54345617e-01 8.13225061e-02 -1.16485149e-01 -2.40341842e-01 2.05697179e-01 8.75843763e-01 -8.09952736e-01 -6.06709123e-01 -7.41707802e-01 -1.16880938e-01 -1.40235603e-01 -4.08657968e-01 1.18361712e-02 -2.86666960e-01 -5.93597114e-01 5.38089216e-01 -1.07342303e+00 -8.20940316e-01 1.34879500e-01 -3.85215104e-01 -2.90711969e-01 1.13928996e-01 -4.23235625e-01 1.47949946e+00 -2.15427494e+00 4.37095940e-01 6.43698454e-01 5.24783075e-01 1.76733673e-01 -1.38306826e-01 5.97678781e-01 -1.69192255e-01 8.21736455e-02 -4.62897003e-01 -3.44290167e-01 4.31844443e-02 1.62096530e-01 -3.38897079e-01 6.92096472e-01 -4.02474523e-01 4.37994063e-01 -5.98918021e-01 -1.87907234e-01 -1.91946551e-01 1.81105524e-01 -8.14809501e-01 -3.18364143e-01 -1.42732516e-01 -1.73214018e-01 -3.31051677e-01 5.05113840e-01 9.42193687e-01 -3.32991868e-01 4.50132459e-01 -1.43434897e-01 -1.54369771e-01 1.37677774e-01 -2.06496978e+00 1.86154878e+00 -3.05018127e-01 2.37317190e-01 4.69718367e-01 -1.54052806e+00 8.65648866e-01 5.17110946e-03 8.68789196e-01 -3.19783598e-01 3.04464307e-02 5.14265537e-01 -2.78059989e-01 -3.45075339e-01 2.76877761e-01 -4.13598776e-01 -2.81369478e-01 7.53994405e-01 -2.26621300e-01 -2.14351192e-02 3.74256283e-01 4.33560133e-01 1.49326968e+00 -2.87876606e-01 1.56550318e-01 -1.55485258e-01 4.46374238e-01 9.45319459e-02 7.25608587e-01 6.44158363e-01 2.81266391e-01 3.48107874e-01 7.64479160e-01 -7.94247746e-01 -1.04796135e+00 -8.42489839e-01 -1.30602866e-01 1.39842129e+00 -2.47216433e-01 -7.34921157e-01 -6.59007251e-01 -5.10331929e-01 3.28165412e-01 2.54677624e-01 -6.39939487e-01 1.50863767e-01 -7.84055233e-01 -8.76219630e-01 3.30767870e-01 4.46927935e-01 -3.71610001e-02 -7.17406988e-01 -2.84737170e-01 1.48497403e-01 2.04686588e-03 -5.41572690e-01 -4.60328609e-01 5.42831659e-01 -1.24837554e+00 -9.89718974e-01 -5.76624155e-01 -6.87634349e-01 8.75523806e-01 4.95658934e-01 8.69338036e-01 -8.65333900e-03 -3.93039078e-01 -6.97003826e-02 -1.36957511e-01 -3.34457040e-01 1.11342870e-01 1.02521647e-02 4.35888588e-01 -2.04246596e-01 3.96940440e-01 -8.73922348e-01 -8.29676449e-01 1.29949361e-01 -1.05051005e+00 -7.34074786e-02 6.41586423e-01 8.86846662e-01 9.48681295e-01 4.76035848e-02 3.34757656e-01 -1.25829315e+00 3.38353574e-01 -5.47669351e-01 -7.39830256e-01 -1.50212329e-02 -6.83911145e-01 2.64584810e-01 1.04330420e+00 -5.99910617e-01 -3.13638866e-01 3.23746830e-01 -2.12471306e-01 -5.36308587e-01 4.21791881e-01 7.34561503e-01 2.07645908e-01 -4.33049574e-02 8.60722780e-01 3.53526741e-01 2.38192767e-01 -8.13152969e-01 5.52080810e-01 5.90073347e-01 2.33877108e-01 -6.37331903e-01 9.04015541e-01 7.48101532e-01 1.93686530e-01 -6.39074385e-01 -6.16468668e-01 -5.14488816e-01 -2.54249662e-01 3.86456668e-01 2.79270522e-02 -9.11033988e-01 -1.34935045e+00 3.15862894e-02 -5.89638710e-01 -2.85355031e-01 -5.49243033e-01 3.33188444e-01 -4.91706967e-01 2.63849467e-01 -3.72956127e-01 -7.90202200e-01 -7.16092527e-01 -7.50829160e-01 8.27520669e-01 -1.76480666e-01 -1.70624241e-01 -4.82139558e-01 1.91584885e-01 4.48734462e-01 3.72134477e-01 6.35544360e-02 9.15249705e-01 -7.18583584e-01 -4.71695781e-01 -3.97030592e-01 -5.03502935e-02 4.86606419e-01 -1.39754474e-01 -2.46065184e-01 -4.18828696e-01 -4.40378129e-01 7.47230602e-03 -1.36910051e-01 9.89780724e-01 3.29959184e-01 1.39607501e+00 -8.31543863e-01 -3.07603717e-01 9.16091323e-01 1.67403221e+00 -2.62575418e-01 5.33644140e-01 -1.23288587e-01 5.22419930e-01 3.95608097e-01 3.84887367e-01 1.00852776e+00 2.01065809e-01 2.71860778e-01 2.16681495e-01 -1.87014401e-01 3.49179268e-01 1.96795389e-02 2.82747000e-01 9.88811255e-01 -2.28846610e-01 3.95638347e-01 -7.13269770e-01 4.89353418e-01 -1.86197233e+00 -1.08905387e+00 -2.46776462e-01 2.51024437e+00 1.20623159e+00 1.43028110e-01 4.99695182e-01 3.94389033e-01 2.25470513e-01 -1.01822792e-02 -6.50132000e-01 -6.44653618e-01 -1.37897030e-01 9.49959755e-01 8.10779989e-01 3.52931976e-01 -8.17248166e-01 7.11549342e-01 5.99159193e+00 1.29508042e+00 -9.91521060e-01 -3.18066142e-02 5.95483422e-01 -7.19325662e-01 -5.28633595e-01 -2.99090315e-02 -7.42062747e-01 4.16043192e-01 8.49122107e-01 -2.56336987e-01 8.42793822e-01 1.09646475e+00 1.20543458e-01 -2.18891561e-01 -1.06528866e+00 1.29183435e+00 -6.71199560e-02 -1.70451152e+00 -3.77585888e-01 3.10735673e-01 7.19928861e-01 8.57657567e-02 4.38988283e-02 2.14754045e-01 4.59250748e-01 -9.62073028e-01 2.62229592e-01 3.85855436e-02 9.42951441e-01 -8.02154839e-01 4.33750629e-01 5.08675933e-01 -1.16106808e+00 -2.94048786e-01 -6.24666154e-01 -6.85841218e-02 -1.31126881e-01 9.41854656e-01 -4.08419669e-01 2.79359490e-01 8.06712210e-01 3.89761776e-01 2.82405730e-04 6.38011217e-01 3.56567591e-01 6.32219493e-01 -9.90851283e-01 -1.61475196e-01 2.33025581e-01 -5.33361256e-01 2.85738289e-01 1.08359754e+00 4.35294479e-01 4.27165389e-01 6.37735426e-02 3.91519517e-01 -2.82338172e-01 4.70760286e-01 -5.25210083e-01 8.95590484e-02 4.45328772e-01 1.23122299e+00 -5.03311098e-01 -3.86123210e-01 -5.16454279e-01 5.82370877e-01 3.46345186e-01 1.59011349e-01 -7.57059216e-01 -5.74710369e-01 8.42902899e-01 4.55666155e-01 6.07088268e-01 -4.26966339e-01 -4.36289907e-01 -1.22997928e+00 1.66430935e-01 -1.22721279e+00 6.39094830e-01 -1.13877758e-01 -1.28047502e+00 2.56969094e-01 -2.40385532e-01 -1.18311954e+00 1.14784196e-01 -5.75607121e-01 -2.13383466e-01 7.10958660e-01 -1.06906056e+00 -6.87676549e-01 9.40694809e-02 8.94427061e-01 2.51545101e-01 -2.26096045e-02 9.69396055e-01 3.82237941e-01 -5.80391943e-01 7.04391360e-01 6.97435200e-01 3.89960520e-02 6.29009426e-01 -1.16299784e+00 1.23188041e-01 3.78044426e-01 2.38695964e-01 9.55632567e-01 7.80118585e-01 -3.15988153e-01 -2.01858449e+00 -5.96208811e-01 9.17838216e-01 2.74444018e-02 7.60998070e-01 -3.61620873e-01 -8.17086220e-01 7.49053299e-01 -3.02390426e-01 2.69870251e-01 1.05365753e+00 5.14041603e-01 -5.56759775e-01 -7.70369530e-01 -1.14083207e+00 5.23986340e-01 1.27412570e+00 -1.25732258e-01 -8.22077990e-02 6.29904747e-01 4.29359674e-01 -3.65418881e-01 -9.63409543e-01 2.53037900e-01 6.48831487e-01 -9.03105855e-01 1.22935379e+00 -9.40185010e-01 3.43705297e-01 -3.98690738e-02 -3.43467385e-01 -8.86721671e-01 -1.83678925e-01 -1.08684242e+00 -3.40544343e-01 5.91316223e-01 5.54166079e-01 -5.87271273e-01 1.08134377e+00 5.56661308e-01 3.21128309e-01 -1.20695889e+00 -1.10222971e+00 -6.34523749e-01 4.00943130e-01 -5.56268632e-01 5.69905460e-01 9.71496403e-01 2.77463943e-01 1.20408930e-01 -3.94109249e-01 -2.58331627e-01 7.29856193e-01 3.96273106e-01 1.06336486e+00 -1.13921535e+00 -5.15168250e-01 -3.48721862e-01 -1.41028792e-01 -1.12794685e+00 -2.57210314e-01 -1.08361506e+00 -3.63046646e-01 -1.06720757e+00 3.41552496e-01 -9.25118983e-01 -3.69489402e-01 6.10294521e-01 3.26700732e-02 1.33792415e-01 2.02870205e-01 1.28764793e-01 -6.06364489e-01 2.34903097e-01 9.06417549e-01 2.03937641e-03 -2.76228249e-01 9.95674953e-02 -1.03150415e+00 7.16118991e-01 5.76215863e-01 -7.22575188e-01 -4.12373766e-02 -4.77839589e-01 8.88807893e-01 3.63779396e-01 -1.59382880e-01 -6.91404402e-01 2.40727663e-01 -3.79229844e-01 1.26521140e-01 -4.57695812e-01 2.51570940e-01 -7.94109285e-01 2.41739184e-01 5.72014451e-01 -3.26695383e-01 1.42104104e-02 3.86168100e-02 4.26319629e-01 1.00219063e-01 -2.38314390e-01 6.18896067e-01 -3.34763914e-01 -6.61274344e-02 5.09551704e-01 -9.12328809e-02 1.02943420e-01 9.38535213e-01 -4.53596301e-02 -2.28785366e-01 -3.07043195e-01 -7.14335144e-01 1.82884529e-01 1.32727489e-01 -2.26800904e-01 5.68258822e-01 -1.21574616e+00 -8.69672358e-01 2.79244274e-01 -3.73196661e-01 4.03756410e-01 3.88888419e-01 1.06214142e+00 -5.00460029e-01 2.51678228e-01 2.29646638e-01 -4.40604866e-01 -1.21044111e+00 6.42802656e-01 -1.43557861e-01 -3.96053582e-01 -4.92208719e-01 1.27655423e+00 -9.19854492e-02 -2.84237862e-01 3.62076253e-01 -2.80132324e-01 4.72253770e-01 2.83451080e-01 7.64952064e-01 6.49379075e-01 9.77082644e-03 -5.77255860e-02 -4.64047700e-01 5.90211332e-01 -1.69856086e-01 -3.25555094e-02 1.91544020e+00 8.95759016e-02 -4.22616273e-01 2.07867026e-01 1.43628836e+00 3.44122708e-01 -9.09415066e-01 -6.14282608e-01 -2.07415178e-01 -3.97510231e-01 -2.48989254e-01 -3.36953878e-01 -1.32537210e+00 6.73041344e-01 4.06857789e-01 2.77177900e-01 1.32096720e+00 1.30411059e-01 1.10867488e+00 6.67125583e-01 5.72707832e-01 -1.18887210e+00 -1.50664032e-01 2.04795718e-01 7.03381896e-01 -1.16781020e+00 3.95979375e-01 -2.64792949e-01 -5.34944654e-01 1.07910192e+00 2.50766158e-01 -5.38426578e-01 8.31546903e-01 4.76688713e-01 -3.54951024e-01 -1.04733594e-01 -8.91362071e-01 -3.57159413e-02 -2.57573891e-02 8.99514034e-02 3.37084681e-01 1.62361220e-01 -5.80092728e-01 9.52099979e-01 -3.91615957e-01 1.88011393e-01 1.77040607e-01 9.56970513e-01 -7.02111721e-01 -1.30936861e+00 -4.16131735e-01 8.56446862e-01 -4.72003818e-01 -2.94815838e-01 8.36332664e-02 5.56577384e-01 2.27756739e-01 5.22166133e-01 5.42397005e-03 -4.08701777e-01 2.42314667e-01 1.19911283e-01 4.57512200e-01 -3.95062625e-01 -6.93630874e-01 2.32031152e-01 -1.65002510e-01 -7.43692636e-01 -1.08403109e-01 -7.95901716e-01 -1.48985016e+00 -1.12467766e+00 -1.11103505e-01 1.81721643e-01 8.61898601e-01 7.30177701e-01 4.09711927e-01 -6.17545694e-02 7.83126771e-01 -5.12385130e-01 -8.11820507e-01 -6.97172642e-01 -7.60823548e-01 3.63233119e-01 1.48754254e-01 -1.95752978e-01 -5.91546476e-01 4.51954715e-02]
[6.739251136779785, 4.638258934020996]
5b13a2b9-37de-4042-b244-4aec0a253dda
partial-3d-object-retrieval-using-local
2107.03368
null
https://arxiv.org/abs/2107.03368v1
https://arxiv.org/pdf/2107.03368v1.pdf
Partial 3D Object Retrieval using Local Binary QUICCI Descriptors and Dissimilarity Tree Indexing
A complete pipeline is presented for accurate and efficient partial 3D object retrieval based on Quick Intersection Count Change Image (QUICCI) binary local descriptors and a novel indexing tree. It is shown how a modification to the QUICCI query descriptor makes it ideal for partial retrieval. An indexing structure called Dissimilarity Tree is proposed which can significantly accelerate searching the large space of local descriptors; this is applicable to QUICCI and other binary descriptors. The index exploits the distribution of bits within descriptors for efficient retrieval. The retrieval pipeline is tested on the artificial part of SHREC'16 dataset with near-ideal retrieval results.
['Theoharis Theoharis', 'Bart Iver van Blokland']
2021-07-07
null
null
null
null
['3d-object-retrieval']
['computer-vision']
[ 1.91553310e-01 -7.54856229e-01 -4.11688417e-01 -2.98696339e-01 -1.08152127e+00 -6.75608218e-01 7.79318631e-01 5.70210218e-01 -3.77372503e-01 2.15253368e-01 -4.40413654e-02 -6.55357018e-02 -6.25753462e-01 -8.39144528e-01 -1.84450358e-01 -4.99875039e-01 -4.40543264e-01 5.60049951e-01 8.93400669e-01 -1.76858544e-01 8.74920487e-01 1.28351724e+00 -2.02106929e+00 2.56201863e-01 -1.51220217e-01 1.29626715e+00 1.03507712e-01 9.72494006e-01 -1.00724079e-01 6.46095872e-02 -9.82316971e-01 1.61633685e-01 6.73801184e-01 1.14402108e-01 -9.14417624e-01 -6.82568967e-01 8.82275105e-01 -6.63517714e-01 -2.88539380e-01 5.72577953e-01 9.56943870e-01 4.24111038e-02 9.26674545e-01 -1.01142538e+00 -9.81822908e-02 -1.49233341e-01 -3.86219054e-01 6.35692954e-01 9.08533633e-01 -3.49786699e-01 1.01241565e+00 -1.09873509e+00 1.11162031e+00 1.39402866e+00 6.67182267e-01 -2.25937292e-01 -9.58569407e-01 -3.39409739e-01 -5.51521420e-01 4.36013788e-01 -2.09860969e+00 -1.72381382e-02 3.94261807e-01 -2.00555921e-02 1.69704926e+00 6.18689060e-01 8.45055521e-01 6.69712201e-02 6.35143280e-01 6.84380174e-01 6.71218753e-01 -6.42915308e-01 2.93950677e-01 -4.11642015e-01 4.21781391e-01 5.30930281e-01 1.44491881e-01 4.47547287e-01 -5.92367291e-01 -4.00618047e-01 7.52753198e-01 -5.39284013e-02 2.16611966e-01 -5.43292344e-01 -1.22934580e+00 5.95346630e-01 5.84598303e-01 3.99725646e-01 -2.92235911e-01 2.34585062e-01 7.03842103e-01 6.68604136e-01 2.09797487e-01 1.83115706e-01 -3.40908468e-01 -3.31354886e-01 -1.03498411e+00 7.42750049e-01 9.58798707e-01 1.20628893e+00 7.87990570e-01 -4.25219387e-01 -3.13811094e-01 7.90590703e-01 1.40895694e-01 8.24416757e-01 2.74090797e-01 -8.59989941e-01 4.82805856e-02 4.88545388e-01 -1.35334149e-01 -1.29129088e+00 -3.67287129e-01 9.59475115e-02 -3.65695655e-01 4.29748833e-01 -1.85656548e-01 9.56408739e-01 -9.71780002e-01 8.29318345e-01 5.11792421e-01 -3.12943906e-01 -2.36994714e-01 6.45758092e-01 9.44003165e-01 6.76153243e-01 -3.88386339e-01 1.18531890e-01 1.13000417e+00 -4.58784789e-01 -4.06225890e-01 6.35761142e-01 7.66562402e-01 -1.07925367e+00 2.80719012e-01 1.84424028e-01 -1.06247735e+00 -6.11389399e-01 -1.50120080e+00 -5.21075189e-01 -8.12160194e-01 -4.16006088e-01 5.16814172e-01 5.17705441e-01 -1.36801076e+00 5.03331602e-01 -6.43129647e-01 -4.14224386e-01 1.81274444e-01 8.08294952e-01 -2.95690209e-01 -3.15777600e-01 -8.57545793e-01 8.41446161e-01 4.92886633e-01 -3.87473375e-01 -6.44589961e-01 -4.91657197e-01 -6.56828344e-01 -1.49873793e-02 -3.28257799e-01 -2.65858948e-01 9.25784051e-01 9.85031575e-02 -1.26541579e+00 1.18351436e+00 -5.12203649e-02 -4.34438646e-01 -2.80291960e-02 1.29519865e-01 -1.07690632e-01 9.10909116e-01 1.29537955e-01 1.09468865e+00 8.78625989e-01 -6.45140231e-01 -5.84962249e-01 -3.77530485e-01 -3.49860698e-01 2.77488589e-01 1.60875469e-01 3.29907358e-01 -5.91174781e-01 -6.93771243e-01 3.03149998e-01 -7.41236091e-01 2.85573125e-01 4.16985512e-01 1.85727984e-01 -5.89215457e-01 1.37758017e+00 -1.10779889e-01 1.31239831e+00 -2.30217481e+00 -2.58754492e-01 7.60146856e-01 -3.53001244e-02 4.61259514e-01 -3.73415977e-01 6.76766694e-01 2.97714900e-02 -7.33649135e-02 1.55219421e-01 5.28255515e-02 7.73549974e-02 3.91430020e-01 -3.00629586e-01 6.42821133e-01 -5.73183112e-02 9.38603163e-01 -6.00336432e-01 -8.49392593e-01 5.03868163e-01 5.19027710e-01 -3.97421390e-01 -9.05741230e-02 2.87513405e-01 -5.29606104e-01 -5.72740197e-01 1.08680749e+00 1.14138997e+00 1.80482030e-01 -3.97131532e-01 -1.52173713e-01 -4.27094519e-01 3.47095251e-01 -1.22926700e+00 1.78992414e+00 -2.79771406e-02 6.00130379e-01 -1.49821714e-01 -8.51926982e-01 1.35333455e+00 1.06348775e-01 6.01516306e-01 -1.19242609e+00 -1.28729925e-01 5.51737726e-01 -6.46638095e-01 1.16421036e-01 7.38777876e-01 2.76335299e-01 -2.17777744e-01 4.69016224e-01 -3.18308808e-02 -8.14525127e-01 4.15939212e-01 2.55518347e-01 1.20199299e+00 -4.28710766e-02 4.03087258e-01 -6.64768934e-01 7.09843695e-01 1.64282680e-01 -9.44759026e-02 9.67710197e-01 -3.30943406e-01 8.77026320e-01 -9.73981544e-02 -6.46094441e-01 -9.59635377e-01 -1.21770227e+00 -8.15500200e-01 8.55653346e-01 6.84819400e-01 -9.78345096e-01 -1.26199514e-01 -2.42065862e-01 3.69876951e-01 2.65371241e-02 -3.81039232e-01 -3.80843058e-02 -5.94775140e-01 -9.16630551e-02 5.18255770e-01 3.82385671e-01 3.86652231e-01 -7.84663856e-01 -1.16200864e+00 1.11024581e-01 4.48070228e-01 -5.92773795e-01 -5.09974062e-01 3.27286869e-01 -1.09597397e+00 -8.98390591e-01 -6.80709481e-01 -9.43173170e-01 3.21823090e-01 7.92292893e-01 1.14193130e+00 4.28039491e-01 -1.15985537e+00 8.94704044e-01 -5.06620228e-01 -2.69943416e-01 -1.88664526e-01 2.39419360e-02 -2.10614726e-01 -1.08713651e+00 5.02339840e-01 -3.42214614e-01 -7.47379184e-01 5.66719532e-01 -1.15656984e+00 -4.97250021e-01 3.43866229e-01 8.64743114e-01 1.02258384e+00 -1.13511622e-01 -6.50157854e-02 9.74575952e-02 3.97719353e-01 -7.42051974e-02 -9.32815075e-01 1.44795835e-01 -7.90762126e-01 3.58337581e-01 -2.70430464e-02 -3.17975491e-01 -3.82945031e-01 -1.46866024e-01 -6.78638816e-02 -3.12816381e-01 1.48936898e-01 1.89876884e-01 3.04520160e-01 -6.78104162e-01 6.26306474e-01 1.11619964e-01 1.02523915e-01 -5.76824069e-01 1.41981542e-01 7.83293307e-01 4.09277141e-01 -5.58089614e-01 6.75648868e-01 5.54128349e-01 5.59627831e-01 -1.04631710e+00 1.31200636e-02 -1.08720374e+00 -9.60930586e-01 1.42291322e-01 3.55435640e-01 -5.97613811e-01 -7.69088626e-01 4.47871178e-01 -1.09857965e+00 1.29444236e-02 -7.28830278e-01 2.55890608e-01 -6.64759994e-01 3.89717281e-01 -3.68739814e-01 -3.43010843e-01 -5.53010404e-01 -1.17878032e+00 1.72452343e+00 2.90110502e-02 -1.00196324e-01 -6.42368138e-01 5.77553928e-01 -3.63383204e-01 6.09608710e-01 9.05284844e-03 7.48526871e-01 -6.44975781e-01 -9.47414279e-01 -7.40317762e-01 -6.36863112e-01 -1.34417070e-02 -1.51423976e-01 1.69754013e-01 -5.59720695e-01 -5.11600256e-01 -4.66944486e-01 -3.22322041e-01 6.25033081e-01 1.84986368e-01 7.50083327e-01 1.50302857e-01 -6.47187769e-01 6.14237249e-01 1.63207245e+00 4.38456267e-01 7.92237341e-01 5.29539049e-01 -9.68046635e-02 -2.55477857e-02 9.32154894e-01 5.18671036e-01 1.06953286e-01 1.08612359e+00 1.75325543e-01 2.67441303e-01 -5.78101993e-01 9.52849239e-02 1.65662300e-02 8.06670904e-01 2.79164970e-01 9.97023284e-02 -8.78584385e-01 6.48641348e-01 -1.24734128e+00 -9.28643644e-01 8.60806555e-02 2.30496716e+00 6.83974504e-01 -3.99364866e-02 -2.54363060e-01 2.66579658e-01 4.93650109e-01 2.40471080e-01 -2.09237397e-01 -1.04440355e+00 -2.27747262e-01 9.58127081e-01 6.25277162e-01 4.70898062e-01 -1.05748999e+00 6.54518068e-01 8.06187248e+00 1.15755212e+00 -8.56626809e-01 -7.02988356e-02 6.60586283e-02 1.28262907e-01 2.18447857e-02 7.27921277e-02 -1.25659263e+00 4.82281782e-02 9.92146373e-01 -3.33941489e-01 -4.71396744e-02 7.54058301e-01 -5.18309712e-01 -5.59925795e-01 -8.49699318e-01 1.25774837e+00 1.94250584e-01 -1.29820991e+00 4.17056382e-01 1.63508788e-01 5.72642386e-01 2.93775201e-01 3.64492014e-02 2.79664006e-02 -2.43238404e-01 -6.70031250e-01 5.38372040e-01 5.13800919e-01 1.17903411e+00 -9.09184992e-01 7.08213091e-01 -2.17400774e-01 -1.57445657e+00 1.61970630e-01 -6.65157974e-01 2.33701587e-01 -3.35328490e-01 2.37823561e-01 -9.99058008e-01 4.48908150e-01 8.52626503e-01 6.49817705e-01 -8.74258578e-01 1.52154815e+00 4.89144325e-01 -1.57897145e-01 -1.21922004e+00 -2.48383000e-01 1.54763401e-01 1.77634820e-01 7.62818635e-01 1.52679622e+00 4.59798247e-01 1.29772738e-01 9.54284891e-03 1.31775707e-01 2.50061065e-01 2.24487171e-01 -8.59545171e-01 2.53084928e-01 6.48471594e-01 8.86887670e-01 -8.07503223e-01 -3.00152838e-01 -4.85627651e-02 9.20571029e-01 -2.81111002e-01 -1.05389372e-01 -1.65908411e-01 -1.09400976e+00 4.53750670e-01 -5.72058111e-02 7.96662092e-01 -3.60018522e-01 1.28804862e-01 -5.53389251e-01 -1.82582080e-01 -5.31421959e-01 6.42736375e-01 -7.34636068e-01 -7.67084002e-01 2.91206419e-01 6.33496702e-01 -1.49147010e+00 -6.00214064e-01 -7.72740006e-01 -1.36185333e-01 6.97609305e-01 -1.43661225e+00 -1.00376058e+00 -2.94189632e-01 7.93931663e-01 2.50016809e-01 -7.10714310e-02 1.01875198e+00 3.44438255e-01 2.76496649e-01 6.76290154e-01 4.68070447e-01 -3.61807615e-01 7.43888378e-01 -1.03036189e+00 4.44532335e-01 2.54335463e-01 7.72218853e-02 6.35552883e-01 3.73182535e-01 -4.81244802e-01 -1.84128892e+00 -3.28946829e-01 6.76383793e-01 -1.79619342e-01 2.79119134e-01 1.15820520e-01 -6.40635848e-01 1.22556232e-01 -7.28580505e-02 2.71287948e-01 3.35474372e-01 -5.56133270e-01 -5.80576837e-01 -4.61228222e-01 -1.33311296e+00 1.40385598e-01 9.62062955e-01 -7.34943986e-01 -5.44599295e-01 4.82723176e-01 4.58136886e-01 -6.11655533e-01 -1.17008412e+00 5.25124729e-01 8.43931615e-01 -8.30036044e-01 1.57264149e+00 2.48140246e-01 -3.99903446e-01 -2.77252972e-01 -5.11464894e-01 -3.18644762e-01 -1.12908632e-01 -6.85624480e-01 -1.17197022e-01 6.02462530e-01 -1.64233744e-01 -4.30792779e-01 3.71053129e-01 2.68434584e-01 1.70112520e-01 -5.46420336e-01 -1.48520064e+00 -9.85749304e-01 -1.45803332e-01 -3.98928463e-01 5.56182921e-01 2.64587790e-01 -3.41149718e-02 -4.38974462e-02 3.92068684e-01 -2.50395924e-01 5.36566377e-01 4.10180688e-01 7.82799244e-01 -1.31367171e+00 2.28204280e-01 -5.99932015e-01 -1.39263117e+00 -1.36186457e+00 -3.75416964e-01 -1.00800049e+00 -9.56695154e-02 -1.22629714e+00 1.82525292e-01 -6.87226176e-01 -2.19163954e-01 2.86999345e-01 5.06110430e-01 6.57359421e-01 1.66850433e-01 4.02713597e-01 -8.61305654e-01 3.32300752e-01 1.05335653e+00 -2.68076807e-01 -1.48458019e-01 -2.15385064e-01 2.38541424e-01 -1.87207311e-02 3.24404091e-01 -5.46215415e-01 -7.73444325e-02 -1.18663743e-01 2.90340304e-01 -6.86002448e-02 4.51613665e-01 -1.11778247e+00 4.81863767e-01 4.93244797e-01 2.85065830e-01 -1.63539743e+00 5.25130391e-01 -7.73316443e-01 -6.24073250e-03 5.59888661e-01 -1.33141056e-01 5.93628228e-01 4.57636505e-01 3.66909206e-01 -6.05800271e-01 -4.66881543e-01 7.41199255e-01 9.43969116e-02 -8.52294505e-01 1.62084267e-01 -1.46835685e-01 -3.72900188e-01 9.82977927e-01 -7.08772004e-01 -2.11480007e-01 -1.10481292e-01 -3.64725411e-01 -3.97144770e-03 5.54068923e-01 2.30361372e-01 1.13272750e+00 -1.68513238e+00 -5.17956376e-01 6.07070208e-01 3.20616782e-01 -1.03514574e-01 -1.27823830e-01 4.76407737e-01 -1.23372209e+00 9.50606406e-01 -2.06067204e-01 -1.04908013e+00 -1.65787756e+00 5.66647172e-01 3.39628803e-03 -2.90751666e-01 -7.66453028e-01 7.23410308e-01 -3.08745265e-01 -1.04625501e-01 1.93200171e-01 -3.24397534e-01 2.56489575e-01 1.43649176e-01 8.20618391e-01 7.49290407e-01 4.59926963e-01 -6.63619697e-01 -6.74030066e-01 1.25653934e+00 -3.35407019e-01 -3.14275354e-01 1.19656587e+00 -2.99205035e-01 -4.75996435e-01 7.86809176e-02 2.12834477e+00 -3.45372796e-01 -5.84619522e-01 -1.67327702e-01 7.56068081e-02 -8.29165041e-01 3.34479481e-01 -4.62527722e-01 -5.91696799e-01 7.94488847e-01 1.25486255e+00 1.59700856e-01 1.27917850e+00 2.16944162e-02 8.27946782e-01 9.35602605e-01 8.10181379e-01 -9.89629686e-01 -3.22294571e-02 7.12278068e-01 9.79763448e-01 -7.70659387e-01 7.49342263e-01 3.08827199e-02 2.18857780e-01 1.44589186e+00 -2.22216889e-01 -3.52384508e-01 1.11900485e+00 4.46982682e-01 -2.34455660e-01 -4.80230063e-01 -5.71129441e-01 -1.21985950e-01 3.59715790e-01 5.66736460e-01 -1.06912799e-01 -3.51422369e-01 -6.07942581e-01 -7.71059930e-01 1.38091490e-01 -6.29886910e-02 -1.59177527e-01 1.51520061e+00 -5.27143776e-01 -1.28401279e+00 -6.22371078e-01 2.02522650e-01 -2.88632631e-01 -3.55629474e-02 -2.13670731e-01 1.03261173e+00 -3.34947258e-01 4.51991171e-01 3.79501343e-01 -8.16812441e-02 2.02978164e-01 -1.74141914e-01 7.26639986e-01 -2.23658234e-01 -5.97014606e-01 1.59331173e-01 -3.43743354e-01 -1.10498977e+00 -5.24062753e-01 -5.92629313e-01 -1.26116705e+00 -1.30836666e-01 -4.62631166e-01 1.15401298e-01 1.07577610e+00 4.19663191e-01 6.10280812e-01 -1.88448608e-01 8.09992015e-01 -1.25676966e+00 -4.66769129e-01 -5.33942997e-01 -6.32754803e-01 1.77880391e-01 3.46416384e-01 -7.60756195e-01 -3.41445208e-01 -2.30410442e-01]
[10.615935325622559, 0.2708977162837982]
a8b263b6-5056-4d00-8be6-b37a58f7aa93
dopa-a-fast-and-comprehensive-cnn-defense
1905.08790
null
https://arxiv.org/abs/1905.08790v4
https://arxiv.org/pdf/1905.08790v4.pdf
DoPa: A Comprehensive CNN Detection Methodology against Physical Adversarial Attacks
Recently, Convolutional Neural Networks (CNNs) demonstrate a considerable vulnerability to adversarial attacks, which can be easily misled by adversarial perturbations. With more aggressive methods proposed, adversarial attacks can be also applied to the physical world, causing practical issues to various CNN powered applications. To secure CNNs, adversarial attack detection is considered as the most critical approach. However, most existing works focus on superficial patterns and merely search a particular method to differentiate the adversarial inputs and natural inputs, ignoring the analysis of CNN inner vulnerability. Therefore, they can only target to specific physical adversarial attacks, lacking expected versatility to different attacks. To address this issue, we propose DoPa -- a comprehensive CNN detection methodology for various physical adversarial attacks. By interpreting the CNN's vulnerability, we find that non-semantic adversarial perturbations can activate CNN with significantly abnormal activations and even overwhelm other semantic input patterns' activations. Therefore, we add a self-verification stage to analyze the semantics of distinguished activation patterns, which improves the CNN recognition process. We apply such a detection methodology into both image and audio CNN recognition scenarios. Experiments show that DoPa can achieve an average rate of 90% success for image attack detection and 92% success for audio attack detection. Announcement:[The original DoPa draft on arXiv was modified and submitted to a conference already, while this short abstract was submitted only for a presentation at the KDD 2019 AIoT Workshop.]
['Fuxun Yu', 'Zirui Xu', 'Xiang Chen']
2019-05-21
null
null
null
null
['adversarial-attack-detection', 'adversarial-attack-detection']
['computer-vision', 'knowledge-base']
[ 3.17231834e-01 9.39810649e-02 1.30837396e-01 -1.00826100e-01 -6.36066437e-01 -1.12565339e+00 6.17479086e-01 -1.44739464e-01 -1.18398860e-01 3.69761139e-01 -9.46479365e-02 -4.63368744e-01 1.87328503e-01 -9.25895631e-01 -8.43915999e-01 -6.48047984e-01 -1.56084420e-02 -2.49240041e-01 4.44145709e-01 -3.74472797e-01 4.34291689e-03 8.12501550e-01 -1.29035926e+00 6.67127907e-01 2.48026162e-01 1.18555772e+00 -4.95609820e-01 7.35177875e-01 -1.47058042e-02 6.92387104e-01 -1.13303709e+00 -6.65535271e-01 4.99164283e-01 -2.99848288e-01 -8.37662101e-01 -5.69353461e-01 3.73698831e-01 -1.96629137e-01 -9.51949239e-01 1.61228180e+00 6.29934490e-01 -3.11478764e-01 3.78926039e-01 -1.53815758e+00 -5.98615766e-01 9.19200242e-01 -2.10582495e-01 4.03749466e-01 3.40811491e-01 4.39371377e-01 6.10975623e-01 -5.85866451e-01 6.23330213e-02 1.38052416e+00 5.11974514e-01 1.01904106e+00 -7.94838727e-01 -1.24493766e+00 2.43026480e-01 8.62700492e-02 -1.25603378e+00 -3.82730484e-01 9.43558276e-01 -2.28365496e-01 7.57341981e-01 6.17489755e-01 3.34212482e-01 1.65046012e+00 2.10215867e-01 6.99782550e-01 7.73047388e-01 -6.25945628e-02 1.84667259e-01 -1.85405351e-02 -1.40017226e-01 4.07339931e-01 1.20954178e-01 3.08053374e-01 -3.11567336e-01 -2.52291262e-01 3.68185490e-01 -7.05352426e-02 -2.77213693e-01 3.07327747e-01 -9.13424551e-01 4.79767114e-01 6.33942485e-01 2.68360227e-01 -3.01983766e-02 4.26708400e-01 7.47129202e-01 6.32419348e-01 1.80151053e-02 6.89189672e-01 -2.95610338e-01 -6.99520782e-02 -4.12291020e-01 2.20214590e-01 5.41325092e-01 8.28586996e-01 3.63236129e-01 5.37037253e-01 1.60791688e-02 5.32786012e-01 6.76301569e-02 6.15272641e-01 3.04172873e-01 -4.12348628e-01 4.61404800e-01 4.87055004e-01 -3.96624565e-01 -1.29273415e+00 -3.23721379e-01 -2.75334299e-01 -9.86382544e-01 5.13035119e-01 4.75246102e-01 -3.72980386e-01 -8.85940671e-01 1.70591080e+00 1.14129797e-01 2.17408910e-01 3.71687084e-01 8.64060044e-01 8.96327615e-01 4.89336818e-01 1.33805364e-01 2.13848740e-01 1.29393852e+00 -4.48047340e-01 -5.14856339e-01 -1.59200728e-01 4.78698641e-01 -8.09046805e-01 9.79042947e-01 5.18978000e-01 -9.62712705e-01 -5.31688273e-01 -1.36191726e+00 6.02101564e-01 -8.10326397e-01 -3.98388952e-01 3.74695808e-01 1.16255474e+00 -7.49854386e-01 5.09980679e-01 -5.02821386e-01 -5.40666580e-02 5.94294727e-01 4.03792679e-01 -3.40377152e-01 6.22063316e-02 -1.80162370e+00 5.87692738e-01 4.85064954e-01 9.89350453e-02 -1.26906121e+00 -5.97442389e-01 -4.73160416e-01 -1.42436534e-01 1.52271017e-01 -1.02194667e-01 1.01886630e+00 -1.31473005e+00 -1.31095672e+00 7.65058041e-01 5.81782222e-01 -5.54352880e-01 5.16867042e-01 -3.43489684e-02 -1.03015876e+00 2.62852311e-01 -5.34564197e-01 3.61727208e-01 1.26674771e+00 -1.12474084e+00 -2.45562911e-01 -9.90098193e-02 5.20706415e-01 -2.30827585e-01 -8.42807770e-01 4.45719510e-01 -1.79919735e-01 -1.07061303e+00 -8.53814371e-03 -8.92712355e-01 -2.42673978e-02 -2.92311367e-02 -8.71994972e-01 1.96854204e-01 1.17935598e+00 -1.32009566e-01 1.12586117e+00 -2.59307933e+00 -2.32297465e-01 4.32919055e-01 2.92211890e-01 7.15445936e-01 -9.51079726e-02 1.56483158e-01 -6.25577033e-01 5.52327693e-01 -6.57159314e-02 7.72048831e-02 1.19474083e-01 3.52612487e-03 -9.23001528e-01 5.17491341e-01 5.11683166e-01 8.92576277e-01 -8.45059276e-01 -5.39933098e-03 3.84869166e-02 2.71474868e-01 -4.51221168e-01 1.58105686e-01 7.30629712e-02 2.12969840e-01 -5.72691858e-01 9.37708616e-01 8.64996970e-01 3.37894559e-01 -2.25024194e-01 -3.73346984e-01 3.66052747e-01 -4.53995019e-02 -1.10359740e+00 9.44158852e-01 -1.41196966e-01 6.82605386e-01 -4.27008457e-02 -1.05822015e+00 9.66834843e-01 5.05168140e-01 1.78984046e-01 -5.28516948e-01 3.19657564e-01 2.74204433e-01 2.16715783e-01 -2.48378292e-01 1.39855579e-01 3.16307321e-02 -5.46379685e-01 2.92365849e-01 4.46416531e-03 -6.86131418e-02 -5.71716130e-01 3.66927624e-01 1.53550148e+00 -5.68110228e-01 -5.51003702e-02 -2.28143647e-01 5.62333584e-01 -1.46399364e-01 3.85004103e-01 1.02296495e+00 -3.82696360e-01 5.72553575e-01 6.56342983e-01 -6.40790462e-01 -8.13541830e-01 -1.28972507e+00 -7.53955692e-02 7.25193799e-01 2.70674109e-01 -2.31731370e-01 -9.43501711e-01 -1.02368689e+00 -1.30056262e-01 2.30893537e-01 -6.33935809e-01 -9.39925432e-01 -5.52121520e-01 -6.81330919e-01 1.72026443e+00 5.30372500e-01 7.88767934e-01 -1.31197369e+00 -2.29856506e-01 -8.89279842e-02 2.32610375e-01 -1.29555702e+00 -2.72913456e-01 3.11055332e-01 -3.44310433e-01 -1.01063132e+00 -4.29813236e-01 -6.13860488e-01 6.23938441e-01 -4.29111794e-02 7.45633960e-01 3.88121665e-01 -4.59739745e-01 2.95682728e-01 -3.61216336e-01 -8.03738177e-01 -8.18079710e-01 -9.82320588e-03 5.27566433e-01 7.66321272e-02 7.59713799e-02 -6.67586327e-01 -4.69995111e-01 3.98571819e-01 -1.16273487e+00 -7.11413443e-01 4.09637302e-01 5.48854291e-01 1.19969040e-01 3.58229935e-01 8.02195072e-01 -7.30121613e-01 6.37839735e-01 -5.04049540e-01 -3.78938258e-01 3.21001969e-02 -2.07357347e-01 -2.60853916e-01 1.20382500e+00 -1.01184356e+00 -6.66759789e-01 -4.29347567e-02 -3.53613615e-01 -8.36162150e-01 -4.69294399e-01 2.51062870e-01 -8.37749898e-01 -5.08675396e-01 9.12097514e-01 1.79472446e-01 -2.37132326e-01 -1.07981063e-01 9.68122557e-02 4.04285222e-01 8.36110115e-01 -5.78526855e-01 1.48330605e+00 3.65049988e-01 5.68026863e-02 -6.75151765e-01 -4.88144070e-01 1.82995573e-02 -2.81065285e-01 -4.40690428e-01 6.33952796e-01 -7.20599174e-01 -7.74197102e-01 1.13454211e+00 -1.19263494e+00 -1.93159059e-01 -5.46855405e-02 1.35880873e-01 -3.55383679e-02 4.88508224e-01 -6.43207848e-01 -5.75029969e-01 -1.53858632e-01 -1.19401896e+00 5.76826453e-01 7.13963881e-02 -1.26492113e-01 -9.02663052e-01 -2.30683625e-01 -8.56659785e-02 4.69354182e-01 4.51138437e-01 8.72949123e-01 -1.08792925e+00 -4.33772147e-01 -6.85515165e-01 -2.05410182e-01 6.05868876e-01 -2.28147134e-02 2.42900252e-01 -1.52293134e+00 -2.05235839e-01 -7.89330825e-02 -5.36457181e-01 7.17945874e-01 -1.92370698e-01 1.48544121e+00 -4.72736895e-01 -2.31877908e-01 7.27048576e-01 1.23305964e+00 3.14571083e-01 1.07885373e+00 3.14238757e-01 8.87954593e-01 3.01780909e-01 2.25214407e-01 2.88562775e-01 -5.93938172e-01 6.22583628e-01 1.11304796e+00 -2.41665348e-01 -7.55778626e-02 -1.91515103e-01 7.73669124e-01 3.93038988e-01 2.19908938e-01 -5.66539407e-01 -9.25542533e-01 2.76047051e-01 -1.34453738e+00 -1.03094113e+00 -1.08207479e-01 1.95441341e+00 6.89294577e-01 6.77429497e-01 -1.71889644e-02 5.92920959e-01 7.72316575e-01 2.12439239e-01 -5.97539604e-01 -7.14322329e-01 -4.21668112e-01 4.23084557e-01 6.50393784e-01 1.68799683e-01 -1.25206661e+00 9.83582973e-01 6.81975174e+00 1.10184443e+00 -1.30644667e+00 -2.67650709e-02 4.18264031e-01 -1.94896832e-02 -2.72689134e-01 -2.45949358e-01 -5.05448341e-01 5.01256526e-01 8.55855227e-01 -7.23119068e-04 2.59950250e-01 8.36667061e-01 -2.29609683e-01 5.58270216e-01 -9.20161843e-01 8.14457774e-01 -2.15468884e-01 -1.20928502e+00 3.55011523e-01 8.16479325e-03 5.27798653e-01 -2.28321031e-01 3.98541600e-01 2.88896054e-01 3.37947249e-01 -1.27443254e+00 8.54538441e-01 3.20170760e-01 8.08591127e-01 -9.64651704e-01 8.10580194e-01 8.30938220e-02 -1.16511738e+00 -2.72099555e-01 -3.91814321e-01 -1.78984012e-02 -3.20845991e-01 5.15712738e-01 -5.00478983e-01 4.81707662e-01 8.39915574e-01 2.65103340e-01 -5.58191836e-01 6.28745258e-01 -1.63416564e-01 8.59487653e-01 -1.97113842e-01 9.18944553e-03 2.92814910e-01 6.37588918e-01 7.89729714e-01 1.26957560e+00 1.57427967e-01 3.08549870e-02 -3.47505957e-02 7.08152175e-01 -3.50734115e-01 -1.46114200e-01 -1.00116169e+00 -7.94923753e-02 6.65988564e-01 1.02625704e+00 -7.34867334e-01 -8.03202391e-02 -1.05019324e-01 1.09846902e+00 -6.35390505e-02 3.83371264e-01 -1.19595838e+00 -7.49216795e-01 1.12450433e+00 -2.66845226e-01 -5.08756749e-02 6.35815933e-02 -2.72423416e-01 -8.46716285e-01 1.11042880e-01 -1.15253413e+00 3.89650106e-01 -4.80679274e-01 -1.40361500e+00 7.81565785e-01 -1.56902090e-01 -1.32097375e+00 2.06402689e-01 -7.44766831e-01 -1.11890459e+00 7.46103406e-01 -1.01255679e+00 -1.05625916e+00 -2.57781297e-01 1.04113770e+00 2.33813047e-01 -5.74413300e-01 9.53121185e-01 4.89870131e-01 -7.48491228e-01 1.39909399e+00 -3.25131267e-01 6.76151276e-01 6.62212551e-01 -9.24573600e-01 7.14396238e-01 1.12631750e+00 1.23229502e-02 3.97160530e-01 6.47631824e-01 -4.32644606e-01 -1.33894086e+00 -1.35403800e+00 3.48563403e-01 -4.11983281e-01 9.27946150e-01 -5.02174497e-01 -1.03691792e+00 4.32470143e-01 9.39515382e-02 2.66290039e-01 5.02262831e-01 -3.93911451e-01 -7.82179117e-01 -1.52842358e-01 -1.28111494e+00 8.72733831e-01 1.03003013e+00 -8.68485928e-01 -2.72091657e-01 2.22829238e-01 1.02614570e+00 -4.22480941e-01 -6.70268238e-01 4.72305089e-01 3.72981817e-01 -7.85445571e-01 1.23190022e+00 -8.23453009e-01 4.38266665e-01 -3.87624383e-01 -2.76769847e-01 -1.02426207e+00 -5.97676858e-02 -8.35681498e-01 -3.04888226e-02 1.34726560e+00 4.58886445e-01 -7.92064548e-01 4.74346876e-01 2.55849779e-01 -3.06530565e-01 -5.19492149e-01 -1.08113062e+00 -9.52095270e-01 2.17156455e-01 -8.69784832e-01 7.95558810e-01 1.24896967e+00 5.12080118e-02 -3.31608206e-01 -3.86297852e-01 7.20660686e-01 4.23969597e-01 -6.51040435e-01 7.52476215e-01 -8.34249854e-01 -1.48499846e-01 -6.87545359e-01 -9.58509922e-01 -4.92901862e-01 2.15671226e-01 -9.36254799e-01 -2.04277020e-02 -5.92229962e-01 -3.11278939e-01 -4.41250354e-01 -6.19401872e-01 7.09746778e-01 -7.62231946e-02 8.70173156e-01 9.13891941e-02 1.06282555e-01 -2.21090794e-01 1.07676975e-01 8.85042071e-01 -4.95449394e-01 2.82768875e-01 1.67108938e-01 -6.62799060e-01 7.33385086e-01 1.07554340e+00 -6.59775078e-01 -4.39154804e-01 -2.20326588e-01 2.88763434e-01 -3.56251210e-01 7.21584380e-01 -1.34725583e+00 9.08519998e-02 -1.78332403e-01 2.98248768e-01 -1.51184157e-01 1.83962092e-01 -6.92262948e-01 8.71690214e-02 6.29827917e-01 -3.65118027e-01 4.69630025e-02 6.73417151e-01 5.67288876e-01 -2.89540768e-01 -1.57836914e-01 9.88532245e-01 -1.90096460e-02 -5.85157037e-01 4.19208974e-01 -7.48456180e-01 8.07164386e-02 1.15902424e+00 -2.69877672e-01 -4.31218088e-01 -4.09772754e-01 -7.45406628e-01 -1.56972840e-01 2.85294026e-01 6.20430887e-01 8.37372601e-01 -1.43959343e+00 -4.46525365e-01 3.35161120e-01 9.23180208e-02 -2.76454210e-01 3.10645759e-01 4.53315318e-01 -4.28894520e-01 -1.18284207e-03 -3.32388639e-01 -3.80507112e-01 -1.30395865e+00 9.12665009e-01 6.18053198e-01 3.83060612e-02 -5.48538983e-01 9.78753924e-01 2.89045751e-01 -1.98579207e-01 4.73908603e-01 -6.67188019e-02 6.61977613e-03 -2.47723535e-01 6.66682839e-01 -5.70900925e-03 1.79210842e-01 -3.85062099e-01 -5.70598483e-01 3.23921472e-01 1.61200147e-02 2.24183947e-01 6.86392188e-01 3.08867514e-01 9.74970981e-02 9.95610058e-02 1.33763099e+00 1.67532772e-01 -8.04830134e-01 -6.91275075e-02 -2.53849566e-01 -3.04307550e-01 -2.13835120e-01 -6.56380713e-01 -1.43096673e+00 1.03395200e+00 6.08682930e-01 6.17802501e-01 1.33926368e+00 -4.56442721e-02 7.90297747e-01 4.02872920e-01 9.52895060e-02 -6.87406659e-01 4.24090266e-01 5.97290874e-01 7.81746387e-01 -8.67527962e-01 -3.78523678e-01 -3.36785555e-01 -4.48355943e-01 1.28387189e+00 8.27720344e-01 -2.37199664e-01 7.98683643e-01 6.67508364e-01 2.99180210e-01 -1.60763457e-01 -5.06144822e-01 2.93570638e-01 1.36208177e-01 7.70694733e-01 -1.34693056e-01 -5.09653464e-02 3.99873137e-01 7.61030018e-01 -2.86610454e-01 -7.43219018e-01 5.21111071e-01 7.25777924e-01 -1.46033302e-01 -1.10506475e+00 -5.37980139e-01 4.30182628e-02 -8.23462307e-01 -1.85185045e-01 -8.91315401e-01 6.38927996e-01 1.73711225e-01 8.17460954e-01 -9.41087604e-02 -1.00969112e+00 4.59627777e-01 8.13225657e-03 1.32165670e-01 -2.72635490e-01 -1.04938114e+00 -4.97483492e-01 -1.07978329e-01 -6.29973531e-01 -4.41329665e-02 -3.54660362e-01 -1.16291022e+00 -5.38620710e-01 -2.48662785e-01 -1.11820392e-01 5.09321153e-01 8.98365796e-01 1.82357237e-01 7.70605385e-01 1.03747201e+00 -5.72177589e-01 -6.42160118e-01 -6.27367616e-01 -4.37727779e-01 4.56777960e-01 3.94163698e-01 -4.84933764e-01 -7.47994900e-01 -1.87438756e-01]
[5.59773063659668, 7.898801803588867]
b2a4bf14-fe86-428a-8234-1b1730e5f3d0
black-box-testing-of-deep-neural-networks
2112.12591
null
https://arxiv.org/abs/2112.12591v5
https://arxiv.org/pdf/2112.12591v5.pdf
Black-Box Testing of Deep Neural Networks Through Test Case Diversity
Deep Neural Networks (DNNs) have been extensively used in many areas including image processing, medical diagnostics, and autonomous driving. However, DNNs can exhibit erroneous behaviours that may lead to critical errors, especially when used in safety-critical systems. Inspired by testing techniques for traditional software systems, researchers have proposed neuron coverage criteria, as an analogy to source code coverage, to guide the testing of DNN models. Despite very active research on DNN coverage, several recent studies have questioned the usefulness of such criteria in guiding DNN testing. Further, from a practical standpoint, these criteria are white-box as they require access to the internals or training data of DNN models, which is in many contexts not feasible or convenient. In this paper, we investigate black-box input diversity metrics as an alternative to white-box coverage criteria. To this end, we first select and adapt three diversity metrics and study, in a controlled manner, their capacity to measure actual diversity in input sets. We then analyse their statistical association with fault detection using four datasets and five DNN models. We further compare diversity with state-of-the-art white-box coverage criteria. Our experiments show that relying on the diversity of image features embedded in test input sets is a more reliable indicator than coverage criteria to effectively guide the testing of DNNs. Indeed, we found that one of our selected black-box diversity metrics far outperforms existing coverage criteria in terms of fault-revealing capability and computational time. Results also confirm the suspicions that state-of-the-art coverage metrics are not adequate to guide the construction of test input sets to detect as many faults as possible with natural inputs.
['Mojtaba Bagherzadeh', 'Ramesh S', 'Lionel Briand', 'Manel Abdellatif', 'Zohreh Aghababaeyan']
2021-12-20
null
null
null
null
['dnn-testing']
['adversarial']
[ 3.46518815e-01 2.33185459e-02 1.22346513e-01 -4.06465530e-01 -1.51906639e-01 -4.13120180e-01 4.59198266e-01 2.60525227e-01 -5.25727332e-01 8.67931247e-01 -4.40985262e-01 -7.44091153e-01 -5.66357374e-01 -9.54426050e-01 -7.74518132e-01 -4.73676831e-01 -1.47956824e-02 3.49147797e-01 5.53962529e-01 -6.65737540e-02 3.64626884e-01 7.71520436e-01 -2.28916454e+00 2.12743655e-01 1.03379333e+00 1.03405333e+00 9.85088870e-02 4.36677665e-01 -2.81042755e-02 6.49967015e-01 -1.11856520e+00 -2.60536015e-01 4.04741541e-02 -6.76086903e-01 -6.35483325e-01 -7.21430928e-02 1.36028782e-01 -1.85203165e-01 -6.10329909e-03 1.38945007e+00 3.64384711e-01 -2.64987588e-01 3.88928503e-01 -1.49524415e+00 -1.21016823e-01 5.60799956e-01 2.90425420e-01 4.26544815e-01 2.13886023e-01 3.87224048e-01 6.64270759e-01 -4.49726582e-01 5.27506948e-01 9.27879632e-01 5.81565797e-01 6.12846911e-01 -9.73118901e-01 -4.35630947e-01 -4.19314988e-02 2.56727576e-01 -1.05915916e+00 -3.59436542e-01 5.85836828e-01 -4.96302396e-01 1.16336799e+00 3.96953762e-01 9.04966474e-01 1.22455609e+00 4.72030878e-01 4.41830426e-01 8.86116564e-01 -6.46757841e-01 7.75062978e-01 1.41677544e-01 1.16588987e-01 4.65247095e-01 1.11398065e+00 3.10958236e-01 -2.19354779e-01 1.63139384e-02 6.29007161e-01 -1.96030661e-01 -2.51145929e-01 -1.72393322e-01 -9.12675619e-01 8.20436716e-01 2.28845254e-02 7.53121376e-01 -2.69916862e-01 1.53450787e-01 6.07577562e-01 6.30257308e-01 1.56776160e-01 6.91558003e-01 -4.30141062e-01 -3.96852612e-01 -7.68766165e-01 1.56776741e-01 8.54626596e-01 6.56877935e-01 4.92654324e-01 4.45644289e-01 -1.72765218e-02 5.23131430e-01 2.38957271e-01 1.52969718e-01 6.92908585e-01 -5.41451871e-01 1.22997299e-01 1.02915263e+00 -2.92083114e-01 -7.28330553e-01 -5.04338086e-01 -6.52797222e-01 -7.24748254e-01 6.05574369e-01 2.73013800e-01 -1.81203187e-01 -8.06469381e-01 1.51654649e+00 -7.05756322e-02 -9.96586680e-02 1.73582718e-01 6.67391598e-01 5.08856773e-01 1.37526430e-02 -4.15174991e-01 -1.08644865e-01 9.18842793e-01 -3.81014764e-01 -5.42218804e-01 -3.16663444e-01 9.76575196e-01 -2.86135852e-01 1.03750455e+00 8.80159557e-01 -8.25937510e-01 -5.06893873e-01 -1.49112129e+00 7.39826560e-01 -3.76448214e-01 -5.06646745e-02 4.04433250e-01 1.11259460e+00 -1.15515423e+00 7.99522936e-01 -7.60269821e-01 -3.66336435e-01 4.52853888e-01 4.03283447e-01 -1.81841522e-01 -1.76308647e-01 -1.12926757e+00 1.17858255e+00 5.11668682e-01 1.45575359e-01 -1.16661453e+00 -2.22458526e-01 -6.58653319e-01 -6.59414902e-02 2.79350340e-01 -2.98967868e-01 1.05685329e+00 -9.76748526e-01 -1.03604174e+00 4.47498590e-01 3.90715510e-01 -6.55655921e-01 6.57353938e-01 5.83606735e-02 -8.76162827e-01 9.21345353e-02 -1.73637658e-01 4.44878250e-01 4.86401349e-01 -1.11947918e+00 -5.11455059e-01 -1.26402229e-01 1.87256023e-01 -3.94501805e-01 -2.00702414e-01 -1.80499285e-01 8.07852373e-02 -4.77442980e-01 3.00460532e-02 -6.51863217e-01 -9.02413651e-02 -5.95178641e-02 -4.18952048e-01 -1.46869585e-01 6.89462066e-01 -1.59108922e-01 1.33873510e+00 -2.14832711e+00 -2.40240902e-01 5.15606344e-01 1.97453365e-01 4.43739384e-01 -9.46623981e-02 2.19392881e-01 -7.44012073e-02 2.91087091e-01 -3.75644535e-01 2.83746213e-01 8.48614052e-02 4.86184180e-01 5.42097688e-02 5.02822042e-01 5.94408810e-01 6.04020000e-01 -6.62280798e-01 -2.06515580e-01 3.04670841e-01 4.88117158e-01 -4.73250836e-01 -1.46426097e-01 -2.29637310e-01 1.12575427e-01 -2.15333939e-01 6.72294259e-01 2.64930844e-01 1.18699230e-01 -2.94254627e-02 1.16453744e-01 1.62080437e-01 3.83541763e-01 -1.04394674e+00 1.04864657e+00 -1.33412391e-01 9.90360975e-01 -6.66397512e-01 -1.24647248e+00 1.27436900e+00 3.62653255e-01 1.41235828e-01 -1.02881587e+00 3.60552460e-01 6.24685884e-01 7.77921617e-01 -5.62195659e-01 1.92094386e-01 2.87314225e-02 1.14705980e-01 2.18794778e-01 2.85265028e-01 5.82698733e-02 4.99197841e-01 -4.78950709e-01 1.63555479e+00 -2.16571555e-01 1.52846038e-01 -3.34947705e-01 5.04936159e-01 -6.30738437e-02 4.34679896e-01 7.92608857e-01 -2.28477135e-01 5.84579945e-01 1.01380026e+00 -3.48954678e-01 -1.02263427e+00 -6.08888865e-01 -4.13732320e-01 2.45713577e-01 -2.17376396e-01 -1.63350999e-01 -1.03425670e+00 -7.98819780e-01 -1.44111842e-01 9.51549292e-01 -7.35811174e-01 -5.25862992e-01 -2.09104404e-01 -6.21927798e-01 9.85156596e-01 4.81071055e-01 3.35779607e-01 -1.33831620e+00 -1.42374063e+00 3.08390588e-01 4.73660171e-01 -8.01132262e-01 4.32922781e-01 7.31284380e-01 -1.11324525e+00 -1.38755190e+00 -4.14320678e-01 -3.03497463e-01 7.55585551e-01 -1.62083313e-01 1.17992198e+00 5.82283497e-01 -2.78482854e-01 1.88285753e-01 -7.45352328e-01 -4.74176139e-01 -8.63408744e-01 -1.68364853e-01 8.76285285e-02 -4.03422266e-01 5.21535695e-01 -5.46755850e-01 -2.39640996e-01 4.80733842e-01 -1.46794415e+00 -5.07511556e-01 9.58213568e-01 6.87743902e-01 3.07392061e-01 4.24976379e-01 8.89071703e-01 -8.43313634e-01 9.13654447e-01 -3.13576221e-01 -6.82037175e-01 3.40612173e-01 -9.81413245e-01 2.12669194e-01 6.37096763e-01 -5.85576653e-01 -2.09771201e-01 -4.15004909e-01 -5.11382580e-01 -4.09341007e-01 -4.53413963e-01 6.67671382e-01 -3.44708234e-01 -2.41043493e-01 9.75026727e-01 1.01829395e-01 1.01428896e-01 -2.12514549e-01 -5.09127498e-01 3.55779767e-01 1.70682803e-01 -3.92137468e-01 3.13641876e-01 -9.55314040e-02 8.66250768e-02 -7.04984546e-01 -5.46029694e-02 7.36453831e-02 -3.71571392e-01 -4.22445863e-01 3.66859168e-01 -3.26474160e-01 -5.35527050e-01 4.54381108e-01 -1.10924768e+00 -3.20736021e-01 -5.04193664e-01 4.20723110e-01 -3.76185030e-01 1.08079009e-01 -2.11168423e-01 -8.17546546e-01 2.83940993e-02 -1.46529150e+00 7.02662468e-01 -6.46580383e-02 -3.59439999e-01 -9.02647853e-01 -1.07799232e-01 -3.31951976e-01 3.92943889e-01 3.69965911e-01 1.05009532e+00 -1.07199252e+00 -1.38713986e-01 -5.19424021e-01 1.59218282e-01 7.90443122e-01 -4.53638248e-02 1.21324196e-01 -9.95535016e-01 -1.56622872e-01 2.67683774e-01 -3.20624225e-02 6.16387963e-01 2.88610518e-01 1.04890168e+00 -2.03359246e-01 -9.53158364e-02 1.64931342e-01 1.67243135e+00 6.96904659e-01 9.63877439e-01 6.34700596e-01 2.77445227e-01 7.25783348e-01 4.44950372e-01 4.56362933e-01 -2.85968721e-01 3.11848313e-01 7.77806520e-01 2.07826361e-01 1.11017995e-01 2.82634348e-01 5.28793514e-01 3.71320695e-01 1.64548770e-01 -5.74373126e-01 -1.10522187e+00 4.87886161e-01 -1.55284739e+00 -7.39726424e-01 -2.52536654e-01 2.22636771e+00 5.44068158e-01 9.15996790e-01 6.88099489e-02 9.49497581e-01 5.73891938e-01 -2.59450585e-01 -5.76135278e-01 -6.56528115e-01 -2.31658444e-01 1.31168514e-01 3.17585886e-01 5.11381356e-03 -5.85771680e-01 1.54055029e-01 6.04260397e+00 5.65522730e-01 -1.13596272e+00 1.47483284e-02 6.54189229e-01 -5.08004874e-02 -5.42108595e-01 -2.05537990e-01 -7.25811958e-01 6.22459590e-01 1.19765043e+00 1.51138932e-01 4.73661125e-02 8.82345021e-01 -4.36645523e-02 -3.90377611e-01 -1.36712062e+00 7.43741512e-01 5.11765741e-02 -1.12071013e+00 -9.59814712e-02 2.03757465e-01 5.62228262e-01 -3.10851075e-02 -1.72238141e-01 1.10124640e-01 -3.39793041e-02 -1.19795644e+00 1.04873776e+00 4.77501065e-01 6.05918705e-01 -9.20343220e-01 1.37608850e+00 3.17147255e-01 -4.87514675e-01 -6.49323314e-02 -3.63194942e-01 -2.05598876e-01 -1.51330441e-01 1.04429758e+00 -7.60977447e-01 3.74134839e-01 7.07991660e-01 1.08116038e-01 -8.32197070e-01 1.35609245e+00 -2.05105335e-01 6.19326234e-01 -1.80359423e-01 -4.63623375e-01 2.16850579e-01 2.66569585e-01 3.77326250e-01 1.05619621e+00 5.97347200e-01 -5.59691787e-01 -5.00887632e-01 9.72968221e-01 3.31940681e-01 -2.70800710e-01 -8.84522200e-01 -1.58654258e-01 4.91598248e-01 7.58103251e-01 -1.04082775e+00 -1.25287294e-01 -4.03838962e-01 6.95850790e-01 -1.80121243e-01 1.90007836e-01 -8.22521150e-01 -4.35755879e-01 5.89409351e-01 6.57098815e-02 3.14545989e-01 8.92139599e-03 -5.51215649e-01 -6.22888863e-01 3.06785524e-01 -1.05684638e+00 -1.83966979e-01 -4.37371910e-01 -1.01536107e+00 1.21739721e+00 1.34018913e-01 -1.49575233e+00 -4.57911819e-01 -9.89239872e-01 -4.48942095e-01 6.79743409e-01 -1.07501352e+00 -3.80214751e-01 -2.63832420e-01 2.40909204e-01 2.69303083e-01 -2.86992520e-01 7.75735915e-01 3.24687123e-01 -6.80945754e-01 4.21055019e-01 -5.97123057e-02 3.85376774e-02 1.61213845e-01 -9.37443674e-01 3.67615879e-01 1.25185692e+00 9.22977403e-02 4.47105199e-01 9.47564304e-01 -6.67285144e-01 -1.08855617e+00 -9.69679832e-01 6.97245955e-01 -1.38989970e-01 3.93578500e-01 -2.28930399e-01 -9.59312916e-01 2.65657127e-01 -4.81298193e-02 4.99429554e-02 4.63615656e-01 -1.28038332e-01 -2.82506123e-02 -1.17995739e-01 -1.24779928e+00 3.70819688e-01 9.49514508e-01 -2.46196166e-01 -5.08942902e-01 -7.17772841e-02 5.30274272e-01 -1.23942234e-01 -7.19794035e-01 6.94787323e-01 4.13230926e-01 -1.53643072e+00 4.02127415e-01 -1.94725916e-01 3.56872320e-01 -2.54674554e-01 -2.26700559e-01 -1.18585193e+00 1.25635982e-01 -7.78348073e-02 2.65205111e-02 1.01043868e+00 5.46905875e-01 -8.73440683e-01 8.22395504e-01 4.41709608e-01 -5.73193192e-01 -1.02632356e+00 -1.17585790e+00 -1.05232704e+00 -1.81394160e-01 -9.84202385e-01 7.37917840e-01 6.81882024e-01 -1.97147220e-01 -5.99036068e-02 2.74282068e-01 -1.26339734e-01 1.49018154e-01 -7.12438226e-01 3.17575753e-01 -1.61054242e+00 -2.00516090e-01 -8.91102135e-01 -1.08945775e+00 -2.69611865e-01 2.45039957e-03 -3.84136021e-01 2.54636973e-01 -1.54783523e+00 -3.50109041e-01 -5.08257389e-01 -3.76905203e-01 5.77999651e-01 1.81060940e-01 2.12697610e-01 -1.25864744e-01 -7.64509216e-02 -2.25911617e-01 1.41600817e-01 7.91550100e-01 -1.81209937e-01 3.45278904e-02 8.04442093e-02 -6.30967200e-01 5.62880516e-01 8.27528417e-01 -5.35124481e-01 -7.24224150e-01 -2.48560622e-01 7.18902588e-01 -2.90871561e-01 6.20993078e-01 -1.73884511e+00 9.00877565e-02 1.31595790e-01 2.42554486e-01 -4.52344298e-01 -1.60265550e-01 -1.05950451e+00 3.44608784e-01 9.05258477e-01 -1.10856712e-01 3.38137627e-01 4.75073665e-01 2.46184334e-01 -4.11246002e-01 -9.82626975e-01 3.97910595e-01 -1.42516650e-03 -9.19295132e-01 -1.45392165e-01 -5.72806120e-01 -1.94582567e-01 1.08108878e+00 -8.20795238e-01 -4.15290534e-01 5.74549241e-03 -2.27468342e-01 -1.98568374e-01 5.76071441e-01 3.52669269e-01 8.92089367e-01 -1.08386445e+00 -4.51220334e-01 5.08988380e-01 2.08784521e-01 4.96452078e-02 1.61715433e-01 9.29651797e-01 -9.22789812e-01 5.54931283e-01 -4.82974589e-01 -7.70570993e-01 -9.69562829e-01 5.25566638e-01 4.71882224e-01 3.12120449e-02 -4.49635953e-01 7.69191504e-01 -2.42604926e-01 -1.59314826e-01 4.18351084e-01 -8.73275340e-01 -1.80167869e-01 4.55365144e-03 3.96656245e-01 3.56147707e-01 8.26283872e-01 -1.16415955e-01 -5.51742554e-01 8.78134742e-02 2.79818892e-01 -1.65812284e-01 1.16677809e+00 3.80014658e-01 2.33907960e-02 7.28099763e-01 9.31096375e-01 -4.55680698e-01 -1.06464183e+00 3.78383011e-01 3.85561079e-01 -1.96336657e-01 -6.58653155e-02 -8.75285089e-01 -9.37581956e-01 8.41314912e-01 7.70552337e-01 7.22445905e-01 1.38696373e+00 -2.27876663e-01 1.65753737e-01 4.48612571e-01 4.67096835e-01 -9.60862219e-01 5.69531880e-03 4.28399414e-01 7.13414848e-01 -9.86335814e-01 -3.04621071e-01 -3.06778355e-03 -4.14378464e-01 1.19912219e+00 7.00325370e-01 -7.42027909e-02 4.16050673e-01 5.99532664e-01 -1.24874309e-01 -1.43925115e-01 -9.32184339e-01 -9.06151980e-02 1.84669510e-01 7.19877779e-01 3.89800608e-01 -1.40063748e-01 -3.90581220e-01 5.32585502e-01 -1.68442875e-01 1.92025051e-01 5.97498417e-01 1.01980197e+00 -6.59983218e-01 -1.13826215e+00 -3.86168122e-01 8.51324260e-01 -3.12391639e-01 -6.32826388e-02 -3.90091449e-01 1.10149443e+00 4.56110954e-01 1.11179972e+00 1.49030745e-01 -9.35385525e-01 4.45744216e-01 1.21305309e-01 5.80751836e-01 -5.33555865e-01 -6.96474791e-01 -3.86437744e-01 3.20106208e-01 -3.98929685e-01 -4.91007358e-01 -6.04356527e-01 -9.80773032e-01 -2.92250395e-01 -6.12437010e-01 8.72287601e-02 7.68315077e-01 1.25917971e+00 1.62135154e-01 1.03100729e+00 2.30355576e-01 -4.23751116e-01 -5.71797669e-01 -1.18620801e+00 -4.86795574e-01 4.19355743e-02 3.15029234e-01 -7.39324212e-01 -4.00766343e-01 -3.51415813e-01]
[6.562127113342285, 7.635762691497803]
17fa26e4-c1dd-41b3-9d9f-8bc917b752a1
hierarchical-pronunciation-assessment-with
2211.08102
null
https://arxiv.org/abs/2211.08102v2
https://arxiv.org/pdf/2211.08102v2.pdf
Hierarchical Pronunciation Assessment with Multi-Aspect Attention
Automatic pronunciation assessment is a major component of a computer-assisted pronunciation training system. To provide in-depth feedback, scoring pronunciation at various levels of granularity such as phoneme, word, and utterance, with diverse aspects such as accuracy, fluency, and completeness, is essential. However, existing multi-aspect multi-granularity methods simultaneously predict all aspects at all granularity levels; therefore, they have difficulty in capturing the linguistic hierarchy of phoneme, word, and utterance. This limitation further leads to neglecting intimate cross-aspect relations at the same linguistic unit. In this paper, we propose a Hierarchical Pronunciation Assessment with Multi-aspect Attention (HiPAMA) model, which hierarchically represents the granularity levels to directly capture their linguistic structures and introduces multi-aspect attention that reflects associations across aspects at the same level to create more connotative representations. By obtaining relational information from both the granularity- and aspect-side, HiPAMA can take full advantage of multi-task learning. Remarkable improvements in the experimental results on the speachocean762 datasets demonstrate the robustness of HiPAMA, particularly in the difficult-to-assess aspects.
['Gary Geunbae Lee', 'Yunsu Kim', 'Heejin Do']
2022-11-15
null
null
null
null
['phone-level-pronunciation-scoring', 'word-level-pronunciation-scoring', 'utterance-level-pronounciation-scoring']
['speech', 'speech', 'speech']
[-1.86862215e-01 -1.77531630e-01 -3.53920639e-01 -5.72288871e-01 -1.17163575e+00 -5.81770420e-01 4.28414851e-01 4.53803688e-01 -2.35909954e-01 5.65996170e-01 6.97267592e-01 -3.22532684e-01 -1.91297919e-01 -7.50220597e-01 -3.22507024e-01 -3.13828737e-01 5.73569596e-01 6.68893874e-01 -8.93028975e-02 -2.07337037e-01 1.08337142e-01 1.31728008e-01 -1.45946777e+00 6.83593035e-01 1.26559651e+00 1.05915797e+00 1.36256412e-01 6.07349932e-01 -4.83392864e-01 6.19417369e-01 -6.94031000e-01 -6.69219971e-01 -4.50824291e-01 -3.19085956e-01 -6.88016891e-01 1.36586502e-01 3.46283734e-01 -2.53206968e-01 2.66607046e-01 1.16154051e+00 3.90207916e-01 2.74950303e-02 7.10385859e-01 -8.73090267e-01 -7.21376956e-01 8.47101629e-01 -3.54413450e-01 9.77791771e-02 3.09018224e-01 7.08256438e-02 1.44995058e+00 -1.02368546e+00 -3.80513102e-01 1.39171660e+00 5.45486331e-01 2.67741472e-01 -9.72475529e-01 -8.25388432e-01 6.31798387e-01 3.63180488e-01 -1.09952915e+00 -2.45787352e-01 6.64372623e-01 -4.86863762e-01 9.27086115e-01 2.13075906e-01 8.15911353e-01 7.63979793e-01 1.26238927e-01 9.10800636e-01 1.39892006e+00 -8.50185826e-02 -1.21905424e-01 2.22423479e-01 5.01211345e-01 5.14495254e-01 8.82283151e-02 -7.90817589e-02 -6.94680214e-01 1.58209652e-01 5.90420604e-01 -1.30173683e-01 -1.34457618e-01 1.53330803e-01 -1.14123225e+00 5.51748514e-01 2.31795311e-01 4.25395310e-01 -3.59245926e-01 -1.82274744e-01 4.02356684e-01 2.69775480e-01 4.72768128e-01 6.35445774e-01 -6.34012580e-01 -5.38390219e-01 -7.32627332e-01 -2.84291226e-02 4.25138682e-01 7.63541758e-01 5.47113121e-01 1.98933840e-01 -7.07667589e-01 1.12790215e+00 3.12199891e-01 3.48120123e-01 6.87599957e-01 -6.84300482e-01 7.01206923e-01 8.85712445e-01 -2.30130106e-01 -5.53824365e-01 -2.29199156e-01 -7.82216549e-01 -8.00866067e-01 8.58039185e-02 1.84938312e-01 6.96990862e-02 -7.81356156e-01 2.04019713e+00 2.62807816e-01 -1.69046134e-01 -1.11034445e-01 7.50617564e-01 9.48993981e-01 5.93298376e-01 4.84547704e-01 -2.73778051e-01 1.77818894e+00 -1.04457343e+00 -1.04799581e+00 -2.16333553e-01 5.04229784e-01 -6.98543251e-01 1.72578442e+00 6.18366241e-01 -1.22504878e+00 -9.46586251e-01 -1.02674437e+00 -2.23200142e-01 -3.98566991e-01 3.11739534e-01 6.25036001e-01 4.99588132e-01 -7.58660257e-01 3.19573253e-01 -4.54187334e-01 3.95205438e-01 1.70607686e-01 3.16471010e-01 -1.19626649e-01 -3.67839076e-02 -1.46715057e+00 7.87461042e-01 2.27194965e-01 -1.77812099e-01 -6.62792206e-01 -8.53516936e-01 -8.09635758e-01 4.55712438e-01 1.08192138e-01 -5.04468560e-01 1.29572880e+00 -6.32846534e-01 -1.45214283e+00 4.92261767e-01 -2.53479213e-01 1.33840412e-01 3.09593529e-02 -4.68443066e-01 -6.08383834e-01 -8.57174248e-02 7.75658116e-02 4.45911795e-01 6.54763997e-01 -9.15517807e-01 -7.41735816e-01 -5.71655989e-01 1.84584096e-01 8.26333165e-01 -5.99879146e-01 -1.75171706e-04 -3.41659039e-01 -7.77547479e-01 -6.57331124e-02 -3.38527262e-01 1.88851342e-01 -5.14896810e-01 -3.14248532e-01 -6.43336058e-01 3.94975066e-01 -7.66679108e-01 1.58024430e+00 -2.18331861e+00 2.08665594e-01 -2.61772692e-01 2.41260335e-01 4.58660811e-01 -4.35378775e-02 5.30303232e-02 1.25017002e-01 7.03290626e-02 -1.00457624e-01 -6.67098105e-01 4.08703424e-02 2.99081393e-02 -5.46425134e-02 -1.25501126e-01 2.39539579e-01 9.34160471e-01 -9.61914599e-01 -4.65310007e-01 2.35690847e-01 3.74849945e-01 -7.85937309e-01 3.74023378e-01 -1.79380670e-01 4.08131242e-01 -3.56101513e-01 6.36449277e-01 3.14768016e-01 -1.08645923e-01 -2.48970106e-01 -1.89494103e-01 -3.78222838e-02 9.61637914e-01 -8.44352424e-01 1.44975185e+00 -1.01581216e+00 2.66198397e-01 7.88481068e-03 -8.23456407e-01 8.15647840e-01 5.87972939e-01 5.19967675e-02 -7.15937197e-01 3.55550088e-02 2.19276100e-01 3.20032090e-01 -2.17216566e-01 6.29513204e-01 -5.25707603e-01 -4.50141877e-01 4.88532007e-01 2.58268565e-01 -3.06854844e-01 3.25454660e-02 -4.17121537e-02 4.94442284e-01 -2.16705427e-01 6.08355701e-01 -2.49930218e-01 6.21395051e-01 -4.44402426e-01 4.45284843e-01 4.66919601e-01 -3.29486877e-01 5.58818102e-01 4.30461675e-01 4.38727532e-03 -6.29381001e-01 -1.14104891e+00 -1.79165840e-01 1.32045197e+00 -6.40818179e-02 -5.87282479e-01 -6.82011008e-01 -6.12977445e-01 -1.06608465e-01 9.12656605e-01 -4.72653180e-01 -4.30133045e-01 -3.31308246e-01 -4.25291747e-01 4.39267606e-01 8.64273131e-01 5.41769922e-01 -1.15789783e+00 -2.10944768e-02 3.45564187e-01 -4.77628171e-01 -1.18642581e+00 -7.52896428e-01 8.67553875e-02 -6.28999889e-01 -7.01366544e-01 -3.81380647e-01 -6.39929533e-01 1.59293383e-01 -8.53918567e-02 1.25194836e+00 -1.03487916e-01 2.80076683e-01 -7.60184303e-02 -2.07870364e-01 -4.01502192e-01 -2.87773818e-01 1.63367242e-01 9.73454267e-02 7.86723793e-02 5.67374587e-01 -7.04199433e-01 -3.75797778e-01 1.74397036e-01 -4.79194432e-01 1.70294985e-01 8.09027076e-01 8.07694316e-01 7.91823387e-01 1.53524112e-02 7.76246965e-01 -8.68443251e-01 9.88169372e-01 -3.53548914e-01 -8.69500488e-02 2.61573017e-01 -5.99067569e-01 -6.44721836e-02 8.82364869e-01 -5.09927392e-01 -9.52306330e-01 -4.66812968e-01 -6.71379566e-01 -2.04053938e-01 -3.88271451e-01 5.63949883e-01 -7.43884683e-01 5.84713817e-01 6.62938952e-02 4.17890429e-01 -1.86676875e-01 -4.95893151e-01 3.88571620e-01 1.01382828e+00 5.12061000e-01 -7.27761984e-01 2.52734005e-01 -2.41571039e-01 -3.45452309e-01 -6.80080175e-01 -1.36552894e+00 -5.28911352e-01 -4.31715012e-01 -7.80493990e-02 7.38360226e-01 -1.03799093e+00 -7.24283159e-01 5.84445357e-01 -1.03412187e+00 -4.70120274e-02 -3.53181958e-01 6.17721796e-01 -5.13973832e-01 3.78292650e-02 -7.78540790e-01 -7.45100796e-01 -3.21304798e-01 -1.41424811e+00 1.22472334e+00 9.26674381e-02 -4.65506256e-01 -9.88930941e-01 -2.31307760e-01 6.70181215e-01 4.42435116e-01 -4.35788095e-01 1.41169190e+00 -7.80543089e-01 -3.53339702e-01 8.01798403e-02 -2.18640000e-01 5.39517522e-01 4.17533576e-01 -3.94135952e-01 -1.07182074e+00 8.05060342e-02 1.40339598e-01 -4.83535975e-01 8.37637722e-01 2.34643906e-01 1.46000433e+00 -4.90881473e-01 2.04988554e-01 4.88263220e-01 1.02268231e+00 2.01573089e-01 3.89498323e-01 -9.37656164e-02 8.24941695e-01 6.05555356e-01 6.58950984e-01 1.57590374e-01 7.00728595e-01 7.46981561e-01 5.18296063e-02 4.47512940e-02 -4.43054587e-01 -4.53994989e-01 3.73495102e-01 1.47383475e+00 2.47302845e-01 -9.22295749e-02 -6.42947078e-01 6.83131158e-01 -1.22692919e+00 -6.75211906e-01 1.55316532e-01 2.17212462e+00 1.21108735e+00 2.75602102e-01 1.19504862e-01 4.96311456e-01 6.90574825e-01 1.85463682e-01 -4.52385575e-01 -6.55856788e-01 1.09286323e-01 9.84135643e-02 -3.44739407e-01 8.19941103e-01 -7.34446704e-01 1.04479408e+00 5.49291229e+00 1.18066204e+00 -9.13048923e-01 2.25406274e-01 7.57964849e-01 5.62797301e-03 -6.42825365e-01 -3.40970814e-01 -1.11180806e+00 5.60693026e-01 8.02757144e-01 -5.48349693e-02 3.43227148e-01 6.37359560e-01 -8.24912190e-02 1.99099779e-01 -1.13996816e+00 7.63072193e-01 1.87501982e-02 -6.34570181e-01 6.45703077e-01 -4.45981584e-02 4.89421636e-01 -3.70432824e-01 4.33527410e-01 7.83217907e-01 1.91380486e-01 -1.13146627e+00 7.24050701e-01 3.13998610e-01 1.15347850e+00 -1.08876514e+00 5.86543560e-01 4.44170415e-01 -1.49299049e+00 8.26499313e-02 -1.09179996e-01 -1.71826854e-01 2.73545474e-01 5.81431806e-01 -5.76183319e-01 4.70552236e-01 2.99896002e-01 3.65172684e-01 -3.99969906e-01 6.84782267e-01 -4.78893071e-01 7.59725988e-01 5.17434590e-02 -2.54932076e-01 1.23246208e-01 -1.65725201e-01 3.81371379e-01 1.14401293e+00 3.56901109e-01 2.51162291e-01 1.20552808e-01 6.36686385e-01 -6.71644360e-02 3.75490516e-01 -2.96863645e-01 -2.63845384e-01 6.36770129e-01 9.99399006e-01 -1.97636560e-01 -5.00217915e-01 -4.80316222e-01 6.19341016e-01 7.16687739e-01 2.73577958e-01 -5.42240858e-01 -3.99362564e-01 1.01506197e+00 4.39748205e-02 1.24316551e-01 -1.59974635e-01 -8.42834532e-01 -1.08524632e+00 1.57926641e-02 -1.02159762e+00 3.83557379e-01 -5.61289072e-01 -1.15779567e+00 8.45778823e-01 -2.58749783e-01 -1.18574941e+00 -2.77671665e-01 -6.11904860e-01 -5.46667933e-01 1.17141449e+00 -1.45484149e+00 -1.12867272e+00 -8.37181434e-02 5.50174654e-01 9.14939344e-01 -1.12762898e-01 8.72378111e-01 2.57680595e-01 -3.45695734e-01 1.04412639e+00 -4.65455711e-01 3.71500105e-02 4.81669843e-01 -1.50977409e+00 1.70193508e-01 6.16831720e-01 2.82794446e-01 6.87662721e-01 2.68627971e-01 -4.82835799e-01 -6.77668512e-01 -1.11801326e+00 1.37190723e+00 -5.05161643e-01 6.09069586e-01 -4.11829412e-01 -1.10985708e+00 3.71291667e-01 5.65682836e-02 -4.49618787e-01 1.02452254e+00 7.71733046e-01 -3.83509547e-01 -3.20358366e-01 -7.19952524e-01 6.35277390e-01 8.59950662e-01 -9.47894931e-01 -9.15542424e-01 -3.74752060e-02 1.07458973e+00 -2.21060306e-01 -9.57726896e-01 5.73717475e-01 4.49086875e-01 -9.23732281e-01 7.93449283e-01 -5.44106364e-01 6.54992044e-01 -5.72691113e-02 -3.54007214e-01 -1.73975778e+00 -4.15441245e-01 -1.60192072e-01 -3.40859413e-01 1.55780184e+00 5.93391657e-01 -4.69332695e-01 1.94991529e-01 1.52727097e-01 -4.78797615e-01 -1.16131222e+00 -8.66617858e-01 -4.83888388e-01 3.10375065e-01 -7.59218752e-01 1.15054822e+00 7.02405572e-01 2.96044469e-01 9.54037189e-01 -2.95988560e-01 6.47296980e-02 3.40647250e-01 3.23196143e-01 1.50130928e-01 -1.15414023e+00 -6.25748217e-01 -8.67829025e-01 -1.08876608e-01 -1.26082838e+00 8.55891183e-02 -9.44382966e-01 -3.05446535e-02 -1.58762527e+00 7.83118904e-02 -5.30505955e-01 -6.36710584e-01 2.28953108e-01 -8.93981278e-01 1.38437189e-02 3.25325057e-02 -6.50016591e-02 -6.04802012e-01 8.68736923e-01 1.65092385e+00 -1.41500086e-01 5.85248582e-02 2.08252206e-01 -1.09091830e+00 7.75699854e-01 8.77418816e-01 -9.26313698e-02 -5.17348111e-01 -6.05734587e-01 -1.32316530e-01 2.95935512e-01 -3.38045985e-01 -1.03266358e+00 4.79692295e-02 -1.47976145e-01 3.77787948e-01 -6.72972441e-01 7.09942937e-01 -3.91573668e-01 -5.35455227e-01 1.77616686e-01 -5.63662350e-01 1.50846601e-01 1.44506991e-01 2.22950459e-01 -7.17294633e-01 9.92557481e-02 7.64777541e-01 -9.09459218e-02 -3.91739726e-01 5.07125795e-01 -2.36992240e-01 3.12640905e-01 5.05348861e-01 1.52083829e-01 -2.64499456e-01 -3.07103723e-01 -6.80315137e-01 3.01799506e-01 6.53887019e-02 5.66844165e-01 5.42597890e-01 -1.62152565e+00 -6.78037822e-01 5.31674802e-01 3.17195058e-01 5.45875132e-02 2.65041232e-01 5.92687190e-01 2.24700600e-01 6.52338862e-01 -1.38778329e-01 -4.05786276e-01 -1.11767197e+00 4.02851701e-01 2.27562994e-01 -5.92672110e-01 -8.63691866e-02 1.00660467e+00 5.69160342e-01 -6.35830522e-01 3.50243539e-01 -5.09351552e-01 -6.51045918e-01 2.66234368e-01 7.02664793e-01 9.57452431e-02 2.56324142e-01 -7.95998812e-01 -7.15560317e-02 7.75760770e-01 -1.38559073e-01 -2.56684005e-01 7.92279243e-01 -2.08540931e-01 5.68445399e-02 9.08993483e-01 9.70407844e-01 2.61545271e-01 -1.05076528e+00 -2.29943946e-01 -2.08525479e-01 -2.91880310e-01 2.85401251e-02 -1.07827127e+00 -7.58448839e-01 1.51031494e+00 -2.00106669e-02 3.18608359e-02 1.12359631e+00 8.13547671e-02 9.18926120e-01 5.29525094e-02 2.50715554e-01 -1.06898916e+00 2.48380944e-01 7.77462840e-01 9.41615343e-01 -1.03581715e+00 -4.58062649e-01 -4.74916995e-01 -8.39204967e-01 6.31966174e-01 1.03419900e+00 3.31643015e-01 5.11008620e-01 2.12565288e-01 3.70995849e-02 -2.16536783e-03 -9.48605418e-01 -2.92013615e-01 6.59674466e-01 4.06998903e-01 8.36037517e-01 6.19413376e-01 -2.89546579e-01 1.25188017e+00 -5.82728326e-01 -4.50723976e-01 1.55641139e-01 2.80907422e-01 -6.58828139e-01 -1.08787334e+00 -1.23221807e-01 6.55457556e-01 -6.39413893e-01 -5.15267432e-01 -2.44347110e-01 4.61739421e-01 1.67310610e-01 9.60265994e-01 -1.97043014e-03 -5.62265754e-01 4.72784936e-01 3.25596333e-01 3.22734773e-01 -9.43435788e-01 -6.93986475e-01 -5.06901853e-02 1.82926789e-01 -3.27690154e-01 1.34040743e-01 -5.56918025e-01 -1.10712588e+00 -5.88947609e-02 -1.46066010e-01 4.84931916e-01 5.01521349e-01 1.27048779e+00 2.54861385e-01 8.84359181e-01 5.74628651e-01 -3.19242239e-01 -7.03232110e-01 -1.42575705e+00 -6.07138515e-01 4.09765750e-01 5.34245133e-01 -8.06719899e-01 -3.39613169e-01 -5.50296068e-01]
[14.31932258605957, 6.821648597717285]
63a243b4-2771-452c-8fcb-2122639056cb
learning-an-animatable-detailed-3d-face-model
2012.04012
null
https://arxiv.org/abs/2012.04012v2
https://arxiv.org/pdf/2012.04012v2.pdf
Learning an Animatable Detailed 3D Face Model from In-The-Wild Images
While current monocular 3D face reconstruction methods can recover fine geometric details, they suffer several limitations. Some methods produce faces that cannot be realistically animated because they do not model how wrinkles vary with expression. Other methods are trained on high-quality face scans and do not generalize well to in-the-wild images. We present the first approach that regresses 3D face shape and animatable details that are specific to an individual but change with expression. Our model, DECA (Detailed Expression Capture and Animation), is trained to robustly produce a UV displacement map from a low-dimensional latent representation that consists of person-specific detail parameters and generic expression parameters, while a regressor is trained to predict detail, shape, albedo, expression, pose and illumination parameters from a single image. To enable this, we introduce a novel detail-consistency loss that disentangles person-specific details from expression-dependent wrinkles. This disentanglement allows us to synthesize realistic person-specific wrinkles by controlling expression parameters while keeping person-specific details unchanged. DECA is learned from in-the-wild images with no paired 3D supervision and achieves state-of-the-art shape reconstruction accuracy on two benchmarks. Qualitative results on in-the-wild data demonstrate DECA's robustness and its ability to disentangle identity- and expression-dependent details enabling animation of reconstructed faces. The model and code are publicly available at https://deca.is.tue.mpg.de.
['Timo Bolkart', 'Michael J. Black', 'Haiwen Feng', 'Yao Feng']
2020-12-07
null
null
null
null
['3d-face-animation', '3d-face-modeling', 'face-alignment', 'face-model', 'face-reconstruction']
['computer-vision', 'computer-vision', 'computer-vision', 'computer-vision', 'computer-vision']
[ 6.78335354e-02 -4.88325320e-02 1.02926977e-01 -7.92182624e-01 -5.78035235e-01 -7.33120084e-01 7.83139944e-01 -7.73512006e-01 1.31419823e-01 5.30281186e-01 2.96065688e-01 2.52132595e-01 3.68129402e-01 -6.65616751e-01 -7.61881530e-01 -8.06147695e-01 4.84618433e-02 5.33557475e-01 -5.67695498e-01 -1.59908831e-01 -3.18865806e-01 1.11434197e+00 -1.66892874e+00 1.84155554e-01 2.74419218e-01 9.42914546e-01 -4.64392066e-01 7.59193540e-01 1.62702844e-01 3.60021561e-01 -4.86455321e-01 -4.89780813e-01 6.71023607e-01 -5.78890264e-01 -2.94075191e-01 4.62863147e-01 1.26678085e+00 -6.30553067e-01 -3.06109935e-01 6.66738510e-01 5.26522517e-01 -1.27370313e-01 8.06279719e-01 -1.33004534e+00 -7.55213082e-01 -5.02238274e-01 -6.62656009e-01 -4.82274652e-01 5.06965637e-01 4.85778332e-01 5.87484539e-01 -9.81639802e-01 1.01847494e+00 1.76454437e+00 7.48727620e-01 9.63301361e-01 -1.93006551e+00 -9.99314487e-01 -1.37879634e-02 -4.94925678e-01 -1.45196593e+00 -8.58192742e-01 1.01438904e+00 -6.35334253e-01 5.65383554e-01 4.00131106e-01 8.64571095e-01 1.50856805e+00 6.95210919e-02 2.26403803e-01 1.42591727e+00 -1.89182341e-01 -5.42360134e-02 2.74926256e-02 -5.67866862e-01 9.15887594e-01 -7.10972175e-02 4.12107736e-01 -5.70267558e-01 -2.72967279e-01 1.30448973e+00 -2.44341940e-01 -4.70911533e-01 -7.12653220e-01 -8.07517648e-01 7.34366477e-01 3.09451967e-01 -2.59552985e-01 -1.57309175e-01 1.88123927e-01 -8.60332996e-02 3.22535694e-01 6.91636384e-01 3.05185795e-01 -3.43850911e-01 1.50097147e-01 -9.62826967e-01 6.48596942e-01 6.43882096e-01 7.63743341e-01 1.09756744e+00 3.82167637e-01 -9.42521244e-02 8.09790075e-01 2.49557406e-01 9.06456113e-01 -1.45449772e-01 -1.49464047e+00 -7.44418353e-02 3.52235764e-01 3.10342517e-02 -1.00418067e+00 -2.47295663e-01 -6.08429983e-02 -6.67528152e-01 1.03279245e+00 4.45163876e-01 -2.25190997e-01 -1.17055774e+00 2.12129259e+00 5.77968717e-01 2.24415168e-01 -2.63344079e-01 1.01173604e+00 8.55775297e-01 4.42359239e-01 -4.61965986e-02 -2.17673868e-01 1.14336097e+00 -6.42837703e-01 -6.29752159e-01 -2.59141356e-01 7.03902030e-03 -8.48371089e-01 1.00631702e+00 2.66281575e-01 -1.42263460e+00 -3.98002416e-01 -7.26284385e-01 -2.67730147e-01 -1.12158079e-02 3.77245471e-02 5.16561270e-01 6.90017343e-01 -1.26077044e+00 4.74529803e-01 -6.85240149e-01 -1.67555317e-01 5.51511526e-01 3.93902272e-01 -7.89415300e-01 9.36726555e-02 -7.01780081e-01 7.97202766e-01 -4.43925589e-01 -2.79086758e-03 -9.53044474e-01 -1.14343357e+00 -1.16565871e+00 -2.43819714e-01 2.02325452e-02 -1.01223719e+00 1.09432507e+00 -1.42446935e+00 -1.84524155e+00 1.37443542e+00 -2.57272094e-01 1.59396455e-01 7.56261706e-01 7.05978870e-02 -2.36846775e-01 2.46803328e-01 -1.28337651e-01 9.19795632e-01 1.26075816e+00 -1.79947460e+00 2.48241618e-01 -5.60578287e-01 -1.65640041e-01 2.05152929e-01 5.47509976e-02 1.08309537e-01 -5.16470909e-01 -8.02223325e-01 -1.85093001e-01 -9.54365969e-01 2.04029784e-01 1.02609503e+00 -2.83663392e-01 4.61655974e-01 9.77446377e-01 -7.94994056e-01 4.17673796e-01 -1.92481971e+00 3.58313143e-01 5.92505857e-02 1.09038740e-01 4.66132052e-02 -4.38874453e-01 8.53463821e-03 -1.83112681e-01 1.71502560e-01 -2.77677894e-01 -9.00276780e-01 -1.00636315e-02 2.51594335e-01 -1.56069815e-01 6.77873909e-01 5.46142876e-01 8.83457243e-01 -5.71605206e-01 -2.87112951e-01 3.37385237e-01 1.20994985e+00 -6.51372492e-01 4.26092178e-01 -3.44124466e-01 9.22900021e-01 -1.25266641e-01 7.61954188e-01 1.03712237e+00 8.78165737e-02 1.11970514e-01 -3.95974457e-01 -2.19735280e-02 -2.01794967e-01 -9.25925314e-01 1.64851606e+00 -6.36301339e-01 7.56098092e-01 7.95531332e-01 -3.64570141e-01 9.80586290e-01 2.34822839e-01 5.20804048e-01 -4.72196549e-01 2.67550319e-01 2.93693785e-02 -4.39336270e-01 -3.00946236e-01 -1.44782988e-02 -5.98250747e-01 1.64811164e-01 3.05166274e-01 7.18665794e-02 -7.38149285e-01 -5.04912257e-01 -1.40095368e-01 6.23664260e-01 5.34264088e-01 -6.35153893e-03 -3.16813886e-02 3.31673115e-01 -5.07301450e-01 5.47814965e-01 -8.77191573e-02 3.70281190e-02 1.04290843e+00 3.48959386e-01 -4.88388628e-01 -1.33334935e+00 -1.37852299e+00 -2.49994040e-01 6.30433261e-01 -1.33466914e-01 -1.81775197e-01 -7.63998389e-01 -3.56695145e-01 2.85295486e-01 5.52514017e-01 -1.02394271e+00 1.01821996e-01 -5.57311714e-01 -3.65534097e-01 5.91725707e-01 7.18340129e-02 5.01087546e-01 -7.96375632e-01 -3.14623833e-01 -2.42084503e-01 1.04459934e-01 -1.23717415e+00 -6.23299956e-01 -5.69703698e-01 -4.10539448e-01 -1.01358485e+00 -6.66516960e-01 -4.14520323e-01 9.55169499e-01 -3.28836553e-02 1.23319113e+00 1.03516437e-01 -5.70393562e-01 4.90656078e-01 1.61268473e-01 -2.58714944e-01 -5.24157047e-01 -5.93577266e-01 1.06364891e-01 4.20651257e-01 -2.13178948e-01 -9.50223327e-01 -6.73049808e-01 4.21401203e-01 -6.49982691e-01 3.47968966e-01 1.05187848e-01 6.45493627e-01 6.87191308e-01 -5.42275012e-01 4.66408161e-03 -8.22382390e-01 1.50677845e-01 -1.11024193e-01 -6.06652498e-01 3.78019363e-02 -3.82856041e-01 -7.14950711e-02 3.99801105e-01 -5.71162283e-01 -1.29259813e+00 3.13793659e-01 -1.18682981e-01 -9.20574307e-01 -2.77399480e-01 -3.66747588e-01 -4.41795051e-01 -4.73435074e-01 6.99028611e-01 2.17819456e-02 4.11663115e-01 -5.27328193e-01 3.01235259e-01 2.29497746e-01 7.23292768e-01 -8.12048435e-01 1.25390291e+00 8.19550157e-01 4.00352329e-01 -1.00376201e+00 -7.60750175e-01 2.45857090e-01 -6.74008310e-01 -4.03256744e-01 8.63633156e-01 -1.20322478e+00 -7.14300394e-01 6.90637708e-01 -1.16565979e+00 -7.19505966e-01 -2.72667021e-01 1.13569342e-01 -6.90985084e-01 8.13351274e-02 -4.12726402e-01 -7.90109158e-01 -2.50457913e-01 -1.01302755e+00 1.53183722e+00 1.40712082e-01 -2.23960295e-01 -8.24811399e-01 9.21180248e-02 4.18525249e-01 3.72156113e-01 1.13513923e+00 6.66394413e-01 3.81951451e-01 -5.27607501e-01 1.09669991e-01 -1.60921767e-01 3.60441417e-01 2.92520881e-01 3.77865911e-01 -1.26473558e+00 -3.58712792e-01 -1.72378570e-01 -4.28866863e-01 4.17054385e-01 3.24166894e-01 1.01118207e+00 -6.90303564e-01 6.60992786e-02 1.33778071e+00 1.33920264e+00 -4.48798746e-01 6.00399077e-01 -1.71502650e-01 9.19631004e-01 7.83414543e-01 1.85772061e-01 3.31288338e-01 2.90367424e-01 1.17493892e+00 4.69302833e-01 -3.09947044e-01 -6.62734926e-01 -4.61992532e-01 5.09174705e-01 1.88409373e-01 -3.42275560e-01 1.18730679e-01 -5.55341303e-01 1.64369926e-01 -1.20284963e+00 -9.85666513e-01 -7.06925476e-03 2.05280042e+00 1.03902614e+00 -4.55301404e-01 1.75155222e-01 -3.34408581e-01 3.62918943e-01 2.01925561e-01 -7.27086604e-01 -4.80665475e-01 -3.99726570e-01 4.49848026e-01 2.65695602e-01 9.69738901e-01 -6.51588440e-01 1.01126528e+00 5.95583153e+00 4.08716142e-01 -1.38325655e+00 -1.30098602e-02 8.56109262e-01 -5.23151398e-01 -7.35987306e-01 -1.75360501e-01 -5.44464529e-01 1.47730067e-01 6.47758901e-01 8.37445706e-02 8.15387607e-01 6.19563401e-01 3.99060369e-01 2.52889365e-01 -1.14276493e+00 1.15244901e+00 3.52761775e-01 -1.31307352e+00 2.48384118e-01 1.63195476e-01 8.89894128e-01 -4.05606151e-01 3.40127766e-01 -4.84426245e-02 1.81373954e-01 -1.44963670e+00 8.70513141e-01 6.89328432e-01 1.37666154e+00 -6.54961407e-01 -2.59844493e-02 7.33486637e-02 -8.48550260e-01 3.61842662e-01 -5.79506159e-02 -1.53745031e-02 2.26215020e-01 2.15766564e-01 -6.12514853e-01 2.27144867e-01 6.35742068e-01 5.10780692e-01 -3.92466873e-01 2.69461513e-01 -4.08404142e-01 2.60386199e-01 -3.02592903e-01 5.71567774e-01 -2.43341357e-01 -4.84952956e-01 6.45501375e-01 1.09825194e+00 4.07710969e-01 3.43367457e-01 -7.22854584e-02 1.26571119e+00 -2.16369778e-01 -4.37173955e-02 -6.49330199e-01 2.25086063e-01 3.29686224e-01 1.24669921e+00 -2.52658594e-02 4.84495237e-02 -2.00221464e-01 1.15896094e+00 3.16207588e-01 5.45294464e-01 -6.46007657e-01 3.35884333e-01 1.43359160e+00 6.70074522e-01 1.77043125e-01 -3.45667630e-01 -1.56473055e-01 -1.18253088e+00 -1.18234165e-01 -9.44439352e-01 -2.04876840e-01 -1.18496704e+00 -1.34188724e+00 7.14036763e-01 1.43660426e-01 -8.70532632e-01 -4.74581748e-01 -7.06961215e-01 -5.88441610e-01 1.01701808e+00 -1.57150805e+00 -1.77276075e+00 -6.33992612e-01 7.81266749e-01 2.83778906e-01 2.32249275e-02 1.04628932e+00 4.05311584e-04 -5.56013703e-01 8.29516232e-01 -3.82306099e-01 1.74449369e-01 8.88345540e-01 -9.68147874e-01 3.88285518e-01 5.64662516e-01 5.63409664e-02 5.54958701e-01 8.98376822e-01 -4.21727777e-01 -1.64639568e+00 -1.09140801e+00 5.31768203e-01 -7.06684232e-01 1.99409902e-01 -7.09390402e-01 -7.61468768e-01 8.30299020e-01 -7.95617700e-02 6.03575230e-01 5.12263060e-01 -1.49915725e-01 -8.79767299e-01 -1.73854440e-01 -1.55662358e+00 7.50833452e-01 1.22060716e+00 -6.80223703e-01 -2.10218728e-02 1.21090695e-01 3.11934322e-01 -5.85863113e-01 -9.41661417e-01 3.39951247e-01 9.64762330e-01 -1.21139801e+00 1.23502636e+00 -4.61102843e-01 3.59790474e-01 -2.61265725e-01 -1.55180067e-01 -1.24836123e+00 -1.41877264e-01 -1.03019905e+00 -1.58912897e-01 1.25443327e+00 2.37542987e-02 -5.44149637e-01 8.85462046e-01 8.89714479e-01 2.47051328e-01 -6.17169142e-01 -8.79120529e-01 -6.68944955e-01 3.52210939e-01 -5.52574731e-02 8.37996542e-01 1.02854693e+00 -6.72648132e-01 3.20276320e-01 -5.83900630e-01 2.13064849e-01 8.63457501e-01 2.67405599e-01 1.23831630e+00 -1.24227476e+00 -2.47381791e-01 -3.05484831e-01 -3.18635613e-01 -6.79806113e-01 5.89954197e-01 -6.85611427e-01 -2.02469215e-01 -1.00305521e+00 9.73718092e-02 -3.36629897e-01 5.89771688e-01 6.54908180e-01 1.76060900e-01 6.53510928e-01 2.67764807e-01 1.65857114e-02 3.52878094e-01 6.81615710e-01 1.67077827e+00 7.91619122e-02 -6.05183728e-02 -2.32942268e-01 -5.30111849e-01 7.55775809e-01 6.51801109e-01 -1.05584539e-01 -4.05635357e-01 -4.99792933e-01 -1.67954981e-01 1.07171617e-01 8.49865079e-01 -7.01531589e-01 -2.74135679e-01 -3.97049576e-01 8.58549595e-01 -3.00936121e-02 1.06941187e+00 -7.83173859e-01 8.46473813e-01 -1.08733000e-02 -1.08211808e-01 -1.48241416e-01 3.51707011e-01 2.78077900e-01 1.89298853e-01 4.19575065e-01 1.14921355e+00 -3.74815986e-02 -1.23795956e-01 7.67614424e-01 1.27253309e-01 1.33145489e-02 7.66486347e-01 -3.65961283e-01 -1.01384588e-01 -6.88289523e-01 -6.02095306e-01 -2.35129178e-01 1.26867390e+00 4.22962278e-01 5.04364192e-01 -1.59759653e+00 -1.08519638e+00 6.84453964e-01 -5.02884462e-02 1.30291104e-01 3.02289128e-01 2.16062427e-01 -6.95145845e-01 -2.19142571e-01 -4.85992104e-01 -4.77650642e-01 -1.68961811e+00 2.37091452e-01 7.44562924e-01 1.59793973e-01 -5.50698221e-01 9.16290879e-01 5.33387184e-01 -7.53069937e-01 -8.07274207e-02 1.25243649e-01 3.54661077e-01 -2.38338709e-01 4.67691451e-01 -5.09261116e-02 -2.71399528e-01 -1.32346010e+00 -2.30477825e-01 1.20889962e+00 3.88156712e-01 -3.54203284e-01 1.47832286e+00 -6.80434629e-02 -6.81003556e-02 1.65370390e-01 1.43633580e+00 2.75004655e-01 -2.02859354e+00 4.56289463e-02 -9.04091001e-01 -9.10584629e-01 1.43325612e-01 -9.25954282e-01 -1.43026137e+00 8.91473234e-01 4.87214297e-01 -5.41274667e-01 1.23810458e+00 -1.63859278e-01 5.33831179e-01 -1.76215306e-01 2.50996739e-01 -6.92848742e-01 2.02527747e-01 1.67340338e-01 1.47470522e+00 -1.02075219e+00 1.86955109e-01 -7.06324875e-01 -4.64626223e-01 1.02065814e+00 6.47144198e-01 -4.55219857e-02 6.16879642e-01 5.89439571e-01 2.89709896e-01 -2.32032195e-01 -5.89522183e-01 2.28673145e-01 4.92079854e-01 8.87308419e-01 2.87143320e-01 9.18461233e-02 3.97531152e-01 1.45923957e-01 -5.69642305e-01 -1.46956772e-01 2.82131881e-01 4.18865442e-01 1.33330598e-01 -1.12813699e+00 -5.41372061e-01 1.24133527e-02 -1.60563618e-01 3.26988190e-01 -5.94786465e-01 8.53236139e-01 2.69557208e-01 5.61704814e-01 2.24485591e-01 -1.86957210e-01 3.67607236e-01 -1.87201216e-03 1.08598053e+00 -5.09193420e-01 -3.79339159e-01 1.17210530e-01 1.83903977e-01 -7.50048161e-01 -2.84459829e-01 -9.01637793e-01 -9.52546239e-01 -6.71248734e-01 2.12461188e-01 -4.01845723e-01 7.62217343e-01 3.91484618e-01 5.10833263e-01 3.42102647e-02 8.26926291e-01 -1.38491356e+00 -1.92526475e-01 -6.08344078e-01 -5.07804751e-01 6.47325814e-01 7.54595339e-01 -7.87353158e-01 -5.92159390e-01 2.72805572e-01]
[12.755997657775879, -0.3493427336215973]
5438dc8a-f314-4409-a708-07245d97653a
safe-exploration-for-identifying-linear
1711.11165
null
http://arxiv.org/abs/1711.11165v1
http://arxiv.org/pdf/1711.11165v1.pdf
Safe Exploration for Identifying Linear Systems via Robust Optimization
Safely exploring an unknown dynamical system is critical to the deployment of reinforcement learning (RL) in physical systems where failures may have catastrophic consequences. In scenarios where one knows little about the dynamics, diverse transition data covering relevant regions of state-action space is needed to apply either model-based or model-free RL. Motivated by the cooling of Google's data centers, we study how one can safely identify the parameters of a system model with a desired accuracy and confidence level. In particular, we focus on learning an unknown linear system with Gaussian noise assuming only that, initially, a nominal safe action is known. Define safety as satisfying specific linear constraints on the state space (e.g., requirements on process variable) that must hold over the span of an entire trajectory, and given a Probably Approximately Correct (PAC) style bound on the estimation error of model parameters, we show how to compute safe regions of action space by gradually growing a ball around the nominal safe action. One can apply any exploration strategy where actions are chosen from such safe regions. Experiments on a stylized model of data center cooling dynamics show how computing proper safe regions can increase the sample efficiency of safe exploration.
['Tyler Lu', 'Craig Boutilier', 'Binz Roy', 'Dale Schuurmans', 'Martin Zinkevich']
2017-11-30
null
null
null
null
['safe-exploration']
['robots']
[ 3.46869081e-02 2.77190000e-01 -2.48558074e-01 3.02990645e-01 -8.81889820e-01 -7.02254713e-01 3.61134201e-01 3.50691378e-01 -3.92016858e-01 9.53516066e-01 -4.23328727e-01 -6.64770722e-01 -5.50145268e-01 -5.44662476e-01 -9.42608356e-01 -9.03091311e-01 -3.62757921e-01 5.94836771e-01 2.01221153e-01 -2.17310172e-02 3.52307320e-01 7.08683074e-01 -1.09048498e+00 -5.04601359e-01 7.89539337e-01 7.89385200e-01 7.74271265e-02 7.31674910e-01 4.13618833e-01 5.64803779e-01 -7.30183423e-01 5.54971933e-01 4.21686232e-01 -3.75543445e-01 -7.18995690e-01 8.94628465e-02 -3.57358992e-01 -1.80167928e-01 -1.53419271e-01 1.20467508e+00 4.72836271e-02 5.46242476e-01 5.04754245e-01 -1.43059993e+00 1.41204432e-01 4.96399015e-01 -2.92376935e-01 1.89832717e-01 7.28019625e-02 8.51598144e-01 5.26835620e-01 -6.20805994e-02 3.46197873e-01 9.38726604e-01 2.17932105e-01 5.33391654e-01 -1.40243876e+00 -6.10102296e-01 5.93664110e-01 -1.72782734e-01 -1.45623112e+00 -1.23304449e-01 3.79363120e-01 -4.28346664e-01 8.00053775e-01 1.56735063e-01 6.18797481e-01 8.06609750e-01 9.23672915e-01 2.18147248e-01 9.69591439e-01 -3.42911154e-01 9.94134724e-01 8.92501771e-02 -3.52724046e-02 5.70012748e-01 2.95225412e-01 4.70085829e-01 -2.92133003e-01 -4.57579374e-01 8.12510788e-01 -1.50301635e-01 -2.07118034e-01 -7.03165412e-01 -9.31799412e-01 6.48120224e-01 3.50377038e-02 -1.53077111e-01 -2.61878550e-01 5.00252306e-01 2.54696965e-01 5.08941174e-01 1.16710454e-01 9.93991494e-01 -4.66494083e-01 -4.55701649e-01 -5.39724588e-01 3.24251264e-01 8.36877882e-01 7.35313118e-01 4.89378035e-01 3.51706088e-01 3.74216884e-02 1.38773218e-01 -9.46668908e-02 6.35294855e-01 -8.39504600e-02 -1.11053360e+00 3.04731011e-01 1.88325420e-01 7.85248816e-01 -2.12018013e-01 -2.78058916e-01 -3.38372827e-01 -3.15057099e-01 6.32979631e-01 5.79106569e-01 -6.25751376e-01 -9.82334495e-01 1.77302599e+00 4.62080956e-01 4.19519871e-01 -3.06473859e-02 6.71021521e-01 -7.61038363e-01 8.28637898e-01 -9.36009884e-02 -6.64831758e-01 6.63957298e-01 -2.28541180e-01 -3.61041248e-01 -2.62732536e-01 6.40545666e-01 -2.09026694e-01 1.20987189e+00 8.44603360e-01 -1.11075056e+00 -2.22203925e-01 -1.24488854e+00 7.96775460e-01 -1.71100609e-02 -2.57967561e-01 6.88170940e-02 5.93654394e-01 -6.45197093e-01 7.89374471e-01 -1.42740476e+00 -5.66215478e-02 6.54037595e-02 5.06496131e-01 4.73100767e-02 1.28212005e-01 -8.31650674e-01 1.06132650e+00 4.43873644e-01 8.00778344e-02 -1.69548881e+00 -9.05905962e-01 -3.59620810e-01 2.04734445e-01 1.17185545e+00 -3.16481978e-01 1.39015245e+00 -3.58648717e-01 -1.52883947e+00 -2.20596269e-01 3.04155827e-01 -6.22967899e-01 4.20117348e-01 -3.59349519e-01 -2.25768819e-01 8.93939957e-02 -3.89695823e-01 2.52668206e-02 1.15715837e+00 -1.10888338e+00 -4.42709237e-01 -1.63608789e-01 1.90761104e-01 2.91240126e-01 -2.96117691e-03 -3.19641829e-01 -5.75674623e-02 5.54606169e-02 -3.04999501e-02 -1.53437197e+00 -6.13213539e-01 -6.57589734e-02 -3.63893211e-01 -1.28373981e-01 6.65038526e-01 -4.92476970e-01 1.00998545e+00 -1.99933457e+00 5.92079945e-02 6.72252119e-01 -1.44394547e-01 -1.51963249e-01 2.34946251e-01 6.26420021e-01 5.84114939e-02 1.78057373e-01 -2.14507371e-01 4.48336191e-02 -1.64577141e-02 2.43591443e-01 -7.41961837e-01 8.05883527e-01 8.22930336e-02 2.03108385e-01 -8.42936993e-01 5.04725575e-02 2.61694133e-01 9.73412767e-02 -3.39180022e-01 2.53357977e-01 -6.61474168e-01 5.04713297e-01 -8.00797462e-01 1.70093462e-01 2.18950331e-01 5.85020408e-02 1.40992820e-01 4.77650762e-01 3.14594842e-02 8.17044824e-02 -1.49333954e+00 9.62536752e-01 -8.36425543e-01 2.49908939e-01 2.96229869e-01 -9.76221144e-01 7.95911431e-01 4.91966307e-02 5.51936150e-01 -2.04609811e-01 5.24119847e-02 -9.68355238e-02 1.70707613e-01 -1.74239576e-01 2.70111203e-01 -3.77637058e-01 -2.21631110e-01 7.91917622e-01 -3.03997964e-01 -6.67090774e-01 -9.03450176e-02 7.61090443e-02 1.45033300e+00 7.28593320e-02 5.78305013e-02 -6.04751647e-01 1.62822872e-01 1.17187612e-01 6.23709321e-01 7.70048141e-01 -1.28268272e-01 4.74216044e-02 1.03075016e+00 -5.07978871e-02 -1.08986533e+00 -1.01883864e+00 -7.49329031e-02 4.52333212e-01 3.09158593e-01 -2.85738736e-01 -7.48580158e-01 -6.57012701e-01 4.17713262e-02 1.13875341e+00 -6.35902643e-01 -6.38161838e-01 -5.75415909e-01 -4.83222157e-01 -1.54662114e-02 3.75380635e-01 4.34750356e-02 -8.76922309e-01 -1.12950516e+00 1.83899298e-01 6.33829474e-01 -7.04334259e-01 -4.43687856e-01 6.38220787e-01 -8.28905165e-01 -1.17999065e+00 -5.10075828e-03 -7.93950036e-02 8.69979143e-01 -2.82819301e-01 6.91386700e-01 -2.17557579e-01 -2.12554872e-01 5.78885853e-01 1.25354454e-01 -2.80533105e-01 -7.22664595e-01 -6.55526891e-02 3.35391581e-01 -5.04079998e-01 -2.79975951e-01 -1.25738934e-01 -2.89261967e-01 1.54194415e-01 -6.45099521e-01 -2.93900996e-01 1.40112862e-01 7.12288022e-01 5.21704674e-01 7.96919048e-01 4.55311835e-01 -5.94218075e-01 8.92443538e-01 -5.50827742e-01 -1.16358912e+00 4.75060970e-01 -9.22447860e-01 5.49407780e-01 1.00483739e+00 -7.40994990e-01 -7.47627079e-01 -1.67109668e-02 4.27873254e-01 -5.45615554e-01 -7.16047958e-02 3.27639997e-01 -1.64720267e-01 1.32217020e-01 5.94881713e-01 1.17637880e-01 2.04740107e-01 -1.01659216e-01 1.60158098e-01 -2.54316442e-02 3.84528756e-01 -1.22950971e+00 7.17240870e-01 6.22647405e-02 3.16679180e-01 -6.15207195e-01 -4.03822273e-01 -5.65142483e-02 -3.68925273e-01 -3.98996741e-01 2.56097674e-01 -5.14905512e-01 -1.18921888e+00 1.00285426e-01 -3.23094130e-01 -9.05708432e-01 -5.95263004e-01 2.82574564e-01 -9.71652925e-01 -2.52886657e-02 -3.65573227e-01 -1.38327420e+00 -1.33153913e-03 -1.06923711e+00 6.92493498e-01 2.41375342e-01 -3.24176252e-01 -7.81447291e-01 1.40629351e-01 -3.97185475e-01 3.26219022e-01 4.62183237e-01 1.10670924e+00 -4.95260835e-01 -6.29878163e-01 -1.74772143e-01 6.48944914e-01 2.51116365e-01 4.61387970e-02 2.31185973e-01 -4.15675938e-01 -7.93144941e-01 2.61342227e-01 -3.16367179e-01 4.09266144e-01 4.50570285e-01 1.32514977e+00 -4.71602231e-01 -5.16702771e-01 5.39015196e-02 1.26393211e+00 7.00375974e-01 1.16086565e-01 2.41786510e-01 2.10056126e-01 3.26244622e-01 1.04191804e+00 8.53540599e-01 -1.44568622e-01 3.83620828e-01 4.75881338e-01 3.50935221e-01 7.55558491e-01 -3.11607063e-01 5.01391709e-01 1.75498445e-02 5.55140436e-01 -1.12021662e-01 -1.07788181e+00 5.09370565e-01 -1.70843577e+00 -5.87868989e-01 6.92260385e-01 3.02479458e+00 8.66010725e-01 8.93363118e-01 2.53783077e-01 5.01106046e-02 4.14199203e-01 -2.68190205e-01 -1.42002356e+00 -6.30474329e-01 5.34323990e-01 1.90544739e-01 9.42484140e-01 8.57342720e-01 -6.59500718e-01 5.45759678e-01 5.91700888e+00 7.25726128e-01 -1.09341335e+00 -3.52426112e-01 7.79940724e-01 -3.72477919e-01 -2.47516390e-02 2.60695964e-01 -8.61355960e-01 4.75115657e-01 1.65306485e+00 -6.80013776e-01 7.39520133e-01 9.61507022e-01 5.15287101e-01 -6.14622533e-01 -1.27773333e+00 4.53593105e-01 -6.35394752e-01 -9.73763108e-01 -6.88930452e-01 2.10797653e-01 6.59848392e-01 -3.19110543e-01 1.61919639e-01 6.52076066e-01 9.35233474e-01 -9.99846578e-01 6.43343151e-01 7.04949737e-01 4.77937400e-01 -1.32682574e+00 3.31642002e-01 8.29649568e-01 -7.82786131e-01 -5.16452670e-01 -1.09239310e-01 -2.21615599e-04 1.15264043e-01 5.23368537e-01 -1.07488251e+00 4.09055129e-02 4.09298927e-01 1.65818185e-01 -1.95815101e-01 9.90774155e-01 -1.44139394e-01 9.05997932e-01 -8.64372075e-01 -1.40735671e-01 2.31434450e-01 -3.45694870e-01 3.93267155e-01 4.70850438e-01 3.07656825e-01 7.14674667e-02 5.32427669e-01 9.14510846e-01 4.87429559e-01 -4.99127030e-01 -6.66811407e-01 -8.01705644e-02 8.75963271e-01 9.12852883e-01 -9.12900686e-01 -1.20040566e-01 2.44020075e-01 4.17601109e-01 1.07138865e-01 4.92979735e-01 -9.54406798e-01 -1.26625791e-01 7.23809659e-01 8.87047872e-02 9.39183906e-02 -5.87520301e-01 -4.59852248e-01 -4.35203701e-01 -3.88111211e-02 -8.29922557e-01 3.36768746e-01 -4.42570329e-01 -7.78510630e-01 2.08709136e-01 4.06611443e-01 -9.23870623e-01 -6.27296150e-01 -2.20562860e-01 -7.21197069e-01 9.10997689e-01 -7.07452714e-01 -2.97618687e-01 3.67019206e-01 2.92207837e-01 5.64873397e-01 1.74712002e-01 5.12319744e-01 -8.10209274e-01 -9.07931685e-01 1.64530247e-01 3.87405276e-01 -6.32987618e-01 4.37534958e-01 -1.40600240e+00 3.94969702e-01 9.20681119e-01 -3.09140354e-01 7.00793147e-01 1.28747773e+00 -9.22877073e-01 -1.71456850e+00 -9.05157030e-01 -1.74241140e-01 -5.42740822e-01 8.57176840e-01 -3.61873507e-01 -1.14824784e+00 6.34177029e-01 -8.13004524e-02 4.94110882e-02 -1.22307390e-01 -2.57525663e-03 1.44951865e-01 -6.64730594e-02 -1.27353609e+00 7.17499018e-01 4.37229246e-01 -4.94849473e-01 -1.09698378e-01 3.18837523e-01 7.07805753e-01 -5.58422923e-01 -9.22501087e-01 2.80150175e-01 1.72082320e-01 -3.18594426e-01 5.63405931e-01 -8.47079933e-01 -2.85152853e-01 -4.03840333e-01 1.25400379e-01 -1.61897182e+00 -3.46069100e-05 -1.10817957e+00 -6.31388307e-01 7.88533211e-01 3.45078290e-01 -3.81929040e-01 7.89246619e-01 1.28862059e+00 -6.74965382e-02 -1.07992601e+00 -1.12938750e+00 -1.16182089e+00 5.09830475e-01 -3.96007776e-01 4.72633511e-01 4.19888884e-01 4.15903419e-01 -9.60041583e-02 -1.78638205e-01 6.49673104e-01 5.94228625e-01 -2.60318555e-02 5.63396037e-01 -6.81018054e-01 -5.32472074e-01 -2.53382623e-01 3.17519486e-01 -7.15877891e-01 3.56924772e-01 4.05070707e-02 4.69667912e-01 -9.01546240e-01 4.08133976e-02 -7.30853558e-01 -4.16931689e-01 3.67770225e-01 -1.67651862e-01 -8.25117052e-01 1.88975886e-01 -6.62776232e-02 -5.83794475e-01 6.80962801e-01 8.54877591e-01 1.73780292e-01 -5.10727882e-01 3.94247681e-01 -3.13065678e-01 2.37020522e-01 1.07208800e+00 -4.16626722e-01 -9.46616948e-01 3.40123057e-01 1.72704369e-01 6.75213456e-01 9.14527327e-02 -1.00977206e+00 2.15870902e-01 -9.86514449e-01 -5.51132895e-02 -5.53832114e-01 2.74884671e-01 -8.23688507e-01 3.45438331e-01 9.62215960e-01 -7.09419191e-01 1.16186664e-01 4.91916388e-01 8.18100989e-01 3.70884597e-01 -1.79862410e-01 9.16498601e-01 1.62688211e-01 -4.70498055e-01 1.04793631e-01 -6.79223239e-01 1.08145826e-01 1.31599641e+00 5.84465750e-02 -1.97931394e-01 -4.92107719e-01 -7.98068583e-01 6.44490480e-01 5.28273940e-01 3.40300649e-01 3.27223569e-01 -7.05017686e-01 -1.16254650e-01 2.20298156e-01 -1.31408200e-01 -1.36964247e-01 1.91120431e-01 6.15830719e-01 -1.48396358e-01 3.66139978e-01 4.15630154e-02 -3.82447183e-01 -8.92762840e-01 9.44213271e-01 6.82219028e-01 -1.34780854e-01 -6.81720316e-01 4.68623877e-01 -1.55753121e-01 3.06102302e-04 2.45117620e-01 -7.51920521e-01 5.13797402e-01 -5.77409506e-01 3.99680257e-01 4.42036182e-01 1.45076215e-01 3.46799195e-01 -2.53550112e-01 6.35609403e-02 -1.61810473e-01 -3.47163141e-01 1.02160466e+00 -1.16900839e-02 4.17321503e-01 7.15170503e-01 5.40896535e-01 -2.72350162e-01 -2.10263848e+00 3.53289023e-02 1.80434287e-01 -2.10712448e-01 2.50371754e-01 -8.50489259e-01 -7.42205143e-01 6.49228871e-01 6.37037575e-01 4.22213107e-01 9.55227852e-01 -2.14476138e-01 3.34488034e-01 4.58184808e-01 7.05396116e-01 -1.42555535e+00 -1.05499201e-01 3.30745578e-01 6.08369827e-01 -7.58903921e-01 1.23524532e-01 7.08776861e-02 -7.32476115e-01 1.08317959e+00 9.87129748e-01 -3.05021346e-01 6.49596035e-01 6.29440129e-01 -5.15977442e-01 8.18245485e-02 -1.36437237e+00 2.76646435e-01 -1.57786936e-01 3.59375983e-01 -2.71854877e-01 2.14034960e-01 1.33045942e-01 1.66175500e-01 -1.20629929e-02 -1.44111440e-01 8.67559314e-01 1.15213192e+00 -7.70053029e-01 -8.09061170e-01 -6.13101721e-01 4.01635021e-01 -1.12938844e-01 3.27209204e-01 3.40761282e-02 1.08720672e+00 -4.38534647e-01 8.41956973e-01 -3.01234331e-03 -3.41872089e-02 1.49140745e-01 1.23638779e-01 4.91569221e-01 -5.91671586e-01 -2.04839200e-01 -5.43784648e-02 8.89390707e-02 -6.32296026e-01 6.21977687e-01 -9.10970509e-01 -1.51815164e+00 -3.72462869e-01 -3.24179709e-01 2.84727663e-01 5.17356992e-01 1.08336985e+00 2.25301012e-01 4.39123541e-01 8.53283048e-01 -2.92788982e-01 -1.37250018e+00 -7.13850856e-01 -7.31472671e-01 -3.03014666e-01 6.18381143e-01 -7.45210052e-01 -5.51209927e-01 -2.36295685e-01]
[4.660819053649902, 2.264647960662842]
548787cb-136a-4346-90c0-a24acc0e8890
weakly-supervised-contrastive-learning-for
2109.12242
null
https://arxiv.org/abs/2109.12242v1
https://arxiv.org/pdf/2109.12242v1.pdf
Weakly Supervised Contrastive Learning for Chest X-Ray Report Generation
Radiology report generation aims at generating descriptive text from radiology images automatically, which may present an opportunity to improve radiology reporting and interpretation. A typical setting consists of training encoder-decoder models on image-report pairs with a cross entropy loss, which struggles to generate informative sentences for clinical diagnoses since normal findings dominate the datasets. To tackle this challenge and encourage more clinically-accurate text outputs, we propose a novel weakly supervised contrastive loss for medical report generation. Experimental results demonstrate that our method benefits from contrasting target reports with incorrect but semantically-close ones. It outperforms previous work on both clinical correctness and text generation metrics for two public benchmarks.
['Chun-Nan Hsu', 'Julian McAuley', 'Amilcare Gentili', 'Eric Chang', 'Jiang Du', 'Xing Lu', 'Zexue He', 'An Yan']
2021-09-25
null
https://aclanthology.org/2021.findings-emnlp.336
https://aclanthology.org/2021.findings-emnlp.336.pdf
findings-emnlp-2021-11
['medical-report-generation']
['medical']
[ 7.96318889e-01 1.00270164e+00 -1.71873033e-01 -6.44923031e-01 -1.85383344e+00 -1.50599301e-01 5.81304371e-01 5.30983210e-01 -3.01350176e-01 1.13135791e+00 9.78268862e-01 -3.23097587e-01 2.04375699e-01 -6.20353103e-01 -7.72399664e-01 -3.81377727e-01 -5.43891005e-02 6.46534503e-01 -1.87027499e-01 1.26455739e-01 4.91750613e-03 -7.45379478e-02 -9.70089138e-01 9.01848674e-01 7.21612930e-01 6.79732323e-01 2.90513009e-01 8.77090037e-01 1.54184923e-01 1.41894150e+00 -6.86056018e-01 -9.99286532e-01 3.98482149e-03 -9.41997945e-01 -1.13777459e+00 8.73844326e-02 5.17833352e-01 -5.32866597e-01 -3.44338238e-01 9.37024415e-01 8.15954030e-01 -2.38483429e-01 9.43484426e-01 -4.47109759e-01 -8.70594740e-01 1.12755108e+00 -5.83421648e-01 3.75790834e-01 5.79013288e-01 1.66715875e-01 1.04504609e+00 -7.35945821e-01 1.08454454e+00 9.35686231e-01 6.32683456e-01 9.04735208e-01 -1.25840878e+00 -3.35831404e-01 -1.67903975e-01 -2.78798461e-01 -1.26554120e+00 -3.36154133e-01 4.51518804e-01 -2.40097106e-01 7.21280336e-01 5.74452579e-01 3.10642123e-01 1.49675155e+00 4.78076339e-01 9.65661287e-01 7.79494405e-01 -4.58373539e-02 -4.19708677e-02 1.79190367e-01 -3.73223305e-01 8.02984834e-01 4.25437957e-01 -1.26170218e-01 -3.49866062e-01 -2.33763233e-01 5.69012761e-01 -3.14964205e-01 -5.27878940e-01 2.70791471e-01 -1.60174680e+00 9.44767118e-01 5.27203679e-01 1.41460776e-01 -5.38032413e-01 1.48941070e-01 5.80211878e-01 -5.67350723e-02 6.84659004e-01 7.50974774e-01 9.90598649e-02 -6.10385761e-02 -9.90167439e-01 4.36155349e-01 7.06211686e-01 1.01263893e+00 -2.21018493e-02 -1.84467241e-01 -1.04550767e+00 7.78503120e-01 -2.01930273e-02 6.22148275e-01 7.61917651e-01 -6.17476046e-01 7.91359007e-01 2.12221205e-01 -1.25671923e-01 -7.55108237e-01 -3.95616740e-01 -5.54226398e-01 -1.00247502e+00 -3.94731641e-01 -1.35208756e-01 -7.38784522e-02 -1.03700876e+00 1.45684886e+00 -7.59612769e-02 -3.73150036e-02 2.42117628e-01 9.50026870e-01 1.33913147e+00 4.32326078e-01 2.20638916e-01 -2.89915383e-01 1.42179716e+00 -9.21286583e-01 -8.00626099e-01 -1.63144231e-01 8.61558676e-01 -8.19466949e-01 1.04452205e+00 -8.95856023e-02 -1.64247203e+00 -1.24118067e-01 -7.23071814e-01 -1.18692786e-01 2.38705009e-01 4.75923359e-01 3.74346316e-01 1.78390145e-01 -1.02912796e+00 1.38452530e-01 -7.64926910e-01 7.82094449e-02 7.97947407e-01 9.55759361e-03 -3.77269059e-01 -1.58688292e-01 -9.85870481e-01 9.52242672e-01 5.83294868e-01 -1.83904022e-01 -8.40868354e-01 -1.09310293e+00 -1.10475469e+00 -8.66100341e-02 2.82454818e-01 -1.23482704e+00 1.78549206e+00 -5.18088579e-01 -1.09334111e+00 1.33345509e+00 3.38991061e-02 -9.73626316e-01 1.06423330e+00 7.74629563e-02 -3.34426582e-01 4.76961255e-01 4.36616659e-01 1.15441716e+00 6.01202071e-01 -1.24876451e+00 -3.93809676e-01 1.94569360e-02 -1.12210423e-01 2.10907221e-01 8.23634788e-02 -3.53575230e-01 -5.80151156e-02 -1.19265628e+00 -1.51427656e-01 -7.71190047e-01 -6.74908042e-01 -1.05410665e-01 -9.17637706e-01 -6.13471642e-02 5.12956977e-02 -8.61638963e-01 1.08921409e+00 -1.86671472e+00 -3.25515091e-01 -1.55442357e-01 5.88524103e-01 -1.68090433e-01 4.94330265e-02 1.31262094e-01 -2.71263361e-01 3.05568993e-01 -7.01270878e-01 -4.39679831e-01 -2.56083965e-01 2.70863418e-02 -4.80056137e-01 1.83102041e-01 7.92656362e-01 1.08720303e+00 -1.13133788e+00 -9.45356071e-01 -2.06057727e-01 2.80914783e-01 -7.88728833e-01 4.22947615e-01 -2.33259425e-01 6.55505657e-01 -3.61125886e-01 4.69380260e-01 3.77626210e-01 -7.88043976e-01 5.15364967e-02 -1.54282898e-01 4.78085190e-01 8.10359359e-01 -2.96423793e-01 1.85519040e+00 -6.66434169e-01 5.03171682e-01 -5.57734489e-01 -6.59507990e-01 6.99293733e-01 5.38840294e-01 5.67020893e-01 -7.61398673e-01 -5.93199283e-02 2.36813903e-01 -1.03255212e-01 -7.72188365e-01 5.55599928e-01 -5.09060621e-01 -1.95101708e-01 6.59726620e-01 -1.64974302e-01 -4.81328845e-01 1.11237459e-01 4.78543699e-01 1.54131687e+00 -2.72642970e-01 4.30654585e-01 2.60459542e-01 3.84534895e-01 1.24729082e-01 1.74378008e-01 1.02340674e+00 7.83292130e-02 1.25476050e+00 4.43464309e-01 -2.77993172e-01 -1.29896498e+00 -1.48157144e+00 -2.87883639e-01 1.93808377e-01 -2.03619301e-01 -2.96699494e-01 -5.36886811e-01 -1.29107440e+00 -2.73282349e-01 1.06775260e+00 -7.03447044e-01 -4.61441964e-01 -6.42991364e-01 -6.96497142e-01 6.44929647e-01 6.10099316e-01 1.60730422e-01 -1.03508854e+00 -4.78506863e-01 2.72937089e-01 -7.79355288e-01 -1.44592261e+00 -7.49868810e-01 -6.93900436e-02 -8.32648039e-01 -9.04083848e-01 -1.24511385e+00 -6.73943162e-01 1.14072382e+00 -2.17673883e-01 1.69355512e+00 -9.72872302e-02 -7.48684466e-01 5.09714901e-01 -4.40107524e-01 -5.82749724e-01 -1.17590523e+00 2.41086930e-01 -5.44813335e-01 -4.11046594e-01 9.96218249e-03 -4.24697995e-02 -7.94925570e-01 -4.36586410e-01 -1.19237614e+00 5.22838295e-01 1.06139731e+00 1.22445023e+00 7.71434605e-01 -5.61019778e-01 6.40587568e-01 -1.10087216e+00 9.73142385e-01 -6.96533322e-01 1.86196283e-01 2.48833984e-01 -6.23161435e-01 5.02089500e-01 5.52614927e-01 -4.03543785e-02 -1.09844828e+00 -1.13957174e-01 -4.51105863e-01 -1.12311736e-01 -1.48961069e-02 3.53791058e-01 7.25602329e-01 4.48764026e-01 8.28937590e-01 5.58158219e-01 1.24043506e-02 -1.08851176e-02 1.88871175e-01 5.64477861e-01 6.67333364e-01 -1.61109447e-01 4.80872869e-01 5.56320071e-01 -5.93376048e-02 -1.54025525e-01 -1.50435722e+00 -2.38538116e-01 1.30011672e-02 -7.26089850e-02 9.34558809e-01 -9.43649054e-01 -3.66186261e-01 -5.33554554e-01 -1.33919454e+00 2.55199194e-01 -7.85083175e-01 6.45239770e-01 -7.95846403e-01 1.63758248e-01 -7.33408034e-01 -4.83376622e-01 -7.78126121e-01 -1.37855458e+00 1.69322777e+00 -2.06361353e-01 -4.99666244e-01 -8.61453950e-01 1.58546388e-01 2.30841309e-01 7.99251273e-02 4.41661805e-01 7.60282040e-01 -6.72479987e-01 -3.73151541e-01 -2.23743141e-01 -4.18662667e-01 3.26739132e-01 2.06153154e-01 -4.25918758e-01 -8.49446774e-01 -7.06367195e-02 -1.14162654e-01 -5.41829586e-01 1.22328365e+00 4.31439668e-01 1.62761331e+00 -6.69370711e-01 -3.51231158e-01 3.11155975e-01 1.15309131e+00 -2.40671158e-01 4.97795045e-01 -9.93642211e-02 4.75628376e-01 5.42856038e-01 4.58083272e-01 5.50418735e-01 7.87407219e-01 5.05852818e-01 2.96406120e-01 -3.69886547e-01 -4.35500145e-01 -4.54059422e-01 1.06931947e-01 7.81158209e-01 2.32422650e-01 -2.74388522e-01 -9.49737608e-01 6.62766337e-01 -1.50945687e+00 -9.50098991e-01 5.93143553e-02 1.74074769e+00 1.56251264e+00 9.88625363e-02 -1.65619984e-01 -2.89028585e-01 4.19898331e-01 1.26018316e-01 -3.80533278e-01 -2.26395443e-01 -4.01154086e-02 3.44401866e-01 6.08212948e-01 2.60380894e-01 -1.12933767e+00 3.25424045e-01 7.07283449e+00 6.63027942e-01 -9.62730229e-01 3.26047182e-01 1.18868768e+00 -3.18515480e-01 -6.75163150e-01 -6.03019178e-01 -6.61731601e-01 6.11420870e-01 1.03928518e+00 -3.17075104e-01 -5.31184316e-01 7.54411757e-01 2.67417043e-01 1.11278538e-02 -1.38379776e+00 1.05433464e+00 4.97094959e-01 -1.95994496e+00 6.07240200e-01 -1.20523050e-02 9.14796770e-01 -1.57273501e-01 2.14506149e-01 1.08261250e-01 4.51482594e-01 -1.12104106e+00 5.96230030e-01 6.70464396e-01 1.07671404e+00 -4.90953475e-01 1.05552983e+00 -1.12147823e-01 -6.57235146e-01 2.23046154e-01 -2.39977047e-01 5.69744468e-01 3.89109761e-01 7.89401174e-01 -1.85885191e+00 6.76422298e-01 2.72631645e-01 6.17660105e-01 -5.46221375e-01 1.07427490e+00 -2.03156359e-02 5.42329967e-01 3.15280706e-02 2.57366188e-02 4.10581172e-01 3.25244606e-01 5.85879087e-01 1.66567659e+00 4.61131155e-01 8.04734677e-02 2.97493458e-01 1.32199550e+00 -6.02980614e-01 1.96824342e-01 -9.57921147e-01 -1.67737808e-02 5.76981828e-02 1.04351747e+00 -7.87932277e-01 -6.66297436e-01 -1.71991408e-01 1.05362415e+00 1.80389851e-01 -9.25043225e-02 -9.81679618e-01 4.37422246e-02 1.89395875e-01 2.38520950e-01 -1.93856191e-02 3.26193273e-01 -6.34883881e-01 -1.13658965e+00 3.51796389e-01 -7.13570356e-01 5.23808241e-01 -8.57140124e-01 -1.31717801e+00 8.58858585e-01 -2.04711348e-01 -1.68515277e+00 -7.13230312e-01 -1.47146016e-01 -2.63497591e-01 5.93456566e-01 -1.64309680e+00 -1.14790225e+00 -2.41898879e-01 6.56875968e-02 7.03224182e-01 -5.55633754e-02 9.07086015e-01 3.33719313e-01 -5.30347694e-03 9.55893457e-01 -2.97552526e-01 1.79731607e-01 8.61520529e-01 -1.61319101e+00 4.47761357e-01 5.43317199e-01 1.73937529e-01 2.05427885e-01 7.73466110e-01 -6.77182734e-01 -6.92385435e-01 -1.40619564e+00 9.29112613e-01 -5.09061337e-01 3.48816365e-01 -8.34880322e-02 -6.13375902e-01 5.36459625e-01 2.44943261e-01 -1.07227422e-01 9.51327920e-01 -6.91586375e-01 -1.28539026e-01 1.47327363e-01 -1.29350030e+00 8.55685055e-01 1.05460930e+00 -3.54252815e-01 -6.20144963e-01 9.10676837e-01 1.14472938e+00 -8.00845861e-01 -1.02620697e+00 4.76808816e-01 1.59916684e-01 -5.88976264e-01 9.57096040e-01 -7.50558496e-01 1.40408885e+00 1.55227304e-01 2.36662954e-01 -1.25137305e+00 1.05176613e-01 -4.00274217e-01 8.47685784e-02 7.35395312e-01 8.85078430e-01 -2.64152557e-01 6.98319972e-01 3.41974437e-01 -4.34196174e-01 -1.03685951e+00 -7.61922002e-01 -5.31866193e-01 7.97990039e-02 -4.17565703e-01 4.22975779e-01 7.65792966e-01 -7.25141913e-02 1.25489265e-01 -1.61580950e-01 -1.40064836e-01 5.19066095e-01 -2.70726085e-02 2.45323032e-01 -6.09176993e-01 -2.24846572e-01 -5.20696759e-01 -4.37273026e-01 -9.51934338e-01 1.04738936e-01 -1.44380069e+00 3.18239629e-01 -1.97194791e+00 6.79815233e-01 -2.13212252e-01 1.47141486e-01 4.13394272e-01 -5.75976193e-01 6.74900055e-01 -9.23888907e-02 1.09530263e-01 -6.77054286e-01 6.03835344e-01 1.66784537e+00 -5.66961646e-01 2.07348168e-01 1.38340637e-01 -9.51457202e-01 4.98818994e-01 3.08964133e-01 -8.08845222e-01 -3.35360944e-01 -3.94923747e-01 3.54436994e-01 3.56015295e-01 4.95025963e-01 -7.91697502e-01 -1.34773687e-01 1.62103951e-01 3.21717262e-01 -7.70670354e-01 -5.60508743e-02 -4.24807489e-01 -2.55689502e-01 6.71578765e-01 -1.07782197e+00 1.90011621e-01 -9.46403965e-02 5.96999407e-01 -5.76201439e-01 -2.60905564e-01 6.47964597e-01 -4.68491137e-01 9.77784991e-02 2.87703723e-01 -3.12791109e-01 5.27544916e-01 1.03208065e+00 5.32550476e-02 -2.74205685e-01 -5.71821451e-01 -9.27207887e-01 1.87686637e-01 -9.63333398e-02 4.43046480e-01 1.08461308e+00 -1.36635268e+00 -1.60635388e+00 -1.25126302e-01 5.07531524e-01 3.02841425e-01 2.58986950e-01 1.03225756e+00 -6.95441604e-01 5.57128191e-01 4.02944535e-01 -7.33165324e-01 -1.11812854e+00 1.10838085e-01 3.66466343e-01 -9.47977185e-01 -1.09651196e+00 9.25863504e-01 4.03826058e-01 -3.00011158e-01 6.67451173e-02 -7.12307096e-01 5.62493280e-02 -3.22812468e-01 7.30256617e-01 -2.52468526e-01 2.24437475e-01 -2.16904208e-01 6.72928058e-04 -1.35414585e-01 -6.72414839e-01 -1.49029374e-01 1.30603480e+00 -1.34898603e-01 1.02609023e-01 2.40248382e-01 1.21875858e+00 -9.09130424e-02 -7.53098249e-01 -2.42206842e-01 1.17529362e-01 -2.40415186e-01 -9.74549949e-02 -1.04604959e+00 -8.61637473e-01 5.31657159e-01 5.07689834e-01 1.33392334e-01 1.01712072e+00 4.61529016e-01 1.11278653e+00 4.08725291e-01 -4.15985584e-02 -7.72209525e-01 4.33175981e-01 1.35742009e-01 1.20502687e+00 -1.58025002e+00 -3.00207529e-02 -3.02592278e-01 -1.05393815e+00 9.84558821e-01 4.19495583e-01 2.41159461e-02 3.53703856e-01 4.50784117e-01 -2.47944649e-02 -3.23012054e-01 -8.70434225e-01 -8.18249136e-02 5.21904886e-01 5.08678854e-01 9.27132130e-01 2.02119797e-01 -5.71301222e-01 5.21246314e-01 -5.79223037e-01 -5.59276566e-02 8.62199366e-01 8.31891656e-01 3.41054685e-02 -7.74277031e-01 -1.87073469e-01 1.09132552e+00 -9.73495007e-01 -4.58928823e-01 -2.21524939e-01 4.87362802e-01 -2.30823711e-01 4.61150289e-01 3.34559865e-02 -6.39373139e-02 3.26741397e-01 -1.77151173e-01 5.16061664e-01 -1.07565165e+00 -6.14482045e-01 -1.93812340e-01 3.62208962e-01 -4.33323681e-01 -2.99401462e-01 -5.47324061e-01 -1.41712141e+00 1.07087269e-01 -2.76439171e-02 9.01519954e-02 5.71864784e-01 6.83411658e-01 4.66435671e-01 1.19858479e+00 6.95692301e-01 -2.07080498e-01 -8.63497734e-01 -1.02651453e+00 -2.10373208e-01 7.42447793e-01 4.70252842e-01 -2.52970457e-01 -4.48147804e-02 5.46036661e-01]
[15.03172492980957, -1.4003006219863892]
275a9994-0336-4a2b-a110-391587dee18a
graph-conditioned-sparse-attention-for
2112.00663
null
https://arxiv.org/abs/2112.00663v2
https://arxiv.org/pdf/2112.00663v2.pdf
Graph Conditioned Sparse-Attention for Improved Source Code Understanding
Transformer architectures have been successfully used in learning source code representations. The fusion between a graph representation like Abstract Syntax Tree (AST) and a source code sequence makes the use of current approaches computationally intractable for large input sequence lengths. Source code can have long-range dependencies that require larger sequence lengths to model effectively. Current approaches have a quadratic growth in computational and memory costs with respect to the sequence length. Using such models in practical scenarios is difficult. In this work, we propose the conditioning of a source code snippet with its graph modality by using the graph adjacency matrix as an attention mask for a sparse self-attention mechanism and the use of a graph diffusion mechanism to model longer-range token dependencies. Our model reaches state-of-the-art results in BLEU, METEOR, and ROUGE-L metrics for the code summarization task and near state-of-the-art accuracy in the variable misuse task. The memory use and inference time of our model have linear growth with respect to the input sequence length as compared to the quadratic growth of previous works.
['Barry Boehm', 'Iordanis Fostiropoulos', 'Junyan Cheng']
2021-12-01
null
null
null
null
['variable-misuse']
['computer-code']
[ 3.10119420e-01 2.14509368e-01 -1.89130083e-01 -1.46005765e-01 -7.71100640e-01 -6.05984867e-01 4.45073247e-01 9.34469163e-01 -2.72416621e-01 2.87521631e-01 4.62024957e-01 -7.74942875e-01 8.45155343e-02 -5.85510850e-01 -9.46020246e-01 -1.30059332e-01 -3.02409559e-01 2.08727926e-01 2.84781456e-01 -1.36452943e-01 5.97190917e-01 -3.49757932e-02 -1.39348543e+00 4.88878399e-01 1.08232033e+00 4.63551700e-01 5.60344577e-01 1.23173606e+00 -6.76817834e-01 1.39050293e+00 -6.39536560e-01 -5.45553446e-01 -1.06506772e-01 -5.45461595e-01 -9.57156122e-01 -1.15204826e-01 5.92427552e-01 2.23371789e-01 -5.33971548e-01 1.19589257e+00 3.25852603e-01 -1.91924088e-02 5.76309323e-01 -1.08104813e+00 -8.24135482e-01 1.27824986e+00 -8.26987743e-01 5.18357575e-01 5.32833397e-01 1.89905874e-02 1.24751532e+00 -4.26796943e-01 6.28992915e-01 1.13727856e+00 8.33601356e-01 1.75469041e-01 -1.25476325e+00 -2.71254182e-01 2.63478875e-01 3.71970564e-01 -1.05954659e+00 -2.09141284e-01 5.12664914e-01 -7.72534668e-01 1.89325809e+00 -6.40161429e-03 4.99759823e-01 6.80547237e-01 5.31373084e-01 6.58193111e-01 2.68611491e-01 -6.01954639e-01 1.32593825e-01 -3.57583165e-01 4.57663924e-01 1.21599472e+00 3.78920436e-01 -6.23627603e-01 -3.57915312e-01 -2.59487748e-01 4.24562693e-01 -1.84009135e-01 -1.14999153e-01 -1.87030733e-01 -1.06082380e+00 1.11551619e+00 6.03400826e-01 4.39880729e-01 -1.73991308e-01 8.82224500e-01 8.10045123e-01 5.96871674e-01 5.41882336e-01 5.70319414e-01 -3.34629595e-01 -5.30132234e-01 -1.25380278e+00 -3.58291417e-02 1.05149436e+00 1.25414789e+00 6.63655341e-01 -4.25372310e-02 -2.07251668e-01 8.24718058e-01 3.43330443e-01 3.42252105e-01 7.05048442e-01 -6.35553837e-01 9.42894757e-01 9.27302599e-01 -4.05777395e-01 -1.03129101e+00 -4.13774073e-01 -4.81825292e-01 -6.80109680e-01 -8.74266550e-02 2.93679208e-01 7.16396272e-02 -9.01239753e-01 1.47593331e+00 -1.85412154e-01 7.64626935e-02 -1.24216042e-01 2.31249630e-01 5.69244325e-01 7.09342539e-01 -2.48031393e-02 -1.30915999e-01 1.17065775e+00 -1.18954802e+00 -4.39636797e-01 -6.85741782e-01 1.13516641e+00 -7.72047162e-01 1.12552786e+00 8.46376792e-02 -1.06981874e+00 -3.14941734e-01 -1.12328589e+00 -2.48108193e-01 -1.20577566e-01 -8.44273195e-02 6.38855457e-01 7.78357327e-01 -1.27050531e+00 7.04482198e-01 -1.03847110e+00 -4.86029238e-01 4.28190291e-01 2.50028342e-01 -2.10980251e-01 -3.00650865e-01 -5.93864143e-01 7.86408722e-01 4.43465471e-01 -4.07532334e-01 -7.29003787e-01 -8.33479166e-01 -1.09172750e+00 4.22860593e-01 1.83168992e-01 -6.83689058e-01 1.10588205e+00 -9.96867597e-01 -1.17085636e+00 6.37093306e-01 -2.53476143e-01 -7.38864124e-01 1.83133915e-01 -2.73272306e-01 -5.82055300e-02 -5.45231774e-02 -1.04157433e-01 2.22510070e-01 9.06389892e-01 -6.76104486e-01 -3.43702585e-01 -8.20265934e-02 2.94250041e-01 8.37138593e-02 -3.06431741e-01 5.69255799e-02 -3.69635135e-01 -6.25654459e-01 -2.25316942e-01 -1.00493133e+00 -2.98998088e-01 -4.07634020e-01 -2.62330085e-01 -2.59862751e-01 5.06958067e-01 -9.48387861e-01 1.79156804e+00 -2.08418632e+00 5.22087157e-01 1.39862550e-02 3.53674710e-01 2.59474248e-01 -4.76293534e-01 8.19796681e-01 -1.95732251e-01 4.07936692e-01 -5.16830444e-01 -2.47265682e-01 -5.60249649e-02 1.07901677e-01 -9.76869464e-02 5.71002483e-01 1.41660601e-01 9.64412153e-01 -1.13125598e+00 -4.07631606e-01 -2.25902274e-01 2.59956986e-01 -9.94784057e-01 3.00735235e-01 -4.26537097e-01 -2.41094396e-01 -1.71177208e-01 3.38324383e-02 1.90206811e-01 -7.36197412e-01 2.92541653e-01 1.91001862e-01 1.25129446e-01 6.44533932e-01 -7.38562524e-01 2.21800065e+00 -8.63031089e-01 8.31310987e-01 -1.93404943e-01 -9.42944109e-01 6.79317474e-01 8.74468386e-02 7.46423230e-02 -3.14817041e-01 -1.99124783e-01 1.79247215e-01 2.87195772e-01 -6.48732662e-01 5.46487510e-01 3.02724004e-01 -1.65300444e-01 5.80694437e-01 3.31976861e-01 -1.87439188e-01 7.41400659e-01 8.13813806e-01 1.73527467e+00 -5.11393808e-02 5.56847751e-01 -4.07982469e-01 5.31578183e-01 -1.88394383e-01 6.00082837e-02 8.74223590e-01 4.75734591e-01 2.98770696e-01 1.17245114e+00 -2.80269802e-01 -1.19521558e+00 -5.04491925e-01 4.59592611e-01 1.30140007e+00 -2.71319419e-01 -1.08857071e+00 -1.00760901e+00 -9.17234063e-01 5.14682569e-02 8.59335601e-01 -5.52546382e-01 -2.71476150e-01 -7.84075677e-01 -5.73629975e-01 8.45146239e-01 5.52632332e-01 3.30097526e-02 -9.50699568e-01 -6.18874609e-01 4.50927138e-01 -1.42406568e-01 -8.33304644e-01 -7.09298790e-01 2.06413180e-01 -8.42751443e-01 -1.08089590e+00 -4.41505939e-01 -5.99502087e-01 7.13207304e-01 -1.66537821e-01 1.49075651e+00 6.87752068e-01 -3.62366229e-01 3.99706870e-01 -5.81023932e-01 -1.57695040e-01 -9.11910474e-01 5.78658283e-01 -6.01100624e-01 -4.30567622e-01 1.47673404e-02 -6.05807543e-01 -3.57404083e-01 -3.80674362e-01 -1.04023314e+00 3.12890336e-02 6.59668088e-01 9.15020525e-01 -5.77909201e-02 -3.72859746e-01 4.04406697e-01 -1.12381303e+00 7.29285896e-01 -7.85862267e-01 -6.49089098e-01 3.28564703e-01 -6.39293253e-01 5.68343401e-01 6.95063114e-01 -3.07536870e-01 -7.56285608e-01 7.78929368e-02 -9.60655361e-02 -1.21846393e-01 2.63516456e-01 1.08384669e+00 3.93588990e-01 -3.38990544e-03 7.51607537e-01 1.46741018e-01 -1.32356007e-02 -3.16583276e-01 6.26297832e-01 6.48907423e-01 1.54431999e-01 -4.56346512e-01 5.87651789e-01 -5.54409809e-02 7.06734194e-04 -9.08539832e-01 -6.22694790e-01 -5.67491055e-01 -7.81626999e-01 9.60224494e-02 7.27229536e-01 -7.32976019e-01 -2.43033379e-01 3.26003879e-01 -1.49393618e+00 -3.51101816e-01 -1.02389574e-01 1.35976836e-01 -4.36576843e-01 8.28342974e-01 -7.66465485e-01 -4.82805461e-01 -4.26339537e-01 -1.29155111e+00 9.69477415e-01 -1.59147471e-01 -2.93719798e-01 -1.24177384e+00 1.47170573e-01 1.55624002e-01 7.10514605e-01 1.50156513e-01 1.36829209e+00 -5.81968129e-01 -7.15012074e-01 -1.16451249e-01 -2.99435377e-01 1.73568651e-01 -6.78859651e-02 -3.71734612e-02 -5.79663515e-01 -5.51913977e-01 -1.41886666e-01 -1.49931163e-01 1.09645987e+00 1.50177181e-01 1.05223489e+00 -6.02771878e-01 -3.26231688e-01 5.00986934e-01 1.63262868e+00 -1.81087270e-01 6.31842017e-01 -3.07496823e-02 1.10411870e+00 2.33063444e-01 6.76612258e-02 6.08072698e-01 3.65457386e-01 3.92067671e-01 5.67465305e-01 3.48709732e-01 -4.42924857e-01 -1.69725522e-01 5.42654753e-01 1.43264401e+00 3.11794102e-01 -5.61721265e-01 -1.34516513e+00 8.20276260e-01 -1.92008913e+00 -9.70302582e-01 -4.79660064e-01 2.11937213e+00 8.20342720e-01 1.38823763e-01 -1.14606202e-01 -2.58120652e-02 5.33405364e-01 3.48179251e-01 -3.75362128e-01 -7.12204456e-01 3.55410188e-01 -7.34753860e-03 8.53569865e-01 6.57641172e-01 -7.68826604e-01 6.78159475e-01 6.01040649e+00 8.25967371e-01 -1.04608929e+00 2.65891045e-01 3.44426364e-01 -9.23020393e-02 -5.35047889e-01 2.16187224e-01 -4.59711492e-01 6.00140035e-01 1.51743877e+00 -4.42189366e-01 7.82559991e-01 8.34704220e-01 -3.62951428e-01 -1.03665367e-01 -1.38258052e+00 1.09546232e+00 3.99346352e-01 -1.46210301e+00 -8.41021538e-02 -1.22130446e-01 6.50187910e-01 5.99424243e-01 -4.23323095e-01 4.32581723e-01 6.07625484e-01 -1.05184281e+00 7.28974819e-01 3.61297637e-01 8.83447289e-01 -6.18076861e-01 6.51122987e-01 3.57375413e-01 -1.43743980e+00 -2.91450024e-01 -4.23117876e-01 -1.94912717e-01 3.60068232e-02 7.12108493e-01 -8.94229949e-01 5.68104506e-01 2.96451628e-01 8.87977183e-01 -1.11161506e+00 1.08669925e+00 -1.72925070e-01 8.71352613e-01 -2.52314121e-01 -2.87942559e-01 3.29974204e-01 1.47139609e-01 5.18237710e-01 1.83657300e+00 3.81974012e-01 -4.54509199e-01 1.87379703e-01 7.04716444e-01 -4.10438120e-01 1.99464902e-01 -6.67613983e-01 -4.56072569e-01 3.17900836e-01 8.89891744e-01 -8.41775835e-01 -5.31053960e-01 -6.90086126e-01 8.71001124e-01 7.70746708e-01 2.33464777e-01 -7.40742207e-01 -6.52951121e-01 3.11237961e-01 1.26315325e-01 7.14210153e-01 -4.10861045e-01 -3.21616203e-01 -1.32445896e+00 5.33670112e-02 -9.04514968e-01 4.52474207e-01 -4.13864285e-01 -9.78011191e-01 6.81324720e-01 -5.22075146e-02 -7.80582547e-01 -4.12527591e-01 -3.40558857e-01 -5.38760483e-01 7.88797140e-01 -1.36261666e+00 -9.20211554e-01 -1.09293923e-01 2.86207438e-01 7.07985342e-01 -2.35583648e-01 7.92775989e-01 3.36778909e-01 -3.00643384e-01 5.53634524e-01 1.60877451e-01 1.77064776e-01 3.14957500e-01 -1.46983647e+00 9.84604955e-01 1.17201030e+00 3.57414842e-01 8.38762760e-01 7.64341235e-01 -7.30245054e-01 -1.73612034e+00 -1.06749547e+00 9.53624010e-01 -4.21495020e-01 1.04600382e+00 -4.51213896e-01 -1.07831585e+00 8.93377066e-01 5.54790080e-01 1.02445692e-01 5.82574189e-01 2.99098134e-01 -7.88430929e-01 2.71002322e-01 -6.23552263e-01 2.73534089e-01 1.13862205e+00 -7.53030777e-01 -5.93332231e-01 4.45550561e-01 8.82384002e-01 -4.69165593e-01 -8.66553962e-01 -1.52357876e-01 2.02165678e-01 -7.69616425e-01 4.24565047e-01 -4.70812678e-01 7.04349041e-01 -1.75208151e-01 7.23198578e-02 -1.48361146e+00 -4.69000787e-01 -7.44428337e-01 -3.47508878e-01 1.19838822e+00 5.37811399e-01 -2.34332815e-01 3.91448855e-01 1.67424992e-01 -2.50681251e-01 -5.42549968e-01 -8.56468737e-01 -7.59023547e-01 6.06651306e-02 -3.88839573e-01 3.12882602e-01 8.56789708e-01 5.24276614e-01 7.15235174e-01 -2.65080668e-02 -7.13935494e-02 4.58498508e-01 3.45714800e-02 5.39165616e-01 -1.12626171e+00 -5.63829839e-01 -6.28603458e-01 -6.01260602e-01 -1.06674194e+00 4.53641355e-01 -1.51668036e+00 -1.06247598e-02 -1.93622100e+00 3.08412343e-01 -1.95569143e-01 -3.49319018e-02 4.04254645e-01 -3.06925654e-01 -4.34216887e-01 3.26247066e-01 1.02941103e-01 -7.00960875e-01 3.04203540e-01 4.27162975e-01 -3.86698127e-01 1.33488595e-01 -4.20881599e-01 -4.46591526e-01 4.56594080e-01 7.05700994e-01 -8.75856876e-01 -6.08526230e-01 -8.32458496e-01 7.95071483e-01 3.81006569e-01 -1.35772243e-01 -9.18312490e-01 3.75177890e-01 2.50070065e-01 -2.67196536e-01 -1.79337114e-01 -2.16351539e-01 -4.98234928e-01 -2.28281952e-02 6.42875552e-01 -6.18231714e-01 5.47375381e-01 3.11669677e-01 7.84594774e-01 -1.43111765e-01 -7.07383692e-01 6.03767812e-01 -2.72999734e-01 -4.77535307e-01 1.23557998e-02 -5.21272957e-01 4.11116362e-01 7.37114370e-01 1.95125863e-01 -4.68855530e-01 -3.09035838e-01 -1.41208157e-01 5.81644364e-02 6.10571742e-01 6.86477959e-01 5.04250467e-01 -8.30515385e-01 -8.62617195e-01 1.03838064e-01 5.42164892e-02 -3.69620562e-01 8.73893425e-02 7.22911835e-01 -9.01244581e-01 3.07320148e-01 -1.42633691e-01 -4.45837468e-01 -1.50150454e+00 9.05248404e-01 1.65240005e-01 -7.53201783e-01 -6.53839469e-01 9.10356641e-01 -8.69213715e-02 -1.69689044e-01 7.95956329e-02 -6.40362799e-01 -1.72567651e-01 4.92845736e-02 5.27465165e-01 3.71432364e-01 3.50601047e-01 -4.08646256e-01 -3.69945765e-01 3.45871001e-01 -1.49164602e-01 2.68809944e-01 1.32985890e+00 2.54561659e-02 -6.55028880e-01 5.07768571e-01 1.49991965e+00 1.33894011e-01 -9.77548361e-01 -1.94979921e-01 4.82973605e-01 -3.62452120e-01 3.62210013e-02 -5.29237330e-01 -9.46644902e-01 8.58035326e-01 2.20104828e-01 3.89771283e-01 6.32698476e-01 6.10567890e-02 5.70265532e-01 3.88564825e-01 3.04489285e-01 -6.13648713e-01 1.24013312e-01 1.02827871e+00 8.56610775e-01 -9.28138196e-01 7.51526281e-02 -2.22884208e-01 -3.61989945e-01 1.21046901e+00 2.13063598e-01 -3.23590636e-01 3.99309188e-01 6.14675045e-01 -2.80733258e-01 -2.24824056e-01 -1.07312369e+00 7.63901994e-02 4.06806976e-01 3.95587772e-01 9.77084100e-01 -1.26125231e-01 -3.69940042e-01 1.27008975e-01 -9.87984538e-02 -2.59760112e-01 9.71148908e-01 9.05640066e-01 -4.48523849e-01 -1.24280238e+00 1.53503530e-02 9.13223684e-01 -7.16669738e-01 -7.05763102e-01 -1.09512530e-01 3.84481072e-01 -4.97257620e-01 9.58900750e-01 5.63705750e-02 -1.13358259e-01 1.17367081e-01 2.65031785e-01 6.15336537e-01 -1.09516311e+00 -1.02599919e+00 -4.27470952e-01 2.02679455e-01 -6.72193289e-01 -1.69817820e-01 -6.12893760e-01 -1.22525787e+00 -3.41271371e-01 -2.28490293e-01 1.16317563e-01 6.30103767e-01 7.63181448e-01 6.10554457e-01 8.62598836e-01 -1.71742588e-02 -6.76344991e-01 -6.04757845e-01 -1.17096329e+00 -2.12472215e-01 4.96475488e-01 4.60054398e-01 -1.72003105e-01 -3.15246820e-01 3.30908537e-01]
[7.55947732925415, 7.9532623291015625]
b30cf50a-6bfd-4758-902d-a7b903cc8239
dscribe-library-of-descriptors-for-machine
1904.08875
null
http://arxiv.org/abs/1904.08875v1
http://arxiv.org/pdf/1904.08875v1.pdf
DScribe: Library of Descriptors for Machine Learning in Materials Science
DScribe is a software package for machine learning that provides popular feature transformations ("descriptors") for atomistic materials simulations. DScribe accelerates the application of machine learning for atomistic property prediction by providing user-friendly, off-the-shelf descriptor implementations. The package currently contains implementations for Coulomb matrix, Ewald sum matrix, sine matrix, Many-body Tensor Representation (MBTR), Atom-centered Symmetry Function (ACSF) and Smooth Overlap of Atomic Positions (SOAP). Usage of the package is illustrated for two different applications: formation energy prediction for solids and ionic charge prediction for atoms in organic molecules. The package is freely available under the open-source Apache License 2.0.
['Adam S. Foster', 'Patrick Rinke', 'Filippo Federici Canova', 'Marc O. J. Jäger', 'David Z. Gao', 'Yashasvi S. Ranawat', 'Eiaki V. Morooka', 'Lauri Himanen']
2019-04-18
null
null
null
null
['formation-energy']
['miscellaneous']
[-2.02148050e-01 -4.50399101e-01 -1.32809877e-01 -4.03710991e-01 -4.49505806e-01 -3.13572377e-01 5.93530297e-01 7.02695072e-01 -3.36394787e-01 1.02975607e+00 3.56786884e-02 -5.05999744e-01 1.18537468e-03 -9.18978035e-01 -6.10348105e-01 -1.19655073e+00 -3.57441276e-01 6.81737900e-01 2.44684249e-01 -4.93975282e-01 6.91169918e-01 9.34256971e-01 -1.55050564e+00 4.89429832e-01 7.68597126e-01 7.64939427e-01 1.82461292e-01 4.21256512e-01 1.86683342e-01 3.48947167e-01 5.19420616e-02 -3.19858402e-01 -5.39700575e-02 -3.52702476e-02 -9.36851203e-01 -9.38422203e-01 -7.94018432e-02 3.17114264e-01 1.29536286e-01 5.35403311e-01 6.82392836e-01 3.20233166e-01 1.40441787e+00 -1.08335185e+00 -8.04285944e-01 2.68687397e-01 -1.67074159e-01 -2.30594456e-01 8.55672240e-01 2.55195826e-01 1.00430632e+00 -1.29740191e+00 7.76906908e-01 1.05357027e+00 7.81187475e-01 3.54018122e-01 -1.22925055e+00 -6.19674921e-01 -7.56811440e-01 5.49573421e-01 -1.40143692e+00 -7.20139891e-02 4.63540167e-01 -4.12447900e-01 2.18241191e+00 6.20307267e-01 5.34362674e-01 7.64698267e-01 1.06706929e+00 3.27265292e-01 8.96742105e-01 -5.15514851e-01 6.42535031e-01 -4.53072004e-02 3.36395115e-01 3.68252546e-01 4.91227359e-01 1.36055529e-01 -6.35043919e-01 -9.55732346e-01 2.94008851e-01 -5.15838601e-02 4.00721252e-01 -8.28221977e-01 -1.24563932e+00 8.23968351e-01 3.37305456e-01 2.70371903e-02 -7.10066497e-01 4.45555188e-02 6.37919009e-01 2.96180367e-01 3.63681018e-01 5.68268478e-01 -7.84137487e-01 -4.23308998e-01 -3.50144058e-01 9.57297444e-01 1.00546598e+00 7.64266014e-01 9.43049610e-01 1.00478604e-02 4.72806334e-01 7.13687837e-01 5.64994633e-01 3.74452978e-01 4.09984857e-01 -5.65392733e-01 -3.26275289e-01 2.73419529e-01 3.28032702e-01 -4.23641950e-01 -5.00976801e-01 3.66525769e-01 -7.01018870e-01 2.93547064e-01 -4.04789038e-02 2.32722580e-01 -5.00696182e-01 9.17878449e-01 5.72184086e-01 -1.53052807e-01 1.72272980e-01 5.20170748e-01 1.16501021e+00 6.72714472e-01 3.58867496e-01 -4.12622780e-01 1.07359827e+00 -7.62741208e-01 -5.21958053e-01 7.75798500e-01 1.09020269e+00 -9.05520320e-01 5.55719197e-01 5.80473542e-01 -1.21112692e+00 -4.54816937e-01 -7.85929620e-01 -1.99932709e-01 -9.76598859e-01 -3.83948296e-01 1.22241664e+00 4.74637479e-01 -7.09058881e-01 1.18768966e+00 -1.07239878e+00 -2.68964648e-01 4.61276732e-02 9.26293254e-01 -8.38557959e-01 3.23116183e-01 -1.29452038e+00 1.10372007e+00 2.91923404e-01 -6.91310227e-01 -4.63768750e-01 -1.03054273e+00 -8.72857094e-01 -3.97364885e-01 -2.34665707e-01 -6.41043007e-01 9.69576478e-01 -3.71517777e-01 -1.65039134e+00 8.63620281e-01 -3.55528831e-01 -1.79013267e-01 6.78138658e-02 -3.79778966e-02 -3.59954894e-01 -1.70762181e-01 -4.03456837e-02 4.83189195e-01 3.98291916e-01 -6.15836263e-01 2.94191033e-01 -1.86061472e-01 -6.91421866e-01 -1.31394742e-02 2.96246409e-01 3.26491535e-01 5.21123707e-01 -4.16737378e-01 -9.37309191e-02 -8.85966778e-01 -3.76617640e-01 -1.59993857e-01 -2.85392940e-01 -6.10375047e-01 7.66787827e-01 -3.79545927e-01 9.08357918e-01 -1.61070585e+00 3.32906187e-01 5.29841304e-01 -3.01672339e-01 5.04335724e-02 1.01054676e-01 1.44611657e+00 -9.20516133e-01 -1.30945519e-01 -4.87004034e-02 1.72247887e-01 -4.71542440e-02 -1.66065156e-01 1.92186251e-01 6.88511968e-01 4.09961939e-02 7.08264709e-01 -7.59777844e-01 -1.57899320e-01 5.23088098e-01 6.38069868e-01 -7.31196702e-01 1.03321142e-01 -2.32598811e-01 2.15829000e-01 -4.75347549e-01 8.33894670e-01 1.13409400e+00 -7.11945817e-02 -2.28996240e-02 -3.62556338e-01 -7.68525600e-01 5.42995989e-01 -1.30274320e+00 1.43163645e+00 4.94203903e-02 -2.23857194e-01 -3.58574599e-01 -5.31628907e-01 1.06225121e+00 2.20825821e-01 8.19701493e-01 -3.20983887e-01 -3.50668058e-02 5.01924574e-01 2.78492093e-01 -4.50114161e-01 2.46170834e-01 -1.69795379e-01 3.20100427e-01 5.52489698e-01 -3.28380130e-02 -4.73478228e-01 7.26891160e-02 1.28048629e-01 9.25399780e-01 5.55325866e-01 9.07082736e-01 -6.91169441e-01 6.05869055e-01 1.06419392e-01 -1.89440519e-01 1.75018072e-01 1.78850070e-01 3.07384074e-01 1.56690180e-01 -8.45720053e-01 -1.55021334e+00 -9.02597308e-01 -8.00549865e-01 1.20870578e+00 -1.66315988e-01 -9.30201054e-01 -6.26204789e-01 1.83134049e-01 6.45597875e-01 7.11083174e-01 -4.65027690e-01 -2.59276867e-01 -1.95730507e-01 -8.76255691e-01 -3.24558020e-02 3.61858904e-01 -1.62527144e-01 -1.61661470e+00 -2.46962294e-01 2.38235310e-01 5.71387112e-01 -3.35526139e-01 -7.14132562e-02 6.25877440e-01 -6.62629187e-01 -9.83939886e-01 -1.46929696e-01 -2.87078708e-01 2.52444863e-01 1.78405985e-01 8.97413790e-01 1.23650104e-01 -6.80741787e-01 1.78077310e-01 -4.00713921e-01 -5.33872306e-01 -5.03039241e-01 -2.18937192e-02 5.76950192e-01 -4.64137971e-01 8.35128963e-01 -7.05128372e-01 -6.61567569e-01 1.51976407e-01 -6.77613318e-01 2.65849959e-02 2.26482078e-01 7.54876852e-01 8.78632307e-01 -4.91592258e-01 3.92100155e-01 -4.64519829e-01 6.72870278e-01 -6.03071809e-01 -2.33066872e-01 -4.20497879e-02 -7.15056181e-01 2.74199545e-01 6.35998785e-01 -6.28454760e-02 -6.15153253e-01 2.39535183e-01 -6.85534537e-01 9.35998261e-02 -2.51900136e-01 4.37989235e-01 -1.94323868e-01 -5.47293544e-01 5.43865263e-01 4.25346494e-01 1.78623512e-01 -7.38872290e-01 -1.29455747e-02 9.21028018e-01 9.31968763e-02 -9.69644427e-01 3.84616464e-01 2.83842325e-01 4.04049039e-01 -1.20618439e+00 -6.68402463e-02 -5.60853541e-01 -1.18414128e+00 1.48279443e-01 7.89531052e-01 -6.82703197e-01 -1.49903929e+00 4.81413394e-01 -8.58240604e-01 -2.97973640e-02 -5.12769353e-03 5.33024132e-01 -7.84717262e-01 4.55425471e-01 -3.40209424e-01 -5.44511974e-01 -8.64804029e-01 -1.21820164e+00 1.08990717e+00 1.91015415e-02 -5.85767627e-01 -9.98277664e-01 6.46618962e-01 1.23881251e-01 4.30884421e-01 3.92972767e-01 1.12795722e+00 -1.01122129e+00 -2.01496869e-01 -3.23446631e-01 2.10585877e-01 2.66035527e-01 1.17171928e-01 5.78628778e-01 -6.48759067e-01 -3.64284962e-01 -7.49312758e-01 -3.73265773e-01 6.76107347e-01 4.80342329e-01 1.19607592e+00 -5.83064146e-02 -5.79930663e-01 6.67613685e-01 1.27619874e+00 4.16626185e-01 8.16209793e-01 5.39180338e-01 6.91555798e-01 2.82742441e-01 6.57199681e-01 8.33193541e-01 -5.70950694e-02 8.07223558e-01 5.60765326e-01 7.76053369e-02 4.79408264e-01 1.42967775e-01 3.24054182e-01 6.86151981e-01 -7.99725950e-01 1.46407396e-01 -1.10125911e+00 -1.08458765e-01 -1.63403463e+00 -1.30800951e+00 -8.58739197e-01 2.25619602e+00 1.03960598e+00 -1.45779759e-01 2.60260582e-01 4.89507020e-02 1.72101349e-01 -1.97901443e-01 -7.59147823e-01 -1.09114933e+00 9.20939445e-02 8.23210180e-01 4.48566496e-01 6.86867416e-01 -1.01710451e+00 1.11439681e+00 7.44100332e+00 8.68343651e-01 -1.17525995e+00 6.96030855e-02 6.18295791e-03 2.40365535e-01 -4.00852680e-01 1.99167103e-01 -8.16067338e-01 5.30316889e-01 1.29044187e+00 -1.33151487e-01 1.58442691e-01 1.11877942e+00 5.57873368e-01 -2.00515375e-01 -9.98061419e-01 6.69526160e-01 -5.82295895e-01 -1.71570253e+00 9.34567899e-02 1.91630527e-01 4.31830168e-01 2.95695424e-01 -1.47910297e-01 -1.03107162e-01 -2.18606785e-01 -1.20456743e+00 4.26403254e-01 6.02895737e-01 9.11060989e-01 -1.02847803e+00 6.94957018e-01 -1.04898661e-01 -1.16181254e+00 5.18059015e-01 -6.85431182e-01 -3.03135067e-01 -6.17794916e-02 4.80183125e-01 -7.26159394e-01 4.54454243e-01 4.74932671e-01 9.64579880e-01 -4.78263289e-01 1.01145768e+00 6.21737361e-01 3.16072524e-01 -2.48082206e-01 -3.81328851e-01 2.05874562e-01 -8.02583754e-01 3.60010028e-01 1.39654291e+00 -1.56361852e-02 1.88493520e-01 -1.19406663e-01 5.36589146e-01 4.40986633e-01 4.60581422e-01 -4.95279133e-01 9.62163433e-02 5.59835017e-01 1.02774298e+00 -4.98378128e-01 -8.83400589e-02 -2.61659294e-01 7.99247622e-01 1.97709482e-02 -6.34087324e-02 -8.09293211e-01 -7.21824765e-01 1.20655930e+00 5.17206490e-01 2.42684245e-01 -3.25535774e-01 2.57275730e-01 -7.70787716e-01 -5.47658443e-01 -9.58480120e-01 -5.78006310e-03 -1.08595002e+00 -1.20037532e+00 4.75576371e-02 3.19867373e-01 -8.39842200e-01 -7.42863193e-02 -1.41632915e+00 -8.69317651e-01 8.83194625e-01 -1.13797796e+00 -1.04252577e+00 5.36323451e-02 5.81843972e-01 1.15753882e-01 -2.55076647e-01 1.43503964e+00 7.74459988e-02 -4.03191000e-01 2.14979544e-01 9.40102458e-01 -6.82707846e-01 9.83628392e-01 -1.33171940e+00 6.69975281e-01 -4.42005932e-01 -5.55170298e-01 1.04178572e+00 1.17273533e+00 -8.55133891e-01 -1.88033926e+00 -7.03215599e-01 8.36307764e-01 -2.81225026e-01 8.62447321e-01 -2.91895151e-01 -1.03075874e+00 2.96165287e-01 1.80142596e-01 -1.14484718e-02 1.28298676e+00 -8.92550871e-02 -3.69396001e-01 2.09075332e-01 -1.16331518e+00 1.98711038e-01 7.31130958e-01 -3.61508608e-01 -2.74069756e-01 1.02788126e+00 3.07712495e-01 -2.53762990e-01 -1.53557539e+00 2.58094639e-01 8.36368799e-01 -1.12910879e+00 1.34745455e+00 -1.07502484e+00 1.06549881e-01 -9.04988423e-02 -1.05745129e-01 -8.56100738e-01 -6.16299331e-01 -8.40540707e-01 -1.33862644e-01 3.87817740e-01 4.23147649e-01 -7.71807551e-01 4.97596741e-01 6.19464397e-01 -4.35563415e-01 -1.04463327e+00 -1.01703477e+00 -5.78964651e-01 3.70327771e-01 -2.94731438e-01 7.69944072e-01 8.35473180e-01 4.30484354e-01 1.10180579e-01 -3.55438322e-01 -2.81910896e-01 3.56595188e-01 -2.52399910e-02 6.81560457e-01 -1.51570606e+00 -1.40146166e-01 -1.96557641e-02 -6.08368874e-01 -3.48329782e-01 2.70359606e-01 -1.05471182e+00 -7.65468836e-01 -1.24629271e+00 5.34973443e-01 -4.06330854e-01 -2.68098235e-01 8.73410881e-01 3.99839908e-01 -1.05288923e-01 -3.72815162e-01 3.39401275e-01 -3.69814157e-01 6.24250352e-01 1.11132085e+00 8.99513811e-02 -1.23704113e-01 1.02873996e-01 -2.93429375e-01 3.79545540e-01 7.48843253e-01 -6.73472047e-01 6.74423799e-02 6.58041835e-01 5.11831105e-01 -4.17544574e-01 3.06020230e-01 -8.03383291e-01 -1.68407664e-01 -4.04693484e-01 3.39967638e-01 -9.65815306e-01 5.74911356e-01 -6.34071767e-01 5.44264555e-01 8.74521673e-01 1.63356606e-02 3.53006542e-01 2.07287639e-01 3.70615534e-02 1.61757678e-01 -4.30005670e-01 9.54645038e-01 -1.47747204e-01 -1.62109092e-01 3.41622621e-01 -5.40094972e-01 -6.84223413e-01 1.23792815e+00 -6.18121862e-01 -2.42379025e-01 -1.31956801e-01 -6.90592706e-01 -4.33055937e-01 8.28324616e-01 9.61603373e-02 6.98568881e-01 -1.41122687e+00 -5.24109423e-01 5.61000049e-01 1.27002984e-01 -5.92445612e-01 1.19069293e-01 9.38720465e-01 -1.12014294e+00 7.48591661e-01 -5.27383685e-01 -2.68127024e-01 -1.66110134e+00 6.06195688e-01 4.10632640e-01 1.62169024e-01 -4.19806808e-01 7.26578891e-01 -3.80936295e-01 -4.27060425e-01 -3.30568641e-01 -2.56308079e-01 -1.49817578e-02 -9.64164734e-02 5.93581975e-01 6.75855935e-01 4.11602765e-01 -8.59770834e-01 -8.06966484e-01 7.26924717e-01 -4.02588516e-01 4.85396087e-01 1.92103100e+00 5.95305502e-01 -7.24620521e-01 2.45434120e-01 1.45748949e+00 -3.53183560e-02 -7.06157506e-01 4.10162449e-01 -1.35221347e-01 -2.24928800e-02 -1.84419155e-01 -5.69377244e-01 -1.43031150e-01 6.18465364e-01 6.85690403e-01 4.90159430e-02 3.77131224e-01 -1.22819193e-01 5.84548712e-01 9.76527572e-01 3.53533536e-01 -1.00460100e+00 -1.45466864e-01 7.36736119e-01 9.05852735e-01 -1.09924757e+00 7.42871106e-01 -3.86762708e-01 -5.84872544e-01 1.39719760e+00 3.04592550e-01 -2.92174816e-01 1.06687427e+00 3.07844907e-01 -4.93466794e-01 -3.61556441e-01 -9.14999008e-01 2.12308854e-01 3.63864332e-01 7.03178942e-01 1.14215052e+00 2.06596002e-01 -6.89106643e-01 3.13620120e-01 -3.09410989e-01 -3.55687171e-01 4.19194818e-01 1.24636543e+00 -5.78430712e-01 -1.73679352e+00 -5.81102073e-01 3.12901646e-01 -3.15739691e-01 -3.98707986e-01 -6.95730209e-01 7.89594233e-01 1.68811783e-01 3.76437575e-01 -4.42207605e-02 -2.29847088e-01 1.04466282e-01 3.92488033e-01 7.16799676e-01 -6.26012743e-01 -7.67034769e-01 -1.47133395e-02 -3.28022824e-03 -6.86701179e-01 -4.53949809e-01 -9.18076634e-01 -1.65545619e+00 -5.96228600e-01 -5.07359803e-01 6.06742203e-01 9.20798540e-01 7.34202206e-01 6.89455867e-01 -6.59326538e-02 3.28012556e-01 -1.41234410e+00 -3.12314987e-01 -9.97613490e-01 -7.76053011e-01 2.03871846e-01 -6.62894994e-02 -9.30641055e-01 -1.87383831e-01 -3.42183784e-02]
[5.120254993438721, 5.380608081817627]
2db6145d-2251-4257-827d-e3dce423530b
i-2r-net-intra-and-inter-human-relation
2206.10892
null
https://arxiv.org/abs/2206.10892v2
https://arxiv.org/pdf/2206.10892v2.pdf
I^2R-Net: Intra- and Inter-Human Relation Network for Multi-Person Pose Estimation
In this paper, we present the Intra- and Inter-Human Relation Networks (I^2R-Net) for Multi-Person Pose Estimation. It involves two basic modules. First, the Intra-Human Relation Module operates on a single person and aims to capture Intra-Human dependencies. Second, the Inter-Human Relation Module considers the relation between multiple instances and focuses on capturing Inter-Human interactions. The Inter-Human Relation Module can be designed very lightweight by reducing the resolution of feature map, yet learn useful relation information to significantly boost the performance of the Intra-Human Relation Module. Even without bells and whistles, our method can compete or outperform current competition winners. We conduct extensive experiments on COCO, CrowdPose, and OCHuman datasets. The results demonstrate that the proposed model surpasses all the state-of-the-art methods. Concretely, the proposed method achieves 77.4% AP on CrowPose dataset and 67.8% AP on OCHuman dataset respectively, outperforming existing methods by a large margin. Additionally, the ablation study and visualization analysis also prove the effectiveness of our model.
['Ming Zeng', 'Dong Chen', 'Jianmin Bao', 'Xuan Cheng', 'Meihong Wang', 'PengFei Liu', 'Yinglin Zheng', 'Wenjin Deng', 'Yiwei Ding']
2022-06-22
null
null
null
null
['multi-person-pose-estimation']
['computer-vision']
[-3.30113381e-01 9.10478830e-02 1.78397447e-01 -3.02695185e-01 -3.39204878e-01 -2.68809915e-01 5.29443562e-01 -9.50979143e-02 -4.85320956e-01 7.03947544e-01 2.89235294e-01 3.75497252e-01 -2.29748964e-01 -4.93788093e-01 -5.00996709e-01 -3.40838909e-01 -3.83923471e-01 7.38921881e-01 5.00443161e-01 -5.11338174e-01 -2.01368541e-01 1.54886439e-01 -1.58051169e+00 1.60244629e-01 6.49480045e-01 9.45715249e-01 -2.15135232e-01 5.76803982e-01 5.41932225e-01 7.92172730e-01 -6.81534171e-01 -8.11598897e-01 2.05968887e-01 -1.85756221e-01 -6.62728190e-01 -2.37709969e-01 4.39721376e-01 -1.55061170e-01 -4.25744802e-01 7.11566567e-01 8.98438275e-01 2.32191175e-01 4.87832844e-01 -1.40903676e+00 -3.57168704e-01 5.94121397e-01 -9.89690781e-01 2.76971668e-01 8.87602806e-01 -5.97856306e-02 9.55127954e-01 -9.74984288e-01 4.77652043e-01 1.56461513e+00 8.88353109e-01 3.06404501e-01 -6.65020347e-01 -8.62118185e-01 2.99983084e-01 1.31877899e-01 -1.76550639e+00 -2.60214627e-01 4.31243479e-01 -3.85920703e-01 8.02086949e-01 4.18278039e-01 7.65322685e-01 1.04353285e+00 7.86886811e-02 9.54649568e-01 8.45971167e-01 -2.89702505e-01 -3.43073249e-01 -3.17020901e-02 3.71594131e-01 7.45427668e-01 3.48221421e-01 -6.87964186e-02 -9.72024858e-01 -5.73644303e-02 7.78452992e-01 -7.77297467e-02 -3.30550492e-01 -2.19430342e-01 -1.21948969e+00 4.04363394e-01 5.91866016e-01 5.31134615e-03 -1.32854268e-01 -1.02204792e-02 4.70777303e-01 -1.38661787e-01 2.31359571e-01 7.94480816e-02 -1.42601177e-01 -2.29928821e-01 -5.50297439e-01 4.37332392e-01 7.81738162e-01 1.34251750e+00 2.74469316e-01 -5.27482390e-01 -3.61308932e-01 7.93194056e-01 2.17511564e-01 2.61739165e-01 1.39837742e-01 -5.35986900e-01 7.94611931e-01 7.54854441e-01 3.03146899e-01 -1.38278902e+00 -8.69334102e-01 -4.81510282e-01 -1.08799970e+00 -3.56010407e-01 3.42364997e-01 -1.47366688e-01 -5.42631924e-01 1.53619015e+00 4.11583006e-01 1.57104015e-01 -2.02521682e-01 1.01151299e+00 1.22048080e+00 1.96723208e-01 1.53184801e-01 -1.11908413e-01 1.62481058e+00 -1.39019656e+00 -8.45321655e-01 -3.51699173e-01 2.79475749e-01 -7.13441551e-01 8.10571015e-01 2.66464919e-01 -1.06608868e+00 -9.41580534e-01 -9.88547742e-01 5.83540872e-02 -2.84804910e-01 4.26940858e-01 7.40230083e-01 5.31330824e-01 -5.99340975e-01 3.46778810e-01 -4.77616549e-01 -6.51003480e-01 2.52427876e-01 6.95369840e-01 -6.26823545e-01 7.69435428e-03 -1.44774246e+00 9.17156339e-01 3.91974807e-01 4.68279600e-01 -3.71674657e-01 -2.50666380e-01 -7.17712283e-01 -1.53618798e-01 6.15611076e-01 -7.27572620e-01 9.51760888e-01 -4.00873214e-01 -1.20607305e+00 7.32044458e-01 -5.99996746e-02 -1.98741183e-01 8.24140251e-01 -9.17413950e-01 -6.04276001e-01 2.62620319e-02 9.13723335e-02 4.40019757e-01 2.69731432e-01 -1.14608812e+00 -8.74390125e-01 -4.21850592e-01 9.44257975e-02 5.72300494e-01 -1.94706321e-01 1.60692990e-01 -1.16399086e+00 -5.50911009e-01 -1.30043700e-01 -1.14718914e+00 3.32406946e-02 -3.27916265e-01 -7.16635108e-01 -4.11448210e-01 6.93590224e-01 -6.01779163e-01 1.54217243e+00 -1.96498001e+00 2.87905365e-01 2.29899317e-01 4.73144919e-01 3.79245460e-01 1.05162300e-01 5.63690841e-01 1.13797393e-02 -1.39437154e-01 2.59347647e-01 -5.15923083e-01 1.48297949e-02 -7.54980324e-03 3.81868511e-01 6.56789899e-01 -2.82552063e-01 1.09428144e+00 -6.23240173e-01 -7.80210078e-01 1.61449537e-01 6.67631388e-01 -2.69035608e-01 3.14236492e-01 4.67619330e-01 3.76768351e-01 -2.48428345e-01 7.24125564e-01 6.68538868e-01 -2.95115173e-01 3.28383595e-01 -3.67561907e-01 1.18210718e-01 -2.65295655e-01 -1.49620080e+00 1.66288686e+00 3.76741923e-02 2.59912729e-01 -1.52094439e-01 -4.81453240e-01 9.15732205e-01 1.41346335e-01 2.98621625e-01 -5.50214350e-01 2.02499300e-01 8.86954833e-03 6.17180765e-02 -6.17304862e-01 4.16880608e-01 3.66491139e-01 -3.01709086e-01 1.32014215e-01 -5.25126830e-02 7.09540844e-01 2.96142042e-01 2.48519287e-01 9.15138006e-01 3.33770663e-01 5.62326193e-01 -2.48110577e-01 7.67176867e-01 -7.46204793e-01 6.64870083e-01 6.73283339e-01 -4.86193776e-01 5.05273700e-01 4.77739602e-01 -8.63121629e-01 -6.07224345e-01 -1.00572288e+00 2.30872914e-01 1.26579177e+00 7.23670721e-01 -7.33992815e-01 -7.06184864e-01 -8.60807121e-01 3.95245962e-02 7.23817647e-02 -9.31675911e-01 2.29521677e-01 -6.18079662e-01 -7.22672880e-01 7.36467719e-01 7.90750325e-01 9.78049874e-01 -8.28202069e-01 -6.01683795e-01 -1.22641370e-01 -5.54492891e-01 -1.41865468e+00 -5.65968275e-01 -3.85522693e-01 -1.03572950e-01 -1.22740722e+00 -7.23716795e-01 -7.59251535e-01 3.73701364e-01 6.79976121e-02 1.22419548e+00 1.97971180e-01 -2.56011069e-01 1.66120991e-01 -4.23674196e-01 -3.19849163e-01 5.24495184e-01 3.88162374e-01 3.92833412e-01 8.95589516e-02 6.11603439e-01 -5.85520089e-01 -7.84849703e-01 8.81937861e-01 -2.07590871e-02 8.84302855e-02 6.69891477e-01 6.37681127e-01 3.21318030e-01 8.78733695e-02 3.98770541e-01 -9.27249610e-01 6.39441967e-01 -1.88650295e-01 2.75252834e-02 4.37029511e-01 -4.17613477e-01 -2.92338192e-01 3.21920544e-01 -3.47308159e-01 -9.21604276e-01 1.43905342e-01 2.55714893e-01 1.93718132e-02 -5.21960296e-02 1.95075393e-01 -4.66552079e-01 -3.40908021e-02 6.61709845e-01 -8.43241438e-02 -3.31522048e-01 -3.94561708e-01 1.81619935e-02 7.84749806e-01 8.24125767e-01 -7.08790123e-01 7.92961121e-01 3.49884510e-01 1.79749861e-01 -5.78887403e-01 -9.95624721e-01 -6.30477488e-01 -9.87254024e-01 -5.79682112e-01 9.14876640e-01 -1.21919131e+00 -1.37081063e+00 4.30904776e-01 -1.21769381e+00 1.63966969e-01 1.89881340e-01 2.84329534e-01 -1.78217173e-01 3.11675996e-01 -6.81198537e-01 -9.96078253e-01 -4.20764148e-01 -9.08989310e-01 1.16654146e+00 4.01482791e-01 -4.48397815e-01 -7.10846066e-01 -7.92749375e-02 7.11418390e-01 6.92139342e-02 4.99786854e-01 1.10503800e-01 -7.83401430e-01 -1.63724646e-01 -5.11831403e-01 -3.77974927e-01 -1.78238034e-01 -2.60091219e-02 -2.61226714e-01 -8.32987785e-01 -3.19512546e-01 -7.69042909e-01 -3.84450823e-01 5.31884730e-01 -2.63476297e-02 8.08701694e-01 -7.98035339e-02 -6.10924900e-01 3.32297415e-01 8.23706329e-01 -4.97802682e-02 7.16268897e-01 3.96302819e-01 1.04261386e+00 7.92433918e-01 7.58954406e-01 4.87793237e-01 9.33887422e-01 1.05303049e+00 1.62332207e-01 -2.15731710e-01 3.58223692e-02 -4.20564502e-01 -2.81663723e-02 6.78279996e-01 -8.19888532e-01 -3.07363570e-01 -1.06656206e+00 1.32854074e-01 -2.51237154e+00 -9.49686885e-01 -3.93503368e-01 1.98249567e+00 3.62536848e-01 2.43219420e-01 7.84027100e-01 3.63612473e-02 9.26353872e-01 7.04391720e-03 -2.54991263e-01 1.73244968e-01 -3.26890498e-02 -1.54382944e-01 3.01917374e-01 4.41615373e-01 -1.34825552e+00 9.17639673e-01 6.14817572e+00 6.56124234e-01 -2.03660250e-01 -5.16827703e-02 5.22762537e-01 -2.20671788e-01 6.17732167e-01 -4.99420971e-01 -1.11819267e+00 4.02372539e-01 4.00850117e-01 1.04925320e-01 2.61911750e-01 7.80585825e-01 -5.09103425e-02 -1.79515168e-01 -1.31243944e+00 1.43256974e+00 3.44995975e-01 -7.43489385e-01 -8.38631615e-02 -1.71480384e-02 3.60374063e-01 -6.34499907e-01 -1.76917791e-01 5.28628886e-01 1.11260870e-02 -1.35159171e+00 5.65836966e-01 6.74633265e-01 6.07197106e-01 -1.31284845e+00 1.35354197e+00 4.58285779e-01 -1.81861198e+00 1.80943497e-02 -1.36155099e-01 -4.39488977e-01 2.02032924e-01 3.25510889e-01 -1.37173444e-01 9.21601117e-01 1.09831285e+00 6.37716651e-01 -9.40067828e-01 7.80682087e-01 -3.48495960e-01 -5.72882928e-02 -3.07401836e-01 -1.43000513e-01 -2.08163947e-01 1.02310441e-01 3.24410707e-01 1.47339845e+00 -6.09072447e-02 3.66133064e-01 5.23172379e-01 2.12679386e-01 4.13882993e-02 1.71935990e-01 -2.92418718e-01 4.81240571e-01 6.61786258e-01 1.36375880e+00 -4.95340884e-01 -2.34793231e-01 -3.94018322e-01 1.11013114e+00 6.79797649e-01 1.65715575e-01 -1.17969835e+00 -4.18802649e-01 3.47955734e-01 2.42720529e-01 1.66275486e-01 2.45216861e-02 -1.24286115e-01 -1.03746057e+00 3.06929946e-01 -9.72208977e-01 6.27318501e-01 -5.83193004e-01 -1.23584926e+00 8.31966341e-01 1.34415656e-01 -1.18920410e+00 -1.50521785e-01 -5.23712814e-01 -4.01081264e-01 7.22274125e-01 -8.43334675e-01 -1.69477570e+00 -7.39164472e-01 6.44153476e-01 2.13218927e-01 -2.93790877e-01 7.24136353e-01 4.41791832e-01 -1.03192425e+00 1.09294689e+00 -6.94600284e-01 5.16248763e-01 7.82572091e-01 -9.74902868e-01 3.48334342e-01 7.79967725e-01 -1.64605081e-01 1.00061584e+00 6.69393539e-01 -6.88066125e-01 -1.08932769e+00 -8.08426440e-01 8.89752746e-01 -6.48931682e-01 2.71286637e-01 -7.03922391e-01 -3.89107674e-01 8.60829949e-01 2.14400753e-01 2.70499796e-01 8.80485296e-01 6.75272942e-01 -4.54814732e-01 -9.00178477e-02 -1.03315508e+00 7.23276496e-01 1.44878089e+00 -1.32161677e-01 -4.25649613e-01 1.38870180e-01 4.40333843e-01 -7.76746571e-01 -1.03826046e+00 5.81621826e-01 1.00132227e+00 -1.26428676e+00 1.25861382e+00 -4.63607371e-01 3.96569341e-01 -3.49643290e-01 -2.25405425e-01 -8.81606102e-01 -6.43133521e-01 -7.74436831e-01 -5.84320545e-01 1.37800252e+00 2.46223718e-01 -3.67767394e-01 7.06202805e-01 7.98315704e-01 3.48359138e-01 -1.02979422e+00 -8.03492010e-01 -8.60811234e-01 -3.46732676e-01 -1.01306990e-01 6.84470415e-01 8.12306404e-01 2.16297939e-01 7.42382944e-01 -9.91171658e-01 2.29042172e-01 8.62939179e-01 -2.03012183e-01 1.27241659e+00 -1.29933596e+00 -6.13232374e-01 -1.18374459e-01 -6.51932299e-01 -1.10720861e+00 -1.23735704e-01 -3.34759444e-01 -9.79608446e-02 -1.41789448e+00 6.78673267e-01 -1.92810893e-01 -2.41919458e-01 4.41385835e-01 -5.87095439e-01 3.86882305e-01 4.39645439e-01 2.80099362e-01 -1.26383507e+00 4.67926979e-01 1.25796139e+00 2.06157804e-01 -1.76409155e-01 -1.68424714e-02 -6.99799299e-01 9.90794957e-01 5.87146044e-01 -1.05421461e-01 -2.79557794e-01 -1.51633054e-01 1.94611058e-01 -1.74221188e-01 4.24060792e-01 -1.35007465e+00 6.35875463e-01 1.01169422e-01 7.51944542e-01 -6.26525342e-01 5.31940520e-01 -6.73716068e-01 2.36870602e-01 2.87925601e-01 -1.01026729e-01 1.93201676e-01 3.81308012e-02 6.91836536e-01 -8.16022232e-02 4.28885072e-01 5.82948625e-01 -7.34161362e-02 -5.69804311e-01 2.91948527e-01 2.29368374e-01 2.66229749e-01 1.28519785e+00 -1.40525684e-01 -4.15899515e-01 -4.67289001e-01 -7.02282786e-01 5.85805476e-01 -2.04521809e-02 5.89596689e-01 4.45460707e-01 -1.59580767e+00 -6.75328970e-01 -8.99183527e-02 2.93451786e-01 8.08509812e-02 3.45507771e-01 8.36522520e-01 -3.02791566e-01 3.55473757e-01 -2.25201696e-01 -5.25247097e-01 -1.61653209e+00 3.23006064e-01 2.70161062e-01 -5.18266678e-01 -5.24629295e-01 1.21814561e+00 1.63362622e-01 -4.69109714e-01 4.78895813e-01 4.08776700e-01 -4.37689215e-01 5.79921380e-02 7.26260126e-01 7.77105153e-01 -3.77311677e-01 -1.09039581e+00 -8.94942045e-01 8.28097284e-01 -2.75987834e-01 8.83596689e-02 1.20496595e+00 -1.43154338e-01 4.00399864e-02 3.34906846e-01 8.59240413e-01 6.07148372e-02 -1.10910189e+00 -1.04905702e-01 -2.30634958e-01 -4.99208480e-01 -5.34809351e-01 -8.93643975e-01 -8.92936468e-01 5.34177721e-01 4.62429881e-01 -3.25789154e-02 9.66745615e-01 1.91630423e-02 7.23464966e-01 2.24977240e-01 6.68002963e-01 -1.23042035e+00 2.34731004e-01 3.43736529e-01 9.57021415e-01 -1.26562929e+00 4.92489040e-01 -9.33281660e-01 -9.29066658e-01 8.58555257e-01 1.27161574e+00 -1.65998582e-02 5.22528470e-01 2.52591610e-01 -1.24866851e-01 -3.91721070e-01 -3.90278280e-01 -2.44989485e-01 7.05762804e-01 6.29418075e-01 5.91581404e-01 2.07762808e-01 -3.12448740e-01 1.03864563e+00 -3.81835401e-01 -2.11400334e-02 1.66629199e-02 7.49819398e-01 -1.83439270e-01 -8.75877380e-01 -4.50851053e-01 1.62109330e-01 -2.19309986e-01 1.71397746e-01 -6.59006953e-01 1.33568645e+00 4.93857831e-01 1.05697334e+00 -1.52841225e-01 -9.19326425e-01 8.48267794e-01 -1.89264491e-01 4.16347682e-01 -2.87524134e-01 -7.33475268e-01 -1.21941604e-02 4.16304141e-01 -7.97699213e-01 -4.66188580e-01 -6.24700785e-01 -1.08320618e+00 -6.85401857e-01 -2.72402108e-01 -1.91559624e-02 -8.61174837e-02 9.50008512e-01 3.05586815e-01 5.99193990e-01 2.01312855e-01 -9.66883600e-01 -4.08432968e-02 -1.14337301e+00 -5.25626659e-01 5.65914869e-01 -9.45729241e-02 -1.06569719e+00 -1.06126070e-01 -2.09114611e-01]
[7.190167427062988, -0.757649302482605]
61da4ca4-9a28-43a8-8526-8fc27c1d4f45
momentum-contrastive-voxel-wise
2105.07059
null
https://arxiv.org/abs/2105.07059v4
https://arxiv.org/pdf/2105.07059v4.pdf
Momentum Contrastive Voxel-wise Representation Learning for Semi-supervised Volumetric Medical Image Segmentation
Contrastive learning (CL) aims to learn useful representation without relying on expert annotations in the context of medical image segmentation. Existing approaches mainly contrast a single positive vector (i.e., an augmentation of the same image) against a set of negatives within the entire remainder of the batch by simply mapping all input features into the same constant vector. Despite the impressive empirical performance, those methods have the following shortcomings: (1) it remains a formidable challenge to prevent the collapsing problems to trivial solutions; and (2) we argue that not all voxels within the same image are equally positive since there exist the dissimilar anatomical structures with the same image. In this work, we present a novel Contrastive Voxel-wise Representation Learning (CVRL) method to effectively learn low-level and high-level features by capturing 3D spatial context and rich anatomical information along both the feature and the batch dimensions. Specifically, we first introduce a novel CL strategy to ensure feature diversity promotion among the 3D representation dimensions. We train the framework through bi-level contrastive optimization (i.e., low-level and high-level) on 3D images. Experiments on two benchmark datasets and different labeled settings demonstrate the superiority of our proposed framework. More importantly, we also prove that our method inherits the benefit of hardness-aware property from the standard CL approaches.
['James S. Duncan', 'Lawrence Staib', 'Ruihan Zhao', 'Chenyu You']
2021-05-14
null
null
null
null
['volumetric-medical-image-segmentation']
['medical']
[ 4.35750395e-01 8.09733495e-02 -2.99985617e-01 -4.50514138e-01 -1.11275434e+00 -5.69030225e-01 3.47357780e-01 2.01472774e-01 -3.76186937e-01 6.07894957e-01 1.73349097e-01 -1.97438329e-01 -4.93767083e-01 -4.92691755e-01 -7.47705877e-01 -9.72353280e-01 -2.51891434e-01 2.25559875e-01 1.51356339e-01 -3.00865304e-02 1.28014028e-01 6.70529187e-01 -1.16322196e+00 2.10412666e-01 8.84203196e-01 1.32417500e+00 3.21556777e-01 1.88546404e-01 -5.98552488e-02 6.42756462e-01 -4.36107576e-01 -7.75040612e-02 5.57352364e-01 -3.61468256e-01 -8.43493760e-01 4.16241497e-01 6.05353355e-01 -3.28178331e-02 -1.60518765e-01 1.09140933e+00 4.40744638e-01 7.36549944e-02 7.10954368e-01 -1.17239630e+00 -7.09217608e-01 2.98291028e-01 -1.14757276e+00 4.13462788e-01 -1.07025504e-01 1.49893522e-01 1.00230110e+00 -9.91635203e-01 5.69989622e-01 9.85181570e-01 3.94457847e-01 4.51921910e-01 -1.31017292e+00 -4.98905033e-01 5.62179387e-01 -1.14331529e-01 -1.35070825e+00 2.69226059e-02 1.04381168e+00 -5.60402691e-01 5.12244344e-01 5.02344251e-01 6.74423635e-01 6.50555849e-01 3.22426468e-01 8.20868552e-01 1.38471031e+00 -1.35706097e-01 2.47322083e-01 1.09193489e-01 3.40001762e-01 8.40481758e-01 2.21694142e-01 3.47312703e-03 -2.14511573e-01 -3.26488137e-01 8.65352452e-01 1.99441269e-01 -3.76317024e-01 -8.78245056e-01 -1.28315473e+00 9.05166566e-01 8.22206020e-01 4.37654078e-01 -5.31622112e-01 -2.21972298e-02 4.49447513e-01 8.26713070e-03 3.52921039e-01 4.45467740e-01 -4.38715309e-01 3.97863716e-01 -6.94868028e-01 1.66087866e-01 2.11443454e-01 7.59503543e-01 7.60936260e-01 -9.24745798e-02 -3.78373235e-01 7.67774761e-01 3.25919427e-02 1.74665049e-01 6.06027365e-01 -5.08380473e-01 4.82186943e-01 7.00412452e-01 -2.77123958e-01 -1.21030807e+00 -6.26574814e-01 -9.02109683e-01 -1.22185576e+00 1.02912903e-01 1.92703083e-01 9.04962048e-03 -1.00938368e+00 1.83604074e+00 4.17133719e-01 2.67743409e-01 -2.21689105e-01 1.06332898e+00 9.78036344e-01 4.58479166e-01 3.76660712e-02 -4.46098238e-01 1.11237454e+00 -1.05398226e+00 -5.30284524e-01 -1.80365577e-01 5.22549272e-01 -3.82268041e-01 1.11196005e+00 2.43082225e-01 -1.23098862e+00 -4.14293349e-01 -1.09555948e+00 6.52729273e-02 -1.37192130e-01 2.18882896e-02 9.78813469e-01 5.23343563e-01 -8.77172291e-01 2.45191291e-01 -6.94245994e-01 2.46313840e-01 7.52059460e-01 5.56061685e-01 -5.86346209e-01 -1.31096721e-01 -7.10867465e-01 5.18930197e-01 1.56938314e-01 2.20176175e-01 -7.24159360e-01 -1.06116819e+00 -9.68678534e-01 -1.11410394e-01 4.85059917e-01 -5.90348005e-01 7.66178310e-01 -9.54367459e-01 -9.11728621e-01 1.02017152e+00 -1.18093595e-01 -1.01907313e-01 6.83057904e-01 -4.92226072e-02 -1.12371787e-03 1.70803145e-01 3.04364920e-01 5.55535316e-01 7.63538897e-01 -1.62163949e+00 -6.17956638e-01 -7.53245533e-01 1.77937329e-01 3.28515619e-01 -4.80594307e-01 -4.44938064e-01 -5.66078842e-01 -8.81570697e-01 6.36267364e-01 -6.85087681e-01 -6.65417075e-01 3.43686700e-01 -5.64461231e-01 -8.15588236e-02 4.45815325e-01 -3.94063562e-01 1.12039101e+00 -2.23263860e+00 4.14535075e-01 5.24531901e-01 6.36150301e-01 -2.37277523e-02 -1.14986062e-01 -1.62870258e-01 -3.31592083e-01 1.72761440e-01 -6.25899673e-01 -1.68303862e-01 -3.47583234e-01 2.95781374e-01 -1.07113913e-01 7.74146616e-01 4.31643128e-01 9.92184281e-01 -1.14526176e+00 -7.19794035e-01 2.60371327e-01 4.18109804e-01 -6.89691246e-01 -5.69150317e-03 2.99262017e-01 7.71140933e-01 -6.31388426e-01 6.38680875e-01 8.92921448e-01 -4.53716755e-01 1.59532987e-02 -4.06629592e-01 1.00208826e-01 -5.09822778e-02 -1.26924706e+00 1.75025535e+00 -4.16334897e-01 -3.36154141e-02 -5.85661195e-02 -1.42472589e+00 6.65148437e-01 9.89976339e-04 8.92098308e-01 -8.72185588e-01 8.86716917e-02 3.17523032e-01 7.43396254e-03 -3.79538536e-01 1.87405869e-01 -2.76013315e-01 -1.46833822e-01 4.12745416e-01 4.13563885e-02 3.36428881e-02 -1.45954669e-01 3.42392921e-02 8.89364719e-01 -2.38437861e-01 4.67554718e-01 -4.32555079e-01 6.62125468e-01 -2.97835648e-01 8.00619304e-01 7.97165453e-01 -4.41402316e-01 7.08691716e-01 6.25446200e-01 -3.79241675e-01 -7.16057599e-01 -1.28993261e+00 -3.97186249e-01 7.80359864e-01 4.25695628e-01 -1.76165476e-01 -3.96371394e-01 -1.07110476e+00 3.13988701e-02 1.19223692e-01 -1.07171345e+00 -1.64550170e-01 -7.89907753e-01 -9.77170050e-01 1.20435432e-01 7.53090024e-01 2.65179932e-01 -6.40206754e-01 -5.33501446e-01 1.91099960e-02 -1.76041864e-03 -9.06647265e-01 -5.80002129e-01 3.72846574e-01 -9.68161821e-01 -9.05575395e-01 -8.58301163e-01 -1.20263588e+00 1.07220089e+00 5.34753680e-01 1.04278469e+00 1.58128187e-01 -4.88987595e-01 2.69553363e-01 -1.67217538e-01 -7.63751268e-02 1.13956705e-01 -4.13931645e-02 -3.26809824e-01 7.23624602e-02 -1.46570265e-01 -4.45006728e-01 -8.11315179e-01 1.88318178e-01 -8.68944108e-01 3.67134996e-02 6.43215895e-01 1.31419337e+00 1.34655499e+00 3.15513313e-02 6.52491331e-01 -9.53347623e-01 4.77568299e-01 -6.36591673e-01 -2.53757656e-01 3.84589523e-01 -3.84447306e-01 -2.74683423e-02 5.20660818e-01 -4.41633135e-01 -5.54563880e-01 3.05583954e-01 -5.70017584e-02 -4.59260732e-01 1.30297929e-01 4.26430643e-01 -1.32916734e-01 -1.93840772e-01 5.49997032e-01 4.11598355e-01 -3.87917086e-02 -1.91726789e-01 3.67517710e-01 1.94777846e-01 5.23720264e-01 -6.41547441e-01 8.77453148e-01 7.41129696e-01 2.42272511e-01 -6.15739405e-01 -9.96224463e-01 -5.29371142e-01 -6.93885922e-01 -1.20831214e-01 6.67241454e-01 -7.21134722e-01 -5.56743622e-01 1.34768844e-01 -6.05678022e-01 -4.56704982e-02 -6.32834971e-01 5.04242003e-01 -5.54351568e-01 5.39439261e-01 -3.59309494e-01 -6.10279202e-01 -4.60286409e-01 -1.42141461e+00 1.08859062e+00 1.74293235e-01 1.49816155e-01 -7.98663318e-01 -9.31601226e-02 2.11731866e-01 1.59986645e-01 5.46098173e-01 1.18098545e+00 -4.78940248e-01 -4.70154524e-01 -9.95901451e-02 -3.52013648e-01 3.58905196e-01 4.18497384e-01 -3.51863474e-01 -7.92569578e-01 -3.82440627e-01 1.92779765e-01 -4.57586795e-01 8.54060531e-01 6.75744653e-01 1.69157684e+00 -2.60190964e-01 -2.50480324e-01 7.58733094e-01 1.45755374e+00 -8.48949794e-03 3.49142164e-01 1.81851104e-01 7.74805486e-01 6.36996746e-01 5.80965877e-01 4.30249065e-01 2.24821523e-01 4.92204696e-01 6.12041712e-01 -5.38262129e-01 1.66771887e-03 -2.60122996e-02 -2.46869147e-01 7.75781333e-01 3.71072479e-02 1.92878023e-01 -6.13689542e-01 6.20571852e-01 -1.69651473e+00 -6.86129510e-01 7.67104561e-03 2.29775381e+00 8.69402289e-01 1.18633449e-01 6.28218874e-02 2.07408115e-01 5.69425821e-01 2.19949931e-01 -7.83460379e-01 -3.81033123e-03 -1.76445201e-01 3.45692486e-01 3.68117779e-01 3.18272561e-01 -1.30680537e+00 4.95731086e-01 5.42934465e+00 8.30379963e-01 -1.30307972e+00 1.48844346e-01 1.02276874e+00 -2.78680712e-01 -5.39371610e-01 -4.48038697e-01 -4.59472418e-01 3.69436979e-01 1.05935149e-02 -1.32055745e-01 2.00769588e-01 7.50109434e-01 -1.48754627e-01 2.31691554e-01 -1.05359685e+00 1.15867293e+00 1.82819679e-01 -1.41393006e+00 1.03255741e-01 8.17165375e-02 8.72573078e-01 -1.43563062e-01 4.55228984e-01 2.61966497e-01 -2.39034757e-01 -1.22223961e+00 6.96518779e-01 1.82053059e-01 8.02972615e-01 -8.10227871e-01 6.64530277e-01 1.74055845e-01 -1.28714001e+00 -1.22227646e-01 -3.68258506e-01 3.98404717e-01 -7.52656832e-02 4.33538318e-01 -4.59093451e-01 7.31528103e-01 6.59069479e-01 7.71829844e-01 -6.27532899e-01 1.14052141e+00 1.24245636e-01 2.40710482e-01 -1.36194184e-01 2.61680841e-01 6.19532406e-01 -1.02357864e-01 4.13059711e-01 1.03718174e+00 -2.13893838e-02 4.27076407e-02 4.86091524e-01 7.55513072e-01 -6.69889823e-02 5.26577055e-01 -5.38263023e-01 3.69574994e-01 1.84643269e-01 1.40085888e+00 -7.55290329e-01 -8.94765854e-02 -5.88188589e-01 7.91749835e-01 5.08637249e-01 1.40868008e-01 -7.43099809e-01 1.96629111e-02 4.49902982e-01 -4.46608514e-02 2.72417992e-01 1.16653524e-01 -7.45840788e-01 -9.88092363e-01 4.27412808e-01 -7.32349753e-01 6.02166772e-01 -1.95563272e-01 -1.64074337e+00 7.11868584e-01 -1.06967561e-01 -1.37034941e+00 -1.65400878e-02 -4.57746983e-01 -3.40181202e-01 8.72624695e-01 -1.67313778e+00 -1.24756098e+00 -3.66942644e-01 8.09498608e-01 3.95184815e-01 1.11677930e-01 6.32580578e-01 3.48127842e-01 -4.98656452e-01 9.22135115e-01 1.17224464e-02 1.68709859e-01 3.92033190e-01 -1.36253774e+00 -1.18542269e-01 4.88613218e-01 -1.60461560e-01 6.43335462e-01 2.37155423e-01 -3.89226198e-01 -1.33189034e+00 -1.14088976e+00 4.57028508e-01 -1.11911081e-01 4.69261259e-01 -2.83256620e-01 -9.79321718e-01 5.55645585e-01 -3.27187151e-01 5.79540908e-01 8.42054188e-01 -9.06816497e-03 -2.42030755e-01 -1.92548767e-01 -1.35640597e+00 6.09118760e-01 1.03886521e+00 -2.48834401e-01 -5.37651181e-01 5.01077354e-01 6.03012860e-01 -4.57062095e-01 -1.01204991e+00 7.39551127e-01 5.73106647e-01 -6.92291737e-01 1.16641200e+00 -8.05463433e-01 4.78143752e-01 -2.11627454e-01 -4.02447253e-01 -1.04990280e+00 -3.88515234e-01 -2.07622826e-01 -9.12199318e-02 8.85274827e-01 3.34517598e-01 -6.53463602e-01 7.08048522e-01 4.87378597e-01 -3.09633404e-01 -1.46409035e+00 -1.16183531e+00 -8.39662611e-01 4.95205730e-01 -2.54726112e-01 4.96941239e-01 1.12073040e+00 -1.75782308e-01 2.62364447e-01 -3.44312191e-01 2.70408362e-01 6.52465522e-01 3.95219237e-01 5.00433683e-01 -1.17458403e+00 -4.39357489e-01 -5.86898267e-01 -6.83776498e-01 -1.06828630e+00 4.58894856e-02 -1.24763048e+00 7.16022477e-02 -1.43038809e+00 7.47155845e-01 -7.85846889e-01 -7.24377394e-01 4.80133504e-01 -4.20218945e-01 3.85511994e-01 1.80459600e-02 1.88923180e-01 -3.89170170e-01 5.23200572e-01 1.82906508e+00 -3.06690961e-01 -2.17408821e-01 -1.60830274e-01 -1.10621655e+00 6.31757498e-01 5.86462319e-01 -3.61940295e-01 -6.49965584e-01 -3.79221559e-01 1.63483676e-02 -3.16033885e-02 5.74886501e-01 -6.64660037e-01 -9.00785401e-02 -1.61143050e-01 6.06877685e-01 -5.59302986e-01 2.23909944e-01 -8.57180297e-01 -3.20302397e-01 5.51566482e-01 -5.14600754e-01 -8.82279128e-02 9.95584950e-02 6.78529024e-01 -1.35016784e-01 -9.46531817e-02 8.83607745e-01 -1.51286960e-01 -5.57197809e-01 7.38681674e-01 2.19645768e-01 2.64658272e-01 1.18876171e+00 -2.41607860e-01 -2.18423158e-01 7.48020262e-02 -7.09462345e-01 2.77737319e-01 2.65891701e-01 2.52202719e-01 7.22178757e-01 -1.28728950e+00 -7.29418635e-01 3.57876301e-01 3.86705101e-01 2.98934758e-01 5.57753325e-01 1.10693133e+00 -2.81984091e-01 2.62827218e-01 -3.03638369e-01 -9.28956091e-01 -1.03339851e+00 6.10948920e-01 5.16932487e-01 -5.12528658e-01 -9.31401610e-01 9.55799460e-01 1.00274587e+00 -5.01286149e-01 4.31378335e-01 -3.13518256e-01 -3.18963170e-01 -6.83527216e-02 6.08431041e-01 -9.40625835e-03 2.49172032e-01 -7.63744891e-01 -5.33014894e-01 7.02621400e-01 -4.02455777e-01 1.58271953e-01 1.51721656e+00 6.12520464e-02 -5.72594702e-02 3.79711032e-01 1.52435231e+00 -8.86854678e-02 -1.21514022e+00 -3.68628591e-01 -2.20875338e-01 -6.40922189e-01 2.09906578e-01 -6.82637036e-01 -1.47503042e+00 8.38050306e-01 9.31252420e-01 -9.67306495e-02 1.27332520e+00 2.61709481e-01 7.21950114e-01 5.51833734e-02 3.73328328e-01 -8.77885282e-01 1.31930187e-01 -5.68295158e-02 1.07315183e+00 -1.35008335e+00 1.67674318e-01 -6.68646991e-01 -7.29603410e-01 8.00920606e-01 6.30889654e-01 -3.15706819e-01 7.64433265e-01 8.28274712e-02 -1.61549628e-01 -4.56925720e-01 -4.34537530e-01 -1.47299856e-01 5.44896305e-01 5.57957411e-01 3.71592283e-01 1.66275978e-01 -5.52351058e-01 7.80367017e-01 5.29111065e-02 -2.20815152e-01 1.69687226e-01 1.06624866e+00 -1.20567530e-01 -6.77875280e-01 -1.39041796e-01 8.01216900e-01 -4.81969595e-01 -2.21605483e-03 -1.41277239e-01 8.39418292e-01 9.11208168e-02 6.11771882e-01 -9.30822734e-03 -2.27716506e-01 4.56079811e-01 -5.15136957e-01 6.50518417e-01 -7.03185618e-01 -6.37622058e-01 1.69550017e-01 -5.31260848e-01 -6.25727296e-01 -2.04221472e-01 -6.39990628e-01 -1.31044948e+00 1.72313929e-01 -2.45430022e-01 -1.10755831e-01 2.79384881e-01 7.67762303e-01 1.32684872e-01 6.65477812e-01 9.50128257e-01 -7.27573037e-01 -6.90414906e-01 -4.88080174e-01 -7.49763310e-01 5.68133712e-01 4.70132947e-01 -8.48614872e-01 -3.60673815e-01 -2.30401427e-01]
[14.723199844360352, -2.134382963180542]
49e8cda7-c125-4660-8841-30fb346a57d4
leveraging-instance-image-and-dataset-level
null
null
https://ieeexplore.ieee.org/abstract/document/9193980
http://mftp.mmcheng.net/Papers/21PAMI_InsImgDatasetWSIS.pdf
Leveraging Instance-, Image- and Dataset-Level Information for Weakly Supervised Instance Segmentation
Weakly supervised semantic instance segmentation with only image-level supervision, instead of relying on expensive pixel wise masks or bounding box annotations, is an important problem to alleviate the data-hungry nature of deep learning. In this paper, we tackle this challenging problem by aggregating the image-level information of all training images into a large knowledge graph and exploiting semantic relationships from this graph. Specifically, our effort starts with some generic segment-based object proposals (SOP) without category priors. We propose a multiple instance learning (MIL) framework, which can be trained in an end-to-end manner using training images with image-level labels. For each proposal, this MIL framework can simultaneously compute probability distributions and category-aware semantic features, with which we can formulate a large undirected graph. The category of background is also included in this graph to remove the massive noisy object proposals. An optimal multi-way cut of this graph can thus assign a reliable category label to each proposal. The denoised SOP with assigned category labels can be viewed as pseudo instance segmentation of training images, which are used to train fully supervised models. The proposed approach achieves state-of-the-art performance for both weakly supervised instance segmentation and semantic segmentation.
['Ming-Ming Cheng', 'Yu Qiu', 'Yu-Jun Shi', 'Pei-Song Wen', 'Yu-Huan Wu', 'Yun Liu']
2020-09-10
null
null
null
null
['weakly-supervised-instance-segmentation', 'image-level-supervised-instance-segmentation']
['computer-vision', 'computer-vision']
[ 6.59197390e-01 6.15674615e-01 -2.59329826e-01 -6.19983077e-01 -1.20981276e+00 -4.65152562e-01 3.88457686e-01 3.15683693e-01 -5.11869788e-01 4.95314062e-01 -3.72633189e-01 8.39924961e-02 -6.49713427e-02 -8.81625295e-01 -1.04066789e+00 -8.90886486e-01 3.46567661e-01 7.65265584e-01 5.09653211e-01 2.55570233e-01 1.09573089e-01 1.45456135e-01 -1.47637737e+00 3.04377764e-01 9.72800553e-01 1.24218953e+00 5.52354932e-01 3.64117086e-01 -3.85477871e-01 6.49730802e-01 -4.45486426e-01 -3.93350840e-01 1.71140462e-01 -2.75082201e-01 -9.39500213e-01 7.97684968e-01 4.92217660e-01 5.93931042e-02 2.91903913e-01 1.33196592e+00 9.50169563e-02 1.95376515e-01 6.40824676e-01 -1.08759308e+00 -1.23543769e-01 5.37052214e-01 -8.67700279e-01 -1.54545829e-01 -2.82776922e-01 8.21434781e-02 1.20340896e+00 -8.66414368e-01 6.17211282e-01 1.16009820e+00 3.51866126e-01 3.03414077e-01 -1.40617025e+00 -3.47475678e-01 5.34670651e-01 8.51454586e-02 -1.27100778e+00 -1.15477115e-01 1.11933804e+00 -3.53974730e-01 2.90978283e-01 5.17876297e-02 5.73801696e-01 6.59948409e-01 -4.25737053e-01 1.02872729e+00 1.14374745e+00 -3.16230267e-01 5.79284847e-01 2.39641771e-01 3.47401172e-01 8.29503834e-01 1.57710508e-01 -5.54989874e-01 -1.91942155e-01 2.44326964e-01 6.80786014e-01 1.55517176e-01 -1.46411240e-01 -7.20061779e-01 -9.59640026e-01 7.02771068e-01 8.41482341e-01 2.17138126e-01 -4.76945817e-01 3.02119344e-01 1.08276933e-01 -3.20976257e-01 7.58438110e-01 -9.98234004e-02 -6.08778179e-01 5.18484175e-01 -1.08158517e+00 -6.45562112e-02 6.70403540e-01 7.60087609e-01 1.30655563e+00 -2.30519116e-01 -1.34348869e-01 1.00648928e+00 5.06878376e-01 2.01288342e-01 -1.04119144e-01 -1.06557345e+00 4.53690618e-01 8.24477673e-01 7.32036382e-02 -7.24531949e-01 -2.52802730e-01 -8.67379963e-01 -6.23922706e-01 3.66759002e-02 6.15048289e-01 3.90559845e-02 -1.52008438e+00 1.64040208e+00 8.01287234e-01 4.84776109e-01 -1.07950695e-01 1.00156248e+00 7.44640827e-01 6.74744129e-01 3.58206242e-01 -1.18802696e-01 1.44730055e+00 -1.27539134e+00 -4.16788548e-01 -6.42292619e-01 3.73792827e-01 -5.52292109e-01 8.24542105e-01 3.04753631e-01 -9.51822698e-01 -5.24952650e-01 -7.31718957e-01 -9.27209258e-02 -1.44277349e-01 2.07091182e-01 3.36085230e-01 3.89447749e-01 -8.50936532e-01 3.84177148e-01 -9.04289424e-01 -7.32043013e-02 9.88459051e-01 3.14501107e-01 -2.23369107e-01 -3.91583353e-01 -6.73701167e-01 4.89897698e-01 7.59626806e-01 1.86594427e-01 -1.10518897e+00 -6.16449475e-01 -1.04006469e+00 9.25731286e-03 1.00868833e+00 -5.37906706e-01 8.75037670e-01 -1.19281816e+00 -1.05484867e+00 1.14250445e+00 -2.39125803e-01 -3.62158120e-01 4.40143704e-01 -1.95357483e-03 2.70093292e-01 3.93711448e-01 4.80316758e-01 7.97473073e-01 1.32331955e+00 -1.91061461e+00 -8.80426884e-01 -6.09608829e-01 -1.78089608e-02 2.53374904e-01 1.17072843e-01 -3.36587548e-01 -1.04621077e+00 -5.50831676e-01 5.61619520e-01 -8.21412623e-01 -5.16776741e-01 -8.35847631e-02 -7.18326688e-01 -2.52666801e-01 7.60748267e-01 -7.89537251e-01 6.17307067e-01 -1.98519778e+00 3.98530662e-01 3.81589204e-01 1.85602576e-01 2.81672906e-02 -8.00506100e-02 -2.56074756e-01 2.76144832e-01 1.18829170e-03 -8.54720235e-01 -7.33403504e-01 -8.20460171e-02 4.87088948e-01 7.25485831e-02 5.02629101e-01 4.17948991e-01 9.85384762e-01 -9.72104132e-01 -7.12405622e-01 4.32941258e-01 4.34633166e-01 -5.23382545e-01 2.97684878e-01 -7.66230762e-01 7.13458359e-01 -6.03058457e-01 7.22932816e-01 8.58237565e-01 -3.26633722e-01 1.22798860e-01 -3.36044610e-01 2.05442488e-01 -8.13485160e-02 -1.30621481e+00 1.90392601e+00 -4.98588711e-01 5.07552624e-02 3.80755544e-01 -1.50771368e+00 7.68430471e-01 -8.73209536e-02 3.95057827e-01 -4.81663555e-01 2.02727050e-01 2.61690259e-01 -4.13715065e-01 -2.71285802e-01 1.30585087e-02 -2.59766519e-01 -2.51848221e-01 3.63554768e-02 3.50314051e-01 -3.78582865e-01 1.62414372e-01 2.24915624e-01 6.93851113e-01 3.36463928e-01 6.87339753e-02 -1.74065024e-01 6.83347404e-01 7.39850253e-02 8.30462813e-01 7.65112519e-01 6.47921697e-04 7.32658088e-01 5.19499600e-01 -1.62989676e-01 -8.30893457e-01 -1.15490770e+00 -1.40840456e-01 9.52275693e-01 4.72341239e-01 9.10518877e-03 -1.14767957e+00 -1.02764368e+00 -7.64856637e-02 5.26185334e-01 -4.72870976e-01 9.03542638e-02 -5.08696020e-01 -7.17657208e-01 -2.54265070e-02 3.80799741e-01 6.23352289e-01 -1.05864477e+00 -2.79946417e-01 3.47639918e-01 -3.68595958e-01 -1.31932211e+00 -2.34298155e-01 5.73929667e-01 -8.46551836e-01 -1.17168128e+00 -7.22062647e-01 -1.05666852e+00 1.02833331e+00 1.67787611e-01 1.14799142e+00 1.11774273e-01 -2.24772662e-01 2.43406907e-01 -2.63451666e-01 -5.51921539e-02 -2.10293174e-01 -1.56625599e-01 -4.86798257e-01 6.74849570e-01 -3.43464650e-02 -3.13860446e-01 -6.74411058e-01 2.82961339e-01 -1.04161346e+00 1.18881531e-01 8.51802886e-01 8.34061503e-01 1.33930612e+00 2.09591255e-01 5.86307883e-01 -1.25552368e+00 -2.22058609e-01 -4.54874545e-01 -7.10118473e-01 2.40081638e-01 -8.83187652e-02 5.68470620e-02 4.83951837e-01 -7.39653111e-02 -1.26582742e+00 4.63141948e-01 -1.85011879e-01 -3.89563769e-01 -4.93884623e-01 4.53815728e-01 -6.70501173e-01 -1.47737134e-02 1.52417585e-01 1.35901883e-01 -2.32133299e-01 -6.73971832e-01 6.79256618e-01 5.15036702e-01 7.51437306e-01 -6.41717613e-01 7.74615884e-01 7.25301325e-01 -2.36548409e-02 -7.91216612e-01 -1.55001688e+00 -8.72112453e-01 -8.15232873e-01 -2.63194263e-01 1.23905849e+00 -1.05342746e+00 -1.11262709e-01 4.50683475e-01 -9.91305292e-01 -5.73379159e-01 -4.39613163e-01 1.32775247e-01 -6.04531884e-01 2.41279855e-01 -4.32754964e-01 -6.95878446e-01 -8.05008598e-03 -1.27152812e+00 1.53895533e+00 3.26058984e-01 3.55981797e-01 -9.24414158e-01 -4.04979527e-01 1.00025046e+00 -1.67770237e-01 4.86152381e-01 7.53993750e-01 -5.21774113e-01 -9.02599454e-01 -2.87167817e-01 -6.07976019e-01 8.29881191e-01 -6.81685656e-02 -4.14318889e-01 -1.14000428e+00 -1.22694872e-01 9.50817689e-02 -3.17979127e-01 1.26607013e+00 5.91733992e-01 1.39261496e+00 -2.88530350e-01 -4.32676017e-01 5.77317536e-01 1.71604884e+00 -3.33100289e-01 2.61006594e-01 -5.15284436e-03 1.23294425e+00 8.60135019e-01 7.84668684e-01 2.24232778e-01 4.47151184e-01 5.96482456e-01 7.45151818e-01 -2.82426536e-01 -2.19427124e-01 -2.13799044e-01 -2.25484576e-02 3.65194827e-01 1.64883614e-01 -2.75870055e-01 -6.92746222e-01 7.70561934e-01 -1.94651175e+00 -4.65627104e-01 -3.57506365e-01 2.04178953e+00 7.32802749e-01 3.01129431e-01 -5.28411455e-02 1.86146736e-01 9.04468060e-01 2.59530008e-01 -6.62880540e-01 1.25476137e-01 9.80230644e-02 2.22137794e-01 4.74830985e-01 5.24995506e-01 -1.37633932e+00 1.09819114e+00 3.91802406e+00 1.12440169e+00 -7.42002726e-01 3.98769379e-01 1.03058767e+00 2.71839172e-01 -1.93899885e-01 1.81594759e-01 -7.24851310e-01 5.68675876e-01 4.44283456e-01 3.85380864e-01 8.92717242e-02 9.49621260e-01 1.53427556e-01 -4.35203642e-01 -1.02522540e+00 7.71009028e-01 3.31211127e-02 -1.15245092e+00 1.57061685e-02 1.02398628e-02 1.05015659e+00 8.17936379e-03 -1.12112030e-01 1.76780373e-01 2.43359312e-01 -6.51878655e-01 9.16526437e-01 2.66554266e-01 5.66947937e-01 -8.01111162e-01 7.70654976e-01 5.30286551e-01 -1.25919986e+00 -3.54354605e-02 -3.00647765e-01 3.27328771e-01 2.71783561e-01 1.14467633e+00 -5.21768570e-01 6.54173315e-01 5.89832425e-01 7.49099910e-01 -3.90890867e-01 9.36535060e-01 -5.08559167e-01 7.61633515e-01 -5.07976055e-01 4.87151027e-01 4.88594949e-01 -3.72886419e-01 3.18948478e-01 1.01109672e+00 -1.24706313e-01 6.29870370e-02 7.11351156e-01 1.01820683e+00 -4.10683334e-01 -8.77612606e-02 -8.56400132e-02 3.32986653e-01 6.63187355e-02 1.56565714e+00 -1.45277882e+00 -5.52989781e-01 -3.74551058e-01 1.22845852e+00 4.80407476e-01 4.34290171e-01 -7.11116433e-01 9.99230072e-02 3.63908708e-01 6.76970137e-03 5.26876390e-01 9.66173112e-02 -5.18770874e-01 -1.05649030e+00 1.66923061e-01 -2.82454699e-01 4.52283859e-01 -5.53884149e-01 -1.25434005e+00 1.41319692e-01 -1.30588472e-01 -8.80901039e-01 2.60808676e-01 -2.74113715e-01 -6.31069541e-01 7.20590115e-01 -1.85984528e+00 -1.48015666e+00 -5.35401106e-01 4.62246835e-01 7.86128700e-01 5.19833446e-01 2.92629123e-01 3.27307016e-01 -7.03684092e-01 1.14603303e-01 -8.39132518e-02 4.47496735e-02 3.06359380e-01 -1.57901132e+00 -8.68728086e-02 9.08496380e-01 2.81144589e-01 1.61978200e-01 5.27136385e-01 -7.14663029e-01 -8.78154457e-01 -1.61652803e+00 5.02644122e-01 -4.16198075e-01 4.15145189e-01 -3.94548297e-01 -1.11054277e+00 4.05426085e-01 -3.30879062e-01 5.57981849e-01 1.47344589e-01 -2.65177339e-01 -1.18155301e-01 -9.59482044e-02 -1.23711050e+00 2.27413327e-01 1.02021301e+00 -2.61054128e-01 -4.45296526e-01 6.03488088e-01 8.45669150e-01 -3.41992378e-01 -6.14579737e-01 4.18014497e-01 -8.08492452e-02 -6.76935375e-01 1.10717022e+00 -1.85079515e-01 3.89616191e-01 -5.88407218e-01 9.47900042e-02 -1.09623170e+00 5.54708950e-02 -1.35557979e-01 1.67563841e-01 1.40899491e+00 3.32892805e-01 -1.81495473e-01 9.84342813e-01 3.93059105e-01 -2.92431593e-01 -7.25950480e-01 -8.92444372e-01 -5.53029895e-01 -1.16170868e-01 -6.84159577e-01 1.51851729e-01 6.04570150e-01 -6.71485066e-01 2.93164998e-01 -1.17709726e-01 4.45519418e-01 1.34219897e+00 3.12904298e-01 5.63084662e-01 -1.17542982e+00 -3.59328300e-01 -2.21813649e-01 -4.22069103e-01 -8.76850307e-01 3.43846768e-01 -1.05989361e+00 4.97179270e-01 -1.85634172e+00 3.92182261e-01 -7.15711534e-01 -3.53128582e-01 5.07068336e-01 -4.66227293e-01 7.41074443e-01 8.20417404e-02 1.18447199e-01 -1.07056308e+00 4.81812328e-01 1.22487080e+00 -3.78463507e-01 -9.80036929e-02 3.72529738e-02 -5.44770777e-01 9.67804790e-01 6.16295993e-01 -7.06192136e-01 -3.20083380e-01 -2.66903758e-01 -2.34140426e-01 -7.60717466e-02 8.31861317e-01 -7.65045643e-01 9.91062671e-02 -2.25134119e-02 2.69199997e-01 -5.20198584e-01 2.35903218e-01 -8.83972764e-01 -1.16805933e-01 2.25249961e-01 -3.05057943e-01 -1.10416234e+00 -1.49254665e-01 1.00609815e+00 -2.91161150e-01 -2.85378605e-01 1.13398099e+00 -3.94631058e-01 -9.36586022e-01 4.25254196e-01 1.74223408e-01 2.07348615e-01 1.14973938e+00 -4.97697480e-02 1.26710236e-01 9.78436098e-02 -1.09360015e+00 4.52298373e-01 5.10213554e-01 1.72439784e-01 4.70942110e-01 -8.10835063e-01 -5.16080439e-01 2.88003776e-02 1.87567919e-01 9.52072263e-01 5.13648510e-01 7.04299688e-01 -2.92827308e-01 5.69474287e-02 1.63745627e-01 -1.03364480e+00 -1.05502713e+00 4.60418701e-01 1.85244188e-01 -3.48332852e-01 -6.17757618e-01 1.14727712e+00 6.58702016e-01 -4.84800100e-01 2.94635475e-01 -3.29371721e-01 -2.00209916e-01 2.37444967e-01 1.75390705e-01 -6.13442697e-02 1.34953082e-01 -8.39473248e-01 -4.38073009e-01 7.15143263e-01 4.69822483e-03 1.48269171e-02 1.41407228e+00 -2.46685803e-01 -2.21091151e-01 2.33013272e-01 1.37500072e+00 -3.14711124e-01 -1.60984254e+00 -4.06205922e-01 1.69790328e-01 -3.63566935e-01 3.47790748e-01 -7.88631201e-01 -1.45809281e+00 8.15806329e-01 4.13404524e-01 -6.05519116e-02 1.02795517e+00 5.85681200e-01 8.22678089e-01 -6.11004718e-02 4.80748147e-01 -1.14169157e+00 2.23973051e-01 7.58837610e-02 4.74969000e-01 -1.56409109e+00 -1.95790917e-01 -8.66478801e-01 -5.66757321e-01 6.98706567e-01 4.42206413e-01 -2.73366868e-01 5.29792070e-01 -8.68914276e-02 -8.23453516e-02 -2.65641481e-01 -2.28689447e-01 -7.86891460e-01 5.34514308e-01 5.43905437e-01 -1.30552739e-01 2.13851407e-01 -1.21119402e-01 7.09908962e-01 3.90139371e-01 -2.55266219e-01 2.19129279e-01 7.22869515e-01 -6.89104319e-01 -1.01834500e+00 -4.24346119e-01 5.59251130e-01 -4.02470261e-01 9.02248994e-02 -1.65775508e-01 3.75693113e-01 5.39891422e-01 9.41168129e-01 5.83863668e-02 2.37104390e-02 2.26046503e-01 6.23419248e-02 3.65465820e-01 -1.08944309e+00 -2.47100890e-01 3.60870987e-01 3.55134573e-04 -6.37693763e-01 -5.25052249e-01 -6.04321241e-01 -1.56213045e+00 4.31231558e-01 -4.26773369e-01 -3.37526165e-02 8.16198647e-01 1.20632839e+00 -5.58737852e-02 6.92266941e-01 4.90113020e-01 -1.20365953e+00 -9.35099721e-02 -6.20904267e-01 -6.54290020e-01 7.55424142e-01 3.18363428e-01 -6.99160814e-01 -5.16640365e-01 3.66466701e-01]
[9.503747940063477, 0.5350584983825684]
ed1d1dc9-5e8a-474d-88d3-652bdf2e5dc0
self-training-for-domain-adaptive-scene-text
2005.11487
null
https://arxiv.org/abs/2005.11487v1
https://arxiv.org/pdf/2005.11487v1.pdf
Self-Training for Domain Adaptive Scene Text Detection
Though deep learning based scene text detection has achieved great progress, well-trained detectors suffer from severe performance degradation for different domains. In general, a tremendous amount of data is indispensable to train the detector in the target domain. However, data collection and annotation are expensive and time-consuming. To address this problem, we propose a self-training framework to automatically mine hard examples with pseudo-labels from unannotated videos or images. To reduce the noise of hard examples, a novel text mining module is implemented based on the fusion of detection and tracking results. Then, an image-to-video generation method is designed for the tasks that videos are unavailable and only images can be used. Experimental results on standard benchmarks, including ICDAR2015, MSRA-TD500, ICDAR2017 MLT, demonstrate the effectiveness of our self-training method. The simple Mask R-CNN adapted with self-training and fine-tuned on real data can achieve comparable or even superior results with the state-of-the-art methods.
['Dongbao Yang', 'Yudi Chen', 'Yu Zhou', 'Fei Yang', 'Wei Wang', 'Weiping Wang']
2020-05-23
null
null
null
null
['scene-text-detection', 'image-to-video']
['computer-vision', 'computer-vision']
[ 2.03395605e-01 -3.29904974e-01 -9.25492197e-02 -4.07455087e-01 -6.31019890e-01 -1.00816861e-01 6.18413508e-01 -1.56429306e-01 -5.55582285e-01 3.74545336e-01 -2.22707078e-01 1.34101659e-02 5.14324307e-01 -7.41745472e-01 -6.72452688e-01 -5.90956211e-01 4.55029517e-01 5.04463613e-01 5.56361139e-01 1.51551530e-01 4.33841087e-02 -1.12783864e-01 -1.70926309e+00 5.44661582e-01 8.06045890e-01 1.17231798e+00 6.49334490e-01 3.75832796e-01 -2.72931993e-01 8.13767970e-01 -7.11087346e-01 -2.97811568e-01 3.38143528e-01 -3.85906756e-01 -3.44197273e-01 6.35430992e-01 3.95700037e-01 -6.23282075e-01 -6.05213821e-01 1.09272063e+00 5.42312324e-01 -4.29593846e-02 4.80575502e-01 -1.12059808e+00 -4.70625609e-01 6.86124623e-01 -8.49322736e-01 7.86311626e-02 6.97062239e-02 3.02419543e-01 5.54854572e-01 -1.31355226e+00 5.08050203e-01 9.97039676e-01 4.63102669e-01 7.21892715e-01 -7.94315636e-01 -1.01266336e+00 3.37352961e-01 2.46059269e-01 -1.53082061e+00 -4.58485067e-01 6.08014166e-01 -4.03472602e-01 5.55434346e-01 -3.42216976e-02 4.60760385e-01 1.39705861e+00 -1.91694707e-01 1.29126918e+00 9.13391113e-01 -3.95781934e-01 9.19792727e-02 3.75268161e-01 -1.19183019e-01 8.87834132e-01 4.31188971e-01 -1.24451883e-01 -5.83989084e-01 1.81506872e-01 6.01988912e-01 2.56769508e-01 -2.31577501e-01 -2.41878316e-01 -1.20772111e+00 6.58417344e-01 1.09355867e-01 2.81310230e-01 -5.55817112e-02 -1.45339683e-01 6.79029047e-01 1.09767251e-01 5.52939057e-01 2.85486355e-02 -4.30214316e-01 1.73068807e-01 -1.25390649e+00 -2.39862148e-02 4.75633442e-01 1.08692741e+00 5.36325037e-01 2.40352571e-01 -4.24528986e-01 9.78775084e-01 1.60941347e-01 5.56027293e-01 7.58208513e-01 -1.97719857e-01 7.63747394e-01 8.30237508e-01 8.83279275e-03 -9.15860891e-01 -2.83054620e-01 -6.18215561e-01 -1.09499979e+00 -1.80484969e-02 2.56086618e-01 -2.31229484e-01 -1.22723043e+00 1.13691127e+00 4.05077755e-01 3.26996416e-01 -2.73077041e-02 1.10118949e+00 9.29905593e-01 7.96610892e-01 -1.84722990e-02 -3.16205323e-01 1.20014024e+00 -1.20011282e+00 -7.72218704e-01 -5.60947120e-01 7.06933796e-01 -7.56137431e-01 1.07060635e+00 5.12308896e-01 -6.16407394e-01 -8.22028637e-01 -1.12697959e+00 1.49814263e-01 -1.90709934e-01 8.98560464e-01 2.50173688e-01 4.26229179e-01 -4.37631369e-01 3.23342770e-01 -6.46238625e-01 -3.65518570e-01 8.15917253e-01 1.76374823e-01 -3.20087671e-01 -2.67775029e-01 -8.44052017e-01 3.86600703e-01 8.49499285e-01 1.96638346e-01 -1.22843790e+00 -2.43023232e-01 -7.32772470e-01 -8.77783149e-02 8.30973625e-01 -4.08284128e-01 1.15349078e+00 -1.18978727e+00 -1.25549066e+00 9.40052509e-01 2.77645051e-01 -6.25692964e-01 9.32500124e-01 -4.51488614e-01 -4.88572538e-01 9.53691825e-02 3.76610428e-01 6.66482210e-01 1.31912482e+00 -1.05857646e+00 -9.49021041e-01 -2.86836058e-01 -3.54723990e-01 2.46476635e-01 -7.92341053e-01 7.35715255e-02 -9.71901834e-01 -8.62345457e-01 -1.13979600e-01 -7.86845565e-01 -1.63523450e-01 4.88625988e-02 -6.54234827e-01 -3.24529380e-01 1.28040338e+00 -4.83363658e-01 1.20343828e+00 -2.23172188e+00 -2.00256363e-01 -2.12080806e-01 1.50463104e-01 7.09314048e-01 -6.54016212e-02 5.15918024e-02 3.67648602e-02 -9.75120887e-02 -1.30059972e-01 -4.20773596e-01 -2.48523697e-01 -6.02066964e-02 -2.52921849e-01 5.38916469e-01 3.18074524e-01 5.84922910e-01 -7.17918217e-01 -9.73112285e-01 5.09884000e-01 7.44266361e-02 -2.53001660e-01 3.89056534e-01 -5.63028038e-01 2.44154096e-01 -6.00356758e-01 7.54738450e-01 6.46281660e-01 -5.62930226e-01 1.49600720e-02 -2.09865004e-01 3.27072665e-02 -9.85873416e-02 -1.27816033e+00 1.73217487e+00 -1.66003510e-01 7.68309593e-01 8.93642530e-02 -1.11844468e+00 8.06958735e-01 1.08481020e-01 2.97666878e-01 -7.67244875e-01 5.08897483e-01 -3.83664761e-03 -2.40400538e-01 -8.15276682e-01 3.36553842e-01 3.20761055e-01 8.66847560e-02 3.70027661e-01 1.95186976e-02 2.60737509e-01 4.37955737e-01 1.09863929e-01 1.15738690e+00 2.89817840e-01 5.71084693e-02 9.98180285e-02 6.32270694e-01 2.55970955e-01 8.15127552e-01 7.42309868e-01 -1.30077600e-01 6.89833224e-01 1.00240789e-01 -5.37046313e-01 -1.02962589e+00 -5.25614381e-01 -1.48225307e-01 1.27063072e+00 2.72797585e-01 -5.13356626e-01 -9.20931697e-01 -1.03965402e+00 -1.38305947e-01 4.31692064e-01 -6.67925775e-01 -1.17626250e-01 -4.65626419e-01 -9.51179087e-01 4.65193897e-01 5.90840280e-01 9.43616629e-01 -1.20532417e+00 -6.74022198e-01 2.54524529e-01 -2.34125882e-01 -1.49878395e+00 -4.71426159e-01 1.50696680e-01 -7.43469477e-01 -1.01588595e+00 -7.56884694e-01 -9.91242945e-01 7.90610194e-01 7.17976987e-01 8.02196622e-01 2.67335534e-01 -6.70043528e-01 -4.64751087e-02 -5.57778358e-01 -4.37028795e-01 -3.47421885e-01 -7.10705668e-02 -4.88659032e-02 3.35335612e-01 3.92482728e-01 9.85860080e-02 -7.00825632e-01 5.11443198e-01 -9.39370096e-01 3.45967740e-01 9.14635539e-01 9.85028684e-01 6.16259098e-01 4.00103569e-01 5.26770294e-01 -9.46964622e-01 2.28205651e-01 -2.48865828e-01 -7.10920811e-01 1.67254373e-01 -5.36299765e-01 -1.29554182e-01 7.57116377e-01 -7.22742856e-01 -1.17325318e+00 6.78084373e-01 1.28191069e-01 -7.22662449e-01 -2.06983969e-01 2.77162910e-01 -2.74690568e-01 2.22484812e-01 7.30401337e-01 5.62832952e-01 -2.56845206e-01 -4.50217217e-01 1.93996057e-01 1.00676191e+00 5.54763854e-01 -2.05294639e-01 8.65987480e-01 6.27208114e-01 -4.88481849e-01 -8.03241849e-01 -1.09980917e+00 -7.29345977e-01 -4.72840160e-01 -2.19060242e-01 9.65734839e-01 -1.33276975e+00 -1.39346048e-01 6.35607362e-01 -9.87250566e-01 -2.40144417e-01 3.04925162e-02 3.45078140e-01 -3.12469423e-01 5.56464911e-01 -4.27908719e-01 -7.83494055e-01 -6.72887862e-01 -1.07380152e+00 1.28815651e+00 2.28925809e-01 3.44003439e-01 -3.79821062e-01 -2.73239464e-01 4.99076337e-01 -5.82618900e-02 -6.88981637e-02 3.65975916e-01 -7.86654949e-01 -7.30880082e-01 -4.57199723e-01 -5.05724549e-01 4.40836191e-01 -6.03532605e-02 -5.57718538e-02 -1.01386619e+00 -2.67431825e-01 -1.61426738e-01 -6.07751489e-01 1.15478945e+00 9.58021954e-02 1.47792923e+00 -2.68006504e-01 -5.87622166e-01 4.68455702e-01 1.35966861e+00 2.77987476e-02 3.36379111e-01 4.14839238e-01 8.71082783e-01 4.35934663e-01 1.05422366e+00 6.33426368e-01 1.57838792e-01 6.31755114e-01 5.08681595e-01 -1.97545528e-01 -1.37836218e-01 -3.39206904e-01 4.77725774e-01 4.91423070e-01 3.69844943e-01 -4.06622976e-01 -8.52719545e-01 5.05328119e-01 -2.13104296e+00 -8.95692885e-01 -2.34048292e-01 2.08614111e+00 7.27751672e-01 4.91544932e-01 1.16129786e-01 2.75162309e-01 9.71443474e-01 -7.93233961e-02 -6.67542100e-01 5.34675777e-01 -1.42464176e-01 -1.23851016e-01 5.53183734e-01 -3.09534669e-01 -1.64177978e+00 1.23279679e+00 5.38232946e+00 1.07591248e+00 -1.17132044e+00 2.65627712e-01 6.53843582e-01 -2.37387437e-02 5.32882214e-01 -2.61372656e-01 -9.79245603e-01 6.32487416e-01 5.01693606e-01 2.89936922e-02 2.75025051e-02 1.27824974e+00 2.25854158e-01 -6.54968768e-02 -9.53719616e-01 1.38072348e+00 2.76657134e-01 -1.31205130e+00 -6.47940189e-02 -2.20852271e-01 8.25115621e-01 1.32936284e-01 -9.07660052e-02 4.36813474e-01 1.77672908e-01 -6.86896503e-01 7.63524055e-01 1.67492613e-01 8.85909736e-01 -5.54973841e-01 8.03900719e-01 6.97422326e-01 -1.23460472e+00 -3.15149099e-01 -7.25062490e-01 3.24696720e-01 -1.14977382e-01 8.26839268e-01 -9.29406345e-01 4.00067359e-01 9.51490521e-01 9.86273289e-01 -8.10118854e-01 1.17031884e+00 -2.32627109e-01 6.78144991e-01 -1.93681479e-01 -2.25750059e-01 8.43874589e-02 6.89068362e-02 2.88770050e-01 1.37512386e+00 2.06210136e-01 -2.19002560e-01 6.50108278e-01 7.29268432e-01 -2.58207649e-01 4.10420239e-01 -6.71099544e-01 -5.48533574e-02 3.32441598e-01 1.63712168e+00 -9.84532237e-01 -5.87435186e-01 -5.85458457e-01 1.10193884e+00 2.00397521e-01 1.20333485e-01 -1.05047393e+00 -2.57151067e-01 1.23683855e-01 1.35770649e-01 4.74102288e-01 -4.17577382e-03 -1.09914981e-01 -1.42597926e+00 1.31444454e-01 -9.96206939e-01 3.87548119e-01 -7.20826566e-01 -1.23813117e+00 5.60529411e-01 -3.19809616e-01 -1.56566143e+00 -2.68506147e-02 -6.32267535e-01 -5.73189259e-01 2.38125533e-01 -1.25818694e+00 -1.11080801e+00 -6.95895970e-01 6.65455580e-01 1.06656742e+00 -4.11804140e-01 3.19410354e-01 5.35802305e-01 -1.13613570e+00 7.05830097e-01 9.11200568e-02 6.42540693e-01 8.73067558e-01 -8.51161599e-01 3.51481736e-01 9.11908925e-01 3.10628295e-01 -3.26032601e-02 4.61136729e-01 -7.61514246e-01 -1.34932303e+00 -1.64108264e+00 1.14432476e-01 -2.30412051e-01 5.22395372e-01 -7.43514955e-01 -1.02253759e+00 4.05669302e-01 1.08747259e-01 2.30013430e-01 2.69272208e-01 -3.38541299e-01 -1.86395466e-01 -2.24991277e-01 -7.95681357e-01 5.38889110e-01 1.10790586e+00 -2.09943622e-01 -4.10297662e-01 7.58324206e-01 6.56899989e-01 -4.06509250e-01 -3.22240472e-01 4.37360317e-01 3.40610296e-01 -9.05393183e-01 6.82534039e-01 -3.36646348e-01 3.28949541e-01 -4.54245836e-01 2.94768363e-02 -7.41753161e-01 -2.58792881e-02 -4.40527141e-01 -4.63338569e-02 1.39797699e+00 2.48120531e-01 -1.50934801e-01 1.03963459e+00 8.04031938e-02 -1.15238771e-01 -4.95979100e-01 -6.98079646e-01 -8.39035332e-01 -3.95025909e-01 -4.45105970e-01 2.40677804e-01 9.37864780e-01 -2.43301451e-01 6.59236073e-01 -6.19272768e-01 8.85429755e-02 7.52394140e-01 1.31833643e-01 1.16384625e+00 -1.26325083e+00 -2.46459708e-01 -1.84570819e-01 -3.86230677e-01 -1.12705469e+00 1.48087278e-01 -5.86493134e-01 3.89341593e-01 -1.43740320e+00 4.33581173e-01 -3.82725805e-01 -1.00098841e-01 5.77999175e-01 -3.43148321e-01 2.32780382e-01 1.32625520e-01 4.27111417e-01 -1.17284584e+00 6.52927101e-01 1.13518786e+00 -4.13191080e-01 -1.42969921e-01 -7.39121884e-02 -3.39073807e-01 9.54915106e-01 7.47038484e-01 -7.00414419e-01 -3.37384343e-01 -4.16504711e-01 1.76757351e-02 -3.75749171e-02 2.51329154e-01 -1.31910014e+00 3.70682597e-01 -6.09226078e-02 6.94689929e-01 -1.10797298e+00 1.13990828e-01 -8.86130691e-01 -1.77461773e-01 4.52264220e-01 -2.38308042e-01 -2.43042529e-01 5.94010465e-02 8.51182342e-01 -1.49864271e-01 -3.46770018e-01 8.30355585e-01 -1.31396145e-01 -8.80704045e-01 4.64289755e-01 -2.57995933e-01 1.29839644e-01 1.28341532e+00 -5.85113056e-02 -3.58476251e-01 -7.58598670e-02 -3.65622640e-01 4.39176410e-01 3.60410124e-01 6.02275431e-01 8.51289928e-01 -1.15360558e+00 -8.15155685e-01 1.70099363e-01 5.14526665e-01 2.70900190e-01 2.01898694e-01 5.94883084e-01 -3.27654213e-01 2.98532724e-01 -5.60581975e-04 -8.68920863e-01 -1.28766882e+00 8.40825498e-01 1.77371547e-01 -3.06069434e-01 -8.38939369e-01 5.92964590e-01 4.61872399e-01 -1.47512868e-01 5.60164452e-01 -1.03178024e-01 -1.65233418e-01 -6.08259030e-02 6.91438019e-01 1.17327556e-01 2.43516527e-02 -4.56136853e-01 -2.59120136e-01 3.37525904e-01 -4.47171569e-01 2.29152024e-01 1.08940959e+00 -7.65432045e-02 3.64245296e-01 1.93030909e-01 7.80203998e-01 -3.26401800e-01 -1.42253685e+00 -4.05161500e-01 -1.07702374e-01 -3.91617984e-01 1.93542287e-01 -6.34588897e-01 -1.15886045e+00 1.00537825e+00 7.73285866e-01 6.43390492e-02 1.13345122e+00 -7.25203753e-02 8.21042478e-01 6.38654888e-01 2.30745181e-01 -1.54052877e+00 5.52707314e-01 2.44341090e-01 5.49583972e-01 -1.65075827e+00 9.74568445e-03 -3.94607186e-01 -6.48903012e-01 9.86490965e-01 1.14588702e+00 -2.51345765e-02 4.29425567e-01 2.24673226e-01 1.15341425e-01 7.95066506e-02 -7.63437450e-01 -3.57064009e-01 1.98653698e-01 4.49472666e-01 1.73523784e-01 -2.01983273e-01 -1.20167159e-01 5.83667636e-01 2.96825171e-01 1.27262533e-01 3.74861866e-01 9.57779586e-01 -7.20620453e-01 -8.44093263e-01 -4.41263199e-01 7.33862996e-01 -5.33257842e-01 -1.81710441e-02 -3.89651030e-01 8.94778728e-01 2.16073588e-01 9.56975579e-01 -8.22605491e-02 -5.53623855e-01 9.33907703e-02 -8.54659081e-02 1.77368090e-01 -8.71760070e-01 -3.43134433e-01 4.16669011e-01 -6.19017472e-03 -3.34573656e-01 -5.12897611e-01 -5.02609134e-01 -1.24908245e+00 1.10113554e-01 -8.78075182e-01 -1.83716685e-01 6.28053486e-01 9.52101171e-01 3.67688656e-01 5.19765913e-01 7.01383770e-01 -6.52529180e-01 -5.88492632e-01 -1.15920675e+00 -3.89408469e-01 4.15484816e-01 5.45770898e-02 -6.19621098e-01 -1.34692922e-01 4.12929267e-01]
[11.760824203491211, 2.138737678527832]
798553ce-74c4-4eb7-baf2-db79729f1ee5
dual-swin-transformer-based-mutual
2206.03105
null
https://arxiv.org/abs/2206.03105v1
https://arxiv.org/pdf/2206.03105v1.pdf
Dual Swin-Transformer based Mutual Interactive Network for RGB-D Salient Object Detection
Salient Object Detection is the task of predicting the human attended region in a given scene. Fusing depth information has been proven effective in this task. The main challenge of this problem is how to aggregate the complementary information from RGB modality and depth modality. However, conventional deep models heavily rely on CNN feature extractors, and the long-range contextual dependencies are usually ignored. In this work, we propose Dual Swin-Transformer based Mutual Interactive Network. We adopt Swin-Transformer as the feature extractor for both RGB and depth modality to model the long-range dependencies in visual inputs. Before fusing the two branches of features into one, attention-based modules are applied to enhance features from each modality. We design a self-attention-based cross-modality interaction module and a gated modality attention module to leverage the complementary information between the two modalities. For the saliency decoding, we create different stages enhanced with dense connections and keep a decoding memory while the multi-level encoding features are considered simultaneously. Considering the inaccurate depth map issue, we collect the RGB features of early stages into a skip convolution module to give more guidance from RGB modality to the final saliency prediction. In addition, we add edge supervision to regularize the feature learning process. Comprehensive experiments on five standard RGB-D SOD benchmark datasets over four evaluation metrics demonstrate the superiority of the proposed DTMINet method.
['Sam Kwong', 'Chao Zeng']
2022-06-07
null
null
null
null
['rgb-d-salient-object-detection']
['computer-vision']
[ 3.37528855e-01 -4.99991104e-02 -1.54573128e-01 -6.11847281e-01 -4.08338726e-01 -3.47467251e-02 4.68093306e-01 2.79960898e-03 -5.03991127e-01 4.94331568e-01 3.70582163e-01 -9.00298506e-02 7.06357881e-02 -8.05937409e-01 -6.83973134e-01 -8.33710730e-01 4.16395217e-01 -3.62912238e-01 7.44238973e-01 -2.46472940e-01 3.12247992e-01 2.34995335e-01 -1.75338781e+00 5.42446911e-01 1.00172424e+00 1.43218398e+00 7.09570229e-01 4.54086483e-01 -3.40432614e-01 8.62428963e-01 -4.11713347e-02 1.52523350e-02 8.85840803e-02 -3.61869276e-01 -6.45363629e-01 1.08968988e-01 2.07046047e-02 -4.43265736e-01 -5.37633181e-01 1.03267181e+00 6.65766537e-01 1.16934456e-01 2.01723188e-01 -1.26836431e+00 -5.93576133e-01 4.26368505e-01 -9.31670785e-01 6.24206185e-01 3.39606553e-01 3.79644364e-01 8.09335470e-01 -1.02108169e+00 2.69900441e-01 1.06802773e+00 1.36385158e-01 4.07193214e-01 -7.20891654e-01 -5.46242416e-01 5.60956836e-01 5.01914740e-01 -1.10096514e+00 -3.11893106e-01 1.20232999e+00 -1.34838447e-01 8.80352199e-01 1.16422385e-01 8.56615543e-01 7.71320701e-01 1.85060471e-01 1.27809298e+00 1.07954085e+00 -1.83640838e-01 -2.07470387e-01 9.95087475e-02 1.66097835e-01 6.93712950e-01 -1.12067640e-01 4.29488234e-02 -8.69737029e-01 3.94060075e-01 8.61713529e-01 6.25203609e-01 -3.46932679e-01 -2.26551712e-01 -1.26168501e+00 6.24677777e-01 1.24514878e+00 2.43989557e-01 -4.58349079e-01 5.97716458e-02 2.39096954e-01 6.80784211e-02 4.55091506e-01 -6.56306818e-02 -5.10557532e-01 2.26943880e-01 -7.68668234e-01 -2.28933841e-01 8.90930463e-03 8.92811179e-01 1.12914038e+00 -2.60754794e-01 -4.47602004e-01 5.79278469e-01 5.04742205e-01 3.82566571e-01 5.71909070e-01 -3.25229019e-01 6.81146741e-01 1.15937495e+00 -1.51551366e-01 -9.98652816e-01 -5.21286905e-01 -5.45956671e-01 -8.10356200e-01 2.16755122e-01 3.21587957e-02 6.54011071e-02 -1.31490088e+00 1.63692451e+00 4.56074059e-01 2.04920962e-01 -9.88383219e-02 1.54905081e+00 1.19518244e+00 6.03225350e-01 2.05883458e-01 1.73177309e-02 1.31961513e+00 -1.23859572e+00 -4.81540978e-01 -5.39122462e-01 3.79368842e-01 -5.78607261e-01 1.07897067e+00 -1.41289726e-01 -1.02109730e+00 -6.80318594e-01 -1.13652432e+00 -7.16369390e-01 -4.05778706e-01 9.94948521e-02 8.08643818e-01 -2.91046835e-02 -8.01302731e-01 3.37767541e-01 -9.09643471e-01 7.97145534e-03 7.27011919e-01 6.10447288e-01 -3.54948610e-01 -2.23176241e-01 -1.33963859e+00 7.27879167e-01 3.75517696e-01 5.81693232e-01 -7.26014316e-01 -3.69746774e-01 -9.51151013e-01 6.04846589e-02 2.27690548e-01 -6.68691397e-01 9.10981357e-01 -1.24083233e+00 -1.15340507e+00 7.48366654e-01 -4.46073771e-01 6.66747466e-02 9.08629522e-02 -2.48160452e-01 -2.41261184e-01 7.31850415e-02 1.46465942e-01 8.39470804e-01 8.31905723e-01 -1.06051242e+00 -1.07685292e+00 -5.50330162e-01 2.37964734e-01 7.63824821e-01 -4.90121692e-01 -1.71816498e-01 -6.60728574e-01 -3.62753838e-01 5.63498676e-01 -4.62431133e-01 -6.82943314e-02 1.18209198e-02 -5.65427244e-01 -1.19625270e-01 8.56530011e-01 -5.34487307e-01 1.13615489e+00 -2.11757851e+00 4.97910142e-01 4.11535464e-02 5.04219413e-01 -3.63681950e-02 3.40976492e-02 -1.93251014e-01 -8.71633962e-02 -2.92869598e-01 -2.06947386e-01 -4.65192497e-01 -2.86712915e-01 1.67961940e-01 -2.06636876e-01 3.55136603e-01 6.42180324e-01 1.05744541e+00 -1.07365096e+00 -6.50376022e-01 3.11190158e-01 5.57557344e-01 -4.06865031e-01 3.48362684e-01 -5.40840104e-02 6.02122307e-01 -8.03150535e-01 8.05216491e-01 8.09007227e-01 -3.51666421e-01 -4.50505912e-01 -5.65775931e-01 -2.92106867e-01 3.88467133e-01 -8.03929329e-01 2.23177910e+00 -5.11316121e-01 4.79493976e-01 -1.60233721e-01 -7.59173155e-01 8.63925457e-01 -4.32971120e-02 1.67346328e-01 -9.75920200e-01 5.35025775e-01 1.54399708e-01 1.12686493e-01 -4.98577535e-01 3.41085762e-01 -6.46589398e-02 3.36435996e-02 4.32572693e-01 2.30063856e-01 2.25457907e-01 -3.54999512e-01 1.23333201e-01 8.95445049e-01 3.39700818e-01 8.87374394e-03 -3.98529880e-02 6.50960267e-01 -2.38357127e-01 7.01106966e-01 1.96490675e-01 -3.00969452e-01 7.75562048e-01 3.65094036e-01 -4.03456926e-01 -5.78443706e-01 -8.13088298e-01 1.09559841e-01 1.44456768e+00 8.30367863e-01 -9.03115124e-02 -3.51275325e-01 -8.88480365e-01 -9.93625596e-02 2.04579324e-01 -1.04889965e+00 -5.17936587e-01 -3.33464652e-01 -6.82219923e-01 -4.35754880e-02 7.66495943e-01 8.91554177e-01 -1.25128424e+00 -9.74026084e-01 4.88040857e-02 -1.67189747e-01 -9.12261188e-01 -5.93369722e-01 5.66714227e-01 -8.21862221e-01 -8.81221831e-01 -7.38924444e-01 -9.42293048e-01 7.45405138e-01 5.97857654e-01 7.29500413e-01 3.10479194e-01 -7.29130581e-02 6.22172616e-02 -3.70249748e-01 -3.35636258e-01 4.56921875e-01 3.71322036e-01 -3.82124573e-01 3.17166150e-01 5.44851601e-01 -7.06000924e-01 -1.06593871e+00 2.30339468e-01 -9.61986184e-01 6.32476151e-01 8.80797684e-01 8.81451309e-01 6.55063570e-01 -2.86661446e-01 4.79078472e-01 -4.84277636e-01 3.81138362e-02 -6.02416217e-01 -2.09147051e-01 2.09097117e-01 -1.72568724e-01 1.77144274e-01 3.00497591e-01 -2.10782692e-01 -1.22230268e+00 3.99240494e-01 -9.07881409e-02 -5.87576210e-01 -6.57480508e-02 4.00750697e-01 -6.53011203e-01 -1.13820042e-02 1.56502679e-01 3.83312494e-01 -2.51254648e-01 -3.62457603e-01 1.89981639e-01 6.18481040e-01 4.53569889e-01 -9.70070958e-02 5.69853544e-01 4.54849750e-01 -2.09200501e-01 -3.89821529e-01 -1.18608463e+00 -3.72833848e-01 -6.73032939e-01 -1.86516970e-01 1.01137400e+00 -1.09128320e+00 -5.68536162e-01 6.49284542e-01 -1.13076329e+00 -2.03347459e-01 -1.78098425e-01 3.44216913e-01 -2.10612565e-01 -2.22172141e-02 -5.02254188e-01 -4.84900028e-01 -3.44244599e-01 -1.42632282e+00 1.37549007e+00 8.87692511e-01 5.50511897e-01 -8.17390025e-01 -2.42461249e-01 1.45623103e-01 3.38063926e-01 1.10569254e-01 5.41201711e-01 -2.81152755e-01 -7.56705642e-01 -2.39816550e-02 -8.79206240e-01 1.38169691e-01 2.97605306e-01 -2.86847383e-01 -1.29004264e+00 1.50863007e-02 7.91796744e-02 -3.37868690e-01 1.39955282e+00 1.87385276e-01 1.24275970e+00 1.75173506e-01 -4.44433212e-01 7.55126536e-01 1.30418193e+00 -1.43113598e-01 5.95777273e-01 4.39593226e-01 1.22013223e+00 4.88443226e-01 7.04668999e-01 3.78378421e-01 7.19598711e-01 3.13568562e-01 7.49374270e-01 -4.23738569e-01 -7.82984197e-02 -2.86466628e-01 1.21252902e-01 5.84848881e-01 -9.91668105e-02 2.84496490e-02 -8.31527770e-01 5.41119456e-01 -1.95789242e+00 -6.53549135e-01 -4.66017611e-02 2.01926541e+00 8.86204481e-01 4.03867036e-01 -3.75489108e-02 1.23355716e-01 7.56240368e-01 1.50769487e-01 -7.56830871e-01 -9.43225902e-03 -3.84510666e-01 -7.22153932e-02 2.76889890e-01 2.99972206e-01 -1.26097584e+00 8.41008961e-01 4.39390182e+00 6.87789857e-01 -1.49763644e+00 1.76433697e-01 7.61920929e-01 -3.63763452e-01 -5.21939814e-01 8.66988301e-02 -6.61750376e-01 5.19964576e-01 2.55293429e-01 2.56932646e-01 8.01014528e-02 6.10589266e-01 -5.75553533e-03 -4.65421766e-01 -9.72359717e-01 1.05482590e+00 4.79069576e-02 -9.67329741e-01 -8.83800238e-02 -1.03018731e-01 6.60937369e-01 2.20309332e-01 2.31673285e-01 1.86439812e-01 -2.01238289e-01 -8.42773557e-01 7.40938962e-01 8.42755258e-01 5.62193036e-01 -9.20463443e-01 8.64398539e-01 1.79292038e-01 -1.45788908e+00 -2.72864491e-01 -4.60748345e-01 -1.93475604e-01 1.00759126e-01 7.95915544e-01 -1.24834813e-01 6.01182818e-01 8.59673977e-01 1.24612415e+00 -8.26051354e-01 1.10072148e+00 -3.82185787e-01 1.17116146e-01 -3.63619238e-01 1.81003243e-01 3.17401022e-01 1.52966455e-01 3.50938201e-01 9.53779757e-01 1.23804405e-01 2.14883924e-01 1.35371536e-01 7.61602461e-01 -3.36673041e-03 -1.57537341e-01 -3.23494464e-01 3.66491824e-01 2.01045707e-01 1.47936082e+00 -7.43376255e-01 -2.55871981e-01 -6.49907231e-01 1.41729867e+00 4.94877934e-01 2.70358145e-01 -9.58817244e-01 -5.43062210e-01 5.25565326e-01 -4.03573141e-02 5.02663136e-01 7.91784152e-02 -6.27192736e-01 -1.22259283e+00 7.39372224e-02 -2.95667380e-01 3.35982889e-01 -9.46394920e-01 -1.01752067e+00 7.86365867e-01 -3.11778635e-01 -1.41485608e+00 3.71507287e-01 -3.98160636e-01 -8.95526230e-01 1.27225697e+00 -2.15934253e+00 -1.46256149e+00 -8.14005554e-01 8.90564382e-01 3.54780614e-01 2.56075203e-01 2.77579755e-01 3.16416562e-01 -8.75044882e-01 5.20470202e-01 -3.81161332e-01 -2.14631855e-02 5.25086462e-01 -1.10911477e+00 4.51608002e-02 8.78487527e-01 -3.12360346e-01 4.62974966e-01 3.10070038e-01 -6.21009648e-01 -1.26381421e+00 -1.13460755e+00 6.35288239e-01 -1.95782989e-01 4.38012242e-01 -2.85975665e-01 -9.58193600e-01 5.27739286e-01 1.80846483e-01 4.98305321e-01 3.92810196e-01 -1.30715325e-01 -2.47891903e-01 -1.99172810e-01 -6.69719338e-01 3.04721862e-01 1.18005240e+00 -7.07906485e-01 -7.03537583e-01 -1.78396143e-03 1.01645875e+00 -5.53360403e-01 -5.30186772e-01 5.26926696e-01 4.96524721e-01 -1.13713169e+00 7.63772488e-01 -3.41571808e-01 7.97129273e-01 -5.98542333e-01 -1.39195710e-01 -1.06562650e+00 -2.80798554e-01 -6.51195422e-02 -1.13290235e-01 1.25943339e+00 3.63161534e-01 -3.46467406e-01 7.85498619e-01 5.65756083e-01 -4.43680316e-01 -1.14378047e+00 -7.90904462e-01 -5.94924092e-02 -4.29807186e-01 -3.21615160e-01 6.42730474e-01 7.25143671e-01 2.05098162e-03 6.57947063e-01 -2.68198967e-01 1.48128092e-01 3.90174270e-01 5.34293115e-01 3.97501379e-01 -1.01493299e+00 -1.18389226e-01 -3.45419526e-01 -5.14904737e-01 -1.35024786e+00 -6.87773228e-02 -8.43144536e-01 2.56510109e-01 -1.64388943e+00 4.51132208e-01 -5.11547983e-01 -8.76907945e-01 7.48159587e-01 -7.65294194e-01 3.13148201e-01 1.09439939e-01 7.13980943e-02 -8.24210525e-01 1.01597393e+00 1.50537276e+00 -1.53358325e-01 -3.77496541e-01 -2.24697411e-01 -7.52666652e-01 7.98866212e-01 6.92677677e-01 -2.60947078e-01 -5.62874556e-01 -8.55201423e-01 6.06381223e-02 -1.08951233e-01 7.17030823e-01 -1.00474799e+00 6.11212134e-01 -1.44802019e-01 8.28564584e-01 -8.11465800e-01 3.76414269e-01 -8.50346982e-01 -5.37719071e-01 7.67058060e-02 -1.95797682e-01 -7.73805380e-02 2.88677931e-01 5.08467257e-01 -3.60995412e-01 2.30649441e-01 5.23866296e-01 3.88233960e-02 -1.11231101e+00 6.64327860e-01 1.19702257e-01 -7.17355013e-02 9.46554303e-01 -2.76432663e-01 -3.07747096e-01 -9.94001050e-03 -7.42300034e-01 4.95033950e-01 4.91511345e-01 5.24298668e-01 1.02989316e+00 -1.46320331e+00 -4.05974448e-01 4.89382982e-01 2.54385322e-01 3.92074496e-01 7.01650739e-01 1.16853487e+00 -9.58333984e-02 1.52402401e-01 -5.72270691e-01 -8.06820989e-01 -8.88681114e-01 5.66799343e-01 3.05077732e-01 -7.84012228e-02 -3.70353162e-01 1.22692955e+00 6.03694737e-01 -3.22488099e-02 1.33830801e-01 -3.21350217e-01 -4.47677940e-01 1.11947522e-01 6.20755434e-01 1.82040303e-03 1.08549379e-04 -7.67001390e-01 -6.45201921e-01 5.45095086e-01 -2.51547337e-01 7.45294094e-02 1.37784851e+00 -5.64775050e-01 -1.82909921e-01 4.98750895e-01 1.28850961e+00 -3.80192757e-01 -1.63350654e+00 -4.63219762e-01 -4.40594167e-01 -4.46633697e-01 4.98864204e-01 -5.76661766e-01 -1.56711793e+00 1.30913365e+00 7.31155217e-01 -1.29551440e-01 1.72571707e+00 5.87333888e-02 8.89901817e-01 -9.42615941e-02 1.17837815e-02 -7.78152764e-01 1.13171570e-01 4.08296287e-01 8.11150193e-01 -1.55423093e+00 -1.06057845e-01 -4.30023581e-01 -7.88312674e-01 8.81753623e-01 1.10550082e+00 -1.61736637e-01 7.41707921e-01 -3.00763571e-03 -1.39406070e-01 -2.31968060e-01 -6.12672985e-01 -6.16308451e-01 4.84791428e-01 3.88042897e-01 3.08014065e-01 -2.43372768e-01 1.92300491e-02 9.54168677e-01 2.64156342e-01 7.93142617e-02 1.24019973e-01 1.17343307e+00 -5.96185625e-01 -7.69905806e-01 -6.48925230e-02 5.82894385e-01 -1.18175745e-01 -3.61248434e-01 -1.81899026e-01 3.91643763e-01 5.27980924e-01 7.77105808e-01 1.05243139e-01 -8.01754177e-01 3.13468426e-01 -1.25127971e-01 2.81189203e-01 -5.17337739e-01 -6.42262280e-01 1.45490080e-01 -3.96217406e-01 -7.25033402e-01 -6.35543883e-01 -5.93745768e-01 -1.46204281e+00 -6.53253943e-02 -5.43737650e-01 -1.02244318e-01 3.74132246e-01 1.08714020e+00 5.34345686e-01 9.68212664e-01 8.32960010e-01 -1.22696745e+00 8.13118666e-02 -8.71507227e-01 -4.62097764e-01 1.59530774e-01 7.04373777e-01 -9.75196838e-01 -2.94471234e-01 -2.31224477e-01]
[9.747942924499512, -0.7042974829673767]
dff35287-ca63-4c1e-8410-4c9590ed87ea
co-design-hardware-and-algorithm-for-vector
2306.11182
null
https://arxiv.org/abs/2306.11182v3
https://arxiv.org/pdf/2306.11182v3.pdf
Co-design Hardware and Algorithm for Vector Search
Vector search has emerged as the foundation for large-scale information retrieval and machine learning systems, with search engines like Google and Bing processing tens of thousands of queries per second on petabyte-scale document datasets by evaluating vector similarities between encoded query texts and web documents. As performance demands for vector search systems surge, accelerated hardware offers a promising solution in the post-Moore's Law era. We introduce \textit{FANNS}, an end-to-end and scalable vector search framework on FPGAs. Given a user-provided recall requirement on a dataset and a hardware resource budget, \textit{FANNS} automatically co-designs hardware and algorithm, subsequently generating the corresponding accelerator. The framework also supports scale-out by incorporating a hardware TCP/IP stack in the accelerator. \textit{FANNS} attains up to 23.0$\times$ and 37.2$\times$ speedup compared to FPGA and CPU baselines, respectively, and demonstrates superior scalability to GPUs, achieving 5.5$\times$ and 7.6$\times$ speedup in median and 95\textsuperscript{th} percentile (P95) latency within an eight-accelerator configuration. The remarkable performance of \textit{FANNS} lays a robust groundwork for future FPGA integration in data centers and AI supercomputers.
['Gustavo Alonso', 'Torsten Hoefler', 'Theodoros Rekatsinas', 'Shuai Zhang', 'Cedric Renggli', 'Runbin Shi', 'Zhenhao He', 'Johannes De Fine Licht', 'Yu Zhu', 'Shigang Li', 'Wenqi Jiang']
2023-06-19
null
null
null
null
['information-retrieval']
['natural-language-processing']
[ 1.56140745e-01 -9.10359919e-01 -4.74285901e-01 -3.39381635e-01 -7.71544039e-01 -7.22157776e-01 6.05329990e-01 5.01653254e-01 -6.40577555e-01 2.16915086e-01 -1.58384189e-01 -1.04478538e+00 -3.62005889e-01 -9.54175174e-01 -2.51630127e-01 -3.27872574e-01 -1.39484033e-01 4.68932062e-01 3.42745423e-01 -3.07392895e-01 6.01656199e-01 5.30880749e-01 -1.71789646e+00 4.55409199e-01 2.65679955e-01 1.69003320e+00 1.95803151e-01 1.00988543e+00 -1.82443529e-01 3.67409408e-01 -7.67514050e-01 -3.94061387e-01 5.81648767e-01 2.44181380e-01 -3.34150255e-01 -6.72336459e-01 6.62890255e-01 -8.12660098e-01 -6.97103858e-01 7.81447530e-01 5.45071959e-01 -1.58305869e-01 3.80846262e-01 -1.38778746e+00 1.24973796e-01 2.89922208e-01 -7.12833345e-01 7.64546275e-01 5.46723783e-01 2.93341964e-01 1.26714218e+00 -7.92437434e-01 4.41297710e-01 7.45671511e-01 4.16366100e-01 -3.77991170e-01 -9.23680425e-01 -8.45055044e-01 -4.51618403e-01 1.87351316e-01 -1.64022470e+00 -1.58759683e-01 4.72100526e-02 -2.67517030e-01 1.75949848e+00 8.46822083e-01 6.92536473e-01 6.97427154e-01 6.57868207e-01 7.10970640e-01 5.77853739e-01 -4.10772562e-01 2.24137217e-01 6.03133030e-02 6.31350696e-01 5.99154353e-01 6.31694794e-01 2.06952706e-01 -6.79609299e-01 -7.02248216e-01 6.74447417e-01 1.55054227e-01 7.82122388e-02 2.42608100e-01 -1.45338154e+00 8.53150368e-01 3.62909019e-01 2.74466872e-01 -3.80216837e-01 5.32672703e-01 8.62787247e-01 9.67447683e-02 -3.90118137e-02 4.37967569e-01 -3.28708649e-01 -4.39718604e-01 -1.44127727e+00 5.47478318e-01 7.62143731e-01 1.14502263e+00 4.71651763e-01 2.58178145e-01 -5.96760631e-01 5.70575237e-01 7.03311898e-03 8.88272643e-01 5.85081816e-01 -3.78414929e-01 6.83124721e-01 4.71961081e-01 -1.60106778e-01 -1.07221365e+00 -5.77720284e-01 -7.35992372e-01 -8.47656548e-01 -3.80438536e-01 1.01415135e-01 1.39650807e-01 -5.68153739e-01 9.31076348e-01 1.89367846e-01 -2.28959545e-01 -4.17901188e-01 7.97763288e-01 3.91808569e-01 1.11855090e+00 -9.61831436e-02 -1.55749038e-01 2.02324986e+00 -9.16516304e-01 -4.51731354e-01 -2.01217026e-01 7.47514963e-01 -1.14479899e+00 1.11890566e+00 5.36287785e-01 -1.00907576e+00 -6.13018274e-01 -1.44159830e+00 -9.08861160e-02 -4.11507249e-01 2.37096086e-01 9.50498521e-01 9.54547703e-01 -1.01492846e+00 3.13374132e-01 -9.39697504e-01 -3.50533456e-01 1.81982040e-01 7.41318464e-01 3.49172562e-01 4.22764430e-03 -8.23550701e-01 2.70870179e-01 4.64479506e-01 -3.27086538e-01 -2.24717796e-01 -1.15773463e+00 -2.10531622e-01 6.00369334e-01 8.10412765e-02 -6.79878175e-01 1.11450398e+00 1.62418447e-02 -1.08397353e+00 4.64177042e-01 7.43954629e-02 -6.34851515e-01 9.74125043e-02 4.77336757e-02 -8.09476435e-01 1.18669614e-01 9.82829090e-03 3.74720663e-01 6.33077085e-01 4.81041111e-02 -9.62573707e-01 -5.86396396e-01 -4.48717713e-01 -7.45194703e-02 -9.66050208e-01 2.51286358e-01 -9.34671342e-01 -7.76624441e-01 -7.14731365e-02 -8.50078940e-01 -2.62973487e-01 -1.45702288e-01 -3.32323730e-01 -1.99266642e-01 7.94499397e-01 -1.59225672e-01 1.81341624e+00 -1.89822769e+00 -6.56884134e-01 7.45372593e-01 3.18289816e-01 5.42264998e-01 2.12896571e-01 5.29133081e-01 3.45859408e-01 -3.89435649e-01 6.69699728e-01 3.75432163e-01 3.04391056e-01 -3.79914999e-01 -7.07235098e-01 4.47582603e-01 -4.45104301e-01 7.70193100e-01 -5.08056045e-01 -2.17475235e-01 2.91330695e-01 4.38626289e-01 -7.39898622e-01 -2.09235370e-01 -1.04238421e-01 -4.72670734e-01 -8.38527203e-01 7.53040254e-01 6.33486390e-01 -7.62652278e-01 1.13394275e-01 -4.93112653e-01 -3.00760537e-01 3.35764587e-01 -1.16081214e+00 1.50622511e+00 -4.43179846e-01 8.42945695e-01 2.65578032e-02 -6.76245809e-01 8.66284370e-01 -4.51936275e-02 6.57479286e-01 -1.17320907e+00 4.32673484e-01 4.75848973e-01 -2.28438765e-01 1.17629431e-01 1.16796279e+00 5.02546370e-01 -4.59735304e-01 7.28141725e-01 -1.31999120e-01 -2.94924319e-01 5.64305246e-01 3.89618784e-01 1.55359435e+00 -7.42622137e-01 7.56670013e-02 -7.74339676e-01 4.46987540e-01 3.87310445e-01 -1.72910944e-01 8.82001638e-01 8.71017203e-02 -1.83830217e-01 4.34577577e-02 -6.29065275e-01 -1.19065619e+00 -9.66290772e-01 -6.55315697e-01 1.59588373e+00 -9.35001224e-02 -1.21920002e+00 -4.78234679e-01 -1.84956081e-02 3.49281698e-01 5.94489515e-01 3.57118934e-01 8.97007212e-02 -5.22218049e-01 -8.84233594e-01 9.02605057e-01 5.43449759e-01 5.94366968e-01 -3.95773411e-01 -1.24910820e+00 5.37327468e-01 5.73880613e-01 -1.26636517e+00 -6.21873081e-01 2.64330566e-01 -6.19001508e-01 -5.22445738e-01 -1.86659813e-01 -4.60243583e-01 2.55039036e-01 3.58861297e-01 1.00621641e+00 1.48978040e-01 -1.14726460e+00 1.41570464e-01 3.79905514e-02 -3.30014050e-01 9.72382072e-03 2.11854130e-01 3.11162055e-01 -6.33921206e-01 6.70427740e-01 -3.59164804e-01 -1.07145905e+00 4.49851155e-01 -8.89615059e-01 -1.19176872e-01 4.31790829e-01 8.58827531e-01 5.67702830e-01 1.28176510e-01 -4.35566083e-02 -5.15751064e-01 7.56259680e-01 -1.71366781e-01 -1.36883664e+00 -1.31924134e-02 -1.14436221e+00 1.28212959e-01 9.70296621e-01 -1.60504490e-01 -3.34826082e-01 -1.60521671e-01 -2.77254194e-01 -2.08875775e-01 4.26186562e-01 9.45111662e-02 3.95868391e-01 -1.44950554e-01 8.60532522e-01 2.45322555e-01 -3.48428518e-01 1.39099851e-01 1.36964336e-01 8.86852264e-01 7.50701666e-01 -6.42917693e-01 5.77637672e-01 3.93360049e-01 4.00592946e-02 -8.27062964e-01 -1.62630677e-01 -7.79796362e-01 1.64284900e-01 2.50451803e-01 4.32115346e-01 -1.00912035e+00 -1.44235432e+00 -1.13214917e-01 -7.99209714e-01 3.98076139e-02 6.46995530e-02 5.05500913e-01 -3.02226990e-01 4.89985868e-02 -6.17185354e-01 -3.67091715e-01 -1.21657550e+00 -1.42598939e+00 1.36146855e+00 3.01336378e-01 -5.41775465e-01 -4.34730083e-01 -3.62458825e-01 3.05134535e-01 9.82399702e-01 -3.38534385e-01 9.12093639e-01 -8.22087526e-01 -7.99456060e-01 -7.34298885e-01 -5.93191922e-01 -6.55851290e-02 -4.60695148e-01 -4.94216830e-02 -5.75517118e-01 -8.09593499e-01 -1.95277825e-01 -4.87321615e-02 4.72187072e-01 1.36737436e-01 1.40526736e+00 -2.66535908e-01 -8.87681842e-01 8.99574518e-01 1.61787033e+00 4.59477663e-01 3.09061259e-01 1.93208203e-01 1.57861844e-01 -3.88306439e-01 6.30457520e-01 1.02230620e+00 -1.20025061e-01 9.94434714e-01 -3.13092992e-02 2.43576869e-01 1.70861021e-01 -1.67324036e-01 1.07305445e-01 7.88775325e-01 4.73480076e-01 -5.76021612e-01 -1.13736451e+00 1.82690267e-02 -1.09196472e+00 -7.36996233e-01 2.91713085e-02 2.36717319e+00 4.95936394e-01 5.23913324e-01 -1.68101028e-01 2.32849583e-01 2.72421449e-01 1.71554163e-01 -7.59361207e-01 -8.24160993e-01 2.21056044e-01 7.90975213e-01 1.06811309e+00 7.29341656e-02 -7.80058682e-01 8.03783059e-01 5.42227316e+00 1.51826656e+00 -1.39032614e+00 -1.42578065e-01 4.49883044e-01 -4.86641377e-01 6.70007840e-02 -2.20116690e-01 -1.52361381e+00 6.14863753e-01 1.60696447e+00 -7.48689473e-01 3.29960734e-01 1.19065487e+00 -2.16243595e-01 -1.87711149e-01 -1.17806900e+00 1.42488372e+00 -6.56761527e-02 -1.79173577e+00 -2.76435670e-02 4.68147129e-01 4.45752501e-01 2.08965644e-01 4.06403631e-01 4.56337422e-01 1.91513419e-01 -6.84505463e-01 3.45013022e-01 -3.52819622e-01 1.28922021e+00 -7.41383851e-01 3.87150615e-01 1.62732065e-01 -1.44094276e+00 -2.04226971e-01 -2.92789459e-01 1.83648437e-01 9.35500339e-02 6.69855714e-01 -1.33394885e+00 2.36175209e-01 7.24951923e-01 -8.53560194e-02 -6.43316627e-01 8.09724391e-01 5.88929653e-01 4.49887782e-01 -8.49291205e-01 -7.73012817e-01 1.67971507e-01 4.65987213e-02 3.24235380e-01 1.50779307e+00 5.51198065e-01 2.91683972e-01 2.00277746e-01 2.37289369e-01 -7.76190087e-02 3.62847120e-01 -2.26848036e-01 -2.65353441e-01 7.41297066e-01 1.33009708e+00 -9.90667939e-01 -7.68569231e-01 -2.57959455e-01 1.08678877e+00 -2.19027176e-01 -8.40559006e-02 -9.89277899e-01 -9.28774118e-01 8.84069920e-01 9.87832546e-02 3.47018898e-01 -4.21103984e-01 -5.65121114e-01 -9.24032509e-01 9.70750749e-02 -1.02730811e+00 5.42133749e-01 -5.13164937e-01 -7.63576031e-01 7.57394075e-01 -1.32035747e-01 -1.06716669e+00 -5.65064192e-01 -8.07103038e-01 -3.87804694e-02 6.97267771e-01 -7.99546301e-01 -5.77139139e-01 -3.60323846e-01 6.90411031e-01 4.80258942e-01 -3.49733829e-01 9.11748827e-01 6.08220398e-01 -3.38665664e-01 1.25743103e+00 6.89864516e-01 -2.51589417e-01 4.10265625e-01 -4.99042332e-01 7.62285531e-01 4.82559800e-01 2.50173390e-01 9.67049062e-01 6.94946110e-01 -5.21191478e-01 -2.72147727e+00 -7.87663281e-01 3.34626019e-01 -6.42046705e-02 1.05628657e+00 -6.02014661e-01 -3.69982630e-01 3.81794631e-01 6.97560981e-02 2.03468099e-01 7.51438677e-01 -6.70892596e-02 -4.99438584e-01 -5.26727378e-01 -1.11976635e+00 6.16155028e-01 7.73356676e-01 -4.87747043e-01 2.88955212e-01 9.45990503e-01 6.50678098e-01 -8.35325539e-01 -9.35713828e-01 2.05979630e-01 6.72735274e-01 -7.94863641e-01 1.33114386e+00 -2.76078850e-01 -1.00024983e-01 1.19424514e-01 -6.19784117e-01 -4.50838149e-01 -1.78500384e-01 -8.21515918e-01 -7.60725141e-02 5.96651435e-01 3.19877505e-01 -5.21988869e-01 1.09420729e+00 5.52688479e-01 1.34078683e-02 -6.21495664e-01 -9.18534338e-01 -8.17008972e-01 -2.94480979e-01 -8.61818135e-01 6.63877606e-01 5.66699147e-01 1.74587458e-01 4.07590777e-01 1.42854303e-01 6.67171776e-02 5.10813355e-01 1.87364981e-01 7.60678649e-01 -7.97005773e-01 -4.50265944e-01 -7.83159137e-01 -7.84152329e-01 -1.38406801e+00 -3.96698982e-01 -1.08251381e+00 -5.62125146e-01 -9.67769086e-01 -1.17171049e-01 -6.07473910e-01 -2.82467276e-01 3.25237125e-01 1.77590191e-01 4.15747851e-01 1.98869258e-01 7.60856569e-02 -5.32293499e-01 2.96469983e-02 6.01683617e-01 -2.30238616e-01 -7.28917494e-02 -1.06570490e-01 -4.07253921e-01 9.26445127e-02 4.91654366e-01 9.13240388e-03 -3.61841261e-01 -5.07314861e-01 3.59921724e-01 2.87329167e-01 2.80880202e-02 -1.29612255e+00 9.36760604e-01 1.90224454e-01 5.72228670e-01 -1.00135660e+00 5.32405794e-01 -7.64745474e-01 1.25940695e-01 8.23759317e-01 -1.74232852e-02 8.73992622e-01 6.83239639e-01 2.44617313e-01 1.58164557e-02 2.09573463e-01 4.03030515e-01 4.94597524e-01 -4.81677830e-01 3.14740837e-01 -4.36395735e-01 -9.64780897e-02 1.11169255e+00 -8.50296393e-02 -6.47694409e-01 1.68757349e-01 1.89748138e-01 2.99547940e-01 -2.84412969e-03 3.54136348e-01 5.40452361e-01 -1.18936944e+00 -4.65339839e-01 8.73464346e-01 -5.01636080e-02 -5.83187044e-01 2.36678451e-01 3.48221987e-01 -1.02742708e+00 1.47585833e+00 -3.85327823e-02 -9.29233551e-01 -1.37142992e+00 5.37453413e-01 -3.71843189e-01 -4.78845835e-01 -4.29233432e-01 1.03230715e+00 -3.94662350e-01 2.08911940e-01 1.89728171e-01 -4.49136496e-01 6.44664049e-01 -5.60353212e-02 9.68988597e-01 5.51452577e-01 8.36003482e-01 -1.00061195e-02 -5.26285648e-01 2.77993888e-01 -5.86606801e-01 -1.33752465e-01 8.98704648e-01 4.27457094e-01 -1.48403600e-01 -3.36652905e-01 1.68796813e+00 -1.15329020e-01 -2.81458080e-01 -2.09970281e-01 4.05440293e-03 -6.91902459e-01 5.62873721e-01 -6.10067368e-01 -9.36041474e-01 5.99886417e-01 9.26563740e-01 1.00694239e-01 1.21164715e+00 -2.24988773e-01 1.18624103e+00 6.80950224e-01 7.48521447e-01 -1.11203277e+00 -1.48383707e-01 4.03092414e-01 2.36625209e-01 -7.37964988e-01 6.65728450e-01 -1.27121001e-01 -3.04067899e-02 1.24276245e+00 4.21596318e-01 5.92437536e-02 6.95998251e-01 8.92021179e-01 -5.04677653e-01 -3.73966604e-01 -1.06599379e+00 4.92532194e-01 2.47744843e-01 -7.37666711e-02 3.60629171e-01 1.76430210e-01 -5.63635230e-01 3.88773561e-01 -7.68273294e-01 1.53947383e-01 -5.79593554e-02 1.24272907e+00 -4.69182581e-01 -1.09711623e+00 -5.04952431e-01 1.06204307e+00 -3.57024938e-01 -5.25404096e-01 4.94687110e-01 5.71519434e-01 -5.08253455e-01 7.79469013e-01 5.46499431e-01 -6.26111686e-01 3.43001634e-01 -2.27575362e-01 2.09162757e-01 -1.44024864e-01 -1.08839619e+00 1.39185563e-01 -9.85898823e-02 -9.23700213e-01 8.30590963e-01 -3.81039053e-01 -1.34969282e+00 -8.60592782e-01 -2.29667500e-01 2.55775452e-01 1.27080452e+00 3.01089019e-01 8.46429884e-01 5.57429552e-01 5.66492140e-01 -4.82642442e-01 -8.91817093e-01 -4.40753877e-01 -3.66059154e-01 -3.48296672e-01 -1.28265470e-01 -2.06442237e-01 -1.09887376e-01 -3.82066220e-01]
[8.547296524047852, 3.2680563926696777]
84748616-27f5-4bd3-a097-e68314a8d5d6
a-data-and-analysis-resourcea-fora-ana
null
null
https://aclanthology.org/L12-1629
https://aclanthology.org/L12-1629.pdf
A data and analysis resource for an experiment in text mining a collection of micro-blogs on a political topic.
The analysis of a corpus of micro-blogs on the topic of the 2011 UK referendum about the Alternative Vote has been undertaken as a joint activity by text miners and social scientists. To facilitate the collaboration, the corpus and its analysis is managed in a Web-accessible framework that allows users to upload their own textual data for analysis and to manage their own text annotation resources used for analysis. The framework also allows annotations to be searched, and the analysis to be re-run after amending the analysis resources. The corpus is also doubly human-annotated stating both whether each tweet is overall positive or negative in sentiment and whether it is for or against the proposition of the referendum.
['Steven Gray', 'Sophia Ananiadou', 'William Black', 'Rob Procter']
2012-05-01
null
null
null
lrec-2012-5
['text-annotation']
['natural-language-processing']
[ 2.37909764e-01 7.78519094e-01 -5.54578662e-01 -3.65138024e-01 -9.40365016e-01 -8.84873271e-01 9.22624826e-01 1.05699551e+00 -6.11921966e-01 8.15602958e-01 1.02531660e+00 -8.54340017e-01 1.09770358e-01 -3.95469069e-01 -2.17270091e-01 -3.48986804e-01 4.54419851e-01 5.60045898e-01 1.56143352e-01 -1.20413877e-01 6.72637701e-01 -3.77078503e-01 -1.19081926e+00 2.04930767e-01 4.36260253e-01 7.79351473e-01 1.58857167e-01 3.65385830e-01 -2.31957123e-01 1.05731142e+00 -5.94278038e-01 -7.23657250e-01 1.25659198e-01 -1.98378444e-01 -1.10928476e+00 -2.15720788e-01 2.27603540e-01 2.88433909e-01 3.57570022e-01 1.06030810e+00 3.05590659e-01 -2.00171083e-01 3.35094988e-01 -8.52220893e-01 7.91842565e-02 7.80147314e-01 -3.39503467e-01 2.98866272e-01 5.34308076e-01 -1.68482527e-01 1.42022145e+00 -6.36037827e-01 1.13377869e+00 1.11927521e+00 4.89600003e-01 -5.55297099e-02 -6.36464059e-01 -8.59974682e-01 -1.94485724e-01 -1.22592367e-01 -7.47490525e-01 -5.63975394e-01 3.79963100e-01 -7.86698103e-01 8.61960709e-01 4.21034157e-01 7.85921276e-01 8.14162433e-01 3.81161243e-01 2.35199273e-01 1.17154717e+00 -8.14877272e-01 2.46558696e-01 2.99637735e-01 2.67893910e-01 1.41374201e-01 3.86254251e-01 -6.36957049e-01 -4.82271641e-01 -5.59511602e-01 -2.08533153e-01 -4.75650698e-01 2.55402625e-01 1.58162713e-01 -1.04805756e+00 1.08102083e+00 1.28927454e-01 5.81075609e-01 -5.74899077e-01 -1.58252001e-01 8.90553534e-01 2.42317140e-01 1.11521554e+00 2.39927039e-01 -6.42091751e-01 -4.25818026e-01 -1.00999820e+00 4.26374227e-01 1.19282138e+00 2.31695428e-01 5.02983332e-01 -7.02694476e-01 2.16140196e-01 8.02943408e-01 9.20433223e-01 1.09744616e-01 4.18152243e-01 -9.01639581e-01 6.95529222e-01 1.11732388e+00 4.58705276e-01 -1.01458037e+00 -3.01833034e-01 -1.64276823e-01 -3.89971167e-01 1.27035335e-01 4.93526220e-01 -3.62465411e-01 -3.74877453e-01 1.10687280e+00 5.99150062e-01 -6.55532062e-01 6.67319596e-02 5.35082817e-01 1.02455723e+00 3.84584278e-01 4.15240854e-01 -4.33974326e-01 1.65885639e+00 -3.22441489e-01 -1.01871192e+00 -1.10023715e-01 9.37697828e-01 -1.30520427e+00 5.16260028e-01 1.69089690e-01 -1.03799760e+00 -3.46997404e-04 -1.09239042e+00 -2.74369180e-01 -3.96920145e-01 -4.15154129e-01 2.79900730e-01 3.62039149e-01 -8.27644408e-01 3.01872402e-01 -5.68894207e-01 -6.44238770e-01 4.99319851e-01 1.21555157e-01 -2.12483466e-01 4.45273221e-01 -1.34586835e+00 1.13159680e+00 5.13709128e-01 -3.52046862e-02 8.55622962e-02 -4.11195815e-01 -9.06812787e-01 -4.79489744e-01 4.62582469e-01 -2.44141370e-01 1.33768737e+00 -1.08275247e+00 -9.32052195e-01 1.53723216e+00 -3.46471399e-01 -5.35631597e-01 6.91805005e-01 4.04982939e-02 -6.34436667e-01 -2.72561401e-01 9.64256108e-01 -6.58994392e-02 4.09911036e-01 -5.43877482e-01 -1.15146434e+00 -6.03298187e-01 -1.05379291e-01 3.72606814e-02 -2.53951520e-01 8.49097729e-01 -9.50519890e-02 -5.46349823e-01 2.81178635e-02 -8.93321633e-01 4.39456068e-02 -6.06655180e-01 -4.78689700e-01 -6.92552924e-01 1.04024839e+00 -8.26096654e-01 1.51917672e+00 -1.81856143e+00 -4.62453246e-01 3.57197255e-01 2.15141997e-01 -1.78909302e-01 6.58152819e-01 1.01824236e+00 8.37825164e-02 6.35986686e-01 1.48087561e-01 -2.17972204e-01 1.04223289e-01 2.47720867e-01 -4.44026321e-01 7.88212836e-01 -3.55687201e-01 5.28656185e-01 -9.71300125e-01 -7.64754713e-01 -1.40517160e-01 5.83088212e-02 -4.26326059e-02 -6.77437007e-01 -6.44657481e-03 3.52156758e-01 -5.41381478e-01 4.44679320e-01 3.39121938e-01 -3.39142442e-01 8.67384732e-01 -1.04593106e-01 -9.36647654e-01 1.05677354e+00 -9.56153691e-01 9.76672173e-01 -3.07426631e-01 1.19994867e+00 3.67463022e-01 -5.87681711e-01 9.27120209e-01 5.29472113e-01 2.23795116e-01 -5.07441401e-01 3.85311842e-01 6.79130614e-01 -1.81472078e-01 -1.77521154e-01 9.72527802e-01 -8.24540108e-02 -2.97574610e-01 1.00156283e+00 -4.83365476e-01 3.31432782e-02 4.95697021e-01 4.12931532e-01 8.60065460e-01 1.12810932e-01 5.99601388e-01 -6.55665576e-01 6.94073498e-01 2.96104878e-01 5.50065100e-01 3.05614710e-01 1.60182580e-01 -2.32877240e-01 5.34088612e-01 -4.81271714e-01 -1.30532920e+00 -6.49398342e-02 -5.93700767e-01 1.18963754e+00 -1.46234185e-01 -8.40155244e-01 -3.65821749e-01 -5.36107719e-01 -2.19311297e-01 7.14415133e-01 -7.84276128e-01 8.98242235e-01 -2.73135245e-01 -4.35090154e-01 7.31912479e-02 -6.92072213e-02 4.17596191e-01 -1.04592657e+00 -1.03515196e+00 2.09227771e-01 -5.93667328e-01 -7.11146235e-01 -5.44278286e-02 3.47469956e-01 -2.97069907e-01 -1.30584240e+00 -7.68356696e-02 -7.19172001e-01 4.84062046e-01 -3.87087643e-01 7.98166633e-01 2.36634478e-01 3.16565007e-01 -1.70542583e-01 -4.02492017e-01 -9.89188671e-01 -7.64903486e-01 3.33951622e-01 -3.12712312e-01 -3.12646359e-01 4.47656214e-01 -2.59650856e-01 -3.14775735e-01 1.26248896e-01 -9.12245095e-01 -8.03914741e-02 -1.62078291e-01 5.19000292e-01 2.39965588e-01 -4.96910140e-02 7.94120729e-01 -1.29850590e+00 8.56193304e-01 -8.48249972e-01 -5.66418529e-01 -1.73874646e-01 -8.50306451e-01 -5.60726762e-01 -1.36867046e-01 4.04767871e-01 -8.79025459e-01 -3.95199984e-01 2.61045303e-02 7.08176911e-01 1.85304448e-01 1.26265168e+00 7.17025027e-02 5.21162212e-01 5.05817175e-01 -5.94281375e-01 5.13575710e-02 -4.05332357e-01 1.47519380e-01 1.06449020e+00 4.55664158e-01 -3.26860324e-02 8.07400286e-01 3.54210615e-01 -3.30957472e-01 -6.54105842e-01 -1.10772169e+00 -8.37815404e-01 -7.56654263e-01 -5.13978839e-01 7.95466244e-01 -1.00631058e+00 -6.15723312e-01 2.81358033e-01 -8.58922958e-01 -3.94880414e-01 -7.98472837e-02 2.63462245e-01 1.09058954e-01 3.10536385e-01 -1.16183102e-01 -9.81474876e-01 -4.03967589e-01 -5.82021534e-01 3.48092854e-01 4.27828617e-02 -1.37493253e+00 -1.41141462e+00 4.12957013e-01 8.11595321e-01 2.39236340e-01 6.48777127e-01 7.04597712e-01 -1.29289150e+00 3.19736570e-01 -6.43181443e-01 -7.86413997e-02 -2.67563127e-02 5.44664711e-02 4.27309364e-01 -6.42894089e-01 -1.16253272e-01 4.05291170e-02 -5.83944559e-01 3.15811962e-01 -2.98342537e-02 3.88690174e-01 -9.84672070e-01 -4.29175436e-01 -4.35556889e-01 1.24308228e+00 -1.59730986e-01 5.54418862e-01 1.30044711e+00 -3.22666243e-02 9.28634942e-01 6.52693689e-01 5.44949830e-01 6.16864741e-01 6.53835654e-01 3.80736619e-01 3.24554443e-01 3.98623109e-01 -6.55971244e-02 2.81684756e-01 7.27950811e-01 3.15510742e-02 -3.23182762e-01 -1.10372925e+00 8.77741694e-01 -2.06225705e+00 -1.12430465e+00 -1.04041255e+00 1.85818636e+00 6.80990696e-01 5.34631371e-01 1.02699280e-01 5.93203425e-01 8.30020308e-01 4.64750171e-01 1.07297219e-01 -6.06554985e-01 -1.55102491e-01 2.61694968e-01 6.35081947e-01 6.47813141e-01 -7.70681798e-01 7.50059545e-01 6.10765648e+00 4.81504858e-01 -6.05515540e-01 3.62934798e-01 6.82908475e-01 -5.04352190e-02 -5.38322628e-01 5.41173577e-01 -6.67025745e-01 4.61503357e-01 9.97821629e-01 -5.03408551e-01 -3.98352325e-01 5.83486736e-01 6.04607105e-01 -8.06947589e-01 -4.01554078e-01 1.37198836e-01 -2.07726777e-01 -1.76252520e+00 -5.40899515e-01 4.98264015e-01 8.66229415e-01 4.25704330e-01 -3.52626830e-01 -1.56160453e-02 6.75029993e-01 -6.79797590e-01 1.38836539e+00 1.65547073e-01 6.99598849e-01 -3.32707614e-01 1.02179849e+00 5.00225306e-01 -8.57814014e-01 -5.36108436e-03 1.81636363e-01 -6.35338604e-01 4.28892970e-01 6.31318569e-01 -9.24056053e-01 5.19977510e-01 8.08685780e-01 6.42174661e-01 -4.58206594e-01 4.30533201e-01 -4.11181867e-01 9.33858156e-01 -2.60100514e-01 -3.02769810e-01 4.14005160e-01 -9.35002416e-02 7.74235249e-01 1.16065514e+00 -8.01984742e-02 -1.30749688e-01 1.37790386e-02 1.87294781e-01 -1.95497215e-01 4.51589674e-01 -1.46832377e-01 -1.80062830e-01 6.49085641e-01 1.57972610e+00 -1.03443229e+00 -5.21583140e-01 -2.55049318e-01 2.45457098e-01 6.68023829e-04 -5.21452427e-02 -1.60067886e-01 -3.79701525e-01 1.31836891e-01 6.55740917e-01 2.79936522e-01 1.07363999e-01 -3.52805972e-01 -6.27384186e-01 7.07191899e-02 -5.87636650e-01 6.37262702e-01 -7.80333042e-01 -8.11917663e-01 5.06805718e-01 -1.46827504e-01 -8.02772641e-01 -3.96632910e-01 -6.28598854e-02 -8.99838507e-01 8.90798390e-01 -1.02539217e+00 -8.38774025e-01 -2.04458367e-02 2.13535830e-01 9.76368263e-02 2.69336440e-02 6.92225039e-01 9.91709530e-02 -2.86576152e-01 -1.35734499e-01 -2.12344085e-03 1.35575250e-01 6.97348714e-01 -1.11269283e+00 1.68540508e-01 4.45084363e-01 -2.85183191e-01 5.27934372e-01 1.03396177e+00 -1.24012709e+00 -5.84555089e-01 -7.87305057e-01 2.02463508e+00 -6.66515946e-01 1.47218204e+00 -9.99969691e-02 -6.02403462e-01 5.94300747e-01 6.80518985e-01 -5.06726563e-01 1.21725059e+00 2.72157282e-01 1.50934890e-01 2.90071726e-01 -9.56604421e-01 1.98883191e-01 2.18443513e-01 -5.37027776e-01 -1.03064978e+00 5.59247255e-01 7.60061070e-02 -2.53000796e-01 -7.94201732e-01 8.66077188e-03 7.75126874e-01 -4.78232592e-01 4.23143879e-02 -3.65643680e-01 6.06116056e-01 -4.32556689e-01 -1.78531840e-01 -6.35835230e-01 8.24931860e-02 -8.47445786e-01 5.02448261e-01 1.54179835e+00 7.22303033e-01 -6.73570573e-01 3.94926906e-01 7.75304019e-01 -6.12030290e-02 -3.17249477e-01 -1.20590270e+00 3.62180322e-02 -2.86261588e-01 -8.46642375e-01 1.39909327e-01 1.20232069e+00 6.43496215e-01 5.79394758e-01 -5.00711380e-03 -4.88668263e-01 2.39174366e-01 -1.20062754e-01 8.83687615e-01 -1.62639034e+00 3.98251534e-01 -5.13976812e-01 -8.73909891e-02 -2.96343774e-01 5.11576869e-02 -1.15952432e+00 -2.99001038e-01 -1.87485659e+00 1.15206204e-01 -3.43505591e-01 5.96139356e-02 4.62255299e-01 3.69154066e-01 4.53671753e-01 3.64719331e-02 1.02839184e+00 -5.85684955e-01 -8.96347836e-02 7.13294029e-01 9.15980190e-02 -2.54412830e-01 -9.61952955e-02 -1.13305795e+00 9.36028779e-01 5.92212260e-01 -6.83657706e-01 2.69070297e-01 2.41155684e-01 1.38993883e+00 -3.22285354e-01 4.26208287e-01 -4.65512991e-01 2.47263774e-01 4.68184240e-03 1.65422231e-01 -8.91163945e-01 -2.38465935e-01 -6.08120501e-01 4.89748776e-01 5.28682530e-01 -6.14810705e-01 5.17123081e-02 1.80059984e-01 4.92075831e-01 -3.76696408e-01 -5.22515595e-01 3.38534415e-01 -1.30292118e-01 -1.36246100e-01 -1.58748120e-01 -8.01094830e-01 2.60381788e-01 8.17593038e-01 -3.41705441e-01 -4.08879042e-01 -6.44717634e-01 -7.49098361e-01 3.32908690e-01 7.08380401e-01 1.97773919e-01 -2.35115632e-01 -1.22299004e+00 -9.81448829e-01 -5.33454299e-01 3.81838754e-02 -1.11040972e-01 -4.43768799e-02 1.07774198e+00 -5.08737624e-01 3.41103107e-01 1.56593487e-01 3.48207206e-02 -1.28393638e+00 -4.36140858e-02 -2.68220633e-01 -7.52417624e-01 -5.93506455e-01 2.17960462e-01 -2.31635541e-01 -1.73986137e-01 -1.63219854e-01 -1.08940385e-01 -7.86157250e-01 9.51079488e-01 5.34098506e-01 4.48782235e-01 4.40448970e-02 -1.27845466e+00 -4.16960537e-01 -1.53560877e-01 1.58523411e-01 -5.15751779e-01 1.85739243e+00 -6.03314757e-01 -6.52893603e-01 9.64018166e-01 9.51735914e-01 6.01086259e-01 -6.14405990e-01 -1.26027033e-01 4.88985449e-01 -1.75299451e-01 1.64886862e-01 -9.50036943e-01 -4.94437039e-01 -1.01135209e-01 -2.07236648e-01 9.70084667e-01 2.53932804e-01 3.09148937e-01 4.15150583e-01 -6.19239137e-02 -4.24860448e-01 -1.49781632e+00 -4.58273828e-01 6.70097828e-01 1.11991930e+00 -1.00493407e+00 5.93864322e-01 -2.67412633e-01 -5.87062836e-01 1.35584199e+00 -1.12831607e-01 1.69393480e-01 8.93583477e-01 1.66518763e-01 2.43457630e-01 -8.87827754e-01 -6.57751381e-01 3.04235131e-01 3.14409435e-01 9.83655732e-03 8.07874680e-01 -1.32704735e-01 -1.14837444e+00 7.25305319e-01 -7.20996797e-01 -9.15861428e-02 5.84161639e-01 7.51298189e-01 -3.19544286e-01 -1.19053257e+00 -3.82098377e-01 5.66379488e-01 -1.26365626e+00 -1.02393478e-01 -7.08462596e-01 9.75530148e-01 1.00088678e-01 1.07080495e+00 2.58009672e-01 -6.44777715e-02 -8.75911415e-02 1.12652078e-01 -2.48782530e-01 -5.50358295e-01 -1.13752425e+00 1.52645230e-01 1.08062339e+00 9.90299042e-03 -1.07121122e+00 -1.10298598e+00 -1.26746833e+00 -2.73188263e-01 -5.72644830e-01 5.21088362e-01 1.06211019e+00 1.31205845e+00 -8.64783488e-03 8.38992279e-03 3.15255523e-01 -3.44989538e-01 1.03797287e-01 -1.26647997e+00 -3.31302524e-01 1.39183961e-02 -7.29899621e-03 -2.79462159e-01 -5.21662474e-01 1.80620253e-01]
[9.101018905639648, 9.798480033874512]
92bb9a5b-5189-49e5-983f-bc7d37a3801b
unsupervised-gaze-prediction-in-egocentric
2001.11580
null
https://arxiv.org/abs/2001.11580v2
https://arxiv.org/pdf/2001.11580v2.pdf
Unsupervised Gaze Prediction in Egocentric Videos by Energy-based Surprise Modeling
Egocentric perception has grown rapidly with the advent of immersive computing devices. Human gaze prediction is an important problem in analyzing egocentric videos and has primarily been tackled through either saliency-based modeling or highly supervised learning. We quantitatively analyze the generalization capabilities of supervised, deep learning models on the egocentric gaze prediction task on unseen, out-of-domain data. We find that their performance is highly dependent on the training data and is restricted to the domains specified in the training annotations. In this work, we tackle the problem of jointly predicting human gaze points and temporal segmentation of egocentric videos without using any training data. We introduce an unsupervised computational model that draws inspiration from cognitive psychology models of event perception. We use Grenander's pattern theory formalism to represent spatial-temporal features and model surprise as a mechanism to predict gaze fixation points. Extensive evaluation on two publicly available datasets - GTEA and GTEA+ datasets-shows that the proposed model can significantly outperform all unsupervised baselines and some supervised gaze prediction baselines. Finally, we show that the model can also temporally segment egocentric videos with a performance comparable to more complex, fully supervised deep learning baselines.
['Arunkumar Bagavathi', 'Sathyanarayanan N. Aakur']
2020-01-30
null
null
null
null
['eye-tracking']
['computer-vision']
[ 2.02785566e-01 2.22694069e-01 -2.27298081e-01 -5.00133276e-01 -1.51454851e-01 -1.91322371e-01 3.61443371e-01 -1.51451141e-01 -4.38835561e-01 3.72799903e-01 2.41436616e-01 4.66168206e-03 -1.48045510e-01 -1.74791202e-01 -8.65842819e-01 -6.17214322e-01 -6.85335919e-02 7.04953671e-02 4.15011644e-01 -1.87889159e-01 7.25992560e-01 -4.53583375e-02 -2.24897027e+00 2.12293401e-01 7.91348279e-01 1.05292392e+00 2.97896624e-01 8.42477441e-01 3.76431346e-01 1.14976144e+00 -7.91514888e-02 -1.45181865e-01 1.04886275e-02 -4.89530116e-01 -1.00259805e+00 -1.17447106e-02 5.81965089e-01 -6.66808262e-02 -2.71569490e-01 8.96350086e-01 3.72739345e-01 4.60708469e-01 8.77141953e-01 -1.53780830e+00 -5.68207145e-01 1.17021956e-01 -8.55197847e-01 8.37432384e-01 4.88386214e-01 1.70941696e-01 1.02325797e+00 -7.65767157e-01 9.63654876e-01 9.48092520e-01 4.27368730e-01 7.83720016e-01 -1.09978008e+00 -3.24761301e-01 2.92142600e-01 9.40051734e-01 -1.14546871e+00 -4.89662021e-01 8.73944223e-01 -7.31521130e-01 1.04299355e+00 1.13873526e-01 5.00793040e-01 1.25286865e+00 2.60069042e-01 1.15393806e+00 9.21588600e-01 -5.62596679e-01 2.77267098e-01 2.67247157e-03 1.10296711e-01 7.08558381e-01 -2.93401122e-01 1.81821793e-01 -9.83418286e-01 4.45255727e-01 3.78047585e-01 6.84627369e-02 -3.95114690e-01 -8.00042868e-01 -1.20060563e+00 6.82115018e-01 8.61129284e-01 4.34035324e-02 -4.22271371e-01 1.81306183e-01 3.64850640e-01 -3.26584466e-02 8.31174076e-01 4.86767530e-01 -4.00144160e-01 -3.32504153e-01 -9.79211688e-01 2.94492573e-01 3.45232159e-01 1.12689483e+00 9.27729666e-01 -1.87854528e-01 -2.10150853e-01 4.19882476e-01 3.00939858e-01 3.51498753e-01 6.23156846e-01 -1.10053718e+00 9.27288234e-02 4.13472980e-01 1.64807931e-01 -1.05090642e+00 -7.37667799e-01 -3.75286415e-02 -1.34456456e-01 1.77440181e-01 5.47966719e-01 -2.42919400e-01 -8.19234490e-01 1.93535316e+00 2.25108922e-01 4.27826911e-01 -1.56575486e-01 1.33139610e+00 7.90678203e-01 2.31956646e-01 4.15075183e-01 -2.56280065e-01 1.22188139e+00 -1.02073836e+00 -6.69137716e-01 -1.46673217e-01 8.71657431e-01 -4.54136431e-01 1.24130881e+00 3.44729304e-01 -1.17332542e+00 -6.03378475e-01 -7.69495547e-01 -4.79334652e-01 -4.13455606e-01 -7.27183819e-02 5.32198012e-01 5.72921693e-01 -1.29578090e+00 4.27578509e-01 -7.28194833e-01 -8.95931482e-01 6.11340761e-01 4.78311986e-01 -1.08309187e-01 2.89218485e-01 -8.78898799e-01 7.40343034e-01 3.71601999e-01 -1.78070188e-01 -8.49579751e-01 -7.35843718e-01 -9.90553319e-01 3.11044544e-01 3.96148235e-01 -8.42514932e-01 1.49007165e+00 -1.48798585e+00 -1.36712623e+00 1.17963290e+00 -8.39290559e-01 -6.85154796e-01 1.66806951e-01 -3.67389858e-01 -2.66738087e-01 3.82923365e-01 6.93759993e-02 1.27285779e+00 1.09127736e+00 -9.19240236e-01 -8.70155156e-01 -4.46228862e-01 2.77988732e-01 3.35534036e-01 -2.74793148e-01 1.53720841e-01 -2.18232095e-01 -3.04066002e-01 -1.67117029e-01 -1.12552488e+00 4.06386256e-02 -3.73383671e-01 -3.64470333e-01 -6.22715533e-01 8.76638412e-01 -2.56721765e-01 1.17802680e+00 -2.08143449e+00 3.93458545e-01 -2.23955676e-01 6.08703434e-01 -9.51054096e-02 1.04076207e-01 -4.25849855e-02 -4.52584684e-01 -1.54934809e-01 1.83494180e-01 -5.74448287e-01 6.96499646e-02 -2.66154826e-01 -2.84687638e-01 4.16987687e-01 1.88858226e-01 1.19164908e+00 -1.17885315e+00 -6.07691765e-01 1.84587449e-01 1.36932313e-01 -8.95309210e-01 1.25182018e-01 -3.92857611e-01 4.99546885e-01 -1.03989176e-01 5.74519932e-01 3.31941873e-01 -5.54679334e-01 -8.17957819e-02 -1.10689268e-01 -1.94657892e-01 2.33866408e-01 -2.38600075e-01 1.95537198e+00 -7.86692128e-02 1.44021833e+00 -7.60455310e-01 -9.28030372e-01 4.68688190e-01 6.19453043e-02 6.10688865e-01 -9.53256190e-01 4.05724227e-01 -2.58127034e-01 1.06247270e-03 -9.91772115e-01 8.32604885e-01 1.47614539e-01 4.35134843e-02 6.11715734e-01 4.84764785e-01 4.66869801e-01 3.84714693e-01 2.10426554e-01 7.03961194e-01 6.38801932e-01 1.40072331e-01 -4.21179116e-01 3.99709195e-01 2.26688042e-01 3.94736081e-01 5.80050945e-01 -7.29292214e-01 7.62176812e-01 8.77920806e-01 -2.76349813e-01 -9.32816029e-01 -7.91960359e-01 9.25281718e-02 1.79715073e+00 5.79380751e-01 -5.39229453e-01 -1.20297444e+00 -8.11369061e-01 -4.51540679e-01 8.42371047e-01 -1.08676434e+00 -3.01459581e-01 -4.52470720e-01 -5.10253191e-01 4.23904434e-02 6.40895665e-01 1.59281641e-01 -1.37700200e+00 -9.88253832e-01 -4.26269859e-01 -3.12618017e-01 -1.09399951e+00 -2.79975504e-01 -1.00625753e-01 -6.44189239e-01 -1.37309611e+00 -7.24694014e-01 -7.70263731e-01 6.13339245e-01 5.75470865e-01 1.17652977e+00 -3.17064315e-01 -1.45349398e-01 8.61156225e-01 -3.57861042e-01 -8.55256498e-01 3.29592913e-01 2.80436069e-01 3.05539727e-01 1.84909701e-01 1.36352074e+00 -3.40446204e-01 -1.02612281e+00 3.64578098e-01 -3.33517522e-01 1.48809791e-01 2.30688229e-02 4.92977291e-01 3.10781240e-01 -6.42142951e-01 4.18286860e-01 -9.25039649e-01 2.82726437e-01 -8.16217363e-01 -4.13163275e-01 -8.82116407e-02 -4.43876624e-01 -2.01805353e-01 7.01512694e-02 -2.11568132e-01 -1.22837567e+00 -1.66855752e-02 3.97132516e-01 -8.13680768e-01 -5.55025876e-01 3.90017569e-01 2.02993363e-01 3.47864367e-02 9.54592466e-01 4.51349467e-03 -1.03664331e-01 -1.98482916e-01 2.81176537e-01 4.50643599e-01 4.77350414e-01 1.14759319e-02 3.37805450e-01 7.03543901e-01 -1.99444639e-03 -8.86508226e-01 -1.04959249e+00 -6.58338070e-01 -9.70359564e-01 -6.38718426e-01 1.26788425e+00 -9.24024343e-01 -1.13554716e+00 2.68822104e-01 -9.50504482e-01 -5.32763898e-01 -3.30850989e-01 5.45398831e-01 -1.15483999e+00 2.76699394e-01 -1.26647130e-01 -7.23964036e-01 6.99800476e-02 -9.11755860e-01 1.16273892e+00 6.03591800e-01 -3.79682004e-01 -1.18923426e+00 1.81238905e-01 3.09945405e-01 2.21548639e-02 1.30678430e-01 5.77411175e-01 -5.27742028e-01 -8.16559494e-01 1.63846493e-01 -3.43409866e-01 -9.20144245e-02 -1.22896150e-01 -2.54954677e-03 -1.25627494e+00 -1.75822061e-02 9.71354693e-02 -3.91835719e-01 1.00042677e+00 8.02902520e-01 1.36047208e+00 9.11391675e-02 -5.79383433e-01 7.39523828e-01 1.04835463e+00 -6.24183640e-02 7.24345267e-01 3.99769843e-01 8.19228649e-01 1.06163144e+00 1.01336348e+00 2.23131433e-01 7.45056033e-01 4.33047265e-01 6.80633783e-01 1.50828913e-01 2.18197033e-01 -1.45251960e-01 1.55784786e-01 2.49198571e-01 -4.88607466e-01 -4.41722125e-01 -1.00090468e+00 7.74168909e-01 -2.22063136e+00 -1.37494516e+00 -2.57582635e-01 1.79297948e+00 2.42220178e-01 2.33415589e-02 5.58326364e-01 -3.80947329e-02 6.88533068e-01 1.25388414e-01 -6.72254503e-01 -5.30284107e-01 9.87780616e-02 -1.49090542e-02 3.40201169e-01 1.33338764e-01 -1.36315024e+00 1.21040940e+00 6.68572521e+00 1.84821695e-01 -1.30488324e+00 2.87755728e-01 3.99652541e-01 -4.97419357e-01 1.66735142e-01 -1.71145156e-01 -6.90352023e-01 6.29086494e-01 1.09214973e+00 -1.76667482e-01 2.01010481e-01 9.94579077e-01 5.43492973e-01 -4.32261378e-01 -1.48731422e+00 1.20652437e+00 4.83629048e-01 -1.23041415e+00 -4.03119981e-01 1.09275900e-01 9.79232192e-01 3.30925345e-01 5.00155807e-01 2.83465475e-01 -9.99802426e-02 -9.80283439e-01 6.48043215e-01 9.76250768e-01 5.33737421e-01 -4.63741392e-01 4.24935341e-01 3.22360218e-01 -6.63445950e-01 -3.87102336e-01 -4.03928280e-01 -2.16342971e-01 1.23659670e-01 -2.90357500e-01 -8.41630638e-01 8.14123675e-02 1.26714516e+00 1.33816874e+00 -9.01153564e-01 1.30719483e+00 -1.55230045e-01 5.55789888e-01 -4.17451411e-02 -1.24930546e-01 2.53055245e-01 1.29542276e-01 6.58378243e-01 9.99198198e-01 8.33211616e-02 8.38633068e-03 -4.81606454e-01 9.16374683e-01 8.06868970e-02 4.00530919e-03 -5.10638833e-01 1.67496726e-01 -1.61130562e-01 9.33376014e-01 -5.28292775e-01 -2.51896959e-02 -5.16846836e-01 7.88365424e-01 5.32135248e-01 7.60951817e-01 -8.80735755e-01 -3.97140086e-01 7.31975198e-01 3.27846110e-01 5.96282840e-01 -1.16025042e-02 -2.27478608e-01 -1.35077441e+00 -2.23882243e-01 -2.92327255e-01 3.82728666e-01 -1.39447355e+00 -9.80324984e-01 3.76807630e-01 6.79390505e-02 -1.44191599e+00 -8.48673940e-01 -7.21327484e-01 -7.16855586e-01 7.97826469e-01 -1.72378111e+00 -1.16183662e+00 -7.77332187e-01 8.26526642e-01 7.35648632e-01 -2.49211952e-01 3.87341321e-01 -9.61026922e-02 -6.60519719e-01 5.00339150e-01 -5.34992926e-02 -1.22531615e-01 8.90919805e-01 -1.52815115e+00 2.20136359e-01 8.98658216e-01 1.85481776e-02 5.51121593e-01 9.85239923e-01 -2.11252496e-01 -8.45970571e-01 -8.25004101e-01 9.99898255e-01 -9.90374386e-01 5.92601418e-01 -4.15476650e-01 -7.46337771e-01 1.10653627e+00 4.69048858e-01 -1.26445413e-01 9.24315929e-01 3.42338145e-01 -4.33563143e-02 2.01176003e-01 -7.55706251e-01 6.93956614e-01 1.30338049e+00 -6.37623847e-01 -7.92014122e-01 3.45295906e-01 5.17547548e-01 -3.78763467e-01 -3.12440515e-01 3.42779279e-01 4.68194634e-01 -1.40238237e+00 7.24473953e-01 -1.12730217e+00 7.96435118e-01 -1.06489711e-01 2.80819029e-01 -1.09398305e+00 -3.89124304e-01 -6.80544198e-01 -4.84393924e-01 8.16322803e-01 1.34280354e-01 -2.83897430e-01 1.13472867e+00 6.21478796e-01 -8.18063095e-02 -4.48948264e-01 -6.19882464e-01 -2.33866125e-01 -2.25001946e-01 -4.15482342e-01 2.48650610e-01 8.38430047e-01 3.99852306e-01 4.08128083e-01 -1.89196736e-01 -4.34716865e-02 5.75690746e-01 -4.49661873e-02 1.05009687e+00 -1.26469612e+00 2.86412269e-01 -4.83042330e-01 -8.74244809e-01 -1.24232042e+00 5.11232555e-01 -4.99080360e-01 1.38061005e-03 -1.03968549e+00 1.27706274e-01 1.60462726e-02 -5.98314166e-01 9.17177498e-02 -1.39998823e-01 5.65829635e-01 3.52153145e-02 2.90930510e-01 -1.15229273e+00 4.37488228e-01 1.09269953e+00 2.10098445e-01 -3.66209775e-01 2.32250988e-02 -6.56970561e-01 1.03143084e+00 6.26443088e-01 -2.86936462e-01 -7.26797640e-01 -3.91530752e-01 4.90515381e-01 -3.38519186e-01 7.00544417e-01 -9.94887471e-01 5.43144822e-01 -1.30444154e-01 2.61565566e-01 -7.41614580e-01 3.71712893e-01 -5.53729355e-01 -7.47533083e-01 -1.40323684e-01 -4.17951465e-01 -1.69126138e-01 1.06247380e-01 8.36503804e-01 -1.38940707e-01 -1.40897661e-01 6.22920752e-01 1.76816955e-01 -1.26484799e+00 2.25850388e-01 -3.89481515e-01 1.36301994e-01 1.29751527e+00 -6.50963008e-01 -4.47719395e-01 -4.59192961e-01 -1.10451245e+00 3.76620948e-01 6.78631723e-01 7.87942946e-01 4.44873542e-01 -8.61806929e-01 -2.48899400e-01 1.19032390e-01 4.60055828e-01 -2.67243564e-01 4.48939443e-01 1.41503215e+00 -4.65260625e-01 7.84843326e-01 -5.01421332e-01 -1.30952382e+00 -1.16152012e+00 1.15652680e+00 3.17572236e-01 4.42657501e-01 -2.51023144e-01 9.04513955e-01 9.06021297e-01 1.53559148e-01 8.87427852e-02 -3.20466846e-01 -8.37097883e-01 1.95894808e-01 4.01551485e-01 4.89998937e-01 -2.70287126e-01 -1.06741560e+00 -2.94812679e-01 5.21706820e-01 -1.57024205e-01 1.95294723e-01 1.17190087e+00 -6.23761475e-01 1.54363185e-01 6.06920063e-01 1.04575491e+00 -3.82154584e-01 -1.44833076e+00 -2.45032936e-01 1.59509525e-01 -5.10801077e-01 1.21333979e-01 -3.41737747e-01 -8.05867791e-01 1.27362216e+00 7.25330830e-01 2.93079615e-01 1.20524323e+00 1.57518372e-01 5.35806358e-01 2.63477117e-01 7.68676177e-02 -1.14250576e+00 3.72558743e-01 4.70441550e-01 4.59204555e-01 -1.68966675e+00 -3.16230685e-01 -3.05686325e-01 -1.01453066e+00 8.39283228e-01 9.47497189e-01 -3.66443723e-01 8.36483240e-01 -6.84681892e-01 -7.04308674e-02 -3.76326561e-01 -8.00430357e-01 -7.16563225e-01 7.01682985e-01 8.12260091e-01 3.60096872e-01 -3.20630968e-01 1.72406454e-02 5.67544043e-01 -3.53295296e-01 2.81251639e-01 5.30786932e-01 7.60447323e-01 -3.69445622e-01 -3.66156399e-01 -6.07577600e-02 3.51534754e-01 -4.74404633e-01 -3.78374266e-03 -1.86912373e-01 7.60273755e-01 1.02932960e-01 7.90207624e-01 6.93445742e-01 -3.35538894e-01 8.13429952e-02 1.45826086e-01 5.41892409e-01 -6.34428442e-01 -3.10047001e-01 -4.07139622e-02 -2.84893870e-01 -8.38252485e-01 -1.01062381e+00 -1.05414915e+00 -8.07832599e-01 -3.64502132e-01 -2.30793402e-01 -1.02899179e-01 4.66551572e-01 1.11375344e+00 6.51407361e-01 4.48717475e-01 4.76975381e-01 -1.29509389e+00 1.35948494e-01 -9.40787673e-01 -4.80438858e-01 3.86134505e-01 5.29915035e-01 -1.06076014e+00 -3.50533783e-01 5.14300942e-01]
[13.928386688232422, 0.05947059392929077]
a7ab6e05-7c22-48fa-b235-760b250343ad
a-comprehensive-review-of-image-line-segment
2305.00264
null
https://arxiv.org/abs/2305.00264v1
https://arxiv.org/pdf/2305.00264v1.pdf
A Comprehensive Review of Image Line Segment Detection and Description: Taxonomies, Comparisons, and Challenges
Detection and description of line segments lay the basis for numerous vision tasks. Although many studies have aimed to detect and describe line segments, a comprehensive review is lacking, obstructing their progress. This study fills the gap by comprehensively reviewing related studies on detecting and describing two-dimensional image line segments to provide researchers with an overall picture and deep understanding. Based on their mechanisms, two taxonomies for line segment detection and description are presented to introduce, analyze, and summarize these studies, facilitating researchers to learn about them quickly and extensively. The key issues, core ideas, advantages and disadvantages of existing methods, and their potential applications for each category are analyzed and summarized, including previously unknown findings. The challenges in existing methods and corresponding insights for potentially solving them are also provided to inspire researchers. In addition, some state-of-the-art line segment detection and description algorithms are evaluated without bias, and the evaluation code will be publicly available. The theoretical analysis, coupled with the experimental results, can guide researchers in selecting the best method for their intended vision applications. Finally, this study provides insights for potentially interesting future research directions to attract more attention from researchers to this field.
['Ce Zhu', 'Yipeng Liu', 'Yingjie Zhou', 'Xinyu Lin']
2023-04-29
null
null
null
null
['line-segment-detection']
['computer-vision']
[ 2.65275538e-01 -5.32717228e-01 -6.42162740e-01 -2.50668317e-01 -3.35042626e-01 -6.93443954e-01 8.78077149e-02 1.39540043e-02 -4.75365994e-03 4.20139790e-01 -2.83009589e-01 -3.06070447e-01 1.14193805e-01 -4.98778731e-01 -3.30265552e-01 -6.90064251e-01 -1.61539808e-01 -5.51148541e-02 5.30983210e-01 1.51409544e-02 9.30398703e-01 9.48970079e-01 -1.50964618e+00 4.35429662e-02 8.96420479e-01 9.18288469e-01 1.46037072e-01 5.48283875e-01 -2.65393078e-01 3.34627122e-01 -6.84767962e-01 -4.45422709e-01 3.07673186e-01 -6.42850101e-01 -4.40093040e-01 6.07086718e-01 8.31592679e-01 -4.41066891e-01 -2.86094517e-01 1.01321387e+00 6.46778107e-01 -1.20349303e-01 8.02151620e-01 -1.27430904e+00 -8.10680807e-01 4.36704844e-01 -1.06964469e+00 2.10025474e-01 6.05718195e-01 -3.03965062e-02 7.80709863e-01 -9.20521438e-01 4.22478259e-01 5.99938393e-01 9.32174444e-01 3.29747975e-01 -6.97126687e-01 -3.42832118e-01 2.05054149e-01 2.38407448e-01 -1.40313232e+00 -2.16640130e-01 1.08111870e+00 -4.28586751e-01 8.22545409e-01 3.42985332e-01 1.08967972e+00 6.76260889e-01 -6.73492700e-02 1.30274570e+00 9.36416149e-01 -7.11556613e-01 -1.67397961e-01 1.65805191e-01 5.42005599e-01 8.37931812e-01 5.56496143e-01 -4.93204221e-02 -4.13783073e-01 2.51133114e-01 1.30735183e+00 -3.84482920e-01 -4.07379836e-01 -8.39218855e-01 -1.28554654e+00 6.78168595e-01 4.00241524e-01 2.43018299e-01 -7.82198235e-02 -3.10098439e-01 3.78021896e-01 -6.76207840e-02 5.50294705e-02 4.63285983e-01 5.59968464e-02 -1.92366272e-01 -1.10709882e+00 2.28715822e-01 5.45225084e-01 1.59670711e+00 5.30870259e-01 8.85279998e-02 -1.59066662e-01 1.01577568e+00 2.30748668e-01 6.80503488e-01 1.30079433e-01 -8.13531280e-01 2.66082793e-01 3.01845700e-01 8.27084333e-02 -1.14405358e+00 -7.25404799e-01 -5.70155025e-01 -6.23356938e-01 8.25102702e-02 3.35228205e-01 -2.04372257e-01 -6.91109598e-01 8.33836913e-01 -2.55406704e-02 1.14554159e-01 -2.97049165e-01 1.07699811e+00 9.91395056e-01 6.08701825e-01 -6.03821456e-01 -4.31049526e-01 1.60273433e+00 -1.52698684e+00 -8.50348592e-01 -2.56077886e-01 3.64487767e-01 -1.38591588e+00 8.97839487e-01 2.52965093e-01 -1.25819361e+00 -5.73543966e-01 -1.30861723e+00 -2.44451746e-01 -2.29204167e-02 5.93277156e-01 1.00738621e+00 6.13564610e-01 -8.15217018e-01 1.26639128e-01 -8.54129314e-01 -8.91087532e-01 3.74188572e-01 2.26938799e-02 2.57759988e-01 8.15915223e-03 -7.16216803e-01 7.11958468e-01 5.62862456e-02 3.19899559e-01 1.07369453e-01 -3.28327924e-01 -7.44921982e-01 -4.92979854e-01 3.99075657e-01 -8.85479271e-01 1.50836682e+00 -3.73734355e-01 -1.42098033e+00 9.13644016e-01 -5.14439821e-01 -4.19866771e-01 5.55465996e-01 -1.32931903e-01 -3.61064970e-01 5.41132689e-01 1.98335901e-01 6.17860496e-01 5.33341885e-01 -1.38907683e+00 -1.07104707e+00 -1.31285831e-01 -8.64083022e-02 5.09635389e-01 -4.17277887e-02 2.24846169e-01 -1.18573642e+00 -7.83430994e-01 5.67457616e-01 -8.19786966e-01 -1.98552702e-02 2.91440606e-01 -8.32816601e-01 -2.47189164e-01 7.38876939e-01 -5.65563329e-02 1.60577536e+00 -1.80074561e+00 -4.63725716e-01 1.18232191e-01 1.09542318e-01 3.63855600e-01 1.41235128e-01 8.39552760e-01 7.60104656e-02 -6.85536191e-02 -2.36356676e-01 -2.16112778e-01 -3.33198786e-01 -3.52102667e-01 -4.33666378e-01 7.40419030e-01 -1.36270955e-01 7.80809402e-01 -6.41920030e-01 -5.42635441e-01 6.44862294e-01 2.91755706e-01 1.91841483e-01 -9.66185257e-02 3.84248316e-01 1.33196056e-01 -6.63333654e-01 9.09206927e-01 8.93608987e-01 -1.73419386e-01 -4.11133796e-01 -6.35030508e-01 -7.78043449e-01 -3.22430059e-02 -1.11376905e+00 1.35033834e+00 1.97742224e-01 1.15373790e+00 -4.17308241e-01 -9.26270306e-01 1.26086557e+00 -4.23262045e-02 6.71455920e-01 -6.29207671e-01 -3.68316695e-02 3.63545299e-01 -4.24291305e-02 -6.16945446e-01 5.79253852e-01 5.55021405e-01 6.67679533e-02 2.49383911e-01 -6.02266967e-01 -4.70653176e-01 5.74177861e-01 9.58021432e-02 3.27533245e-01 1.12586237e-01 5.86428642e-01 1.13866977e-01 6.17631674e-01 4.28687096e-01 3.07401687e-01 1.03035712e+00 -7.29410529e-01 6.91391230e-01 1.80021867e-01 -3.91662538e-01 -1.04133272e+00 -1.13357401e+00 -5.60422361e-01 4.90236431e-01 9.90281522e-01 -4.36524808e-01 -8.15824807e-01 -1.82021707e-01 -1.54463172e-01 3.59621942e-01 -3.65552306e-01 3.98298144e-01 -6.39339387e-01 -4.88093615e-01 6.97733164e-01 8.63981485e-01 9.03356373e-01 -8.34159732e-01 -7.46674299e-01 -1.17490292e-01 -2.31958658e-01 -1.18259501e+00 -5.28764248e-01 -4.14109975e-01 -1.11618876e+00 -1.29032147e+00 -1.21942449e+00 -1.47778010e+00 8.67121756e-01 1.28403342e+00 7.55345762e-01 9.51578170e-02 -4.66549784e-01 6.69108152e-01 -2.27223814e-01 -4.74967480e-01 3.49051803e-01 -1.44509941e-01 -2.01782566e-02 -4.08677429e-01 7.09417224e-01 3.62277776e-02 -8.04438233e-01 7.00141191e-01 -3.93240094e-01 3.46607476e-01 5.82528651e-01 5.65554440e-01 6.63124382e-01 -1.54718950e-01 2.54936725e-01 -7.60953605e-01 8.50239456e-01 2.55749226e-01 -8.24013650e-01 2.22148404e-01 -8.10778499e-01 -5.29792666e-01 2.08036259e-01 -1.40168622e-01 -9.58759248e-01 -6.31857961e-02 -2.57709380e-02 1.60017833e-01 -3.38500559e-01 2.97743410e-01 2.93549299e-01 -4.37353671e-01 6.51250422e-01 3.39594483e-01 -2.12075949e-01 -2.72771567e-01 4.97343183e-01 6.24323845e-01 8.33013773e-01 -1.45884424e-01 7.54815876e-01 7.20559537e-01 5.67326602e-03 -1.35848427e+00 -7.85272598e-01 -1.02046311e+00 -8.97835314e-01 -4.23643857e-01 5.72358072e-01 -5.11656344e-01 -5.19665003e-01 1.01255047e+00 -1.08775985e+00 -6.32084385e-02 1.96163207e-02 7.15062380e-01 -9.94021475e-01 9.61804330e-01 -8.52990627e-01 -7.53540933e-01 -2.37025488e-02 -1.24548542e+00 8.10778379e-01 9.49081242e-01 -7.89813101e-02 -1.14775395e+00 -6.67783991e-02 4.81288791e-01 2.56407022e-01 -3.02828420e-02 5.47262311e-01 -1.67406052e-01 -8.23727787e-01 -4.06937033e-01 -5.58472693e-01 -1.18587926e-01 6.81105852e-02 5.78747094e-01 -6.88373208e-01 -1.21952333e-01 -2.91094303e-01 -2.99937502e-02 7.38460362e-01 9.89571095e-01 1.05563641e+00 1.34513825e-01 -9.34856176e-01 6.92743719e-01 1.26433814e+00 5.26685536e-01 6.94400370e-01 6.61020517e-01 5.64997435e-01 6.52306914e-01 8.98173809e-01 4.13907588e-01 5.78986943e-01 8.48245680e-01 7.79422524e-04 -5.27177751e-01 -5.12992442e-01 -1.32284403e-01 -1.54142365e-01 8.58417094e-01 -3.68785918e-01 -4.05164629e-01 -7.91495144e-01 2.72805184e-01 -1.72889185e+00 -1.20472729e+00 -8.76775861e-01 2.11256814e+00 4.00685191e-01 1.51956141e-01 5.25164962e-01 1.48701042e-01 1.04985833e+00 1.20631546e-01 -5.63800812e-01 -1.71241537e-01 -5.60672164e-01 -5.39470792e-01 6.10754907e-01 2.58962095e-01 -1.40268922e+00 1.18489790e+00 7.63647413e+00 6.21318460e-01 -1.42879438e+00 -8.87485087e-01 2.28927523e-01 6.49274111e-01 9.05252099e-02 9.07021984e-02 -9.43627715e-01 4.21693735e-02 6.31651878e-02 -3.85679454e-01 1.30932033e-01 7.57580280e-01 4.68950033e-01 -4.40903753e-01 -8.60328138e-01 1.54119003e+00 4.75002170e-01 -1.33841836e+00 -1.82622090e-01 -2.20591545e-01 7.74065316e-01 -4.53096256e-02 2.70525396e-01 -2.49570340e-01 -4.81438100e-01 -4.79198098e-01 5.39152801e-01 5.07006884e-01 5.16748846e-01 -4.28090125e-01 5.62286317e-01 1.47252291e-01 -1.39965928e+00 2.32812807e-01 -5.93988597e-01 -1.97258472e-01 4.28641468e-01 4.45663095e-01 -8.17738831e-01 5.80584168e-01 4.26228493e-01 1.03771698e+00 -6.79689586e-01 2.05881739e+00 -3.78858238e-01 6.78706884e-01 -3.02982181e-01 -2.50477195e-01 2.24579945e-01 -5.45099080e-01 5.86616039e-01 1.37901080e+00 2.78744400e-01 -6.81039318e-02 2.30889499e-01 7.49823093e-01 4.45445567e-01 3.82194996e-01 -7.35634923e-01 -2.23634709e-02 7.78004646e-01 1.13610482e+00 -1.35794318e+00 -1.42289490e-01 -1.05328429e+00 1.13568532e+00 -5.35207707e-03 6.44919872e-01 -7.16232598e-01 -9.92472768e-01 4.39210474e-01 -6.12885840e-02 1.85099587e-01 -7.30382979e-01 -8.52039039e-01 -1.04646409e+00 3.55189778e-02 -3.43620211e-01 2.04714984e-01 -7.66003788e-01 -1.28428423e+00 4.36265826e-01 2.33822912e-01 -1.67758846e+00 -2.15319380e-01 -6.22600257e-01 -6.77355886e-01 5.81993163e-01 -1.66159308e+00 -1.05469608e+00 -6.44264460e-01 3.11715692e-01 9.36234474e-01 -1.11838728e-01 5.32276511e-01 -4.52463813e-02 -8.75326633e-01 7.55673826e-01 3.08910966e-01 5.16198277e-01 9.15279448e-01 -9.76178885e-01 6.71763718e-01 8.19508433e-01 1.72864541e-01 8.25279355e-01 7.47026503e-01 -5.71677446e-01 -1.30805123e+00 -5.94416082e-01 7.50932932e-01 -8.34477022e-02 4.84566927e-01 -2.58165635e-02 -8.53590429e-01 5.99611819e-01 2.73729056e-01 -3.11481774e-01 7.25564301e-01 -3.03170085e-02 2.77705528e-02 1.68181375e-01 -7.09113538e-01 9.48008120e-01 1.24969685e+00 -1.33769542e-01 -5.29141188e-01 2.06775486e-01 2.83709671e-02 -6.92094862e-01 -3.95242900e-01 2.58173674e-01 8.28235626e-01 -1.48842680e+00 1.03721118e+00 -9.02808458e-02 -5.66051826e-02 -5.25716186e-01 4.21461284e-01 -9.50948179e-01 -4.66704130e-01 -5.07763267e-01 6.77436441e-02 1.09513831e+00 3.43369365e-01 -5.39054751e-01 8.10826838e-01 2.50196606e-01 -4.49841112e-01 -9.74990189e-01 -2.83081234e-01 -7.60738850e-01 -1.89015448e-01 -5.78998744e-01 1.64060444e-01 7.01264203e-01 2.17677318e-02 3.09044123e-01 -2.38309637e-01 2.23057196e-01 8.13297808e-01 6.15290642e-01 1.03624403e+00 -9.04015422e-01 4.80481356e-01 -1.01786792e+00 -6.76485240e-01 -2.09519768e+00 -2.91193396e-01 -4.94223863e-01 -2.29044244e-01 -2.12842393e+00 3.40653919e-02 -3.12434882e-01 3.33536595e-01 1.57038867e-02 1.59923658e-02 5.56803703e-01 2.25000113e-01 6.55444801e-01 -5.58809817e-01 1.75627083e-01 1.62060010e+00 -6.27740547e-02 -2.94852436e-01 4.82496679e-01 -8.23303521e-01 1.09573913e+00 8.18781018e-01 1.97334737e-01 -5.21136999e-01 -4.51520413e-01 -1.78442895e-01 -1.41547874e-01 4.73650433e-02 -1.08288598e+00 7.17390895e-01 -2.22879753e-01 5.23123682e-01 -1.45235813e+00 3.60166460e-01 -5.54865777e-01 -6.60876572e-01 3.57298136e-01 -2.25854263e-01 1.78812549e-01 2.01358974e-01 2.95823038e-01 -2.99514502e-01 -5.70358455e-01 8.30752850e-01 1.93770871e-01 -1.29486370e+00 3.18802714e-01 -3.73251647e-01 -2.07730457e-01 1.53009856e+00 -1.12683856e+00 -4.44047898e-01 -6.95333242e-01 -4.15196359e-01 3.81131589e-01 6.63199842e-01 3.20829749e-01 9.74721909e-01 -1.33914351e+00 -6.62443161e-01 4.60748702e-01 4.14553583e-01 -2.43645877e-01 3.36846203e-01 7.83450663e-01 -1.03974807e+00 9.09733295e-01 -3.65129501e-01 -8.27156663e-01 -1.22559679e+00 6.81093514e-01 3.11503351e-01 5.04431427e-01 -7.97329664e-01 9.14287150e-01 2.57792950e-01 3.35426718e-01 4.83813822e-01 -4.20275718e-01 -4.64761674e-01 -9.78809446e-02 4.88981277e-01 6.98263407e-01 -3.19857955e-01 -6.22378707e-01 -3.22771668e-01 1.43420184e+00 -1.43358544e-01 1.37597486e-01 8.24935198e-01 -6.45282447e-01 1.45775050e-01 6.34385407e-01 8.61904383e-01 4.33576316e-01 -1.18315339e+00 -1.62292704e-01 -1.41159028e-01 -6.45238042e-01 -2.69497663e-01 -4.69542354e-01 -8.59741867e-01 9.55938160e-01 4.51172739e-01 4.11583751e-01 1.28779876e+00 4.84683216e-02 8.00923944e-01 1.59317955e-01 4.45863903e-01 -1.03721738e+00 1.26063883e-01 4.61856991e-01 8.21968436e-01 -1.21724200e+00 1.93376854e-01 -1.18007839e+00 -7.55735457e-01 1.82308304e+00 6.34312391e-01 -1.54095769e-01 3.88567179e-01 2.62264550e-01 4.05453414e-01 -1.17352501e-01 2.65201367e-02 -3.05513114e-01 4.52764004e-01 1.12435162e+00 6.97725475e-01 -1.65482000e-01 -5.55072486e-01 1.46550432e-01 -2.55055040e-01 -1.63731247e-01 6.18996561e-01 8.62348974e-01 -6.36606991e-01 -1.13012731e+00 -7.16327131e-01 3.49190205e-01 1.41016766e-01 1.08371042e-01 -4.69106346e-01 1.00865340e+00 -3.05899203e-01 1.11310959e+00 1.26973316e-01 -8.22961517e-03 5.86287618e-01 -6.69758260e-01 6.96051836e-01 -3.17246914e-01 1.48259223e-01 2.99390644e-01 -6.30909801e-02 -2.53087968e-01 -5.51166892e-01 -7.21318424e-01 -1.27140915e+00 -2.43742168e-01 -6.49879098e-01 8.93998612e-03 6.12453818e-01 4.12417322e-01 1.38392508e-01 2.32414410e-01 5.74230194e-01 -6.39956534e-01 -2.37780899e-01 -5.72379470e-01 -6.70001090e-01 1.53953344e-01 3.15702617e-01 -7.82831907e-01 -8.00881088e-02 8.74161348e-02]
[8.357903480529785, -1.588034987449646]
713e1b40-fe66-4d2d-8b92-1c236e4080d9
multimodal-adaptive-distillation-for
2204.10496
null
https://arxiv.org/abs/2204.10496v2
https://arxiv.org/pdf/2204.10496v2.pdf
Multimodal Adaptive Distillation for Leveraging Unimodal Encoders for Vision-Language Tasks
Cross-modal encoders for vision-language (VL) tasks are often pretrained with carefully curated vision-language datasets. While these datasets reach an order of 10 million samples, the labor cost is prohibitive to scale further. Conversely, unimodal encoders are pretrained with simpler annotations that are less cost-prohibitive, achieving scales of hundreds of millions to billions. As a result, unimodal encoders have achieved state-of-art (SOTA) on many downstream tasks. However, challenges remain when applying to VL tasks. The pretraining data is not optimal for cross-modal architectures and requires heavy computational resources. In addition, unimodal architectures lack cross-modal interactions that have demonstrated significant benefits for VL tasks. Therefore, how to best leverage pretrained unimodal encoders for VL tasks is still an area of active research. In this work, we propose a method to leverage unimodal vision and text encoders for VL tasks that augment existing VL approaches while conserving computational complexity. Specifically, we propose Multimodal Adaptive Distillation (MAD), which adaptively distills useful knowledge from pretrained encoders to cross-modal VL encoders. Second, to better capture nuanced impacts on VL task performance, we introduce an evaluation protocol that includes Visual Commonsense Reasoning (VCR), Visual Entailment (SNLI-VE), and Visual Question Answering (VQA), across a variety of data constraints and conditions of domain shift. Experiments demonstrate that MAD leads to consistent gains in the low-shot, domain-shifted, and fully-supervised conditions on VCR, SNLI-VE, and VQA, achieving SOTA performance on VCR compared to other single models pretrained with image-text data. Finally, MAD outperforms concurrent works utilizing pretrained vision encoder from CLIP. Code will be made available.
['Lu Yuan', 'Shih-Fu Chang', 'Kai-Wei Chang', 'Haoxuan You', 'Jianwei Yang', 'Bin Xiao', 'Xiyang Dai', 'Luowei Zhou', 'Yen-Chun Chen', 'Noel Codella', 'Zhecan Wang']
2022-04-22
null
null
null
null
['visual-commonsense-reasoning', 'visual-entailment']
['reasoning', 'reasoning']
[ 5.25574759e-02 -7.99935609e-02 -2.50699699e-01 -3.51126939e-01 -8.83038044e-01 -6.72697484e-01 8.24108720e-01 -2.62336105e-01 -5.39874136e-01 5.94202638e-01 2.31107578e-01 -6.05182469e-01 3.30541700e-01 -5.64923882e-01 -9.71172929e-01 -2.92052805e-01 4.47713256e-01 3.92961144e-01 3.19210067e-02 -2.15477586e-01 1.41718024e-02 1.47023693e-01 -1.42659354e+00 7.38976121e-01 9.99035895e-01 9.49119568e-01 4.12606418e-01 6.44222319e-01 -3.05719256e-01 1.00522220e+00 -3.73123527e-01 -8.46237540e-01 2.03235939e-01 -4.22815770e-01 -8.15728962e-01 -5.45710437e-02 9.46305990e-01 -5.96724331e-01 -5.12873948e-01 9.86042917e-01 4.77101177e-01 -4.44184616e-02 8.56535971e-01 -1.44719362e+00 -1.69185030e+00 5.67982376e-01 -5.71408272e-01 5.75979203e-02 2.79716045e-01 6.89118326e-01 1.31758654e+00 -1.19824672e+00 6.12576962e-01 1.53082919e+00 4.27527457e-01 7.27343023e-01 -1.36605740e+00 -6.70641899e-01 2.10013255e-01 3.49876374e-01 -1.27045155e+00 -5.17908037e-01 6.14890754e-01 -4.78564829e-01 1.22168088e+00 -5.37753627e-02 3.63415360e-01 1.56414795e+00 -1.03104100e-01 1.11843932e+00 1.09411490e+00 -2.84926951e-01 -2.25478429e-02 3.53201687e-01 4.46622521e-02 8.54357660e-01 3.17825943e-01 1.34036932e-02 -5.37909627e-01 3.91525060e-01 6.15216851e-01 2.67165843e-02 -4.31541026e-01 -3.39136213e-01 -1.22924745e+00 9.43921447e-01 6.06615186e-01 7.37322196e-02 -1.61599338e-01 2.70396531e-01 6.53744876e-01 3.66488039e-01 1.94610655e-01 5.41218519e-01 -3.55229139e-01 -4.22773249e-02 -8.46734822e-01 -4.58193803e-03 4.50433880e-01 1.11716926e+00 8.80687237e-01 2.36788690e-01 -4.70766872e-01 9.32268977e-01 2.76695341e-01 8.14849138e-01 4.02742445e-01 -9.49547470e-01 6.75074518e-01 6.62292361e-01 -3.89391363e-01 -7.16967881e-01 -1.80012807e-02 -1.21048115e-01 -8.40882957e-01 3.70416306e-02 4.01305825e-01 3.72576229e-02 -1.22496557e+00 1.93696523e+00 -1.79750711e-01 -1.48870826e-01 4.62208271e-01 9.48060334e-01 1.27331710e+00 7.06275940e-01 3.25585037e-01 1.40340850e-01 1.49629259e+00 -1.18444574e+00 -6.24275863e-01 -6.77702248e-01 4.97962028e-01 -5.64919472e-01 1.94178391e+00 4.91954312e-02 -1.06381977e+00 -7.26342678e-01 -8.91788900e-01 -6.46770477e-01 -4.76991951e-01 1.69296578e-01 5.55748880e-01 3.50330144e-01 -1.16185963e+00 -6.37705103e-02 -5.19860506e-01 -4.87240821e-01 6.83761775e-01 -6.85021430e-02 -3.93426120e-01 -5.06069243e-01 -1.28299558e+00 9.74696517e-01 3.82302701e-01 -1.15949698e-01 -1.28956342e+00 -9.09231067e-01 -1.26593888e+00 5.17889373e-02 4.17527854e-01 -9.45084989e-01 1.14429021e+00 -1.06170952e+00 -1.09603429e+00 1.08519268e+00 -2.19535947e-01 -4.73001182e-01 4.13385957e-01 -1.58238918e-01 -2.71438450e-01 3.60678613e-01 2.27207750e-01 1.20172858e+00 8.64119887e-01 -1.47344351e+00 -3.31745088e-01 -2.38156214e-01 5.51967740e-01 3.21633190e-01 -4.85867471e-01 -3.74560803e-01 -7.22073972e-01 -4.75882590e-01 -4.13497567e-01 -7.20187128e-01 1.96711853e-01 8.00850913e-02 -3.40633094e-01 -3.93547118e-01 7.88113415e-01 -6.60834789e-01 9.12470102e-01 -2.21827459e+00 1.89189836e-01 -4.26377296e-01 2.32098296e-01 2.36547083e-01 -5.94782293e-01 2.68588334e-01 2.26640210e-01 2.22147435e-01 -2.72533387e-01 -4.82646555e-01 1.90505445e-01 4.34183449e-01 -4.89934385e-01 -3.54355909e-02 5.34465134e-01 1.47190177e+00 -7.58548677e-01 -7.15718687e-01 2.90735126e-01 4.62670445e-01 -7.20105946e-01 2.36882150e-01 -4.31455463e-01 1.49264723e-01 4.35626581e-02 9.69739139e-01 4.74730223e-01 -5.12636185e-01 -1.25511065e-01 -5.62241614e-01 1.70346662e-01 -2.17001997e-02 -4.56473887e-01 1.83667576e+00 -5.82517266e-01 1.00869250e+00 -1.74028456e-01 -9.47546721e-01 6.49142861e-01 -8.68525167e-05 -6.29803464e-02 -1.13777578e+00 1.12014525e-01 -6.85309991e-02 -9.93301570e-02 -6.92325532e-01 5.24909735e-01 -3.48637849e-01 -8.65911543e-02 1.64259270e-01 2.92765468e-01 -1.83054149e-01 3.28252137e-01 5.56110144e-01 8.51821065e-01 1.23418264e-01 1.10877171e-01 2.19953403e-01 4.03558403e-01 3.19993049e-01 1.88836843e-01 6.47870660e-01 -5.06752074e-01 5.01107037e-01 4.38880891e-01 -2.51663779e-03 -1.04007781e+00 -1.34535444e+00 -4.49669212e-02 1.31144536e+00 2.65034825e-01 -2.43367299e-01 -6.02633476e-01 -7.47620344e-01 2.05269977e-01 9.59508479e-01 -6.08583331e-01 -3.48860741e-01 -2.23696917e-01 -2.92103887e-01 1.00867724e+00 8.53420794e-01 6.85852230e-01 -1.06392217e+00 -4.32535172e-01 -4.15634453e-01 -3.57301265e-01 -1.61292911e+00 -5.05586207e-01 -1.16205208e-01 -5.43441415e-01 -8.81869376e-01 -8.63423884e-01 -8.08006287e-01 5.21075845e-01 5.46206534e-01 1.37753403e+00 -3.24085623e-01 -2.54524022e-01 8.25680315e-01 -3.34307820e-01 -2.91346192e-01 -2.88588345e-01 -2.44213939e-01 -1.17171139e-01 -2.35703960e-01 5.77567816e-01 -1.68357685e-01 -4.98207897e-01 -3.77373882e-02 -9.95290816e-01 2.26358443e-01 9.79462743e-01 1.00060415e+00 6.33278310e-01 -4.30443555e-01 6.58931077e-01 -6.49598479e-01 7.27941990e-01 -4.61510867e-01 -3.02644551e-01 5.27805448e-01 -5.44518590e-01 1.84810653e-01 7.51341522e-01 -6.32411301e-01 -1.26707435e+00 -1.04435764e-01 1.18203357e-01 -1.03199017e+00 -1.87172607e-01 6.02148294e-01 -1.90801173e-01 1.81511894e-01 6.42076433e-01 4.51957971e-01 6.87090773e-03 -1.13151036e-01 1.02941763e+00 6.61671638e-01 7.74677098e-01 -6.83927298e-01 6.47800207e-01 5.14805615e-01 -3.95816028e-01 -7.75719285e-01 -9.48388398e-01 -3.29660088e-01 -4.47462648e-01 -5.81568815e-02 1.19623113e+00 -1.33500481e+00 -6.26964986e-01 1.26473531e-01 -1.19606459e+00 -4.68012005e-01 -1.54807895e-01 3.30375761e-01 -4.37908769e-01 5.36057234e-01 -6.13759875e-01 -5.21726847e-01 -4.03589666e-01 -1.24624515e+00 1.03683484e+00 6.11073188e-02 -1.28208831e-01 -1.06889999e+00 -1.42558590e-01 8.22649240e-01 2.80032158e-01 -2.43850332e-02 1.16330099e+00 -4.53369766e-01 -6.42919242e-01 2.84227073e-01 -9.36195493e-01 5.67217946e-01 -8.37874040e-02 -1.41971543e-01 -1.09122634e+00 -3.65879267e-01 -4.76239949e-01 -1.02817798e+00 1.19666815e+00 2.81991839e-01 1.16829240e+00 -2.29943156e-01 -1.77310869e-01 7.48809814e-01 1.43051398e+00 3.75489146e-02 6.18627310e-01 2.19833091e-01 1.05899644e+00 5.11578619e-01 5.16259670e-01 1.56123668e-01 9.24710929e-01 3.17648530e-01 5.48467457e-01 -2.06373438e-01 -5.99155009e-01 -4.66261387e-01 6.79623127e-01 5.51756918e-01 2.05523632e-02 -2.18770042e-01 -1.03996623e+00 7.69932806e-01 -1.58849323e+00 -1.04306531e+00 1.10188156e-01 1.74774289e+00 9.88281071e-01 -1.31480888e-01 -6.08851649e-02 -3.03903848e-01 3.81265253e-01 1.98986799e-01 -8.87548029e-01 -4.55422044e-01 -2.16045350e-01 -2.36400142e-02 3.02393019e-01 3.76590520e-01 -8.88914526e-01 1.25586450e+00 5.62200308e+00 6.81965768e-01 -1.13348341e+00 1.65808916e-01 4.16054666e-01 -1.51467204e-01 -6.37243986e-01 -7.15615973e-02 -7.16248274e-01 2.66086906e-01 7.12403059e-01 -4.68902402e-02 4.93301034e-01 9.87701595e-01 -1.53259844e-01 -6.43733889e-03 -1.40885639e+00 1.36886132e+00 4.78614122e-01 -1.20858121e+00 6.55754089e-01 -7.48987272e-02 6.90546691e-01 1.79980755e-01 4.27023709e-01 9.14985001e-01 3.90289396e-01 -1.19223297e+00 7.21792102e-01 2.24401146e-01 1.32084441e+00 -5.56550145e-01 4.47981179e-01 1.29288763e-01 -1.13227320e+00 -7.03035146e-02 -5.11198401e-01 1.39868498e-01 1.27726614e-01 1.91035300e-01 -7.96535611e-01 4.08908874e-01 7.40156651e-01 8.65306377e-01 -7.84470797e-01 3.89342040e-01 -2.26647571e-01 4.31025296e-01 1.48552150e-01 1.14853807e-01 4.53331590e-01 1.47233024e-01 2.70468235e-01 1.42631507e+00 1.11676037e-01 -6.44141063e-02 8.71083438e-02 1.24452138e+00 -3.57007146e-01 -1.11424252e-01 -9.44357276e-01 -4.12861437e-01 4.60389793e-01 1.04063547e+00 -1.93761140e-01 -4.26986903e-01 -8.46419990e-01 1.30587864e+00 6.50945008e-01 7.38742709e-01 -1.11125624e+00 -2.27578297e-01 7.74719000e-01 -1.65883496e-01 2.71575272e-01 -2.02023119e-01 -3.54420751e-01 -1.48013556e+00 -1.91185549e-01 -1.06595683e+00 5.62355280e-01 -1.02084494e+00 -1.67702520e+00 5.16075492e-01 -2.15627402e-02 -1.03476560e+00 -1.65081397e-01 -1.07850111e+00 -2.75624841e-01 7.49650240e-01 -1.89172626e+00 -1.60128498e+00 -5.24629414e-01 9.60414231e-01 8.67710471e-01 -2.91216731e-01 4.89252567e-01 2.59219676e-01 -5.82138598e-01 9.17304814e-01 -1.95281744e-01 3.11236560e-01 1.01564610e+00 -1.14125121e+00 7.61344880e-02 8.65699112e-01 2.37370282e-01 6.83451772e-01 3.20576370e-01 -5.70385575e-01 -1.59582829e+00 -1.24510324e+00 5.49088895e-01 -5.77305555e-01 7.30728745e-01 -3.38671386e-01 -9.55724299e-01 9.29081619e-01 5.23097217e-01 4.91008116e-03 6.75853193e-01 1.84320286e-01 -9.28659379e-01 4.97910194e-02 -8.39607000e-01 8.60854566e-01 1.10747778e+00 -1.07059932e+00 -7.70116627e-01 8.99969116e-02 1.11165142e+00 -1.38103083e-01 -6.82247519e-01 4.36825573e-01 4.79016304e-01 -8.56857777e-01 1.29246891e+00 -7.26827443e-01 9.29489970e-01 -1.71844304e-01 -5.15635312e-01 -1.18188667e+00 -2.87182927e-01 -2.46009529e-02 -2.00247347e-01 1.22675109e+00 4.51761723e-01 -3.96723926e-01 3.25548977e-01 5.82053721e-01 -2.33410180e-01 -6.11113191e-01 -5.82490027e-01 -8.18205893e-01 3.12079668e-01 -5.89347959e-01 1.91909075e-01 1.21761119e+00 -1.47799686e-01 8.12264621e-01 -3.90595138e-01 7.72887915e-02 5.98073542e-01 9.94591117e-02 8.34624171e-01 -7.27842510e-01 -3.63258958e-01 -5.11945605e-01 -1.22116052e-01 -1.17280269e+00 4.45721209e-01 -1.28038526e+00 -4.60151061e-02 -1.83697295e+00 5.73783696e-01 -2.32597645e-02 -2.92950273e-01 7.76411176e-01 -3.09760898e-01 3.30632269e-01 4.39379483e-01 2.08460301e-01 -7.38998771e-01 6.88729525e-01 1.52840054e+00 -4.51384842e-01 6.08296040e-03 -8.27027440e-01 -8.19526553e-01 6.82945907e-01 6.05224252e-01 1.63009584e-01 -7.97700226e-01 -9.93931592e-01 5.24347723e-02 -1.37220263e-01 6.69800699e-01 -6.75688028e-01 1.29745468e-01 -2.10258096e-01 4.37860817e-01 -6.35376453e-01 4.56907839e-01 -6.46357179e-01 -3.89051408e-01 2.30533347e-01 -5.12878120e-01 2.45736502e-02 4.02374685e-01 6.86178148e-01 -4.07713532e-01 4.98034246e-02 7.73371577e-01 -1.79067507e-01 -1.12624025e+00 2.70778388e-01 -1.67828854e-02 4.60543603e-01 1.05021882e+00 -2.20474169e-01 -8.11122954e-01 -4.35340822e-01 -3.16289544e-01 4.97993827e-01 4.95335281e-01 5.89357495e-01 9.58710134e-01 -1.34341466e+00 -7.01892853e-01 -6.52445331e-02 6.13848388e-01 -2.84672044e-02 3.90731514e-01 7.52318680e-01 -2.88038194e-01 6.35071516e-01 -3.55910152e-01 -7.69849181e-01 -1.23544276e+00 8.66060615e-01 1.76169306e-01 -3.99441905e-02 -4.40534502e-01 8.91575396e-01 7.78064609e-01 -4.03412163e-01 2.44810790e-01 -4.10075665e-01 -1.04108199e-01 1.30256832e-01 4.32448268e-01 1.38003960e-01 -4.11241680e-01 -5.39524734e-01 -3.02687407e-01 4.73898023e-01 -1.87938899e-01 -3.90370414e-02 8.99643123e-01 -2.70151407e-01 9.37186107e-02 3.54377061e-01 1.30436456e+00 -3.07928592e-01 -1.41918981e+00 -3.95443380e-01 -3.37605625e-01 -2.98038810e-01 8.55091959e-03 -1.02218854e+00 -1.01407790e+00 1.40100658e+00 4.21414673e-01 -2.70071238e-01 1.20070183e+00 3.43408912e-01 8.35937560e-01 5.48135817e-01 5.23997545e-02 -9.18756783e-01 6.23842061e-01 6.93501711e-01 9.89995480e-01 -1.67845881e+00 -4.19505179e-01 -9.11862478e-02 -1.20827770e+00 7.90565550e-01 1.05366874e+00 2.15439156e-01 1.26479596e-01 1.25972271e-01 6.69741184e-02 -1.65954575e-01 -8.83319676e-01 -7.41990149e-01 5.09077191e-01 7.89747298e-01 4.09715742e-01 5.06738154e-03 1.22032657e-01 5.82328141e-01 5.21213375e-02 -8.16908106e-02 2.09320724e-01 5.43871939e-01 -2.35137597e-01 -6.05060339e-01 -1.55966729e-01 2.35961586e-01 -1.36695594e-01 -5.71209192e-01 -4.74053830e-01 9.57605958e-01 1.03409469e-01 8.93072844e-01 9.71228853e-02 -3.35757375e-01 3.52044493e-01 2.20909551e-01 5.93984008e-01 -5.11794686e-01 -6.72416389e-02 -2.55359352e-01 1.18065700e-01 -5.10359108e-01 -2.48834282e-01 -3.02222520e-01 -1.34110761e+00 -3.14161867e-01 -1.23771252e-02 -5.22252262e-01 3.33627582e-01 9.02536690e-01 4.75613356e-01 5.58100104e-01 1.23635037e-02 -5.57055771e-01 -4.68217134e-01 -7.89121509e-01 -5.83236292e-02 7.61912167e-01 2.65568346e-01 -7.16998637e-01 -3.14765632e-01 2.68047780e-01]
[10.795576095581055, 1.6715483665466309]
bfe9efc8-1659-498c-94e2-c6219c443f01
comparing-in-context-improving-cosine
2203.14996
null
https://arxiv.org/abs/2203.14996v1
https://arxiv.org/pdf/2203.14996v1.pdf
Comparing in context: Improving cosine similarity measures with a metric tensor
Cosine similarity is a widely used measure of the relatedness of pre-trained word embeddings, trained on a language modeling goal. Datasets such as WordSim-353 and SimLex-999 rate how similar words are according to human annotators, and as such are often used to evaluate the performance of language models. Thus, any improvement on the word similarity task requires an improved word representation. In this paper, we propose instead the use of an extended cosine similarity measure to improve performance on that task, with gains in interpretability. We explore the hypothesis that this approach is particularly useful if the word-similarity pairs share the same context, for which distinct contextualized similarity measures can be learned. We first use the dataset of Richie et al. (2020) to learn contextualized metrics and compare the results with the baseline values obtained using the standard cosine similarity measure, which consistently shows improvement. We also train a contextualized similarity measure for both SimLex-999 and WordSim-353, comparing the results with the corresponding baselines, and using these datasets as independent test sets for the all-context similarity measure learned on the contextualized dataset, obtaining positive results for a number of tests.
['Adriana D. Correia', 'Mara D. Fennema', 'Ghislaine L. van den Boogerd', 'Isa M. Apallius de Vos']
2022-03-28
null
https://aclanthology.org/2021.icon-main.17
https://aclanthology.org/2021.icon-main.17.pdf
icon-2021-12
['word-similarity']
['natural-language-processing']
[ 1.08912528e-01 -2.05158621e-01 -2.22289085e-01 -5.21381378e-01 -7.32364118e-01 -7.35839903e-01 1.09391677e+00 8.88657451e-01 -1.19652379e+00 4.50278521e-01 6.77602470e-01 -3.11399490e-01 -1.83953613e-01 -6.32326484e-01 -1.48437008e-01 -4.07826930e-01 -2.31071133e-02 5.12192190e-01 2.89652139e-01 -4.96718347e-01 4.62621272e-01 1.79184005e-01 -1.40425086e+00 1.25203580e-01 8.03992152e-01 5.63651383e-01 3.35828364e-01 5.86274326e-01 -2.60128975e-01 1.52990267e-01 -5.37037671e-01 -4.45980430e-01 -5.19888625e-02 -3.13130051e-01 -9.36342835e-01 -5.10082364e-01 5.33940434e-01 5.27998269e-01 -1.06566951e-01 8.94144177e-01 3.95118356e-01 6.59502566e-01 8.83498609e-01 -9.58889484e-01 -7.47662187e-01 5.58840513e-01 -5.49710914e-02 5.58894157e-01 5.37272751e-01 -1.89047351e-01 1.61808443e+00 -1.04311442e+00 6.25769019e-01 1.18252707e+00 7.79116511e-01 3.39119434e-01 -1.34225607e+00 -4.43110406e-01 1.32631630e-01 3.96645814e-01 -1.53917134e+00 -1.88759416e-01 2.21178338e-01 -1.86799675e-01 1.40840685e+00 2.24874333e-01 1.32642016e-01 1.02089226e+00 -7.21874759e-02 2.68527001e-01 9.54063773e-01 -8.52337599e-01 1.20638914e-01 3.42862666e-01 4.46689188e-01 3.71978194e-01 4.33171064e-01 1.40920244e-02 -1.75167099e-01 -2.21478373e-01 8.51950943e-02 -1.52464658e-01 -2.12787926e-01 -2.33034462e-01 -1.36117375e+00 1.12563300e+00 4.20632631e-01 1.09274006e+00 1.62986107e-02 -2.27412730e-02 7.17172205e-01 3.58496964e-01 6.77435756e-01 1.12564504e+00 -4.54878688e-01 -2.27921605e-01 -8.26635659e-01 5.02288938e-01 7.37276435e-01 6.94285572e-01 5.87949932e-01 -4.00713831e-01 -2.57888943e-01 1.14498770e+00 3.52478653e-01 2.66380519e-01 1.04423440e+00 -5.52538455e-01 3.96919250e-01 3.62887323e-01 -9.93561745e-02 -1.01044357e+00 -4.30889428e-01 -3.00029665e-01 -2.49990508e-01 -2.17730984e-01 2.44862154e-01 4.96434122e-02 -7.34824061e-01 1.89697337e+00 -1.02560660e-02 2.30140477e-01 2.28879139e-01 6.55712247e-01 7.10883498e-01 5.22680700e-01 3.72208655e-01 7.93252140e-02 1.41325641e+00 -8.84453118e-01 -5.77534854e-01 -5.12842953e-01 1.36238492e+00 -1.06713915e+00 1.47901773e+00 -6.72982782e-02 -6.59134448e-01 -6.07874990e-01 -1.17344213e+00 -1.29203394e-01 -8.02416384e-01 -3.28810096e-01 4.08287704e-01 4.47324693e-01 -9.30055916e-01 7.60575056e-01 -5.22361159e-01 -8.44407082e-01 -2.31224626e-01 -3.82569954e-02 -5.37431717e-01 -2.23785684e-01 -1.52284884e+00 1.66381490e+00 5.82061768e-01 -7.55211592e-01 -1.20569721e-01 -7.76190341e-01 -1.06610990e+00 5.91867678e-02 -1.21080704e-01 -3.56683522e-01 1.08923995e+00 -5.30209124e-01 -7.14002550e-01 1.27007198e+00 -1.09461218e-01 -4.18541729e-01 9.55714937e-03 -2.01488622e-02 -7.38846898e-01 -2.82472938e-01 2.97505200e-01 5.19861162e-01 1.49238989e-01 -9.95286405e-01 -6.50771081e-01 -1.30841985e-01 1.55878579e-02 2.60618836e-01 -6.87406540e-01 3.36044550e-01 -1.98928937e-01 -8.91840696e-01 -4.06488746e-01 -1.04989469e+00 -1.80474833e-01 -2.11957023e-01 1.99105129e-01 -7.00746238e-01 3.67236942e-01 -5.16571105e-01 1.66410351e+00 -2.19952607e+00 4.92621027e-02 3.27471942e-01 -1.02297097e-01 6.10538483e-01 -6.74247444e-01 7.00577557e-01 -4.15542781e-01 2.46731013e-01 -4.21414316e-01 -3.60216677e-01 -3.76916416e-02 4.02058691e-01 1.38733452e-02 2.18560189e-01 2.53254354e-01 6.97773039e-01 -1.16059864e+00 -4.12518114e-01 3.19736004e-02 4.38494682e-01 -5.66201270e-01 2.41556451e-01 6.58783689e-02 -2.15358496e-01 8.09677318e-03 -1.57594338e-01 1.72521517e-01 2.88361292e-02 2.76596248e-01 -3.24224800e-01 1.15215331e-01 5.81845164e-01 -9.33130145e-01 1.66539085e+00 -1.07179010e+00 8.95907104e-01 -5.65565288e-01 -1.10914886e+00 1.03981113e+00 2.85986871e-01 2.52021044e-01 -8.37940097e-01 1.63643644e-03 3.53358388e-01 3.67389113e-01 -3.93671542e-01 7.40652561e-01 -3.78884643e-01 -2.15309575e-01 4.91384417e-01 2.64052838e-01 -4.59793687e-01 4.81537253e-01 2.26359054e-01 1.23393667e+00 -2.76966840e-01 4.63529110e-01 -6.42446518e-01 5.90101302e-01 -6.55212849e-02 1.20165646e-02 4.19252813e-01 -1.07942514e-01 6.89214826e-01 -8.85523914e-04 -5.71921989e-02 -1.13393497e+00 -1.03634989e+00 -4.07875150e-01 1.27043009e+00 1.30156730e-03 -1.03496921e+00 -5.72913408e-01 -8.07880998e-01 1.15602002e-01 1.29090786e+00 -7.08170593e-01 -5.36629856e-01 -5.20304859e-01 -6.06429160e-01 5.69140673e-01 7.11318076e-01 -1.61821306e-01 -8.30423594e-01 -2.45958492e-01 1.11963138e-01 -2.26089768e-02 -1.09980261e+00 -6.50417805e-01 4.75573748e-01 -6.08902931e-01 -9.88902390e-01 -6.07220054e-01 -9.32700276e-01 2.46404871e-01 3.55145991e-01 1.49394488e+00 3.10113102e-01 -2.37648025e-01 5.66131532e-01 -6.37415111e-01 -3.39297473e-01 -5.23975849e-01 1.63695216e-01 2.91368783e-01 -4.58856493e-01 8.37410510e-01 -3.62105101e-01 -2.27730617e-01 2.44451940e-01 -1.13667440e+00 -5.17346501e-01 1.56199887e-01 9.87510800e-01 1.95364967e-01 -3.13654900e-01 5.56584477e-01 -7.16242313e-01 1.03992343e+00 -5.96050620e-01 -1.09621167e-01 2.89196372e-01 -1.01019883e+00 5.33970296e-01 6.61907434e-01 -4.40706849e-01 -5.63784301e-01 -6.21022820e-01 -2.09191784e-01 -9.91746262e-02 -4.44157906e-02 6.60160244e-01 1.31788999e-01 6.95482418e-02 9.58982110e-01 -1.97703958e-01 -1.45369083e-01 -5.42221129e-01 6.11816108e-01 8.04124594e-01 3.59973073e-01 -5.68406343e-01 6.66907132e-01 -1.20077200e-01 -2.24582806e-01 -8.22194755e-01 -9.42183375e-01 -8.66937399e-01 -4.42491263e-01 4.63228166e-01 8.12610745e-01 -7.09781826e-01 -1.37466684e-01 -2.59744287e-01 -1.25873804e+00 -1.23772305e-03 -2.98769355e-01 7.99680054e-01 -4.64111894e-01 4.46600735e-01 -1.71908274e-01 -1.97015360e-01 -1.27430320e-01 -8.52604687e-01 1.01527858e+00 -1.30844474e-01 -1.04270685e+00 -1.74423754e+00 6.60371423e-01 6.89268410e-02 6.38403714e-01 -1.08208984e-01 1.29339027e+00 -1.52923715e+00 5.30633748e-01 -2.92298198e-01 -1.68537363e-01 5.50931513e-01 4.47461993e-01 -2.61659116e-01 -7.88783669e-01 -3.75650585e-01 -2.59834945e-01 -1.97895885e-01 8.68403137e-01 -2.80274004e-01 9.86789405e-01 5.32246046e-02 -3.43606919e-01 2.24341437e-01 1.33278453e+00 1.85724243e-03 3.26740056e-01 5.17266452e-01 5.73886216e-01 5.51052094e-01 7.14490294e-01 2.84737460e-02 3.11744213e-01 1.00052059e+00 2.29127128e-02 4.45612893e-02 -1.55466050e-01 -2.38097578e-01 2.43778720e-01 1.29846752e+00 1.87161207e-01 -3.62061232e-01 -1.08824956e+00 8.43272448e-01 -1.64847720e+00 -8.56240869e-01 2.28972808e-02 2.41320801e+00 1.01786423e+00 1.28522471e-01 -3.17831822e-02 2.38145649e-01 5.69876611e-01 3.74772310e-01 8.16544145e-02 -9.82645690e-01 -2.90162601e-02 7.58710325e-01 3.85797709e-01 7.72615254e-01 -8.72225940e-01 9.83998656e-01 6.61317253e+00 9.97404873e-01 -8.30188870e-01 1.49750039e-01 1.42467916e-01 -4.35082391e-02 -5.86614013e-01 2.03604507e-03 -6.90780818e-01 5.74315310e-01 1.40553117e+00 -7.13931620e-01 1.82387143e-01 6.34404659e-01 4.56008092e-02 -7.91611671e-02 -1.48644841e+00 9.66639817e-01 5.24581075e-01 -9.70035911e-01 8.92880559e-02 -2.45115846e-01 6.77537024e-01 8.07914883e-02 -1.51336193e-01 4.54025477e-01 2.08596528e-01 -1.18112528e+00 2.37557843e-01 3.36057425e-01 5.83597839e-01 -8.03517044e-01 8.95990729e-01 3.60235125e-02 -1.11959302e+00 2.66034722e-01 -4.62110341e-01 6.85463622e-02 7.45505691e-02 5.00105500e-01 -9.93303597e-01 3.94212812e-01 3.58232290e-01 8.13483357e-01 -9.03926015e-01 8.94291162e-01 -1.85593054e-01 5.61075747e-01 -1.77164003e-01 -3.13175380e-01 6.27126336e-01 -7.91918412e-02 4.58608031e-01 1.70178688e+00 3.11806262e-01 -4.20930713e-01 4.83657829e-02 4.56165195e-01 2.81840246e-02 4.22582358e-01 -9.21536028e-01 -2.81785131e-01 4.97538447e-01 1.24191821e+00 -2.95329690e-01 -4.24776107e-01 -6.27639592e-01 9.17660892e-01 5.06466269e-01 -1.33234859e-01 -6.98842645e-01 -7.48143077e-01 1.15872514e+00 4.22297977e-02 1.93433717e-01 -4.79907215e-01 -9.30375606e-03 -9.62031245e-01 -1.96926340e-01 -8.21667135e-01 5.38967013e-01 -6.43490136e-01 -1.72575736e+00 7.21456528e-01 2.64615029e-01 -1.09926975e+00 -5.90661168e-01 -8.84665668e-01 -6.78871691e-01 1.14954650e+00 -1.48670185e+00 -6.60856128e-01 -1.79073572e-01 3.33478153e-01 4.79189903e-01 -1.99660316e-01 1.22585809e+00 3.44863981e-01 -3.08284070e-02 8.27584326e-01 2.18423322e-01 6.96105137e-02 1.18900847e+00 -1.31968582e+00 6.42053187e-01 5.23408234e-01 7.63442099e-01 8.64937782e-01 8.41564953e-01 -4.01854277e-01 -7.71715105e-01 -1.05339944e+00 1.43540430e+00 -6.39187813e-01 1.06649566e+00 -1.59046173e-01 -1.11767399e+00 3.89928132e-01 3.71590823e-01 6.33475482e-02 1.08203840e+00 3.19771916e-01 -7.20090508e-01 2.01036349e-01 -9.60299194e-01 6.51130438e-01 1.25323033e+00 -8.26454580e-01 -1.19767785e+00 4.16178644e-01 9.33887720e-01 1.99198157e-01 -1.12208283e+00 2.88859636e-01 3.57753605e-01 -6.10612869e-01 9.30651963e-01 -1.10650063e+00 3.99582356e-01 -1.39677450e-01 -6.45762265e-01 -2.03063440e+00 -4.50738132e-01 7.00912923e-02 4.00855482e-01 1.28297770e+00 6.89590573e-01 -6.11939549e-01 2.77593702e-01 3.46088558e-01 1.89964585e-02 -6.40018165e-01 -8.39224637e-01 -1.32831275e+00 5.12673557e-01 -4.07505870e-01 3.81672740e-01 1.23506618e+00 2.34743819e-01 5.99594533e-01 3.29559654e-01 -2.88961709e-01 1.79445595e-01 -2.98717916e-01 3.70555162e-01 -1.16961837e+00 -1.42005011e-01 -6.05715454e-01 -8.30755651e-01 -5.10573983e-01 5.62724173e-01 -1.45204997e+00 -3.71109471e-02 -1.36618662e+00 1.04489654e-01 -5.33084810e-01 -5.29612601e-01 2.89477050e-01 -5.39810359e-01 2.22270414e-01 2.97410488e-01 -1.32232532e-01 -4.26413685e-01 5.05587161e-01 6.66798174e-01 -9.74603593e-02 1.36561126e-01 -4.28231120e-01 -6.50532305e-01 6.30382836e-01 7.06268191e-01 -4.88949656e-01 -5.03233492e-01 -5.40911376e-01 1.83550075e-01 -6.82622850e-01 5.40384166e-02 -9.95130241e-01 4.39376831e-02 -5.64257707e-03 -2.14362727e-03 1.39263853e-01 2.69354433e-01 -8.62951458e-01 -3.38171840e-01 3.19125503e-01 -7.89043546e-01 6.83319569e-01 2.37989843e-01 3.73981357e-01 -3.97582561e-01 -7.73949981e-01 7.07877994e-01 1.95373036e-02 -7.71594882e-01 -6.93925619e-02 -2.33110920e-01 6.97155178e-01 8.94586086e-01 -5.24643660e-02 -9.19235945e-02 -3.08469534e-01 -4.79445457e-01 7.45359994e-03 5.51285505e-01 1.02467930e+00 3.40216875e-01 -1.64112532e+00 -8.81055057e-01 -1.65054556e-02 8.76358747e-01 -6.58250868e-01 -5.13632178e-01 4.80593354e-01 -1.50250658e-01 5.14877379e-01 8.67624357e-02 -3.98970246e-01 -1.47861314e+00 6.25559568e-01 1.42951593e-01 -4.46519464e-01 -1.71932489e-01 5.95981717e-01 -2.95732543e-02 -5.72445035e-01 -5.28516173e-02 -3.73672068e-01 -3.48490298e-01 3.28685939e-01 6.05064631e-01 2.13139534e-01 4.02549773e-01 -7.55039394e-01 -6.07105911e-01 7.78509080e-01 -1.19774193e-01 -2.78669119e-01 1.30028808e+00 6.04819618e-02 5.36710061e-02 6.29574537e-01 1.82406569e+00 1.07911438e-01 -1.43134743e-01 -5.55267990e-01 6.50390983e-01 -6.43602073e-01 -1.89972743e-02 -5.77045202e-01 -6.58785820e-01 8.11713576e-01 6.04536891e-01 2.03802034e-01 5.76791644e-01 4.82180528e-02 5.93098521e-01 4.35791135e-01 1.61818996e-01 -1.02961171e+00 -1.63627621e-02 8.97437930e-01 8.23479414e-01 -1.23075962e+00 -1.08345728e-02 -4.98178527e-02 -6.61132872e-01 1.03872001e+00 4.26241219e-01 -2.26439044e-01 5.86454272e-01 1.07453443e-01 -4.37423065e-02 -6.62332103e-02 -7.36699462e-01 -3.63580674e-01 6.12843931e-01 6.23088598e-01 1.11587679e+00 7.38603249e-02 -8.81415725e-01 4.24350888e-01 -3.03886443e-01 -5.49114943e-01 1.76949725e-01 8.56739819e-01 -3.95462841e-01 -1.51041245e+00 -2.02797409e-02 4.28783447e-01 -1.63829699e-01 -5.56293488e-01 -4.32206213e-01 7.22342372e-01 -3.34942862e-02 9.44438040e-01 3.78028214e-01 -4.26891446e-01 3.77148777e-01 3.06265533e-01 4.02546942e-01 -1.04993832e+00 -7.07671881e-01 -6.71232522e-01 5.10631919e-01 -3.97064418e-01 -2.98462480e-01 -4.51808840e-01 -1.15465260e+00 -1.29772976e-01 -3.66852403e-01 5.14053047e-01 8.11767459e-01 1.09608865e+00 2.28633463e-01 3.29796046e-01 6.62145436e-01 -5.38305223e-01 -6.19370043e-01 -1.16884184e+00 -2.59955823e-01 1.10595632e+00 3.56279756e-03 -7.06257284e-01 -6.37391269e-01 -3.56711924e-01]
[10.472134590148926, 8.863777160644531]
47c5f576-2931-48d5-9df1-f99310d5661d
self-supervised-representation-learning-on
2004.10605
null
https://arxiv.org/abs/2004.10605v2
https://arxiv.org/pdf/2004.10605v2.pdf
Self-Supervised Representation Learning on Document Images
This work analyses the impact of self-supervised pre-training on document images in the context of document image classification. While previous approaches explore the effect of self-supervision on natural images, we show that patch-based pre-training performs poorly on document images because of their different structural properties and poor intra-sample semantic information. We propose two context-aware alternatives to improve performance on the Tobacco-3482 image classification task. We also propose a novel method for self-supervision, which makes use of the inherent multi-modality of documents (image and text), which performs better than other popular self-supervised methods, including supervised ImageNet pre-training, on document image classification scenarios with a limited amount of data.
['Michael Panaitescu-Liess', 'Mihai Ghidoveanu', 'Marius Popescu', 'Adrian Cosma']
2020-04-18
null
null
null
null
['document-image-classification']
['computer-vision']
[ 6.97845340e-01 2.40756404e-02 -4.95903432e-01 -6.03031933e-01 -6.64516211e-01 -6.29686892e-01 1.20241666e+00 4.05723870e-01 -6.77179337e-01 4.96579260e-01 3.30765367e-01 -2.00425997e-01 4.98139448e-02 -5.65883458e-01 -8.67004812e-01 -5.76828897e-01 2.56157875e-01 4.67030466e-01 3.07723463e-01 -2.04982162e-02 3.37164193e-01 1.03246830e-01 -1.65961409e+00 7.71430731e-01 8.56853724e-01 8.14545751e-01 2.88649023e-01 8.13621759e-01 -2.01343849e-01 8.58334422e-01 -6.71402037e-01 -2.53644347e-01 -5.57129271e-03 -6.61357760e-01 -1.12556612e+00 6.56568229e-01 1.16926253e+00 -4.04160708e-01 -1.33794084e-01 1.05330956e+00 3.16476017e-01 1.01400323e-01 9.97692168e-01 -9.86196756e-01 -8.35030079e-01 6.69481754e-01 -4.70746189e-01 5.25210440e-01 1.22313276e-01 -1.60915814e-02 1.03149951e+00 -5.95347703e-01 8.43391180e-01 1.12428606e+00 7.50670254e-01 3.91877949e-01 -1.44883943e+00 -4.05836970e-01 7.18543455e-02 2.97132939e-01 -1.06154978e+00 -4.90528464e-01 8.01885724e-01 -2.79691815e-01 1.01052237e+00 2.71800682e-02 3.22806984e-01 1.43703496e+00 -2.29208469e-02 1.17498541e+00 1.45883310e+00 -8.75426650e-01 7.93183222e-02 5.90301812e-01 3.19293737e-01 7.42047369e-01 -5.57410866e-02 8.80936757e-02 -5.03268719e-01 -3.77195105e-02 4.96944070e-01 -2.00910997e-02 -4.95472103e-02 -4.28380042e-01 -1.11301458e+00 7.30496168e-01 5.14660478e-01 7.38704920e-01 -4.36842218e-02 -1.38697967e-01 6.71086550e-01 2.16432586e-01 7.64151812e-01 1.77929088e-01 -3.89949471e-01 2.17396885e-01 -1.43923080e+00 -8.57951306e-03 5.89559495e-01 8.23710203e-01 8.63342047e-01 2.78193466e-02 -3.96904200e-01 9.30744946e-01 3.05798929e-02 4.50680196e-01 8.36150169e-01 -6.17905736e-01 3.97325367e-01 4.06613022e-01 -4.51835752e-01 -8.19065750e-01 -2.50731856e-01 -5.15903294e-01 -9.26964223e-01 2.84412913e-02 4.33521509e-01 2.50326335e-01 -1.37405121e+00 1.31857800e+00 -8.86537433e-02 -1.45434648e-01 3.36464345e-01 6.21233582e-01 7.64046907e-01 5.80242753e-01 2.28020087e-01 -7.17547014e-02 1.17042983e+00 -1.46652269e+00 -6.71319306e-01 -1.71393722e-01 7.64701188e-01 -7.33556509e-01 1.28434396e+00 1.98097751e-01 -8.62007618e-01 -7.07643509e-01 -9.35129225e-01 3.04624215e-02 -6.63133860e-01 3.26637745e-01 5.23953557e-01 7.70555496e-01 -9.82500672e-01 6.43927753e-01 -4.07407314e-01 -7.82490551e-01 6.44543529e-01 7.48978406e-02 -4.34756190e-01 -1.98064923e-01 -8.07643116e-01 8.25104475e-01 4.77690607e-01 -4.97931600e-01 -8.38409662e-01 -7.06275821e-01 -7.61181533e-01 1.43503025e-01 3.06214690e-01 -3.38656545e-01 1.16145563e+00 -1.65915644e+00 -1.26871872e+00 1.23260415e+00 -1.05753608e-01 -6.28266513e-01 5.29793918e-01 7.95219988e-02 -3.65445346e-01 6.74653769e-01 1.87068731e-01 9.90738332e-01 1.32384968e+00 -1.62946272e+00 -4.10664290e-01 -5.16261816e-01 -1.36489496e-01 2.53332585e-01 -7.72427976e-01 -9.01416987e-02 -4.36235875e-01 -8.31437349e-01 -3.38790506e-01 -1.04228711e+00 -8.83419961e-02 -2.69194603e-01 -6.80330753e-01 1.68259663e-03 1.17235100e+00 -4.66324598e-01 5.51477671e-01 -2.16250062e+00 -2.46485904e-01 2.78763801e-01 -3.45316470e-01 4.22002465e-01 -4.45413679e-01 3.06412697e-01 -2.35446990e-01 1.75481271e-02 -4.74186510e-01 -6.78398967e-01 -3.83909047e-01 2.87934601e-01 -2.31108814e-01 3.82830501e-01 1.77221984e-01 9.25231278e-01 -6.96443141e-01 -9.88196373e-01 6.98324516e-02 4.86995846e-01 -4.32472497e-01 -5.22176549e-02 -4.59277004e-01 2.97710449e-01 -1.73023090e-01 6.70884430e-01 6.21017277e-01 -4.28508282e-01 9.75763276e-02 -4.05942738e-01 8.10193121e-02 4.84485142e-02 -8.30294371e-01 1.88106418e+00 -4.28207308e-01 7.28232384e-01 1.68252632e-01 -1.40000153e+00 5.72633624e-01 2.27116391e-01 4.63842779e-01 -8.85023654e-01 1.43801749e-01 -5.36686741e-02 -2.21604750e-01 -3.16773623e-01 6.03795409e-01 -1.62417635e-01 5.11520386e-01 7.08024204e-01 4.98835206e-01 -2.06203908e-01 5.75105250e-01 4.60049778e-01 9.46458578e-01 5.36381379e-02 -5.94066978e-02 -5.17656386e-01 6.10977590e-01 3.86929303e-01 -2.16180906e-01 9.86083984e-01 -8.57314765e-02 8.85534823e-01 6.23931214e-02 -3.55008841e-02 -1.20771623e+00 -6.00651562e-01 -3.68385196e-01 1.32140374e+00 -4.55913134e-02 -5.05182981e-01 -9.22935367e-01 -1.14296091e+00 1.03103504e-01 5.27476609e-01 -6.96752131e-01 -8.59289430e-03 -1.28432065e-01 -8.13087344e-01 5.45647383e-01 5.98888159e-01 7.57619441e-01 -1.05580854e+00 -2.31163442e-01 -4.26714085e-02 2.68749967e-02 -1.42366087e+00 -4.46738571e-01 6.26433194e-01 -1.02920234e+00 -9.58172500e-01 -1.06191015e+00 -9.40220654e-01 9.74812984e-01 5.11702836e-01 1.09588051e+00 2.62140363e-01 -4.63557214e-01 9.94129002e-01 -4.47115541e-01 -2.51321077e-01 -7.84179449e-01 2.72464186e-01 -3.02833140e-01 8.60782415e-02 1.30474567e-01 -2.68755972e-01 -4.27400321e-01 1.48001283e-01 -1.31327021e+00 -2.92503219e-02 7.12601423e-01 1.18151402e+00 4.92382377e-01 4.28675115e-01 2.02859730e-01 -1.33472347e+00 5.53195298e-01 -1.55567229e-01 -1.45607308e-01 3.95145357e-01 -1.01458979e+00 5.95951490e-02 6.75340891e-01 -4.12201673e-01 -1.31324124e+00 2.19198719e-01 1.88381374e-01 -3.11408877e-01 -5.99930823e-01 3.98698151e-01 1.71898618e-01 -7.13215694e-02 9.68323588e-01 4.50079292e-01 3.22390884e-01 -4.56784666e-01 3.05050641e-01 7.58211672e-01 7.27126598e-01 -5.13411105e-01 6.74994290e-01 7.82658398e-01 -1.16425738e-01 -1.09575284e+00 -1.25369132e+00 -6.96771383e-01 -9.19639289e-01 6.79896921e-02 9.29422438e-01 -9.33339655e-01 1.69892460e-02 3.91464829e-01 -8.93462300e-01 -3.89110535e-01 -4.09325004e-01 3.41445655e-01 -5.43356240e-01 7.81866074e-01 -6.04177296e-01 -5.34465730e-01 -3.16130608e-01 -8.79953682e-01 1.25295103e+00 -1.36101514e-01 -8.48013535e-02 -1.24322736e+00 -3.72216068e-02 7.13827252e-01 3.73925775e-01 -2.54580826e-01 9.07903194e-01 -6.97388291e-01 -1.63788155e-01 -1.28512010e-01 -5.13770521e-01 8.23457658e-01 2.59909809e-01 1.20206615e-02 -1.18818188e+00 -4.06004310e-01 -1.84504509e-01 -6.60867453e-01 1.37821651e+00 3.86187226e-01 1.21685743e+00 -3.22917402e-01 -2.51855761e-01 3.85406703e-01 1.51808846e+00 -1.94834962e-01 5.47990561e-01 5.82077980e-01 8.75329852e-01 7.55422056e-01 3.60051394e-01 1.23441882e-01 3.16675931e-01 4.93750691e-01 3.22492480e-01 -5.84477127e-01 -4.45342481e-01 -2.07075939e-01 2.73730844e-01 4.29820478e-01 7.24180564e-02 -3.65154326e-01 -7.95419276e-01 5.78238308e-01 -1.72908688e+00 -9.75839674e-01 -9.16176885e-02 1.70871329e+00 1.01763511e+00 1.96658015e-01 3.24891955e-01 4.32020307e-01 6.84275985e-01 2.43144408e-01 -3.94893438e-01 -3.02415550e-01 -5.71281731e-01 4.39799167e-02 8.21197093e-01 2.20110223e-01 -1.50735807e+00 9.92033780e-01 7.74057198e+00 8.79938424e-01 -9.30711508e-01 1.76051974e-01 6.98257565e-01 3.56215417e-01 -8.42784494e-02 -4.96286452e-02 -8.76988888e-01 4.34536785e-01 9.24603999e-01 3.48685890e-01 7.28714913e-02 9.99974489e-01 -1.24419332e-01 -4.36236709e-01 -1.19035602e+00 7.01099157e-01 6.22707784e-01 -1.60460114e+00 3.12034726e-01 -1.24753587e-01 1.02324963e+00 7.86115974e-02 1.80581465e-01 -1.10506333e-01 1.13264970e-01 -8.23777378e-01 5.39848983e-01 4.72184643e-02 6.46479905e-01 -5.51034808e-01 5.91159582e-01 3.71975571e-01 -5.43479741e-01 -1.21877372e-01 -1.89742938e-01 2.52893120e-01 -2.89726198e-01 7.27306008e-01 -8.16961288e-01 3.31370980e-01 8.84904206e-01 1.02229941e+00 -1.08603776e+00 6.61284745e-01 -7.62972087e-02 1.02294958e+00 -2.13359550e-01 2.04895228e-01 4.95179802e-01 -7.66600594e-02 2.93092638e-01 1.75446522e+00 -2.17196941e-01 -4.52311933e-01 2.35536918e-01 7.30865359e-01 1.91665515e-01 3.35169613e-01 -5.82079291e-01 -1.11897036e-01 -1.75278649e-01 1.05440569e+00 -1.09019518e+00 -8.41622651e-01 -5.89881182e-01 1.25532961e+00 6.06817603e-02 3.66554052e-01 -2.22985357e-01 6.52696490e-02 -1.32340759e-01 1.35441586e-01 6.08200729e-01 -1.43795878e-01 -2.08431914e-01 -1.14770842e+00 -5.60630374e-02 -1.01566267e+00 5.85858405e-01 -7.58432686e-01 -1.25246644e+00 5.33733487e-01 5.63823245e-02 -9.76627111e-01 -2.54398704e-01 -6.15175366e-01 -5.41960061e-01 3.01201314e-01 -1.77890849e+00 -1.52624476e+00 -2.26342469e-01 8.03813994e-01 6.55622602e-01 -4.71850932e-01 9.48352754e-01 3.24379206e-02 -3.81473005e-01 6.06267691e-01 6.38550639e-01 -1.28006963e-02 9.51879621e-01 -1.42827606e+00 -3.72141856e-03 6.59446299e-01 6.57963216e-01 3.56729209e-01 5.64598382e-01 -5.37408292e-01 -1.14003706e+00 -9.65276539e-01 4.08468455e-01 -3.66993427e-01 4.35770214e-01 -2.99327552e-01 -9.88370180e-01 4.97813195e-01 5.72066486e-01 5.34084328e-02 7.53187776e-01 5.51753193e-02 -5.94811141e-01 2.23337747e-02 -1.19269705e+00 3.64570737e-01 7.87613988e-01 -9.12942767e-01 -5.58957636e-01 7.47695744e-01 3.43698621e-01 6.07904270e-02 -6.96198285e-01 3.60492229e-01 1.81076750e-01 -1.03380299e+00 1.09297013e+00 -4.59608167e-01 7.65564859e-01 6.96283206e-02 2.79532745e-02 -1.29433596e+00 -2.00949475e-01 -2.49791235e-01 9.69379023e-02 1.42524219e+00 2.78930753e-01 -1.65009275e-01 1.13663328e+00 -1.27772139e-02 1.75780982e-01 -1.91737294e-01 -5.80497205e-01 -1.09677732e+00 1.05392464e-01 -1.46246031e-01 4.49246615e-02 1.19933081e+00 -6.36881292e-02 6.60480857e-01 -3.21085602e-01 -2.92362720e-01 8.21142673e-01 2.25663379e-01 7.41020441e-01 -9.82283711e-01 -2.77699769e-01 -3.29927266e-01 -3.20635229e-01 -7.66309917e-01 4.88826483e-01 -1.16529775e+00 2.20717434e-02 -1.38842797e+00 6.20879889e-01 -2.26177692e-01 -3.37224454e-01 5.94447553e-01 -6.46289438e-02 6.25059247e-01 1.13973320e-01 4.80683893e-01 -7.51728535e-01 3.91393274e-01 1.31652200e+00 -6.58432662e-01 8.61226618e-02 -2.32351691e-01 -4.52965528e-01 5.81876993e-01 5.93743086e-01 -4.89235461e-01 -4.59938049e-01 -3.77375990e-01 -3.65711093e-01 -4.26007509e-01 4.30326998e-01 -9.49917078e-01 1.58498660e-01 1.46000490e-01 6.21395767e-01 -8.77331436e-01 6.41745031e-02 -1.00728428e+00 -5.21925509e-01 5.06233513e-01 -8.19886804e-01 -2.89512604e-01 2.89769202e-01 7.46041000e-01 -5.01591086e-01 -8.11724484e-01 8.29785526e-01 -4.62960273e-01 -6.89216316e-01 -6.77465945e-02 -6.41573787e-01 -1.28773361e-01 6.59542382e-01 -3.15765053e-01 -3.31940234e-01 -3.93149346e-01 -6.29279435e-01 -7.09077641e-02 6.10747516e-01 4.00046706e-01 2.89090574e-01 -9.69246030e-01 -5.12074172e-01 1.35888875e-01 2.72577792e-01 -2.26380959e-01 1.17495187e-01 6.33609295e-01 -2.45076761e-01 6.38125837e-01 -3.24016690e-01 -9.92641509e-01 -1.58174884e+00 6.91451013e-01 9.11302771e-03 -4.31546062e-01 -7.85649657e-01 5.85130692e-01 1.70676708e-01 -3.41021657e-01 4.46476668e-01 -1.63228333e-01 -2.42411211e-01 1.50715888e-01 3.67055446e-01 1.36151547e-02 3.35287988e-01 -7.28947163e-01 -1.91968262e-01 4.50672090e-01 -5.61588168e-01 -1.67888358e-01 1.31543517e+00 6.13499992e-03 3.47277299e-02 2.71981061e-01 1.52931583e+00 -2.02203646e-01 -1.06206501e+00 -6.27671480e-01 1.22546755e-01 -4.58905160e-01 4.98155057e-01 -9.16289628e-01 -1.05131745e+00 6.58787608e-01 6.88531637e-01 2.15035632e-01 1.04756308e+00 -4.40098941e-02 3.77353519e-01 7.58732855e-01 -3.04330498e-01 -1.63828683e+00 6.52985334e-01 4.53729391e-01 7.14549303e-01 -1.76470029e+00 4.62185234e-01 -4.09629494e-01 -8.33121598e-01 9.89053071e-01 5.06215215e-01 -8.07132572e-02 7.32494175e-01 1.94611713e-01 1.80627286e-01 -2.72436470e-01 -5.05152047e-01 -4.43459392e-01 4.43320900e-01 8.02221894e-01 4.94681567e-01 -4.49119687e-01 -1.51353125e-02 1.06802166e-01 -9.35692191e-02 -9.69088227e-02 3.50751430e-01 1.34134281e+00 -2.46379059e-03 -1.06770444e+00 -5.19726038e-01 6.63376510e-01 -5.37138760e-01 -2.12196589e-01 -8.21498632e-01 6.04333520e-01 -6.38504326e-02 6.89900637e-01 2.73168236e-01 2.53892876e-02 4.98008402e-03 1.77842245e-01 7.24653661e-01 -7.87726939e-01 -7.70094872e-01 1.62349433e-01 1.63811773e-01 -1.47836998e-01 -1.16865683e+00 -7.16286182e-01 -9.22462165e-01 2.23618284e-01 -6.38349056e-01 -5.30803613e-02 1.02269924e+00 1.07316422e+00 2.11760908e-01 4.63466227e-01 6.66920006e-01 -9.28501070e-01 -4.65982527e-01 -1.07730687e+00 -6.16188765e-01 8.05222869e-01 3.45173746e-01 -3.63408089e-01 -4.36984897e-01 6.22095406e-01]
[9.622245788574219, 2.5111751556396484]
1f9db6a5-b596-4d0d-8b67-1aba1d238650
statistics-and-samples-in-distributional
1902.08102
null
http://arxiv.org/abs/1902.08102v1
http://arxiv.org/pdf/1902.08102v1.pdf
Statistics and Samples in Distributional Reinforcement Learning
We present a unifying framework for designing and analysing distributional reinforcement learning (DRL) algorithms in terms of recursively estimating statistics of the return distribution. Our key insight is that DRL algorithms can be decomposed as the combination of some statistical estimator and a method for imputing a return distribution consistent with that set of statistics. With this new understanding, we are able to provide improved analyses of existing DRL algorithms as well as construct a new algorithm (EDRL) based upon estimation of the expectiles of the return distribution. We compare EDRL with existing methods on a variety of MDPs to illustrate concrete aspects of our analysis, and develop a deep RL variant of the algorithm, ER-DQN, which we evaluate on the Atari-57 suite of games.
['Rémi Munos', 'Saurabh Kumar', 'Robert Dadashi', 'Mark Rowland', 'Will Dabney', 'Marc G. Bellemare']
2019-02-21
null
null
null
null
['distributional-reinforcement-learning']
['methodology']
[-3.19427580e-01 -4.08713222e-02 -2.05284834e-01 -2.80357957e-01 -1.19956470e+00 -8.54981184e-01 7.77161062e-01 1.44066766e-01 -6.97834134e-01 1.15366268e+00 1.32691428e-01 -8.45101833e-01 -6.26539946e-01 -9.20833051e-01 -7.77703762e-01 -7.85684526e-01 -4.72924501e-01 7.71266997e-01 9.82879698e-02 -3.71133476e-01 3.36117864e-01 4.07442302e-01 -1.42733562e+00 3.85676920e-02 6.85209930e-01 8.77781451e-01 1.87138356e-02 9.68836546e-01 1.02522723e-01 1.14491796e+00 -6.81130588e-01 -3.30120742e-01 1.07911803e-01 -5.51831067e-01 -7.14158356e-01 -4.14365947e-01 4.62643653e-02 -5.67753136e-01 -1.87463477e-01 9.98207450e-01 6.18200481e-01 3.33958745e-01 1.01148248e+00 -1.36393523e+00 -3.35162491e-01 9.95199263e-01 -6.57997012e-01 4.08667982e-01 3.27380955e-01 1.42410785e-01 1.40388262e+00 -2.78485209e-01 3.16272378e-01 1.33621573e+00 7.53411472e-01 4.06005800e-01 -1.47998893e+00 -3.55546534e-01 -1.13471985e-01 2.16854494e-02 -8.68595243e-01 -2.10017934e-01 3.60497415e-01 -3.52311224e-01 9.28819895e-01 5.58487922e-02 2.99321324e-01 1.21109664e+00 4.61413532e-01 1.15751958e+00 1.40072465e+00 -3.81214112e-01 7.06541181e-01 -2.11772084e-01 4.89200577e-02 4.56990600e-01 2.42481649e-01 7.77741671e-01 -3.14881563e-01 -4.69458044e-01 7.04058707e-01 -4.14392471e-01 2.98162997e-01 -7.65152752e-01 -6.94210887e-01 1.29289043e+00 -1.98742107e-01 -1.42771915e-01 -4.70426828e-01 6.91635907e-01 5.20989120e-01 3.99603337e-01 5.59000254e-01 1.56990722e-01 -5.15150428e-01 -5.56474626e-01 -7.42368519e-01 9.78404403e-01 1.01017964e+00 6.76132917e-01 6.14793777e-01 2.28445798e-01 -4.57211554e-01 5.28035581e-01 2.64303058e-01 7.50714421e-01 3.86956155e-01 -1.11531091e+00 4.77886766e-01 -3.59721988e-01 4.58459854e-01 -3.29066187e-01 -4.03354585e-01 -4.56124395e-01 -3.02441418e-01 4.02777642e-01 6.97949767e-01 -6.27511561e-01 -5.06495893e-01 2.05307674e+00 -6.32397383e-02 1.75638840e-01 3.34962666e-01 4.16015834e-01 2.01514393e-01 4.75690871e-01 7.70182982e-02 -7.23870769e-02 7.07731605e-01 -2.17444658e-01 -4.13716465e-01 9.93231162e-02 8.54377210e-01 -2.05655992e-01 1.08035278e+00 5.62071562e-01 -1.10723352e+00 -2.33247787e-01 -7.64773965e-01 4.52809036e-01 -9.80829448e-02 -5.96499182e-02 1.00065804e+00 8.07767987e-01 -1.28842962e+00 8.33866835e-01 -8.23484123e-01 1.56257730e-02 4.00617838e-01 2.33952090e-01 2.24378988e-01 2.77677625e-01 -1.09030521e+00 7.60838687e-01 4.00077581e-01 -4.37585413e-01 -1.26286662e+00 -7.71453202e-01 -7.74815559e-01 3.33619006e-02 3.56452584e-01 -3.67444187e-01 1.85427368e+00 -7.79297829e-01 -1.61164939e+00 5.86505830e-01 3.65193576e-01 -9.48940158e-01 6.01178944e-01 -2.16315359e-01 -1.50581673e-01 -2.07843095e-01 1.06014401e-01 1.56428382e-01 7.14708865e-01 -1.10634768e+00 -9.90895510e-01 -1.02835849e-01 1.73201695e-01 4.90327887e-02 2.56787837e-01 -2.42261946e-01 3.57512355e-01 -4.87795442e-01 -6.95662081e-01 -6.15141690e-01 -2.73125648e-01 -8.25052798e-01 -3.44746321e-01 -5.87208211e-01 2.55404323e-01 -3.17300647e-01 1.17874515e+00 -1.90210354e+00 1.44858703e-01 4.86716330e-01 2.47074276e-01 -3.90885323e-02 -3.38529438e-01 6.57296479e-01 -2.78778225e-01 -1.34545386e-01 -3.12353283e-01 -1.51763946e-01 5.84960222e-01 4.14823383e-01 -8.67541134e-01 5.99465489e-01 -1.58724729e-02 8.09441626e-01 -1.16019249e+00 -9.15037841e-02 4.93965894e-02 -1.79759949e-01 -9.31811929e-01 3.88164818e-01 -5.40182412e-01 -8.22509602e-02 -4.84141827e-01 2.60002375e-01 5.88264585e-01 2.09736332e-01 1.61489844e-01 4.26200420e-01 -1.83420911e-01 3.31397951e-01 -1.08643126e+00 1.33177972e+00 -3.64269465e-01 2.84883887e-01 -1.89333677e-01 -1.34151876e+00 7.19761312e-01 -2.85170943e-01 6.81905746e-01 -6.40445173e-01 1.54706582e-01 1.41319200e-01 8.72442648e-02 -3.12108576e-01 4.20835525e-01 -5.80727935e-01 -4.08765823e-01 1.06083822e+00 3.19055915e-01 -8.44561532e-02 4.32148397e-01 1.39780432e-01 1.42065239e+00 5.03041923e-01 3.16351563e-01 -6.04689598e-01 1.73741549e-01 -2.39489987e-01 1.22181185e-01 1.40040982e+00 -1.80706367e-01 -1.06792487e-01 1.18128097e+00 -4.45311517e-01 -9.82246280e-01 -1.86242557e+00 -1.76101550e-01 1.35464394e+00 -4.89772379e-01 -3.98187488e-01 -6.60117269e-01 -1.00407124e+00 4.26705837e-01 1.22145748e+00 -7.41921842e-01 -1.05739973e-01 -1.27234325e-01 -1.01294935e+00 8.84817898e-01 5.07530749e-01 2.39723459e-01 -1.25167608e+00 -7.23707497e-01 1.11845426e-01 1.81573227e-01 -5.89804292e-01 -9.29476097e-02 6.15233362e-01 -5.05088091e-01 -1.16134608e+00 -3.68980765e-01 -3.03979725e-01 -8.07031766e-02 -3.40770751e-01 1.55274761e+00 -2.40541175e-01 -7.77477585e-03 7.26554155e-01 -2.36741856e-01 -6.48335516e-01 -7.15332210e-01 1.24152131e-01 2.95976996e-01 -4.74317968e-01 5.59173703e-01 -6.70005441e-01 -1.74340904e-01 -2.31151506e-01 -1.04494190e+00 -5.55589736e-01 4.67812806e-01 7.70481110e-01 2.82828510e-01 2.57594109e-01 8.21926653e-01 -9.73401546e-01 1.18621075e+00 -7.99285650e-01 -1.03540802e+00 3.78285982e-02 -4.08295393e-01 6.73836768e-01 6.85874701e-01 -1.88776180e-01 -1.02958822e+00 -2.66865045e-01 -4.09994185e-01 -9.48182121e-03 -8.16364214e-02 5.33865869e-01 1.12322368e-01 5.18088341e-01 5.15822470e-01 3.44598293e-01 2.38677949e-01 -2.32354149e-01 6.67918444e-01 5.30199111e-01 5.05925953e-01 -1.27827919e+00 8.55199814e-01 1.23626634e-01 1.56426579e-01 -6.44172311e-01 -1.15935016e+00 -4.64265160e-02 -3.17346334e-01 -9.53565538e-02 5.09883344e-01 -6.61853552e-01 -1.19564724e+00 3.74896526e-01 -6.20943308e-01 -1.01948380e+00 -7.03792274e-01 4.96531248e-01 -1.47029817e+00 1.80043593e-01 -5.62188447e-01 -1.16498113e+00 1.52419016e-01 -9.99733090e-01 8.78924251e-01 -6.23669587e-02 -1.21618398e-01 -1.44280791e+00 7.79392660e-01 -3.63187820e-01 3.47767144e-01 8.03558826e-02 1.23607874e+00 -9.63527858e-01 -1.91665795e-02 2.80305237e-01 -1.28043219e-01 4.79521930e-01 -2.95374077e-02 -1.77188724e-01 -7.65147328e-01 -4.57329720e-01 -2.55398564e-02 -7.20860362e-01 9.54631805e-01 7.76461422e-01 1.34887362e+00 -1.15237229e-01 1.89446941e-01 4.95882690e-01 1.52186155e+00 -2.63288175e-03 5.41120350e-01 5.95725596e-01 1.00772277e-01 3.16729665e-01 6.73056662e-01 9.88536239e-01 3.31120491e-01 2.93643475e-01 3.57503295e-01 2.39939928e-01 3.42613518e-01 -6.20544136e-01 7.84328401e-01 2.73892343e-01 3.55196605e-03 -1.80044875e-01 -7.67208934e-01 2.93313146e-01 -1.99521840e+00 -1.40738821e+00 3.67170185e-01 2.24358535e+00 1.00342810e+00 2.09112108e-01 9.37010169e-01 -1.54106930e-01 4.49988097e-01 2.22564504e-01 -8.08099866e-01 -7.37662375e-01 8.20060596e-02 8.38952243e-01 8.37204278e-01 7.34849036e-01 -9.37548518e-01 9.33003306e-01 8.06825924e+00 1.29904962e+00 -4.67646748e-01 -1.92029044e-01 6.04722440e-01 1.27012432e-01 -5.56459606e-01 -1.16008706e-01 -7.69658923e-01 3.43264401e-01 1.22501183e+00 -3.99086505e-01 7.92890728e-01 9.55991864e-01 1.43846631e-01 -3.24600905e-01 -1.30152774e+00 6.77974164e-01 -2.33072937e-01 -1.29194951e+00 -1.34429578e-02 2.96320766e-01 5.02496362e-01 1.87375963e-01 3.91641468e-01 8.96490693e-01 1.46327257e+00 -1.28013468e+00 6.40811384e-01 6.36421978e-01 6.31605864e-01 -1.43948519e+00 8.55893254e-01 4.76864785e-01 -6.76729858e-01 -2.46372268e-01 -5.93146861e-01 -4.67906177e-01 -4.44818169e-01 4.00455236e-01 -4.70667094e-01 4.30887818e-01 5.55274844e-01 7.75053084e-01 -2.33134434e-01 7.15270996e-01 -2.18089342e-01 8.50424409e-01 -2.37701461e-01 -1.86441511e-01 4.66554999e-01 -3.80051911e-01 4.20181066e-01 1.02663922e+00 3.71998191e-01 -9.51339379e-02 1.66742966e-01 1.11733866e+00 5.88835701e-02 -2.47861385e-01 -7.22573757e-01 2.20315624e-02 3.24104220e-01 1.02433455e+00 -3.45527142e-01 -1.14246428e-01 -1.99434727e-01 3.01018298e-01 7.16862082e-01 3.72724175e-01 -1.10391581e+00 -3.67793411e-01 1.04158998e+00 -2.06498668e-01 5.36681950e-01 -1.59934014e-01 1.20004177e-01 -9.15789187e-01 -2.92361736e-01 -1.02868354e+00 5.54533124e-01 -4.66405153e-01 -1.64383662e+00 1.59626186e-01 3.93906176e-01 -9.14129853e-01 -8.42149079e-01 -7.84135282e-01 -5.67481279e-01 6.72957778e-01 -1.52054846e+00 -4.32390869e-01 6.15795732e-01 5.86437643e-01 1.40624180e-01 -5.81704438e-01 7.97857344e-01 -3.07364464e-01 -4.16329205e-01 4.81499255e-01 5.86297691e-01 -6.66448027e-02 3.69333297e-01 -1.92669475e+00 2.37818167e-01 4.60383445e-01 2.20874518e-01 1.50481641e-01 1.02163637e+00 -4.09844518e-01 -1.50274909e+00 -8.80295575e-01 6.35646954e-02 -7.84085393e-01 1.20893478e+00 -2.20865309e-01 -3.03164810e-01 9.19673383e-01 1.94572523e-01 -3.20591599e-01 7.30307519e-01 2.95210719e-01 -1.93541884e-01 -5.45207001e-02 -1.09211588e+00 5.60221791e-01 7.55813122e-01 -3.32332820e-01 -7.67116249e-01 1.39836684e-01 4.63030815e-01 -3.45525831e-01 -6.20324552e-01 -5.62888309e-02 3.44267219e-01 -1.39748287e+00 9.80516195e-01 -9.62615550e-01 6.61792397e-01 8.51749703e-02 -4.41597730e-01 -1.82237589e+00 -1.93570942e-01 -9.15040135e-01 -1.34698689e-01 1.08830500e+00 1.54795364e-01 -6.96520805e-01 6.86282277e-01 1.65802419e-01 1.18995026e-01 -8.27200592e-01 -9.41901505e-01 -1.04351747e+00 7.92419016e-01 -9.70032334e-01 6.04612589e-01 3.29792678e-01 1.84764490e-01 5.53510599e-02 -3.10383558e-01 4.46511284e-02 1.03301561e+00 3.73663269e-02 7.73440123e-01 -9.62453902e-01 -6.37085915e-01 -6.19824708e-01 -1.81519806e-01 -1.05395603e+00 6.90091848e-01 -9.58950996e-01 1.23358957e-01 -1.06310713e+00 3.49103421e-01 -4.53935832e-01 -4.94711548e-01 4.08657432e-01 2.27598518e-01 -1.59240618e-01 -1.35846987e-01 -3.62637818e-01 -8.68838012e-01 7.92685330e-01 7.14383841e-01 2.44593471e-01 -7.19747245e-02 3.16598266e-01 -7.79104829e-01 8.19253743e-01 9.41960275e-01 -5.71826041e-01 -5.91696858e-01 1.44017681e-01 5.77392757e-01 3.15912902e-01 4.28049654e-01 -7.42378414e-01 -1.81013897e-01 -4.26697016e-01 4.26330268e-01 -7.52595365e-01 -2.72812784e-01 -2.72031635e-01 -5.94478071e-01 5.14426112e-01 -7.74756670e-01 4.15510356e-01 3.13452542e-01 4.90442276e-01 4.67107892e-02 -3.23705912e-01 6.25230134e-01 4.07627858e-02 -6.21028304e-01 1.71469063e-01 -7.94397593e-01 7.96344757e-01 8.01375329e-01 3.87356967e-01 -2.46967912e-01 -7.48911142e-01 -3.96693766e-01 2.59938627e-01 1.99057505e-01 -2.23928280e-02 5.99020600e-01 -1.25891292e+00 -8.51060212e-01 1.34854823e-01 -1.27201647e-01 -4.78607476e-01 -2.07963511e-02 5.37460566e-01 -3.82389039e-01 3.50660890e-01 -2.05451742e-01 -2.17885569e-01 -6.46707177e-01 4.56538051e-01 4.19437140e-01 -7.52635062e-01 -4.50179815e-01 5.33999085e-01 9.49896127e-02 -5.67895949e-01 1.99847862e-01 -3.75739366e-01 -3.22256275e-02 -7.16087446e-02 5.47471344e-01 2.36066908e-01 -3.01319361e-01 -1.06511220e-01 -2.28508189e-01 1.50158793e-01 1.11911930e-01 -3.65916908e-01 1.67646325e+00 -1.62435099e-02 2.54979413e-02 5.89745581e-01 7.80628622e-01 4.27933931e-02 -1.48895705e+00 -7.98281804e-02 2.00430647e-01 -1.44029945e-01 -1.27000967e-02 -7.62245119e-01 -8.39532852e-01 6.10879540e-01 2.17254981e-01 4.21675265e-01 1.01174986e+00 1.09316647e-01 4.74747390e-01 5.81554711e-01 5.60750723e-01 -1.20463324e+00 1.20065555e-01 7.89910793e-01 5.76602876e-01 -8.96438897e-01 1.46923199e-01 5.39243102e-01 -8.10158491e-01 1.22754085e+00 1.75607860e-01 -6.56631291e-01 7.13434637e-01 6.80181265e-01 -3.54651064e-01 -1.57291099e-01 -9.70388114e-01 -5.79189181e-01 6.68985164e-03 1.00939584e+00 3.43632281e-01 1.61868408e-01 -3.18409242e-02 8.90484989e-01 -7.23492086e-01 -1.97105020e-01 7.18682051e-01 6.27148211e-01 -5.52707911e-01 -1.27569282e+00 5.74783832e-02 5.31326413e-01 -5.77997327e-01 -8.72120112e-02 8.61795768e-02 1.16020942e+00 -3.91217053e-01 7.36349165e-01 2.72064418e-01 -3.00981253e-01 7.63003156e-02 -1.16352499e-01 7.47412145e-01 -5.39362967e-01 -1.14205554e-01 -1.39057547e-01 2.06060365e-01 -6.27840817e-01 -2.60649502e-01 -8.69462132e-01 -1.10017133e+00 -6.20871782e-01 2.63032883e-01 3.82966161e-01 2.94133753e-01 1.29807055e+00 -2.52611369e-01 4.11740780e-01 7.46804833e-01 -7.84718394e-01 -1.51577270e+00 -5.95303833e-01 -1.30505037e+00 1.47751212e-01 3.46759707e-01 -6.58452153e-01 -5.18569231e-01 -6.13159299e-01]
[4.020107746124268, 2.646852970123291]
1ab67cac-9d62-421b-a0c9-30a3a1bed33a
dependency-parsing-past-present-and-future
null
null
https://aclanthology.org/C14-3006
https://aclanthology.org/C14-3006.pdf
Dependency Parsing: Past, Present, and Future
null
['Zhenghua Li', 'Wenliang Chen', 'Min Zhang']
2014-08-01
dependency-parsing-past-present-and-future-1
https://aclanthology.org/C14-3006
https://aclanthology.org/C14-3006.pdf
coling-2014-8
['lexical-analysis']
['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]
[-7.174190521240234, 3.425464153289795]
c9ec9f42-33ed-46ab-967d-d039f1ad6594
rotation-equivariant-cnns-for-digital
1806.03962
null
http://arxiv.org/abs/1806.03962v1
http://arxiv.org/pdf/1806.03962v1.pdf
Rotation Equivariant CNNs for Digital Pathology
We propose a new model for digital pathology segmentation, based on the observation that histopathology images are inherently symmetric under rotation and reflection. Utilizing recent findings on rotation equivariant CNNs, the proposed model leverages these symmetries in a principled manner. We present a visual analysis showing improved stability on predictions, and demonstrate that exploiting rotation equivariance significantly improves tumor detection performance on a challenging lymph node metastases dataset. We further present a novel derived dataset to enable principled comparison of machine learning models, in combination with an initial benchmark. Through this dataset, the task of histopathology diagnosis becomes accessible as a challenging benchmark for fundamental machine learning research.
['Taco Cohen', 'Bastiaan S. Veeling', 'Max Welling', 'Jasper Linmans', 'Jim Winkens']
2018-06-08
null
null
null
null
['breast-tumour-classification']
['medical']
[ 4.50964391e-01 6.50376081e-01 -5.81355572e-01 -1.14835225e-01 -7.72606075e-01 -6.85860217e-01 6.02853358e-01 -1.30173657e-02 -3.10772717e-01 2.33267158e-01 3.17221642e-01 -6.67196929e-01 -1.78226814e-01 -2.87964702e-01 -5.30546904e-01 -9.30555880e-01 -1.44357942e-02 2.05985606e-01 -9.75317359e-02 -2.61865139e-01 -1.35354295e-01 1.13400853e+00 -5.20020008e-01 3.03309441e-01 1.71364665e-01 6.15252376e-01 -2.50851274e-01 7.55104899e-01 5.42975605e-01 5.22540092e-01 7.10185915e-02 -3.52882445e-01 5.27753830e-01 -2.56091386e-01 -1.11528850e+00 -1.25912605e-02 5.92485666e-01 -1.36416033e-01 -5.85248351e-01 9.22291756e-01 3.47169399e-01 -2.59435266e-01 7.38394380e-01 -9.28715646e-01 -4.40106481e-01 4.27314669e-01 -3.66329342e-01 2.90019035e-01 -1.54696271e-01 1.33207843e-01 1.24517655e+00 -4.33954716e-01 1.50742352e+00 7.87750483e-01 9.44845140e-01 8.69273305e-01 -1.34537280e+00 -1.25301376e-01 2.75392784e-03 -1.23753950e-01 -1.00496948e+00 -3.62291157e-01 7.16131687e-01 -2.13949323e-01 6.61424816e-01 5.57648897e-01 9.58260059e-01 1.28350806e+00 8.64069760e-01 6.32472098e-01 1.10845566e+00 -2.40692168e-01 -1.64971557e-02 -1.50607675e-01 1.07637174e-01 1.21489763e+00 4.24595922e-01 -5.46724424e-02 -3.14212710e-01 9.50143486e-03 9.65624452e-01 2.25791812e-01 -4.55160111e-01 -8.08464468e-01 -1.49709284e+00 7.16509283e-01 8.47321272e-01 2.61078179e-02 1.12848550e-01 1.91674724e-01 2.32695475e-01 5.29521108e-02 2.61154473e-01 3.57324570e-01 -2.23112613e-01 3.76623809e-01 -6.65079534e-01 -3.85009684e-02 6.50368750e-01 6.01500511e-01 1.41416922e-01 -2.74019651e-02 -1.73228160e-02 3.60949963e-01 4.32058781e-01 3.97374779e-01 6.20957613e-01 -7.81736970e-01 -1.61938071e-01 5.87061226e-01 -5.53514957e-01 -8.56773317e-01 -1.07006454e+00 -9.07851338e-01 -1.16665912e+00 2.09761769e-01 5.12966990e-01 4.19781208e-01 -1.13457692e+00 1.60408604e+00 3.98318797e-01 1.65680069e-02 7.96260014e-02 7.13169932e-01 8.22409928e-01 -3.50760460e-01 -4.40214481e-03 1.01739839e-02 1.33445752e+00 -6.44778371e-01 -5.07024109e-01 3.57663125e-01 1.23740613e+00 -4.85390127e-01 6.03472233e-01 2.42479786e-01 -7.73650110e-01 3.26708466e-01 -1.00553477e+00 -4.59348351e-01 -2.86083907e-01 1.75991371e-01 9.37615812e-01 6.08019948e-01 -1.43562531e+00 5.26193798e-01 -1.44335604e+00 -7.83825517e-01 8.78187358e-01 4.25183147e-01 -8.38022172e-01 1.28209546e-01 -4.52476859e-01 7.75842249e-01 -2.11380105e-02 2.58748859e-01 -9.03280199e-01 -1.20859551e+00 -9.95975137e-01 -2.63435066e-01 -2.73728907e-01 -1.01135802e+00 1.32080662e+00 -5.43511033e-01 -1.48548305e+00 1.36000168e+00 -6.92191049e-02 -7.39385068e-01 1.15914142e+00 5.80179989e-01 -2.72269733e-02 5.59467912e-01 -3.22832197e-01 9.25166368e-01 5.05080879e-01 -1.01191819e+00 -2.58831769e-01 -4.40865397e-01 -1.42196551e-01 4.70304191e-02 -3.18205982e-01 -4.99292642e-01 -2.35768497e-01 -5.84852695e-01 6.39082909e-01 -1.23672628e+00 -6.15050077e-01 4.00165766e-01 -8.78012538e-01 4.56006259e-01 9.45889175e-01 -5.95013022e-01 3.96000683e-01 -1.89770067e+00 1.36755720e-01 5.63174844e-01 8.08931947e-01 -2.02434897e-01 -4.29917350e-02 -1.71652094e-01 -4.57956702e-01 3.95858943e-01 -8.57291371e-02 -2.09530354e-01 -1.92800894e-01 9.99760404e-02 -2.39417762e-01 1.07038915e+00 1.31110877e-01 1.50268137e+00 -7.78405130e-01 -6.73676610e-01 2.00621262e-01 5.89396358e-01 -7.79012620e-01 -3.63975346e-01 2.27893680e-01 5.51422000e-01 -3.48954767e-01 8.71168435e-01 5.05717695e-01 -5.99794269e-01 7.06076145e-01 -5.87800384e-01 2.14523852e-01 -5.86617459e-03 -4.52412039e-01 1.65715706e+00 -2.64225096e-01 7.98120379e-01 1.12610407e-01 -9.62638736e-01 4.40612376e-01 2.01291427e-01 9.33136344e-01 -4.80725676e-01 3.35695565e-01 -3.54184881e-02 1.44696727e-01 -3.23758006e-01 7.95387179e-02 -2.59993166e-01 2.20634311e-01 3.95405322e-01 2.46027365e-01 -2.75960028e-01 -1.24655493e-01 4.15895104e-01 1.30390573e+00 -5.12883924e-02 4.85558510e-01 -5.36630154e-01 2.07955092e-01 2.06698179e-01 3.86704683e-01 5.57739139e-01 -6.54044330e-01 8.76630068e-01 8.66312861e-01 -7.28842735e-01 -1.00663972e+00 -1.26555681e+00 -5.52056134e-01 4.84901458e-01 -1.54951699e-02 -1.07877634e-01 -6.98523521e-01 -1.02135146e+00 -5.74662276e-02 -1.93185180e-01 -1.22561264e+00 -9.50122848e-02 -6.37654841e-01 -1.20167530e+00 6.79444909e-01 4.96981114e-01 1.90419510e-01 -4.77786362e-01 -4.09333676e-01 -3.29676837e-01 1.06221147e-01 -1.41447151e+00 -1.56852417e-02 2.84756452e-01 -1.18393350e+00 -1.51328731e+00 -8.18339765e-01 -9.32463706e-01 1.21027660e+00 4.82771337e-01 8.07733774e-01 2.25167945e-01 -1.07439053e+00 7.38334715e-01 -7.26014376e-03 -4.04407591e-01 -5.15070498e-01 3.42621446e-01 -1.84582233e-01 -1.71099901e-01 -1.87955916e-01 -3.65526706e-01 -8.36670697e-01 1.53083771e-01 -9.46542084e-01 3.58563453e-01 8.07941198e-01 8.58301938e-01 9.64440346e-01 -5.84458292e-01 2.52861157e-02 -1.09249687e+00 1.20861322e-01 -2.78273463e-01 -2.83651084e-01 1.52843386e-01 -2.19674423e-01 2.93610245e-02 4.43872094e-01 -1.34758558e-02 -8.38246047e-01 3.01836550e-01 -5.49002066e-02 -1.98835507e-01 -1.22947417e-01 3.93614501e-01 3.27076823e-01 -8.22444379e-01 8.06903780e-01 3.72454599e-02 4.68777299e-01 -5.55587076e-02 2.84468532e-01 7.19308183e-02 6.15734220e-01 -1.97039559e-01 9.78769600e-01 1.46579647e+00 9.34562027e-01 -8.55816960e-01 -7.66705811e-01 -3.82370949e-01 -1.00281394e+00 -9.32158828e-02 9.75239992e-01 -7.66318023e-01 -8.31359506e-01 2.99713671e-01 -7.92181075e-01 -4.48608696e-01 -2.57075012e-01 4.43581462e-01 -6.21699035e-01 4.16858643e-01 -8.09316218e-01 9.29355621e-02 -3.70442867e-01 -1.16628885e+00 1.13009548e+00 -1.78240929e-02 -2.92369366e-01 -1.51015151e+00 2.10084423e-01 2.56231755e-01 3.83489788e-01 7.06065118e-01 1.09054530e+00 -8.12922537e-01 -7.01824009e-01 -3.40674222e-01 -1.70097500e-01 -1.48297787e-01 5.86985052e-03 3.75761449e-01 -8.32371056e-01 -5.03446698e-01 -2.72200704e-01 -1.97358847e-01 1.16084790e+00 3.48828286e-01 1.30692971e+00 4.35898006e-02 -7.17108667e-01 1.20085239e+00 1.24361491e+00 -6.00561917e-01 3.32615137e-01 4.60800350e-01 9.46333528e-01 4.79960591e-01 -1.34750009e-01 -1.51894584e-01 3.63681585e-01 4.33403552e-01 7.10519433e-01 -6.92393541e-01 -6.42631531e-01 -6.10286668e-02 -1.35939464e-01 8.15716147e-01 -2.74101973e-01 2.08981723e-01 -1.18230486e+00 4.12642807e-01 -1.58352458e+00 -5.95268667e-01 -4.71764691e-02 1.80062532e+00 5.49695611e-01 5.09030418e-03 -2.13622317e-01 -5.06501310e-02 3.56785625e-01 8.20232928e-02 -3.56454104e-01 -8.99728239e-02 -2.65706807e-01 2.80141056e-01 8.52769136e-01 2.94770777e-01 -1.32921469e+00 6.54611170e-01 8.12357521e+00 4.99244243e-01 -1.63178623e+00 -1.60380565e-02 7.98392951e-01 5.39135821e-02 -5.02902210e-01 -1.97364628e-01 -5.24133027e-01 -3.75046432e-01 5.99079549e-01 -1.84467196e-01 -1.62502974e-01 6.41056240e-01 2.05288649e-01 4.00924981e-01 -1.34492981e+00 6.17864907e-01 1.22912511e-01 -1.78086627e+00 4.83458519e-01 4.57574010e-01 9.60514963e-01 1.81970045e-01 7.26188421e-01 -3.17071855e-01 3.58796984e-01 -1.23751783e+00 1.97853938e-01 2.59958118e-01 8.52732837e-01 -4.31065589e-01 7.04586565e-01 -2.07175210e-01 -8.25259566e-01 1.54139265e-01 -1.40047818e-01 5.34293234e-01 -1.79318339e-01 3.14262390e-01 -1.40210247e+00 8.26227725e-01 4.13869172e-01 1.26059902e+00 -8.75020623e-01 9.72962141e-01 -2.11246192e-01 4.61583495e-01 -1.12027831e-01 4.27609652e-01 1.40188336e-01 1.55418172e-01 6.28382325e-01 1.45343685e+00 1.89806774e-01 -1.78952917e-01 -1.88793093e-01 5.79799235e-01 -3.00566703e-01 9.38716084e-02 -6.67053342e-01 2.67720282e-01 -1.85781240e-01 1.92607784e+00 -1.32577538e+00 9.90193412e-02 -1.04593776e-01 5.18013299e-01 3.87616426e-01 4.94783580e-01 -6.84605300e-01 -1.78278815e-02 5.88448346e-01 5.43139242e-02 1.39081702e-01 -2.70247340e-01 -5.29465199e-01 -1.41090965e+00 -2.69003540e-01 -7.14951158e-01 4.86515462e-01 -4.28741992e-01 -8.81188810e-01 1.39995053e-01 -3.31103176e-01 -1.29168880e+00 2.64193285e-02 -1.33348787e+00 -6.74850583e-01 1.49059922e-01 -1.86518502e+00 -1.59056270e+00 -4.76453334e-01 6.23082638e-01 6.69530453e-03 -8.92728791e-02 8.86011899e-01 -2.50422329e-01 -7.39731014e-01 7.83701241e-01 1.83584228e-01 4.34995621e-01 7.27083921e-01 -1.39511323e+00 6.24821447e-02 7.64352262e-01 2.11368546e-01 7.00781941e-01 4.88648653e-01 -2.84888685e-01 -1.68815351e+00 -1.25539815e+00 2.58185029e-01 -7.18472481e-01 8.32483590e-01 -2.76077360e-01 -4.52366441e-01 1.12553799e+00 -1.37968421e-01 5.88177323e-01 1.07749999e+00 -2.19941691e-01 -5.60535491e-01 1.08708732e-03 -1.31215465e+00 8.92651021e-01 1.06750703e+00 -5.70814371e-01 -1.25939474e-01 6.63105071e-01 5.40504754e-01 -7.84830153e-01 -9.46335137e-01 5.81162274e-01 8.12375069e-01 -6.51506186e-01 1.08585489e+00 -9.31018353e-01 6.48266554e-01 5.19666821e-02 8.86747539e-02 -1.28055048e+00 -2.36176223e-01 -6.04098558e-01 4.50854674e-02 2.78782278e-01 5.09246290e-01 -7.81696081e-01 1.05298996e+00 2.36077800e-01 -3.05009991e-01 -8.75351489e-01 -1.16469991e+00 -6.07303858e-01 4.13397491e-01 -1.57313377e-01 -1.33951455e-01 1.09496248e+00 2.17866972e-01 -4.71935481e-01 2.60211140e-01 1.56571791e-01 7.52977252e-01 -7.90170804e-02 5.27643621e-01 -1.05877423e+00 -5.85445203e-02 -7.38567472e-01 -1.04778349e+00 -6.05525315e-01 2.46951163e-01 -1.70790577e+00 -3.92328084e-01 -1.16236091e+00 5.82120419e-01 3.61179151e-02 -5.42211235e-01 6.94983423e-01 3.03742290e-01 9.45343673e-01 9.72526707e-03 3.07232887e-01 -6.81211591e-01 3.83496657e-02 1.67865813e+00 -3.38389546e-01 2.31153414e-01 -2.79894143e-01 -9.28167343e-01 9.32504594e-01 7.39669323e-01 -2.93415695e-01 -2.04795357e-02 -1.51324421e-01 2.39393458e-01 -2.42247999e-01 8.17113936e-01 -8.09748471e-01 1.53577670e-01 -7.69613981e-02 7.80711532e-01 -6.87262788e-02 1.17944546e-01 -5.75705826e-01 -1.85306162e-01 9.91212130e-01 -4.92802769e-01 -1.83120728e-01 1.96147025e-01 7.09715843e-01 5.82409799e-02 3.70180249e-01 7.96296954e-01 -6.00408986e-02 -2.51965016e-01 5.43529093e-01 -3.18661362e-01 -1.54920802e-01 1.18199325e+00 -1.20239787e-01 -7.07204878e-01 -6.98586926e-02 -1.07122445e+00 -1.04740754e-01 6.19449914e-01 7.50083700e-02 4.96813506e-01 -1.15872753e+00 -6.32927716e-01 1.64731637e-01 2.20719278e-01 -3.43593769e-02 3.45451176e-01 1.29249620e+00 -1.10853064e+00 6.48767710e-01 -4.40211982e-01 -9.55960631e-01 -1.39052308e+00 9.59896147e-02 8.58185828e-01 -6.62785947e-01 -7.21387744e-01 6.25808358e-01 2.93680638e-01 -6.14678681e-01 -8.84587616e-02 -5.70167720e-01 -3.22040409e-01 -3.56496394e-01 8.89423266e-02 1.43667370e-01 3.58984500e-01 -5.98992169e-01 -4.08004880e-01 6.05521500e-01 -4.57339376e-01 5.71633969e-03 1.30955589e+00 1.67723045e-01 -1.48899391e-01 -9.20202583e-02 1.42071235e+00 -9.73400176e-02 -1.12432694e+00 -8.67745131e-02 -2.05978975e-01 -1.37695059e-01 2.04816950e-03 -5.35127163e-01 -1.18733144e+00 8.74846816e-01 4.78449941e-01 -3.48863788e-02 7.63662875e-01 9.79157016e-02 4.18476611e-01 7.71008909e-01 -7.77262226e-02 -6.84466124e-01 2.16001853e-01 4.77003336e-01 6.97627246e-01 -1.28714085e+00 2.65715837e-01 -7.16621041e-01 -1.54255286e-01 1.48506844e+00 3.67668867e-01 -2.08567187e-01 7.70326197e-01 3.35373998e-01 4.90020156e-01 -3.67041975e-01 -5.80301046e-01 1.27421826e-01 4.09450799e-01 7.97045469e-01 5.09084821e-01 1.65370271e-01 -1.21335655e-01 2.33298540e-01 -5.13907850e-01 -2.92959392e-01 8.22792351e-01 8.68605614e-01 -7.74908885e-02 -8.55139375e-01 -1.46318460e-02 2.85572737e-01 -7.71474659e-01 2.64939904e-01 -5.92651427e-01 1.13288510e+00 -4.60361183e-01 1.43026307e-01 9.24200490e-02 1.41169932e-02 1.05067498e-04 -1.16683535e-01 6.53633058e-01 -4.64488715e-01 -2.10365295e-01 -1.83056936e-01 -3.67476404e-01 -7.19687760e-01 -4.34307486e-01 -6.20120943e-01 -1.27835524e+00 -1.84192047e-01 1.30863070e-01 -4.29755867e-01 8.16129863e-01 9.22414124e-01 2.54648894e-01 7.61204541e-01 5.42063177e-01 -7.93528080e-01 -5.79597890e-01 -3.78306001e-01 -3.73874694e-01 4.85523969e-01 5.94321012e-01 -2.06662640e-01 -3.94383639e-01 2.44667009e-01]
[15.061229705810547, -2.9261105060577393]
18ab874d-550a-449c-8ab7-c341ef34e882
bert-rankers-are-brittle-a-study-using
2206.11724
null
https://arxiv.org/abs/2206.11724v1
https://arxiv.org/pdf/2206.11724v1.pdf
BERT Rankers are Brittle: a Study using Adversarial Document Perturbations
Contextual ranking models based on BERT are now well established for a wide range of passage and document ranking tasks. However, the robustness of BERT-based ranking models under adversarial inputs is under-explored. In this paper, we argue that BERT-rankers are not immune to adversarial attacks targeting retrieved documents given a query. Firstly, we propose algorithms for adversarial perturbation of both highly relevant and non-relevant documents using gradient-based optimization methods. The aim of our algorithms is to add/replace a small number of tokens to a highly relevant or non-relevant document to cause a large rank demotion or promotion. Our experiments show that a small number of tokens can already result in a large change in the rank of a document. Moreover, we find that BERT-rankers heavily rely on the document start/head for relevance prediction, making the initial part of the document more susceptible to adversarial attacks. More interestingly, we find a small set of recurring adversarial words that when added to documents result in successful rank demotion/promotion of any relevant/non-relevant document respectively. Finally, our adversarial tokens also show particular topic preferences within and across datasets, exposing potential biases from BERT pre-training or downstream datasets.
['Avishek Anand', 'Lijun Lyu', 'Yumeng Wang']
2022-06-23
null
null
null
null
['document-ranking']
['natural-language-processing']
[ 3.01633179e-01 -6.39630258e-02 -2.10998915e-02 -1.81416199e-01 -1.51922154e+00 -1.42708778e+00 1.27806556e+00 5.47699571e-01 -4.59665149e-01 7.70009279e-01 6.15428030e-01 -3.33225489e-01 -4.86559391e-01 -9.19995546e-01 -1.04006529e+00 -6.77383602e-01 -3.31690460e-01 7.93296933e-01 5.42002976e-01 -7.89372027e-01 5.49765229e-01 6.91779792e-01 -1.06673050e+00 4.82554764e-01 6.64161921e-01 5.47160089e-01 -3.17732871e-01 1.02653134e+00 2.36201540e-01 6.36040509e-01 -1.36498821e+00 -8.38934720e-01 5.24482310e-01 1.95556045e-01 -1.10117865e+00 -8.68608952e-01 7.54718363e-01 -5.09675384e-01 -5.06714940e-01 9.19250250e-01 9.29170430e-01 4.63438541e-01 9.92373288e-01 -1.02127135e+00 -6.76733196e-01 1.06886411e+00 -1.43690854e-01 5.32243967e-01 4.99235719e-01 -3.51299495e-02 1.33118927e+00 -5.54050326e-01 6.92490160e-01 1.37626791e+00 2.44759813e-01 5.37609458e-01 -8.26401889e-01 -5.41292727e-01 3.06931168e-01 1.93487599e-01 -9.71460581e-01 -2.76003361e-01 7.71424294e-01 -2.02778596e-02 6.57377720e-01 9.76443052e-01 2.16921419e-01 1.52416456e+00 3.97603601e-01 7.47452974e-01 8.59726369e-01 -2.37581119e-01 2.70357579e-01 9.21493620e-02 8.97302553e-02 -2.20602117e-02 1.27453491e-01 3.95536810e-01 -4.63223666e-01 -6.97219193e-01 -2.18681186e-01 -6.04575351e-02 4.13726596e-03 2.79654413e-02 -8.43697727e-01 9.26513672e-01 6.64834738e-01 3.46542984e-01 -6.94004372e-02 3.89304817e-01 6.62113070e-01 5.15698433e-01 5.67640603e-01 1.14692736e+00 -4.83665347e-01 -7.72350132e-02 -7.26229370e-01 6.61621511e-01 6.58255577e-01 6.48750603e-01 3.63865495e-01 -4.25349712e-01 -9.88062441e-01 5.43329954e-01 -7.95111284e-02 6.43858671e-01 3.74688268e-01 -5.71056545e-01 6.89971089e-01 9.32656527e-02 2.13149473e-01 -1.16755331e+00 -1.95700731e-02 -5.69546878e-01 -5.09466052e-01 -2.83793285e-02 3.24710876e-01 4.79258262e-02 -7.36071944e-01 1.79707718e+00 6.76234290e-02 -5.41400462e-02 1.38714552e-01 7.59730518e-01 5.81203282e-01 6.40385568e-01 5.11435866e-02 3.48730423e-02 1.12770903e+00 -6.05275571e-01 -2.91620165e-01 -3.79521132e-01 4.99937743e-01 -9.48588073e-01 1.29017711e+00 3.37008774e-01 -1.10289490e+00 -6.71508089e-02 -9.29408193e-01 7.17564598e-02 -8.40064704e-01 -7.65097797e-01 5.16273737e-01 6.99694872e-01 -8.81677449e-01 7.01121151e-01 -7.18567446e-02 -1.73260123e-01 1.93141580e-01 4.50224012e-01 -1.00900225e-01 -1.23104133e-01 -2.07164097e+00 9.54336882e-01 -5.58897574e-03 -7.21087009e-02 -1.40049696e+00 -7.14303017e-01 -1.96509480e-01 7.07696564e-03 4.75142539e-01 -6.04988933e-01 1.06705093e+00 -4.72197562e-01 -1.11254585e+00 5.19144475e-01 1.20357975e-01 -5.88986278e-01 9.12919223e-01 -8.18298221e-01 -4.44291323e-01 2.18273729e-01 1.23381630e-01 3.10122490e-01 1.11670125e+00 -1.37829316e+00 -4.36307847e-01 -3.26402932e-01 5.17241240e-01 5.26188970e-01 -7.17589378e-01 1.88039660e-01 -3.13909739e-01 -1.02349639e+00 -4.25882190e-01 -8.61485720e-01 -2.77599931e-01 -7.46836245e-01 -9.03829515e-01 -1.35802060e-01 6.62916422e-01 -3.25262010e-01 1.30536282e+00 -1.92061830e+00 1.21724661e-02 6.79402411e-01 1.00902356e-01 3.67737293e-01 -3.26293826e-01 8.23225379e-01 1.68096393e-01 8.13871205e-01 1.91099539e-01 6.31420314e-02 1.06930651e-01 4.16595973e-02 -1.18374503e+00 3.04397702e-01 -7.64439628e-02 9.91025865e-01 -1.29943609e+00 -1.80493027e-01 -1.79461122e-01 1.18582837e-01 -6.41760528e-01 3.53723159e-03 -2.58363158e-01 8.10249522e-02 -6.62622809e-01 4.27978784e-01 5.00653446e-01 3.25496137e-01 -2.93794125e-01 1.88515589e-01 5.09076357e-01 7.02647746e-01 -7.53741205e-01 1.00787115e+00 -4.05762464e-01 6.26550019e-01 -2.80932516e-01 -5.88154137e-01 5.08582950e-01 1.06909104e-01 6.22549802e-02 -5.39592147e-01 -3.51388082e-02 1.91400483e-01 -2.55350351e-01 -2.90743764e-02 1.10792434e+00 -8.01245347e-02 -4.91643816e-01 6.87739372e-01 -6.00129962e-01 -5.08308113e-01 2.60831773e-01 1.02887356e+00 1.56160378e+00 -7.05059350e-01 -3.67285758e-01 -1.07560158e-02 4.61035728e-01 -1.52498513e-01 1.28242567e-01 1.57242703e+00 -7.55998418e-02 5.80137074e-01 7.08435297e-01 7.58295357e-02 -8.68539929e-01 -1.26050556e+00 -1.05558895e-01 1.67111814e+00 2.69503057e-01 -3.39558244e-01 -4.09120798e-01 -1.18279636e+00 2.96772838e-01 9.01044369e-01 -8.06477606e-01 -7.37743735e-01 -6.21053815e-01 -5.48996627e-01 9.90486026e-01 2.36102208e-01 -1.96791831e-02 -1.23370063e+00 3.98193002e-02 8.43831748e-02 -5.96805476e-02 -5.76504469e-01 -7.32479870e-01 3.18180978e-01 -6.33404732e-01 -8.90761614e-01 -6.94116712e-01 -3.23486984e-01 6.17477834e-01 1.87933430e-01 1.09657609e+00 2.70046648e-02 1.54817611e-01 4.93783295e-01 -5.56445241e-01 -4.25476253e-01 -6.95829749e-01 3.37304682e-01 1.75251812e-01 -4.27449584e-01 -1.80557802e-01 -2.46463463e-01 -6.50667369e-01 5.07969379e-01 -1.33699417e+00 -8.84794831e-01 3.57040733e-01 7.05781400e-01 2.03403220e-01 2.47880891e-01 6.80334687e-01 -1.15916479e+00 1.27873623e+00 -5.16904652e-01 -1.57580018e-01 3.90364170e-01 -7.06447959e-01 1.64687499e-01 8.50581646e-01 -7.74952888e-01 -7.39253044e-01 -7.26539671e-01 1.07942156e-01 -1.97391093e-01 3.56008202e-01 5.03116548e-01 -1.22952357e-01 2.75612950e-01 1.22067928e+00 -3.31298523e-02 -6.70539677e-01 -1.93807214e-01 7.52664447e-01 4.14558679e-01 4.84692007e-01 -7.87229836e-01 1.38823009e+00 5.17614007e-01 -7.06828982e-02 -1.66507497e-01 -6.60164475e-01 -5.81097364e-01 -1.77509263e-01 -1.78015098e-01 2.64319301e-01 -4.82702285e-01 -3.88042659e-01 2.25958735e-01 -9.68996465e-01 -1.59864336e-01 -4.97895360e-01 2.34107599e-02 -1.05907314e-01 4.05611932e-01 -4.80315000e-01 -4.32512760e-01 -6.41421795e-01 -1.08292687e+00 9.76508379e-01 -8.06430355e-02 -3.43460202e-01 -1.04873574e+00 2.19826937e-01 3.22764575e-01 4.29691255e-01 1.25496581e-01 1.06051862e+00 -1.32005906e+00 -4.99627978e-01 -9.42870378e-01 2.20390081e-01 2.37537771e-01 -1.77003637e-01 1.50733247e-01 -9.68331337e-01 -5.63076377e-01 -3.81486088e-01 -3.60620379e-01 9.53514636e-01 6.88724294e-02 8.85161757e-01 -8.08516562e-01 -3.89973104e-01 2.79018968e-01 8.50950122e-01 -5.81986196e-02 5.47916293e-01 7.33618557e-01 4.36277032e-01 5.63733578e-01 9.14950669e-01 2.29128763e-01 -2.97004372e-01 7.20902383e-01 6.94293380e-01 3.32568586e-01 2.73796409e-01 -4.17879283e-01 7.23266721e-01 7.45478868e-02 2.79344797e-01 -7.83006728e-01 -7.22914398e-01 6.08700454e-01 -1.49672329e+00 -1.20027173e+00 4.12529074e-02 2.43951988e+00 1.10182333e+00 6.92494690e-01 -6.98599815e-02 1.50533654e-02 7.14379489e-01 4.79716688e-01 -4.77760881e-01 -6.91374898e-01 -2.99999923e-01 1.25624865e-01 7.29433835e-01 6.79476440e-01 -1.08991551e+00 1.27021372e+00 6.54289532e+00 1.21614480e+00 -8.39141965e-01 -1.59170464e-01 6.15047693e-01 -6.81267619e-01 -9.19032395e-01 1.72050167e-02 -7.52132535e-01 3.92593741e-01 8.63056719e-01 -6.39377236e-01 5.24011493e-01 8.62029672e-01 -3.04426439e-02 2.10271955e-01 -1.11895370e+00 1.02428436e-01 -5.41792773e-02 -1.13877273e+00 5.26005268e-01 1.20028697e-01 1.00216186e+00 -1.47993460e-01 6.98819518e-01 4.98149037e-01 1.04224801e+00 -1.04859686e+00 8.02222610e-01 3.31410706e-01 6.19833469e-01 -1.05239105e+00 8.44824493e-01 1.21781819e-01 -5.22302210e-01 -2.06381753e-01 -5.37209213e-01 4.27619159e-01 -3.85851488e-02 8.05782020e-01 -1.12778389e+00 4.31430936e-01 6.49493873e-01 1.47943065e-01 -9.17950928e-01 6.96001649e-01 -5.67475080e-01 6.62040651e-01 -2.57872969e-01 -1.92249194e-01 4.00019467e-01 4.58399773e-01 1.12643278e+00 1.12047184e+00 -1.16257379e-02 -1.81722000e-01 -1.99040711e-01 1.95940450e-01 -4.26878899e-01 1.71546310e-01 -6.85405552e-01 6.06437726e-03 7.26522326e-01 9.88859653e-01 -2.50625193e-01 -3.95480961e-01 4.15473849e-01 1.03477561e+00 2.96018302e-01 5.76528430e-01 -7.49418080e-01 -4.33456123e-01 6.53269947e-01 1.19529236e-02 -5.59879467e-02 3.69757593e-01 -2.70773888e-01 -7.11692572e-01 4.22292612e-02 -1.23664165e+00 5.66915929e-01 -6.59494460e-01 -1.42606878e+00 4.18974936e-01 1.19637258e-01 -9.94762003e-01 -4.83233273e-01 -1.18274063e-01 -9.75085974e-01 7.90439010e-01 -1.24182022e+00 -8.48752618e-01 1.48499981e-01 6.38278604e-01 2.83392310e-01 -4.10234571e-01 6.48263097e-01 -7.35512525e-02 -2.59053767e-01 1.08454597e+00 5.69463909e-01 5.00859469e-02 1.30185831e+00 -1.46630061e+00 5.29876769e-01 1.00040257e+00 1.33154109e-01 1.16638541e+00 1.08038127e+00 -9.20852602e-01 -1.08815539e+00 -1.10288513e+00 1.05345404e+00 -9.97628391e-01 8.08519959e-01 -3.88547033e-01 -7.25172460e-01 4.21859264e-01 -1.41369864e-01 -2.57623136e-01 5.02426147e-01 4.27367747e-01 -6.39260411e-01 -1.68961138e-01 -1.19829810e+00 1.16828740e+00 9.10202801e-01 -7.86125481e-01 -6.46151662e-01 7.64908254e-01 8.86768281e-01 -2.52419740e-01 -3.83133292e-01 4.90144432e-01 4.67332512e-01 -7.02507377e-01 1.44410920e+00 -1.03946030e+00 7.68379927e-01 1.38103310e-02 -1.77310079e-01 -1.49946642e+00 -3.05569082e-01 -1.07268977e+00 -2.27846369e-01 1.26608908e+00 4.89053696e-01 -4.66289610e-01 7.02788174e-01 5.88970482e-01 -3.29170227e-02 -5.03636777e-01 -9.81819093e-01 -8.63227308e-01 5.85496187e-01 -4.94032472e-01 6.60363972e-01 6.83220148e-01 2.15748325e-03 8.33503082e-02 -4.16978925e-01 2.67928988e-01 2.22928643e-01 -2.73576051e-01 9.13010955e-01 -9.04159129e-01 -2.77459949e-01 -6.03443682e-01 -1.60834625e-01 -8.04594100e-01 8.81965160e-02 -1.05340111e+00 -8.10153689e-03 -1.54006207e+00 1.02805763e-01 -5.56723416e-01 -6.73231602e-01 2.80709624e-01 -7.30153024e-01 2.63692856e-01 1.86160073e-01 4.87534165e-01 -6.98773980e-01 3.80045116e-01 1.10389555e+00 -5.90736330e-01 -1.63448170e-01 4.75214452e-01 -1.05675948e+00 3.40001613e-01 7.86049247e-01 -8.36064935e-01 -4.75988299e-01 -2.03042462e-01 8.17759335e-01 -1.93230465e-01 2.91469872e-01 -4.11342353e-01 1.36735052e-01 -8.03511217e-02 2.20262423e-01 -5.65854609e-01 1.58957601e-01 -6.87152445e-01 -4.28847790e-01 5.40279031e-01 -1.06652832e+00 4.63705473e-02 9.19113159e-02 8.87084484e-01 1.10237271e-01 -5.83930552e-01 4.34294105e-01 5.88622317e-03 -2.31928006e-01 1.23000279e-01 -4.95411068e-01 4.05814350e-01 9.20157194e-01 1.08914077e-01 -7.66357839e-01 -7.86618054e-01 -5.75379252e-01 2.04811156e-01 2.17820659e-01 5.95924079e-01 4.37759757e-01 -1.11595082e+00 -8.46157193e-01 -2.58623064e-01 1.34946108e-01 -1.67723179e-01 9.95835736e-02 6.99903369e-02 -2.67897397e-01 5.26241899e-01 3.56248796e-01 8.24188534e-03 -1.33337545e+00 6.53913796e-01 7.14056343e-02 -9.03066635e-01 -1.09616749e-01 1.18529880e+00 1.51090816e-01 -3.88240576e-01 3.47381771e-01 1.80196688e-01 -1.39551803e-01 3.29749167e-01 2.94824839e-01 6.39954329e-01 2.76913255e-01 -2.62050033e-01 -4.42977786e-01 1.67732000e-01 -7.88436472e-01 -6.02561116e-01 9.40457523e-01 1.86210915e-01 6.12016022e-02 -5.49327433e-02 1.31129777e+00 7.90241778e-01 -8.07786584e-01 -2.24632680e-01 -1.23142965e-01 -6.24857962e-01 -8.05129111e-02 -1.05315340e+00 -7.27627695e-01 5.71078598e-01 3.13471526e-01 4.31552678e-01 8.27870429e-01 1.11273736e-01 6.50318801e-01 7.46275961e-01 1.51817143e-01 -1.10408211e+00 5.66372931e-01 6.91715837e-01 1.30973208e+00 -9.75494027e-01 3.09347719e-01 6.62842989e-02 -7.34332323e-01 5.00502527e-01 3.63119841e-01 4.09298241e-02 4.06776309e-01 -1.66971788e-01 2.66964659e-02 -1.63981915e-01 -8.23941648e-01 2.10475475e-01 5.87487161e-01 6.06918156e-01 -3.35740298e-02 -1.89356711e-02 -2.34917983e-01 2.32533410e-01 -6.78683937e-01 -8.73813927e-01 4.18975353e-01 8.70366037e-01 -4.40889746e-01 -1.20943475e+00 -6.66620076e-01 5.84559679e-01 -9.67206717e-01 -5.86056948e-01 -9.54450309e-01 5.25702178e-01 -3.19296092e-01 1.20831442e+00 -2.59552747e-01 -5.89509785e-01 3.33175808e-01 4.99575213e-03 -4.45455983e-02 -6.34072483e-01 -1.24076378e+00 -2.05058828e-01 7.86494240e-02 -3.39470625e-01 4.17669803e-01 -6.75201893e-01 -9.68997002e-01 -4.88244861e-01 -3.71576101e-01 4.32372391e-01 3.92940789e-01 7.27549314e-01 2.36532569e-01 4.72143263e-01 9.34811413e-01 -4.59642649e-01 -1.15630376e+00 -9.65936422e-01 -4.60100949e-01 8.22863877e-01 2.67910331e-01 -2.53053099e-01 -9.22541380e-01 -3.48841310e-01]
[6.043377876281738, 8.131400108337402]
268f92d2-0d3f-4409-a8a8-494a608aadc2
an-iterative-multi-path-fully-convolutional
null
null
https://doi.org/10.1002/mp.13859
https://doi.org/10.1002/mp.13859
An Iterative Multi‐path Fully Convolutional Neural Network for Automatic Cardiac Segmentation in Cine MR Images
Purpose: Segmentation of the left ventricle (LV), right ventricle (RV) cavities and the myocardium (MYO) from cine cardiac magnetic resonance (MR) images is an important step for diagnosis and monitoring cardiac diseases. Spatial context information may be highly beneficial for segmentation performance improvement. To this end, this paper proposes an iterative multi-path fully convolutional network (IMFCN) to effectively leverage spatial context for automatic cardiac segmentation in cine MR images. Methods: To effectively leverage spatial context information, the proposed IMFCN explicitly models the interslice spatial correlations using a multi-path late fusion strategy. First, the contextual inputs including both the adjacent slices and the already predicted mask of the above adjacent slice are processed by independent feature-extraction paths. Then, an atrous spatial pyramid pooling (ASPP) module is employed at the feature fusion process to combine the extracted high-level contextual features in a more effective way. Finally, deep supervision (DS) and batch-wise class re-weighting mechanism are utilized to enhance the training of the proposed network. Results: The proposed IMFCN was evaluated and analyzed on the MICCAI 2017 automatic cardiac diagnosis challenge (ACDC) dataset. On the held-out training dataset reserved for testing, our method effectively improved its counterparts that without spatial context and that with spatial context but using an early fusion strategy. On the 50 subjects test dataset, our method achieved Dice similarity coefficient of 0.935, 0.920, and 0.905, and Hausdorff distance of 7.66, 12.10, and 8.80 mm for LV, RV, and MYO, respectively, which are comparable or even better than the state-of-the-art methods of ACDC Challenge. In addition, to explore the applicability to other datasets, the proposed IMFCN was retrained on the Sunnybrook dataset for LV segmentation and also produced comparable performance to the state-of-the-art methods. Conclusions: We have presented an automatic end-to-end fully convolutional architecture for accurate cardiac segmentation. The proposed method provides an effective way to leverage spatial context in a two-dimensional manner and results in precise and consistent segmentation results. Keywords: automatic segmentation; cardiac MRI; deep learning; fully convolutional network.
['Jiliu Zhou', 'Yan Wang 1', 'Kunlin Cao 3', 'Youbing Yin 3', 'Qi Song 3', 'Xin Wang 3', 'Xi Wu 2', 'Zongqing Ma 1']
2019-11-01
null
null
null
med-phy-2019-11
['cardiac-segmentation']
['medical']
[ 2.57256150e-01 -1.27954140e-01 2.27284193e-01 -4.96820867e-01 -7.80781567e-01 -3.68752301e-01 1.61203057e-01 2.54670233e-01 -4.96396244e-01 6.30352795e-01 4.61827666e-02 -1.24110542e-01 -2.24076092e-01 -3.66434574e-01 -2.90639877e-01 -9.56912637e-01 -4.87985492e-01 3.97318959e-01 2.83191383e-01 1.96919143e-01 1.39525369e-01 6.81812823e-01 -1.02214682e+00 3.09044123e-01 8.50719571e-01 1.02459466e+00 3.62488776e-01 9.19262528e-01 1.93904981e-01 8.29162478e-01 -3.49133790e-01 1.80497155e-01 1.84658676e-01 -4.26143765e-01 -9.96463180e-01 1.53588548e-01 3.83265167e-01 -3.93386781e-01 8.66178274e-02 7.20820844e-01 8.96428227e-01 -1.32620819e-02 4.97516721e-01 -4.97725725e-01 -1.31690845e-01 5.67559302e-01 -7.24490821e-01 7.99144030e-01 -3.80248278e-01 1.26465365e-01 5.32054007e-01 -8.41299832e-01 6.93792939e-01 5.66202104e-01 8.66403878e-01 2.74545193e-01 -1.00990367e+00 -5.28419316e-01 -2.23930433e-01 1.06686642e-02 -1.10344398e+00 1.34051694e-02 7.97834098e-01 -6.05059981e-01 7.10232258e-01 2.27094218e-01 7.09093809e-01 2.85054982e-01 3.65785718e-01 6.07196212e-01 1.07769608e+00 -2.75933087e-01 2.56899856e-02 -1.90240324e-01 2.94648349e-01 6.74395859e-01 -1.20843109e-02 1.93678476e-02 5.77387847e-02 -1.22472495e-01 8.88996422e-01 1.37082055e-01 -3.70229453e-01 -3.12108427e-01 -1.63068879e+00 6.23616636e-01 6.81025445e-01 8.09902251e-01 -7.09164143e-01 -2.32972465e-02 7.67779768e-01 -8.91999751e-02 4.22691524e-01 4.15668100e-01 -4.91048008e-01 3.38755064e-02 -1.30599427e+00 1.53931111e-01 2.88692594e-01 3.77535254e-01 2.34751910e-01 -7.58382455e-02 -4.36247379e-01 8.82968605e-01 1.91185564e-01 3.52911532e-01 6.00274920e-01 -1.10526550e+00 4.49564993e-01 4.08549905e-01 -3.43176275e-01 -9.41700995e-01 -8.30837965e-01 -9.14985836e-01 -1.30010986e+00 6.04355484e-02 3.93200636e-01 -3.00906479e-01 -1.14579678e+00 1.48258746e+00 5.23683429e-01 4.04845178e-01 -2.52912585e-02 1.34951758e+00 8.59755039e-01 3.99081975e-01 1.79729462e-01 -3.78164083e-01 1.51678765e+00 -8.77230406e-01 -7.21405566e-01 1.56615213e-01 1.03612840e+00 -6.75314486e-01 6.84020460e-01 1.81289107e-01 -1.09981525e+00 -7.14519322e-01 -1.07672763e+00 3.26696515e-01 9.90411341e-02 4.09981281e-01 4.63919818e-01 4.70504075e-01 -1.17223048e+00 9.19822395e-01 -1.14816511e+00 3.05240564e-02 6.36361122e-01 4.35961694e-01 -4.22981620e-01 -1.70858547e-01 -1.10765958e+00 6.04500473e-01 3.65509897e-01 4.50317532e-01 -5.70974290e-01 -9.95158851e-01 -6.61288023e-01 -1.98833451e-01 -1.87399853e-02 -7.70501435e-01 7.86519349e-01 -8.08635473e-01 -1.13631344e+00 9.28912878e-01 1.58915415e-01 -6.28243089e-01 3.69308323e-01 -2.36487538e-02 -1.25199944e-01 7.39551783e-01 2.47299552e-01 8.21835160e-01 5.72687984e-01 -9.37695503e-01 -5.02123594e-01 -5.94659328e-01 -3.08146626e-01 3.12496901e-01 9.14889276e-02 1.08887978e-01 -2.72419661e-01 -9.41361427e-01 4.31323111e-01 -9.81709421e-01 -4.44504112e-01 -2.07293317e-01 -1.85275778e-01 1.85355633e-01 8.43684316e-01 -1.26737607e+00 1.31946230e+00 -2.20111585e+00 1.60674274e-01 2.88118541e-01 6.32829010e-01 4.60396856e-01 2.06344903e-01 -3.29414189e-01 -2.56390631e-01 1.94194838e-01 -7.75047183e-01 -1.71687588e-01 -7.58379400e-01 -1.47009697e-02 3.90270412e-01 6.48070455e-01 7.60905370e-02 8.71104479e-01 -7.06232309e-01 -8.22136700e-01 3.83422226e-01 6.36252403e-01 -4.69584733e-01 1.22329704e-01 4.49925601e-01 1.05413795e+00 -2.66815931e-01 5.88704407e-01 7.51925945e-01 -3.03216130e-01 2.32903063e-01 -3.99107397e-01 3.04880049e-02 -2.78671056e-01 -9.52416897e-01 1.77730501e+00 -3.05334836e-01 4.37588096e-01 1.66847527e-01 -1.17735636e+00 8.53776157e-01 6.92900598e-01 1.06298232e+00 -6.55752182e-01 1.15283363e-01 4.70155925e-01 4.42472279e-01 -4.96609330e-01 -6.00888804e-02 -2.39766911e-01 1.74158111e-01 2.23530307e-01 5.87966517e-02 3.00657228e-02 1.04015388e-01 1.72865763e-01 9.65779424e-01 3.36398110e-02 7.03101382e-02 -5.98879755e-01 8.96869898e-01 -1.14389263e-01 8.05656552e-01 5.84247351e-01 -5.73660791e-01 1.02634645e+00 2.53887266e-01 -7.60208786e-01 -1.00991929e+00 -8.30381632e-01 -4.60795790e-01 3.54890853e-01 -9.39452723e-02 -1.33872375e-01 -9.77966428e-01 -7.41263032e-01 -3.63443732e-01 2.11938873e-01 -6.46423340e-01 2.13497579e-01 -1.11179066e+00 -8.26139867e-01 6.14348710e-01 9.26100850e-01 7.83136308e-01 -1.10954797e+00 -1.04913235e+00 4.93429065e-01 -3.85075033e-01 -1.12717450e+00 -5.29845119e-01 9.33985971e-03 -1.47332847e+00 -1.04374576e+00 -1.16748571e+00 -7.96826184e-01 5.63275635e-01 -9.77324992e-02 1.06541312e+00 3.47847849e-01 -5.23442030e-01 7.81705719e-04 -2.39786357e-01 4.01033834e-02 -2.34417841e-01 9.61347390e-03 -3.06350976e-01 -2.24153012e-01 -2.61547148e-01 -5.56461811e-01 -1.03666866e+00 2.13862821e-01 -7.87986577e-01 1.50959134e-01 5.48826396e-01 1.24037015e+00 8.55761170e-01 -1.23360261e-01 6.99287534e-01 -9.12626386e-01 2.63490885e-01 -3.79157573e-01 -3.50938976e-01 1.01007350e-01 -6.46109343e-01 -2.88362175e-01 5.57970643e-01 -6.91950992e-02 -8.51526558e-01 2.13704780e-01 -2.52640784e-01 -5.40361762e-01 -2.32900307e-01 5.07002413e-01 2.34731302e-01 1.06834061e-02 4.01494503e-01 8.28485712e-02 1.53750762e-01 -2.95314968e-01 2.29241773e-02 5.19964457e-01 6.03815079e-01 -4.49850082e-01 1.74582809e-01 4.95887458e-01 3.54705334e-01 -5.24130166e-01 -4.68286216e-01 -6.20668113e-01 -1.21482146e+00 -2.11197481e-01 1.43486917e+00 -6.99836612e-01 -1.57978684e-01 6.53957307e-01 -9.97599959e-01 -2.72875607e-01 -2.47818768e-01 7.07285583e-01 -4.68905210e-01 5.62374711e-01 -9.49376047e-01 -3.23305607e-01 -9.46826100e-01 -1.47578573e+00 8.94772768e-01 1.26325458e-01 -6.93248436e-02 -1.20400476e+00 -2.85178907e-02 6.65675104e-01 6.01230264e-01 7.70635247e-01 8.97472382e-01 -7.90258169e-01 -2.99706578e-01 -3.06396913e-02 -2.59781569e-01 6.82026744e-01 1.86097696e-01 -2.56696790e-01 -7.56435037e-01 -3.82094264e-01 4.53410782e-02 2.75820345e-02 7.75217116e-01 9.41921949e-01 1.21780491e+00 9.95661840e-02 -1.28665224e-01 8.08628678e-01 1.22469008e+00 3.45246136e-01 4.71025854e-01 1.41986176e-01 8.19644988e-01 4.05315429e-01 6.46015882e-01 3.31758201e-01 3.06200653e-01 5.15400290e-01 2.30168477e-01 -6.29251540e-01 -2.63336122e-01 5.11792839e-01 -2.48975202e-01 9.94514346e-01 -3.59660894e-01 3.76955330e-01 -1.19432938e+00 7.20728159e-01 -1.59866345e+00 -7.54260957e-01 -3.23771089e-01 1.98266280e+00 7.94963598e-01 8.38119164e-03 -2.44106855e-02 3.19649369e-01 7.62585640e-01 -5.39310202e-02 -3.46286863e-01 -3.77160609e-01 -4.47327532e-02 5.19216299e-01 4.52376992e-01 2.30661467e-01 -1.57231569e+00 4.85692561e-01 4.91806221e+00 5.41577101e-01 -1.35417306e+00 2.93996274e-01 1.17223346e+00 1.60408631e-01 2.40335628e-01 -1.32322952e-01 -3.06531340e-01 3.57389182e-01 8.56829405e-01 3.88035774e-01 3.57109345e-02 4.44015354e-01 2.38876119e-01 -9.12674740e-02 -7.08082199e-01 9.49419737e-01 -9.94803607e-02 -1.53030682e+00 -4.54061300e-01 2.53226999e-02 8.71668756e-01 1.52554318e-01 -1.30274966e-01 1.09237157e-01 -5.94124019e-01 -1.00959718e+00 3.13682646e-01 5.66207230e-01 8.93908739e-01 -8.19428146e-01 1.37092924e+00 2.06563100e-01 -1.16970420e+00 -5.80555089e-02 5.74154481e-02 3.51119578e-01 2.06806496e-01 8.27296495e-01 -9.17311907e-01 7.22903371e-01 7.90334225e-01 7.08547831e-01 -5.58538318e-01 9.12474096e-01 2.34426677e-01 8.45391512e-01 -1.99905708e-01 5.52298605e-01 3.49368870e-01 -4.34390735e-03 6.76544309e-01 1.40225434e+00 2.32927427e-01 3.54764283e-01 2.45040655e-01 6.86706722e-01 4.22739536e-02 3.30773205e-01 -2.57907003e-01 3.83345813e-01 1.29041076e-01 1.56580496e+00 -1.09009135e+00 -6.89873517e-01 -2.16073290e-01 6.87400460e-01 -9.77029204e-02 1.94160759e-01 -8.74873340e-01 -2.74136782e-01 4.69517894e-02 3.70524466e-01 3.04169118e-01 -8.13635252e-03 -6.44809127e-01 -9.38296795e-01 7.03569651e-02 -8.07318151e-01 5.63877702e-01 -4.47720170e-01 -1.06615913e+00 7.99014926e-01 -7.13484827e-03 -1.11317837e+00 -2.38911495e-01 -2.73870826e-01 -6.15376353e-01 1.14232600e+00 -1.37566566e+00 -1.05825984e+00 -2.11253375e-01 4.04045314e-01 4.32174891e-01 -1.26390710e-01 7.82633126e-01 5.45537651e-01 -3.53815228e-01 3.54795605e-01 -2.28222683e-01 4.83859837e-01 5.23276448e-01 -1.39842093e+00 -1.51247540e-02 9.19237733e-01 -3.20151985e-01 6.75733745e-01 1.22021496e-01 -6.82986438e-01 -7.29523361e-01 -1.26431882e+00 6.70323730e-01 -2.22802795e-02 5.40934242e-02 2.44677976e-01 -1.05745661e+00 2.64521301e-01 1.19906053e-01 6.25689983e-01 6.05319977e-01 -4.11367327e-01 2.88457185e-01 -5.18789589e-02 -1.43612504e+00 1.49953514e-01 5.95437467e-01 -1.61569104e-01 -5.31170130e-01 9.27171856e-02 5.30568123e-01 -6.75347149e-01 -1.46800995e+00 1.00153518e+00 4.71492887e-01 -9.89561558e-01 1.00101995e+00 -2.43320867e-01 5.58059335e-01 -4.46418315e-01 -4.24650349e-02 -9.63912785e-01 -3.15344483e-01 -1.83966920e-01 -1.89847481e-02 9.09248233e-01 2.63899118e-01 -4.37669665e-01 5.13693929e-01 3.48017216e-01 -5.69350302e-01 -1.30596924e+00 -1.00042748e+00 -2.11634189e-01 3.03141445e-01 -2.58591622e-01 2.95187473e-01 1.08000124e+00 -4.09779429e-01 -1.13886334e-02 -3.41379270e-02 3.69229726e-02 7.47869670e-01 -8.17944482e-03 1.12924680e-01 -1.14370406e+00 -2.26848170e-01 -4.13305610e-01 -5.56742311e-01 -4.89195049e-01 -1.38936564e-01 -1.12240112e+00 -1.67162135e-01 -1.51609099e+00 3.18734944e-01 -6.36810362e-01 -7.30486095e-01 2.21202895e-01 -4.28753823e-01 4.37845916e-01 2.91632235e-01 3.39672953e-01 -2.90930063e-01 4.99216653e-02 1.72708881e+00 5.04203737e-02 -2.46742249e-01 8.76445994e-02 -4.19330269e-01 6.38496280e-01 1.01194561e+00 -2.51470506e-01 -3.31664830e-01 -1.99546739e-01 -3.95915806e-01 5.59248984e-01 6.15336061e-01 -1.31562543e+00 6.91632181e-03 4.60805953e-01 7.26287961e-01 -8.68842602e-01 -6.45522997e-02 -6.20274365e-01 -2.69463402e-03 6.54109538e-01 -2.90522426e-01 1.60612941e-01 1.09936163e-01 1.03007555e-01 -3.90391618e-01 6.57768920e-02 1.10945451e+00 -3.10610473e-01 -3.04944605e-01 4.13684100e-01 -2.52259552e-01 1.51661843e-01 1.04423761e+00 -2.56215394e-01 1.78243324e-01 7.96355158e-02 -1.10765791e+00 1.78723544e-01 -4.82483245e-02 5.58916330e-02 7.72370756e-01 -1.10984600e+00 -9.26953077e-01 1.58342287e-01 -2.99089998e-01 3.23065758e-01 6.95860386e-01 1.69863558e+00 -9.41935003e-01 4.50086594e-01 -2.21099764e-01 -1.26349282e+00 -1.29268610e+00 3.02751511e-01 7.53427744e-01 -7.97696233e-01 -1.05372131e+00 6.94289982e-01 7.95549378e-02 -5.17146528e-01 1.54909752e-02 -7.66631246e-01 -3.69100362e-01 -1.40529424e-01 3.93052369e-01 2.56545126e-01 2.92873859e-01 -8.41234744e-01 -5.37423909e-01 7.65495956e-01 1.25934249e-02 3.11095640e-02 1.29090834e+00 -1.90963402e-01 -1.83952108e-01 1.78320631e-01 1.29309487e+00 -2.80545801e-01 -1.15790176e+00 -2.01849610e-01 4.11393829e-02 -1.72212571e-01 3.96921664e-01 -9.90596533e-01 -1.62423146e+00 1.23122323e+00 1.21527696e+00 -2.02767760e-01 1.34494627e+00 -2.36844406e-01 1.06338573e+00 -2.92365789e-01 -4.48315777e-02 -9.79386926e-01 -1.56341434e-01 2.63446957e-01 7.00887144e-01 -1.15580845e+00 4.27435748e-02 -2.27683783e-01 -9.88894343e-01 1.34595966e+00 3.98545265e-01 -1.49892271e-01 8.16809297e-01 3.31047893e-01 2.91357398e-01 -2.62826562e-01 -4.01799232e-01 1.34001374e-01 3.77667129e-01 3.34244907e-01 7.89380729e-01 1.58067152e-01 -4.28620785e-01 5.99132001e-01 3.77283782e-01 1.36557847e-01 2.25433797e-01 8.87137711e-01 -2.00665176e-01 -8.19194853e-01 -3.96208674e-01 4.79783773e-01 -9.62045014e-01 -2.21891329e-01 2.97958404e-01 6.88634634e-01 4.98133183e-01 6.70163214e-01 -1.08662874e-01 -1.91232920e-01 1.17581740e-01 5.35930693e-02 3.09585929e-01 -4.07900482e-01 -1.03561676e+00 3.86167586e-01 -1.45054519e-01 -4.27196592e-01 -6.36496723e-01 -7.57954061e-01 -1.78578949e+00 1.87842026e-01 -1.81759030e-01 -5.21358214e-02 6.18415177e-01 9.35591578e-01 3.88314158e-01 9.43565130e-01 6.95309162e-01 -7.98704922e-01 -3.31791818e-01 -1.02535546e+00 -4.92903382e-01 5.53030431e-01 3.53581876e-01 -4.23024237e-01 -6.84850514e-02 2.75497586e-01]
[14.287158966064453, -2.4278767108917236]
9f50088f-88d8-4842-b5ab-be085aa92bf8
afd-net-aggregated-feature-difference
null
null
http://openaccess.thecvf.com/content_ICCV_2019/html/Quan_AFD-Net_Aggregated_Feature_Difference_Learning_for_Cross-Spectral_Image_Patch_Matching_ICCV_2019_paper.html
http://openaccess.thecvf.com/content_ICCV_2019/papers/Quan_AFD-Net_Aggregated_Feature_Difference_Learning_for_Cross-Spectral_Image_Patch_Matching_ICCV_2019_paper.pdf
AFD-Net: Aggregated Feature Difference Learning for Cross-Spectral Image Patch Matching
Image patch matching across different spectral domains is more challenging than in a single spectral domain. We consider the reason is twofold: 1. the weaker discriminative feature learned by conventional methods; 2. the significant appearance difference between two images domains. To tackle these problems, we propose an aggregated feature difference learning network (AFD-Net). Unlike other methods that merely rely on the high-level features, we find the feature differences in other levels also provide useful learning information. Thus, the multi-level feature differences are aggregated to enhance the discrimination. To make features invariant across different domains, we introduce a domain invariant feature extraction network based on instance normalization (IN). In order to optimize the AFD-Net, we borrow the large margin cosine loss which can minimize intra-class distance and maximize inter-class distance between matching and non-matching samples. Extensive experiments show that AFD-Net largely outperforms the state-of-the-arts on the cross-spectral dataset, meanwhile, demonstrates a considerable generalizability on a single spectral dataset.
[' Licheng Jiao', ' Ning Huyan', ' Yanfeng Li', ' Shaowei Wei', ' Shuang Wang', ' Xuefeng Liang', 'Dou Quan']
2019-10-01
null
null
null
iccv-2019-10
['patch-matching']
['computer-vision']
[ 5.44482589e-01 -4.46729034e-01 -3.33557308e-01 -5.54264307e-01 -8.75942707e-01 -5.13042510e-01 4.89089072e-01 -1.36094481e-01 -2.57714272e-01 4.57068205e-01 4.51013260e-02 2.36770824e-01 -3.64479899e-01 -7.37861633e-01 -5.25529027e-01 -8.19914877e-01 8.57376829e-02 -2.42625639e-01 3.74253124e-01 -8.99620950e-02 9.49515104e-02 5.03930926e-01 -1.67817903e+00 1.98609129e-01 1.05312192e+00 1.48802853e+00 2.91472077e-01 -2.21288905e-01 -2.97277775e-02 2.44841918e-01 -1.85568824e-01 -1.01364991e-02 7.93113589e-01 -3.71441275e-01 -4.78592515e-01 3.01102996e-01 8.61767352e-01 -2.56434679e-01 -4.45897788e-01 1.49477375e+00 6.34695888e-01 1.72292322e-01 7.46090651e-01 -1.36961472e+00 -9.02303219e-01 -3.62312086e-02 -8.46166074e-01 7.79948011e-03 1.92123324e-01 2.05709696e-01 1.13241529e+00 -9.20778573e-01 3.54535818e-01 1.12481010e+00 8.32869530e-01 -9.06632096e-03 -1.39160740e+00 -8.10511112e-01 2.59957433e-01 2.74584502e-01 -1.59680295e+00 -2.80029893e-01 9.89055812e-01 -5.77834129e-01 4.20289397e-01 7.55482689e-02 4.44786608e-01 7.46178567e-01 -1.15505964e-01 7.06609547e-01 1.41479635e+00 -3.38096589e-01 -1.77822843e-01 -1.77747663e-02 -9.24672782e-02 5.40246844e-01 8.83933455e-02 2.70128518e-01 -6.07559502e-01 -4.73542251e-02 6.84726954e-01 2.46621057e-01 -4.26410794e-01 -6.11985803e-01 -1.30171418e+00 7.45358884e-01 5.89069903e-01 2.39771545e-01 -1.90548584e-01 -3.35009098e-01 3.98646265e-01 5.43301344e-01 5.33003747e-01 1.96654275e-01 -5.18558323e-01 3.99286687e-01 -7.81226337e-01 2.12597474e-01 1.33111641e-01 7.80520916e-01 1.13282752e+00 -7.48627335e-02 -3.07001382e-01 1.25412047e+00 1.29838333e-01 5.23715734e-01 3.91337812e-01 -6.68845594e-01 3.66582811e-01 8.16576421e-01 -1.17588036e-01 -1.31871188e+00 -3.23989064e-01 -5.84647536e-01 -1.07050633e+00 2.08089232e-01 4.08546805e-01 9.45043489e-02 -8.18382978e-01 1.99514627e+00 3.61218542e-01 1.53416857e-01 -2.36501023e-01 1.16161740e+00 8.29317391e-01 3.94720048e-01 -1.84006736e-01 -1.55039892e-01 1.15372419e+00 -8.73594284e-01 -4.39658582e-01 -3.84107381e-01 3.17533016e-01 -9.35600519e-01 1.16562974e+00 1.04481474e-01 -6.27832055e-01 -9.09353256e-01 -1.27785110e+00 2.54305620e-02 -4.74251390e-01 3.63197327e-01 4.58202034e-01 3.08130205e-01 -6.23645902e-01 6.60147309e-01 -2.17205226e-01 -3.41445565e-01 3.46498877e-01 2.35967711e-01 -5.45286477e-01 -2.67348945e-01 -1.25830162e+00 5.54913461e-01 3.28290224e-01 -1.77304864e-01 -4.73170191e-01 -9.26566005e-01 -9.06953335e-01 -1.33123741e-01 4.71696436e-01 -4.22092468e-01 8.67243171e-01 -1.30698681e+00 -1.30211020e+00 1.05903721e+00 5.94383292e-02 -8.26034248e-02 3.55458289e-01 1.07295237e-01 -8.12738538e-01 6.13168627e-02 4.43778992e-01 5.08149922e-01 1.04512572e+00 -1.08329260e+00 -7.93381155e-01 -4.84440833e-01 -6.26894310e-02 2.66653180e-01 -4.20585036e-01 -2.36774608e-01 -4.26816165e-01 -9.75963354e-01 4.59897786e-01 -9.44866657e-01 1.58155575e-01 5.53385079e-01 -3.01959991e-01 -1.68480113e-01 9.15687144e-01 -5.60766876e-01 1.11011016e+00 -2.45775151e+00 -3.83744091e-02 2.31973261e-01 1.91441383e-02 3.09037089e-01 -4.62617338e-01 1.74833789e-01 -3.79256040e-01 -3.09742689e-01 -4.83492464e-01 2.68071741e-01 8.52011740e-02 -9.40236300e-02 -1.41223878e-01 6.27912104e-01 3.63494396e-01 5.14962792e-01 -6.86983943e-01 -4.17715549e-01 1.80922270e-01 1.75726607e-01 -2.96872646e-01 1.41383022e-01 1.61179721e-01 2.13788092e-01 -4.78020817e-01 7.52664030e-01 1.26512504e+00 -2.06277713e-01 9.56588145e-03 -8.77343237e-01 -1.61509737e-01 1.09794987e-02 -1.47687471e+00 1.98774171e+00 -2.92702973e-01 4.46460426e-01 1.48172647e-01 -1.35460091e+00 1.01575089e+00 -8.55647251e-02 6.93243682e-01 -1.05002308e+00 -1.58528030e-01 4.83703494e-01 6.29844666e-02 -3.69459540e-01 1.38232693e-01 -2.57121354e-01 -7.40993489e-03 1.25641257e-01 1.77350909e-01 4.04461324e-02 -5.00521697e-02 -2.09701031e-01 7.14962244e-01 4.00107820e-03 4.17752802e-01 -5.08339524e-01 8.40077281e-01 -3.29035729e-01 9.38849032e-01 6.76721931e-01 -4.74433780e-01 7.11461484e-01 2.25200832e-01 -4.24157172e-01 -7.86665142e-01 -1.11928475e+00 -4.76266772e-01 9.62587059e-01 6.96700633e-01 -2.13288158e-01 -3.80840987e-01 -8.57907832e-01 4.80418652e-01 2.66160648e-02 -5.52426755e-01 -2.89961815e-01 -1.51131853e-01 -7.59880066e-01 2.86930293e-01 4.69918460e-01 1.00202250e+00 -5.70903957e-01 1.47285294e-02 4.55859900e-02 -8.24345797e-02 -1.04307938e+00 -7.80086458e-01 9.13543850e-02 -5.60587585e-01 -1.10061061e+00 -7.73632526e-01 -9.52077091e-01 3.89544010e-01 8.07834208e-01 1.07640123e+00 -2.01589331e-01 -4.63660151e-01 1.63140029e-01 -4.07421440e-01 -1.10232361e-01 2.61413187e-01 1.01893328e-01 1.44421384e-01 2.63075620e-01 6.38637364e-01 -7.29056358e-01 -8.19020689e-01 6.44737542e-01 -9.11227703e-01 -9.38366503e-02 7.28626192e-01 1.11184084e+00 8.96883965e-01 2.03363433e-01 6.73738480e-01 -5.16766548e-01 3.90526533e-01 -4.14823174e-01 -4.97594118e-01 3.46224248e-01 -7.82438278e-01 -2.75960919e-02 6.28128827e-01 -4.91383433e-01 -8.61212909e-01 2.32721224e-01 1.77869827e-01 -2.20377028e-01 -1.32252187e-01 4.21004444e-01 -4.07645583e-01 -5.84164560e-01 6.16312087e-01 3.28949630e-01 6.49836734e-02 -5.16373217e-01 1.66799456e-01 7.30933070e-01 5.23869395e-01 -5.88436663e-01 1.21071124e+00 4.70898151e-01 1.30222827e-01 -5.91176927e-01 -1.05699039e+00 -6.30874515e-01 -6.98348880e-01 1.42330462e-02 7.34388053e-01 -1.18546307e+00 -3.94733101e-01 6.50168180e-01 -7.57450938e-01 -4.10691053e-02 -2.51257807e-01 4.85805452e-01 -3.67504567e-01 5.93056738e-01 -2.83874661e-01 -4.02780652e-01 -5.45393452e-02 -1.05470049e+00 1.15328395e+00 4.03267771e-01 3.36483032e-01 -7.35924304e-01 3.36911380e-02 1.76884457e-02 3.49965423e-01 2.45397046e-01 9.04814601e-01 -4.75087911e-01 -4.13190961e-01 -9.15209576e-02 -8.58181119e-01 6.48131669e-01 6.19630635e-01 -1.82379991e-01 -9.50612724e-01 -5.19343257e-01 -1.46205381e-01 -3.75616521e-01 9.60094333e-01 2.78322995e-01 1.46691871e+00 -4.43076082e-02 -1.69544294e-01 9.51370597e-01 1.66389978e+00 1.07578440e-02 5.20334661e-01 4.54342932e-01 5.69560826e-01 6.03776693e-01 8.31206977e-01 3.29391092e-01 1.62719280e-01 9.07178044e-01 3.21663946e-01 -4.49981004e-01 -1.96882516e-01 -2.08896205e-01 3.19670409e-01 5.68243623e-01 1.45041004e-01 2.44404852e-01 -5.03680408e-01 5.41075110e-01 -1.90914106e+00 -9.87374961e-01 3.10239345e-02 2.23615718e+00 7.08140910e-01 -1.01819023e-01 2.57269323e-01 -4.83120605e-02 9.50344980e-01 5.31299591e-01 -8.61537695e-01 2.37004668e-01 -5.71313381e-01 1.04544230e-01 6.61976159e-01 7.61511922e-02 -1.40418386e+00 5.81326663e-01 6.00303078e+00 1.35748112e+00 -1.24249816e+00 1.56494156e-02 3.99625659e-01 1.07492030e-01 -1.49969652e-01 -5.56327105e-02 -4.43947822e-01 6.18539691e-01 1.36582106e-01 -1.38610184e-01 5.33355176e-01 8.12765241e-01 -1.43403366e-01 1.19000077e-01 -8.56959999e-01 1.28112268e+00 5.58122359e-02 -1.01618016e+00 7.94923156e-02 2.44614854e-02 8.04709315e-01 9.24581438e-02 1.86104745e-01 3.30346584e-01 -1.52109817e-01 -7.62764037e-01 6.57294333e-01 3.63005728e-01 1.02313781e+00 -7.67859042e-01 5.57585955e-01 1.19337745e-01 -1.70862806e+00 -5.88111803e-02 -7.26057887e-01 8.85138214e-02 -1.33714423e-01 8.86513174e-01 -5.82158305e-02 1.01847351e+00 8.33548963e-01 1.09689927e+00 -7.38730073e-01 1.10020185e+00 9.62482989e-02 1.96153477e-01 -2.00095549e-01 5.71553767e-01 2.41612628e-01 -5.82896650e-01 4.77423519e-01 1.12685418e+00 4.69915986e-01 -1.98201329e-01 5.97636223e-01 8.17454338e-01 -6.51690215e-02 2.52790898e-01 -5.19831419e-01 1.19751975e-01 4.55505729e-01 1.48334670e+00 -2.83703327e-01 -3.39407027e-02 -8.29655111e-01 1.19167185e+00 1.06215514e-01 3.12989503e-01 -7.45971441e-01 -6.58649385e-01 1.03121197e+00 1.49527565e-01 3.62660795e-01 -1.45810014e-02 -5.50352596e-02 -1.29152799e+00 3.81418496e-01 -1.01889789e+00 4.61636364e-01 -6.06682241e-01 -2.05819297e+00 3.76567870e-01 -2.43163005e-01 -1.82045841e+00 2.77440965e-01 -8.47719729e-01 -5.03666222e-01 1.05034280e+00 -2.03501225e+00 -1.39023817e+00 -5.66682458e-01 8.44991863e-01 3.21905732e-01 -3.39528114e-01 5.05213201e-01 6.84592664e-01 -5.11738241e-01 7.86365271e-01 3.10633302e-01 2.27494836e-01 1.28711092e+00 -1.03239202e+00 6.50283471e-02 7.97191381e-01 -1.72457740e-01 4.17767256e-01 2.53180653e-01 -4.35196280e-01 -1.04905152e+00 -1.13413262e+00 4.23071802e-01 1.36827856e-01 6.88635886e-01 -1.78847298e-01 -1.11959553e+00 2.31677249e-01 -1.50062218e-02 4.79069442e-01 8.17723155e-01 3.80329527e-02 -9.22524929e-01 -6.34456694e-01 -1.20727277e+00 2.40617335e-01 1.37630951e+00 -1.01772869e+00 -3.02337378e-01 2.45667160e-01 6.12028480e-01 -2.99221370e-02 -1.02258956e+00 8.15132022e-01 6.22326732e-01 -1.13707554e+00 1.11251986e+00 -4.54347163e-01 2.57197559e-01 -7.59216905e-01 -6.10970259e-01 -1.35923243e+00 -7.49025643e-01 -3.22644085e-01 2.91352838e-01 1.45609260e+00 2.30863437e-01 -7.88207710e-01 5.17701983e-01 1.83646217e-01 -7.80370086e-02 -5.10898888e-01 -8.39376807e-01 -1.23952389e+00 6.09436221e-02 -1.27666831e-01 8.11329365e-01 1.23700345e+00 -3.22862774e-01 2.14449614e-01 -3.91387612e-01 2.96413809e-01 7.80882418e-01 4.97264236e-01 6.10472322e-01 -1.53447747e+00 -3.08349818e-01 -4.12291348e-01 -6.21779323e-01 -1.14502072e+00 3.02489460e-01 -1.02177286e+00 2.56451853e-02 -1.05284429e+00 5.33444524e-01 -5.47006905e-01 -5.91511965e-01 4.53861088e-01 -2.81156600e-01 4.24715310e-01 1.64007500e-01 4.35355932e-01 -5.05589366e-01 7.40438998e-01 1.37652922e+00 -4.51173156e-01 2.03859098e-02 -2.69545287e-01 -8.62427652e-01 6.57882810e-01 6.23821020e-01 -3.70745987e-01 -3.18048120e-01 -1.92342892e-01 -6.31519854e-02 -3.68387282e-01 4.67934340e-01 -1.19346356e+00 1.40829429e-01 -3.92844290e-01 5.25402784e-01 -5.13063192e-01 7.39471167e-02 -8.92680705e-01 -9.60989017e-03 1.78417072e-01 -1.91347778e-01 -2.93889314e-01 2.60774288e-02 6.57686412e-01 -6.33164644e-01 -6.27395604e-03 1.02463508e+00 6.63160756e-02 -1.05405033e+00 6.22800708e-01 5.82017079e-02 1.54322401e-01 8.26485932e-01 -3.56600434e-01 -2.80711681e-01 -2.16715261e-01 -3.52897197e-01 4.61053587e-02 5.83509147e-01 4.51765507e-01 2.47875705e-01 -1.88284028e+00 -6.38837993e-01 3.06762666e-01 6.94813609e-01 -1.55146196e-01 4.77522016e-01 9.03007567e-01 5.04425950e-02 1.89030215e-01 -5.01275897e-01 -6.45372808e-01 -1.08124387e+00 5.09762883e-01 6.00465178e-01 -1.39059946e-01 -2.60104418e-01 6.43025577e-01 6.97309315e-01 -7.38802433e-01 -5.80431893e-02 4.66267318e-02 1.15182661e-01 2.76884973e-01 4.82835084e-01 2.52823830e-01 1.90514341e-01 -7.15089798e-01 -4.82532024e-01 1.00688720e+00 -4.18959968e-02 2.81669319e-01 1.33133745e+00 -2.13855937e-01 -7.45714903e-02 1.52687520e-01 1.53827953e+00 -8.40471908e-02 -1.59941900e+00 -8.87189031e-01 -1.84634924e-01 -8.92416239e-01 1.12465389e-01 -7.38385677e-01 -1.23204815e+00 7.23652005e-01 1.10371614e+00 2.75134891e-01 1.56957650e+00 -1.87685207e-01 7.26445735e-01 1.80310428e-01 1.97967634e-01 -1.42597246e+00 2.11159185e-01 4.32738036e-01 8.42430651e-01 -1.55194879e+00 2.46164843e-01 -6.48401737e-01 -5.22066474e-01 1.15583277e+00 7.09815383e-01 -2.09296048e-01 7.31926978e-01 -1.75380632e-01 4.04033661e-02 -9.11927894e-02 -1.15848579e-01 -5.78641713e-01 7.91329741e-01 8.42284083e-01 4.83461767e-01 2.79186517e-02 -2.93010890e-01 6.30393982e-01 2.98454314e-01 -7.85596669e-02 -1.14613399e-01 6.28953278e-01 -4.78783190e-01 -1.29174054e+00 -2.63148397e-01 5.16872168e-01 -1.97727531e-01 -6.64335536e-03 -5.53527117e-01 7.74388433e-01 5.63167095e-01 8.24902713e-01 1.53015405e-02 -6.61123931e-01 5.73235452e-01 -2.21859083e-01 4.31472421e-01 -3.99139404e-01 -3.93141329e-01 -3.43443565e-02 -1.99615255e-01 -6.44686460e-01 -7.56494880e-01 -6.76283598e-01 -7.33333528e-01 -1.37724116e-01 -4.07868534e-01 -2.93786526e-01 2.08958566e-01 7.58496284e-01 4.90543187e-01 2.19726652e-01 1.01325774e+00 -7.48922288e-01 -7.57048786e-01 -7.74281681e-01 -1.04714310e+00 8.12735736e-01 4.39666301e-01 -9.60429430e-01 -5.40998697e-01 -1.25046492e-01]
[9.78939437866211, 2.0874929428100586]
ebeacdb8-2eed-4c58-87a1-1e21d5534ad9
improving-question-answering-performance-1
2212.08897
null
https://arxiv.org/abs/2212.08897v1
https://arxiv.org/pdf/2212.08897v1.pdf
Improving Question Answering Performance through Manual Annotation: Costs, Benefits and Strategies
Recently proposed systems for open-domain question answering (OpenQA) require large amounts of training data to achieve state-of-the-art performance. However, data annotation is known to be time-consuming and therefore expensive to acquire. As a result, the appropriate datasets are available only for a handful of languages (mainly English and Chinese). In this work, we introduce and publicly release PolQA, the first Polish dataset for OpenQA. It consists of 7,000 questions, 87,525 manually labeled evidence passages, and a corpus of over 7,097,322 candidate passages. Each question is classified according to its formulation, type, as well as entity type of the answer. This resource allows us to evaluate the impact of different annotation choices on the performance of the QA system and propose an efficient annotation strategy that increases the passage retrieval performance by 10.55 p.p. while reducing the annotation cost by 82%.
['Maciej Ogrodniczuk', 'Piotr Przybyła', 'Piotr Rybak']
2022-12-17
null
null
null
null
['passage-retrieval', 'open-domain-question-answering']
['natural-language-processing', 'natural-language-processing']
[-3.79178017e-01 1.00999914e-01 1.59721673e-01 -1.55083537e-01 -1.65458369e+00 -1.03025997e+00 4.75577414e-01 6.55463517e-01 -7.02630937e-01 1.17366898e+00 1.60929292e-01 -3.53115201e-01 -2.56785691e-01 -8.40231478e-01 -6.03190839e-01 -3.04530084e-01 4.07496452e-01 1.08780551e+00 7.41939127e-01 -4.53576297e-01 2.43953049e-01 -2.29500979e-02 -1.56202424e+00 4.71392095e-01 1.56130135e+00 1.06469810e+00 3.14837068e-01 6.83900058e-01 -4.86402094e-01 6.98025167e-01 -8.93880367e-01 -9.05599356e-01 -1.94235474e-01 -4.32274908e-01 -1.47341776e+00 -3.38800728e-01 6.27509594e-01 -9.70175862e-02 -1.24870121e-01 8.02991688e-01 6.74341619e-01 2.33479187e-01 5.81119895e-01 -6.01539195e-01 -6.86347485e-01 4.55860347e-01 2.01206967e-01 3.31898957e-01 6.17414713e-01 -4.31806259e-02 1.58081186e+00 -7.55652308e-01 9.11840498e-01 5.46191752e-01 2.68209785e-01 4.91202623e-01 -7.55505383e-01 -5.77905551e-02 -3.92900050e-01 6.00330710e-01 -1.25851035e+00 -3.05844486e-01 3.52687836e-01 -1.67575940e-01 6.88157201e-01 4.35692042e-01 2.66170233e-01 7.97224402e-01 -2.92573422e-01 5.88827491e-01 9.00655806e-01 -6.22233152e-01 2.78952092e-01 2.89385170e-01 5.49331963e-01 5.84385812e-01 3.81841809e-01 -8.45946610e-01 -2.31942102e-01 -4.26900804e-01 3.37216593e-02 -6.90958500e-01 -4.22082067e-01 -4.33602184e-02 -9.49479103e-01 8.23688686e-01 1.86357141e-01 5.39922595e-01 -4.00001556e-01 -4.68221962e-01 4.27279979e-01 4.76442724e-01 3.82186085e-01 9.38283205e-01 -8.53642702e-01 -4.72386122e-01 -5.91004193e-01 5.20165265e-01 1.14718175e+00 6.31455958e-01 7.27407038e-01 -7.03149915e-01 -5.39220929e-01 1.17500687e+00 1.39248550e-01 6.06373906e-01 3.82098883e-01 -1.03240931e+00 8.93073618e-01 8.79140973e-01 5.36095440e-01 -7.53314435e-01 -3.69434923e-01 -2.57657230e-01 -1.39690861e-01 -7.63612449e-01 9.82902467e-01 -2.88845181e-01 -3.78108263e-01 1.33817756e+00 3.50390941e-01 -5.00720799e-01 2.60859668e-01 8.80353272e-01 1.30016947e+00 8.05367231e-01 -1.77598950e-02 4.07798402e-03 1.87884581e+00 -1.27712858e+00 -8.72950017e-01 7.27559179e-02 8.05824816e-01 -1.09613085e+00 1.46192455e+00 2.06904903e-01 -1.16733003e+00 -2.89433539e-01 -6.62135005e-01 -4.62707490e-01 -4.26448613e-01 3.36413056e-01 2.13297009e-01 5.71752846e-01 -6.89282715e-01 1.06142864e-01 -2.60837436e-01 -2.40324527e-01 1.30541176e-01 -6.47477340e-03 -1.59328520e-01 -4.11669403e-01 -1.48126543e+00 9.45287228e-01 2.15105355e-01 -2.45703727e-01 -5.18077016e-01 -6.45737886e-01 -5.21844327e-01 3.42017531e-01 5.80783248e-01 -6.45709932e-01 1.54489100e+00 -4.73272026e-01 -1.42559171e+00 8.12338710e-01 -1.75481051e-01 -5.85082591e-01 3.31300139e-01 -3.98445785e-01 -6.11881912e-01 7.91240096e-01 3.29375744e-01 2.76846647e-01 2.94153750e-01 -8.22867274e-01 -6.03871286e-01 -2.95049995e-01 4.58306879e-01 1.90162674e-01 -4.34719115e-01 2.10594594e-01 -6.84017539e-01 -2.37019628e-01 -2.01994330e-01 -7.25440979e-01 -3.23773846e-02 -4.56219643e-01 -1.46132722e-01 -6.98965311e-01 3.70493263e-01 -1.19641960e+00 1.44430184e+00 -1.86106563e+00 -1.78136840e-01 -1.18545972e-01 -6.64507747e-02 5.22448957e-01 -2.14781329e-01 7.42094040e-01 6.54654026e-01 -4.09794040e-02 -3.17448914e-01 1.97814424e-02 1.41778991e-01 1.98806807e-01 -3.52563560e-01 -9.19177383e-03 4.16252427e-02 8.76117885e-01 -9.19140518e-01 -8.29963744e-01 -1.94392309e-01 1.75568253e-01 -6.35074317e-01 4.17857170e-01 -7.21877873e-01 4.83702362e-01 -8.08880746e-01 6.71778917e-01 3.52821678e-01 -3.99801314e-01 -7.62649402e-02 1.47280291e-01 -5.44644659e-03 8.74696434e-01 -8.33629549e-01 1.71897662e+00 -5.42600870e-01 5.07837892e-01 -1.44042581e-01 -5.07659018e-01 9.63923633e-01 6.37986124e-01 2.75805295e-01 -7.95501471e-01 1.32307112e-01 6.57191157e-01 1.94741040e-02 -9.16290939e-01 8.34512353e-01 3.29818815e-01 -2.90778995e-01 4.88596916e-01 2.86206156e-01 -1.99347243e-01 8.11637938e-01 1.54191539e-01 1.27293873e+00 -2.63806671e-01 8.15217271e-02 -3.02723706e-01 1.02317035e+00 4.50332046e-01 2.31079638e-01 7.35162377e-01 -1.05408877e-01 5.29136181e-01 4.24210906e-01 -2.82204628e-01 -8.88585150e-01 -7.88094461e-01 -3.29592258e-01 9.98130560e-01 -1.18758202e-01 -3.73043329e-01 -1.11440516e+00 -7.23443925e-01 -3.24041277e-01 7.32729375e-01 -2.17551902e-01 9.72002298e-02 -6.93757474e-01 -3.42912346e-01 9.37709749e-01 -7.48074725e-02 6.59811735e-01 -1.18440747e+00 -4.33992535e-01 2.41001114e-01 -1.04246724e+00 -1.27499151e+00 -2.81773776e-01 -2.63760090e-01 -7.43566394e-01 -1.39495122e+00 -8.27135444e-01 -8.61117125e-01 4.35465515e-01 -1.00093141e-01 1.52659714e+00 1.22684248e-01 7.04041198e-02 4.15225744e-01 -8.60382974e-01 -2.49976754e-01 -2.93311805e-01 5.96061051e-01 -4.87075269e-01 -1.28966436e-01 3.88965279e-01 1.00748286e-01 -6.03680313e-01 4.09356296e-01 -8.03126514e-01 -6.39713109e-01 4.57924217e-01 8.39432716e-01 5.72595954e-01 -2.15303227e-01 8.76704276e-01 -8.20285618e-01 1.05057192e+00 -5.60244322e-01 -6.83032870e-01 7.67146051e-01 -3.10455590e-01 2.16090921e-02 7.65286267e-01 -1.09443203e-01 -1.39098966e+00 -4.57316190e-01 -5.95228851e-01 2.23191366e-01 -2.25496158e-01 7.05825806e-01 -1.14451181e-02 -6.58652708e-02 7.74517953e-01 -2.26943358e-03 -3.87674958e-01 -8.04692209e-01 4.58763659e-01 9.17081237e-01 3.81650418e-01 -7.55568504e-01 4.50103045e-01 1.36484161e-01 -5.22555768e-01 -8.26029718e-01 -1.26524353e+00 -7.50253677e-01 -5.03152847e-01 -6.49518296e-02 8.90543580e-01 -6.97454214e-01 -4.70503390e-01 1.80666849e-01 -1.15169311e+00 -1.16814204e-01 -3.86288464e-01 4.14812922e-01 -2.50233859e-01 4.54626054e-01 -6.76681101e-01 -5.90625703e-01 -6.90232456e-01 -8.59215021e-01 7.20932662e-01 2.21023723e-01 -1.92216754e-01 -8.00986052e-01 3.38192582e-01 1.23588753e+00 3.47343653e-01 -3.34382862e-01 9.67709184e-01 -1.01424444e+00 -5.17258883e-01 -4.41644520e-01 -9.94508117e-02 3.09732378e-01 -2.39791974e-01 -2.37494692e-01 -8.03327560e-01 1.29940748e-01 9.10840854e-02 -6.62138879e-01 6.83320463e-01 -9.35289450e-03 9.56723928e-01 -2.73549616e-01 2.00602144e-01 -2.59389728e-01 1.20386028e+00 4.62618610e-03 6.69195890e-01 4.94643241e-01 2.18005255e-01 8.28587234e-01 1.05691159e+00 2.97592103e-01 5.93588591e-01 5.22612751e-01 2.65155643e-01 4.67777938e-01 -1.78121263e-04 -2.46841878e-01 2.80528404e-02 1.20587707e+00 -5.52116632e-02 -4.07983452e-01 -1.01339912e+00 8.94574523e-01 -1.44868743e+00 -7.52567828e-01 -4.04367298e-01 2.10571814e+00 1.03197372e+00 -4.40363064e-02 -7.24376962e-02 1.96210980e-01 4.64989394e-01 5.67436628e-02 -9.77990031e-02 -1.46210626e-01 -2.62848139e-01 4.52408791e-01 6.02592155e-02 4.64853913e-01 -7.76840925e-01 7.93798685e-01 5.73184347e+00 9.63709474e-01 -3.90459657e-01 2.61697114e-01 3.86515260e-01 2.58675903e-01 -3.46124917e-01 -1.90030523e-02 -8.14336181e-01 4.54097748e-01 1.20636189e+00 -3.69477607e-02 1.94706485e-01 6.87221706e-01 -1.52375355e-01 -2.73133934e-01 -6.09797478e-01 5.69838226e-01 2.64491946e-01 -1.34942043e+00 1.60870329e-01 -3.17360014e-01 7.57395625e-01 9.44726840e-02 -2.93078542e-01 5.58202446e-01 2.86053866e-02 -5.91286123e-01 3.73865098e-01 4.06171143e-01 3.52356106e-01 -5.67578137e-01 1.20946062e+00 5.80072343e-01 -7.57757545e-01 -1.63210601e-01 -5.43027997e-01 1.85938515e-02 2.92306364e-01 4.80768621e-01 -5.50946832e-01 7.17840016e-01 8.82048905e-01 -6.83453232e-02 -7.14760244e-01 1.19832373e+00 -5.67687750e-01 9.46880460e-01 -2.86382824e-01 -5.06583035e-01 4.23913568e-01 -2.50327617e-01 3.59502465e-01 7.94557214e-01 2.86443889e-01 3.93689454e-01 -1.77860454e-01 3.88603210e-01 -4.00953114e-01 6.74018621e-01 -1.29574209e-01 -1.36829495e-01 6.56851947e-01 1.10860932e+00 -4.42732990e-01 -4.63834167e-01 -5.25935233e-01 6.53559864e-01 5.80787718e-01 1.95806444e-01 -7.16076732e-01 -6.57051861e-01 2.06400216e-01 -2.30989307e-02 4.58702892e-01 -4.87219393e-02 3.30908783e-02 -1.27172887e+00 3.85884911e-01 -1.02639711e+00 9.74564016e-01 -6.56682611e-01 -1.33395231e+00 7.67347693e-01 -3.00346911e-01 -1.03113222e+00 -1.36405274e-01 -3.72356802e-01 -1.17041476e-01 7.87020087e-01 -1.75512910e+00 -6.79271102e-01 -1.98619112e-01 5.66008389e-01 6.48753703e-01 -2.56812596e-03 1.05320799e+00 8.05288434e-01 -3.49023998e-01 4.90192175e-01 4.54420805e-01 2.66538441e-01 9.14385676e-01 -1.13151133e+00 -9.10559371e-02 3.63140225e-01 4.08206969e-01 4.71212536e-01 5.21235406e-01 -2.02135876e-01 -1.24168289e+00 -6.63755894e-01 1.61947846e+00 -6.98804975e-01 8.90451550e-01 1.01666460e-02 -1.15875065e+00 2.75361359e-01 3.82146239e-01 -1.00641802e-01 7.99951315e-01 1.81737274e-01 -1.90783456e-01 -1.18267722e-01 -1.23557293e+00 4.05130714e-01 4.67484295e-01 -6.51557028e-01 -1.16279817e+00 4.50214356e-01 8.58655035e-01 -5.10020852e-01 -1.21140301e+00 1.68518886e-01 1.15292288e-01 -6.75483525e-01 9.03035045e-01 -6.36748195e-01 3.43257427e-01 -1.85086727e-01 -1.58926204e-01 -9.81806278e-01 7.54657611e-02 -1.64740175e-01 -2.47327700e-01 1.30690026e+00 9.07607257e-01 -7.66691685e-01 4.82781917e-01 6.18332744e-01 -2.74976701e-01 -8.75323951e-01 -1.18122137e+00 -5.59757411e-01 1.57105967e-01 -2.68458247e-01 6.07357919e-01 8.20742786e-01 2.99364738e-02 6.06544077e-01 2.13019475e-02 1.45416036e-01 4.40983288e-02 4.31243658e-01 5.31525195e-01 -1.14593291e+00 -1.61197767e-01 -2.00758427e-01 1.40626639e-01 -1.13102162e+00 -5.23040257e-03 -6.28753960e-01 -8.45506266e-02 -1.78329527e+00 -1.35062292e-01 -6.00149155e-01 -7.41476119e-02 1.20571099e-01 -3.36068392e-01 2.20494166e-01 -7.89462551e-02 3.12965393e-01 -1.07302046e+00 5.92183590e-01 1.29189456e+00 -5.45956716e-02 6.13488033e-02 6.01359680e-02 -4.91111368e-01 5.70518196e-01 9.80224192e-01 -2.73873419e-01 -1.97919920e-01 -7.80991256e-01 4.20292109e-01 1.38414159e-01 -3.77769396e-02 -8.61776412e-01 1.50568143e-01 1.80925533e-01 -1.98203310e-01 -6.60033405e-01 3.82220149e-01 -7.37885952e-01 -3.97314340e-01 3.82581770e-01 -3.66128981e-01 1.93211008e-02 8.54144543e-02 3.98335755e-01 -7.38056540e-01 -9.36674893e-01 3.26063871e-01 -1.78529635e-01 -5.29080689e-01 7.67824352e-02 -3.02324206e-01 8.58201921e-01 7.50949562e-01 4.66120869e-01 -6.31587684e-01 -3.59633446e-01 -3.83437216e-01 4.35153753e-01 1.07014343e-01 3.18267018e-01 1.98969707e-01 -1.00718904e+00 -7.57856488e-01 -3.41096461e-01 3.30795646e-01 -3.76939960e-02 5.94134450e-01 5.28075635e-01 -8.06718826e-01 1.01792562e+00 5.54086491e-02 -2.21674591e-01 -1.19573927e+00 2.60618657e-01 -7.13218972e-02 -8.56395781e-01 -3.75597328e-01 7.15342641e-01 -6.21716976e-01 -6.91452801e-01 6.89128339e-02 -1.55864388e-01 -8.26356411e-01 2.25162268e-01 5.97964466e-01 6.15411341e-01 3.69223982e-01 -5.26196182e-01 -3.85005102e-02 4.68534291e-01 1.25321686e-01 -1.28576174e-01 9.72927511e-01 -1.85973331e-01 -3.19093645e-01 2.62416482e-01 9.35555875e-01 2.63046503e-01 -4.97145951e-01 -3.47394913e-01 2.61852086e-01 -4.04280603e-01 -3.55080545e-01 -1.02390385e+00 -4.96452868e-01 7.47515082e-01 2.44965702e-01 4.18910086e-01 9.46578979e-01 3.42238992e-01 1.29612744e+00 8.85000288e-01 4.74203974e-01 -1.19940174e+00 1.02372803e-01 9.12587881e-01 8.06935191e-01 -1.20325387e+00 -2.89287806e-01 -4.39462483e-01 -5.22663951e-01 8.45565379e-01 5.85825205e-01 1.18333675e-01 2.89823383e-01 -5.30581594e-01 3.35674375e-01 -4.41143394e-01 -6.41932130e-01 -2.66333997e-01 4.35576916e-01 1.93392500e-01 3.90073121e-01 -7.52809122e-02 -1.02315867e+00 7.78762639e-01 -3.64727020e-01 -2.40214616e-01 2.92563438e-01 7.82509923e-01 -5.86817205e-01 -1.16545439e+00 -4.09581631e-01 4.92521197e-01 -8.36854815e-01 -1.14047416e-01 -2.57601321e-01 6.09440267e-01 -1.02352113e-01 1.29705572e+00 -4.90713269e-02 3.86439353e-01 4.94999945e-01 2.76065201e-01 3.11481714e-01 -5.96845746e-01 -7.18297124e-01 -5.58504939e-01 7.56323993e-01 -2.94909603e-03 -5.01345813e-01 -5.85159719e-01 -1.11349499e+00 2.56762560e-02 -6.08941197e-01 9.87027109e-01 5.27345181e-01 9.34065819e-01 4.81009394e-01 1.64393157e-01 2.80474484e-01 1.62764683e-01 -7.59650111e-01 -1.16372049e+00 -2.70978451e-01 3.93790245e-01 -2.53765434e-02 -5.44522405e-01 -2.90615857e-01 -1.04818888e-01]
[11.317534446716309, 8.00059700012207]
e3e461ca-0a51-484e-a74e-6f534eb4b6a1
a-gaussian-process-regression-based-dynamical
2211.14162
null
https://arxiv.org/abs/2211.14162v1
https://arxiv.org/pdf/2211.14162v1.pdf
A Gaussian Process Regression based Dynamical Models Learning Algorithm for Target Tracking
Maneuvering target tracking is a challenging problem for sensor systems because of the unpredictability of the targets' motions. This paper proposes a novel data-driven method for learning the dynamical motion model of a target. Non-parametric Gaussian process regression (GPR) is used to learn a target's naturally shift invariant motion (NSIM) behavior, which is translationally invariant and does not need to be constantly updated as the target moves. The learned Gaussian processes (GPs) can be applied to track targets within different surveillance regions from the surveillance region of the training data by being incorporated into the particle filter (PF) implementation. The performance of our proposed approach is evaluated over different maneuvering scenarios by being compared with commonly used interacting multiple model (IMM)-PF methods and provides around $90\%$ performance improvement for a multi-target tracking (MTT) highly maneuvering scenario.
['James R. Hopgood', 'Ian K. Proudler', 'Mike E. Davies', 'Mengwei Sun']
2022-11-25
null
null
null
null
['gpr', 'gpr']
['computer-vision', 'miscellaneous']
[ 3.43529791e-01 -2.76062727e-01 2.14274839e-01 -9.70470235e-02 -6.19562566e-01 -5.23204744e-01 7.40836680e-01 1.63492739e-01 -3.13001275e-01 5.79607308e-01 -2.12617934e-01 -2.93362170e-01 -4.51057434e-01 -5.77423990e-01 -6.02451861e-01 -1.25204206e+00 -3.95469248e-01 5.64100802e-01 7.80937910e-01 -8.32056180e-02 3.28342104e-03 7.76807785e-01 -1.31137896e+00 -5.00398457e-01 5.78684330e-01 8.48706722e-01 4.53471750e-01 1.24104810e+00 3.35709721e-01 6.44307315e-01 -6.05068266e-01 2.08782107e-01 3.92822385e-01 -1.92193612e-02 2.27919593e-01 1.28944919e-01 2.57831782e-01 1.17142476e-01 -2.75936812e-01 9.65101957e-01 2.87071258e-01 4.47849482e-01 7.27606714e-01 -1.05369771e+00 1.60194010e-01 -2.82554448e-01 -4.88851041e-01 5.73974788e-01 -1.08339109e-01 3.50489110e-01 1.33124128e-01 -2.96517193e-01 1.56129792e-01 1.44074357e+00 9.01866257e-01 4.38300401e-01 -9.44713235e-01 -3.34026664e-01 3.33447874e-01 1.21066809e-01 -1.16753221e+00 -3.26515995e-02 4.85714614e-01 -7.04422832e-01 5.41613519e-01 4.08738136e-01 6.40491664e-01 9.60181832e-01 1.04242074e+00 2.47998118e-01 8.89701307e-01 1.88685536e-01 4.84547138e-01 -2.41938338e-01 2.53707439e-01 3.62074077e-01 6.80348396e-01 5.24243832e-01 -3.05945873e-02 -5.40438354e-01 5.10812223e-01 2.38573954e-01 -2.46415302e-01 -3.71380746e-01 -1.08901048e+00 6.76812232e-01 1.61697447e-01 -1.44088613e-02 -8.30338717e-01 2.45677948e-01 1.28533185e-01 1.18106581e-01 4.77686584e-01 -1.55342445e-01 -5.38845420e-01 -2.89889514e-01 -8.43840837e-01 4.91741568e-01 1.05267060e+00 1.00461268e+00 3.86305213e-01 5.60627460e-01 -5.04885733e-01 -2.75367238e-02 7.83520639e-01 1.58899462e+00 5.14353476e-02 -4.52571481e-01 1.97793290e-01 3.38433802e-01 1.01081705e+00 -1.20370984e+00 -5.31579494e-01 -5.81602633e-01 -7.77915537e-01 3.68922502e-01 1.67863995e-01 -7.28944004e-01 -9.58225191e-01 1.41287732e+00 8.47248375e-01 5.58828890e-01 2.98959076e-01 5.52985609e-01 2.72171140e-01 9.98667955e-01 3.62351567e-01 -4.60844100e-01 1.13433456e+00 -3.96127850e-01 -7.36630142e-01 -3.76913518e-01 2.14206547e-01 -6.80167079e-01 -3.96970958e-02 2.05088601e-01 -3.83569866e-01 -7.89089501e-01 -6.57274902e-01 1.19671583e+00 -3.39297615e-02 -8.89041796e-02 2.01628171e-03 6.40007913e-01 -9.30915475e-01 3.51553500e-01 -1.61399150e+00 -4.87640083e-01 4.48278226e-02 3.36669564e-01 4.22390290e-02 -2.58329604e-03 -7.13826716e-01 1.08003116e+00 3.50884229e-01 3.07543516e-01 -1.66441607e+00 -7.27486253e-01 -7.44064331e-01 -3.72327626e-01 2.52133071e-01 -6.44955754e-01 9.11289573e-01 -3.14895272e-01 -1.57250261e+00 -1.31525910e-02 -4.75476652e-01 -7.06804335e-01 4.07736003e-01 -4.92219001e-01 -6.88107729e-01 -1.00847483e-01 -3.15133333e-01 1.06782421e-01 1.37697375e+00 -1.10215712e+00 -1.03596509e+00 -3.70193839e-01 -6.47915125e-01 3.74722093e-01 3.57894331e-01 -1.17712624e-01 -5.90721406e-02 -5.53484678e-01 2.63557602e-02 -1.45693254e+00 -5.50492644e-01 -3.25947165e-01 -2.71337688e-01 3.08550671e-02 1.63662612e+00 -7.22775221e-01 8.50476146e-01 -2.12572145e+00 2.13766649e-01 1.03592709e-01 -2.52204746e-01 5.26317835e-01 1.47147998e-01 6.91155672e-01 4.02252108e-01 -7.27330387e-01 -2.54046600e-02 -3.27573657e-01 -2.84539819e-01 3.87616694e-01 -3.46309930e-01 8.93501043e-01 1.10533722e-02 6.21200860e-01 -7.63830304e-01 9.63187516e-02 4.12736088e-01 7.01814413e-01 2.02042565e-01 3.35915476e-01 -3.53255808e-01 1.05414605e+00 -7.36846745e-01 3.23611736e-01 8.77909124e-01 2.49745622e-01 -1.91576675e-01 -5.06074019e-02 -4.75172669e-01 -5.26730895e-01 -1.33264053e+00 7.00927079e-01 3.31857204e-02 4.06588703e-01 2.84600526e-01 -7.89285541e-01 1.26374316e+00 1.98089600e-01 6.66083455e-01 8.41989648e-03 2.16130272e-01 -2.22670078e-01 2.49651998e-01 -4.37733769e-01 2.66727537e-01 -1.09974340e-01 -3.39441523e-02 -7.23841935e-02 -2.94216037e-01 -5.72129786e-02 -3.25537354e-01 -4.88580674e-01 1.62526524e+00 -3.07896994e-02 1.54943824e-01 -5.50857782e-01 6.39848948e-01 4.30262119e-01 9.26647067e-01 8.42841506e-01 -3.36911023e-01 -2.74509311e-01 -4.21456188e-01 -6.44961178e-01 -6.62476301e-01 -1.26209152e+00 1.34129807e-01 6.68223619e-01 1.94883600e-01 3.23472530e-01 -1.92209974e-01 -1.22252509e-01 1.34545907e-01 6.89092815e-01 -6.32033944e-01 -1.70639858e-01 -8.00627649e-01 -1.05156600e+00 5.03422208e-02 2.64252812e-01 2.68903852e-01 -7.16953516e-01 -1.10153496e+00 6.86583936e-01 3.81747007e-01 -1.06185365e+00 -2.69008223e-02 9.47590545e-02 -8.26099157e-01 -1.24566841e+00 -4.89675105e-01 -2.83612370e-01 5.29316604e-01 3.68015498e-01 4.03856128e-01 -7.59329140e-01 -7.15802684e-02 1.23716128e+00 -2.18763366e-01 -1.09649360e+00 -6.33148372e-01 -5.99648535e-01 3.48771513e-01 3.42339128e-01 2.95204371e-01 -1.44430906e-01 -5.25486052e-01 4.49673176e-01 -7.07692266e-01 -4.74583745e-01 4.81500268e-01 4.15966928e-01 7.24448919e-01 6.16212904e-01 -5.30602299e-02 -2.32142851e-01 4.54123855e-01 -4.62841064e-01 -1.26535070e+00 1.32194072e-01 2.83337720e-02 -2.56191224e-01 4.87626076e-01 -9.72332358e-01 -9.53815937e-01 2.42227495e-01 4.46222812e-01 -6.56785071e-01 -3.72560263e-01 1.27796352e-01 -3.83009017e-02 -4.97664243e-01 2.76833206e-01 5.69598019e-01 2.31555313e-01 -1.68313622e-01 8.77466649e-02 7.26265162e-02 4.99868810e-01 -1.03200272e-01 1.44733882e+00 9.03152823e-01 6.94811761e-01 -1.19771361e+00 -5.24489582e-01 -7.26147592e-01 -3.49473119e-01 -5.07555306e-01 9.48941410e-01 -1.03130567e+00 -7.90700316e-01 6.92803562e-01 -9.06609118e-01 -3.10408354e-01 -3.17490071e-01 8.51637781e-01 -4.83448118e-01 3.62324953e-01 -3.23871493e-01 -1.46940565e+00 -4.36830282e-01 -8.96655560e-01 1.01275992e+00 4.66524363e-01 8.91099870e-02 -1.44324970e+00 5.65590143e-01 -1.60808608e-01 7.13851154e-01 8.65542471e-01 2.20947698e-01 -8.35225880e-01 -6.37997210e-01 -5.96414983e-01 4.24006134e-01 2.38363162e-01 2.52603024e-01 -1.14083536e-01 -3.96056086e-01 -9.30943549e-01 6.90513790e-01 6.85711980e-01 2.28520721e-01 9.67438161e-01 -1.00033633e-01 -3.66948247e-01 -8.68750811e-01 2.17472091e-01 1.51090539e+00 7.28031814e-01 2.01439247e-01 2.39849389e-01 7.27834046e-01 4.68789876e-01 1.25477791e+00 6.47350371e-01 4.68900025e-01 6.28276110e-01 7.93884099e-01 2.99894601e-01 4.88063842e-01 6.70227110e-02 7.03964770e-01 2.92711079e-01 1.21510595e-01 -2.50687122e-01 -9.67234492e-01 4.36593503e-01 -2.07270026e+00 -1.12625825e+00 -6.30359292e-01 2.41489768e+00 -1.81244686e-01 8.45426321e-02 -6.54061185e-03 -3.72858554e-01 8.68470788e-01 -3.33045945e-02 -7.37374067e-01 1.12346992e-01 1.15014732e-01 -3.27510312e-02 1.09282506e+00 6.48050785e-01 -1.48109245e+00 5.70142150e-01 6.01069021e+00 5.65193176e-01 -1.06425703e+00 9.30562094e-02 -2.02720448e-01 3.98370802e-01 5.47111452e-01 -5.97240254e-02 -1.51141214e+00 5.25656641e-01 1.51185930e+00 -9.93148685e-02 -3.74990314e-01 6.64053679e-01 8.48163724e-01 -3.43676627e-01 -6.10648274e-01 7.09025800e-01 -1.50636837e-01 -7.69980609e-01 -2.78286844e-01 1.85470730e-01 4.31749314e-01 4.50941890e-01 2.44264707e-01 3.51449519e-01 6.16825938e-01 -5.00116765e-01 3.13895553e-01 1.00388718e+00 -2.88745075e-01 -6.44903004e-01 6.77514374e-01 8.99820209e-01 -1.43127012e+00 -2.63962477e-01 -4.78441656e-01 1.48750558e-01 6.91754937e-01 4.90686446e-01 -8.84448469e-01 6.61461473e-01 7.23453164e-01 5.51678777e-01 -3.05837780e-01 1.46723557e+00 1.72859624e-01 7.67946482e-01 -6.50344610e-01 -1.61735713e-01 5.12131512e-01 -2.25182861e-01 1.58459508e+00 9.34322119e-01 7.27801144e-01 -6.26956746e-02 6.02854848e-01 2.80526340e-01 1.13197887e+00 -3.08858842e-01 -6.89741313e-01 1.56859934e-01 2.59102166e-01 1.12845755e+00 -4.71791863e-01 -1.70122847e-01 -1.28059134e-01 4.45550174e-01 -4.19186711e-01 2.70132780e-01 -1.03932977e+00 2.27618873e-01 8.18653226e-01 6.07339330e-02 9.01775658e-01 -7.76900887e-01 6.21653855e-01 -6.30684137e-01 -2.68688768e-01 -3.14424485e-01 2.92812079e-01 -3.65191728e-01 -1.30617762e+00 6.24974787e-01 2.75473386e-01 -1.70738029e+00 -3.57139587e-01 -4.30390596e-01 -8.30315351e-01 1.09541166e+00 -1.21142113e+00 -1.24587429e+00 -2.56161511e-01 5.36081910e-01 4.86889660e-01 -2.09637269e-01 6.84781790e-01 -2.50651240e-01 -2.04988182e-01 -3.57248068e-01 3.69932026e-01 -2.95048714e-01 2.06944942e-01 -9.43727911e-01 3.13379347e-01 1.12532604e+00 -3.14803004e-01 2.42842257e-01 1.46268451e+00 -1.31082809e+00 -1.83842111e+00 -1.71606863e+00 1.04205869e-01 -6.94418371e-01 9.28115249e-01 -5.86253032e-02 -8.93036902e-01 6.62209392e-01 -3.49905826e-02 -9.70302429e-03 6.06612146e-01 -6.05616212e-01 2.34214693e-01 -5.77711798e-02 -1.17423153e+00 2.64568835e-01 3.27368587e-01 3.92873198e-01 -5.47633588e-01 2.13668346e-01 5.01099765e-01 -4.28399414e-01 -9.49789524e-01 9.55176175e-01 1.27113789e-01 -3.83248150e-01 8.45402598e-01 -4.39019114e-01 -1.15113056e+00 -8.78902972e-01 -3.14801663e-01 -1.44043469e+00 -6.01093054e-01 -8.13615263e-01 -5.03076077e-01 8.43118906e-01 3.63547727e-02 -7.72928894e-01 8.36389542e-01 3.49097252e-01 -1.20888487e-01 -3.82628322e-01 -1.32433987e+00 -1.14728761e+00 -3.63820881e-01 -3.48290831e-01 3.93745601e-02 2.92832762e-01 -7.16486216e-01 -6.06986694e-02 -7.16207325e-01 1.34359336e+00 1.25124073e+00 -2.04421222e-01 9.46954250e-01 -1.36882913e+00 -4.93944824e-01 3.06664467e-01 -8.95990551e-01 -9.14031148e-01 -8.00229385e-02 4.94862124e-02 4.14112806e-01 -1.29052234e+00 -2.60020792e-01 -2.11522654e-01 6.32995218e-02 -3.10062885e-01 -1.54068425e-01 -4.72504348e-01 9.40317959e-02 2.39147142e-01 -4.91319716e-01 7.47586966e-01 8.82356286e-01 -1.07863836e-01 -2.34907597e-01 9.99360085e-01 -4.68352139e-02 5.84848702e-01 7.47310996e-01 -6.56048656e-01 -4.94544894e-01 -1.35214150e-01 -4.07474995e-01 3.45517993e-01 4.79152471e-01 -1.83430684e+00 5.10721684e-01 -3.58294636e-01 3.65150720e-01 -1.28004646e+00 6.62059307e-01 -1.38851261e+00 8.71850610e-01 1.06984997e+00 3.96985680e-01 4.93936718e-01 6.05576396e-01 1.44712710e+00 1.63483806e-02 8.46669674e-02 9.14099872e-01 1.83050692e-01 -7.77443886e-01 7.30203331e-01 -8.12961519e-01 -3.30737948e-01 1.59100723e+00 -1.48259759e-01 -1.65031284e-01 -3.43770415e-01 -1.00357246e+00 2.26415098e-01 6.44427612e-02 4.10366237e-01 5.37066817e-01 -8.80583346e-01 -5.67653954e-01 -1.54661521e-01 -1.31588623e-01 -1.27418339e-01 6.04086399e-01 1.00792372e+00 -3.13710451e-01 6.19016469e-01 6.49897009e-02 -1.03268611e+00 -1.33193874e+00 6.45092845e-01 4.68386054e-01 -3.37931871e-01 -7.23844051e-01 4.33215499e-01 -5.75644411e-02 -1.55388460e-01 3.55447605e-02 -3.67251694e-01 -2.80695200e-01 -3.98948580e-01 8.13486755e-01 5.06728411e-01 -3.71008277e-01 -9.99923170e-01 -2.45376885e-01 7.43845046e-01 1.57864496e-01 -7.67231956e-02 1.18834126e+00 -1.57146960e-01 2.47068927e-01 6.62104607e-01 4.58176941e-01 -3.61429274e-01 -2.01012897e+00 -1.66241363e-01 2.92141289e-01 -8.31111595e-02 -1.39607102e-01 -2.93071628e-01 -3.00587863e-01 3.08483303e-01 1.16263902e+00 2.10950747e-01 7.67745614e-01 -1.68804154e-01 3.51651192e-01 3.10717583e-01 7.13426948e-01 -5.80766320e-01 -3.00899267e-01 6.48624778e-01 5.82146108e-01 -1.00965083e+00 -3.15406144e-01 -3.89274806e-01 -6.24391198e-01 9.55348730e-01 3.39961886e-01 -5.75599849e-01 1.10141551e+00 4.28741187e-01 1.35049611e-01 -1.94140658e-01 -8.00457537e-01 -2.22024083e-01 2.49510288e-01 1.23599052e+00 -4.35553461e-01 2.51684159e-01 1.26206756e-01 -1.29410312e-01 2.55163342e-01 -5.40090740e-01 2.17757776e-01 1.45169926e+00 -9.63405669e-01 -6.85678184e-01 -1.03753281e+00 2.22508341e-01 -7.91129768e-02 3.20331752e-01 3.27102900e-01 8.72300744e-01 -7.20287338e-02 1.17969167e+00 1.22100636e-02 -1.13149710e-01 5.58051288e-01 -1.41014889e-01 2.44002223e-01 -4.92068768e-01 -3.57221484e-01 1.90363511e-01 -3.20849478e-01 -4.94511068e-01 -5.20710409e-01 -1.11782336e+00 -8.08018565e-01 3.15883197e-02 1.02836631e-01 3.21012735e-01 7.37834632e-01 9.87560153e-01 4.64624345e-01 5.02678216e-01 5.38935363e-01 -1.01661217e+00 -7.91004062e-01 -1.11284161e+00 -3.46438199e-01 -5.62167764e-02 4.02723491e-01 -9.97454345e-01 -4.63257998e-01 -1.79221760e-02]
[6.553560256958008, -1.995778203010559]
81d15ea3-fdb4-4398-a79c-6dc701e80b5d
improving-the-naturalness-and-diversity-of
null
null
https://aclanthology.org/2020.inlg-1.7
https://aclanthology.org/2020.inlg-1.7.pdf
Improving the Naturalness and Diversity of Referring Expression Generation models using Minimum Risk Training
In this paper we consider the problem of optimizing neural Referring Expression Generation (REG) models with sequence level objectives. Recently reinforcement learning (RL) techniques have been adopted to train deep end-to-end systems to directly optimize sequence-level objectives. However, there are two issues associated with RL training: (1) effectively applying RL is challenging, and (2) the generated sentences lack in diversity and naturalness due to deficiencies in the generated word distribution, smaller vocabulary size, and repetitiveness of frequent words and phrases. To alleviate these issues, we propose a novel strategy for training REG models, using minimum risk training (MRT) with maximum likelihood estimation (MLE) and we show that our approach outperforms RL w.r.t naturalness and diversity of the output. Specifically, our approach achieves an increase in CIDEr scores between 23%-57% in two datasets. We further demonstrate the robustness of the proposed method through a detailed comparison with different REG models.
['Dimitra Gkatzia', 'Emma Hart', 'Nikolaos Panagiaris']
null
null
null
null
inlg-acl-2020-12
['referring-expression-generation']
['computer-vision']
[ 2.98411340e-01 2.42857546e-01 -1.37428259e-02 -2.94174641e-01 -1.17690289e+00 -5.82204878e-01 5.17701924e-01 -1.85744986e-02 -5.55918097e-01 1.17535925e+00 4.75410849e-01 -1.03685118e-01 1.72188075e-03 -8.01907241e-01 -7.77056932e-01 -4.33208913e-01 8.12107977e-03 2.21776694e-01 -3.63048911e-01 -4.50599164e-01 3.47887307e-01 3.47982407e-01 -1.53355575e+00 2.98122138e-01 1.21044135e+00 8.13090920e-01 4.15753901e-01 7.24803925e-01 -4.14167851e-01 1.14312029e+00 -1.01001346e+00 -5.54009438e-01 -2.44861376e-02 -7.30552375e-01 -8.34559739e-01 -1.67928874e-01 1.59763515e-01 -2.14019850e-01 -1.55325513e-02 9.84806716e-01 7.43903041e-01 5.08151650e-01 6.09709680e-01 -1.02273571e+00 -7.36035347e-01 7.92680383e-01 -3.68130714e-01 -2.79763341e-02 4.24038738e-01 7.95960054e-02 1.27310157e+00 -8.05753887e-01 7.58376539e-01 1.36810923e+00 3.34103495e-01 8.86007547e-01 -9.95230377e-01 -6.63445771e-01 1.66395873e-01 -4.05487157e-02 -1.43227768e+00 -5.02912939e-01 6.28097892e-01 2.30878219e-02 1.48732698e+00 1.55056208e-01 3.45811903e-01 1.26944721e+00 -1.49089815e-02 1.09969056e+00 8.28888535e-01 -6.09072626e-01 3.53106737e-01 1.73040941e-01 -2.60481238e-01 7.46190548e-01 -7.95252919e-02 1.62290141e-01 -6.51216328e-01 6.84608966e-02 5.58109164e-01 -7.28540123e-01 -1.51099920e-01 3.27999629e-02 -1.02076089e+00 1.07495022e+00 2.66420573e-01 4.60983902e-01 -4.84625697e-01 3.84369910e-01 5.04449069e-01 5.81133887e-02 2.83878863e-01 8.40379596e-01 -3.34078550e-01 -4.59787816e-01 -8.54107559e-01 4.28755164e-01 6.78020179e-01 1.27951205e+00 3.71488839e-01 4.29967493e-01 -5.20637095e-01 1.18771696e+00 1.87697843e-01 3.16636831e-01 7.82850266e-01 -9.32994246e-01 7.55859673e-01 1.62647575e-01 7.57942125e-02 -7.98362613e-01 -2.04256356e-01 -5.32033443e-01 -7.75272191e-01 -1.34738326e-01 1.18855081e-01 -4.73335236e-01 -6.98013544e-01 2.25438666e+00 -1.11323476e-01 -1.19630001e-01 4.52706188e-01 5.44278741e-01 8.31280529e-01 9.24406111e-01 3.54004145e-01 -3.30299735e-01 1.16396844e+00 -9.52953875e-01 -1.00871730e+00 -3.03061038e-01 9.17926371e-01 -5.39148569e-01 1.24915385e+00 3.73659164e-01 -1.24852931e+00 -5.21074951e-01 -9.90153074e-01 -7.84816444e-02 -2.74836421e-01 3.81940961e-01 3.93764734e-01 7.44052470e-01 -1.08583009e+00 5.72723985e-01 -3.82000566e-01 -1.83952972e-01 3.52816522e-01 3.97341102e-01 -1.57832041e-01 2.49427944e-01 -1.46444964e+00 9.19898331e-01 5.94743490e-01 -1.59398690e-01 -6.32843792e-01 -5.45010030e-01 -9.48315203e-01 1.28984183e-01 1.79547638e-01 -6.13199472e-01 1.35687768e+00 -1.04198635e+00 -1.98338282e+00 6.61014795e-01 -3.47509593e-01 -6.06955528e-01 4.62333947e-01 -5.52902341e-01 -2.22847119e-01 3.34383287e-02 -8.79927352e-03 9.31064487e-01 4.93443608e-01 -1.08628690e+00 -5.71151972e-01 6.30581826e-02 4.62155119e-02 2.23171517e-01 -2.97952294e-01 3.75681408e-02 -2.44780570e-01 -9.56811428e-01 -5.18316805e-01 -7.31559455e-01 -1.51574567e-01 -5.70574462e-01 -5.00882745e-01 -4.81464416e-01 3.04537535e-01 -7.11300433e-01 1.41016817e+00 -2.07078028e+00 1.74126193e-01 -9.09916218e-03 -2.19784379e-01 4.59150702e-01 -4.07087654e-01 3.70806932e-01 -3.56824733e-02 4.51495677e-01 -2.97751456e-01 -4.69018847e-01 3.64392921e-02 2.38242060e-01 -5.19561291e-01 -1.75628334e-01 6.93241000e-01 1.07448065e+00 -9.93259788e-01 -4.77806717e-01 -2.17215326e-02 3.67001146e-01 -6.30120158e-01 5.19944370e-01 -4.80179220e-01 3.68700922e-01 -3.83065611e-01 4.70505536e-01 2.90815443e-01 -2.46794671e-02 2.28388593e-01 7.76140690e-02 7.45580881e-04 5.22900522e-01 -9.76442456e-01 1.66975152e+00 -8.75281215e-01 5.11901021e-01 -5.12062788e-01 -1.00469434e+00 1.11324930e+00 4.65182573e-01 7.60385618e-02 -8.56168032e-01 1.49885729e-01 2.50939667e-01 -7.76688308e-02 -5.86642981e-01 8.73867929e-01 -3.79358113e-01 -2.22213313e-01 4.78020221e-01 3.58839810e-01 -1.48164660e-01 3.14099103e-01 8.16348717e-02 8.81056130e-01 4.08326060e-01 4.45049256e-01 -9.86500755e-02 4.99357402e-01 -1.95155203e-01 5.47350526e-01 8.46508503e-01 2.05074269e-02 4.92406785e-01 6.28236711e-01 -2.58926093e-03 -1.10191691e+00 -8.02144468e-01 3.11262041e-01 1.07725978e+00 -1.32979244e-01 -3.10737640e-01 -8.19156170e-01 -6.25216126e-01 -4.49020535e-01 1.47158194e+00 -4.72622603e-01 -3.71763289e-01 -9.24930930e-01 -7.40847349e-01 1.07931590e+00 6.12770140e-01 5.01309693e-01 -1.39265251e+00 -5.86079657e-01 4.48435724e-01 -5.35247445e-01 -1.25910699e+00 -3.20962161e-01 1.16315216e-01 -9.43113983e-01 -3.65629315e-01 -8.28110874e-01 -6.77040994e-01 4.01104331e-01 -2.82914758e-01 1.34560621e+00 -2.65694335e-02 -9.74213555e-02 -1.28836960e-01 -4.82729524e-01 -3.49514842e-01 -7.27233946e-01 2.99214035e-01 -1.36974484e-01 -2.36938626e-01 6.21238165e-02 -3.47422481e-01 -2.26035565e-01 -8.69701207e-02 -9.76832032e-01 4.72511686e-02 8.18859279e-01 9.16892350e-01 4.98126358e-01 -1.45968243e-01 1.16460943e+00 -8.05051565e-01 1.20401382e+00 -4.44449574e-01 -5.19327104e-01 4.95109320e-01 -5.35574138e-01 5.05306602e-01 7.12530017e-01 -4.46444929e-01 -1.18692553e+00 -1.41754225e-01 -5.47908366e-01 -1.86627388e-01 -6.96942732e-02 4.23306584e-01 -2.26923093e-01 3.29179317e-01 6.73759937e-01 3.65172654e-01 -1.55544085e-02 -1.88596815e-01 5.26675403e-01 6.49715781e-01 2.71705061e-01 -7.48759449e-01 3.39937389e-01 -1.38905585e-01 -2.35657915e-01 -6.42888665e-01 -8.69755983e-01 8.18425119e-02 -1.69679940e-01 -1.08378276e-01 7.11928308e-01 -8.44724417e-01 -7.03682661e-01 1.41632199e-01 -1.35633099e+00 -3.65993142e-01 -3.02343816e-01 3.47270221e-01 -8.51758182e-01 3.93967688e-01 -5.49423456e-01 -1.21279538e+00 -7.74800122e-01 -1.21615398e+00 1.12115419e+00 2.61194438e-01 -5.81353247e-01 -9.17258561e-01 -9.91628692e-02 2.18380541e-01 5.15638947e-01 3.15683007e-01 1.26118279e+00 -6.64684057e-01 -3.85505438e-01 2.14179396e-03 -7.03887120e-02 5.10182381e-01 7.14971349e-02 -7.38238692e-02 -8.71639609e-01 1.77161098e-02 -1.83394745e-01 -7.02695429e-01 4.76977438e-01 2.07854643e-01 1.20376921e+00 -4.68241125e-01 2.40627397e-03 3.14908594e-01 1.44805837e+00 3.03363651e-01 7.90457785e-01 3.25145751e-01 3.78995568e-01 7.36171484e-01 8.58261108e-01 5.19738436e-01 2.26879388e-01 7.09984183e-01 2.10034013e-01 1.62146851e-01 -1.45811886e-01 -5.31247199e-01 5.65575004e-01 6.55617833e-01 6.21758103e-02 -7.51131177e-01 -6.70526266e-01 6.52663887e-01 -1.78960550e+00 -9.26149487e-01 2.33515576e-01 2.04752803e+00 1.17495143e+00 7.47890398e-03 1.36685967e-01 -5.89722069e-03 5.64038098e-01 2.06983206e-03 -5.06630063e-01 -6.99874699e-01 -4.67410415e-01 4.74235535e-01 2.59766579e-01 5.81756294e-01 -6.56245649e-01 1.28692341e+00 6.65845728e+00 1.11885142e+00 -1.00249779e+00 -1.50096804e-01 6.78218305e-01 -1.81916907e-01 -4.83615726e-01 -3.87391448e-01 -9.64507282e-01 3.29666495e-01 1.06601894e+00 -2.23479018e-01 4.56426173e-01 6.15889549e-01 1.98982239e-01 1.33409604e-01 -9.70692575e-01 9.15915072e-01 1.62209809e-01 -1.25277781e+00 3.42475325e-01 -2.12890774e-01 7.02044785e-01 -2.96240568e-01 2.97547858e-02 6.85961366e-01 3.73977423e-01 -1.37338102e+00 8.60352576e-01 4.19167310e-01 6.55499756e-01 -1.14477265e+00 7.10237443e-01 4.39823359e-01 -7.31371522e-01 2.16241851e-02 -2.39340648e-01 1.76919580e-01 3.11980456e-01 4.92692709e-01 -9.96767759e-01 6.42424703e-01 2.73761243e-01 3.50381285e-01 -2.63582349e-01 5.66764057e-01 -5.33682585e-01 5.04989326e-01 -1.74048364e-01 -6.44177914e-01 4.84215915e-01 -2.96137519e-02 5.17178118e-01 1.44675982e+00 5.12265444e-01 -5.51512726e-02 -3.56625408e-01 1.09456301e+00 -3.19970876e-01 3.48538131e-01 -6.27682149e-01 -2.59800673e-01 4.82921392e-01 9.24067020e-01 -3.19979966e-01 -2.97311068e-01 -8.68993253e-02 9.98572946e-01 6.47747278e-01 3.40853691e-01 -7.88385630e-01 -6.15663469e-01 5.86308658e-01 -3.90004724e-01 3.77289683e-01 -2.15096608e-01 -2.84500808e-01 -9.72394168e-01 8.17556400e-03 -1.17456400e+00 2.34707355e-01 -8.43330324e-01 -1.19824111e+00 9.45127249e-01 -7.95781519e-03 -1.03475142e+00 -9.02575970e-01 -4.45297152e-01 -2.99385250e-01 9.66675878e-01 -1.55515444e+00 -7.36636400e-01 9.73215699e-02 3.51044506e-01 8.30793977e-01 -4.04916048e-01 9.77966309e-01 1.88693374e-01 -5.27067363e-01 1.01559699e+00 -8.50901529e-02 1.07387580e-01 3.01575243e-01 -1.20310962e+00 3.57365251e-01 7.21591830e-01 2.30571017e-01 5.89932263e-01 8.77020538e-01 -3.99311900e-01 -1.19335341e+00 -1.16770339e+00 1.22550189e+00 -6.19067848e-02 3.66526097e-01 -3.34293306e-01 -7.24479854e-01 3.90032142e-01 3.81701112e-01 -5.48853278e-01 7.85365283e-01 -5.87885734e-03 -2.14617386e-01 2.61348814e-01 -1.19893324e+00 8.82115841e-01 1.00368679e+00 -3.90530497e-01 -5.74111521e-01 2.78094381e-01 9.68629777e-01 -3.82077992e-01 -7.40018368e-01 4.77971643e-01 4.01371151e-01 -7.49742627e-01 6.82909787e-01 -6.91781282e-01 7.03609943e-01 -6.28087744e-02 -3.69616210e-01 -1.46734500e+00 8.21800008e-02 -7.23153651e-01 -1.40498608e-01 1.44597268e+00 6.28329992e-01 -2.97862113e-01 5.79426825e-01 4.64991808e-01 1.58291161e-02 -1.02295446e+00 -8.38631570e-01 -8.95382345e-01 4.39864963e-01 -4.34176385e-01 6.44644916e-01 5.48081160e-01 4.88648377e-02 6.22986674e-01 -5.68685412e-01 -1.88706845e-01 1.49085343e-01 -1.42367848e-03 4.58657831e-01 -5.66298962e-01 -4.58766788e-01 -4.71456528e-01 2.55695414e-02 -1.21917498e+00 5.10437012e-01 -7.88582563e-01 3.21448505e-01 -1.34560549e+00 -8.11700597e-02 -3.61514419e-01 -4.10220623e-02 2.46946484e-01 -2.90416807e-01 -1.57616019e-01 3.07902873e-01 -1.18101925e-01 -4.29337591e-01 9.48253751e-01 1.04414916e+00 1.81676373e-01 -2.48076960e-01 -1.50488198e-01 -7.77763069e-01 3.08448672e-01 1.11775386e+00 -4.01835710e-01 -4.52906251e-01 -5.94609976e-01 3.59073848e-01 1.86151683e-01 -7.41119087e-02 -6.68184936e-01 -1.74158797e-01 -8.02352326e-04 3.15233544e-02 -4.97120470e-01 3.69859993e-01 -3.31630230e-01 -2.22668394e-01 3.72114658e-01 -8.38613987e-01 2.40669385e-01 4.62458640e-01 4.40018952e-01 -3.52501720e-01 -5.01108229e-01 7.33552575e-01 -1.52451426e-01 -5.27039170e-01 -1.43579856e-01 -6.38636768e-01 2.84578204e-01 8.57289433e-01 1.20088318e-02 -8.42679590e-02 -7.10076511e-01 -2.98064530e-01 1.90294757e-01 -3.69493142e-02 5.95134676e-01 7.37524509e-01 -1.29931986e+00 -7.96163797e-01 -5.06152622e-02 1.54635549e-01 -5.70419431e-02 2.22636741e-02 4.49606895e-01 -3.93687785e-01 5.66004574e-01 -7.80154988e-02 -1.87266007e-01 -1.15515339e+00 2.59161919e-01 3.74638736e-01 -6.79650486e-01 -2.73867995e-01 1.02050519e+00 -7.14606196e-02 -6.18727028e-01 4.02511120e-01 -1.71418101e-01 -3.50860715e-01 -4.60202917e-02 3.90616953e-01 1.90936908e-01 7.77877644e-02 -4.74841446e-01 -1.26034737e-01 3.46325636e-01 -2.03578621e-01 -4.39881265e-01 1.16982687e+00 6.00511692e-02 1.47489473e-01 2.69385070e-01 1.26972663e+00 -1.80345736e-02 -8.14380109e-01 -2.05435932e-01 3.14467758e-01 -1.10644430e-01 -3.64733413e-02 -8.97534609e-01 -8.48436713e-01 9.36295390e-01 3.57632548e-01 -3.65170836e-02 1.10355163e+00 -1.60769433e-01 1.05528629e+00 5.44138432e-01 1.83966666e-01 -1.18895483e+00 2.13244751e-01 7.02313244e-01 9.14411426e-01 -7.69922554e-01 -3.89361143e-01 -2.36531898e-01 -1.02838814e+00 1.03427577e+00 5.67156076e-01 -8.64315107e-02 5.45296073e-02 1.37935415e-01 3.34277726e-03 1.51079580e-01 -1.05493653e+00 -2.41134763e-01 -1.63123989e-03 5.08335471e-01 8.28548312e-01 -9.02586356e-02 -4.99916434e-01 5.45966744e-01 -4.95263159e-01 1.25458702e-01 3.80478054e-01 7.75018692e-01 -3.17609340e-01 -1.33256412e+00 -3.39074899e-03 2.11070120e-01 -7.25713909e-01 -3.87895465e-01 -3.89607996e-01 6.42432451e-01 -1.40053898e-01 1.13041008e+00 1.14521198e-02 -3.03353727e-01 4.46100622e-01 2.00125039e-01 4.62523073e-01 -5.77869773e-01 -5.99769473e-01 -1.28379509e-01 4.93008137e-01 -4.46159095e-01 -3.81192952e-01 -6.03387415e-01 -1.39654171e+00 2.65559903e-03 -3.06435943e-01 1.51891336e-01 6.08439863e-01 1.07143867e+00 5.29001296e-01 6.96123123e-01 5.89957893e-01 -2.85117060e-01 -8.32426727e-01 -1.12424934e+00 -3.85195285e-01 3.36858153e-01 1.93250373e-01 -4.46087301e-01 -4.27659117e-02 -7.76270181e-02]
[11.885564804077148, 9.175776481628418]
0c5ec7fd-97d6-4724-9e07-c68916335205
towards-automatic-neural-architecture-search
2305.18030
null
https://arxiv.org/abs/2305.18030v1
https://arxiv.org/pdf/2305.18030v1.pdf
Towards Automatic Neural Architecture Search within General Super-Networks
Existing neural architecture search (NAS) methods typically rely on pre-specified super deep neural networks (super-networks) with handcrafted search spaces beforehand. Such requirements make it challenging to extend them onto general scenarios without significant human expertise and manual intervention. To overcome the limitations, we propose the third generation of Only-Train-Once (OTOv3). OTOv3 is perhaps the first automated system that trains general super-networks and produces high-performing sub-networks in the one shot manner without pretraining and fine-tuning. Technologically, OTOv3 delivers three noticeable contributions to minimize human efforts: (i) automatic search space construction for general super-networks; (ii) a Hierarchical Half-Space Projected Gradient (H2SPG) that leverages the dependency graph to ensure the network validity during optimization and reliably produces a solution with both high performance and hierarchical group sparsity; and (iii) automatic sub-network construction based on the super-network and the H2SPG solution. Numerically, we demonstrate the effectiveness of OTOv3 on a variety of super-networks, including RegNet, StackedUnets, SuperResNet, and DARTS, over benchmark datasets such as CIFAR10, Fashion-MNIST, ImageNet, STL-10, and SVNH. The sub-networks computed by OTOv3 achieve competitive even superior performance compared to the super-networks and other state-of-the-arts. The library will be released at https://github.com/tianyic/only_train_once.
['Ilya Zharkov', 'Tianyu Ding', 'Luming Liang', 'Tianyi Chen']
2023-05-25
null
null
null
null
['architecture-search']
['methodology']
[-1.51655778e-01 -4.27800678e-02 -2.12471291e-01 -3.80604118e-01 -5.22487879e-01 -3.95929337e-01 1.48560688e-01 -6.18088901e-01 -4.87796396e-01 7.25937724e-01 1.75039414e-02 -4.29538608e-01 -4.84967113e-01 -5.09046674e-01 -8.46844852e-01 -6.99726880e-01 -1.25581622e-01 5.37579894e-01 2.53143664e-02 -4.68270361e-01 -9.79040340e-02 4.47092980e-01 -1.28836226e+00 9.47988182e-02 8.06282699e-01 1.39774275e+00 3.61943334e-01 3.07458282e-01 2.55378038e-01 4.65602219e-01 -9.04606935e-03 -2.28943437e-01 7.84515023e-01 -1.73598081e-01 -7.99861372e-01 -3.07180375e-01 6.41764879e-01 -1.97296694e-01 -4.19462442e-01 1.09919190e+00 7.80716836e-01 2.33781293e-01 9.68226641e-02 -1.20177734e+00 -5.79638541e-01 9.79334176e-01 -2.20397666e-01 2.98541218e-01 -4.98353332e-01 3.77879053e-01 1.41432965e+00 -1.26752770e+00 5.14759123e-01 1.17384863e+00 8.24616730e-01 6.65737569e-01 -1.26699340e+00 -1.21549094e+00 4.00391817e-01 2.47962311e-01 -1.53833878e+00 -6.28182411e-01 5.99169314e-01 -7.31416866e-02 1.17442918e+00 2.38132983e-01 6.36220694e-01 1.35099721e+00 -2.37035856e-01 8.63362849e-01 6.65716469e-01 -1.70826167e-01 4.31329787e-01 -9.01384428e-02 4.06044364e-01 9.13442194e-01 1.80561841e-01 2.46702820e-01 -7.12507665e-01 -1.60801753e-01 9.74934876e-01 -1.26060516e-01 -3.45180273e-01 -3.60464811e-01 -1.08805561e+00 8.42747509e-01 7.67878532e-01 3.98457944e-01 -5.69907308e-01 3.98998022e-01 3.37804109e-01 3.91792893e-01 3.96387428e-01 8.79365623e-01 -9.09709990e-01 1.80437386e-01 -1.18246436e+00 2.27279097e-01 7.16756821e-01 9.97523963e-01 5.91311336e-01 6.86582148e-01 -8.15038010e-02 1.05213642e+00 3.75454612e-02 2.80676838e-02 7.17837155e-01 -9.12840486e-01 5.11772394e-01 6.08124852e-01 -3.43337089e-01 -7.23753810e-01 -5.46342552e-01 -1.34316945e+00 -1.07778823e+00 -6.87963367e-02 1.58906668e-01 -4.09686267e-01 -1.21976054e+00 2.00648642e+00 1.75633684e-01 4.07712281e-01 -1.29837198e-02 9.96466517e-01 1.11306238e+00 3.37112457e-01 -2.24726200e-01 1.85168862e-01 1.10805047e+00 -1.44039607e+00 -1.59038335e-01 -4.64135319e-01 6.00480437e-01 -1.10153198e-01 1.05106890e+00 4.22310889e-01 -1.19121552e+00 -5.34892499e-01 -1.14253664e+00 1.39549732e-01 -2.69507527e-01 2.27751017e-01 7.09278345e-01 3.15311432e-01 -1.59713960e+00 7.33600259e-01 -8.60951841e-01 -1.90007091e-01 7.27582514e-01 9.52268124e-01 -3.11911404e-01 1.63637012e-01 -1.24833095e+00 6.23257101e-01 7.04896629e-01 4.08017218e-01 -1.14351726e+00 -9.80824888e-01 -5.94592333e-01 4.27127719e-01 5.86951494e-01 -8.47663581e-01 1.07570493e+00 -1.13736010e+00 -1.38248074e+00 5.36812544e-01 8.19094828e-04 -7.79463291e-01 3.20357561e-01 -2.90272474e-01 -1.07460923e-01 -7.11009651e-02 -1.85638577e-01 9.70450640e-01 7.21285284e-01 -9.18293595e-01 -5.15381217e-01 -4.37879294e-01 -1.91240847e-01 2.59887129e-01 -6.33331656e-01 -1.52407825e-01 -7.80065596e-01 -8.21768045e-01 2.90156752e-01 -1.12427711e+00 -6.40914083e-01 -2.55375564e-01 -8.90577018e-01 -3.44847351e-01 7.00580180e-01 -4.15879250e-01 1.26819670e+00 -2.02358484e+00 3.86921287e-01 4.80741978e-01 5.56615889e-01 5.51630020e-01 -6.08575821e-01 -1.66109160e-01 -3.37730259e-01 1.32961541e-01 1.31400630e-01 -3.98879021e-01 -5.56257255e-02 8.92738774e-02 -3.53929624e-02 1.78978801e-01 -1.23503484e-01 1.21663678e+00 -6.63999498e-01 -1.57558903e-01 -2.03980565e-01 4.14275825e-01 -8.15532565e-01 -2.51821727e-01 -2.02000648e-01 1.05675906e-01 -6.44177020e-01 7.81301558e-01 2.19806939e-01 -7.63112009e-01 2.00556338e-01 -4.72085685e-01 -2.00287648e-03 3.21389556e-01 -1.06260908e+00 1.72157300e+00 -2.43985891e-01 4.84755188e-01 2.60772049e-01 -1.11081398e+00 8.86565328e-01 1.92020029e-01 4.47823972e-01 -6.99209690e-01 2.72243381e-01 5.38453698e-01 4.66773063e-02 -2.84368601e-02 2.44760841e-01 4.45238352e-01 2.78906286e-01 2.54762322e-01 5.53819478e-01 4.90663052e-01 2.07397535e-01 4.75738160e-02 1.04320729e+00 -2.19125912e-01 -1.55583605e-01 -6.36701822e-01 2.73388475e-01 -1.68983310e-01 7.81685114e-01 8.65259945e-01 -7.02013308e-03 5.23403108e-01 2.03930885e-01 -7.56870151e-01 -1.08402085e+00 -8.71830463e-01 2.33040266e-02 1.36916399e+00 -6.59912676e-02 -1.93234310e-01 -7.52836049e-01 -5.53391218e-01 -1.64321959e-02 7.10157096e-01 -5.01778543e-01 -1.67749658e-01 -7.22223461e-01 -8.07760358e-01 7.27549493e-01 7.35006988e-01 7.66028941e-01 -1.01738060e+00 -2.87878215e-01 1.68351278e-01 5.66242002e-02 -1.10198843e+00 -7.19975412e-01 5.16632438e-01 -1.01301169e+00 -8.54563773e-01 -6.24492526e-01 -1.15222740e+00 6.75418496e-01 9.57614854e-02 1.09162366e+00 3.34700793e-01 -2.42094189e-01 -2.38179669e-01 -8.06203559e-02 -6.80278167e-02 2.84873307e-01 6.03236973e-01 3.74290049e-01 -6.79951087e-02 1.42651275e-01 -9.76693332e-01 -6.97788715e-01 4.00368005e-01 -4.23277199e-01 3.12311769e-01 7.05002666e-01 1.10784280e+00 7.18024015e-01 -4.63077463e-02 6.39485180e-01 -6.10307813e-01 5.79528213e-01 -4.85744864e-01 -7.99703717e-01 2.66736716e-01 -9.68392134e-01 2.55715102e-01 5.20605147e-01 -4.99836236e-01 -6.56518102e-01 1.60369664e-01 -1.75533786e-01 -9.08959150e-01 2.50596792e-01 6.38129771e-01 5.81540503e-02 -3.50383013e-01 9.16263461e-01 2.91643798e-01 -7.93108717e-02 -6.24091625e-01 1.05117343e-01 1.77255109e-01 5.20794511e-01 -4.19905037e-01 8.25559735e-01 1.92980260e-01 -1.07545465e-01 -5.99467039e-01 -1.10423434e+00 -3.09994519e-01 -4.08263683e-01 1.29915088e-01 6.60300970e-01 -1.00449574e+00 -6.06982529e-01 3.71998936e-01 -8.20800960e-01 -7.69231498e-01 -5.81689030e-02 4.90439862e-01 -8.76693353e-02 -2.27088615e-01 -6.96086586e-01 -3.58680934e-01 -1.03890216e+00 -1.36888623e+00 6.35447025e-01 2.75655299e-01 -5.41588627e-02 -1.03999269e+00 -2.67038256e-01 3.47339958e-01 6.38860404e-01 8.53977650e-02 7.02253044e-01 -9.82703567e-01 -6.71786487e-01 6.38460144e-02 -3.52690995e-01 5.01365662e-01 -1.44902602e-01 -3.83000582e-01 -8.37895870e-01 -5.12280583e-01 -2.45729253e-01 -6.09075129e-01 1.05471396e+00 6.61122501e-01 1.44215667e+00 -6.10594630e-01 -3.34298432e-01 1.32827294e+00 1.36651230e+00 1.22891717e-01 2.93754935e-01 4.80144083e-01 8.23435962e-01 1.35230988e-01 3.17713339e-03 1.47238195e-01 2.46440738e-01 6.44017100e-01 6.16465926e-01 -1.38487801e-01 -1.07738473e-01 -1.40746832e-01 6.32588491e-02 8.76910746e-01 -1.70045272e-01 -1.49718329e-01 -9.93054152e-01 7.46395946e-01 -2.01247740e+00 -7.68582046e-01 1.85101584e-01 1.72531748e+00 9.90300000e-01 1.64144546e-01 -1.41954422e-01 -2.50554413e-01 5.80736697e-01 2.09833160e-01 -1.14573598e+00 6.41746297e-02 -2.59621501e-01 3.51932406e-01 6.72142446e-01 3.18158358e-01 -1.03867912e+00 1.12280178e+00 5.22776222e+00 1.05394578e+00 -1.29234827e+00 2.74962425e-01 9.91603494e-01 -5.80983162e-01 -1.52491882e-01 -2.52554357e-01 -1.19695568e+00 1.36636496e-01 7.47282386e-01 7.56571516e-02 8.35940242e-01 1.26814604e+00 -7.85765499e-02 5.72727263e-01 -1.08495784e+00 1.22482276e+00 -2.08785012e-01 -1.83481145e+00 -6.06659912e-02 4.53166366e-02 9.43747461e-01 8.80232692e-01 2.11929053e-01 5.96559227e-01 5.98512173e-01 -1.23823607e+00 7.17513919e-01 9.03732404e-02 7.76278079e-01 -6.60379827e-01 6.67173862e-01 2.62145817e-01 -1.17232096e+00 -3.62337857e-01 -3.73040587e-01 3.15607995e-01 -7.75192603e-02 5.91408670e-01 -7.58864343e-01 3.48904669e-01 9.55336511e-01 6.35319769e-01 -5.89731276e-01 9.83505130e-01 -4.87621464e-02 7.97132611e-01 -6.41985297e-01 -1.31117806e-01 6.60008967e-01 -2.95987241e-02 6.58450902e-01 9.39932644e-01 3.31529856e-01 -6.50652498e-02 1.58354715e-01 1.08650684e+00 -4.36671376e-01 -1.81108758e-01 -1.24790654e-01 -6.32288530e-02 7.29761124e-01 1.34509242e+00 -5.31270146e-01 -1.69983923e-01 2.92955525e-02 5.37383437e-01 5.75745106e-01 6.15862727e-01 -6.10832214e-01 -1.74728423e-01 7.02666759e-01 -6.80609718e-02 4.87083733e-01 9.52364132e-03 -6.73800051e-01 -1.10533011e+00 -3.94240431e-02 -1.27858424e+00 3.79189253e-01 -7.02092886e-01 -1.14078152e+00 1.30562913e+00 -2.11820945e-01 -6.95338190e-01 -2.16923088e-01 -6.22473300e-01 -5.11051476e-01 7.48432577e-01 -1.32596719e+00 -1.07973146e+00 -3.32282275e-01 7.07019866e-01 7.56656170e-01 -7.14011610e-01 5.67157090e-01 4.49395686e-01 -1.03333151e+00 1.04940200e+00 -7.64269754e-02 2.27699876e-01 2.24083975e-01 -8.64636362e-01 6.55981302e-01 6.74055457e-01 1.80111155e-01 7.53786623e-01 5.60464263e-01 -3.82953614e-01 -1.27184415e+00 -1.28333759e+00 5.26651561e-01 -1.36101961e-01 7.51782179e-01 -4.47486848e-01 -8.00428987e-01 7.23035574e-01 8.59351887e-04 1.93656072e-01 1.98210582e-01 4.46582258e-01 -4.05133575e-01 -4.06813443e-01 -6.52556300e-01 7.76639402e-01 1.45491457e+00 -3.31966877e-01 -2.14947850e-01 5.41493237e-01 1.01990724e+00 -6.04383528e-01 -5.52546561e-01 7.42019594e-01 5.12878180e-01 -8.13960075e-01 1.22528434e+00 -7.62548566e-01 1.86067477e-01 4.55044731e-02 -1.91731676e-01 -1.26377535e+00 -4.72790301e-01 -7.26740420e-01 -1.24953203e-01 6.79050505e-01 1.00713062e+00 -8.19941700e-01 1.30747473e+00 6.05311155e-01 -4.90417302e-01 -1.39226687e+00 -9.21769857e-01 -9.48513806e-01 -1.18029416e-01 -3.76094192e-01 6.85248077e-01 1.10535979e+00 -5.55760860e-01 3.22614729e-01 -3.05695266e-01 1.60720259e-01 5.69971085e-01 -2.00179562e-01 4.85594660e-01 -1.28206587e+00 -4.16207701e-01 -7.61351585e-01 -3.82862724e-02 -1.09107292e+00 2.12895289e-01 -9.98821497e-01 -2.17884965e-02 -1.16707218e+00 4.20899689e-02 -7.63319373e-01 -6.72559857e-01 9.77713525e-01 1.15027148e-02 1.89304322e-01 3.69544439e-02 3.15800458e-01 -6.84381902e-01 6.25482321e-01 1.14294076e+00 9.88653377e-02 -5.34675837e-01 -3.23252343e-02 -8.48888457e-01 8.27251852e-01 1.19496930e+00 -4.65252310e-01 -7.23643303e-01 -6.17920995e-01 3.76784712e-01 -3.31952423e-01 6.54209435e-01 -1.06235588e+00 5.57262599e-01 -3.72631922e-02 2.11836815e-01 -4.33347613e-01 3.26529950e-01 -4.81277794e-01 9.75570306e-02 4.45609868e-01 -4.83005136e-01 3.11508387e-01 3.21209192e-01 2.21852928e-01 -1.25103578e-01 -1.62472248e-01 7.43791044e-01 -2.85570681e-01 -7.27425814e-01 7.50033438e-01 3.67785931e-01 2.23143220e-01 6.32190108e-01 -2.93718159e-01 -3.87795836e-01 -2.88187742e-01 -6.68446958e-01 6.90689504e-01 -7.51447380e-02 3.06235999e-01 7.14531541e-01 -1.32060432e+00 -5.70805252e-01 2.62456685e-01 -3.47367823e-01 2.89604783e-01 2.42600396e-01 9.35147107e-01 -4.33649272e-01 6.68447196e-01 -6.79304376e-02 -5.44646740e-01 -1.11892390e+00 2.38061592e-01 6.61180496e-01 -5.58451295e-01 -5.20085454e-01 1.53412092e+00 3.84626120e-01 -7.93343067e-01 5.36669314e-01 -1.14359953e-01 7.61201009e-02 -3.11577857e-01 3.19422871e-01 1.30332112e-01 1.57097533e-01 -2.59374499e-01 -4.04491484e-01 2.03339770e-01 -3.39442998e-01 8.94583389e-03 1.76313758e+00 2.87425876e-01 5.01457034e-05 -1.33264601e-01 1.25398362e+00 -5.97771466e-01 -1.39430976e+00 -5.63822687e-01 2.37207697e-03 1.31876647e-01 5.87471366e-01 -7.30243981e-01 -1.72164202e+00 5.14838994e-01 3.97173077e-01 -2.18200639e-01 1.27030981e+00 -1.15870312e-02 1.02315426e+00 8.77279222e-01 3.37584227e-01 -1.18567252e+00 2.44638026e-01 7.33876348e-01 1.15945292e+00 -9.08357799e-01 -4.69300412e-02 -6.54618517e-02 -5.43820024e-01 9.11664426e-01 1.03506863e+00 -2.17564419e-01 8.83921683e-01 1.59061879e-01 -1.64031893e-01 -4.57080156e-01 -1.21377933e+00 4.27022204e-02 6.24221921e-01 1.17838316e-01 1.63571686e-02 -3.09709489e-01 2.58363724e-01 7.35708177e-01 -5.24062693e-01 4.34516110e-02 -2.02993020e-01 6.34987593e-01 -3.03470135e-01 -6.81737721e-01 9.58109498e-02 7.36857116e-01 -4.05235797e-01 -6.40571356e-01 -2.31679067e-01 7.95062482e-01 -4.25896533e-02 5.00391901e-01 -1.08161189e-01 -7.50043094e-01 2.69597739e-01 -2.23012622e-02 -2.75284518e-02 -4.65319812e-01 -8.37162316e-01 -4.81022224e-02 3.29024702e-01 -9.46022868e-01 3.75280231e-02 -1.87478244e-01 -1.06634390e+00 -2.32979313e-01 -5.34690976e-01 1.71889678e-01 7.38389134e-01 8.94504607e-01 7.97453582e-01 6.38039052e-01 2.26401329e-01 -1.06656992e+00 -8.74888599e-01 -9.11338568e-01 -2.06274912e-01 -2.98625827e-02 4.46011454e-01 -5.84997714e-01 -4.59442139e-01 -3.99641722e-01]
[8.615863800048828, 3.2816951274871826]
300f6188-9fc0-464f-b8fc-4ec5d7188644
multi-slice-dense-sparse-learning-for
2108.06761
null
https://arxiv.org/abs/2108.06761v1
https://arxiv.org/pdf/2108.06761v1.pdf
Multi-Slice Dense-Sparse Learning for Efficient Liver and Tumor Segmentation
Accurate automatic liver and tumor segmentation plays a vital role in treatment planning and disease monitoring. Recently, deep convolutional neural network (DCNNs) has obtained tremendous success in 2D and 3D medical image segmentation. However, 2D DCNNs cannot fully leverage the inter-slice information, while 3D DCNNs are computationally expensive and memory intensive. To address these issues, we first propose a novel dense-sparse training flow from a data perspective, in which, densely adjacent slices and sparsely adjacent slices are extracted as inputs for regularizing DCNNs, thereby improving the model performance. Moreover, we design a 2.5D light-weight nnU-Net from a network perspective, in which, depthwise separable convolutions are adopted to improve the efficiency. Extensive experiments on the LiTS dataset have demonstrated the superiority of the proposed method.
['Pierce KH Chow', 'Zeng Zeng', 'Yanjie Liu', 'Zeyu Ma', 'Ziyuan Zhao']
2021-08-15
null
null
null
null
['automatic-liver-and-tumor-segmentation', 'sparse-learning']
['medical', 'methodology']
[-1.00885155e-02 1.07889883e-02 -5.17494977e-01 -4.96656299e-01 -2.06483826e-01 -7.81062320e-02 2.63258427e-01 8.26643407e-02 -3.59685570e-01 4.59547102e-01 4.44338888e-01 -4.70871776e-01 -6.45123422e-02 -8.66683543e-01 -2.66516477e-01 -8.10079396e-01 7.06300559e-03 1.74664930e-02 2.34341159e-01 2.97333032e-01 -1.85432434e-01 6.99560881e-01 -4.74373668e-01 -3.37222777e-02 9.94273305e-01 1.15868199e+00 2.40321159e-01 1.70078218e-01 -3.49100977e-01 6.14579499e-01 -9.15502161e-02 8.60702693e-02 4.28869516e-01 -4.33849186e-01 -6.97720587e-01 4.29682970e-01 -2.62336396e-02 -5.07082283e-01 -7.48317838e-01 1.19817519e+00 5.34526587e-01 -4.66270000e-02 2.58577347e-01 -5.48239648e-01 -4.90975887e-01 5.42472899e-01 -7.20537245e-01 3.11496615e-01 -2.74088293e-01 3.94863755e-01 3.17736715e-01 -6.12210393e-01 3.40066314e-01 8.05121422e-01 5.96556306e-01 4.79967535e-01 -9.69606638e-01 -5.81907928e-01 1.52046308e-01 -2.17027515e-01 -1.31924319e+00 -4.47598659e-02 9.27587986e-01 -2.24966243e-01 5.79175651e-01 1.86090708e-01 1.05038631e+00 6.33567512e-01 2.30711609e-01 1.08326137e+00 9.07331228e-01 -7.75257498e-02 -8.42408165e-02 -3.35390836e-01 -4.34700735e-02 9.91194606e-01 3.18675697e-01 9.35261697e-02 1.49182975e-01 2.25936115e-01 1.51451099e+00 5.41958809e-01 -6.38011217e-01 -4.37426984e-01 -1.42432916e+00 8.88396502e-01 1.17955351e+00 5.27966678e-01 -4.70144540e-01 1.12848841e-01 5.70261478e-01 -1.91148445e-01 5.35319626e-01 1.76730573e-01 -2.54210949e-01 2.32845619e-01 -1.03648245e+00 -3.61130059e-01 3.04298282e-01 6.91237092e-01 3.20100158e-01 -5.86001389e-03 -4.45999414e-01 6.47846580e-01 3.64022374e-01 3.86004969e-02 7.33188331e-01 -5.73550820e-01 2.80420601e-01 8.68376970e-01 -2.04062089e-01 -7.93100536e-01 -8.01045656e-01 -7.95857847e-01 -1.63435495e+00 -1.54025927e-01 3.03575426e-01 -1.54093176e-01 -1.37449801e+00 1.25128305e+00 5.08940697e-01 4.93551522e-01 -6.01767227e-02 1.40593553e+00 1.10888684e+00 4.12416220e-01 1.83131337e-01 -4.44036633e-01 1.29455304e+00 -1.28787923e+00 -7.15590477e-01 -4.41889390e-02 8.72438312e-01 -6.94220960e-01 8.64536941e-01 -5.43997288e-02 -1.12658870e+00 -2.60026485e-01 -8.34189355e-01 -7.12278411e-02 2.64889121e-01 1.94354504e-01 1.19973743e+00 4.42961246e-01 -7.65822768e-01 2.56327569e-01 -1.37511075e+00 3.34693231e-02 9.74655688e-01 4.89354402e-01 -9.12010521e-02 -3.96862805e-01 -9.96850133e-01 5.12781501e-01 4.15130526e-01 4.80074853e-01 -9.86327767e-01 -9.37394321e-01 -9.77432668e-01 1.31051198e-01 3.91845733e-01 -7.41065860e-01 1.19572830e+00 -5.87583303e-01 -1.65798116e+00 6.74226224e-01 1.65484305e-02 -4.45704341e-01 5.37018359e-01 3.19882452e-01 -1.96605995e-01 3.71455759e-01 -1.10021599e-01 6.50229633e-01 3.35102081e-01 -7.58481741e-01 -2.83175796e-01 -3.81671846e-01 6.16712943e-02 3.52603614e-01 -3.02413493e-01 -3.64133239e-01 -7.42646575e-01 -8.94194782e-01 5.02167761e-01 -5.36345184e-01 -8.49661827e-01 3.26134801e-01 -6.57959104e-01 2.86945999e-02 7.29371905e-01 -6.07142389e-01 1.18592262e+00 -2.00471783e+00 1.28811702e-01 2.25816250e-01 5.57969451e-01 3.78336161e-01 1.60722271e-01 -4.36332673e-01 3.06204036e-02 -4.57252972e-02 -4.55718070e-01 -1.42467961e-01 -4.17410493e-01 4.33590651e-01 3.84111047e-01 5.95211387e-01 -2.79703755e-02 1.22505021e+00 -8.43382180e-01 -6.63011253e-01 4.11938846e-01 5.11781454e-01 -5.88047564e-01 2.70871371e-01 -2.19914876e-02 7.93263137e-01 -8.45766604e-01 6.82751834e-01 8.86213005e-01 -6.56981468e-01 -1.27959670e-03 -5.53176343e-01 -2.66628295e-01 1.69058144e-01 -5.34413993e-01 2.24121070e+00 -6.46592975e-01 2.05184206e-01 3.53311867e-01 -1.19405162e+00 6.70051992e-01 3.27549189e-01 8.62335742e-01 -8.61812353e-01 3.97057325e-01 2.67100751e-01 1.81345791e-01 -5.28056681e-01 -1.38804261e-02 -3.47410619e-01 1.74502581e-01 3.06314111e-01 -2.30835930e-01 -1.39477655e-01 -1.01482365e-02 5.09165488e-02 8.50133002e-01 -2.32053414e-01 2.25715622e-01 -3.97917718e-01 7.42174208e-01 8.26050341e-02 8.74639094e-01 1.66749418e-01 -2.70843536e-01 6.33430183e-01 5.27899086e-01 -9.01541829e-01 -5.40260077e-01 -7.27774799e-01 -2.98609197e-01 1.29092559e-01 5.55467129e-01 -7.05055296e-02 -6.74545050e-01 -9.92957413e-01 -1.27941951e-01 9.05686691e-02 -6.02460146e-01 -1.24428257e-01 -7.76365161e-01 -1.05213904e+00 2.26634100e-01 6.43878639e-01 8.22605669e-01 -7.23629713e-01 -7.34401286e-01 3.79513413e-01 -4.92829271e-02 -1.01549840e+00 -7.96226561e-01 1.47053033e-01 -1.31695569e+00 -1.12561047e+00 -1.25750768e+00 -9.73346174e-01 1.16363323e+00 4.53805208e-01 9.42014635e-01 3.96117717e-01 -5.31648755e-01 -2.57068247e-01 -7.46615604e-02 3.02906707e-02 2.27724984e-02 2.78357178e-01 -3.56936663e-01 -2.40962386e-01 2.58710086e-01 -6.00111246e-01 -1.07597399e+00 2.09132239e-01 -9.88530397e-01 6.15452826e-01 9.84629273e-01 1.04970622e+00 8.01710606e-01 8.31001997e-03 2.06166059e-01 -1.04093134e+00 3.18303972e-01 -3.68384719e-01 -6.47150636e-01 1.30808949e-01 -3.06006491e-01 -1.60173878e-01 5.93050122e-01 -2.84013003e-01 -9.59534109e-01 2.82027304e-01 -3.98796618e-01 -6.02174699e-01 -2.99628042e-02 8.17707837e-01 -3.27416062e-02 -4.07059699e-01 1.55676290e-01 3.77552897e-01 1.26283512e-01 -4.66406763e-01 -1.26853539e-02 3.06013733e-01 2.51248002e-01 -2.96795815e-01 6.42996371e-01 4.86167520e-01 1.88636586e-01 -3.40668529e-01 -1.04119432e+00 -3.10123146e-01 -5.20601511e-01 4.64127436e-02 1.11200476e+00 -8.63250494e-01 -5.80411434e-01 5.72511017e-01 -1.04115319e+00 -5.25264144e-01 -3.11427116e-01 9.11694109e-01 -9.46874768e-02 3.27150166e-01 -1.02549207e+00 -9.26530659e-02 -5.60270846e-01 -1.63447952e+00 7.23453045e-01 7.77628005e-01 3.14290136e-01 -1.18862224e+00 -3.28935117e-01 3.28186825e-02 6.24951243e-01 3.36305887e-01 9.38897967e-01 -4.48514551e-01 -7.74979591e-01 -1.01204820e-01 -8.87603819e-01 2.15427041e-01 5.18503010e-01 -4.71138924e-01 -3.59585762e-01 -2.75837094e-01 2.06400286e-02 -6.90430254e-02 8.09654295e-01 8.66508186e-01 1.73409665e+00 -1.10812858e-01 -4.38928694e-01 1.28446209e+00 1.34378922e+00 2.36609295e-01 2.29613066e-01 -2.67866552e-02 9.66946483e-01 6.57631978e-02 1.19610295e-01 3.32897067e-01 3.41597378e-01 2.88057536e-01 6.42477930e-01 -8.02851081e-01 -3.33538592e-01 -6.23353720e-02 -4.26923275e-01 1.10870183e+00 1.35352135e-01 1.53773978e-01 -8.94175828e-01 5.45396864e-01 -1.59876227e+00 -3.47112209e-01 3.05075175e-03 1.86336255e+00 7.67665625e-01 1.30426437e-01 -2.72723913e-01 -2.89580494e-01 4.71726954e-01 2.84448922e-01 -6.89461410e-01 2.31608897e-01 2.42841870e-01 2.30455086e-01 5.10215521e-01 3.98763388e-01 -1.22985220e+00 5.53582668e-01 5.25165987e+00 9.45571303e-01 -1.43314409e+00 2.38101050e-01 1.05567729e+00 -5.59202656e-02 -4.08970922e-01 -3.25505286e-01 -4.00439233e-01 4.82798487e-01 -1.57023743e-02 -4.05648947e-02 5.73428757e-02 7.44558334e-01 3.87695968e-01 -1.18136741e-01 -6.92485809e-01 9.99112725e-01 -2.15452597e-01 -1.73141325e+00 -2.86815763e-02 1.73625633e-01 9.01041925e-01 3.92144583e-02 -1.15287185e-01 1.52828068e-01 2.30944425e-01 -1.10377693e+00 -1.99872572e-02 3.30255359e-01 8.95448029e-01 -9.51457322e-01 9.30417955e-01 3.19987535e-01 -1.22532773e+00 2.96529979e-01 -5.06854415e-01 2.40782231e-01 2.62900323e-01 9.63096738e-01 -5.65290213e-01 6.74449682e-01 3.49206835e-01 1.02297914e+00 -2.48616830e-01 1.30107331e+00 -2.16272563e-01 5.07656515e-01 -4.24213469e-01 1.50039822e-01 7.31289744e-01 -3.03485960e-01 2.69635320e-01 1.05153275e+00 3.45929563e-01 5.06122947e-01 5.52724421e-01 8.31287026e-01 -3.59120399e-01 -1.16744809e-01 -1.33178487e-01 1.53094754e-01 2.04982683e-02 1.58932602e+00 -1.09548938e+00 -2.88912416e-01 -4.86265182e-01 8.79964828e-01 2.24334709e-02 3.17019254e-01 -7.06227362e-01 -2.32999399e-01 5.72728395e-01 2.74474584e-02 1.34699136e-01 -1.73899740e-01 -4.21286702e-01 -1.41076887e+00 -2.68960029e-01 -3.65806162e-01 3.83254349e-01 -1.57579079e-01 -1.22720599e+00 6.17700517e-01 -4.54355747e-01 -1.27650321e+00 2.43402377e-01 -4.12540972e-01 -8.72477293e-01 9.51064467e-01 -1.98336422e+00 -1.10548365e+00 -7.88000107e-01 6.75537527e-01 4.43162143e-01 2.32122570e-01 5.41537523e-01 5.56191742e-01 -7.58510888e-01 4.47353125e-01 -1.93695486e-01 5.59879541e-01 2.30563581e-01 -9.81675863e-01 3.04873697e-02 7.71265745e-01 -3.15995008e-01 4.79794443e-01 -6.90983161e-02 -4.72295344e-01 -1.40269482e+00 -1.37760675e+00 3.40233892e-01 4.12069738e-01 3.81310880e-01 2.09009256e-02 -8.82647932e-01 4.44542944e-01 1.16579048e-01 8.76058221e-01 7.03746915e-01 -2.95009822e-01 2.28290960e-01 -4.13294286e-02 -1.34463179e+00 5.05140960e-01 1.03819621e+00 -1.83960065e-01 -2.56523758e-01 6.02671921e-01 9.18773830e-01 -1.09801435e+00 -1.08615875e+00 7.14537561e-01 1.84309751e-01 -8.63093376e-01 1.03310871e+00 -3.21746796e-01 5.88433385e-01 -3.92309189e-01 2.66012698e-01 -1.30988097e+00 -2.87409872e-01 -3.02557528e-01 -2.13353589e-01 4.56600577e-01 1.33956164e-01 -5.01590371e-01 1.12491608e+00 4.58263248e-01 -7.68636882e-01 -1.38601518e+00 -9.01648462e-01 -3.86426985e-01 4.27179299e-02 -2.50234008e-01 6.82116389e-01 9.85545516e-01 -3.05078059e-01 1.81548166e-04 -7.57498965e-02 1.77496657e-01 5.87053835e-01 3.69207799e-01 2.83234179e-01 -9.76099253e-01 -1.21307641e-01 -6.72217131e-01 -3.21541429e-01 -1.74927211e+00 -1.59059837e-01 -1.03582335e+00 -1.32225528e-01 -1.77781177e+00 2.47831702e-01 -7.86321461e-01 -4.82694864e-01 3.92470092e-01 -2.72923201e-01 3.85448307e-01 -5.70027828e-02 2.13996843e-01 -4.51982886e-01 6.96086645e-01 2.11013436e+00 -2.69008666e-01 -1.98423266e-01 4.78236340e-02 -5.01170576e-01 8.26532066e-01 5.82848549e-01 -2.07403168e-01 -3.65473211e-01 -8.00556362e-01 -4.00762618e-01 4.58907694e-01 2.64303178e-01 -8.85246694e-01 4.09777194e-01 -2.03486100e-01 8.56968462e-01 -6.74369574e-01 4.29844372e-02 -1.03691351e+00 -1.70658559e-01 9.32230771e-01 -1.62406325e-01 -3.24039280e-01 9.23022032e-02 2.50981957e-01 -5.30573666e-01 -1.05961179e-02 8.46824348e-01 -4.73657161e-01 -4.10143018e-01 1.23575461e+00 -4.08915989e-02 -1.60727531e-01 1.13084483e+00 4.39765136e-05 1.81320265e-01 2.05886304e-01 -7.15706348e-01 5.42495012e-01 2.59230942e-01 -4.08395678e-02 7.32578456e-01 -1.53748262e+00 -4.44491029e-01 4.32420999e-01 -2.41757780e-01 9.84489858e-01 6.67241693e-01 1.44098425e+00 -8.26844156e-01 5.42876303e-01 -7.05443844e-02 -7.58940041e-01 -7.38498032e-01 4.25943762e-01 6.23525321e-01 -6.45791054e-01 -9.81715620e-01 1.05940974e+00 5.56123078e-01 -3.94419491e-01 1.74151376e-01 -5.78909457e-01 -8.36510211e-02 -3.04744571e-01 5.51476538e-01 -1.42683417e-01 5.04053384e-02 -2.37030908e-01 -2.75510967e-01 4.41069454e-01 -3.20275903e-01 4.82969284e-01 1.41606450e+00 -6.29263669e-02 -1.52857095e-01 -1.85674369e-01 1.21588433e+00 -3.01697075e-01 -1.48681080e+00 -5.00038445e-01 -1.56078786e-01 -5.48269033e-01 7.04592228e-01 -6.25784934e-01 -1.97646058e+00 1.10775745e+00 3.75016749e-01 -2.97927158e-03 1.40786052e+00 -2.90676862e-01 1.49191368e+00 -7.34511912e-02 5.78534454e-02 -4.55633044e-01 -1.65281687e-02 2.58664399e-01 4.71201688e-01 -1.26529360e+00 1.54447451e-01 -6.69116795e-01 -3.49654853e-01 1.37675929e+00 8.04679215e-01 -1.68859288e-01 8.21853459e-01 3.09717357e-01 7.45839551e-02 -2.66350657e-01 -1.37998432e-01 -9.69929323e-02 2.20154598e-01 2.37992540e-01 5.00732422e-01 1.48278296e-01 -3.26327294e-01 7.45182574e-01 4.47076976e-01 3.90610725e-01 2.83044368e-01 7.91455746e-01 -1.19253576e-01 -8.32770050e-01 1.15837872e-01 5.65498292e-01 -4.01514143e-01 -2.54977375e-01 2.36264437e-01 7.53423870e-01 1.01638816e-01 2.23121688e-01 4.52310629e-02 -1.01482667e-01 2.22069427e-01 -4.34581429e-01 3.93405527e-01 -6.15567505e-01 -6.59713209e-01 3.58336806e-01 -4.41401273e-01 -4.81040388e-01 -4.19622511e-01 -2.73764163e-01 -1.47062254e+00 -2.36727327e-01 -1.89531520e-01 2.52867103e-01 5.24423957e-01 9.44828570e-01 1.47080377e-01 8.98160875e-01 9.12850618e-01 -8.40514958e-01 -4.03352588e-01 -7.64533222e-01 -3.84240270e-01 5.09915990e-04 3.21983039e-01 -3.93171191e-01 -1.51484767e-02 -2.81777322e-01]
[14.632064819335938, -2.56558895111084]
5c43aab5-2bad-4d54-9b00-683669a3d2fd
cross-node-federated-graph-neural-network-for-1
2106.05223
null
https://arxiv.org/abs/2106.05223v1
https://arxiv.org/pdf/2106.05223v1.pdf
Cross-Node Federated Graph Neural Network for Spatio-Temporal Data Modeling
Vast amount of data generated from networks of sensors, wearables, and the Internet of Things (IoT) devices underscores the need for advanced modeling techniques that leverage the spatio-temporal structure of decentralized data due to the need for edge computation and licensing (data access) issues. While federated learning (FL) has emerged as a framework for model training without requiring direct data sharing and exchange, effectively modeling the complex spatio-temporal dependencies to improve forecasting capabilities still remains an open problem. On the other hand, state-of-the-art spatio-temporal forecasting models assume unfettered access to the data, neglecting constraints on data sharing. To bridge this gap, we propose a federated spatio-temporal model -- Cross-Node Federated Graph Neural Network (CNFGNN) -- which explicitly encodes the underlying graph structure using graph neural network (GNN)-based architecture under the constraint of cross-node federated learning, which requires that data in a network of nodes is generated locally on each node and remains decentralized. CNFGNN operates by disentangling the temporal dynamics modeling on devices and spatial dynamics on the server, utilizing alternating optimization to reduce the communication cost, facilitating computations on the edge devices. Experiments on the traffic flow forecasting task show that CNFGNN achieves the best forecasting performance in both transductive and inductive learning settings with no extra computation cost on edge devices, while incurring modest communication cost.
['Yan Liu', 'Sirisha Rambhatla', 'Chuizheng Meng']
2021-06-09
cross-node-federated-graph-neural-network-for
https://openreview.net/forum?id=HWX5j6Bv_ih
https://openreview.net/pdf?id=HWX5j6Bv_ih
null
['spatio-temporal-forecasting']
['time-series']
[-2.08762720e-01 7.20428899e-02 -6.93976998e-01 -1.73946604e-01 -1.12364709e-01 -4.68939126e-01 4.16912615e-01 -2.41366066e-02 4.23727632e-02 7.15162218e-01 7.14150891e-02 -8.51400435e-01 -6.07436419e-01 -1.08678257e+00 -8.11986387e-01 -6.77975953e-01 -6.85365856e-01 4.88856435e-01 -5.81325665e-02 -1.62846327e-03 -6.17859423e-01 4.49263960e-01 -1.21433163e+00 1.32491663e-01 7.47463346e-01 1.56436038e+00 7.85012320e-02 5.34148753e-01 -2.09007278e-01 8.86986971e-01 4.29836288e-02 -4.82495636e-01 4.45511431e-01 7.95859769e-02 -4.55456853e-01 -1.85195848e-01 -1.71695556e-02 -2.67852098e-01 -7.85068929e-01 6.80787206e-01 3.59551698e-01 -1.45595938e-01 2.41813630e-01 -1.82383776e+00 -7.90346861e-01 6.58103228e-01 -1.78046077e-01 3.10838014e-01 -2.33332470e-01 1.31595731e-01 8.37267101e-01 -3.54331851e-01 5.45208812e-01 7.54364371e-01 8.69421721e-01 3.53748381e-01 -1.10920227e+00 -6.40284121e-01 5.31810582e-01 1.45157889e-01 -1.26727545e+00 -2.49545693e-01 1.02924836e+00 -3.35785031e-01 9.77064073e-01 2.92236537e-01 7.92536020e-01 1.14612186e+00 4.36071187e-01 6.05992138e-01 7.31490314e-01 1.02831960e-01 3.73778552e-01 -1.88002288e-01 -9.77102444e-02 6.50947452e-01 2.81341136e-01 2.70111352e-01 -7.02530622e-01 -3.22877675e-01 5.66888213e-01 6.79479837e-01 1.42996311e-01 -3.01138848e-01 -7.07822680e-01 5.29628038e-01 7.47178435e-01 1.78051189e-01 -8.55622828e-01 3.89619946e-01 5.67131698e-01 2.98861533e-01 1.04950786e+00 -5.94499409e-01 -9.36540782e-01 -2.65414920e-02 -9.11938548e-01 -3.74787420e-01 1.01497579e+00 1.00290060e+00 8.64210188e-01 2.76124418e-01 2.35680006e-02 6.67705238e-02 2.24130198e-01 6.26661956e-01 3.09504308e-02 -6.67897761e-01 7.04903781e-01 1.06641912e+00 -1.75740898e-01 -1.13968527e+00 -6.69496298e-01 -4.91657197e-01 -1.45749450e+00 -4.89810884e-01 2.73372736e-02 -5.49154937e-01 -5.77913284e-01 1.75239193e+00 7.53836334e-01 7.93453574e-01 -2.70262390e-01 6.74824178e-01 4.96001244e-01 6.35264874e-01 2.37654209e-01 -3.46725196e-01 1.00315332e+00 -8.22561920e-01 -5.84095001e-01 9.64270532e-02 1.01604998e+00 1.08746380e-01 5.34855545e-01 -4.93503213e-02 -7.96230972e-01 -1.88850582e-01 -4.79553550e-01 2.08711207e-01 -8.30904722e-01 -7.26330057e-02 1.36748719e+00 6.21621013e-01 -1.22522330e+00 3.68705809e-01 -1.10593736e+00 -2.80656159e-01 7.14797437e-01 7.04145193e-01 -2.43844450e-01 -1.92332923e-01 -1.11436403e+00 2.85630584e-01 1.33022428e-01 2.66580224e-01 -7.46157467e-01 -1.12407291e+00 -6.17825687e-01 2.46442795e-01 3.57163846e-01 -1.07486773e+00 6.50282204e-01 -6.79561019e-01 -1.09039533e+00 2.02737093e-01 1.84515417e-02 -5.98530054e-01 2.58841157e-01 4.02943760e-01 -9.95213926e-01 -7.23649487e-02 -1.45468876e-01 2.54974782e-01 7.11944818e-01 -6.83500707e-01 -6.43807471e-01 -6.33647323e-01 1.16351008e-01 -2.17220977e-01 -8.10989201e-01 -3.86596471e-01 -2.26092696e-01 -6.23388469e-01 -1.55659392e-01 -9.80856061e-01 -2.28063405e-01 1.94972575e-01 -1.67355388e-01 -6.05203271e-01 1.51589620e+00 -6.36228144e-01 1.50157309e+00 -1.93641746e+00 -1.47206664e-01 2.48046815e-01 4.65037256e-01 1.35350809e-01 -1.65337220e-01 6.35124326e-01 2.76757777e-01 -6.94413111e-02 1.47777081e-01 -4.91198570e-01 -4.94138077e-02 5.37101746e-01 -2.55134851e-01 3.56977642e-01 -2.65031755e-01 1.43518066e+00 -9.34986353e-01 -3.48525137e-01 6.88240156e-02 6.06024265e-01 -5.03505111e-01 2.88863736e-03 -4.24621701e-01 5.11890411e-01 -7.95378268e-01 7.56031692e-01 5.76459944e-01 -1.02282059e+00 5.92399478e-01 -3.45069408e-01 1.61173150e-01 2.46654898e-01 -9.99215007e-01 1.72560549e+00 -7.61983991e-01 1.89332515e-01 4.10036772e-01 -1.18868780e+00 5.99902272e-01 4.13297355e-01 1.24958134e+00 -1.01333022e+00 -4.11028601e-02 -5.29279709e-02 -4.10200387e-01 -5.62155843e-01 -1.87301282e-02 3.11196625e-01 -3.45160216e-02 7.36694038e-01 -1.38421535e-01 9.71997738e-01 -1.82534978e-01 4.32829618e-01 1.44489920e+00 -5.13458885e-02 -3.48356634e-01 -2.81253904e-01 8.41902345e-02 -2.44979784e-02 6.56751394e-01 4.50733185e-01 -6.10580258e-02 -3.53200734e-01 4.93658260e-02 -1.04949903e+00 -6.57710493e-01 -1.01023924e+00 3.33268702e-01 1.24855268e+00 8.78377184e-02 -5.49082518e-01 -4.19987440e-01 -9.15650368e-01 3.40081513e-01 2.07651928e-01 -4.33965623e-01 -6.96065426e-02 -4.11599576e-01 -8.66769135e-01 3.43155235e-01 5.92405260e-01 4.12531763e-01 -5.59106112e-01 -4.07078326e-01 2.99051940e-01 -1.31093740e-01 -1.17266321e+00 -5.27596533e-01 8.34409669e-02 -1.03123736e+00 -1.10224509e+00 -5.01971580e-02 -4.32314813e-01 7.79054940e-01 5.08016706e-01 1.19444621e+00 -2.99103614e-02 -2.25673050e-01 7.75697410e-01 5.81206987e-03 -2.00809836e-01 1.22757852e-01 3.47863287e-01 2.48717472e-01 3.64130348e-01 2.79517859e-01 -1.09007013e+00 -8.36333990e-01 3.01970094e-01 -8.85571361e-01 1.30414531e-01 3.46376717e-01 6.08650565e-01 5.24846792e-01 2.83686697e-01 9.08667743e-01 -8.86712849e-01 2.75698483e-01 -1.12605095e+00 -7.40949929e-01 4.96484995e-01 -1.01711166e+00 -1.17025366e-02 9.05977011e-01 -7.38807440e-01 -7.74632037e-01 8.37277845e-02 6.64633870e-01 -7.86605358e-01 2.99279898e-01 7.12195814e-01 -1.13206774e-01 -2.23506778e-01 2.83601910e-01 9.22706947e-02 -1.72943130e-01 -4.80570555e-01 4.15062964e-01 7.19027877e-01 2.16281295e-01 -4.60884809e-01 6.49558544e-01 8.59317958e-01 3.94399822e-01 -5.60242414e-01 -5.82290471e-01 -2.94071138e-01 -1.81197196e-01 -2.93254554e-01 6.32706583e-01 -1.19381177e+00 -1.16988397e+00 1.37754679e-01 -1.01256692e+00 -5.94127715e-01 -4.45735246e-01 4.29835081e-01 -1.56260237e-01 -1.33474767e-01 -5.02154589e-01 -9.79211926e-01 -6.50079310e-01 -4.51711357e-01 9.69380736e-01 -1.29330993e-01 2.22465277e-01 -1.58767760e+00 -1.29768297e-01 3.87759238e-01 8.55325162e-01 3.14181328e-01 9.59660411e-01 -3.94843400e-01 -1.15815032e+00 -4.16323394e-01 -3.47954661e-01 -9.49362814e-02 3.97039860e-01 -5.26514113e-01 -8.07299435e-01 -4.47501481e-01 -2.68605322e-01 1.06726311e-01 3.33276331e-01 5.34058571e-01 1.39994442e+00 -8.88463199e-01 -6.87015653e-01 8.68747354e-01 1.31190705e+00 -2.56706625e-01 9.81690735e-02 -3.17432970e-01 1.17203152e+00 4.87139016e-01 -1.09775729e-01 5.66025794e-01 1.26618040e+00 3.42470050e-01 7.62224555e-01 -1.65182441e-01 -1.59902424e-01 -6.32165551e-01 1.23740077e-01 9.34451640e-01 -4.11663810e-03 -4.24027503e-01 -8.72982681e-01 6.39018178e-01 -2.28382659e+00 -8.50175917e-01 -6.77239075e-02 2.06982136e+00 1.39961556e-01 -1.96638301e-01 4.05733734e-01 -1.06620640e-01 5.67358613e-01 2.96036810e-01 -1.11092877e+00 -2.42892541e-02 -1.58309668e-01 -1.09597549e-01 9.95415211e-01 8.31192285e-02 -7.13741958e-01 5.76398671e-01 5.66222668e+00 6.16067946e-01 -1.39828253e+00 5.89514613e-01 7.54329383e-01 -3.17982823e-01 -4.01889563e-01 1.38469100e-01 -6.06158435e-01 8.49486172e-01 1.53330541e+00 -1.64184391e-01 1.05829108e+00 8.82291079e-01 3.53565395e-01 4.62994426e-01 -9.45857882e-01 1.24974430e+00 -4.28617954e-01 -1.82511294e+00 -1.19946808e-01 5.39925098e-01 9.03199315e-01 8.18519950e-01 -1.75085619e-01 1.51371434e-01 5.02194703e-01 -6.92973018e-01 2.78338373e-01 7.45252550e-01 9.04245317e-01 -3.80538464e-01 2.05345884e-01 7.28947282e-01 -1.69491589e+00 -5.26964366e-01 -7.18256366e-03 -2.41524264e-01 2.79149801e-01 8.85413229e-01 -3.30934465e-01 8.20884407e-01 9.38161254e-01 8.77401292e-01 -2.40627348e-01 5.42770088e-01 4.68517989e-01 7.98952818e-01 -7.55382419e-01 1.16288841e-01 2.55755968e-02 -2.83473551e-01 2.73738652e-01 7.65976548e-01 4.74291533e-01 3.80456299e-02 4.98002529e-01 7.25758851e-01 -5.43947458e-01 -2.45619878e-01 -9.75888669e-01 -1.96282491e-01 6.18119001e-01 1.42100179e+00 -4.40943062e-01 -1.42364755e-01 -8.01599562e-01 6.21803105e-01 3.83736223e-01 6.39585614e-01 -6.23593211e-01 2.68614262e-01 5.54811478e-01 5.12410223e-01 2.54120290e-01 -3.28458518e-01 -2.69520640e-01 -1.13824034e+00 2.23428190e-01 -3.09054166e-01 8.03665519e-01 -5.11210799e-01 -1.69322133e+00 4.48557824e-01 -2.52173632e-01 -1.14847434e+00 -1.97160259e-01 -3.65138054e-01 -6.74033403e-01 6.39720261e-01 -1.60760546e+00 -1.79816079e+00 -2.28296593e-01 1.09139526e+00 -7.73208588e-02 -7.90959504e-03 6.20011210e-01 6.66141331e-01 -6.62364900e-01 3.99746597e-01 2.31216885e-02 -6.32929951e-02 -5.56700751e-02 -6.43277824e-01 4.40708250e-01 5.78196883e-01 1.80040807e-01 2.11663857e-01 -8.32369104e-02 -8.33458722e-01 -2.12778711e+00 -1.59066486e+00 1.11946964e+00 -4.29892570e-01 9.95322227e-01 -5.59463322e-01 -5.82492113e-01 7.76085675e-01 -7.22131431e-02 1.02606571e+00 6.38047874e-01 1.60271585e-01 -3.26355219e-01 -7.74804056e-01 -1.09696686e+00 3.24252725e-01 1.58318520e+00 -8.61413658e-01 5.31687081e-01 7.37380207e-01 8.59909534e-01 -3.41507457e-02 -1.10611522e+00 3.97565007e-01 3.36656243e-01 -4.48138803e-01 7.90556490e-01 -9.59919214e-01 -4.05970782e-01 -2.71652281e-01 -2.69892931e-01 -9.67774570e-01 -2.68179297e-01 -1.14277792e+00 -1.12554514e+00 1.11144698e+00 3.24100554e-01 -1.02530575e+00 1.12197280e+00 1.07487047e+00 5.95585592e-02 -9.33222711e-01 -1.24581230e+00 -1.03112102e+00 -4.21173602e-01 -6.40779316e-01 1.07607853e+00 1.07701850e+00 -7.12466538e-02 3.97367388e-01 -6.20159686e-01 3.86544555e-01 6.47239625e-01 2.06345797e-01 6.82732105e-01 -1.54277670e+00 -2.53466070e-01 9.99848917e-02 -4.34464097e-01 -9.92351472e-01 3.09035212e-01 -1.08772123e+00 -8.20803523e-01 -1.52280092e+00 -2.41451085e-01 -9.80182171e-01 -5.32481790e-01 8.11223686e-01 4.24014926e-01 -5.27789630e-02 9.82255936e-02 3.20379496e-01 -8.84743571e-01 5.78988969e-01 8.60278666e-01 -3.48524511e-01 -2.19527110e-01 1.22912608e-01 -6.04779840e-01 2.53428608e-01 6.33162439e-01 -4.27905649e-01 -8.17460716e-01 -8.80020857e-01 5.78992665e-01 3.61005366e-01 6.39466465e-01 -7.29608476e-01 7.96340466e-01 -2.13336170e-01 2.98259966e-02 -5.58389723e-01 1.49130434e-01 -1.50286925e+00 7.60671198e-01 1.79059178e-01 2.27190226e-01 5.58057845e-01 -3.22909765e-02 1.10780835e+00 2.82073170e-01 8.60799193e-01 -1.56872272e-01 2.64575243e-01 -3.66262406e-01 1.23688817e+00 1.81849256e-01 -6.95298687e-02 1.06300485e+00 -2.79075466e-02 -5.28429270e-01 -4.61857677e-01 -6.76720321e-01 5.42720675e-01 2.12925553e-01 4.85956639e-01 1.09030463e-01 -1.38424063e+00 -2.65818119e-01 3.57390672e-01 -8.87501165e-02 2.92664636e-02 6.04859054e-01 9.62491274e-01 2.78293282e-01 4.45989579e-01 3.81196409e-01 -6.59887552e-01 -5.71746051e-01 8.43884110e-01 2.69125402e-01 -5.56002796e-01 -5.10423720e-01 3.47424567e-01 -1.85791299e-01 -6.16845131e-01 2.17719465e-01 -2.29297355e-01 5.75766504e-01 -1.48387730e-01 9.72068980e-02 7.98309207e-01 3.55020851e-01 -3.51936728e-01 -3.77700925e-01 1.56898499e-01 4.32369351e-01 5.75400352e-01 1.60227990e+00 -3.77094954e-01 -1.52940035e-01 2.84030229e-01 1.21284688e+00 -4.47706968e-01 -1.50816834e+00 -5.05308926e-01 -5.67334332e-02 7.04745203e-02 5.15452445e-01 -8.13863635e-01 -1.66749203e+00 4.97171372e-01 6.04423225e-01 6.97772622e-01 1.35255253e+00 3.13737541e-02 1.24172676e+00 3.35775316e-01 9.03934240e-01 -9.44283783e-01 -4.05676425e-01 1.17970228e-01 9.57314745e-02 -1.11808956e+00 -1.58958212e-01 -3.80866706e-01 -1.66214168e-01 5.97314775e-01 3.40739429e-01 8.25805441e-02 1.55462563e+00 3.62966865e-01 -3.30389202e-01 -4.26853716e-01 -1.13862419e+00 8.61795917e-02 1.87846661e-01 6.36097252e-01 -2.14676827e-01 3.52409631e-01 1.48581535e-01 7.56427944e-01 3.25093716e-01 3.98987293e-01 -2.94229448e-01 8.92143905e-01 3.83077592e-01 -1.10361993e+00 2.70436585e-01 8.04640591e-01 -1.91427886e-01 -4.45817076e-02 1.44789040e-01 4.83220935e-01 2.27243662e-01 9.98277187e-01 2.31480002e-01 -5.21382570e-01 5.39691970e-02 -2.63506547e-02 4.83067036e-02 -1.81260452e-01 -5.52083969e-01 -2.64385223e-01 -1.64650172e-01 -7.81330466e-01 -5.22299945e-01 -2.30490535e-01 -1.05912042e+00 -6.49916172e-01 -2.62050956e-01 1.05443463e-01 9.77911830e-01 1.00652230e+00 1.45251322e+00 4.21235472e-01 9.68318224e-01 -8.45711708e-01 -3.04872572e-01 -5.12680054e-01 -6.64247930e-01 2.00494096e-01 3.55358034e-01 -4.38847780e-01 -3.06149811e-01 -4.63286042e-02]
[6.711524963378906, 2.736300230026245]
b924ec88-3e23-453a-9668-d31b487b3272
a-hungarian-sentiment-corpus-manually
null
null
https://aclanthology.org/L16-1459
https://aclanthology.org/L16-1459.pdf
A Hungarian Sentiment Corpus Manually Annotated at Aspect Level
In this paper we present a Hungarian sentiment corpus manually annotated at aspect level. Our corpus consists of Hungarian opinion texts written about different types of products. The main aim of creating the corpus was to produce an appropriate database providing possibilities for developing text mining software tools. The corpus is a unique Hungarian database: to the best of our knowledge, no digitized Hungarian sentiment corpus that is annotated on the level of fragments and targets has been made so far. In addition, many language elements of the corpus, relevant from the point of view of sentiment analysis, got distinct types of tags in the annotation. In this paper, on the one hand, we present the method of annotation, and we discuss the difficulties concerning text annotation process. On the other hand, we provide some quantitative and qualitative data on the corpus. We conclude with a description of the applicability of the corpus.
["Katalin Ilona Simk{\\'o}", "Martina Katalin Szab{\\'o}", 'Viktor Hangya', 'Veronika Vincze', 'Viktor Varga']
2016-05-01
a-hungarian-sentiment-corpus-manually-1
https://aclanthology.org/L16-1459
https://aclanthology.org/L16-1459.pdf
lrec-2016-5
['text-annotation']
['natural-language-processing']
[-4.69185412e-03 6.76190555e-01 -6.73286989e-02 -6.96161330e-01 -3.13954443e-01 -6.78607702e-01 5.58765829e-01 7.83778727e-01 -4.97992426e-01 6.83442891e-01 3.19186896e-01 -1.78963006e-01 -6.92891181e-02 -9.35870171e-01 -1.28874362e-01 -5.53345323e-01 5.24346173e-01 8.07379544e-01 2.27982864e-01 -6.72739208e-01 4.67256278e-01 1.67114899e-01 -1.71210170e+00 5.40259361e-01 6.34263039e-01 9.63519037e-01 4.23513591e-01 1.86839867e-02 -4.51867580e-01 7.01570332e-01 -6.01572096e-01 -1.12817287e+00 -5.21063916e-02 -5.64058065e-01 -1.07379329e+00 2.73017883e-01 -2.03586817e-01 2.42335990e-01 6.26110792e-01 1.13663542e+00 3.70061606e-01 -1.73715919e-01 7.24231958e-01 -7.62602568e-01 -2.54795432e-01 1.12356460e+00 -6.99050501e-02 -1.92938179e-01 4.38352764e-01 -5.37795961e-01 1.15033066e+00 -7.57936478e-01 1.06749535e+00 7.36384332e-01 2.97880948e-01 2.61883408e-01 -5.17875671e-01 -2.10157007e-01 1.06242366e-01 6.69892356e-02 -8.88950288e-01 -1.55628324e-01 1.11185408e+00 -6.43244565e-01 1.00949228e+00 -2.43482422e-02 9.76463377e-01 9.14873660e-01 3.10698271e-01 6.36187494e-01 1.31923676e+00 -1.09871674e+00 3.31242353e-01 1.08640969e+00 3.25117856e-01 4.11958992e-01 5.42636216e-01 -8.08703482e-01 -3.59729290e-01 -1.37959486e-02 1.48653805e-01 -5.23404896e-01 7.03665763e-02 -4.74739581e-01 -7.27941692e-01 1.00576746e+00 -3.63828778e-01 1.21875715e+00 -1.69921577e-01 -6.26450002e-01 9.77508903e-01 2.66585708e-01 6.17600620e-01 3.99847120e-01 -1.09921980e+00 -2.38414109e-01 -5.17325938e-01 5.07572256e-02 1.24882746e+00 1.11689007e+00 5.82510412e-01 -2.48612553e-01 5.02844572e-01 7.22272873e-01 3.59743029e-01 5.15702903e-01 7.41625845e-01 -1.54904678e-01 4.86154586e-01 9.18909073e-01 -3.73880528e-02 -1.23811901e+00 -2.89956450e-01 -1.28497079e-01 -1.91287130e-01 3.92168500e-02 3.63017648e-01 -4.30741429e-01 -3.99063379e-01 1.12582469e+00 3.03173214e-01 -1.15935540e+00 5.38352191e-01 6.12850368e-01 1.15578592e+00 3.83012682e-01 7.75645971e-02 -5.97041845e-01 1.73329341e+00 -6.40001297e-01 -1.34745693e+00 2.34974667e-01 9.62007523e-01 -1.35106099e+00 1.12268639e+00 8.11416864e-01 -9.70700204e-01 -1.98159605e-01 -7.96327591e-01 4.18485552e-02 -1.20158815e+00 2.97979861e-01 7.97185361e-01 1.05822480e+00 -6.76509857e-01 2.17487574e-01 -4.80461091e-01 -8.73137057e-01 -1.67079985e-01 1.40006721e-01 -2.84246713e-01 5.68875849e-01 -1.23852932e+00 9.72761154e-01 4.76699263e-01 -1.09384485e-01 2.47322127e-01 1.27146039e-02 -8.60503793e-01 -1.98663279e-01 3.99691194e-01 -1.40797034e-01 1.38826442e+00 -1.68067372e+00 -1.54424810e+00 1.40320230e+00 -9.28158462e-02 -2.79386312e-01 4.12354112e-01 6.22635661e-03 -8.31110537e-01 1.29919827e-01 3.90632719e-01 -5.06438352e-02 3.51641655e-01 -1.02178538e+00 -8.91915381e-01 -5.94418824e-01 -1.85255826e-01 1.33865569e-02 -5.67732036e-01 3.71236622e-01 -2.95233727e-01 -6.84291482e-01 -1.35531723e-01 -4.73135203e-01 -1.90008983e-01 -8.66175711e-01 1.83357432e-01 -3.81850600e-01 6.40050292e-01 -4.23894405e-01 1.36712372e+00 -2.13927412e+00 -1.75703093e-01 4.06858176e-01 -1.00947015e-01 -1.95627678e-02 3.94114316e-01 9.12561178e-01 -5.33209071e-02 3.30518931e-01 8.98239166e-02 7.04432884e-03 2.33986959e-01 4.05631751e-01 -2.56331176e-01 3.87549400e-01 -7.83760920e-02 3.50648075e-01 -9.25457716e-01 -1.00397122e+00 2.45684236e-01 2.13581502e-01 -1.74245194e-01 -3.41755241e-01 -1.56460002e-01 -1.52811274e-01 -1.00161433e+00 6.60593390e-01 4.29493904e-01 3.60305250e-01 4.05014306e-01 -2.68809497e-01 -5.95315814e-01 3.56872559e-01 -9.91195619e-01 1.41083956e+00 -4.26169842e-01 5.84358454e-01 -1.62445739e-01 -1.06554425e+00 1.29499292e+00 5.98680913e-01 6.30549788e-01 -7.54886031e-01 7.74117827e-01 7.76131034e-01 -1.37075007e-01 -4.05885816e-01 1.00411618e+00 -3.08814436e-01 -4.65510547e-01 1.83813557e-01 3.05402339e-01 -5.72336614e-01 6.80905759e-01 -2.39247411e-01 4.19942677e-01 1.38765112e-01 8.33534598e-01 -5.05621731e-01 8.94584298e-01 6.49865568e-01 3.34977269e-01 9.80060454e-03 -1.75171807e-01 3.79326731e-01 8.22543859e-01 -8.74193728e-01 -1.14982736e+00 -2.17650235e-01 -6.50575101e-01 7.69228160e-01 -3.17529172e-01 -8.06226671e-01 -1.10489035e+00 -7.39155948e-01 -6.19905233e-01 6.15579903e-01 -5.98133981e-01 6.51842296e-01 -3.01317662e-01 -7.30527341e-01 -6.40003709e-03 -1.89108595e-01 3.54567349e-01 -1.40413857e+00 -9.30208802e-01 2.54726678e-01 -2.79499799e-01 -1.05311692e+00 2.98810989e-01 6.30017459e-01 -7.90856421e-01 -1.04874277e+00 -9.81572643e-02 -1.09162271e+00 5.08186996e-01 -2.26599351e-01 1.31224108e+00 -2.80830979e-01 1.02726214e-01 8.90222341e-02 -1.22446275e+00 -1.15131557e+00 -5.68991125e-01 5.42601824e-01 -3.39957535e-01 -3.69299918e-01 1.00327981e+00 -1.11787327e-01 7.36806169e-02 2.92286843e-01 -1.01113951e+00 -4.47923630e-01 5.39865494e-01 4.38774794e-01 5.94360173e-01 4.59991276e-01 5.00762224e-01 -1.47414386e+00 6.44551873e-01 -1.43706083e-01 -6.66465044e-01 -6.34796619e-02 -5.89385569e-01 -2.94838905e-01 6.70299113e-01 1.20620579e-01 -1.26071382e+00 8.82519633e-02 -4.97072905e-01 6.29584789e-01 -4.95938659e-01 5.53368986e-01 -3.88186812e-01 1.93010107e-01 7.80802488e-01 -2.82278389e-01 -2.26217121e-01 -5.12664795e-01 1.58705339e-01 9.63824689e-01 -6.89764023e-02 -4.10336047e-01 2.02808037e-01 1.90091148e-01 -1.60774678e-01 -1.04198563e+00 -8.99510384e-01 -7.86446333e-01 -8.92421484e-01 -2.98548490e-01 9.72781122e-01 -6.03763342e-01 -1.96188882e-01 1.57648906e-01 -9.06432390e-01 -1.57100752e-01 -6.94440603e-01 3.58557254e-01 -8.21736753e-01 2.16829643e-01 -3.28417659e-01 -1.02044976e+00 -3.71893704e-01 -8.35228980e-01 8.24500680e-01 -2.38352016e-01 -6.23246729e-01 -1.16816783e+00 2.28318438e-01 2.93963909e-01 -1.92535058e-01 3.52446377e-01 7.35693216e-01 -7.83911645e-01 4.36344743e-01 -4.99039739e-01 4.16297317e-01 2.28532225e-01 1.80471703e-01 2.66433120e-01 -9.19171393e-01 2.05920652e-01 5.21705210e-01 -2.96312541e-01 4.47945952e-01 2.16089979e-01 5.46660960e-01 -2.03166381e-01 -1.57176197e-01 -2.05173835e-01 1.71829915e+00 5.41109860e-01 6.74923122e-01 8.25422287e-01 3.01494487e-02 1.18179023e+00 1.28133929e+00 5.61688423e-01 4.19998728e-02 6.09989345e-01 2.07199916e-01 6.79580644e-02 2.22664878e-01 -2.97626313e-02 3.40890527e-01 9.85433578e-01 -1.11192748e-01 -3.70553792e-01 -8.85744691e-01 7.70200372e-01 -1.81751251e+00 -8.74855042e-01 -6.25549734e-01 1.72845662e+00 5.19198120e-01 3.71649861e-01 9.37416479e-02 5.34836709e-01 4.63593841e-01 -1.30903013e-02 4.88254279e-01 -9.59508657e-01 -3.83433342e-01 3.35949630e-01 3.44283521e-01 5.37298799e-01 -1.19454122e+00 1.17765570e+00 6.02247477e+00 7.17660129e-01 -7.23778903e-01 1.91847414e-01 3.43046367e-01 2.00391769e-01 -2.65134454e-01 -1.27654657e-01 -7.78297007e-01 2.26616770e-01 8.20191562e-01 1.09806426e-01 -3.64483356e-01 1.32203686e+00 2.03616872e-01 -3.29292208e-01 -5.26139915e-01 6.08495414e-01 3.26348871e-01 -8.33455026e-01 5.09483702e-02 1.89222187e-01 7.06005633e-01 -1.78091809e-01 -2.04346895e-01 -5.28923422e-03 9.37820897e-02 -4.49558944e-01 1.25680137e+00 2.39555314e-02 2.58084893e-01 -1.06870675e+00 1.77298427e+00 1.40481154e-02 -8.91590655e-01 1.97920337e-01 -4.75963712e-01 -3.63129497e-01 2.71422178e-01 8.48876953e-01 -5.77646494e-01 8.12034428e-01 9.68142867e-01 6.10322714e-01 -4.91226166e-01 4.09214884e-01 -1.85066521e-01 4.48101312e-01 -1.19086392e-01 -7.82314897e-01 4.70446110e-01 -6.79539979e-01 2.35913321e-01 1.42806017e+00 2.91456074e-01 5.77557422e-02 3.32740285e-02 2.79006839e-01 4.32771176e-01 1.29986560e+00 -9.38480377e-01 -1.71338126e-01 -6.56702369e-02 1.57389688e+00 -1.29035366e+00 -2.16605410e-01 -6.91399634e-01 5.72538555e-01 -7.05451742e-02 -1.03698142e-01 -1.92647323e-01 -6.29912913e-01 2.30628550e-01 1.79717883e-01 2.75148451e-01 2.29628429e-01 -5.18986642e-01 -1.14248896e+00 7.81210735e-02 -7.02455342e-01 4.20771778e-01 -5.39808333e-01 -1.22658527e+00 9.98924255e-01 -5.69633245e-02 -1.00676596e+00 -3.15864235e-01 -9.59760725e-01 3.53464223e-02 6.85001552e-01 -9.18152750e-01 -1.06713557e+00 -3.39162047e-03 2.96049893e-01 5.02803862e-01 -3.22823256e-01 1.15881264e+00 1.36012426e-02 -1.77522480e-01 -6.16631769e-02 -1.72693238e-01 1.26427233e-01 7.97092438e-01 -1.30961323e+00 -2.21528396e-01 4.50985610e-01 -5.09735793e-02 2.35750571e-01 1.16250610e+00 -6.31871939e-01 -7.82502234e-01 -4.30139899e-01 1.85746527e+00 -5.85373700e-01 1.03638756e+00 -2.98746288e-01 -3.99314612e-01 3.76429856e-01 6.90930009e-01 -7.04025805e-01 1.07613146e+00 2.00549215e-01 9.72188041e-02 -7.69601613e-02 -1.22489917e+00 3.60095948e-01 3.52468371e-01 -2.64287114e-01 -9.25897598e-01 3.18885207e-01 3.46127063e-01 -8.52591470e-02 -9.26104009e-01 1.50943100e-01 4.74676639e-01 -1.04004872e+00 1.66337162e-01 -1.49919853e-01 5.60030580e-01 -6.42083362e-02 -2.15924397e-01 -1.01326668e+00 6.76617920e-02 -3.49500120e-01 8.35684538e-01 1.61726761e+00 8.28304231e-01 -4.98555452e-01 8.23908031e-01 3.02519202e-01 -3.47931594e-01 -4.99037832e-01 -5.42893648e-01 -1.17007807e-01 2.09071591e-01 -7.33858824e-01 4.27262902e-01 1.12168050e+00 8.67137909e-01 5.30199826e-01 1.56902730e-01 -5.05369544e-01 -4.33913134e-02 1.85106665e-01 5.93795776e-01 -1.41411722e+00 4.97937538e-02 -6.17800355e-01 -4.50429231e-01 1.65101290e-02 1.42853796e-01 -5.01155496e-01 -7.72142857e-02 -1.61453009e+00 4.72316593e-02 -1.47149548e-01 3.49917471e-01 2.60769188e-01 4.78020072e-01 1.49612203e-01 -7.01878816e-02 -8.96282941e-02 -2.91461736e-01 -1.11032911e-02 1.03839958e+00 1.93476439e-01 -3.08328897e-01 -1.74363241e-01 -8.13456476e-01 1.01951706e+00 1.30337989e+00 -3.05558741e-01 -4.39988405e-01 1.38285483e-04 1.00626409e+00 -6.28980219e-01 -5.00391424e-01 -4.62148666e-01 -2.63622850e-01 -1.00387886e-01 2.09522337e-01 -7.61120021e-01 -3.51289175e-02 -1.42646098e+00 1.60506263e-01 3.29685241e-01 -2.96547543e-03 1.72961697e-01 5.00913821e-02 1.89052392e-02 -6.02635086e-01 -8.59458387e-01 5.17210245e-01 -4.83411014e-01 -7.26587713e-01 -3.84061128e-01 -9.56377923e-01 -1.13302700e-01 1.09393001e+00 -2.62237370e-01 3.00157126e-02 -1.56512797e-01 -6.60423219e-01 -1.50636911e-01 8.63343835e-01 1.15529083e-01 2.85146460e-02 -9.63696480e-01 -1.34674355e-01 -2.25933827e-02 5.02205431e-01 -2.86689669e-01 -4.27820645e-02 4.98929024e-01 -8.39756727e-01 7.06558764e-01 -4.71197039e-01 3.24648619e-02 -1.24015152e+00 8.57141078e-01 -2.48504169e-02 -4.14838314e-01 -3.63611668e-01 -1.25724122e-01 -1.02605253e-01 -2.63687491e-01 -1.54758677e-01 -4.16961402e-01 -1.15062261e+00 8.41289520e-01 3.86113197e-01 1.07241511e-01 5.74151337e-01 -1.14468896e+00 -1.61241293e-01 6.99182570e-01 2.24857941e-01 -4.04891968e-01 1.32583463e+00 -3.63221675e-01 -6.26221180e-01 8.69426787e-01 8.44727755e-01 4.26669627e-01 1.74316112e-02 3.26702595e-01 1.24782018e-01 -2.33157650e-01 -4.30618227e-03 -7.86508918e-01 -8.70634079e-01 4.44922954e-01 2.44241774e-01 8.04022014e-01 1.28705931e+00 1.14089422e-01 2.71597683e-01 5.03545702e-01 5.59398234e-01 -1.69018161e+00 -3.75429064e-01 5.52874148e-01 6.92164421e-01 -1.03716767e+00 -2.78867453e-01 -1.00487244e+00 -8.89577806e-01 1.35079920e+00 2.92072207e-01 1.28508553e-01 8.59470606e-01 5.26935041e-01 5.00399172e-01 -5.99555612e-01 -3.84296238e-01 -5.91172040e-01 1.87241361e-02 7.09423482e-01 9.93852615e-01 2.58791670e-02 -1.54371464e+00 8.52199018e-01 -8.73049200e-01 2.92461030e-02 7.79693425e-01 1.02820551e+00 -6.11912906e-01 -1.56085587e+00 -4.10619140e-01 8.48313347e-02 -1.04283822e+00 8.54461342e-02 -1.01053798e+00 1.21697593e+00 4.37970668e-01 1.28669930e+00 -1.53465003e-01 -8.22939649e-02 6.22283220e-01 1.44014776e-01 3.32456291e-01 -6.72089219e-01 -8.95622253e-01 5.64325213e-01 6.80824518e-01 -9.05031040e-02 -1.07354712e+00 -9.39105988e-01 -9.29972470e-01 -3.19751143e-01 -3.71645480e-01 9.35832620e-01 9.38798010e-01 8.81510317e-01 -4.78651792e-01 3.09896737e-01 2.65044898e-01 -5.58075726e-01 1.05446249e-01 -1.09109986e+00 -1.08650529e+00 3.44442755e-01 -2.03843459e-01 -4.58422661e-01 -2.35521078e-01 3.56001407e-01]
[11.04214859008789, 6.985107421875]
38c7dc59-37aa-4fc0-8f5d-5d969fe8f621
federated-semi-supervised-classification-of
2205.00550
null
https://arxiv.org/abs/2205.00550v1
https://arxiv.org/pdf/2205.00550v1.pdf
Federated Semi-Supervised Classification of Multimedia Flows for 3D Networks
Automatic traffic classification is increasingly becoming important in traffic engineering, as the current trend of encrypting transport information (e.g., behind HTTP-encrypted tunnels) prevents intermediate nodes from accessing end-to-end packet headers. However, this information is crucial for traffic shaping, network slicing, and Quality of Service (QoS) management, for preventing network intrusion, and for anomaly detection. 3D networks offer multiple routes that can guarantee different levels of QoS. Therefore, service classification and separation are essential to guarantee the required QoS level to each traffic sub-flow through the appropriate network trunk. In this paper, a federated feature selection and feature reduction learning scheme is proposed to classify network traffic in a semi-supervised cooperative manner. The federated gateways of 3D network help to enhance the global knowledge of network traffic to improve the accuracy of anomaly and intrusion detection and service identification of a new traffic flow.
['Alberto Gotta', 'Pietro Cassarà', 'Lorenzo Valerio', 'Achilles Machumilane', 'Saira Bano']
2022-05-01
null
null
null
null
['traffic-classification']
['miscellaneous']
[-5.42191230e-02 -4.50776905e-01 -6.69095159e-01 -5.46448529e-01 -4.38764542e-02 -8.15404654e-01 3.75083014e-02 1.15033768e-01 -1.04698636e-01 6.50101960e-01 -4.37056929e-01 -1.04474974e+00 -3.96910280e-01 -1.04495156e+00 2.98850775e-01 -7.21009612e-01 -3.95915955e-01 3.64446342e-01 5.38313150e-01 1.98841821e-02 5.04617572e-01 1.30848956e+00 -1.47368169e+00 1.21929772e-01 4.57757264e-01 1.70030463e+00 -2.44954348e-01 3.45370889e-01 -7.99672663e-01 2.29954630e-01 -6.45935893e-01 -3.60466450e-01 6.38175607e-01 1.61383018e-01 -7.49740541e-01 1.78686127e-01 -1.74670637e-01 -2.46855155e-01 -5.76318502e-01 9.85026836e-01 1.05664041e-03 -6.80220276e-02 4.52394068e-01 -1.93809688e+00 -4.12027538e-02 6.59127012e-02 -2.65637755e-01 5.80809057e-01 -8.66710097e-02 3.25966775e-01 9.64994192e-01 -2.26863861e-01 2.88024485e-01 9.21772480e-01 -6.27712980e-02 4.23339844e-01 -1.05105281e+00 -7.30165064e-01 3.50736916e-01 4.46356177e-01 -1.07691550e+00 -6.16210401e-01 8.08537066e-01 -1.58231452e-01 5.14116228e-01 6.59473777e-01 3.66711110e-01 6.03260636e-01 1.01195037e-01 4.35085624e-01 3.29771191e-01 6.70922473e-02 3.41034114e-01 4.32939172e-01 1.16333768e-01 4.25988108e-01 2.88809448e-01 -1.70215685e-02 3.95928249e-02 -3.61656487e-01 4.17413861e-01 -2.04662681e-02 -2.14076087e-01 -4.25032258e-01 -5.82910180e-01 5.38905919e-01 1.61456198e-01 3.09137315e-01 -4.69017684e-01 -4.66539115e-01 7.45758533e-01 8.22831988e-01 3.09224457e-01 9.98803601e-02 -6.89008474e-01 -2.67096311e-01 -5.40022314e-01 -1.93527475e-01 8.31055999e-01 9.48152125e-01 8.80511880e-01 1.05218418e-01 2.72137582e-01 5.35973430e-01 1.56761542e-01 5.83885670e-01 -9.21535492e-02 -1.34361494e+00 2.34942451e-01 9.10622358e-01 -2.55057484e-01 -1.07344222e+00 -3.45673263e-01 -4.76471573e-01 -9.47328329e-01 5.10581613e-01 4.77837443e-01 1.77890107e-01 -3.30368549e-01 1.47702682e+00 5.85400879e-01 2.22544685e-01 -8.86757970e-02 8.07088435e-01 -1.12239895e-02 4.45062816e-01 2.00074553e-01 -2.88689673e-01 1.19677830e+00 -4.31250095e-01 -5.25214553e-01 1.96119651e-01 6.51722968e-01 -8.01280499e-01 4.27773356e-01 5.74149378e-02 -8.17183733e-01 -2.84202814e-01 -4.80321199e-01 5.22871077e-01 -5.26692092e-01 -4.75867659e-01 5.22816479e-01 9.99047041e-01 -4.77268845e-01 1.96790099e-01 -3.05550367e-01 -3.25258762e-01 6.23242021e-01 5.85222244e-01 -4.87569749e-01 -6.71144202e-02 -1.24638677e+00 4.02157456e-01 3.12191844e-01 -1.57987088e-01 -4.81917918e-01 -8.60013843e-01 -6.17078245e-01 9.76860002e-02 4.92681891e-01 -1.56747654e-01 8.63843679e-01 -3.53297830e-01 -1.02427208e+00 5.12307048e-01 -1.01383766e-02 -4.14124817e-01 1.56693637e-01 3.69242013e-01 -1.16940701e+00 4.43407863e-01 2.06245705e-01 -2.08412912e-02 9.07778084e-01 -1.00938892e+00 -1.32543337e+00 -4.09166634e-01 1.19587079e-01 -4.58427519e-01 -5.58817327e-01 2.19309822e-01 -2.18208253e-01 -1.83572933e-01 3.05363059e-01 -5.34565985e-01 -1.35367185e-01 3.21473658e-01 -4.78341013e-01 -2.73590863e-01 1.68001854e+00 -2.45082378e-01 1.28868866e+00 -2.10212374e+00 -4.67311502e-01 1.03836131e+00 5.28923750e-01 5.53948939e-01 -1.20753450e-02 1.22370444e-01 1.43823319e-03 3.04018915e-01 9.98091474e-02 1.60868198e-01 -1.40034601e-01 2.30070546e-01 -2.75736958e-01 3.46325099e-01 7.13810995e-02 1.12096928e-01 -7.05780745e-01 -5.25418997e-01 4.63727355e-01 1.20956535e-02 -5.08489132e-01 7.80993178e-02 1.98930465e-02 7.05366552e-01 -8.75059247e-01 8.54724109e-01 9.45586324e-01 5.46245873e-02 1.65086329e-01 -4.20237333e-01 -4.28377748e-01 4.30481941e-01 -1.27583456e+00 9.33676124e-01 -6.34407759e-01 3.57706815e-01 3.67469966e-01 -1.22859883e+00 1.02493358e+00 4.87899870e-01 1.05785871e+00 -7.31192946e-01 3.43575180e-01 4.31005299e-01 -2.85274796e-02 -2.55336016e-01 2.74904594e-02 3.64578336e-01 -8.09438229e-02 7.05470979e-01 -3.12167495e-01 7.55797327e-01 4.21125323e-01 8.54206607e-02 1.24069512e+00 -6.09151959e-01 -1.50443047e-01 1.00897253e-01 1.42497301e+00 -3.25901568e-01 8.01849186e-01 4.66327846e-01 -6.42928720e-01 -2.51896530e-01 6.89908266e-01 -5.50861061e-01 -8.43775749e-01 -1.28216350e+00 -2.23582953e-01 6.13600969e-01 7.92562217e-02 -4.80137430e-02 -3.83902848e-01 -1.00440025e+00 2.77668446e-01 7.08563030e-01 2.31301561e-01 -7.67442286e-01 -5.57612360e-01 -5.68364784e-02 6.05287135e-01 -8.04106891e-02 6.13484621e-01 -4.32627738e-01 -7.21644163e-02 5.05701542e-01 -1.01610050e-01 -1.66518867e+00 -3.64738256e-01 8.11419040e-02 -6.53207600e-01 -1.33396614e+00 1.99164316e-01 -5.48624873e-01 4.88352448e-01 6.25029325e-01 6.75516248e-01 2.75419146e-01 -2.40538314e-01 1.13002457e-01 -1.39991894e-01 2.64192671e-01 -7.90513754e-01 4.13639724e-01 3.70426804e-01 6.99800909e-01 6.94101214e-01 -9.37470794e-01 -3.18467677e-01 8.15279543e-01 -6.19585156e-01 -5.79364419e-01 3.81991804e-01 1.05342969e-01 2.33031452e-01 6.75388277e-01 6.21783733e-01 -5.34906209e-01 3.75644237e-01 -7.19141543e-01 -7.39369869e-01 5.06884269e-02 -5.90826571e-01 -1.64228991e-01 9.82438028e-01 2.43637636e-02 -8.73392403e-01 -7.19298661e-01 -1.43024981e-01 -4.67176616e-01 -6.19150460e-01 -1.15107708e-01 -7.56657481e-01 -3.33207279e-01 2.06556052e-01 1.31665021e-01 2.26947471e-01 -7.25549996e-01 3.98278758e-02 1.13987827e+00 6.85210079e-02 -4.91777450e-01 1.33748436e+00 5.92994928e-01 6.34409368e-01 -9.47019815e-01 -4.74337518e-01 -5.73915541e-01 -5.34066498e-01 -4.74451959e-01 4.85206783e-01 -3.05171758e-01 -1.29668593e+00 2.32260257e-01 -9.52966988e-01 6.83975220e-01 -2.06803605e-01 4.53225076e-01 -1.99522272e-01 4.16174740e-01 -3.70979875e-01 -8.04258943e-01 6.71099871e-02 -1.01151395e+00 4.42130029e-01 -2.19358946e-03 2.18093581e-02 -7.86780715e-01 -4.04859483e-01 4.20187712e-01 6.03604555e-01 -1.59690753e-01 1.43825638e+00 -9.16749835e-01 -7.33094573e-01 -4.06985074e-01 -6.19945467e-01 6.31148279e-01 7.44760156e-01 2.60476977e-01 -6.82411194e-01 -1.13396063e-01 -4.28473912e-02 1.48169279e-01 3.25744450e-01 -6.46181852e-02 1.42021060e+00 -1.32486477e-01 -1.52109206e-01 7.42468238e-01 9.64241445e-01 3.72800678e-01 4.14623797e-01 1.52548462e-01 3.61013919e-01 9.00736094e-01 5.86907208e-01 4.43419576e-01 7.60384724e-02 4.11063582e-01 7.37853110e-01 2.95633346e-01 7.50250444e-02 1.06367253e-01 1.58318818e-01 2.72383928e-01 1.86982766e-01 -4.37870026e-01 -5.18896699e-01 9.72782671e-02 -1.31382632e+00 -1.21684396e+00 -1.46527275e-01 2.16715765e+00 2.17154622e-02 5.83788037e-01 2.66503930e-01 8.41430902e-01 1.04391873e+00 4.88390550e-02 -5.07471621e-01 -6.04244113e-01 8.49063396e-02 6.64208503e-03 5.20884693e-01 3.06412280e-01 -7.40629554e-01 7.13404357e-01 4.68509579e+00 1.02349854e+00 -1.14970648e+00 -2.07195133e-01 3.53609771e-01 3.93893540e-01 -1.06677637e-01 8.13300759e-02 -6.98615789e-01 6.57661021e-01 9.87238407e-01 -2.40491927e-01 4.40107852e-01 7.91113436e-01 6.93672597e-01 1.89551637e-01 -5.97658157e-01 6.21332705e-01 -8.72833848e-01 -1.31215322e+00 4.74963278e-01 3.89287055e-01 3.47251445e-02 -3.43092144e-01 -7.15560839e-02 -1.34740444e-02 -2.84862846e-01 -2.13869274e-01 1.37810662e-01 4.18706909e-02 6.29335642e-01 -1.13153648e+00 4.78556484e-01 9.61849019e-02 -1.22804606e+00 -5.36029994e-01 9.97700021e-02 2.66630828e-01 4.41301167e-01 7.78497338e-01 -3.68531138e-01 6.18216753e-01 4.58123088e-01 4.01773363e-01 -1.56256765e-01 9.93814588e-01 1.58892885e-01 5.29772460e-01 -2.48467743e-01 3.07588816e-01 1.86153665e-01 -3.80874723e-01 1.16803336e+00 5.43361723e-01 1.40622223e-03 -1.99798435e-01 3.80883008e-01 5.54130435e-01 -2.36285388e-01 5.28019108e-02 -3.75535905e-01 -3.93983573e-02 8.30244720e-01 1.20648527e+00 -6.29626453e-01 1.56095430e-01 -7.21917331e-01 6.66318357e-01 -2.77564883e-01 4.21800584e-01 -7.55535305e-01 -9.07140434e-01 1.94903433e+00 5.08069038e-01 2.36549497e-01 -5.69310844e-01 -4.38975655e-02 -8.09230208e-01 -1.42126590e-01 -7.20603347e-01 5.09366751e-01 4.82443087e-02 -1.35086381e+00 3.53874862e-01 -4.20323521e-01 -1.48780024e+00 2.98067421e-01 -6.22152269e-01 -7.13895559e-01 5.00261307e-01 -1.90599799e+00 -7.01962233e-01 -1.12893887e-01 1.09141195e+00 1.00518972e-01 -6.19564116e-01 8.45288336e-01 9.85942066e-01 -6.54807150e-01 5.84118724e-01 -1.01122722e-01 2.02839151e-01 3.79661888e-01 -4.27979469e-01 1.46152064e-01 7.84767389e-01 -2.66011298e-01 2.67932773e-01 2.69874483e-01 -5.17514110e-01 -1.31483746e+00 -1.07442570e+00 8.04127336e-01 2.77258277e-01 6.37328029e-01 -3.45902629e-02 -8.03457260e-01 1.64662942e-01 -5.59304297e-01 1.40563354e-01 1.12283683e+00 -1.95312053e-01 -3.72903973e-01 -7.72637963e-01 -1.67829227e+00 6.12577975e-01 8.75866294e-01 -4.40554827e-01 1.93732962e-01 7.81471431e-02 6.71855748e-01 2.91636825e-01 -7.58154690e-01 1.50425613e-01 3.99539262e-01 -8.42602313e-01 8.98180068e-01 -1.18653357e+00 -5.87140083e-01 -6.27724111e-01 -5.31898320e-01 -6.24768078e-01 -3.25465381e-01 -7.35013306e-01 -3.81610900e-01 1.41948628e+00 1.52617201e-01 -9.23111618e-01 1.05680108e+00 3.88817072e-01 2.34759804e-02 -4.21414107e-01 -1.37116051e+00 -9.87939954e-01 -5.52136958e-01 -1.04322672e+00 1.10509694e+00 6.36263430e-01 -5.04483402e-01 7.15797395e-02 -1.41004458e-01 6.01347387e-01 1.22521603e+00 1.63753163e-02 8.27200294e-01 -1.58822596e+00 5.17863274e-01 -8.35866272e-01 -9.75841641e-01 -9.57087219e-01 5.50316095e-01 -9.84260321e-01 -9.03396606e-01 -6.86963499e-01 -7.62345612e-01 -9.22458470e-01 -5.70639908e-01 5.54349199e-02 7.23127306e-01 1.22574858e-01 -1.32159710e-01 1.28517419e-01 -2.84771532e-01 3.81053299e-01 1.12885702e+00 1.70704573e-02 -1.27630770e-01 9.94630933e-01 -4.92025554e-01 5.26746213e-01 1.25912559e+00 -5.39393902e-01 -5.75629771e-01 -2.89773792e-02 -3.50785404e-01 1.58488542e-01 5.09356678e-01 -8.33639324e-01 1.84472561e-01 -5.67279041e-01 -2.73689684e-02 -5.85332453e-01 1.19454212e-01 -1.70473886e+00 -2.99495667e-01 3.98312241e-01 -6.07239082e-02 -3.28520238e-01 -2.42430776e-01 6.12510324e-01 -1.97767913e-01 5.86120486e-02 9.68537211e-01 1.20242752e-01 -7.29314506e-01 9.83926892e-01 -1.03379858e+00 4.84123379e-02 1.21609831e+00 -4.05641526e-01 -1.21715792e-01 -3.07469398e-01 -7.18191743e-01 4.86490488e-01 3.29481989e-01 6.42104745e-01 7.02490091e-01 -1.13542438e+00 -3.25679541e-01 8.17875385e-01 1.67167217e-01 -6.49146557e-01 1.88410148e-01 7.01611161e-01 -2.70064443e-01 5.00265121e-01 -3.63003403e-01 -4.41459477e-01 -1.20118725e+00 4.46549356e-01 2.99277067e-01 -6.22630678e-02 -8.23847651e-02 1.09671533e-01 -5.29349566e-01 -3.30111384e-01 4.91695374e-01 3.49949092e-01 4.97565828e-02 -1.79632276e-01 5.68084359e-01 1.03480673e+00 2.32859090e-01 -5.51782846e-01 -5.06876528e-01 2.45172292e-01 -1.90607011e-01 3.90360594e-01 8.85585964e-01 -6.57073617e-01 -4.01561648e-01 -2.30346352e-01 1.39636099e+00 -2.55230013e-02 -4.34184074e-01 -3.38658035e-01 1.65544659e-01 -1.02816367e+00 4.01288599e-01 -2.69638062e-01 -1.56712127e+00 8.48414898e-01 4.48975921e-01 7.49687254e-01 1.23373973e+00 -3.45547497e-01 1.30815005e+00 4.31757689e-01 5.79660833e-01 -8.71946573e-01 -3.29201192e-01 5.43016493e-01 -2.67851830e-01 -8.03461552e-01 -4.93423760e-01 -6.83646262e-01 -1.19318724e-01 1.44738376e+00 8.12216818e-01 4.74920005e-01 1.15941179e+00 1.51065230e-01 -2.51619052e-02 -1.86587542e-01 -6.34903848e-01 -2.33917072e-01 -1.91237971e-01 8.10415924e-01 -2.37603709e-01 -2.75730163e-01 1.92796335e-01 6.72344267e-02 4.12260473e-01 -5.24359941e-01 2.28973836e-01 7.84805894e-01 -8.26841116e-01 -1.63158250e+00 -1.27626076e-01 5.57660639e-01 -3.83551240e-01 2.43548110e-01 1.58052772e-01 5.28389633e-01 9.93968919e-02 1.25306189e+00 2.13120759e-01 -6.22806191e-01 3.40631634e-01 2.41145998e-01 -1.23155534e-01 -1.77648276e-01 -6.17937483e-02 -5.29854834e-01 1.60042122e-01 -7.15088069e-01 2.85488758e-02 -3.01737189e-01 -1.31584501e+00 -1.03554022e+00 -2.04601079e-01 7.02885628e-01 7.86158741e-01 7.61068940e-01 4.42599386e-01 3.17494929e-01 1.79866457e+00 -1.59177929e-01 -3.18493098e-01 1.05225325e-01 -8.39429319e-01 2.56752789e-01 5.19442022e-01 -7.59203196e-01 -7.20234215e-01 -4.78738636e-01]
[5.070310115814209, 7.231068134307861]
87667442-83c1-4c51-a626-af48b682d07b
zju-reler-submission-for-epic-kitchen
2307.02010
null
https://arxiv.org/abs/2307.02010v2
https://arxiv.org/pdf/2307.02010v2.pdf
ZJU ReLER Submission for EPIC-KITCHEN Challenge 2023: Semi-Supervised Video Object Segmentation
The Associating Objects with Transformers (AOT) framework has exhibited exceptional performance in a wide range of complex scenarios for video object segmentation. In this study, we introduce MSDeAOT, a variant of the AOT series that incorporates transformers at multiple feature scales. Leveraging the hierarchical Gated Propagation Module (GPM), MSDeAOT efficiently propagates object masks from previous frames to the current frame using a feature scale with a stride of 16. Additionally, we employ GPM in a more refined feature scale with a stride of 8, leading to improved accuracy in detecting and tracking small objects. Through the implementation of test-time augmentations and model ensemble techniques, we achieve the top-ranking position in the EPIC-KITCHEN VISOR Semi-supervised Video Object Segmentation Challenge.
['Yueting Zhuang', 'Yi Yang', 'Zongxin Yang', 'Yuanyou Xu', 'Jiahao Li']
2023-07-05
null
null
null
null
['semi-supervised-video-object-segmentation', 'video-object-segmentation', 'video-semantic-segmentation']
['computer-vision', 'computer-vision', 'computer-vision']
[ 4.77737755e-01 -2.97469735e-01 -1.15515016e-01 -1.64521426e-01 -6.53651893e-01 -4.70808804e-01 5.05836785e-01 -5.48326150e-02 -5.37278056e-01 1.55539230e-01 -1.94083109e-01 -1.32612020e-01 -1.30441055e-01 -5.20977795e-01 -6.88700855e-01 -5.17686605e-01 -2.20378324e-01 3.82095486e-01 9.57012057e-01 -2.51163244e-02 -1.23449363e-01 3.93015146e-01 -1.39959979e+00 5.31227052e-01 6.60025060e-01 1.26743281e+00 7.81040937e-02 9.83958721e-01 1.06088005e-01 5.43441117e-01 -2.71204025e-01 -4.11287069e-01 5.86048245e-01 -2.16820855e-02 -8.55179012e-01 4.46126163e-01 9.94531453e-01 -2.94274479e-01 -2.93374568e-01 6.72448754e-01 3.13951164e-01 2.85357773e-01 3.67008835e-01 -1.29514515e+00 -1.37661070e-01 5.89562416e-01 -7.12412000e-01 6.45762086e-01 -1.21111777e-02 3.19564253e-01 1.02381051e+00 -9.58562016e-01 4.20841843e-01 1.12595081e+00 1.03208923e+00 2.20756575e-01 -1.44140553e+00 -6.05290413e-01 5.46416998e-01 1.80221617e-01 -1.31338227e+00 -2.10299402e-01 3.57668251e-01 -5.23232818e-01 1.19089067e+00 2.06886768e-01 8.63638520e-01 7.83382535e-01 -3.24005745e-02 1.04192698e+00 1.02281773e+00 -9.71050113e-02 9.69302841e-03 -1.95648149e-01 1.37727156e-01 7.69889951e-01 -5.96666709e-02 2.82665482e-03 -5.68825185e-01 7.46402740e-02 7.47933149e-01 -2.29521897e-02 9.00101438e-02 -3.82469326e-01 -1.21102715e+00 6.84790254e-01 4.77459341e-01 5.56213185e-02 -5.23163974e-01 4.57098961e-01 4.56956685e-01 -8.92890021e-02 5.79360008e-01 3.41435164e-01 -6.68948948e-01 -4.05652113e-02 -1.23086512e+00 2.42892414e-01 2.76841700e-01 9.89513218e-01 4.18553412e-01 6.20468371e-02 -7.33330488e-01 6.36688590e-01 2.41086677e-01 2.25568861e-01 -1.89689025e-02 -1.05934012e+00 3.84885371e-01 6.20468795e-01 -7.98955485e-02 -5.93499839e-01 -5.70277333e-01 -7.28999972e-01 -3.83348227e-01 3.08168232e-01 4.32709783e-01 -3.48564908e-02 -1.34547257e+00 1.47291887e+00 7.89133728e-01 5.91146171e-01 -1.92079857e-01 9.31302547e-01 7.54359961e-01 5.79103827e-01 4.32079613e-01 1.59592003e-01 1.36339247e+00 -1.27135694e+00 -1.86288327e-01 -2.16838792e-01 4.42141384e-01 -6.15081966e-01 1.00655234e+00 5.90955734e-01 -9.98620629e-01 -7.60889888e-01 -8.08559060e-01 -3.97491977e-02 -2.24004030e-01 1.64169550e-01 9.51156437e-01 7.64293909e-01 -1.22915721e+00 6.57392144e-01 -1.08526886e+00 -3.75438422e-01 9.24748242e-01 4.68952477e-01 -2.03933388e-01 6.47931080e-03 -5.08487701e-01 5.19406319e-01 6.44616902e-01 2.57822871e-01 -8.88843358e-01 -9.28615928e-01 -7.53292859e-01 2.45263670e-02 6.65179372e-01 -7.10237861e-01 1.11754465e+00 -7.47923613e-01 -1.34321058e+00 5.73779941e-01 -8.94071185e-04 -8.49677205e-01 7.54592299e-01 -5.33858180e-01 -8.45638365e-02 2.60777503e-01 3.33238915e-02 1.16580248e+00 1.01741576e+00 -8.01121354e-01 -9.91461635e-01 -2.33300298e-01 3.37644592e-02 1.90304950e-01 -3.04405510e-01 1.52275592e-01 -9.78830695e-01 -8.00992370e-01 1.52970731e-01 -9.00462747e-01 -3.35333824e-01 -3.78974862e-02 -4.05681044e-01 -2.43464276e-01 1.14238453e+00 -6.65959835e-01 1.16651344e+00 -2.11394811e+00 4.34641510e-01 2.23920628e-01 4.81435388e-01 3.63536894e-01 -2.53784269e-01 -1.52623802e-01 1.66369289e-01 -7.95167536e-02 -3.18771988e-01 -6.06593072e-01 -8.72227401e-02 2.04814583e-01 -2.13196734e-03 2.09239125e-01 5.47833979e-01 1.09892392e+00 -7.47409225e-01 -6.72153592e-01 3.02605599e-01 4.77861404e-01 -7.68654108e-01 -1.95391253e-01 -3.89847666e-01 3.59230578e-01 -3.66917998e-01 7.80241489e-01 5.31701148e-01 -4.16754425e-01 -2.74736047e-01 -3.10782611e-01 -2.04861075e-01 2.18365546e-02 -1.14781797e+00 1.66882682e+00 -1.47146154e-02 6.17712796e-01 -1.78152714e-02 -4.98349488e-01 4.30383921e-01 1.55194581e-01 7.69906640e-01 -5.68052173e-01 1.18970260e-01 -8.22687447e-02 -4.66745123e-02 -2.60350257e-01 6.01899862e-01 1.34427994e-01 1.25931501e-01 7.32016340e-02 2.91103601e-01 1.10433079e-01 5.96983671e-01 3.29061687e-01 1.15465903e+00 5.13699532e-01 -2.33095080e-01 -2.11361483e-01 3.14106435e-01 9.69172269e-02 6.19639099e-01 9.28240597e-01 -3.02544504e-01 7.60974765e-01 3.64073664e-01 -3.21009457e-01 -1.03930497e+00 -1.04531121e+00 -1.56130761e-01 1.47658491e+00 6.17743060e-02 -5.11930585e-01 -7.82920420e-01 -1.06530833e+00 1.39672086e-01 3.93429369e-01 -6.76064491e-01 7.33343363e-02 -7.05243707e-01 -8.29457104e-01 4.34962124e-01 9.54876125e-01 5.95075309e-01 -1.03252923e+00 -8.09863091e-01 2.91260600e-01 6.95924759e-02 -1.54722309e+00 -6.77257836e-01 3.14451456e-01 -9.02171135e-01 -1.04247153e+00 -7.11509824e-01 -6.37169898e-01 4.37302053e-01 7.11328834e-02 1.12531304e+00 -1.03780948e-01 -3.86181355e-01 4.26636130e-01 -4.65379149e-01 -2.73340315e-01 -1.81103885e-01 1.49913058e-01 -2.55664676e-01 9.32800323e-02 1.49477543e-02 -2.48163253e-01 -6.18574023e-01 2.43719816e-01 -8.31759453e-01 3.74166250e-01 4.49309349e-01 5.00151515e-01 7.11132288e-01 -5.07523119e-02 2.84791350e-01 -9.18544769e-01 -1.23737633e-01 -1.94249168e-01 -7.86043942e-01 8.19378868e-02 -3.61707151e-01 -2.73795545e-01 1.97393864e-01 -5.83870471e-01 -8.26163888e-01 1.14036739e-01 -2.59524167e-01 -6.28282249e-01 -5.01416214e-02 9.32375193e-02 -8.14097840e-03 -3.28417629e-01 3.23665559e-01 1.00733534e-01 -2.16565847e-01 -4.33500290e-01 3.22004378e-01 6.32929336e-03 6.06653571e-01 -4.45561826e-01 7.71455944e-01 4.27571297e-01 -6.20617578e-03 -5.22727132e-01 -9.39309716e-01 -5.51159263e-01 -6.73351884e-01 -3.56543064e-01 1.11864734e+00 -8.71664166e-01 -4.77064699e-01 7.01303422e-01 -7.28739083e-01 -7.11781144e-01 -4.98469889e-01 4.97082382e-01 -3.94330144e-01 1.59538358e-01 -7.14790940e-01 -5.83450854e-01 -3.36337745e-01 -1.16910398e+00 1.23715246e+00 4.10718620e-01 -7.05039427e-02 -6.72502995e-01 -3.76209766e-01 5.04108548e-01 3.52871627e-01 3.59903157e-01 5.18349051e-01 -5.68747699e-01 -8.23244452e-01 -7.93496445e-02 -3.76874864e-01 5.49029708e-01 -2.65463024e-01 2.95767896e-02 -8.39923203e-01 -3.98690820e-01 -5.08515000e-01 -1.86047852e-02 1.27126098e+00 6.34852529e-01 1.12803769e+00 1.03245713e-01 -3.12485784e-01 8.68584931e-01 1.13579965e+00 1.65241748e-01 4.07354176e-01 3.24678808e-01 1.05912924e+00 1.92262903e-01 6.12703145e-01 2.77011633e-01 3.76526207e-01 7.61120200e-01 4.58716005e-01 -4.65453923e-01 -3.85298461e-01 8.36042129e-03 1.39085427e-01 2.12284282e-01 -1.22166395e-01 -1.76446080e-01 -8.30605388e-01 5.18050849e-01 -1.68507612e+00 -6.86677337e-01 -2.50444710e-01 2.05278444e+00 4.50819105e-01 5.68225682e-01 5.15812993e-01 5.20785674e-02 4.92510229e-01 3.67792398e-02 -5.91849625e-01 4.57245596e-02 -1.24642134e-01 3.22853267e-01 7.56201863e-01 2.20995098e-01 -1.51396716e+00 1.19722474e+00 6.60207605e+00 8.69047761e-01 -9.94469464e-01 2.97716320e-01 5.74896455e-01 -1.04958951e-01 2.85755157e-01 5.57215884e-04 -1.06993139e+00 3.68797451e-01 5.80051303e-01 4.93247211e-01 2.73203194e-01 6.96240544e-01 2.02615503e-02 -3.14261734e-01 -9.90454078e-01 6.28266215e-01 -1.02352187e-01 -1.28067577e+00 -4.37983200e-02 -2.07538866e-02 7.86531210e-01 4.28123146e-01 2.73298979e-01 5.42198420e-01 8.71587172e-02 -7.59249091e-01 1.00352645e+00 9.90960151e-02 4.66545165e-01 -6.19970918e-01 4.60449636e-01 8.16838369e-02 -1.44462037e+00 -2.65250891e-01 6.42436147e-02 2.49162868e-01 3.46044153e-01 3.55349571e-01 -1.05782962e+00 5.07292449e-01 9.87642646e-01 7.30956078e-01 -9.53448474e-01 1.45397210e+00 1.57240227e-01 9.82669473e-01 -6.47620201e-01 4.89183754e-01 4.75563824e-01 -1.28224760e-01 7.96827555e-01 1.53298271e+00 1.86451495e-01 -3.89868580e-02 3.83045137e-01 6.61351025e-01 2.01345701e-02 -1.91410229e-01 1.23958841e-01 3.51924784e-02 1.07551798e-01 1.46253872e+00 -1.27155066e+00 -4.60222423e-01 -2.95561373e-01 9.73925114e-01 4.08723839e-02 4.45658386e-01 -1.15694380e+00 -2.47158147e-02 5.60803831e-01 2.59646446e-01 1.02298105e+00 -2.62360722e-01 -1.70921892e-01 -8.03251147e-01 -1.29160568e-01 -6.82925224e-01 5.62862754e-01 -6.45190120e-01 -1.01548064e+00 5.79818428e-01 1.10067263e-01 -1.05257368e+00 2.49031633e-01 -5.82479417e-01 -5.40797114e-01 5.59851766e-01 -1.36241031e+00 -1.51210427e+00 -2.93332100e-01 5.08199990e-01 7.19981372e-01 5.72607629e-02 2.86910146e-01 5.38494229e-01 -7.80107379e-01 5.65156341e-01 -3.32538754e-01 1.60719380e-01 2.38583535e-01 -1.34305024e+00 6.93900466e-01 1.04423392e+00 3.10708523e-01 1.53202057e-01 7.19030976e-01 -6.54146314e-01 -1.06829762e+00 -1.62107980e+00 3.15548718e-01 -6.56351686e-01 4.82097715e-01 -4.23438936e-01 -9.95339692e-01 9.00162697e-01 -3.00861858e-02 3.61287713e-01 3.95795822e-01 9.76469666e-02 -3.43752861e-01 -7.81737566e-02 -1.02671707e+00 3.83366883e-01 1.15508592e+00 -2.43797153e-01 -1.94388211e-01 3.06464255e-01 8.57647419e-01 -7.49420524e-01 -8.74821842e-01 6.91475511e-01 4.97986406e-01 -8.41800809e-01 1.09972441e+00 -4.16716099e-01 9.11957584e-03 -5.83261669e-01 -1.31405979e-01 -8.09282660e-01 -5.18655419e-01 -7.17764735e-01 -3.44558567e-01 1.20847130e+00 3.34747404e-01 -3.84243309e-01 9.76243615e-01 3.03251415e-01 -3.67450058e-01 -9.08670723e-01 -9.67459679e-01 -6.72860563e-01 -3.42497677e-01 -6.88990176e-01 3.72448385e-01 3.40014398e-01 -8.00276756e-01 9.11655873e-02 -2.10576251e-01 4.08619881e-01 8.07198167e-01 -4.43642028e-02 6.87655389e-01 -1.13247025e+00 -6.46746516e-01 -5.36655605e-01 -6.06562078e-01 -1.19638467e+00 -2.40134999e-01 -8.63995731e-01 1.20056406e-01 -1.59948385e+00 3.82736444e-01 -4.60803032e-01 -4.94931996e-01 5.31270862e-01 -4.43499655e-01 9.16115880e-01 4.33772147e-01 -8.98330584e-02 -1.19693124e+00 3.03772837e-01 1.12034345e+00 -7.73865134e-02 -3.29352915e-01 1.05574094e-01 -2.79173672e-01 8.05788398e-01 5.63262999e-01 -4.72903252e-01 -1.10866114e-01 -3.93382013e-01 -1.15507185e-01 -2.79643059e-01 7.01558888e-01 -1.17358696e+00 -1.91655960e-02 1.84762269e-01 7.07904696e-01 -9.48931336e-01 5.29687166e-01 -7.23140121e-01 1.32934779e-01 3.39716762e-01 -1.56671986e-01 9.73734409e-02 6.01070523e-01 5.96141934e-01 1.08711228e-01 1.51679441e-01 7.40891814e-01 2.74264514e-01 -1.06798530e+00 5.10226727e-01 -2.73629099e-01 -3.08467932e-02 1.23344886e+00 -6.40886903e-01 2.85834067e-05 2.78505802e-01 -8.91643882e-01 5.97246289e-01 2.49685854e-01 5.80948353e-01 4.40955937e-01 -1.01480699e+00 -5.95363557e-01 2.69630343e-01 9.60384160e-02 1.39874503e-01 4.34295475e-01 1.30854058e+00 -3.66804600e-01 1.89040631e-01 -3.45856808e-02 -1.13040769e+00 -1.51780999e+00 1.84435040e-01 2.72702426e-01 -3.95391047e-01 -9.40463245e-01 1.23500764e+00 2.62044609e-01 -1.71058968e-01 2.11789936e-01 -5.14619470e-01 -2.80715227e-01 1.23026716e-02 3.37733209e-01 4.76914018e-01 1.44451097e-01 -7.40037084e-01 -3.37791026e-01 5.40471733e-01 -2.66238064e-01 -5.79461753e-02 1.25069022e+00 -5.96095547e-02 2.67058909e-01 1.30538493e-01 9.19012666e-01 -2.10475773e-01 -2.04738975e+00 -2.66726196e-01 1.85419649e-01 -4.44414675e-01 3.16620111e-01 -8.38302970e-01 -1.18344891e+00 4.76035774e-01 7.98729777e-01 2.56248750e-02 1.21404171e+00 2.25926071e-01 8.29152584e-01 -3.17750052e-02 2.76416391e-01 -9.01379406e-01 2.08685309e-01 5.58010280e-01 4.39748466e-01 -1.13774657e+00 -7.45892152e-02 -4.54071432e-01 -6.54885411e-01 7.85227954e-01 7.05960870e-01 -5.72341308e-02 3.02878797e-01 4.58873361e-01 -9.80754942e-02 -1.59002662e-01 -6.40111148e-01 -4.35315132e-01 6.32631481e-01 5.07688999e-01 1.64975196e-01 -1.31854430e-01 2.11867258e-01 3.25800478e-01 6.72893822e-02 4.33615223e-02 8.89561176e-02 8.71662617e-01 -4.43445593e-01 -7.28338838e-01 -2.79343605e-01 5.63165784e-01 -5.11944771e-01 -1.84691399e-01 8.40483159e-02 7.56682515e-01 4.18510616e-01 7.86258340e-01 1.83529854e-01 -3.75496209e-01 3.90035450e-01 2.46255044e-02 6.43421590e-01 -5.68479538e-01 -9.86292660e-01 5.03499866e-01 -1.07866526e-01 -8.75507534e-01 -5.64489841e-01 -8.36652458e-01 -1.30054641e+00 2.59792898e-02 -4.18889344e-01 -4.28460762e-02 4.30511475e-01 9.79085565e-01 4.17015851e-01 9.68444169e-01 3.68006453e-02 -1.15960240e+00 -1.65655106e-01 -8.98331642e-01 -3.65985811e-01 1.54522270e-01 3.53267998e-01 -5.96055448e-01 1.92149043e-01 1.17495298e-01]
[9.03046989440918, -0.16447462141513824]
5287aced-dc66-40b2-a018-f4bac35ab143
touch-and-go-learning-from-human-collected
2211.12498
null
https://arxiv.org/abs/2211.12498v2
https://arxiv.org/pdf/2211.12498v2.pdf
Touch and Go: Learning from Human-Collected Vision and Touch
The ability to associate touch with sight is essential for tasks that require physically interacting with objects in the world. We propose a dataset with paired visual and tactile data called Touch and Go, in which human data collectors probe objects in natural environments using tactile sensors, while simultaneously recording egocentric video. In contrast to previous efforts, which have largely been confined to lab settings or simulated environments, our dataset spans a large number of "in the wild" objects and scenes. To demonstrate our dataset's effectiveness, we successfully apply it to a variety of tasks: 1) self-supervised visuo-tactile feature learning, 2) tactile-driven image stylization, i.e., making the visual appearance of an object more consistent with a given tactile signal, and 3) predicting future frames of a tactile signal from visuo-tactile inputs.
['Andrew Owens', 'Wenzhen Yuan', 'Jing Zhu', 'Jiacheng Zhang', 'Chenyang Ma', 'Fengyu Yang']
2022-11-22
null
null
null
null
['image-stylization']
['computer-vision']
[ 4.80145633e-01 -4.05480415e-01 1.26298696e-01 -2.71414459e-01 -2.07094803e-01 -8.20838630e-01 5.77262580e-01 -3.29325736e-01 -2.76221901e-01 4.57301468e-01 1.04639336e-01 3.20458077e-02 5.41463643e-02 -4.72769886e-01 -9.84829724e-01 -3.20679605e-01 6.31199628e-02 1.43983096e-01 1.54132813e-01 2.94750363e-01 7.37616837e-01 4.54326361e-01 -1.96964109e+00 1.26222536e-01 2.62847960e-01 1.26920044e+00 5.91251016e-01 1.11328614e+00 -1.67557612e-01 3.73058468e-01 -1.74706846e-01 4.35184985e-02 2.65049160e-01 -1.64505482e-01 -5.14003158e-01 1.63947389e-01 7.50871956e-01 -6.56189799e-01 -2.48110294e-01 8.04162443e-01 4.40989941e-01 6.13799617e-02 7.42492974e-01 -1.32482743e+00 -1.04463124e+00 2.90295649e-02 -6.64783597e-01 -2.02143833e-01 9.98381138e-01 3.55473250e-01 5.71745098e-01 -8.04212570e-01 8.72435093e-01 1.46188557e+00 3.36194485e-01 7.64462709e-01 -1.48059261e+00 -4.55403328e-01 1.99003428e-01 4.50963937e-02 -9.03979659e-01 -6.00616336e-01 9.62425768e-01 -6.12150848e-01 7.51149535e-01 3.77918392e-01 6.73333585e-01 1.75721276e+00 9.10829231e-02 8.96351337e-01 1.24402452e+00 -4.76922423e-01 3.98420900e-01 1.19720079e-01 -3.00755978e-01 4.59025890e-01 1.29295081e-01 2.06550330e-01 -1.24895000e+00 -1.74529236e-02 1.47341883e+00 4.66810130e-02 1.75893446e-03 -8.89609218e-01 -1.52097023e+00 1.93047196e-01 3.02213132e-01 9.51677095e-03 -4.30806875e-01 4.72224236e-01 -6.78202882e-02 2.88703352e-01 1.79955512e-01 5.37463546e-01 -3.56623054e-01 -4.85621005e-01 -2.77379721e-01 1.31702781e-01 6.34616435e-01 1.16930187e+00 7.52604306e-01 -1.00511104e-01 7.61664808e-02 3.89602214e-01 3.69090319e-01 9.66787457e-01 2.54682004e-01 -1.32550120e+00 2.92003661e-01 2.07136497e-01 7.18331635e-01 -7.47695744e-01 -3.15036625e-01 6.86419666e-01 -1.32506713e-01 6.46525323e-01 6.10846221e-01 -2.32585773e-01 -9.23456252e-01 1.75613260e+00 2.63964564e-01 4.10068184e-01 -2.57833004e-01 1.07361126e+00 7.35325158e-01 1.70349553e-01 1.92686781e-01 1.58265948e-01 9.94470239e-01 -3.33283782e-01 -6.16349995e-01 -5.36529422e-01 -1.02391906e-01 -8.85657966e-01 1.65509999e+00 4.52817529e-01 -9.13375318e-01 -5.74293494e-01 -6.35508299e-01 -2.22631320e-01 -4.36734468e-01 -5.25957458e-02 1.06540883e+00 3.76326799e-01 -1.00128663e+00 5.30638456e-01 -8.73904407e-01 -1.03489995e+00 3.69568288e-01 4.92526330e-02 -5.59373498e-01 2.67947465e-01 -3.91007632e-01 6.86612368e-01 -7.56216273e-02 -1.44839725e-02 -6.52835429e-01 -4.61872101e-01 -8.02922487e-01 -3.04096133e-01 3.14540982e-01 -7.57413507e-01 1.14919734e+00 -8.17220867e-01 -1.71525371e+00 1.09922338e+00 -5.33526361e-01 7.66124129e-02 5.93186080e-01 -7.30871439e-01 -1.84843749e-01 2.06485882e-01 -9.03249253e-03 9.21096981e-01 1.10026288e+00 -1.74901497e+00 -2.98783481e-01 -6.68505490e-01 -1.79553553e-01 4.62521017e-01 -2.42650911e-01 -2.87686616e-01 -4.50045049e-01 -4.14784700e-01 2.71458089e-01 -7.43326724e-01 1.06964372e-01 8.04737210e-01 -5.60412109e-01 1.32788029e-02 1.12657976e+00 -1.44540697e-01 4.24507231e-01 -2.14231896e+00 1.08312361e-01 2.92602479e-01 2.19418243e-01 2.50946190e-02 -2.04108208e-01 4.51041371e-01 3.54213297e-01 -8.21769014e-02 5.52622199e-01 -5.96782684e-01 9.68916491e-02 -4.27401140e-02 -5.52603304e-01 2.15248421e-01 2.26224097e-03 1.24331462e+00 -1.11179972e+00 -4.92207080e-01 8.30774665e-01 1.69550687e-01 -2.99756944e-01 4.89518881e-01 -2.24443644e-01 7.72315621e-01 -6.58659041e-01 9.57400978e-01 5.48669338e-01 -1.48905246e-02 -2.30970122e-02 -2.25714967e-01 -8.15151110e-02 3.25286239e-02 -9.94195461e-01 2.06066704e+00 -4.72269207e-01 7.07325757e-01 1.15778796e-01 -4.33982164e-01 9.54736769e-01 -8.67683664e-02 4.79301631e-01 -1.06046462e+00 2.00051636e-01 -1.10055141e-01 -6.66427612e-01 -1.12392974e+00 5.13371885e-01 2.21749052e-01 -9.55165401e-02 5.21133900e-01 -1.16453707e-01 -3.56115490e-01 -3.14279258e-01 4.84434702e-02 9.72338438e-01 7.57746279e-01 9.80442688e-02 3.79391052e-02 -4.91022259e-01 -3.33660364e-01 -1.17127553e-01 1.08961761e+00 -3.36172342e-01 6.98122978e-01 1.02587476e-01 -3.29033881e-01 -9.88206744e-01 -2.10310817e+00 -7.51467496e-02 1.31924748e+00 7.60174930e-01 2.58382559e-01 -4.23625529e-01 -1.64864168e-01 5.79466879e-01 6.09270334e-01 -9.81418192e-01 2.25815222e-01 2.90303566e-02 3.35174620e-01 2.47631967e-01 8.78593624e-01 2.73831218e-01 -1.65366375e+00 -1.23833156e+00 -1.38392881e-01 3.13462764e-02 -1.01873815e+00 -4.44982141e-01 2.35258445e-01 -5.71059465e-01 -1.05894053e+00 -5.15599430e-01 -5.87319672e-01 5.56331515e-01 5.75160086e-01 9.43089724e-01 -5.70483208e-01 -6.73516273e-01 1.18043733e+00 -3.09307605e-01 -8.13533485e-01 3.17551434e-01 -3.57513785e-01 3.25617880e-01 -3.16691399e-02 6.26010239e-01 -6.12343490e-01 -6.61595941e-01 2.65111864e-01 -7.21611679e-01 2.33072102e-01 4.09460008e-01 4.81318831e-01 3.66263658e-01 -1.08109939e+00 5.27901411e-01 -5.01928091e-01 6.79462254e-01 -2.42152095e-01 -2.54072219e-01 3.29967052e-01 1.33088663e-01 -2.06448272e-01 2.80934364e-01 -9.30239320e-01 -1.07454515e+00 2.12638140e-01 5.12682915e-01 -5.87799907e-01 -5.20801425e-01 -4.82340083e-02 2.66896244e-02 -1.51757717e-01 9.46489036e-01 -5.69197834e-02 1.95689440e-01 -2.13584140e-01 5.44194162e-01 7.86612809e-01 8.66991639e-01 -7.22433388e-01 7.23408043e-01 7.96175241e-01 -1.51669115e-01 -1.01323044e+00 -5.29298306e-01 -3.92708898e-01 -9.53347623e-01 -5.20876467e-01 8.65309417e-01 -3.96739900e-01 -1.53241861e+00 6.39308631e-01 -8.17665458e-01 -7.54334390e-01 -3.14094931e-01 7.03314483e-01 -9.87135231e-01 2.09557697e-01 -2.27755249e-01 -1.13007975e+00 7.96541423e-02 -6.60571396e-01 1.46404350e+00 5.80169022e-01 -7.33726025e-01 -8.82730484e-01 -1.84480324e-02 5.58965690e-02 4.38696116e-01 4.35147643e-01 4.97924358e-01 1.86503127e-01 -4.40362960e-01 3.60730104e-02 -4.32818592e-01 -1.11869499e-01 6.99375272e-01 9.88581479e-02 -1.44744313e+00 -2.13009641e-01 -4.42347854e-01 -9.51863527e-01 3.42681944e-01 5.32275736e-01 1.27574778e+00 2.81588525e-01 -4.82838988e-01 5.61180770e-01 1.12549174e+00 3.86348575e-01 5.71584105e-01 5.40521629e-02 7.61714816e-01 5.90586603e-01 6.97914600e-01 6.68541729e-01 1.83140501e-01 5.33821821e-01 4.80047613e-01 -1.47745296e-01 1.85407586e-02 -5.39979458e-01 -4.51500266e-04 4.33216542e-02 -7.14911148e-02 -2.42684796e-01 -5.28753638e-01 4.65929240e-01 -1.75931895e+00 -7.45486557e-01 3.41303319e-01 2.32455707e+00 4.73800123e-01 6.80853054e-03 4.89916727e-02 -3.15308839e-01 4.96862620e-01 1.05403969e-02 -1.36179662e+00 -4.51778650e-01 9.48077068e-02 1.47154912e-01 3.47582489e-01 1.32687271e-01 -1.04224455e+00 9.77617145e-01 7.25410938e+00 -1.47835866e-01 -1.36320424e+00 -6.21118605e-01 7.32224435e-02 -3.13496500e-01 -2.91398793e-01 -2.84669131e-01 -1.57884210e-01 3.30021054e-01 1.08576201e-01 -7.90357813e-02 1.06263745e+00 6.36965692e-01 1.80462092e-01 -7.40052223e-01 -1.63708210e+00 1.31355369e+00 1.39065431e-02 -8.15166950e-01 -1.16029091e-01 -2.08767429e-01 3.73420656e-01 -1.40449002e-01 5.98731399e-01 -1.75451458e-01 4.74198788e-01 -1.11356008e+00 7.67577052e-01 1.02449048e+00 1.22657537e+00 2.09448431e-02 -1.36314735e-01 1.02604926e-01 -1.08243287e+00 6.28743991e-02 -2.55270183e-01 -3.48285228e-01 1.50046542e-01 2.02609524e-01 -7.26409256e-01 -9.00800452e-02 1.12861931e+00 6.91109598e-01 -4.76244569e-01 1.03850138e+00 1.32011712e-01 2.24918485e-01 -4.82695371e-01 -3.67952764e-01 -1.12193525e-01 -7.30134770e-02 2.74454147e-01 9.50646698e-01 3.22621793e-01 -1.21316239e-01 -1.00001730e-01 1.00531793e+00 4.58757393e-02 -1.98610783e-01 -1.43745863e+00 -8.40273499e-02 7.92851686e-01 9.90102887e-01 -6.12333119e-01 -4.88514043e-02 -1.64559647e-01 1.10810637e+00 3.44347656e-01 7.11486995e-01 -4.06256557e-01 -4.56873149e-01 7.65902698e-01 -1.87180832e-01 2.52558917e-01 -4.16786909e-01 -7.19072700e-01 -1.03868043e+00 4.03496921e-01 -1.67524323e-01 -2.27158487e-01 -1.54847813e+00 -1.45040321e+00 -2.26418879e-02 -2.09613487e-01 -1.50073421e+00 -3.56705099e-01 -6.38082325e-01 -5.00399709e-01 8.41497600e-01 -1.09075880e+00 -1.23498499e+00 -8.65187705e-01 7.16146827e-01 3.26245874e-01 2.26792991e-01 9.92714643e-01 -3.05223048e-01 1.49012908e-01 4.51227844e-01 8.90228078e-02 -1.96442187e-01 1.28925455e+00 -1.18621731e+00 6.22442365e-01 1.89252555e-01 9.56200659e-02 8.23722661e-01 7.14283943e-01 -7.76967764e-01 -1.99572372e+00 -5.38709462e-01 1.46372274e-01 -1.08534491e+00 7.07617283e-01 -9.08544898e-01 -6.50927484e-01 8.72349977e-01 -1.92507990e-02 1.19015248e-02 3.56073618e-01 1.88565820e-01 -4.24333006e-01 -1.53750228e-02 -1.28474498e+00 9.67749715e-01 1.58493865e+00 -9.71445441e-01 -6.55346334e-01 -5.25758453e-02 3.22982728e-01 -5.50038397e-01 -4.23898131e-01 -3.54657806e-02 1.40512681e+00 -8.68954360e-01 1.08356273e+00 -6.94638133e-01 3.96951258e-01 -1.40471488e-01 -2.30185077e-01 -1.40843034e+00 -3.58020514e-01 -5.00426531e-01 7.98889901e-03 1.07045174e+00 -1.44892901e-01 -5.34343958e-01 8.50380063e-01 7.82348990e-01 2.55113363e-01 -1.44145310e-01 -8.47151339e-01 -6.21336162e-01 -3.26280832e-01 -5.67985654e-01 2.38856152e-01 7.33785987e-01 2.89051086e-01 -7.67979100e-02 -4.92430151e-01 -9.83607545e-02 7.52872109e-01 1.48849517e-01 1.17204416e+00 -1.41737711e+00 -1.01657221e-02 -3.36713374e-01 -3.99968594e-01 -1.11479366e+00 4.77273669e-03 -3.37120742e-01 3.13659519e-01 -1.54119205e+00 1.30976796e-01 -3.16011757e-01 -1.15560934e-01 3.98813814e-01 -9.18680131e-02 2.89500505e-01 1.02682494e-01 4.58773710e-02 -7.18877554e-01 3.44598621e-01 1.56977069e+00 1.80739105e-01 -2.92898208e-01 -2.21025258e-01 -7.37296045e-01 7.40718007e-01 4.74315047e-01 2.05361977e-01 -5.58525085e-01 -7.40039110e-01 -4.80824354e-04 1.76929697e-01 7.63545513e-01 -8.84249628e-01 2.55152255e-01 -6.83924794e-01 7.74955392e-01 -4.74421561e-01 7.75025487e-01 -8.64126921e-01 -2.92668287e-02 8.79063010e-02 -4.96202081e-01 -1.63418889e-01 3.94487679e-01 6.84336066e-01 4.12873298e-01 2.27684781e-01 3.51348758e-01 -5.84642738e-02 -1.09255743e+00 -4.76090331e-03 -4.04478312e-01 -1.02412939e-01 1.24377894e+00 -5.99989116e-01 -7.16958344e-01 -5.53643167e-01 -6.87482834e-01 1.61696315e-01 8.54927897e-01 7.84717977e-01 9.36812639e-01 -1.39834249e+00 -7.22082034e-02 5.06839871e-01 3.76784831e-01 -2.62832195e-01 3.94907326e-01 3.26334894e-01 -1.90777674e-01 8.11004117e-02 -5.87628961e-01 -8.51476312e-01 -1.16325760e+00 6.21076584e-01 -9.66213793e-02 7.20670223e-01 -5.79161942e-01 7.21000195e-01 5.27088583e-01 -3.07544291e-01 4.19198692e-01 -5.16029000e-01 7.33174160e-02 -3.46731693e-01 3.47401530e-01 2.32940882e-01 -3.81289035e-01 -1.22486673e-01 -2.33935729e-01 8.63931715e-01 1.58553898e-01 -3.58536452e-01 9.77484107e-01 -3.24934036e-01 4.06287789e-01 1.19816685e+00 9.81406331e-01 -9.78898406e-02 -2.02791667e+00 -3.96330833e-01 -4.69047725e-01 -8.90385687e-01 -3.05522144e-01 -8.62103999e-01 -3.97830963e-01 8.86471450e-01 7.36817837e-01 8.82615000e-02 8.48186433e-01 1.95031509e-01 4.90842491e-01 8.85010302e-01 9.12862420e-01 -1.34585857e+00 8.27371597e-01 3.54026079e-01 8.43047917e-01 -1.31811583e+00 -3.91193718e-01 -2.81986415e-01 -8.55204165e-01 8.85205805e-01 8.08164179e-01 -8.60780627e-02 4.52811122e-01 7.00677931e-01 3.09737712e-01 -1.74675643e-01 -6.13484085e-01 -2.70818561e-01 1.12490967e-01 1.21686208e+00 -5.47092117e-04 2.53757060e-01 5.65361142e-01 1.06737137e-01 -1.60342321e-01 4.92426425e-01 2.94586957e-01 1.32157230e+00 -5.22546768e-01 -4.57942963e-01 -4.29647386e-01 7.65442193e-01 2.48813719e-01 2.67278254e-01 -7.01128960e-01 4.92503136e-01 -2.36125514e-01 9.92371321e-01 3.89043212e-01 -6.16508186e-01 5.30051708e-01 3.83367874e-02 1.04392052e+00 -5.77870965e-01 9.38739628e-03 -2.01051682e-02 -1.82882383e-01 -9.59318280e-01 -3.91544580e-01 -8.39993656e-01 -1.04002166e+00 -2.84273714e-01 1.41920850e-01 -6.03236735e-01 7.75633335e-01 8.74406874e-01 3.23784202e-01 2.61084706e-01 6.66956604e-01 -1.68657887e+00 -9.83869433e-02 -9.09583926e-01 -9.13041711e-01 8.04900050e-01 3.12572569e-01 -1.02031875e+00 -1.59764037e-01 2.84084499e-01]
[5.947760105133057, -0.8843041062355042]
0de596b1-5dd7-4eb8-87c6-f41cba1ddedd
iterative-document-representation-learning
1809.10324
null
https://arxiv.org/abs/1809.10324v2
https://arxiv.org/pdf/1809.10324v2.pdf
Iterative Document Representation Learning Towards Summarization with Polishing
In this paper, we introduce Iterative Text Summarization (ITS), an iteration-based model for supervised extractive text summarization, inspired by the observation that it is often necessary for a human to read an article multiple times in order to fully understand and summarize its contents. Current summarization approaches read through a document only once to generate a document representation, resulting in a sub-optimal representation. To address this issue we introduce a model which iteratively polishes the document representation on many passes through the document. As part of our model, we also introduce a selective reading mechanism that decides more accurately the extent to which each sentence in the model should be updated. Experimental results on the CNN/DailyMail and DUC2002 datasets demonstrate that our model significantly outperforms state-of-the-art extractive systems when evaluated by machines and by humans.
['Yan Song', 'Shen Gao', 'Xiuying Chen', 'Chongyang Tao', 'Rui Yan', 'Dongyan Zhao']
2018-09-27
iterative-document-representation-learning-1
https://aclanthology.org/D18-1442
https://aclanthology.org/D18-1442.pdf
emnlp-2018-10
['extractive-document-summarization']
['natural-language-processing']
[ 6.32973552e-01 6.03021026e-01 -1.99219316e-01 -2.73056269e-01 -8.54725003e-01 -6.78712726e-01 8.21684718e-01 7.87776291e-01 -5.86353600e-01 6.93781078e-01 1.10431266e+00 -1.30606383e-01 1.62571207e-01 -5.73008955e-01 -6.81545556e-01 -1.51169986e-01 3.77694964e-01 6.52444541e-01 1.66370228e-01 -2.83893824e-01 8.02551389e-01 1.87633067e-01 -1.20961010e+00 7.49326348e-01 1.30926538e+00 3.11356932e-01 2.20639303e-01 1.01713526e+00 -4.27591473e-01 8.56737018e-01 -1.04659903e+00 -5.70511162e-01 -3.17190588e-02 -9.77906704e-01 -1.41764987e+00 3.57787579e-01 6.95484102e-01 -5.88833153e-01 -2.29447573e-01 1.01560950e+00 3.18421096e-01 2.02374563e-01 8.35144937e-01 -4.67305988e-01 -5.23330212e-01 1.44017982e+00 -5.61869800e-01 4.03906345e-01 4.00934815e-01 -1.46351710e-01 1.27009094e+00 -6.23089135e-01 8.24502766e-01 1.05021453e+00 2.90915340e-01 4.66976106e-01 -1.11284113e+00 -3.10719945e-02 4.29832906e-01 9.58451778e-02 -5.39239228e-01 -5.74624181e-01 7.60053217e-01 6.81792647e-02 1.24628580e+00 4.41596389e-01 9.15593565e-01 8.71139169e-01 5.18253982e-01 1.30424118e+00 3.02430838e-01 -6.51806176e-01 1.76665783e-01 -3.13543975e-01 7.51318693e-01 5.79799354e-01 7.23939002e-01 -7.04658031e-01 -7.56939232e-01 -6.86235800e-02 1.08104169e-01 -3.06508303e-01 -4.61399168e-01 9.81390551e-02 -1.16622663e+00 7.67435431e-01 2.29742184e-01 4.15689796e-01 -8.30970466e-01 -1.76883116e-02 8.03972483e-01 2.46633917e-01 7.64487147e-01 8.76519620e-01 -1.15843065e-01 -2.50499994e-01 -1.60484111e+00 4.89039391e-01 1.13955104e+00 7.46634960e-01 4.27771717e-01 -6.85343891e-02 -6.87241733e-01 7.54148424e-01 -1.45212144e-01 6.89480528e-02 7.69497931e-01 -8.85503769e-01 7.18634546e-01 7.52880454e-01 1.21830568e-01 -7.67242908e-01 -2.05664665e-01 -5.68873167e-01 -1.04001033e+00 -3.29612851e-01 -6.00957535e-02 -3.81918140e-02 -8.74055743e-01 1.17168152e+00 -1.14726663e-01 -2.35624790e-01 1.45950451e-01 5.19943655e-01 1.04491770e+00 8.16838622e-01 -1.81986392e-01 -5.14065444e-01 1.25425601e+00 -1.37026405e+00 -9.09526229e-01 -3.70238483e-01 5.67695439e-01 -6.33323967e-01 8.06347907e-01 5.43852508e-01 -1.51317585e+00 -3.17691982e-01 -1.27419317e+00 -3.79695803e-01 -4.05084677e-02 3.59989583e-01 2.17683911e-01 1.97406679e-01 -9.76987958e-01 9.30788636e-01 -9.65588748e-01 -6.67689741e-01 4.83492672e-01 5.03027923e-02 -1.50822490e-01 1.58476844e-01 -7.91818857e-01 9.57269192e-01 8.64957452e-01 2.78665908e-02 -5.55256367e-01 -4.73204345e-01 -6.84009135e-01 5.29640019e-01 3.38567138e-01 -1.07972896e+00 1.83270872e+00 -8.96303773e-01 -1.62164843e+00 6.89020455e-01 -5.62844992e-01 -1.07504189e+00 5.27584553e-01 -7.04840958e-01 9.78949070e-02 5.34308314e-01 -3.03145219e-02 4.88841832e-01 7.11501539e-01 -1.11283386e+00 -6.72864020e-01 -2.94939727e-01 8.34865794e-02 3.86439025e-01 -3.17720801e-01 -3.24805714e-02 -5.45070589e-01 -7.12327778e-01 -8.78351629e-02 -5.78881383e-01 -2.46409029e-01 -7.98860610e-01 -9.69626009e-01 -3.70437056e-01 5.08876681e-01 -1.09266758e+00 1.65837002e+00 -1.55455649e+00 6.32955730e-01 -1.81394219e-01 4.11797345e-01 4.84711558e-01 -3.41484874e-01 9.35043514e-01 2.54303426e-01 2.63624340e-01 -5.08803070e-01 -7.13941872e-01 -5.45302825e-03 3.79099287e-02 -6.89540923e-01 8.54199007e-02 1.34494901e-01 1.02301550e+00 -9.69918132e-01 -5.65827668e-01 -1.12313144e-01 5.50149418e-02 -5.82776785e-01 1.38319910e-01 -5.06249130e-01 6.75917566e-02 -4.38008964e-01 1.35223180e-01 5.71180701e-01 -2.46142402e-01 1.17228709e-01 -1.99558035e-01 -1.63383320e-01 6.91218078e-01 -6.91894829e-01 1.81681585e+00 -2.28032812e-01 8.71782899e-01 -2.06834927e-01 -1.08437002e+00 5.23291409e-01 1.89927697e-01 1.65186226e-01 -5.10189712e-01 2.09131271e-01 -2.11530197e-02 -2.22419381e-01 -3.19995373e-01 1.49895430e+00 3.17023695e-01 -8.52447841e-03 9.29062903e-01 5.89845441e-02 -3.32539648e-01 9.91158724e-01 7.76871800e-01 1.14176857e+00 -1.31094366e-01 6.96827769e-01 -2.17524230e-01 6.10768318e-01 2.67651677e-01 3.15993130e-01 1.16619551e+00 2.57702112e-01 8.47349823e-01 7.70681322e-01 -3.10818702e-01 -1.10447693e+00 -7.06239581e-01 4.60837185e-01 9.52887654e-01 -2.11685169e-02 -9.06144083e-01 -1.21216834e+00 -9.11757231e-01 -3.29698354e-01 1.32467961e+00 -6.42241538e-01 -2.72269428e-01 -8.83315563e-01 -4.34039026e-01 5.10285497e-01 4.46300566e-01 6.29834175e-01 -1.19656622e+00 -9.01000261e-01 4.07909542e-01 -6.27589941e-01 -8.04159701e-01 -6.74598038e-01 -1.51020754e-02 -1.08951855e+00 -9.75112438e-01 -6.69955969e-01 -6.10000730e-01 8.04968953e-01 3.41439098e-01 1.22444928e+00 1.81151718e-01 1.11719936e-01 3.74286503e-01 -6.73401296e-01 -6.69399381e-01 -9.45935071e-01 9.16631818e-01 -4.42453265e-01 -2.00729504e-01 1.12990364e-01 -2.48800665e-01 -4.60746139e-01 -5.27082562e-01 -1.30105710e+00 3.88522863e-01 7.06423342e-01 6.97987914e-01 4.11782473e-01 -1.00806132e-01 5.45224309e-01 -1.24386322e+00 1.39886773e+00 -1.04276992e-01 -6.23426251e-02 5.66657007e-01 -4.41733062e-01 4.59850073e-01 7.54026055e-01 -2.42796019e-01 -1.23930693e+00 -2.82685310e-01 -8.87126550e-02 2.08392873e-01 9.12096128e-02 8.90937686e-01 2.45610207e-01 7.68102646e-01 6.88862503e-01 6.68165147e-01 -7.68151879e-02 -5.28005123e-01 5.24197578e-01 7.78324425e-01 7.08828628e-01 -2.26460338e-01 6.53105319e-01 4.21578228e-01 -4.37954277e-01 -1.13337922e+00 -1.25094569e+00 -5.26827216e-01 -8.58679891e-01 -1.01836398e-01 7.11658537e-01 -5.54040790e-01 -2.06045374e-01 4.47944105e-01 -1.66316724e+00 -1.49977162e-01 -5.77432692e-01 1.40211657e-01 -4.57208872e-01 7.49955714e-01 -5.91056168e-01 -3.35611403e-01 -1.22025716e+00 -6.22048199e-01 1.16338837e+00 4.53459620e-01 -8.30203414e-01 -8.31738353e-01 2.32257620e-01 2.65973449e-01 2.79014051e-01 3.92211194e-04 8.46830666e-01 -9.69400465e-01 -2.76642084e-01 -2.32191250e-01 3.21683586e-02 4.74842936e-01 1.76086709e-01 1.74803823e-01 -4.62340176e-01 -3.09578747e-01 -1.09407827e-01 -2.60346830e-01 1.59200728e+00 3.72526407e-01 1.00187850e+00 -7.78967798e-01 -2.38230407e-01 1.29836857e-01 9.43684995e-01 -2.20978901e-01 7.96831131e-01 3.44310313e-01 4.37522709e-01 5.14789164e-01 3.81403714e-01 5.08735538e-01 3.44962984e-01 2.91656051e-02 1.27931476e-01 1.89311236e-01 -2.67465293e-01 -4.31580484e-01 2.75343031e-01 1.29414356e+00 6.68327138e-02 -7.63136029e-01 -4.91766542e-01 6.65151358e-01 -1.95736694e+00 -1.24965608e+00 -1.55327812e-01 1.92094421e+00 9.97025549e-01 2.37660751e-01 -1.10216355e-02 9.32241306e-02 4.75692034e-01 6.03967547e-01 -4.87533122e-01 -7.85242081e-01 -8.07748586e-02 2.06084982e-01 3.20816517e-01 5.69918811e-01 -8.16718221e-01 1.09469211e+00 6.36762857e+00 5.19869864e-01 -9.80927765e-01 -3.63781393e-01 3.79664510e-01 -1.84814885e-01 -3.93370420e-01 -9.33510717e-03 -8.08293581e-01 2.19871566e-01 7.99632907e-01 -7.96223998e-01 1.94893718e-01 3.93142790e-01 3.28650296e-01 -5.02535403e-01 -1.16249263e+00 4.57523018e-01 6.52018726e-01 -1.65360820e+00 7.31894970e-01 -1.72367051e-01 8.40830266e-01 -1.43959364e-02 -3.63082081e-01 2.26336226e-01 3.69602948e-01 -6.79138243e-01 8.50165427e-01 5.87530255e-01 2.13329762e-01 -7.17615664e-01 6.12432301e-01 6.16624713e-01 -7.92875230e-01 9.45170447e-02 -4.79768723e-01 6.41072020e-02 3.13811272e-01 5.38866937e-01 -9.30054963e-01 8.28106940e-01 2.96881318e-01 7.84454525e-01 -8.28552067e-01 1.01045251e+00 -4.62062031e-01 7.70346761e-01 -4.27212790e-02 -2.33796135e-01 3.79616588e-01 -1.01417519e-01 9.61106777e-01 1.52217603e+00 2.38198698e-01 1.45641938e-01 -1.17812283e-01 7.16958284e-01 -5.27828276e-01 1.96223065e-01 -2.40192577e-01 -3.90146405e-01 2.08376944e-01 1.07019424e+00 -7.70889461e-01 -6.89854085e-01 7.42250457e-02 1.37199080e+00 5.22908926e-01 3.83800328e-01 -4.27620649e-01 -7.20865488e-01 -8.49812175e-04 -2.96053886e-02 5.38934529e-01 -2.85083085e-01 -3.19277644e-01 -1.41994715e+00 1.75134271e-01 -9.50386524e-01 3.17017168e-01 -6.11352444e-01 -7.47461140e-01 6.74273431e-01 -2.91062519e-02 -1.01092064e+00 -5.34250259e-01 6.64402843e-02 -9.72019851e-01 4.94707257e-01 -1.36274493e+00 -8.77413332e-01 -1.86300755e-01 -1.91740394e-01 1.21632183e+00 4.28598002e-02 6.48518503e-01 -4.40159202e-01 -5.93400836e-01 2.27649167e-01 1.91310346e-01 1.35890365e-01 6.22278273e-01 -1.42597973e+00 9.00395274e-01 1.11841524e+00 3.36004019e-01 8.43747854e-01 1.17110527e+00 -8.15813780e-01 -1.14819658e+00 -1.02645850e+00 1.23422277e+00 -2.00549096e-01 2.97505677e-01 -3.49312909e-02 -1.10696685e+00 7.57631898e-01 1.00206590e+00 -9.68569040e-01 4.91879940e-01 -5.94498543e-03 -5.52203357e-02 6.09929077e-02 -6.99773252e-01 8.06439400e-01 8.03475797e-01 -1.68847248e-01 -1.47841811e+00 2.92713940e-01 8.74429345e-01 -4.16661352e-01 -2.88093507e-01 6.86566830e-02 3.48147124e-01 -9.33656394e-01 4.23696667e-01 -6.75379455e-01 8.99091303e-01 -7.71971643e-02 4.58743542e-01 -1.87898922e+00 -2.85741657e-01 -8.39925587e-01 -4.37577695e-01 1.11831510e+00 4.25742060e-01 -4.05631840e-01 6.50139511e-01 2.17853457e-01 -3.50531340e-01 -5.00262678e-01 -6.68458045e-01 -4.58094329e-01 8.96165594e-02 -9.25807934e-03 5.45063615e-01 4.43485677e-01 3.18430811e-01 7.93958604e-01 -7.41846412e-02 -2.23025069e-01 5.39367974e-01 1.80578411e-01 9.18014348e-01 -1.20642292e+00 -2.58795060e-02 -8.61742437e-01 1.86862066e-01 -1.58084285e+00 3.39251339e-01 -8.87990773e-01 1.63999021e-01 -2.45275402e+00 5.77487409e-01 6.54243112e-01 -2.97882059e-03 2.61541158e-01 -3.56827378e-01 -2.17217922e-01 2.83083022e-01 3.01105499e-01 -1.00329423e+00 7.82882154e-01 1.09261882e+00 -4.74772066e-01 -5.03727436e-01 4.45433185e-02 -1.14136446e+00 5.78793406e-01 7.06822276e-01 -3.14086407e-01 -3.29622954e-01 -6.79497242e-01 2.59907693e-01 8.00695736e-03 -1.36369243e-02 -8.57125878e-01 6.68123662e-01 2.01061293e-01 2.10105538e-01 -1.24611068e+00 -1.70942828e-01 -2.02414215e-01 -4.59679961e-01 5.54406106e-01 -9.35671270e-01 3.54942195e-02 1.65418595e-01 5.68144321e-01 -3.29268008e-01 -5.94722331e-01 5.72965264e-01 -1.43262148e-01 -3.37306976e-01 -9.83147696e-02 -7.05786288e-01 1.57036811e-01 4.80385184e-01 -4.56795692e-02 -5.31869352e-01 -5.96530259e-01 -3.06677520e-01 4.36861634e-01 3.60251695e-01 3.82523447e-01 6.94981873e-01 -8.01863194e-01 -1.19451702e+00 -1.74473077e-01 -4.75030160e-03 1.52015209e-01 1.85801074e-01 5.14497936e-01 -7.92030990e-01 6.65938914e-01 5.71019016e-02 -2.09020078e-01 -1.34129155e+00 1.79217085e-01 8.42966959e-02 -7.51733422e-01 -9.42006707e-01 3.63145322e-01 -9.85940620e-02 -1.17845915e-01 9.42537785e-02 -7.01542735e-01 -5.21879673e-01 3.16802293e-01 9.63347852e-01 5.35691321e-01 2.66988486e-01 -3.46364141e-01 1.61080703e-01 2.49826998e-01 -8.53856325e-01 -3.14512044e-01 1.41816521e+00 -2.39789665e-01 -4.31192189e-01 2.99266934e-01 9.02706265e-01 1.52810335e-01 -9.74951684e-01 -3.36419106e-01 2.61590213e-01 6.10502250e-02 -2.66478322e-02 -8.65981698e-01 -4.46348757e-01 5.92337012e-01 -4.22720522e-01 4.59259391e-01 1.04467237e+00 1.12834377e-02 1.15467215e+00 9.49136913e-01 -2.18395948e-01 -1.44635975e+00 1.70357570e-01 9.11281526e-01 1.24146676e+00 -8.74741971e-01 5.20235121e-01 -1.67085513e-01 -7.61087418e-01 1.34949756e+00 2.83922344e-01 -1.84723780e-01 9.75054223e-03 -1.44118324e-01 -3.10163409e-01 -1.84068382e-01 -9.45058048e-01 1.31518573e-01 4.39440906e-01 1.16208836e-01 4.14828569e-01 -1.06170692e-01 -7.25132644e-01 6.00601137e-01 -5.78601420e-01 -1.57487020e-02 1.02483249e+00 1.14450490e+00 -9.41012919e-01 -9.19985533e-01 -1.64331630e-01 7.70698071e-01 -4.73568648e-01 -1.09656572e-01 -9.65913475e-01 4.34825212e-01 -7.11355209e-01 1.01700974e+00 1.82150021e-01 1.04589425e-02 4.99718398e-01 2.04402924e-01 5.52254021e-01 -9.30193901e-01 -8.77004266e-01 -3.96364592e-02 3.20290565e-01 -2.69191384e-01 -2.56556958e-01 -6.62601769e-01 -1.43713725e+00 -2.07036793e-01 -9.99098644e-02 4.80344385e-01 5.97621202e-01 1.06947577e+00 4.44236308e-01 7.62078762e-01 3.84701759e-01 -9.26715255e-01 -9.14150536e-01 -1.42645872e+00 -2.93201745e-01 2.22377822e-01 4.29463118e-01 3.02801996e-01 -3.23407829e-01 3.78733188e-01]
[12.512659072875977, 9.49335765838623]
99247d94-0094-445c-8702-e8fc27db83a7
boosting-video-object-segmentation-via-space
2304.06211
null
https://arxiv.org/abs/2304.06211v1
https://arxiv.org/pdf/2304.06211v1.pdf
Boosting Video Object Segmentation via Space-time Correspondence Learning
Current top-leading solutions for video object segmentation (VOS) typically follow a matching-based regime: for each query frame, the segmentation mask is inferred according to its correspondence to previously processed and the first annotated frames. They simply exploit the supervisory signals from the groundtruth masks for learning mask prediction only, without posing any constraint on the space-time correspondence matching, which, however, is the fundamental building block of such regime. To alleviate this crucial yet commonly ignored issue, we devise a correspondence-aware training framework, which boosts matching-based VOS solutions by explicitly encouraging robust correspondence matching during network learning. Through comprehensively exploring the intrinsic coherence in videos on pixel and object levels, our algorithm reinforces the standard, fully supervised training of mask segmentation with label-free, contrastive correspondence learning. Without neither requiring extra annotation cost during training, nor causing speed delay during deployment, nor incurring architectural modification, our algorithm provides solid performance gains on four widely used benchmarks, i.e., DAVIS2016&2017, and YouTube-VOS2018&2019, on the top of famous matching-based VOS solutions.
['Wenjun Zhang', 'Li Song', 'Rong Xie', 'Wenguan Wang', 'Liulei Li', 'Yurong Zhang']
2023-04-13
null
http://openaccess.thecvf.com//content/CVPR2023/html/Zhang_Boosting_Video_Object_Segmentation_via_Space-Time_Correspondence_Learning_CVPR_2023_paper.html
http://openaccess.thecvf.com//content/CVPR2023/papers/Zhang_Boosting_Video_Object_Segmentation_via_Space-Time_Correspondence_Learning_CVPR_2023_paper.pdf
cvpr-2023-1
['video-object-segmentation', 'video-semantic-segmentation']
['computer-vision', 'computer-vision']
[ 5.68620622e-01 2.00908110e-01 -6.22761607e-01 -2.44499624e-01 -6.87224329e-01 -4.66319114e-01 3.87969315e-01 -1.38933018e-01 -2.64327377e-01 3.95828396e-01 -1.08711012e-01 -2.82739282e-01 -5.18652573e-02 -4.90591109e-01 -9.99854386e-01 -5.33089995e-01 2.47794151e-01 3.88512820e-01 7.64326930e-01 9.48557481e-02 8.13799202e-02 2.61846125e-01 -1.57394409e+00 3.20346981e-01 6.34829760e-01 1.39688885e+00 3.52744967e-01 3.53662997e-01 -1.48771390e-01 7.28820682e-01 -1.72373876e-01 -2.78202921e-01 6.50382936e-01 -4.39215243e-01 -9.90670562e-01 5.74006438e-01 6.87610805e-01 -1.93958893e-01 -6.10175431e-01 9.91406381e-01 3.94359790e-03 -4.24279980e-02 3.53208691e-01 -1.14601994e+00 -7.34545738e-02 5.29163897e-01 -7.47909606e-01 2.00001165e-01 -3.90023626e-02 4.98782784e-01 1.15963888e+00 -8.25663447e-01 7.69573450e-01 9.01275218e-01 6.91153049e-01 6.25092506e-01 -1.37440777e+00 -5.48002005e-01 6.14010453e-01 2.07274169e-01 -1.37886143e+00 -7.47893035e-01 9.00048852e-01 -6.01238012e-01 4.40866768e-01 2.99938709e-01 7.37587988e-01 8.73154104e-01 -1.98993236e-01 1.30449533e+00 9.93934691e-01 -2.28650659e-01 1.69277519e-01 3.40771191e-02 4.64995354e-02 7.46205986e-01 1.31388865e-02 -6.37150854e-02 -6.53443277e-01 3.09309334e-01 7.19171703e-01 -7.36559033e-02 -5.20465374e-01 -6.34160280e-01 -1.26576054e+00 3.95187974e-01 3.17473531e-01 9.49725322e-03 -2.28944123e-01 2.55425751e-01 3.45606863e-01 1.42920479e-01 5.56477308e-01 1.88987121e-01 -6.12383544e-01 9.69579965e-02 -1.55187368e+00 -4.61213775e-02 5.29163480e-01 1.17898452e+00 1.03157341e+00 1.70644075e-01 -4.85565543e-01 4.95518386e-01 4.73449290e-01 6.18207566e-02 1.24497883e-01 -1.24274874e+00 4.76722389e-01 6.93561673e-01 4.81486171e-02 -8.79427016e-01 -2.24581480e-01 -7.03561604e-01 -6.03193223e-01 -5.35761975e-02 5.51721096e-01 8.97185057e-02 -1.19403434e+00 1.58314037e+00 5.55348992e-01 7.71871269e-01 -2.70046026e-01 8.81339252e-01 7.21540093e-01 4.16968703e-01 2.18911376e-02 -3.10995579e-01 1.10049474e+00 -1.30852985e+00 -4.49169546e-01 -3.71088415e-01 5.17836750e-01 -6.39797688e-01 1.12376451e+00 4.08227116e-01 -1.18876910e+00 -6.47191584e-01 -1.11174488e+00 1.51310518e-01 1.29784510e-01 -1.86937097e-02 5.74950218e-01 5.57421029e-01 -1.04233682e+00 8.93749297e-01 -8.93401086e-01 -1.33598804e-01 9.09574747e-01 5.12046695e-01 -3.00769601e-02 -4.92332987e-02 -7.89720833e-01 3.83042216e-01 3.16878170e-01 2.66564608e-01 -9.73795295e-01 -8.91132414e-01 -7.34370410e-01 -1.12817533e-01 1.00292051e+00 -6.54495239e-01 9.99807537e-01 -1.07135439e+00 -1.43767047e+00 9.21470404e-01 -2.50450373e-01 -6.10064626e-01 8.25090289e-01 -1.16823733e-01 -2.32743606e-01 3.37769479e-01 -1.43150026e-02 1.16579711e+00 1.18083680e+00 -1.55519676e+00 -6.42531157e-01 3.32506420e-03 3.06994393e-02 -1.61890462e-02 -3.35068226e-01 -2.62595117e-01 -9.86434281e-01 -6.61413670e-01 4.51460898e-01 -8.42216730e-01 -3.84880960e-01 1.99898735e-01 -6.21721923e-01 -1.84302032e-01 8.83584023e-01 -5.27123332e-01 1.17889130e+00 -2.25694919e+00 1.27337411e-01 2.23801211e-01 4.29632664e-01 4.66478497e-01 -2.05820516e-01 -1.14910610e-01 1.28304213e-01 1.48812085e-01 -4.39218372e-01 -7.56846309e-01 -1.91228632e-02 3.84272009e-01 -3.09724152e-01 6.54408097e-01 3.80857080e-01 1.20501399e+00 -8.17607820e-01 -8.26202810e-01 5.02366602e-01 4.78458330e-02 -8.23596001e-01 3.34453464e-01 -6.32090807e-01 8.50889444e-01 -2.61295378e-01 6.80841327e-01 4.23185050e-01 -5.13141572e-01 9.68696177e-02 -5.13361812e-01 2.50923578e-02 2.42904052e-01 -1.17354131e+00 2.03404665e+00 -2.27632821e-01 6.22954071e-01 2.73452163e-01 -1.29289675e+00 6.52897418e-01 2.02694044e-01 8.95842850e-01 -6.20611012e-01 1.55679643e-01 1.52762488e-01 -1.47272751e-01 -4.36130881e-01 3.15127164e-01 3.02154005e-01 2.25560680e-01 1.58034369e-01 1.67933509e-01 6.12549633e-02 2.63854265e-01 1.88170478e-01 9.87088203e-01 6.26330018e-01 -7.74217322e-02 -4.35677886e-01 4.99065846e-01 3.73308547e-02 8.99722576e-01 7.34327912e-01 -3.21090758e-01 7.57632375e-01 4.04943109e-01 -3.71000528e-01 -7.76315451e-01 -9.97004867e-01 -1.15045041e-01 8.89171422e-01 5.56283534e-01 -4.84158188e-01 -1.02021170e+00 -9.93548095e-01 -1.79309651e-01 2.06111729e-01 -5.08394420e-01 -5.96401021e-02 -7.62282431e-01 -4.02242124e-01 3.48063380e-01 3.92073035e-01 4.86719549e-01 -9.91763890e-01 -5.42649329e-01 2.49572173e-01 -2.10241824e-01 -1.56823993e+00 -6.85368657e-01 2.24339232e-01 -9.52134550e-01 -1.11866808e+00 -5.73677003e-01 -6.91411912e-01 5.60880244e-01 3.89259875e-01 1.15764487e+00 2.74399161e-01 -2.97012448e-01 4.14327770e-01 -6.93270788e-02 6.37148097e-02 -2.23263085e-01 1.42460957e-01 -1.43730313e-01 4.10623789e-01 5.48552424e-02 -7.36644149e-01 -8.59710693e-01 4.70363230e-01 -7.41618395e-01 3.88295829e-01 6.78229094e-01 7.16858327e-01 7.58289337e-01 -1.68683693e-01 5.29264033e-01 -8.76563370e-01 -4.75941300e-01 -3.43462735e-01 -7.68401623e-01 2.60881901e-01 -6.30621552e-01 -1.36075541e-01 3.88710111e-01 -3.91035318e-01 -5.74757338e-01 4.41571265e-01 -1.26508370e-01 -7.68382847e-01 -3.29284728e-01 6.69364110e-02 -3.76582742e-01 -2.85296112e-01 2.33457610e-01 3.33724260e-01 -1.87987447e-01 -4.82299566e-01 3.54338586e-01 3.92022282e-01 9.97503221e-01 -5.43071806e-01 9.59010363e-01 7.59007752e-01 -8.98904260e-03 -5.63918591e-01 -1.05309033e+00 -7.42926538e-01 -8.75368774e-01 -3.95369291e-01 1.05834031e+00 -9.72453117e-01 -7.37519205e-01 2.50828117e-01 -1.07681322e+00 -5.47310770e-01 -5.38889110e-01 1.17437340e-01 -6.98184729e-01 4.77144986e-01 -4.85973150e-01 -5.66002786e-01 -4.87656109e-02 -1.29640388e+00 1.05552340e+00 9.41733941e-02 -2.38841042e-01 -8.28887999e-01 -5.11922121e-01 6.40657842e-01 2.52447665e-01 2.74839938e-01 7.46729910e-01 -4.18664068e-01 -1.18095195e+00 1.78940505e-01 -4.40347165e-01 4.26964581e-01 1.14611268e-01 -1.54054567e-01 -1.11590672e+00 -1.15289837e-01 -5.02573103e-02 -1.88303679e-01 1.12290025e+00 3.66042584e-01 1.40094602e+00 -1.17948249e-01 -3.54683131e-01 8.89317811e-01 1.29416502e+00 4.56714481e-02 3.36627454e-01 1.69727936e-01 1.15042603e+00 6.80364668e-01 6.37275100e-01 2.25430429e-01 3.82522881e-01 7.49888778e-01 7.19724536e-01 -2.96515822e-01 -4.75143820e-01 -2.91887730e-01 3.16760808e-01 6.82989657e-01 2.24249184e-01 -3.00748199e-01 -6.94228113e-01 6.25911534e-01 -2.06101918e+00 -6.81838870e-01 -1.90303117e-01 2.25372291e+00 8.70266259e-01 3.88718188e-01 8.61221254e-02 2.07989693e-01 6.78773165e-01 5.14449596e-01 -7.15226531e-01 2.04499602e-01 3.11058797e-02 1.89932495e-01 5.54938614e-01 4.17141348e-01 -1.43062675e+00 1.27127516e+00 5.54557371e+00 9.59057450e-01 -1.19691980e+00 2.52773017e-01 8.88369262e-01 -2.16035098e-01 -3.76543343e-01 9.22437087e-02 -7.87545025e-01 4.89918143e-01 5.84435403e-01 3.23582947e-01 3.99782658e-01 6.09303117e-01 3.63943905e-01 -1.97378203e-01 -1.45840549e+00 1.07355106e+00 -1.84458867e-01 -1.81025076e+00 -1.34263217e-01 -9.52623487e-02 9.56962705e-01 1.03349604e-01 1.71770528e-02 1.17108360e-01 -1.70514826e-02 -8.70123148e-01 1.10436618e+00 3.16370040e-01 7.20723450e-01 -3.13281089e-01 4.33832526e-01 2.85518050e-01 -1.22038746e+00 1.54568776e-01 -1.75043810e-02 1.06318697e-01 3.89088333e-01 5.85155964e-01 -4.00061101e-01 6.77040756e-01 6.56911314e-01 1.05213821e+00 -5.38352013e-01 1.02144504e+00 -5.32163307e-02 9.48496878e-01 -2.57028371e-01 6.25322759e-01 4.28263724e-01 -1.13416180e-01 5.21747053e-01 1.17505324e+00 -6.18613996e-02 -1.71837613e-01 7.16260195e-01 8.64950895e-01 -2.58796453e-01 -1.28514513e-01 -1.60746142e-01 1.27381757e-01 3.50090683e-01 1.20958853e+00 -1.21427977e+00 -4.68108863e-01 -4.32015866e-01 9.27711904e-01 -3.29676196e-02 3.71529281e-01 -9.88748431e-01 3.84173572e-01 5.59246242e-01 4.11262244e-01 6.52327716e-01 -1.64443314e-01 -5.42951643e-01 -9.53881204e-01 2.13515550e-01 -7.18488276e-01 1.84173718e-01 -7.98589885e-02 -1.01731336e+00 4.55899537e-01 -9.18098390e-02 -1.38350129e+00 8.36694390e-02 -3.04444343e-01 -5.02299249e-01 5.53422570e-01 -1.82560074e+00 -1.07133782e+00 -2.69525021e-01 6.01078629e-01 7.48616099e-01 4.37585004e-02 1.78294569e-01 5.75382650e-01 -7.26098895e-01 5.62759876e-01 -2.95554876e-01 4.61546071e-02 5.09619772e-01 -1.09611821e+00 4.84196723e-01 9.39806581e-01 5.52155554e-01 2.27250278e-01 5.47331333e-01 -5.40899873e-01 -1.28495169e+00 -1.44675684e+00 4.77640778e-01 -3.34774941e-01 6.08763695e-01 -6.15872622e-01 -9.18128312e-01 5.57789326e-01 8.80010501e-02 6.40460372e-01 1.82925344e-01 -2.92576283e-01 -2.11027637e-01 -2.36054018e-01 -7.91161358e-01 6.96829081e-01 1.53361714e+00 -3.55746716e-01 -1.07693210e-01 4.83571082e-01 9.97931063e-01 -6.15396559e-01 -6.37963295e-01 7.42005229e-01 3.00227225e-01 -1.03437161e+00 1.06788969e+00 -4.63379771e-01 3.64111573e-01 -5.92790961e-01 -1.54037639e-01 -5.29514730e-01 -3.97517756e-02 -1.06552446e+00 -4.13522363e-01 1.27726722e+00 2.82252073e-01 -1.79452062e-01 1.38924325e+00 4.27880466e-01 -6.29454374e-01 -1.09168887e+00 -1.04203880e+00 -9.74115252e-01 -3.26068103e-01 -7.37017334e-01 4.57205564e-01 9.35470402e-01 -6.46146476e-01 -1.79503914e-02 -5.84310710e-01 3.98524553e-01 6.95029736e-01 1.49998426e-01 9.38447833e-01 -1.02568090e+00 -7.02121437e-01 -5.06612182e-01 -3.41874003e-01 -1.65448797e+00 1.68795377e-01 -8.93496275e-01 4.12017077e-01 -1.38591504e+00 -6.40054420e-02 -6.28355324e-01 -3.10473859e-01 3.72334361e-01 -2.43010715e-01 7.60080874e-01 3.81609648e-01 4.13706630e-01 -1.07638729e+00 3.96865964e-01 1.34671378e+00 -2.14037076e-01 -2.32548863e-01 6.67261407e-02 -5.02823472e-01 1.00610626e+00 5.12455881e-01 -5.91265082e-01 -5.39745867e-01 -4.99828815e-01 -1.30528837e-01 -3.91610600e-02 5.64585686e-01 -1.16969121e+00 3.99171025e-01 -3.11799645e-01 5.41347079e-02 -4.83667076e-01 3.37378144e-01 -9.62865412e-01 1.44419849e-01 4.13561165e-01 -2.87461609e-01 -3.11238050e-01 6.62485808e-02 7.42389619e-01 -1.71017796e-01 -1.58082590e-01 8.75358820e-01 7.86355734e-02 -1.10582316e+00 7.41417825e-01 -4.52720597e-02 1.54735491e-01 1.20298886e+00 -5.71279347e-01 7.31547400e-02 -6.90758079e-02 -7.86301136e-01 3.18977803e-01 4.39581215e-01 3.52415383e-01 3.66539627e-01 -9.85118389e-01 -3.52453977e-01 1.82568669e-01 4.69236672e-02 5.86388588e-01 3.37932527e-01 1.18899298e+00 -2.21113801e-01 2.06861168e-01 2.76134610e-01 -1.16587520e+00 -1.00792551e+00 6.13655329e-01 1.63194433e-01 -1.54940337e-01 -9.23880577e-01 1.02093971e+00 5.36879659e-01 7.55068064e-02 5.03869116e-01 -3.54722589e-01 7.71685690e-02 -1.95570827e-01 1.64915577e-01 1.00193627e-01 4.66656201e-02 -5.62263608e-01 -2.53251314e-01 6.01155519e-01 -1.52772143e-01 1.64775163e-01 1.02779365e+00 -3.41230899e-01 8.03862587e-02 4.78230298e-01 1.04894900e+00 -6.90109879e-02 -1.96433437e+00 -5.06485879e-01 4.36618567e-01 -5.64830363e-01 5.74796572e-02 -4.87937093e-01 -1.51119089e+00 7.31951833e-01 2.80859739e-01 3.36912759e-02 1.12254703e+00 2.32606724e-01 1.18542063e+00 1.06952153e-03 3.73421788e-01 -1.03541839e+00 2.35818312e-01 1.20229416e-01 3.93754482e-01 -1.36549401e+00 -6.54013380e-02 -8.86098564e-01 -3.22747469e-01 8.11491132e-01 6.52936876e-01 -2.23094553e-01 5.52916646e-01 2.46299967e-01 4.33269776e-02 -9.36744064e-02 -6.65188313e-01 -1.96252078e-01 5.30941784e-01 4.64431047e-01 7.90493041e-02 -3.01078826e-01 1.97648630e-02 3.62993389e-01 1.51255667e-01 -2.33327702e-01 1.49184212e-01 5.99377155e-01 -3.66659433e-01 -1.04988968e+00 -9.05824453e-02 5.22327900e-01 -3.01186383e-01 -1.37392014e-01 -9.69334990e-02 7.87386537e-01 3.89381647e-01 8.17629993e-01 1.58655807e-01 -3.48501712e-01 6.65507391e-02 -1.00139529e-01 4.27494258e-01 -6.58317327e-01 -4.55791980e-01 2.32747674e-01 -7.15114698e-02 -1.02099240e+00 -7.24103749e-01 -8.12520564e-01 -1.24470317e+00 -1.57203928e-01 -3.50246668e-01 -8.94071385e-02 2.76841760e-01 1.42509115e+00 4.89972979e-01 6.68282866e-01 6.12157226e-01 -1.10705364e+00 -2.75202692e-01 -4.17663425e-01 -2.14543194e-01 3.24315757e-01 3.40608567e-01 -6.21284366e-01 -4.58277352e-02 2.71076649e-01]
[9.137992858886719, -0.09593158960342407]
87247fe2-a043-49cc-996a-061fed83564d
combining-evaluation-metrics-via-the
1401.4590
null
http://arxiv.org/abs/1401.4590v1
http://arxiv.org/pdf/1401.4590v1.pdf
Combining Evaluation Metrics via the Unanimous Improvement Ratio and its Application to Clustering Tasks
Many Artificial Intelligence tasks cannot be evaluated with a single quality criterion and some sort of weighted combination is needed to provide system rankings. A problem of weighted combination measures is that slight changes in the relative weights may produce substantial changes in the system rankings. This paper introduces the Unanimous Improvement Ratio (UIR), a measure that complements standard metric combination criteria (such as van Rijsbergen's F-measure) and indicates how robust the measured differences are to changes in the relative weights of the individual metrics. UIR is meant to elucidate whether a perceived difference between two systems is an artifact of how individual metrics are weighted. Besides discussing the theoretical foundations of UIR, this paper presents empirical results that confirm the validity and usefulness of the metric for the Text Clustering problem, where there is a tradeoff between precision and recall based metrics and results are particularly sensitive to the weighting scheme used to combine them. Remarkably, our experiments show that UIR can be used as a predictor of how well differences between systems measured on a given test bed will also hold in a different test bed.
['Enrique Amigó', 'Julio Gonzalo', 'Javier Artiles', 'Felisa Verdejo']
2014-01-18
null
null
null
null
['text-clustering']
['natural-language-processing']
[ 2.17620879e-01 -2.00415492e-01 -9.82370451e-02 -5.66066027e-01 -6.61305428e-01 -7.05708921e-01 7.38159478e-01 6.45835817e-01 -7.40694642e-01 4.60896075e-01 3.46342951e-01 -3.03519785e-01 -7.49507666e-01 -5.47525406e-01 -1.66890025e-01 -5.10780990e-01 1.85429037e-01 3.38462800e-01 2.51810908e-01 -3.63722920e-01 6.96132958e-01 3.91545087e-01 -1.69286966e+00 2.63490260e-01 7.25311637e-01 7.47998595e-01 6.92647770e-02 6.79085076e-01 -1.06371820e-01 5.98597944e-01 -1.07740355e+00 -3.11283082e-01 3.23259771e-01 -4.93225724e-01 -8.58991981e-01 -4.30507138e-02 3.86489093e-01 1.32681161e-01 6.49786415e-03 9.95336711e-01 2.10731715e-01 4.47316580e-02 9.12820518e-01 -1.40082920e+00 -4.08033282e-01 6.76586628e-01 -4.79977638e-01 2.82320529e-01 5.69451272e-01 -9.22595114e-02 1.43110585e+00 -6.33940399e-01 4.20361966e-01 1.19264865e+00 6.61843479e-01 -7.69159347e-02 -1.25282335e+00 -2.51020461e-01 1.25066727e-01 8.13278779e-02 -1.37733245e+00 -3.04563463e-01 3.35617989e-01 -5.25331259e-01 6.80505037e-01 6.61092520e-01 1.21550225e-01 1.05020560e-01 2.67591715e-01 3.20959806e-01 1.13254058e+00 -6.83490276e-01 2.79051393e-01 5.52733839e-01 6.27785385e-01 4.09093797e-01 8.52755070e-01 -3.10037792e-01 -3.35694224e-01 -3.19572300e-01 1.75515801e-01 -1.27608418e-01 -3.90132844e-01 -4.31377858e-01 -1.41789007e+00 6.38668776e-01 4.33776855e-01 8.92481685e-01 -8.84873699e-03 -1.26337141e-01 3.07593524e-01 6.65721238e-01 2.46932209e-01 1.39359665e+00 -4.81282651e-01 -2.35653058e-01 -8.31132352e-01 2.17033610e-01 7.47703910e-01 5.04340708e-01 5.07366538e-01 -3.65321666e-01 -4.28341776e-01 9.43957269e-01 1.01862542e-01 1.83473855e-01 8.54974687e-01 -1.01845157e+00 1.97168157e-01 8.49329650e-01 3.39272112e-01 -1.19456267e+00 -3.90823960e-01 -1.64411664e-01 -2.08501935e-01 3.17570478e-01 6.20160580e-01 -3.81024666e-02 -5.20563900e-01 1.69975102e+00 -1.36952728e-01 -7.02713430e-01 -7.14237392e-02 7.26084769e-01 6.47558868e-01 2.46255487e-01 -1.93139344e-01 -3.42262894e-01 1.13379812e+00 -7.28820443e-01 -6.01039231e-01 -2.11404264e-01 9.35008466e-01 -1.04440081e+00 1.29939055e+00 2.00550303e-01 -1.06133616e+00 -6.10391378e-01 -1.30942333e+00 2.27136374e-01 -5.31261206e-01 -2.04382077e-01 3.15568358e-01 8.19269240e-01 -1.20141494e+00 8.46745431e-01 -4.31781530e-01 -5.97510040e-01 -4.23942000e-01 3.51255238e-01 -2.98051685e-01 1.29454628e-01 -8.00029933e-01 1.35869396e+00 2.87240863e-01 -2.83794492e-01 -2.71513443e-02 -4.36635047e-01 -4.11096066e-01 2.34897271e-01 2.78121263e-01 -2.40843818e-01 1.20720959e+00 -9.06846046e-01 -8.63098264e-01 6.35040402e-01 5.11218309e-02 -3.48615646e-02 4.06885833e-01 1.86255038e-01 -4.06149179e-01 -2.91799992e-01 1.61547273e-01 2.76669621e-01 1.36074975e-01 -1.36589766e+00 -9.10214722e-01 -4.11318153e-01 1.90582305e-01 4.74995524e-01 -5.69583595e-01 2.43947022e-02 -3.41602206e-01 -4.96719509e-01 1.88830286e-01 -8.53601158e-01 -6.44436926e-02 -3.03440630e-01 7.07390136e-04 -4.03159708e-01 3.76084834e-01 -1.59673527e-01 1.67045188e+00 -2.04553032e+00 4.95022684e-02 4.49593782e-01 2.53641456e-01 4.00530994e-02 -3.22075486e-01 4.95359480e-01 -7.82138482e-02 3.39082628e-01 -1.41495436e-01 3.44819456e-01 1.67788446e-01 -2.19548843e-03 2.90226012e-01 2.74044663e-01 -8.15344825e-02 3.29116046e-01 -9.63469744e-01 -2.45082051e-01 1.37648955e-01 1.76552936e-01 -1.17983527e-01 -1.85345355e-02 4.22991753e-01 -4.56723064e-01 -3.28250439e-03 1.74761847e-01 2.91472107e-01 -2.68826008e-01 3.97652298e-01 -2.42893085e-01 -1.27942950e-01 3.67360920e-01 -1.25656044e+00 9.11591887e-01 -3.31978977e-01 8.66385758e-01 -4.63015229e-01 -8.03853571e-01 1.04273689e+00 1.90968439e-01 4.64013875e-01 -8.10556173e-01 7.18255490e-02 3.44729513e-01 4.83507901e-01 -2.17925087e-01 9.10634339e-01 1.97640300e-01 -4.48137783e-02 6.38520837e-01 -3.07854950e-01 -2.37515733e-01 6.19115770e-01 2.38007769e-01 1.38843739e+00 -4.79054093e-01 3.96850318e-01 -6.31343365e-01 4.45100546e-01 1.31388092e-02 4.42787230e-01 9.11449909e-01 -2.86970228e-01 5.64251304e-01 6.05280757e-01 -1.90430284e-01 -1.00803244e+00 -8.75439286e-01 -3.79475087e-01 1.04720891e+00 3.54536891e-01 -6.12612545e-01 -6.18099689e-01 -7.64990091e-01 2.12812513e-01 7.91609943e-01 -6.44026041e-01 -3.23873490e-01 -1.37915149e-01 -1.13050056e+00 3.54214698e-01 4.66271877e-01 1.27095535e-01 -5.98794758e-01 -8.11386228e-01 -1.74038801e-02 -1.14430673e-01 -8.21912289e-01 -2.92557031e-01 9.48779956e-02 -8.45364213e-01 -1.18542504e+00 -5.92734396e-01 -5.71765423e-01 6.98409975e-01 7.60716915e-01 1.32869577e+00 4.85631496e-01 1.36318475e-01 4.61728066e-01 -5.88109851e-01 -3.36177230e-01 -4.25683618e-01 1.38502911e-01 1.88933462e-01 -4.43470955e-01 5.45216084e-01 -3.63318026e-02 -5.02828598e-01 7.27397740e-01 -9.54129398e-01 -4.00437891e-01 5.31737149e-01 6.01316690e-01 1.31732956e-01 4.57913876e-01 5.57447851e-01 -1.02221513e+00 1.30849469e+00 -2.98932344e-01 -1.78849697e-01 7.31916487e-01 -1.51890695e+00 3.49865198e-01 3.42754811e-01 -2.32287198e-01 -7.41698086e-01 -3.67647141e-01 5.03001690e-01 3.95514190e-01 8.44048485e-02 6.93895757e-01 1.70849308e-01 6.28248695e-03 8.74121845e-01 -4.68118608e-01 -1.04749009e-01 -3.25376183e-01 1.34594172e-01 9.60082293e-01 2.13217974e-01 -4.09683228e-01 5.67716479e-01 7.80833587e-02 -2.10747600e-01 -4.02113199e-01 -6.59189463e-01 -6.25675917e-01 -5.10851860e-01 -2.40815401e-01 3.56309593e-01 -4.58378166e-01 -4.42734241e-01 9.59555805e-03 -6.45691216e-01 -1.99141413e-01 -1.66924611e-01 4.83237833e-01 -2.05742374e-01 3.75602037e-01 -2.04092205e-01 -6.27008855e-01 -7.99237862e-02 -1.26491380e+00 6.17679834e-01 1.41315997e-01 -8.40435982e-01 -1.00833261e+00 8.02425221e-02 4.21390116e-01 4.06712621e-01 1.90460071e-01 9.97672379e-01 -6.80376530e-01 2.41552517e-01 -5.32922029e-01 -3.37133646e-01 3.19863200e-01 5.03107846e-01 4.39082414e-01 -6.69103444e-01 -4.12651002e-01 -4.19955747e-03 2.22096592e-01 6.75252020e-01 2.82784522e-01 8.61679792e-01 -2.71248251e-01 -2.02357873e-01 -1.83560580e-01 1.52735496e+00 4.91356671e-01 4.43530858e-01 6.78228736e-01 4.89571542e-01 6.58345103e-01 6.03341997e-01 3.62045281e-02 3.08333427e-01 8.14000964e-01 -1.43061921e-01 6.71308339e-02 5.57679683e-03 2.18923971e-01 3.57270539e-01 1.04249203e+00 -5.91403805e-02 -2.80982286e-01 -1.18691909e+00 4.74439323e-01 -1.84602427e+00 -6.06345475e-01 -2.02772677e-01 2.67828751e+00 5.96069038e-01 5.45446157e-01 3.60497802e-01 6.99225307e-01 7.51356184e-01 -1.24221236e-01 -2.10180685e-01 -6.23077929e-01 8.06824863e-02 -2.82852538e-02 5.87342143e-01 6.16755605e-01 -5.66379607e-01 3.47183168e-01 7.51304150e+00 5.81541717e-01 -7.54957020e-01 -2.62691855e-01 5.87536871e-01 -9.99761894e-02 -2.42599025e-01 -5.04169054e-02 -1.76896989e-01 5.91798961e-01 1.08837318e+00 -7.48268187e-01 1.56308919e-01 4.77604240e-01 -1.51828770e-02 -3.03330719e-01 -1.31792152e+00 6.59700751e-01 3.19897719e-02 -8.63379657e-01 -1.63135946e-01 1.30968124e-01 7.46068954e-01 -1.52750492e-01 2.28802264e-01 1.15234181e-01 6.02995217e-01 -8.96251321e-01 5.15258729e-01 3.76257986e-01 3.06084216e-01 -7.58579433e-01 1.03187335e+00 -4.48912615e-03 -8.06790650e-01 -3.98455970e-02 -4.24666524e-01 -2.73814797e-01 -4.32949930e-01 7.06198335e-01 -9.01557624e-01 2.84885794e-01 7.29865074e-01 4.21623886e-02 -9.77303207e-01 1.09091401e+00 1.55424029e-01 2.54421979e-01 -9.59433243e-03 -2.71524876e-01 1.23629868e-01 -9.41214487e-02 3.30703199e-01 1.26382148e+00 2.56503791e-01 -2.35661894e-01 -1.25331700e-01 3.34750772e-01 3.11845429e-02 3.62786174e-01 -5.44075668e-01 -4.70579751e-02 7.29849815e-01 1.21001828e+00 -1.12799239e+00 -3.12579691e-01 -5.14869630e-01 5.36551237e-01 1.09255992e-01 1.57944143e-01 -4.17703062e-01 -6.79751992e-01 5.14002502e-01 9.95816588e-02 -7.81979188e-02 -1.43136606e-02 -7.00082541e-01 -8.47779930e-01 6.50965646e-02 -9.53798354e-01 3.75406176e-01 -7.56424665e-01 -1.31643260e+00 6.75634801e-01 9.11866426e-02 -1.18587422e+00 -1.01734921e-01 -5.08309245e-01 -3.83466870e-01 5.81053078e-01 -8.59983385e-01 -2.77049392e-01 -4.77927178e-01 1.63667053e-02 2.21510619e-01 -2.03795001e-01 6.41414523e-01 1.17363013e-01 -5.27836978e-01 7.16875553e-01 4.61771458e-01 3.87265272e-02 1.04403305e+00 -1.68455517e+00 2.13325426e-01 7.93536007e-01 2.52205878e-01 6.42934859e-01 1.11269546e+00 -2.13977218e-01 -8.14588189e-01 -6.11105502e-01 9.21219587e-01 -9.64991152e-01 4.89767820e-01 4.14977759e-01 -8.73219967e-01 3.17144215e-01 1.32194459e-01 -5.54168761e-01 8.99292111e-01 5.20699322e-01 -7.17263222e-01 -3.90345544e-01 -1.34093666e+00 5.32337010e-01 7.37155616e-01 -2.92075634e-01 -5.43234110e-01 2.67785162e-01 4.55310643e-01 9.46285427e-02 -1.09066772e+00 5.61565578e-01 8.05717766e-01 -1.10340309e+00 6.70691609e-01 -4.64488029e-01 3.89815360e-01 -4.04910177e-01 -5.00830650e-01 -1.53895080e+00 -6.95472062e-01 1.90631792e-01 3.42945486e-01 1.05491686e+00 7.33154714e-01 -5.09598613e-01 3.64877224e-01 1.16861343e+00 2.59730130e-01 -6.01348341e-01 -4.15501922e-01 -9.68690217e-01 2.15511531e-01 -1.11199811e-01 8.06328893e-01 1.09365404e+00 5.63328624e-01 5.56820035e-01 3.12677264e-01 -2.91265965e-01 3.49230081e-01 -1.81860492e-01 6.90144539e-01 -1.44664550e+00 -9.76794362e-02 -9.67178762e-01 -8.47786009e-01 -2.80709863e-01 -2.88837940e-01 -7.92668343e-01 9.23694372e-02 -1.59621739e+00 2.28678226e-01 -4.32548732e-01 -9.56693411e-01 3.37694913e-01 -4.70502704e-01 3.21645290e-01 5.74218690e-01 5.28459728e-01 -7.43164062e-01 -1.82555676e-01 6.36645675e-01 -1.26637816e-01 -3.83884817e-01 -1.42027840e-01 -1.27498257e+00 7.20241427e-01 7.70801187e-01 -4.11969602e-01 -4.26389366e-01 -5.50818086e-01 4.93790805e-01 -3.07143629e-01 -2.31670603e-01 -1.26798558e+00 2.48269290e-01 -1.42286420e-01 1.58319071e-01 7.63517842e-02 -1.10703580e-01 -8.42761517e-01 1.09616682e-01 3.90120983e-01 -8.37936282e-01 6.69398308e-01 2.04424877e-02 2.94598430e-01 -2.88173407e-01 -5.20578980e-01 7.93623805e-01 -6.14981130e-02 -6.79484665e-01 -4.04194057e-01 -3.68824035e-01 -8.29148740e-02 9.03056026e-01 -4.12344038e-01 -3.89419079e-01 -2.11869374e-01 -2.62537450e-01 1.63305894e-01 7.65357912e-01 5.39893031e-01 1.59074411e-01 -1.31338263e+00 -7.43948400e-01 -4.53164607e-01 5.12271225e-01 -6.48340464e-01 -2.82129645e-01 6.82507873e-01 -4.69815314e-01 4.51130241e-01 -2.22823903e-01 -4.33625758e-01 -1.66952825e+00 3.14797312e-01 3.13316762e-01 -3.70187819e-01 -5.64301573e-02 5.87558210e-01 -1.48745716e-01 -2.95975357e-01 3.15341145e-01 -2.71306664e-01 -8.35980773e-02 3.05333555e-01 5.20498574e-01 5.93277931e-01 5.07889152e-01 -6.11855745e-01 -4.80268568e-01 4.09479111e-01 -3.16687316e-01 -3.91988218e-01 1.12856889e+00 -1.79188013e-01 -1.39314562e-01 7.99392581e-01 9.56964135e-01 -1.94762781e-01 -5.84299028e-01 -3.42477143e-01 3.20414692e-01 -6.64194942e-01 2.37628728e-01 -9.22215343e-01 -8.96128535e-01 4.03502375e-01 7.26698458e-01 8.00814450e-01 1.07555211e+00 -3.05819243e-01 2.64031380e-01 6.15095198e-01 2.90146798e-01 -1.45528364e+00 5.53825609e-02 1.55355483e-01 7.98286498e-01 -1.25791454e+00 3.20850194e-01 -1.95105225e-02 -7.92766929e-01 1.01223993e+00 5.67787290e-01 1.03144966e-01 6.07538223e-01 1.11598954e-01 2.64485836e-01 -1.80443287e-01 -7.85445273e-01 -2.06044868e-01 4.51772869e-01 3.88685107e-01 1.05174887e+00 3.78968120e-01 -1.02187479e+00 1.05411775e-01 -4.74928319e-01 -2.68879533e-01 7.50001252e-01 8.49569619e-01 -6.33995593e-01 -1.11597085e+00 -3.49584550e-01 8.22251618e-01 -3.04676861e-01 2.63715327e-01 -8.77453744e-01 8.28318238e-01 1.40139228e-02 1.26662374e+00 2.09233627e-01 -8.69099259e-01 4.24429119e-01 7.51575083e-02 4.58681077e-01 -6.92185283e-01 -7.18235970e-01 -3.24590176e-01 2.95500383e-02 -3.43088627e-01 -5.72183073e-01 -6.94176853e-01 -1.01100504e+00 -6.79463506e-01 -5.73929131e-01 5.29740632e-01 6.32579684e-01 9.75582242e-01 1.16897933e-01 4.82469320e-01 7.34418929e-01 -8.86003301e-02 -5.46469986e-01 -1.01306868e+00 -5.21110594e-01 7.83068299e-01 1.92739144e-01 -6.36214793e-01 -5.66915214e-01 -2.76731461e-01]
[9.017226219177246, 4.7619309425354]
621db338-31b0-4f05-a2d6-be39f01ec77d
srpol-dialogue-systems-at-semeval-2021-task-5
null
null
https://aclanthology.org/2021.semeval-1.133
https://aclanthology.org/2021.semeval-1.133.pdf
SRPOL DIALOGUE SYSTEMS at SemEval-2021 Task 5: Automatic Generation of Training Data for Toxic Spans Detection
This paper presents a system used for SemEval-2021 Task 5: Toxic Spans Detection. Our system is an ensemble of BERT-based models for binary word classification, trained on a dataset extended by toxic comments modified and generated by two language models. For the toxic word classification, the prediction threshold value was optimized separately for every comment, in order to maximize the expected F1 value.
['Piotr Andruszkiewicz', 'Pawe{\\l} Bujnowski', 'Christian Goltz', 'Zuzanna Bordzicka', 'Katarzyna Beksa', 'Klaudia Firl{\\k{a}}g', 'Joanna Kolis', 'Jaros{\\l}aw Piersa', "Katarzyna Zam{\\l}y{\\'n}ska", 'Micha{\\l} Sat{\\l}awa']
2021-08-01
null
null
null
semeval-2021
['toxic-spans-detection']
['natural-language-processing']
[ 2.57113036e-02 1.40573338e-01 -5.48332810e-01 -1.51559457e-01 -9.02709842e-01 -4.85679865e-01 6.92646742e-01 7.34351039e-01 -7.48925805e-01 9.96742249e-01 5.07701576e-01 -3.57935011e-01 1.16600409e-01 -6.45332694e-01 1.96014307e-02 -4.31995392e-01 -3.91722843e-02 1.46986395e-01 1.74620003e-01 -1.44801453e-01 5.45645773e-01 1.96949035e-01 -9.92032528e-01 7.66738653e-01 8.98246169e-01 8.93508792e-01 -6.07222393e-02 9.23368275e-01 -2.04374641e-01 1.00572240e+00 -1.06377852e+00 -8.60208333e-01 6.62356094e-02 -3.62407029e-01 -5.25569439e-01 -3.01831186e-01 5.00770807e-02 2.47848198e-01 -2.03464553e-01 1.09087825e+00 5.69490314e-01 -2.34739180e-03 1.40876746e+00 -1.18276823e+00 -9.46974397e-01 1.13729477e+00 -7.51334876e-02 4.94188249e-01 3.52461755e-01 3.67754757e-01 1.42248654e+00 -8.96497726e-01 5.70902467e-01 1.27116990e+00 5.80746055e-01 5.92728913e-01 -1.12130785e+00 -6.91220284e-01 5.46327466e-03 3.16486746e-01 -1.45228088e+00 -7.40694776e-02 3.17000329e-01 -9.26845074e-01 1.41745067e+00 3.26683074e-01 5.38975716e-01 1.50896335e+00 9.60105300e-01 5.36303639e-01 9.18501854e-01 -3.08919996e-01 3.03225040e-01 4.70410675e-01 6.69179797e-01 4.05820280e-01 3.25132519e-01 -1.77969784e-01 -7.25019932e-01 -5.87119937e-01 -7.38726020e-01 -3.43182743e-01 -1.43112749e-01 6.39021099e-01 -5.67140758e-01 1.35428548e+00 -3.36754625e-03 3.11893910e-01 -4.87542868e-01 1.85412049e-01 7.68114269e-01 2.11370751e-01 1.29112303e+00 5.21941304e-01 -5.81795990e-01 -1.37877867e-01 -9.25080478e-01 4.09186661e-01 8.13262343e-01 5.96923709e-01 1.53154612e-01 -1.61183506e-01 -9.21389580e-01 7.16424286e-01 7.41173565e-01 5.15375197e-01 7.50693917e-01 -2.22800687e-01 4.36933517e-01 5.93067169e-01 -1.85157195e-01 -6.68796182e-01 -5.80995977e-01 -5.90859652e-01 -4.34024811e-01 -4.01555374e-02 -1.60240918e-01 -4.48216915e-01 -1.01731181e+00 1.05719471e+00 -3.03117186e-01 -1.90535814e-01 6.55297190e-03 1.77273452e-01 9.92307782e-01 6.78904355e-01 7.03683496e-01 -3.59427303e-01 1.20404887e+00 -9.04737294e-01 -1.00175333e+00 -3.38847823e-02 1.12240088e+00 -7.10019946e-01 7.68406570e-01 5.98282218e-01 -4.60181326e-01 -1.58204623e-02 -9.67440605e-01 2.59252250e-01 -9.99277115e-01 -4.34375912e-01 -1.19460419e-01 1.16564167e+00 -6.48178816e-01 4.92701858e-01 1.45042136e-01 -3.46524894e-01 4.86160576e-01 4.39977497e-02 -1.05797745e-01 3.03855240e-01 -1.84937668e+00 1.17811000e+00 4.18311059e-01 -4.85718101e-01 -7.66102791e-01 -6.11868382e-01 -7.82954395e-01 -1.03050731e-01 -3.58333886e-01 -4.03742701e-01 1.21965194e+00 -4.11149979e-01 -1.09663415e+00 8.62701714e-01 1.81083664e-01 -8.69722247e-01 5.91352403e-01 -1.34408340e-01 -4.75485146e-01 -1.01007558e-01 1.75401658e-01 2.36959025e-01 5.50073147e-01 -9.14488435e-01 -5.09180844e-01 -7.79125392e-02 -1.80676639e-01 -7.22244605e-02 -7.70475388e-01 7.00698733e-01 2.10842639e-02 -8.08255672e-01 -9.58682358e-01 -5.78176796e-01 -3.67477477e-01 -3.32087219e-01 -6.95393503e-01 -5.96590161e-01 5.93170643e-01 -9.15469050e-01 1.89060438e+00 -1.89808297e+00 -3.70508552e-01 2.05642134e-01 8.76385123e-02 1.84584409e-01 -2.71201819e-01 4.42719042e-01 -1.79674298e-01 8.28078568e-01 -1.78938374e-01 -7.08906949e-01 7.19531924e-02 -2.13118747e-01 -2.19607025e-01 5.06300628e-01 5.30357584e-02 7.21343637e-01 -8.91263962e-01 -5.62960505e-01 -2.14356035e-01 1.71987653e-01 -3.20270211e-01 7.73241222e-02 -4.91467148e-01 -3.50665629e-01 -3.24618578e-01 4.78550345e-01 4.27816063e-01 3.17980230e-01 -4.21623230e-01 2.53859520e-01 -2.74535954e-01 3.18397760e-01 -4.94739950e-01 9.69260812e-01 -3.75299990e-01 5.62453687e-01 -3.23534995e-01 -3.66181433e-01 9.93278980e-01 4.12836730e-01 6.10422909e-01 -4.44576502e-01 5.30212164e-01 -1.11927651e-01 -2.07127720e-01 -8.00878048e-01 6.17321789e-01 -2.57966608e-01 -4.59755987e-01 5.01368880e-01 -5.91313653e-02 -2.65307099e-01 4.28565681e-01 3.71960342e-01 1.45370269e+00 -5.35292149e-01 4.08152044e-01 -1.86296657e-01 4.20925051e-01 1.16649136e-01 2.90616423e-01 5.74028850e-01 -2.38545612e-01 7.06122994e-01 5.26396930e-01 -4.70650494e-02 -6.77787423e-01 -6.90301120e-01 -3.24012190e-01 1.05798948e+00 -3.90821278e-01 -9.22703385e-01 -5.78809559e-01 -1.22456813e+00 9.53882635e-02 1.63350558e+00 -8.30252528e-01 -3.64755958e-01 -1.05020998e-03 -1.17434657e+00 7.70017505e-01 3.05840727e-02 -1.09394506e-01 -1.18540478e+00 -4.42817777e-01 3.82984161e-01 -7.41253495e-02 -7.51979947e-01 -7.06583202e-01 7.58034885e-01 -2.46186256e-01 -1.03328955e+00 -4.43779290e-01 -2.36984089e-01 8.57519954e-02 -1.75667271e-01 1.00405276e+00 7.17334896e-02 -2.40482330e-01 2.27981098e-02 -8.68053436e-01 -8.59495521e-01 -6.53841436e-01 1.58605233e-01 -2.02841908e-02 -1.60067976e-02 5.13223529e-01 8.56897309e-02 -2.14994222e-01 -2.50539541e-01 -9.94219303e-01 -3.84270996e-01 1.67442128e-01 5.47658920e-01 -7.43659213e-02 -1.46734238e-01 7.60451198e-01 -1.05496335e+00 1.34860861e+00 -1.13518906e+00 1.00526102e-02 1.78158447e-01 -9.12776172e-01 -2.57063389e-01 4.51267153e-01 -5.26105940e-01 -7.45920360e-01 -1.95773095e-02 -5.69707513e-01 2.35439129e-02 1.41396955e-01 9.01136041e-01 -7.48005733e-02 2.66752273e-01 1.00340140e+00 -6.57776445e-02 -5.30469060e-01 -6.62986636e-01 5.72991014e-01 1.50196218e+00 1.01699218e-01 1.42472014e-02 5.61210752e-01 -2.60951102e-01 -5.31016827e-01 -6.88446045e-01 -1.02714694e+00 -7.58469582e-01 -8.12318981e-01 -5.16399086e-01 1.12500834e+00 -8.01637769e-01 -2.30351798e-02 5.74470520e-01 -1.60465848e+00 -1.21353582e-01 -2.90065467e-01 1.92534506e-01 -8.29328671e-02 2.04239875e-01 -5.16781807e-01 -1.16925490e+00 -9.86168861e-01 -5.67840695e-01 5.43878198e-01 -7.21311569e-02 -9.20703709e-01 -1.12844086e+00 5.06529689e-01 2.02514678e-01 2.58901179e-01 3.34594727e-01 1.08057594e+00 -1.34111333e+00 5.03659725e-01 -6.65688157e-01 -1.91945825e-02 5.68969846e-01 -1.65255815e-02 3.05417091e-01 -7.34023929e-01 1.30426437e-01 -2.53789961e-01 -4.08828348e-01 1.11658323e+00 -5.26773445e-02 1.14599049e+00 -5.95190167e-01 -3.13840300e-01 -1.89480931e-01 1.18443882e+00 1.52848035e-01 3.60223979e-01 4.19490159e-01 2.84689784e-01 3.49192381e-01 5.68692803e-01 9.10340667e-01 1.54998377e-01 4.70251083e-01 6.28368497e-01 3.71260256e-01 -1.62797645e-01 -2.32794043e-02 7.43368447e-01 4.52108920e-01 6.32944584e-01 -1.09657931e+00 -9.75924611e-01 7.50696480e-01 -1.59768009e+00 -9.74259675e-01 -6.65953636e-01 1.59901464e+00 8.46948981e-01 4.82858509e-01 1.56073421e-01 2.78792560e-01 6.61835790e-01 3.46914202e-01 -2.92478025e-01 -1.04324532e+00 -2.46364638e-01 1.79651931e-01 6.29698217e-01 5.47937691e-01 -9.18010235e-01 7.48953700e-01 7.76659298e+00 1.02920473e+00 -6.73371375e-01 5.01514077e-01 6.17790639e-01 -2.89782733e-01 -6.59210026e-01 -3.72001410e-01 -8.90489399e-01 8.88882041e-01 1.58052933e+00 -6.80623353e-01 -4.10298347e-01 7.01923549e-01 5.89546680e-01 -2.58890718e-01 -6.18291676e-01 4.80947256e-01 6.11619353e-01 -1.07200158e+00 1.62259623e-01 -1.99177749e-02 6.44855559e-01 3.92350912e-01 1.07410811e-01 4.41582590e-01 4.82415855e-01 -9.88279700e-01 1.13411784e+00 6.62489891e-01 6.77893281e-01 -6.14091992e-01 1.04294109e+00 7.65171468e-01 -5.47627807e-01 -3.07285637e-01 -5.08701280e-02 5.43193817e-02 3.36031824e-01 1.04372942e+00 -7.08904445e-01 1.71811789e-01 6.31438255e-01 9.50500488e-01 -8.98048282e-01 1.37259698e+00 -3.57699633e-01 8.63854051e-01 1.51411891e-01 -7.01304436e-01 3.43556494e-01 2.81731904e-01 7.72358477e-01 2.01844335e+00 2.71163613e-01 -1.32593066e-01 -3.32105532e-03 4.61030483e-01 -2.47320220e-01 6.35292828e-01 -6.30893350e-01 -1.39690891e-01 2.55090296e-01 1.19271946e+00 -3.35801274e-01 -3.67223918e-01 -3.29098135e-01 8.42684746e-01 1.43061206e-02 -1.66690037e-01 -9.75205779e-01 -5.75745463e-01 2.53541768e-01 2.32367009e-01 -3.43288854e-02 1.01817653e-01 -7.99530864e-01 -8.12502027e-01 -5.88559866e-01 -5.39395690e-01 6.84059381e-01 -7.95362175e-01 -1.66839063e+00 8.92868042e-01 2.67319079e-03 -1.13023067e+00 -5.51562533e-02 -7.64264584e-01 -1.01048493e+00 8.66231322e-01 -1.13965416e+00 -9.48144734e-01 2.02587768e-01 1.27359122e-01 8.49742532e-01 -3.72452796e-01 8.50525379e-01 2.80323684e-01 -6.87997758e-01 7.11643100e-01 1.55284733e-01 -1.71129674e-01 9.31220889e-01 -1.18573129e+00 3.54615033e-01 5.67671597e-01 -1.83412790e-01 8.56754929e-02 1.02543235e+00 -1.02920187e+00 -2.38545567e-01 -1.72857320e+00 1.75845981e+00 -6.85248852e-01 1.11897552e+00 -4.49426472e-01 -6.34732425e-01 1.79120585e-01 5.84877610e-01 -5.69710374e-01 1.24647915e+00 -1.79766551e-01 -3.83220971e-01 3.41194391e-01 -1.39428687e+00 4.52733368e-01 4.90436941e-01 -3.18890125e-01 -6.99101329e-01 8.81601691e-01 8.73726368e-01 3.64839345e-01 -8.42944443e-01 1.38205826e-01 1.96395293e-01 -2.58094281e-01 4.04489905e-01 -9.08112168e-01 5.87506652e-01 1.04378983e-01 -2.68222898e-01 -1.67053759e+00 -3.18429470e-01 -4.33610260e-01 -2.16544271e-01 1.33580017e+00 1.06638265e+00 -2.49668166e-01 4.76629168e-01 6.69705033e-01 -9.90293398e-02 -8.18741977e-01 -1.11055160e+00 -7.14230359e-01 4.72931474e-01 -8.46572399e-01 1.37464494e-01 4.98040587e-01 3.59945476e-01 7.37602592e-01 -3.65252107e-01 -2.30806738e-01 3.87682259e-01 -5.94064474e-01 9.98956487e-02 -1.28874981e+00 2.42219403e-01 -5.63944161e-01 -1.93832785e-01 -3.66018951e-01 5.46683669e-01 -1.01307333e+00 2.32389644e-01 -1.53354669e+00 5.84307671e-01 3.29065844e-02 -4.86724377e-01 5.90930283e-01 -1.85243309e-01 6.88052058e-01 2.91851133e-01 -1.38922960e-01 -6.43297970e-01 6.33359313e-01 3.37315232e-01 -5.65949142e-01 1.06661662e-01 2.39033222e-01 -8.90480697e-01 6.86715782e-01 1.10880721e+00 -1.16063631e+00 5.65859489e-02 1.35980293e-01 6.02881253e-01 -5.22510648e-01 -6.06439635e-02 -7.06107438e-01 8.79528895e-02 -3.26109350e-01 3.28783458e-03 -9.50475335e-01 9.52406041e-03 -5.50343752e-01 -2.51235105e-02 7.94975698e-01 -8.05233479e-01 1.93042353e-01 -6.04508407e-02 6.10174179e-01 1.62751704e-01 -9.87736523e-01 9.28626835e-01 -6.12116829e-02 -1.29935667e-01 2.07101628e-01 -1.33711958e+00 -9.26697403e-02 1.24025905e+00 3.08853865e-01 -5.65514445e-01 -3.20199579e-01 -7.42313564e-01 2.66103029e-01 3.04990057e-02 6.18991435e-01 6.28040254e-01 -1.11200857e+00 -1.20898116e+00 -5.83456159e-01 5.73642850e-01 -9.53690410e-01 -1.15377046e-01 5.12523770e-01 -2.05921367e-01 3.15063298e-01 1.59813568e-01 2.67721176e-01 -1.71238708e+00 7.01734185e-01 3.23202938e-01 -6.27783120e-01 -1.90829694e-01 8.16333890e-01 -4.92830992e-01 -1.27799943e-01 2.28668615e-01 -1.69752121e-01 -8.87156844e-01 6.24206960e-01 8.14614177e-01 5.64743638e-01 2.39787906e-01 -8.31488431e-01 -2.21169516e-01 3.09897456e-02 1.56486835e-02 -1.90376043e-01 1.19307351e+00 -1.34166434e-01 -2.82843739e-01 5.61697841e-01 1.31896269e+00 1.74512476e-01 -2.72083193e-01 -6.78602140e-03 3.55488122e-01 2.48727836e-02 2.70789951e-01 -1.29757512e+00 -6.43112719e-01 6.85854793e-01 4.98960108e-01 4.44955915e-01 6.47662759e-01 -8.13831910e-02 7.32958794e-01 5.06234393e-02 -2.51336813e-01 -1.14960051e+00 1.77441955e-01 9.64978397e-01 1.08917809e+00 -9.26484704e-01 1.86066031e-01 -2.82642603e-01 -7.45988011e-01 1.00404763e+00 4.43795979e-01 1.31832764e-01 1.03395605e+00 3.12002420e-01 1.18359722e-01 -2.60349125e-01 -1.23215628e+00 -1.09766722e-01 4.72381979e-01 5.33526242e-01 3.96023273e-01 8.37546438e-02 -1.08133841e+00 1.23462832e+00 -1.31266505e-01 -3.95474136e-01 8.08634579e-01 5.08598804e-01 -7.04213977e-01 -9.00358796e-01 -2.10775405e-01 8.94968152e-01 -9.46584880e-01 -4.77401197e-01 -9.63854969e-01 1.87800586e-01 4.77357388e-01 1.48144460e+00 -2.18478680e-01 -8.18655014e-01 3.83749455e-01 2.66647965e-01 -4.56689268e-01 -1.11850071e+00 -1.36112356e+00 -3.21369171e-01 6.70880854e-01 -1.10206110e-02 1.63399465e-02 -7.62052417e-01 -1.06557727e+00 -7.60706663e-02 -4.99549776e-01 5.06575763e-01 7.75945783e-01 8.89303327e-01 9.22725163e-03 5.63713491e-01 8.80428910e-01 -1.32065430e-01 -7.29731560e-01 -1.53966534e+00 -6.62391305e-01 2.33152658e-01 -4.78909314e-02 -2.23228410e-01 -7.51612306e-01 1.92142531e-01]
[8.970885276794434, 10.644497871398926]
dd5e6407-dedf-46fd-9e5f-f15592bf714d
transformer-based-multilingual-document
2008.08567
null
https://arxiv.org/abs/2008.08567v2
https://arxiv.org/pdf/2008.08567v2.pdf
Transformer based Multilingual document Embedding model
One of the current state-of-the-art multilingual document embedding model LASER is based on the bidirectional LSTM neural machine translation model. This paper presents a transformer-based sentence/document embedding model, T-LASER, which makes three significant improvements. Firstly, the BiLSTM layers is replaced by the attention-based transformer layers, which is more capable of learning sequential patterns in longer texts. Secondly, due to the absence of recurrence, T-LASER enables faster parallel computations in the encoder to generate the text embedding. Thirdly, we augment the NMT translation loss function with an additional novel distance constraint loss. This distance constraint loss would further bring the embeddings of parallel sentences close together in the vector space; we call the T-LASER model trained with distance constraint, cT-LASER. Our cT-LASER model significantly outperforms both BiLSTM-based LASER and the simpler transformer-based T-LASER.
['Wei Li', 'Brian Mak']
2020-08-19
null
null
null
null
['document-embedding']
['methodology']
[ 2.70978548e-03 1.27456546e-01 -3.38963270e-01 -1.96845055e-01 -8.94286931e-01 -2.90359795e-01 8.60189974e-01 -2.19710190e-02 -5.83013237e-01 7.17054844e-01 4.92578715e-01 -8.68390203e-01 4.44897830e-01 -7.43058681e-01 -9.95688379e-01 -5.03319919e-01 2.83906192e-01 4.44397807e-01 3.28577422e-02 -3.04486454e-01 3.64719145e-02 6.17982307e-03 -7.20084965e-01 4.48976994e-01 9.48813856e-01 5.53045452e-01 2.13360712e-01 5.33375442e-01 -4.22337770e-01 6.32944047e-01 -2.84990668e-01 -6.86254978e-01 1.60139233e-01 -5.42849243e-01 -6.99772418e-01 -5.80078065e-01 5.71481645e-01 -4.72625732e-01 -6.81144655e-01 7.05368280e-01 4.91002828e-01 -4.61134352e-02 6.32874250e-01 -9.69956219e-01 -1.36266923e+00 9.59638596e-01 -5.31981587e-01 1.90696388e-01 1.49025530e-01 -1.40832260e-01 1.43171501e+00 -1.36373127e+00 4.98978734e-01 1.30298233e+00 7.85544813e-01 4.47294146e-01 -1.33779860e+00 -4.59990650e-01 1.85167328e-01 3.48606169e-01 -1.07214677e+00 -1.87726378e-01 6.92887962e-01 -1.40962943e-01 1.76351273e+00 2.90506650e-02 5.17154455e-01 1.19051468e+00 8.83403003e-01 8.58167291e-01 1.11944735e+00 -8.13203335e-01 -2.99475580e-01 1.19734943e-01 -6.65622652e-02 8.88708651e-01 1.07807657e-02 1.12904347e-01 -5.77675104e-01 2.13550180e-02 6.32911086e-01 1.40273988e-01 -5.95767647e-02 -2.80773729e-01 -1.51563489e+00 9.89416420e-01 6.44775212e-01 5.38111269e-01 -1.19883165e-01 3.45481843e-01 6.46350980e-01 6.77757740e-01 9.61985171e-01 3.19320351e-01 -5.55495918e-01 -1.37987226e-01 -9.48999286e-01 -7.81459138e-02 4.88726139e-01 9.39132631e-01 8.09762478e-01 9.39727351e-02 -3.46422225e-01 7.37728238e-01 3.35290194e-01 5.21067023e-01 8.45268309e-01 -2.98203230e-01 9.66380715e-01 4.74240154e-01 -4.00260299e-01 -6.16600215e-01 -5.23748249e-02 -5.76504290e-01 -6.68855250e-01 -3.20017040e-01 -2.03501508e-01 -9.00341719e-02 -7.90841162e-01 1.67547321e+00 -1.44459307e-01 -9.39416960e-02 1.61944568e-01 5.60142100e-01 4.24975693e-01 1.05284274e+00 -3.32569510e-01 6.09747507e-02 9.71220911e-01 -1.68203104e+00 -7.93464601e-01 -2.31236145e-01 1.24702525e+00 -8.92050982e-01 1.33773303e+00 -2.66485423e-01 -1.02637923e+00 -4.62837666e-01 -1.18442523e+00 -6.43602073e-01 -5.19436479e-01 3.54512483e-01 4.55004424e-01 1.31409228e-01 -1.32259512e+00 5.61476231e-01 -1.04247010e+00 -4.13843215e-01 -1.12237725e-02 3.00269902e-01 -3.29566866e-01 -2.23893628e-01 -1.51500261e+00 1.55605233e+00 9.39339846e-02 2.20464095e-01 -5.78406930e-01 -5.78997612e-01 -1.06609941e+00 1.52199075e-01 -2.72776693e-01 -8.57299209e-01 1.27241182e+00 -6.28688097e-01 -1.83901441e+00 5.90944827e-01 -6.10802054e-01 -4.78047609e-01 3.54672194e-01 -3.19937080e-01 -2.07173288e-01 -1.76264271e-01 1.61619753e-01 7.14227915e-01 6.23492539e-01 -8.29920948e-01 -2.25345090e-01 -2.25410715e-01 -4.11853269e-02 2.53003389e-01 -7.65118122e-01 7.49314949e-02 -1.20206833e-01 -7.57472634e-01 -1.57027930e-01 -1.06585979e+00 -1.15038827e-02 4.47309203e-02 -3.17651480e-01 -5.78591704e-01 7.63118565e-01 -9.17958677e-01 1.41986573e+00 -1.81805134e+00 5.00965536e-01 -3.00291687e-01 -3.75566706e-02 2.36314520e-01 -5.63719451e-01 1.06926155e+00 1.23713464e-02 1.62663102e-01 -2.21822500e-01 -7.76780069e-01 2.06376538e-01 4.44124490e-01 -6.05222046e-01 1.75206885e-01 4.70618397e-01 1.31128943e+00 -8.50198388e-01 -4.79717702e-01 -5.19906636e-03 6.34810567e-01 -5.64721465e-01 4.53164876e-02 -2.07516968e-01 9.47787985e-02 -9.23986360e-02 2.52255082e-01 3.91791582e-01 1.16388373e-01 1.64241016e-01 -1.00813977e-01 -4.27389503e-01 1.04266346e+00 6.87182322e-02 2.14808178e+00 -9.98916566e-01 9.38752234e-01 -4.69208717e-01 -8.33688378e-01 8.92065406e-01 6.01525664e-01 9.26940888e-02 -8.97313833e-01 -9.37047154e-02 5.72706938e-01 5.51235378e-02 -1.76886678e-01 6.43394530e-01 -1.76333964e-01 9.11724344e-02 8.49773288e-01 1.05601899e-01 6.48581833e-02 6.73709288e-02 2.41875574e-01 9.65791106e-01 5.70256889e-01 -3.48504066e-01 -2.35040963e-01 2.98121393e-01 -3.03464741e-01 4.59664881e-01 2.54982382e-01 1.86506137e-01 2.50468910e-01 2.74296045e-01 -4.39220279e-01 -1.42526793e+00 -1.05271673e+00 -2.83683278e-02 9.43778634e-01 -3.14381301e-01 -4.27087367e-01 -5.40317118e-01 -8.58899951e-01 -4.55757193e-02 8.75856757e-01 -5.62254667e-01 -3.58306170e-01 -1.02446055e+00 -5.60863078e-01 7.11979449e-01 6.99021459e-01 3.70987862e-01 -7.65942991e-01 -4.65232395e-02 3.67349684e-01 -5.69971263e-01 -7.91830242e-01 -1.18264771e+00 1.64242223e-01 -1.17589521e+00 -3.66495192e-01 -7.00230777e-01 -1.23205018e+00 7.83363044e-01 1.45861164e-01 9.16791260e-01 -5.88282496e-02 1.39422491e-01 -3.05780858e-01 -3.66859853e-01 -3.23043197e-01 -3.77054423e-01 4.70913112e-01 5.75220808e-02 -2.93262869e-01 6.43983543e-01 -6.71317041e-01 -3.16212684e-01 -5.84537834e-02 -6.74548566e-01 2.09703997e-01 7.32753873e-01 1.25809836e+00 3.12724769e-01 -4.71041441e-01 3.82622570e-01 -4.09405053e-01 9.30701733e-01 -9.13547426e-02 -1.37025461e-01 3.59542340e-01 -9.34610605e-01 4.83214080e-01 6.82487786e-01 -5.98176897e-01 -6.34307921e-01 -5.27712464e-01 -2.09347233e-01 -3.98194522e-01 7.95172095e-01 7.69204199e-01 1.54235631e-01 1.28926486e-01 2.13423163e-01 5.07775128e-01 5.65993860e-02 -7.15681374e-01 4.71383780e-01 7.92751133e-01 6.54375851e-02 -4.28772360e-01 7.82919645e-01 -4.29693423e-02 -2.27187887e-01 -4.57069188e-01 -7.44642258e-01 -7.96108246e-02 -7.13573992e-01 3.09579492e-01 8.15435171e-01 -8.19749832e-01 -1.18089251e-01 2.92883605e-01 -1.75436115e+00 -3.97912234e-01 -1.36862099e-01 8.30959797e-01 -4.09664810e-01 3.47158045e-01 -1.28676045e+00 -3.95163476e-01 -8.18808734e-01 -1.12890589e+00 1.09734702e+00 -4.64736700e-01 -1.03521876e-01 -1.33866334e+00 3.71523291e-01 3.05783525e-02 6.76655352e-01 -2.27793500e-01 1.34949362e+00 -3.60247672e-01 -3.21961701e-01 -2.62266070e-01 -1.84646830e-01 7.24636078e-01 1.24644153e-01 -8.42433199e-02 -6.08540833e-01 -6.59969091e-01 -2.50724778e-02 -1.33586109e-01 1.08280742e+00 5.17274514e-02 6.04351342e-01 -5.85351646e-01 -3.27670306e-01 6.00335836e-01 1.26178575e+00 -1.84061192e-02 5.95540404e-01 4.93559510e-01 8.17085385e-01 2.47596279e-01 2.90977001e-01 -3.11992884e-01 8.43399107e-01 8.96500945e-01 9.78314206e-02 -5.24458528e-01 -1.72943443e-01 -7.12699831e-01 1.01682317e+00 1.95575321e+00 8.34608003e-02 -1.92729279e-01 -6.94761038e-01 5.79014122e-01 -1.88787472e+00 -6.24474943e-01 3.94964442e-02 2.00557995e+00 9.99209523e-01 2.37288892e-01 -2.38168910e-01 1.56795289e-02 4.25201654e-01 1.53054476e-01 -2.10238114e-01 -1.16345704e+00 5.67134749e-03 4.55748171e-01 6.11945570e-01 7.52585649e-01 -8.26894939e-01 1.04456258e+00 6.19218826e+00 5.44313431e-01 -1.38654804e+00 5.48877418e-01 4.39445972e-02 -1.28101543e-01 -6.19983196e-01 2.44062245e-01 -9.01253104e-01 5.08307099e-01 1.20046186e+00 -1.34480193e-01 2.99926043e-01 2.48191327e-01 2.09672734e-01 4.90472853e-01 -1.48107326e+00 4.50035602e-01 2.97124267e-01 -1.24364650e+00 5.89829504e-01 3.08575392e-01 5.29035628e-01 3.54744136e-01 1.56087890e-01 5.79118252e-01 1.46923840e-01 -7.32599378e-01 7.57097602e-01 3.54810208e-01 1.03180182e+00 -7.86939919e-01 8.45856369e-01 3.69083673e-01 -1.02335107e+00 1.12212121e-01 -6.19287133e-01 -1.20192833e-01 3.15198481e-01 4.23878193e-01 -7.47637331e-01 7.70640373e-01 3.10645044e-01 9.04142022e-01 -3.46406221e-01 4.96719867e-01 -5.01387477e-01 6.93853438e-01 -2.89113950e-02 -2.06839263e-01 6.29542351e-01 -3.87252897e-01 4.70458806e-01 1.30780101e+00 5.93140066e-01 -6.80190027e-01 7.90986270e-02 8.30654562e-01 -2.64075994e-01 -9.29154977e-02 -5.67409515e-01 -3.05496901e-01 4.27116692e-01 9.09314096e-01 1.60310879e-01 -2.03427762e-01 -6.58702254e-01 1.40010762e+00 6.59976482e-01 3.85761350e-01 -7.86578238e-01 -6.83286011e-01 5.87877095e-01 8.03397819e-02 3.67281526e-01 -4.74481165e-01 -6.17033690e-02 -1.42278838e+00 3.23921949e-01 -6.66817486e-01 -2.45651037e-01 -6.84805691e-01 -1.28124392e+00 9.46249723e-01 -3.87768388e-01 -1.22304094e+00 -3.21090579e-01 -6.33688450e-01 -6.65267885e-01 1.30827618e+00 -2.01724935e+00 -1.68836451e+00 6.55103862e-01 1.99846700e-01 5.97359300e-01 -1.16458893e-01 1.30828202e+00 4.88872468e-01 -6.80893004e-01 1.00086582e+00 4.68123019e-01 3.67084980e-01 9.28014815e-01 -1.25030005e+00 9.47748721e-01 8.49399328e-01 1.88086614e-01 9.73247588e-01 2.82647729e-01 -4.46510613e-01 -1.43765831e+00 -1.33916938e+00 1.86961961e+00 -5.23764670e-01 8.97382557e-01 -7.45903969e-01 -7.76528180e-01 1.12461317e+00 6.18910313e-01 -3.41110051e-01 7.24884212e-01 7.94774517e-02 -7.31093943e-01 -2.14774713e-01 -6.82456493e-01 6.98953807e-01 6.46807194e-01 -1.06463802e+00 -8.66434872e-01 4.15753603e-01 1.26598942e+00 -9.24708545e-02 -9.05598700e-01 2.16230348e-01 5.78795254e-01 -3.46096486e-01 6.72925770e-01 -5.06096363e-01 8.32556486e-01 5.98229282e-02 4.45630625e-02 -1.68124628e+00 -4.03021753e-01 -4.16957915e-01 -2.60964781e-01 1.04547155e+00 8.37501884e-01 -1.02619624e+00 2.86288381e-01 -7.06670049e-04 -5.00838339e-01 -1.25391555e+00 -1.29123604e+00 -1.04788446e+00 8.45185816e-01 9.36475024e-02 5.00946224e-01 9.57358778e-01 3.77959877e-01 7.21119523e-01 -4.55724835e-01 -3.51272613e-01 4.07631099e-01 -1.47453863e-02 3.44043911e-01 -8.00650120e-01 -3.28374445e-01 -4.88938630e-01 -1.39665022e-01 -1.49089646e+00 2.67860442e-01 -1.44579446e+00 -1.18765287e-01 -1.96848583e+00 2.02544123e-01 -2.62154013e-01 -2.95164764e-01 6.39070034e-01 -1.24885984e-01 2.76154101e-01 2.40386963e-01 3.35481375e-01 -5.59581704e-02 1.14436245e+00 1.41237986e+00 -3.17750871e-01 1.02509826e-01 -3.52503240e-01 -3.54044378e-01 1.50412634e-01 6.42751873e-01 -5.58853805e-01 -1.85127661e-01 -1.26051545e+00 2.33642459e-01 -2.45525047e-01 -6.83849305e-02 -3.32196325e-01 1.39788061e-01 3.27982008e-01 1.26227453e-01 -5.47561169e-01 4.54015136e-01 -5.13104796e-01 -5.68302035e-01 6.22235298e-01 -5.08448303e-01 8.07418764e-01 1.21727452e-01 2.48503178e-01 -3.44867200e-01 1.00282766e-01 3.74359876e-01 3.35601754e-02 1.68920442e-01 3.04694861e-01 -5.16715586e-01 -2.87306100e-01 5.53642869e-01 -6.76341578e-02 -4.41611111e-01 -1.33241475e-01 -2.57601261e-01 3.57868135e-01 2.55714506e-01 7.47472703e-01 6.24131560e-01 -1.71698070e+00 -9.32357848e-01 2.89230555e-01 3.17580476e-02 -2.68728673e-01 -3.96087468e-01 1.11628580e+00 -4.16349828e-01 1.00379705e+00 1.33846745e-01 -3.29191804e-01 -1.07334042e+00 4.77251679e-01 1.95679918e-01 -7.25607038e-01 -6.95991695e-01 7.76265144e-01 -2.83715948e-02 -9.60166216e-01 2.51867268e-02 -5.23456097e-01 3.28442842e-01 -3.67002964e-01 2.33932137e-01 2.24893659e-01 2.42442980e-01 -5.75890183e-01 -2.32710317e-01 4.84416246e-01 -4.22677934e-01 -3.64012182e-01 1.28927433e+00 -1.48429662e-01 -6.81583941e-01 6.98280692e-01 1.48364055e+00 -1.73635811e-01 -7.91734695e-01 -6.74566984e-01 9.01671201e-02 -1.15601838e-01 1.00160368e-01 -7.16304421e-01 -6.24936104e-01 1.48826540e+00 1.93165749e-01 -3.46902162e-01 7.53403366e-01 -4.37627286e-01 1.38362813e+00 4.70858872e-01 3.03163946e-01 -9.55272734e-01 1.09811045e-01 1.02323604e+00 8.94455731e-01 -8.09182227e-01 -3.33045363e-01 9.68368202e-02 -1.89690217e-01 1.19988763e+00 5.48821092e-01 -6.61155805e-02 4.16450918e-01 2.70056754e-01 2.05025718e-01 2.56436527e-01 -1.13523352e+00 3.46680701e-01 3.24571371e-01 1.62171751e-01 9.07187283e-01 9.27604642e-03 -7.14256167e-01 1.42411232e-01 -3.93608809e-01 -2.11655647e-01 3.51661235e-01 8.91653955e-01 -2.07211688e-01 -1.62164760e+00 1.17669031e-01 4.27720785e-01 -2.99895912e-01 -7.69549251e-01 -4.31567580e-01 4.18160349e-01 -7.85922185e-02 7.72449613e-01 2.47370437e-01 -6.61012053e-01 2.59102881e-01 4.54584956e-01 6.64639413e-01 -7.41846383e-01 -6.29040003e-01 -4.22765240e-02 1.69994887e-02 -2.58632749e-01 -1.29111677e-01 -3.69128734e-01 -1.24402225e+00 -4.14130986e-01 -6.58537149e-01 3.65441322e-01 9.93528128e-01 9.88810897e-01 5.33842385e-01 6.00403368e-01 7.28530645e-01 -8.04158032e-01 -1.04863536e+00 -1.47680366e+00 -2.03865051e-01 -8.27774853e-02 5.36249161e-01 -3.14796239e-01 -2.82129228e-01 -2.08936229e-01]
[11.551376342773438, 10.037117004394531]
7bc2b8a6-834f-4d16-bbab-904ae9e0ddc9
an-enactivist-account-of-mind-reading-in
2111.06179
null
https://arxiv.org/abs/2111.06179v5
https://arxiv.org/pdf/2111.06179v5.pdf
An Enactivist account of Mind Reading in Natural Language Understanding
In this paper we apply our understanding of the radical enactivist agenda to the classic AI-hard problem of Natural Language Understanding. When Turing devised his famous test the assumption was that a computer could use language and the challenge would be to mimic human intelligence. It turned out playing chess and formal logic were easy compared to understanding what people say. The techniques of good old-fashioned AI (GOFAI) assume symbolic representation is the core of reasoning and by that paradigm human communication consists of transferring representations from one mind to another. However, one finds that representations appear in another's mind, without appearing in the intermediary language. People communicate by mind reading it seems. Systems with speech interfaces such as Alexa and Siri are of course common, but they are limited. Rather than adding mind reading skills, we introduced a "cheat" that enabled our systems to fake it. The cheat is simple and only slightly interesting to computer scientists and not at all interesting to philosophers. However, reading about the enactivist idea that we "directly perceive" the intentions of others, our cheat took on a new light and in this paper look again at how natural language understanding might actually work between humans.
['Peter Wallis']
2021-11-11
null
null
null
null
['formal-logic']
['reasoning']
[ 2.39941522e-01 9.60739791e-01 2.94343770e-01 -3.35637420e-01 -2.45759003e-02 -8.39046061e-01 1.17333126e+00 -2.73855682e-02 -2.23314986e-01 6.12441421e-01 2.11596116e-01 -9.08296764e-01 1.89232796e-01 -1.02642179e+00 -5.16207933e-01 -2.09112421e-01 3.76678407e-01 7.77727067e-01 3.74742776e-01 -9.91410673e-01 5.32918632e-01 1.28422678e-02 -1.16246819e+00 3.48966897e-01 4.27152932e-01 1.98742598e-01 -3.06941513e-02 7.31946111e-01 -2.78334945e-01 1.94819057e+00 -7.44052172e-01 -5.09658217e-01 6.72731325e-02 -8.31737757e-01 -1.87102365e+00 -2.93149725e-02 -1.96869355e-02 -1.35718629e-01 -2.73170590e-01 1.24950325e+00 -3.99166048e-01 -1.40125155e-01 2.93861091e-01 -1.32948565e+00 -8.52298200e-01 1.05014849e+00 -1.05427086e-01 -3.43592726e-02 8.66761148e-01 5.25538981e-01 7.28324175e-01 -2.17020418e-03 4.44969803e-01 1.59500098e+00 5.88480771e-01 8.07475865e-01 -1.26840127e+00 -2.81990558e-01 -4.02330281e-03 1.50465444e-01 -1.32046139e+00 -4.01599348e-01 4.40738410e-01 -6.50545120e-01 1.01675022e+00 5.25378406e-01 6.47099912e-01 8.04627657e-01 3.21509898e-01 5.49556017e-01 1.54537416e+00 -7.64234722e-01 5.07205278e-02 5.93109131e-01 6.77413702e-01 8.09805512e-01 3.05320889e-01 9.47892293e-02 -4.73293036e-01 -2.73745596e-01 7.14408576e-01 -2.55902678e-01 -2.63222575e-01 1.49398312e-01 -1.67105627e+00 9.30747330e-01 1.51557207e-01 9.13768470e-01 -2.35445887e-01 3.42687786e-01 3.72114778e-01 9.07207072e-01 -1.16955005e-01 7.55680561e-01 -1.52889654e-01 -3.96048576e-01 -3.98216814e-01 2.03453034e-01 1.40333867e+00 5.17738760e-01 4.60867465e-01 -2.13496387e-01 6.76838100e-01 9.11747012e-03 5.36521494e-01 2.69891500e-01 7.12324440e-01 -1.43013048e+00 -1.54063404e-01 7.52659142e-01 1.58198997e-01 -1.41432512e+00 -2.54893750e-01 7.35137165e-02 -4.28953618e-01 5.54520309e-01 9.65180576e-01 -9.45182592e-02 -2.18611956e-01 1.54636753e+00 2.35790126e-02 -1.25211835e-01 4.35168296e-01 1.09660220e+00 4.22840804e-01 6.96329236e-01 -5.37909865e-02 3.37315612e-02 1.74074221e+00 -6.08402669e-01 -6.05002761e-01 -4.29776818e-01 9.88499224e-01 -5.37654340e-01 1.16840208e+00 6.13178313e-01 -1.30109143e+00 -1.94676816e-01 -1.23797774e+00 -3.54057252e-01 -4.46484774e-01 -6.30234122e-01 1.06329000e+00 6.51905358e-01 -1.03888464e+00 5.35079002e-01 -6.24972105e-01 -6.24597669e-01 -7.86545128e-02 3.54825377e-01 -2.24877894e-01 3.63670647e-01 -1.32165003e+00 1.38301468e+00 5.03688455e-01 -7.36371279e-02 -5.24623811e-01 1.25960827e-01 -7.50519156e-01 -9.31314155e-02 6.62162840e-01 -8.90518904e-01 1.64779162e+00 -1.88973534e+00 -1.87197840e+00 1.33017433e+00 -8.77841115e-02 -9.19348121e-01 4.12280768e-01 7.05227926e-02 -4.30687457e-01 7.38179758e-02 7.28812963e-02 3.33195627e-01 3.99297327e-01 -1.24976027e+00 -8.96363333e-02 -2.89799243e-01 8.26064587e-01 -1.61227345e-01 5.81927299e-01 3.06394190e-01 5.15913963e-01 3.17464657e-02 3.66984159e-01 -6.78233683e-01 1.55462861e-01 -6.00002483e-02 -1.49900541e-01 -3.47901583e-01 4.32836503e-01 -2.21410364e-01 9.14944768e-01 -2.05233312e+00 -1.61882773e-01 3.21403533e-01 7.43820846e-01 2.32736230e-01 3.55102837e-01 7.03943491e-01 -1.37341738e-01 3.70480418e-01 2.44892910e-02 5.94922543e-01 5.34003913e-01 3.65469575e-01 -5.89998126e-01 3.65887135e-01 -1.18264630e-01 9.27495956e-01 -1.05805540e+00 -4.50202495e-01 1.23647921e-01 1.27027720e-01 -4.59230512e-01 -6.46661893e-02 -2.78943568e-01 2.44910389e-01 -4.89756137e-01 8.67540017e-02 1.71047002e-01 -8.98402095e-01 5.52227974e-01 4.54626232e-01 -2.45584786e-01 7.98746884e-01 -9.04847920e-01 1.26997268e+00 -1.54175952e-01 9.33084965e-01 2.34846443e-01 -1.04624653e+00 6.53228045e-01 5.50587595e-01 -3.73274386e-01 -3.73156518e-01 4.67933774e-01 1.58640698e-01 8.45117807e-01 -5.52081466e-01 3.32802385e-01 -8.30317974e-01 -1.63161978e-01 1.04688358e+00 -5.68618774e-01 -7.52866149e-01 -2.91583657e-01 5.03538787e-01 1.05201709e+00 6.46675006e-02 7.56067991e-01 -4.49061066e-01 9.36211050e-01 7.17663765e-01 9.48818922e-02 8.42033386e-01 -1.92656025e-01 3.31463665e-02 5.47816753e-01 -1.01659727e+00 -6.75696671e-01 -9.02051449e-01 3.31495643e-01 9.69346583e-01 1.70366868e-01 -5.73956609e-01 -1.02926958e+00 -1.51578024e-01 -4.43999141e-01 1.40911746e+00 -2.69905269e-01 -2.30830461e-01 -4.77335066e-01 2.08219700e-02 8.38992655e-01 -1.02989197e-01 7.67316282e-01 -1.02965009e+00 -1.20536518e+00 1.22391522e-01 -2.62481749e-01 -8.39358687e-01 1.69576630e-01 2.94519104e-02 -5.42361736e-01 -8.18012774e-01 -1.06525578e-01 -5.28761446e-01 5.36214948e-01 2.87548274e-01 1.22210097e+00 7.59612799e-01 1.38735667e-01 4.76440668e-01 -2.84985304e-01 -8.44496846e-01 -1.25902462e+00 -3.92288953e-01 -1.64966509e-02 -3.79826665e-01 8.90367806e-01 -5.93504846e-01 -1.27775878e-01 3.03506970e-01 -8.52101684e-01 5.11988819e-01 8.10706690e-02 5.07025003e-01 -5.01398861e-01 1.89000461e-02 1.09861240e-01 -1.05371332e+00 5.94829321e-01 -1.93751857e-01 -2.50289440e-01 2.25251883e-01 -1.80647060e-01 2.47341126e-01 7.83125341e-01 -2.67423749e-01 -8.38995159e-01 -6.10366344e-01 2.55114555e-01 3.76934409e-01 -3.76423776e-01 1.94740191e-01 9.10937190e-02 5.13148829e-02 8.32080960e-01 1.52068332e-01 3.12746555e-01 3.50362174e-02 3.11685294e-01 8.38156581e-01 5.97384393e-01 -9.03605282e-01 8.21256578e-01 5.27138412e-01 -1.62591875e-01 -1.00588572e+00 -6.64462805e-01 6.47742748e-02 -2.21640766e-01 2.47171968e-02 7.95060158e-01 -6.43521786e-01 -1.48728883e+00 1.01578243e-01 -1.59774399e+00 -5.49368918e-01 -5.22957504e-01 3.12473744e-01 -8.12849462e-01 6.58884585e-01 -4.05600965e-01 -9.63158548e-01 1.44145906e-01 -8.80956709e-01 4.39859480e-01 5.40109177e-04 -1.13731420e+00 -1.00224984e+00 3.55647248e-03 7.24374890e-01 5.67120552e-01 -1.61893681e-01 1.08326638e+00 -9.29186285e-01 -4.25018281e-01 -3.94710042e-02 -9.82574522e-02 3.25700343e-01 1.78664163e-01 -2.48266369e-01 -8.69696558e-01 1.83437958e-01 5.04024923e-01 -4.10476685e-01 -1.29620343e-01 -3.43292981e-01 4.95767117e-01 -6.16124868e-01 -1.04183055e-01 -2.79261172e-01 1.12948239e+00 4.38895911e-01 1.00142801e+00 4.51279581e-01 1.70707956e-01 8.62187862e-01 1.29360612e-02 -2.35232729e-02 7.09055424e-01 4.73538399e-01 -2.13716432e-01 1.09890243e-02 2.54759103e-01 -2.55790681e-01 5.16226530e-01 6.44743621e-01 -4.14772749e-01 4.04921435e-02 -1.42480266e+00 5.67619354e-02 -1.78353477e+00 -1.38673413e+00 -4.11111653e-01 1.87258196e+00 9.13321793e-01 4.91746038e-01 -2.99983956e-02 1.39873207e-01 6.55791104e-01 -1.58181757e-01 5.22818565e-02 -1.17443180e+00 5.31531014e-02 5.09014986e-02 1.38363585e-01 8.76487553e-01 -4.47218895e-01 1.00670540e+00 6.39288235e+00 1.44834787e-01 -9.98166859e-01 1.55477077e-01 3.15537930e-01 4.74682212e-01 -3.29077125e-01 6.36570871e-01 -4.14197594e-02 2.81197906e-01 1.21947563e+00 -4.77519661e-01 7.79057384e-01 3.61187190e-01 2.20456347e-01 -6.23832166e-01 -1.44378448e+00 6.98416710e-01 5.26858456e-02 -1.20281291e+00 6.43700659e-02 6.59784675e-02 -1.90027636e-02 -3.13685715e-01 -4.33443218e-01 1.43931881e-01 8.66562486e-01 -1.48699534e+00 7.92781949e-01 6.95038974e-01 -1.50262825e-02 -2.50609040e-01 7.94136405e-01 9.82532442e-01 -4.03277934e-01 1.21507369e-01 -1.77287400e-01 -1.09734380e+00 8.74077678e-02 -2.66255647e-01 -6.31767213e-01 9.86723006e-02 -4.87897471e-02 -8.43949020e-02 -3.18005204e-01 3.06199372e-01 -1.29827499e-01 3.85857582e-01 -4.30756032e-01 -4.08273935e-01 4.97825921e-01 -3.02813619e-01 4.80400443e-01 8.47647488e-01 -1.54824361e-01 8.13854277e-01 -2.83357576e-02 1.15588200e+00 4.46852773e-01 -3.92165273e-01 -8.75731826e-01 -3.78913999e-01 1.50163826e-02 5.47503531e-01 -6.82485521e-01 -7.28177488e-01 -5.38514197e-01 8.29549253e-01 -2.45048463e-01 -6.19803108e-02 -6.77962780e-01 -2.07515880e-01 4.16096330e-01 5.03116548e-01 -4.36033696e-01 -3.91696215e-01 -1.74799070e-01 -1.29707694e+00 -4.53278035e-01 -1.39190805e+00 -8.89835954e-02 -1.21617734e+00 -1.04932666e+00 3.71487349e-01 1.07302509e-01 -6.77359223e-01 -5.55180728e-01 -7.40465462e-01 -8.50346386e-01 7.55376101e-01 -7.96619296e-01 -1.03029978e+00 8.23151022e-02 4.22866315e-01 4.14993733e-01 9.30739939e-02 1.10388303e+00 -4.92300570e-01 1.50211558e-01 5.87533079e-02 -5.35537839e-01 1.93967894e-01 1.94937497e-01 -9.38421607e-01 4.75055337e-01 4.63560730e-01 5.00200570e-01 1.29580879e+00 1.26188123e+00 -8.34583044e-02 -1.60828841e+00 4.82227765e-02 1.25535107e+00 -6.57216072e-01 1.14340031e+00 -8.40222687e-02 -1.13435578e+00 1.17296934e+00 7.49207616e-01 -5.86993814e-01 3.66138577e-01 -2.40440309e-01 -4.89763588e-01 4.48094517e-01 -1.08076692e+00 7.61701524e-01 7.06825316e-01 -9.92926002e-01 -1.57342625e+00 4.98572201e-01 5.86178780e-01 2.36674361e-02 -2.38469750e-01 -3.09269488e-01 6.25015855e-01 -1.14168298e+00 6.91718817e-01 -9.19413388e-01 2.07695171e-01 -7.10186660e-01 -1.53722003e-01 -8.57584059e-01 -4.96054217e-02 -1.15303552e+00 6.13457799e-01 9.42369640e-01 4.10031021e-01 -1.13330770e+00 5.51692724e-01 1.22016573e+00 2.98525751e-01 9.36108902e-02 -6.50620639e-01 -5.87136328e-01 4.11508203e-01 -5.70655286e-01 5.13934910e-01 1.18561435e+00 1.18182230e+00 7.72611022e-01 1.59751371e-01 1.00688100e-01 3.13438535e-01 1.90766733e-02 1.04519010e+00 -1.25386262e+00 -7.70700991e-01 -3.75049174e-01 -5.66755593e-01 -1.25330901e+00 1.31088138e-01 -9.70414460e-01 2.08703475e-03 -1.20482278e+00 -3.81257124e-02 2.48088136e-01 3.05024922e-01 4.31780070e-01 5.54506898e-01 -8.29678327e-02 5.79925656e-01 4.21860993e-01 -4.95349675e-01 -2.25972667e-01 1.06540298e+00 -2.52020717e-01 4.74022850e-02 -3.26204538e-01 -1.02394402e+00 1.42370427e+00 9.41068351e-01 -3.35988373e-01 -5.06524801e-01 -2.98863113e-01 6.80405021e-01 2.40911648e-01 1.08468187e+00 -9.89775538e-01 5.92200339e-01 -3.95603031e-01 -2.30461121e-01 1.07703649e-01 1.63553745e-01 -8.85964513e-01 4.40610439e-01 7.87012756e-01 -4.97195363e-01 1.15287416e-01 -5.66976592e-02 8.92232731e-02 -1.35103017e-01 -4.98948425e-01 8.53678644e-01 -8.52285981e-01 -6.89878225e-01 -1.02544808e+00 -1.20127559e+00 7.29687586e-02 1.03833759e+00 -5.40081382e-01 -5.34018219e-01 -7.36084819e-01 -8.24789345e-01 4.98640724e-02 8.43084693e-01 -1.92909837e-01 4.47821707e-01 -5.17157853e-01 -4.36900526e-01 1.12588249e-01 -4.97508720e-02 -3.19133967e-01 -2.33032465e-01 9.04255927e-01 -1.04754162e+00 5.17373443e-01 -2.10377768e-01 -2.97164619e-01 -1.22304010e+00 5.58415294e-01 6.48961186e-01 1.75341383e-01 -7.77103603e-01 6.55689120e-01 6.69559658e-01 -2.85071164e-01 -2.45133743e-01 -3.88988823e-01 4.87994142e-02 -5.93431890e-01 8.36191416e-01 4.79739718e-02 -6.34189904e-01 -7.39045680e-01 -3.93733948e-01 4.80355233e-01 -1.66983351e-01 -3.84942919e-01 9.86137211e-01 -4.51728940e-01 -7.50834644e-01 9.62818921e-01 6.15526259e-01 -1.18635811e-01 -1.46299228e-01 -1.81176886e-01 1.59952432e-01 -2.67053485e-01 -2.98055708e-01 -9.31882381e-01 3.23277675e-02 9.95706856e-01 -5.76298945e-02 9.20685828e-01 5.26346684e-01 2.79850870e-01 6.45498872e-01 1.07631302e+00 9.94116426e-01 -8.82592082e-01 -2.45668739e-01 6.55763447e-01 7.32755542e-01 -1.04875398e+00 1.67420879e-01 -3.37547719e-01 -8.97518277e-01 1.40206718e+00 2.85067022e-01 -2.31675938e-01 1.05915584e-01 2.74940014e-01 2.46298134e-01 -7.33167470e-01 -9.72258389e-01 1.31609693e-01 -4.82162893e-01 4.68871415e-01 6.81570530e-01 1.34354234e-01 -4.79206622e-01 3.41211140e-01 -6.91229522e-01 4.38228011e-01 1.14513683e+00 9.90788519e-01 -7.98471212e-01 -9.46698964e-01 -7.65583336e-01 -2.53856987e-01 -5.40502012e-01 1.36462286e-01 -1.00573111e+00 1.15435219e+00 2.31054559e-01 1.38962626e+00 -3.74953188e-02 -2.65114069e-01 4.77008000e-02 3.65865439e-01 4.57495600e-01 -7.29482591e-01 -9.21212077e-01 -5.59716165e-01 4.42099601e-01 -5.13293862e-01 -6.52209163e-01 -2.50658423e-01 -1.64935935e+00 -1.18965232e+00 -8.25646222e-02 6.97374523e-01 2.58624375e-01 1.26602745e+00 -1.67700201e-01 -4.46103662e-02 1.61420554e-01 -2.57732093e-01 -5.94401181e-01 -5.74582219e-01 -4.88727659e-01 1.66826889e-01 3.07188958e-01 -6.86050281e-02 -5.06757021e-01 2.18230620e-01]
[9.36721134185791, 7.146583080291748]
dcc5b68a-38ee-4bce-b051-39827cd6cb0c
weakly-supervised-learning-for-tool
1806.05573
null
http://arxiv.org/abs/1806.05573v2
http://arxiv.org/pdf/1806.05573v2.pdf
Weakly-Supervised Learning for Tool Localization in Laparoscopic Videos
Surgical tool localization is an essential task for the automatic analysis of endoscopic videos. In the literature, existing methods for tool localization, tracking and segmentation require training data that is fully annotated, thereby limiting the size of the datasets that can be used and the generalization of the approaches. In this work, we propose to circumvent the lack of annotated data with weak supervision. We propose a deep architecture, trained solely on image level annotations, that can be used for both tool presence detection and localization in surgical videos. Our architecture relies on a fully convolutional neural network, trained end-to-end, enabling us to localize surgical tools without explicit spatial annotations. We demonstrate the benefits of our approach on a large public dataset, Cholec80, which is fully annotated with binary tool presence information and of which 5 videos have been fully annotated with bounding boxes and tool centers for the evaluation.
['Didier Mutter', 'Armine Vardazaryan', 'Nicolas Padoy', 'Jacques Marescaux']
2018-06-14
null
null
null
null
['surgical-tool-detection']
['computer-vision']
[ 2.10769966e-01 2.13788286e-01 -4.70256478e-01 -8.87108296e-02 -7.18174517e-01 -8.45127106e-01 2.47142002e-01 1.88483834e-01 -7.56290674e-01 1.09254830e-01 -1.69721823e-02 -3.37162584e-01 1.46593347e-01 -3.56419355e-01 -7.95742512e-01 -2.86379993e-01 -7.92178139e-03 6.35079294e-03 4.75663334e-01 1.87965959e-01 1.26888633e-01 4.56944376e-01 -1.36019039e+00 2.77492911e-01 6.93090618e-01 1.25149107e+00 3.85477245e-01 6.09975696e-01 4.79898788e-02 8.22052360e-01 -4.01655376e-01 -4.41232435e-02 4.55191165e-01 -2.90135086e-01 -7.58795619e-01 7.93110728e-02 4.52432185e-01 -6.81645453e-01 -4.59626138e-01 9.29125130e-01 2.13758871e-01 -3.29536110e-01 4.62427467e-01 -7.47378290e-01 -1.42668307e-01 4.48690206e-01 -3.26120555e-01 4.52295728e-02 2.65718162e-01 2.36142680e-01 6.46927834e-01 -7.18817592e-01 1.02933240e+00 6.33218706e-01 9.73139942e-01 4.42747384e-01 -8.00077200e-01 -6.89782619e-01 -1.76525310e-01 -1.50667191e-01 -1.28458488e+00 -1.52870595e-01 7.28571892e-01 -8.84969532e-01 5.10488451e-01 -1.67257525e-02 1.03200030e+00 1.03106689e+00 9.37934592e-02 8.84078205e-01 5.79867482e-01 -6.46737337e-01 9.91707295e-02 1.24154195e-01 -1.23135872e-01 1.20840883e+00 2.81632483e-01 1.97326481e-01 -3.56685162e-01 9.53256711e-02 1.15226185e+00 5.71559310e-01 -3.17469209e-01 -9.94361281e-01 -1.45585299e+00 8.04599226e-01 9.10896957e-01 5.20500124e-01 -3.36712956e-01 3.58967990e-01 6.46395743e-01 -1.57719582e-01 2.29050562e-01 6.00581706e-01 -2.70879716e-01 -1.55179158e-01 -1.05213940e+00 -2.96297610e-01 7.69993067e-01 1.10360634e+00 3.30946058e-01 -5.32380104e-01 -3.47612679e-01 4.30240422e-01 5.81534766e-02 1.27993226e-01 5.48760474e-01 -8.50644171e-01 2.32887045e-01 1.11393929e+00 2.57956624e-01 -6.55357242e-01 -8.15529943e-01 -4.90471691e-01 -3.78481925e-01 2.33841166e-01 5.95940232e-01 -3.02392960e-01 -1.15412128e+00 1.30264843e+00 3.75544578e-02 1.75794750e-01 -2.71379292e-01 8.76486540e-01 1.02914000e+00 -1.84454545e-01 3.81863900e-02 3.41816902e-01 1.39717185e+00 -1.47366655e+00 -6.00970328e-01 -7.58415386e-02 1.27551794e+00 -8.79461586e-01 1.10782695e+00 3.55106890e-01 -7.42147803e-01 -3.13098401e-01 -1.02193451e+00 -2.11822867e-01 -3.86495084e-01 8.90655935e-01 8.63156736e-01 4.38813746e-01 -8.99359226e-01 5.94264388e-01 -1.27300441e+00 -4.60333407e-01 7.67788410e-01 5.79575181e-01 -6.62405550e-01 -4.62498516e-02 -5.73706746e-01 9.00472820e-01 5.01420736e-01 3.16767067e-01 -9.15162921e-01 -5.25203049e-01 -1.08601034e+00 -1.02425948e-01 5.69398463e-01 -6.78518355e-01 1.33766294e+00 -1.08994257e+00 -1.36661839e+00 1.16206002e+00 1.97122738e-01 -6.00228608e-01 9.71687257e-01 -3.61301303e-01 5.29608876e-02 3.93986344e-01 1.21732242e-02 6.80101693e-01 5.06975591e-01 -9.75099862e-01 -6.71106875e-01 -2.60541648e-01 3.30872864e-01 -1.77748173e-01 -4.49744612e-01 4.82896455e-02 -7.46833563e-01 -7.36368597e-01 -1.56868801e-01 -1.31152344e+00 -3.95533532e-01 7.02522874e-01 -6.05734468e-01 1.22566886e-01 7.49464869e-01 -8.21919262e-01 1.04189181e+00 -2.28146458e+00 1.12709574e-01 -5.65312011e-03 1.80496484e-01 1.59544215e-01 -1.70961112e-01 9.99080688e-02 2.59666085e-01 -1.75687805e-01 -1.34840086e-01 -4.44931835e-01 -3.28005105e-01 4.90205809e-02 2.85123438e-01 8.16318214e-01 -8.15916359e-02 9.81416047e-01 -9.81394708e-01 -7.02581048e-01 6.62264705e-01 5.29682159e-01 -7.78232455e-01 3.17961663e-01 -9.27857533e-02 7.12254763e-01 -3.33805233e-01 9.94997680e-01 4.18418884e-01 -2.24730954e-01 2.60674000e-01 -4.49843466e-01 -1.23058625e-01 -1.52102545e-01 -8.07959259e-01 2.55132842e+00 -1.04078519e+00 6.28784716e-01 3.54594231e-01 -4.21580046e-01 4.20151979e-01 3.15117091e-01 8.10262263e-01 -3.41540664e-01 5.14492333e-01 5.25114834e-01 1.66463166e-01 -5.46441078e-01 2.83903122e-01 2.87201732e-01 -2.52996862e-01 -8.66043791e-02 5.36621988e-01 -6.30126242e-03 2.73716956e-01 -1.08298615e-01 1.35337305e+00 2.09195748e-01 3.32869619e-01 -1.29688799e-01 2.90102184e-01 3.78247797e-01 1.42892942e-01 7.65768588e-01 -3.35623980e-01 5.69140851e-01 6.05470896e-01 -4.82535571e-01 -7.84793735e-01 -7.56890953e-01 -2.10275248e-01 1.00159311e+00 3.41224223e-01 -2.85534620e-01 -8.35747838e-01 -1.24691105e+00 -1.35233179e-02 4.86723967e-02 -1.05449510e+00 -6.73119128e-02 -6.60416424e-01 -3.68386740e-03 3.96554410e-01 8.53537321e-01 1.06252439e-01 -1.04571211e+00 -1.12528098e+00 3.03769410e-02 1.73181087e-01 -1.41364324e+00 -5.02836704e-01 5.30464053e-01 -7.54360676e-01 -1.55769336e+00 -8.92567873e-01 -9.14118588e-01 1.00614512e+00 1.40129626e-02 1.00741220e+00 9.07715037e-02 -7.28238344e-01 3.37262869e-01 -2.85784245e-01 -4.86873657e-01 -2.78990984e-01 4.85689849e-01 -3.58401775e-01 -4.05585110e-01 1.35029376e-01 -2.10421920e-01 -1.08816242e+00 2.75534868e-01 -7.95760334e-01 8.28801543e-02 1.05844474e+00 1.02026975e+00 4.47353035e-01 -6.70391381e-01 1.33921420e-02 -1.03408408e+00 1.04908444e-01 -3.13559681e-01 -8.43835175e-01 1.14618754e-02 8.44580233e-02 -1.17199630e-01 4.96867180e-01 -3.87319237e-01 -5.05209923e-01 9.35956299e-01 -1.66241243e-01 -8.17698777e-01 -2.30453297e-01 3.69890988e-01 1.05703428e-01 -7.06594646e-01 5.64142942e-01 -2.00058728e-01 1.44432575e-01 -5.25786102e-01 4.49352741e-01 6.44613862e-01 6.08467400e-01 2.52611656e-02 3.70801151e-01 4.49160844e-01 1.34690657e-01 -2.99876243e-01 -1.02370298e+00 -8.92187417e-01 -1.11830235e+00 -3.69606540e-02 1.04717743e+00 -9.17626083e-01 -6.06685936e-01 -6.84963390e-02 -1.19004309e+00 -3.97588313e-01 -2.58276224e-01 7.39222944e-01 -5.92752874e-01 1.82749331e-01 -6.46146059e-01 -4.36269492e-01 -4.68780071e-01 -1.42524195e+00 1.41604602e+00 1.30803473e-02 -2.18213797e-01 -1.01070857e+00 -7.12234750e-02 1.10204563e-01 3.54874551e-01 7.25169957e-01 3.05118352e-01 -8.20239186e-01 -6.30523741e-01 -9.02158022e-01 -2.82032847e-01 3.17914814e-01 3.36842299e-01 -3.53829473e-01 -9.13361490e-01 -2.32521415e-01 -3.59502643e-01 -1.29604444e-01 9.99655306e-01 4.81655061e-01 1.43233013e+00 -5.88787645e-02 -8.63930047e-01 8.87515843e-01 1.33310425e+00 -1.68618128e-01 2.34850179e-02 4.93590325e-01 9.09323096e-01 2.59205520e-01 7.08182037e-01 1.94252282e-01 3.32000013e-03 7.49078572e-01 8.45322251e-01 -7.11652100e-01 -2.88914144e-01 -2.19117358e-01 -3.36357355e-02 3.50566328e-01 -7.25348815e-02 2.15030849e-01 -9.21352684e-01 7.77514875e-01 -1.65048409e+00 -3.62977803e-01 2.23210499e-01 2.05088329e+00 7.02627540e-01 -1.45103171e-01 1.56893790e-01 -1.09742180e-01 4.65639949e-01 -1.88960969e-01 -4.09527063e-01 7.62299448e-02 4.33138460e-01 1.93394825e-01 9.67235565e-01 1.42664000e-01 -1.84604466e+00 8.38788450e-01 6.16377592e+00 4.56646919e-01 -1.33406270e+00 2.22971305e-01 2.06931010e-02 -2.62959659e-01 5.27450025e-01 -3.52978706e-01 -2.62492150e-01 4.92657512e-01 5.60681760e-01 4.63524163e-01 -1.27457619e-01 1.34406018e+00 -5.26372641e-02 -1.17099918e-01 -1.49029577e+00 1.04481101e+00 1.46588013e-01 -1.40454018e+00 -3.42434257e-01 6.75626472e-02 6.18214309e-01 1.47587016e-01 -9.28000286e-02 2.16285482e-01 5.64857870e-02 -9.94918048e-01 7.24617004e-01 3.25346142e-01 1.04909468e+00 -5.18074870e-01 1.12282109e+00 2.18278587e-01 -1.03972316e+00 -2.45466635e-01 2.88024414e-02 3.91062737e-01 -9.65162180e-03 1.29531771e-01 -1.16736317e+00 4.32542026e-01 4.34818327e-01 8.30455005e-01 -7.21413851e-01 1.41344929e+00 -3.07201236e-01 2.11482048e-01 -5.13263881e-01 9.00175348e-02 3.43019217e-01 1.95006549e-01 2.22336099e-01 1.49554992e+00 5.00615478e-01 -5.61497808e-01 6.77727580e-01 6.54143274e-01 -3.19675058e-01 2.67537236e-01 -7.65277863e-01 1.78729780e-02 2.58632243e-01 1.35860586e+00 -9.64389145e-01 -1.83534250e-02 -5.61016202e-01 1.20082855e+00 1.25289172e-01 -5.63851930e-02 -7.87827671e-01 -6.80601954e-01 1.63706511e-01 2.84471095e-01 3.09328109e-01 -1.84566826e-01 -6.48799986e-02 -1.00541103e+00 1.83698043e-01 -4.43646342e-01 4.35769856e-01 -4.65604484e-01 -7.75470555e-01 7.35500634e-01 -5.67054808e-01 -1.69865036e+00 -3.61226350e-01 -1.10868788e+00 -4.65759307e-01 5.97413540e-01 -1.50341249e+00 -1.75965977e+00 -9.04101789e-01 5.37726879e-01 4.14509356e-01 1.43086821e-01 9.26066935e-01 4.97881621e-01 -3.54311466e-01 7.27065980e-01 -4.93542738e-02 7.06764460e-01 7.35484719e-01 -1.23760402e+00 -2.82805860e-01 6.26427293e-01 6.99273944e-02 5.67768514e-01 3.08017313e-01 -3.35518330e-01 -1.37682784e+00 -1.22838962e+00 -5.34222238e-02 -7.86597371e-01 8.54659617e-01 -5.13791621e-01 -4.07323420e-01 9.28330958e-01 8.62942729e-03 6.46946013e-01 7.19313800e-01 -3.77456099e-02 -5.55045940e-02 3.80279183e-01 -1.17677402e+00 3.31011057e-01 1.10048223e+00 -4.52369809e-01 -2.44716078e-01 7.00400829e-01 4.22077864e-01 -1.03193140e+00 -1.03655267e+00 5.86667001e-01 5.58094203e-01 -7.49200523e-01 9.12042260e-01 -3.42865348e-01 5.31195223e-01 -2.24917620e-01 2.07626984e-01 -1.05654323e+00 3.04204822e-01 -1.94476768e-01 -1.29637301e-01 6.77710414e-01 2.62718141e-01 -1.56582803e-01 9.07962680e-01 1.29106835e-01 -4.25336927e-01 -1.03075337e+00 -1.01831663e+00 -5.80847919e-01 -3.19006033e-02 -1.46205574e-01 2.39287719e-01 7.34816194e-01 -2.03693751e-02 -2.21531957e-01 -1.70475751e-01 -3.21300887e-02 1.70037955e-01 2.24226564e-01 7.44509816e-01 -1.18030906e+00 -1.20971434e-01 -5.98683774e-01 -7.49021471e-01 -8.78077865e-01 1.29631370e-01 -1.10465753e+00 2.20316187e-01 -1.72536123e+00 2.69299090e-01 -3.67959529e-01 -3.14547539e-01 6.91535294e-01 9.54154357e-02 6.07293606e-01 1.17826559e-01 3.25118631e-01 -9.15266335e-01 1.35207489e-01 1.17486930e+00 -6.59857169e-02 -1.83281645e-01 1.68264844e-02 -7.84741621e-03 1.09368706e+00 4.05053020e-01 -5.17673612e-01 8.14110637e-02 -2.60160357e-01 -2.60427177e-01 -5.96450083e-02 6.14618301e-01 -1.05132842e+00 1.10862359e-01 3.77227962e-01 6.31094277e-01 -5.19851744e-01 2.54882574e-01 -1.33488488e+00 -3.54520321e-01 9.04902101e-01 -3.52259934e-01 -3.71873081e-01 3.18687469e-01 5.20877838e-01 -3.94593090e-01 -4.54794288e-01 6.34344757e-01 -2.59814888e-01 -5.88889599e-01 3.97579521e-01 -9.40772146e-02 -3.17217350e-01 1.32308507e+00 -9.77944806e-02 7.76502341e-02 1.72063470e-01 -9.07141984e-01 8.05437416e-02 8.46028388e-01 4.66897607e-01 3.03644896e-01 -1.06290162e+00 -2.88820684e-01 -8.85520130e-02 6.47118509e-01 1.70307517e-01 2.99213648e-01 1.27643955e+00 -1.07720721e+00 4.88251567e-01 -1.89037040e-01 -9.21116829e-01 -1.10702419e+00 9.90964651e-01 7.56653190e-01 -3.56633306e-01 -1.02087796e+00 6.25438690e-01 1.64739251e-01 -5.09363472e-01 5.80190241e-01 -9.40169334e-01 -3.23464423e-01 -1.86284073e-02 2.54616618e-01 -1.91625714e-01 2.30661094e-01 -3.59407187e-01 -4.86162454e-01 7.87176490e-01 1.55296158e-02 4.45661038e-01 1.04790199e+00 2.97760904e-01 7.47215822e-02 2.01394409e-01 1.18721330e+00 1.64483398e-01 -1.25556827e+00 -9.79193002e-02 1.79110229e-01 -5.08876145e-01 2.21745268e-01 -8.82900894e-01 -1.26027966e+00 9.50158417e-01 1.14252484e+00 -1.55388623e-01 8.44802856e-01 1.41760454e-01 7.86202133e-01 2.97487229e-01 5.36830366e-01 -7.12124348e-01 -3.47195342e-02 1.08098403e-01 8.10795486e-01 -1.55862796e+00 -3.92361671e-01 -6.42200351e-01 -2.98260242e-01 1.21847022e+00 5.82571149e-01 -2.49507084e-01 5.16946077e-01 7.04815626e-01 3.46498966e-01 -1.94756746e-01 1.96111463e-02 -2.85537928e-01 4.71177548e-01 4.45014209e-01 5.68175197e-01 -5.25397882e-02 -1.61667407e-01 7.94173062e-01 8.50325376e-02 4.06300366e-01 3.98785889e-01 1.21034932e+00 -5.87336384e-02 -6.89161599e-01 9.97040123e-02 4.64266956e-01 -6.55741096e-01 -3.06074955e-02 -3.75410914e-01 1.07201111e+00 3.46646935e-01 5.59558511e-01 9.61265434e-03 7.84177929e-02 5.40313125e-01 -2.53137529e-01 6.22638226e-01 -8.04124534e-01 -9.40817952e-01 -1.38924628e-01 5.36421575e-02 -7.14331627e-01 -1.29665941e-01 -2.67473280e-01 -1.17367542e+00 3.94682437e-01 -4.55054522e-01 5.25258062e-03 7.05550313e-01 6.59576297e-01 3.34257573e-01 1.14342904e+00 1.85228556e-01 -1.09452462e+00 -3.15939039e-01 -1.13734353e+00 -4.20673043e-01 4.37180698e-01 5.00326991e-01 -9.84978914e-01 -7.83801228e-02 6.46439791e-02]
[14.099921226501465, -3.300269365310669]
da4bb28a-9ee2-4c29-94cb-cb7d1d08a43a
frustratingly-simple-but-effective-zero-shot
2302.07319
null
https://arxiv.org/abs/2302.07319v1
https://arxiv.org/pdf/2302.07319v1.pdf
Frustratingly Simple but Effective Zero-shot Detection and Segmentation: Analysis and a Strong Baseline
Methods for object detection and segmentation often require abundant instance-level annotations for training, which are time-consuming and expensive to collect. To address this, the task of zero-shot object detection (or segmentation) aims at learning effective methods for identifying and localizing object instances for the categories that have no supervision available. Constructing architectures for these tasks requires choosing from a myriad of design options, ranging from the form of the class encoding used to transfer information from seen to unseen categories, to the nature of the function being optimized for learning. In this work, we extensively study these design choices, and carefully construct a simple yet extremely effective zero-shot recognition method. Through extensive experiments on the MSCOCO dataset on object detection and segmentation, we highlight that our proposed method outperforms existing, considerably more complex, architectures. Our findings and method, which we propose as a competitive future baseline, point towards the need to revisit some of the recent design trends in zero-shot detection / segmentation.
['Leonid Sigal', 'Jayan Eledath', 'Behjat Siddiquie', 'Anirudth Nambirajan', 'Siddhesh Khandelwal']
2023-02-14
null
null
null
null
['zero-shot-object-detection']
['computer-vision']
[ 6.14913166e-01 4.89033349e-02 -3.69674385e-01 -5.08738756e-01 -9.79535282e-01 -5.15751123e-01 6.16422236e-01 2.04036444e-01 -3.74872357e-01 2.38306284e-01 -1.27929002e-01 -1.11831628e-01 8.59963894e-02 -6.38889909e-01 -6.38455808e-01 -6.71778083e-01 4.70055155e-02 5.00268936e-01 6.31970525e-01 1.35151863e-01 3.24950337e-01 2.05834299e-01 -2.09019375e+00 3.39206398e-01 3.78071994e-01 1.25090313e+00 1.59313217e-01 8.80707681e-01 -1.20433852e-01 8.25984776e-01 -5.12310803e-01 -2.03117445e-01 2.12239668e-01 -4.31875616e-01 -9.70734119e-01 5.64908028e-01 8.17073166e-01 -3.16171616e-01 9.04940814e-02 1.01062071e+00 5.42295277e-01 5.60628772e-01 9.38634276e-01 -1.09791040e+00 -5.46466470e-01 5.57111263e-01 -3.36880475e-01 6.43408477e-01 -1.18700759e-02 4.15182143e-01 1.18902171e+00 -1.14524901e+00 7.03401208e-01 1.01735735e+00 5.95732868e-01 8.86605740e-01 -1.22671366e+00 -2.90846497e-01 2.21120864e-01 2.37075076e-01 -1.37300420e+00 -8.69236410e-01 4.69334066e-01 -7.29166746e-01 9.48704064e-01 1.58396453e-01 4.46228951e-01 1.03229797e+00 -3.27746451e-01 1.23764670e+00 7.84306407e-01 -6.06293261e-01 5.39702773e-01 2.82904863e-01 6.88301146e-01 6.66512668e-01 2.07056701e-01 -1.15957022e-01 -3.45823199e-01 2.61354446e-03 5.04886329e-01 -8.61780196e-02 -1.42607793e-01 -7.41531968e-01 -8.99592280e-01 7.59025753e-01 4.88543719e-01 2.52702713e-01 -1.91018563e-02 2.14327037e-01 3.77650976e-01 1.03625290e-01 3.61566871e-01 5.87599993e-01 -3.14370304e-01 -8.69024545e-02 -1.00065756e+00 -2.45823078e-02 8.10998738e-01 1.12647080e+00 9.50855732e-01 -4.74996772e-03 -5.07725418e-01 8.76275897e-01 1.63276151e-01 -8.44858140e-02 4.15385962e-01 -8.33080709e-01 2.52278239e-01 3.95287484e-01 5.09296395e-02 -4.76206779e-01 -3.03618670e-01 -3.36998880e-01 -3.27774495e-01 7.89622311e-03 4.72533494e-01 -5.10478094e-02 -1.47487533e+00 1.41368866e+00 5.02582431e-01 4.88231748e-01 -9.98829976e-02 9.69023764e-01 1.03538799e+00 4.55456883e-01 3.61954957e-01 -6.32755179e-03 1.50603032e+00 -1.00478339e+00 -4.14064288e-01 -4.83914405e-01 7.55864561e-01 -5.16827941e-01 1.18936503e+00 4.97370884e-02 -1.07651711e+00 -5.52547812e-01 -1.15074730e+00 -2.96859771e-01 -7.02310860e-01 5.57250641e-02 7.58823097e-01 6.93064570e-01 -7.80494392e-01 6.26326144e-01 -9.37498629e-01 -5.30483663e-01 8.66992593e-01 2.76364148e-01 8.88432637e-02 -5.46689471e-03 -9.15795386e-01 7.33311653e-01 6.61974728e-01 -1.03582337e-01 -1.24344075e+00 -7.27765739e-01 -9.06983972e-01 2.68610299e-01 6.78270578e-01 -4.08773601e-01 1.65995872e+00 -9.82090712e-01 -1.14029658e+00 1.26576340e+00 1.24218181e-01 -7.09065199e-01 2.64936984e-01 -3.69724445e-02 -1.26463221e-02 2.07457811e-01 1.99288532e-01 1.01649046e+00 9.76541042e-01 -1.15583229e+00 -9.45120633e-01 -2.15458781e-01 1.20799810e-01 3.06929611e-02 -3.14853936e-01 1.47756130e-01 -5.48280716e-01 -4.42670584e-01 5.38286567e-02 -7.10262895e-01 -3.22763503e-01 1.49073377e-01 -2.91279405e-01 -4.63621408e-01 9.79287446e-01 -2.32283935e-01 1.09359097e+00 -2.20193982e+00 -1.49784982e-02 -2.04416439e-01 2.26498395e-01 3.59760135e-01 -5.83327599e-02 -2.56460067e-02 3.58686484e-02 4.72700335e-02 -5.71175873e-01 -5.30508697e-01 6.91255033e-02 1.52744710e-01 -3.52154523e-01 5.11465013e-01 5.21431684e-01 1.10643828e+00 -8.72922063e-01 -7.45966017e-01 3.60498369e-01 2.17259720e-01 -3.85881811e-01 1.95421919e-01 -4.46451873e-01 -1.94865856e-02 -3.57246518e-01 1.00437796e+00 1.82725817e-01 -4.79757041e-01 -2.03027278e-01 -2.52091587e-02 -1.59151614e-01 1.95635006e-01 -1.16102457e+00 1.59222150e+00 -1.16524257e-01 6.83510005e-01 -6.13943897e-02 -1.11207616e+00 6.27046943e-01 2.12695137e-01 4.12185669e-01 -4.51300412e-01 4.01182562e-01 7.87776932e-02 -4.58224267e-02 -4.83356774e-01 4.02698070e-01 -1.40877053e-01 -1.51899621e-01 3.89561743e-01 5.14155388e-01 -1.74079426e-02 4.69432116e-01 2.13335991e-01 1.11203074e+00 -4.46545407e-02 4.14200455e-01 -2.27791697e-01 1.55814901e-01 9.17086080e-02 5.01073122e-01 1.14906669e+00 -7.03736603e-01 7.98517108e-01 3.15114915e-01 -4.40803409e-01 -9.39954221e-01 -8.34384143e-01 -4.00605738e-01 1.45645642e+00 2.08311692e-01 -7.61892125e-02 -9.27527189e-01 -7.52460718e-01 -7.79679120e-02 7.08529234e-01 -6.68751061e-01 -2.21563905e-01 -3.64376843e-01 -6.71375215e-01 3.05542409e-01 8.22088718e-01 1.90918013e-01 -1.08463538e+00 -9.37515318e-01 1.96107015e-01 8.29607099e-02 -1.05444860e+00 -2.68597424e-01 5.72408438e-01 -7.39091516e-01 -1.05386496e+00 -6.99066460e-01 -9.66389358e-01 5.54517567e-01 5.48041463e-01 1.16401529e+00 1.20793551e-01 -8.41389179e-01 5.46650052e-01 -4.35281157e-01 -6.42350078e-01 -5.36017604e-02 2.81102568e-01 -2.71772504e-01 1.14782572e-01 7.00851798e-01 -1.82541549e-01 -5.44026077e-01 2.89015144e-01 -9.32917476e-01 -1.11062340e-01 4.93316263e-01 6.35275424e-01 5.67888319e-01 -2.86492109e-01 5.45835078e-01 -1.20308721e+00 2.48607039e-01 -5.48171461e-01 -5.63491762e-01 3.69091928e-01 -3.60337287e-01 -1.08373009e-01 1.57283291e-01 -4.21808749e-01 -9.97243762e-01 3.12469602e-01 3.52266356e-02 -7.09167957e-01 -4.59034830e-01 2.49568801e-02 -1.34530783e-01 -2.23667473e-01 9.71275389e-01 8.82478803e-02 -4.16109622e-01 -5.97149432e-01 6.27642989e-01 7.11452901e-01 5.80575228e-01 -3.80246699e-01 4.94573355e-01 5.69992125e-01 -3.27622890e-01 -1.00263739e+00 -1.24299741e+00 -1.08506274e+00 -8.69742990e-01 -9.37163159e-02 1.06953406e+00 -7.48894334e-01 -5.29108047e-02 3.52210641e-01 -1.02186990e+00 -4.46582347e-01 -6.31047904e-01 8.67246389e-02 -6.81147754e-01 1.44468457e-01 -5.19750714e-01 -9.66094971e-01 -2.32054651e-01 -1.00953615e+00 1.32299042e+00 3.01269770e-01 -1.73133537e-01 -8.81076932e-01 -1.21767551e-01 3.91675830e-01 2.33391881e-01 1.48775458e-01 8.56844723e-01 -7.54853368e-01 -8.59012485e-01 -3.09969336e-01 -3.83341432e-01 2.43289620e-01 -1.09151907e-01 5.35131283e-02 -1.22521317e+00 -2.83257395e-01 -1.45487497e-02 -7.99977064e-01 1.23547256e+00 4.50515985e-01 1.37344193e+00 -1.08103730e-01 -5.21906793e-01 6.32691145e-01 1.31402802e+00 2.54840739e-02 3.78847927e-01 4.73681726e-02 6.94476485e-01 6.07781827e-01 8.00424159e-01 3.73370618e-01 1.36707589e-01 6.17316604e-01 3.14777136e-01 -4.39469367e-02 -4.10203665e-01 -3.26254703e-02 4.56839129e-02 3.63465965e-01 4.00662050e-02 -2.33588248e-01 -1.04332054e+00 8.05815041e-01 -1.76162791e+00 -1.07732236e+00 1.90713704e-01 2.22398543e+00 7.80083537e-01 2.31267035e-01 2.88759947e-01 1.05106927e-01 7.45226204e-01 2.93635458e-01 -6.72900736e-01 -1.28851250e-01 2.23856941e-01 1.59548089e-01 3.38820726e-01 3.31811696e-01 -1.64803922e+00 1.15675473e+00 6.69584560e+00 8.75349700e-01 -1.00881791e+00 2.65432626e-01 6.25521362e-01 -1.92915991e-01 2.14557797e-01 4.06269766e-02 -1.13184536e+00 3.20420653e-01 9.95158136e-01 5.23585938e-02 1.68858171e-01 1.18490815e+00 -2.76996791e-01 -4.51074764e-02 -1.56969249e+00 8.11578810e-01 1.99177101e-01 -1.38643026e+00 -6.02603592e-02 -1.92632198e-01 8.33390415e-01 1.67241022e-01 -4.41330634e-02 6.40188992e-01 2.12327138e-01 -8.50351632e-01 7.36425936e-01 2.27680624e-01 5.74891448e-01 -3.41863364e-01 3.63096684e-01 3.59297276e-01 -1.29605472e+00 -2.86062896e-01 -5.19241810e-01 9.41571593e-02 5.98063655e-02 1.30834997e-01 -9.46946323e-01 -8.18469971e-02 6.28910005e-01 6.50373697e-01 -5.63059688e-01 1.39009869e+00 -1.44277930e-01 7.20751405e-01 -1.29669085e-01 -1.59699768e-02 4.47518706e-01 3.19151819e-01 5.21414518e-01 1.48125243e+00 -3.15704569e-02 3.36455107e-01 5.37766039e-01 8.98158908e-01 -1.77617133e-01 -1.38594821e-01 -5.04285395e-01 -2.32662290e-01 4.85807866e-01 1.34253502e+00 -1.22062838e+00 -7.13153183e-01 -5.80333829e-01 6.15628123e-01 4.11093861e-01 2.79773951e-01 -7.52970576e-01 -5.34675956e-01 5.86627543e-01 1.28767312e-01 5.28093755e-01 -1.10175915e-01 -2.84842819e-01 -1.15964460e+00 -1.71218291e-01 -6.66856766e-01 7.10571647e-01 -4.46092933e-01 -1.24668276e+00 2.11331576e-01 1.07013665e-01 -1.11106431e+00 -1.64646909e-01 -7.14724004e-01 -6.74104691e-01 2.45569587e-01 -1.47062075e+00 -1.07836747e+00 -2.22765803e-01 3.00752223e-01 1.14469218e+00 7.08670393e-02 5.64137995e-01 3.29279035e-01 -6.53480113e-01 5.16520917e-01 -2.03675926e-01 2.94928282e-01 2.69784927e-01 -1.28209758e+00 4.94333923e-01 7.98439980e-01 5.40851831e-01 3.69593710e-01 7.23702133e-01 -4.09194797e-01 -1.27713084e+00 -1.14562416e+00 5.84439456e-01 -6.90757871e-01 7.13300228e-01 -5.91463804e-01 -1.06482482e+00 8.33194852e-01 -1.92430541e-01 2.49999151e-01 7.35887408e-01 1.72149852e-01 -3.76899272e-01 1.63399160e-01 -1.07041574e+00 3.31883997e-01 1.14318883e+00 -5.38663685e-01 -7.59598792e-01 4.53344464e-01 7.84527600e-01 -2.27854550e-01 -5.35504520e-01 3.31351578e-01 3.37794036e-01 -7.49454141e-01 9.51126039e-01 -8.65223646e-01 2.22134531e-01 -1.19478069e-01 -2.12185219e-01 -8.10784638e-01 -3.80722970e-01 -3.55297506e-01 -4.25198346e-01 1.21463001e+00 3.63564253e-01 -2.43938431e-01 9.38419640e-01 9.43818808e-01 -2.67769605e-01 -8.77735078e-01 -8.60974789e-01 -7.27020741e-01 -1.66536450e-01 -4.43424791e-01 1.53318495e-01 9.17554975e-01 -1.83666080e-01 5.49980283e-01 -1.23639151e-01 6.89495727e-02 8.44774425e-01 2.08627149e-01 6.75727189e-01 -1.31162107e+00 -2.69898355e-01 -5.51620841e-01 -6.86588407e-01 -9.87807333e-01 -5.37559856e-03 -8.58576536e-01 5.26656210e-01 -1.40735340e+00 4.02428329e-01 -3.48363131e-01 -4.54776704e-01 6.96538627e-01 -1.42736718e-01 3.39982778e-01 2.11742416e-01 3.64048779e-01 -1.07040441e+00 3.63695651e-01 8.51722240e-01 -3.19683135e-01 -2.09535569e-01 1.96641847e-01 -5.81033051e-01 8.47635984e-01 5.34787774e-01 -4.55476612e-01 -5.04791319e-01 -4.72285479e-01 -3.49393457e-01 -2.70870030e-01 4.35658365e-01 -1.20068777e+00 3.94985288e-01 -2.27749467e-01 1.58594325e-01 -5.78597307e-01 5.38765311e-01 -5.52075505e-01 -5.93091905e-01 2.22375348e-01 -5.07732093e-01 -5.12655199e-01 -1.43282628e-02 7.60829449e-01 -1.52729645e-01 -5.90787768e-01 1.13944757e+00 -4.19405252e-01 -1.40415704e+00 5.04519522e-01 -1.70463234e-01 3.39447945e-01 1.34357178e+00 -4.28777695e-01 -2.26826802e-01 8.70246887e-02 -9.09422457e-01 4.26955730e-01 3.97006363e-01 4.77031678e-01 5.34830511e-01 -1.00417149e+00 -3.51205349e-01 5.00984043e-02 3.97283584e-01 3.91536877e-02 1.31709322e-01 6.64205432e-01 -1.13035023e-01 4.36962962e-01 -7.24989176e-02 -7.54253387e-01 -1.21429551e+00 7.66883910e-01 4.15603369e-01 1.63215309e-01 -7.36979246e-01 1.20714676e+00 3.24407488e-01 -1.19845577e-01 5.73259652e-01 -2.44069546e-01 -1.85274482e-01 4.09886360e-01 6.38404906e-01 5.41816056e-01 6.23253398e-02 -4.98018622e-01 -3.00501943e-01 4.19355005e-01 -2.87412286e-01 9.65752900e-02 1.18852735e+00 -7.46724308e-02 1.85188442e-01 7.00716496e-01 1.06568527e+00 -9.44206297e-01 -1.54903924e+00 -3.82650554e-01 4.70968992e-01 -4.03422982e-01 1.85224652e-01 -3.89816582e-01 -8.66173983e-01 1.14580309e+00 6.92512870e-01 3.80606562e-01 7.91091979e-01 3.56728017e-01 6.46094382e-01 5.79839349e-01 4.00192708e-01 -1.43146682e+00 2.34402642e-01 4.83009994e-01 3.96822721e-01 -1.55977106e+00 -1.50657266e-01 -4.20865119e-01 -4.89748538e-01 8.32290411e-01 8.67008626e-01 -1.26027077e-01 5.66397667e-01 1.00396410e-01 -4.28723991e-02 -4.62953806e-01 -7.97649741e-01 -7.79480696e-01 4.31820393e-01 5.69253564e-01 3.24199200e-01 -8.83651972e-02 -5.91802597e-02 3.73060912e-01 2.56175995e-01 -1.06320426e-01 3.03907782e-01 1.23697746e+00 -1.11022353e+00 -7.82079041e-01 -1.65044084e-01 6.43494904e-01 -2.73701698e-01 5.13578653e-02 -5.56969941e-01 6.35460913e-01 2.58772820e-01 8.54401112e-01 2.49617741e-01 1.17487364e-01 2.04166085e-01 3.64323080e-01 3.03673506e-01 -1.28205228e+00 -3.52726460e-01 -1.99672416e-01 -6.42343462e-02 -5.70820272e-01 -3.34042609e-01 -7.26441264e-01 -1.25238693e+00 3.97398412e-01 -5.23130000e-01 -1.10034399e-01 5.21501839e-01 1.06945300e+00 2.21085355e-01 4.90333438e-01 1.81967303e-01 -1.21939552e+00 -7.56651282e-01 -7.58168578e-01 -5.33442080e-01 3.83034438e-01 4.62399364e-01 -8.94824982e-01 -3.43766987e-01 1.68833256e-01]
[9.570816993713379, 1.5918128490447998]
8aee189e-fee7-4474-a6f5-3f89b1ce7f79
generating-multiple-hypotheses-for-3d-human
1904.05547
null
http://arxiv.org/abs/1904.05547v1
http://arxiv.org/pdf/1904.05547v1.pdf
Generating Multiple Hypotheses for 3D Human Pose Estimation with Mixture Density Network
3D human pose estimation from a monocular image or 2D joints is an ill-posed problem because of depth ambiguity and occluded joints. We argue that 3D human pose estimation from a monocular input is an inverse problem where multiple feasible solutions can exist. In this paper, we propose a novel approach to generate multiple feasible hypotheses of the 3D pose from 2D joints.In contrast to existing deep learning approaches which minimize a mean square error based on an unimodal Gaussian distribution, our method is able to generate multiple feasible hypotheses of 3D pose based on a multimodal mixture density networks. Our experiments show that the 3D poses estimated by our approach from an input of 2D joints are consistent in 2D reprojections, which supports our argument that multiple solutions exist for the 2D-to-3D inverse problem. Furthermore, we show state-of-the-art performance on the Human3.6M dataset in both best hypothesis and multi-view settings, and we demonstrate the generalization capacity of our model by testing on the MPII and MPI-INF-3DHP datasets. Our code is available at the project website.
['Chen Li', 'Gim Hee Lee']
2019-04-11
generating-multiple-hypotheses-for-3d-human-1
http://openaccess.thecvf.com/content_CVPR_2019/html/Li_Generating_Multiple_Hypotheses_for_3D_Human_Pose_Estimation_With_Mixture_CVPR_2019_paper.html
http://openaccess.thecvf.com/content_CVPR_2019/papers/Li_Generating_Multiple_Hypotheses_for_3D_Human_Pose_Estimation_With_Mixture_CVPR_2019_paper.pdf
cvpr-2019-6
['multi-hypotheses-3d-human-pose-estimation', 'monocular-3d-human-pose-estimation']
['computer-vision', 'computer-vision']
[-2.85866112e-01 2.82579809e-01 6.96729347e-02 -3.05292130e-01 -8.39906633e-01 -5.16041815e-01 4.56783682e-01 -8.40611517e-01 -3.87554944e-01 8.85913372e-01 2.90557444e-01 1.52057841e-01 -1.49714410e-01 -3.38123024e-01 -1.00905120e+00 -4.15592909e-01 1.41827539e-01 1.28750980e+00 -2.29959353e-03 -1.21979550e-01 -5.57933934e-02 4.35164213e-01 -1.53080022e+00 9.63470712e-02 3.78897160e-01 7.22420394e-01 -3.07128597e-02 9.42225397e-01 4.48843658e-01 1.82003543e-01 -4.87961203e-01 -3.13046694e-01 5.69883943e-01 -4.15541470e-01 -7.55071580e-01 3.17071587e-01 6.93524599e-01 -6.38293505e-01 -3.32916200e-01 7.52178192e-01 8.03434253e-01 6.12129755e-02 1.05051160e+00 -1.58181357e+00 -4.27107990e-01 -1.12066761e-01 -6.01500690e-01 -4.67521340e-01 8.33845556e-01 1.29685402e-02 7.15452611e-01 -1.20177710e+00 8.62945318e-01 1.59426618e+00 6.91298008e-01 6.03535891e-01 -1.22185016e+00 -2.35190988e-01 -4.40134890e-02 -2.31281482e-02 -1.44966257e+00 -2.73953199e-01 5.91351092e-01 -6.78315759e-01 7.69514859e-01 4.78442684e-02 8.29901576e-01 1.45431995e+00 4.45087224e-01 8.04224491e-01 1.11948252e+00 -3.65592033e-01 7.17678517e-02 -2.90264159e-01 -3.39122176e-01 7.13445425e-01 2.33800441e-01 1.53255194e-01 -6.72583342e-01 -3.10376495e-01 1.34191358e+00 -1.87583953e-01 -1.25897869e-01 -1.00856066e+00 -1.41515720e+00 8.12158763e-01 1.40765548e-01 -3.98996651e-01 -3.47246796e-01 3.27261865e-01 -6.24593720e-02 -2.96308566e-03 3.75164092e-01 2.73246825e-01 -5.47858834e-01 -2.14027017e-01 -4.97864753e-01 8.64298463e-01 9.07567382e-01 1.04101491e+00 5.03563523e-01 -4.06468153e-01 3.22248101e-01 7.05812871e-01 5.80531538e-01 7.43382573e-01 1.51879087e-01 -1.65873265e+00 5.14632761e-01 2.30039328e-01 4.42868054e-01 -8.56565416e-01 -7.87070692e-01 -3.27729732e-01 -6.60154164e-01 3.77639890e-01 6.97283924e-01 -3.71762961e-01 -9.31159735e-01 1.92875719e+00 5.62565684e-01 -4.29944426e-01 -7.31819570e-02 1.49238718e+00 5.64279377e-01 2.53276229e-01 -4.70604867e-01 2.93984026e-01 1.21120822e+00 -6.80724680e-01 -5.34790576e-01 -5.13531148e-01 2.25097820e-01 -8.09397995e-01 6.52065456e-01 5.37245929e-01 -1.19016421e+00 -6.59625173e-01 -7.66362906e-01 -2.54076064e-01 1.57862991e-01 2.30113924e-01 3.48510683e-01 3.79644334e-01 -8.54508698e-01 4.29197878e-01 -9.50192392e-01 -5.03797293e-01 -1.98584497e-01 3.51224512e-01 -9.18011129e-01 -1.35545596e-01 -9.65790153e-01 1.26342392e+00 2.06967220e-01 2.75355399e-01 -6.48938954e-01 -2.64802843e-01 -1.04725003e+00 -4.91400599e-01 3.92347991e-01 -1.41564441e+00 1.18944705e+00 -3.10711831e-01 -1.42875147e+00 1.04241216e+00 -9.94624794e-02 -1.57035738e-01 1.05013442e+00 -6.68523371e-01 3.39938849e-01 4.28115398e-01 1.49148718e-01 1.16712689e+00 6.46191299e-01 -1.54652333e+00 -1.93056166e-01 -7.74030507e-01 2.92581543e-02 7.24562585e-01 5.03109634e-01 -5.94698012e-01 -4.59027469e-01 -3.72570395e-01 6.81758463e-01 -1.42295635e+00 -2.52229452e-01 2.86551148e-01 -7.14609146e-01 3.75739718e-03 2.33124077e-01 -8.88364077e-01 3.71039629e-01 -1.58901680e+00 9.03288245e-01 1.64150462e-01 5.43084219e-02 -4.41564441e-01 8.50430802e-02 4.06844646e-01 4.84368540e-02 -1.48221478e-01 -1.32752433e-01 -5.64916670e-01 5.15023530e-01 2.90095627e-01 1.84795707e-01 9.12543356e-01 3.34329382e-02 7.94482648e-01 -5.82051337e-01 -5.23096979e-01 2.67417401e-01 7.17728734e-01 -5.91556013e-01 3.42544615e-01 -2.44066007e-02 6.35619402e-01 -6.56097382e-02 4.71015394e-01 6.50168538e-01 -2.07366973e-01 2.69288987e-01 -3.69179100e-01 2.28994146e-01 -3.38880308e-02 -1.43791878e+00 2.03254485e+00 -1.94731578e-01 2.01271325e-01 1.03879929e-01 -7.54363835e-01 6.66984677e-01 4.61983919e-01 5.42712331e-01 -2.16017157e-01 1.00435153e-01 2.34553203e-01 -1.41075581e-01 -4.26316947e-01 4.13272679e-01 -4.34892386e-01 -2.99448669e-01 3.72962177e-01 2.71455020e-01 -4.05947000e-01 -1.87662229e-01 -2.75462270e-02 7.77877212e-01 8.67857873e-01 9.72301662e-02 -6.58499897e-02 2.26332769e-02 -1.41492113e-01 3.41866255e-01 5.75542033e-01 -1.11557104e-01 1.25684345e+00 6.66606665e-01 -4.02720273e-01 -1.36791790e+00 -1.65801108e+00 -1.53190158e-02 3.75306666e-01 1.87399238e-01 3.00791841e-02 -6.28901362e-01 -4.53960210e-01 1.75922900e-01 4.64117378e-01 -6.14942193e-01 1.75023466e-01 -5.53917766e-01 -6.18457377e-01 3.74574989e-01 5.13242126e-01 3.40921342e-01 -5.77357173e-01 -8.50445330e-01 -1.43313572e-01 -6.90827549e-01 -1.31311893e+00 -3.99464637e-01 -9.52821132e-03 -7.35352337e-01 -1.23776114e+00 -1.34083736e+00 -6.32185817e-01 6.76399648e-01 -1.54282777e-02 1.19216156e+00 -4.40335542e-01 -2.22730502e-01 7.45167434e-01 -1.21102100e-02 -4.53946181e-02 -2.69203097e-01 -6.49580136e-02 4.96120125e-01 -4.15567845e-01 8.25880915e-02 -5.03097773e-01 -6.66932166e-01 7.68621087e-01 -4.26115066e-01 3.53656977e-01 4.35255110e-01 7.66121268e-01 6.05863392e-01 -2.33530208e-01 3.87927651e-01 -1.42216563e-01 3.56143415e-01 -3.86108160e-01 -3.94850552e-01 -8.63950253e-02 1.77889496e-01 7.48180076e-02 2.94166543e-02 -3.60286444e-01 -1.02085161e+00 6.90719187e-01 -2.42897198e-01 -6.23098195e-01 -5.49948335e-01 2.92710841e-01 -2.25895718e-01 2.91150063e-01 7.82300115e-01 -1.04381807e-01 5.88302493e-01 -5.81449747e-01 5.29247761e-01 4.10597116e-01 5.45456350e-01 -8.45212102e-01 7.07603335e-01 7.14445591e-01 3.26984584e-01 -7.26446092e-01 -6.72071218e-01 -2.76106805e-01 -1.12119758e+00 -4.53114003e-01 1.18942082e+00 -1.20464754e+00 -8.88197124e-01 5.41506171e-01 -1.42005575e+00 -1.76777408e-01 1.33176222e-01 7.54342556e-01 -1.25874674e+00 6.51817858e-01 -5.98310053e-01 -8.14192057e-01 5.41084334e-02 -1.34456766e+00 1.52931714e+00 -1.41299665e-01 -7.31576443e-01 -7.87824810e-01 1.30318463e-01 5.28719008e-01 -3.34258258e-01 5.90451479e-01 6.04308248e-01 -2.26945594e-01 -3.60746980e-01 -2.11783513e-01 1.02998927e-01 2.00766742e-01 -9.33875069e-02 -3.73088211e-01 -6.85066938e-01 -2.64816970e-01 -3.59748770e-03 -5.55861235e-01 4.83614266e-01 7.06689000e-01 4.28038716e-01 -9.46268998e-03 -2.18476191e-01 4.03597087e-01 1.03979242e+00 -3.41473132e-01 5.64356506e-01 2.43974432e-01 6.93418920e-01 1.06100249e+00 5.85135996e-01 6.03847444e-01 6.65695071e-01 9.36210871e-01 6.11462057e-01 1.28195509e-01 6.51249066e-02 -3.95960063e-01 2.02579364e-01 3.85058492e-01 -1.24396421e-01 -1.72868639e-01 -8.76418412e-01 4.76870507e-01 -2.01927805e+00 -6.99817002e-01 -1.26564533e-01 2.14692283e+00 5.69384098e-01 6.62215352e-02 5.37643015e-01 -5.99360131e-02 7.14067459e-01 -3.07032585e-01 -5.18697739e-01 2.99742371e-02 1.36568556e-02 -6.89643025e-02 2.91196823e-01 6.76823318e-01 -9.97614622e-01 4.93717581e-01 6.71592283e+00 3.44765484e-01 -4.53762442e-01 -1.64113581e-01 3.67443085e-01 -2.87086040e-01 -1.52325407e-01 -2.31188864e-01 -8.44074309e-01 2.18437523e-01 3.87116730e-01 4.23925459e-01 1.61933541e-01 6.82940066e-01 -1.35842571e-02 -4.53080684e-01 -1.39433384e+00 1.17058349e+00 2.17937663e-01 -8.91496062e-01 -1.03002086e-01 4.54397470e-01 5.74181676e-01 -7.83146843e-02 3.22041884e-02 -1.50056062e-02 1.85648710e-01 -9.94315445e-01 1.01590037e+00 6.11134112e-01 6.42402887e-01 -8.36698711e-01 5.74995816e-01 7.98719883e-01 -8.12930465e-01 4.07629222e-01 -2.29248419e-01 -1.60510078e-01 5.42065382e-01 4.60062921e-01 -8.73700678e-01 6.31402552e-01 6.71918809e-01 2.88719237e-01 -2.78449148e-01 8.85097146e-01 -3.87162536e-01 -2.08393589e-01 -5.45873821e-01 2.17481658e-01 -2.25444928e-01 -1.67956486e-01 5.83828390e-01 6.41400695e-01 4.41408694e-01 -2.14142632e-02 2.25526229e-01 1.04871988e+00 3.43047172e-01 -3.46115232e-01 -6.26515925e-01 3.06644499e-01 5.46788052e-02 9.92618442e-01 -6.00026548e-01 3.47179100e-02 -2.88138594e-02 1.27463531e+00 2.66363561e-01 2.96092778e-01 -9.60731983e-01 -4.42300923e-02 5.44035494e-01 1.26743272e-01 2.42227778e-01 -7.29340017e-01 -2.27181271e-01 -1.29805911e+00 3.54250908e-01 -8.90830934e-01 1.08339936e-01 -1.11371398e+00 -1.46232653e+00 4.43312854e-01 5.83606184e-01 -1.26901615e+00 -7.44851947e-01 -9.43293095e-01 4.23546731e-02 1.01496851e+00 -7.29573071e-01 -1.05690122e+00 -1.25898182e-01 3.02106649e-01 2.39522830e-01 2.01643616e-01 7.34351039e-01 3.38146128e-02 1.40518710e-01 1.12185068e-01 -3.92717689e-01 6.43736720e-02 8.82511735e-01 -1.28155804e+00 6.66819274e-01 5.03122449e-01 6.67367652e-02 5.18727779e-01 9.51007843e-01 -7.76579082e-01 -1.55601954e+00 -3.70483667e-01 6.82054579e-01 -8.95965457e-01 2.23455682e-01 -4.26522464e-01 -3.96810621e-01 8.49934816e-01 -5.37120402e-02 -1.07101277e-01 4.68720824e-01 2.78894939e-02 -9.91890281e-02 6.18974745e-01 -1.20410061e+00 5.49625874e-01 1.30101907e+00 -2.80799031e-01 -7.72021592e-01 4.40472633e-01 4.16598678e-01 -9.66634154e-01 -9.95649397e-01 5.62544763e-01 1.00431931e+00 -1.13188779e+00 1.32358372e+00 -5.06353498e-01 6.18655086e-01 -2.44504154e-01 -5.02377331e-01 -1.42449272e+00 -7.82898590e-02 -3.90147030e-01 -1.69722930e-01 4.81253594e-01 1.24223329e-01 -4.09483880e-01 1.02263570e+00 5.57346582e-01 4.80247512e-02 -6.87702715e-01 -1.26712239e+00 -8.47247958e-01 2.52090722e-01 -4.22561646e-01 7.93417916e-02 3.32930237e-01 -2.94608027e-01 3.27738851e-01 -6.74586236e-01 3.31249982e-01 9.83603537e-01 3.27706560e-02 1.25901484e+00 -1.23833418e+00 -6.22141719e-01 6.07349537e-02 -3.49485159e-01 -1.51815641e+00 3.88012558e-01 -4.90494430e-01 3.27865928e-01 -1.63535190e+00 1.49029136e-01 -4.13853675e-02 5.11256278e-01 4.64472845e-02 1.39235169e-01 3.89623523e-01 2.73034006e-01 5.95795289e-02 -4.14462954e-01 6.60143018e-01 1.45855868e+00 2.10919037e-01 1.41290933e-01 2.94062532e-02 -2.91513175e-01 9.81208384e-01 4.94467765e-01 -2.89778888e-01 -2.46534184e-01 -4.74652171e-01 4.42203581e-01 4.68665034e-01 8.47737432e-01 -1.00755620e+00 -6.51490763e-02 -3.53013128e-02 9.56837595e-01 -1.06211257e+00 1.18717849e+00 -6.26412630e-01 5.13340116e-01 4.65164095e-01 -1.29295792e-02 1.88585460e-01 -6.64992109e-02 5.97419620e-01 1.89087242e-01 -4.01258059e-02 6.05522156e-01 -6.00160778e-01 -4.21366483e-01 1.81305036e-01 -4.20491755e-01 2.48286515e-01 6.99679136e-01 -2.71453351e-01 -7.26989880e-02 -8.86790216e-01 -1.13583136e+00 1.51893169e-01 7.64171064e-01 4.50036258e-01 8.24890256e-01 -1.56759882e+00 -8.06258321e-01 1.70082822e-01 -1.22879259e-01 2.60319859e-01 2.73089051e-01 7.64240384e-01 -7.10048735e-01 3.16980094e-01 -3.63880426e-01 -1.11123633e+00 -1.06505620e+00 1.58379987e-01 6.37067914e-01 5.95781952e-02 -4.84662026e-01 7.99667478e-01 2.72174269e-01 -1.06374133e+00 9.51818600e-02 6.11000769e-02 3.51907790e-01 -2.14121029e-01 -1.09624639e-02 5.19326389e-01 -2.16525406e-01 -1.02022636e+00 -4.24283057e-01 8.58350158e-01 4.67170656e-01 -8.15241337e-01 1.18221784e+00 -2.33011022e-01 8.72095525e-02 3.76312792e-01 1.31414640e+00 -2.57505417e-01 -1.43409204e+00 8.92919153e-02 -5.59191346e-01 -4.52062935e-01 -6.15145564e-01 -6.54589832e-01 -6.62967741e-01 7.70142138e-01 4.05993432e-01 -3.33919257e-01 5.11041760e-01 2.32064813e-01 5.74404657e-01 4.03655589e-01 4.95000154e-01 -1.05499721e+00 3.56028527e-01 5.69750845e-01 1.36260080e+00 -1.25105917e+00 1.96596801e-01 -4.99932468e-01 -6.08522832e-01 1.13710070e+00 8.78553152e-01 -3.05704951e-01 6.15277886e-01 1.06379174e-01 1.83867425e-01 -1.05795071e-01 -4.71868843e-01 7.40033686e-02 5.91195047e-01 7.63206422e-01 3.64984930e-01 5.57834730e-02 -1.47602931e-01 3.39199156e-01 -5.12397766e-01 -2.78907746e-01 5.56271493e-01 9.63821828e-01 -2.97500700e-01 -1.03864205e+00 -9.51231897e-01 -6.70113713e-02 -1.40585929e-01 4.92885709e-01 -5.26469290e-01 1.06316388e+00 -8.90435651e-03 7.77473569e-01 7.83008933e-02 -3.72552365e-01 3.91889334e-01 1.92867175e-01 1.28757215e+00 -5.45246005e-01 1.43108010e-01 3.65845978e-01 1.55137330e-01 -5.80580473e-01 -4.54375625e-01 -7.09385157e-01 -1.22400212e+00 -7.59137943e-02 -1.07650444e-01 -1.98639110e-01 7.73077846e-01 1.05722451e+00 3.43808085e-01 5.01062796e-02 2.81185121e-03 -1.58224177e+00 -7.15650558e-01 -8.87579381e-01 -5.07284880e-01 4.63157207e-01 2.11808532e-01 -1.13716674e+00 -4.80609506e-01 -1.41179502e-01]
[7.028101921081543, -1.018384575843811]
6f166c88-cb79-4a15-bd3e-7fb940594ec3
peakconv-learning-peak-receptive-field-for
null
null
http://openaccess.thecvf.com//content/CVPR2023/html/Zhang_PeakConv_Learning_Peak_Receptive_Field_for_Radar_Semantic_Segmentation_CVPR_2023_paper.html
http://openaccess.thecvf.com//content/CVPR2023/papers/Zhang_PeakConv_Learning_Peak_Receptive_Field_for_Radar_Semantic_Segmentation_CVPR_2023_paper.pdf
PeakConv: Learning Peak Receptive Field for Radar Semantic Segmentation
The modern machine learning-based technologies have shown considerable potential in automatic radar scene understanding. Among these efforts, radar semantic segmentation (RSS) can provide more refined and detailed information including the moving objects and background clutters within the effective receptive field of the radar. Motivated by the success of convolutional networks in various visual computing tasks, these networks have also been introduced to solve RSS task. However, neither the regular convolution operation nor the modified ones are specific to interpret radar signals. The receptive fields of existing convolutions are defined by the object presentation in optical signals, but these two signals have different perception mechanisms. In classic radar signal processing, the object signature is detected according to a local peak response, i.e., CFAR detection. Inspired by this idea, we redefine the receptive field of the convolution operation as the peak receptive field (PRF) and propose the peak convolution operation (PeakConv) to learn the object signatures in an end-to-end network. By incorporating the proposed PeakConv layers into the encoders, our RSS network can achieve better segmentation results compared with other SoTA methods on a multi-view real-measured dataset collected from an FMCW radar. Our code for PeakConv is available at https://github.com/zlw9161/PKC.
['Zhe Ma', 'Xuhui Huang', 'Yuanpei Chen', 'Yufei Guo', 'Youcheng Zhang', 'Xinyan Zhang', 'Liwen Zhang']
2023-01-01
null
null
null
cvpr-2023-1
['scene-understanding']
['computer-vision']
[ 5.15010893e-01 -5.41574538e-01 2.80927420e-01 -7.32954443e-01 -2.25526899e-01 -4.93045866e-01 4.76676375e-01 -4.51816797e-01 -3.98144424e-01 3.32306772e-01 -4.87479232e-02 -3.83890659e-01 -2.99113333e-01 -8.28881323e-01 -5.39883494e-01 -8.04846585e-01 -8.18443149e-02 7.16028512e-02 3.31712186e-01 -1.16324022e-01 4.05487567e-01 7.56405711e-01 -1.37839305e+00 4.99386221e-01 8.61697733e-01 1.39482415e+00 5.61233580e-01 6.39661670e-01 -3.10648918e-01 7.21477509e-01 -7.42882907e-01 3.69307622e-02 4.98370349e-01 -2.71241277e-01 -1.85351908e-01 1.00240521e-01 5.01271665e-01 -2.47280970e-01 -5.95394731e-01 1.36544108e+00 3.43287200e-01 7.99318030e-02 4.81654465e-01 -1.07858062e+00 -7.14021444e-01 4.50061470e-01 -6.07082307e-01 6.67610109e-01 -2.47183323e-01 -5.85766807e-02 6.47244394e-01 -9.98817921e-01 2.67416418e-01 1.27761149e+00 3.59699100e-01 2.72148311e-01 -8.04241836e-01 -8.39456499e-01 8.83790478e-02 4.77240413e-01 -1.46262360e+00 -3.57831657e-01 8.69120657e-01 -3.78783643e-01 3.15893352e-01 4.72216189e-01 5.62548518e-01 6.64817810e-01 3.45833838e-01 8.32922339e-01 1.17490089e+00 -5.26191154e-03 7.16090128e-02 -1.76805735e-01 4.66217697e-01 3.91695470e-01 3.09176981e-01 4.40697819e-01 -3.04689184e-02 2.88423866e-01 9.40698147e-01 4.80537564e-01 -6.31790280e-01 -7.36026093e-02 -1.18933833e+00 8.27872872e-01 7.77119398e-01 2.94412315e-01 -3.70239854e-01 3.80399793e-01 -7.41864294e-02 -5.95545163e-03 2.50748813e-01 8.94068182e-02 -2.38389626e-01 5.13240576e-01 -1.04748070e+00 2.24772394e-01 3.70014668e-01 5.80128193e-01 7.76593506e-01 4.65632439e-01 -3.20604533e-01 6.10145748e-01 4.51362282e-01 9.34116006e-01 -2.52006743e-02 -6.24596655e-01 1.87225461e-01 4.47687268e-01 1.61996245e-01 -8.84651959e-01 -4.71948564e-01 -1.03111136e+00 -9.54311848e-01 4.46857244e-01 2.57041782e-01 -1.72098100e-01 -1.69512212e+00 1.18358302e+00 1.42901614e-01 4.83389765e-01 1.89068630e-01 1.55321229e+00 1.02131653e+00 9.00328100e-01 -5.62865734e-02 1.81616694e-01 1.54075527e+00 -6.02779746e-01 -6.53537571e-01 -5.96558690e-01 -1.30846977e-01 -8.43749106e-01 4.35152680e-01 3.47741127e-01 -4.72523034e-01 -9.10840869e-01 -1.20188820e+00 4.28315938e-01 -2.63814986e-01 3.34170341e-01 8.08792055e-01 5.05934298e-01 -5.50674736e-01 3.00631970e-01 -5.20383537e-01 -7.22098649e-02 8.61328006e-01 -1.17202409e-01 1.21523470e-01 -2.58580804e-01 -1.23349488e+00 6.12219870e-01 4.40789729e-01 9.49997127e-01 -9.98158336e-01 -6.76825821e-01 -5.49326301e-01 -1.13957170e-02 4.05255377e-01 -3.18065375e-01 9.65243340e-01 -1.11516511e+00 -9.83047307e-01 3.92162770e-01 1.53614685e-01 -6.33254290e-01 1.92349553e-01 -2.89255619e-01 -7.52561510e-01 2.42667317e-01 -1.55430660e-01 5.49857020e-01 1.29414904e+00 -1.36018968e+00 -7.04098821e-01 -1.04738340e-01 -5.11260889e-02 1.68776244e-01 4.65895295e-01 1.62368447e-01 -2.15626359e-01 -7.08497524e-01 3.04297417e-01 -3.09590340e-01 -3.06073785e-01 -2.19341546e-01 -3.02910268e-01 8.78446326e-02 9.94918764e-01 -6.63895726e-01 7.90741384e-01 -2.23826838e+00 -2.50047088e-01 4.03433472e-01 2.78714418e-01 5.90120792e-01 -3.73382084e-02 -4.69233915e-02 -1.91225976e-01 -2.13072956e-01 -5.28703809e-01 5.89878619e-01 -1.28172696e-01 -9.35389623e-02 -7.32098281e-01 5.01202941e-01 2.34062180e-01 9.70336437e-01 -5.35129368e-01 -3.50887984e-01 3.85217309e-01 3.69495958e-01 -1.24133736e-01 8.09676722e-02 -2.29702979e-01 4.64795411e-01 -6.72405064e-01 8.45283091e-01 1.34042728e+00 6.50606379e-02 -2.14917228e-01 -5.58830440e-01 -3.47370982e-01 -3.05580407e-01 -1.15501833e+00 1.21560836e+00 4.18450646e-02 8.92872453e-01 3.32448810e-01 -1.23999918e+00 1.25521839e+00 -1.25385180e-01 3.19881171e-01 -1.02366185e+00 2.27568403e-01 4.56419103e-02 4.26944643e-01 -3.99624497e-01 2.60577559e-01 -3.34036708e-01 -3.75411399e-02 1.77881587e-02 -1.55722693e-01 1.53868064e-01 -1.33864522e-01 -7.32147023e-02 8.96949887e-01 7.75545090e-02 -3.80207114e-02 -1.58342957e-01 6.00578368e-01 2.32762694e-01 7.19186842e-01 8.94064426e-01 -1.30280226e-01 5.68087876e-01 -2.04852045e-01 -6.09694242e-01 -5.78117251e-01 -1.41105497e+00 -5.70407033e-01 8.30361366e-01 4.59730744e-01 1.82076439e-01 -4.56276119e-01 -4.38999444e-01 -1.72358062e-02 6.39467120e-01 -2.13173717e-01 7.04152137e-02 -7.04744697e-01 -7.92969108e-01 5.00897944e-01 5.25325358e-01 9.02145028e-01 -1.28096509e+00 -1.12881470e+00 2.80694753e-01 -5.69618419e-02 -1.27276599e+00 -2.56421626e-01 1.68868154e-01 -6.35037005e-01 -9.46792841e-01 -5.73174179e-01 -6.79366112e-01 6.32179260e-01 7.02728868e-01 7.38507330e-01 -4.88295667e-02 -9.09325778e-01 3.06287169e-01 -4.75214690e-01 -7.39227593e-01 2.95172632e-01 -6.12326384e-01 -3.33858073e-01 4.00046796e-01 5.20473540e-01 -3.21586967e-01 -7.46502757e-01 1.98872536e-01 -1.03641343e+00 1.71730831e-01 1.14924693e+00 4.98684943e-01 3.53740692e-01 9.27384421e-02 4.23215002e-01 -5.80825090e-01 2.89718866e-01 -5.75475581e-02 -8.09993625e-01 1.83155745e-01 -1.34391841e-02 1.07953046e-02 5.34253240e-01 -1.40918210e-01 -1.12598515e+00 -3.72568332e-02 4.62871939e-02 -5.93620837e-01 -5.11917114e-01 6.21521473e-01 -1.53917372e-01 -1.58212692e-01 3.44126105e-01 4.81077850e-01 -3.08271289e-01 -3.90745819e-01 3.19571555e-01 7.88164258e-01 6.78116739e-01 -9.46094468e-02 1.15281892e+00 6.18251204e-01 -2.62297273e-01 -1.19778752e+00 -9.72380340e-01 -4.84542400e-01 -3.82779688e-01 -5.40090322e-01 1.06483400e+00 -8.00529659e-01 -6.69879854e-01 4.92125154e-01 -1.15482771e+00 1.22288978e-02 -7.58146048e-02 7.80504405e-01 -2.66932458e-01 2.77950317e-01 -6.37687221e-02 -1.02731502e+00 -5.29155970e-01 -7.67246544e-01 8.94286513e-01 6.95391715e-01 5.37512720e-01 -5.37680745e-01 -2.99546361e-01 2.76309967e-01 5.80336034e-01 9.96650085e-02 8.34632576e-01 -5.46056867e-01 -1.10816073e+00 -1.33542970e-01 -6.63142622e-01 4.82710361e-01 -1.34754181e-01 -3.25681031e-01 -9.69557762e-01 -2.38373280e-01 1.35176435e-01 2.83423871e-01 1.33170986e+00 8.71293783e-01 1.24502373e+00 8.44937563e-02 -3.68861049e-01 9.79696572e-01 1.63852990e+00 7.30651915e-01 1.02420151e+00 4.11443450e-02 5.82526684e-01 4.06368822e-01 8.21793914e-01 3.72122884e-01 -2.15693548e-01 5.20236135e-01 5.37366331e-01 -3.54755223e-01 -1.34834483e-01 2.18337759e-01 4.77761775e-01 2.01961875e-01 -2.26013243e-01 -1.62474170e-01 -8.32009733e-01 2.39584073e-01 -1.54485714e+00 -1.19636536e+00 -1.61923364e-01 2.01857853e+00 2.17669308e-01 2.15532050e-01 -4.57771510e-01 -2.16316566e-01 9.45064843e-01 5.68531036e-01 -4.89536464e-01 -1.84440672e-01 -1.87908992e-01 6.60019457e-01 7.81566501e-01 4.14003760e-01 -1.39162755e+00 9.11371112e-01 5.02284622e+00 8.45791340e-01 -1.37792420e+00 -7.59562552e-02 1.08972140e-01 1.44370630e-01 -1.45801932e-01 1.55347452e-01 -8.38725150e-01 3.43406588e-01 5.30811012e-01 1.51871473e-01 4.99215186e-01 5.08712113e-01 7.38980398e-02 -1.28264502e-01 -5.38601160e-01 9.85106170e-01 1.09464377e-01 -1.34258914e+00 1.70654226e-02 -1.25246614e-01 6.52239472e-02 -1.67080127e-02 5.44137359e-02 2.97500670e-01 2.39545312e-02 -1.28150856e+00 7.40979612e-01 9.80263591e-01 5.66450298e-01 -7.46365130e-01 8.37272406e-01 1.85806245e-01 -1.51242125e+00 -1.41417325e-01 -6.49956405e-01 1.95555344e-01 1.92541152e-01 7.67058372e-01 -8.86286914e-01 7.81615615e-01 6.76485896e-01 5.22777438e-01 -4.61412489e-01 1.28168619e+00 -2.08333313e-01 6.46431148e-01 -2.18701959e-02 1.30336091e-01 4.38085496e-01 -4.29734707e-01 8.33546996e-01 1.31143522e+00 3.37091833e-01 2.68813640e-01 2.88576752e-01 1.17333388e+00 3.81708831e-01 -3.70937556e-01 -2.89253891e-01 -2.60601878e-01 3.22002172e-01 1.62794948e+00 -8.54905009e-01 -7.23255798e-02 -2.53363192e-01 6.71150208e-01 -3.46511036e-01 6.98396742e-01 -1.24052036e+00 -9.68962252e-01 7.65128911e-01 5.90891540e-02 7.82604277e-01 -3.11683238e-01 -2.75160279e-02 -7.81942785e-01 -1.44252300e-01 -6.33786440e-01 1.00409277e-01 -9.09808278e-01 -1.23841631e+00 7.25666642e-01 8.04346651e-02 -1.45463192e+00 3.63507271e-01 -9.48121727e-01 -7.69172251e-01 1.08415353e+00 -1.76041663e+00 -1.08586884e+00 -7.36398101e-01 5.82038641e-01 5.03196001e-01 -1.58107802e-01 2.84004807e-01 2.93415219e-01 -3.35550040e-01 3.22201550e-02 -1.61765680e-01 5.86165249e-01 4.46510375e-01 -9.05427754e-01 3.77405316e-01 1.20261252e+00 1.34753361e-01 3.56122762e-01 6.09403551e-01 -6.88907683e-01 -1.34780347e+00 -1.27238584e+00 2.77736515e-01 1.58377141e-01 3.82416219e-01 -4.49409842e-01 -7.90340483e-01 4.55240399e-01 1.98744565e-01 1.48125157e-01 3.20068747e-01 -4.73753810e-01 -1.27817273e-01 -4.14900035e-01 -7.99007356e-01 2.97848314e-01 1.01313651e+00 -1.52131692e-01 -8.49292517e-01 5.56509756e-02 5.00762165e-01 -3.12260896e-01 -2.23065689e-01 6.60902381e-01 3.37867588e-01 -8.74078691e-01 1.29743695e+00 -3.62009674e-01 4.58997823e-02 -8.88709843e-01 -3.91671926e-01 -1.03699744e+00 -5.62018216e-01 -9.42765400e-02 8.12262595e-02 8.62237334e-01 -6.31380593e-04 -6.88705742e-01 5.98234475e-01 -4.30891931e-01 -5.66405892e-01 -5.14033318e-01 -8.02194595e-01 -6.06659591e-01 -4.64646757e-01 -6.30105257e-01 3.59643906e-01 5.25077701e-01 -7.72577882e-01 2.01676339e-01 -3.18506211e-01 7.75464892e-01 1.06222761e+00 8.97983015e-01 4.08449620e-01 -1.60659087e+00 -1.32735521e-01 -3.12929124e-01 -5.95677853e-01 -1.32660329e+00 -1.36730686e-01 -1.05896747e+00 3.43519837e-01 -1.74444032e+00 -1.13181397e-02 -3.95244777e-01 -4.54620242e-01 2.16727778e-01 8.45344588e-02 2.67391413e-01 3.63517672e-01 1.40101552e-01 -3.89251024e-01 4.14659500e-01 1.40124619e+00 -3.70091140e-01 -6.49579689e-02 1.77692175e-01 -5.85869133e-01 7.63642251e-01 8.45916510e-01 -3.21774065e-01 -1.90294739e-02 -4.04262990e-01 -2.16831446e-01 -4.23548743e-03 8.64812076e-01 -1.30989170e+00 4.63086098e-01 -3.11430752e-01 9.23529446e-01 -1.18747687e+00 3.26251984e-01 -9.22734618e-01 6.82219863e-02 5.60234487e-01 5.00301011e-02 -3.83731842e-01 1.50989771e-01 6.54461682e-01 -3.10920119e-01 -2.20518872e-01 7.62349546e-01 -1.49496630e-01 -1.25656867e+00 2.89612532e-01 -5.27790368e-01 -9.52113271e-02 9.36837733e-01 -5.46781778e-01 -5.83886445e-01 -1.85978860e-01 -5.67418158e-01 2.50344694e-01 -3.09355974e-01 4.22793031e-01 1.16267991e+00 -1.25373292e+00 -8.88472855e-01 5.44864714e-01 -7.07178116e-02 -1.07820831e-01 6.41527653e-01 9.32243764e-01 -5.15570045e-01 5.84378242e-01 -5.48612654e-01 -6.90684199e-01 -1.02541637e+00 1.98846087e-01 6.41704738e-01 9.38633084e-02 -7.85238206e-01 8.33895564e-01 6.06214643e-01 -3.04905176e-02 1.19991973e-01 -3.88126075e-01 -4.23077524e-01 -1.18940044e-02 6.11881077e-01 7.14471862e-02 -1.82801858e-01 -3.95528316e-01 -6.26044989e-01 6.46063805e-01 -5.69420233e-02 -3.42968218e-02 1.34007764e+00 2.38464504e-01 -7.72903487e-03 3.79882865e-02 6.61950827e-01 -1.88755505e-02 -1.30917847e+00 -3.54439408e-01 -1.10974684e-01 -5.26295722e-01 4.18457150e-01 -8.70017767e-01 -1.30487180e+00 1.03997684e+00 1.00467288e+00 5.89134693e-02 1.35452163e+00 -8.93684924e-02 6.65898681e-01 3.07954848e-01 3.05588812e-01 -8.89296055e-01 -1.05976999e-01 7.05509603e-01 8.70374501e-01 -8.85602057e-01 4.25448082e-02 -4.45611775e-01 -8.08642805e-01 1.38318944e+00 4.24104542e-01 -2.97329098e-01 6.20232046e-01 4.66102064e-01 2.14863956e-01 -4.52010363e-01 -2.28460282e-01 -6.84596360e-01 4.57388997e-01 6.62079990e-01 7.53831342e-02 1.77919656e-01 -7.54085928e-02 8.14844489e-01 1.04037456e-01 -2.38258138e-01 2.85940677e-01 8.37853134e-01 -1.05743372e+00 -5.38283765e-01 -6.51712418e-01 5.36823273e-01 -1.67962328e-01 -2.24437669e-01 -1.79646671e-01 5.76122165e-01 4.31616485e-01 9.12541687e-01 3.84974390e-01 -5.75364590e-01 3.57275426e-01 -7.88655952e-02 4.71043438e-01 -4.52490598e-01 -2.76856214e-01 2.85629034e-01 -3.02447379e-01 -3.54017645e-01 -4.74842072e-01 -4.24490243e-01 -1.58071089e+00 1.13842227e-01 -2.04043418e-01 1.81927845e-01 5.87751269e-01 9.34908032e-01 6.27604052e-02 8.90794694e-01 6.94531858e-01 -7.34018624e-01 -5.09709477e-01 -8.25639725e-01 -7.23338246e-01 -2.03340296e-02 2.93675900e-01 -6.65641546e-01 -3.04868013e-01 -8.16835761e-02]
[8.294219970703125, -1.203277349472046]
071cdca0-694e-4d1b-9545-dad8712fb959
ousiometrics-and-telegnomics-the-essence-of
2110.06847
null
https://arxiv.org/abs/2110.06847v2
https://arxiv.org/pdf/2110.06847v2.pdf
Ousiometrics and Telegnomics: The essence of meaning conforms to a two-dimensional powerful-weak and dangerous-safe framework with diverse corpora presenting a safety bias
We define `ousiometrics' to be the study of essential meaning in whatever context that meaningful signals are communicated, and `telegnomics' as the study of remotely sensed knowledge. From work emerging through the middle of the 20th century, the essence of meaning has become generally accepted as being well captured by the three orthogonal dimensions of evaluation, potency, and activation (EPA). By re-examining first types and then tokens for the English language, and through the use of automatically annotated histograms -- `ousiograms' -- we find here that: 1. The essence of meaning conveyed by words is instead best described by a compass-like power-danger (PD) framework, and 2. Analysis of a disparate collection of large-scale English language corpora -- literature, news, Wikipedia, talk radio, and social media -- shows that natural language exhibits a systematic bias toward safe, low danger words -- a reinterpretation of the Pollyanna principle's positivity bias for written expression. To help justify our choice of dimension names and to help address the problems with representing observed ousiometric dimensions by bipolar adjective pairs, we introduce and explore `synousionyms' and `antousionyms' -- ousiometric counterparts of synonyms and antonyms. We further show that the PD framework revises the circumplex model of affect as a more general model of state of mind. Finally, we use our findings to construct and test a prototype `ousiometer', a telegnomic instrument that measures ousiometric time series for temporal corpora. We contend that our power-danger ousiometric framework provides a complement for entropy-based measurements, and may be of value for the study of a wide variety of communication across biological and artificial life.
['C. M. Danforth', 'A. J. Reagan', 'M. V. Arnold', 'J. R. Minot', 'S. Beaulieu', 'J. Lovato', 'J. W. Zimmerman', 'M. I. Fudolig', 'T. Alshaabi', 'P. S. Dodds']
2021-10-13
null
null
null
null
['artificial-life']
['miscellaneous']
[ 3.21944684e-01 7.08769634e-03 -3.53185773e-01 -2.77885914e-01 2.15816759e-02 -7.47202456e-01 9.20524776e-01 5.38692296e-01 -6.00579083e-01 5.52740395e-01 7.67733097e-01 -3.64795506e-01 -4.42300916e-01 -7.76414394e-01 3.71931121e-02 -6.42350614e-01 -1.74087465e-01 1.04239449e-01 -3.52736801e-01 -6.74577892e-01 4.69572008e-01 4.20812547e-01 -1.14269495e+00 -1.06816785e-02 2.27643892e-01 1.03582728e+00 -2.24650651e-01 3.13723624e-01 9.46380496e-02 5.28161526e-01 -6.11255646e-01 -4.44052875e-01 -1.43086061e-01 -5.35591841e-01 -7.76634276e-01 -4.24868077e-01 -2.84646809e-01 1.62198335e-01 4.10479144e-04 9.12771046e-01 4.69158739e-01 -1.96187258e-01 6.51012778e-01 -1.01066470e+00 -9.13636744e-01 4.95179951e-01 -3.95600080e-01 4.00762439e-01 5.88585258e-01 7.82677010e-02 1.29190803e+00 -3.03729415e-01 7.13506520e-01 1.48188114e+00 7.93873370e-01 4.50039148e-01 -1.57309484e+00 -5.78214586e-01 -2.07707379e-02 -2.79601872e-01 -1.20746446e+00 -3.16953212e-01 5.29469609e-01 -6.93972707e-01 9.16324735e-01 6.19491458e-01 1.24475670e+00 1.40731430e+00 4.98746127e-01 3.37572306e-01 1.46753526e+00 -1.27320752e-01 2.82103747e-01 2.16309309e-01 1.50533319e-01 1.51665434e-01 4.01813149e-01 4.04632986e-01 -7.49609113e-01 -6.04325354e-01 5.47744811e-01 -1.95541129e-01 -3.57555360e-01 3.20844531e-01 -1.39705646e+00 8.62764418e-01 1.53568044e-01 6.87978983e-01 -6.14142954e-01 1.72757819e-01 6.94929481e-01 4.42595363e-01 6.84026837e-01 8.52269530e-01 -5.91131508e-01 -5.27763963e-01 -5.14588654e-01 2.30555564e-01 9.21699464e-01 1.85939848e-01 2.52902299e-01 -8.59204978e-02 3.69820088e-01 7.95834363e-01 4.40962672e-01 5.48538029e-01 7.85298049e-01 -8.35850894e-01 -2.19951108e-01 3.33239853e-01 4.84463610e-02 -1.25455391e+00 -8.26299608e-01 -2.60692567e-01 -6.50783002e-01 -1.75246596e-01 1.86464012e-01 -1.95625037e-01 -2.16548875e-01 2.30548382e+00 -2.78491843e-02 -4.08473372e-01 5.92225455e-02 8.21810663e-01 5.09479940e-01 5.46742260e-01 5.93919694e-01 -6.66832328e-01 1.68544030e+00 4.45880622e-01 -6.49944484e-01 -2.28331670e-01 5.64293921e-01 -3.75944465e-01 1.38843203e+00 5.23210049e-01 -8.32668185e-01 1.31865665e-01 -9.75926936e-01 9.53567922e-02 -5.34876645e-01 -7.71507323e-01 8.77813041e-01 9.01562512e-01 -1.01236773e+00 6.87384009e-01 -5.53373456e-01 -6.86251938e-01 5.36880381e-02 -1.90721750e-01 -3.43708813e-01 8.15600574e-01 -1.33287764e+00 1.17220151e+00 2.21470669e-01 2.02566199e-02 -2.52684444e-01 -5.39904475e-01 -6.04211748e-01 -1.70561850e-01 -1.03303559e-01 -5.04002929e-01 1.06593728e+00 -1.17997134e+00 -1.32217467e+00 1.29731071e+00 2.95174003e-01 -7.62133002e-02 -1.88354000e-01 -1.05441278e-02 -6.97075069e-01 8.22234303e-02 3.62888351e-03 3.27245116e-01 3.87080163e-01 -1.06509709e+00 4.12836447e-02 -5.56850433e-01 8.66666436e-02 2.43117899e-01 -3.84929746e-01 3.79100889e-01 5.65182090e-01 -8.72990847e-01 4.08520728e-01 -8.29018354e-01 1.02202423e-01 -2.82177571e-02 -2.04946071e-01 -8.11861977e-02 2.99265265e-01 -5.24242997e-01 1.46171582e+00 -2.22227335e+00 1.23113811e-01 4.89758164e-01 4.44046378e-01 -3.17743093e-01 -1.88765638e-02 7.74857461e-01 -5.74566543e-01 5.60804784e-01 -4.87076014e-01 3.88494015e-01 2.79413521e-01 3.50823820e-01 -4.11354423e-01 8.77849162e-01 -8.98074359e-02 8.68373394e-01 -1.19156086e+00 -1.71416879e-01 -8.83522034e-02 4.68222320e-01 -4.64504719e-01 -2.89032549e-01 1.02274530e-01 1.17829554e-01 -2.09600046e-01 6.26344383e-01 -1.23204002e-02 -7.59645700e-02 4.00132924e-01 -2.40667220e-02 -1.59484580e-01 5.50673306e-01 -3.87791812e-01 1.41674471e+00 -3.04539025e-01 9.73337531e-01 -8.66734982e-03 -7.46087015e-01 8.56460810e-01 4.37349558e-01 4.39720243e-01 -8.27937424e-01 3.80263954e-01 1.29831865e-01 3.76213700e-01 -6.50841355e-01 4.50989276e-01 -9.01182592e-01 -4.57422942e-01 6.19437516e-01 -2.25633234e-01 -5.32741904e-01 -2.33411387e-01 1.55882940e-01 9.27867174e-01 -7.94300213e-02 6.53873622e-01 -7.34161377e-01 1.45326003e-01 -2.53331065e-01 4.85607117e-01 1.48339763e-01 -4.39679742e-01 4.75580432e-02 9.00193691e-01 -6.27703071e-02 -8.26834679e-01 -1.24526882e+00 -6.64548039e-01 1.21258032e+00 -8.20134431e-02 -5.35495579e-01 -4.38871562e-01 2.06434071e-01 7.75424181e-04 1.14586794e+00 -7.78518915e-01 -2.76768655e-01 7.88286775e-02 -1.34465873e+00 9.24508095e-01 2.61623472e-01 -6.30358458e-02 -9.39111173e-01 -1.14509046e+00 6.00585435e-03 -2.36348689e-01 -7.63348162e-01 1.12762503e-01 3.68525743e-01 -6.47773027e-01 -6.31064653e-01 -2.73885965e-01 -8.56896639e-02 5.41368090e-02 -2.78316766e-01 1.20062613e+00 -1.39479935e-01 -2.31663600e-01 6.84131086e-01 -4.72863525e-01 -6.22528255e-01 -4.10565972e-01 -5.63571393e-01 2.17865407e-01 -2.92275935e-01 5.67246616e-01 -1.17868781e+00 -6.17745459e-01 -2.10452870e-01 -1.17750800e+00 -4.00065899e-01 3.36076528e-01 6.17094159e-01 3.13995481e-02 -4.32626605e-01 7.49859929e-01 -6.85392976e-01 1.15933979e+00 -8.91091287e-01 1.28566641e-02 -2.58465528e-01 -6.77647650e-01 -1.04224816e-01 1.45822629e-01 -4.48686540e-01 -7.34315753e-01 -9.55323994e-01 -2.14573264e-01 2.85679370e-01 -4.51934449e-02 8.81251097e-01 9.64425579e-02 1.68206155e-01 9.68779624e-01 1.08319707e-01 -5.08531965e-02 9.40540284e-02 5.27234137e-01 9.18500602e-01 4.90675181e-01 -8.27613890e-01 3.42827320e-01 6.16508305e-01 1.45999804e-01 -1.34977043e+00 -5.83150983e-01 -3.60424399e-01 -1.92615286e-01 -3.05948257e-01 9.79742408e-01 -7.91727126e-01 -1.23274553e+00 2.18833350e-02 -1.07012284e+00 -1.52660072e-01 -4.30800229e-01 5.86559057e-01 -7.11124957e-01 3.48559350e-01 -6.89438999e-01 -1.17145634e+00 -2.57641464e-01 -4.39639479e-01 9.98152137e-01 -2.00102091e-01 -1.07195044e+00 -1.20768690e+00 5.31650603e-01 -1.12320155e-01 4.36450601e-01 6.83970392e-01 1.03490055e+00 -7.91659474e-01 4.23180819e-01 -3.76859158e-02 -8.04591551e-02 1.13262206e-01 9.17290375e-02 -1.67224020e-01 -1.18102849e+00 1.60389170e-01 7.72445261e-01 -4.96722072e-01 4.11910683e-01 3.95002663e-02 6.02168083e-01 -4.36771095e-01 5.86698949e-03 3.30679059e-01 1.52778852e+00 2.41252422e-01 6.61149561e-01 2.79671431e-01 -9.84322838e-03 9.57495749e-01 2.38227725e-01 7.08852828e-01 2.94393927e-01 2.91177154e-01 2.09252208e-01 3.73198062e-01 6.40352428e-01 -1.13596275e-01 4.86601382e-01 1.06723106e+00 -1.25460848e-01 -2.57383883e-01 -8.89624536e-01 5.86062670e-01 -1.35558856e+00 -1.23545778e+00 -1.14969127e-01 2.10696363e+00 1.03154957e+00 1.30085588e-01 3.78658473e-01 2.00545356e-01 3.47094029e-01 4.57253247e-01 -1.11719593e-01 -9.25793946e-01 -1.78069681e-01 2.41718724e-01 4.43810493e-01 4.05933887e-01 -5.33665240e-01 5.53721189e-01 7.18916655e+00 4.34098125e-01 -1.23560178e+00 -4.81046662e-02 5.24224937e-01 -6.63538054e-02 -7.17974544e-01 1.81118064e-02 1.12095706e-01 4.59558696e-01 1.34781909e+00 -4.78265762e-01 3.60867918e-01 4.49226081e-01 4.43375379e-01 -3.72116446e-01 -1.13652039e+00 9.55977499e-01 5.85011765e-02 -5.57464778e-01 -3.27844858e-01 -3.14600803e-02 9.07380581e-02 4.31714095e-02 1.40833244e-01 -3.18727434e-01 2.27159008e-01 -9.51554060e-01 9.86712158e-01 4.67044681e-01 9.23650980e-01 -2.89056718e-01 2.94752628e-01 4.62456383e-02 -6.74833536e-01 7.53933983e-03 -3.00591677e-01 -6.35822415e-01 2.47484192e-01 6.38454676e-01 -4.12206978e-01 2.11899459e-01 3.38686794e-01 5.52808940e-01 -5.84229641e-02 4.56110150e-01 -1.59278646e-01 5.95332146e-01 -4.81153816e-01 -4.08812732e-01 1.69338316e-01 -2.43887663e-01 7.66925037e-01 1.52496934e+00 1.64141089e-01 4.43448454e-01 -5.20805478e-01 9.77514505e-01 4.56557482e-01 2.58900523e-01 -8.38804483e-01 -7.24289536e-01 4.45578277e-01 1.29921126e+00 -1.02887654e+00 -2.74009138e-01 -3.25461447e-01 7.69473672e-01 -1.85654551e-01 2.63748258e-01 -6.48287714e-01 -2.97070384e-01 9.13062036e-01 -1.76934619e-02 -4.06807572e-01 -2.59135216e-01 -5.88318706e-01 -1.11922956e+00 -2.88406461e-01 -7.35582352e-01 7.01795146e-02 -9.32023883e-01 -1.43944681e+00 3.27823907e-01 2.20178559e-01 -7.54337966e-01 -5.24318069e-02 -7.18251824e-01 -3.42060000e-01 7.86441922e-01 -7.64313638e-01 -6.98959291e-01 1.76854029e-01 2.33107075e-01 -1.16770603e-01 4.94127095e-01 1.11597419e+00 -4.51235697e-02 -3.26131701e-01 1.78345919e-01 1.46243768e-03 -2.97057360e-01 5.04037499e-01 -1.32156837e+00 3.21159601e-01 1.26411557e-01 -1.50009006e-01 8.99911523e-01 1.15916741e+00 -5.79634607e-01 -1.25684655e+00 -3.67124885e-01 1.00682902e+00 -6.67693615e-01 1.25596464e+00 -3.50254685e-01 -5.61412275e-01 6.29902542e-01 2.08696812e-01 -7.64011681e-01 1.20564580e+00 2.93584049e-01 -4.77229774e-01 1.44749478e-01 -1.07869959e+00 7.87209272e-01 1.14132690e+00 -8.85672569e-01 -1.00609779e+00 3.87147695e-01 7.47105300e-01 2.25473017e-01 -1.17748737e+00 1.46391600e-01 1.08405674e+00 -1.01153409e+00 7.33932018e-01 -4.89774138e-01 2.97824144e-01 2.05141962e-01 -5.41372120e-01 -1.34280682e+00 -3.78563017e-01 -6.83956385e-01 6.15973711e-01 9.34063971e-01 2.16032282e-01 -9.77151155e-01 1.42822087e-01 6.45729363e-01 -8.81588981e-02 -7.30054855e-01 -9.47315454e-01 -5.65020919e-01 2.12446794e-01 -9.03643072e-01 3.68392080e-01 1.36378074e+00 1.10371089e+00 6.24855876e-01 7.22162277e-02 -4.82648700e-01 2.68988729e-01 -4.42556441e-01 1.25673577e-01 -1.56926918e+00 -3.48984510e-01 -7.30548978e-01 -7.29244590e-01 -5.45929432e-01 9.71495360e-02 -9.28992867e-01 -6.92460313e-02 -9.20971513e-01 1.18683316e-01 -9.46885571e-02 -1.69209644e-01 2.06717476e-01 3.82874846e-01 2.44923130e-01 1.64709106e-01 8.00575688e-02 6.09457307e-03 4.10014808e-01 9.12828743e-01 3.82768214e-01 -2.86148965e-01 -7.12868512e-01 -1.31129682e+00 9.49957132e-01 7.35724509e-01 -3.04132164e-01 -5.37835360e-01 5.98212369e-02 1.13383150e+00 1.30883053e-01 6.18598998e-01 -4.50446218e-01 -4.10537153e-01 -5.97764730e-01 1.52485251e-01 1.74206421e-01 5.20994663e-01 -7.22577095e-01 3.58095229e-01 6.08592093e-01 -3.59236956e-01 3.08876991e-01 3.15454870e-01 4.21901643e-01 2.51732051e-01 6.00529015e-02 6.26091421e-01 -2.39558026e-01 -2.46713087e-01 -3.56069982e-01 -7.82600105e-01 2.14364499e-01 7.32600629e-01 -1.59950644e-01 -5.71147203e-01 -4.61438209e-01 -5.83348215e-01 -1.67805478e-01 6.39051020e-01 1.65901393e-01 3.93620431e-01 -1.20905113e+00 -4.84472901e-01 -7.67412484e-02 -3.71864811e-03 -9.52820003e-01 1.61306504e-02 1.12671578e+00 -4.33794051e-01 4.01447326e-01 -2.12621465e-01 -2.86930710e-01 -6.90801382e-01 4.66559589e-01 2.04430923e-01 3.80295038e-01 -4.02781934e-01 7.46671915e-01 6.27630234e-01 -1.19533129e-01 -3.12680215e-01 -5.22758722e-01 -6.39604926e-02 5.32651603e-01 3.22140664e-01 3.75487119e-01 -5.76963186e-01 -9.20000255e-01 -5.14971852e-01 3.32168460e-01 5.36916018e-01 -7.80852437e-01 1.30005693e+00 -3.32982898e-01 -7.33952880e-01 1.29853249e+00 1.05559492e+00 -4.08114344e-02 -2.73411512e-01 3.23880374e-01 1.05704233e-01 -9.46866199e-02 -1.77531660e-01 -9.22388375e-01 -2.32714161e-01 6.22600138e-01 3.67070913e-01 6.27629638e-01 1.04752326e+00 3.59591916e-02 4.04936224e-01 2.59506464e-01 3.71646099e-02 -1.08405066e+00 -2.05901995e-01 2.52075225e-01 1.15285766e+00 -3.99852306e-01 3.01443562e-02 -1.49724543e-01 -6.29792929e-01 8.67321253e-01 1.11901358e-01 3.64758223e-02 8.74070287e-01 1.50488958e-01 6.47995546e-02 -6.07613444e-01 -1.04174554e+00 -7.63656273e-02 -1.11059532e-01 3.81680042e-01 9.47324932e-01 3.51197779e-01 -1.18118119e+00 8.44323397e-01 -8.13691199e-01 -2.00991228e-01 6.48925424e-01 7.43120015e-01 -4.91683304e-01 -7.85030067e-01 -4.66721416e-01 5.29375613e-01 -6.38913274e-01 -2.13977590e-01 -9.26553786e-01 7.48123586e-01 1.53258905e-01 9.52026784e-01 2.44044468e-01 -6.41752124e-01 1.16654010e-02 3.49148721e-01 4.17408615e-01 -4.08970177e-01 -6.78448200e-01 3.14623326e-01 3.98058593e-01 -5.71687460e-01 -8.59518290e-01 -6.04807734e-01 -1.08639145e+00 -5.52399457e-01 -1.84331909e-01 2.56779432e-01 6.28889501e-01 9.90355134e-01 -2.57330462e-02 2.77672082e-01 2.16148466e-01 -6.25891447e-01 -4.52042133e-01 -9.89321649e-01 -1.11796045e+00 4.03608590e-01 1.52473375e-01 -4.69674855e-01 -4.56252962e-01 -1.51268959e-01]
[9.882827758789062, 8.699287414550781]
c69cb337-9ad5-41f0-b15e-7e272b42f921
learning-to-remove-clutter-in-real-world-gpr
2205.08135
null
https://arxiv.org/abs/2205.08135v1
https://arxiv.org/pdf/2205.08135v1.pdf
Learning to Remove Clutter in Real-World GPR Images Using Hybrid Data
The clutter in the ground-penetrating radar (GPR) radargram disguises or distorts subsurface target responses, which severely affects the accuracy of target detection and identification. Existing clutter removal methods either leave residual clutter or deform target responses when facing complex and irregular clutter in the real-world radargram. To tackle the challenge of clutter removal in real scenarios, a clutter-removal neural network (CR-Net) trained on a large-scale hybrid dataset is presented in this study. The CR-Net integrates residual dense blocks into the U-Net architecture to enhance its capability in clutter suppression and target reflection restoration. The combination of the mean absolute error (MAE) loss and the multi-scale structural similarity (MS-SSIM) loss is used to effectively drive the optimization of the network. To train the proposed CR-Net to remove complex and diverse clutter in real-world radargrams, the first large-scale hybrid dataset named CLT-GPR dataset containing clutter collected by different GPR systems in multiple scenarios is built. The CLT-GPR dataset significantly improves the generalizability of the network to remove clutter in real-world GPR radargrams. Extensive experimental results demonstrate that the CR-Net achieves superior performance over existing methods in removing clutter and restoring target responses in diverse real-world scenarios. Moreover, the CR-Net with its end-to-end design does not require manual parameter tuning, making it highly suitable for automatically producing clutter-free radargrams in GPR applications. The CLT-GPR dataset and the code implemented in the paper can be found at https://haihan-sun.github.io/GPR.html.
['Zheng Fan', 'Weixia Cheng', 'Hai-Han Sun']
2022-05-17
null
null
null
null
['gpr', 'ms-ssim', 'gpr']
['computer-vision', 'computer-vision', 'miscellaneous']
[ 4.01943386e-01 -4.64904875e-01 7.19710350e-01 -3.66273195e-01 -1.04661834e+00 -2.93442488e-01 7.76794776e-02 -3.22493941e-01 -1.66755900e-01 4.88886058e-01 1.44668579e-01 -4.66893584e-01 -2.87062764e-01 -9.38158751e-01 -4.27170753e-01 -1.05137479e+00 -3.96149904e-01 1.10522851e-01 1.39214635e-01 -4.10915077e-01 3.83492783e-02 9.99961376e-01 -1.25328326e+00 2.46974543e-01 9.25122440e-01 1.23744059e+00 5.07492602e-01 4.19073492e-01 4.35037553e-01 2.75495529e-01 -4.93333369e-01 8.12698156e-02 6.88359737e-01 5.21817207e-02 2.12883592e-01 -2.81157136e-01 5.54200709e-01 -2.46636301e-01 -5.04218459e-01 1.16467130e+00 1.11178458e+00 2.13279411e-01 5.99930227e-01 -6.06391013e-01 -6.68528199e-01 2.87016004e-01 -8.58099461e-01 5.11004925e-01 2.76968889e-02 -4.21478823e-02 4.74256068e-01 -1.38489830e+00 -1.10371672e-01 1.34189332e+00 1.08831513e+00 1.42483637e-01 -1.04736042e+00 -1.06943595e+00 2.21236408e-01 -1.02171347e-01 -1.46131861e+00 -3.99457753e-01 7.06479013e-01 -4.21859056e-01 4.80799377e-01 5.53055704e-01 6.43905878e-01 1.05953085e+00 4.60602254e-01 6.25767767e-01 8.17726254e-01 -2.87193358e-01 3.59110720e-02 -5.91167510e-01 4.81540680e-01 2.01174021e-01 9.62428927e-01 7.56823659e-01 -9.65723023e-03 -3.15384626e-01 9.14850235e-01 1.09886363e-01 -8.79447579e-01 -1.24985263e-01 -7.07064211e-01 8.79034519e-01 7.91646957e-01 3.70581150e-02 -5.81934333e-01 -1.39002070e-01 2.71685004e-01 2.59944052e-01 5.44377267e-01 4.80368465e-01 -1.93554416e-01 6.69290245e-01 -8.04272354e-01 4.04939920e-01 6.38953269e-01 7.45839059e-01 1.93003729e-01 9.36839640e-01 -3.20456445e-01 1.10981095e+00 3.89821798e-01 1.26029873e+00 -1.91479847e-02 -1.75173387e-01 3.86386305e-01 -9.17544868e-03 2.78448284e-01 -1.25397384e+00 -6.26208425e-01 -1.09769654e+00 -1.19680464e+00 1.22276701e-01 -2.44963914e-02 -4.59077984e-01 -1.34286475e+00 1.42826617e+00 3.16823125e-01 3.11894983e-01 3.89442831e-01 1.28826535e+00 1.05514753e+00 5.62633812e-01 -7.01873675e-02 -1.37265474e-01 1.23904097e+00 -3.29452187e-01 -6.35165095e-01 -7.08692014e-01 1.70489587e-02 -9.42141056e-01 5.15836000e-01 3.50323260e-01 -5.59717238e-01 -5.22945702e-01 -1.18878472e+00 9.29593146e-01 2.22022552e-02 6.09058961e-02 6.54222608e-01 6.78441942e-01 -5.39006770e-01 2.51893066e-02 -6.74944341e-01 1.37397289e-01 4.51629192e-01 -8.58470276e-02 4.95004840e-02 -7.42815793e-01 -1.51641858e+00 8.71453047e-01 1.88599572e-01 1.12852550e+00 -1.08460188e+00 -1.04248798e+00 -8.82635951e-01 -2.51844265e-02 3.23689610e-01 -2.95658380e-01 8.79176199e-01 -5.36471426e-01 -9.02584791e-01 1.59781560e-01 3.82495582e-01 -2.33546451e-01 4.57745671e-01 -7.11515784e-01 -9.31764364e-01 1.51672335e-02 1.04355112e-01 -6.74528405e-02 1.14657021e+00 -1.66617560e+00 -4.26845491e-01 -3.15504700e-01 -6.87383413e-01 8.85963887e-02 4.21227485e-01 -1.61283240e-01 -4.94297147e-02 -1.15903687e+00 4.25075531e-01 -5.60697198e-01 -3.83643866e-01 -5.36096752e-01 -3.13912511e-01 7.55777717e-01 8.60235691e-01 -7.36783087e-01 7.46977568e-01 -2.15237927e+00 -6.21776342e-01 4.74476814e-01 9.27198902e-02 6.12172484e-01 -4.60967481e-01 2.92215437e-01 -3.92287195e-01 -5.22250473e-01 -5.21723747e-01 4.55247790e-01 -3.47114921e-01 -1.69089094e-01 -5.99740982e-01 7.20049798e-01 3.24605554e-01 6.53585613e-01 -6.59739733e-01 3.44416231e-01 6.93770796e-02 6.81278586e-01 -3.24988931e-01 1.75747067e-01 1.37982458e-01 5.07971406e-01 -7.02593267e-01 1.03994620e+00 1.62888551e+00 3.98317844e-01 -1.94856256e-01 -5.78620672e-01 -1.38767123e-01 -4.12113786e-01 -1.41227329e+00 7.18619764e-01 -3.07577074e-01 5.03380299e-01 6.63022578e-01 -9.00391698e-01 1.27118039e+00 1.12256363e-01 2.21428573e-01 -1.16833127e+00 1.59866974e-01 3.68569911e-01 3.65876079e-01 -2.63093233e-01 3.61280799e-01 -3.17208946e-01 -1.05647206e-01 -1.43697523e-02 -5.05984128e-01 1.89639293e-02 -2.54472613e-01 -1.04370572e-01 1.22788346e+00 -4.65520322e-01 4.42731492e-02 -5.11286974e-01 3.59588534e-01 2.32072920e-02 7.87052453e-01 1.19321382e+00 1.74912646e-01 6.12353623e-01 -4.34696019e-01 -5.70468903e-01 -6.09047174e-01 -1.37796319e+00 -5.78845739e-01 8.60767305e-01 9.42367688e-02 6.19247496e-01 -3.68712023e-02 4.91959974e-02 2.22007051e-01 6.85166597e-01 -5.15956938e-01 -4.14983094e-01 -7.69016385e-01 -1.31295145e+00 7.50963509e-01 5.05101621e-01 5.36351562e-01 -1.19526601e+00 -5.84671855e-01 5.45917451e-01 -2.36052960e-01 -1.02114201e+00 -2.60925353e-01 4.31852609e-01 -6.22031808e-01 -1.11349750e+00 -5.81052244e-01 -5.75914502e-01 5.79318285e-01 8.24773371e-01 9.96258080e-01 -6.15948588e-02 -8.36281419e-01 2.77987242e-01 -6.81680083e-01 -7.18407810e-01 9.43899751e-02 -5.54007173e-01 8.95573720e-02 9.63296220e-02 2.51491249e-01 -5.05543113e-01 -4.72518682e-01 3.87715131e-01 -9.44987297e-01 -3.78327459e-01 9.63199794e-01 1.15993631e+00 4.14605260e-01 1.15025081e-01 9.02410150e-01 -5.91321766e-01 6.81137145e-01 -3.10633183e-01 -7.44737089e-01 -3.00840661e-02 -3.14345993e-02 -3.61572504e-01 4.03008997e-01 -5.31644881e-01 -1.33890796e+00 -3.34627897e-01 -2.49475002e-01 -3.44902158e-01 1.67092942e-02 9.62094724e-01 -3.90081614e-01 -4.08868670e-01 7.85455585e-01 2.66943187e-01 -5.02460480e-01 -4.25868064e-01 -2.64426935e-02 4.09858942e-01 9.04554427e-01 -5.46185195e-01 1.20961785e+00 3.14534664e-01 -1.21910721e-01 -1.28860462e+00 -1.13338661e+00 -5.16947806e-01 -1.07296027e-01 -2.52875656e-01 2.01449499e-01 -1.25763142e+00 -2.94210494e-01 6.17828310e-01 -7.77626693e-01 -3.21336895e-01 8.00110772e-02 7.96625555e-01 7.27300942e-02 2.69584268e-01 -4.61688876e-01 -1.25782621e+00 -9.10810351e-01 -6.51137412e-01 8.28010201e-01 1.74656153e-01 2.67319977e-01 -5.85418820e-01 -4.84052189e-02 7.03833578e-03 7.19298482e-01 4.82140571e-01 6.25270724e-01 -4.32365566e-01 -4.88296688e-01 -4.46388781e-01 -5.44397533e-01 2.97337532e-01 1.98654726e-01 -4.26663250e-01 -8.47330868e-01 -8.69404256e-01 2.00938478e-01 -3.44781280e-02 1.21824574e+00 9.61904049e-01 7.60323226e-01 -7.59814819e-03 -2.32141092e-01 9.49600697e-01 1.71055603e+00 4.81431782e-01 9.31911349e-01 3.15461695e-01 6.52893960e-01 2.50217706e-01 9.73387480e-01 5.23587108e-01 -3.30737352e-01 2.40306064e-01 5.31549990e-01 -4.28334892e-01 4.77396548e-02 3.94972831e-01 3.02100331e-01 4.48967963e-01 8.69789049e-02 -2.15382993e-01 -9.92666662e-01 6.04649663e-01 -1.51278520e+00 -1.06776035e+00 -2.15234369e-01 1.99320436e+00 3.24264199e-01 7.24486485e-02 -6.42737091e-01 -1.76764697e-01 8.66075635e-01 7.32382163e-02 -4.89202797e-01 1.42003924e-01 -3.05804729e-01 4.84098464e-01 7.13795483e-01 5.89079022e-01 -1.06954741e+00 6.56445384e-01 5.49722624e+00 6.33100808e-01 -1.40279889e+00 -1.65793121e-01 -4.53295037e-02 1.40132681e-01 -8.25794339e-02 -4.54472780e-01 -7.45347559e-01 8.52370411e-02 5.51422656e-01 2.33488262e-01 4.05045748e-02 4.54461396e-01 3.97205263e-01 -1.17628321e-01 -4.97832835e-01 7.91833878e-01 -6.10716734e-03 -9.06251371e-01 1.22795403e-01 -2.55537659e-01 2.03883246e-01 3.41697067e-01 -1.59746930e-02 4.88444209e-01 6.50157034e-01 -1.01222372e+00 6.43751562e-01 6.78882897e-01 7.77657211e-01 -9.00894284e-01 1.01639509e+00 9.23396647e-02 -1.40276968e+00 -1.71748802e-01 -6.42116010e-01 1.72572285e-01 3.41889471e-01 1.21653366e+00 -6.37348652e-01 1.05823517e+00 8.32710743e-01 3.20804298e-01 -3.02519530e-01 1.52304471e+00 -2.73691975e-02 6.34728074e-01 -3.51095855e-01 4.55060363e-01 2.96170235e-01 -2.55693495e-01 1.04581153e+00 1.69539666e+00 6.05146885e-01 5.04540205e-01 5.51011026e-01 5.80252409e-01 4.81121361e-01 -3.09419036e-01 -4.98245835e-01 2.75823772e-01 6.99405372e-01 1.32417536e+00 -4.68385339e-01 1.86520636e-01 -7.23696202e-02 -2.57791467e-02 -3.01010370e-01 7.06456065e-01 -8.64911735e-01 -5.74568272e-01 7.57417142e-01 1.87579274e-01 6.37021720e-01 -1.48225669e-02 -2.11289182e-01 -8.97193313e-01 2.41574124e-02 -9.69159484e-01 2.50000179e-01 -7.24098384e-01 -1.82477486e+00 1.04743612e+00 1.00379735e-01 -1.50198936e+00 3.88494909e-01 -5.75696230e-01 -8.38794053e-01 1.10174060e+00 -1.70035458e+00 -1.17990720e+00 -9.59007502e-01 5.17339349e-01 2.07404569e-01 -3.57382149e-01 6.29121542e-01 5.20445645e-01 -5.48412085e-01 3.47237527e-01 7.10117221e-02 2.76469260e-01 7.52617657e-01 -6.08898818e-01 5.86855233e-01 1.27031648e+00 -6.89392269e-01 5.26458561e-01 1.11088336e+00 -1.18311512e+00 -1.48277628e+00 -1.63350224e+00 -2.12556839e-01 2.82182336e-01 5.23126304e-01 -4.47398812e-01 -1.24006605e+00 4.78350610e-01 -3.37980121e-01 2.88514823e-01 5.99953473e-01 -7.14808330e-02 -4.64457601e-01 -1.70693696e-01 -1.08332789e+00 4.58492130e-01 8.96190763e-01 7.90407583e-02 -6.69493735e-01 1.18369542e-01 4.84208256e-01 -5.65931380e-01 -6.34629190e-01 1.07735980e+00 4.79532272e-01 -5.40739715e-01 1.28504789e+00 -2.26998582e-01 -5.18465713e-02 -4.52133685e-01 -7.10902274e-01 -1.60434008e+00 -9.35754955e-01 -8.80982280e-02 7.18068704e-02 8.21249068e-01 1.93835407e-01 -9.84471381e-01 5.07586420e-01 -7.73288682e-02 -7.82242894e-01 -3.87549371e-01 -8.51389408e-01 -7.82539368e-01 -1.40226245e-01 -4.71408397e-01 2.68091083e-01 8.11172605e-01 -8.57959270e-01 2.17403442e-01 -3.34208161e-01 1.03949690e+00 1.28111792e+00 2.14724690e-01 6.11003518e-01 -1.68455338e+00 -1.60860166e-01 -5.20212501e-02 -1.17488019e-01 -8.02507162e-01 -2.15172842e-01 -8.19371819e-01 6.13251209e-01 -1.47895992e+00 -1.56245515e-01 -7.99863517e-01 -1.27072573e-01 3.78322870e-01 -3.38967025e-01 1.02131918e-01 4.87815822e-03 1.25187472e-01 1.04315504e-01 8.61925900e-01 1.35891902e+00 -3.06057751e-01 -2.12223038e-01 1.87323093e-01 -8.68617177e-01 5.72938919e-01 7.06411302e-01 -6.20747387e-01 -2.76257880e-02 -5.27656376e-01 -2.24369422e-01 2.68371642e-01 5.72958589e-01 -1.24359572e+00 1.14464231e-01 -1.78700000e-01 1.02400637e+00 -1.03917086e+00 2.69414186e-01 -8.49271595e-01 2.36911714e-01 5.31892538e-01 3.44822109e-01 -1.94081485e-01 6.56843364e-01 7.47904599e-01 -3.23361397e-01 -4.67851087e-02 1.20740843e+00 -2.55762246e-02 -5.78508496e-01 3.45541328e-01 -5.42626560e-01 -3.13721597e-02 5.99549413e-01 -3.73820156e-01 -3.79622102e-01 -2.07692385e-01 -5.92129767e-01 2.58239180e-01 -2.68380284e-01 2.69066900e-01 1.01004124e+00 -1.23365772e+00 -1.16704619e+00 4.90920693e-01 1.02346756e-01 7.68586546e-02 7.49843597e-01 8.13648582e-01 -3.77055347e-01 2.79717565e-01 -4.28456903e-01 -5.11015177e-01 -9.15553927e-01 -6.65979227e-03 6.50425255e-01 -2.12364703e-01 -1.12005794e+00 7.33621538e-01 6.39999568e-01 -5.83326757e-01 8.13757032e-02 -1.40771344e-01 -1.31152242e-01 -7.98145011e-02 9.69797432e-01 1.45865262e-01 5.26232004e-01 -3.91521215e-01 -4.32167977e-01 3.27909380e-01 -3.21940541e-01 2.00117424e-01 1.73904812e+00 2.68159062e-01 7.91444108e-02 -1.86648577e-01 6.80186868e-01 7.75010213e-02 -1.27109444e+00 -4.25410390e-01 -2.92994618e-01 -6.17421567e-01 4.39237326e-01 -9.00265813e-01 -1.37902153e+00 3.97474051e-01 8.03730547e-01 -1.03548557e-01 1.38252056e+00 -5.46194971e-01 6.04353786e-01 6.57510996e-01 3.41116875e-01 -5.49340308e-01 1.58422962e-01 1.08704686e+00 1.37187099e+00 -7.99841225e-01 1.65251687e-01 -4.93681222e-01 -5.70124090e-01 9.21005011e-01 6.49495840e-01 -5.57111025e-01 7.12254047e-01 7.65290320e-01 4.25921530e-01 -5.28284848e-01 -2.68604100e-01 -5.53172901e-02 8.09124932e-02 9.69618499e-01 -1.43249584e-02 1.52719095e-01 1.58234149e-01 9.21332598e-01 -1.73493892e-01 -4.09996003e-01 5.82989335e-01 1.14547157e+00 -8.10966909e-01 -4.47773397e-01 -1.09674978e+00 7.77696311e-01 -2.91221589e-01 -2.84117013e-01 7.67592415e-02 8.30094814e-01 -3.13883722e-01 9.41070914e-01 -1.29938677e-01 -1.51819065e-01 7.79354095e-01 -4.60681111e-01 3.92532676e-01 -5.36789060e-01 -4.70593840e-01 5.84687650e-01 9.06504467e-02 -2.90965438e-01 -4.73957397e-02 -5.66014826e-01 -9.77302134e-01 -3.92717756e-02 -7.04202056e-01 2.38590524e-01 2.24176064e-01 5.94226718e-01 3.61562781e-02 9.97382939e-01 4.41048592e-01 -1.24855101e+00 -6.02425635e-01 -1.18290555e+00 -9.35967207e-01 -1.91404536e-01 6.10106528e-01 -1.11991191e+00 -5.52614391e-01 -3.80132914e-01]
[6.923391342163086, 1.1690303087234497]
a7213808-0dd1-4553-b43e-86e6f4d57ff0
a-deep-pyramid-deformable-part-model-for-face
1508.04389
null
http://arxiv.org/abs/1508.04389v1
http://arxiv.org/pdf/1508.04389v1.pdf
A Deep Pyramid Deformable Part Model for Face Detection
We present a face detection algorithm based on Deformable Part Models and deep pyramidal features. The proposed method called DP2MFD is able to detect faces of various sizes and poses in unconstrained conditions. It reduces the gap in training and testing of DPM on deep features by adding a normalization layer to the deep convolutional neural network (CNN). Extensive experiments on four publicly available unconstrained face detection datasets show that our method is able to capture the meaningful structure of faces and performs significantly better than many competitive face detection algorithms.
['Rajeev Ranjan', 'Vishal M. Patel', 'Rama Chellappa']
2015-08-18
null
null
null
null
['robust-face-recognition']
['computer-vision']
[-2.94239428e-02 1.45720184e-01 6.07866123e-02 -4.64590073e-01 -4.54844050e-02 -5.26404500e-01 5.16552031e-01 -7.16507018e-01 -1.66721418e-01 1.01412974e-01 -2.93085605e-01 3.28182839e-02 3.59779149e-01 -9.04240072e-01 -7.29988992e-01 -5.90879798e-01 -3.00702065e-01 3.51704657e-01 4.76035863e-01 1.69816807e-01 1.43779516e-01 1.23047209e+00 -1.82686794e+00 2.86642551e-01 -6.53086081e-02 1.30444074e+00 2.09400244e-02 4.69455779e-01 1.99467555e-01 3.61637861e-01 -5.14416218e-01 -6.04024887e-01 7.29084074e-01 2.84253567e-01 -5.79695284e-01 5.65859415e-02 1.13998747e+00 -8.88377845e-01 -6.94830835e-01 1.07370853e+00 7.02960908e-01 -1.78410679e-01 7.89086640e-01 -1.07026279e+00 -5.94393730e-01 1.47439033e-01 -8.83406103e-01 4.80793774e-01 5.25788605e-01 -2.40080729e-02 3.86136234e-01 -1.55373967e+00 5.85560262e-01 2.10174656e+00 1.05179501e+00 8.62198114e-01 -9.70237136e-01 -9.93056536e-01 -7.59709775e-02 4.14138213e-02 -1.73008966e+00 -7.68153429e-01 4.91947830e-01 -2.72058904e-01 1.22772193e+00 6.35188818e-02 4.71955478e-01 1.00998557e+00 -2.66590603e-02 6.61456704e-01 5.24965584e-01 -4.86818820e-01 -1.27172664e-01 -1.95368290e-01 -6.41582683e-02 1.47807288e+00 5.77163100e-01 8.41568410e-02 -3.30158979e-01 -3.51758599e-01 1.21838880e+00 3.46567035e-02 3.95516381e-02 -1.08685747e-01 -2.46045738e-01 9.56056595e-01 4.26631033e-01 3.05695813e-02 -1.02031231e-03 4.88880910e-02 4.18852866e-01 3.81414779e-02 5.34492731e-01 -2.01345921e-01 -3.56465995e-01 5.11725783e-01 -9.42895532e-01 1.98903814e-01 7.40713537e-01 7.95147836e-01 5.50561845e-01 4.89953235e-02 -4.34434384e-01 7.67628908e-01 5.57608604e-01 9.36005175e-01 1.25102192e-01 -8.45701218e-01 2.34718397e-01 9.14462090e-01 -5.97613081e-02 -1.17155457e+00 -4.72400308e-01 2.56404698e-01 -4.69373494e-01 5.11216521e-01 2.00190574e-01 -3.20376039e-01 -1.07835007e+00 1.50436699e+00 5.31149983e-01 1.65191516e-01 -5.05547166e-01 6.11822069e-01 1.41534603e+00 2.90608257e-01 4.84666461e-03 -1.00784265e-02 1.39346433e+00 -7.52432764e-01 -4.10276413e-01 -1.61270723e-01 1.36718035e-01 -8.71089995e-01 2.42763743e-01 2.11478055e-01 -1.01915920e+00 -9.26259100e-01 -9.95005071e-01 -6.03159284e-03 -3.22891772e-01 8.30739915e-01 6.84149981e-01 1.17658985e+00 -1.29320204e+00 5.45308769e-01 -6.93163037e-01 -4.72316116e-01 1.14467371e+00 9.67769504e-01 -8.10174942e-01 -2.05403328e-01 -6.06172919e-01 5.20793855e-01 1.87286988e-01 3.50425214e-01 -1.21118498e+00 -2.99288929e-01 -1.02250385e+00 4.51070480e-02 1.54382691e-01 -3.50284934e-01 1.01777315e+00 -8.27592134e-01 -1.50473225e+00 1.27356791e+00 -1.64124653e-01 -2.09271267e-01 5.17781317e-01 -3.58047128e-01 -3.60676169e-01 6.10622644e-01 -1.58646166e-01 1.06579316e+00 1.43146634e+00 -6.23624146e-01 -5.14095068e-01 -6.97443604e-01 -8.24423358e-02 -2.11665884e-01 -5.08607447e-01 6.98095381e-01 -7.45273829e-01 -4.30920124e-01 -4.02809307e-02 -7.74457514e-01 2.27117136e-01 7.03392625e-01 -2.93069601e-01 -6.93601966e-01 1.70189619e+00 -4.30535972e-01 6.31651580e-01 -1.97994661e+00 -6.17574267e-02 1.79879919e-01 3.27182472e-01 8.94721568e-01 -5.32433331e-01 -1.68000355e-01 7.35342428e-02 -1.55623779e-01 3.46028268e-01 -6.36531234e-01 -5.87029085e-02 1.24738142e-01 -4.92624678e-02 8.82479906e-01 4.95959908e-01 7.74884820e-01 -3.53896320e-01 -5.83610535e-01 2.99072444e-01 8.27033699e-01 -5.97410560e-01 2.07392588e-01 9.30229574e-02 -5.08418605e-02 -2.95269042e-01 1.12529945e+00 1.52935433e+00 2.02698410e-01 4.93007489e-02 -9.56385285e-02 -1.03234258e-02 -1.93043500e-01 -1.29824078e+00 1.13912857e+00 -9.01664235e-03 8.21031690e-01 5.53913414e-01 -8.25197220e-01 9.38629448e-01 3.35818738e-01 1.94557030e-02 -2.54669875e-01 5.05103946e-01 8.26069787e-02 -1.83336705e-01 -4.79616463e-01 -7.16667250e-02 2.08599955e-01 5.56738675e-01 1.09470330e-01 5.69676399e-01 4.16841149e-01 1.29328609e-01 -3.35190654e-01 9.57866907e-01 -1.19366869e-01 7.83511326e-02 -5.17820477e-01 9.23821688e-01 -8.13852191e-01 6.57157004e-01 6.63906395e-01 -4.58088040e-01 4.20077503e-01 4.52671111e-01 -9.32139754e-01 -5.50482333e-01 -1.00284827e+00 -5.94333053e-01 1.33937395e+00 -9.02619362e-02 -7.40444362e-02 -1.08944690e+00 -1.08137369e+00 5.30008078e-01 -4.41631824e-01 -1.08940542e+00 9.36579853e-02 -7.10563660e-01 -6.42289519e-01 9.55135465e-01 8.74077857e-01 6.82682335e-01 -1.23040867e+00 -4.28386480e-01 -2.46099621e-01 4.35375839e-01 -1.34655321e+00 -3.68345559e-01 -1.62307292e-01 -5.45458615e-01 -1.60689843e+00 -6.79146349e-01 -1.37824309e+00 9.59622741e-01 4.62037593e-01 8.63099098e-01 3.28434736e-01 -1.10337341e+00 3.40945870e-01 -1.04412682e-01 -5.66522002e-01 -7.93192908e-02 -3.00415933e-01 4.31145072e-01 1.48205096e-02 7.98322439e-01 -9.40280333e-02 -7.85349607e-01 5.16952515e-01 -3.50764453e-01 -7.93487549e-01 4.07229304e-01 4.57056463e-01 2.10161239e-01 1.04848213e-01 2.21036702e-01 -5.39074063e-01 1.37992159e-01 -2.20648237e-02 -9.55232918e-01 1.42632857e-01 8.12583715e-02 -2.47525722e-01 1.96938843e-01 -4.12137151e-01 -9.94883299e-01 5.08594453e-01 -3.55591446e-01 -6.28245592e-01 -2.07248598e-01 -5.90528369e-01 -2.67743081e-01 -1.08443570e+00 5.61758280e-01 1.75636299e-02 2.17194054e-02 -6.73317850e-01 3.52781825e-02 6.04533613e-01 4.03851479e-01 -3.37315142e-01 9.85975027e-01 7.84333527e-01 2.08411694e-01 -1.07293379e+00 -5.35270810e-01 -3.67332160e-01 -1.05597663e+00 -2.77153969e-01 6.54506624e-01 -9.43095446e-01 -9.55731928e-01 8.07356179e-01 -1.30653179e+00 9.04293582e-02 1.88339487e-01 1.24831453e-01 -9.68525931e-02 4.47339952e-01 -8.04317594e-01 -7.87973642e-01 -6.82638288e-01 -1.03639138e+00 1.51272714e+00 3.71932477e-01 4.01920259e-01 -8.02241743e-01 -2.25711334e-02 -8.24995413e-02 1.76288709e-01 2.77307928e-01 2.61693716e-01 -4.08744603e-01 -2.49931410e-01 -5.60407281e-01 -5.43475866e-01 4.98562157e-01 1.51732221e-01 6.21461987e-01 -1.38399100e+00 -6.86865270e-01 -1.65423423e-01 -4.24489856e-01 1.24930501e+00 5.22040248e-01 1.29771483e+00 -2.87737012e-01 -7.73766160e-01 7.20282495e-01 1.32637608e+00 -9.95983556e-02 5.93353152e-01 -2.59428382e-01 5.03012836e-01 4.77488726e-01 2.04894871e-01 5.05266428e-01 -2.73379177e-01 5.85698128e-01 5.83800912e-01 -1.79830417e-01 -3.71541023e-01 2.93539576e-02 4.44210052e-01 -2.04384983e-01 -2.92685062e-01 6.63221553e-02 -7.12692559e-01 1.51442990e-01 -1.37990367e+00 -9.91146624e-01 1.16630428e-01 1.77404964e+00 3.10053498e-01 2.54829247e-02 3.76986951e-01 1.48745447e-01 1.01214266e+00 -2.36252874e-01 -3.12012553e-01 -4.28235024e-01 -8.67216587e-02 7.41850734e-01 3.40318680e-01 2.85638005e-01 -1.84150958e+00 9.95197415e-01 7.86636448e+00 7.35006928e-01 -1.03216255e+00 1.27313510e-01 4.27659035e-01 -4.37096693e-02 8.48125041e-01 -8.84328127e-01 -1.55437660e+00 1.25274569e-01 4.44065034e-01 2.25775108e-01 6.12142012e-02 1.30742025e+00 -2.79208332e-01 2.36729622e-01 -1.26657844e+00 1.18554270e+00 3.09805512e-01 -1.17990255e+00 2.25970775e-01 3.15713650e-03 6.35800719e-01 2.01461427e-02 3.60845059e-01 2.71704644e-01 4.73995022e-02 -1.42278910e+00 6.51821375e-01 1.47047058e-01 9.89754736e-01 -8.59588742e-01 8.76555681e-01 9.90255084e-03 -1.63375092e+00 -5.23071885e-01 -1.05251169e+00 -2.08767518e-01 -7.00608492e-01 1.17465958e-01 -8.29094291e-01 -6.90733120e-02 8.39180589e-01 4.53350395e-01 -7.50149190e-01 9.86472964e-01 -1.70101553e-01 1.55784637e-01 -4.21746552e-01 2.65500128e-01 -7.74395615e-02 4.12814796e-01 2.31471956e-01 1.56360960e+00 2.14929104e-01 -1.06408251e-02 1.95168734e-01 7.51645148e-01 -4.89808083e-01 -1.15755955e-02 -7.03656852e-01 2.10514531e-01 4.79441792e-01 1.77019072e+00 -9.58949208e-01 -3.54604162e-02 -6.04339540e-01 9.19491827e-01 4.45898205e-01 -9.30665806e-02 -7.49670744e-01 -2.49248818e-01 8.60907495e-01 1.31257698e-01 1.01546037e+00 3.80552225e-02 4.49276656e-01 -6.89327121e-01 -2.17693448e-02 -5.26101768e-01 4.29784030e-01 -4.03884798e-02 -1.32818162e+00 6.31697714e-01 -2.74165750e-01 -8.59475195e-01 1.62550375e-01 -1.51090884e+00 -8.88708353e-01 5.61434209e-01 -1.45701456e+00 -1.41834939e+00 -4.38499749e-01 9.03547168e-01 4.30698365e-01 -4.11807954e-01 8.83820236e-01 4.31391269e-01 -9.80960488e-01 1.05246139e+00 -1.49386913e-01 7.72685409e-01 4.46154535e-01 -8.13100517e-01 6.79050386e-01 7.22597659e-01 3.21402475e-02 6.08924866e-01 1.96038276e-01 -3.66956085e-01 -1.48999727e+00 -1.40016735e+00 3.88509244e-01 -4.43930626e-01 3.02192774e-02 -6.72124982e-01 -5.09180903e-01 6.81303024e-01 -7.01601654e-02 8.08731496e-01 5.79630673e-01 -1.04579851e-01 -6.58191681e-01 -1.61819495e-02 -1.79185867e+00 -2.24862732e-02 1.23089051e+00 -4.52130258e-01 -5.33264995e-01 5.86482704e-01 7.79852122e-02 -3.76521558e-01 -4.85010982e-01 6.39147043e-01 8.11457694e-01 -9.39316392e-01 1.29408574e+00 -3.55278671e-01 -6.33259937e-02 -8.88875872e-02 -1.96265094e-02 -3.77249122e-01 -6.68963492e-01 -4.65249002e-01 -6.06345117e-01 1.02993369e+00 -1.60009384e-01 -5.78552604e-01 9.38025415e-01 1.77928373e-01 3.77962172e-01 -5.47291994e-01 -1.32493103e+00 -9.51446831e-01 -7.01391324e-02 1.50599182e-01 5.07973969e-01 4.88399714e-01 -4.96604562e-01 -1.40357465e-01 1.32090285e-01 6.09363794e-01 7.16798484e-01 5.18934205e-02 5.46167910e-01 -1.72774422e+00 8.19633454e-02 -5.30658484e-01 -1.01615489e+00 -8.17182183e-01 5.36477149e-01 -7.09968567e-01 -1.93248838e-01 -1.00447881e+00 5.94502091e-01 1.98920742e-01 -6.10239692e-02 8.55751574e-01 -2.56095398e-02 1.18083417e+00 -1.18250124e-01 -1.38027638e-01 -5.66687644e-01 3.13297302e-01 1.03087211e+00 -2.40461573e-01 1.33895829e-01 5.09057604e-02 -1.83037832e-01 1.18394518e+00 4.70839083e-01 -3.53945225e-01 2.05736496e-02 -3.79852563e-01 -3.76790792e-01 -5.67878783e-01 5.31563222e-01 -1.27660394e+00 -2.39005052e-02 4.03420717e-01 1.26134408e+00 -5.93268573e-01 4.11657512e-01 -5.67392588e-01 -4.52583998e-01 8.20335686e-01 3.33404273e-01 -2.58858263e-01 7.40278602e-01 3.68358672e-01 1.25709802e-01 -2.82269537e-01 1.19247568e+00 -1.04443938e-01 -7.21996427e-01 6.91687346e-01 -1.85161054e-01 -2.84066796e-01 1.19973445e+00 -3.48587513e-01 -2.21285403e-01 4.05429631e-01 -6.13987327e-01 -6.43708259e-02 3.16205978e-01 6.08374536e-01 8.27018142e-01 -1.40494537e+00 -7.59603202e-01 8.59269857e-01 -2.65552461e-01 -4.21364784e-01 -1.75841022e-02 3.50737572e-01 -8.65180731e-01 7.42677689e-01 -5.61501265e-01 -6.23747766e-01 -1.99205899e+00 4.71468121e-01 5.77326357e-01 2.84463376e-01 -6.40128255e-01 1.61651635e+00 3.18974227e-01 -3.85894656e-01 6.01757109e-01 -7.26573467e-02 -4.75492328e-01 1.26547083e-01 1.19115746e+00 5.01281321e-01 9.50374454e-02 -1.02997768e+00 -9.26107526e-01 7.89755404e-01 -3.32633853e-01 7.17291594e-01 1.30034328e+00 3.97009104e-01 -2.47821227e-01 -5.82595468e-01 1.30193770e+00 -1.07113726e-01 -1.51899886e+00 -5.24472736e-04 -3.69961083e-01 -7.30418503e-01 1.15733027e-01 -1.95410222e-01 -1.56760287e+00 8.58309627e-01 1.11730909e+00 3.68383676e-02 9.80712950e-01 1.03617385e-01 4.87202287e-01 6.82624578e-01 4.20638114e-01 -9.45056558e-01 3.72157544e-01 5.17875254e-01 1.02429008e+00 -1.21544945e+00 8.10595825e-02 -8.63840103e-01 3.32448751e-01 1.68699205e+00 9.34347212e-01 -3.89499694e-01 9.25228834e-01 5.26340127e-01 -1.88637093e-01 -3.04058760e-01 -2.98371285e-01 -2.40894169e-01 4.47177142e-01 9.54449058e-01 1.91733465e-01 -1.29927143e-01 1.29975140e-01 2.85029352e-01 2.24942893e-01 -9.70772207e-02 -1.25920117e-01 8.37465405e-01 -7.61250556e-01 -9.13074493e-01 -5.31728983e-01 2.35226974e-01 -6.83628201e-01 2.56931484e-01 -7.26870596e-01 7.82000899e-01 5.35830200e-01 7.36612320e-01 3.02515179e-01 -1.58006847e-01 1.55224446e-02 2.59872479e-03 1.13716781e+00 -8.12260032e-01 -3.17947596e-01 -2.02635303e-01 -3.90073419e-01 -7.99238503e-01 -4.17296529e-01 -6.89743698e-01 -1.06441355e+00 -4.55770344e-01 -5.67039251e-01 -6.00227237e-01 2.58698106e-01 7.42776215e-01 2.97602236e-01 6.55115396e-02 9.19302642e-01 -1.44871926e+00 -3.86723965e-01 -1.00720239e+00 -7.46289790e-01 1.82808712e-02 3.00646424e-01 -1.23875296e+00 -3.55747432e-01 -1.84240028e-01]
[13.342473983764648, 0.6957021355628967]
ec26e01c-3cec-4105-bb49-16aeea65934e
saving-dense-retriever-from-shortcut
2202.07280
null
https://arxiv.org/abs/2202.07280v2
https://arxiv.org/pdf/2202.07280v2.pdf
Saving Dense Retriever from Shortcut Dependency in Conversational Search
Conversational search (CS) needs a holistic understanding of conversational inputs to retrieve relevant passages. In this paper, we demonstrate the existence of a retrieval shortcut in CS, which causes models to retrieve passages solely relying on partial history while disregarding the latest question. With in-depth analysis, we first show that naively trained dense retrievers heavily exploit the shortcut and hence perform poorly when asked to answer history-independent questions. To build more robust models against shortcut dependency, we explore various hard negative mining strategies. Experimental results show that training with the model-based hard negatives effectively mitigates the dependency on the shortcut, significantly improving dense retrievers on recent CS benchmarks. In particular, our retriever outperforms the previous state-of-the-art model by 11.0 in Recall@10 on QReCC.
['Gangwoo Kim', 'Sungdong Kim']
2022-02-15
null
null
null
null
['conversational-search']
['natural-language-processing']
[ 1.74202934e-01 3.48670632e-02 -4.35289979e-01 -1.94397420e-01 -1.26866436e+00 -8.67704928e-01 6.08728230e-01 1.58934399e-01 -5.99727929e-01 6.84644699e-01 5.18483400e-01 -7.42217839e-01 -1.98746845e-01 -7.87156343e-01 -5.92395246e-01 -2.97226012e-01 1.10759892e-01 5.25852799e-01 6.32578254e-01 -7.56242812e-01 5.91932595e-01 -8.74736067e-03 -1.41586506e+00 9.04333830e-01 8.80359948e-01 7.23593473e-01 2.61650622e-01 1.02857530e+00 -2.51081139e-01 8.60834301e-01 -7.83496261e-01 -6.33127987e-01 -1.49747217e-03 -5.35969794e-01 -1.29043162e+00 -5.83179414e-01 3.69919330e-01 -4.08378065e-01 -6.12094820e-01 4.77104992e-01 3.98583710e-01 3.14593405e-01 4.56414610e-01 -8.27263713e-01 -5.46450555e-01 8.31880450e-01 -2.26675957e-01 7.16545999e-01 8.17920804e-01 1.17091753e-01 1.61486948e+00 -9.08682048e-01 9.23772991e-01 9.83981967e-01 2.17208937e-01 7.50768661e-01 -1.08520830e+00 -4.69519526e-01 4.28249121e-01 5.05352259e-01 -1.12258267e+00 -3.55367184e-01 4.77026314e-01 1.28390029e-01 1.68920326e+00 8.90050113e-01 5.39912045e-01 1.21959245e+00 -3.48790325e-02 1.08835518e+00 8.41882885e-01 -6.18459821e-01 -4.17923406e-02 1.68583632e-01 7.08732307e-01 4.08747822e-01 5.02750389e-02 -1.63767561e-02 -8.63073170e-01 -4.27223355e-01 -1.15455270e-01 -1.68527126e-01 -5.97855985e-01 1.05450012e-01 -6.62993193e-01 8.71755421e-01 1.90918505e-01 3.08649123e-01 7.27838948e-02 -1.24038294e-01 2.40291119e-01 8.57700765e-01 3.71792436e-01 8.37955773e-01 -6.64219737e-01 -3.63542795e-01 -7.16610253e-01 3.36495697e-01 1.43117392e+00 1.13285172e+00 6.04988277e-01 -7.95521796e-01 -5.36884248e-01 1.10714948e+00 -2.86073893e-01 3.44681203e-01 3.77425700e-01 -1.03009927e+00 6.45914674e-01 5.42104900e-01 5.52557558e-02 -8.69782209e-01 -1.40613481e-01 -6.80241227e-01 -2.55271405e-01 -7.69311011e-01 7.33294785e-02 8.17555338e-02 -5.59638321e-01 1.63395858e+00 -1.16432132e-03 -1.98388666e-01 2.56627034e-02 7.53221214e-01 8.45257461e-01 6.05379879e-01 -4.71818261e-02 -3.28866720e-01 1.02991843e+00 -1.27987528e+00 -4.99619335e-01 -4.17540878e-01 8.38986397e-01 -6.97807193e-01 1.38441896e+00 4.78768110e-01 -1.11650479e+00 -4.38156128e-02 -9.81927872e-01 -2.68390983e-01 -2.44547069e-01 -2.71894485e-01 6.79584801e-01 3.57228488e-01 -1.32031775e+00 4.85724807e-01 -4.97935385e-01 -5.19033730e-01 -1.15794584e-01 7.57605359e-02 9.23106819e-02 -2.51687855e-01 -1.48107779e+00 9.20477808e-01 -1.51118830e-01 -1.41040444e-01 -8.40705574e-01 -7.24061489e-01 -4.06060696e-01 3.22391838e-01 7.54038751e-01 -7.42177069e-01 1.77707553e+00 -4.95001882e-01 -1.20965183e+00 8.54922295e-01 -5.68950176e-01 -6.94854558e-01 4.36929137e-01 -5.42708397e-01 -3.94693702e-01 6.32410645e-01 2.99051963e-02 3.59467834e-01 4.53871697e-01 -1.10094285e+00 -6.56511724e-01 -8.08617845e-02 5.44132888e-01 3.26962233e-01 -4.45234686e-01 -2.70612296e-02 -1.13792360e+00 -2.95096189e-01 -1.00318015e-01 -9.89368141e-01 -4.28289734e-02 -5.19731879e-01 -5.39301276e-01 -5.62466264e-01 3.62890035e-01 -3.91592681e-01 1.81042647e+00 -1.64844596e+00 -2.21351702e-02 3.52959305e-01 1.87605411e-01 9.58962142e-02 -3.37732017e-01 9.81535256e-01 4.53268528e-01 4.19581413e-01 -9.41390395e-02 -3.78492147e-01 -2.35205531e-01 1.26684368e-01 -6.33838236e-01 3.09858266e-02 -2.01483481e-02 9.81914580e-01 -1.02134907e+00 -5.80619395e-01 -3.27919394e-01 -2.79103704e-02 -8.02747667e-01 2.41192266e-01 -5.34116566e-01 -1.79439008e-01 -5.42933643e-01 5.74093819e-01 2.52168775e-01 -5.57465315e-01 3.03560346e-01 3.80918980e-01 1.84932530e-01 1.04917455e+00 -4.44625765e-01 1.67787659e+00 -6.90905035e-01 8.32596540e-01 1.88063055e-01 -5.19081652e-01 4.12814468e-01 1.58227831e-01 7.09703639e-02 -1.16000056e+00 -3.04011971e-01 1.95104212e-01 -2.58220166e-01 -5.42464674e-01 7.31761277e-01 3.69588356e-03 7.85743468e-05 8.66819382e-01 -1.14278682e-01 2.68017706e-02 3.42950046e-01 1.00285995e+00 1.62160146e+00 -4.43797827e-01 -2.83475727e-01 -1.27808958e-01 6.76544487e-01 2.33241811e-01 2.40699947e-01 1.47132277e+00 6.18415438e-02 6.22904181e-01 4.98987138e-01 3.20384018e-02 -4.43257779e-01 -8.70140731e-01 9.34996754e-02 1.65185332e+00 2.47261003e-01 -7.17800796e-01 -4.87837255e-01 -9.89058733e-01 2.29187515e-02 1.08366656e+00 -5.24270117e-01 -4.15663153e-01 -8.32075417e-01 -5.44481158e-01 4.69972640e-01 3.80004883e-01 1.27672460e-02 -1.19977546e+00 -2.37062588e-01 1.75811440e-01 -7.24678576e-01 -1.02444816e+00 -6.26778007e-01 1.58282310e-01 -8.40224862e-01 -1.26512718e+00 -5.75250506e-01 -7.05518663e-01 1.66515768e-01 7.91997552e-01 1.89747453e+00 9.13529158e-01 -9.74282175e-02 8.36707413e-01 -6.61567986e-01 -1.49810821e-01 -2.94273734e-01 8.44251096e-01 -5.46427667e-01 -6.33912265e-01 7.32056677e-01 -5.49221516e-01 -9.08697665e-01 3.83011401e-01 -7.83032835e-01 -2.31174469e-01 4.63803649e-01 8.35662127e-01 1.73718333e-01 -5.55409074e-01 1.01592290e+00 -1.13170040e+00 1.07098258e+00 -6.82222068e-01 -2.55963981e-01 5.95778048e-01 -9.16860878e-01 2.03068897e-01 3.63683313e-01 -2.98393101e-01 -1.03624535e+00 -6.32168293e-01 -2.19804928e-01 5.75260678e-03 3.83762896e-01 6.25500441e-01 4.24205452e-01 2.25646973e-01 9.24188077e-01 3.25271368e-01 -3.14690232e-01 -5.82675934e-01 3.15223932e-01 7.35478640e-01 3.92461360e-01 -7.70038307e-01 4.42287207e-01 4.45474923e-01 -6.71324253e-01 -6.47486150e-01 -1.29081607e+00 -1.25235057e+00 1.04793971e-02 -1.97954983e-01 4.14162576e-01 -7.68452525e-01 -7.86563635e-01 -1.35861903e-01 -1.18286192e+00 -3.50066632e-01 -1.25327557e-02 9.08444971e-02 -2.91000962e-01 4.38760638e-01 -1.04506087e+00 -9.63070512e-01 -8.19020867e-01 -6.42485023e-01 9.14280593e-01 5.55760488e-02 -3.66028905e-01 -7.15108871e-01 3.48135501e-01 6.73440933e-01 4.44042891e-01 -5.66627145e-01 1.02282548e+00 -1.08709538e+00 -9.05863702e-01 -1.91124678e-01 4.69048433e-02 2.51842171e-01 -3.83658558e-01 -3.23341250e-01 -1.05919611e+00 -2.58087635e-01 6.68288320e-02 -5.54107845e-01 1.54636264e+00 -2.10463494e-01 9.96540725e-01 -3.92324924e-01 -5.24181545e-01 1.38325259e-01 1.33848047e+00 -7.35317683e-03 5.14201820e-01 4.66401488e-01 1.95717104e-02 7.12682188e-01 5.76675057e-01 1.90741152e-01 3.84017617e-01 3.01383197e-01 1.99317902e-01 3.19205463e-01 5.97776026e-02 -3.86585593e-01 3.23929518e-01 1.03680944e+00 3.27239484e-01 -6.24960065e-01 -6.75163746e-01 9.96934950e-01 -1.77605629e+00 -9.35112298e-01 -3.13732997e-02 1.98414516e+00 1.34976161e+00 4.83920008e-01 -1.18469708e-01 -6.85835704e-02 3.62481296e-01 2.50000507e-01 -4.57886249e-01 -4.86726463e-01 -2.40323633e-01 6.13428175e-01 2.19492972e-01 9.52470303e-01 -6.22478962e-01 9.73991752e-01 6.72765970e+00 7.87033081e-01 -6.71142399e-01 -4.31345217e-02 4.16244149e-01 -6.95407808e-01 -7.79085875e-01 9.84816626e-02 -1.04176259e+00 1.63393945e-01 9.91898477e-01 -2.34032795e-01 5.26646972e-01 7.86986291e-01 -1.78529724e-01 -3.71934444e-01 -1.18730879e+00 4.20468807e-01 2.52532005e-01 -1.41832292e+00 8.14425275e-02 -1.99550658e-01 8.23976755e-01 2.41675481e-01 -1.31670311e-01 8.52275372e-01 4.51275378e-01 -9.29009140e-01 3.49529654e-01 4.82704103e-01 2.58005559e-01 -6.80447996e-01 7.28357911e-01 6.27023697e-01 -7.31744051e-01 -1.07326180e-01 -2.75591910e-01 -3.29762627e-03 1.38452008e-01 4.70893800e-01 -9.34323132e-01 4.59632546e-01 8.34802270e-01 3.80769432e-01 -5.75986624e-01 9.61740851e-01 -2.66855180e-01 8.95016670e-01 -4.61888254e-01 -4.17826384e-01 3.96706164e-01 9.92614105e-02 6.23922884e-01 1.63122058e+00 7.79803246e-02 4.73484218e-01 -1.40480787e-01 6.28114104e-01 -3.10019970e-01 1.54630765e-01 -5.06029665e-01 7.12873489e-02 5.21006882e-01 1.02204597e+00 -2.30591357e-01 -4.75973338e-01 -4.29567039e-01 8.39511931e-01 6.43465579e-01 6.76022887e-01 -5.07513344e-01 -2.39751026e-01 3.79782468e-01 1.76785424e-01 5.22308469e-01 3.87659930e-02 -2.07938582e-01 -1.05385339e+00 4.59783942e-01 -1.10705197e+00 7.82836199e-01 -5.09501576e-01 -1.32793581e+00 6.50646806e-01 -1.61205724e-01 -7.39762425e-01 -5.56757390e-01 -9.37609822e-02 -6.87038124e-01 7.68248439e-01 -1.82029855e+00 -5.29177308e-01 -1.04601502e-01 5.29121339e-01 7.58543491e-01 2.24379867e-01 6.85712576e-01 1.82090923e-01 -2.68855721e-01 6.83480322e-01 3.77447084e-02 -5.17170727e-02 8.81485820e-01 -1.18288124e+00 2.10711330e-01 5.54889560e-01 2.85656214e-01 1.31261790e+00 7.28803039e-01 -5.71052372e-01 -1.52378130e+00 -5.63303173e-01 1.27014709e+00 -9.00849998e-01 6.28418803e-01 -3.43499631e-01 -9.88560677e-01 5.42661428e-01 5.45051932e-01 -5.86360753e-01 6.93481028e-01 7.10366130e-01 -6.21294618e-01 -9.15401950e-02 -6.01041079e-01 6.88086808e-01 1.17480087e+00 -1.02397799e+00 -8.79594326e-01 4.96681660e-01 1.13063300e+00 -1.09858930e-01 -4.13157165e-01 4.23298568e-01 6.42778993e-01 -1.04628992e+00 1.03355682e+00 -9.34929430e-01 4.42069292e-01 2.94079512e-01 2.54151765e-02 -8.35846126e-01 1.05747737e-01 -9.15583074e-01 -4.70142275e-01 8.04286182e-01 1.03979182e+00 -3.17848831e-01 8.37184608e-01 6.52098179e-01 -1.66725039e-01 -1.03951120e+00 -7.04424322e-01 -7.01843083e-01 1.26695201e-01 -6.55378640e-01 3.03569883e-01 6.12582862e-01 4.30024594e-01 6.28264904e-01 -8.66560638e-02 -2.07815841e-01 7.41085187e-02 4.65066165e-01 5.71913660e-01 -9.29493785e-01 -5.20859241e-01 -5.15606225e-01 6.28297746e-01 -1.81206203e+00 4.12534364e-02 -7.86205351e-01 2.43814826e-01 -1.66071188e+00 5.42241871e-01 -3.14031035e-01 -4.38073009e-01 7.19039664e-02 -5.29308796e-01 -1.14916954e-02 -1.17957229e-02 3.15308928e-01 -1.26225924e+00 4.65502471e-01 1.22843730e+00 -1.91004202e-01 -3.04081291e-01 2.31869593e-01 -1.04357171e+00 2.25781694e-01 6.71569645e-01 -6.95398510e-01 -6.11687839e-01 -2.97861218e-01 7.20362127e-01 2.99780101e-01 8.96974653e-02 -5.31933367e-01 7.32158124e-01 -2.82359906e-02 -2.37061933e-01 -8.75870228e-01 4.70094472e-01 -3.39059979e-01 -4.86700445e-01 2.16841534e-01 -1.05215693e+00 1.24713898e-01 1.34695664e-01 9.36023116e-01 -3.01594853e-01 -3.85928154e-01 1.64231405e-01 -3.42369437e-01 -3.06708544e-01 -1.37887880e-01 -4.03905272e-01 6.02836728e-01 2.08479866e-01 3.38266522e-01 -6.04005754e-01 -8.93135369e-01 -5.06366730e-01 6.49261594e-01 2.89241880e-01 5.88389695e-01 5.44920743e-01 -7.19776273e-01 -5.80001175e-01 -1.05110325e-01 3.22511613e-01 -2.25869030e-01 2.52281636e-01 6.91273034e-01 -2.21419394e-01 9.73477066e-01 7.00490534e-01 -3.71046185e-01 -1.28202951e+00 5.11388898e-01 2.39889577e-01 -7.28147209e-01 -3.90347391e-01 1.02643764e+00 1.13726985e-02 -4.45379674e-01 6.38135135e-01 -3.85063618e-01 -1.98973149e-01 4.83130328e-02 6.09314263e-01 3.04305762e-01 5.68706870e-01 2.24404991e-01 -3.51045579e-01 1.41199201e-01 -5.52948833e-01 -5.35642266e-01 1.10765934e+00 -2.02717587e-01 -9.03080925e-02 1.84983373e-01 1.40280354e+00 2.66440541e-01 -6.35323644e-01 -5.01075566e-01 3.54139835e-01 -3.08776319e-01 -8.24045539e-02 -1.28530324e+00 -7.65242815e-01 7.52313435e-01 -1.09285749e-01 4.60133284e-01 1.05641484e+00 2.13631526e-01 1.20863116e+00 1.10941589e+00 2.94042021e-01 -1.29269278e+00 2.15242416e-01 7.69780040e-01 1.04450023e+00 -1.07922184e+00 -1.95609629e-02 -3.42977673e-01 -6.02020800e-01 7.88530231e-01 6.56655192e-01 9.55671892e-02 5.97430170e-01 -4.59014103e-02 -6.17115125e-02 -4.69280571e-01 -1.66674936e+00 -3.39422196e-01 3.94774497e-01 3.57858650e-02 3.09618205e-01 -2.84927100e-01 -5.72600245e-01 6.54359162e-01 -2.99032360e-01 -1.85906842e-01 2.65134603e-01 1.16375875e+00 -6.13053441e-01 -1.11084235e+00 1.51023552e-01 5.63318789e-01 -5.92657268e-01 -7.48931706e-01 -9.24089491e-01 8.04695249e-01 -7.78970897e-01 1.54779840e+00 -2.55600512e-02 -4.69942123e-01 4.47182953e-01 2.97070503e-01 2.50617623e-01 -6.29364729e-01 -1.20725119e+00 -1.74392655e-01 7.63089061e-01 -7.28623867e-01 -1.03724040e-01 -4.29767162e-01 -9.53664780e-01 -4.17171419e-01 -6.49759233e-01 5.64502776e-01 2.30686188e-01 8.31297278e-01 6.20033145e-01 2.33699046e-02 7.05751300e-01 1.63528413e-01 -9.55141425e-01 -1.13911915e+00 -1.02050364e-01 2.31096789e-01 5.37373602e-01 -1.27263322e-01 -9.17114258e-01 -4.47734594e-01]
[11.783440589904785, 7.802545547485352]
37c6401b-1f44-45a5-8588-ee4c8bb75d96
hybridqa-a-dataset-of-multi-hop-question
2004.07347
null
https://arxiv.org/abs/2004.07347v3
https://arxiv.org/pdf/2004.07347v3.pdf
HybridQA: A Dataset of Multi-Hop Question Answering over Tabular and Textual Data
Existing question answering datasets focus on dealing with homogeneous information, based either only on text or KB/Table information alone. However, as human knowledge is distributed over heterogeneous forms, using homogeneous information alone might lead to severe coverage problems. To fill in the gap, we present HybridQA https://github.com/wenhuchen/HybridQA, a new large-scale question-answering dataset that requires reasoning on heterogeneous information. Each question is aligned with a Wikipedia table and multiple free-form corpora linked with the entities in the table. The questions are designed to aggregate both tabular information and text information, i.e., lack of either form would render the question unanswerable. We test with three different models: 1) a table-only model. 2) text-only model. 3) a hybrid model that combines heterogeneous information to find the answer. The experimental results show that the EM scores obtained by two baselines are below 20\%, while the hybrid model can achieve an EM over 40\%. This gap suggests the necessity to aggregate heterogeneous information in HybridQA. However, the hybrid model's score is still far behind human performance. Hence, HybridQA can serve as a challenging benchmark to study question answering with heterogeneous information.
['Hong Wang', 'Wenhan Xiong', 'Zhiyu Chen', 'Wenhu Chen', 'Hanwen Zha', 'William Wang']
2020-04-15
null
https://aclanthology.org/2020.findings-emnlp.91
https://aclanthology.org/2020.findings-emnlp.91.pdf
findings-of-the-association-for-computational
['multi-hop-question-answering']
['knowledge-base']
[-2.82846153e-01 5.00380039e-01 -1.18811086e-01 -1.10489376e-01 -1.55298638e+00 -8.31709146e-01 3.04476887e-01 4.21407253e-01 -4.52653915e-01 1.03594685e+00 3.36352617e-01 -4.69490111e-01 -7.01421797e-02 -1.00934553e+00 -7.03468382e-01 -2.06193864e-01 7.92523623e-01 9.48804259e-01 8.48800480e-01 -7.81391144e-01 -1.70481671e-02 -5.20792902e-01 -1.39909446e+00 8.43915641e-01 1.65034115e+00 1.01528263e+00 2.49610871e-01 4.40879524e-01 -7.39030778e-01 1.38041270e+00 -7.67846346e-01 -9.85817373e-01 1.63432449e-01 -5.13445675e-01 -1.23749888e+00 -6.68645561e-01 6.81813836e-01 -2.69015461e-01 -3.90185475e-01 1.07892478e+00 4.38805044e-01 -2.07386389e-01 4.60988790e-01 -1.11683810e+00 -7.05204248e-01 6.88217402e-01 -1.45208374e-01 1.06337331e-01 9.43370402e-01 2.27751676e-02 1.46744037e+00 -8.11046481e-01 8.07654440e-01 1.09469080e+00 5.30479312e-01 3.65265250e-01 -7.97937870e-01 -2.38872081e-01 1.56796556e-02 5.30923903e-01 -1.36693883e+00 -2.18782842e-01 4.87472266e-01 -1.96750745e-01 1.01966357e+00 6.96559370e-01 4.85663228e-02 7.26629555e-01 -1.36905849e-01 9.71470892e-01 1.02811289e+00 -2.54152954e-01 3.53185236e-02 3.28204751e-01 3.83831948e-01 6.50878549e-01 4.32456434e-01 -7.11939752e-01 -3.40292484e-01 -3.00629497e-01 3.38184312e-02 -4.34087187e-01 -4.59925026e-01 -9.65432823e-02 -1.23624420e+00 6.99094474e-01 5.58900356e-01 2.78020799e-01 -4.07847971e-01 -3.93822879e-01 2.63901263e-01 5.01496434e-01 -4.52330932e-02 8.89343679e-01 -6.90033376e-01 -4.03443649e-02 -6.11705661e-01 7.34820247e-01 1.32455635e+00 1.22013700e+00 1.01429296e+00 -8.08082402e-01 -5.54764986e-01 8.89414907e-01 1.81945354e-01 6.89996183e-01 3.99434000e-01 -1.11720479e+00 1.26182115e+00 1.14994025e+00 4.03755873e-01 -1.02393377e+00 -3.89983624e-01 -1.20507576e-01 -5.56616545e-01 -6.89070582e-01 9.56564188e-01 -3.46676379e-01 -4.68920082e-01 1.62589753e+00 5.69056571e-01 -5.03666401e-01 3.16396564e-01 9.04771864e-01 1.60232115e+00 6.89625263e-01 -1.92358449e-01 4.25910614e-02 1.62995660e+00 -1.31474698e+00 -1.25350237e+00 -2.62272477e-01 1.01492059e+00 -8.51390779e-01 1.31982815e+00 1.06379054e-02 -1.08297396e+00 -3.49294960e-01 -8.89907420e-01 -4.61271584e-01 -5.60833335e-01 -1.11275092e-01 1.38194039e-01 4.86770689e-01 -8.95658374e-01 -8.60398263e-02 -3.20944011e-01 -3.25880945e-01 -1.02328427e-01 -9.25546810e-02 -2.87237585e-01 -5.89690447e-01 -1.82870734e+00 9.99442279e-01 4.69127536e-01 -1.12478919e-01 -1.44451976e-01 -7.00011194e-01 -7.71293223e-01 1.13038823e-01 9.55622554e-01 -1.03911614e+00 1.45841801e+00 -3.92482489e-01 -9.53061342e-01 4.84264940e-01 -3.89203280e-01 -2.18941063e-01 4.73980993e-01 -2.09469810e-01 -3.72284293e-01 2.49416023e-01 4.87403452e-01 3.18087369e-01 -4.34765173e-03 -1.24467742e+00 -6.63758636e-01 -4.75085407e-01 6.07160449e-01 4.55234081e-01 -7.78366998e-02 -1.60851493e-01 -8.94002855e-01 -2.20926300e-01 1.65294081e-01 -7.21756041e-01 1.56206354e-01 -3.10198754e-01 -5.68148732e-01 -5.90787351e-01 4.95053500e-01 -1.10648429e+00 1.75286067e+00 -1.48733127e+00 -5.51656820e-02 -1.63869426e-01 3.08292568e-01 1.77594796e-01 -1.71937361e-01 8.68899763e-01 4.79250580e-01 2.09011748e-01 -3.18440944e-01 1.87172502e-01 2.31849700e-01 2.06348568e-01 -2.41871327e-01 -2.42293864e-01 3.73296142e-02 1.25499618e+00 -9.39972460e-01 -7.87029743e-01 -4.37623888e-01 -1.27693713e-01 -7.22774148e-01 2.90683210e-01 -6.75014794e-01 -1.79712828e-02 -6.69199944e-01 9.10218477e-01 5.74404359e-01 -5.16676545e-01 1.78844273e-01 -2.94493496e-01 3.56570125e-01 7.24231958e-01 -1.15786767e+00 1.62164617e+00 -2.53551006e-01 1.80612847e-01 8.18632022e-02 -4.41587865e-01 5.91028392e-01 4.21588719e-01 1.84176087e-01 -1.07038033e+00 -2.26424173e-01 5.40251970e-01 1.48006707e-01 -9.45408881e-01 5.71554184e-01 2.05124617e-01 -2.01543495e-01 3.51488054e-01 -2.74857488e-02 -2.87338138e-01 5.64712644e-01 6.14930034e-01 1.57060337e+00 -1.33476377e-01 2.90503837e-02 -2.11099491e-01 6.42167807e-01 3.15988243e-01 4.67696130e-01 9.00584757e-01 -1.23048164e-01 5.86386979e-01 7.28489578e-01 -6.45992458e-02 -6.47832811e-01 -9.63936388e-01 1.57196000e-02 1.05065298e+00 5.56088567e-01 -7.67845750e-01 -9.44947839e-01 -1.16023076e+00 -3.08427084e-02 7.88236380e-01 -4.97252882e-01 7.44130313e-02 -6.34358585e-01 -4.66260761e-01 5.93504488e-01 3.09009284e-01 8.95555258e-01 -8.28783810e-01 -1.42255843e-01 8.96702707e-02 -1.12510812e+00 -1.25207090e+00 -3.37252408e-01 -2.09984943e-01 -4.35326576e-01 -1.37248230e+00 -5.28309584e-01 -6.42040074e-01 3.71294886e-01 1.99119374e-01 1.75511289e+00 3.25088024e-01 4.80513632e-01 4.38258797e-01 -7.26332307e-01 -2.99250811e-01 -1.25724882e-01 5.92910230e-01 -6.27711654e-01 -3.69899690e-01 6.30322337e-01 -6.09642304e-02 -7.62591958e-01 5.88621795e-01 -1.06932151e+00 -8.28521773e-02 5.05134821e-01 7.45095193e-01 3.62369031e-01 -1.80984721e-01 9.00088608e-01 -1.22368741e+00 8.59397173e-01 -9.26253259e-01 -2.91710436e-01 8.76980662e-01 -4.22440261e-01 1.74803033e-01 6.74887002e-01 -7.64943361e-02 -1.09104788e+00 -3.40543747e-01 -5.21939278e-01 2.33566165e-01 1.70194238e-01 9.00812089e-01 -5.54652154e-01 4.26386923e-01 8.81903470e-01 -6.64526448e-02 -2.89844602e-01 -4.37631011e-01 3.15604091e-01 8.12928438e-01 3.29310924e-01 -7.97538817e-01 6.93779647e-01 3.73259783e-02 -6.18548155e-01 -4.17058557e-01 -1.30572236e+00 -7.25577176e-01 -2.68312365e-01 6.74474705e-03 8.33191752e-01 -7.45244920e-01 -5.24326205e-01 1.28627762e-01 -1.23641872e+00 -1.78070575e-01 -1.06767803e-01 2.29247585e-01 -2.63610691e-01 4.45226222e-01 -6.13302767e-01 -4.91165251e-01 -2.90043622e-01 -1.03238618e+00 9.47748005e-01 1.92924976e-01 -3.11232537e-01 -8.82258356e-01 1.80784523e-01 1.23671138e+00 3.53645533e-01 -1.02042854e-01 1.15842831e+00 -1.12662065e+00 -7.94674814e-01 -3.29442978e-01 -3.03867698e-01 2.42158305e-02 -8.77806768e-02 -4.41445678e-01 -8.47600460e-01 8.55821371e-03 1.11479223e-01 -7.00883508e-01 6.95497334e-01 -3.79665256e-01 9.25210178e-01 -5.55992603e-01 -4.59792502e-02 -9.56758335e-02 1.23819768e+00 5.03836870e-02 7.98436284e-01 5.39731503e-01 6.23370051e-01 8.48600805e-01 6.04423702e-01 -3.31954323e-02 1.26928186e+00 7.57265747e-01 4.56903219e-01 2.23956093e-01 -1.34224772e-01 -5.10798156e-01 1.50650084e-01 1.27144921e+00 3.36587161e-01 -5.35757363e-01 -1.30053020e+00 6.47439659e-01 -1.96317172e+00 -8.04587960e-01 -4.06009734e-01 1.98893034e+00 1.23111725e+00 -4.90555819e-03 -6.20895661e-02 -5.60565032e-02 4.73428071e-01 -2.41649617e-02 -6.46163702e-01 4.71259765e-02 -4.06417727e-01 -4.13352773e-02 1.76985011e-01 4.97421741e-01 -6.99800313e-01 7.41612136e-01 5.61814547e+00 1.02229249e+00 -3.58642578e-01 1.03788063e-01 2.99758375e-01 1.98393285e-01 -9.74226177e-01 1.29128247e-01 -8.81599903e-01 5.28439522e-01 9.19538379e-01 -3.03234905e-01 1.74606502e-01 4.82064724e-01 -6.26973450e-01 -2.54804581e-01 -8.99120152e-01 8.01275671e-01 1.78247020e-01 -1.12302363e+00 3.65828037e-01 -2.50545084e-01 7.43030012e-01 3.22938338e-02 -1.78301916e-01 1.00476468e+00 4.13654208e-01 -8.71684372e-01 6.02693796e-01 4.93766606e-01 3.07402849e-01 -4.76099819e-01 1.35476637e+00 8.95534754e-01 -1.09531653e+00 4.77826260e-02 -4.60446477e-01 3.61258797e-02 9.90565419e-02 4.23387229e-01 -5.64489245e-01 9.25759554e-01 6.63682580e-01 1.27343833e-02 -8.72138679e-01 9.02316689e-01 -2.75234967e-01 5.81338882e-01 -3.62417400e-01 -3.49177182e-01 2.55594999e-01 -8.86092633e-02 1.86153114e-01 8.41782808e-01 2.94185489e-01 3.37112963e-01 1.66558176e-01 7.68640280e-01 -5.02315879e-01 6.31369770e-01 -5.30465901e-01 3.04778609e-02 6.79291487e-01 1.13589311e+00 -1.99791789e-01 -4.99268115e-01 -9.64068413e-01 6.92696989e-01 6.48483992e-01 4.73311245e-01 -7.59548247e-01 -7.37767458e-01 1.03951342e-01 1.78418815e-01 1.35163799e-01 1.62614450e-01 -3.31769556e-01 -1.53767920e+00 6.54277503e-01 -1.26464534e+00 9.81933355e-01 -9.24621224e-01 -1.36443210e+00 7.92823076e-01 -8.87651071e-02 -8.51733923e-01 -4.06470090e-01 -2.66070694e-01 -1.44282997e-01 8.77303004e-01 -1.68483722e+00 -8.65568280e-01 -4.87706304e-01 5.94205737e-01 4.37711269e-01 1.70813248e-01 5.78165829e-01 5.87162793e-01 -5.45158565e-01 8.88000250e-01 -9.03029367e-03 3.02046150e-01 9.77271676e-01 -1.47871721e+00 1.56367332e-01 5.36982238e-01 1.04770370e-01 7.57673204e-01 5.54443359e-01 -6.35207951e-01 -1.53939605e+00 -8.72852087e-01 1.30270004e+00 -1.17673838e+00 6.76683664e-01 -2.00401574e-01 -1.48840261e+00 5.58463335e-01 5.84298909e-01 -2.52082348e-01 7.99986601e-01 2.60226965e-01 -6.93550289e-01 -1.11194618e-01 -1.19049811e+00 5.63086331e-01 6.38284028e-01 -6.57796800e-01 -1.09235525e+00 3.68285000e-01 1.06380177e+00 -6.11456573e-01 -1.10751319e+00 7.29842901e-01 2.34452218e-01 -1.01602268e+00 6.76760077e-01 -5.24569929e-01 5.58791876e-01 -4.77461219e-01 -4.73777860e-01 -1.10363615e+00 9.64592397e-02 -2.12140724e-01 -3.72269511e-01 1.34229863e+00 1.13248956e+00 -6.86633825e-01 6.11016154e-01 9.99139845e-01 -1.66879401e-01 -1.02390778e+00 -9.69789982e-01 -6.04378164e-01 3.87427568e-01 -2.12375462e-01 8.80095780e-01 9.17011797e-01 2.52506286e-01 7.63830781e-01 2.16227517e-01 1.38315752e-01 1.07393466e-01 2.54057080e-01 8.06396008e-01 -8.44592273e-01 -1.38008475e-01 -2.99873650e-01 1.47478268e-01 -1.44955242e+00 3.09374165e-02 -8.97968173e-01 3.09937121e-03 -2.16090250e+00 3.75035107e-01 -4.11494285e-01 1.51677579e-02 3.43970418e-01 -7.33402669e-01 -5.81718981e-02 7.21259713e-02 2.15070277e-01 -1.13421774e+00 5.88876307e-01 1.38222754e+00 -2.31147051e-01 1.67912558e-01 -2.54974753e-01 -1.03686976e+00 6.09383881e-01 7.61170268e-01 -2.55956620e-01 -5.33018470e-01 -7.59894311e-01 8.55933487e-01 3.93254489e-01 -8.10193345e-02 -9.07978952e-01 7.25728989e-01 -3.50459591e-02 -1.96650416e-01 -5.72142720e-01 2.72902250e-01 -7.68179357e-01 -3.42039019e-01 1.41640514e-01 -3.93713027e-01 1.94196567e-01 1.53571874e-01 3.53249311e-01 -7.50617206e-01 -4.28535521e-01 3.18046331e-01 -2.97156245e-01 -4.06883836e-01 9.16219652e-02 -3.98666710e-02 9.70322132e-01 5.55051208e-01 2.52662063e-01 -9.52042699e-01 -6.52729034e-01 -2.37583637e-01 7.65282869e-01 2.88988054e-01 3.60200644e-01 2.59091854e-01 -1.37755311e+00 -8.21224868e-01 -3.13177764e-01 5.67159176e-01 1.98430911e-01 5.01402736e-01 7.36974299e-01 -6.73907995e-01 8.17689061e-01 2.30726808e-01 -3.76606554e-01 -8.85701358e-01 5.31779885e-01 3.65511060e-01 -7.23082066e-01 -5.72992004e-02 7.02358723e-01 1.98862121e-01 -9.87593055e-01 1.46643937e-01 -3.33498389e-01 -4.65631723e-01 2.67814606e-01 5.91328204e-01 2.87535787e-01 4.74766463e-01 -4.96016026e-01 -1.68884337e-01 3.99768323e-01 -1.05038121e-01 2.70969886e-02 6.78498268e-01 -3.11458021e-01 -3.43185455e-01 4.49510068e-01 9.46848333e-01 2.87905872e-01 -3.99088293e-01 -6.05040848e-01 2.81956226e-01 -3.09048086e-01 -3.42330307e-01 -1.12619388e+00 -6.99454010e-01 6.62081957e-01 -2.79553533e-02 5.79198599e-01 1.00220442e+00 2.25588068e-01 1.19798672e+00 1.01676822e+00 4.50180411e-01 -9.39067125e-01 -6.04395121e-02 8.64298344e-01 9.95442808e-01 -1.49377251e+00 -3.31243366e-01 -5.07680476e-01 -8.10347259e-01 6.19991124e-01 1.20130765e+00 5.69907010e-01 3.89568835e-01 -1.29194424e-01 4.58827257e-01 -3.62741917e-01 -1.18137991e+00 -5.94006896e-01 5.94574451e-01 2.53300399e-01 4.58824217e-01 -3.02321613e-01 -4.65612322e-01 9.46165085e-01 -4.57204998e-01 -3.29387993e-01 2.39390031e-01 8.85131478e-01 -5.33477306e-01 -9.05724585e-01 -2.98729002e-01 7.35887289e-01 -5.83576739e-01 -2.03879416e-01 -6.89139783e-01 7.89632380e-01 -2.99545135e-02 1.41843688e+00 -2.36473575e-01 -2.94940591e-01 7.65341699e-01 2.13553816e-01 1.93087533e-01 -6.50606871e-01 -7.39536047e-01 -4.06340718e-01 4.27878737e-01 -5.13647020e-01 -1.21350467e-01 -2.95257926e-01 -1.18812144e+00 -2.40757704e-01 -4.39088821e-01 7.90149689e-01 2.48759568e-01 9.31208432e-01 4.08405155e-01 2.18835071e-01 2.24284559e-01 4.95178998e-01 -6.09856606e-01 -1.16328681e+00 -2.06256002e-01 4.56171036e-01 1.48818314e-01 -4.56808209e-01 -3.84542316e-01 -3.09829712e-01]
[10.752554893493652, 7.945376396179199]
70b89555-9ad6-41e8-b94a-4f17c92e506b
coil-sketching-for-computationally-efficient
2305.06482
null
https://arxiv.org/abs/2305.06482v2
https://arxiv.org/pdf/2305.06482v2.pdf
Coil Sketching for computationally-efficient MR iterative reconstruction
Purpose: Parallel imaging and compressed sensing reconstructions of large MRI datasets often have a prohibitive computational cost that bottlenecks clinical deployment, especially for 3D non-Cartesian acquisitions. One common approach is to reduce the number of coil channels actively used during reconstruction as in coil compression. While effective for Cartesian imaging, coil compression inherently loses signal energy, producing shading artifacts that compromise image quality for 3D non-Cartesian imaging. We propose coil sketching, a general and versatile method for computationally-efficient iterative MR image reconstruction. Theory and Methods: We based our method on randomized sketching algorithms, a type of large-scale optimization algorithms well established in the fields of machine learning and big data analysis. We adapt the sketching theory to the MRI reconstruction problem via a structured sketching matrix that, similar to coil compression, considers high-energy virtual coils obtained from principal component analysis. But, unlike coil compression, it also considers random linear combinations of the remaining low-energy coils, effectively leveraging information from all coils. Results: First, we performed ablation experiments to validate the sketching matrix design on both Cartesian and non-Cartesian datasets. The resulting design yielded both improved computational efficiency and preserved signal-to-noise ratio (SNR) as measured by the inverse g-factor. Then, we verified the efficacy of our approach on high-dimensional non-Cartesian 3D cones datasets, where coil sketching yielded up to three-fold faster reconstructions with equivalent image quality. Conclusion: Coil sketching is a general and versatile reconstruction framework for computationally fast and memory-efficient reconstruction.
['Shreyas S. Vasanawala', 'Mert Pilanci', 'Daniel B. Ennis', 'Batu Ozturkler', 'Christopher M. Sandino', 'Zhitao Li', 'Siddharth S. Iyer', 'Frank Ong', 'Julio A. Oscanoa']
2023-05-10
null
null
null
null
['image-reconstruction', 'mri-reconstruction']
['computer-vision', 'computer-vision']
[ 5.87299347e-01 5.81588633e-02 8.08152258e-02 -1.68759212e-01 -9.43434477e-01 -6.06427550e-01 1.74820960e-01 -1.26797929e-01 -4.27832395e-01 4.59452331e-01 5.16158640e-01 -4.55587447e-01 -4.94305402e-01 -2.49548212e-01 -6.37741983e-01 -9.42112505e-01 -6.75217092e-01 3.55945557e-01 -1.16917253e-01 1.56500578e-01 -4.99347262e-02 6.23938143e-01 -8.49878669e-01 1.41987041e-01 7.17052698e-01 9.04695332e-01 5.85143507e-01 5.32409251e-01 3.09877008e-01 7.44601905e-01 -7.98292831e-02 -6.01219982e-02 3.23151290e-01 -4.26865786e-01 -8.45915496e-01 1.61491200e-01 2.28196606e-01 -3.83133143e-01 -5.96290231e-01 9.16551352e-01 9.17778075e-01 -5.77914380e-02 3.46504927e-01 -5.46918511e-01 -9.62396059e-03 8.00775766e-01 -5.72267056e-01 2.30920494e-01 1.59186602e-01 1.60263643e-01 2.77284414e-01 -8.30129325e-01 8.19456935e-01 4.89848763e-01 9.80873048e-01 1.86706588e-01 -1.68321764e+00 -4.00402278e-01 -4.52341884e-01 -4.77019385e-05 -1.27211165e+00 -4.74793077e-01 8.90561163e-01 -4.48010862e-01 6.37004673e-01 6.80763662e-01 1.01550555e+00 6.54295146e-01 5.69193542e-01 6.51887715e-01 1.47209513e+00 -3.74655813e-01 3.02009970e-01 -3.63524526e-01 -1.89790472e-01 5.58825076e-01 7.58963153e-02 1.39791191e-01 -3.07513028e-01 -5.11866570e-01 1.05886590e+00 4.04504538e-02 -6.73372805e-01 -7.27717042e-01 -1.77627277e+00 7.15570986e-01 3.70441765e-01 6.03473365e-01 -8.28587413e-01 9.67751071e-02 4.21316266e-01 -8.15373473e-03 1.01943098e-01 7.39427626e-01 8.94430876e-02 -4.16150428e-02 -1.29653358e+00 7.43320137e-02 5.05963504e-01 6.33541644e-01 1.73303232e-01 1.93093438e-02 -2.93627053e-01 8.07405651e-01 -1.14359036e-01 6.14211738e-01 4.95933294e-01 -1.00672531e+00 8.70776847e-02 -7.84473866e-02 -2.52191722e-01 -1.03481400e+00 -8.85239303e-01 -8.61330271e-01 -1.37707508e+00 -2.21838683e-01 2.19442740e-01 -4.55717295e-02 -5.04840016e-01 1.62350714e+00 2.89123654e-01 1.47633851e-01 -3.58560890e-01 1.40581632e+00 5.94538212e-01 1.45046800e-01 -4.42793667e-02 -6.99295998e-01 1.60206270e+00 -4.92268264e-01 -9.18328226e-01 -6.75225258e-02 5.05097687e-01 -7.65836358e-01 8.18628132e-01 4.57846075e-01 -1.41618609e+00 -4.30708006e-02 -9.89996970e-01 4.32222605e-01 5.74333429e-01 4.68504569e-03 8.66982818e-01 7.30254233e-01 -1.06643522e+00 5.87054014e-01 -9.38709438e-01 2.48258501e-01 4.88515854e-01 3.16690981e-01 -5.81239343e-01 -5.10489762e-01 -8.75569344e-01 9.06354189e-01 -1.14649765e-01 7.46457875e-02 -6.71595871e-01 -1.17888844e+00 -5.76103091e-01 -1.87085211e-01 2.90967554e-01 -7.72463262e-01 1.01472080e+00 -4.46849495e-01 -1.29696524e+00 7.03979015e-01 2.67545320e-02 -5.01667261e-01 4.35631722e-01 5.43642044e-02 -2.86864907e-01 6.43248916e-01 1.05708279e-02 2.48864383e-01 1.04025638e+00 -9.07994390e-01 5.64722896e-01 -5.27181864e-01 -5.00025094e-01 2.58070379e-01 2.67134812e-02 -1.51128203e-01 -1.38942018e-01 -8.47855687e-01 8.10624957e-01 -1.06252778e+00 -6.68484688e-01 4.31789719e-02 -3.12021583e-01 7.13486671e-01 1.54655293e-01 -9.83304620e-01 1.08842027e+00 -2.06329346e+00 2.60463536e-01 4.26749349e-01 9.13734555e-01 -5.38738705e-02 -1.79741587e-02 3.09288561e-01 -4.87446010e-01 -3.34228307e-01 -6.83927834e-01 2.02027902e-01 -4.90456700e-01 1.03311967e-02 -7.07030967e-02 9.81232584e-01 -3.72677058e-01 1.01147223e+00 -1.15136933e+00 -4.63663340e-01 3.55143577e-01 5.68203092e-01 -7.24495292e-01 2.22005352e-01 6.01571620e-01 9.02208090e-01 -3.67372245e-01 3.17399919e-01 9.19999838e-01 -4.20203716e-01 7.05074549e-01 -8.02736342e-01 -3.67587083e-03 -6.78306306e-03 -1.11605132e+00 2.27435732e+00 -5.25003731e-01 3.48087609e-01 6.06401443e-01 -1.12563074e+00 5.48541784e-01 4.83089477e-01 1.30856681e+00 -9.40672636e-01 1.23320632e-01 3.96083176e-01 7.48832151e-02 -5.70011199e-01 2.38763988e-01 -5.89366674e-01 1.38294920e-01 6.09751701e-01 -1.36744827e-01 -4.92581517e-01 -2.58938938e-01 4.79865789e-01 1.32593977e+00 -2.14350566e-01 1.18339844e-01 -6.46924555e-01 1.16946228e-01 -4.12064865e-02 1.17859922e-01 8.14444542e-01 -7.71528333e-02 7.49299645e-01 2.66476333e-01 -4.74601507e-01 -1.20308995e+00 -9.15068150e-01 -6.69391811e-01 2.99307853e-01 -6.17285073e-02 -4.45472479e-01 -7.53055334e-01 -1.72998250e-01 -2.41085857e-01 3.23202670e-01 -3.46541286e-01 2.06460338e-02 -8.36544752e-01 -9.53228712e-01 4.02435452e-01 1.72222778e-02 1.04540348e-01 -5.15465498e-01 -1.11509109e+00 6.02814198e-01 -5.62503159e-01 -1.13115692e+00 -5.91403902e-01 2.90075898e-01 -1.30202067e+00 -1.00370097e+00 -1.08494651e+00 -2.05930620e-01 6.11590207e-01 3.46004069e-01 1.22408712e+00 -5.29853813e-02 -7.00802267e-01 4.46358562e-01 -1.86483607e-01 1.87490895e-01 -1.71301171e-01 -3.73916537e-01 1.32327914e-01 -1.15878746e-01 -5.15022516e-01 -9.11746085e-01 -1.10013270e+00 1.66439027e-01 -1.01082289e+00 3.91859353e-01 8.83300960e-01 1.04612386e+00 8.23995411e-01 -2.10255325e-01 2.40926310e-01 -7.00852036e-01 5.53813934e-01 -3.57006550e-01 -2.12567225e-01 6.64653443e-03 -4.84395802e-01 3.81953657e-01 5.05339146e-01 -4.82073873e-01 -4.19721961e-01 2.41800681e-01 -3.73671412e-01 -5.03869593e-01 2.83024818e-01 4.46671009e-01 3.06580037e-01 -5.06999552e-01 7.01236129e-01 6.28289044e-01 5.15849888e-01 -4.29707825e-01 4.13177013e-01 2.63255626e-01 5.43104589e-01 -5.29789329e-01 3.43184739e-01 7.04329431e-01 3.69160116e-01 -9.33273792e-01 -9.74493846e-02 -1.69409350e-01 -3.58624309e-01 -3.90135318e-01 6.30308807e-01 -5.48775375e-01 -7.28276372e-01 2.16435090e-01 -7.73157179e-01 -1.85869843e-01 -3.96543950e-01 9.83655989e-01 -6.34678841e-01 6.87367260e-01 -5.85811377e-01 -3.96244884e-01 -7.45152056e-01 -1.59198916e+00 9.68923867e-01 -4.48587060e-01 -1.87507391e-01 -7.25376308e-01 1.30472347e-01 1.16149426e-01 9.54555035e-01 3.55568320e-01 1.02048910e+00 4.54946235e-02 -4.11448568e-01 3.23778205e-02 7.62168467e-02 -8.87890682e-02 -2.53180981e-01 -1.02875626e+00 -5.43955207e-01 -5.15700340e-01 5.79095125e-01 -1.53960213e-01 4.15965706e-01 9.56892669e-01 1.35027826e+00 -2.52479315e-01 -2.66532660e-01 9.38486278e-01 1.44659710e+00 -1.45308882e-01 8.74844313e-01 -4.98092435e-02 4.89185303e-01 3.44963133e-01 3.45907919e-03 4.78085428e-01 4.06753831e-02 9.93777633e-01 1.13344967e-01 -2.88690954e-01 -3.32327545e-01 -1.40183702e-01 -2.15750292e-01 1.16505718e+00 -7.53640383e-03 6.34886384e-01 -9.00717795e-01 4.47513640e-01 -1.27270043e+00 -6.95069671e-01 -2.28091151e-01 2.47852683e+00 8.10117006e-01 -5.52334249e-01 9.31791589e-02 2.70111024e-01 3.27908665e-01 7.22442567e-02 -5.03631771e-01 1.01494648e-01 -1.06984250e-01 4.16627139e-01 7.99147189e-01 4.57692176e-01 -7.11678743e-01 8.98547694e-02 6.72654390e+00 8.37632298e-01 -1.34338558e+00 7.16313899e-01 4.33012545e-01 -3.67064357e-01 -6.41747773e-01 -9.04854089e-02 2.17674106e-01 3.49384636e-01 8.78463268e-01 1.44290417e-01 9.10444915e-01 4.61816400e-01 7.90066049e-02 -3.06690007e-01 -7.93200970e-01 1.45117748e+00 -1.79865941e-01 -1.70318520e+00 -4.48457062e-01 1.09717786e-01 4.29664016e-01 1.97724432e-01 -1.09616034e-01 -2.83524215e-01 -2.48011500e-01 -1.06592941e+00 6.59268141e-01 4.02751654e-01 1.39387131e+00 -4.83830690e-01 3.48570138e-01 2.94241399e-01 -9.47149992e-01 1.90942019e-01 -1.26310930e-01 3.50658685e-01 6.12257659e-01 1.14002049e+00 -5.97457230e-01 6.55161738e-01 4.97440994e-01 4.31482404e-01 -1.82692915e-01 1.04123354e+00 1.96624190e-01 4.40202355e-01 -4.52435434e-01 4.94324952e-01 8.71045813e-02 -3.49923700e-01 8.39892745e-01 1.02579713e+00 2.89881110e-01 7.04139948e-01 -2.04700902e-02 6.48715079e-01 3.15862685e-01 7.42103159e-02 -6.29922152e-01 2.35188186e-01 4.29260045e-01 1.43132043e+00 -7.86892891e-01 -2.23684400e-01 -2.09767818e-01 7.69354343e-01 5.65532520e-02 3.05166155e-01 -5.85489631e-01 -5.64908683e-02 1.22852176e-01 6.38122320e-01 5.26407734e-02 -4.10957277e-01 -4.41722065e-01 -1.22334623e+00 2.98649427e-02 -1.04503739e+00 1.17173918e-01 -6.51911676e-01 -1.00739133e+00 8.27391505e-01 2.01192424e-01 -1.19469237e+00 -1.47665694e-01 -2.26447180e-01 -3.74307558e-02 8.55802715e-01 -1.16289997e+00 -7.54353344e-01 -1.66565418e-01 7.74129212e-01 6.24979055e-03 2.31448129e-01 9.34037328e-01 5.77086508e-01 8.53258669e-02 4.16592151e-01 1.68576568e-01 -2.20433965e-01 2.81506330e-01 -8.08613062e-01 -2.44359691e-02 6.64764285e-01 -2.30226591e-02 7.97719300e-01 7.12276876e-01 -5.99851727e-01 -2.29960752e+00 -4.96245742e-01 3.20330977e-01 7.28149936e-02 4.97961193e-01 -2.82768488e-01 -8.20911407e-01 3.30186248e-01 -2.57528331e-02 3.08683127e-01 7.44672000e-01 -1.45672888e-01 3.75631563e-02 2.07396615e-02 -1.48688698e+00 3.01244080e-01 1.02604759e+00 -5.25225997e-01 -3.54142070e-01 5.71659267e-01 3.02454203e-01 -7.20964372e-01 -1.35061049e+00 3.21206599e-01 7.21578598e-01 -8.77290368e-01 1.28416479e+00 -1.84074268e-02 2.99628913e-01 -6.93784580e-02 -2.05408111e-01 -1.37142408e+00 -5.53204477e-01 -8.49973857e-01 -6.14391565e-02 5.05667776e-02 -1.41268075e-01 -7.05720186e-01 6.99644268e-01 6.00942731e-01 -3.11747313e-01 -9.49366093e-01 -1.38164186e+00 -5.38019121e-01 1.05618164e-01 -6.20130599e-01 3.66009742e-01 9.67962682e-01 3.38444740e-01 7.55774304e-02 -5.26413858e-01 -1.12824604e-01 1.17497778e+00 2.33557239e-01 1.43295914e-01 -7.68115580e-01 -7.13116348e-01 -3.02497745e-01 -2.63852119e-01 -1.15957880e+00 -3.16381842e-01 -1.35803974e+00 -1.99178264e-01 -1.21235609e+00 4.49327588e-01 -9.99561012e-01 1.78801846e-02 1.55474037e-01 1.05200969e-02 3.77176762e-01 2.57529140e-01 4.64457601e-01 -1.90597907e-01 3.79625380e-01 1.78661096e+00 4.97294739e-02 1.51305199e-02 -4.22710449e-01 -5.46449900e-01 2.03956500e-01 2.86138356e-01 -5.11307418e-01 -4.06473279e-01 -4.60355371e-01 2.02311650e-01 9.15647864e-01 4.19171721e-01 -9.90034759e-01 1.66423023e-01 2.26227492e-01 2.72085816e-01 -3.29546273e-01 2.40020990e-01 -1.04746282e+00 9.18823123e-01 7.99057662e-01 -5.77191263e-02 8.70079559e-04 1.86849311e-01 1.00286774e-01 8.70509669e-02 5.61043806e-02 8.19077492e-01 -2.93705672e-01 -6.49058744e-02 2.50978857e-01 -4.24844086e-01 -3.04140374e-02 7.19431520e-01 -7.06772879e-02 2.12605223e-01 -3.02564085e-01 -9.46725368e-01 -3.90668333e-01 2.86605299e-01 -1.30567923e-01 9.06008899e-01 -1.41124725e+00 -8.42249691e-01 6.28111720e-01 -3.06487381e-01 -3.18093657e-01 9.49718893e-01 1.83806479e+00 -6.77606404e-01 4.38342154e-01 -3.11963707e-01 -1.01873553e+00 -8.73222768e-01 5.94239712e-01 6.43937230e-01 -5.16857922e-01 -1.29853404e+00 5.56227386e-01 1.09472059e-01 -3.47474247e-01 -2.33481482e-01 -1.95947900e-01 2.39372149e-01 -3.73452932e-01 7.16335535e-01 2.22504154e-01 4.49348360e-01 -5.41624606e-01 -5.80721796e-01 6.77305579e-01 1.95102960e-01 -3.38371813e-01 1.51320887e+00 -1.13704704e-01 -1.76844776e-01 -9.60095972e-02 1.28057766e+00 3.26876864e-02 -1.10153723e+00 -3.46307606e-01 -4.52005178e-01 -6.72351003e-01 6.58223331e-01 -7.87462592e-01 -1.29319835e+00 7.47944832e-01 8.88479173e-01 -1.12090044e-01 1.30732930e+00 -4.54711840e-02 8.14947307e-01 -2.00633809e-01 9.51943338e-01 -4.56918061e-01 -2.75162578e-01 -1.05747096e-02 1.14902544e+00 -7.60830700e-01 3.92070204e-01 -3.02091867e-01 -6.72260940e-01 7.93291986e-01 -3.48178715e-01 -2.15245970e-02 6.54202580e-01 7.09302485e-01 -2.49301612e-01 -6.03978276e-01 -1.67847261e-01 4.41259503e-01 2.40612000e-01 4.69205052e-01 5.07259667e-01 4.85960275e-01 -5.43042243e-01 5.28088689e-01 -7.34030977e-02 2.12130770e-01 3.48287612e-01 9.37855542e-01 5.62756285e-02 -9.47981536e-01 -6.10812664e-01 6.67622566e-01 -4.47910100e-01 -3.24769586e-01 4.43489611e-01 2.24603102e-01 -2.34866738e-01 6.51995540e-01 -2.85990536e-01 -2.77019262e-01 2.68936485e-01 -2.62606472e-01 1.09015942e+00 -1.13174945e-01 -7.50326872e-01 3.04596543e-01 -5.41184843e-02 -9.80364859e-01 -2.19343349e-01 -7.17454195e-01 -1.12684095e+00 -3.17622334e-01 -5.44624701e-02 2.67342865e-01 1.05628645e+00 6.57262325e-01 5.28828859e-01 6.60807431e-01 5.87759495e-01 -1.16522920e+00 -6.91659808e-01 -8.01419675e-01 -8.75465930e-01 4.31150496e-01 3.22898418e-01 -5.95658302e-01 6.90837437e-03 -3.11235696e-01]
[13.466877937316895, -2.425082206726074]
71837f7e-1907-4785-8a40-c4abc804bf45
known-unknowns-uncertainty-quality-in
1612.01251
null
http://arxiv.org/abs/1612.01251v2
http://arxiv.org/pdf/1612.01251v2.pdf
Known Unknowns: Uncertainty Quality in Bayesian Neural Networks
We evaluate the uncertainty quality in neural networks using anomaly detection. We extract uncertainty measures (e.g. entropy) from the predictions of candidate models, use those measures as features for an anomaly detector, and gauge how well the detector differentiates known from unknown classes. We assign higher uncertainty quality to candidate models that lead to better detectors. We also propose a novel method for sampling a variational approximation of a Bayesian neural network, called One-Sample Bayesian Approximation (OSBA). We experiment on two datasets, MNIST and CIFAR10. We compare the following candidate neural network models: Maximum Likelihood, Bayesian Dropout, OSBA, and --- for MNIST --- the standard variational approximation. We show that Bayesian Dropout and OSBA provide better uncertainty information than Maximum Likelihood, and are essentially equivalent to the standard variational approximation, but much faster.
['Pedro Tabacof', 'Ramon Oliveira', 'Eduardo Valle']
2016-12-05
null
null
null
null
['known-unknowns']
['miscellaneous']
[-9.15763974e-02 4.57607567e-01 -8.23977962e-03 -8.58523965e-01 -8.04647744e-01 -2.98339009e-01 5.01900136e-01 -1.26485387e-02 -6.89244568e-01 8.50098014e-01 -2.38459921e-04 -2.44319424e-01 -3.41160595e-01 -5.69175780e-01 -9.84461963e-01 -6.51125669e-01 -1.90931223e-02 8.55898261e-01 6.84116662e-01 4.89118010e-01 2.06976846e-01 6.62191510e-01 -1.31829381e+00 1.89442128e-01 8.86284649e-01 1.02871120e+00 -3.30658942e-01 5.57145476e-01 9.77367535e-02 4.82893139e-01 -4.85951781e-01 -5.84652662e-01 1.38732299e-01 -4.07161832e-01 -6.31951571e-01 -3.27468306e-01 8.50010693e-01 -5.68673134e-01 -6.34453237e-01 1.63707793e+00 -2.57669538e-02 3.58587593e-01 1.28664279e+00 -1.37142110e+00 -3.07577550e-01 1.25205302e+00 -2.07132310e-01 3.84572238e-01 -2.86993980e-01 -8.99967477e-02 1.08488631e+00 -7.23272502e-01 2.98839450e-01 1.70002794e+00 6.83590472e-01 8.89197230e-01 -1.72188818e+00 -8.37922752e-01 4.07274663e-01 1.91742554e-01 -1.27148151e+00 -6.86624110e-01 5.92203379e-01 -6.13624632e-01 7.79558480e-01 -2.17759326e-01 2.42216736e-01 1.44039845e+00 2.38029391e-01 1.05200565e+00 7.66580224e-01 -7.73144737e-02 8.01920474e-01 2.01843157e-01 9.62053597e-01 8.01241279e-01 5.06326735e-01 3.95156682e-01 -4.46035981e-01 -6.46818340e-01 5.76256514e-01 -4.78040893e-03 -3.33106428e-01 -2.54729956e-01 -5.89920998e-01 9.54916894e-01 4.46231477e-02 -8.05317611e-02 -2.70945996e-01 5.75093448e-01 1.85773328e-01 3.15878481e-01 4.33549792e-01 2.57822186e-01 -5.15816867e-01 2.96825264e-03 -1.07534802e+00 5.10794878e-01 9.54112470e-01 7.48914599e-01 6.77907050e-01 4.50277001e-01 -4.59425479e-01 6.82990968e-01 8.60750556e-01 4.84244376e-01 2.78117955e-01 -1.28013110e+00 1.57309145e-01 -9.76305753e-02 1.21382125e-01 -2.97529995e-01 -3.15919459e-01 -3.45320672e-01 -7.05718994e-01 4.72889334e-01 7.46590495e-01 -3.38284671e-01 -1.12593126e+00 2.10698318e+00 -2.37571031e-01 5.62214732e-01 -1.90545265e-02 5.00771999e-01 4.40526277e-01 5.83496153e-01 -2.04572290e-01 -2.56151021e-01 7.69650161e-01 -5.12170732e-01 -7.50144362e-01 -1.68979093e-01 4.63018477e-01 -6.90071881e-02 5.94744325e-01 9.11163688e-01 -9.32066441e-01 -2.62875944e-01 -1.16925371e+00 4.83393848e-01 -8.95616040e-02 -8.97714496e-02 3.81038606e-01 8.80208552e-01 -9.82216835e-01 1.28929007e+00 -1.40596831e+00 1.33882642e-01 6.57621443e-01 3.42507392e-01 -1.04513004e-01 1.42259493e-01 -1.06409907e+00 8.25390935e-01 9.50852752e-01 -1.54980868e-01 -1.29630744e+00 -5.35485864e-01 -6.58628583e-01 1.94743246e-01 1.40968055e-01 -1.99814677e-01 1.51886201e+00 -8.62502813e-01 -1.47509754e+00 3.53419721e-01 -1.67253047e-01 -9.71346140e-01 3.05190742e-01 -3.44484419e-01 -2.41841108e-01 -1.48991585e-01 -6.33749545e-01 7.44544923e-01 1.01627612e+00 -9.85808074e-01 -6.64906919e-01 -4.48037118e-01 -4.55274016e-01 -3.66867959e-01 -1.11224828e-02 -6.82565346e-02 -5.07230163e-02 -3.40719640e-01 5.07931113e-01 -8.61559212e-01 -8.37983564e-02 2.19222307e-02 -8.97819221e-01 -5.53844392e-01 5.38470864e-01 -5.74614704e-01 1.08173120e+00 -2.02025199e+00 -5.67918904e-02 7.31690824e-01 3.89006585e-01 4.96719070e-02 -9.04805809e-02 -5.09603262e-01 -1.32993758e-01 1.33341581e-01 -3.80917817e-01 -3.44542891e-01 2.86705911e-01 5.35155535e-01 -3.77860188e-01 5.12441397e-01 3.37815344e-01 3.26764226e-01 -6.27195835e-01 -2.20765620e-01 -2.27116883e-01 2.09220618e-01 -8.64494443e-01 -1.43768623e-01 -4.66028601e-01 1.46697462e-02 -3.84573430e-01 3.31482381e-01 7.01711953e-01 1.48660943e-01 -2.09222957e-01 -3.82477604e-02 3.85120720e-01 3.35234404e-01 -1.42900705e+00 1.23730254e+00 1.56830534e-01 9.49609816e-01 -2.68492162e-01 -9.72990155e-01 1.12693989e+00 3.16238731e-01 -5.23973852e-02 1.74622312e-01 2.39822656e-01 2.75967121e-01 3.26215059e-01 6.58804476e-02 2.54032761e-01 1.60678729e-01 2.20899194e-01 2.25473881e-01 6.75648093e-01 2.64631957e-01 1.59603298e-01 2.41377354e-01 1.17764854e+00 9.77348834e-02 -1.31659910e-01 -3.72938037e-01 -1.45253897e-01 -6.25807285e-01 9.85366285e-01 1.54131651e+00 -4.77977067e-01 6.22438908e-01 1.24476218e+00 -1.25553131e-01 -9.45206046e-01 -1.77628326e+00 -5.12084424e-01 6.46739006e-01 -4.03977662e-01 -2.30962411e-01 -8.45859766e-01 -8.33286166e-01 1.45024955e-01 1.34013605e+00 -3.71993572e-01 -6.05187654e-01 -5.45146577e-02 -9.73989129e-01 7.22881138e-01 5.79153180e-01 2.57732064e-01 -7.22646177e-01 -9.88332480e-02 1.44842118e-01 2.65756905e-01 -8.08135450e-01 -1.71229869e-01 6.45646214e-01 -1.28563464e+00 -7.92364657e-01 -2.58434117e-01 -1.80982858e-01 3.68098587e-01 -7.93721080e-01 1.09804654e+00 -4.44661796e-01 1.19641066e-01 9.44528729e-02 4.07176390e-02 -5.07570624e-01 -6.21419489e-01 -2.25073934e-01 7.81880260e-01 7.11772777e-03 6.97353423e-01 -6.40820742e-01 -4.67442088e-02 5.12040395e-04 -7.26382554e-01 -6.78869724e-01 5.31170547e-01 8.46293092e-01 6.08579159e-01 2.04687133e-01 4.12645996e-01 -1.16321564e+00 5.53180754e-01 -3.66313517e-01 -1.12681246e+00 2.36707166e-01 -9.08336103e-01 8.26594651e-01 2.65763342e-01 -5.93534052e-01 -1.19321084e+00 1.27687380e-01 -3.35511804e-01 -7.71749854e-01 -5.00435710e-01 3.19806248e-01 -1.17728889e-01 2.75759578e-01 9.88009155e-01 -2.77762145e-01 -3.08552921e-01 -7.85788834e-01 2.08884180e-01 4.47505534e-01 6.91404641e-01 -7.29155540e-01 6.55578792e-01 1.20485306e-01 -1.80135608e-01 -6.56960726e-01 -1.06876850e+00 -4.15380485e-03 -5.89361191e-01 5.67411371e-02 7.05493927e-01 -5.43618262e-01 -5.36691010e-01 5.99822223e-01 -1.45286119e+00 -1.75386205e-01 -2.92204589e-01 1.04846787e+00 -3.24584752e-01 1.22833580e-01 -8.46049249e-01 -1.19283700e+00 3.58968824e-02 -1.22960997e+00 4.34255749e-01 4.82688546e-01 -1.75630078e-01 -9.22012746e-01 2.65947014e-01 -3.79099488e-01 1.17196754e-01 -2.21124411e-01 9.02557373e-01 -1.45733178e+00 -4.26746607e-01 -2.59341300e-01 -2.53025591e-01 6.68986142e-01 -4.99355793e-01 6.47001326e-01 -1.55032754e+00 -1.62997887e-01 -1.63608164e-01 -3.72959077e-01 1.70238519e+00 9.27340090e-01 1.34972262e+00 -2.46559709e-01 -3.72604191e-01 6.02930605e-01 1.04116929e+00 2.44925395e-01 5.86534858e-01 -2.31985018e-01 4.29407984e-01 3.23254019e-01 9.01099816e-02 3.68880242e-01 -3.14992487e-01 3.66118759e-01 3.44066322e-01 6.69526815e-01 4.68136042e-01 -3.37745160e-01 8.12393785e-01 5.80862880e-01 2.51652718e-01 -1.33510962e-01 -8.72036994e-01 3.06945503e-01 -1.95888829e+00 -1.09917724e+00 -1.18924864e-01 2.47407842e+00 9.26475048e-01 9.74525034e-01 9.33229178e-02 -4.38096486e-02 6.59911633e-01 -1.72224194e-01 -7.73080170e-01 -4.89728242e-01 4.88109924e-02 1.16973609e-01 5.25625408e-01 7.43420839e-01 -1.28891802e+00 8.15006673e-01 6.85641432e+00 1.02116263e+00 -4.12364662e-01 2.16944646e-02 8.08695018e-01 -2.83563018e-01 -1.67614266e-01 1.41157717e-01 -1.35986042e+00 3.57011199e-01 1.54081273e+00 2.05378756e-01 2.08663940e-01 1.22102308e+00 -1.22927032e-01 -2.68479027e-02 -1.69124913e+00 8.05227399e-01 -2.43182227e-01 -1.27923012e+00 -2.14589313e-02 -5.37428260e-02 6.88594103e-01 6.15456223e-01 4.94891927e-02 5.54587841e-01 8.78464103e-01 -8.15953076e-01 7.43979931e-01 1.03991878e+00 1.67734951e-01 -9.47752118e-01 8.69970143e-01 3.78268480e-01 -5.86996913e-01 6.70927018e-03 -9.55372870e-01 3.71814787e-01 -2.25395381e-01 1.06758070e+00 -4.22617286e-01 -1.66900724e-01 7.08765745e-01 4.36194897e-01 -4.68820691e-01 1.27705872e+00 -1.73458576e-01 1.06954598e+00 -7.25501359e-01 -8.20388272e-02 2.26863399e-01 -3.44097853e-01 7.90197849e-01 9.62310255e-01 1.38318539e-01 -2.70463824e-01 1.15488440e-01 1.60945332e+00 -1.76668614e-01 -5.61106324e-01 -5.72126031e-01 -9.48741212e-02 8.08827281e-01 7.23377764e-01 -5.56335032e-01 -3.80505681e-01 -1.27979115e-01 7.05369174e-01 3.98245484e-01 5.01228571e-01 -6.08823538e-01 -7.13393986e-01 7.16637194e-01 -5.83804011e-01 2.13870004e-01 2.57795602e-01 2.72025876e-02 -1.24158347e+00 -3.01784188e-01 -6.34618759e-01 4.14659798e-01 -6.03365242e-01 -1.49173701e+00 6.01394713e-01 3.74087393e-01 -7.08548725e-01 -5.23002207e-01 -1.15213203e+00 -7.39899933e-01 6.78886294e-01 -1.07470274e+00 -3.29096429e-02 4.10425514e-01 3.21762174e-01 3.07403386e-01 -3.68358970e-01 7.01334596e-01 -3.51602286e-02 -9.66495037e-01 8.27581167e-01 6.40064359e-01 3.57106239e-01 4.39220279e-01 -1.49398351e+00 3.94453466e-01 9.40894067e-01 5.83848298e-01 4.05498207e-01 8.84520948e-01 -5.22718251e-01 -6.34096205e-01 -1.02369344e+00 4.99436319e-01 -6.11585140e-01 6.11359417e-01 -1.41459122e-01 -1.37475836e+00 9.13703501e-01 -3.53334427e-01 1.39575414e-02 3.62158656e-01 4.89231706e-01 -6.87277555e-01 -1.57935262e-01 -1.26145566e+00 5.55813015e-01 6.84029818e-01 -5.65678418e-01 -9.16644692e-01 3.09084713e-01 8.48682940e-01 1.68700833e-02 -5.71714342e-01 1.16699353e-01 5.52613378e-01 -1.16857958e+00 6.42021656e-01 -9.91564810e-01 -3.67690995e-02 7.13056698e-02 -3.58659208e-01 -1.47414351e+00 -3.05625081e-01 -5.87501287e-01 -5.85942686e-01 1.11647022e+00 9.24186468e-01 -7.11457729e-01 1.09931409e+00 8.22195411e-01 -1.24347389e-01 -3.62378985e-01 -1.14356971e+00 -1.18544579e+00 3.86285454e-01 -9.92748380e-01 1.62247643e-01 7.05365062e-01 -1.85915068e-01 7.01693445e-02 -1.24261759e-01 4.37574059e-01 1.17548096e+00 -5.59217036e-01 1.96921043e-02 -2.03948569e+00 -5.44370592e-01 -7.87327647e-01 -6.20026231e-01 -1.02326989e+00 6.38405144e-01 -8.50390196e-01 2.41404444e-01 -8.62507880e-01 2.65033424e-01 1.36116296e-01 -8.23247373e-01 4.98591244e-01 1.10608757e-01 -1.55342653e-01 -3.61726433e-01 -5.38491979e-02 -5.52704513e-01 6.41243041e-01 2.08345145e-01 -5.97487204e-02 -1.97724044e-01 4.34181750e-01 -1.49978563e-01 1.38548267e+00 8.51251423e-01 -1.05535901e+00 -3.18499207e-01 -1.04931757e-01 8.96407440e-02 -2.16808114e-02 4.05942082e-01 -1.25547552e+00 1.53015733e-01 2.93451045e-02 5.51392734e-01 -8.06481659e-01 2.90046424e-01 -5.14573514e-01 -3.35562944e-01 6.17528558e-01 -8.35845411e-01 -3.38180304e-01 -2.69983709e-02 8.61341000e-01 3.71867456e-02 -9.60075855e-01 1.13514137e+00 1.36290491e-01 -3.97974759e-01 5.09777904e-01 -6.38599396e-01 2.94337660e-01 2.58424491e-01 2.08808154e-01 -3.54469925e-01 -3.64277780e-01 -1.15660667e+00 1.59246713e-01 2.95741186e-02 -4.55228500e-02 8.89343202e-01 -1.28629661e+00 -7.36367404e-01 2.78307647e-01 -3.25952917e-02 -2.38684922e-01 -2.71316946e-01 7.68330991e-01 -2.81893790e-01 2.49359280e-01 4.96724285e-02 -7.68795669e-01 -8.52108657e-01 9.37600881e-02 7.71442592e-01 -6.08177297e-02 -4.36739922e-01 1.20718908e+00 -3.27913873e-02 -5.61023414e-01 9.52929735e-01 -7.27920890e-01 -2.99109370e-02 -1.65367231e-01 5.06338418e-01 3.90795708e-01 -9.35017020e-02 5.43765537e-02 -2.39528432e-01 -1.51670232e-01 -4.50995296e-01 -4.92332101e-01 9.73192692e-01 3.07648450e-01 1.34097964e-01 9.40258861e-01 8.93524945e-01 -5.20200849e-01 -1.64264309e+00 -7.32345521e-01 6.52772486e-01 -1.32782504e-01 6.10363841e-01 -4.98299211e-01 -1.00964034e+00 1.09985769e+00 9.74431634e-01 1.58287197e-01 5.69533706e-01 -4.99246791e-02 8.68548378e-02 1.27596927e+00 5.54416236e-03 -1.23023379e+00 -2.24959314e-01 7.57567942e-01 3.88163835e-01 -1.12528431e+00 -1.14820361e-01 3.52117717e-01 -3.76433939e-01 1.23880255e+00 7.51007974e-01 -3.65049183e-01 1.11400831e+00 3.08237344e-01 -3.14843029e-01 7.57044777e-02 -1.09163225e+00 6.17325939e-02 5.71778774e-01 6.13341033e-01 1.35886699e-01 -1.26727268e-01 4.28705394e-01 9.38226521e-01 8.98619071e-02 -2.54852593e-01 6.27560496e-01 3.77531558e-01 -9.87668097e-01 -5.28725922e-01 -2.65314192e-01 1.17261922e+00 -5.04648805e-01 -1.94589466e-01 -3.15169632e-01 4.24629390e-01 -2.75767356e-01 7.61774063e-01 4.37833846e-01 -3.77080232e-01 1.32831961e-01 7.40376353e-01 5.36190033e-01 -4.74782646e-01 9.76803154e-02 1.05353273e-01 1.54920474e-01 -7.43863404e-01 7.76673108e-02 -8.03284943e-01 -1.10076034e+00 -2.07784668e-01 -6.95884943e-01 3.24333817e-01 5.87611616e-01 1.21144938e+00 -9.33792219e-02 4.42633659e-01 3.03259134e-01 -6.36208355e-01 -1.35620165e+00 -1.20919406e+00 -1.00891674e+00 -5.48220463e-02 4.70459163e-01 -7.31639862e-01 -9.47348177e-01 -3.21656406e-01]
[7.4073638916015625, 3.7843267917633057]
338bdd44-41e6-42f1-8de4-703f56b9fafb
combining-graph-and-sequence-information-to
null
null
https://openreview.net/forum?id=Skx73lBFDS
https://openreview.net/pdf?id=Skx73lBFDS
Combining graph and sequence information to learn protein representations
Computational methods that infer the function of proteins are key to understanding life at the molecular level. In recent years, representation learning has emerged as a powerful paradigm to discover new patterns among entities as varied as images, words, speech, molecules. In typical representation learning, there is only one source of data or one level of abstraction at which the learned representation occurs. However, proteins can be described by their primary, secondary, tertiary, and quaternary structure or even as nodes in protein-protein interaction networks. Given that protein function is an emergent property of all these levels of interactions in this work, we learn joint representations from both amino acid sequence and multilayer networks representing tissue-specific protein-protein interactions. Using these representations, we train machine learning models that outperform existing methods on the task of tissue-specific protein function prediction on 10 out of 13 tissues. Furthermore, we outperform existing methods by 19% on average.
['Ali Abdalla', 'Pelkins Ajanoh', 'Mohamed Coulibali', 'Hassan Kané']
2019-09-25
null
null
null
null
['protein-function-prediction']
['medical']
[ 6.78020239e-01 2.06235111e-01 -4.33497161e-01 -2.80223668e-01 -5.37726462e-01 -4.02034223e-01 7.14854121e-01 6.78146720e-01 -1.93204060e-01 1.18533516e+00 3.91701072e-01 -3.99467707e-01 -1.07665300e-01 -7.13557005e-01 -9.40832853e-01 -8.96390319e-01 -2.62569398e-01 4.96843606e-01 7.54684210e-02 -1.31853774e-01 8.36061779e-03 6.23655319e-01 -1.36159790e+00 6.74168289e-01 6.91684186e-01 8.38505447e-01 2.40815178e-01 7.36781776e-01 -4.61628079e-01 9.55042720e-01 -6.38706803e-01 -1.68888882e-01 -3.88590455e-01 -3.68589014e-01 -9.55400050e-01 -1.62210345e-01 -3.42492126e-02 4.13785785e-01 -5.23020566e-01 8.79902840e-01 1.47655889e-01 4.45823446e-02 1.07375157e+00 -7.47205973e-01 -1.02660024e+00 3.70077610e-01 -4.10272717e-01 8.38506892e-02 4.47151572e-01 -7.67517164e-02 1.23252487e+00 -9.29513276e-01 8.76257956e-01 1.34995019e+00 3.95100772e-01 5.70517063e-01 -1.88195920e+00 -2.93103099e-01 1.25165358e-01 1.52453437e-01 -1.04565907e+00 -1.60383880e-01 3.18364948e-01 -6.89269185e-01 1.39655149e+00 1.81174457e-01 5.20208657e-01 1.04175234e+00 6.29622817e-01 7.06344008e-01 7.83550739e-01 -3.00812811e-01 9.64839831e-02 -2.72518337e-01 3.53299022e-01 9.00263667e-01 1.81600198e-01 -3.71878564e-01 -5.98447680e-01 -6.95844948e-01 4.85563219e-01 4.34178859e-01 -4.48388577e-01 -1.85154513e-01 -1.34299183e+00 7.29656518e-01 5.60255051e-01 5.39982200e-01 -5.32273889e-01 4.33764309e-01 3.77627611e-01 2.05408409e-01 6.69827461e-01 6.94956362e-01 -8.94490480e-01 1.68920666e-01 -2.90041417e-01 1.39665633e-01 9.42471504e-01 4.57064211e-01 9.85059321e-01 -2.34782591e-01 1.09470010e-01 1.01714718e+00 2.78565377e-01 1.24576963e-01 8.00265551e-01 -4.04971957e-01 -2.19935700e-01 8.73177350e-01 -1.12529241e-01 -8.64476025e-01 -2.73644537e-01 -2.40878224e-01 -9.58594024e-01 -5.95453978e-02 2.69220084e-01 2.65079290e-01 -1.07254529e+00 1.80270302e+00 9.27186906e-02 3.35686475e-01 1.97797060e-01 5.25665522e-01 8.87826264e-01 9.95758653e-01 4.63681072e-01 -2.09477261e-01 1.41166568e+00 -6.55966520e-01 -7.19595015e-01 3.23379450e-02 8.33575428e-01 -4.32175428e-01 1.90055862e-01 1.74996346e-01 -8.82762015e-01 -2.20306739e-01 -9.73586023e-01 -8.59894827e-02 -8.14005733e-01 -1.43546730e-01 1.05378687e+00 -5.44552803e-02 -7.91350842e-01 7.82415211e-01 -8.22281480e-01 -4.51315939e-01 3.55032444e-01 5.73324025e-01 -6.93649769e-01 6.85599521e-02 -1.21427143e+00 1.11727440e+00 4.06928778e-01 -1.86508968e-01 -4.71979290e-01 -8.70687604e-01 -8.01820457e-01 3.05031657e-01 -3.09741676e-01 -8.06188226e-01 9.80513632e-01 -8.58078778e-01 -1.14183438e+00 1.13907194e+00 -6.11945450e-01 -5.98968744e-01 -4.75260586e-01 3.55027393e-02 -3.18177640e-01 1.07816547e-01 -2.87378520e-01 5.93166471e-01 3.47340971e-01 -1.08768249e+00 -2.92205513e-01 -3.20221752e-01 -2.88414173e-02 -4.71787415e-02 5.91829270e-02 -4.69558574e-02 1.44962534e-01 -3.76857936e-01 2.14469239e-01 -7.72589445e-01 -5.36895752e-01 2.65864253e-01 -3.49746406e-01 -5.21645010e-01 5.48462987e-01 -5.95220387e-01 7.31986105e-01 -1.98737395e+00 6.22807205e-01 1.21941097e-01 6.23495221e-01 2.58445382e-01 -5.61774056e-03 5.29068410e-01 -5.87529182e-01 3.30913126e-01 -2.78246045e-01 1.00287639e-01 -2.21509993e-01 3.13332289e-01 -1.52193561e-01 6.50858700e-01 3.75797987e-01 1.04242861e+00 -1.00505698e+00 -2.45799392e-01 1.36124730e-01 7.78644919e-01 -1.86537519e-01 2.07305863e-01 -5.43484569e-01 3.47932130e-01 -7.02144563e-01 6.13191605e-01 2.72061616e-01 -9.18498278e-01 7.47605205e-01 -2.51974732e-01 2.43756264e-01 3.94103974e-01 -2.76081592e-01 1.45861042e+00 -2.16470778e-01 7.83310711e-01 -2.07686290e-01 -1.61475217e+00 9.80830789e-01 6.25271320e-01 8.73272896e-01 -2.41902411e-01 -2.22807080e-01 -4.20114920e-02 3.19661111e-01 -2.87137181e-01 -9.00012404e-02 -9.73664820e-02 1.64095968e-01 2.92142421e-01 2.81928360e-01 3.51280540e-01 9.45379883e-02 9.20962989e-02 1.48257852e+00 2.46670526e-02 7.60423779e-01 -7.83287287e-02 5.47235429e-01 -1.27049610e-01 6.51987255e-01 3.23566824e-01 6.42535463e-02 3.55453610e-01 6.47350073e-01 -7.96592116e-01 -9.68662977e-01 -1.05778468e+00 -3.09460580e-01 1.20234692e+00 -2.33334675e-01 -4.61608201e-01 -2.66891301e-01 -3.83154064e-01 2.35945329e-01 6.75182790e-02 -8.26765299e-01 -4.05849874e-01 -3.15775722e-01 -9.88872290e-01 4.15124655e-01 3.61189991e-01 -9.44237784e-02 -1.31189787e+00 -1.31903589e-01 3.16346288e-01 8.44550412e-03 -8.59733999e-01 -1.48455858e-01 7.20482647e-01 -9.53861535e-01 -1.31215596e+00 -6.87151074e-01 -9.38467085e-01 8.80061328e-01 6.16567172e-02 1.23876786e+00 3.67770970e-01 -7.69814432e-01 -2.17550405e-05 -1.55537307e-01 -3.32853168e-01 -5.10769844e-01 -7.40491822e-02 6.35483190e-02 -2.17300728e-02 3.61390501e-01 -8.33268881e-01 -6.33717716e-01 -4.64675277e-02 -8.71994793e-01 2.29116231e-01 8.76308739e-01 1.11230099e+00 7.99277008e-01 -3.44484419e-01 6.60374165e-01 -1.24502099e+00 5.97322583e-01 -8.39267254e-01 -1.66205928e-01 6.43939734e-01 -4.15509403e-01 4.86915350e-01 5.08927882e-01 -3.99572343e-01 -7.25460410e-01 4.49698478e-01 5.51289469e-02 -7.67968502e-03 -3.58121902e-01 8.96671832e-01 1.09049290e-01 6.79674372e-02 8.40514660e-01 5.89985371e-01 1.48243397e-01 -4.43572789e-01 3.50264281e-01 5.52642465e-01 1.73148289e-01 -7.73374975e-01 2.54998058e-01 2.54094392e-01 3.83300811e-01 -1.01064765e+00 -8.44956100e-01 -5.18660486e-01 -8.69075716e-01 1.44086957e-01 9.86011565e-01 -6.99689925e-01 -9.60676908e-01 -8.54332931e-03 -1.28100419e+00 1.20767742e-01 2.03531727e-01 3.39139998e-01 -5.40449262e-01 3.35730791e-01 -8.40641379e-01 -6.41284347e-01 -3.25740069e-01 -9.34723794e-01 1.09940338e+00 1.69131801e-01 -4.56336081e-01 -1.25202787e+00 2.85997033e-01 2.42587626e-01 1.80678442e-01 6.66517079e-01 1.62740958e+00 -8.40025425e-01 -4.62431222e-01 -1.40106872e-01 -1.93713069e-01 -8.92926976e-02 5.60126066e-01 1.01375610e-01 -7.03502774e-01 -1.60459220e-01 -6.86272323e-01 -3.84983808e-01 1.08043635e+00 3.62904817e-01 1.41079915e+00 -4.05179709e-01 -8.56089175e-01 4.11192596e-01 1.12743986e+00 3.43688011e-01 5.21349669e-01 -5.26491404e-02 4.69508499e-01 7.50470996e-01 -2.25701723e-02 5.95671274e-02 -1.06553644e-01 4.98483509e-01 2.62797713e-01 -2.32007533e-01 -4.04769294e-02 -1.37531623e-01 1.26433000e-01 5.72436750e-01 -2.39829287e-01 -6.97176531e-02 -8.85405481e-01 4.00311828e-01 -2.13039064e+00 -1.07187223e+00 -5.14966473e-02 1.92713642e+00 1.04205000e+00 -5.27345017e-02 -1.63945809e-01 -3.58363718e-01 8.23757470e-01 -2.04768050e-02 -9.78527009e-01 -3.70029539e-01 -1.64651453e-01 3.81204575e-01 2.40205914e-01 2.11326659e-01 -7.94030488e-01 7.97157705e-01 7.42339945e+00 3.05677176e-01 -1.08372879e+00 -3.85836929e-01 7.70323455e-01 3.04084748e-01 -3.24283510e-01 -8.20914954e-02 -3.77823889e-01 3.20897877e-01 1.17008388e+00 -2.79605418e-01 3.50260079e-01 7.80495465e-01 -1.29965410e-01 2.97198147e-01 -1.30437672e+00 8.66424382e-01 -1.89108580e-01 -1.82541728e+00 2.83265084e-01 2.41773158e-01 2.82762975e-01 3.82762216e-02 -5.00383005e-02 1.29600525e-01 5.80485582e-01 -1.54873502e+00 -1.31681502e-01 8.86141956e-01 6.54054761e-01 -5.70353746e-01 5.15652895e-01 3.41101915e-01 -1.07857299e+00 2.66149044e-01 -5.83485544e-01 -2.30359975e-02 -1.88690066e-01 6.52291596e-01 -1.05984938e+00 3.69501382e-01 4.90401804e-01 1.05454409e+00 -2.90337205e-01 7.01136589e-01 -2.05591455e-01 6.90937579e-01 1.32556364e-01 -2.18815640e-01 -3.39494646e-02 -1.39525294e-01 1.46921113e-01 1.31096637e+00 -3.97031009e-02 2.14250013e-01 2.95035869e-01 7.95788109e-01 -3.57391745e-01 8.19776207e-02 -8.65964115e-01 -6.46008551e-01 1.97799578e-01 1.37358630e+00 -6.02331519e-01 -4.27713752e-01 -4.78836566e-01 9.06065702e-01 8.22680950e-01 5.68583965e-01 -5.07954299e-01 -2.74029076e-01 1.47731507e+00 -3.92729044e-02 9.68823284e-02 -2.10354224e-01 2.97655672e-01 -1.05324507e+00 -4.72422242e-01 -6.67068005e-01 2.03705698e-01 -7.21739233e-01 -1.58391500e+00 3.24466735e-01 -4.14702982e-01 -6.89343989e-01 -2.29550242e-01 -1.14004397e+00 -3.06638360e-01 8.64974618e-01 -1.17268586e+00 -9.48703468e-01 2.76969463e-01 1.55690923e-01 3.60706568e-01 -2.18317181e-01 1.55197215e+00 -6.51861578e-02 -4.60246325e-01 6.25333861e-02 4.86369282e-01 1.46752745e-01 5.11178851e-01 -1.42084849e+00 4.90874499e-01 7.28335558e-03 3.66933465e-01 1.15414500e+00 6.99811220e-01 -4.15798157e-01 -1.38863301e+00 -8.55557144e-01 9.84936416e-01 -3.34496558e-01 5.75588942e-01 -5.91958284e-01 -1.27724254e+00 7.44089842e-01 -1.56398129e-03 2.82362223e-01 1.34568346e+00 5.86088836e-01 -6.18465722e-01 2.04579115e-01 -1.05756104e+00 5.65944970e-01 8.81412208e-01 -7.99012840e-01 -5.81737161e-01 7.48893976e-01 8.46900880e-01 -6.92309737e-02 -1.31229174e+00 3.15029025e-01 7.40251660e-01 -4.42880303e-01 1.19941640e+00 -1.40354085e+00 4.72624093e-01 -3.76296580e-01 -2.63774633e-01 -1.35087597e+00 -7.79558599e-01 -3.17309737e-01 -4.66659307e-01 5.85019886e-01 8.53303015e-01 -6.51902735e-01 8.13340008e-01 5.06456614e-01 -9.45099071e-02 -1.29692090e+00 -7.22715914e-01 -2.57003009e-01 1.07962266e-01 1.98066130e-01 4.08440799e-01 9.81861830e-01 4.30501878e-01 6.79049671e-01 -1.44520849e-01 -1.44742191e-01 4.03995603e-01 3.79817158e-01 3.89603257e-01 -1.59102571e+00 -4.51875448e-01 -3.71758521e-01 -8.71575415e-01 -9.98769760e-01 7.46428251e-01 -1.27219129e+00 1.13282725e-02 -1.76821935e+00 6.15575612e-01 8.40068534e-02 -8.82363975e-01 5.79416335e-01 -1.77600801e-01 -8.91986042e-02 -3.40777844e-01 3.27089995e-01 -5.50648153e-01 3.01820874e-01 1.11861980e+00 -5.10009289e-01 2.77164340e-01 -1.62723482e-01 -8.44020605e-01 6.27140284e-01 7.58092999e-01 -3.78156126e-01 -1.32171050e-01 -1.93176102e-02 2.65137583e-01 1.35901377e-01 1.19857445e-01 -5.59199929e-01 1.52807221e-01 -3.25398654e-01 8.19163084e-01 -2.06553325e-01 6.77676857e-01 -5.64047933e-01 1.39085069e-01 8.19885910e-01 -6.73311293e-01 -1.94789208e-02 -1.64549679e-01 1.06613779e+00 -2.13771269e-01 6.15534000e-02 6.92422748e-01 -4.30147052e-01 -5.25947988e-01 3.63530427e-01 -8.72472405e-01 -4.49313730e-01 9.16735888e-01 -4.75514010e-02 -5.15745163e-01 -3.26611370e-01 -9.61133182e-01 -3.18633243e-02 2.23752663e-01 3.35050881e-01 7.94681907e-01 -1.09288323e+00 -4.68870342e-01 -1.30184352e-01 2.84785986e-01 -3.74240637e-01 -7.74762705e-02 4.35368270e-01 -4.44246948e-01 7.54824936e-01 -2.58116156e-01 -5.01493931e-01 -1.41927969e+00 5.10588765e-01 4.09542710e-01 -4.43737477e-01 -4.40002531e-01 8.22891593e-01 6.55542374e-01 -5.39884448e-01 -4.11519222e-02 -1.58308610e-01 -4.06893373e-01 -3.03753704e-01 3.85976255e-01 -2.54016787e-01 -2.63454258e-01 -7.63635635e-01 -4.37623739e-01 3.17838073e-01 -5.90126574e-01 4.09574330e-01 1.52011156e+00 3.85477006e-01 -5.91852605e-01 7.94077754e-01 1.32505417e+00 -5.46280682e-01 -8.92240822e-01 -2.88160503e-01 3.76168549e-01 -7.30107278e-02 -3.99084657e-01 -8.33920896e-01 -5.90655804e-01 6.06529593e-01 2.72850841e-01 8.55303109e-02 6.52770638e-01 4.35986280e-01 6.53359115e-01 9.34671283e-01 3.22674811e-01 -5.10020137e-01 1.49193155e-02 5.85913837e-01 6.52747452e-01 -1.18409765e+00 1.87266082e-01 -4.43974912e-01 -1.43779382e-01 1.23424351e+00 4.55841243e-01 -1.31296396e-01 8.53705764e-01 1.63624391e-01 -3.73490632e-01 -4.14401591e-01 -1.34668839e+00 -7.18230084e-02 4.68991697e-01 7.38756895e-01 1.55743742e+00 3.20227772e-01 -3.10314924e-01 1.90338448e-01 1.77895948e-01 -3.48100215e-01 1.56336665e-01 8.98973048e-01 -7.05987811e-01 -1.46043110e+00 2.15449452e-01 8.38864148e-01 -7.92786896e-01 -2.24640921e-01 -8.62984478e-01 2.62701094e-01 -2.95401275e-01 5.93590677e-01 -3.23559926e-03 -2.31589600e-01 -1.28579900e-01 4.72868711e-01 5.06164730e-01 -9.95365441e-01 -3.79145026e-01 -4.76327568e-01 -5.81457317e-02 -5.09596407e-01 -5.34539044e-01 -4.65205342e-01 -1.79220879e+00 -1.27124384e-01 -2.04610139e-01 4.16816503e-01 7.06923008e-01 9.30032253e-01 7.15579033e-01 8.06140244e-01 2.59541571e-01 -4.67072248e-01 -9.08569470e-02 -1.00799656e+00 -5.47735214e-01 4.85965341e-01 4.24966753e-01 -6.73092961e-01 -1.14148118e-01 4.37150151e-01]
[4.973657608032227, 5.686740875244141]
d55e39d4-7941-4805-8f78-892259baa28e
adaptively-accumulated-knowledge-transfer-for
2008.11873
null
https://arxiv.org/abs/2008.11873v1
https://arxiv.org/pdf/2008.11873v1.pdf
Adaptively-Accumulated Knowledge Transfer for Partial Domain Adaptation
Partial domain adaptation (PDA) attracts appealing attention as it deals with a realistic and challenging problem when the source domain label space substitutes the target domain. Most conventional domain adaptation (DA) efforts concentrate on learning domain-invariant features to mitigate the distribution disparity across domains. However, it is crucial to alleviate the negative influence caused by the irrelevant source domain categories explicitly for PDA. In this work, we propose an Adaptively-Accumulated Knowledge Transfer framework (A$^2$KT) to align the relevant categories across two domains for effective domain adaptation. Specifically, an adaptively-accumulated mechanism is explored to gradually filter out the most confident target samples and their corresponding source categories, promoting positive transfer with more knowledge across two domains. Moreover, a dual distinct classifier architecture consisting of a prototype classifier and a multilayer perceptron classifier is built to capture intrinsic data distribution knowledge across domains from various perspectives. By maximizing the inter-class center-wise discrepancy and minimizing the intra-class sample-wise compactness, the proposed model is able to obtain more domain-invariant and task-specific discriminative representations of the shared categories data. Comprehensive experiments on several partial domain adaptation benchmarks demonstrate the effectiveness of our proposed model, compared with the state-of-the-art PDA methods.
['Haifeng Xia', 'Zhengming Ding', 'Taotao Jing']
2020-08-27
null
null
null
null
['partial-domain-adaptation']
['methodology']
[ 2.19145745e-01 -1.07675165e-01 -4.80116218e-01 -6.30011320e-01 -6.38922036e-01 -4.24636275e-01 3.94066125e-01 1.84827924e-01 -3.97113353e-01 7.48951614e-01 6.92505911e-02 1.49986714e-01 -4.89182830e-01 -6.78314567e-01 -5.34176648e-01 -8.82034779e-01 3.49668682e-01 4.58578467e-01 4.54692125e-01 -1.58891827e-01 2.02077135e-01 2.75781929e-01 -1.33143508e+00 2.13695914e-01 1.34323525e+00 1.18185186e+00 4.98674452e-01 -1.39665395e-01 -3.53085041e-01 3.35295528e-01 -4.29523081e-01 -2.60897845e-01 3.13312382e-01 -3.53719145e-01 -5.10583937e-01 2.00442091e-01 3.21884125e-01 3.02716102e-02 -2.40022130e-02 1.22155368e+00 3.95617694e-01 2.71847963e-01 1.04033673e+00 -1.13469183e+00 -8.80639732e-01 3.07788610e-01 -8.31991255e-01 3.07412416e-01 -1.36727795e-01 2.41552945e-02 7.93456197e-01 -9.53631938e-01 3.74184847e-01 1.13874686e+00 3.64854634e-01 5.19541621e-01 -1.29350090e+00 -8.07851732e-01 6.12416327e-01 4.32844579e-01 -1.42135000e+00 -1.45722851e-01 1.28269160e+00 -4.94795531e-01 4.40455109e-01 -3.21542360e-02 2.83375770e-01 1.01066458e+00 -9.37059075e-02 8.50501120e-01 9.19615507e-01 -4.62415010e-01 3.99349302e-01 8.60425234e-01 2.86261410e-01 1.01802826e-01 1.60654426e-01 -1.12883233e-01 -3.07387084e-01 -1.34136304e-01 6.17275655e-01 1.52755305e-01 -2.10188597e-01 -1.10496891e+00 -1.05927026e+00 8.22288811e-01 5.16372979e-01 3.33140492e-01 -4.72827047e-01 -6.22309387e-01 5.26276290e-01 2.36299634e-01 5.33959031e-01 3.31648499e-01 -7.99792111e-01 1.50760949e-01 -6.10760450e-01 1.05322950e-01 3.46231371e-01 1.10362482e+00 8.63842189e-01 -9.71089229e-02 -1.90076768e-01 1.38285315e+00 3.26630384e-01 3.11018795e-01 7.85614014e-01 -5.41171372e-01 6.53647244e-01 9.78131771e-01 3.30910273e-02 -1.02049839e+00 -1.35389835e-01 -7.36925364e-01 -9.32818651e-01 -7.14435205e-02 4.12284434e-01 -1.04620485e-02 -5.78198552e-01 1.88368726e+00 5.85351884e-01 1.62817284e-01 1.41399547e-01 9.80541408e-01 3.69648039e-01 5.31365812e-01 5.03823698e-01 -2.59945422e-01 1.27351344e+00 -9.27052200e-01 -4.13853824e-01 -5.83884537e-01 4.15056199e-01 -5.30758917e-01 1.14568257e+00 1.80475190e-01 -7.27724850e-01 -8.30970109e-01 -1.04666400e+00 2.03521103e-01 -3.45864594e-01 3.10491413e-01 2.24012777e-01 4.33333248e-01 -4.06127125e-01 3.40872854e-01 -3.75009865e-01 -3.28409046e-01 6.92839265e-01 3.49594057e-01 -3.29703271e-01 -2.19947189e-01 -1.27158928e+00 7.47271776e-01 7.40599096e-01 -3.13721597e-01 -5.40350735e-01 -9.02668655e-01 -7.58494258e-01 5.31798154e-02 2.13316381e-01 -5.36957681e-01 9.22612607e-01 -1.37345791e+00 -1.44861281e+00 8.22737098e-01 -7.32952952e-02 -2.78273582e-01 4.16934282e-01 -1.49111852e-01 -6.17982090e-01 -4.50539924e-02 1.23541392e-01 6.54818237e-01 1.02504480e+00 -1.23029363e+00 -1.00884867e+00 -5.92213154e-01 -3.42550904e-01 6.27881885e-01 -1.03355742e+00 -3.12795758e-01 -2.84713537e-01 -8.01559865e-01 1.27932087e-01 -6.66423798e-01 -4.22866531e-02 -4.04822752e-02 5.33214249e-02 -3.30507517e-01 8.85374367e-01 -5.72400451e-01 1.28852522e+00 -2.39975166e+00 3.19888622e-01 1.11859187e-01 1.73133463e-02 4.69963819e-01 -2.30538487e-01 -9.44517925e-02 -1.53765947e-01 -4.68077213e-01 -4.05649364e-01 8.84872861e-04 -9.19986293e-02 1.65321693e-01 -3.76913726e-01 3.35451275e-01 4.48574215e-01 4.29621279e-01 -8.18158984e-01 -5.10508239e-01 2.30956972e-01 1.74494371e-01 -4.58612949e-01 3.20104510e-01 -1.26474231e-01 5.39113760e-01 -7.62409389e-01 6.41226888e-01 1.00552440e+00 -2.12786481e-01 2.68769979e-01 -1.55187652e-01 2.73116946e-01 -2.70869546e-02 -1.22086060e+00 1.64099193e+00 -3.69419187e-01 1.52350217e-01 8.48996788e-02 -1.47291732e+00 1.49341607e+00 -8.79353583e-02 4.50730622e-01 -9.06062722e-01 6.75846860e-02 3.36029530e-01 -1.23851120e-01 -2.14722231e-01 2.62910426e-01 -3.08628887e-01 -1.06873989e-01 2.27715783e-02 1.28128290e-01 1.75477788e-01 -2.74125993e-01 -2.79869527e-01 6.54412448e-01 5.44393919e-02 5.72926879e-01 -4.77209091e-01 9.56936002e-01 9.20336470e-02 8.69681180e-01 2.42196843e-01 -6.99680388e-01 3.55613112e-01 5.28362840e-02 -3.37276608e-01 -9.57768619e-01 -1.20353961e+00 -2.58346468e-01 1.41726971e+00 6.09246433e-01 2.88921297e-01 -5.48130989e-01 -1.04747379e+00 2.84344107e-01 8.15716207e-01 -4.75535065e-01 -6.88625813e-01 -6.02623641e-01 -5.98035216e-01 1.20254882e-01 7.20064104e-01 6.29356265e-01 -9.08694327e-01 -2.12740645e-01 3.14164251e-01 -7.46130943e-02 -7.73245573e-01 -5.89065373e-01 3.23486358e-01 -1.06101108e+00 -8.90102804e-01 -1.10727215e+00 -1.22803843e+00 7.59670317e-01 4.39182103e-01 8.74970794e-01 -6.97629392e-01 1.63553774e-01 1.26720920e-01 -3.60760868e-01 -2.60861397e-01 -2.34919906e-01 1.90173775e-01 2.04184562e-01 2.40507618e-01 9.81308043e-01 -6.59118116e-01 -6.25685453e-01 6.99874580e-01 -7.11968958e-01 -1.87617496e-01 7.84692228e-01 1.12123442e+00 7.35559225e-01 1.86234951e-01 1.02661371e+00 -7.27804303e-01 6.93172038e-01 -8.70907485e-01 -4.05834973e-01 3.31179321e-01 -6.11763179e-01 2.88952840e-03 1.00096214e+00 -8.58599365e-01 -1.41041172e+00 1.91125777e-02 3.66345763e-01 -5.38175762e-01 -3.76323730e-01 2.15985104e-01 -6.53746367e-01 1.82656839e-01 8.69005620e-01 5.66481769e-01 2.01388244e-02 -5.31815350e-01 3.31189036e-01 7.92595685e-01 4.17247564e-01 -7.91894436e-01 6.69429421e-01 3.63126934e-01 -4.42729175e-01 -5.91570675e-01 -7.35742211e-01 -7.26657927e-01 -8.83825481e-01 2.02508435e-01 4.99089509e-01 -1.05424523e+00 -5.73028028e-02 4.80052501e-01 -8.07902753e-01 -1.22635014e-01 -4.61702704e-01 5.83121419e-01 -4.35379893e-01 5.10089815e-01 -1.11659348e-01 -5.25575817e-01 -2.29637265e-01 -9.60751712e-01 8.13789725e-01 4.68524873e-01 -1.15938336e-01 -1.03244758e+00 1.14504516e-01 2.12778747e-01 3.64979744e-01 -1.55531213e-01 1.10523391e+00 -1.13655591e+00 -9.24866423e-02 -8.60404968e-03 -4.90260988e-01 7.39414573e-01 5.49276233e-01 -5.76468468e-01 -7.79996336e-01 -4.08337057e-01 4.81959842e-02 -2.01473489e-01 7.77163208e-01 4.02205646e-01 1.12554038e+00 -2.77806401e-01 -6.07936621e-01 4.03532356e-01 1.27903759e+00 4.08861279e-01 2.48755157e-01 5.17938375e-01 5.62197745e-01 7.13551939e-01 1.05309498e+00 5.41538477e-01 3.39739263e-01 7.42001593e-01 1.07875034e-01 1.90811247e-01 4.43275869e-02 -2.39664599e-01 2.51477480e-01 6.34592175e-01 4.67246860e-01 8.36535171e-02 -7.57732451e-01 8.43322635e-01 -1.87186575e+00 -6.22895300e-01 3.29078645e-01 2.24620318e+00 9.96441722e-01 1.79937676e-01 2.77571321e-01 -7.15552419e-02 1.14811456e+00 -1.10351458e-01 -9.53121305e-01 -9.18412879e-02 -7.52415061e-02 -4.40316275e-02 3.18973899e-01 -6.61150040e-03 -1.41454077e+00 7.42567658e-01 5.24353552e+00 1.34595656e+00 -1.05286205e+00 -4.83919233e-02 5.08959353e-01 1.24731213e-01 -1.90934807e-01 -3.08375061e-01 -9.06717002e-01 6.56464756e-01 5.47430336e-01 -3.65505040e-01 8.44158530e-02 1.44673002e+00 -9.01234597e-02 2.29196385e-01 -9.90151882e-01 8.88065934e-01 -1.25758350e-01 -9.12058234e-01 1.44361585e-01 -1.82502586e-02 6.53468132e-01 -3.37527364e-01 2.57429659e-01 5.86380243e-01 1.72557905e-01 -4.88308340e-01 5.69530010e-01 3.80304128e-01 7.47953534e-01 -8.25117350e-01 6.22185349e-01 4.43034321e-01 -1.21099079e+00 -4.50215995e-01 -6.89063370e-01 1.00928001e-01 -2.12880120e-01 6.22202456e-01 -9.07708049e-01 5.93901098e-01 6.69602990e-01 9.55965698e-01 -4.94234860e-01 9.02370334e-01 1.02308162e-01 4.47945356e-01 -1.02169849e-01 7.72215202e-02 9.04123411e-02 -3.67715180e-01 5.60090065e-01 1.12653172e+00 3.12670410e-01 4.64457236e-02 1.84815124e-01 6.98188722e-01 7.39219934e-02 3.10592622e-01 -5.43251157e-01 3.06961000e-01 8.36729705e-01 9.64622736e-01 -3.31589937e-01 -3.53511989e-01 -4.35723543e-01 9.62632418e-01 5.72394907e-01 2.62882143e-01 -7.10069835e-01 -3.39089096e-01 8.94432485e-01 1.21623166e-01 4.92517322e-01 -3.74818896e-03 -5.83765984e-01 -1.05892706e+00 1.55536175e-01 -8.18783700e-01 8.94704938e-01 -2.49835491e-01 -1.91715920e+00 3.19663286e-01 1.03119373e-01 -1.63186097e+00 5.53289428e-02 -5.58687210e-01 -5.25894344e-01 9.64940667e-01 -1.80352986e+00 -1.05686319e+00 -3.62881571e-01 9.27757621e-01 6.76543832e-01 -4.96141672e-01 6.07565880e-01 4.79882687e-01 -6.02493346e-01 8.96991968e-01 6.08416677e-01 -4.80816402e-02 1.15550530e+00 -1.08787763e+00 -9.98390391e-02 5.98551691e-01 -4.19953585e-01 7.10424721e-01 4.67705280e-01 -5.26949704e-01 -9.21199322e-01 -1.38883805e+00 4.06286269e-01 -1.32089764e-01 4.66150969e-01 -7.87540898e-02 -1.44874358e+00 4.28008467e-01 -1.86476216e-01 3.68829109e-02 7.33302474e-01 4.17922176e-02 -5.41188776e-01 -6.80588961e-01 -1.33585691e+00 3.50277036e-01 9.18243885e-01 -2.56073058e-01 -9.16931868e-01 8.38737339e-02 6.97314501e-01 -8.02288279e-02 -7.90210426e-01 4.94576633e-01 3.97248358e-01 -6.55255497e-01 1.04231405e+00 -5.98256707e-01 2.87463963e-01 -3.58319312e-01 -2.77433455e-01 -1.50928199e+00 -6.69777155e-01 1.01844326e-01 -1.30785242e-01 1.42725241e+00 8.06325004e-02 -7.28918493e-01 8.53661597e-01 4.50681210e-01 -1.64144859e-01 -5.14051199e-01 -9.46565449e-01 -9.65820074e-01 3.72326404e-01 -7.31776133e-02 5.54723024e-01 1.26381302e+00 2.30163634e-02 3.50780964e-01 -1.70949295e-01 1.90984845e-01 6.14221036e-01 2.32440293e-01 6.59157276e-01 -1.37395036e+00 -1.80919543e-01 -5.52060664e-01 -3.18618715e-01 -1.20863903e+00 2.18324929e-01 -8.11572075e-01 -6.87090540e-03 -1.05572855e+00 2.46370614e-01 -7.52034962e-01 -8.64049494e-01 4.03637648e-01 -3.40361208e-01 -2.15840951e-01 -1.03989281e-01 4.74363446e-01 -6.50996923e-01 9.56460416e-01 1.29645479e+00 -1.89508155e-01 -4.64573830e-01 4.50052992e-02 -1.05007350e+00 6.30108297e-01 7.92847812e-01 -3.82972211e-01 -6.55245602e-01 -2.23270223e-01 -3.99216563e-01 -3.32587510e-01 2.42345482e-01 -9.91231799e-01 2.29351744e-01 -5.13592720e-01 5.08807421e-01 -4.24553871e-01 1.66841015e-01 -1.04009938e+00 -1.31474391e-01 2.87724942e-01 -3.50007117e-01 -5.60164511e-01 2.82341123e-01 8.82670820e-01 -4.27375078e-01 3.29572596e-02 1.31233358e+00 1.31580457e-01 -1.10912764e+00 2.09758401e-01 1.36442006e-01 5.59117272e-02 1.34952509e+00 -3.58846754e-01 -2.43329525e-01 1.44842550e-01 -5.05034804e-01 3.69530320e-01 4.40797091e-01 6.42665982e-01 5.41760743e-01 -1.53721511e+00 -6.43897116e-01 2.82151550e-01 5.59754372e-01 5.27434126e-02 5.42086065e-01 5.42837560e-01 1.24256872e-01 4.45952266e-01 -5.91471732e-01 -7.14561820e-01 -9.79513407e-01 8.15759361e-01 2.85472900e-01 -3.14348310e-01 -4.01354432e-01 9.01181340e-01 8.22345376e-01 -7.05507636e-01 1.28201857e-01 -5.32302745e-02 -4.40885603e-01 1.55166045e-01 4.83177394e-01 3.64067376e-01 -7.84108415e-03 -4.21516031e-01 -6.03493929e-01 5.66162229e-01 -4.39307809e-01 3.72821003e-01 1.24761033e+00 -3.47899318e-01 2.64579266e-01 1.80830747e-01 1.15829384e+00 -2.71812230e-01 -1.67937398e+00 -8.08768690e-01 9.53822136e-02 -5.39180934e-01 -2.86553472e-01 -8.45822811e-01 -7.75827408e-01 8.10787618e-01 7.69758940e-01 -1.43383160e-01 1.49251044e+00 -8.82439874e-03 6.69004619e-01 3.42300422e-02 3.08113158e-01 -1.36987531e+00 2.92273045e-01 3.71964782e-01 7.02794075e-01 -1.31499064e+00 -1.62873238e-01 -4.01591778e-01 -9.75446105e-01 9.46518719e-01 1.00410390e+00 -1.86915100e-01 5.76073349e-01 -3.78685951e-01 -1.73024058e-01 2.27524459e-01 -4.58814651e-01 1.39302805e-01 4.72720057e-01 9.85180676e-01 1.01886488e-01 2.04468638e-01 -3.40834439e-01 1.19719303e+00 2.65041530e-01 -8.34260508e-02 -9.51675847e-02 7.61260808e-01 -6.37126386e-01 -1.12018180e+00 -3.97892952e-01 3.30390900e-01 3.14622261e-02 1.91408545e-01 -2.29854912e-01 7.43468761e-01 2.45965689e-01 6.29995883e-01 2.37815809e-02 -3.66614580e-01 4.93346125e-01 1.84555337e-01 2.35462368e-01 -5.82002819e-01 -1.80345401e-01 -3.35006253e-03 -4.21218187e-01 -1.65210471e-01 -2.54636616e-01 -6.42600775e-01 -1.00492907e+00 5.90738617e-02 -3.11943948e-01 1.21964805e-01 2.84991145e-01 8.39179218e-01 6.98306441e-01 4.13390249e-01 8.30762506e-01 -4.59404707e-01 -9.32135224e-01 -9.57822859e-01 -8.25270593e-01 5.43898821e-01 1.10413700e-01 -1.02585137e+00 -2.70732760e-01 3.73050645e-02]
[10.348881721496582, 3.0766618251800537]
07122283-9fea-4b7b-be35-9985ef8aa3f7
combination-of-single-and-multi-frame-image
2303.03212
null
https://arxiv.org/abs/2303.03212v1
https://arxiv.org/pdf/2303.03212v1.pdf
Combination of Single and Multi-frame Image Super-resolution: An Analytical Perspective
Super-resolution is the process of obtaining a high-resolution image from one or more low-resolution images. Single image super-resolution (SISR) and multi-frame super-resolution (MFSR) methods have been evolved almost independently for years. A neglected study in this field is the theoretical analysis of finding the optimum combination of SISR and MFSR. To fill this gap, we propose a novel theoretical analysis based on the iterative shrinkage and thresholding algorithm. We implement and compare several approaches for combining SISR and MFSR, and simulation results support the finding of our theoretical analysis, both quantitatively and qualitatively.
['Aliazam Abbasfar', 'Reshad Hosseini', 'Mohammad Mahdi Afrasiabi']
2023-03-06
null
null
null
null
['multi-frame-super-resolution']
['computer-vision']
[ 6.83610439e-01 -2.09321350e-01 -2.92047765e-02 -2.54878908e-01 -9.36654270e-01 -7.26084262e-02 2.97963738e-01 -5.10779560e-01 -1.74261987e-01 9.67755079e-01 3.21843535e-01 3.19715559e-01 -1.44983456e-01 -3.99337947e-01 -2.13202059e-01 -7.04099953e-01 4.52999808e-02 -1.53378975e-02 7.05571651e-01 -2.74935126e-01 4.22840893e-01 4.81791586e-01 -1.53743947e+00 3.48915279e-01 7.84260631e-01 6.03947699e-01 6.83172107e-01 6.97642744e-01 2.20823050e-01 7.12063253e-01 -2.84358948e-01 1.11252755e-01 1.72816679e-01 -7.68000245e-01 -9.33984876e-01 7.36382231e-02 2.98337579e-01 -4.05132681e-01 -1.17663831e-01 1.19856155e+00 5.83924711e-01 1.68057755e-01 1.88012466e-01 -4.03402358e-01 -8.30700994e-01 4.53877002e-01 -1.15119004e+00 1.01590943e+00 7.43418574e-01 -4.45127308e-01 4.23022538e-01 -1.02317202e+00 9.17493165e-01 1.23466039e+00 4.91170168e-01 5.78313947e-01 -1.30068672e+00 -3.55949312e-01 -2.70635992e-01 -1.52609386e-02 -1.70469487e+00 -6.10197604e-01 6.78690910e-01 -1.83204532e-01 5.93492687e-01 2.52976447e-01 2.38196179e-01 6.10363603e-01 1.61284149e-01 2.83669919e-01 1.84179246e+00 -6.75674081e-01 1.31912157e-02 2.90561952e-02 4.37732562e-02 4.66420829e-01 4.41556633e-01 2.76109904e-01 -8.09531450e-01 -7.98831508e-02 1.83416045e+00 -3.52903485e-01 -3.70788336e-01 1.26541585e-01 -1.09669065e+00 5.38160861e-01 -3.74663398e-02 7.80254900e-01 -4.27765757e-01 -2.10720524e-01 -5.64279892e-02 2.11314276e-01 7.37679303e-01 2.54965931e-01 1.82440132e-01 3.37559842e-02 -1.39181149e+00 -7.51285702e-02 1.14214756e-01 5.85886121e-01 6.28789544e-01 2.37812191e-01 1.69952154e-01 8.61331165e-01 -7.42044821e-02 3.38845581e-01 5.96491039e-01 -1.57036471e+00 -1.08734213e-01 -7.47888982e-02 5.18191040e-01 -8.61308455e-01 -2.29341701e-01 -9.42231566e-02 -8.80767941e-01 4.03594553e-01 2.80224741e-01 1.10836454e-01 -7.11452067e-01 1.30770731e+00 4.31999326e-01 4.22951996e-01 1.30188167e-01 1.40800333e+00 8.68909895e-01 6.48764074e-01 -3.24005872e-01 -1.00409114e+00 1.22467971e+00 -5.81165850e-01 -1.13736844e+00 2.24002618e-02 -5.37487678e-02 -8.97678435e-01 7.08099902e-01 2.30010286e-01 -1.56521797e+00 -5.83904803e-01 -9.23829079e-01 -3.11977625e-01 2.47532725e-01 -3.84259410e-02 2.93577343e-01 1.35881543e-01 -1.27006519e+00 7.59302378e-01 -7.10271120e-01 -2.56939650e-01 1.68335363e-01 1.13067046e-01 -4.92857665e-01 -8.84818435e-02 -1.15095687e+00 1.05523026e+00 1.49332657e-01 -1.34704039e-01 -4.53379422e-01 -6.67435646e-01 -4.32937860e-01 -2.80203193e-01 2.94942737e-01 -6.13183200e-01 1.12040865e+00 -8.87345195e-01 -1.55392182e+00 1.11694455e+00 -7.99559832e-01 -4.07729983e-01 7.33412653e-02 -1.01244099e-01 -4.90363836e-01 8.15121055e-01 9.89967659e-02 1.48113221e-01 1.04843748e+00 -1.65580571e+00 -7.19772220e-01 -4.47907537e-01 -1.89108580e-01 2.50643760e-01 3.34641516e-01 7.74749041e-01 8.05105865e-02 -4.97832924e-01 5.59163988e-01 -3.23562175e-01 -5.33355594e-01 -4.25749391e-01 1.03887632e-01 2.86182761e-01 7.50913441e-01 -1.00427079e+00 1.38808692e+00 -1.97587478e+00 2.76993006e-01 -2.73571134e-01 5.05734384e-01 2.35907137e-01 1.65856048e-01 4.85207587e-02 -1.64735317e-01 2.53749013e-01 -1.63777798e-01 -2.15092331e-01 -9.60706949e-01 -1.42548069e-01 -3.52858275e-01 6.79212034e-01 -8.16292390e-02 7.05050468e-01 -8.07064235e-01 -9.01059687e-01 3.59817624e-01 9.82632399e-01 1.06520273e-01 3.72280665e-02 4.46535707e-01 9.92607117e-01 -4.54109251e-01 6.95664287e-01 8.26561332e-01 -4.51013416e-01 2.46710390e-01 -4.85307932e-01 -7.82672346e-01 -1.53755039e-01 -1.27932286e+00 1.28322184e+00 -2.36219764e-01 5.88307023e-01 4.22130555e-01 -7.05839336e-01 1.01521921e+00 4.98409122e-01 8.45198095e-01 -9.02414083e-01 -1.03994958e-01 3.36251527e-01 -5.21267354e-01 -4.42283273e-01 7.50660002e-01 -7.37458348e-01 2.84059972e-01 3.74041498e-01 -9.70966294e-02 1.99832544e-01 1.33176416e-01 1.60743862e-01 7.66499639e-01 9.44464654e-02 8.03679585e-01 -4.19867098e-01 7.10009158e-01 -1.56299472e-01 5.06099045e-01 6.34141743e-01 -4.14408624e-01 1.00605154e+00 -1.84474781e-01 -3.91213328e-01 -1.72883487e+00 -9.44109619e-01 -3.89864326e-01 6.78281009e-01 5.65816522e-01 -1.85515836e-01 -8.01020741e-01 2.99944520e-01 -5.53219378e-01 2.20897540e-01 -5.70621312e-01 4.72322017e-01 -9.05435264e-01 -7.68593967e-01 -1.27742328e-02 3.88700843e-01 5.41221917e-01 -8.58832181e-01 -1.04702222e+00 1.57832518e-01 -6.03041053e-01 -1.69692183e+00 -1.48594335e-01 -2.39764035e-01 -1.16127932e+00 -6.54457808e-01 -1.03069210e+00 -6.50266469e-01 6.14701390e-01 9.22955573e-01 9.72493410e-01 9.10245441e-03 -3.30960333e-01 4.28361684e-01 -5.73161662e-01 3.94580930e-01 -3.54502201e-01 -3.86789352e-01 2.90729254e-01 1.72176227e-01 -5.54251857e-02 -9.81528938e-01 -5.47937989e-01 4.62622225e-01 -9.00271952e-01 2.79654711e-01 5.95764339e-01 4.92827296e-01 1.26906216e+00 4.21125114e-01 5.74635744e-01 -6.07948601e-01 6.65593982e-01 -9.27986428e-02 -6.49542391e-01 1.39138848e-01 -5.86762667e-01 -1.22297928e-01 5.30321360e-01 -3.83726865e-01 -1.53055155e+00 -2.68105771e-02 6.44284040e-02 -5.46102881e-01 -7.97802433e-02 1.49784520e-01 2.52574980e-01 -5.27933955e-01 5.91657639e-01 5.68567812e-01 -5.84922768e-02 -7.03418970e-01 2.60163635e-01 5.70971727e-01 9.85711813e-01 -2.76602775e-01 7.72356629e-01 1.20732939e+00 1.55539900e-01 -1.24444342e+00 -1.08702266e+00 -5.62598825e-01 -1.00100303e+00 -3.78387690e-01 9.81755733e-01 -8.99437964e-01 -2.17137083e-01 1.35677949e-01 -9.63359058e-01 -1.03189573e-01 -3.48894984e-01 5.83645940e-01 -8.78615320e-01 6.42372668e-01 -5.57152212e-01 -1.06635034e+00 -3.80092174e-01 -9.71941054e-01 1.17779493e+00 6.52336299e-01 3.32549177e-02 -8.18396509e-01 1.35406077e-01 4.40351069e-01 6.65837705e-01 4.46768790e-01 1.30484551e-01 2.13728070e-01 -6.52590454e-01 1.89960182e-01 -5.82323670e-01 1.37463212e-01 1.01544127e-01 -1.63334131e-01 -5.93683064e-01 -4.39145744e-01 5.31339467e-01 3.64002422e-03 7.20478117e-01 1.01000035e+00 7.41067350e-01 2.68877279e-02 -2.62963206e-01 8.39449465e-01 2.08865070e+00 1.84569076e-01 9.93907571e-01 4.65362102e-01 3.56863976e-01 3.05623055e-01 8.62001896e-01 4.49161917e-01 1.07536810e-02 9.58846927e-01 7.84793198e-02 -1.25511840e-01 -5.87268114e-01 -1.34993847e-02 2.09529534e-01 6.74661756e-01 -1.01017356e+00 4.37772930e-01 -4.93972957e-01 4.03417438e-01 -1.58276129e+00 -1.36205244e+00 -2.54125804e-01 2.27995920e+00 7.53609478e-01 -2.89441466e-01 7.94021934e-02 -7.90882390e-03 1.16924810e+00 2.86605150e-01 -3.08903456e-01 -3.01369689e-02 -5.45046508e-01 1.98740289e-01 4.75490063e-01 8.54607046e-01 -8.51402581e-01 1.03599989e+00 7.93199348e+00 7.75530636e-01 -1.08828735e+00 4.60687906e-01 4.54785466e-01 3.08507867e-02 -7.85567462e-02 7.37127196e-03 -1.01389575e+00 2.18311980e-01 9.38834131e-01 -5.76825619e-01 6.73938274e-01 3.66906673e-01 6.90628171e-01 -4.66715574e-01 -4.49549675e-01 1.18275034e+00 1.27851382e-01 -1.37694335e+00 2.26148795e-02 -1.49180487e-01 8.40176404e-01 -2.08821014e-01 5.06622158e-02 -4.72104967e-01 1.54644683e-01 -1.08957815e+00 4.45674628e-01 7.30178714e-01 1.23095584e+00 -5.26308119e-01 3.89653563e-01 9.66052189e-02 -1.47063136e+00 -3.80567238e-02 -6.38211429e-01 4.07406054e-02 7.35723615e-01 7.23543942e-01 -2.31047664e-02 7.23451912e-01 9.68668461e-01 6.14512682e-01 -2.96258122e-01 7.89170682e-01 -6.88132420e-02 3.60959589e-01 -1.65978059e-01 7.70500720e-01 -2.53594965e-01 -4.41805035e-01 1.00413966e+00 1.01776206e+00 3.40379626e-01 7.69546866e-01 -1.19735338e-01 9.49225366e-01 3.95684958e-01 -4.20190170e-02 -2.89900124e-01 2.17150658e-01 5.49947679e-01 1.34326160e+00 -9.46593881e-01 -3.53258044e-01 -4.76648241e-01 8.24252188e-01 3.64716835e-02 4.38357055e-01 -6.24495089e-01 1.92070883e-02 2.19714507e-01 5.93302786e-01 3.50432605e-01 -4.66297746e-01 -5.59299409e-01 -1.36841404e+00 -9.10716727e-02 -5.91304064e-01 3.26990187e-01 -1.10388649e+00 -8.63275647e-01 6.66703224e-01 2.94546634e-01 -1.24391437e+00 -4.33995202e-02 1.55838579e-01 -2.30852112e-01 1.01935363e+00 -1.83824551e+00 -8.46409857e-01 -3.10261786e-01 5.57893991e-01 7.64528573e-01 1.92970783e-01 4.16033834e-01 1.65962890e-01 -4.68881428e-01 -1.34046953e-02 1.72376096e-01 -2.46909589e-01 4.87248838e-01 -9.03025210e-01 7.24293441e-02 1.17452979e+00 -3.61371070e-01 5.32804310e-01 1.12357759e+00 -7.40099847e-01 -1.04468656e+00 -5.35771310e-01 7.56913066e-01 -2.06179112e-01 4.33288157e-01 2.27520794e-01 -1.22553492e+00 6.27500117e-01 7.86166638e-02 1.56296074e-01 4.38436329e-01 -3.78760368e-01 -1.63312964e-02 1.24446258e-01 -1.46030927e+00 3.52273196e-01 9.30835605e-01 -3.50998133e-01 -6.86547458e-01 -9.21076462e-02 6.48190379e-01 -4.93031442e-01 -1.36276937e+00 6.01638854e-01 6.16310537e-01 -1.37771571e+00 1.25478077e+00 9.93512496e-02 5.96148849e-01 -5.11320829e-01 -3.90394151e-01 -9.20405030e-01 -6.98673487e-01 -7.00420558e-01 -2.48093188e-01 9.26038384e-01 -1.37895197e-01 -3.86130244e-01 4.65689868e-01 2.70417631e-01 1.53820738e-01 -6.06394231e-01 -1.13849998e+00 -8.81050169e-01 -1.90242708e-01 2.64211923e-01 2.31631070e-01 9.24407601e-01 -1.21846981e-01 3.11815113e-01 -6.56141222e-01 4.82777417e-01 1.33341753e+00 2.94606954e-01 1.76150396e-01 -1.18676221e+00 8.83245394e-02 -4.25890163e-02 -2.12530240e-01 -7.53981829e-01 -2.48562440e-01 -3.11110973e-01 -1.14451170e-01 -1.71112406e+00 6.38845563e-01 4.32937481e-02 -1.36683449e-01 -7.50679448e-02 -1.58189386e-01 4.96330976e-01 7.25751519e-02 8.56012464e-01 -7.05928564e-01 8.95331874e-02 1.61658382e+00 7.82521844e-01 -3.11398566e-01 -2.26926520e-01 -7.81563699e-01 8.25505674e-01 6.91057146e-01 -2.64048696e-01 -1.68803573e-01 -1.68337449e-01 7.51925185e-02 5.81326544e-01 2.76256830e-01 -8.35541844e-01 6.58128038e-02 -3.46399009e-01 3.55037630e-01 -5.43222666e-01 4.90660608e-01 -5.00347733e-01 3.61555874e-01 1.69306874e-01 -2.39693552e-01 -1.46167472e-01 -2.75842637e-01 3.82291615e-01 -2.59090662e-01 -2.56514639e-01 1.62745786e+00 -6.22335374e-01 -9.21236932e-01 -9.25200467e-04 -3.70430171e-01 -1.95498243e-01 8.06938827e-01 -6.33018196e-01 -4.07779545e-01 -3.67269725e-01 -1.06281638e+00 -2.21246496e-01 7.09565938e-01 -2.19080374e-02 1.15725839e+00 -1.24484730e+00 -9.80423272e-01 1.28819030e-02 -5.48658550e-01 -7.58659989e-02 5.66341281e-01 1.07109225e+00 -4.42815691e-01 4.28695172e-01 -3.57066035e-01 -4.35279042e-01 -1.55959082e+00 6.13550305e-01 5.46864748e-01 -4.55560267e-01 -1.10804248e+00 3.91764998e-01 6.55294955e-02 3.86683643e-01 -4.77126032e-01 2.70364195e-01 -5.16216516e-01 -3.20601672e-01 1.19896483e+00 7.89425552e-01 -5.81745923e-01 -1.16652083e+00 -1.16582312e-01 1.14123166e+00 -5.32033145e-02 -4.36682940e-01 1.41850269e+00 -9.97143090e-01 -3.82716447e-01 4.65053707e-01 7.57743180e-01 -8.57966170e-02 -1.19548368e+00 -3.40889961e-01 -3.05512697e-01 -8.30897570e-01 3.70033622e-01 -4.50719446e-01 -9.49697256e-01 4.45100367e-01 5.46033859e-01 3.25466722e-01 1.45847821e+00 2.12683514e-01 7.34246612e-01 -3.76659840e-01 6.98374093e-01 -1.16356564e+00 1.10853754e-01 1.56588033e-01 9.02922750e-01 -1.09935570e+00 5.00858128e-01 -8.47904325e-01 -4.86814022e-01 1.00252545e+00 3.60960245e-01 -3.22438985e-01 5.12704968e-01 5.50049424e-01 -1.80331990e-01 -1.58306882e-01 -6.66806161e-01 -3.37453485e-01 -8.22635815e-02 7.26000488e-01 5.21581590e-01 -2.76249826e-01 -8.65617275e-01 4.77282971e-01 1.06437989e-01 5.68033457e-01 9.11762059e-01 8.17470968e-01 -1.04481208e+00 -7.19015718e-01 -8.53363752e-01 4.00144197e-02 -8.22369277e-01 4.26559597e-02 3.84487733e-02 3.70604932e-01 -2.31948316e-01 8.44657898e-01 -1.34483084e-01 -2.37890348e-01 4.87466119e-02 -3.16142142e-01 8.52882683e-01 -3.66923749e-01 5.93428202e-02 4.76918131e-01 -1.59311429e-01 -7.19947636e-01 -1.20564651e+00 -6.74784184e-01 -1.30607867e+00 -1.19722597e-01 -3.53420526e-01 1.15341552e-01 4.86156434e-01 8.15018594e-01 1.81959927e-01 3.86913955e-01 7.77100563e-01 -1.10136354e+00 -2.67059147e-01 -6.81602180e-01 -1.10339117e+00 1.97558165e-01 4.49176222e-01 -5.66469848e-01 -5.48925281e-01 2.64500558e-01]
[11.020029067993164, -2.2026050090789795]
ad32b296-9bbc-4dd7-9b62-fbaf14da7adf
implicit-3d-human-mesh-recovery-using-1
2306.17651
null
https://arxiv.org/abs/2306.17651v2
https://arxiv.org/pdf/2306.17651v2.pdf
Implicit 3D Human Mesh Recovery using Consistency with Pose and Shape from Unseen-view
From an image of a person, we can easily infer the natural 3D pose and shape of the person even if ambiguity exists. This is because we have a mental model that allows us to imagine a person's appearance at different viewing directions from a given image and utilize the consistency between them for inference. However, existing human mesh recovery methods only consider the direction in which the image was taken due to their structural limitations. Hence, we propose "Implicit 3D Human Mesh Recovery (ImpHMR)" that can implicitly imagine a person in 3D space at the feature-level via Neural Feature Fields. In ImpHMR, feature fields are generated by CNN-based image encoder for a given image. Then, the 2D feature map is volume-rendered from the feature field for a given viewing direction, and the pose and shape parameters are regressed from the feature. To utilize consistency with pose and shape from unseen-view, if there are 3D labels, the model predicts results including the silhouette from an arbitrary direction and makes it equal to the rotated ground-truth. In the case of only 2D labels, we perform self-supervised learning through the constraint that the pose and shape parameters inferred from different directions should be the same. Extensive evaluations show the efficacy of the proposed method.
['Junmo Kim', 'Jaesung Ahn', 'Yooshin Cho', 'Hanbyel Cho']
2023-06-30
implicit-3d-human-mesh-recovery-using
http://openaccess.thecvf.com//content/CVPR2023/html/Cho_Implicit_3D_Human_Mesh_Recovery_Using_Consistency_With_Pose_and_CVPR_2023_paper.html
http://openaccess.thecvf.com//content/CVPR2023/papers/Cho_Implicit_3D_Human_Mesh_Recovery_Using_Consistency_With_Pose_and_CVPR_2023_paper.pdf
cvpr-2023-1
['self-supervised-learning', 'human-mesh-recovery']
['computer-vision', 'computer-vision']
[ 1.00124605e-01 3.43950510e-01 -1.96165852e-02 -5.20271897e-01 -2.70330638e-01 -3.19997281e-01 4.49727654e-01 -3.26243371e-01 -1.55221477e-01 5.13934851e-01 2.13949651e-01 3.50396752e-01 2.19895363e-01 -8.57492030e-01 -8.51186752e-01 -4.71789896e-01 5.52675545e-01 7.72579074e-01 -1.93204403e-01 -1.58720203e-02 1.32023573e-01 5.15441477e-01 -1.49396765e+00 2.86954287e-02 4.49089527e-01 8.91681850e-01 2.45786794e-02 3.70864123e-01 2.18812600e-01 2.30232954e-01 -3.89694691e-01 -4.99702215e-01 5.38324535e-01 -2.79313594e-01 -7.62216747e-01 9.47891235e-01 9.36346054e-01 -8.01679194e-01 -5.71363449e-01 1.03450751e+00 1.69757247e-01 3.21209654e-02 7.32340455e-01 -9.38671589e-01 -6.78554773e-01 -2.47156054e-01 -7.05161154e-01 -4.21651840e-01 1.00440943e+00 1.65687099e-01 7.42039800e-01 -8.76312137e-01 1.05565298e+00 1.49763310e+00 4.77865905e-01 6.71950936e-01 -1.32045031e+00 -4.70960736e-01 4.12875444e-01 -1.44256353e-01 -1.65221798e+00 -3.09243858e-01 1.12881589e+00 -6.78426683e-01 4.87182856e-01 7.61375055e-02 1.14890420e+00 1.06580138e+00 2.36922055e-01 4.44739342e-01 1.11434364e+00 -4.15157199e-01 1.04249462e-01 2.13610634e-01 -1.45983234e-01 1.07137322e+00 2.28677556e-01 3.83584276e-02 -5.65234065e-01 -6.87957406e-02 1.40426266e+00 2.62035131e-01 -3.20536256e-01 -7.68367648e-01 -1.35235429e+00 5.33238530e-01 3.92956257e-01 -2.87133425e-01 -3.88641685e-01 -3.49940956e-02 -1.52526930e-01 -5.55647276e-02 4.26241279e-01 1.00183703e-01 -1.77416474e-01 3.84001374e-01 -6.11807942e-01 5.40928304e-01 6.36068463e-01 1.09895575e+00 9.85483348e-01 -8.76601115e-02 9.29904878e-02 4.51160133e-01 4.39297199e-01 8.15500438e-01 1.98873460e-01 -1.29732823e+00 3.20878893e-01 8.37209761e-01 3.63196760e-01 -1.36533380e+00 -1.93738684e-01 -1.88072294e-01 -8.65306735e-01 3.06949466e-01 5.64525843e-01 8.61873850e-02 -1.19324410e+00 1.80719090e+00 6.31820083e-01 4.47667651e-02 -2.55829334e-01 1.37081778e+00 7.68361509e-01 2.19580472e-01 -4.26946342e-01 -4.21358794e-02 1.41885054e+00 -4.86628085e-01 -6.08760834e-01 -4.03990895e-01 9.57766175e-02 -4.29296792e-01 7.71209478e-01 3.35564375e-01 -1.07240319e+00 -7.63633430e-01 -1.07191837e+00 -2.65011162e-01 -5.52956238e-02 4.72638607e-02 3.59026521e-01 2.30670378e-01 -7.12921441e-01 4.15517509e-01 -7.84861624e-01 -3.75896037e-01 3.01312320e-02 2.10210800e-01 -6.37794375e-01 -1.19310729e-01 -1.02621186e+00 9.13213670e-01 -1.74065074e-03 3.14444065e-01 -7.40979254e-01 -2.36566782e-01 -1.13934767e+00 -3.49502563e-01 3.83790284e-01 -1.17762160e+00 9.26683247e-01 -1.08566821e+00 -1.21280551e+00 1.40344965e+00 -5.30483007e-01 8.07403997e-02 7.74543166e-01 -2.38910392e-01 -1.00416988e-01 2.41340205e-01 2.94120550e-01 8.03773642e-01 1.15870297e+00 -1.68520379e+00 -5.42890094e-02 -7.26236463e-01 3.99089605e-01 5.62857687e-01 2.86507159e-01 -6.68161094e-01 -6.83134735e-01 -4.41774964e-01 7.35946178e-01 -1.06718242e+00 -1.00273483e-01 5.39569497e-01 -7.58994818e-01 5.85503019e-02 4.64938998e-01 -8.59180689e-01 6.42334938e-01 -2.10318780e+00 2.74281502e-01 4.15102869e-01 4.24421370e-01 -3.87750536e-01 1.58496514e-01 1.33681819e-02 -1.65019426e-02 -5.75084984e-02 -8.21810886e-02 -5.87065935e-01 -1.84908405e-01 2.81819373e-01 -1.67605281e-01 8.40642273e-01 1.08252041e-01 7.75795162e-01 -9.00326312e-01 -7.25532055e-01 4.48377699e-01 6.92752242e-01 -6.69981539e-01 5.25895417e-01 -1.18280143e-01 9.60291982e-01 -6.72534406e-01 5.21888256e-01 7.50377536e-01 -4.49810505e-01 6.81845844e-02 -6.27586246e-01 1.47799298e-01 5.88192372e-03 -1.39897311e+00 2.07503986e+00 -3.56145978e-01 2.08904237e-01 -8.26785490e-02 -6.24046326e-01 7.55920291e-01 4.14994329e-01 4.90056694e-01 -5.89475930e-01 5.62267080e-02 -7.89698437e-02 -1.69600829e-01 -6.79048836e-01 2.36921996e-01 -2.50309914e-01 1.44606069e-01 2.89427400e-01 -2.61694640e-01 -2.26055920e-01 -3.03692967e-01 4.19611335e-02 4.63816077e-01 5.49944103e-01 3.13452870e-01 2.15158835e-01 4.05899674e-01 -3.06039125e-01 6.25123024e-01 3.37569207e-01 1.37821749e-01 1.13285577e+00 1.54997006e-01 -7.42808938e-01 -1.21179461e+00 -1.35085964e+00 -2.41688043e-01 3.33069772e-01 5.28062284e-01 -2.62678981e-01 -7.92310596e-01 -4.75236535e-01 1.31532237e-01 4.47239548e-01 -9.15357053e-01 6.55850172e-02 -6.03342831e-01 -1.42230839e-01 -2.85433792e-02 2.84343630e-01 5.75482726e-01 -8.76584470e-01 -6.84604526e-01 -1.59925461e-01 -5.23853064e-01 -1.17723095e+00 -6.73123360e-01 -6.09150529e-01 -6.65356934e-01 -1.09968305e+00 -7.43785203e-01 -7.49062359e-01 1.40834093e+00 1.55968904e-01 9.01311576e-01 2.30032787e-01 -2.68258482e-01 4.66479808e-01 5.31323478e-02 2.01326311e-01 -9.41657946e-02 -4.76816505e-01 2.11058006e-01 3.66666287e-01 -1.88266039e-02 -6.64024770e-01 -9.23863173e-01 3.29731017e-01 -5.05750358e-01 5.18131256e-01 2.73567021e-01 5.98709226e-01 8.65666151e-01 -1.28224745e-01 2.05503717e-01 -8.95527422e-01 4.94216941e-02 -1.42888278e-01 -2.11948469e-01 1.24959745e-01 -3.79527926e-01 -3.94881777e-02 3.05314332e-01 -5.85561037e-01 -1.12560987e+00 2.21520677e-01 4.48324159e-02 -8.00472796e-01 -4.34847385e-01 1.07682355e-01 -4.23760593e-01 3.21020365e-01 4.85863656e-01 1.39794752e-01 1.90000623e-01 -4.52705264e-01 2.45814815e-01 5.00304699e-01 5.77579916e-01 -7.05595434e-01 7.68469751e-01 9.51378822e-01 1.21790498e-01 -6.56803727e-01 -1.07852197e+00 -2.01255977e-01 -1.10627306e+00 -3.12988967e-01 1.01008189e+00 -1.11806929e+00 -7.72671998e-01 4.35957879e-01 -1.31891465e+00 8.72597992e-02 -1.47917688e-01 4.34903175e-01 -5.42676151e-01 4.82734501e-01 -3.45964372e-01 -7.76753843e-01 -1.07710436e-01 -1.18646264e+00 1.49184215e+00 1.67207979e-02 -5.33285618e-01 -9.79616344e-01 -2.29454443e-01 4.89486217e-01 -1.34882644e-01 5.79333901e-01 8.48627746e-01 -1.26737624e-01 -7.34102905e-01 -2.58670807e-01 -6.87535480e-02 2.17034221e-01 2.38602072e-01 -2.75622666e-01 -9.57671225e-01 -2.91162789e-01 1.04856141e-01 -9.88377333e-02 1.93599880e-01 4.58786935e-01 1.11797357e+00 -4.67615992e-01 -3.79136950e-01 7.99691200e-01 1.27638566e+00 -1.07264757e-01 4.42909241e-01 1.02957180e-02 1.14861524e+00 7.70871699e-01 3.67996663e-01 4.06788141e-01 5.66164315e-01 7.43540585e-01 4.85393345e-01 -8.68622139e-02 -1.91656664e-01 -7.87521005e-01 7.02747777e-02 5.58073223e-01 -4.60922062e-01 4.82524745e-02 -5.89550495e-01 2.87322551e-01 -1.51018524e+00 -6.58259630e-01 1.22620940e-01 2.63247228e+00 6.44910872e-01 1.69227764e-01 -1.59754053e-01 -2.01357067e-01 7.43872225e-01 1.53657217e-02 -9.18833137e-01 1.11021474e-01 1.44258603e-01 -1.47682473e-01 8.94962922e-02 6.46375835e-01 -6.61760509e-01 6.70197725e-01 5.81177044e+00 1.68910667e-01 -1.04833233e+00 -1.92392126e-01 5.53321123e-01 -1.68133080e-02 -4.27846849e-01 1.69089034e-01 -7.01043010e-01 2.41780818e-01 1.36211395e-01 6.98089153e-02 4.30975586e-01 5.80308676e-01 1.57393694e-01 -9.78958234e-02 -1.55115199e+00 1.26805556e+00 4.55676675e-01 -8.73033941e-01 3.65087122e-01 4.07437384e-01 5.74159801e-01 -6.25606656e-01 9.35163349e-02 -6.87795877e-02 -1.34181917e-01 -1.06184959e+00 9.31515038e-01 1.08792400e+00 1.18337369e+00 -4.77500081e-01 4.62387860e-01 6.02668643e-01 -1.01096988e+00 4.05125350e-01 -2.43430540e-01 -2.37297207e-01 4.59195048e-01 5.51847160e-01 -8.56664717e-01 4.33194757e-01 4.80874509e-01 7.52614021e-01 -4.10076559e-01 5.06663978e-01 -1.77476332e-01 -2.04449892e-03 -2.71129131e-01 5.90919375e-01 -4.10302192e-01 -4.79889035e-01 5.83684862e-01 6.14534140e-01 3.63083065e-01 1.66307896e-01 4.83625472e-01 1.17904592e+00 -7.53614074e-03 -1.33319646e-01 -7.07989275e-01 4.30302262e-01 3.97635430e-01 9.86212552e-01 -5.47573328e-01 -2.80355036e-01 -3.12518209e-01 1.30535436e+00 3.70884955e-01 5.58939278e-01 -6.60695374e-01 1.91910535e-01 4.79870021e-01 8.24202538e-01 3.50459702e-02 -1.75604090e-01 -2.37574369e-01 -1.33161080e+00 2.88933605e-01 -5.71469307e-01 7.32750772e-03 -1.28028917e+00 -1.31962764e+00 5.37261665e-01 3.08259577e-01 -1.48535109e+00 -4.09749061e-01 -5.09249032e-01 -2.83976644e-01 9.34969962e-01 -9.98339474e-01 -1.34776926e+00 -4.45989102e-01 4.50676560e-01 4.44894791e-01 2.72588909e-01 8.01560462e-01 -1.06684364e-01 -1.78532109e-01 5.58972538e-01 -5.72501540e-01 4.72302049e-01 7.69020617e-01 -1.08847773e+00 2.62452066e-01 4.12339181e-01 1.67105421e-02 9.30328190e-01 5.79373300e-01 -8.91987085e-01 -1.44575477e+00 -9.18650448e-01 7.18725622e-01 -7.35985160e-01 3.35194655e-02 -4.24074739e-01 -8.54450822e-01 1.01375318e+00 -1.30190030e-01 2.44773284e-01 2.50719398e-01 1.92613378e-02 -4.09168720e-01 4.69293538e-03 -1.26895761e+00 7.79175818e-01 1.32461274e+00 -6.28098249e-01 -5.98114967e-01 3.30435067e-01 5.76552868e-01 -8.61464679e-01 -9.76564825e-01 2.78463870e-01 8.01780224e-01 -9.13080692e-01 1.14618170e+00 -5.07823110e-01 4.76382077e-01 -3.67495149e-01 -1.84952572e-01 -1.28251851e+00 -3.06495100e-01 -2.47917756e-01 -1.16938837e-01 9.37495589e-01 -4.11590077e-02 -5.60152769e-01 7.63330996e-01 1.03956485e+00 1.97990626e-01 -8.34395409e-01 -8.61500680e-01 -4.12062466e-01 -1.29268900e-01 -2.55576044e-01 7.39728510e-01 8.81683230e-01 -1.53208077e-01 4.57381457e-01 -6.17838740e-01 3.96952331e-01 8.52616310e-01 3.39462340e-01 9.37798679e-01 -1.33186150e+00 -2.38646790e-01 2.65432864e-01 -5.21799982e-01 -1.53403234e+00 3.19242775e-01 -8.10487688e-01 -4.26322855e-02 -1.43404615e+00 4.08158630e-01 -2.44931772e-01 1.71083137e-01 3.34892929e-01 -7.42766261e-02 1.59648418e-01 1.45901870e-02 3.21582168e-01 -2.49137595e-01 5.38394928e-01 1.93916523e+00 3.22979726e-02 8.04524124e-02 -1.22311272e-01 -4.65546340e-01 1.20650899e+00 3.49784970e-01 -1.94815174e-01 -4.23669040e-01 -6.13525748e-01 2.41611436e-01 4.83022273e-01 8.45621169e-01 -6.38942242e-01 5.63280284e-02 -3.35619450e-01 8.97579670e-01 -6.30445182e-01 8.42196047e-01 -9.05015230e-01 5.05789101e-01 1.33256584e-01 -3.86112273e-01 1.29517406e-01 -4.73539650e-01 6.86051130e-01 2.50370085e-01 5.55755794e-02 4.31514293e-01 -4.58311319e-01 -3.43953192e-01 6.73055232e-01 7.43720829e-02 4.75595593e-02 7.14706779e-01 -6.39319241e-01 2.08743304e-01 -5.79534411e-01 -1.03842831e+00 -1.35963917e-01 1.09403765e+00 3.01583081e-01 1.07083488e+00 -1.52671313e+00 -7.03549623e-01 6.20114326e-01 1.44548655e-01 7.01059520e-01 3.50890011e-01 6.45727754e-01 -2.91989475e-01 6.11865595e-02 -9.10539106e-02 -8.14747989e-01 -1.12903965e+00 6.01199687e-01 5.75737953e-01 9.63598192e-02 -9.74361539e-01 3.89787495e-01 7.04381645e-01 -5.89338899e-01 1.29299968e-01 -2.03030437e-01 2.12507229e-02 -4.39925045e-01 3.53391260e-01 5.47585189e-02 -1.75836429e-01 -1.28769672e+00 -1.77385852e-01 1.06693637e+00 -3.82104665e-02 -3.15869927e-01 9.42111790e-01 -3.39863986e-01 4.29899246e-03 6.31586254e-01 1.26440239e+00 2.56123662e-01 -1.65109026e+00 -3.91448557e-01 -9.28928256e-01 -9.09827769e-01 -3.34070146e-01 -5.62041759e-01 -1.19183004e+00 8.36471379e-01 5.11019766e-01 -4.04866546e-01 6.82675421e-01 9.43750739e-02 6.16105080e-01 1.89287841e-01 6.66867256e-01 -8.23979616e-01 1.49453998e-01 1.44305483e-01 1.05127370e+00 -1.15011990e+00 3.21958601e-01 -7.12335169e-01 -5.18884420e-01 8.97273898e-01 8.76420438e-01 -3.13735157e-01 7.30380833e-01 -2.00003058e-01 -6.63417876e-02 -3.92772436e-01 -3.21168393e-01 1.70389161e-01 6.73795402e-01 7.29866266e-01 2.84927011e-01 2.23355502e-01 1.42613530e-01 2.52911180e-01 -4.90405172e-01 -8.05962160e-02 4.24074203e-01 6.19660676e-01 -1.21081233e-01 -6.61627471e-01 -5.65438330e-01 2.89505810e-01 -7.50944763e-02 3.32845211e-01 -2.05296978e-01 7.37699926e-01 5.15587807e-01 5.93300819e-01 2.97093123e-01 -3.33845973e-01 4.95414257e-01 -9.93090868e-02 9.07534659e-01 -8.97390306e-01 1.66630954e-01 1.40985265e-01 -4.26904000e-02 -4.38762933e-01 -4.51382428e-01 -6.39917254e-01 -1.52791250e+00 -1.62365675e-01 -6.00980632e-02 -1.97102740e-01 2.92348534e-01 1.07459283e+00 1.77780434e-01 1.66761741e-01 5.09048939e-01 -1.10393739e+00 -2.32848093e-01 -7.64191687e-01 -7.33596265e-01 1.12663186e+00 4.71839428e-01 -1.06827354e+00 -4.72264767e-01 4.88220215e-01]
[7.191318511962891, -1.3257355690002441]
fa22dc2f-02b0-4e42-a783-768996471e0c
n-gram-opcode-analysis-for-android-malware
1612.01445
null
http://arxiv.org/abs/1612.01445v1
http://arxiv.org/pdf/1612.01445v1.pdf
N-gram Opcode Analysis for Android Malware Detection
Android malware has been on the rise in recent years due to the increasing popularity of Android and the proliferation of third party application markets. Emerging Android malware families are increasingly adopting sophisticated detection avoidance techniques and this calls for more effective approaches for Android malware detection. Hence, in this paper we present and evaluate an n-gram opcode features based approach that utilizes machine learning to identify and categorize Android malware. This approach enables automated feature discovery without relying on prior expert or domain knowledge for pre-determined features. Furthermore, by using a data segmentation technique for feature selection, our analysis is able to scale up to 10-gram opcodes. Our experiments on a dataset of 2520 samples showed an f-measure of 98% using the n-gram opcode based approach. We also provide empirical findings that illustrate factors that have probable impact on the overall n-gram opcodes performance trends.
['Kieran McLaughlin', 'Suleiman Y. Yerima', 'Sakir Sezer', 'BooJoong Kang']
2016-12-05
null
null
null
null
['android-malware-detection']
['miscellaneous']
[ 2.68854320e-01 -1.53831407e-01 -6.40786469e-01 -1.30869329e-01 -5.90408564e-01 -7.02633560e-01 6.96891427e-01 3.02166581e-01 -4.24196690e-01 3.36079359e-01 -5.26678003e-02 -8.43366385e-01 -1.68959141e-01 -4.33014691e-01 -2.73406863e-01 -4.08377200e-01 -3.56942832e-01 5.81265613e-02 4.47781533e-01 1.61194131e-02 7.55899787e-01 4.14861441e-01 -1.89561510e+00 2.16377720e-01 1.04592764e+00 1.03862309e+00 -9.50951781e-03 1.03892553e+00 1.21825968e-03 5.11940837e-01 -8.59037101e-01 -4.65777218e-01 3.22543561e-01 -2.53860742e-01 -4.89125609e-01 -2.50059307e-01 -5.56782894e-02 -2.60008723e-01 3.07401985e-01 1.05917323e+00 5.42170443e-02 -2.33152509e-01 7.06775188e-01 -1.04784548e+00 -1.15485869e-01 2.39430994e-01 -7.34248638e-01 4.04503465e-01 6.75213277e-01 -1.56849585e-02 7.07809746e-01 -3.95469636e-01 2.55810827e-01 8.54622602e-01 8.21152151e-01 1.88065171e-01 -9.19673800e-01 -6.47449017e-01 -1.73056766e-01 2.24442512e-01 -1.41268671e+00 -2.60427862e-01 6.74486458e-01 -6.76978111e-01 1.14580810e+00 3.65971953e-01 7.43896663e-01 8.43251646e-01 6.02847278e-01 2.64455050e-01 1.35017526e+00 -4.40177500e-01 3.67506444e-01 3.84609789e-01 2.37487629e-01 6.82464123e-01 6.46465361e-01 -9.33482945e-02 4.02587429e-02 -8.27722967e-01 -2.50546224e-02 1.61538005e-01 2.44836897e-01 -8.88060182e-02 -3.68302077e-01 1.15359235e+00 -1.92788407e-01 4.35732275e-01 -2.71319151e-01 -1.50651351e-01 7.53880322e-01 -5.70372455e-02 4.30858880e-01 5.58719754e-01 -5.98227561e-01 -9.78710353e-01 -9.88501608e-01 -1.15305468e-01 8.14661145e-01 4.19586271e-01 6.83563769e-01 -1.10963173e-01 5.68058550e-01 7.77538896e-01 4.71934944e-01 3.68110627e-01 9.57114875e-01 -9.38085973e-01 1.48938345e-02 9.61594582e-01 -3.02454561e-01 -1.03170097e+00 -2.15373542e-02 1.06689157e-02 -1.61147118e-01 8.77699405e-02 2.93977737e-01 -1.22066945e-01 -9.08826411e-01 1.22206104e+00 2.28772357e-01 -7.09580956e-03 -2.32313365e-01 -1.99448273e-01 8.06251839e-02 5.02176702e-01 9.44755301e-02 -4.49985087e-01 1.58548605e+00 -1.93180218e-01 -3.26534778e-01 9.70779732e-02 7.37505376e-01 -7.98557758e-01 1.21746063e+00 6.70787871e-01 -2.97441244e-01 -1.82477862e-01 -1.07367194e+00 4.15264547e-01 -7.81125367e-01 1.20271081e-02 5.68497896e-01 1.66789460e+00 -7.59344280e-01 1.99362770e-01 -1.00958526e+00 -5.41846991e-01 6.73860967e-01 8.68554473e-01 -1.12215914e-01 3.24239671e-01 -3.74520540e-01 4.74249214e-01 3.73143554e-01 -5.06183088e-01 -7.13622093e-01 -2.72761881e-01 -5.40044546e-01 -1.24030747e-01 5.67629933e-01 2.40770519e-01 1.22615457e+00 -9.73151922e-01 -1.50971377e+00 5.34900367e-01 -2.78238714e-01 -5.44026911e-01 2.97562294e-02 -3.16751212e-01 -6.06580079e-01 2.06358626e-01 8.93994048e-02 9.91720185e-02 1.20286739e+00 -1.10566211e+00 -7.41983354e-01 -3.04282129e-01 1.23306349e-01 -3.68984997e-01 -7.51839578e-01 2.97556639e-01 -1.49370924e-01 -3.77251685e-01 -4.65535253e-01 -1.31486380e+00 6.20112829e-02 -9.69925165e-01 -3.16802889e-01 -1.75607100e-01 1.42338312e+00 -7.47784495e-01 1.57830369e+00 -2.24220562e+00 -3.09453487e-01 5.61846316e-01 3.37882698e-01 5.95954955e-01 2.80759901e-01 2.58985162e-01 6.27591461e-02 6.39081776e-01 -3.44631016e-01 5.02486750e-02 -3.38545233e-01 1.86821297e-01 -1.60046458e-01 3.55994850e-01 1.51172861e-01 4.58748549e-01 -6.60902917e-01 -5.23788869e-01 1.12388782e-01 6.49513841e-01 -7.70101368e-01 -5.63340522e-02 -1.45801112e-01 6.11812286e-02 -4.93643045e-01 1.00397170e+00 4.28677529e-01 -3.81743498e-02 2.73257047e-01 2.63859153e-01 -1.88017100e-01 3.96606922e-01 -4.23717380e-01 7.40284085e-01 -5.43836653e-01 6.49034023e-01 -1.69272438e-01 -5.45376062e-01 6.71655595e-01 -6.28350526e-02 6.75860524e-01 -4.53863382e-01 6.08618915e-01 5.20890176e-01 6.82982266e-01 -3.55740815e-01 5.17038286e-01 4.64146793e-01 -2.64468770e-02 4.10192996e-01 -3.73508483e-01 -9.88367386e-03 1.52832702e-01 1.45602182e-01 1.30891502e+00 -2.22102419e-01 7.21947074e-01 -4.82071787e-01 9.21427727e-01 4.26273495e-02 1.64716259e-01 3.93986613e-01 -4.58117425e-01 -1.19347190e-02 8.20764065e-01 -1.91848546e-01 -5.97418487e-01 -6.47831917e-01 -5.88657595e-02 1.06033933e+00 -2.39473075e-01 -8.80393922e-01 -1.23613834e+00 -1.31620193e+00 -1.98358595e-01 4.13047403e-01 -6.87927842e-01 -2.46778131e-01 -6.35364115e-01 -7.93945134e-01 5.77567518e-01 1.17549310e-02 3.97670597e-01 -7.45536447e-01 -1.00167203e+00 -1.91475406e-01 3.27539742e-01 -9.13223505e-01 -2.37560406e-01 3.08033019e-01 -9.97017980e-01 -1.40771413e+00 -4.08417523e-01 -4.36750263e-01 7.05397785e-01 1.53581962e-01 6.16863549e-01 2.55045891e-01 -3.93552631e-01 5.26757419e-01 -6.76087081e-01 -4.90875155e-01 -7.35530257e-01 5.61299384e-01 1.11049451e-01 -5.45159020e-02 7.77802765e-01 -3.69241893e-01 -3.83338332e-01 2.69779593e-01 -7.51186073e-01 -9.35740292e-01 6.50029898e-01 2.59415001e-01 4.39396083e-01 5.23421884e-01 4.32508886e-01 -1.02736449e+00 9.60610986e-01 -8.05836797e-01 -5.36897361e-01 -1.74783945e-01 -8.25026870e-01 -1.85477182e-01 6.63000345e-01 -8.50476325e-01 -7.54599035e-01 2.62757063e-01 -3.77176255e-01 9.48031098e-02 -1.37614056e-01 3.19291085e-01 -2.20397055e-01 -4.55021441e-01 6.03304446e-01 5.93694225e-02 1.51971251e-01 -2.41512552e-01 5.04464321e-02 1.07712638e+00 -1.25895012e-02 -1.56577229e-01 6.51098490e-01 3.21923494e-01 -2.10116461e-01 -1.38453746e+00 -1.66406497e-01 -3.91624153e-01 -3.65397036e-01 2.53176019e-02 9.44397926e-01 -4.33439761e-01 -8.30123305e-01 4.41815078e-01 -6.64310634e-01 -1.18773423e-01 1.53205618e-01 2.82220542e-01 -2.21940026e-01 7.12335825e-01 -3.33757013e-01 -1.03405404e+00 -1.31938055e-01 -1.53908157e+00 8.72979045e-01 2.15746388e-01 -7.18845487e-01 -8.75288129e-01 7.56146982e-02 3.76472980e-01 4.59467918e-01 1.38485720e-02 9.04354990e-01 -1.14660358e+00 -1.10647276e-01 -4.55044061e-01 -2.08524480e-01 4.30975109e-01 6.25646651e-01 3.81797463e-01 -8.55450153e-01 -1.74797133e-01 3.88308227e-01 2.73720443e-01 4.87855673e-01 3.77234131e-01 1.26217794e+00 -5.33838272e-01 -5.55756271e-01 2.24887058e-01 1.12827206e+00 7.66229391e-01 5.87391675e-01 4.51019198e-01 8.36174905e-01 6.62086129e-01 7.62071729e-01 3.13639969e-01 2.96225935e-01 4.82712269e-01 6.35093570e-01 5.86523533e-01 2.38852352e-01 -3.28877643e-02 6.96695507e-01 5.25662839e-01 -1.62634179e-01 -8.77216160e-02 -1.05799139e+00 3.15832853e-01 -1.12647963e+00 -7.09385097e-01 -1.16501689e-01 2.42560434e+00 5.23752213e-01 4.82333809e-01 7.11752892e-01 4.09889251e-01 5.76373160e-01 -1.60064865e-02 -3.40234846e-01 -1.01508665e+00 4.75283623e-01 3.40123266e-01 7.99380839e-01 5.11906207e-01 -1.06832886e+00 6.98109925e-01 6.65879393e+00 1.12844062e+00 -1.27685070e+00 2.52003968e-01 6.94543779e-01 2.11643323e-01 -4.74682860e-02 -6.76534250e-02 -8.10894310e-01 9.24029469e-01 1.39846849e+00 7.98023418e-02 3.88087183e-01 1.08753705e+00 9.07387584e-03 -6.32666051e-01 -4.79916632e-01 1.12564433e+00 1.51635408e-01 -8.54288280e-01 -2.13564143e-01 1.07726276e+00 6.67276323e-01 2.52761580e-02 3.15558076e-01 2.84547687e-01 -7.57376030e-02 -8.85034800e-01 1.34926543e-01 -8.30414817e-02 5.92390120e-01 -1.06047642e+00 7.17086732e-01 1.85448423e-01 -1.11843777e+00 -6.16251528e-01 1.78629696e-01 -1.38933659e-01 -2.30253279e-01 4.86423165e-01 -1.20013654e+00 1.39739603e-01 6.39871120e-01 5.03616869e-01 -1.03695750e+00 8.14316750e-01 1.82907999e-01 1.04129803e+00 -3.79332215e-01 -2.95804888e-01 3.58580463e-02 -2.72959657e-02 3.73396218e-01 1.11285114e+00 4.37441915e-01 -4.28834468e-01 -3.73223759e-02 1.53570563e-01 2.05377817e-01 2.63883233e-01 -7.24148631e-01 -6.65408313e-01 3.22021365e-01 1.28080821e+00 -1.33603346e+00 -2.96852887e-01 -4.38713849e-01 6.95105374e-01 -3.40816289e-01 -1.13045163e-01 -8.57838869e-01 -5.43357790e-01 8.79485250e-01 5.51276028e-01 3.85049999e-01 -5.15644550e-01 -2.30507329e-01 -6.64521813e-01 -1.86112255e-01 -1.39888942e+00 5.95686771e-02 2.54332572e-01 -7.37542033e-01 7.48955905e-01 1.51796460e-01 -9.85337079e-01 -7.51948833e-01 -8.78695011e-01 -5.41717112e-01 2.91166037e-01 -1.08039200e+00 -6.61305547e-01 6.35076687e-02 2.06293777e-01 4.71655399e-01 -4.10096914e-01 6.61570430e-01 2.96005338e-01 -5.26823759e-01 5.97235978e-01 8.27388316e-02 -1.51130170e-01 2.60458469e-01 -1.05509412e+00 2.41147935e-01 8.11787724e-01 2.33889055e-02 1.08404541e+00 5.78170359e-01 -1.11470127e+00 -1.37769341e+00 -7.28022218e-01 5.34159362e-01 -6.91834033e-01 8.20557237e-01 -4.72049654e-01 -6.37178302e-01 4.58209842e-01 5.50218821e-02 -4.53068435e-01 1.07166314e+00 -6.25950247e-02 -5.03857732e-01 9.71753672e-02 -1.38909757e+00 4.49574083e-01 5.56082666e-01 -6.74147129e-01 -1.42568797e-01 -2.12877430e-02 2.59248644e-01 1.25229985e-01 -6.73424423e-01 5.54304600e-01 7.67934859e-01 -1.29314828e+00 7.04280674e-01 1.76404100e-02 -2.36297445e-03 -1.97174713e-01 -2.76521295e-01 -5.93707085e-01 3.07510644e-01 -8.20344269e-01 -5.33833146e-01 1.10072875e+00 4.78959590e-01 -9.35876489e-01 9.85466838e-01 2.06487343e-01 1.50186568e-01 -1.02328324e+00 -8.63795519e-01 -8.91672730e-01 -2.25304157e-01 -5.01568079e-01 3.92451465e-01 5.73307812e-01 -1.50640547e-01 -6.36210069e-02 -6.25767857e-02 -2.30143502e-01 2.33550131e-01 -5.08851469e-01 8.66867661e-01 -1.35866714e+00 -5.31863034e-01 -6.73443496e-01 -7.07103908e-01 -6.39137506e-01 1.49391040e-01 -4.79157507e-01 -2.69060135e-01 -7.53344238e-01 3.06915700e-01 -2.31106386e-01 1.05811603e-01 3.45470160e-01 5.08649163e-02 6.60216630e-01 -8.76557156e-02 1.69561327e-01 -4.30600435e-01 -4.42317761e-02 3.89790893e-01 1.41542152e-01 -6.03731573e-01 2.33301774e-01 -7.89207697e-01 9.42276061e-01 1.16244292e+00 -4.86379087e-01 -6.87287807e-01 3.86199534e-01 2.91925520e-01 -9.62310195e-01 -5.40465340e-02 -1.07888329e+00 -4.27419662e-01 2.85221450e-02 1.84192508e-01 -3.02619159e-01 4.29896384e-01 -9.58202362e-01 1.63354762e-02 9.98750329e-01 3.10214520e-01 3.23685378e-01 4.50844914e-01 5.28654397e-01 1.11838356e-01 -3.71040225e-01 6.19125545e-01 1.74651191e-01 -5.58973193e-01 -2.07144886e-01 -9.90212142e-01 -3.08337629e-01 1.46487713e+00 -7.61529148e-01 -2.59573847e-01 -2.60763854e-01 -2.09995285e-01 -5.58056176e-01 8.48774850e-01 5.04198134e-01 3.24728131e-01 -6.21727705e-01 7.53232837e-02 3.59948307e-01 1.44087166e-01 -9.62694049e-01 -2.21024364e-01 8.83427620e-01 -6.78891838e-01 3.98143530e-01 -1.15360327e-01 -5.96960843e-01 -1.91163933e+00 5.78730226e-01 -3.66540588e-02 -2.61821389e-01 -1.51855350e-02 3.91835928e-01 -3.68112355e-01 -1.65133879e-01 -6.40843362e-02 -3.31550062e-01 -3.66592079e-01 1.80618122e-01 6.96754575e-01 8.16700697e-01 1.01501442e-01 -1.12703800e+00 -8.01190913e-01 6.23185098e-01 -2.93177366e-01 8.73981975e-03 8.36399019e-01 9.17282850e-02 -3.58267754e-01 2.67139316e-01 1.47442734e+00 9.07021046e-01 -6.65180504e-01 8.05554211e-01 5.23551822e-01 -5.78169227e-01 -2.96107858e-01 -6.04586959e-01 -7.42998004e-01 5.14820993e-01 9.09399450e-01 9.45303738e-01 1.16393363e+00 7.75209591e-02 6.54825449e-01 1.88470334e-01 2.58749902e-01 -8.24954689e-01 9.83404666e-02 4.61443841e-01 1.17063887e-01 -1.30854332e+00 2.12590888e-01 -6.84654474e-01 -2.14479268e-01 8.92232537e-01 3.75138313e-01 -9.47511569e-02 9.33102071e-01 2.62942940e-01 -1.06042296e-01 -1.72390983e-01 -9.63845253e-02 -4.53063026e-02 3.15524906e-01 8.32899570e-01 4.84971255e-01 2.31499955e-01 -7.29708135e-01 3.22717667e-01 -3.80960733e-01 -3.58676940e-01 5.01062930e-01 1.07912099e+00 -6.27626061e-01 -1.75785363e+00 -3.57062340e-01 1.05918992e+00 -1.14687431e+00 1.68354455e-02 -1.02988422e+00 8.86651218e-01 3.42790455e-01 1.25500596e+00 -2.29643330e-01 -9.55796063e-01 -2.04488620e-01 1.59133393e-02 2.35825315e-01 -7.58107603e-01 -8.51458490e-01 1.16975084e-02 6.97116032e-02 -4.39075142e-01 -1.49681270e-01 -8.16294730e-01 -1.21898055e+00 -2.36038640e-01 -4.03094202e-01 2.07557037e-01 1.11155534e+00 8.45047951e-01 8.20970953e-01 2.04656914e-01 5.51596522e-01 -6.80295825e-01 -1.75411135e-01 -8.32132161e-01 -2.81092571e-03 -1.07143469e-01 2.60124236e-01 -7.52421379e-01 -3.80332828e-01 -9.68087018e-02]
[14.423629760742188, 9.68017578125]
ea27ab7f-3709-457e-94ea-5276d56778f2
advancing-pico-element-detection-in-medical
1810.12780
null
https://arxiv.org/abs/1810.12780v4
https://arxiv.org/pdf/1810.12780v4.pdf
Advancing PICO Element Detection in Biomedical Text via Deep Neural Networks
In evidence-based medicine (EBM), defining a clinical question in terms of the specific patient problem aids the physicians to efficiently identify appropriate resources and search for the best available evidence for medical treatment. In order to formulate a well-defined, focused clinical question, a framework called PICO is widely used, which identifies the sentences in a given medical text that belong to the four components typically reported in clinical trials: Participants/Problem (P), Intervention (I), Comparison (C) and Outcome (O). In this work, we propose a novel deep learning model for recognizing PICO elements in biomedical abstracts. Based on the previous state-of-the-art bidirectional long-short term memory (biLSTM) plus conditional random field (CRF) architecture, we add another layer of biLSTM upon the sentence representation vectors so that the contextual information from surrounding sentences can be gathered to help infer the interpretation of the current one. In addition, we propose two methods to further generalize and improve the model: adversarial training and unsupervised pre-training over large corpora. We tested our proposed approach over two benchmark datasets. One is the PubMed-PICO dataset, where our best results outperform the previous best by 5.5%, 7.9%, and 5.8% for P, I, and O elements in terms of F1 score, respectively. And for the other dataset named NICTA-PIBOSO, the improvements for P/I/O elements are 2.4%, 13.6%, and 1.0% in F1 score, respectively. Overall, our proposed deep learning model can obtain unprecedented PICO element detection accuracy while avoiding the need for any manual feature selection.
['Peter Szolovits', 'Di Jin']
2018-10-30
null
null
null
null
['pico']
['natural-language-processing']
[ 5.46413183e-01 2.08912604e-02 -2.88193703e-01 -2.72403389e-01 -1.27340627e+00 -3.30709994e-01 5.47039509e-01 9.15273786e-01 -6.49989665e-01 9.71738815e-01 3.77364457e-01 -5.65651953e-01 -1.02504149e-01 -6.24364316e-01 -9.48640764e-01 -7.31539726e-01 1.49910077e-01 3.82502317e-01 7.11605623e-02 1.30055755e-01 3.16390753e-01 2.44528934e-01 -8.97776484e-01 6.53226912e-01 1.02749813e+00 9.18077528e-01 3.72158915e-01 5.41915894e-01 -1.74914330e-01 7.06792057e-01 -7.33071089e-01 -2.46574998e-01 -5.54586411e-01 -4.33247417e-01 -9.62630451e-01 -2.82008588e-01 -1.45729128e-02 -1.23540290e-01 -1.86972678e-01 9.81004000e-01 7.25407124e-01 9.45610702e-02 7.31586337e-01 -6.59790576e-01 -4.91171092e-01 6.90067947e-01 -4.89534855e-01 2.40307853e-01 3.95743906e-01 1.28869534e-01 6.85782015e-01 -7.66086698e-01 8.15771341e-01 9.56716061e-01 5.35600007e-01 5.66268742e-01 -8.48228872e-01 -6.21131063e-01 -1.44505706e-02 2.26555467e-01 -1.14689732e+00 -3.36850494e-01 5.05863667e-01 -5.23318887e-01 9.48329687e-01 3.86875421e-01 2.35366046e-01 1.37237585e+00 8.37584913e-01 6.31297469e-01 9.18910503e-01 -5.24894357e-01 2.94107050e-01 -2.19563078e-02 2.79110312e-01 5.50229728e-01 1.07378803e-01 -2.56913602e-01 -1.11166172e-01 -3.78692448e-01 2.27576941e-01 2.17921987e-01 -3.07562679e-01 5.18576741e-01 -1.10301304e+00 6.63717926e-01 4.66990590e-01 5.93393743e-01 -5.12844324e-01 -6.03790171e-02 6.65173471e-01 -3.78662020e-01 3.84978801e-01 2.99129993e-01 -6.33807182e-01 -7.44989216e-02 -9.57851112e-01 1.95998445e-01 5.70411861e-01 7.20211744e-01 2.28822842e-01 -5.30930042e-01 -6.57263041e-01 8.63389969e-01 3.54452968e-01 2.14851692e-01 8.02430630e-01 -5.06320238e-01 7.73258090e-01 6.55571640e-01 6.31507412e-02 -9.39168811e-01 -5.83214045e-01 -4.85025764e-01 -1.08220100e+00 -5.05009115e-01 3.05158552e-02 -1.14345312e-01 -1.18347502e+00 1.75570142e+00 2.94899493e-01 3.16840053e-01 3.04143131e-01 5.41442633e-01 1.32725966e+00 8.26119661e-01 4.60507482e-01 -2.59130567e-01 1.76593530e+00 -8.39112461e-01 -1.06862187e+00 -3.11090946e-01 9.33282852e-01 -9.66037869e-01 7.60509908e-01 1.57374978e-01 -8.70075881e-01 -2.91638464e-01 -9.77452815e-01 -2.12911472e-01 -3.58068109e-01 1.93162948e-01 2.13608712e-01 1.22503251e-01 -5.69858789e-01 5.20856917e-01 -7.71375120e-01 -1.40995443e-01 5.92966378e-01 2.46038958e-01 -3.72081399e-01 -1.67805493e-01 -1.56400919e+00 9.54761147e-01 5.17214656e-01 3.10956568e-01 -9.95923460e-01 -7.99330294e-01 -7.21963406e-01 1.07759252e-01 3.69491190e-01 -8.63569379e-01 1.05980861e+00 -2.88123369e-01 -1.20078671e+00 8.01300287e-01 -3.56889457e-01 -4.35065091e-01 1.60214365e-01 -3.85038406e-01 -5.13551235e-01 3.15632790e-01 1.91354796e-01 3.97879124e-01 2.68565595e-01 -9.07349229e-01 -4.01743323e-01 -5.70575297e-01 -9.17990208e-02 -8.13312456e-02 -1.84803903e-01 2.26064458e-01 -4.82024699e-01 -7.77912617e-01 -2.90532827e-01 -8.48426342e-01 -4.49410826e-01 -2.89817750e-01 -8.14088345e-01 -4.37957734e-01 3.03316504e-01 -1.04453409e+00 1.49851418e+00 -1.99074864e+00 5.54617075e-03 -1.39573947e-01 3.34572464e-01 5.12608230e-01 1.48728609e-01 4.69026953e-01 -1.04717925e-01 4.15187240e-01 -4.68228400e-01 -2.03721985e-01 -2.84566164e-01 5.59106432e-02 -4.12513223e-03 2.80533671e-01 3.11147898e-01 8.78122091e-01 -8.53991747e-01 -8.75099003e-01 -7.59412572e-02 5.35872519e-01 -4.06841874e-01 2.98222780e-01 -2.34332129e-01 6.45978153e-01 -6.86262310e-01 4.19842362e-01 5.51940143e-01 -4.95406270e-01 2.86867887e-01 -9.65489298e-02 4.82676551e-02 7.73908198e-01 -7.27409959e-01 1.82353699e+00 -5.67785144e-01 2.82734007e-01 -8.54694843e-02 -1.11144769e+00 7.13743448e-01 7.39627540e-01 4.37543750e-01 -5.39965034e-01 1.69217438e-01 3.41101140e-01 1.33157372e-01 -9.08076286e-01 1.45943791e-01 -2.49652475e-01 -1.28735200e-01 5.61242215e-02 1.16413675e-01 4.23018932e-01 -1.68393459e-03 2.01403797e-01 1.13066947e+00 -1.60068944e-01 5.79700351e-01 -1.76880181e-01 7.82913506e-01 -2.41947845e-01 7.97601700e-01 6.61731303e-01 -1.43523902e-01 5.25549114e-01 5.55555820e-01 -1.64703548e-01 -6.75201178e-01 -6.94252074e-01 -5.54605246e-01 4.39061046e-01 -1.52823806e-01 -3.62606913e-01 -7.75673449e-01 -7.09136903e-01 -3.72580111e-01 7.87778497e-01 -7.40169466e-01 -2.12199688e-01 -6.51608586e-01 -1.00285459e+00 6.39695764e-01 5.69008410e-01 3.46581817e-01 -1.28359509e+00 -5.02600074e-01 3.60149622e-01 -4.08718199e-01 -1.12767041e+00 -7.16302872e-01 8.63594934e-02 -8.13998878e-01 -1.05847597e+00 -8.47510934e-01 -9.46987450e-01 5.09758830e-01 -4.34602767e-01 9.38642263e-01 -1.26713246e-01 -3.34890038e-01 -1.65308759e-01 -3.09384644e-01 -3.77958894e-01 -5.42287052e-01 -4.49478850e-02 -2.94054180e-01 -5.75543158e-02 3.13074619e-01 -3.34028840e-01 -9.84575689e-01 -4.89641614e-02 -1.07491004e+00 1.20895647e-01 8.06359828e-01 1.08175945e+00 8.01263630e-01 -3.74739319e-01 7.60264456e-01 -1.20283437e+00 5.23557961e-01 -7.97621787e-01 -1.47199288e-01 4.60215449e-01 -6.26431346e-01 9.46500376e-02 7.21471667e-01 -4.33151454e-01 -9.79103982e-01 -1.78718016e-01 -5.41272879e-01 -5.33851534e-02 -2.42766917e-01 1.01682508e+00 -2.56940037e-01 4.41291749e-01 5.49426198e-01 2.09462583e-01 -3.16111267e-01 -5.85180283e-01 8.98630619e-02 9.16748583e-01 3.57292622e-01 -3.58688146e-01 1.19074732e-01 1.29528016e-01 -3.93535271e-02 -5.30180275e-01 -9.15862799e-01 -3.64536703e-01 -1.99838251e-01 2.50287235e-01 1.11546659e+00 -7.21304059e-01 -7.30445683e-01 1.82691738e-01 -1.45155966e+00 1.21983632e-01 8.80225822e-02 6.59060776e-01 -1.71514899e-01 5.37234247e-01 -7.19304919e-01 -4.80469882e-01 -9.36223269e-01 -1.59969151e+00 1.08405328e+00 1.76991835e-01 -3.44246536e-01 -1.02500510e+00 1.42674297e-01 5.32813072e-01 1.29334018e-01 5.09425461e-01 1.26579082e+00 -1.05282021e+00 -8.79668146e-02 -3.79891336e-01 -1.24729061e-02 2.56188333e-01 1.80541009e-01 -2.47636512e-01 -7.27915347e-01 -4.66359593e-02 2.64800906e-01 -1.50939673e-01 8.31063509e-01 5.07602394e-01 1.47797489e+00 -4.62955177e-01 -6.78991973e-01 3.41454744e-01 1.26004195e+00 6.15832984e-01 6.76819623e-01 1.63285404e-01 6.00616753e-01 3.51876169e-01 5.19053280e-01 2.73252249e-01 2.31748357e-01 6.43292189e-01 1.64525568e-01 -8.69174600e-02 -6.53118119e-02 -1.70915052e-01 2.20232263e-01 9.54380512e-01 1.94370627e-01 -4.01959807e-01 -1.03721964e+00 6.56923175e-01 -1.74495959e+00 -5.97058773e-01 -1.25007004e-01 2.00115657e+00 1.31573021e+00 3.18067729e-01 -4.25933838e-01 -5.16785681e-02 7.94999897e-01 7.09072426e-02 -4.34999198e-01 -4.71956134e-01 1.65033653e-01 5.01484215e-01 2.83143908e-01 3.38178754e-01 -1.16441166e+00 4.76548284e-01 5.42565584e+00 1.05924213e+00 -1.16962993e+00 2.42769182e-01 9.99626160e-01 1.11044452e-01 -2.50409603e-01 -1.29830003e-01 -8.44340205e-01 7.22301006e-01 1.35359931e+00 3.93781066e-03 -2.70389616e-01 4.44896609e-01 4.55426127e-01 -5.12049720e-02 -1.07056940e+00 7.78462231e-01 1.88860286e-03 -1.62969720e+00 5.87258562e-02 -6.44631833e-02 5.35317659e-01 -1.59262955e-01 -1.59100685e-02 1.44458562e-01 -1.44168273e-01 -1.18940687e+00 2.13383675e-01 6.98681116e-01 8.14423919e-01 -4.58816648e-01 1.12012911e+00 4.66895372e-01 -8.59131217e-01 2.58813381e-01 -9.34086591e-02 3.46764773e-01 1.14054196e-01 8.29673767e-01 -1.07346535e+00 1.03384864e+00 5.63267529e-01 5.85612118e-01 -3.20569605e-01 9.33977008e-01 -2.28117272e-01 9.30903971e-01 -5.55945411e-02 -3.12087387e-01 2.97295958e-01 6.80580884e-02 4.16288793e-01 1.31885457e+00 2.44330198e-01 1.37507826e-01 9.69043225e-02 7.14331925e-01 -5.58208168e-01 3.88080180e-01 -2.47913241e-01 -1.00443020e-01 3.42361808e-01 1.14124691e+00 -5.57589471e-01 -4.95497465e-01 -2.99837053e-01 6.75526321e-01 1.95018411e-01 1.46725699e-01 -8.39381516e-01 -6.19637251e-01 1.34245083e-01 5.24442531e-02 1.66076481e-01 3.47401202e-01 -2.37731412e-01 -9.57742393e-01 2.11606055e-01 -9.08266068e-01 4.84645396e-01 -4.78599221e-01 -1.10717285e+00 7.28497624e-01 -2.01523304e-01 -1.12372541e+00 -1.56918466e-01 -4.83880907e-01 -4.51217294e-01 1.18450963e+00 -1.32380211e+00 -1.02620959e+00 8.84528682e-02 1.78643882e-01 5.14055610e-01 -1.89292291e-03 1.07369375e+00 5.66037476e-01 -7.88955271e-01 6.78213418e-01 3.39237094e-01 3.24679077e-01 8.71728361e-01 -1.08869636e+00 5.98535053e-02 6.22080266e-01 -1.20118730e-01 9.01483715e-01 5.03989637e-01 -6.67997479e-01 -1.07693040e+00 -1.27354968e+00 1.39048338e+00 -3.83876204e-01 4.94641662e-01 -1.05954416e-01 -1.07900846e+00 4.99970436e-01 1.55765280e-01 -2.63202302e-02 9.23016191e-01 -7.32717589e-02 -1.66013345e-01 -6.31160066e-02 -1.14269876e+00 5.35061598e-01 5.73195875e-01 -3.53989363e-01 -7.98878074e-01 6.25054359e-01 1.03780377e+00 -5.63596964e-01 -1.31304801e+00 5.64730883e-01 3.14200073e-01 -2.73879737e-01 8.58440638e-01 -9.15598691e-01 9.91684973e-01 -2.11466569e-02 1.67955533e-02 -1.02052248e+00 -1.29609391e-01 -5.02087593e-01 8.71887710e-03 1.20421052e+00 7.20571995e-01 -5.49455583e-01 4.05893296e-01 3.46992910e-01 -4.79267031e-01 -1.33293891e+00 -1.10103488e+00 -1.92585647e-01 3.75312746e-01 -3.26892465e-01 4.51621026e-01 8.70472848e-01 8.56040977e-03 5.07382631e-01 -1.81121305e-01 6.22758791e-02 2.54241228e-01 1.79368973e-01 9.54255238e-02 -9.66533899e-01 -2.49463916e-01 -2.56206125e-01 -1.61829770e-01 -8.68999302e-01 7.49302432e-02 -1.14733136e+00 1.29952490e-01 -1.88336337e+00 5.51906884e-01 -2.26323530e-01 -7.12087810e-01 4.78146613e-01 -6.55051172e-01 -2.09962219e-01 -1.52380615e-01 5.03839040e-03 -4.39311087e-01 4.67161059e-01 1.27413893e+00 -4.04174268e-01 -8.41417685e-02 1.05818398e-02 -7.53126800e-01 5.20954072e-01 7.09084392e-01 -8.09963465e-01 -2.23065391e-02 -4.06196356e-01 5.78403613e-03 3.93625319e-01 1.52781546e-01 -5.97161174e-01 1.90607488e-01 -7.91745558e-02 3.32018018e-01 -5.95534682e-01 5.17393872e-02 -4.24164385e-01 -9.88977328e-02 8.26032937e-01 -7.98089087e-01 -7.34561086e-02 3.63022476e-01 6.34414613e-01 -3.31216604e-01 -4.59256142e-01 5.55054009e-01 -8.26294199e-02 -2.06812635e-01 2.53325105e-01 -3.68208498e-01 2.29406923e-01 8.34479690e-01 4.34968889e-01 -4.51631218e-01 -2.80248765e-02 -7.71592438e-01 3.80186409e-01 -2.45533004e-01 3.18117261e-01 8.12816858e-01 -1.06237388e+00 -9.10663188e-01 -3.78657132e-01 1.98416367e-01 1.54574946e-01 6.18234098e-01 9.84695852e-01 -5.18954933e-01 7.43787885e-01 2.50803530e-01 -4.51329112e-01 -1.19541013e+00 7.12350190e-01 1.41981706e-01 -9.41927433e-01 -5.98763049e-01 7.10105956e-01 3.27985495e-01 -3.63966584e-01 1.06880322e-01 -5.76819181e-01 -4.79531467e-01 -1.40563771e-01 5.16443849e-01 7.87457526e-02 4.20868367e-01 -4.62472051e-01 -6.97959423e-01 3.55739444e-01 -4.11159933e-01 2.92775687e-02 1.36302531e+00 3.85779947e-01 -2.86637574e-01 4.22012508e-01 1.46671724e+00 1.37793296e-03 -3.89898479e-01 -1.94465443e-01 1.53875172e-01 2.42867440e-01 8.45591649e-02 -1.07660592e+00 -7.90686548e-01 1.07410669e+00 6.85733557e-01 -1.35453701e-01 1.14450276e+00 6.78380653e-02 1.07042003e+00 6.09420091e-02 -1.15061074e-01 -8.00103903e-01 -2.16641158e-01 3.17304075e-01 9.49517250e-01 -1.05351198e+00 -2.13566460e-02 -2.92610079e-01 -4.78441745e-01 9.47165728e-01 2.57796198e-01 1.02309488e-01 7.61411130e-01 1.70397386e-01 -1.53561890e-01 -2.71117955e-01 -8.12352598e-01 3.32663417e-01 4.48509991e-01 1.73664149e-02 7.21812844e-01 1.50834337e-01 -8.62486541e-01 1.11885202e+00 2.37934873e-01 1.92747742e-01 2.56776750e-01 7.97239542e-01 -1.23499073e-01 -1.20124614e+00 -2.89606482e-01 7.13956416e-01 -1.06206286e+00 -3.92303377e-01 -4.56473008e-02 4.45716649e-01 1.06696270e-01 9.22657311e-01 -1.94307417e-01 -2.56209165e-01 3.74489546e-01 2.41224036e-01 1.26831204e-01 -6.70727670e-01 -8.96655798e-01 2.67366558e-01 1.09203830e-01 -3.04058075e-01 -5.75890422e-01 -3.85393828e-01 -1.44884741e+00 2.60745674e-01 -3.31600457e-01 4.76726443e-01 5.88205993e-01 1.21657670e+00 6.85213029e-01 1.04069686e+00 3.70761454e-01 -1.03476733e-01 -5.19263864e-01 -1.33459318e+00 -7.75576755e-02 4.34043676e-01 3.27294171e-01 -4.75557089e-01 -1.69280112e-01 9.53577831e-02]
[8.525064468383789, 8.707735061645508]
593aaf33-ca49-4985-b1b7-94003223c2bc
semantic-answer-type-prediction-task-smart-at
2012.00555
null
https://arxiv.org/abs/2012.00555v1
https://arxiv.org/pdf/2012.00555v1.pdf
SeMantic AnsweR Type prediction task (SMART) at ISWC 2020 Semantic Web Challenge
Each year the International Semantic Web Conference accepts a set of Semantic Web Challenges to establish competitions that will advance the state of the art solutions in any given problem domain. The SeMantic AnsweR Type prediction task (SMART) was part of ISWC 2020 challenges. Question type and answer type prediction can play a key role in knowledge base question answering systems providing insights that are helpful to generate correct queries or rank the answer candidates. More concretely, given a question in natural language, the task of SMART challenge is, to predict the answer type using a target ontology (e.g., DBpedia or Wikidata).
['Ricardo Usbeck', 'Axel-Cyrille Ngonga Ngomo', 'Jens Lehmann', 'Alfio Gliozzo', 'Mohnish Dubey', 'Nandana Mihindukulasooriya']
2020-12-01
null
null
null
null
['type-prediction', 'knowledge-base-question-answering']
['computer-code', 'natural-language-processing']
[-5.34227863e-02 7.16864109e-01 -7.12143257e-02 -4.29566085e-01 -8.06122065e-01 -7.10233808e-01 6.66842341e-01 5.40800512e-01 -4.18994933e-01 8.66830945e-01 5.12433112e-01 -2.78636754e-01 -4.77425218e-01 -1.26032400e+00 -4.91241366e-01 5.06252885e-01 5.43657601e-01 9.78047311e-01 1.08009028e+00 -9.79066789e-01 5.72882891e-01 -2.63238579e-01 -2.08026457e+00 1.11339366e+00 1.01395583e+00 1.41874886e+00 1.18478775e-01 6.00845814e-01 -1.14315438e+00 1.33746123e+00 -6.05512559e-01 -6.50825024e-01 -2.16340378e-01 -2.00782657e-01 -2.14511800e+00 -8.15654337e-01 5.19940138e-01 3.54653507e-01 2.68654823e-02 9.91836727e-01 1.94262549e-01 2.77116179e-01 2.67266035e-01 -1.32135296e+00 -6.16305113e-01 3.45709473e-01 6.02007806e-01 2.56742746e-01 1.07843959e+00 -6.05973244e-01 1.51316762e+00 -6.29785061e-01 1.02775490e+00 1.40093744e+00 3.43985498e-01 7.93941498e-01 -7.82780170e-01 3.34671419e-03 -1.66279674e-01 1.19211471e+00 -1.01377189e+00 -1.97156206e-01 4.09951717e-01 -2.53862977e-01 1.02775455e+00 7.54891872e-01 1.87345147e-01 3.74931335e-01 -5.15924335e-01 6.84797883e-01 9.53729510e-01 -5.96881211e-01 4.90863621e-01 3.21035117e-01 7.49687016e-01 6.56437814e-01 8.92690048e-02 -5.94717622e-01 -5.55039525e-01 -4.30150062e-01 -3.01899791e-01 -3.96809906e-01 -1.61053076e-01 7.17509016e-02 -6.44339740e-01 1.05051875e+00 6.50965035e-01 3.54519546e-01 -2.76718974e-01 -1.01014733e-01 2.70524353e-01 5.68782210e-01 1.31751135e-01 1.43176734e+00 -8.28526437e-01 -1.09339342e-01 -2.43151158e-01 1.06134593e+00 1.45285022e+00 8.85869086e-01 9.18435335e-01 -8.93754005e-01 -3.32581162e-01 1.04347110e+00 3.14065576e-01 1.64358541e-01 4.55183119e-01 -1.48232949e+00 5.45595229e-01 1.48231637e+00 5.60123622e-01 -7.75826931e-01 -5.10368645e-01 8.33323449e-02 1.74097762e-01 -3.81099790e-01 5.66966891e-01 -1.59160141e-02 -4.50023055e-01 1.24098706e+00 7.56339371e-01 -5.63599914e-02 4.32478666e-01 7.72143841e-01 1.74005163e+00 6.69643044e-01 3.38362381e-02 3.18952352e-01 1.89497983e+00 -7.86575317e-01 -7.64118791e-01 -5.46128392e-01 7.40995526e-01 -7.15565801e-01 8.48357022e-01 -5.62229678e-02 -5.90736866e-01 -3.19610059e-01 -8.19081306e-01 -4.15129215e-01 -1.09503686e+00 -5.43258905e-01 2.34967828e-01 1.69486880e-01 -7.05296934e-01 9.92475823e-02 1.43799499e-01 -7.69229650e-01 7.58256018e-02 -4.37848531e-02 -3.23510319e-02 -6.16023064e-01 -1.89848495e+00 1.31569433e+00 6.34809554e-01 -5.88110030e-01 -2.24972829e-01 -1.05499363e+00 -6.01204097e-01 -3.57769690e-02 6.89830124e-01 -8.97060812e-01 1.38240182e+00 -5.95255613e-01 -9.34940755e-01 1.07889736e+00 -3.14153969e-01 -8.12508821e-01 -2.73007423e-01 -2.82484572e-02 -9.67060864e-01 3.53983611e-01 4.58389759e-01 7.73569822e-01 2.60854512e-01 -9.36564505e-01 -1.27816868e+00 -5.33864796e-01 8.38631034e-01 1.12110361e-01 -2.10279524e-01 4.08297896e-01 -3.19089830e-01 1.67611420e-01 -1.01122253e-01 -4.54481453e-01 9.72256996e-03 -3.31047118e-01 5.05748689e-02 -9.93251085e-01 8.12512755e-01 -1.03576696e+00 1.23310339e+00 -1.35791564e+00 -1.57547385e-01 2.12052707e-02 2.66997129e-01 3.26774418e-01 -2.22104397e-02 4.88721520e-01 3.45110029e-01 2.11749926e-01 2.38573492e-01 8.73774350e-01 2.07612291e-01 3.18811864e-01 -3.68952721e-01 -7.74281085e-01 -1.10859223e-01 9.23848569e-01 -1.06587505e+00 -3.03404331e-01 -3.50708216e-01 -2.45448381e-01 -6.40750885e-01 3.97243559e-01 -1.26444280e+00 3.95541452e-02 -8.45812261e-01 3.86169404e-01 4.86784540e-02 -4.51060593e-01 3.33944745e-02 -3.13984960e-01 2.21871659e-01 9.07783568e-01 -9.94726300e-01 1.50459015e+00 -8.18930984e-01 3.65311265e-01 -9.58100036e-02 -8.93036962e-01 1.12166345e+00 4.60399568e-01 3.71978402e-01 -1.17081296e+00 -3.64454627e-01 7.97174990e-01 -1.78855896e-01 -1.19192183e+00 6.83803737e-01 -1.93479687e-01 -2.08186194e-01 1.05778433e-01 1.06575467e-01 -3.33582252e-01 5.08033395e-01 1.34038106e-01 1.51110792e+00 1.05817601e-01 2.76749492e-01 -4.97314274e-01 9.27163720e-01 7.78467059e-01 3.21120143e-01 6.97063088e-01 1.63869500e-01 2.73742318e-01 2.49482036e-01 -7.75034904e-01 -6.84545338e-01 -6.45221949e-01 6.30245283e-02 1.14838731e+00 2.62783378e-01 -6.98875546e-01 -7.54451632e-01 -9.76792812e-01 1.84089795e-01 1.00589287e+00 -4.01707351e-01 -1.26821607e-01 -5.78421891e-01 -1.85337409e-01 5.64748526e-01 1.32990494e-01 5.31821012e-01 -1.23433888e+00 -6.51264489e-01 3.87405097e-01 -1.15397477e+00 -1.32813132e+00 2.58847475e-01 -1.69468716e-01 -4.54705060e-01 -1.75846744e+00 -3.77813667e-01 -9.16107655e-01 1.29122928e-01 -4.09037955e-02 1.72972226e+00 2.80700505e-01 -1.03152379e-01 7.81715393e-01 -8.84365737e-01 -6.78546190e-01 -3.65718633e-01 2.84659147e-01 -5.86766839e-01 -8.99682641e-02 1.01762354e+00 -4.29866500e-02 -7.14195848e-01 4.40435439e-01 -9.28362131e-01 -2.78405309e-01 -4.52653289e-01 3.91404033e-01 2.77628511e-01 -8.87204409e-02 1.14689958e+00 -1.00598991e+00 8.78212988e-01 -8.53265703e-01 -5.73217869e-01 9.65850949e-01 -7.79905796e-01 3.35899651e-01 5.29323220e-01 5.07078767e-01 -1.35756648e+00 -5.51756263e-01 -4.84131962e-01 6.41574681e-01 -1.03549764e-01 8.55030596e-01 -1.87523529e-01 4.05605696e-02 1.03557920e+00 -2.02413529e-01 -2.66633779e-01 -7.93364167e-01 4.59434807e-01 8.33518386e-01 3.56837898e-01 -8.65639925e-01 4.18448806e-01 2.68205971e-01 -2.22216979e-01 -5.12963533e-01 -1.52960622e+00 -1.05513287e+00 -5.50599769e-02 -2.95187742e-01 1.12554812e+00 -4.33992267e-01 -7.95165300e-01 -1.41627625e-01 -1.29526412e+00 3.87416184e-02 -4.47511017e-01 -3.28550726e-01 -5.75727284e-01 -1.95392221e-01 4.63601798e-02 -3.03489298e-01 -4.77746636e-01 -4.35349166e-01 7.25948215e-01 4.54103142e-01 -5.34998238e-01 -1.16079199e+00 1.92887768e-01 1.54848659e+00 6.41286969e-01 -1.41890332e-01 1.49703515e+00 -1.41286802e+00 -3.53931338e-01 -5.03062248e-01 -2.44018734e-01 1.59595668e-01 -1.04719438e-01 -6.64922178e-01 -6.61512315e-01 5.60190916e-01 -1.14756674e-01 -4.11325783e-01 3.73941183e-01 -3.05225670e-01 1.03926957e+00 -7.92115927e-01 -5.17099053e-02 -2.91076779e-01 1.47219944e+00 1.19318686e-01 7.93447018e-01 7.16136515e-01 1.92203403e-01 1.21079838e+00 8.71472776e-01 6.69039190e-02 1.11480641e+00 1.07874084e+00 3.62649679e-01 6.60511911e-01 -4.38774943e-01 -4.31353420e-01 -4.27687913e-01 3.76183212e-01 1.68353796e-01 3.89825068e-02 -1.33329165e+00 7.15193272e-01 -1.95607257e+00 -1.09013176e+00 -4.82285619e-01 2.16137671e+00 1.09844923e+00 -2.96566486e-01 8.82820934e-02 -2.82819755e-02 4.32638198e-01 -2.22050101e-01 -3.19630861e-01 -3.38880271e-01 -9.49877799e-02 7.03265965e-01 1.17844582e-01 6.87323093e-01 -6.03011131e-01 9.21153724e-01 5.29559994e+00 8.13612759e-01 -3.24173272e-01 2.51932323e-01 1.26416460e-01 4.10435736e-01 -6.94571733e-01 3.71871471e-01 -1.13787365e+00 2.49463648e-01 9.86190021e-01 -7.81972051e-01 5.42246878e-01 6.08537972e-01 -3.28268617e-01 -1.97104707e-01 -9.75190580e-01 5.36239147e-01 2.35274062e-01 -2.00014853e+00 1.17252998e-01 -3.49910200e-01 5.27963758e-01 -1.30156562e-01 -5.84416270e-01 6.81101739e-01 5.94021618e-01 -9.51562464e-01 3.07724535e-01 7.80244887e-01 1.12710513e-01 -4.03807223e-01 8.46594930e-01 4.20321584e-01 -1.04380643e+00 -4.15653020e-01 -3.63564521e-01 -7.83956274e-02 1.24454275e-01 2.20145762e-01 -8.32962990e-01 6.54617906e-01 9.34948385e-01 2.07136840e-01 -7.10733593e-01 1.25479054e+00 -5.08557439e-01 4.66866404e-01 -2.34523937e-01 -3.74359280e-01 1.65053442e-01 2.74285704e-01 4.73831981e-01 6.11206293e-01 9.72869620e-02 4.12351131e-01 -3.14101093e-02 6.68192208e-01 -4.15294826e-01 3.18119407e-01 -2.66900361e-01 2.18807608e-01 6.79775298e-01 1.06449473e+00 -8.35005268e-02 -4.99750763e-01 -5.10471404e-01 5.39227664e-01 1.84505075e-01 3.61933142e-01 -2.53781855e-01 -6.20965004e-01 6.60885513e-01 2.93640316e-01 1.24372400e-01 5.71389318e-01 -1.02930576e-01 -1.10486388e+00 1.78333074e-01 -1.24646282e+00 1.17936683e+00 -9.58243370e-01 -1.35160923e+00 2.36973718e-01 -1.66119680e-01 -8.01896751e-01 -4.04091179e-01 -6.22112811e-01 -3.61091763e-01 8.52177083e-01 -1.88129795e+00 -9.10177410e-01 -3.74778926e-01 5.43077409e-01 5.51781416e-01 -1.82181336e-02 1.21847582e+00 5.41699409e-01 1.63473040e-01 4.32019122e-02 -3.40492129e-01 -1.55702757e-03 5.11200070e-01 -1.30709100e+00 7.69263953e-02 3.34131807e-01 1.86405331e-01 2.77006149e-01 9.08088446e-01 -5.06821632e-01 -1.38264740e+00 -1.05222380e+00 1.65664601e+00 -8.68901014e-01 7.70461321e-01 1.89384177e-01 -9.80961978e-01 2.20945731e-01 1.51121050e-01 -6.16477542e-02 8.50893974e-01 3.40449244e-01 -5.57731092e-01 -2.94346750e-01 -1.38349187e+00 2.46797323e-01 6.31964207e-01 -6.66156232e-01 -1.25901699e+00 5.84294617e-01 9.22758579e-01 -2.27180198e-01 -1.06888521e+00 6.13609076e-01 3.24460924e-01 -6.80235326e-01 1.05293667e+00 -1.35418057e+00 5.10907829e-01 -4.66862023e-01 -6.02227032e-01 -1.05383837e+00 2.10991263e-01 -1.73985481e-01 -5.45641005e-01 1.13991702e+00 9.31685805e-01 -4.09653753e-01 9.80120242e-01 1.19272768e+00 8.89665354e-03 -6.34718776e-01 -1.03847837e+00 -7.40325809e-01 5.95496930e-02 -2.67254293e-01 7.81071901e-01 6.95256293e-01 2.96760142e-01 6.73906624e-01 2.86174357e-01 1.26445323e-01 2.37821624e-01 3.62782001e-01 5.50799787e-01 -1.65481555e+00 1.04619063e-01 -3.89758706e-01 -3.37265611e-01 -8.54146183e-01 5.42016886e-02 -1.20015323e+00 -1.34990931e-01 -2.32275343e+00 -2.34511390e-01 -4.60566193e-01 -6.28139600e-02 3.63547951e-01 -4.55152452e-01 -3.20118070e-02 7.90571645e-02 -1.43659294e-01 -1.18265128e+00 3.16400975e-01 9.60284352e-01 -2.81058222e-01 3.86063039e-01 2.75080919e-01 -8.93962204e-01 6.76502526e-01 7.20654607e-01 -4.68343049e-01 -5.10295689e-01 -4.67622906e-01 1.25255179e+00 1.85822189e-01 3.04042399e-01 -8.45316291e-01 4.65787053e-01 -3.95452052e-01 -5.41168153e-01 5.73307369e-03 2.05926880e-01 -8.31918240e-01 -3.35034132e-01 2.70760745e-01 -7.70303130e-01 -1.58268392e-01 -1.31334513e-01 1.72151119e-01 -8.03222775e-01 -9.07128930e-01 3.35099608e-01 -3.24278355e-01 -1.21039307e+00 9.12735164e-02 -1.42575577e-01 1.25351775e+00 6.50071025e-01 9.48879421e-02 -6.94586039e-01 -3.73706102e-01 -5.56385338e-01 6.61370993e-01 -1.19154319e-01 9.99985337e-01 5.13526857e-01 -1.36156094e+00 -7.02713192e-01 -6.94070399e-01 8.18553865e-01 -2.92221665e-01 1.68107450e-01 3.23816687e-01 -5.90569198e-01 8.76496017e-01 5.43375835e-02 1.71787560e-01 -1.11305106e+00 1.91768005e-01 5.52774966e-01 -5.72897196e-01 -1.40663177e-01 9.57583487e-01 -4.54399824e-01 -9.89140332e-01 1.09015286e-01 2.51008958e-01 -1.05353343e+00 1.40473098e-01 1.11000085e+00 5.58800876e-01 3.57794642e-01 2.02579554e-02 -4.65350062e-01 9.75761861e-02 2.94321686e-01 -3.00917849e-02 1.38792253e+00 3.61614823e-02 -5.16139567e-01 1.82795748e-01 1.01987803e+00 -2.99419880e-01 -1.40033007e-01 -7.10593939e-01 1.03257120e+00 -1.99370250e-01 -2.26221547e-01 -1.34106493e+00 -5.95034897e-01 6.05510116e-01 4.14960414e-01 8.10496986e-01 7.48645365e-01 6.32514834e-01 1.20471954e+00 8.03303301e-01 4.69685942e-01 -1.48299205e+00 6.74077794e-02 1.10760641e+00 1.12230742e+00 -1.17829859e+00 -4.73423719e-01 -4.37443078e-01 -2.39946321e-01 1.16256106e+00 8.24810803e-01 3.72652143e-01 5.71477354e-01 -2.37419531e-01 1.13842256e-01 -8.07679415e-01 -9.39455986e-01 -6.47576988e-01 7.77512312e-01 6.46138489e-01 4.88751739e-01 -1.59516960e-01 -4.38473284e-01 7.82757640e-01 -2.54549652e-01 1.02568462e-01 1.49938792e-01 8.58890057e-01 -9.72707331e-01 -1.46435452e+00 -1.84916735e-01 8.35901022e-01 -4.39864129e-01 -2.02417776e-01 -5.76652646e-01 6.42963201e-02 8.08240771e-02 1.59723699e+00 -2.18815640e-01 -2.90208369e-01 6.44256711e-01 5.17332792e-01 3.23902071e-01 -9.96503115e-01 -7.66366899e-01 -1.24473608e+00 9.23896253e-01 -5.50581932e-01 -3.38652581e-01 -2.24439532e-01 -1.39084589e+00 3.01217474e-03 -1.77363623e-02 8.04688156e-01 6.62852645e-01 1.46260154e+00 6.65735602e-01 2.18481272e-01 1.90594241e-01 6.88846290e-01 -2.89885730e-01 -7.23897040e-01 1.16197892e-01 7.34033346e-01 -2.07570970e-01 -4.28685129e-01 1.46693543e-01 2.00390480e-02]
[10.523064613342285, 7.895973205566406]
da079252-90ff-4174-9498-df938ba93401
a-comprehensive-evaluation-of-the-copy
2304.07772
null
https://arxiv.org/abs/2304.07772v2
https://arxiv.org/pdf/2304.07772v2.pdf
A Comprehensive Evaluation of the Copy Mechanism for Natural Language to SPARQL Query Generation
In recent years, the field of neural machine translation (NMT) for SPARQL query generation has witnessed a significant growth. Recently, the incorporation of the copy mechanism with traditional encoder-decoder architectures and the use of pre-trained encoder-decoders have set new performance benchmarks. This paper presents a large variety of experiments that replicate and expand upon recent NMT-based SPARQL generation studies, comparing pre-trained and non-pre-trained models, question annotation formats, and the use of a copy mechanism for non-pre-trained and pre-trained models. Our results show that either adding the copy mechanism or using a question annotation improves performances for nonpre-trained models and for pre-trained models, setting new baselines for three popular datasets.
['Papa Abdou Karim Karou Diallo', 'Amal Zouaq', 'Samuel Reyd']
2023-04-16
null
null
null
null
['nmt']
['computer-code']
[ 1.19868882e-01 5.54174364e-01 -1.39672354e-01 -5.86322069e-01 -1.32106376e+00 -5.08091807e-01 8.64761055e-01 1.76101655e-01 -5.97590268e-01 8.85205865e-01 4.85194623e-01 -4.50714648e-01 4.03809249e-01 -1.04377329e+00 -1.23839593e+00 4.19727862e-01 2.45786846e-01 1.07921267e+00 2.76333451e-01 -6.24925375e-01 -3.76501203e-01 -9.17917397e-03 -1.41434801e+00 9.04024303e-01 7.67514706e-01 8.34385455e-01 -4.41826805e-02 7.86640882e-01 -4.98436779e-01 9.61373866e-01 -6.27771974e-01 -1.00979614e+00 3.05852830e-01 -4.66813028e-01 -1.08219612e+00 -6.52419269e-01 8.59923720e-01 -2.60826468e-01 -3.81537348e-01 6.15412056e-01 7.05619812e-01 -1.78191692e-01 1.62201911e-01 -9.49606955e-01 -1.35022128e+00 8.76979589e-01 7.51665235e-02 -2.21016929e-02 6.22100413e-01 1.63984194e-01 1.23786247e+00 -7.01626182e-01 1.01793349e+00 1.32120180e+00 8.09382379e-01 8.18538785e-01 -1.46787691e+00 -2.55000442e-01 -5.98622501e-01 1.47091702e-01 -1.19783759e+00 -5.16096294e-01 3.11934620e-01 -2.86132321e-02 1.76720953e+00 3.54304612e-01 4.00485635e-01 1.08569992e+00 3.36291730e-01 6.12818658e-01 9.94687378e-01 -5.74128509e-01 -4.07134257e-02 6.24669075e-01 -2.23627120e-01 5.61134219e-01 3.58745217e-01 2.73156106e-01 -4.51982558e-01 -4.15476888e-01 4.04877305e-01 -5.08969009e-01 5.37307002e-02 -2.45925456e-01 -1.12658012e+00 8.91923606e-01 2.95057803e-01 1.75513163e-01 -4.41734403e-01 2.04110369e-01 5.26995182e-01 5.56241691e-01 4.61462140e-01 1.03243089e+00 -8.05769324e-01 -1.01787321e-01 -1.15025091e+00 5.78461587e-01 1.23177481e+00 1.45903265e+00 7.31623054e-01 -7.94546213e-03 -8.82152200e-01 7.07901895e-01 7.05731437e-02 3.23864192e-01 4.93479252e-01 -8.42926621e-01 7.38224387e-01 8.84925544e-01 9.22241807e-02 -1.60324648e-01 -4.58881445e-02 -2.54519969e-01 -2.26792619e-01 -1.16067998e-01 2.27406934e-01 -2.94239312e-01 -1.05471265e+00 1.70882499e+00 8.93604085e-02 -3.44818860e-01 6.27237678e-01 5.81778526e-01 8.91469657e-01 7.10247755e-01 3.49366069e-01 1.48794919e-01 1.52200031e+00 -1.10201693e+00 -5.62596798e-01 -1.75019518e-01 7.67152488e-01 -8.71618509e-01 1.56528974e+00 -2.65265733e-01 -1.40787005e+00 -7.01616585e-01 -8.46378446e-01 -5.79770148e-01 -6.60842836e-01 4.64476906e-02 5.62676847e-01 6.10747218e-01 -1.56293797e+00 3.72527540e-01 -4.68171954e-01 -7.31984913e-01 2.30514333e-01 2.25690737e-01 -4.09046263e-01 -1.73859432e-01 -1.36768472e+00 1.15779221e+00 5.85828006e-01 -5.40024340e-01 -8.52176011e-01 -1.06279957e+00 -7.04454243e-01 4.83521037e-02 2.37361655e-01 -1.42737973e+00 1.72676587e+00 -9.33105469e-01 -1.37054431e+00 9.42764580e-01 -3.15676451e-01 -8.69872034e-01 2.89095193e-01 -3.17192435e-01 -5.54565251e-01 -1.63691223e-01 8.31266493e-02 9.97572839e-01 3.78359884e-01 -8.38739574e-01 -4.00914162e-01 -1.54046118e-02 2.86555707e-01 1.13151945e-01 -2.70587690e-02 3.81801367e-01 -2.39578545e-01 -5.43511033e-01 -6.35437310e-01 -8.59491348e-01 -5.55417947e-02 -3.30847740e-01 -4.58268106e-01 -3.79460335e-01 5.86098492e-01 -7.95372725e-01 1.06285477e+00 -1.52610922e+00 -1.29201218e-01 -2.36271545e-01 -1.94758967e-01 2.77018905e-01 -4.95955735e-01 8.83681238e-01 -9.72117856e-02 3.69727135e-01 -3.39306235e-01 -3.04279327e-01 2.01995030e-01 4.37573731e-01 -2.35999167e-01 -6.78830922e-01 6.98383749e-01 1.40521741e+00 -6.45076752e-01 -6.15208089e-01 -3.51116985e-01 2.90790021e-01 -8.55160773e-01 4.95943189e-01 -1.14795732e+00 6.50816932e-02 -7.12680668e-02 4.51576918e-01 2.68890202e-01 -3.15897495e-01 3.79054695e-02 -8.05353299e-02 9.94241759e-02 1.10567963e+00 -7.04057932e-01 1.93448555e+00 -4.86131012e-01 3.52540314e-01 -3.63197356e-01 -2.18725532e-01 7.97864556e-01 6.68537021e-01 2.23872393e-01 -1.22620308e+00 -3.83706480e-01 4.38257247e-01 -1.26246244e-01 -6.92299962e-01 9.36069250e-01 -2.67498285e-01 -5.62630072e-02 6.57918751e-01 5.40124178e-01 -1.62704006e-01 4.25502837e-01 2.15852037e-01 1.30155349e+00 3.25699598e-01 -4.83834073e-02 -4.30013128e-02 3.79054964e-01 5.56916595e-01 2.72050709e-01 6.26606047e-01 4.14177924e-01 5.96738577e-01 2.79608101e-01 -2.56090194e-01 -1.42569900e+00 -8.35243940e-01 8.15941468e-02 1.22819042e+00 -5.42297065e-01 -7.41738856e-01 -8.11762750e-01 -1.03348196e+00 2.50122130e-01 1.26390398e+00 -4.65653747e-01 -3.33341241e-01 -7.57097244e-01 -5.72095335e-01 1.12936854e+00 5.46798885e-01 3.77736658e-01 -1.17737925e+00 -5.61229825e-01 3.92247498e-01 -1.77227020e-01 -1.18092000e+00 -3.19599718e-01 7.04493597e-02 -9.69266236e-01 -6.62409008e-01 -4.14941877e-01 -6.65703237e-01 2.65043050e-01 -2.46105000e-01 1.93609416e+00 -1.11127764e-01 -6.83003142e-02 3.35296839e-01 -2.03649223e-01 -5.10068178e-01 -8.58113289e-01 5.96158266e-01 -5.46993196e-01 -5.90504646e-01 6.20391965e-01 -2.56108314e-01 -1.70039952e-01 -1.25752345e-01 -1.17357826e+00 3.74285698e-01 7.04153478e-01 6.73323870e-01 7.59396315e-01 -6.73587441e-01 7.70547032e-01 -1.21337616e+00 9.26542103e-01 -4.95632261e-01 -4.58204895e-01 6.75596952e-01 -1.16948593e+00 7.12164044e-01 5.39269686e-01 -1.39187798e-01 -1.31328177e+00 -2.66017497e-01 -3.30497503e-01 -3.16687852e-01 -3.21355872e-02 6.78137362e-01 -2.32017785e-01 2.14367896e-01 1.01721740e+00 3.24521028e-02 -1.03839502e-01 -6.25272214e-01 8.40190530e-01 4.36439514e-01 4.77198303e-01 -7.53680825e-01 7.59154379e-01 -2.96398997e-02 -2.69652188e-01 1.61481611e-02 -5.30306816e-01 1.89649593e-02 -3.51237655e-01 3.39990407e-01 8.40916395e-01 -9.75083411e-01 2.56715305e-02 -1.19416177e-01 -1.41979063e+00 -4.47009981e-01 -9.53139246e-01 1.85715094e-01 -4.33955908e-01 -1.05449378e-01 -7.22451866e-01 -3.88040185e-01 -8.51774752e-01 -1.07071054e+00 1.17285812e+00 2.01406106e-01 -3.51067126e-01 -1.05200171e+00 6.33715689e-01 4.87173796e-01 1.03797352e+00 1.50914704e-02 1.42620778e+00 -9.51573968e-01 -7.45457411e-01 -1.71676859e-01 -3.10550004e-01 4.50422645e-01 -1.20383896e-01 -4.86490071e-01 -8.88625681e-01 -7.35398307e-02 -3.98232043e-01 -5.21958411e-01 5.09595573e-01 -3.81317139e-01 5.52764595e-01 -5.45729876e-01 1.29166236e-02 5.39176166e-01 1.69359994e+00 -2.25843047e-03 7.23876536e-01 3.93451482e-01 4.57065821e-01 3.99043620e-01 8.93296897e-02 -3.90382372e-02 1.10532463e+00 7.59918511e-01 2.88685948e-01 -2.08273362e-02 -5.56426108e-01 -7.15383828e-01 4.04522419e-01 5.83534062e-01 1.32007822e-01 -2.99256027e-01 -9.71757948e-01 5.72388470e-01 -1.55147111e+00 -6.45380259e-01 -1.29291162e-01 2.18851471e+00 1.38264132e+00 -1.44247308e-01 -8.45109299e-02 -6.32964551e-01 1.79171592e-01 3.69546823e-02 -4.62239861e-01 -7.65515983e-01 -2.89709240e-01 1.07408869e+00 6.11340702e-01 3.82886410e-01 -6.98013484e-01 1.24417031e+00 7.17304277e+00 2.54837096e-01 -9.12183106e-01 4.71762210e-01 1.38627589e-01 -2.72562802e-01 -8.36682379e-01 3.89070034e-01 -9.22841847e-01 2.34858200e-01 1.60069013e+00 -5.10137498e-01 4.56470877e-01 8.67190480e-01 -2.46866927e-01 9.07792300e-02 -1.58492291e+00 4.44124699e-01 2.57806093e-01 -1.53707767e+00 6.62262022e-01 -1.45103842e-01 7.77291715e-01 4.34039235e-01 -2.89095640e-01 9.01965678e-01 4.77259934e-01 -1.08574581e+00 6.58717990e-01 9.22532737e-01 8.57677102e-01 -3.04863632e-01 8.83973539e-01 1.03741698e-01 -7.88324952e-01 1.66394323e-01 -4.23265189e-01 1.66118994e-01 1.21584259e-01 2.88840353e-01 -1.15954387e+00 9.00528550e-01 4.97154891e-01 1.02716506e-01 -1.06070149e+00 8.25510919e-01 -2.02102914e-01 5.34928620e-01 -2.89308339e-01 1.04390465e-01 -3.30955200e-02 1.98037118e-01 3.78367275e-01 1.31996834e+00 2.17944175e-01 -3.38023931e-01 -2.40896702e-01 1.14566576e+00 -5.21787345e-01 3.23145092e-01 -4.12657976e-01 -3.40702266e-01 2.68277317e-01 1.02003980e+00 1.65908292e-01 -6.61969960e-01 -6.99213803e-01 9.46181059e-01 4.49558318e-01 3.77596855e-01 -6.85663402e-01 -4.38603699e-01 5.78105807e-01 5.23403525e-01 1.66258991e-01 1.20751178e-02 -4.13471222e-01 -1.28143382e+00 3.30212146e-01 -1.36612093e+00 6.75540090e-01 -1.02861631e+00 -1.00345969e+00 8.82796466e-01 1.66908756e-01 -5.38612425e-01 -8.24411333e-01 -1.19157590e-01 -5.05437791e-01 1.22685039e+00 -1.81619024e+00 -1.40564311e+00 -8.35808069e-02 3.73132229e-01 2.28876382e-01 -3.57947022e-01 1.09892678e+00 6.71941161e-01 -3.53815645e-01 9.23007131e-01 -2.04982877e-01 1.10300012e-01 9.54552770e-01 -1.32498908e+00 7.98255563e-01 7.80280411e-01 2.55680740e-01 7.67026901e-01 4.88584489e-01 -7.60374606e-01 -1.60610926e+00 -1.33442485e+00 1.54440689e+00 -8.92595887e-01 3.20703626e-01 -4.61126268e-01 -1.11707830e+00 1.04011273e+00 8.22723448e-01 -4.02600437e-01 6.16898000e-01 8.37533101e-02 -6.23665810e-01 -2.91056305e-01 -1.08109033e+00 3.41964930e-01 8.30965340e-01 -9.46136892e-01 -8.29051316e-01 4.48098361e-01 1.14867258e+00 -5.48340976e-01 -1.18641508e+00 7.74013281e-01 3.74646515e-01 -6.57964230e-01 9.05667186e-01 -1.10100532e+00 6.17654741e-01 -1.28398955e-01 -2.13250056e-01 -1.15517807e+00 -3.82015072e-02 -5.72737634e-01 -3.32648546e-01 1.30438209e+00 9.83289421e-01 -5.61557353e-01 8.96516562e-01 9.20443475e-01 -3.85495216e-01 -8.48657489e-01 -9.10534143e-01 -5.59888601e-01 2.89423794e-01 -2.53345102e-01 1.05779338e+00 7.23272979e-01 -2.59813339e-01 7.67408073e-01 -6.62463009e-02 -1.67299911e-01 1.33463249e-01 -2.42939424e-02 9.95762110e-01 -9.52469468e-01 -5.87376654e-01 -2.13487953e-01 6.51292875e-02 -8.61799061e-01 1.28059955e-02 -1.40112305e+00 -2.32194185e-01 -2.05797839e+00 1.24106020e-01 -5.22342026e-01 3.35499272e-02 8.21746171e-01 -1.67733222e-01 -1.13536082e-01 1.48026004e-01 2.41364449e-01 -5.18107116e-01 5.57277739e-01 1.00797772e+00 1.13531463e-02 8.13502371e-02 -2.74195939e-01 -8.37585449e-01 -1.15656681e-01 6.76038265e-01 -7.75491595e-01 -5.01263380e-01 -1.14434445e+00 7.42419183e-01 4.95832302e-02 2.84259409e-01 -1.20166707e+00 2.50212163e-01 2.64004558e-01 1.58615455e-01 -2.92284518e-01 3.22806954e-01 -4.91596967e-01 2.93754518e-01 2.81916827e-01 -7.30683684e-01 7.11283922e-01 5.51249385e-01 1.73929453e-01 -2.68237829e-01 -2.27180004e-01 4.31733280e-01 -4.25136775e-01 -4.40816313e-01 8.06310326e-02 -9.51476023e-02 5.04052937e-01 6.54651523e-01 4.68121581e-02 -5.60472012e-01 -3.94448429e-01 -3.44830692e-01 3.50773752e-01 4.85514641e-01 6.66235566e-01 3.80727410e-01 -1.23118532e+00 -1.03065884e+00 2.13317871e-01 4.41244334e-01 -1.03407614e-01 -2.04794571e-01 5.51265538e-01 -4.65523869e-01 9.50716078e-01 -2.05760390e-01 -1.69839218e-01 -9.36591625e-01 4.98119891e-01 5.55778623e-01 -8.18933964e-01 -1.47365212e-01 6.40629292e-01 -3.46712142e-01 -1.00794899e+00 -1.29739091e-01 -5.15528977e-01 1.87968150e-01 -4.02287602e-01 9.45059136e-02 1.27645537e-01 3.50953043e-01 -2.85524517e-01 2.88205240e-02 -1.79515779e-01 -1.83696076e-01 -3.31674755e-01 9.58073616e-01 4.02719229e-01 -3.13799948e-01 3.06816429e-01 1.07622325e+00 -1.53933689e-01 -4.26669329e-01 -5.30136108e-01 4.64156061e-01 6.89290240e-02 -2.40834072e-01 -1.36485338e+00 -7.46196330e-01 8.00408602e-01 4.37660962e-01 -2.26298541e-01 1.01619494e+00 9.08987522e-02 1.15437043e+00 5.84464014e-01 4.79567796e-01 -8.46646488e-01 -1.90151148e-02 8.76008689e-01 9.79213655e-01 -1.01312888e+00 -4.25305665e-01 5.62869050e-02 -5.94345450e-01 8.74034047e-01 9.44399536e-01 1.30625397e-01 2.93128073e-01 3.00337523e-01 2.09663048e-01 -1.66707069e-01 -1.38110971e+00 -2.59060115e-01 3.59200835e-01 4.34501350e-01 9.38503504e-01 -3.74786183e-03 -2.72975266e-01 6.53277993e-01 -5.68389952e-01 3.46955836e-01 2.61371911e-01 1.06593001e+00 2.62754001e-02 -1.84718537e+00 -2.96561215e-02 7.18660772e-01 -5.57819307e-01 -5.96278548e-01 -6.58151388e-01 9.71629143e-01 2.03084022e-01 7.00110793e-01 1.37879610e-01 -5.92796743e-01 8.34876537e-01 9.07700360e-01 4.44354057e-01 -1.13933647e+00 -1.28172779e+00 -6.11905754e-01 7.48292506e-01 -4.54145312e-01 -9.24942791e-02 -4.95888561e-01 -1.01459301e+00 -2.23529115e-01 -2.09485784e-01 5.40617764e-01 7.23363698e-01 5.89285314e-01 1.06644237e+00 4.59668994e-01 -1.06230207e-01 1.59311324e-01 -7.64323831e-01 -1.33925223e+00 2.35167593e-01 3.24256092e-01 -1.49950176e-01 5.60573973e-02 3.02365512e-01 1.18681885e-01]
[11.227594375610352, 8.408230781555176]
0664970b-ba06-4187-8d8c-ce6acedc765f
learning-to-complete-object-shapes-for-object
2208.05067
null
https://arxiv.org/abs/2208.05067v1
https://arxiv.org/pdf/2208.05067v1.pdf
Learning to Complete Object Shapes for Object-level Mapping in Dynamic Scenes
In this paper, we propose a novel object-level mapping system that can simultaneously segment, track, and reconstruct objects in dynamic scenes. It can further predict and complete their full geometries by conditioning on reconstructions from depth inputs and a category-level shape prior with the aim that completed object geometry leads to better object reconstruction and tracking accuracy. For each incoming RGB-D frame, we perform instance segmentation to detect objects and build data associations between the detection and the existing object maps. A new object map will be created for each unmatched detection. For each matched object, we jointly optimise its pose and latent geometry representations using geometric residual and differential rendering residual towards its shape prior and completed geometry. Our approach shows better tracking and reconstruction performance compared to methods using traditional volumetric mapping or learned shape prior approaches. We evaluate its effectiveness by quantitatively and qualitatively testing it in both synthetic and real-world sequences.
['Stefan Leutenegger', 'Andrew J. Davison', 'Binbin Xu']
2022-08-09
null
null
null
null
['object-reconstruction']
['computer-vision']
[ 4.29702759e-01 3.75344515e-01 1.31022528e-01 -2.82749712e-01 -7.19464719e-01 -4.99281138e-01 6.48659706e-01 3.32556516e-01 -3.10882688e-01 3.36744547e-01 -1.90107808e-01 2.68157244e-01 5.58699667e-02 -8.38630199e-01 -9.22192097e-01 -3.76388520e-01 1.60558283e-01 1.27966034e+00 9.81431663e-01 3.09817344e-01 2.43570417e-01 9.07272756e-01 -1.53367126e+00 1.06635205e-01 4.84501839e-01 1.00154340e+00 6.82294667e-01 6.73555434e-01 -8.53667688e-03 5.95205784e-01 -1.22273132e-01 -2.04244494e-01 4.98361051e-01 1.67064127e-02 -5.56901336e-01 5.12402236e-01 6.22791767e-01 -5.17598450e-01 -3.42990249e-01 7.68716812e-01 1.74903125e-01 1.36951879e-01 7.20968843e-01 -8.62413824e-01 -2.21059807e-02 1.33671299e-01 -6.87117159e-01 -1.92484166e-02 4.13601041e-01 2.19947904e-01 5.14366090e-01 -1.02514052e+00 1.01469398e+00 1.17115986e+00 7.43390322e-01 5.17995894e-01 -1.50062537e+00 -5.79484224e-01 1.92116514e-01 -2.61616200e-01 -1.14370131e+00 -3.96899402e-01 8.27583075e-01 -6.92143500e-01 7.11210787e-01 1.22858226e-01 7.40958810e-01 6.16413176e-01 7.93669894e-02 7.09014654e-01 8.06612492e-01 -2.95826524e-01 2.95983225e-01 1.18816294e-01 -2.54299134e-01 7.78272092e-01 1.02922708e-01 2.90305883e-01 -4.47375327e-01 2.22623944e-02 1.33341312e+00 7.28973970e-02 -2.51760751e-01 -1.11041260e+00 -1.49009037e+00 2.94589072e-01 4.80278730e-01 -1.25222698e-01 -6.14596903e-01 4.66880769e-01 -1.77211329e-01 -3.43432665e-01 4.70587283e-01 2.29113206e-01 -1.93784982e-01 2.25881755e-01 -1.14325082e+00 2.93325394e-01 4.29217935e-01 9.78614330e-01 9.58781660e-01 -1.30835637e-01 -2.49415517e-01 3.41281444e-01 5.97645998e-01 6.58716023e-01 -1.47022888e-01 -1.43383312e+00 1.48453340e-01 7.02667594e-01 3.19555521e-01 -8.80035937e-01 -2.99667180e-01 -3.88873875e-01 -2.86819071e-01 5.88381290e-01 3.34441304e-01 3.82646173e-01 -1.22932208e+00 1.28570724e+00 8.95805299e-01 6.01232469e-01 -2.63132989e-01 1.04719996e+00 5.97542107e-01 5.82430840e-01 7.64533207e-02 -1.69627413e-01 1.04605174e+00 -7.34744549e-01 -3.56287748e-01 -3.90127093e-01 1.75386462e-02 -7.60461330e-01 4.97956246e-01 3.29311877e-01 -1.43598771e+00 -7.11858094e-01 -8.77594233e-01 -1.23570323e-01 2.04942316e-01 7.68917799e-02 4.33787137e-01 4.44797367e-01 -8.77558708e-01 5.38919032e-01 -1.24109530e+00 -6.26577884e-02 5.35450101e-01 3.72260481e-01 -2.79214263e-01 -1.86143294e-01 -3.31366986e-01 7.59388268e-01 4.39273179e-01 -1.32736146e-01 -1.41995633e+00 -1.11670005e+00 -8.00681829e-01 -3.45988512e-01 3.40097040e-01 -1.00339103e+00 1.10201776e+00 -5.94634354e-01 -1.29725218e+00 1.14068472e+00 -2.47406103e-02 -3.57651919e-01 6.61607802e-01 -1.19383208e-01 1.61042556e-01 1.45905823e-01 1.48609176e-01 1.10088086e+00 9.41885352e-01 -1.84959233e+00 -7.54323363e-01 -6.45076990e-01 -1.62922442e-01 2.49962613e-01 4.41680402e-01 -3.59630167e-01 -9.22668517e-01 -4.53952909e-01 7.59171128e-01 -7.84235597e-01 -3.53826642e-01 5.57248354e-01 -2.26323992e-01 3.10886323e-01 9.64038193e-01 -8.36896181e-01 4.32688117e-01 -1.98482406e+00 4.56728458e-01 3.37048024e-01 2.30605468e-01 -2.07114145e-02 -4.96593257e-03 -1.34576410e-01 3.58091295e-01 -4.44669425e-01 -3.95895690e-01 -8.55595946e-01 -2.87191838e-01 3.13546538e-01 -2.10110888e-01 7.68146217e-01 8.04009363e-02 9.89666224e-01 -8.95786405e-01 -5.61861873e-01 6.29106462e-01 7.05138445e-01 -7.12838292e-01 1.99532494e-01 -6.48683071e-01 8.92066002e-01 -4.49822158e-01 6.97161615e-01 7.81615674e-01 -1.03258781e-01 1.08901337e-01 -3.49706650e-01 -2.77465641e-01 -8.31153542e-02 -1.43387449e+00 2.00363064e+00 -3.19095194e-01 5.69719672e-01 2.88041383e-01 -5.91238618e-01 8.45050573e-01 8.10032189e-02 8.11060488e-01 -6.18970454e-01 1.90857321e-01 1.22630233e-02 -4.75914687e-01 -2.58926272e-01 5.04155397e-01 -1.40023321e-01 2.02245206e-01 2.19604656e-01 1.15574285e-01 -5.85190773e-01 -2.23174736e-01 2.37892106e-01 8.41083467e-01 5.56808770e-01 -1.94225624e-01 2.24800557e-02 3.78490895e-01 1.49497166e-01 3.84974390e-01 4.85252500e-01 1.26219705e-01 9.44585085e-01 -9.07969549e-02 -2.97775447e-01 -1.40312803e+00 -1.54327500e+00 -3.26048374e-01 6.59709811e-01 6.76529527e-01 1.05585940e-01 -4.72928911e-01 -3.70360404e-01 1.55222028e-01 6.77832067e-01 -6.44085407e-01 2.75644064e-02 -7.37091720e-01 -1.51625142e-01 3.54235657e-02 5.12058973e-01 1.12135038e-01 -1.14282596e+00 -1.13471580e+00 4.09640789e-01 -2.01252729e-01 -1.14339817e+00 -2.18870625e-01 1.48954555e-01 -1.11589158e+00 -1.05392706e+00 -5.00613749e-01 -6.53488934e-01 7.95322776e-01 -3.65990810e-02 9.88639116e-01 -1.39399737e-01 -4.95538980e-01 5.95870614e-01 -7.63724148e-02 -7.94106200e-02 -6.16808355e-01 -3.24689984e-01 -7.44233057e-02 1.39468387e-01 -4.08019513e-01 -4.28971440e-01 -7.54026771e-01 4.05278206e-01 -8.06785822e-01 4.65874225e-01 5.80886543e-01 2.25209355e-01 1.22882009e+00 -1.60247251e-01 -6.94035366e-02 -4.72217590e-01 -2.10061863e-01 -2.34229311e-01 -9.09496725e-01 1.57314092e-01 -3.30095381e-01 1.39223263e-01 -3.01847875e-01 -4.56436425e-01 -1.10440218e+00 7.96609104e-01 1.42696545e-01 -9.25596893e-01 -9.39329639e-02 -2.41435338e-02 -1.21330589e-01 -1.64460212e-01 5.78833103e-01 3.29988807e-01 7.68641010e-02 -4.78780746e-01 4.86975461e-01 5.24993166e-02 8.97870600e-01 -5.89183450e-01 8.30714226e-01 9.55629587e-01 9.42769367e-03 -5.70659161e-01 -6.19975746e-01 -6.71309233e-01 -1.05077243e+00 -7.09714115e-01 1.12344086e+00 -7.41186202e-01 -6.52597070e-01 2.10659564e-01 -1.24881339e+00 -5.43788314e-01 -6.20138884e-01 5.44291675e-01 -8.14510763e-01 2.15744957e-01 -3.07254225e-01 -8.98413539e-01 4.65537095e-03 -1.06898403e+00 1.59748316e+00 4.46311384e-02 8.46200213e-02 -8.53871942e-01 2.35315531e-01 2.36999422e-01 4.51532081e-02 6.51865125e-01 2.65774131e-01 1.81622997e-01 -1.26409340e+00 -1.15901224e-01 -2.82240510e-01 3.50626521e-02 -1.50952742e-01 -6.28575087e-02 -9.33029354e-01 -2.67965138e-01 -2.56958380e-02 6.71705678e-02 7.46879697e-01 5.73462188e-01 8.90742481e-01 1.51191264e-01 -8.31098437e-01 6.89351797e-01 1.65077937e+00 1.47922114e-01 5.39650857e-01 1.40634596e-01 8.21785867e-01 5.97718418e-01 7.85420537e-01 4.22397733e-01 1.61362514e-01 8.37547064e-01 7.46694386e-01 -2.47786772e-02 -4.83018667e-01 -3.46520543e-01 1.57668322e-01 2.86324441e-01 -1.39034599e-01 1.95056126e-02 -1.04131520e+00 5.69242001e-01 -1.66299713e+00 -5.86199045e-01 -3.87576789e-01 2.30463147e+00 5.55990279e-01 2.68819124e-01 7.58776292e-02 -1.56791180e-01 6.23385847e-01 -2.73302764e-01 -6.54566169e-01 3.42076898e-01 -2.82884873e-02 2.30414197e-01 6.11725450e-01 7.42425501e-01 -7.94106841e-01 1.10084844e+00 5.94284534e+00 4.35450584e-01 -7.95168161e-01 3.05594176e-01 4.63528991e-01 -1.71697453e-01 -3.42719108e-01 1.77609861e-01 -7.72438169e-01 1.04117282e-01 4.31996435e-01 2.61783302e-01 2.31190369e-01 6.33427799e-01 -4.25324850e-02 -4.65129226e-01 -1.30315578e+00 9.08981740e-01 -3.90026942e-02 -1.45360208e+00 4.06434052e-02 1.58464596e-01 8.19577754e-01 1.77502915e-01 -6.29286543e-02 -1.61866099e-02 1.92201093e-01 -7.41676688e-01 1.37740731e+00 9.81725454e-01 7.47021735e-01 -5.56100786e-01 3.20445687e-01 5.36433876e-01 -1.34300184e+00 1.95640221e-01 -2.66216636e-01 1.82755455e-01 4.75356877e-01 3.11250836e-01 -9.65022564e-01 5.64301074e-01 5.99108815e-01 4.64815289e-01 -5.51073134e-01 1.25042999e+00 4.25906479e-02 1.25006512e-01 -4.72424120e-01 5.21314919e-01 2.65751611e-02 -1.87851563e-01 6.99699521e-01 9.46998358e-01 1.22822829e-01 2.46839806e-01 4.99396354e-01 1.22989643e+00 3.89714539e-01 -2.30779931e-01 -2.37164885e-01 5.44514358e-01 4.33285177e-01 1.16899335e+00 -1.24294376e+00 -4.40332979e-01 2.21875608e-02 1.07295179e+00 2.45165601e-01 2.07254127e-01 -9.22264576e-01 5.38667142e-01 3.97953451e-01 7.38822997e-01 4.51122820e-01 -4.08574164e-01 -4.29671735e-01 -7.82776237e-01 -9.82119888e-03 -1.19439900e-01 1.62679553e-01 -1.01421833e+00 -7.59962022e-01 3.73297870e-01 1.46855041e-01 -1.06873548e+00 -7.87474811e-02 -3.40952098e-01 -6.43319413e-02 7.01666057e-01 -1.18566370e+00 -1.20909762e+00 -5.54113209e-01 5.87082446e-01 5.27019978e-01 3.66835445e-01 3.77333194e-01 2.67320126e-01 1.20696584e-02 3.41158919e-02 -1.40207127e-01 -4.13602628e-02 2.72700280e-01 -1.03870964e+00 1.96513325e-01 6.55932903e-01 2.86693394e-01 1.39193356e-01 7.01764524e-01 -1.18700993e+00 -1.26075101e+00 -1.31292844e+00 6.66259676e-02 -9.14945185e-01 1.52304083e-01 -4.66505647e-01 -8.15907061e-01 8.32138896e-01 -4.69947428e-01 2.48954415e-01 -2.43617922e-01 -5.43611467e-01 -8.42792727e-03 9.84357297e-02 -1.31114304e+00 2.58867860e-01 1.17124283e+00 -2.77011365e-01 -4.93662566e-01 2.24865571e-01 6.38822436e-01 -9.55460131e-01 -7.40170300e-01 6.62617505e-01 6.26440108e-01 -9.41023231e-01 1.36239898e+00 -4.87088747e-02 1.75165012e-01 -6.35292411e-01 -1.31432950e-01 -8.81087005e-01 -2.78899580e-01 -1.81961298e-01 -2.52963990e-01 8.89761984e-01 6.12341166e-02 -8.71260241e-02 1.16945219e+00 5.03386497e-01 -3.61130834e-01 -6.50020301e-01 -9.99803185e-01 -6.33013248e-01 -2.18114167e-01 -8.01798701e-01 3.16408962e-01 6.90124929e-01 -7.20013559e-01 -6.67378157e-02 -1.11791641e-01 6.03386581e-01 1.13823223e+00 2.45899692e-01 9.67172265e-01 -1.26032960e+00 -3.50605577e-01 -3.06168616e-01 -6.73634529e-01 -1.26059151e+00 -6.11506999e-02 -9.52502191e-01 2.59383768e-01 -1.69980252e+00 1.32511795e-01 -8.75096440e-01 1.39126033e-01 2.21271500e-01 7.27921203e-02 6.74433768e-01 8.99866521e-02 2.47707278e-01 -6.21631563e-01 3.94742221e-01 1.38342178e+00 3.33006531e-02 -3.35473478e-01 -7.12951049e-02 7.34122619e-02 6.53808296e-01 3.65316331e-01 -5.76963782e-01 -2.05552533e-01 -3.96888822e-01 -5.29468246e-02 4.86914426e-01 8.96248221e-01 -1.26767170e+00 3.89952838e-01 1.66664869e-02 7.57197261e-01 -1.18937683e+00 7.95501053e-01 -1.01132739e+00 8.09403241e-01 6.16798699e-01 -1.83174200e-02 -3.48362684e-01 2.55139917e-01 8.01539600e-01 2.84525573e-01 5.35469800e-02 9.32242095e-01 -2.27395430e-01 -7.38766611e-01 5.80121100e-01 4.65587415e-02 -2.16472313e-01 1.32129502e+00 -7.72741199e-01 4.61199969e-01 2.45338231e-02 -1.20984387e+00 2.22801119e-01 8.79869401e-01 4.13373083e-01 1.07551289e+00 -1.12608469e+00 -7.13709474e-01 4.05271918e-01 9.10247266e-02 5.54307699e-01 2.10756481e-01 8.16379905e-01 -7.40902841e-01 1.72704667e-01 -2.27231801e-01 -1.26902032e+00 -1.20740402e+00 6.49143755e-01 6.29914403e-01 1.07608475e-01 -9.87429321e-01 8.70630026e-01 5.56667745e-01 -5.16149282e-01 3.31805289e-01 -5.40741205e-01 1.00777633e-01 -2.51031965e-01 2.37345919e-01 4.52049881e-01 -4.03615162e-02 -8.63547623e-01 -3.29158962e-01 8.77393186e-01 2.33449116e-01 -4.51739907e-01 1.18441224e+00 -2.97365457e-01 1.53856069e-01 4.10703599e-01 7.50159919e-01 -9.95840281e-02 -1.88249314e+00 -4.29611802e-01 -1.69043943e-01 -6.93244517e-01 1.52686790e-01 -6.38754666e-01 -1.13692117e+00 5.04770935e-01 8.91196251e-01 -2.69573390e-01 6.07752681e-01 5.47371447e-01 4.93741930e-01 -2.49108344e-01 7.23516345e-01 -6.34214938e-01 3.92140687e-01 1.39962152e-01 7.71491349e-01 -9.47186947e-01 2.02313498e-01 -6.61693037e-01 -2.84665853e-01 8.95248175e-01 6.40355587e-01 -3.44236910e-01 5.83191454e-01 3.82258832e-01 -2.71946728e-01 -4.83602375e-01 -2.35560909e-01 -1.03784263e-01 4.79459494e-01 7.65321255e-01 -2.06402615e-01 7.26977410e-03 3.83549154e-01 5.98814245e-03 -3.53051201e-02 -1.87242478e-01 1.93180025e-01 8.32665443e-01 -5.96684098e-01 -9.01859999e-01 -7.00200438e-01 2.43858770e-01 -1.76503267e-02 3.47245038e-01 -2.06284061e-01 6.82013571e-01 3.18169415e-01 3.59866679e-01 5.67686558e-01 -1.09006383e-01 4.02773261e-01 -1.96290344e-01 9.67740357e-01 -9.52553034e-01 -2.42019311e-01 2.85981387e-01 -2.23767310e-01 -8.16243410e-01 -5.52761078e-01 -9.98360038e-01 -1.71618724e+00 1.21521376e-01 -3.49664927e-01 -2.95162261e-01 8.79959881e-01 7.74665713e-01 1.65557623e-01 6.64237261e-01 2.08276331e-01 -1.51498532e+00 4.03921008e-02 -5.25777102e-01 -3.16370904e-01 2.69358546e-01 1.77886829e-01 -9.48752046e-01 1.13601405e-02 3.46689135e-01]
[7.43649959564209, -2.4860997200012207]
b047a3cc-91c0-4ed4-96f8-820aa11428ad
a-survey-on-explainability-of-graph-neural
2306.01958
null
https://arxiv.org/abs/2306.01958v1
https://arxiv.org/pdf/2306.01958v1.pdf
A Survey on Explainability of Graph Neural Networks
Graph neural networks (GNNs) are powerful graph-based deep-learning models that have gained significant attention and demonstrated remarkable performance in various domains, including natural language processing, drug discovery, and recommendation systems. However, combining feature information and combinatorial graph structures has led to complex non-linear GNN models. Consequently, this has increased the challenges of understanding the workings of GNNs and the underlying reasons behind their predictions. To address this, numerous explainability methods have been proposed to shed light on the inner mechanism of the GNNs. Explainable GNNs improve their security and enhance trust in their recommendations. This survey aims to provide a comprehensive overview of the existing explainability techniques for GNNs. We create a novel taxonomy and hierarchy to categorize these methods based on their objective and methodology. We also discuss the strengths, limitations, and application scenarios of each category. Furthermore, we highlight the key evaluation metrics and datasets commonly used to assess the explainability of GNNs. This survey aims to assist researchers and practitioners in understanding the existing landscape of explainability methods, identifying gaps, and fostering further advancements in interpretable graph-based machine learning.
['Sourav Medya', 'Charu Aggarwal', 'Kartik Sharma', 'Jaspal Jannu', 'Jaykumar Kakkad']
2023-06-02
null
null
null
null
['drug-discovery']
['medical']
[ 2.17372879e-01 6.88521147e-01 -7.18549192e-01 -3.08748454e-01 3.52612913e-01 -4.09474403e-01 4.12205964e-01 5.54329455e-01 3.72803509e-01 4.67482328e-01 2.36757338e-01 -9.04675066e-01 -5.91736019e-01 -8.58965099e-01 -6.55477166e-01 -1.96040913e-01 -2.62594759e-01 3.48220050e-01 -3.27118248e-01 -3.31144929e-01 2.97176033e-01 4.45093095e-01 -1.06552672e+00 5.80235302e-01 1.07930124e+00 7.54376888e-01 -9.37961563e-02 5.84692061e-01 -4.74635810e-01 1.02053523e+00 -4.26673830e-01 -9.05013800e-01 -6.59189075e-02 -6.08273208e-01 -8.53409350e-01 -2.14334682e-01 1.64223939e-01 -1.38880268e-01 -7.13022470e-01 9.66115415e-01 1.83297731e-02 -2.42020264e-02 6.63860679e-01 -1.72789121e+00 -1.31272578e+00 1.04562235e+00 -3.20699364e-01 9.53535065e-02 2.31724262e-01 -2.08471194e-02 1.54182529e+00 -5.52969873e-01 5.86264372e-01 1.20841396e+00 9.34196591e-01 6.88391149e-01 -9.65093076e-01 -5.58649600e-01 4.24490154e-01 4.68870759e-01 -1.01637411e+00 3.27871218e-02 8.11835766e-01 -2.06524491e-01 1.37830806e+00 3.85464489e-01 9.29726243e-01 1.09302139e+00 5.92431962e-01 7.67540812e-01 5.63749969e-01 -2.00641528e-01 7.83458278e-02 -2.62483478e-01 4.36295807e-01 1.08327067e+00 8.19756866e-01 -3.76276881e-03 -6.03786290e-01 -3.11582506e-01 7.28964031e-01 4.92990822e-01 -1.67342812e-01 -4.64094758e-01 -8.44070613e-01 1.31391263e+00 1.04692233e+00 2.06670493e-01 -2.77121127e-01 2.87033468e-01 3.39101195e-01 2.80962110e-01 5.41652560e-01 7.34607577e-01 -5.55379212e-01 3.54364932e-01 -5.96107244e-01 9.26681012e-02 8.97043943e-01 8.05172861e-01 5.91594696e-01 3.07126164e-01 2.48440608e-01 2.69331962e-01 4.62715089e-01 1.18174583e-01 7.78210908e-02 -3.51280749e-01 3.25947315e-01 1.09432745e+00 -6.17300630e-01 -1.62138021e+00 -8.31152797e-01 -7.40318716e-01 -1.32884741e+00 -1.94292933e-01 9.30488259e-02 1.73396334e-01 -8.95908892e-01 1.55241609e+00 4.94545475e-02 4.50392328e-02 -1.57754794e-01 7.98521101e-01 1.44900036e+00 3.07710648e-01 2.24639758e-01 1.82702392e-01 1.15006948e+00 -1.18816233e+00 -7.17237532e-01 -2.62152731e-01 8.63897800e-01 -4.59378749e-01 9.15643334e-01 2.25993916e-01 -7.61875391e-01 -4.39919770e-01 -1.01982415e+00 -1.90489769e-01 -5.50282001e-01 -2.99810648e-01 1.41710210e+00 6.28934264e-01 -1.13033640e+00 8.95769000e-01 -9.03038919e-01 -4.47362155e-01 6.88276887e-01 7.18952358e-01 -2.34441400e-01 -1.70632631e-01 -1.41465485e+00 7.82748282e-01 3.30393374e-01 1.19801268e-01 -4.92088944e-01 -6.02390826e-01 -8.11456084e-01 4.02868330e-01 1.91710636e-01 -1.13001871e+00 7.07951307e-01 -7.96871662e-01 -8.73934627e-01 3.88448834e-01 1.09805666e-01 -7.00823963e-01 -4.27465849e-02 2.01453939e-01 -4.99891251e-01 -1.18901320e-01 -2.03702733e-01 4.00852263e-01 3.80120873e-01 -9.06318903e-01 -1.66044563e-01 -1.88186511e-01 3.86240929e-01 -2.76081525e-02 -3.88000369e-01 -2.84583539e-01 -2.77744085e-02 -5.88678300e-01 1.97129264e-01 -8.94699335e-01 -4.76300657e-01 8.39916803e-03 -8.11884105e-01 -2.17271313e-01 4.07367259e-01 -3.22345614e-01 1.47937655e+00 -1.64263630e+00 1.20631084e-01 3.80928487e-01 1.42714238e+00 4.44124490e-01 -3.74803275e-01 8.10377479e-01 -3.40752512e-01 6.56945467e-01 1.39413074e-01 9.39090475e-02 5.04101953e-03 3.26585084e-01 -2.20080256e-01 3.15979689e-01 8.22565556e-02 1.41450751e+00 -9.14972484e-01 2.85374466e-02 2.59700328e-01 6.93110943e-01 -6.77417219e-01 4.09856103e-02 -3.31842154e-01 2.36615777e-01 -6.57414794e-01 5.13796449e-01 4.94776219e-01 -9.37669039e-01 6.04154944e-01 -2.41247386e-01 5.09124339e-01 4.81630892e-01 -6.27953231e-01 1.20833790e+00 -1.15996720e-02 7.71339774e-01 -4.29821134e-01 -1.12987900e+00 1.02838171e+00 3.71673852e-02 4.51057464e-01 -4.84478712e-01 4.39844243e-02 1.87746629e-01 4.57147539e-01 -3.13122064e-01 4.13096637e-01 6.92097694e-02 3.34459305e-01 6.84297919e-01 -1.58504054e-01 1.35902062e-01 -5.00909165e-02 5.99792898e-01 1.37057924e+00 -4.86673951e-01 8.31133306e-01 -1.11397609e-01 3.77110243e-01 -1.27239525e-01 2.05877870e-01 9.56721187e-01 -2.16480240e-01 3.23668867e-01 7.74508357e-01 -1.29642785e+00 -9.15253818e-01 -6.72838926e-01 2.99284250e-01 9.10064280e-01 1.81311876e-01 -9.87059534e-01 -7.35564590e-01 -1.04621637e+00 1.34966537e-01 5.01046538e-01 -9.51388538e-01 -6.00063264e-01 -1.64811656e-01 -7.63169408e-01 5.87823153e-01 4.17237967e-01 1.94339082e-01 -1.03764665e+00 1.92031637e-01 7.30877696e-03 -3.57217908e-01 -9.51782346e-01 -2.16917649e-01 1.26981318e-01 -1.14803398e+00 -1.57375646e+00 7.81677291e-02 -5.68944991e-01 9.82301891e-01 6.55008852e-01 1.55334294e+00 1.05936408e+00 3.55970697e-03 2.60934085e-01 -4.63406533e-01 -5.46650350e-01 -4.18898463e-01 4.33545142e-01 1.66798726e-01 -4.12423223e-01 5.65221071e-01 -3.83271724e-01 -6.79106772e-01 2.65520006e-01 -1.13314402e+00 3.08219075e-01 6.40141606e-01 8.50472748e-01 3.94860446e-01 -1.27975404e-01 4.88144785e-01 -1.17552161e+00 1.03299153e+00 -6.42647684e-01 1.31462875e-03 3.83932889e-01 -1.30197024e+00 2.02394679e-01 6.72620833e-01 -1.85194343e-01 -2.13282019e-01 -3.65016013e-01 -2.59797394e-01 -2.27875430e-02 1.71526447e-01 9.72468197e-01 9.28489715e-02 -5.02606511e-01 8.99057984e-01 -8.29244629e-02 1.24716610e-01 -2.37170249e-01 6.09026134e-01 2.64654249e-01 -1.04575865e-01 -1.71875000e-01 7.83076823e-01 2.20904857e-01 2.25447550e-01 -4.70948040e-01 -1.07759678e+00 -4.80729640e-02 -3.87717158e-01 -1.40137210e-01 4.82655436e-01 -4.27524388e-01 -7.96317637e-01 4.19946155e-03 -1.38097668e+00 -4.63408530e-02 -6.37094080e-02 1.86990902e-01 -2.56598353e-01 5.23563862e-01 -9.42129076e-01 -3.58181924e-01 -7.95918345e-01 -1.17268550e+00 4.70942527e-01 2.18855664e-01 -5.92021048e-01 -1.43038476e+00 -1.76475585e-01 5.08836389e-01 5.81728995e-01 2.82976270e-01 1.41491711e+00 -1.11794662e+00 -6.14343047e-01 -1.14277758e-01 -6.19811773e-01 -1.22992009e-01 2.37729937e-01 -1.64003834e-01 -6.39501750e-01 -2.37307191e-01 -3.88780236e-01 -9.42714959e-02 8.23182464e-01 6.85951352e-01 1.24752438e+00 -6.93051040e-01 -5.63786387e-01 7.60787487e-01 1.07021165e+00 -8.30074921e-02 4.85156626e-01 3.78332913e-01 1.13058019e+00 4.75285172e-01 2.24140696e-02 4.63355295e-02 5.38699389e-01 2.89999902e-01 1.05991352e+00 -4.62835968e-01 -1.10223450e-01 -4.72302437e-01 -1.32208252e-02 1.12544894e+00 -3.28862101e-01 -7.11295307e-01 -8.43712449e-01 -6.21522032e-02 -2.05797076e+00 -7.11589754e-01 -5.90810061e-01 1.57885242e+00 2.19505876e-01 7.50386566e-02 -8.28539804e-02 -4.42132689e-02 7.87996233e-01 3.27914059e-01 -8.63037229e-01 -6.35274291e-01 -2.02279761e-01 5.72865680e-02 3.21180403e-01 4.06236529e-01 -8.05377901e-01 1.03293538e+00 7.33014822e+00 5.18925071e-01 -9.92381990e-01 -3.99409980e-01 8.47679019e-01 3.05669129e-01 -8.01797688e-01 -5.39597608e-02 -2.85892218e-01 3.36772613e-02 8.00710022e-01 -2.78615654e-01 8.11309576e-01 8.97744536e-01 2.20114246e-01 7.17533529e-01 -1.14364517e+00 8.25251341e-01 1.09429471e-01 -1.95758402e+00 6.26713872e-01 3.04486454e-01 8.03927600e-01 2.03936085e-01 2.38414034e-01 1.51649997e-01 3.91567111e-01 -1.40187240e+00 1.49017438e-01 3.45941722e-01 5.45244813e-01 -6.73399508e-01 8.15090835e-01 1.36148464e-02 -1.23955941e+00 -3.13199237e-02 -7.34219790e-01 -6.79738879e-01 -1.54578537e-01 6.02520108e-01 -9.87571776e-01 7.89666235e-01 3.41382533e-01 9.53277707e-01 -4.57014740e-01 9.13622260e-01 -5.52229166e-01 6.79123640e-01 6.70760572e-02 -3.71443719e-01 2.73145169e-01 -2.43449569e-01 3.18442851e-01 9.01642501e-01 1.81365609e-01 1.09330982e-01 3.35515570e-03 1.13973749e+00 -3.47477227e-01 1.35941610e-01 -6.63932920e-01 -4.90280539e-01 2.54854381e-01 1.25687575e+00 -1.04250300e+00 -1.20003387e-01 -5.71012855e-01 6.14253521e-01 6.11336648e-01 3.61818492e-01 -9.68151510e-01 -2.01719657e-01 9.21463549e-01 1.55619815e-01 -5.09082116e-02 -2.48410404e-01 -5.32975972e-01 -1.11198568e+00 -4.16149735e-01 -1.23676431e+00 6.07233763e-01 -6.16452336e-01 -1.46718705e+00 8.88317347e-01 -4.17079061e-01 -9.03725088e-01 -7.92930350e-02 -8.23183775e-01 -5.99493146e-01 5.52393436e-01 -1.37516534e+00 -1.35259557e+00 -3.48514259e-01 3.57570469e-01 2.55295873e-01 -2.35358894e-01 9.58714187e-01 1.54550988e-02 -5.65140426e-01 5.53444922e-01 9.48649459e-03 9.22551155e-02 2.67161191e-01 -9.47105587e-01 1.06178284e+00 5.17331004e-01 5.12823284e-01 1.09500921e+00 7.13920236e-01 -8.04579139e-01 -1.47379017e+00 -1.10615230e+00 1.04403543e+00 -5.77849686e-01 8.21911514e-01 -3.48141611e-01 -1.03226113e+00 8.20012867e-01 -2.84511223e-02 -5.97968213e-02 9.50621724e-01 6.72140062e-01 -5.30264735e-01 1.40517429e-01 -8.78578722e-01 7.35936224e-01 1.30354702e+00 -5.10046721e-01 -1.67469665e-01 4.92015779e-01 8.79994452e-01 -3.48692924e-01 -6.16954923e-01 3.08076799e-01 6.04360104e-01 -1.21719217e+00 9.45371568e-01 -1.34681928e+00 7.58064389e-01 -4.72883461e-03 2.89401680e-01 -1.38368750e+00 -8.39669347e-01 -6.59325302e-01 -5.06879747e-01 5.01153409e-01 6.25348985e-01 -7.72488832e-01 1.16728497e+00 9.32959020e-01 -2.49265507e-01 -1.22358811e+00 -3.66411805e-01 -4.60014135e-01 -1.19439535e-01 -5.04025161e-01 9.97712314e-01 1.24314356e+00 3.91387224e-01 4.40332592e-01 -5.66576600e-01 -4.24769968e-02 5.32988131e-01 8.80717710e-02 7.16942608e-01 -1.47798300e+00 -6.87331855e-02 -7.06961930e-01 -7.59934783e-01 -1.01468003e+00 1.44995466e-01 -1.34482336e+00 -7.32167602e-01 -2.13759851e+00 5.80560744e-01 -1.00537077e-01 -4.79563802e-01 7.30918109e-01 -3.38311881e-01 1.57829225e-01 3.09131350e-02 2.57349968e-01 -7.53104150e-01 3.82566243e-01 1.31097579e+00 -3.21119815e-01 3.00510190e-02 -1.10510131e-02 -1.37603343e+00 6.85732901e-01 9.90491748e-01 -5.02263963e-01 -7.64505804e-01 -6.11370206e-01 7.90533900e-01 -1.89643860e-01 2.50288069e-01 -4.94622052e-01 3.06193441e-01 -3.12692970e-01 2.24698722e-01 -2.27149785e-01 -2.34820396e-01 -7.02705801e-01 3.51686269e-01 8.51427317e-01 -5.38017690e-01 2.59846181e-01 7.95931518e-02 8.58665228e-01 -5.00280410e-02 1.29064649e-01 3.20561528e-01 -4.62980419e-02 -4.08983648e-01 8.27399015e-01 -3.40464234e-01 -2.18091145e-01 6.06405616e-01 -2.62139708e-01 -5.49118876e-01 -8.95825624e-01 -5.35101533e-01 4.33399267e-02 1.76551849e-01 6.05916560e-01 8.33703756e-01 -1.32866573e+00 -3.57520998e-01 1.58084482e-01 1.89922497e-01 -3.70967001e-01 1.96575925e-01 5.86592436e-01 -6.18221283e-01 7.73585796e-01 -2.42122322e-01 -2.41470784e-01 -1.18560314e+00 7.42920816e-01 3.81635249e-01 -5.26701868e-01 -4.24250513e-01 8.52272093e-01 4.36426044e-01 -6.97504938e-01 9.93917435e-02 -4.41397041e-01 -3.68829459e-01 -6.16986752e-01 2.78094828e-01 3.25489551e-01 1.62098482e-02 -2.94034421e-01 -3.89662236e-01 2.06366837e-01 -2.43114561e-01 9.10533905e-01 1.47055769e+00 -6.72995076e-02 -4.16083127e-01 -1.87554464e-01 9.33004975e-01 -2.73588389e-01 -7.02435553e-01 -9.47716013e-02 4.64995578e-03 -6.67479187e-02 -1.98936567e-01 -6.97997034e-01 -1.32664204e+00 8.60066295e-01 -4.06046472e-02 6.96665943e-01 7.83609986e-01 5.63938543e-02 8.74754667e-01 4.14214551e-01 2.23100826e-01 -4.43113655e-01 1.64031327e-01 6.34467304e-01 7.35072017e-01 -1.23120654e+00 4.38610524e-01 -7.67084956e-01 -3.95015597e-01 1.52458584e+00 4.90477979e-01 1.10548362e-01 8.23887289e-01 -2.57739842e-01 -1.60179153e-01 -6.96318150e-01 -5.86248457e-01 6.96577355e-02 8.01377118e-01 7.48646438e-01 7.70448506e-01 1.53816879e-01 -3.56405437e-01 9.65009451e-01 -4.15143698e-01 -2.07069069e-01 3.49744469e-01 2.90270358e-01 -2.68305659e-01 -1.27751291e+00 1.45414487e-01 9.25895452e-01 -3.32190692e-01 -5.45918643e-01 -1.00510168e+00 7.57539034e-01 -1.36337101e-01 1.17932332e+00 -4.74285394e-01 -1.01492262e+00 5.40971272e-02 -4.19601202e-01 4.60449979e-02 -4.43749219e-01 -5.35090089e-01 -5.42547226e-01 1.65614188e-01 -6.04874313e-01 -7.86197409e-02 2.39191388e-04 -1.25760639e+00 -1.07013035e+00 -5.09164691e-01 3.26042950e-01 5.94670773e-01 1.06929004e+00 7.76620269e-01 6.76687121e-01 3.23186129e-01 -4.32136536e-01 -2.04737201e-01 -5.47631145e-01 -4.25088614e-01 2.93139964e-01 1.99557871e-01 -5.01518846e-01 -4.33776617e-01 -4.63036120e-01]
[7.585259914398193, 6.306099891662598]
ad51125d-7b37-4ba0-850f-95ba94c48609
hollywood-3d-recognizing-actions-in-3d
null
null
http://openaccess.thecvf.com/content_cvpr_2013/html/Hadfield_Hollywood_3D_Recognizing_2013_CVPR_paper.html
http://openaccess.thecvf.com/content_cvpr_2013/papers/Hadfield_Hollywood_3D_Recognizing_2013_CVPR_paper.pdf
Hollywood 3D: Recognizing Actions in 3D Natural Scenes
Action recognition in unconstrained situations is a difficult task, suffering from massive intra-class variations. It is made even more challenging when complex 3D actions are projected down to the image plane, losing a great deal of information. The recent emergence of 3D data, both in broadcast content, and commercial depth sensors, provides the possibility to overcome this issue. This paper presents a new dataset, for benchmarking action recognition algorithms in natural environments, while making use of 3D information. The dataset contains around 650 video clips, across 14 classes. In addition, two state of the art action recognition algorithms are extended to make use of the 3D data, and five new interest point detection strategies are also proposed, that extend to the 3D data. Our evaluation compares all 4 feature descriptors, using 7 different types of interest point, over a variety of threshold levels, for the Hollywood3D dataset. We make the dataset including stereo video, estimated depth maps and all code required to reproduce the benchmark results, available to the wider community.
['Simon Hadfield', 'Richard Bowden']
2013-06-01
null
null
null
cvpr-2013-6
['interest-point-detection']
['computer-vision']
[ 3.89647424e-01 -3.40854287e-01 -2.14971870e-01 -2.48870373e-01 -5.50082803e-01 -5.66078842e-01 6.50724292e-01 -1.04619153e-01 -5.47359824e-01 5.28116465e-01 4.34359103e-01 4.45374787e-01 -1.21132098e-01 -5.87245584e-01 -2.56527543e-01 -7.70456910e-01 -3.98668438e-01 3.66149545e-01 8.66342306e-01 -5.26651889e-02 5.64284861e-01 1.05348754e+00 -2.10229015e+00 5.03844142e-01 1.28472656e-01 1.31833470e+00 4.37646359e-02 6.56352580e-01 3.25151384e-02 8.19987237e-01 -7.44130373e-01 -2.50186622e-01 5.91911674e-01 -1.42640010e-01 -5.86766541e-01 6.14697635e-01 7.67083764e-01 -7.33760536e-01 -4.51813340e-01 6.93032324e-01 6.01309717e-01 1.73216492e-01 5.66446424e-01 -1.29729080e+00 1.66846573e-01 -2.32157573e-01 -3.65865409e-01 3.44107419e-01 9.74177599e-01 1.19288452e-01 5.86113930e-01 -6.56540036e-01 9.28332865e-01 1.17994833e+00 5.50938070e-01 5.08618593e-01 -8.20464134e-01 -3.24084252e-01 -1.63191423e-01 5.74884176e-01 -1.04300225e+00 -4.62055027e-01 7.16719568e-01 -5.91769278e-01 1.29160094e+00 2.26988882e-01 9.16066766e-01 1.50074446e+00 1.88102543e-01 9.09940720e-01 1.18019187e+00 -3.82140905e-01 3.46198589e-01 -8.02791864e-02 -1.49906561e-01 2.49997035e-01 4.87131402e-02 1.71597704e-01 -8.36979151e-01 -2.90461872e-02 7.52878785e-01 1.12711430e-01 -2.86882937e-01 -9.07381654e-01 -1.21556640e+00 4.25430298e-01 -8.90765246e-03 3.64856660e-01 -3.84206474e-01 -1.89355314e-01 6.24616623e-01 3.51288348e-01 4.92946744e-01 9.46653634e-02 -3.12083840e-01 -1.00167179e+00 -6.72358155e-01 3.69442910e-01 6.78162694e-01 8.57726514e-01 5.74976563e-01 -2.18204767e-01 1.71708483e-02 6.78736687e-01 1.02262609e-01 5.58741152e-01 6.20847344e-01 -1.25907850e+00 5.88371873e-01 7.89585114e-01 2.61454321e-02 -1.14448905e+00 -5.31109273e-01 1.87120363e-01 -6.03456140e-01 6.93400979e-01 6.27306104e-01 2.53918529e-01 -6.70359612e-01 1.08877110e+00 5.62709570e-01 -9.50261503e-02 2.19254359e-03 9.74018991e-01 7.31829882e-01 2.17868403e-01 -3.88433874e-01 -4.65097651e-02 9.82502699e-01 -6.90074205e-01 -4.18776453e-01 -1.48878330e-02 6.29742503e-01 -5.57020724e-01 6.57666266e-01 8.70092332e-01 -8.53486300e-01 -5.42020679e-01 -9.85974967e-01 -8.31622258e-02 -5.43757796e-01 -5.49706817e-03 4.30148274e-01 7.62416422e-01 -8.99829030e-01 7.59901762e-01 -8.71795893e-01 -8.04271817e-01 4.91951793e-01 2.70994455e-01 -1.01660848e+00 -2.30644062e-01 -9.92999315e-01 1.07923841e+00 3.19772243e-01 -1.01932876e-01 -7.35437095e-01 -2.19960585e-01 -8.69258940e-01 -3.77104640e-01 3.09271663e-01 -5.62049560e-02 9.74506021e-01 -7.90721536e-01 -1.57566011e+00 1.30067611e+00 2.35534891e-01 -3.82095933e-01 9.28739130e-01 -3.45608771e-01 -3.52550358e-01 5.06204724e-01 3.88828330e-02 5.56165457e-01 7.92620301e-01 -6.96871996e-01 -7.47206450e-01 -7.92872071e-01 3.51370007e-01 3.11774433e-01 -2.63214558e-01 7.02469498e-02 -4.25931245e-01 -5.02725661e-01 7.42575005e-02 -8.34700346e-01 3.49616520e-02 3.25844496e-01 -1.33304268e-01 -4.95852008e-02 9.09610033e-01 -3.95360589e-01 7.24817991e-01 -2.30481935e+00 4.21965897e-01 2.68937256e-02 -1.23607270e-01 2.67778814e-01 6.06042966e-02 5.04998744e-01 4.45182770e-02 -4.74739224e-01 -2.01006442e-01 -2.45377913e-01 1.63241550e-02 3.61555904e-01 1.14717647e-01 7.30834067e-01 -2.64404416e-02 2.13060066e-01 -7.44080007e-01 -5.20183444e-01 7.69083619e-01 3.94933701e-01 -4.34196413e-01 2.29288973e-02 1.91280112e-01 4.22110558e-01 -5.11480570e-01 8.52701604e-01 6.60665929e-01 2.78430611e-01 -1.81210816e-01 1.47315636e-01 -2.68738478e-01 -9.96623784e-02 -1.52112222e+00 2.01973128e+00 -1.15272533e-02 7.97224045e-01 -9.22724158e-02 -1.06613994e+00 1.14730537e+00 2.99595952e-01 1.02328384e+00 -6.88502848e-01 1.96764488e-02 2.93301284e-01 -3.73971194e-01 -7.48844683e-01 4.39819634e-01 1.41687140e-01 5.97897917e-03 2.25476801e-01 1.70043781e-01 -2.22328186e-01 4.13747936e-01 -2.52550662e-01 1.44352067e+00 4.80597228e-01 4.41417694e-01 3.04371119e-02 7.25563943e-01 5.35050500e-03 3.35684568e-01 5.73439837e-01 -5.24002552e-01 7.98734069e-01 6.69492722e-01 -6.08679831e-01 -8.42197299e-01 -8.76621544e-01 -2.60433167e-01 4.47975010e-01 1.36115819e-01 -3.44132125e-01 -5.85058868e-01 -6.90649748e-01 1.77006781e-01 1.88685924e-01 -7.18549192e-01 -3.93574089e-02 -4.31325972e-01 -3.08838248e-01 4.35354114e-01 4.93745595e-01 9.29821551e-01 -1.04257369e+00 -1.32680213e+00 3.73710841e-02 2.06306446e-02 -1.32581091e+00 1.67962715e-01 1.32940309e-02 -1.17622268e+00 -1.34318244e+00 -9.74013269e-01 -2.87581831e-01 1.64769098e-01 3.33484858e-01 1.00929642e+00 -2.34987721e-01 -5.12285471e-01 9.70084429e-01 -8.24787140e-01 -2.44777203e-01 -1.98552281e-01 -2.58879781e-01 1.45459741e-01 9.04933661e-02 7.69851089e-01 -6.58973813e-01 -3.29936534e-01 5.71865916e-01 -7.58428156e-01 -2.87053585e-01 4.24284399e-01 4.71014470e-01 5.46794415e-01 5.91845252e-03 -1.10169999e-01 -2.63251245e-01 -6.29560798e-02 -1.53910175e-01 -5.10258138e-01 -1.76030844e-01 3.23594473e-02 -2.64721990e-01 2.04684228e-01 -2.60005951e-01 -8.50508928e-01 3.46736401e-01 -2.15123564e-01 -6.27272248e-01 -7.41436005e-01 -1.13840014e-01 -4.85638738e-01 -2.10900471e-01 8.30492616e-01 1.45096123e-01 1.95068985e-01 -7.55034089e-01 -4.02800180e-02 7.92078376e-01 3.28570634e-01 -1.68693215e-01 3.75471294e-01 8.03410113e-01 2.73542672e-01 -1.23541057e+00 -5.94085634e-01 -7.66565263e-01 -1.14968002e+00 -7.12288022e-01 8.91654015e-01 -7.48412192e-01 -3.78202230e-01 1.14306617e+00 -1.02842021e+00 -2.65515655e-01 -5.79001665e-01 9.33021247e-01 -9.99898612e-01 6.27402246e-01 -3.05411667e-01 -6.41207457e-01 3.30211699e-01 -1.13955319e+00 1.23989487e+00 -2.15627681e-02 -3.86600867e-02 -7.18005478e-01 4.11219150e-01 5.47860503e-01 3.64390686e-02 6.75086081e-01 2.68160224e-01 -5.14191031e-01 -3.18990350e-01 -5.04905760e-01 1.19772851e-02 6.70441031e-01 2.46628970e-01 -6.10217415e-02 -1.07548916e+00 2.42237262e-02 2.12826326e-01 -4.80454177e-01 7.88191855e-01 3.29356730e-01 9.92605567e-01 3.07683080e-01 -1.49039865e-01 3.78677547e-01 1.12336493e+00 2.63576686e-01 1.04237366e+00 7.11705148e-01 3.91779482e-01 6.66838884e-01 8.62182617e-01 9.46828604e-01 -2.62250472e-02 1.13165843e+00 6.52852774e-01 2.60547310e-01 -8.51455480e-02 1.64544985e-01 4.55092221e-01 3.08500320e-01 -7.28440344e-01 -1.18885674e-01 -8.95433843e-01 1.98000610e-01 -1.58116603e+00 -1.26318359e+00 -2.06771150e-01 2.48088312e+00 3.29676092e-01 1.84665903e-01 4.00701821e-01 7.20439076e-01 4.99162495e-01 3.57421607e-01 -3.88261110e-01 -1.97132528e-01 -1.35065481e-01 1.03402779e-01 4.96332765e-01 2.30642691e-01 -1.49909592e+00 4.94219989e-01 6.32591963e+00 6.32087588e-01 -1.03794396e+00 -1.80393919e-01 8.58538672e-02 -2.30864853e-01 4.99641120e-01 -2.92532206e-01 -7.22340226e-01 4.85607445e-01 6.78091168e-01 2.13252217e-01 -9.76676792e-02 1.03902292e+00 2.71034598e-01 -7.02610731e-01 -1.11340308e+00 1.41055608e+00 4.63882744e-01 -9.70609307e-01 -1.72185525e-01 2.02215776e-01 4.62526768e-01 -1.56604648e-02 -2.73370951e-01 8.65406990e-02 -3.67408812e-01 -6.09965146e-01 5.80424726e-01 6.36257708e-01 6.01758599e-01 -7.11449683e-01 6.76636934e-01 3.52010578e-01 -9.16838706e-01 -1.64710581e-01 -4.34735447e-01 -4.52291489e-01 1.19812459e-01 4.70726222e-01 -2.68824369e-01 4.16968137e-01 1.02265620e+00 1.46217477e+00 -7.18684614e-01 1.27960038e+00 -1.77779421e-01 -5.85163906e-02 -2.87137240e-01 -2.27027908e-02 1.68739259e-01 -3.22000720e-02 7.69026101e-01 1.06337941e+00 4.74218041e-01 1.05053261e-01 -1.29491597e-01 1.33918181e-01 3.91680926e-01 7.97738284e-02 -1.03471923e+00 2.65477657e-01 -1.46153480e-01 9.11481738e-01 -6.77831352e-01 -2.73717105e-01 -5.53890646e-01 1.36494446e+00 -8.21272656e-02 -2.96885986e-02 -6.60343528e-01 -4.04017448e-01 8.59453261e-01 2.01247752e-01 3.13168257e-01 -5.16854525e-01 3.88012528e-01 -1.26515019e+00 3.27434361e-01 -9.15378809e-01 3.58788133e-01 -8.88243377e-01 -9.52380240e-01 1.68865100e-01 4.46234107e-01 -1.87197626e+00 -4.11954254e-01 -1.16137707e+00 -9.76034105e-02 2.20266566e-01 -1.16667938e+00 -4.32762593e-01 -6.42876983e-01 8.82604539e-01 7.11882830e-01 -2.47307196e-01 8.69947314e-01 5.26236892e-01 -1.34994596e-01 9.81409177e-02 2.86906242e-01 -6.93449797e-03 6.55510247e-01 -9.84688044e-01 1.88866645e-01 4.11047578e-01 2.91128129e-01 -3.10091496e-01 4.60052788e-01 -2.38419935e-01 -1.55947375e+00 -5.98635614e-01 6.98294640e-01 -8.52336168e-01 3.88694137e-01 -4.28676277e-01 -6.46528125e-01 4.56074089e-01 -2.46776968e-01 -1.37995754e-03 5.29921889e-01 -2.32766733e-01 -1.96307689e-01 -2.03564510e-01 -1.26959944e+00 1.76380008e-01 1.42679596e+00 -3.38600516e-01 -6.76082194e-01 4.80285794e-01 -4.86715473e-02 -6.47394896e-01 -9.69396651e-01 4.68768954e-01 8.35792005e-01 -1.73751438e+00 1.03345847e+00 -3.03676337e-01 3.61391813e-01 -1.24094196e-01 -4.37768608e-01 -1.06082904e+00 6.63223118e-02 -2.49277115e-01 -1.79081038e-01 8.15436721e-01 -2.85832807e-02 -4.70498234e-01 1.15507972e+00 4.00383294e-01 -1.26740217e-01 -3.50277543e-01 -1.60509169e+00 -1.03964353e+00 -2.76627243e-01 -6.74836040e-01 1.89160571e-01 5.40009260e-01 2.02077016e-01 -2.56758958e-01 -4.11134571e-01 -2.43314683e-01 5.32065690e-01 -5.07081673e-02 1.08861697e+00 -1.25702167e+00 -2.87105918e-01 -4.26128387e-01 -1.79500020e+00 -1.26463890e+00 -5.14169745e-02 -3.89721572e-01 -2.61803329e-01 -1.28264797e+00 -1.09305985e-01 -8.04347247e-02 5.06734177e-02 2.95245022e-01 5.43863058e-01 5.87701678e-01 2.27668598e-01 2.26797670e-01 -5.15450597e-01 5.32134116e-01 1.21799409e+00 -1.01584449e-01 1.03285678e-01 1.61925718e-01 3.04349631e-01 9.50330019e-01 6.20550573e-01 -3.06141257e-01 -3.86650294e-01 -3.19532573e-01 -1.86211318e-01 3.10841072e-02 5.48300982e-01 -1.84181082e+00 1.90781318e-02 -1.34876773e-01 5.92513859e-01 -7.64559209e-01 9.69119608e-01 -1.14234602e+00 5.66787645e-02 4.24673110e-01 -4.47423086e-02 -2.44686708e-01 2.51024842e-01 4.85440105e-01 -4.95116591e-01 -2.65136153e-01 7.75529563e-01 -3.71162623e-01 -1.25587356e+00 2.07413465e-01 -4.71671999e-01 7.37379044e-02 1.40430319e+00 -9.56589758e-01 1.09012090e-01 -2.83837348e-01 -9.05741692e-01 -2.29686767e-01 7.48236001e-01 5.47595143e-01 6.70634627e-01 -1.44503200e+00 -5.19321024e-01 4.62172955e-01 3.39404106e-01 -3.61328721e-01 5.13353407e-01 9.14843142e-01 -7.42983699e-01 4.20735151e-01 -8.13722968e-01 -8.73387992e-01 -1.47642899e+00 4.59350884e-01 4.27363604e-01 7.97300711e-02 -9.20957685e-01 5.24292469e-01 -1.91759482e-01 -2.42240906e-01 5.05662322e-01 -5.24153888e-01 -3.48493874e-01 4.21577871e-01 7.59567380e-01 7.29852080e-01 2.12561458e-01 -8.70257616e-01 -5.42087793e-01 1.05044031e+00 4.14833814e-01 -1.61884695e-01 1.17862117e+00 2.23831162e-02 2.69000322e-01 5.46042204e-01 1.19179583e+00 -2.76800931e-01 -1.51225603e+00 2.10489128e-02 -3.08531299e-02 -1.12246752e+00 -2.26879761e-01 -3.73090953e-01 -1.08972037e+00 1.04562604e+00 9.11494970e-01 3.82382393e-01 1.25998604e+00 -1.04304001e-01 3.37341368e-01 4.43881840e-01 8.26497197e-01 -1.28804886e+00 1.44132912e-01 5.17709792e-01 9.63128805e-01 -1.38700950e+00 2.96553105e-01 -2.94369251e-01 -5.49703598e-01 1.41166222e+00 3.81554455e-01 -1.96109325e-01 5.74672222e-01 1.60954729e-01 1.78956687e-02 -1.64997846e-01 -3.94810349e-01 -2.37606898e-01 -1.73034519e-01 1.03248978e+00 1.28170893e-01 -2.83453107e-01 -1.81080073e-01 -1.60973206e-01 1.89054266e-01 1.75600037e-01 5.89202583e-01 1.32080126e+00 -4.66748357e-01 -1.19998860e+00 -4.08122063e-01 3.55536729e-01 -3.09138149e-01 4.98021185e-01 -6.42269731e-01 1.15195036e+00 2.73764729e-01 7.91048944e-01 1.59147725e-01 -3.72180939e-01 8.57068300e-01 -6.66867048e-02 7.73920059e-01 -3.06344658e-01 -2.02590287e-01 -3.93229991e-01 2.12799147e-01 -1.23003006e+00 -1.07267189e+00 -1.15588498e+00 -8.31007004e-01 -2.56704539e-01 -1.87772755e-02 -3.10190350e-01 7.07368851e-01 6.83770239e-01 2.60568798e-01 -1.29264966e-01 5.75067461e-01 -1.36413956e+00 -4.35608923e-01 -8.34809959e-01 -1.11580133e+00 5.64494193e-01 1.41823992e-01 -1.20927012e+00 -5.41026831e-01 1.23179927e-01]
[7.872889995574951, 0.24741613864898682]
c117105f-f367-4856-ba53-9a5017118395
a-question-answering-approach-for-emotion
null
null
https://aclanthology.org/D17-1167
https://aclanthology.org/D17-1167.pdf
A Question Answering Approach for Emotion Cause Extraction
Emotion cause extraction aims to identify the reasons behind a certain emotion expressed in text. It is a much more difficult task compared to emotion classification. Inspired by recent advances in using deep memory networks for question answering (QA), we propose a new approach which considers emotion cause identification as a reading comprehension task in QA. Inspired by convolutional neural networks, we propose a new mechanism to store relevant context in different memory slots to model context information. Our proposed approach can extract both word level sequence features and lexical features. Performance evaluation shows that our method achieves the state-of-the-art performance on a recently released emotion cause dataset, outperforming a number of competitive baselines by at least 3.01{\%} in F-measure.
['Qin Lu', 'Jiachen Du', 'Jiannan Hu', 'Ruifeng Xu', 'Lin Gui', 'Yulan He']
2017-09-01
null
null
null
emnlp-2017-9
['emotion-cause-extraction']
['natural-language-processing']
[ 4.56004292e-01 9.26737711e-02 9.43467766e-02 -5.10724187e-01 -9.43344235e-01 -4.29569632e-01 5.34428716e-01 6.67846322e-01 -6.73845947e-01 7.64679253e-01 5.81890345e-01 -1.11003570e-01 5.89865111e-02 -7.10465550e-01 -6.26275480e-01 -2.68617839e-01 1.13755219e-01 1.30705044e-01 -1.34528503e-01 -4.20615107e-01 6.99654579e-01 9.90903936e-03 -1.79997468e+00 8.21703851e-01 1.04000819e+00 1.34886861e+00 5.83298206e-02 6.96681380e-01 -8.35819304e-01 1.34921145e+00 -7.64394462e-01 -4.58229929e-01 -6.81870759e-01 -7.67156482e-01 -1.51898909e+00 -3.99759799e-01 2.44137272e-01 1.64701149e-01 1.08678907e-01 8.01674426e-01 5.87042749e-01 3.63085598e-01 5.79547226e-01 -9.64708865e-01 -6.25628293e-01 3.01380187e-01 -3.56884697e-03 4.69358146e-01 8.11152220e-01 -4.31366265e-01 1.20139384e+00 -8.77031147e-01 6.06915236e-01 1.04595411e+00 2.79634982e-01 8.44335377e-01 -6.26019239e-01 -2.63212204e-01 2.06367448e-01 8.76304448e-01 -8.32165003e-01 -2.68898666e-01 8.44286501e-01 1.18731624e-02 1.69907415e+00 4.10841584e-01 3.35546821e-01 1.09661829e+00 4.95390117e-01 1.10709047e+00 1.25364137e+00 -5.95073283e-01 5.22725642e-01 -2.93763578e-01 7.02151835e-01 7.73408413e-01 -4.58857656e-01 -4.49533761e-01 -8.63784969e-01 -3.58679116e-01 -1.56196117e-01 -3.48958939e-01 -3.27638417e-01 2.21501440e-01 -9.23696637e-01 1.09958076e+00 3.54260504e-01 3.89287502e-01 -7.14736938e-01 3.52645248e-01 6.44857705e-01 4.63028014e-01 4.92849678e-01 7.37297058e-01 -9.05859947e-01 -6.36349857e-01 -4.87238586e-01 2.64234036e-01 1.03007066e+00 1.71805725e-01 4.97252434e-01 -2.70879149e-01 -4.43622559e-01 8.43943477e-01 -1.54390022e-01 2.58659691e-01 6.99931383e-01 -6.56523168e-01 2.84940332e-01 8.72845948e-01 4.88352887e-02 -1.01868069e+00 -8.11806023e-01 -4.57092375e-01 -6.94262385e-01 -4.45332170e-01 1.54940903e-01 -2.58994818e-01 -8.60272527e-01 1.91716754e+00 1.82671964e-01 1.74770683e-01 2.98458815e-01 9.12814558e-01 1.25833964e+00 8.72877479e-01 3.36037308e-01 -3.46643984e-01 1.79314780e+00 -9.33232427e-01 -1.14250338e+00 -4.85896170e-01 6.87621832e-01 -7.54375756e-01 1.08802617e+00 5.79253793e-01 -9.64140832e-01 -2.57912755e-01 -7.70629466e-01 -2.33091906e-01 -6.77242219e-01 -1.90831330e-02 9.08616364e-01 5.16051114e-01 -7.59908676e-01 2.03091860e-01 -8.11618119e-02 -4.48783308e-01 1.65469274e-01 1.77286744e-01 -1.60448238e-01 1.07409149e-01 -1.89399302e+00 9.75327075e-01 1.71332017e-01 4.69664447e-02 -4.65202838e-01 -4.74422783e-01 -8.20593774e-01 2.66236335e-01 3.97355825e-01 -7.58346558e-01 1.22933733e+00 -9.99506176e-01 -1.51476383e+00 9.36218560e-01 -7.97114551e-01 -4.99178946e-01 -3.10568511e-01 -7.47804463e-01 -7.10800707e-01 2.99973905e-01 -1.61039993e-01 4.94609803e-01 7.52176404e-01 -6.66698873e-01 -3.85627389e-01 -3.73114556e-01 -1.47587091e-01 2.27984384e-01 -4.46863979e-01 3.55311394e-01 -1.30736157e-01 -3.30111623e-01 -1.29163519e-01 -5.56045115e-01 -1.17535688e-01 -6.66955709e-01 -1.08343326e-01 -7.67949462e-01 7.12339878e-01 -7.39094257e-01 1.44598055e+00 -1.74105573e+00 1.45150945e-01 -2.81487443e-02 1.16471462e-01 8.99008587e-02 -4.27042365e-01 5.02549529e-01 -2.02866346e-01 6.28412366e-02 -1.98391557e-01 9.23744310e-03 3.06035459e-01 1.49249882e-01 -6.79295361e-01 -1.97563753e-01 5.57970583e-01 1.18759239e+00 -8.35722923e-01 -2.05956504e-01 -2.37493575e-01 2.50603229e-01 -3.93732369e-01 4.62876350e-01 -4.37962413e-01 1.17677590e-02 -3.86512727e-01 4.66691256e-01 3.68975490e-01 -2.79072553e-01 5.79554364e-02 2.03179538e-01 3.28436434e-01 8.59529436e-01 -7.54305243e-01 1.87727773e+00 -5.45479298e-01 5.05161345e-01 -2.33703569e-01 -1.17768991e+00 1.22791791e+00 4.74425822e-01 2.88851634e-02 -1.16412556e+00 3.29827785e-01 1.39618605e-01 2.70395335e-02 -9.12210464e-01 6.37120008e-01 -2.62624741e-01 -5.76095760e-01 4.92420375e-01 2.90055852e-02 2.83245325e-01 -3.05029023e-02 1.99742034e-01 1.50252759e+00 -3.23037982e-01 2.60371864e-01 -1.67435914e-01 7.47205079e-01 -5.21084182e-02 5.88230610e-01 6.38579309e-01 -4.00454372e-01 1.95143059e-01 6.84408009e-01 -6.15668297e-01 -3.80011886e-01 -6.65082514e-01 1.51556328e-01 1.48157561e+00 -6.84030876e-02 -4.82927561e-01 -8.68056357e-01 -8.40485871e-01 -3.85482699e-01 9.19063687e-01 -9.95870471e-01 -3.52800190e-01 -5.11668563e-01 -8.27569664e-01 6.34867728e-01 4.46468025e-01 8.34069371e-01 -1.64325690e+00 -8.23660493e-01 3.35783541e-01 -8.59323621e-01 -1.04456985e+00 1.15978094e-02 5.15806079e-01 -6.70348406e-01 -9.06621337e-01 -4.66252267e-01 -8.46832991e-01 1.69716820e-01 -3.11007082e-01 1.71677685e+00 2.54074812e-01 -3.09596419e-01 3.57710063e-01 -6.82348967e-01 -5.24993598e-01 -5.75529188e-02 3.05044681e-01 -3.82544786e-01 -8.50477666e-02 1.07223833e+00 -1.87150374e-01 -5.99464834e-01 -2.90678382e-01 -8.22623670e-01 -2.10305780e-01 5.85632026e-01 8.61311972e-01 3.68706018e-01 -9.78906080e-02 1.08854461e+00 -5.14140368e-01 1.14448619e+00 -6.63628161e-01 1.85913563e-01 4.29694474e-01 -5.43406844e-01 2.23642945e-01 3.47304851e-01 -7.04745129e-02 -1.18278456e+00 -9.78482142e-02 -4.69313651e-01 2.46080711e-01 -7.35697210e-01 7.58175075e-01 -2.83967406e-01 3.94829512e-01 3.41478646e-01 3.19770187e-01 -1.95862561e-01 -3.52840871e-01 4.82701182e-01 5.44882119e-01 4.51467514e-01 -4.46021736e-01 -1.78037703e-01 1.15031250e-01 -1.63085803e-01 -4.82126534e-01 -1.26110220e+00 -4.63331908e-01 -2.66105235e-01 -1.80776626e-01 1.19741797e+00 -5.28876305e-01 -9.62042093e-01 2.09313840e-01 -1.42218995e+00 -9.08636209e-03 1.94877967e-01 -2.64805779e-02 -4.30776238e-01 1.98051035e-01 -6.82937086e-01 -8.03185403e-01 -7.89667964e-01 -5.93389988e-01 1.09841990e+00 2.74109334e-01 -7.12495625e-01 -8.12862754e-01 1.78956136e-01 6.23399377e-01 5.06953597e-01 1.55575514e-01 1.24461758e+00 -9.74387646e-01 -2.96574663e-02 -1.46067515e-01 -1.39838964e-01 -9.06995833e-02 -1.82772383e-01 -5.82072437e-01 -1.07248664e+00 3.21060151e-01 2.42877111e-01 -7.69638479e-01 1.24810421e+00 3.32018808e-02 1.27504158e+00 -2.58553118e-01 -1.85647942e-02 -1.34113669e-01 1.18174052e+00 3.64992857e-01 9.19301689e-01 4.05291945e-01 2.24767447e-01 8.58635545e-01 6.62326813e-01 4.42470014e-01 6.79668307e-01 3.35020244e-01 5.34959495e-01 5.53843239e-03 2.41925746e-01 7.82513618e-02 3.68773967e-01 6.77101374e-01 3.47498655e-01 -2.38803893e-01 -8.94221187e-01 7.37528682e-01 -1.94209850e+00 -1.10229969e+00 -3.88619691e-01 1.49571311e+00 9.65332031e-01 -1.50715917e-01 -2.31868997e-01 1.64162725e-01 3.71921450e-01 2.60849446e-01 -4.53725308e-01 -1.38822889e+00 -2.44371146e-01 6.94527626e-01 -3.86683822e-01 4.13223624e-01 -1.19249678e+00 1.05549777e+00 6.17075443e+00 8.88233185e-01 -9.64614034e-01 1.53567865e-01 9.04844344e-01 7.72630610e-03 -3.05653036e-01 -2.80216455e-01 -4.36514884e-01 1.97608232e-01 1.17118216e+00 2.33425736e-01 1.25248477e-01 6.39316738e-01 6.21010782e-03 -4.45614249e-01 -7.15259194e-01 9.55381453e-01 4.98429924e-01 -1.42147768e+00 7.31695071e-02 -4.18801844e-01 5.22172570e-01 -2.65964627e-01 -2.32506413e-02 6.20312393e-01 -2.29084730e-01 -1.18817019e+00 2.37521172e-01 8.01240623e-01 5.13862260e-02 -1.19144034e+00 1.02892566e+00 2.59309858e-01 -7.41367996e-01 -1.23902880e-01 -3.61425191e-01 -6.98431134e-01 1.74048141e-01 8.09380352e-01 -5.40792286e-01 3.67408603e-01 8.14703703e-01 2.81366885e-01 -5.59275389e-01 6.54279053e-01 -5.29310763e-01 8.28398466e-01 1.98379725e-01 -6.61741734e-01 1.82952806e-01 4.57611173e-01 2.96597004e-01 1.36111856e+00 1.19696647e-01 3.88808638e-01 -2.38192052e-01 7.31090248e-01 -3.64323258e-01 5.09475112e-01 -2.28639692e-01 -1.39930621e-01 7.73254782e-02 1.36818469e+00 -6.55164659e-01 -4.95190054e-01 -4.47386988e-02 1.49281585e+00 5.78944862e-01 1.51263699e-01 -7.70118952e-01 -7.41569340e-01 6.27891004e-01 -7.11187065e-01 4.64537293e-01 6.09448142e-02 -3.30031127e-01 -1.08344507e+00 5.77993765e-02 -9.08309400e-01 7.06187844e-01 -9.51471806e-01 -1.38964438e+00 8.94617200e-01 -6.06108010e-01 -5.37492394e-01 -5.10614514e-01 -6.62981093e-01 -7.71706104e-01 7.53155470e-01 -1.72086823e+00 -6.95158660e-01 -3.15541476e-01 5.51462293e-01 4.92809743e-01 1.91758070e-02 1.37940764e+00 -5.68566807e-02 -4.03562725e-01 3.74309868e-01 -2.52977699e-01 8.10722783e-02 7.25945830e-01 -1.30799675e+00 9.46601108e-02 7.22315073e-01 3.15804183e-01 5.68909526e-01 7.63547242e-01 -5.09301424e-01 -1.44172800e+00 -5.91336012e-01 1.82177997e+00 -5.38477063e-01 5.58243811e-01 -4.13030237e-01 -1.01922500e+00 8.73988867e-02 9.15050507e-01 -4.76706594e-01 1.15099204e+00 5.21908343e-01 -6.00405216e-01 1.96643919e-01 -1.09771466e+00 2.65686154e-01 8.09032083e-01 -6.56030118e-01 -1.30010080e+00 3.69003825e-02 8.88310790e-01 -1.65742606e-01 -6.12027884e-01 2.91834474e-01 2.71136135e-01 -9.22396779e-01 6.55501604e-01 -9.90689695e-01 9.29741144e-01 1.25231901e-02 -2.08556652e-01 -1.27182913e+00 -1.10641927e-01 -6.15960419e-01 -3.57816309e-01 1.27227902e+00 4.18831587e-01 -3.22879255e-01 6.64130807e-01 7.96624064e-01 3.04394346e-02 -9.62899148e-01 -1.00175571e+00 -4.35201645e-01 2.53358275e-01 -7.04559326e-01 6.84240699e-01 1.06132889e+00 4.46436524e-01 8.88453364e-01 -2.48058319e-01 -1.13794491e-01 6.90932795e-02 6.74500167e-01 9.61997285e-02 -1.08713531e+00 -4.06527109e-02 -6.19903684e-01 -8.75297561e-02 -8.26134145e-01 6.19281232e-01 -8.48687947e-01 1.88601181e-01 -1.77692044e+00 3.80380571e-01 2.45112628e-01 -6.85956895e-01 6.91352069e-01 -6.48180962e-01 1.60907030e-01 -2.37044878e-02 -4.76207972e-01 -1.03208303e+00 8.27760458e-01 8.39257061e-01 -3.08591407e-02 -1.75757054e-02 -5.25067329e-01 -7.61909246e-01 6.46458626e-01 1.06030011e+00 -2.08215252e-01 -9.52810794e-02 -3.27270776e-01 8.95194054e-01 -6.45042136e-02 3.99419636e-01 -7.30252385e-01 9.19109397e-03 -3.03971712e-02 3.28821182e-01 -6.23186707e-01 1.72913253e-01 -4.22841787e-01 -7.69996941e-01 3.53517383e-01 -7.12248147e-01 2.50865340e-01 5.04730463e-01 3.50248516e-01 -5.39207280e-01 -3.50337327e-01 5.93769997e-02 -4.22607251e-02 -1.28711414e+00 -2.03557521e-01 -7.90460169e-01 2.18314230e-01 5.10303259e-01 3.08684170e-01 -6.81186318e-01 -6.60408080e-01 -6.94434643e-01 1.09654807e-01 -3.21389824e-01 7.71610856e-01 8.51024628e-01 -1.22536397e+00 -5.09295344e-01 -2.52331853e-01 3.02209646e-01 -6.69218063e-01 4.52032924e-01 7.15997040e-01 1.60387337e-01 7.26615429e-01 -6.61391094e-02 -1.58407744e-02 -1.42002833e+00 4.58443463e-01 2.36548662e-01 -4.57515180e-01 -1.52739301e-01 1.09501100e+00 -8.39692578e-02 -3.96909058e-01 1.10892534e-01 -1.68009534e-01 -6.28644586e-01 2.94312239e-01 8.14354837e-01 1.38863534e-01 4.00952756e-01 -3.89765143e-01 -5.69759071e-01 2.84350872e-01 -4.16952036e-02 1.86450372e-03 1.25396907e+00 -1.28886029e-01 -6.18216693e-01 4.09663051e-01 1.26921988e+00 -1.99031785e-01 -3.45504820e-01 -1.19053654e-01 5.13028622e-01 1.54788226e-01 1.99890673e-01 -1.61881149e+00 -5.93179703e-01 1.17865610e+00 6.58849835e-01 3.01005840e-01 1.41539168e+00 1.57357767e-01 1.27094960e+00 6.66572213e-01 -4.36803438e-02 -1.34627593e+00 3.98208320e-01 1.13735449e+00 1.13999283e+00 -1.27043653e+00 -5.91787994e-01 -1.24784172e-01 -6.35720193e-01 1.07645822e+00 7.17392445e-01 -1.00216337e-01 3.42639714e-01 -9.64290202e-02 4.40236330e-01 -6.10091925e-01 -1.23125982e+00 -5.96716583e-01 6.86979055e-01 4.09839638e-02 8.63598824e-01 -9.89210159e-02 -8.63649726e-01 1.31991577e+00 -1.22958340e-01 -2.49827951e-02 2.52129167e-01 1.02338874e+00 -7.32242227e-01 -1.28652859e+00 -3.06800246e-01 4.84334856e-01 -7.96171486e-01 -4.39801991e-01 -1.02726340e+00 2.79764831e-01 7.30254278e-02 1.37597477e+00 5.16794845e-02 -3.93883824e-01 1.82548568e-01 9.21436787e-01 3.17619294e-01 -2.77871370e-01 -9.58589673e-01 -8.89637172e-02 3.78213376e-01 -9.21867430e-01 -5.85365593e-01 -6.32643044e-01 -1.68798602e+00 9.60600302e-02 7.10297450e-02 3.18947345e-01 5.65179646e-01 1.33231878e+00 8.04967999e-01 6.75930381e-01 3.85027945e-01 -9.81591456e-03 -7.56285861e-02 -9.68489051e-01 -1.30039647e-01 5.84475636e-01 1.86753452e-01 -4.26770806e-01 -2.15123728e-01 -1.59186885e-01]
[12.667574882507324, 6.21804666519165]
e37c0742-4ab3-4497-a0ab-f8f1b45e259e
active-cost-aware-labeling-of-streaming-data
2304.06808
null
https://arxiv.org/abs/2304.06808v2
https://arxiv.org/pdf/2304.06808v2.pdf
Active Cost-aware Labeling of Streaming Data
We study actively labeling streaming data, where an active learner is faced with a stream of data points and must carefully choose which of these points to label via an expensive experiment. Such problems frequently arise in applications such as healthcare and astronomy. We first study a setting when the data's inputs belong to one of $K$ discrete distributions and formalize this problem via a loss that captures the labeling cost and the prediction error. When the labeling cost is $B$, our algorithm, which chooses to label a point if the uncertainty is larger than a time and cost dependent threshold, achieves a worst-case upper bound of $\widetilde{O}(B^{\frac{1}{3}} K^{\frac{1}{3}} T^{\frac{2}{3}})$ on the loss after $T$ rounds. We also provide a more nuanced upper bound which demonstrates that the algorithm can adapt to the arrival pattern, and achieves better performance when the arrival pattern is more favorable. We complement both upper bounds with matching lower bounds. We next study this problem when the inputs belong to a continuous domain and the output of the experiment is a smooth function with bounded RKHS norm. After $T$ rounds in $d$ dimensions, we show that the loss is bounded by $\widetilde{O}(B^{\frac{1}{d+3}} T^{\frac{d+2}{d+3}})$ in an RKHS with a squared exponential kernel and by $\widetilde{O}(B^{\frac{1}{2d+3}} T^{\frac{2d+2}{2d+3}})$ in an RKHS with a Mat\'ern kernel. Our empirical evaluation demonstrates that our method outperforms other baselines in several synthetic experiments and two real experiments in medicine and astronomy.
['Kirthevasan Kandasamy', 'Ting Cai']
2023-04-13
null
null
null
null
['astronomy']
['miscellaneous']
[ 3.27600569e-01 3.18680882e-01 -1.62944973e-01 -3.84778112e-01 -1.35050678e+00 -6.72015309e-01 -6.19928464e-02 5.29307067e-01 -1.00371420e+00 8.09855759e-01 -3.65163326e-01 -4.29510385e-01 -5.50485373e-01 -7.83452153e-01 -1.07584977e+00 -1.03348517e+00 -9.22641635e-01 5.26708543e-01 2.59898096e-01 8.14717785e-02 -3.43118124e-02 3.41487020e-01 -1.46413875e+00 -1.66646764e-01 6.11698866e-01 1.52764821e+00 -3.86676043e-01 8.66398513e-01 1.86315104e-01 4.76374656e-01 -7.46793747e-01 -2.89674670e-01 5.49227476e-01 -4.22052264e-01 -6.21867299e-01 1.13819852e-01 1.55063480e-01 -5.93349598e-02 -4.72022109e-02 1.23999000e+00 6.11566186e-01 3.85981053e-01 8.91520560e-01 -1.09156477e+00 1.50524184e-01 2.81255305e-01 -8.28013837e-01 2.25706324e-01 2.21951410e-01 -7.16359774e-03 8.16836476e-01 -5.34850538e-01 2.95870095e-01 7.40042508e-01 7.34624922e-01 3.27470571e-01 -1.18970585e+00 -8.87146592e-01 5.12638569e-01 -4.42903697e-01 -1.35668361e+00 -1.62070826e-01 4.35675263e-01 -5.04362106e-01 3.64062965e-01 4.31951404e-01 3.98498237e-01 4.55225825e-01 -5.56490161e-02 9.47548449e-01 9.80146468e-01 -3.21882039e-01 6.51394606e-01 1.43922582e-01 5.51172346e-02 8.11057746e-01 -7.39788115e-02 1.11446783e-01 -1.82284802e-01 -3.36664945e-01 7.37393320e-01 2.24104539e-01 -2.61264265e-01 -7.70674720e-02 -8.15570176e-01 1.03188872e+00 9.85952020e-02 -9.64033231e-02 -2.45289639e-01 3.76624793e-01 2.09764779e-01 3.28049034e-01 3.75890762e-01 2.69875377e-01 -5.35043478e-01 -2.11629659e-01 -1.02568173e+00 2.99866676e-01 6.59607232e-01 1.25428557e+00 5.96302032e-01 -2.75237769e-01 -9.51401591e-02 7.13597953e-01 1.08997576e-01 6.73711717e-01 -5.72326072e-02 -1.17060411e+00 6.11074269e-01 2.86129951e-01 4.26250398e-01 -5.48495054e-01 -5.37142515e-01 -3.27140689e-01 -8.54013860e-01 3.07965755e-01 8.19266021e-01 -4.69060212e-01 -7.36245036e-01 1.92824078e+00 3.98649126e-01 -7.45085403e-02 -2.28848502e-01 6.95281744e-01 1.99452460e-01 5.53131223e-01 1.23821069e-02 -7.58811772e-01 1.20156968e+00 -1.55840799e-01 -2.60018855e-01 1.07284836e-01 6.60784781e-01 -5.67689955e-01 1.23848915e+00 6.95663691e-01 -1.53009903e+00 -2.49071658e-01 -8.87951314e-01 4.53430980e-01 6.56291321e-02 -3.90271187e-01 4.08743680e-01 7.09976375e-01 -6.58660054e-01 5.41761398e-01 -8.17624867e-01 2.84261823e-01 5.72409987e-01 5.59462667e-01 -1.43821388e-01 -1.40525579e-01 -1.10187507e+00 1.44446298e-01 -4.01813909e-03 -2.24122837e-01 -8.98047090e-01 -8.21987271e-01 -5.96421003e-01 -6.22014217e-02 5.50499976e-01 -1.34614527e-01 1.13957882e+00 -6.09021723e-01 -9.65999424e-01 7.32675016e-01 1.53767973e-01 -4.15467769e-01 8.85886014e-01 1.02240279e-01 -3.48803043e-01 2.49726906e-01 1.70246288e-01 2.84065157e-01 5.17302632e-01 -1.03645349e+00 -9.75116789e-01 -7.12429523e-01 1.86318129e-01 1.47065967e-01 -1.61976099e-01 2.88509801e-02 -3.77965778e-01 -7.74529457e-01 1.40373811e-01 -1.00413513e+00 -4.26711828e-01 6.72086403e-02 -2.92012304e-01 -1.06013201e-01 3.44609231e-01 -5.05206585e-02 1.42404735e+00 -2.12941551e+00 -2.71203935e-01 5.80345154e-01 2.35302687e-01 2.91601717e-02 5.01624048e-01 1.74170613e-01 1.28542915e-01 2.52741486e-01 -5.85738778e-01 -3.02179486e-01 7.56497160e-02 7.40626454e-02 -1.13450497e-01 6.54341698e-01 -3.60402942e-01 2.01871589e-01 -8.72612357e-01 -1.73039198e-01 -2.58304119e-01 1.66851759e-01 -4.56550241e-01 -8.13131512e-04 -2.16094017e-01 2.34146744e-01 -6.25273347e-01 3.26850086e-01 5.91035485e-01 -5.15541732e-01 -1.56326517e-01 3.76722455e-01 5.78642227e-02 -1.16952084e-01 -1.78696477e+00 1.40797222e+00 -1.18843339e-01 2.72140950e-02 2.54930377e-01 -1.35584700e+00 8.40929925e-01 3.07138085e-01 9.89026606e-01 -4.77027625e-01 4.93981969e-03 2.14952782e-01 -2.93882847e-01 -2.02409804e-01 -6.99119046e-02 -6.69826388e-01 -3.60435754e-01 6.31719232e-01 -3.65411758e-01 -6.12429976e-02 -1.51203930e-01 1.43616110e-01 1.50072980e+00 -2.85312474e-01 -1.16957866e-01 -3.79069418e-01 2.51084149e-01 -2.16141969e-01 7.55654752e-01 9.86236334e-01 -3.57186437e-01 4.64936972e-01 1.01035535e+00 -2.91532665e-01 -8.35217059e-01 -1.11497045e+00 -4.49460655e-01 9.88058448e-01 1.67259872e-01 2.93003749e-02 -6.15011334e-01 -9.86425698e-01 1.44080585e-02 5.79419851e-01 -7.48475313e-01 3.67503017e-02 -4.22280133e-01 -1.04867041e+00 5.33771098e-01 7.56865919e-01 4.20579076e-01 -8.41839790e-01 -8.64939272e-01 1.59906760e-01 2.51780786e-02 -6.56898379e-01 -7.97051251e-01 8.22558403e-01 -6.83114231e-01 -9.91756260e-01 -8.69823992e-01 -5.08645535e-01 9.11873102e-01 -2.70725280e-01 1.01690209e+00 -2.28833839e-01 -3.77178848e-01 5.43564856e-01 -2.39163369e-01 -9.54289138e-01 1.74599290e-01 -2.26256922e-01 1.11369282e-01 -1.48469850e-01 3.08971375e-01 -3.92894208e-01 -1.05181682e+00 3.10534060e-01 -1.14579737e+00 -5.53391576e-01 7.96412453e-02 8.38933587e-01 1.05642319e+00 3.68971884e-01 6.94762945e-01 -1.08406425e+00 4.01668608e-01 -4.57069725e-01 -9.91846919e-01 4.11225669e-02 -6.09201193e-01 4.81002033e-02 6.97710693e-01 -5.99525154e-01 -5.00322044e-01 2.82331467e-01 -6.42493069e-02 -6.13060892e-01 -8.44805986e-02 3.28667104e-01 -6.87464550e-02 4.46635783e-01 8.28185678e-01 8.19642395e-02 -7.00864792e-02 -3.50736439e-01 -3.13248374e-02 4.11721826e-01 5.58123112e-01 -8.44726145e-01 5.21810949e-01 5.66497028e-01 1.73860654e-01 -4.54583704e-01 -8.64734828e-01 -3.27021986e-01 -2.00004682e-01 -3.51888984e-02 6.70211613e-01 -7.47791171e-01 -1.27843571e+00 -8.69182795e-02 -4.28045452e-01 -5.20998299e-01 -9.46425974e-01 7.20510542e-01 -8.78907561e-01 3.00892472e-01 -3.79164934e-01 -1.27894735e+00 -2.18664184e-01 -1.15142703e+00 8.80643725e-01 1.17455013e-01 -1.39493421e-01 -8.34104061e-01 -1.39184922e-01 2.08484545e-01 8.54602456e-02 4.94239330e-01 9.90209699e-01 -8.74178052e-01 -3.37875485e-01 -6.15349174e-01 2.40061954e-02 5.06096780e-01 -6.51660338e-02 -3.47739041e-01 -6.36428893e-01 -3.76063585e-01 1.94281474e-01 -3.18016708e-01 5.75707257e-01 4.92592543e-01 1.53610122e+00 -5.95898867e-01 -2.26227403e-01 3.81976932e-01 1.23318040e+00 6.93874061e-01 4.61289942e-01 5.14073856e-03 -1.32484064e-02 4.72706586e-01 5.92558622e-01 8.16917241e-01 1.77356735e-01 3.76097113e-01 4.03320163e-01 -1.07719436e-01 4.82294083e-01 1.44807056e-01 1.13511562e-01 1.67109579e-01 1.50533050e-01 -3.04520696e-01 -8.38525474e-01 6.15590453e-01 -1.72073376e+00 -7.50342906e-01 1.59092307e-01 2.87911820e+00 1.14541841e+00 5.96617341e-01 4.29541439e-01 5.05390048e-01 5.84194005e-01 -2.60355920e-01 -6.53484046e-01 -3.97143960e-01 1.26692653e-01 7.88122296e-01 6.68236792e-01 3.77769649e-01 -1.02284396e+00 2.82097667e-01 4.53687477e+00 9.57599044e-01 -9.84254837e-01 -2.57087350e-01 1.06073010e+00 -5.39724290e-01 -1.56917259e-01 -1.53058141e-01 -8.03500652e-01 8.87250543e-01 9.31174457e-01 -1.20756738e-02 7.11771846e-02 9.00597215e-01 -6.92923144e-02 -4.25298959e-01 -1.47073615e+00 8.40721250e-01 -2.66510606e-01 -1.15951002e+00 -7.10093081e-01 3.11525524e-01 6.45329952e-01 -2.22662151e-01 2.76467711e-01 2.85801917e-01 7.03683376e-01 -1.23828197e+00 5.64500511e-01 1.69641793e-01 1.26803267e+00 -9.10880208e-01 5.03833771e-01 6.99228823e-01 -1.10085464e+00 3.26838391e-03 -1.27054513e-01 1.32934272e-01 7.87116438e-02 7.99851418e-01 -7.00333595e-01 3.31878155e-01 9.68798578e-01 -7.38347992e-02 3.99100721e-01 9.09602344e-01 1.19266205e-01 5.34202099e-01 -7.35431015e-01 1.29429102e-01 1.11774519e-01 3.44982892e-02 4.24609661e-01 1.21860576e+00 2.94117957e-01 7.22545147e-01 5.93472898e-01 3.90101701e-01 -2.61763126e-01 2.04474241e-01 -3.31925303e-01 2.51589388e-01 6.11046553e-01 7.42797554e-01 -8.91898930e-01 -2.83890426e-01 -2.50443131e-01 5.71634650e-01 4.19751704e-02 3.91899735e-01 -8.68945122e-01 -8.73715758e-01 4.95027810e-01 5.17561316e-01 4.78867084e-01 1.15707472e-01 -2.44468004e-01 -6.86885417e-01 1.99581578e-01 -5.45104384e-01 9.32132781e-01 -1.00312822e-01 -1.39789438e+00 3.65230799e-01 2.20942363e-01 -1.32919955e+00 -1.39257908e-01 -4.26104277e-01 -2.00827271e-01 9.71304595e-01 -9.23305511e-01 -2.42054909e-01 3.82674448e-02 1.04597723e+00 -8.92943144e-03 -7.59921372e-02 8.79491508e-01 3.58688533e-01 -2.69257188e-01 9.71556306e-01 2.37107024e-01 3.44779760e-01 3.35622221e-01 -1.27629054e+00 -1.56137362e-01 4.43842083e-01 -1.88015491e-01 3.24347168e-01 6.29106343e-01 -4.47860092e-01 -9.61650670e-01 -9.72971797e-01 5.78633606e-01 -3.49991649e-01 2.85556495e-01 -3.71954262e-01 -9.34995830e-01 7.44012773e-01 -5.17307341e-01 6.96639240e-01 9.20486391e-01 -1.51602969e-01 -2.96518743e-01 -3.37901473e-01 -1.69291806e+00 2.62288988e-01 8.10925126e-01 -2.04360411e-01 -3.63236852e-02 4.08310711e-01 4.08280134e-01 -5.65288544e-01 -1.35847807e+00 7.38512039e-01 4.79446113e-01 -9.14996505e-01 6.69348300e-01 -7.28143573e-01 1.15340047e-01 -1.16006404e-01 -4.86078143e-01 -8.05875003e-01 8.31903219e-02 -8.49460185e-01 -4.04447019e-02 8.14095438e-01 7.05649316e-01 -5.75413227e-01 1.07398605e+00 9.38818097e-01 9.31315571e-02 -1.09332538e+00 -1.30353010e+00 -6.79176629e-01 7.00160027e-01 -5.62186420e-01 3.39129120e-01 8.22237253e-01 5.62420785e-02 -1.89089134e-01 -1.05676435e-01 9.06699896e-02 5.54948032e-01 -1.62785068e-01 3.75633299e-01 -1.22868192e+00 -7.19883025e-01 -2.83095449e-01 -1.30935386e-01 -1.01881933e+00 -4.84952867e-01 -5.32740891e-01 1.19683944e-01 -1.00573921e+00 1.11328498e-01 -1.22753775e+00 -6.69718742e-01 2.61005402e-01 -4.81813028e-02 -1.87143326e-01 -2.29764525e-02 -2.87831537e-02 -7.39856720e-01 1.79858834e-01 9.49278235e-01 7.70645514e-02 -3.39278698e-01 4.94529158e-01 -6.05619848e-01 8.02473247e-01 3.82965297e-01 -5.50024092e-01 -6.75199986e-01 -1.78337783e-01 4.04236019e-01 9.18138504e-01 -5.78413457e-02 -6.88790500e-01 2.49820918e-01 -2.64979452e-01 4.23207968e-01 -4.64381456e-01 4.18591887e-01 -7.70593643e-01 4.18032072e-02 4.00457382e-01 -6.43942237e-01 -4.53224918e-03 6.10365607e-02 7.95756459e-01 1.16425395e-01 -9.25916508e-02 1.15258467e+00 -3.06229949e-01 2.32119277e-01 6.63881898e-01 -1.32001042e-01 5.21513164e-01 1.31406283e+00 -1.88976392e-01 -1.10232055e-01 -6.25575185e-01 -1.01714599e+00 4.90509838e-01 2.04885855e-01 -1.61768556e-01 2.92633712e-01 -1.13259363e+00 -6.66731417e-01 2.99299091e-01 6.01966381e-02 8.84477437e-01 1.92314953e-01 8.70666742e-01 -2.37171590e-01 -1.89567357e-01 4.54518199e-01 -7.01852322e-01 -9.13765848e-01 6.14113092e-01 4.14231271e-01 -3.74127328e-01 -4.47345287e-01 1.43710315e+00 2.03526482e-01 -3.29876691e-02 8.52038383e-01 -2.99973488e-01 3.21645409e-01 5.20475172e-02 4.30868357e-01 7.17066705e-01 2.53769875e-01 -1.54715940e-01 -2.81921506e-01 1.56299487e-01 -2.54754514e-01 -3.94358218e-01 1.22485363e+00 -7.78964208e-03 3.71046066e-01 5.90756893e-01 1.38120210e+00 -1.12096807e-02 -1.57167101e+00 -3.71538341e-01 -4.45022769e-02 -3.04158151e-01 -3.95160317e-01 -9.10319328e-01 -1.02921200e+00 8.16150904e-01 8.47147048e-01 4.04152840e-01 1.26040399e+00 9.34451595e-02 5.33237338e-01 -5.66799007e-02 4.60157186e-01 -1.46130848e+00 3.11487734e-01 1.77257583e-01 5.13246477e-01 -1.20436466e+00 -5.11254184e-02 -4.98566441e-02 -7.48892307e-01 8.07346046e-01 2.17306927e-01 -1.17150649e-01 1.10088563e+00 4.29388314e-01 -9.86242890e-02 -2.13803038e-01 -5.36679566e-01 1.26679122e-01 1.45131543e-01 1.62462056e-01 3.74611229e-01 3.03342402e-01 -1.80313513e-01 9.17663872e-01 -2.28840366e-01 2.84715910e-02 3.57005179e-01 1.18127632e+00 -5.02453268e-01 -9.15980995e-01 -2.14942157e-01 6.38431728e-01 -9.21138704e-01 2.16055766e-01 2.64908731e-01 7.53287435e-01 3.28544974e-01 9.50710773e-01 2.17378825e-01 9.76591483e-02 4.97088939e-01 2.26074815e-01 4.56996292e-01 -6.87768221e-01 -4.03232187e-01 6.12567067e-01 -1.94810674e-01 -3.52413207e-01 -1.37801811e-01 -8.20608318e-01 -1.86119568e+00 -9.80720073e-02 4.42800019e-03 3.81483972e-01 3.80996287e-01 7.20566034e-01 7.24788988e-04 2.30660185e-01 9.84554768e-01 -3.98151651e-02 -7.95230150e-01 -5.68618000e-01 -1.11408007e+00 4.38861966e-01 4.64545250e-01 -4.78383124e-01 -4.99404341e-01 -3.69261540e-02]
[6.346229076385498, 4.474951267242432]
c3fd19c3-b65f-41e1-9338-d9c7a6d093ef
learning-equivariant-representations
2012.02771
null
https://arxiv.org/abs/2012.02771v1
https://arxiv.org/pdf/2012.02771v1.pdf
Learning Equivariant Representations
State-of-the-art deep learning systems often require large amounts of data and computation. For this reason, leveraging known or unknown structure of the data is paramount. Convolutional neural networks (CNNs) are successful examples of this principle, their defining characteristic being the shift-equivariance. By sliding a filter over the input, when the input shifts, the response shifts by the same amount, exploiting the structure of natural images where semantic content is independent of absolute pixel positions. This property is essential to the success of CNNs in audio, image and video recognition tasks. In this thesis, we extend equivariance to other kinds of transformations, such as rotation and scaling. We propose equivariant models for different transformations defined by groups of symmetries. The main contributions are (i) polar transformer networks, achieving equivariance to the group of similarities on the plane, (ii) equivariant multi-view networks, achieving equivariance to the group of symmetries of the icosahedron, (iii) spherical CNNs, achieving equivariance to the continuous 3D rotation group, (iv) cross-domain image embeddings, achieving equivariance to 3D rotations for 2D inputs, and (v) spin-weighted spherical CNNs, generalizing the spherical CNNs and achieving equivariance to 3D rotations for spherical vector fields. Applications include image classification, 3D shape classification and retrieval, panoramic image classification and segmentation, shape alignment and pose estimation. What these models have in common is that they leverage symmetries in the data to reduce sample and model complexity and improve generalization performance. The advantages are more significant on (but not limited to) challenging tasks where data is limited or input perturbations such as arbitrary rotations are present.
['Carlos Esteves']
2020-12-04
null
null
null
null
['3d-shape-retrieval']
['computer-vision']
[ 2.67624319e-01 -8.61767158e-02 -1.58174291e-01 -3.59302521e-01 -2.36036897e-01 -1.06182754e+00 8.12676013e-01 -3.56454015e-01 -4.62338805e-01 1.95971876e-01 3.85616094e-01 -1.30622491e-01 -2.65352041e-01 -7.79610515e-01 -9.65578496e-01 -7.95726061e-01 2.75782403e-02 5.62972724e-01 3.70450988e-02 -5.12100279e-01 2.24388450e-01 1.22632909e+00 -1.41426766e+00 2.33971447e-01 4.96399729e-03 8.84943187e-01 -3.37231427e-01 6.65067315e-01 2.81789720e-01 1.76258981e-01 2.02148743e-02 -9.19547826e-02 5.76111197e-01 -1.11546427e-01 -9.07215834e-01 1.44968137e-01 7.76784837e-01 4.89993319e-02 -5.01520038e-01 1.03065419e+00 5.00457168e-01 1.74812615e-01 6.88170075e-01 -1.16564703e+00 -7.99987972e-01 1.72900662e-01 -3.84576887e-01 2.86342949e-02 9.64019150e-02 -3.26917320e-01 1.11415720e+00 -1.09464848e+00 9.91564810e-01 1.30514205e+00 7.29070187e-01 5.37541986e-01 -1.29011154e+00 -3.46378416e-01 -1.41891614e-01 1.36338145e-01 -1.33325398e+00 -4.32177544e-01 8.47090244e-01 -3.01630437e-01 1.07556832e+00 4.36001480e-01 5.99979579e-01 9.69473243e-01 3.27797741e-01 4.30391788e-01 9.08843040e-01 -3.79190594e-01 7.56636038e-02 -1.66584432e-01 -3.47610712e-01 5.84555387e-01 8.09301808e-03 -6.58278987e-02 -5.11232018e-01 1.22422025e-01 1.22490323e+00 2.02864498e-01 -3.12512636e-01 -9.29021478e-01 -1.34921718e+00 8.43266368e-01 6.15522504e-01 3.19793999e-01 -1.96083784e-01 -3.28979008e-02 3.33565116e-01 3.60222310e-01 2.60909915e-01 7.46870160e-01 -5.74798524e-01 1.02876030e-01 -6.33700132e-01 1.74313605e-01 7.24553406e-01 8.30117881e-01 7.29676187e-01 4.28984284e-01 3.16379517e-01 7.06692636e-01 -5.03390841e-02 6.93119109e-01 6.22275352e-01 -9.54543650e-01 1.46765381e-01 5.02108634e-01 -3.79086554e-01 -1.13175213e+00 -7.31600046e-01 -5.60440958e-01 -1.04273498e+00 2.77157843e-01 3.26280594e-01 3.59638751e-01 -9.59127367e-01 1.93204939e+00 1.64547145e-01 -1.38577521e-01 8.88599604e-02 7.52955794e-01 6.73630476e-01 3.85381341e-01 -6.26980543e-01 2.01380655e-01 1.47975206e+00 -4.01510954e-01 -1.54313996e-01 -7.13582113e-02 3.81475002e-01 -9.14373875e-01 8.82176042e-01 2.43905500e-01 -1.13008595e+00 -3.67512047e-01 -1.14366055e+00 -2.55332112e-01 -5.89702308e-01 -1.06742814e-01 4.61386174e-01 4.57112521e-01 -1.15989411e+00 5.97210050e-01 -8.55229795e-01 -4.79885548e-01 2.75632292e-01 7.45575011e-01 -8.94997597e-01 -1.05985701e-02 -9.48996127e-01 7.75887012e-01 8.42642710e-02 -1.39905736e-01 -5.27129233e-01 -9.58152294e-01 -1.06471217e+00 8.34826753e-02 7.44163617e-02 -6.77359939e-01 9.83305931e-01 -1.10523796e+00 -1.37516856e+00 1.07424700e+00 -5.42307943e-02 -3.73870730e-01 3.94170821e-01 8.96701962e-02 -1.80356547e-01 5.73392287e-02 -6.25171512e-02 7.05011189e-01 9.74449694e-01 -9.04294431e-01 -1.17125727e-01 -7.77223408e-01 2.63867676e-01 5.28481483e-01 -2.92346418e-01 -9.43649039e-02 -2.46625736e-01 -6.99837029e-01 6.78346694e-01 -1.27338660e+00 -3.39486487e-02 3.94240580e-02 -1.31393999e-01 -8.67850780e-02 1.30017757e+00 -4.49483395e-01 4.04718846e-01 -2.17281842e+00 5.05290985e-01 3.62309963e-01 2.09737942e-01 -7.43610272e-03 -2.85092115e-01 2.46974632e-01 -6.02250993e-01 -7.81841353e-02 2.63491776e-02 1.32940426e-01 -1.04607612e-01 3.30873251e-01 -4.42020327e-01 9.39028621e-01 3.07882607e-01 9.92408574e-01 -5.12156725e-01 3.92514579e-02 2.84658074e-01 7.65557766e-01 -7.67148137e-01 -2.43279353e-01 1.61087528e-01 3.18299860e-01 -1.21124215e-01 3.98301423e-01 6.28609478e-01 -1.46961227e-01 -2.09267855e-01 -6.30059481e-01 -1.52729139e-01 2.32119724e-01 -1.47487223e+00 1.42921448e+00 -5.03017724e-01 7.41260231e-01 1.73469990e-01 -1.37264442e+00 9.40888226e-01 3.41118872e-01 7.78688431e-01 -6.17274404e-01 1.72739238e-01 1.63064569e-01 1.31494209e-01 -1.48585975e-01 4.85287338e-01 -1.60758168e-01 8.23735893e-02 3.94593596e-01 5.13576329e-01 -5.78507543e-01 -5.92822768e-02 3.80388051e-02 5.91050565e-01 7.53161162e-02 3.58758569e-01 -5.72630644e-01 5.28206289e-01 -4.40006077e-01 4.44793105e-01 3.31884027e-01 2.18386054e-01 9.44522500e-01 5.14731705e-01 -9.68572974e-01 -1.52654755e+00 -1.15494645e+00 -2.84520626e-01 8.72180641e-01 -2.28804454e-01 -1.06592327e-01 -7.56693959e-01 -1.43915311e-01 -1.61820069e-01 5.08766137e-02 -7.02656329e-01 -3.03183883e-01 -8.33444655e-01 -5.01214147e-01 6.53288603e-01 7.08927810e-01 5.87598085e-01 -8.00068378e-01 -6.33885801e-01 -1.65292695e-01 5.65251196e-03 -1.40775704e+00 -5.37652612e-01 2.86813051e-01 -8.97846162e-01 -1.13845301e+00 -6.93939388e-01 -6.94931269e-01 6.98789001e-01 3.86829704e-01 7.26664186e-01 -6.49977326e-01 -3.75723004e-01 8.57274055e-01 5.98124973e-02 -4.40575719e-01 2.41811611e-02 2.24134587e-02 6.39702916e-01 1.21824823e-01 1.36155769e-01 -8.60039651e-01 -3.00482512e-01 5.20552993e-01 -1.14819074e+00 -1.27174705e-01 3.70494157e-01 8.68735313e-01 7.17731655e-01 -3.29259306e-01 -3.27939130e-02 -5.22357821e-01 2.31574491e-01 3.63134444e-02 -4.97028559e-01 -1.31124277e-02 -2.76613116e-01 2.40557268e-01 8.39204967e-01 -5.54650664e-01 -6.13791168e-01 1.21564932e-01 -9.81180090e-03 -6.86521113e-01 -2.04121947e-01 2.36943707e-01 -1.62515745e-01 -4.94354993e-01 8.70296478e-01 3.43962967e-01 1.97709948e-01 -2.31127664e-01 3.77977073e-01 1.44077376e-01 6.34192944e-01 -5.51264882e-01 8.27750742e-01 1.09965646e+00 6.41460836e-01 -1.34379494e+00 -5.41191757e-01 -2.75377393e-01 -9.78118539e-01 -2.37081107e-02 1.04775846e+00 -7.51744270e-01 -7.27968752e-01 6.08852684e-01 -1.11418521e+00 8.91887099e-02 -5.31488419e-01 6.92443490e-01 -8.98557305e-01 2.67381072e-01 -3.24331492e-01 -2.00095281e-01 -1.68462798e-01 -1.31296802e+00 1.00292218e+00 1.27156153e-01 -2.62484103e-01 -1.13135445e+00 -5.26830554e-02 1.52141497e-01 5.18200338e-01 2.86524981e-01 1.00575125e+00 -8.35494161e-01 -4.75734800e-01 -2.23564044e-01 -8.73228651e-04 4.62013781e-01 2.74453666e-02 -8.35125223e-02 -9.37188625e-01 -4.51181948e-01 5.69062829e-02 -2.02825144e-01 7.06336319e-01 6.00420415e-01 1.17242277e+00 -2.24258333e-01 1.20139934e-01 1.12900782e+00 1.05102527e+00 6.15199916e-02 5.14431119e-01 4.66047347e-01 8.91184926e-01 6.18463278e-01 -1.93002194e-01 1.03254974e-01 1.11188635e-01 8.05804551e-01 6.40058398e-01 -3.13955277e-01 -1.16649814e-01 -2.41608158e-01 3.09247911e-01 8.77767682e-01 -5.60389519e-01 2.87225872e-01 -7.01801956e-01 4.75690991e-01 -1.37527037e+00 -9.44533050e-01 7.12984130e-02 2.26622081e+00 4.33101237e-01 -2.36518048e-02 -2.21692786e-01 2.47683838e-01 4.96002316e-01 3.91849518e-01 -6.05461240e-01 -5.71490586e-01 -4.90316421e-01 6.54823363e-01 7.51825511e-01 4.91605431e-01 -1.18660867e+00 9.46153879e-01 5.84805202e+00 6.70094907e-01 -1.72711265e+00 -1.83359817e-01 1.84423178e-01 1.06740631e-01 -2.98091948e-01 -5.74437603e-02 -5.84177077e-01 -3.13442349e-01 3.60802054e-01 1.22265518e-01 6.08610392e-01 8.63898456e-01 -1.38966605e-01 5.70446789e-01 -1.25145268e+00 1.02009046e+00 4.57708985e-01 -1.49004781e+00 2.92427719e-01 3.04148585e-01 8.48592162e-01 3.46394330e-01 4.31786984e-01 -2.30162472e-01 7.87033662e-02 -1.18768752e+00 7.17121243e-01 3.22645575e-01 9.86217141e-01 -6.95153415e-01 4.66936767e-01 4.67526577e-02 -1.12313092e+00 1.61349520e-01 -4.71108019e-01 5.75050600e-02 -9.04019475e-02 2.22323537e-01 -6.30177498e-01 5.47942877e-01 8.59079659e-01 8.82388473e-01 -2.83683747e-01 4.53482062e-01 -5.69076650e-02 1.14526182e-01 -5.78777373e-01 2.18599007e-01 4.54341948e-01 -5.10682821e-01 8.19380879e-01 9.46792364e-01 4.44979668e-01 1.42990872e-01 -2.84932315e-01 8.39016318e-01 -2.33267382e-01 -6.35811388e-02 -9.57993627e-01 2.51197457e-01 1.36307582e-01 1.25894868e+00 -8.07675600e-01 6.88721687e-02 -3.53958905e-01 8.00423741e-01 -1.84125900e-01 5.42226374e-01 -3.86256844e-01 -5.39764166e-01 9.80729759e-01 1.00294888e-01 5.16734838e-01 -5.09301305e-01 -2.36113414e-01 -1.18188298e+00 -1.81699023e-02 -9.81196582e-01 1.79504767e-01 -8.54920924e-01 -7.71338463e-01 3.55079293e-01 1.35429695e-01 -1.15370536e+00 -1.56731457e-01 -1.34653652e+00 -5.85695267e-01 6.37641728e-01 -1.19409680e+00 -1.44320035e+00 -1.88837886e-01 9.74818945e-01 2.75597066e-01 -4.38932508e-01 8.74889493e-01 7.19643533e-02 1.62310645e-01 5.51471591e-01 9.08067375e-02 3.28817129e-01 7.47592330e-01 -1.11452246e+00 3.67412359e-01 5.74908435e-01 5.84685683e-01 7.39057600e-01 5.28719664e-01 1.38759330e-01 -1.68024766e+00 -8.45111966e-01 6.82447076e-01 -4.88597751e-01 5.47811210e-01 -5.20374954e-01 -7.64694273e-01 8.79988194e-01 -1.14970803e-01 3.62274617e-01 5.24863839e-01 -3.15950103e-02 -9.21178222e-01 -2.25918174e-01 -1.02437401e+00 7.20688879e-01 1.04894876e+00 -8.13977957e-01 -4.47298139e-01 2.56117076e-01 5.43968081e-01 -8.44583809e-01 -8.98286104e-01 5.04376054e-01 9.63351667e-01 -9.21081543e-01 1.32059216e+00 -1.04363310e+00 4.73647863e-01 -2.38152996e-01 -4.78532314e-01 -1.39307630e+00 -4.47581947e-01 -3.66168231e-01 4.14136112e-01 5.74883401e-01 2.77001947e-01 -8.41448069e-01 6.89071119e-01 2.58572340e-01 -2.60596395e-01 -6.85463548e-01 -1.24540496e+00 -7.47364759e-01 3.03440660e-01 -3.42910886e-01 5.74058235e-01 1.32747042e+00 -3.95630062e-01 2.44472340e-01 -1.67926058e-01 1.96011662e-01 3.03083450e-01 -8.32081810e-02 7.88010955e-01 -1.24493051e+00 -1.35391220e-01 -5.72909415e-01 -1.02538180e+00 -9.35518444e-01 1.63243651e-01 -1.12169254e+00 -6.13173783e-01 -1.03486598e+00 4.65877325e-04 -1.02843717e-03 -4.48817164e-02 5.31870365e-01 8.58553350e-01 4.90997583e-01 1.97141215e-01 6.27267882e-02 -4.72561456e-02 4.33175743e-01 1.29757392e+00 -6.69548661e-02 -8.18229690e-02 3.84324491e-02 -5.49866915e-01 1.08973789e+00 7.24275172e-01 -1.90488741e-01 -2.52189517e-01 -5.57586074e-01 6.47819877e-01 -3.54734510e-01 5.98316669e-01 -8.61555159e-01 2.94933707e-01 -1.46554098e-01 4.59388524e-01 -2.48270616e-01 5.80786645e-01 -8.54765713e-01 2.43031676e-03 4.32197452e-01 -2.12259099e-01 3.87485504e-01 1.93743736e-01 2.22887903e-01 -2.96042770e-01 2.93891091e-04 1.08054066e+00 -2.54626900e-01 -4.55168635e-01 5.41679382e-01 -1.50307119e-01 1.97982550e-01 7.53417432e-01 -4.49115187e-01 -2.21897423e-01 -6.05139971e-01 -6.90575421e-01 -3.49738896e-01 5.42856395e-01 6.77231252e-01 5.72935700e-01 -1.53451169e+00 -5.06532252e-01 6.99070871e-01 6.82563782e-02 2.33862519e-01 4.74739075e-02 8.49844635e-01 -5.99313736e-01 6.58100843e-01 -6.21802330e-01 -8.88530076e-01 -1.50781524e+00 1.75728932e-01 6.80266619e-01 4.84495088e-02 -3.15646738e-01 7.80413151e-01 4.31036025e-01 -1.05423963e+00 -9.68513191e-02 -4.39372510e-01 -9.25377905e-02 -2.18869671e-02 1.87239349e-01 4.16618407e-01 4.06287253e-01 -1.11867428e+00 -3.65110874e-01 1.17448890e+00 9.73247960e-02 -1.67396292e-01 1.65982771e+00 4.38802570e-01 -4.55421239e-01 1.48715213e-01 1.62269068e+00 2.47118231e-02 -9.63671982e-01 -4.44883257e-01 -4.57432866e-01 -1.08150773e-01 -7.75103271e-02 -1.25224665e-01 -1.08397949e+00 1.19889772e+00 4.54134256e-01 9.91396233e-03 9.13361728e-01 8.99415165e-02 2.60899574e-01 7.04298735e-01 1.68138035e-02 -1.01645255e+00 1.78795874e-01 1.00850713e+00 1.17602038e+00 -7.86671042e-01 8.51082732e-04 -1.08628087e-02 -4.17932361e-01 1.61054897e+00 4.13050354e-01 -3.83237749e-01 9.04142141e-01 1.42113209e-01 -1.01601101e-01 -2.68219203e-01 -3.05433065e-01 2.13474780e-01 5.85194707e-01 6.45659208e-01 3.33502412e-01 8.78719538e-02 7.01493993e-02 1.29506320e-01 -6.37378037e-01 -6.17906690e-01 4.79915619e-01 5.65158546e-01 -2.26960525e-01 -9.24639881e-01 -6.02528334e-01 1.77933499e-01 -5.68895996e-01 1.24126576e-01 -4.24875200e-01 9.70165491e-01 9.15134400e-02 2.36163348e-01 3.71815413e-01 -2.07621217e-01 5.41820943e-01 2.37607546e-02 7.01451480e-01 -3.74293774e-01 -1.90677028e-02 -1.11338515e-02 -3.87729198e-01 -6.31242335e-01 -5.74648559e-01 -7.82976985e-01 -1.06105900e+00 -1.91353559e-01 2.86673345e-02 -2.05419227e-01 9.75702465e-01 9.43121374e-01 4.09563482e-01 3.10218543e-01 6.66284084e-01 -1.14655316e+00 -8.79013538e-01 -7.38431871e-01 -4.88207430e-01 4.73225355e-01 4.08957511e-01 -6.18953645e-01 -4.11158919e-01 1.30264297e-01]
[8.890793800354004, 2.3777015209198]
42cde11b-d6aa-4489-b8a3-e61d04b7a136
learning-accurate-dense-correspondences-and
2101.01710
null
https://arxiv.org/abs/2101.01710v2
https://arxiv.org/pdf/2101.01710v2.pdf
Learning Accurate Dense Correspondences and When to Trust Them
Establishing dense correspondences between a pair of images is an important and general problem. However, dense flow estimation is often inaccurate in the case of large displacements or homogeneous regions. For most applications and down-stream tasks, such as pose estimation, image manipulation, or 3D reconstruction, it is crucial to know when and where to trust the estimated matches. In this work, we aim to estimate a dense flow field relating two images, coupled with a robust pixel-wise confidence map indicating the reliability and accuracy of the prediction. We develop a flexible probabilistic approach that jointly learns the flow prediction and its uncertainty. In particular, we parametrize the predictive distribution as a constrained mixture model, ensuring better modelling of both accurate flow predictions and outliers. Moreover, we develop an architecture and training strategy tailored for robust and generalizable uncertainty prediction in the context of self-supervised training. Our approach obtains state-of-the-art results on multiple challenging geometric matching and optical flow datasets. We further validate the usefulness of our probabilistic confidence estimation for the task of pose estimation. Code and models are available at https://github.com/PruneTruong/PDCNet.
['Radu Timofte', 'Luc van Gool', 'Martin Danelljan', 'Prune Truong']
2021-01-05
null
http://openaccess.thecvf.com//content/CVPR2021/html/Truong_Learning_Accurate_Dense_Correspondences_and_When_To_Trust_Them_CVPR_2021_paper.html
http://openaccess.thecvf.com//content/CVPR2021/papers/Truong_Learning_Accurate_Dense_Correspondences_and_When_To_Trust_Them_CVPR_2021_paper.pdf
cvpr-2021-1
['geometric-matching', 'dense-pixel-correspondence-estimation']
['computer-vision', 'computer-vision']
[-8.85868669e-02 -1.34642437e-01 -3.31117995e-02 -3.64472657e-01 -7.22382367e-01 -5.30560374e-01 4.71993208e-01 2.84652054e-01 -2.50842333e-01 6.20572507e-01 9.48349759e-02 1.96854576e-01 -2.26049185e-01 -4.59601134e-01 -7.77189732e-01 -4.14652407e-01 -2.63790321e-02 6.40457213e-01 3.46118897e-01 3.61531466e-01 3.85011405e-01 6.09157085e-01 -1.49116802e+00 -2.45316476e-01 1.08392048e+00 1.08806336e+00 2.29935333e-01 7.66601443e-01 2.65962571e-01 7.50097096e-01 -1.61486089e-01 -3.73294175e-01 2.96140373e-01 -1.01414579e-03 -7.63999164e-01 1.41713679e-01 8.59153807e-01 -4.77027893e-01 -2.28882462e-01 9.35280085e-01 3.17995101e-01 3.37758988e-01 7.11351693e-01 -1.24073744e+00 -3.95523012e-02 1.85562279e-02 -4.82047021e-01 1.86641455e-01 3.92417043e-01 3.41045767e-01 8.60851765e-01 -6.98446870e-01 7.33892798e-01 9.25482929e-01 6.07332170e-01 3.28238428e-01 -1.40404856e+00 -3.30900818e-01 1.69609696e-01 2.68922329e-01 -1.36578453e+00 -7.39880979e-01 5.75180113e-01 -8.94138515e-01 3.92859131e-01 -7.32242987e-02 4.86234337e-01 8.81042898e-01 9.16912481e-02 7.44742572e-01 6.96272314e-01 -1.38728037e-01 2.62190610e-01 2.29006022e-01 -3.28896314e-01 6.62235558e-01 1.25238329e-01 1.61788806e-01 -6.87056303e-01 -1.02837116e-01 9.12260115e-01 -1.30269438e-01 -5.71541727e-01 -9.39297318e-01 -1.28247654e+00 5.48715651e-01 5.76292992e-01 -7.50516281e-02 -3.67532939e-01 2.83602089e-01 1.06013738e-01 -1.87292755e-01 6.45468235e-01 3.02880764e-01 -2.46892154e-01 -4.20148492e-01 -1.09783649e+00 2.03187570e-01 7.37992227e-01 9.87182915e-01 9.02119756e-01 -2.87463784e-01 -1.88354537e-01 4.62276101e-01 4.51942742e-01 4.26477462e-01 -5.14573790e-02 -1.37234557e+00 3.69178504e-01 8.70021284e-02 3.99357885e-01 -1.24303341e+00 -1.37463912e-01 -4.82315838e-01 -8.30498934e-01 3.67448717e-01 7.43050158e-01 7.24137500e-02 -4.82970685e-01 1.74746251e+00 4.92332518e-01 9.71810400e-01 -2.89724350e-01 1.09774017e+00 2.65633553e-01 4.59655643e-01 -2.72547994e-02 1.72693972e-02 7.81444252e-01 -8.15357924e-01 -2.26197764e-01 -3.17847669e-01 3.25878382e-01 -8.04550350e-01 5.54060638e-01 2.40413144e-01 -9.25432086e-01 -6.43586397e-01 -6.24311805e-01 -1.11147985e-01 1.68258891e-01 2.37254322e-01 4.17308569e-01 3.27159137e-01 -8.19796324e-01 1.03732443e+00 -1.11193001e+00 -2.22209334e-01 4.94274288e-01 2.24057391e-01 -5.39220333e-01 -1.97137609e-01 -6.67101204e-01 8.07195127e-01 1.56330660e-01 3.87948066e-01 -6.01205349e-01 -9.50564265e-01 -1.01564157e+00 -1.02587491e-01 2.21814722e-01 -9.75996912e-01 8.50330651e-01 -4.13587302e-01 -1.42678785e+00 7.57876277e-01 -2.95363337e-01 -4.14938480e-01 9.56570506e-01 -4.11267221e-01 2.84919620e-01 2.76295215e-01 1.97653517e-01 8.22520673e-01 1.09953976e+00 -1.09782398e+00 -4.44051385e-01 -2.17951179e-01 -1.59826636e-01 6.25795424e-02 1.73201054e-01 -3.31146479e-01 -5.47915459e-01 -3.50988925e-01 1.03906572e-01 -9.54518735e-01 -2.87374735e-01 6.01913214e-01 -2.92127967e-01 1.30049571e-01 4.94737208e-01 -5.79606652e-01 8.40585470e-01 -2.06960964e+00 3.34690809e-01 4.10513669e-01 3.15169394e-01 -1.89522505e-02 1.42978117e-01 -7.19338357e-02 3.07775795e-01 -2.00706616e-01 -4.26783115e-01 -8.10663402e-01 6.86625317e-02 8.56231004e-02 -2.48843506e-01 8.44020545e-01 4.75818753e-01 7.22349107e-01 -1.01709533e+00 -5.69773555e-01 6.95493162e-01 6.28562272e-01 -6.76098049e-01 5.84997594e-01 -2.27829397e-01 1.10913026e+00 -4.29991245e-01 3.80172759e-01 6.63789809e-01 -3.89913797e-01 -2.91116625e-01 -4.40060109e-01 -5.49482740e-02 -1.65696181e-02 -1.59151483e+00 2.11814761e+00 -5.01992822e-01 6.28986835e-01 1.88899729e-02 -7.13363528e-01 9.50875401e-01 6.57147169e-02 6.81033909e-01 -9.40125510e-02 9.37922224e-02 1.83267042e-01 -3.86520445e-01 -4.54686880e-01 4.63976145e-01 2.44922824e-02 2.82607585e-01 3.64724636e-01 1.32140517e-01 -5.30772090e-01 1.28698692e-01 2.00728148e-01 7.65616953e-01 7.65465319e-01 1.07104674e-01 -1.34970874e-01 5.85530698e-01 -2.73078173e-01 6.22171342e-01 5.25483251e-01 -4.91788656e-01 1.08570850e+00 4.83577788e-01 -2.70768940e-01 -9.38447714e-01 -8.97668183e-01 -3.87096912e-01 3.47765893e-01 4.32256281e-01 -2.58965194e-01 -4.03974652e-01 -5.77952743e-01 2.66580373e-01 3.62833500e-01 -5.61718702e-01 7.72189274e-02 -3.60508084e-01 -3.35088491e-01 5.82685061e-02 3.51522684e-01 2.68642485e-01 -6.36007607e-01 -5.89142263e-01 1.06832042e-01 -3.57221365e-01 -1.46338093e+00 -6.55371606e-01 -3.35295022e-01 -8.13478768e-01 -1.15837145e+00 -5.05196631e-01 -2.93233782e-01 6.24698043e-01 -5.11481762e-02 1.20110011e+00 -1.24651548e-02 -2.56551147e-01 5.02217174e-01 -9.80239511e-02 1.08846731e-01 -2.95171261e-01 -6.78038195e-05 4.95309420e-02 2.84631640e-01 -1.88611627e-01 -6.99543595e-01 -8.01001668e-01 4.45687503e-01 -5.20059645e-01 -7.26031512e-02 2.90439129e-01 5.36159039e-01 6.60446286e-01 -2.65450895e-01 1.24719121e-01 -6.08718753e-01 1.26125216e-01 -3.50924939e-01 -8.73642027e-01 1.33426949e-01 -3.49587083e-01 2.95852929e-01 3.15391690e-01 -1.25268251e-01 -9.77860332e-01 4.22976613e-01 -7.98279792e-02 -8.73422503e-01 -4.48985845e-01 1.38372347e-01 6.86533600e-02 -2.99365014e-01 6.62762284e-01 -2.28083432e-01 1.99623346e-01 -2.75089771e-01 4.10889983e-01 3.37216109e-01 7.01791167e-01 -8.83510292e-01 8.12836587e-01 6.07025325e-01 4.09848094e-01 -6.46275938e-01 -7.99916744e-01 -7.52392113e-01 -1.03299916e+00 -5.50690532e-01 6.43873811e-01 -9.34066296e-01 -7.91067839e-01 3.96430254e-01 -1.13269687e+00 -3.77806723e-01 -2.49654979e-01 7.37718940e-01 -9.37170625e-01 5.94626129e-01 -3.67557108e-01 -7.79149950e-01 -1.47365630e-01 -1.28022921e+00 1.34540427e+00 1.70841202e-01 -2.90348262e-01 -1.24973309e+00 2.90993124e-01 3.96546215e-01 2.75832355e-01 4.35137331e-01 1.17401384e-01 -7.28582293e-02 -1.01980901e+00 -1.46933764e-01 -2.71677911e-01 3.18196595e-01 -4.40710336e-02 3.76827538e-01 -1.10518754e+00 -1.79803610e-01 -2.68552870e-01 -3.18777323e-01 8.59431028e-01 5.89035273e-01 9.68997180e-01 9.94675010e-02 -1.99434206e-01 7.98280776e-01 1.23411644e+00 -6.39004946e-01 4.66427922e-01 7.54313841e-02 7.35694349e-01 7.63780713e-01 8.28036189e-01 5.77065051e-01 3.64751816e-01 8.11350942e-01 5.92200994e-01 4.24580991e-01 -3.04157864e-02 -2.70115733e-01 3.87915596e-02 3.24321747e-01 -9.55090895e-02 -8.93944055e-02 -8.28687012e-01 4.86845344e-01 -2.06879306e+00 -8.34719241e-01 -3.56566720e-02 2.41510963e+00 7.11973131e-01 1.46037892e-01 -9.74082127e-02 -1.84556782e-01 6.41602635e-01 7.25026205e-02 -4.34920669e-01 2.08090261e-01 2.04711601e-01 5.02792858e-02 3.84427339e-01 9.60877657e-01 -1.18912101e+00 8.55112135e-01 5.08607721e+00 5.60640097e-01 -9.99876857e-01 -1.32917538e-01 6.48570061e-01 -5.55148646e-02 -1.76451847e-01 1.48026228e-01 -6.92663431e-01 5.52658081e-01 5.64718783e-01 1.88989565e-02 1.97764218e-01 5.10389030e-01 3.59200001e-01 -4.74938244e-01 -1.16491258e+00 9.79409158e-01 -2.79617067e-02 -1.37267280e+00 -3.23026508e-01 -7.54871890e-02 6.61025286e-01 4.76310141e-02 -8.77887011e-02 -3.02558392e-01 -1.03316065e-02 -8.15026224e-01 7.51840353e-01 9.49632049e-01 6.83916926e-01 -4.47397768e-01 5.56358755e-01 3.91852498e-01 -9.99258161e-01 3.93364549e-01 -1.91661283e-01 1.37495652e-01 4.34312850e-01 8.96334827e-01 -6.28670573e-01 6.69262588e-01 6.70512915e-01 1.33346462e+00 -3.54864240e-01 1.40119624e+00 -4.21277285e-01 2.24728405e-01 -6.28436267e-01 4.22215164e-01 -1.59716815e-01 -3.99848223e-01 6.93686962e-01 9.26454306e-01 3.62060249e-01 -1.97878376e-01 1.39528781e-01 1.12428677e+00 1.04484215e-01 -1.05468526e-01 -3.37835401e-01 4.23332244e-01 4.20215756e-01 1.26451254e+00 -5.66806972e-01 -3.38366665e-02 -5.28557487e-02 8.73163462e-01 6.02948189e-01 2.35675290e-01 -6.73417509e-01 1.28807530e-01 9.52448487e-01 2.04923853e-01 2.92500347e-01 -4.99765635e-01 -1.97807103e-01 -1.42341065e+00 2.92267740e-01 -1.79006636e-01 1.24233134e-01 -7.22512841e-01 -1.15983880e+00 4.63296354e-01 -1.36760985e-02 -1.51108205e+00 -4.99890029e-01 -4.90621865e-01 -6.27301335e-01 8.59082878e-01 -1.69224238e+00 -9.15498495e-01 -6.01063490e-01 4.09594357e-01 1.08637035e-01 2.86927372e-01 5.58442593e-01 1.50388032e-01 -6.13962710e-01 1.17295139e-01 -1.42150000e-01 4.36551869e-02 1.05186617e+00 -1.17889321e+00 1.54466942e-01 1.06563079e+00 2.10102573e-01 3.27963203e-01 8.72665167e-01 -4.57103074e-01 -1.05996895e+00 -1.12526584e+00 6.59748793e-01 -7.00421214e-01 6.69055402e-01 -1.33680880e-01 -9.31520402e-01 5.81289589e-01 -2.69509554e-01 6.15953147e-01 1.96201831e-01 -5.58385216e-02 -1.91753089e-01 -1.29669204e-01 -1.13528681e+00 1.59018055e-01 9.14210439e-01 -5.31207144e-01 -7.01751113e-02 3.00626993e-01 3.44385594e-01 -7.95336306e-01 -1.07976937e+00 3.83083403e-01 5.68363488e-01 -1.31488717e+00 1.01748788e+00 -1.43641129e-01 3.38905066e-01 -5.92510462e-01 8.32231790e-02 -1.29852748e+00 -1.36893719e-01 -6.68553650e-01 -3.53101730e-01 1.06080925e+00 1.61472768e-01 -3.95117164e-01 9.88283038e-01 8.23318541e-01 1.10828750e-01 -5.28480053e-01 -9.91248846e-01 -5.43759346e-01 -2.25389704e-01 -7.21660435e-01 2.29784265e-01 6.89446449e-01 -2.05943152e-01 -6.82054460e-02 -4.93126184e-01 6.03228986e-01 1.06223166e+00 2.93133885e-01 9.47399795e-01 -1.14865792e+00 -4.31630403e-01 -3.26504409e-01 -8.82634282e-01 -1.34663332e+00 5.00116944e-01 -5.65261185e-01 3.13237131e-01 -1.24249089e+00 -1.14869326e-01 -6.57438993e-01 1.01288103e-01 7.69886747e-02 -3.25678885e-01 2.00546071e-01 1.68019950e-01 3.45245332e-01 -6.07714891e-01 6.87481821e-01 1.05373335e+00 1.01681516e-01 -1.79794863e-01 3.89885932e-01 -1.06871784e-01 3.63349855e-01 8.61083448e-01 -3.23646933e-01 -3.68978411e-01 -4.28967893e-01 2.70875031e-03 1.00295410e-01 7.59963989e-01 -1.16056097e+00 4.76390511e-01 -1.39464304e-01 2.18622893e-01 -3.77431065e-01 4.12118822e-01 -8.93697679e-01 1.31261483e-01 1.52622744e-01 -2.44141266e-01 -3.04923624e-01 1.59294695e-01 7.10461378e-01 -2.29600102e-01 -1.95349261e-01 8.70624781e-01 1.25259072e-01 -8.10229480e-01 7.30420887e-01 2.69175798e-01 2.50287503e-01 9.60355222e-01 -6.85993135e-02 -9.90739912e-02 -5.48891783e-01 -7.81142294e-01 3.52781981e-01 5.41998386e-01 4.04034019e-01 7.31876791e-01 -9.94690895e-01 -7.98489571e-01 4.07955885e-01 2.89241016e-01 5.39885223e-01 3.35185081e-01 1.07534671e+00 -6.23336792e-01 1.59443676e-01 -1.26033232e-01 -1.15878582e+00 -8.91534209e-01 2.19813973e-01 5.21954656e-01 3.72802876e-02 -4.79866922e-01 7.98270524e-01 -1.05568610e-01 -2.27983087e-01 2.18984634e-01 -4.30797368e-01 2.26037763e-02 -1.87131479e-01 3.43168736e-01 4.32896823e-01 -6.08621724e-03 -7.88327217e-01 -3.89505744e-01 8.22925806e-01 3.06747198e-01 4.47717197e-02 1.09498250e+00 -3.61418307e-01 -7.71576613e-02 2.59805977e-01 1.13457692e+00 -5.84518798e-02 -2.05520892e+00 -2.90725708e-01 -1.64510176e-01 -9.40527320e-01 1.99049741e-01 -2.37108052e-01 -1.14014471e+00 8.39703441e-01 3.58938068e-01 -1.85936034e-01 7.37668514e-01 1.25403136e-01 3.60122085e-01 -5.03641665e-02 2.86069512e-01 -6.62159145e-01 -1.80567607e-01 3.11947316e-01 7.56638467e-01 -1.54125154e+00 5.22271954e-02 -6.29391611e-01 -4.62656558e-01 1.17139244e+00 4.69319284e-01 -1.69490799e-01 9.14656281e-01 1.53758347e-01 9.48329456e-03 9.76816937e-02 -6.45139158e-01 -1.77959710e-01 5.48500001e-01 6.19928777e-01 2.10480273e-01 -1.35653451e-01 4.98871118e-01 -2.27008671e-01 -1.35478526e-01 1.79190978e-01 3.90637875e-01 7.43750989e-01 -2.39294991e-01 -8.53104830e-01 -2.21666485e-01 1.13217078e-01 3.53077874e-02 1.51371494e-01 2.48757258e-01 2.86381662e-01 -1.71087235e-02 7.69095480e-01 2.28986233e-01 -1.41937643e-01 3.37932825e-01 -2.46123463e-01 6.15157485e-01 -4.71996188e-01 2.41072048e-02 -6.87043071e-02 -1.40382662e-01 -9.14312065e-01 -8.10748994e-01 -9.30653095e-01 -9.36877310e-01 -4.53794122e-01 -1.32510886e-01 1.01068951e-01 5.42990863e-01 9.00249481e-01 5.39721608e-01 1.97098944e-02 5.64595520e-01 -1.08715987e+00 -4.76420105e-01 -6.76155210e-01 -4.06367898e-01 4.79655445e-01 5.74700236e-01 -6.46749437e-01 -4.94239330e-01 1.35002747e-01]
[8.539700508117676, -2.0436747074127197]
816220ba-fd7a-4936-b999-465428bc4921
toward-characteristic-preserving-image-based
1807.07688
null
http://arxiv.org/abs/1807.07688v3
http://arxiv.org/pdf/1807.07688v3.pdf
Toward Characteristic-Preserving Image-based Virtual Try-On Network
Image-based virtual try-on systems for fitting new in-shop clothes into a person image have attracted increasing research attention, yet is still challenging. A desirable pipeline should not only transform the target clothes into the most fitting shape seamlessly but also preserve well the clothes identity in the generated image, that is, the key characteristics (e.g. texture, logo, embroidery) that depict the original clothes. However, previous image-conditioned generation works fail to meet these critical requirements towards the plausible virtual try-on performance since they fail to handle large spatial misalignment between the input image and target clothes. Prior work explicitly tackled spatial deformation using shape context matching, but failed to preserve clothing details due to its coarse-to-fine strategy. In this work, we propose a new fully-learnable Characteristic-Preserving Virtual Try-On Network(CP-VTON) for addressing all real-world challenges in this task. First, CP-VTON learns a thin-plate spline transformation for transforming the in-shop clothes into fitting the body shape of the target person via a new Geometric Matching Module (GMM) rather than computing correspondences of interest points as prior works did. Second, to alleviate boundary artifacts of warped clothes and make the results more realistic, we employ a Try-On Module that learns a composition mask to integrate the warped clothes and the rendered image to ensure smoothness. Extensive experiments on a fashion dataset demonstrate our CP-VTON achieves the state-of-the-art virtual try-on performance both qualitatively and quantitatively.
['Yimin Chen', 'Liang Lin', 'Bochao Wang', 'Xiaodan Liang', 'Meng Yang', 'Huabin Zheng']
2018-07-20
toward-characteristic-preserving-image-based-1
http://openaccess.thecvf.com/content_ECCV_2018/html/Bochao_Wang_Toward_Characteristic-Preserving_Image-based_ECCV_2018_paper.html
http://openaccess.thecvf.com/content_ECCV_2018/papers/Bochao_Wang_Toward_Characteristic-Preserving_Image-based_ECCV_2018_paper.pdf
eccv-2018-9
['geometric-matching']
['computer-vision']
[ 1.60481915e-01 6.60370439e-02 3.21143091e-01 -2.78308183e-01 -5.34557700e-01 -4.89549845e-01 2.47871920e-01 -5.49533367e-01 1.82119310e-01 2.54811764e-01 -3.70925628e-02 3.20772320e-01 1.32200196e-01 -7.41228104e-01 -1.00010908e+00 -5.20442963e-01 3.56886208e-01 3.83707941e-01 2.22985998e-01 -4.38892275e-01 -2.40460023e-01 4.54092979e-01 -1.19905758e+00 2.16130942e-01 9.12876844e-01 8.44030023e-01 6.48015589e-02 4.57379043e-01 2.34120727e-01 -2.03902021e-01 -2.24282727e-01 -9.15226161e-01 7.71993279e-01 -4.77550745e-01 -3.31676811e-01 5.61219752e-01 1.22131670e+00 -4.14584994e-01 -3.17552537e-01 1.06231368e+00 3.39431196e-01 1.30907461e-01 6.20520473e-01 -1.09955668e+00 -1.18959165e+00 9.92163792e-02 -8.54289591e-01 -8.01750541e-01 4.81800079e-01 4.51693594e-01 6.45824432e-01 -9.95654225e-01 8.34091723e-01 1.40019679e+00 1.13869584e+00 8.50915670e-01 -1.58113325e+00 -5.10112941e-01 2.41582021e-01 -4.61065531e-01 -1.28479481e+00 -2.46632680e-01 1.13421082e+00 -3.88839453e-01 1.25225514e-01 5.27389169e-01 1.19666517e+00 1.22015905e+00 3.53594035e-01 6.02810502e-01 1.13337374e+00 -2.65676856e-01 -1.34830162e-01 1.16692655e-01 -2.00262010e-01 1.05296719e+00 7.98617005e-02 1.45552109e-03 -1.05216458e-01 -2.25031376e-01 1.42634153e+00 2.14939311e-01 -3.21301430e-01 -6.67428792e-01 -1.24822903e+00 3.99574310e-01 5.36530375e-01 1.01287593e-03 -4.11063969e-01 2.42509097e-01 5.51332422e-02 -3.85179892e-02 5.88298142e-01 2.94720590e-01 -3.43008161e-01 5.31598508e-01 -9.74988818e-01 7.85480678e-01 5.55845320e-01 1.12094748e+00 5.13186991e-01 1.77069470e-01 -3.45469236e-01 8.56120586e-01 2.91499138e-01 7.15681791e-01 -2.23826259e-01 -1.04324472e+00 4.44397062e-01 6.76378727e-01 2.68495977e-01 -1.15741456e+00 -1.56132162e-01 -1.84603646e-01 -1.07563937e+00 2.78061062e-01 7.02346385e-01 -2.06541712e-03 -1.23122537e+00 1.58233809e+00 6.03419900e-01 9.76113528e-02 -3.51183116e-01 1.33155596e+00 1.01583850e+00 4.25831527e-01 -9.51917991e-02 7.73851201e-02 1.56671119e+00 -9.84386325e-01 -6.44565761e-01 -1.90319732e-01 -2.88838357e-01 -9.21955705e-01 1.43247080e+00 1.35920510e-01 -1.54941595e+00 -1.00200164e+00 -7.96556056e-01 -1.45775452e-01 9.97924060e-02 2.34522656e-01 5.43329120e-01 5.56407630e-01 -8.88870835e-01 7.96110928e-01 -6.72406256e-01 -3.41587007e-01 3.70983541e-01 2.76343226e-01 -4.73481447e-01 -6.45847842e-02 -8.21457148e-01 4.65571642e-01 -2.00544000e-01 3.74932200e-01 -6.66391492e-01 -1.05660117e+00 -1.05834281e+00 -2.25963861e-01 5.04959524e-01 -1.31613040e+00 7.02296674e-01 -1.08108699e+00 -1.64537776e+00 1.09054160e+00 3.28294523e-02 1.08780287e-01 9.71031904e-01 -7.31359646e-02 -2.76525468e-01 -2.15161279e-01 -9.63642597e-02 6.33348107e-01 1.23214400e+00 -1.80578804e+00 -3.41000818e-02 -4.27537799e-01 -1.39559194e-01 1.02723040e-01 -1.55240536e-01 -1.42870530e-01 -9.89194334e-01 -1.24428427e+00 2.75065362e-01 -1.20299935e+00 -1.93140864e-01 5.86473525e-01 -6.62632644e-01 1.07862465e-01 8.52696478e-01 -1.19799054e+00 6.89023316e-01 -1.98046684e+00 4.05044079e-01 2.86777824e-01 2.38828719e-01 2.54760265e-01 -3.42076689e-01 1.06222123e-01 5.45548312e-02 -3.15819085e-02 -4.15621549e-01 -7.90120482e-01 1.69473037e-01 1.09999985e-01 -2.38049537e-01 5.25613248e-01 1.31881177e-01 1.17272830e+00 -5.86953640e-01 -4.20882553e-01 3.00859392e-01 7.46430457e-01 -5.44927955e-01 4.29944098e-01 -3.11147630e-01 6.50780857e-01 -1.76272467e-01 8.45070302e-01 9.62138176e-01 8.92449468e-02 1.14264235e-01 -7.74576187e-01 1.37572452e-01 -5.74338853e-01 -1.27353752e+00 1.86613536e+00 -2.45546773e-01 -3.55200586e-03 6.08618319e-01 -3.67893666e-01 1.12366378e+00 2.73720175e-01 6.76452994e-01 -5.15035510e-01 5.60057424e-02 9.86884534e-02 -3.48578990e-01 -3.48773718e-01 4.52421725e-01 -1.68168858e-01 -4.94184718e-02 6.15255050e-02 -1.91415608e-01 -4.33338881e-01 -2.64772624e-01 -1.98052257e-01 2.64996171e-01 7.99322069e-01 -2.33074054e-01 -3.58536482e-01 2.54832417e-01 -6.41981289e-02 6.82670832e-01 2.86236733e-01 7.02843070e-02 1.23346984e+00 3.92948315e-02 -6.43048525e-01 -1.41037607e+00 -1.20350969e+00 1.63219646e-02 5.78067243e-01 3.64641607e-01 -1.28318146e-01 -1.33873308e+00 -6.71804905e-01 2.18735352e-01 3.79355609e-01 -7.55877793e-01 -2.00606082e-02 -8.98422956e-01 -3.39208633e-01 3.70307237e-01 4.99512583e-01 6.44371212e-01 -8.41043830e-01 -2.35878244e-01 1.34999260e-01 -2.22464502e-01 -1.06308389e+00 -1.34003437e+00 -7.04980552e-01 -6.14332855e-01 -9.12536442e-01 -1.21112549e+00 -1.00159681e+00 1.03898633e+00 1.59821838e-01 9.91000175e-01 1.17082529e-01 -4.20414507e-01 3.56284648e-01 6.66411817e-02 -1.97789714e-01 -4.30559099e-01 -3.69715065e-01 1.54952526e-01 4.35355127e-01 -3.29084665e-01 -5.17665505e-01 -8.56667161e-01 6.63018823e-01 -7.15784013e-01 5.31935811e-01 1.74111620e-01 7.55146027e-01 9.09974456e-01 -9.01849195e-02 1.00225136e-01 -5.13728380e-01 6.14154100e-01 1.59120306e-01 -3.87573391e-01 4.42634970e-01 -2.14706019e-01 -4.19411659e-01 6.62304461e-01 -9.11111772e-01 -9.46744680e-01 1.77791253e-01 -2.45592758e-01 -8.99779916e-01 -3.33982110e-02 -1.69399649e-01 -4.00987089e-01 -1.77799299e-01 3.07898462e-01 2.81062305e-01 2.57946432e-01 -5.74740350e-01 4.53553706e-01 1.12140313e-01 8.63282263e-01 -8.50130558e-01 1.17531228e+00 5.55325866e-01 -4.92463373e-02 -6.00835383e-01 -6.09407425e-01 -1.36581575e-02 -8.93315077e-01 -2.94883668e-01 1.05800176e+00 -7.08057702e-01 -9.46074784e-01 6.48133576e-01 -1.17413080e+00 -5.47768950e-01 -4.64730620e-01 9.94923641e-04 -5.68471313e-01 4.67160910e-01 -7.15424061e-01 -6.57889903e-01 -6.34044468e-01 -1.11619401e+00 1.34637606e+00 1.71952352e-01 -2.59987175e-01 -9.36577320e-01 -1.59304053e-01 6.78212523e-01 3.67296755e-01 8.11122477e-01 6.85397148e-01 3.01932544e-01 -4.48633611e-01 -8.25864524e-02 -1.19078644e-01 4.48240072e-01 3.33412141e-01 4.50876951e-02 -7.41359949e-01 -5.19357324e-01 -1.74213096e-01 -1.13770403e-02 5.88925660e-01 5.14141798e-01 1.08502495e+00 -5.09887516e-01 -1.14857651e-01 9.47547197e-01 1.27908325e+00 -1.94141477e-01 6.95413709e-01 -1.97479859e-01 1.22665262e+00 9.10556078e-01 5.58973312e-01 1.70911670e-01 4.97930795e-01 1.14357495e+00 3.27899426e-01 -8.67763281e-01 -7.46814847e-01 -5.81037760e-01 4.23938960e-01 6.12486362e-01 -6.49142981e-01 1.11320421e-01 -4.28526610e-01 4.57897276e-01 -1.81213796e+00 -6.98755205e-01 -3.48823994e-01 2.36996722e+00 8.43580782e-01 -1.79848149e-01 4.11910594e-01 -4.22849953e-01 6.18516326e-01 -1.48940951e-01 -6.28184199e-01 -1.62242353e-01 -3.46092992e-02 4.36298326e-02 2.28120059e-01 3.95293802e-01 -9.66251075e-01 9.73932207e-01 5.39704180e+00 6.47511184e-01 -1.07837617e+00 3.55492197e-02 5.90344667e-01 1.54647708e-01 -3.46381754e-01 -2.18887523e-01 -6.13617778e-01 3.82215023e-01 -1.94783658e-02 2.44826704e-01 6.36664689e-01 6.78397715e-01 2.99556911e-01 4.20193046e-01 -9.72909629e-01 1.16583645e+00 9.48227197e-02 -1.18196332e+00 9.50823650e-02 1.52669907e-01 9.45429265e-01 -7.41169989e-01 2.86721915e-01 3.36115807e-03 9.14663672e-02 -9.92366076e-01 1.21568513e+00 9.64787424e-01 1.14388990e+00 -4.25699711e-01 4.06269759e-01 2.01523080e-01 -1.45046353e+00 3.27497840e-01 -2.80027479e-01 3.46098959e-01 3.11371475e-01 4.02589202e-01 -4.84557390e-01 6.90224469e-01 6.92358792e-01 4.29122418e-01 -4.66166019e-01 5.84302187e-01 6.27338216e-02 2.45172903e-01 -1.97412789e-01 4.78502125e-01 -1.26289606e-01 -5.55151880e-01 5.80482006e-01 1.06818056e+00 3.18631798e-01 4.60253358e-02 6.06381118e-01 1.49552965e+00 -1.50263309e-01 1.10582195e-01 -3.54310066e-01 3.17005903e-01 -6.47874475e-02 1.32580447e+00 -5.37437499e-01 -2.87255257e-01 -1.20568417e-01 1.42875826e+00 8.89700279e-02 4.24472570e-01 -9.86997783e-01 1.04527362e-01 7.45437801e-01 7.92989135e-01 2.28917271e-01 -3.35453272e-01 -2.76804507e-01 -1.26464427e+00 4.73308444e-01 -9.98836458e-01 -1.08643711e-01 -7.35823095e-01 -1.65556204e+00 5.54750979e-01 -4.64140251e-02 -1.10410511e+00 3.33126426e-01 -3.35927963e-01 -6.27651989e-01 1.11497962e+00 -9.45863962e-01 -2.24754262e+00 -6.27496421e-01 7.17449069e-01 6.96246266e-01 2.21353695e-01 7.63007641e-01 2.11537808e-01 -5.43050706e-01 7.54643321e-01 -3.07913721e-01 1.68312326e-01 8.85061145e-01 -1.08404410e+00 7.34313369e-01 7.03332007e-01 -1.60090178e-01 8.70285928e-01 6.04162157e-01 -1.11171210e+00 -1.87404954e+00 -1.31518734e+00 3.09180349e-01 -6.01158798e-01 1.06686786e-01 -7.11995602e-01 -1.08134830e+00 6.27284467e-01 1.08387858e-01 3.37220907e-01 1.57996580e-01 -2.49516070e-01 -3.63619447e-01 -2.42367059e-01 -1.33219814e+00 8.21167767e-01 1.27265799e+00 -1.70339763e-01 -3.23546171e-01 2.04138666e-01 5.46445906e-01 -8.63524735e-01 -1.15939164e+00 4.61250007e-01 9.24658239e-01 -7.94787645e-01 1.36766541e+00 -4.20983464e-01 3.31527621e-01 -5.56235313e-01 -4.02143188e-02 -1.31107748e+00 -5.62144160e-01 -9.68901753e-01 -6.55041635e-02 1.53002381e+00 -8.41170102e-02 -3.46752167e-01 7.67147183e-01 1.06178594e+00 -1.27720475e-01 -9.31888342e-01 -5.52927434e-01 -6.94515049e-01 2.41838880e-02 -1.56043753e-01 7.36437798e-01 1.05898798e+00 -6.72423720e-01 -5.95321283e-02 -8.28277707e-01 1.00381069e-01 1.03230643e+00 3.08478922e-01 1.17996109e+00 -1.03718948e+00 -4.04882848e-01 -3.30217749e-01 6.59292471e-03 -1.01939189e+00 -5.95085919e-02 -7.42316723e-01 7.50851333e-02 -1.63953030e+00 2.53306419e-01 -7.20351815e-01 3.64623278e-01 5.72672665e-01 -3.90335470e-01 5.82564473e-01 5.05670011e-01 1.49561405e-01 -1.28547922e-01 5.70288479e-01 2.04933834e+00 -2.49284014e-01 -4.02942210e-01 6.00438118e-02 -6.27844632e-01 8.92520845e-01 4.55624938e-01 -5.99164814e-02 -1.60626695e-01 -4.09511238e-01 -2.52139777e-01 1.26543716e-01 9.18583810e-01 -6.52981341e-01 4.69352165e-03 -3.61908376e-01 6.42528832e-01 -2.42090747e-01 6.54552281e-01 -9.46296513e-01 7.23801136e-01 3.03692251e-01 -7.70115927e-02 -5.46956956e-02 1.63146541e-01 4.45711792e-01 3.38474482e-01 2.42089927e-01 9.50729489e-01 -1.88152075e-01 -2.75646091e-01 6.52091086e-01 3.16358030e-01 -2.29612127e-01 1.02168047e+00 -5.06471574e-01 1.44130841e-01 -1.32984161e-01 -8.42072368e-01 2.91945953e-02 1.08575320e+00 5.52725017e-01 9.01266634e-01 -1.52210450e+00 -9.85128045e-01 4.66153741e-01 -1.09467253e-01 1.90350890e-01 4.59176421e-01 6.66269422e-01 -3.91015142e-01 -2.51147300e-01 -2.44339988e-01 -4.94623989e-01 -1.58908296e+00 5.74650943e-01 4.94307607e-01 -1.70839190e-01 -1.18234062e+00 6.15388930e-01 4.48301524e-01 -7.95876980e-01 1.97036222e-01 -4.20009464e-01 3.29482973e-01 -4.67587024e-01 2.43426204e-01 2.27650911e-01 -1.87066063e-01 -7.83933818e-01 -9.12176296e-02 1.30534041e+00 2.47311905e-01 1.41379714e-01 1.26432967e+00 -4.55307066e-02 -1.32500112e-01 -2.38746312e-03 8.08050394e-01 2.35575408e-01 -1.71613133e+00 -1.98334962e-01 -6.99714005e-01 -7.99987435e-01 -2.73531854e-01 -6.74874127e-01 -1.41194928e+00 5.12187958e-01 5.62727451e-01 -6.54280931e-02 8.99521947e-01 -6.00478239e-02 1.39051831e+00 -3.17789018e-01 4.47168410e-01 -9.31730449e-01 2.07329005e-01 -3.26282158e-02 1.43460107e+00 -9.56153989e-01 1.05246365e-01 -9.80763912e-01 -7.51608193e-01 1.07993615e+00 7.35984147e-01 -3.84779334e-01 4.06724304e-01 2.12206677e-01 3.54413912e-02 -3.38685602e-01 9.68396105e-03 1.77505657e-01 9.37228918e-01 7.53284454e-01 1.65509373e-01 2.07100928e-01 3.66220102e-02 6.81284726e-01 -2.47968003e-01 -1.50469124e-01 -1.57982893e-02 4.56429482e-01 1.87554345e-01 -1.07363975e+00 -8.10117066e-01 1.75166428e-01 -1.76097646e-01 1.88708782e-01 -3.87922734e-01 7.33332992e-01 5.07061064e-01 5.45116127e-01 -7.21108466e-02 -4.22382504e-01 9.69976842e-01 -3.39963704e-01 8.89141619e-01 -3.74635190e-01 -9.63614464e-01 4.52602148e-01 -1.55430183e-01 -5.73073328e-01 -2.12764874e-01 -5.25156200e-01 -1.00400925e+00 -5.28827667e-01 -8.46436098e-02 -3.99759114e-01 5.90720713e-01 5.77395976e-01 3.18009734e-01 5.01412094e-01 4.24331754e-01 -1.27925563e+00 -5.16871452e-01 -6.16771102e-01 -5.62422514e-01 1.04930484e+00 1.21870883e-01 -4.58948761e-01 1.67482570e-01 4.66246754e-01]
[11.92065715789795, -0.8834880590438843]
3c650f63-687f-417a-881a-20392df2a8eb
better-sign-language-translation-with
2304.10844
null
https://arxiv.org/abs/2304.10844v1
https://arxiv.org/pdf/2304.10844v1.pdf
Better Sign Language Translation with Monolingual Data
Sign language translation (SLT) systems, which are often decomposed into video-to-gloss (V2G) recognition and gloss-to-text (G2T) translation through the pivot gloss, heavily relies on the availability of large-scale parallel G2T pairs. However, the manual annotation of pivot gloss, which is a sequence of transcribed written-language words in the order in which they are signed, further exacerbates the scarcity of data for SLT. To address this issue, this paper proposes a simple and efficient rule transformation method to transcribe the large-scale target monolingual data into its pseudo glosses automatically for enhancing the SLT translation. Empirical results show that the proposed approach can significantly improve the performance of SLT, especially achieving state-of-the-art results on two SLT benchmark datasets PHEONIX-WEATHER 2014T and ASLG-PC12. Our code has been released at: https://github.com/pengr/Mono\_SLT.
['Junbo Zhao', 'Yawen Zeng', 'Ru Peng']
2023-04-21
null
null
null
null
['sign-language-translation']
['computer-vision']
[ 4.27223712e-01 -2.47567862e-01 -2.52034903e-01 -3.85633230e-01 -1.27387810e+00 -8.44547153e-01 7.08474100e-01 -6.67019546e-01 -2.78021604e-01 6.53092861e-01 4.23312873e-01 -4.82856572e-01 3.28343302e-01 -2.05014274e-01 -7.26565540e-01 -6.31967068e-01 5.75582266e-01 7.46015787e-01 1.07703745e-01 -2.20416859e-01 -3.67208086e-02 -8.55927691e-02 -1.35124695e+00 4.61333781e-01 1.13722420e+00 6.98538542e-01 1.55557245e-01 6.06038332e-01 -2.84902900e-01 4.71506089e-01 -4.67413604e-01 -7.21797347e-01 5.49952924e-01 -7.20005512e-01 -5.47793865e-01 -1.12553507e-01 8.84127021e-01 -3.86777699e-01 -7.84746781e-02 1.01061130e+00 5.95968425e-01 -3.20964336e-01 5.83108425e-01 -1.19529080e+00 -5.78257918e-01 5.42020798e-01 -4.33011264e-01 -3.62560749e-01 4.90948349e-01 3.02857608e-01 1.15101039e+00 -1.23830926e+00 1.11545587e+00 1.24595749e+00 4.30751503e-01 5.75820208e-01 -6.60140514e-01 -8.92903984e-01 -8.70939866e-02 2.55854815e-01 -1.31786072e+00 -5.66962123e-01 3.84206831e-01 -3.50014955e-01 9.21041906e-01 3.32790136e-01 9.24918294e-01 1.35347629e+00 3.23433280e-02 1.20178640e+00 1.57732940e+00 -5.43716013e-01 -2.28325844e-01 -4.26090449e-01 -9.82639790e-02 7.54044831e-01 1.77129760e-01 1.06643498e-01 -8.97594750e-01 2.93885350e-01 4.16803449e-01 -5.64807057e-01 -2.87044227e-01 7.95380622e-02 -1.46052051e+00 4.16290611e-01 -1.94122463e-01 1.26488507e-01 -9.67383385e-02 1.30326468e-02 5.49049973e-01 4.75026131e-01 3.78926367e-01 -3.11782435e-02 -4.14732218e-01 -5.00611067e-01 -8.66637886e-01 1.76853448e-01 6.39083922e-01 1.45947754e+00 3.72808397e-01 -2.12553311e-02 -2.23583236e-01 8.79218757e-01 4.00293201e-01 1.21477568e+00 5.42165220e-01 -6.59585834e-01 1.06775796e+00 6.10681176e-01 -1.11208588e-01 -3.13816786e-01 4.35902998e-02 -7.93767348e-02 -4.60013866e-01 -2.51474176e-02 6.57843292e-01 -7.25364164e-02 -1.32285035e+00 1.50580907e+00 2.91483432e-01 -1.62326470e-01 7.96824843e-02 1.15418923e+00 8.18915606e-01 6.77198231e-01 -2.06877619e-01 -3.15887302e-01 1.30669904e+00 -1.20550680e+00 -8.01702976e-01 -3.14774901e-01 7.62011468e-01 -1.41262627e+00 1.34944165e+00 3.98438752e-01 -8.39912295e-01 -1.64357424e-01 -6.85077965e-01 -4.34495687e-01 -1.79805905e-01 6.57805383e-01 2.87365675e-01 3.52928817e-01 -8.42017412e-01 5.75222522e-02 -8.61028969e-01 -7.06470788e-01 2.30081171e-01 1.21034332e-01 -4.41775560e-01 -4.07424480e-01 -1.04700089e+00 9.21127796e-01 2.73994595e-01 3.91408175e-01 -5.19503355e-01 -2.51598179e-01 -5.87565422e-01 -6.64297283e-01 5.69949925e-01 -5.13047218e-01 1.50949728e+00 -7.17940629e-01 -1.87445664e+00 1.12621534e+00 -4.40049589e-01 4.86332737e-02 1.11755252e+00 -5.45864940e-01 -2.78282195e-01 -5.92422932e-02 1.87471241e-01 5.65720022e-01 9.12084997e-01 -9.47892308e-01 -9.81404245e-01 -2.83675432e-01 -5.80981791e-01 5.40142655e-01 1.37698531e-01 4.57078934e-01 -8.22267652e-01 -7.83500373e-01 1.10391572e-01 -1.27171040e+00 3.87980819e-01 -1.24341652e-01 -4.52880412e-01 -3.78676891e-01 8.40358794e-01 -1.31029320e+00 1.08264482e+00 -1.85346639e+00 3.46599549e-01 9.94801819e-02 -4.66718376e-02 4.88391966e-01 -2.25944966e-01 7.13015199e-01 3.79181147e-01 -3.81136946e-02 -2.63290465e-01 -5.13218760e-01 1.97858185e-01 5.89513063e-01 -4.19507563e-01 3.04222256e-01 -1.73267931e-01 1.31029785e+00 -9.93752122e-01 -7.79454172e-01 1.52434647e-01 1.94841400e-01 -1.58574775e-01 9.97559354e-02 -3.79637659e-01 7.77379334e-01 -2.59076446e-01 8.78851891e-01 3.82838249e-01 1.25354290e-01 5.51034927e-01 -1.46734059e-01 -3.92846614e-01 5.89315951e-01 -7.03071654e-01 1.68986607e+00 -2.21205294e-01 8.96692157e-01 -1.82522818e-01 -3.66910100e-01 7.44581640e-01 4.58660841e-01 2.79402852e-01 -6.92828715e-01 3.58010918e-01 8.77858222e-01 6.57679699e-03 -6.24031305e-01 5.10070562e-01 4.54675173e-03 -1.87404260e-01 3.79373729e-01 9.35319345e-03 -3.80558312e-01 6.75203383e-01 -5.25830984e-02 7.96193659e-01 6.75874710e-01 4.23533052e-01 4.32980247e-02 1.70137465e-01 1.48366928e-01 6.75811827e-01 3.14003229e-01 -2.14064002e-01 5.26350856e-01 1.75684273e-01 -5.97884767e-02 -1.18416619e+00 -9.93914843e-01 3.67888331e-01 9.89408255e-01 -9.21473727e-02 -5.97820282e-01 -8.02235961e-01 -6.48593843e-01 -2.55031615e-01 7.50601172e-01 -2.79644251e-01 2.22593233e-01 -1.06764722e+00 -3.15045863e-01 9.33364987e-01 4.33779925e-01 5.49823463e-01 -1.05113232e+00 -3.69661331e-01 -1.36664882e-02 -8.40759575e-01 -1.51992607e+00 -1.12173676e+00 -2.95033872e-01 -6.33972824e-01 -1.04107440e+00 -7.60929942e-01 -9.88408685e-01 6.62703037e-01 9.35393050e-02 7.44110167e-01 -1.43554598e-01 4.72106077e-02 -2.51701009e-02 -7.62914419e-01 -2.98114002e-01 -6.86499596e-01 1.04378298e-01 2.32312679e-01 7.91743025e-02 4.65350807e-01 -3.87816161e-01 -2.43360251e-01 4.93108511e-01 -6.96760774e-01 6.25241518e-01 8.36917043e-01 7.61414289e-01 6.94812953e-01 -6.76368773e-01 1.29274465e-02 -4.50398237e-01 5.48146188e-01 2.10711941e-01 -6.61271393e-01 6.74726129e-01 -5.79486430e-01 9.21501070e-02 4.70276266e-01 -6.11738026e-01 -9.25657749e-01 -9.46818292e-02 -1.62719041e-01 -3.74498993e-01 1.49691015e-01 5.64849019e-01 -1.89622417e-01 5.31680696e-02 3.01459074e-01 7.15097785e-01 -1.05876282e-01 -5.83618641e-01 6.49835527e-01 8.83603990e-01 6.49940073e-01 -6.36402130e-01 9.95112658e-01 1.11380905e-01 -2.58651793e-01 -6.83098972e-01 -6.47374928e-01 -3.27852160e-01 -7.14324057e-01 -5.10296166e-01 6.94454253e-01 -6.76366270e-01 -4.14829314e-01 8.90178084e-01 -1.14682782e+00 -6.08512998e-01 -1.68754041e-01 5.03995121e-01 -6.10737324e-01 4.82298851e-01 -5.92571914e-01 -5.19104838e-01 -6.38309002e-01 -1.28261244e+00 1.35694170e+00 -6.02544174e-02 -3.08132619e-01 -4.68973905e-01 2.26323381e-01 8.77898872e-01 -1.89192574e-02 1.49109624e-02 6.70800567e-01 -2.62028396e-01 -6.78394616e-01 -1.57860234e-01 -2.20153704e-01 5.72019756e-01 3.67713034e-01 1.14417560e-01 -4.85651821e-01 -3.80187958e-01 -5.35669208e-01 -4.11290348e-01 5.77121615e-01 6.69896975e-02 2.71589458e-01 -5.73252797e-01 -9.16860253e-02 7.26435602e-01 1.08237505e+00 1.33183643e-01 6.21884406e-01 9.05496161e-03 8.52163434e-01 3.34333748e-01 9.29052651e-01 1.96027048e-02 6.27378762e-01 1.03151095e+00 -1.58596411e-01 3.20173860e-01 -8.80154729e-01 -7.23827124e-01 9.21683550e-01 1.72682703e+00 -5.96048892e-01 -3.40819478e-01 -9.94585633e-01 5.60430348e-01 -1.92069662e+00 -6.85264707e-01 -3.65540117e-01 2.02414775e+00 1.15558052e+00 -1.96657330e-01 -5.80600277e-02 6.96642473e-02 6.30307674e-01 8.46381038e-02 -3.27241063e-01 -2.05253839e-01 -3.95292521e-01 7.06164762e-02 7.10082054e-01 7.09993601e-01 -6.46993101e-01 1.70666838e+00 5.40030670e+00 1.16387689e+00 -1.28105795e+00 3.28558236e-01 -1.74139768e-01 -1.95155367e-01 -8.13737512e-02 -4.13389653e-02 -9.53098536e-01 4.81565356e-01 5.27662039e-01 -6.79326430e-02 6.77909791e-01 4.20054436e-01 4.39003348e-01 9.97399539e-02 -8.60865712e-01 9.64133024e-01 1.76135704e-01 -8.47796917e-01 3.50458443e-01 1.67893216e-01 9.26598310e-01 4.51710403e-01 -8.73784050e-02 3.52698535e-01 4.20899659e-01 -5.04231930e-01 1.20597064e+00 2.52208114e-01 1.49360609e+00 -3.67194973e-02 5.09584248e-01 1.72836155e-01 -1.44595587e+00 2.87724078e-01 1.49947464e-01 1.11556843e-01 5.84419310e-01 1.58121720e-01 -1.13674688e+00 7.27009356e-01 4.86015171e-01 7.34351933e-01 -3.58767718e-01 7.61765659e-01 -1.06058753e+00 1.07057190e+00 -6.21852994e-01 -1.15338162e-01 1.68443456e-01 -4.23338532e-01 9.44998205e-01 1.17467415e+00 6.48072302e-01 -1.30761951e-01 1.08105645e-01 1.36833593e-01 -1.66372612e-01 4.95855242e-01 -5.58721244e-01 -3.12777191e-01 2.46952578e-01 8.71508241e-01 -4.90668863e-01 -4.59415227e-01 -3.81903797e-01 1.24797904e+00 9.54633802e-02 4.76437747e-01 -8.26632261e-01 3.37381326e-02 6.75735235e-01 5.49823977e-02 2.94807196e-01 -5.06843984e-01 -1.09287210e-01 -1.48085463e+00 6.56846762e-01 -1.29227722e+00 1.68913171e-01 -9.27667141e-01 -1.01175117e+00 5.49777210e-01 -5.39264381e-02 -1.64518321e+00 -4.32616800e-01 -6.69219196e-01 -3.30882035e-02 5.26639938e-01 -1.37836087e+00 -1.88745606e+00 -2.78819799e-01 6.10999227e-01 7.63289928e-01 -1.10618897e-01 4.96488750e-01 4.75882143e-01 -3.15717340e-01 9.11480069e-01 2.47413665e-01 3.14143986e-01 1.06510305e+00 -8.87509465e-01 5.76195240e-01 1.22886121e+00 4.10323143e-01 3.79656881e-01 6.50576055e-01 -1.05519414e+00 -1.30913019e+00 -1.17506242e+00 1.43892837e+00 -6.00703001e-01 8.65502119e-01 -5.38135052e-01 -3.57677728e-01 9.53765154e-01 2.16620460e-01 -4.58718061e-01 1.64250448e-01 -4.67473388e-01 -4.96849537e-01 -2.33119354e-02 -5.92260838e-01 1.13578546e+00 1.55195260e+00 -5.29078722e-01 -6.81442916e-01 4.80057210e-01 5.46181858e-01 -7.76261628e-01 -7.24768817e-01 3.62301499e-01 1.00917351e+00 -3.77669126e-01 4.84460264e-01 -3.77444446e-01 4.41064239e-01 -5.34704030e-01 -3.27784389e-01 -1.28222990e+00 3.67618948e-01 -1.07961428e+00 -1.68328844e-02 1.04294407e+00 4.08466309e-01 -7.88503826e-01 4.68798965e-01 2.78230906e-02 -5.39647639e-01 -5.66678643e-01 -1.27159023e+00 -1.03692126e+00 -1.06470622e-01 -5.22343457e-01 4.04640228e-01 7.68177629e-01 -1.47289529e-01 3.64469707e-01 -8.02928090e-01 -7.64197409e-02 7.22884715e-01 2.04684839e-01 1.04058444e+00 -7.19727695e-01 -3.33921053e-02 -5.36529422e-01 -1.80491745e-01 -1.36503708e+00 -1.23259917e-01 -1.14734232e+00 4.47945178e-01 -1.63291633e+00 -2.67754588e-02 -3.44156623e-02 2.88905889e-01 9.18961465e-01 -6.93145767e-02 4.34434652e-01 4.36763793e-01 5.08954763e-01 -4.09135282e-01 6.76317990e-01 1.70958364e+00 -1.33023649e-01 -1.50220707e-01 -2.06512824e-01 -7.67676532e-02 5.04918158e-01 9.72205222e-01 -5.58722258e-01 -1.36556908e-01 -9.30499375e-01 1.88302800e-01 -1.89792231e-01 4.52173725e-02 -6.80726469e-01 1.33017138e-01 -3.33036870e-01 -4.40112084e-01 -7.47697294e-01 3.12725186e-01 -6.24356389e-01 1.73101097e-01 2.96292514e-01 -3.31643000e-02 1.78127542e-01 -4.81172577e-02 1.45466760e-01 -1.75166994e-01 1.45509943e-01 4.66336161e-01 7.02850819e-02 -5.89065194e-01 2.39275753e-01 -4.39174056e-01 1.81373373e-01 6.34087563e-01 -2.69492507e-01 -6.39371276e-01 -3.54675591e-01 -1.74661964e-01 2.52844721e-01 5.44381142e-01 7.14779437e-01 5.39007127e-01 -1.44291019e+00 -9.78960097e-01 1.45610601e-01 4.52088714e-01 -1.72260344e-01 -2.05572993e-01 1.24302053e+00 -7.58898675e-01 5.98265588e-01 -9.89798307e-02 -5.64084888e-01 -1.60510325e+00 -1.19299203e-01 1.07199892e-01 -4.24835503e-01 -7.68257678e-01 6.04224920e-01 -2.04321831e-01 -7.42590606e-01 -4.45785969e-02 -6.97761297e-01 3.81670058e-01 -1.92654476e-01 1.50478169e-01 2.57064372e-01 -2.98317596e-02 -1.03167224e+00 -3.05746943e-01 9.44715023e-01 6.26498088e-02 -5.12406826e-01 9.92999971e-01 -1.61191508e-01 -2.45811030e-01 3.39959532e-01 9.24065769e-01 3.17836642e-01 -9.41217363e-01 -3.19625497e-01 -1.58019718e-02 -3.99602950e-01 -3.55761439e-01 -1.08347905e+00 -9.27992046e-01 7.00747252e-01 2.36869738e-01 -6.66517675e-01 1.03817201e+00 -6.62556216e-02 1.41144788e+00 5.06532311e-01 5.38016200e-01 -1.30606270e+00 -1.97846964e-01 8.12820375e-01 1.22681344e+00 -1.15715516e+00 -2.32945770e-01 -5.40271461e-01 -6.66362822e-01 8.31870198e-01 3.59396756e-01 2.83647060e-01 7.94832483e-02 1.40747681e-01 6.87705815e-01 8.36054757e-02 -5.05816281e-01 -3.36645633e-01 4.12455261e-01 3.56584728e-01 1.75626472e-01 3.51362586e-01 -8.00207496e-01 9.15184990e-02 -7.28711009e-01 2.97568083e-01 9.49569568e-02 9.65113103e-01 -2.00494248e-02 -1.52238250e+00 -3.16681862e-01 2.48693451e-01 -8.18217918e-02 -4.12517250e-01 -7.11441398e-01 9.61465716e-01 7.56162032e-02 8.53248537e-01 -4.71841425e-01 -5.59447527e-01 3.74710441e-01 4.02356982e-01 7.79563427e-01 -3.48153472e-01 -3.77034843e-01 3.35548013e-01 4.77486908e-01 -5.48355937e-01 -4.62330461e-01 -9.35454309e-01 -1.13306677e+00 -3.14194858e-01 -6.76703975e-02 -2.17312858e-01 6.67727292e-01 1.28417945e+00 1.09874591e-01 3.39785963e-02 5.94520494e-02 -5.87955117e-01 -3.96532923e-01 -1.33878851e+00 -2.04276979e-01 5.03901839e-01 4.84246723e-02 -5.14414072e-01 -1.90054268e-01 4.64294434e-01]
[9.213510513305664, -6.535788536071777]
139b26a8-8c9b-4886-b754-f642d6e8876a
exploring-the-effect-of-primitives-for
null
null
http://openaccess.thecvf.com//content/CVPR2023/html/Li_Exploring_the_Effect_of_Primitives_for_Compositional_Generalization_in_Vision-and-Language_CVPR_2023_paper.html
http://openaccess.thecvf.com//content/CVPR2023/papers/Li_Exploring_the_Effect_of_Primitives_for_Compositional_Generalization_in_Vision-and-Language_CVPR_2023_paper.pdf
Exploring the Effect of Primitives for Compositional Generalization in Vision-and-Language
Compositionality is one of the fundamental properties of human cognition (Fodor & Pylyshyn, 1988). Compositional generalization is critical to simulate the compositional capability of humans, and has received much attention in the vision-and-language (V&L) community. It is essential to understand the effect of the primitives, including words, image regions, and video frames, to improve the compositional generalization capability. In this paper, we explore the effect of primitives for compositional generalization in V&L. Specifically, we present a self-supervised learning based framework that equips V&L methods with two characteristics: semantic equivariance and semantic invariance. With the two characteristics, the methods understand primitives by perceiving the effect of primitive changes on sample semantics and ground-truth. Experimental results on two tasks: temporal video grounding and visual question answering, demonstrate the effectiveness of our framework.
['Yuwei Wu', 'Yunde Jia', 'Chenchen Jing', 'Zhen Li', 'Chuanhao Li']
2023-01-01
null
null
null
cvpr-2023-1
['video-grounding']
['computer-vision']
[ 2.65393764e-01 -1.62010029e-01 -1.38698146e-01 -3.93560022e-01 -2.03795191e-02 -5.87545216e-01 8.74618351e-01 7.90469497e-02 -1.98785350e-01 1.08417235e-01 3.79007876e-01 -2.36346886e-01 7.80066997e-02 -6.48299515e-01 -7.49169827e-01 -3.27585369e-01 -7.27173081e-03 -9.85945091e-02 5.41707873e-01 -2.75410622e-01 3.42528790e-01 3.55720729e-01 -1.86329901e+00 9.19585526e-01 9.07553673e-01 9.29357946e-01 4.40407366e-01 5.87122083e-01 -4.00969863e-01 1.01523614e+00 -2.69004047e-01 -2.27804005e-01 2.14368612e-01 -8.09971452e-01 -8.72135341e-01 3.93127441e-01 7.22419441e-01 -2.91431665e-01 -3.49965453e-01 1.31476903e+00 6.55365661e-02 4.05592769e-01 6.09711707e-01 -1.49381566e+00 -1.23015380e+00 6.36534870e-01 -1.98384628e-01 3.61242235e-01 6.99047804e-01 2.70897478e-01 1.23799145e+00 -9.35891092e-01 5.84793985e-01 1.74378276e+00 4.04848367e-01 6.46666050e-01 -1.15631771e+00 -4.36497480e-01 5.71833491e-01 6.72164798e-01 -1.41112399e+00 -2.58884847e-01 8.89210284e-01 -5.88178158e-01 6.01451695e-01 2.29549512e-01 6.98406816e-01 7.09922791e-01 8.05462748e-02 1.03618956e+00 1.33009839e+00 -5.45566499e-01 3.77446443e-01 2.86606178e-02 1.19121097e-01 9.23946083e-01 6.54835254e-03 1.01154059e-01 -7.42730737e-01 1.95845943e-02 9.67875361e-01 5.07102534e-02 -4.47994292e-01 -4.91730511e-01 -1.37976074e+00 5.83740592e-01 5.39856672e-01 2.75687963e-01 -1.03813194e-01 2.79566288e-01 4.81412649e-01 3.68278623e-01 2.02375948e-01 3.63730073e-01 -6.14610165e-02 3.07851970e-01 -5.51160872e-01 3.81473422e-01 4.48434293e-01 1.21031773e+00 7.23316729e-01 9.30004194e-02 -1.63676366e-01 5.40723085e-01 3.14658940e-01 4.96518910e-01 7.57814050e-01 -1.24436951e+00 7.19190240e-02 8.60401869e-01 -2.06934899e-01 -1.06031346e+00 -1.95979491e-01 -8.47576633e-02 -6.06996059e-01 3.61440256e-02 3.53333443e-01 6.42802417e-01 -7.65671790e-01 1.97367203e+00 1.61650136e-01 2.65571475e-01 1.61375180e-01 9.51092541e-01 9.95735466e-01 6.55186892e-01 1.92585185e-01 -2.09774420e-01 1.49730659e+00 -9.95978594e-01 -8.47604156e-01 -1.90695196e-01 5.61613441e-01 -5.72232842e-01 1.75233603e+00 5.46140075e-02 -1.04782403e+00 -9.05639291e-01 -9.53907549e-01 -3.14602911e-01 -4.15017158e-01 -1.96840420e-01 5.59756041e-01 5.25629282e-01 -9.49833095e-01 4.22424018e-01 -5.36160588e-01 -6.09435081e-01 2.08191946e-01 -1.75831437e-01 -2.24396452e-01 -2.48563990e-01 -1.12979293e+00 6.99472964e-01 5.82820177e-01 -1.50112212e-01 -7.74182737e-01 -6.37010276e-01 -1.02538979e+00 -2.22202465e-02 4.10089403e-01 -9.32864547e-01 1.30976534e+00 -1.43547428e+00 -1.25965893e+00 9.65149283e-01 -2.98653722e-01 -3.79871488e-01 2.45000660e-01 -1.49865076e-01 -2.65886366e-01 4.57551152e-01 1.18641071e-01 7.44574249e-01 1.09305489e+00 -1.38799322e+00 -5.22511005e-01 -4.73695636e-01 3.37791681e-01 4.10009801e-01 -7.54155070e-02 -1.77307412e-01 -2.55885065e-01 -9.30129766e-01 4.94990081e-01 -8.38869929e-01 1.18097924e-01 3.22621614e-01 7.95335323e-02 -3.22918504e-01 7.13578999e-01 -7.05570936e-01 9.24735129e-01 -2.41172266e+00 2.01339766e-01 -4.10065092e-02 1.77356705e-01 -3.87017988e-02 -3.32008332e-01 4.79133189e-01 -1.69776052e-01 -5.35829319e-03 -3.03274870e-01 5.54597862e-02 7.65637010e-02 4.04765010e-01 -6.19666815e-01 3.62901360e-01 1.56484008e-01 1.24088323e+00 -1.12906837e+00 -6.16320372e-01 2.26158351e-01 4.25048470e-02 -7.70097852e-01 2.47656822e-01 -5.47470510e-01 5.30721366e-01 -3.00885886e-01 6.79108024e-01 3.14974010e-01 -2.41992444e-01 8.90914053e-02 -5.22419870e-01 -5.14819622e-02 1.53416410e-01 -1.13720894e+00 1.82107389e+00 -7.95461535e-02 6.75786912e-01 -3.61575305e-01 -1.01322281e+00 6.63168430e-01 1.39460132e-01 9.28962678e-02 -9.41242397e-01 -7.13869631e-02 -3.26529890e-02 3.51227559e-02 -8.26359928e-01 5.27349472e-01 -2.62293398e-01 2.11062431e-01 5.88653564e-01 -1.26452222e-01 -6.31554544e-01 2.19683617e-01 4.23604429e-01 5.08550227e-01 3.94903928e-01 5.07856846e-01 -4.64779735e-01 6.96180522e-01 9.19198210e-04 4.56347644e-01 7.12236822e-01 -2.94727772e-01 3.84804994e-01 2.83564210e-01 -5.07959604e-01 -1.05904031e+00 -1.29248405e+00 2.15017006e-01 1.43664682e+00 6.30910158e-01 -5.07469535e-01 -8.42329502e-01 -2.82703221e-01 -1.43730521e-01 9.90322590e-01 -5.90284705e-01 -4.35143590e-01 -6.05446935e-01 -2.35174727e-02 4.18863088e-01 6.53751552e-01 7.30389535e-01 -1.20432127e+00 -8.41782749e-01 -2.08325312e-01 -3.69023234e-01 -1.35316432e+00 -6.80589139e-01 -5.04429400e-01 -9.84710515e-01 -9.57197130e-01 -2.59168565e-01 -9.60204840e-01 6.60545290e-01 7.55946815e-01 1.02701771e+00 3.23289156e-01 -2.47157142e-01 9.04749513e-01 -6.35070562e-01 -2.56293327e-01 -6.00859463e-01 -4.58346456e-01 5.34478202e-03 1.52427167e-01 9.51742195e-03 -4.74062651e-01 -3.87525380e-01 2.98900157e-01 -1.38129544e+00 3.52334827e-01 3.68429601e-01 5.97050190e-01 6.52963638e-01 -6.72351196e-02 2.54300267e-01 -6.76396966e-01 6.77193522e-01 1.82356331e-02 -3.48230839e-01 5.06667554e-01 -2.85407901e-01 2.46769980e-01 4.74815160e-01 -7.46280015e-01 -9.36476827e-01 -1.33137420e-01 3.46593291e-01 -4.54203427e-01 -3.78217585e-02 4.91884619e-01 -3.76341373e-01 1.64540801e-02 6.87269568e-01 4.77738351e-01 1.09425440e-01 -1.45925447e-01 9.73500669e-01 1.51699305e-01 6.96919501e-01 -8.66540670e-01 7.85565078e-01 9.08353329e-01 -3.76843102e-02 -1.01311910e+00 -6.52015984e-01 -4.02304232e-01 -5.82566679e-01 -2.82372624e-01 9.85783994e-01 -8.09154809e-01 -6.75959349e-01 2.64987469e-01 -1.34859645e+00 -1.57852471e-01 -5.63419938e-01 4.08694774e-01 -7.70284295e-01 8.29288423e-01 -3.07954252e-01 -5.68006873e-01 -6.16074493e-03 -1.07616603e+00 8.41137230e-01 1.71773225e-01 -3.07942063e-01 -9.44619656e-01 -2.69261420e-01 2.16991141e-01 1.40205294e-01 5.40922694e-02 1.54087687e+00 -4.54741538e-01 -7.92154729e-01 3.03509831e-01 -3.93290162e-01 3.78053427e-01 1.42413408e-01 -2.27913246e-01 -8.92923355e-01 -1.40030116e-01 2.23638847e-01 -1.24901049e-01 7.41729379e-01 2.24574596e-01 1.33182430e+00 -3.75519663e-01 9.26121771e-02 5.13625562e-01 1.17147696e+00 1.73454002e-01 5.56718946e-01 6.16169572e-02 7.04838991e-01 8.55969965e-01 4.70756501e-01 1.88337848e-01 6.04902208e-01 3.58225405e-01 4.86077070e-01 -6.17356552e-03 -4.10447150e-01 -4.70861763e-01 2.77437478e-01 9.25712228e-01 -2.96743274e-01 1.28635973e-01 -8.80789042e-01 4.02797252e-01 -1.92890787e+00 -1.18863702e+00 4.67020459e-03 1.87103474e+00 7.02414632e-01 -1.97218016e-01 -8.60574096e-02 1.34086743e-01 7.99346864e-01 1.75526470e-01 -4.32941377e-01 -1.81208193e-01 -3.00442755e-01 1.15440600e-02 -4.92754504e-02 5.78177094e-01 -6.90295994e-01 1.23235238e+00 6.55366850e+00 5.08811176e-01 -8.54979277e-01 -8.13408270e-02 1.69688553e-01 3.84727597e-01 -6.02978408e-01 2.64794320e-01 -2.55957186e-01 9.81875360e-02 1.47488073e-01 -2.82933235e-01 8.09540331e-01 6.72487319e-01 1.72804207e-01 -7.67838657e-02 -1.53612185e+00 1.14235294e+00 5.13492942e-01 -1.33467817e+00 6.80473804e-01 -4.76873547e-01 7.39102006e-01 -5.71837068e-01 2.29209200e-01 1.41751871e-01 -3.18505690e-02 -8.38322043e-01 1.11134923e+00 4.80589539e-01 4.74543482e-01 -2.67307609e-01 2.44952247e-01 3.51492405e-01 -1.40397060e+00 -1.37076437e-01 -2.07265332e-01 -2.83574313e-01 1.44976303e-01 9.98879001e-02 -5.29326975e-01 4.54310924e-01 7.67217278e-01 7.22734869e-01 -7.08155990e-01 6.14772797e-01 -3.62227291e-01 2.89472878e-01 1.43312737e-01 -8.98859352e-02 -1.12486109e-02 -2.72440493e-01 4.99888062e-01 9.45266068e-01 -2.79159732e-02 4.60634798e-01 2.68412620e-01 1.04360318e+00 6.22811504e-02 1.35350451e-01 -5.15672445e-01 -2.67400056e-01 4.61574823e-01 5.91993392e-01 -7.56854594e-01 -5.51474690e-01 -6.06011271e-01 9.09762681e-01 1.06143825e-01 7.18356311e-01 -7.33303130e-01 3.94180492e-02 5.59898615e-01 1.77465588e-01 2.68864393e-01 -4.76464897e-01 -2.49715537e-01 -1.30436730e+00 6.10278035e-03 -1.24392331e+00 4.05653149e-01 -1.14279890e+00 -1.32569039e+00 2.73169845e-01 3.33788335e-01 -1.06383252e+00 -2.89410111e-02 -7.31497586e-01 -4.35259610e-01 5.40639639e-01 -1.38776600e+00 -1.45958984e+00 -6.32111609e-01 8.62532973e-01 7.68033683e-01 -6.13752566e-02 5.63752234e-01 -1.73926547e-01 -2.67823320e-02 3.11568946e-01 -5.07327855e-01 -8.12984537e-03 4.91472781e-01 -9.81347799e-01 4.39018637e-01 1.00164843e+00 4.06992912e-01 1.00612879e+00 8.07323337e-01 -4.74593639e-01 -1.55453777e+00 -8.80875766e-01 7.22342014e-01 -3.36419135e-01 7.67840683e-01 -2.50341028e-01 -1.14398384e+00 7.85293818e-01 2.52939790e-01 -1.31036013e-01 6.49721086e-01 -2.71601558e-01 -9.40683722e-01 -2.14308240e-02 -7.63878167e-01 1.00343931e+00 1.35745835e+00 -1.16441214e+00 -1.13876545e+00 2.62513727e-01 1.13740766e+00 -1.67867675e-01 -4.62873161e-01 4.64538068e-01 5.98713875e-01 -1.08760571e+00 1.13898766e+00 -7.14527905e-01 3.38909894e-01 -5.66989005e-01 -6.61719322e-01 -1.01229429e+00 -3.80496413e-01 -3.70127857e-01 7.53724501e-02 8.81228507e-01 -7.29902908e-02 -6.46613538e-01 2.68706203e-01 4.25006390e-01 -1.78748846e-01 -2.67656386e-01 -7.03872383e-01 -8.40572953e-01 3.25586908e-02 -5.37411273e-01 8.15108120e-01 1.10520864e+00 3.25531000e-03 4.34634089e-01 3.47606651e-02 3.06037217e-01 3.78860742e-01 4.53670382e-01 8.00362110e-01 -1.01258242e+00 -3.55505407e-01 -6.16109669e-01 -5.28712690e-01 -1.36081362e+00 3.58383179e-01 -8.80047262e-01 -1.00468874e-01 -1.46716750e+00 1.91098511e-01 1.16713092e-01 -1.13631666e-01 2.66879827e-01 -3.20701212e-01 -2.47636661e-02 4.71074373e-01 4.75298673e-01 -6.08474672e-01 5.65319657e-01 1.47498345e+00 -2.20935673e-01 -1.96654990e-01 -5.12097895e-01 -5.55483818e-01 9.09028947e-01 6.25522912e-01 7.98356682e-02 -8.33530009e-01 -6.49390757e-01 1.75177276e-01 -2.90306449e-01 7.26878047e-01 -6.16885722e-01 1.44567803e-01 -5.10979831e-01 -2.10806374e-02 -5.07019818e-01 2.87131488e-01 -7.40502119e-01 -7.58879185e-02 6.55687094e-01 -5.64715326e-01 4.39774841e-01 2.05237061e-01 5.38748980e-01 -2.79411405e-01 -1.71401367e-01 7.48890102e-01 -3.05024415e-01 -1.37086844e+00 2.04136983e-01 -2.50947863e-01 3.43114942e-01 9.42852318e-01 -3.72531831e-01 -3.46566468e-01 -4.25131649e-01 -4.20241326e-01 -1.55031374e-02 5.63355029e-01 5.09120166e-01 8.92201543e-01 -1.52395225e+00 -5.62865376e-01 2.46917069e-01 5.11340976e-01 -2.53558904e-01 1.56053081e-01 5.88691175e-01 -6.65093601e-01 2.46324375e-01 -2.18087420e-01 -7.88045764e-01 -1.19328046e+00 1.00495017e+00 2.53054321e-01 5.22298992e-01 -6.32358372e-01 8.74033391e-01 9.60449636e-01 -1.23164132e-01 1.80726334e-01 -8.25928092e-01 -2.51544900e-02 -2.69578844e-01 5.79178810e-01 2.79904962e-01 -4.73143339e-01 -6.05849981e-01 -1.97260931e-01 6.72284663e-01 1.49799570e-01 -3.70106012e-01 6.96204305e-01 -3.39685529e-01 -4.66492385e-01 7.66333401e-01 8.77256930e-01 -1.32650629e-01 -1.02516091e+00 -6.26711011e-01 2.01557223e-02 -6.28826618e-01 -2.76921362e-01 -1.91450775e-01 -4.19987738e-01 9.43979084e-01 3.66156459e-01 3.64996970e-01 1.18124282e+00 4.00789261e-01 3.95000428e-01 3.87682110e-01 2.06533507e-01 -9.32931304e-01 6.34944797e-01 5.08973539e-01 1.10749519e+00 -9.95335221e-01 1.13062866e-01 -6.98709130e-01 -6.79657519e-01 9.05184448e-01 5.94592333e-01 -5.93699776e-02 4.17914033e-01 -3.91625434e-01 5.10049425e-02 -9.62483734e-02 -5.58587193e-01 -3.01988214e-01 5.26484311e-01 6.98105395e-01 2.50018567e-01 9.85962451e-02 -2.26703092e-01 2.61682421e-01 -2.26193547e-01 -3.16381574e-01 1.87311411e-01 8.68812203e-01 -6.97925687e-01 -7.61590958e-01 -4.42013830e-01 -8.84983987e-02 2.74371326e-01 -2.62446046e-01 -5.66617846e-01 7.88170397e-01 2.58517295e-01 7.19478846e-01 2.40300912e-02 -1.90005317e-01 3.81762296e-01 2.14853466e-01 8.56833458e-01 -6.60431862e-01 -1.93422154e-01 -2.02298686e-01 -3.31018597e-01 -7.40263879e-01 -8.10361445e-01 -6.42561913e-01 -1.50238550e+00 -2.18621776e-01 -6.19974732e-02 -1.37559503e-01 5.00225902e-01 1.29540741e+00 9.15914997e-02 4.58552092e-01 3.48515391e-01 -4.35302556e-01 -4.80307132e-01 -7.05905616e-01 -5.18123448e-01 9.38853681e-01 2.74216563e-01 -6.28901482e-01 -4.13010120e-01 7.93822706e-01]
[10.549153327941895, 1.5905457735061646]
76c2869c-2803-4005-946c-943c96fcfa77
da-lstm-a-dynamic-drift-adaptive-learning
2305.08767
null
https://arxiv.org/abs/2305.08767v1
https://arxiv.org/pdf/2305.08767v1.pdf
DA-LSTM: A Dynamic Drift-Adaptive Learning Framework for Interval Load Forecasting with LSTM Networks
Load forecasting is a crucial topic in energy management systems (EMS) due to its vital role in optimizing energy scheduling and enabling more flexible and intelligent power grid systems. As a result, these systems allow power utility companies to respond promptly to demands in the electricity market. Deep learning (DL) models have been commonly employed in load forecasting problems supported by adaptation mechanisms to cope with the changing pattern of consumption by customers, known as concept drift. A drift magnitude threshold should be defined to design change detection methods to identify drifts. While the drift magnitude in load forecasting problems can vary significantly over time, existing literature often assumes a fixed drift magnitude threshold, which should be dynamically adjusted rather than fixed during system evolution. To address this gap, in this paper, we propose a dynamic drift-adaptive Long Short-Term Memory (DA-LSTM) framework that can improve the performance of load forecasting models without requiring a drift threshold setting. We integrate several strategies into the framework based on active and passive adaptation approaches. To evaluate DA-LSTM in real-life settings, we thoroughly analyze the proposed framework and deploy it in a real-world problem through a cloud-based environment. Efficiency is evaluated in terms of the prediction performance of each approach and computational cost. The experiments show performance improvements on multiple evaluation metrics achieved by our framework compared to baseline methods from the literature. Finally, we present a trade-off analysis between prediction performance and computational costs.
['Jonas Forsman', 'Andreas Theocharis', 'Andreas Kassler', 'Bestoun S. Ahmed', 'Phil Aupke', 'Firas Bayram']
2023-05-15
null
null
null
null
['change-detection', 'load-forecasting', 'energy-management']
['computer-vision', 'miscellaneous', 'time-series']
[-2.38463908e-01 -7.76033819e-01 -1.36325732e-01 -4.33454067e-01 -1.47969902e-01 -4.74965304e-01 5.32361329e-01 1.24207869e-01 -3.50679368e-01 8.29418898e-01 -1.61161557e-01 -1.53627560e-01 -2.99628615e-01 -9.63189662e-01 -2.09335864e-01 -1.12329173e+00 -2.48894185e-01 5.11482060e-01 7.93223456e-02 -1.76007807e-01 2.91312367e-01 5.51326334e-01 -1.40187287e+00 -5.22969700e-02 1.05542374e+00 1.20644426e+00 4.00107503e-01 3.83607388e-01 -2.55848616e-01 5.53761303e-01 -9.02957499e-01 3.40530992e-01 1.76909208e-01 -2.29007617e-01 -4.26330209e-01 -1.62260592e-01 -5.45091391e-01 -2.19073951e-01 1.57309398e-01 1.08215868e+00 8.32792521e-01 1.31290242e-01 4.69263822e-01 -1.40985727e+00 -3.75895619e-01 7.30251849e-01 -4.83445495e-01 9.42118108e-01 -7.58728832e-02 1.22131430e-01 5.57112277e-01 -4.91862833e-01 -1.06563896e-01 6.86549067e-01 5.23571253e-01 3.47383946e-01 -1.03571177e+00 -7.52558708e-01 5.60867906e-01 8.23361158e-01 -1.22462845e+00 -2.91446060e-01 9.84253168e-01 -4.98085916e-01 1.26970255e+00 1.90758571e-01 6.16485178e-01 8.85851979e-01 6.18225932e-01 6.93007290e-01 7.67837405e-01 -3.48445803e-01 7.36802220e-01 5.55835441e-02 -3.12639624e-02 -1.45181492e-01 2.57590979e-01 3.92954424e-02 -3.60575080e-01 -4.61937860e-02 1.89369351e-01 4.71548643e-03 -2.64546186e-01 -8.80784467e-02 -7.05111742e-01 6.35234535e-01 3.20555270e-01 7.60514915e-01 -6.81366444e-01 1.33227147e-02 7.37526774e-01 3.98830354e-01 8.73083591e-01 -2.65229912e-03 -6.66162729e-01 -3.95078719e-01 -1.06208301e+00 2.85326578e-02 6.96614146e-01 5.72444320e-01 2.07092077e-01 8.32082629e-01 -2.30528429e-01 7.26992846e-01 9.46182311e-02 3.99616748e-01 1.29399443e+00 -2.20049113e-01 1.55570105e-01 5.57943106e-01 2.84115255e-01 -6.30147696e-01 -7.26021826e-01 -8.75261068e-01 -1.06281602e+00 2.59448767e-01 -1.42884612e-01 -2.31037855e-01 -6.23235166e-01 1.68419778e+00 2.21825257e-01 3.24994832e-01 6.58136904e-02 5.03250718e-01 -5.75432777e-02 7.74090469e-01 1.31969452e-02 -7.22468436e-01 1.06424475e+00 -6.82134867e-01 -1.00517595e+00 3.85878347e-02 6.25258684e-01 -5.27910352e-01 8.82111132e-01 3.64527911e-01 -8.36970568e-01 -3.89275789e-01 -1.22090673e+00 7.33598590e-01 -6.00796521e-01 -1.02085322e-01 1.25425607e-01 6.93515122e-01 -9.73616898e-01 6.30718827e-01 -1.07572579e+00 -3.95891190e-01 4.94855829e-02 2.73398191e-01 7.41576076e-01 6.09115481e-01 -1.48914969e+00 1.10171640e+00 8.60817730e-01 3.98311555e-01 -6.47955179e-01 -7.70180821e-01 -3.26057702e-01 2.41089448e-01 2.16786966e-01 -5.37324667e-01 1.55242634e+00 -1.02937746e+00 -1.68778741e+00 4.70578000e-02 -1.62742317e-01 -9.89424467e-01 6.58885896e-01 -7.07124621e-02 -9.35922861e-01 -5.20779788e-01 -4.19284940e-01 -2.26713002e-01 7.36227572e-01 -9.24143076e-01 -9.79910851e-01 -2.59893894e-01 -3.34015429e-01 1.56187534e-01 -6.87867999e-01 -1.30179599e-01 3.87054890e-01 -6.68292284e-01 -2.07765117e-01 -6.11500978e-01 -1.53046876e-01 -8.04301798e-01 5.81582114e-02 -5.50276101e-01 1.32839203e+00 -5.84542513e-01 1.75711346e+00 -1.74467885e+00 -2.59938926e-01 2.52960801e-01 -3.34689140e-01 4.99011934e-01 2.97272652e-01 4.45313126e-01 -1.70606077e-01 -1.35726601e-01 -2.26558790e-01 -1.90224200e-01 2.35994980e-01 3.63444090e-01 -4.19332653e-01 2.54782379e-01 -1.07771173e-01 8.43191624e-01 -8.36402774e-01 2.40323409e-01 4.50032711e-01 4.07983631e-01 1.41697705e-01 1.46352261e-01 -5.02650976e-01 2.40807116e-01 -2.70163298e-01 4.45011139e-01 5.96617997e-01 -1.90839112e-01 1.38811931e-01 -2.87625683e-03 -2.15980589e-01 1.13251038e-01 -1.25560331e+00 1.09084487e+00 -7.38342643e-01 5.02557039e-01 -2.26499662e-01 -1.30390334e+00 1.09782803e+00 4.58879262e-01 7.72999167e-01 -8.40215445e-01 8.39255974e-02 3.92977089e-01 1.83982164e-01 -2.20579252e-01 4.04357582e-01 9.84417349e-02 4.20280784e-01 5.65294266e-01 -4.50037748e-01 4.28011328e-01 4.06501323e-01 -3.77130777e-01 8.84581447e-01 -1.04164757e-01 1.84981078e-01 -5.08658171e-01 8.26706052e-01 -3.30661416e-01 7.88252831e-01 2.25501060e-01 -1.94934085e-01 -2.51475960e-01 3.93213555e-02 -8.18329632e-01 -8.49773228e-01 -5.75812459e-01 -3.14206481e-01 9.49951053e-01 -2.68312812e-01 -6.22833222e-02 -5.67952573e-01 -3.45429063e-01 2.74832621e-02 1.27479911e+00 -4.62739736e-01 -3.10933471e-01 -5.84202647e-01 -1.61698699e+00 4.70216423e-02 6.49002135e-01 6.69060409e-01 -1.18842554e+00 -1.07762289e+00 8.11634243e-01 3.00158501e-01 -8.18467736e-01 -1.82753205e-01 4.47333664e-01 -8.15779209e-01 -7.46025085e-01 -5.62084854e-01 -3.64888459e-01 3.94457698e-01 -1.38851330e-01 1.22466421e+00 -2.41392344e-01 -1.09781757e-01 2.39780396e-01 -1.96364880e-01 -6.89385355e-01 -4.37321246e-01 4.94916499e-01 3.37789088e-01 -5.34567684e-02 5.63854456e-01 -8.75191510e-01 -7.46794283e-01 2.32088879e-01 -1.06626272e+00 -2.15958342e-01 2.80375779e-01 7.82848239e-01 3.33166152e-01 6.90415800e-01 1.23729610e+00 -5.44429421e-01 1.06929350e+00 -8.10682356e-01 -1.16082430e+00 4.50618595e-01 -1.53740168e+00 -4.45355438e-02 8.80559683e-01 -4.26800370e-01 -9.29132462e-01 -2.96067595e-01 5.46884388e-02 -3.67449760e-01 2.26258337e-01 6.65222704e-01 -4.97373007e-02 1.85286820e-01 1.23460054e-01 4.81371135e-01 -4.43687558e-01 -5.33000946e-01 -8.39810818e-02 7.56179988e-01 3.54787558e-01 -3.86281520e-01 5.98256350e-01 3.29308584e-02 -1.64689645e-01 -3.75617683e-01 -5.46339273e-01 -1.45528153e-01 -5.55216670e-01 -3.27338427e-01 2.54725933e-01 -7.60497808e-01 -8.98582339e-01 7.39482641e-01 -8.08335304e-01 -3.50798160e-01 -2.64648020e-01 5.19037433e-02 -2.73823798e-01 -6.39094263e-02 -4.74627137e-01 -1.04612732e+00 -8.69513750e-01 -9.38356400e-01 3.25848639e-01 4.19116288e-01 -1.03251778e-01 -1.50010514e+00 1.89314783e-01 -4.49810237e-01 1.11269677e+00 4.09437448e-01 1.00547624e+00 -6.91011667e-01 -1.60563901e-01 -1.64139807e-01 1.97002888e-01 5.21741033e-01 5.19296050e-01 -1.59798101e-01 -9.23839808e-01 -8.23562801e-01 2.33230218e-01 3.81170630e-01 4.70058322e-01 4.07070637e-01 1.28435540e+00 -4.17103231e-01 -1.97024688e-01 4.72222686e-01 1.57865286e+00 8.38391066e-01 1.87271416e-01 7.32281983e-01 3.54245782e-01 1.05873734e-01 2.02751577e-01 8.77120256e-01 4.65649456e-01 6.24131322e-01 3.20438862e-01 1.75252184e-01 2.98199266e-01 7.33414292e-02 4.76036668e-01 1.15135193e+00 2.29396656e-01 -3.76024038e-01 -1.01044691e+00 6.20950997e-01 -1.94562829e+00 -8.60216141e-01 2.20344007e-01 2.27225757e+00 7.10088432e-01 4.85219270e-01 1.84354201e-01 5.90842664e-01 5.69309235e-01 1.06959820e-01 -1.23862135e+00 -5.28658450e-01 -1.73627928e-01 -3.48547176e-02 4.93300170e-01 3.76517296e-01 -8.39455664e-01 4.25930500e-01 5.71569109e+00 7.21145391e-01 -1.57153010e+00 2.00536221e-01 7.32154548e-01 -3.71295601e-01 -6.26808330e-02 -4.68971133e-01 -7.78335631e-01 1.09171665e+00 1.45477700e+00 -9.48124588e-01 3.99287432e-01 7.83216059e-01 7.80419588e-01 -9.30296257e-02 -1.12659454e+00 9.21467662e-01 -1.84182644e-01 -1.09824812e+00 -1.38063595e-01 -9.65760872e-02 8.65300357e-01 2.87385255e-01 3.60608250e-02 2.62753218e-01 2.15029478e-01 -7.36113966e-01 5.25972009e-01 6.08024597e-01 2.09171593e-01 -1.07019448e+00 1.08369577e+00 5.10205030e-01 -1.36519086e+00 -4.75308955e-01 -1.66844293e-01 -1.57894403e-01 3.28893125e-01 1.01841974e+00 -6.75741494e-01 7.89558411e-01 9.00311887e-01 6.18143618e-01 -3.44886959e-01 9.98766601e-01 1.23324312e-01 8.57443273e-01 -4.78961945e-01 3.13118882e-02 1.20966300e-01 -1.31539449e-01 3.97052914e-01 9.79944468e-01 4.78774130e-01 -2.87039638e-01 2.21498042e-01 7.80874193e-01 9.51630697e-02 -5.45503236e-02 -7.74264187e-02 2.00109154e-01 7.74544418e-01 1.10677779e+00 -6.43484235e-01 -3.65946829e-01 -7.24048987e-02 7.34253109e-01 -2.33986318e-01 4.75060523e-01 -7.88989425e-01 -2.64534742e-01 7.29009807e-01 -1.81285013e-02 2.43711784e-01 -2.12679014e-01 -2.92820632e-01 -9.45537269e-01 2.33940303e-01 -4.72291201e-01 4.53142017e-01 -3.68609428e-01 -1.50988233e+00 7.21675634e-01 1.18577451e-01 -1.17256081e+00 -7.85958290e-01 -1.79596066e-01 -9.20389593e-01 9.52647448e-01 -1.97368455e+00 -6.12650812e-01 -2.22939178e-01 3.66718411e-01 9.01606441e-01 -1.75031632e-01 6.62720919e-01 4.30174351e-01 -1.05713212e+00 3.95176470e-01 7.46093094e-01 -2.51660496e-01 1.84042394e-01 -1.30143595e+00 4.43151206e-01 9.49692845e-01 -2.99154490e-01 2.44536735e-02 8.33270907e-01 -4.26530302e-01 -1.07991111e+00 -1.14972174e+00 4.65489596e-01 1.29430428e-01 6.71805143e-01 -1.89867079e-01 -1.27411354e+00 5.57100594e-01 5.13945043e-01 -3.08487248e-02 4.98852611e-01 -2.73100078e-01 2.71125168e-01 -5.62848508e-01 -1.22435093e+00 1.05156809e-01 3.89412701e-01 -4.50773723e-02 -2.38976642e-01 1.25023782e-01 2.42662743e-01 -2.56205887e-01 -9.67131376e-01 6.51478887e-01 2.18120053e-01 -7.50811040e-01 4.56636429e-01 -1.66672528e-01 -4.46210325e-01 -5.07590294e-01 -7.17547759e-02 -1.81153405e+00 -3.47077757e-01 -6.94690168e-01 -7.88455963e-01 1.36110449e+00 1.59891367e-01 -1.07475984e+00 5.98049819e-01 6.32283449e-01 1.00768041e-02 -1.00705993e+00 -1.09541428e+00 -1.00677598e+00 1.57811671e-01 -3.68144453e-01 1.27100587e+00 9.44227755e-01 -2.21854866e-01 -2.86201034e-02 -1.57884240e-01 3.04129839e-01 4.68268752e-01 1.47368595e-01 1.46294713e-01 -1.10327208e+00 5.53160235e-02 -7.56711245e-01 -2.24683076e-01 -4.61650163e-01 8.07242543e-02 -6.29878879e-01 -1.97895303e-01 -1.41441345e+00 -1.88311428e-01 -3.44648361e-01 -1.04128993e+00 4.48223323e-01 -4.64121550e-02 -2.62518346e-01 1.40003696e-01 3.60922217e-01 -2.69340038e-01 9.37069476e-01 2.76033551e-01 -8.74088109e-02 -4.04310107e-01 3.07202637e-01 -3.91725451e-02 6.46577418e-01 1.37751079e+00 -2.84873754e-01 -6.64848328e-01 -4.48785096e-01 2.80756593e-01 -3.25859219e-01 -1.84202433e-01 -1.22841227e+00 3.93403530e-01 -1.42723709e-01 6.47896230e-01 -7.62794852e-01 -2.22135589e-01 -8.60320807e-01 4.95366514e-01 8.06297600e-01 -2.56565306e-02 8.48185420e-01 3.62645537e-01 3.77521724e-01 -7.47624189e-02 -1.42223328e-01 7.03489244e-01 1.50453761e-01 -8.89780462e-01 2.18668029e-01 -5.14840543e-01 -3.21582317e-01 1.27190840e+00 -3.35048996e-02 -2.09872320e-01 -3.08814228e-01 -5.59544683e-01 7.10628629e-01 1.08647533e-01 7.81347692e-01 2.25906670e-01 -1.35778046e+00 -6.76464677e-01 2.40386084e-01 -1.62800997e-01 -2.73577750e-01 1.67990133e-01 5.80119789e-01 -9.34833437e-02 5.38368464e-01 8.37885663e-02 -6.63919687e-01 -8.04595232e-01 4.86856312e-01 7.99779594e-01 -5.11898637e-01 -4.75212842e-01 2.18773186e-01 -4.28591698e-01 4.93618846e-02 4.47997510e-01 -6.42566979e-01 -2.26553544e-01 2.87303567e-01 7.28391111e-01 6.90790474e-01 7.17989266e-01 -2.84169197e-01 -3.23860765e-01 2.80928999e-01 -1.72022488e-02 3.34106505e-01 1.40106440e+00 -3.87029409e-01 -1.01159275e-01 1.01058912e+00 7.02431023e-01 -6.51754379e-01 -1.02990866e+00 -3.13642979e-01 7.79104590e-01 -8.38098004e-02 4.01633263e-01 -1.11162531e+00 -1.32522964e+00 5.29467940e-01 1.34318316e+00 6.93067491e-01 1.49196637e+00 -7.01730728e-01 9.89851534e-01 1.79931551e-01 3.35527450e-01 -1.30156565e+00 -3.73571992e-01 3.36758763e-01 6.06775820e-01 -1.01192617e+00 -2.23506391e-01 5.65733433e-01 -2.83135504e-01 1.13766146e+00 5.44132769e-01 2.57199079e-01 8.62213969e-01 5.38768888e-01 1.44270584e-01 2.51304269e-01 -1.23977804e+00 -2.81586242e-03 1.22465622e-02 1.92808032e-01 3.64265800e-01 2.48910770e-01 -2.83964306e-01 5.29442549e-01 -1.33635595e-01 2.51342058e-01 2.83213645e-01 1.00730801e+00 -3.43365520e-01 -1.23216629e+00 -3.41099620e-01 4.26316947e-01 -4.89903390e-01 2.52276421e-01 3.05516005e-01 3.65693361e-01 1.45853132e-01 9.51007426e-01 2.86925703e-01 -2.19518930e-01 4.46174890e-01 2.41263419e-01 -5.74908778e-02 -4.43124361e-02 -7.35202014e-01 1.47214662e-02 -4.61127162e-01 -1.93557099e-01 -3.45824271e-01 -7.04148531e-01 -1.20772207e+00 -2.95748472e-01 -3.40090275e-01 3.23567092e-01 8.96833837e-01 1.04545426e+00 3.87617707e-01 8.99686575e-01 1.05189848e+00 -7.18401790e-01 -8.85353327e-01 -1.22222447e+00 -5.38758695e-01 3.56955938e-02 2.71300435e-01 -6.75778389e-01 -4.39648926e-01 -1.14549264e-01]
[6.18073034286499, 2.7780234813690186]
c518076f-0113-48ef-8177-7f70f4420e0a
enhancing-aspect-term-extraction-with-soft
null
null
https://aclanthology.org/2020.emnlp-main.164
https://aclanthology.org/2020.emnlp-main.164.pdf
Enhancing Aspect Term Extraction with Soft Prototypes
Aspect term extraction (ATE) aims to extract aspect terms from a review sentence that users have expressed opinions on. Existing studies mostly focus on designing neural sequence taggers to extract linguistic features from the token level. However, since the aspect terms and context words usually exhibit long-tail distributions, these taggers often converge to an inferior state without enough sample exposure. In this paper, we propose to tackle this problem by correlating words with each other through soft prototypes. These prototypes, generated by a soft retrieval process, can introduce global knowledge from internal or external data and serve as the supporting evidence for discovering the aspect terms. Our proposed model is a general framework and can be combined with almost all sequence taggers. Experiments on four SemEval datasets show that our model boosts the performance of three typical ATE methods by a large margin.
['Tieyun Qian', 'Zhuang Chen']
null
null
null
null
emnlp-2020-11
['extract-aspect']
['natural-language-processing']
[ 9.63646472e-02 1.03992531e-02 -6.68917954e-01 -3.96707863e-01 -1.00815773e+00 -6.95754826e-01 7.91988552e-01 3.36465478e-01 -4.23041344e-01 6.65441394e-01 2.60496080e-01 -2.63881266e-01 2.44513139e-01 -7.48994410e-01 -4.68948275e-01 -6.16603732e-01 2.50179589e-01 4.54468966e-01 9.48416889e-02 -2.52290636e-01 4.73513305e-01 -1.49154529e-01 -1.27273297e+00 4.14678156e-01 9.92469490e-01 8.49654794e-01 -2.85542533e-02 2.38242969e-01 -8.56957138e-01 3.40339482e-01 -7.54834175e-01 -7.24176764e-01 -1.26425192e-01 -3.07925791e-01 -5.66234410e-01 -8.27584043e-02 1.02549076e-01 4.84689735e-02 3.37401122e-01 1.32741785e+00 5.15696704e-01 -4.49097864e-02 7.89574564e-01 -7.36073434e-01 -9.80664909e-01 1.05987096e+00 -6.09514952e-01 8.78585428e-02 4.21943814e-01 8.37726612e-03 1.65158427e+00 -1.24831843e+00 5.20802915e-01 9.65680122e-01 7.42679179e-01 4.76217031e-01 -8.59433651e-01 -3.56414735e-01 5.97322404e-01 1.51411623e-01 -1.21760929e+00 -3.21935713e-01 8.65719020e-01 -1.30769610e-01 1.21625340e+00 2.16208264e-01 8.47487330e-01 1.17014337e+00 2.31882811e-01 1.33255649e+00 9.22546268e-01 -3.65069121e-01 4.92851585e-01 4.01394874e-01 6.63247764e-01 5.45119226e-01 6.68550968e-01 -3.70940328e-01 -7.37816870e-01 -5.43704391e-01 1.91282516e-03 -5.91193885e-02 -1.49006546e-01 8.41447636e-02 -7.26803422e-01 8.27843308e-01 3.47406976e-02 5.21810055e-01 -7.37764716e-01 -4.97574657e-02 4.24591899e-01 2.32133552e-01 7.01825619e-01 5.77803135e-01 -8.21997166e-01 -1.68629855e-01 -7.44050562e-01 3.28829326e-02 9.85309541e-01 1.06649435e+00 7.82734573e-01 -1.31103203e-01 -4.07312304e-01 7.35827506e-01 4.84924912e-01 7.91206956e-01 8.10148299e-01 -1.23982482e-01 3.77341062e-01 8.33944917e-01 -1.40065879e-01 -8.25435042e-01 -2.57092059e-01 -6.68172419e-01 -5.11786819e-01 -4.33961242e-01 -1.98301435e-01 -2.99785376e-01 -1.17092001e+00 1.46389055e+00 3.60201091e-01 -3.97221260e-02 -9.47953388e-03 6.64687455e-01 9.71236527e-01 6.94097698e-01 1.56967416e-01 -2.65541583e-01 1.80473280e+00 -8.76266956e-01 -1.04517186e+00 -6.57002628e-01 6.23804450e-01 -9.44734693e-01 1.00202537e+00 4.47104901e-01 -9.01863337e-01 -2.12770477e-01 -9.91160154e-01 1.69460148e-01 -5.02679765e-01 1.75662160e-01 7.99678862e-01 5.59673667e-01 -7.79058218e-01 2.55520016e-01 -6.41115725e-01 -1.30203709e-01 2.81919897e-01 1.37770399e-01 -1.20571688e-01 4.05310877e-02 -1.39902270e+00 6.81525290e-01 3.75050753e-01 2.94125319e-01 -4.81066227e-01 -3.88373584e-01 -8.28247786e-01 1.51648358e-01 4.51171279e-01 -9.69763637e-01 1.19707251e+00 -1.00348902e+00 -1.29504716e+00 6.11534238e-01 -6.41784191e-01 -1.64997399e-01 -3.11478347e-01 -4.87069219e-01 -4.53415513e-01 -1.46206515e-02 1.48272783e-01 -1.38552883e-03 1.02851617e+00 -1.03058958e+00 -5.65035820e-01 -3.84449929e-01 -1.52757436e-01 8.89834687e-02 -5.71277499e-01 1.31679371e-01 -5.88704586e-01 -7.55375504e-01 -1.16262836e-02 -8.10865402e-01 -4.62981075e-01 -6.17689013e-01 -4.49216247e-01 -4.94386971e-01 5.42206347e-01 -3.83446068e-01 1.48664105e+00 -1.83356595e+00 -8.95783976e-02 4.05449450e-01 2.56220520e-01 6.61315441e-01 -3.12703252e-01 4.78580326e-01 1.45905763e-01 3.17773640e-01 -2.80598581e-01 -2.42966726e-01 8.81896913e-02 1.12845846e-01 -5.16639471e-01 3.08946073e-02 2.97612995e-01 1.28772700e+00 -1.00755560e+00 -5.19727945e-01 -4.90637869e-01 4.31684881e-01 -4.11800236e-01 1.96629673e-01 -3.85721952e-01 -2.94222772e-01 -8.52910042e-01 7.55157828e-01 4.64151353e-01 -3.97290111e-01 2.19055042e-01 2.40581278e-02 2.55833834e-01 7.62934566e-01 -7.52905667e-01 1.47961056e+00 -4.35276508e-01 3.88064206e-01 -3.19972783e-01 -9.59307194e-01 1.03575766e+00 4.72699344e-01 2.14293480e-01 -5.01366973e-01 1.60448104e-01 2.02717394e-01 -1.74712926e-01 -6.08949125e-01 7.19001293e-01 -5.07611215e-01 -2.09464848e-01 6.54145420e-01 1.57361954e-01 3.00070476e-02 2.13156790e-01 1.71453297e-01 1.04307187e+00 -3.84929627e-02 7.06551552e-01 9.56377611e-02 5.64925849e-01 -1.20438226e-01 7.50123799e-01 8.22197378e-01 2.36021858e-02 4.16035175e-01 5.87560356e-01 -2.21131742e-01 -7.97085881e-01 -8.57226908e-01 2.85818931e-02 9.77044821e-01 -7.42189288e-02 -9.24648643e-01 -4.69292372e-01 -1.18025506e+00 -3.12249243e-01 7.47602701e-01 -5.77592790e-01 -4.19451416e-01 -4.20523643e-01 -9.36261177e-01 4.33271676e-01 5.71467757e-01 1.34047121e-01 -1.16484964e+00 -8.40703920e-02 4.19186711e-01 -3.38904738e-01 -1.05712008e+00 -5.84355235e-01 3.47221017e-01 -9.36803758e-01 -6.69824719e-01 -6.07516527e-01 -6.93749249e-01 7.01499045e-01 2.08454892e-01 1.35834861e+00 2.94041812e-01 1.61997452e-02 8.66227522e-02 -7.92114794e-01 -5.74008584e-01 1.30669968e-02 4.32694674e-01 3.01304609e-02 3.17816883e-02 1.23117828e+00 -5.43848634e-01 -4.80130583e-01 -1.15778573e-01 -9.96531427e-01 -4.34120864e-01 1.02335751e+00 9.15993154e-01 6.13398194e-01 -1.85387969e-01 7.35924900e-01 -1.20181847e+00 1.09463131e+00 -5.42618573e-01 -3.18128496e-01 3.43434632e-01 -1.13929534e+00 4.16985154e-01 4.79076147e-01 -4.52336192e-01 -1.09459519e+00 -2.53244728e-01 -2.10685462e-01 6.53961077e-02 3.78297344e-02 1.13233066e+00 -2.42951512e-01 4.95857924e-01 3.94749850e-01 5.26468217e-01 -3.57692391e-01 -5.08763552e-01 5.67102969e-01 7.28665352e-01 2.01812852e-02 -5.62488139e-01 8.00338745e-01 2.69898802e-01 -3.51011127e-01 -7.92105138e-01 -1.36116755e+00 -7.08208740e-01 -4.38450515e-01 8.30343291e-02 4.96197820e-01 -7.55281508e-01 -3.75130743e-01 2.91107625e-01 -1.22281778e+00 3.54158610e-01 -3.54192585e-01 4.05605853e-01 -3.06102727e-02 4.52060819e-01 -5.03202617e-01 -9.53239381e-01 -9.00248826e-01 -7.79325306e-01 1.28608549e+00 3.03588420e-01 -4.98346210e-01 -1.08522463e+00 3.82909000e-01 1.43179163e-01 4.14474487e-01 -3.48353207e-01 8.86431575e-01 -1.11711049e+00 -4.47478324e-01 -5.74777126e-01 1.52708963e-01 4.12700713e-01 2.53986925e-01 -1.16700247e-01 -1.03313756e+00 7.16830045e-02 1.92488119e-01 -5.69107607e-02 1.12438715e+00 1.61792591e-01 5.93994737e-01 -7.50474095e-01 -4.81708735e-01 1.81486443e-01 1.18485916e+00 4.11565900e-02 4.18913603e-01 3.72318983e-01 6.08901322e-01 4.31349874e-01 7.70476997e-01 3.45651507e-01 3.14353317e-01 4.67073321e-01 -1.13154955e-01 2.45997474e-01 1.38181463e-01 -3.23950887e-01 7.97918618e-01 1.45340204e+00 1.27748057e-01 -2.88281739e-01 -7.31469691e-01 7.87334800e-01 -1.82661462e+00 -9.61608171e-01 -6.01531006e-02 1.90740180e+00 1.28437412e+00 4.02696252e-01 -1.13801308e-01 -1.18442722e-01 4.54050541e-01 3.18221182e-01 -4.01813805e-01 -5.70560098e-01 -2.42541969e-01 3.29779208e-01 1.19243667e-01 4.39937115e-01 -9.11989868e-01 1.13280904e+00 6.41972542e+00 8.87746215e-01 -9.79897320e-01 8.24037939e-02 1.86980695e-01 -1.04813397e-01 -9.74961162e-01 2.87122637e-01 -1.13704407e+00 3.45054001e-01 1.02356088e+00 -4.83494878e-01 -8.39862451e-02 9.66104031e-01 -1.98352300e-02 7.62470812e-02 -8.76292527e-01 5.75274467e-01 3.71001422e-01 -9.92076457e-01 4.23809767e-01 8.67348760e-02 8.18507373e-01 7.23278224e-02 8.53291675e-02 4.82942879e-01 2.69563407e-01 -8.10536087e-01 2.90406585e-01 5.93679845e-01 2.17151865e-01 -6.69916391e-01 1.05382645e+00 3.45783055e-01 -1.15654373e+00 2.51916826e-01 -4.10083711e-01 9.32794809e-02 3.61041754e-01 1.21504891e+00 -9.76956964e-01 5.51444948e-01 3.68339211e-01 8.14943433e-01 -5.17234147e-01 9.94458199e-01 -7.65126824e-01 1.03148961e+00 -1.47794738e-01 -6.42496347e-01 4.61311668e-01 -2.78086782e-01 7.42505312e-01 1.30873966e+00 2.00292960e-01 -1.02089897e-01 -6.57263026e-02 6.84317827e-01 -3.26379947e-02 4.54242051e-01 -8.20751965e-01 -3.37545931e-01 1.87090725e-01 1.45746136e+00 -6.20367646e-01 -6.87371194e-01 -4.91302997e-01 7.29844809e-01 3.81492794e-01 4.88546491e-01 -4.68402892e-01 -6.47277951e-01 7.71855354e-01 -3.94568175e-01 7.98619449e-01 -5.82903549e-02 -5.14501452e-01 -1.46865189e+00 4.35334623e-01 -9.46743190e-01 3.45061719e-01 -5.35358846e-01 -1.42149627e+00 8.11945200e-01 -4.36829358e-01 -1.30617797e+00 -4.29112703e-01 -6.22385740e-01 -7.19852805e-01 6.77555263e-01 -1.70667326e+00 -9.97420013e-01 3.28956336e-01 1.53578371e-01 4.79216069e-01 -1.87271193e-01 7.89870739e-01 1.49447739e-01 -4.04164016e-01 7.04477191e-01 -8.55360329e-02 1.95407942e-01 5.67604244e-01 -1.30081940e+00 4.43510324e-01 8.52227032e-01 4.61602181e-01 1.39237082e+00 7.68513024e-01 -8.92827868e-01 -1.38467646e+00 -9.18743014e-01 1.46258152e+00 -6.46356821e-01 7.21408725e-01 -2.37840191e-01 -1.11916697e+00 6.68004632e-01 3.59559536e-01 -3.15103859e-01 1.09389126e+00 7.37267613e-01 -6.49472654e-01 -3.69880982e-02 -7.13092864e-01 6.74906313e-01 8.29383194e-01 -7.62203753e-01 -1.19524276e+00 4.79616970e-02 7.87009776e-01 1.80059969e-01 -5.39768636e-01 3.36279392e-01 5.92843950e-01 -4.75841671e-01 6.69528425e-01 -9.80261683e-01 4.20451730e-01 -4.01016504e-01 3.28752883e-02 -1.53612363e+00 -1.70500800e-01 -6.17588103e-01 -5.76004565e-01 1.60121763e+00 8.85337353e-01 -6.80348992e-01 4.78628755e-01 4.66764897e-01 5.00149652e-02 -8.52845609e-01 -8.28162253e-01 -7.77145267e-01 -5.97087443e-02 -6.20129228e-01 7.86839485e-01 7.19369948e-01 4.24941480e-01 1.05632091e+00 -2.03829959e-01 -5.28089218e-02 5.25184155e-01 4.47141260e-01 4.05893981e-01 -1.06675661e+00 -2.82168597e-01 -6.26905084e-01 -2.98646927e-01 -1.22176516e+00 3.71417910e-01 -1.02301049e+00 4.07875389e-01 -1.55193400e+00 6.37808859e-01 -1.85429037e-01 -5.57650864e-01 5.79876006e-01 -9.32274520e-01 1.29986987e-01 -3.27851295e-01 -6.36464059e-02 -8.86776149e-01 1.01570213e+00 1.04332161e+00 -3.56707543e-01 -3.35146077e-02 1.06098555e-01 -1.18542230e+00 8.82865846e-01 7.53610194e-01 -7.45057404e-01 -3.32036167e-01 -2.60183036e-01 9.69760537e-01 -5.27819335e-01 -2.00808614e-01 -3.38922411e-01 1.85840949e-01 -2.25547757e-02 6.41555861e-02 -6.29502535e-01 9.58682224e-02 -7.68940091e-01 -4.91977364e-01 1.71037391e-01 -4.16018009e-01 3.79914418e-02 -7.34559596e-02 5.63438475e-01 -4.22273457e-01 -5.40695548e-01 8.92445743e-02 -1.44150034e-01 -4.32253301e-01 3.39668632e-01 -6.42402887e-01 2.62937695e-01 4.65411901e-01 2.01947749e-01 -3.74245197e-01 -4.77160335e-01 -3.06434542e-01 1.89158320e-01 8.73311311e-02 7.34573364e-01 7.71383882e-01 -1.46627283e+00 -6.22994542e-01 9.23964083e-02 5.29784203e-01 -1.80473283e-01 -2.01848701e-01 8.35517228e-01 3.36641014e-01 6.56965554e-01 4.74914730e-01 -2.98023194e-01 -1.25668585e+00 7.09863603e-01 -8.42290968e-02 -4.16620165e-01 -5.26020825e-01 7.93294370e-01 9.76607949e-02 -5.30772567e-01 -4.63142283e-02 -2.51593322e-01 -4.99089420e-01 4.15569693e-01 7.74122059e-01 -2.45308608e-01 1.85618579e-01 -4.11870986e-01 -4.29314971e-01 6.17061198e-01 -3.53772014e-01 -2.24760965e-01 1.45964813e+00 -2.23004714e-01 -3.55184674e-01 6.39377356e-01 1.17620373e+00 2.68166482e-01 -4.15682077e-01 -6.68025792e-01 5.69986701e-01 -2.01170042e-01 -1.65829121e-03 -6.48480833e-01 -1.05149031e+00 7.73153722e-01 6.47577345e-02 3.65138322e-01 8.72944832e-01 3.13984752e-01 1.06270480e+00 6.58716917e-01 1.01814546e-01 -1.17337012e+00 -3.02837100e-02 7.92426109e-01 5.35846233e-01 -1.08304751e+00 6.11291267e-03 -1.01172417e-01 -6.67285740e-01 8.42316568e-01 5.83481550e-01 1.20946459e-01 6.72472835e-01 3.03989977e-01 3.33676606e-01 -5.42515695e-01 -1.08313870e+00 -5.38435459e-01 6.74360991e-01 3.73818874e-01 6.84521437e-01 -7.89529383e-02 -8.26661646e-01 9.82491851e-01 -1.63120195e-01 -8.33734274e-02 3.05404812e-01 9.67609704e-01 -7.47962534e-01 -1.44292998e+00 -3.01095541e-03 6.01692855e-01 -7.97844350e-01 -7.25416005e-01 -5.11474729e-01 6.36365041e-02 -2.55638868e-01 9.09946382e-01 -4.08522934e-01 -2.71583498e-01 1.82699695e-01 4.50033098e-01 -2.44229417e-02 -7.40258574e-01 -6.41875207e-01 1.89928606e-01 3.84633601e-01 -3.45132917e-01 -3.75346631e-01 -7.31979012e-01 -1.13019323e+00 2.13276431e-01 -6.28838539e-01 6.40804350e-01 6.40034199e-01 1.32050121e+00 4.86302495e-01 4.74827737e-01 6.73464835e-01 -7.65155032e-02 -5.96302450e-01 -1.15048754e+00 -4.54997987e-01 2.42288873e-01 2.58014709e-01 -2.90563554e-01 -3.66667330e-01 -8.21096525e-02]
[11.35962963104248, 6.749284744262695]
be4f35ed-86e4-40bb-86ff-f00a27efc7a0
dpf-learning-dense-prediction-fields-with
2303.16890
null
https://arxiv.org/abs/2303.16890v1
https://arxiv.org/pdf/2303.16890v1.pdf
DPF: Learning Dense Prediction Fields with Weak Supervision
Nowadays, many visual scene understanding problems are addressed by dense prediction networks. But pixel-wise dense annotations are very expensive (e.g., for scene parsing) or impossible (e.g., for intrinsic image decomposition), motivating us to leverage cheap point-level weak supervision. However, existing pointly-supervised methods still use the same architecture designed for full supervision. In stark contrast to them, we propose a new paradigm that makes predictions for point coordinate queries, as inspired by the recent success of implicit representations, like distance or radiance fields. As such, the method is named as dense prediction fields (DPFs). DPFs generate expressive intermediate features for continuous sub-pixel locations, thus allowing outputs of an arbitrary resolution. DPFs are naturally compatible with point-level supervision. We showcase the effectiveness of DPFs using two substantially different tasks: high-level semantic parsing and low-level intrinsic image decomposition. In these two cases, supervision comes in the form of single-point semantic category and two-point relative reflectance, respectively. As benchmarked by three large-scale public datasets PASCALContext, ADE20K and IIW, DPFs set new state-of-the-art performance on all of them with significant margins. Code can be accessed at https://github.com/cxx226/DPF.
['Ya-Qin Zhang', 'Guyue Zhou', 'Hao Zhao', 'Qiang Zhou', 'Yupeng Zheng', 'Yuhang Zheng', 'Xiaoxue Chen']
2023-03-29
null
http://openaccess.thecvf.com//content/CVPR2023/html/Chen_DPF_Learning_Dense_Prediction_Fields_With_Weak_Supervision_CVPR_2023_paper.html
http://openaccess.thecvf.com//content/CVPR2023/papers/Chen_DPF_Learning_Dense_Prediction_Fields_With_Weak_Supervision_CVPR_2023_paper.pdf
cvpr-2023-1
['scene-parsing', 'intrinsic-image-decomposition', 'semantic-parsing']
['computer-vision', 'computer-vision', 'natural-language-processing']
[ 4.26443666e-01 3.80242288e-01 -2.91258574e-01 -6.31684601e-01 -8.43714535e-01 -6.93975747e-01 7.31345177e-01 1.26417428e-01 -2.34937608e-01 4.65054750e-01 1.00939527e-01 -2.20924631e-01 1.83949359e-02 -1.07347369e+00 -1.24342775e+00 -6.82869613e-01 2.29107112e-01 2.21641511e-01 3.45723897e-01 -2.80974984e-01 1.84135184e-01 4.85221565e-01 -1.86973894e+00 5.59317946e-01 6.88564479e-01 1.23366320e+00 4.52157468e-01 5.21567047e-01 -3.45417470e-01 6.86165810e-01 -1.32739637e-02 -5.03834426e-01 4.66382653e-01 1.51637301e-01 -9.18770730e-01 1.88643843e-01 6.69403255e-01 -2.40036994e-01 -1.91443905e-01 9.44979131e-01 1.11083597e-01 -1.22229038e-02 3.76714736e-01 -1.15959477e+00 -7.31662989e-01 3.31010640e-01 -5.93566239e-01 -4.36288804e-01 3.21167946e-01 3.49291861e-01 1.44406819e+00 -9.62597251e-01 5.71199119e-01 1.15710247e+00 8.00352573e-01 3.94124448e-01 -1.52477527e+00 -1.54811233e-01 3.95807505e-01 1.02629744e-01 -1.13804233e+00 -4.96210784e-01 9.66612875e-01 -3.62420201e-01 9.77839649e-01 4.64053035e-01 3.89857918e-01 1.22718835e+00 -3.16823572e-01 8.76625836e-01 1.26636338e+00 -3.40466112e-01 2.18170762e-01 -1.13845384e-03 1.14422344e-01 6.47375286e-01 -8.42120126e-02 7.73159936e-02 -5.97372055e-01 5.75203076e-02 8.54005039e-01 1.29937962e-01 -3.02738965e-01 -5.76211929e-01 -1.25008142e+00 6.92743957e-01 8.82508755e-01 -1.68559421e-02 -1.19032994e-01 3.00246000e-01 1.95681483e-01 1.04679652e-01 5.26121557e-01 2.93225527e-01 -7.67023504e-01 4.29592840e-02 -5.38758397e-01 1.43851846e-01 5.89970946e-01 9.89041865e-01 1.28867269e+00 -4.46582109e-01 1.01841904e-01 8.51841629e-01 2.84969002e-01 4.28004235e-01 -6.87743053e-02 -1.29930019e+00 4.71356302e-01 6.27917588e-01 1.25183851e-01 -9.22056377e-01 -4.75893855e-01 -3.15000504e-01 -7.45122850e-01 4.27143961e-01 6.27394021e-01 3.14691126e-01 -1.04236138e+00 1.75098968e+00 4.07283932e-01 1.91315249e-01 -1.94389056e-02 9.08077359e-01 8.88605595e-01 7.03568757e-01 2.21805617e-01 2.29528338e-01 1.46360111e+00 -1.03461504e+00 -9.06896591e-02 -4.07520771e-01 5.91618121e-01 -5.07799029e-01 1.54802501e+00 4.20913249e-01 -1.06094635e+00 -6.38021290e-01 -7.16396093e-01 -7.57377625e-01 -5.77313840e-01 1.24784738e-01 1.05052102e+00 3.30635905e-01 -1.22366619e+00 6.50772750e-01 -8.17081273e-01 -4.00279284e-01 6.01466596e-01 1.11540452e-01 -4.96890664e-01 -2.17367023e-01 -9.45631027e-01 4.31269526e-01 1.25652567e-01 4.90999185e-02 -6.27958417e-01 -1.09919477e+00 -1.01538682e+00 7.25317225e-02 3.43038410e-01 -8.45762074e-01 1.01111126e+00 -1.19366753e+00 -1.33628976e+00 1.51997209e+00 -1.38198912e-01 -4.60162640e-01 6.36211872e-01 -2.97298461e-01 5.39261475e-02 2.40750000e-01 3.30591232e-01 1.11779356e+00 8.16370904e-01 -1.45027161e+00 -4.72573280e-01 -5.35728097e-01 6.03076756e-01 1.47986829e-01 -7.71717802e-02 -2.55039215e-01 -5.13165057e-01 -3.70854974e-01 2.67747879e-01 -6.58823431e-01 -2.99066037e-01 5.16919434e-01 -6.76979959e-01 -3.00893009e-01 6.25280857e-01 -4.52516288e-01 3.28465492e-01 -2.37673140e+00 1.71882465e-01 -1.24873444e-01 2.76668876e-01 -1.15258463e-01 -2.62234956e-01 2.62851208e-01 -2.07911402e-01 8.13520551e-02 -5.70640445e-01 -7.42097974e-01 2.58852035e-01 4.27243292e-01 -5.08975983e-01 2.77177423e-01 6.06995761e-01 1.02312005e+00 -8.13762367e-01 -3.03685993e-01 5.38569510e-01 6.30802035e-01 -7.15439558e-01 1.66054130e-01 -6.60042763e-01 4.45004433e-01 -4.60358858e-01 7.71937668e-01 8.31944942e-01 -6.99341834e-01 -1.64635897e-01 -5.49686432e-01 -3.42755288e-01 2.56209552e-01 -9.19249952e-01 2.21187305e+00 -6.07720912e-01 3.87196571e-01 1.30198628e-01 -1.10930276e+00 8.12835813e-01 -2.16823101e-01 4.80199456e-01 -7.62646914e-01 -1.82120129e-01 8.78904983e-02 -6.28060818e-01 -1.57427639e-01 4.49723154e-01 8.57612416e-02 4.22572196e-02 -1.02815241e-01 -4.25088517e-02 -2.44366750e-01 -3.97321358e-02 2.24588394e-01 1.11946177e+00 7.01378107e-01 1.64878279e-01 -1.74330801e-01 4.26536053e-01 2.49487847e-01 5.57903349e-01 6.72701359e-01 2.77635828e-02 1.12242520e+00 6.26815319e-01 -5.13210475e-01 -1.02930677e+00 -1.20366395e+00 -4.32019591e-01 1.17068887e+00 3.75465155e-01 -5.76319337e-01 -6.69131815e-01 -5.25512278e-01 1.83544800e-01 4.84236091e-01 -6.48137033e-01 2.82206476e-01 -4.41164136e-01 -5.88780344e-01 2.03831166e-01 6.26451194e-01 6.76713526e-01 -9.17311370e-01 -6.57347560e-01 -1.68194398e-01 -1.20675527e-02 -1.46913564e+00 2.68234573e-02 4.43983048e-01 -6.77040756e-01 -1.03049481e+00 -5.86371660e-01 -4.99357492e-01 5.29447913e-01 4.46647018e-01 1.50678241e+00 9.72480848e-02 -4.32715565e-01 5.30160427e-01 -3.50470036e-01 -8.48890319e-02 1.01095781e-01 1.36564821e-01 -2.69417942e-01 3.04541495e-02 2.45122164e-01 -8.41981947e-01 -8.57262850e-01 2.08005160e-01 -7.91319489e-01 5.19454658e-01 6.09541774e-01 6.21385694e-01 1.07121158e+00 -5.06502092e-01 2.51358360e-01 -1.02057183e+00 -3.22025955e-01 -3.91067088e-01 -8.08606446e-01 1.55174688e-01 -2.13986114e-01 5.81366848e-03 8.28605413e-01 1.73620170e-03 -1.11344802e+00 4.44220871e-01 -4.21629161e-01 -5.32330334e-01 -6.31576478e-01 1.31990850e-01 -3.75871420e-01 -1.00229219e-01 6.90504432e-01 1.82400659e-01 -2.34666228e-01 -6.51438177e-01 7.91943789e-01 3.31582904e-01 6.95479393e-01 -7.22807646e-01 8.79520237e-01 9.20541167e-01 6.08840585e-02 -7.84674466e-01 -1.19307959e+00 -4.20555145e-01 -6.93862498e-01 8.68765190e-02 1.17871308e+00 -1.16057146e+00 -6.34748042e-01 3.93816173e-01 -1.23463655e+00 -6.43730760e-01 -6.17143810e-01 -6.59731776e-02 -9.28466916e-01 3.10419828e-01 -4.74925280e-01 -5.64815819e-01 1.04895122e-02 -1.04633927e+00 1.76771569e+00 1.67440213e-02 1.82479590e-01 -8.84578228e-01 -1.84815049e-01 5.48643112e-01 1.84501514e-01 3.81530136e-01 8.76893520e-01 -8.17088131e-03 -1.12664378e+00 2.25012064e-01 -7.89289534e-01 3.92574787e-01 -9.48680341e-02 -1.05866097e-01 -1.50696862e+00 1.50477037e-01 -1.54101461e-01 -5.72056055e-01 1.09806991e+00 3.24364930e-01 1.72031152e+00 1.57723371e-02 -2.02012748e-01 1.27235854e+00 1.61825991e+00 -6.04057789e-01 7.83397257e-01 3.96754831e-01 1.08694029e+00 7.99970865e-01 5.17916799e-01 3.15048993e-01 6.44432068e-01 8.28534186e-01 8.04150820e-01 -3.80966514e-01 -4.04495627e-01 -4.28643942e-01 1.15950935e-01 2.80815512e-01 -2.26916939e-01 5.99767417e-02 -8.98974359e-01 2.64167339e-01 -1.87216222e+00 -7.00927317e-01 -3.68482649e-01 2.09980536e+00 7.27362275e-01 -9.51222051e-03 -1.24128662e-01 9.48390514e-02 2.60413498e-01 5.04809976e-01 -6.26870751e-01 -2.53437576e-03 -4.82139677e-01 1.37654558e-01 7.33094871e-01 5.37085354e-01 -1.21954298e+00 9.93911564e-01 4.83069706e+00 7.97934413e-01 -1.02409494e+00 2.16209501e-01 1.16510582e+00 -1.76584776e-02 -5.65183222e-01 1.71712041e-01 -7.41362989e-01 2.77393878e-01 4.84892160e-01 5.22217393e-01 3.87563139e-01 1.04084241e+00 9.04082060e-02 -5.96190430e-02 -1.28424072e+00 1.29746878e+00 -2.31533751e-01 -1.52303600e+00 -7.80560225e-02 -2.62164958e-02 4.81634557e-01 4.70667958e-01 6.36891052e-02 1.64932519e-01 1.09086648e-01 -9.59315121e-01 7.48023033e-01 4.80139941e-01 1.05926669e+00 -3.36912245e-01 1.51890025e-01 2.04677001e-01 -1.10851705e+00 -3.63715999e-02 -6.27830863e-01 -6.94310889e-02 1.37752622e-01 8.85102630e-01 -5.66998683e-02 5.85227311e-01 1.02470982e+00 1.13988531e+00 -6.51272357e-01 4.69331920e-01 -3.28530401e-01 4.00867105e-01 -5.61490715e-01 5.76637208e-01 3.24921221e-01 -3.06570619e-01 3.24778944e-01 1.13904572e+00 -1.30075198e-02 2.64335554e-02 2.00858545e-02 1.14426565e+00 -2.10395798e-01 -3.18872593e-02 -6.19894266e-01 3.71011108e-01 6.15688227e-02 1.50662482e+00 -6.58560991e-01 -1.21255949e-01 -6.22748971e-01 1.16449177e+00 6.82763219e-01 4.06426281e-01 -8.60998809e-01 -5.73523045e-02 9.51782465e-01 4.06112552e-01 3.14367533e-01 -1.48675904e-01 -5.33943594e-01 -1.39808810e+00 3.33218217e-01 -4.99439418e-01 1.01771511e-01 -1.03989995e+00 -1.47645664e+00 3.57390791e-01 -1.64853111e-01 -1.18777561e+00 4.56930026e-02 -1.05152369e+00 -5.49189925e-01 7.46342480e-01 -2.14091849e+00 -1.42501271e+00 -6.51092887e-01 7.18239963e-01 5.65988302e-01 4.68995124e-01 8.04019630e-01 2.17619106e-01 -3.01312029e-01 2.31103525e-01 -1.15248308e-01 5.55877052e-02 5.33984363e-01 -1.50662684e+00 5.98843694e-01 5.95374405e-01 3.20276290e-01 3.72952223e-01 4.26990509e-01 -1.13810137e-01 -1.61563337e+00 -1.17434204e+00 6.32353067e-01 -7.26270556e-01 6.17126882e-01 -8.53482902e-01 -9.13911343e-01 5.42164385e-01 -1.78958952e-01 6.54673994e-01 3.30067903e-01 2.55327493e-01 -7.36515522e-01 -1.92355976e-01 -1.06110215e+00 6.00331903e-01 1.59720957e+00 -6.39355719e-01 -1.89412385e-01 7.48756289e-01 1.13017869e+00 -3.20653349e-01 -9.45970893e-01 5.50138175e-01 2.70139188e-01 -1.45583367e+00 1.39566267e+00 -4.26081479e-01 7.95658588e-01 -4.23984885e-01 -4.96613204e-01 -8.73629034e-01 -2.09213600e-01 -3.89647782e-01 -2.97018383e-02 1.15810537e+00 3.17610294e-01 -5.97364187e-01 9.34585273e-01 6.62123144e-01 -2.55722106e-01 -8.58078599e-01 -8.04858804e-01 -5.60990870e-01 -5.11916028e-03 -7.12009609e-01 5.11523962e-01 9.77684438e-01 -4.66392159e-01 2.58918047e-01 -1.37354881e-01 3.40203017e-01 8.20005238e-01 4.38876510e-01 7.97486782e-01 -1.14808917e+00 -6.24393225e-01 -4.27224070e-01 -5.13074577e-01 -1.51719332e+00 2.30658114e-01 -8.71481955e-01 -5.78537732e-02 -1.53230989e+00 2.41892710e-02 -8.13433588e-01 -1.26445398e-01 6.13892913e-01 -8.41085166e-02 4.80830163e-01 2.04832897e-01 2.80818880e-01 -7.20402896e-01 6.09768689e-01 1.00393689e+00 -8.06905627e-02 3.45065817e-02 -1.81210279e-01 -7.11587548e-01 8.83466959e-01 8.06360304e-01 -3.41775529e-02 -3.78616601e-01 -8.20469379e-01 3.52259725e-01 -1.66985735e-01 9.15563643e-01 -9.18568790e-01 3.37480344e-02 -9.61646140e-02 4.17748421e-01 -4.79881704e-01 6.21207237e-01 -7.33699024e-01 -3.19233388e-02 -1.05315425e-01 -2.75940031e-01 -3.68836105e-01 1.13175297e-02 5.29810727e-01 -2.19769880e-01 1.95131432e-02 6.53023660e-01 -2.75883704e-01 -1.00766659e+00 5.05257845e-01 5.74650168e-01 -6.18381053e-03 8.03297698e-01 -2.93359995e-01 -5.07774413e-01 -4.65051085e-03 -6.28640413e-01 1.55022159e-01 7.69876540e-01 4.39448714e-01 4.73533988e-01 -9.42013323e-01 -5.03657937e-01 1.84609219e-01 4.87132907e-01 6.80411994e-01 3.84342849e-01 7.95935512e-01 -4.02034938e-01 2.46035457e-01 -4.29942040e-03 -9.98647511e-01 -9.24183369e-01 5.86996436e-01 1.95462629e-01 -6.53925985e-02 -8.93482506e-01 9.28266823e-01 8.68122399e-01 -6.27005100e-01 6.08414672e-02 -5.41784406e-01 2.06265166e-01 -2.40845323e-01 3.75181079e-01 -1.94914103e-01 7.34763071e-02 -5.27811766e-01 -3.30624104e-01 9.55508530e-01 1.65391907e-01 3.25259507e-01 1.50198960e+00 -1.15697183e-01 -3.06964904e-01 4.60650265e-01 1.27074218e+00 -5.22269569e-02 -1.78528762e+00 -3.34388614e-01 -8.69767815e-02 -5.94542027e-01 -2.65158024e-02 -7.58775055e-01 -1.09447765e+00 1.06799257e+00 4.22237188e-01 2.54454792e-01 1.23801792e+00 3.56378645e-01 4.55180228e-01 4.13244039e-01 5.69575369e-01 -7.22753286e-01 -1.77328795e-01 3.50227833e-01 8.59763145e-01 -1.68128538e+00 -1.21510684e-01 -9.42559123e-01 -3.66846442e-01 9.31537151e-01 5.94972134e-01 -2.01033890e-01 5.01548767e-01 1.68948770e-01 -3.46763693e-02 -1.70624301e-01 -6.22117460e-01 -3.74649912e-01 2.14425862e-01 6.79799795e-01 4.44595307e-01 7.56495893e-02 2.44858339e-01 3.88915658e-01 -1.70533299e-01 -1.89736128e-01 8.99521932e-02 5.74505806e-01 -3.10070366e-01 -1.03984153e+00 -1.09252401e-01 3.74534488e-01 -7.35400245e-02 -3.06110620e-01 -1.61894083e-01 6.39135063e-01 2.22162530e-01 6.19803786e-01 1.77174732e-01 -1.65159181e-01 3.02767366e-01 -2.02084824e-01 3.61262113e-01 -6.78651571e-01 -3.09524000e-01 -3.31976116e-01 -1.00219510e-02 -1.23395038e+00 -6.07175648e-01 -5.13286948e-01 -1.21866548e+00 -1.63895898e-02 1.83742806e-01 -2.64633358e-01 7.04136848e-01 6.79751694e-01 4.20582980e-01 3.14521939e-01 4.96242076e-01 -1.09662807e+00 -2.84709752e-01 -6.30422950e-01 -3.74251813e-01 7.03183711e-01 3.67275894e-01 -5.26601911e-01 -3.66880089e-01 6.65174946e-02]
[8.459720611572266, -2.873986005783081]
ac6fe3db-f765-4db0-88ac-856ae92c15f0
optimising-game-tactics-for-football
2003.10294
null
https://arxiv.org/abs/2003.10294v1
https://arxiv.org/pdf/2003.10294v1.pdf
Optimising Game Tactics for Football
In this paper we present a novel approach to optimise tactical and strategic decision making in football (soccer). We model the game of football as a multi-stage game which is made up from a Bayesian game to model the pre-match decisions and a stochastic game to model the in-match state transitions and decisions. Using this formulation, we propose a method to predict the probability of game outcomes and the payoffs of team actions. Building upon this, we develop algorithms to optimise team formation and in-game tactics with different objectives. Empirical evaluation of our approach on real-world datasets from 760 matches shows that by using optimised tactics from our Bayesian and stochastic games, we can increase a team chances of winning by up to 16.1\% and 3.4\% respectively.
['Timothy J. Norman', 'Ryan Beal', 'Georgios Chalkiadakis', 'Sarvapali D. Ramchurn']
2020-03-23
null
null
null
null
['game-of-football']
['playing-games']
[-2.60102190e-02 1.76383391e-01 1.39984593e-01 -2.84299850e-01 -2.95977980e-01 -4.74705935e-01 5.28585911e-01 -2.02121790e-02 -7.87207842e-01 9.23389971e-01 1.93161964e-01 -3.19485366e-01 -8.54361176e-01 -1.05232298e+00 -2.93038011e-01 -3.85423034e-01 -3.30199301e-01 1.19886911e+00 6.61761224e-01 -8.58912408e-01 5.86264610e-01 1.86418533e-01 -1.56493378e+00 4.59819227e-01 3.30887705e-01 8.12474489e-01 1.35739639e-01 1.19580340e+00 2.72997878e-02 1.15707362e+00 -5.82356751e-01 -8.75253856e-01 7.53800273e-01 -3.81082982e-01 -8.22476029e-01 -6.98360102e-03 -7.00031817e-01 1.09191285e-02 -2.84955472e-01 5.68667650e-01 5.83794773e-01 5.85537612e-01 4.47456956e-01 -1.02625561e+00 6.82912290e-01 1.14373136e+00 -3.45401675e-01 5.48243523e-01 7.04170287e-01 3.08861315e-01 8.75421226e-01 -2.23743506e-02 3.55905592e-01 1.10762751e+00 6.37594819e-01 2.44954035e-01 -1.19503176e+00 -7.32031345e-01 3.30453403e-02 3.37619573e-01 -1.29986179e+00 -2.93462783e-01 3.13302398e-01 -5.44921637e-01 8.43627214e-01 1.68694347e-01 1.21462858e+00 6.19483232e-01 6.95673347e-01 6.17359042e-01 1.35576153e+00 -5.21184623e-01 3.31054032e-01 -4.41406339e-01 -1.04942068e-01 1.78385079e-01 1.00032296e-02 8.54662180e-01 -7.97109067e-01 -1.70614883e-01 6.97154820e-01 -3.44898283e-01 6.29539788e-01 -7.51342103e-02 -7.07701921e-01 9.16533709e-01 -1.70069546e-01 -1.43246755e-01 -8.08126390e-01 2.52352804e-01 3.80251229e-01 2.38090932e-01 -2.06077322e-02 6.19527280e-01 -7.22436085e-02 -7.70662844e-01 -9.17026520e-01 1.10131288e+00 1.06632626e+00 3.71813446e-01 4.75657433e-01 -2.20401436e-01 -2.93432027e-01 5.25997996e-01 3.28482449e-01 3.62029076e-02 1.69505179e-02 -1.17697525e+00 3.35342914e-01 3.71216238e-01 3.10116649e-01 -1.00712323e+00 -5.09338796e-01 -4.28858638e-01 -1.74799502e-01 3.60856265e-01 5.25450945e-01 -2.72537678e-01 -7.88882732e-01 1.33973289e+00 1.32810883e-02 3.91200632e-01 -1.65240187e-03 5.87392569e-01 2.56707311e-01 5.38816571e-01 1.02161467e-01 -1.86737612e-01 1.14062262e+00 -2.94023514e-01 -6.34818912e-01 -3.22317839e-01 2.02109382e-01 -8.20092559e-01 2.28136212e-01 9.62839663e-01 -1.53650153e+00 -5.50726235e-01 -6.93072438e-01 9.89354908e-01 3.44473392e-01 -6.16001129e-01 7.53390133e-01 9.36005592e-01 -8.91186714e-01 5.85559785e-01 -8.31770718e-01 9.11808908e-02 -1.49464458e-01 7.50085652e-01 5.42613156e-02 4.12766561e-02 -1.33310008e+00 1.06018174e+00 9.86602604e-01 1.08524151e-01 -1.25961161e+00 -1.16547622e-01 -5.55636227e-01 -1.25942871e-01 1.02587450e+00 -4.45090979e-01 1.37562895e+00 -2.83925295e-01 -1.81844115e+00 8.12009692e-01 4.79303777e-01 -7.18794823e-01 4.79305208e-01 -2.76669990e-02 1.99681278e-02 -3.55047017e-01 2.23235711e-01 1.07620947e-01 2.36970469e-01 -1.03952038e+00 -1.25404441e+00 -9.53884721e-02 6.03535414e-01 6.92761719e-01 5.12466729e-01 5.76989889e-01 -3.35097104e-01 -3.59796047e-01 -1.98689494e-02 -1.07228172e+00 -1.05132663e+00 -1.21375012e+00 -9.63184014e-02 3.30706686e-02 -4.70266044e-01 -4.52802569e-01 1.55901539e+00 -1.74322474e+00 4.95253652e-01 7.15625584e-01 1.29965425e-01 -9.16150510e-02 3.11642557e-01 6.95150554e-01 1.58039600e-01 -2.37531304e-01 1.96250379e-01 8.44126381e-03 1.00708820e-01 7.55099297e-01 -4.39899154e-02 6.60573244e-02 -3.53803217e-01 6.06817603e-01 -7.81763971e-01 -3.30770195e-01 3.37190539e-01 -4.22982782e-01 -7.99052119e-01 2.99095213e-01 1.17754281e-01 1.16230577e-01 -4.63480592e-01 2.55220711e-01 2.69752830e-01 7.28525877e-01 7.92561173e-01 7.79523075e-01 -3.25762928e-01 4.33827817e-01 -1.85158706e+00 1.25531745e+00 -1.67788833e-01 1.09029435e-01 9.53172073e-02 -1.07981658e+00 9.56662178e-01 2.68550426e-01 5.23910820e-01 -4.70260680e-01 6.19814575e-01 -6.32781759e-02 4.72506434e-01 -3.89842987e-02 8.20517778e-01 -7.91273355e-01 -8.39085698e-01 7.00089157e-01 1.58043399e-01 -4.01173830e-01 5.18632710e-01 5.07926494e-02 1.10321343e+00 1.70499206e-01 3.53123903e-01 -2.20418781e-01 1.71271026e-01 2.53072560e-01 8.80659819e-01 1.37603080e+00 -1.69397265e-01 4.11992893e-02 7.78078854e-01 -9.04995739e-01 -6.22338951e-01 -1.01669109e+00 4.28895473e-01 1.33011079e+00 1.38548568e-01 -7.08928823e-01 -4.87101048e-01 -1.27283230e-01 -1.48010999e-01 7.26139307e-01 -5.62599301e-01 -2.74814337e-01 -6.24824286e-01 -9.73871887e-01 3.88480902e-01 3.40377897e-01 1.73992202e-01 -9.95918512e-01 -7.92207658e-01 8.42646718e-01 -3.49724919e-01 -9.63373661e-01 -1.58916786e-02 3.85003656e-01 -5.59650242e-01 -1.14446425e+00 9.43497047e-02 -9.49095190e-02 -1.11398824e-01 -2.16600031e-01 1.27310610e+00 3.34729627e-02 -1.13905564e-01 2.01850057e-01 -6.44256115e-01 -8.08178127e-01 -6.39244854e-01 -7.47460499e-02 4.53432679e-01 -8.94637331e-02 3.25882494e-01 -4.81851786e-01 -3.09998065e-01 5.91084838e-01 -5.68769217e-01 2.03135133e-01 4.43020999e-01 6.88388765e-01 1.47412941e-01 7.13096857e-01 -8.99388865e-02 -5.55773020e-01 9.93580401e-01 -4.60340589e-01 -6.97003901e-01 -7.42754489e-02 -3.67090225e-01 2.24203654e-02 -6.67228550e-02 -3.89791697e-01 -1.10853934e+00 1.50772512e-01 -8.10876638e-02 6.53550476e-02 -1.25733182e-01 6.49401128e-01 2.06354827e-01 -5.82480356e-02 7.71996617e-01 7.20413327e-02 -1.21942582e-03 -2.02872440e-01 8.15240592e-02 4.57719237e-01 4.24841553e-01 -1.18166876e+00 6.15165472e-01 1.06318653e-01 7.86869824e-02 -1.66258857e-01 -5.02965450e-01 -5.11345685e-01 -5.19930720e-01 -1.12296450e+00 7.25138426e-01 -7.87390769e-01 -1.33724749e+00 4.83725995e-01 -4.96550918e-01 -5.32014251e-01 -3.48415345e-01 7.09657073e-01 -1.06152272e+00 9.40744653e-02 -5.54243863e-01 -1.37478638e+00 4.02092487e-01 -1.12926233e+00 4.53031331e-01 3.95827562e-01 -4.34442043e-01 -7.34783769e-01 4.03649211e-01 6.93338275e-01 -1.24092260e-02 3.34556788e-01 1.51754364e-01 -5.87027490e-01 -4.31126297e-01 -2.81879842e-01 6.53636932e-01 -2.06237640e-02 -3.14305782e-01 -1.62933975e-01 -4.01113555e-02 -5.76344915e-02 -8.70100558e-02 -2.09541321e-01 3.72043580e-01 6.03743017e-01 2.99940407e-01 2.44362518e-01 -1.87532961e-01 3.52498293e-01 1.17753804e+00 6.42610550e-01 8.02931190e-01 8.38516116e-01 2.18918249e-02 9.61473107e-01 1.39629531e+00 9.93712068e-01 5.62263370e-01 1.18638587e+00 4.34646040e-01 5.53152204e-01 5.20121992e-01 -3.31919909e-01 4.45459485e-01 4.16502476e-01 -1.05228329e+00 -8.99673030e-02 -1.28131986e+00 6.40909016e-01 -2.13287640e+00 -1.32036316e+00 -2.28748992e-01 2.14152646e+00 7.62370169e-01 8.28577518e-01 7.83313930e-01 4.44265276e-01 6.96555018e-01 -1.64242923e-01 3.45890671e-01 -1.08614242e+00 2.11331218e-01 7.84205556e-01 8.37380528e-01 6.75360501e-01 -8.37492585e-01 1.30227232e+00 7.41237736e+00 1.07347524e+00 -2.15766266e-01 -7.04561546e-02 3.56378347e-01 -6.29427254e-01 1.44165829e-01 5.32409132e-01 -9.30337787e-01 4.38441694e-01 1.03854597e+00 -3.53015840e-01 5.36975026e-01 3.77341658e-01 3.45395297e-01 -6.71234012e-01 -4.82587397e-01 6.78603947e-01 -3.08942050e-01 -1.27405882e+00 -3.54895353e-01 4.11557138e-01 6.18594766e-01 -4.53251153e-01 -2.84851730e-01 6.76415086e-01 1.53619683e+00 -1.06066084e+00 1.16763878e+00 7.78967679e-01 8.26377198e-02 -1.46185410e+00 1.05753076e+00 7.61342108e-01 -1.06299067e+00 -4.97474074e-01 -1.29277185e-01 -1.21710551e+00 6.64349854e-01 7.40368590e-02 -8.85266125e-01 8.80637348e-01 8.40325534e-01 9.52172577e-02 1.21974736e-01 1.21135747e+00 -4.08754945e-01 6.63232863e-01 -6.24492109e-01 -8.60422850e-03 4.92917806e-01 -2.86571860e-01 6.63162589e-01 7.26275563e-01 9.70968530e-02 7.62629211e-01 6.30368710e-01 1.88836709e-01 8.40855062e-01 -4.86969352e-02 -2.07929835e-01 4.17408794e-01 3.92476171e-01 6.47650301e-01 -8.24078918e-01 -1.39694137e-03 3.16163749e-01 4.85017985e-01 -6.14095517e-02 -2.38348246e-01 -6.54721618e-01 -3.43993567e-02 8.56798291e-01 2.74430245e-01 2.05287933e-01 -3.17813575e-01 -2.55583048e-01 -7.07139611e-01 -3.11676592e-01 -1.19427121e+00 6.44298851e-01 -4.49528933e-01 -7.39838123e-01 4.75471854e-01 6.42331183e-01 -8.88579965e-01 -7.36931384e-01 -5.04745007e-01 -6.68371201e-01 9.01147842e-01 -7.24774599e-01 -6.29533589e-01 2.68584430e-01 3.29113394e-01 3.50649297e-01 -4.45541322e-01 3.83657604e-01 -1.24276385e-01 -3.61261219e-01 1.68097451e-01 -3.03826362e-01 -1.81399509e-01 1.39878273e-01 -1.13372993e+00 3.93492728e-01 7.52738953e-01 -3.79749713e-03 2.05829322e-01 1.20276535e+00 -7.88528562e-01 -8.21202397e-01 -1.94552407e-01 7.33028054e-01 -6.24012828e-01 6.85208738e-01 -1.26786172e-01 -5.13415597e-02 5.42496264e-01 -1.40806973e-01 -9.20822501e-01 9.19779360e-01 4.94171709e-01 6.09723926e-01 9.45096090e-02 -8.18596303e-01 7.25235820e-01 1.05498815e+00 -6.49682134e-02 -1.17456758e+00 -4.20775041e-02 -2.10608423e-01 -6.92859888e-01 -7.90745199e-01 3.90676051e-01 6.88210547e-01 -1.38105249e+00 1.04942274e+00 -8.05060565e-01 1.50009245e-01 -1.39007956e-01 -4.59801033e-02 -1.29730988e+00 -4.80057031e-01 -9.63581264e-01 5.14298320e-01 6.86819851e-01 3.77517473e-03 -2.21837357e-01 1.05722129e+00 7.13374674e-01 1.60019528e-02 -5.69495201e-01 -1.25472176e+00 -6.06353581e-01 -1.22493058e-01 -1.06052864e+00 4.90755856e-01 4.05503064e-01 6.07235581e-02 -1.59060836e-01 -8.90862823e-01 3.88255925e-03 8.14037621e-01 -2.02778414e-01 9.71517324e-01 -1.30943406e+00 -8.15770745e-01 -5.47710836e-01 -9.75510299e-01 -7.96891570e-01 -1.42524615e-02 -5.05956590e-01 2.85602957e-01 -1.19467938e+00 2.95406044e-01 -5.81933975e-01 -3.71291280e-01 1.42485470e-01 -2.08096772e-01 1.95910320e-01 4.39952880e-01 -8.43168572e-02 -6.23169601e-01 1.09361328e-01 7.00530052e-01 5.28983951e-01 -3.34514886e-01 5.83915651e-01 -7.10084558e-01 7.27116287e-01 7.44850159e-01 -8.43080103e-01 -1.26329854e-01 8.49055201e-02 7.40952671e-01 6.65713251e-01 7.83245191e-02 -9.84123111e-01 2.97709793e-01 -6.96631432e-01 -2.62260973e-01 -4.02565420e-01 4.39233035e-01 -3.37491810e-01 7.01715529e-01 8.29752207e-01 -1.61635086e-01 3.68097052e-03 2.34793440e-01 6.66743159e-01 -2.33295187e-01 -4.97609556e-01 2.14984775e-01 -3.11534226e-01 -7.85676718e-01 -5.50378151e-02 -1.24646604e+00 -1.44170225e-01 1.50563931e+00 -7.35598147e-01 5.40152431e-01 -8.19949329e-01 -1.36689115e+00 6.21440709e-01 2.19463691e-01 1.17818460e-01 3.59817415e-01 -9.75336254e-01 -1.09461355e+00 -8.22064579e-02 -3.05787325e-01 -4.16751832e-01 4.35695589e-01 5.55239201e-01 -8.48711133e-01 1.23231716e-01 -7.94378638e-01 -2.66922921e-01 -1.36239398e+00 2.61795800e-02 4.85993713e-01 -9.80822027e-01 -2.86543779e-02 1.08336163e+00 -2.72765875e-01 -4.60618645e-01 -6.42446205e-02 2.09574893e-01 -5.18514931e-01 -9.92776453e-03 3.28690082e-01 4.52537686e-01 -3.32204074e-01 -7.01457322e-01 -3.86766732e-01 3.67172301e-01 1.13779984e-01 -7.85770118e-01 1.24444807e+00 -7.32078403e-02 9.48062763e-02 3.38602930e-01 -2.95317590e-01 -5.33736832e-02 -1.23672748e+00 -1.54473647e-01 2.74539739e-01 -7.68190563e-01 1.08683750e-01 -6.14547729e-01 -6.14037514e-01 2.36573070e-01 7.48432428e-02 3.55457038e-01 1.07952225e+00 -1.94275916e-01 3.21812570e-01 1.11098826e-01 1.12238634e+00 -1.24329376e+00 -1.38977870e-01 7.64744878e-01 4.03726786e-01 -6.92363679e-01 -3.14639905e-03 -3.76494765e-01 -1.13043725e+00 9.39679146e-01 2.96542346e-01 -3.96017909e-01 5.04375637e-01 4.41675782e-01 -7.54281431e-02 -2.91266471e-01 -1.19703722e+00 -5.33767700e-01 -1.62807360e-01 6.31540895e-01 -2.32148439e-01 4.92345899e-01 -6.73441947e-01 1.02410078e+00 -6.59891784e-01 1.44365832e-01 8.99824023e-01 1.21772623e+00 -5.51249385e-01 -1.62482369e+00 -7.48780310e-01 4.80324656e-01 -7.75571465e-01 1.48371682e-01 -3.28212589e-01 6.69933140e-01 4.37887698e-01 1.19736624e+00 5.43574691e-02 -9.42309201e-01 7.58821487e-01 -8.39744285e-02 5.91654241e-01 -8.53582203e-01 -1.23086667e+00 -5.19547909e-02 6.29839718e-01 -6.47509933e-01 -3.38318527e-01 -1.01035857e+00 -7.24808455e-01 -8.19216073e-01 -1.89132586e-01 5.72186470e-01 3.51217061e-01 1.08979213e+00 -2.35530883e-01 5.93958139e-01 5.35304844e-01 -9.83219028e-01 -6.18170142e-01 -6.77842319e-01 -9.12402153e-01 -6.29815385e-02 -6.69349074e-01 -1.18414021e+00 -6.41103387e-02 -3.39288294e-01]
[3.4814980030059814, 1.4902387857437134]
8f16e554-4a5a-458d-8569-727e8f3eece9
sample-level-cnn-architectures-for-music-auto
1710.10451
null
http://arxiv.org/abs/1710.10451v2
http://arxiv.org/pdf/1710.10451v2.pdf
Sample-level CNN Architectures for Music Auto-tagging Using Raw Waveforms
Recent work has shown that the end-to-end approach using convolutional neural network (CNN) is effective in various types of machine learning tasks. For audio signals, the approach takes raw waveforms as input using an 1-D convolution layer. In this paper, we improve the 1-D CNN architecture for music auto-tagging by adopting building blocks from state-of-the-art image classification models, ResNets and SENets, and adding multi-level feature aggregation to it. We compare different combinations of the modules in building CNN architectures. The results show that they achieve significant improvements over previous state-of-the-art models on the MagnaTagATune dataset and comparable results on Million Song Dataset. Furthermore, we analyze and visualize our model to show how the 1-D CNN operates.
['Jongpil Lee', 'Taejun Kim', 'Juhan Nam']
2017-10-28
null
null
null
null
['music-auto-tagging']
['music']
[-1.60458982e-01 -4.32575792e-02 3.35305184e-02 -2.10222766e-01 -3.53019655e-01 -7.20814824e-01 3.31652522e-01 -2.20095202e-01 -3.98991764e-01 1.80530638e-01 3.86634976e-01 1.82789881e-02 -5.47474623e-02 -8.44486952e-01 -8.39417517e-01 -1.96069390e-01 -4.51748878e-01 1.42182693e-01 2.31421530e-01 -3.81403983e-01 -4.98938374e-03 5.47977090e-01 -1.50204360e+00 8.84403348e-01 -2.16745853e-01 1.41890180e+00 -3.11401218e-01 1.02481675e+00 -2.76340127e-01 9.04130518e-01 -9.91314173e-01 -4.27643806e-01 3.56031865e-01 -1.66350871e-01 -9.01927829e-01 -4.49468881e-01 7.94941306e-01 -1.57858104e-01 -6.75485015e-01 8.71072352e-01 1.07192814e+00 5.85393570e-02 2.62023807e-01 -1.03162313e+00 -7.82413244e-01 1.39446950e+00 -1.61452547e-01 4.37207818e-01 5.61489053e-02 -1.48767874e-01 1.12779105e+00 -7.55744815e-01 3.48216265e-01 1.13220191e+00 1.38687515e+00 4.55150694e-01 -1.11594510e+00 -1.11902106e+00 -2.99775660e-01 3.04920703e-01 -1.34762871e+00 -3.38887334e-01 8.98156881e-01 -4.54289854e-01 1.34446812e+00 4.72802930e-02 8.63167286e-01 1.00018883e+00 -2.32498139e-01 9.09414411e-01 7.18600333e-01 -4.36244547e-01 -2.97478470e-03 -3.01335275e-01 1.37351498e-01 3.96487564e-01 -2.66522080e-01 1.66005462e-01 -9.00576353e-01 8.37788433e-02 1.00998378e+00 -2.06890360e-01 1.45247728e-01 1.52397845e-02 -1.15551102e+00 4.91909921e-01 6.02707386e-01 7.76226997e-01 -1.32877186e-01 6.65163934e-01 7.83731461e-01 4.41130459e-01 4.05020267e-01 6.95371628e-01 -5.88744938e-01 -4.56166387e-01 -1.45302248e+00 3.02971601e-01 6.73302412e-01 7.75559962e-01 3.68567795e-01 7.62971938e-01 -4.11035597e-01 1.07705915e+00 -2.60272205e-01 -6.69943169e-02 6.03493273e-01 -9.02339995e-01 2.40197271e-01 4.03339535e-01 -4.20196265e-01 -8.05937767e-01 -6.01964295e-01 -1.15979099e+00 -8.95113349e-01 3.58397722e-01 1.59911633e-01 6.62302077e-02 -1.07776749e+00 1.66382742e+00 -3.89078647e-01 7.90237308e-01 8.98629054e-03 9.19271946e-01 1.38372636e+00 2.86635637e-01 -3.37595330e-03 8.15518618e-01 1.38526583e+00 -1.13761210e+00 -6.65298462e-01 1.91698417e-01 1.32221982e-01 -9.49497461e-01 1.01157963e+00 6.59696102e-01 -1.32745826e+00 -1.43155849e+00 -1.40365565e+00 -1.34655580e-01 -6.30038202e-01 6.15752876e-01 6.65255249e-01 7.80388892e-01 -1.18392694e+00 1.22515321e+00 -5.35122514e-01 -1.17710643e-01 6.69091582e-01 5.45274317e-01 -2.55125791e-01 7.28640854e-01 -1.22762740e+00 3.84377450e-01 7.82295525e-01 -1.36296883e-01 -1.28934848e+00 -8.68751645e-01 -5.56631088e-01 4.63335216e-01 -1.01091444e-01 -5.36587000e-01 1.56275916e+00 -9.04829800e-01 -1.64409125e+00 6.68777466e-01 6.15200102e-01 -1.09067976e+00 -4.05013561e-03 -4.69233483e-01 -6.57528281e-01 1.09959424e-01 -3.75003546e-01 8.99537563e-01 8.44692826e-01 -7.43003130e-01 -6.46703720e-01 1.59332946e-01 4.14054394e-01 -3.23747456e-01 -5.15994310e-01 6.35639357e-04 -3.82088661e-01 -1.25855947e+00 -3.06114256e-01 -7.84765124e-01 1.03851765e-01 -1.99412405e-01 -3.12845290e-01 -3.82123403e-02 8.28192055e-01 -4.68040884e-01 1.40339839e+00 -2.40803313e+00 -1.03842758e-01 1.11431003e-01 9.56654102e-02 4.15717512e-01 -3.52359742e-01 1.33178234e-01 -5.85642695e-01 3.08344275e-01 5.74059598e-02 -4.80878711e-01 2.09071189e-01 -3.23737338e-02 -3.48794967e-01 -8.04623589e-02 1.53966814e-01 9.81828272e-01 -5.82855999e-01 -8.66920501e-02 2.70818323e-01 5.54944813e-01 -6.57329738e-01 1.65132552e-01 -1.07726343e-01 9.20733735e-02 1.57819390e-01 6.08110309e-01 5.62345266e-01 -9.79697052e-03 4.20760131e-03 -5.87989509e-01 -2.55999506e-01 6.09792352e-01 -1.27245247e+00 2.34295774e+00 -5.97932518e-01 1.04714489e+00 -1.42530292e-01 -9.76888418e-01 9.48706925e-01 5.97488225e-01 5.32435656e-01 -5.11370182e-01 3.23680609e-01 1.36158586e-01 1.84836894e-01 -2.23317787e-01 4.52989519e-01 4.77395467e-02 -2.76504785e-01 3.52698803e-01 9.96674716e-01 6.18820405e-03 9.95953679e-02 -3.27670649e-02 1.05239844e+00 2.67217457e-01 5.61881671e-03 -5.80689237e-02 3.08026403e-01 -1.24001138e-01 3.82189274e-01 8.08128536e-01 2.24011973e-01 9.55344677e-01 2.24254891e-01 -7.76001513e-01 -1.15599608e+00 -9.04887557e-01 -4.50104848e-02 1.07982707e+00 -5.13957381e-01 -8.14303458e-01 -8.25921714e-01 -3.69488180e-01 4.73238453e-02 3.14728618e-01 -7.85901904e-01 -7.34881908e-02 -6.06452763e-01 -5.45057118e-01 1.60745573e+00 9.30323362e-01 9.04996336e-01 -1.17666936e+00 -6.49980426e-01 5.40960312e-01 2.59228691e-04 -1.19103932e+00 -1.88933358e-01 5.16210437e-01 -8.18977475e-01 -8.08949053e-01 -6.27143741e-01 -7.39594758e-01 -2.36085191e-01 -1.06612809e-01 1.63389742e+00 -1.69946030e-01 -3.30069542e-01 1.59671053e-01 -5.17536521e-01 -8.52043271e-01 -1.43735990e-01 5.14002323e-01 3.56870629e-02 -1.38108999e-01 3.31078887e-01 -1.09768164e+00 -5.70632398e-01 -4.05322723e-02 -8.59141767e-01 -1.61441386e-01 4.39239591e-01 5.52392185e-01 4.69414800e-01 1.04417145e-01 5.04955351e-01 -5.23254514e-01 5.93500376e-01 -1.10882737e-01 -1.82858214e-01 -2.15151563e-01 -2.06269681e-01 -8.72738063e-02 5.45828998e-01 -6.94161475e-01 -3.24954778e-01 1.78682640e-01 -6.23000979e-01 -1.00691831e+00 -3.04295599e-01 2.35123500e-01 2.11010024e-01 -1.21517144e-01 6.69537306e-01 -8.38969871e-02 -3.17042679e-01 -9.68606830e-01 4.58136380e-01 8.19211781e-01 9.87747014e-01 -3.74913335e-01 7.26753652e-01 4.39065993e-01 5.06983511e-02 -5.05855143e-01 -1.04917490e+00 -3.93838942e-01 -7.83392727e-01 -4.30148095e-01 9.39384937e-01 -1.19148803e+00 -9.64220941e-01 4.48647201e-01 -1.06829309e+00 -3.03795338e-01 -7.58193791e-01 5.52577078e-01 -6.05206132e-01 -1.49855480e-01 -7.43107736e-01 -5.26859999e-01 -3.97114754e-01 -7.45149910e-01 1.01081038e+00 2.64365256e-01 -3.55710596e-01 -7.09986567e-01 3.34540218e-01 -6.90779239e-02 7.06085205e-01 2.34691054e-01 8.03774595e-01 -8.71409297e-01 -3.34199309e-01 -7.24711120e-02 1.52976941e-02 7.55940497e-01 -3.15496147e-01 -1.69266179e-01 -1.66020858e+00 5.39260246e-02 -5.00252366e-01 -3.05322826e-01 1.23378932e+00 4.45278406e-01 1.65118432e+00 -5.16043417e-02 2.10175768e-01 8.77439737e-01 1.18857300e+00 9.42670405e-02 8.77882004e-01 4.26173031e-01 4.48666513e-01 -8.58891979e-02 1.04085296e-01 3.53308529e-01 5.28601073e-02 7.85935879e-01 6.08789444e-01 -4.55815077e-01 -8.59982312e-01 -3.23432773e-01 2.06091538e-01 9.45424318e-01 -2.60075122e-01 -1.11695893e-01 -6.77821934e-01 6.99589014e-01 -1.74016333e+00 -1.16883874e+00 4.66268100e-02 1.72264838e+00 7.87272990e-01 2.79814124e-01 4.85306412e-01 6.73621237e-01 4.41371411e-01 1.97667941e-01 -6.84759319e-02 -5.39957047e-01 -3.13832849e-01 1.14682102e+00 3.15878630e-01 -1.83440655e-01 -1.45553970e+00 9.39194083e-01 7.44954443e+00 1.05225277e+00 -1.40307510e+00 3.89389932e-01 1.62341058e-01 -3.55188429e-01 3.44574720e-01 -2.73054034e-01 -5.03213584e-01 8.45431909e-02 1.10923982e+00 4.28391337e-01 4.83419478e-01 8.26151550e-01 -1.73571840e-01 5.27651548e-01 -1.13563728e+00 1.16766834e+00 4.44317497e-02 -1.77115965e+00 5.42778447e-02 -3.52656484e-01 5.66671252e-01 6.00635558e-02 1.03524633e-01 5.66956341e-01 1.70462504e-01 -1.09957993e+00 1.15264857e+00 6.41200900e-01 8.71004641e-01 -9.78865325e-01 9.41625595e-01 -2.82152385e-01 -1.68008387e+00 -2.56160051e-01 -3.08056623e-01 -2.47262031e-01 -1.01108095e-02 5.19948065e-01 -8.60708237e-01 6.34518921e-01 1.38296211e+00 8.30831647e-01 -6.86501980e-01 1.23741043e+00 -9.24453419e-03 9.08400357e-01 -1.89760178e-01 2.25780532e-01 3.80796254e-01 4.97277021e-01 4.35656875e-01 1.62585974e+00 5.15608430e-01 -3.33813429e-01 -1.18633151e-01 9.24389660e-01 -4.30861473e-01 -1.60410017e-01 -2.90599495e-01 -2.03465715e-01 2.12313846e-01 1.33378232e+00 -6.07336462e-01 -4.13604259e-01 -2.22888544e-01 6.06440723e-01 1.26048654e-01 -5.76199405e-03 -7.83787549e-01 -8.97513032e-01 8.13750684e-01 -3.25170755e-02 6.64634705e-01 -1.97086409e-01 -5.02272129e-01 -7.27625608e-01 -3.65466207e-01 -7.07432985e-01 3.68831456e-01 -1.00528860e+00 -1.12480450e+00 7.17607558e-01 -3.28691930e-01 -1.66948223e+00 -1.11796416e-01 -7.27418423e-01 -7.54419982e-01 4.83705878e-01 -1.48071706e+00 -1.29796565e+00 -1.99921176e-01 5.99996805e-01 4.68145669e-01 -6.87093973e-01 1.15887833e+00 8.88761163e-01 -1.25784874e-01 7.24590063e-01 -3.50336760e-01 8.02266836e-01 6.87923312e-01 -1.43184376e+00 7.42664218e-01 4.96460110e-01 8.78374636e-01 4.39108670e-01 2.54349113e-01 -2.04381809e-01 -8.69798422e-01 -1.18369794e+00 5.67469954e-01 -1.01902209e-01 6.71468198e-01 -5.39072037e-01 -5.48939764e-01 5.75954556e-01 7.19088674e-01 -3.09802331e-02 1.01174831e+00 4.02592540e-01 -7.75739491e-01 -3.06408376e-01 -7.77743816e-01 2.74926186e-01 1.21274495e+00 -7.54450977e-01 -6.12615347e-01 -1.74852312e-01 6.83970213e-01 -4.95290965e-01 -1.15646982e+00 5.97796023e-01 8.75246763e-01 -9.47137713e-01 1.33882320e+00 -8.57542813e-01 4.98078614e-01 -4.44659472e-01 -1.55137122e-01 -1.34246993e+00 -5.20522714e-01 -7.08756626e-01 -3.26064557e-01 1.26449025e+00 4.94534999e-01 2.38583297e-01 6.22341573e-01 -4.69227552e-01 -6.16326869e-01 -4.28176135e-01 -9.13915455e-01 -8.99593949e-01 -8.52431878e-02 -9.91878450e-01 9.58996832e-01 8.10957432e-01 -1.47005498e-01 3.50420296e-01 -5.76808333e-01 -4.42433916e-02 2.95770913e-01 1.32359847e-01 9.80173111e-01 -1.46867394e+00 -4.93160099e-01 -6.00901008e-01 -8.58973920e-01 -8.34281385e-01 8.80657211e-02 -1.13991427e+00 -3.47413033e-01 -1.25216603e+00 -8.70516598e-02 -1.48298383e-01 -8.63734663e-01 9.19321418e-01 5.85785568e-01 1.05403650e+00 5.70977092e-01 -1.33401379e-01 -4.94923770e-01 1.95653468e-01 9.20052826e-01 -4.59525794e-01 -4.31951582e-02 -7.72652701e-02 -5.25376201e-01 7.96334147e-01 9.72072005e-01 -6.11705959e-01 2.44572293e-02 -5.70853174e-01 4.22392100e-01 -3.13946247e-01 6.55166984e-01 -1.70513308e+00 2.29288593e-01 8.04491282e-01 7.61635482e-01 -9.86991763e-01 7.19609320e-01 -7.62828469e-01 2.58295596e-01 2.68508255e-01 -5.41629672e-01 6.87800813e-04 6.62961066e-01 3.21420617e-02 -6.83631122e-01 -1.00205995e-01 7.68016517e-01 -2.38607451e-01 -6.51496053e-01 6.69915155e-02 -2.31180727e-01 -1.32385388e-01 3.76258522e-01 -9.63711441e-02 -1.61286205e-01 -4.09311175e-01 -1.33424854e+00 -3.62452507e-01 -2.72935420e-01 7.18750536e-01 3.64278764e-01 -1.80570304e+00 -6.05253935e-01 1.31362453e-01 -1.01118535e-01 -5.11975229e-01 3.02014977e-01 4.28969204e-01 -3.80830526e-01 4.76688892e-01 -7.61637211e-01 -6.10620260e-01 -1.32303965e+00 2.17515141e-01 4.89632964e-01 -1.39552057e-01 -3.84147435e-01 9.45375025e-01 -4.75027472e-01 -4.91079301e-01 6.68884933e-01 -6.44573092e-01 -4.21074182e-01 1.80788487e-01 4.57686067e-01 3.08779031e-01 3.73803824e-01 -3.00667137e-01 -1.70161426e-01 7.43458688e-01 2.84934759e-01 -2.17746496e-01 1.74722028e+00 4.99638975e-01 1.46054909e-01 5.56544125e-01 1.06991172e+00 -1.86636616e-02 -8.28708887e-01 -6.41609505e-02 -1.63178295e-01 -3.82329673e-02 2.86359370e-01 -9.25638258e-01 -1.57521451e+00 1.32836401e+00 9.26153183e-01 3.85254234e-01 1.34540474e+00 -1.90470573e-02 8.32180798e-01 4.31345433e-01 7.04659745e-02 -1.22492874e+00 2.83278167e-01 7.68760979e-01 7.71592438e-01 -4.24653947e-01 -1.47148654e-01 2.72120554e-02 -2.08891243e-01 1.53791487e+00 2.97783524e-01 -4.67193872e-01 1.03446603e+00 8.71144056e-01 7.77369365e-02 -2.67413676e-01 -4.69669163e-01 -5.99678636e-01 5.03805459e-01 4.63688642e-01 6.22074962e-01 -3.16120386e-02 1.52865812e-01 1.23521936e+00 -8.53115618e-01 2.25369722e-01 3.11577797e-01 6.72751606e-01 -1.04943886e-01 -1.17893827e+00 -2.21551821e-01 1.43679589e-01 -1.15750051e+00 -2.10656434e-01 -6.55457079e-01 8.97884905e-01 7.25382626e-01 6.77285373e-01 4.38696027e-01 -1.01246321e+00 6.39636874e-01 2.36053228e-01 4.68187064e-01 -5.19337237e-01 -1.53262401e+00 3.17305446e-01 1.82107553e-01 -4.32181925e-01 -8.84835720e-01 -2.91147500e-01 -1.12614799e+00 4.78615910e-02 -2.08412126e-01 -3.43498513e-02 7.61267126e-01 4.27622527e-01 5.94657540e-01 1.30278528e+00 4.40574378e-01 -1.11953676e+00 -1.05468586e-01 -1.30242884e+00 -5.62393546e-01 2.03240618e-01 3.35089117e-01 -4.80445296e-01 1.03058957e-01 1.73658788e-01]
[15.711594581604004, 5.220409393310547]
3355c682-d9c0-4218-83f7-ad9bbbb0e1cd
an-iterative-step-function-estimator-for
1412.2129
null
http://arxiv.org/abs/1412.2129v2
http://arxiv.org/pdf/1412.2129v2.pdf
An iterative step-function estimator for graphons
Exchangeable graphs arise via a sampling procedure from measurable functions known as graphons. A natural estimation problem is how well we can recover a graphon given a single graph sampled from it. One general framework for estimating a graphon uses step-functions obtained by partitioning the nodes of the graph according to some clustering algorithm. We propose an iterative step-function estimator (ISFE) that, given an initial partition, iteratively clusters nodes based on their edge densities with respect to the previous iteration's partition. We analyze ISFE and demonstrate its performance in comparison with other graphon estimation techniques.
['Nathanael Ackerman', 'Diana Cai', 'Cameron Freer']
2014-12-05
null
null
null
null
['graphon-estimation']
['graphs']
[ 1.04000665e-01 5.79454958e-01 -3.58961552e-01 -1.70582741e-01 -6.44059122e-01 -7.05715597e-01 6.33183062e-01 2.51680374e-01 8.65057763e-03 1.03715956e+00 -7.19465241e-02 4.17894498e-02 -4.14805919e-01 -1.21751571e+00 -8.64666581e-01 -6.14374936e-01 -4.03700829e-01 1.13737667e+00 3.46755922e-01 2.83929616e-01 -1.36561900e-01 5.43088913e-01 -9.27693784e-01 -3.38337898e-01 9.26953614e-01 4.89464551e-01 -3.11420299e-02 1.09626281e+00 4.05514911e-02 8.81939411e-01 -4.89059627e-01 -3.51270527e-01 1.40740365e-01 -9.34373379e-01 -1.07718968e+00 5.91775596e-01 4.03877981e-02 2.00821251e-01 -5.33404768e-01 1.54023600e+00 -1.96935669e-01 1.04658119e-01 1.15489531e+00 -1.49964535e+00 -2.32362732e-01 8.51636291e-01 -4.06738639e-01 -9.52470899e-02 3.54487360e-01 -4.29191053e-01 1.13913310e+00 -3.41218710e-01 7.97006965e-01 1.26163328e+00 8.67298365e-01 8.49274546e-02 -2.01294875e+00 -3.02832991e-01 -1.02972724e-01 -1.17355235e-01 -1.51442409e+00 -4.89892811e-02 8.34079146e-01 -3.83011192e-01 2.39403069e-01 4.40800101e-01 7.65508235e-01 4.67494130e-01 2.65842497e-01 5.47306836e-01 1.21951091e+00 -4.15251493e-01 3.70765746e-01 5.83973899e-02 4.00768250e-01 1.01992500e+00 6.96091831e-01 -3.76099288e-01 1.06750876e-01 -5.84961832e-01 7.01705575e-01 -1.13496885e-01 -3.87318522e-01 -8.25696826e-01 -8.96514237e-01 9.18136716e-01 3.97000015e-01 1.71967596e-01 -4.41096902e-01 6.57341063e-01 1.15033843e-01 5.46015382e-01 7.02568054e-01 2.23307073e-01 -6.49118051e-02 2.49930143e-01 -9.08328116e-01 -2.63107806e-01 1.43532920e+00 1.25496972e+00 1.50361764e+00 -3.43497664e-01 2.53619909e-01 3.44270140e-01 3.13618898e-01 4.79487002e-01 -4.08787936e-01 -7.12805331e-01 7.18647316e-02 7.06443250e-01 -1.92310512e-02 -9.49074149e-01 -1.71186537e-01 -1.59809276e-01 -1.01827192e+00 -1.62917182e-01 5.19918501e-01 -1.62745982e-01 -9.63585734e-01 1.65643537e+00 4.57434207e-01 2.26171449e-01 -4.63610202e-01 3.27987522e-01 4.10022825e-01 3.61466914e-01 -3.02881330e-01 -4.43403840e-01 6.40388370e-01 -4.64386344e-01 -4.87381488e-01 1.79883108e-01 3.51035744e-01 -2.17876464e-01 4.00185943e-01 3.33986759e-01 -9.79097486e-01 -2.81425834e-01 -8.28058064e-01 4.84112501e-01 -4.95499820e-02 -1.14573218e-01 6.30260170e-01 9.43452775e-01 -1.47513664e+00 1.00883293e+00 -1.02048790e+00 -4.80049580e-01 1.74656987e-01 4.20080364e-01 -3.71194273e-01 9.38253477e-02 -7.91288316e-01 3.68228436e-01 5.55223405e-01 -1.09655201e-01 -9.48107421e-01 -1.44566998e-01 -9.72582340e-01 4.52034511e-02 4.46747810e-01 -5.19253969e-01 7.26620972e-01 -9.14863706e-01 -1.25331867e+00 7.66966999e-01 -8.37127119e-02 -5.87070644e-01 6.04798377e-01 4.41400558e-01 -1.16860166e-01 5.82493365e-01 1.81781366e-01 1.63230717e-01 1.14211738e+00 -1.20049691e+00 8.76195803e-02 -3.92226726e-01 1.94679461e-02 -7.36785063e-04 7.43144751e-02 -3.50942403e-01 -4.67978626e-01 -1.70354083e-01 3.26867163e-01 -1.29924798e+00 -4.44989055e-01 -2.02560455e-01 -1.10626459e+00 -2.07567036e-01 4.68862355e-01 -3.18991452e-01 1.17837083e+00 -1.79930794e+00 4.58507121e-01 9.54512119e-01 1.11919439e+00 -4.37435389e-01 1.48707017e-01 7.72454023e-01 7.65267247e-03 3.21951091e-01 -3.63253117e-01 -2.16011241e-01 2.02374697e-01 1.41064599e-01 2.15064153e-01 1.29359198e+00 -1.28313303e-01 6.00598633e-01 -1.18386352e+00 -5.70667028e-01 3.40432167e-01 1.58597648e-01 -4.00549918e-01 2.04653502e-01 -3.25496197e-01 2.17950493e-01 -4.44992632e-01 3.21853817e-01 9.60780442e-01 -7.51869977e-01 8.84288311e-01 5.93599528e-02 4.08830762e-01 -6.61354288e-02 -1.34950292e+00 8.00874531e-01 -7.67882019e-02 6.66914582e-01 4.47427452e-01 -1.24938250e+00 1.06225049e+00 2.84575731e-01 6.99500740e-01 4.64332670e-01 2.98327565e-01 8.45495537e-02 7.33307461e-05 1.24270864e-01 2.11452201e-01 -2.59854198e-01 -2.23962054e-01 5.50492704e-01 2.95871198e-01 -2.81181455e-01 3.66531432e-01 7.34457731e-01 1.65824521e+00 -4.26311582e-01 7.23618329e-01 -9.62620318e-01 5.97348750e-01 -3.65907907e-01 7.35740513e-02 9.74629223e-01 -2.67360359e-01 4.97659177e-01 1.12788427e+00 -1.41228199e-01 -8.45448077e-01 -1.23293042e+00 -2.69379050e-01 4.02095377e-01 2.32531115e-01 -6.37507081e-01 -1.01092350e+00 -1.03863919e+00 3.26115757e-01 1.29368722e-01 -1.02831578e+00 6.60248147e-03 6.23184349e-03 -5.26423633e-01 1.33059800e-01 1.62664473e-01 4.41479504e-01 -8.46409559e-01 1.16755143e-01 1.73383906e-01 -2.44507603e-02 -9.27023470e-01 -8.56976688e-01 2.03898907e-01 -8.51625979e-01 -1.48314691e+00 -4.61092532e-01 -8.63936722e-01 1.07723153e+00 9.01309922e-02 1.45208716e+00 1.89702779e-01 1.28916830e-01 8.11428428e-01 -1.71919852e-01 1.40595123e-01 -7.44113803e-01 2.93386370e-01 -2.70597469e-02 3.00611109e-01 1.19270027e-01 -6.11436248e-01 -2.68550694e-01 2.81632692e-01 -6.83393359e-01 -3.74762774e-01 1.43714607e-01 6.12361491e-01 7.83218145e-01 4.10990983e-01 2.07223549e-01 -1.66795409e+00 7.57049859e-01 -7.96897709e-01 -8.75508904e-01 3.51976156e-01 -6.04584754e-01 4.46648806e-01 7.50527143e-01 -1.52583718e-01 -3.79654288e-01 1.95397496e-01 3.72951478e-01 -3.94528180e-01 1.63881421e-01 3.48291218e-01 -2.32949927e-01 -3.91413599e-01 5.20244956e-01 9.60858986e-02 2.01429933e-01 -8.11926723e-02 4.84444618e-01 3.98345530e-01 1.89550802e-01 -4.73753273e-01 8.64586174e-01 5.00951052e-01 3.53562683e-01 -1.16955376e+00 -5.50622761e-01 -6.13358319e-01 -8.00485492e-01 -3.81255925e-01 7.93275297e-01 -5.21647215e-01 -9.62397993e-01 3.81767541e-01 -9.14278090e-01 -3.52334231e-01 -4.49577034e-01 5.25378764e-01 -5.24011731e-01 6.65395677e-01 -7.22392380e-01 -1.01758742e+00 -1.25340559e-03 -6.28847003e-01 8.64661455e-01 -8.57292861e-02 -2.38343343e-01 -1.75400698e+00 4.31579769e-01 -4.66669232e-01 -4.57562767e-02 5.41839361e-01 5.51951706e-01 -6.25866830e-01 -7.61656404e-01 -4.33821946e-01 -1.93868309e-01 3.20539355e-01 2.08862051e-01 1.44183141e-04 -4.51114058e-01 -6.49068117e-01 1.17964409e-01 -5.37085235e-02 7.87209570e-01 8.49352539e-01 9.88315880e-01 -4.51581448e-01 -8.00338745e-01 5.55714190e-01 1.68561184e+00 -2.28612974e-01 5.82716584e-01 -3.08396786e-01 9.10199702e-01 1.12006448e-01 4.57946658e-02 1.48266464e-01 1.81025878e-01 7.30105788e-02 1.17282465e-01 1.78365558e-01 3.57155055e-02 -5.01130164e-01 4.06425148e-02 1.00943530e+00 -5.37608862e-02 -4.79777813e-01 -5.66422880e-01 4.59482223e-01 -1.70522726e+00 -7.38873780e-01 -5.73600054e-01 2.34411526e+00 5.60515702e-01 2.88476586e-01 4.88215744e-01 -4.26015956e-03 1.39665973e+00 2.42870405e-01 -5.42092085e-01 9.88619635e-04 1.83322072e-01 2.00632513e-01 9.22156096e-01 9.06908154e-01 -9.22123909e-01 5.03469646e-01 7.67782879e+00 6.03234708e-01 -3.25145483e-01 -1.22425899e-01 7.05375195e-01 8.01945329e-01 -5.02792835e-01 4.69175905e-01 -4.79784340e-01 6.10144854e-01 1.00337350e+00 -5.31334698e-01 5.96829474e-01 7.59036601e-01 -3.69693458e-01 -2.21986949e-01 -1.18486011e+00 6.50210977e-01 -9.29333791e-02 -9.39104378e-01 -1.93068370e-01 6.59889281e-01 8.40589404e-01 -3.94225456e-02 -4.64743316e-01 6.48724586e-02 1.13394058e+00 -9.05249476e-01 6.81581199e-02 5.56668580e-01 8.60342085e-01 -8.00707400e-01 4.87407386e-01 2.47474238e-01 -1.72631395e+00 6.26758337e-01 -4.84908491e-01 7.92275462e-03 -5.01451455e-02 9.94497716e-01 -1.15331256e+00 5.55721879e-01 1.42844826e-01 9.19394970e-01 -3.05026919e-01 9.82495070e-01 -2.50896364e-01 9.83088851e-01 -4.01419073e-01 -3.28363359e-01 -1.16817042e-01 -1.13005590e+00 8.26921463e-01 9.74361598e-01 1.34393752e-01 -3.05559463e-03 6.81834996e-01 8.95425200e-01 -4.93310004e-01 -9.69271809e-02 -8.71578813e-01 -3.22785199e-01 6.63883805e-01 1.39551628e+00 -1.37321377e+00 -5.29433966e-01 -2.08268300e-01 8.80084097e-01 6.34149909e-01 2.15702146e-01 -7.89744914e-01 -6.48811996e-01 1.54699996e-01 4.57892984e-01 4.21773523e-01 -1.74003139e-01 2.64831603e-01 -1.08467436e+00 -1.99555114e-01 -3.94024760e-01 5.52504659e-01 -4.07061070e-01 -1.44810879e+00 7.30662048e-01 1.56050995e-01 -1.12929654e+00 -1.26152053e-01 -3.36383730e-01 -4.96682554e-01 5.37822545e-01 -6.34994626e-01 -6.47564828e-01 -2.28346780e-01 4.46849138e-01 -2.65779328e-02 3.38721961e-01 6.93798006e-01 -3.31368208e-01 -1.82079583e-01 1.94958195e-01 2.64230698e-01 2.26929441e-01 7.07286410e-03 -2.01204348e+00 3.51496905e-01 8.67024660e-01 3.32513064e-01 3.89856577e-01 7.61825562e-01 -9.56771493e-01 -1.37220538e+00 -1.07115710e+00 6.24963522e-01 -5.60590446e-01 9.83566821e-01 -5.44328392e-01 -8.27817142e-01 1.28543806e+00 2.56624073e-01 2.57565975e-01 2.72041976e-01 3.04921776e-01 -4.07745875e-03 4.07402404e-03 -1.47474897e+00 1.48686007e-01 1.10172868e+00 -6.07289791e-01 -4.09959815e-02 6.14537537e-01 2.49527141e-01 -8.90641287e-02 -1.21303034e+00 3.69010083e-02 8.39836374e-02 -9.56423581e-01 5.78082263e-01 -3.51676106e-01 -2.46135980e-01 -2.61263549e-01 1.38652682e-01 -1.63787103e+00 -6.40604079e-01 -1.30794442e+00 -1.78051040e-01 8.80808890e-01 3.84069718e-02 -1.02903497e+00 1.03538358e+00 2.61388898e-01 4.61937338e-01 -2.97174960e-01 -6.35622442e-01 -9.59661126e-01 -2.39381999e-01 1.76217705e-02 4.90731120e-01 7.82543302e-01 2.12582648e-01 8.11820269e-01 -3.03449869e-01 2.47109309e-01 1.28686035e+00 1.78121775e-01 8.22652340e-01 -1.78918421e+00 -4.75269407e-01 9.54348873e-03 -9.40945148e-01 -9.15061235e-01 4.28761870e-01 -1.30582106e+00 1.63589761e-01 -1.62499309e+00 6.00683510e-01 -1.86237037e-01 3.55388410e-02 -1.68618143e-01 -8.73689502e-02 2.18551606e-01 -1.56267494e-01 -2.32177624e-03 -1.08614361e+00 3.87931585e-01 1.19417894e+00 1.11298844e-01 -3.50898057e-02 3.89896899e-01 -3.66072834e-01 6.88067913e-01 5.96449733e-01 -6.45354271e-01 -5.24389446e-01 4.16067630e-01 4.87667955e-02 5.22702575e-01 8.32461417e-02 -1.01174951e+00 9.45081115e-02 1.41323909e-01 2.38475248e-01 -5.45129716e-01 2.26534996e-02 -7.73493469e-01 8.47192645e-01 6.00062847e-01 -1.67389780e-01 9.38336477e-02 -3.00458968e-01 1.19150448e+00 -5.88275194e-02 -4.04876322e-01 9.22866821e-01 -2.70943791e-01 9.24933515e-03 5.52541912e-01 -5.13471544e-01 5.37644386e-01 1.07479668e+00 -9.13968310e-02 2.98564192e-02 -9.46823657e-01 -1.07039666e+00 -1.76910893e-03 8.60107243e-01 -6.28769040e-01 3.79284948e-01 -1.44954336e+00 -6.48180425e-01 6.59017712e-02 -1.85916707e-01 -2.04788283e-01 -2.99762309e-01 8.40929568e-01 -7.16939390e-01 1.18117174e-03 2.10899651e-01 -6.82677090e-01 -1.17740023e+00 5.73349833e-01 4.41380411e-01 -5.19946456e-01 -5.99278510e-01 5.80331504e-01 1.95362285e-01 -4.65395063e-01 -2.24007905e-01 -3.75980526e-01 2.68819481e-01 -2.54963905e-01 3.28483507e-02 7.28713453e-01 -2.54973114e-01 -5.50354838e-01 -4.37707782e-01 2.32829049e-01 -5.37642874e-02 -5.47492914e-02 1.05469882e+00 -1.00875400e-01 -4.82794791e-01 6.93173587e-01 1.52102995e+00 3.39519113e-01 -1.33108187e+00 -1.41616344e-01 1.04102269e-01 -5.69904625e-01 -3.26711416e-01 9.63882506e-02 -1.06377196e+00 1.34618029e-01 -7.03939945e-02 1.42895365e+00 7.76901960e-01 4.84690696e-01 2.90066481e-01 2.56441593e-01 6.07269049e-01 -9.39848840e-01 -1.38770059e-01 1.42309129e-01 4.27881360e-01 -9.28997397e-01 3.42438251e-01 -8.77459347e-01 -1.75472885e-01 1.04001796e+00 -8.51083696e-02 -9.33903813e-01 1.24561274e+00 1.34337291e-01 -7.13347077e-01 -4.59837079e-01 -5.31576276e-01 -2.81442940e-01 3.99841726e-01 6.52199984e-01 3.32931072e-01 6.72028899e-01 -2.60351956e-01 -1.27607271e-01 -2.31383488e-01 -2.55202681e-01 8.30671191e-01 4.37976688e-01 -4.85716641e-01 -9.75717485e-01 -9.51096863e-02 1.05157161e+00 -6.74536154e-02 1.89982295e-01 -7.86902547e-01 9.42058146e-01 -5.93630135e-01 7.68348217e-01 -3.04405834e-03 -2.88598776e-01 -1.14765897e-01 -4.08187211e-01 8.32490146e-01 -8.28309178e-01 5.60465679e-02 2.10395572e-03 3.93485963e-01 -3.25544327e-01 -4.35349226e-01 -8.51239502e-01 -8.80305231e-01 -6.24924362e-01 -7.34660685e-01 6.87269211e-01 1.40327215e-01 6.54562831e-01 -5.05853556e-02 2.61297017e-01 9.61172998e-01 -6.61659539e-01 -5.22907972e-01 -9.91478682e-01 -1.37950408e+00 3.13229740e-01 2.67863899e-01 -5.24274766e-01 -8.92441154e-01 -2.84646332e-01]
[6.927732944488525, 5.350561618804932]
81dcf421-d7cc-4cdf-af39-7229bfc292da
bpgc-at-semeval-2020-task-11-propaganda
2006.00593
null
https://arxiv.org/abs/2006.00593v2
https://arxiv.org/pdf/2006.00593v2.pdf
BPGC at SemEval-2020 Task 11: Propaganda Detection in News Articles with Multi-Granularity Knowledge Sharing and Linguistic Features based Ensemble Learning
Propaganda spreads the ideology and beliefs of like-minded people, brainwashing their audiences, and sometimes leading to violence. SemEval 2020 Task-11 aims to design automated systems for news propaganda detection. Task-11 consists of two sub-tasks, namely, Span Identification - given any news article, the system tags those specific fragments which contain at least one propaganda technique; and Technique Classification - correctly classify a given propagandist statement amongst 14 propaganda techniques. For sub-task 1, we use contextual embeddings extracted from pre-trained transformer models to represent the text data at various granularities and propose a multi-granularity knowledge sharing approach. For sub-task 2, we use an ensemble of BERT and logistic regression classifiers with linguistic features. Our results reveal that the linguistic features are the strong indicators for covering minority classes in a highly imbalanced dataset.
['Rajaswa Patil', 'Swati Agarwal', 'Somesh Singh']
2020-05-31
null
https://aclanthology.org/2020.semeval-1.226
https://aclanthology.org/2020.semeval-1.226.pdf
semeval-2020
['propaganda-detection']
['natural-language-processing']
[-8.31565186e-02 -3.39586735e-02 -7.22050250e-01 -2.76580572e-01 -7.71689653e-01 -4.83211845e-01 1.15284979e+00 7.70720780e-01 -3.40949625e-01 5.54221332e-01 1.34246194e+00 -3.70567709e-01 1.05474330e-01 -1.00733137e+00 -2.92093635e-01 -4.22310680e-01 1.39565602e-01 3.19635540e-01 -2.51058668e-01 -3.29197675e-01 8.19843054e-01 -3.01988167e-03 -1.13669014e+00 7.90872991e-01 7.88147509e-01 1.00293505e+00 -6.43729687e-01 5.23970902e-01 -5.95252037e-01 1.84521770e+00 -8.82573128e-01 -8.62697423e-01 -2.03525141e-01 -5.07872760e-01 -9.98997867e-01 -1.87083751e-01 5.28104246e-01 -2.19555683e-02 -1.24905996e-01 1.07371974e+00 5.08277774e-01 -2.03370288e-01 1.13173413e+00 -7.33754277e-01 -1.10580254e+00 1.27694118e+00 -6.84610724e-01 6.12005472e-01 5.39237618e-01 -5.86191356e-01 1.17910576e+00 -8.26047599e-01 7.81788945e-01 1.40276480e+00 8.88272703e-01 2.52439946e-01 -1.07167661e+00 -6.23850524e-01 -1.68410420e-01 3.78554016e-01 -9.15051877e-01 -4.44488108e-01 1.00690377e+00 -1.04217124e+00 7.38826454e-01 3.73087823e-01 7.90858090e-01 1.57823920e+00 4.78885531e-01 5.72190702e-01 1.40579236e+00 -4.67289418e-01 2.74353564e-01 4.13745850e-01 6.93688512e-01 7.25334108e-01 2.13772282e-01 -4.00921345e-01 -8.60864758e-01 -5.50985217e-01 -6.77682757e-02 -1.94686249e-01 5.04565565e-03 4.45773602e-01 -1.14151347e+00 1.53352070e+00 1.66048408e-01 5.21801829e-01 -4.50271010e-01 -3.41337286e-02 8.59923780e-01 5.30353904e-01 1.37487078e+00 7.33571410e-01 -5.94778955e-02 -2.27287680e-01 -1.08140707e+00 4.44442123e-01 8.60910773e-01 2.03795731e-01 2.55482048e-01 -2.47642651e-01 -5.89204133e-01 9.68222618e-01 -1.23001479e-01 3.48950595e-01 5.09876847e-01 -2.38946185e-01 7.07687259e-01 7.58888960e-01 -1.38888553e-01 -1.55309153e+00 -6.10290051e-01 -6.57930672e-01 -5.86775064e-01 -2.83597082e-01 1.82829067e-01 -1.55797869e-01 -4.91386592e-01 1.46014512e+00 3.60203832e-01 -4.80620921e-01 -2.22496077e-01 6.93637431e-01 9.03900325e-01 8.27169061e-01 1.45312428e-01 -3.07263017e-01 1.62773323e+00 -9.42125380e-01 -9.34257805e-01 -4.39516902e-02 7.42118478e-01 -9.96409297e-01 7.88138866e-01 9.89386141e-02 -6.02785051e-01 -9.14802104e-02 -8.64377737e-01 -1.43095985e-01 -3.68620217e-01 5.31179048e-02 2.66650140e-01 7.35787868e-01 -8.98648053e-02 2.35889226e-01 -2.55706012e-01 -2.62305021e-01 8.04467082e-01 -7.60512173e-01 -1.69144332e-01 1.04148649e-01 -1.30902135e+00 1.29840922e+00 2.72161901e-01 -4.94227052e-01 -7.67787576e-01 -1.03264201e+00 -8.70412052e-01 -8.38927701e-02 1.75641939e-01 -2.64077246e-01 5.99723995e-01 -1.14657760e+00 -1.10641813e+00 1.29526031e+00 1.36676639e-01 -6.36541963e-01 4.29292262e-01 -5.85958481e-01 -8.39064717e-01 2.25753170e-02 4.16789651e-01 -1.99086770e-01 1.12667263e+00 -7.28712738e-01 -4.79365855e-01 -2.21750736e-01 -1.22837208e-01 3.27343680e-02 -7.58089304e-01 8.08499277e-01 6.49788857e-01 -9.51341152e-01 -2.62521744e-01 -4.47437465e-01 3.70624721e-01 -4.81494159e-01 -7.49103844e-01 -4.89023387e-01 7.76135981e-01 -1.10286009e+00 1.71316004e+00 -2.07955408e+00 1.06416605e-01 -3.03658098e-02 4.42855060e-01 3.75339612e-02 2.12157592e-01 6.54899478e-01 5.81384636e-03 2.18997166e-01 3.65079314e-01 7.06471577e-02 1.38037965e-01 -3.20525169e-01 -7.05173671e-01 8.02395105e-01 1.58359736e-01 5.35281479e-01 -7.08454072e-01 -5.62675059e-01 -2.95404214e-02 1.73624694e-01 -4.04169142e-01 1.14076706e-02 -2.24188626e-01 -3.32617797e-02 -5.57954431e-01 4.46704835e-01 4.01390851e-01 -2.05670789e-01 1.42634615e-01 -2.73572892e-01 -3.80538523e-01 9.68742609e-01 -6.40652180e-01 1.07905245e+00 -3.89267981e-01 8.96198452e-01 -2.00646520e-01 -9.68068540e-01 8.13032627e-01 5.15872166e-02 2.00865373e-01 -8.70412111e-01 6.09909832e-01 6.91790655e-02 -2.05172852e-01 -7.40572691e-01 5.25644720e-01 -3.69943470e-01 -7.37526655e-01 8.44186783e-01 -2.78065987e-02 3.28093678e-01 -4.02144231e-02 3.92170310e-01 1.22614741e+00 -5.36147892e-01 7.21109033e-01 -6.50698721e-01 4.50468779e-01 2.47873917e-01 4.55699772e-01 6.26224875e-01 -2.44765550e-01 9.06224698e-02 7.30930805e-01 -1.04526794e+00 -1.02578413e+00 -9.05009329e-01 -4.10822660e-01 1.67516577e+00 -9.46120992e-02 -5.89994729e-01 -4.46199864e-01 -8.86465847e-01 7.37009495e-02 1.07297170e+00 -1.11632764e+00 1.56832799e-01 -5.00362098e-01 -8.13058853e-01 7.55880117e-01 5.40925451e-02 3.82552922e-01 -8.39974105e-01 -8.65934908e-01 3.87401998e-01 -4.82370734e-01 -7.01563001e-01 -3.73506546e-01 5.44430725e-02 3.45644541e-03 -1.24972665e+00 -3.58070374e-01 -8.13565195e-01 3.07768106e-01 -1.90493986e-01 1.08526099e+00 -8.54272172e-02 -2.64171749e-01 -2.07462862e-01 -7.04650283e-01 -4.68295187e-01 -7.75477469e-01 1.73704237e-01 1.36875138e-01 1.88987240e-01 3.60801518e-01 -1.74701720e-01 -8.29267055e-02 -2.15532690e-01 -3.95684004e-01 3.13325644e-01 1.15396507e-01 8.76213551e-01 -1.20754041e-01 8.11502279e-04 6.83124304e-01 -1.03035164e+00 9.33579743e-01 -7.54089534e-01 1.84958413e-01 3.56386542e-01 -1.93803251e-01 -1.16642289e-01 6.87008262e-01 -5.56103230e-01 -1.09121156e+00 -7.93401182e-01 -1.13207415e-01 4.07332450e-01 2.15396881e-01 6.58936501e-01 3.40794772e-01 4.74415869e-01 1.24309444e+00 5.37547050e-03 -2.22047344e-01 -4.17665154e-01 5.01237392e-01 1.05455899e+00 1.37257069e-01 -5.26003301e-01 5.81489325e-01 4.15886819e-01 -6.49801731e-01 -9.03532207e-01 -1.77565217e+00 -4.85545456e-01 3.36857736e-02 -3.34854215e-01 9.22016859e-01 -9.19973850e-01 -2.42778972e-01 3.90103728e-01 -1.26993263e+00 8.97268504e-02 -8.18583742e-02 4.21511650e-01 -2.25979760e-01 7.02795107e-03 -8.64948988e-01 -9.11087275e-01 -6.26565933e-01 -5.34200191e-01 4.75429773e-01 -1.57579303e-01 -6.18247390e-01 -9.79679704e-01 3.89740855e-01 6.31221592e-01 5.14027655e-01 5.24777949e-01 1.19154894e+00 -1.21438587e+00 4.55191761e-01 -1.05803944e-01 -3.23235482e-01 1.83801368e-01 1.36682048e-01 -1.32812932e-01 -8.28767896e-01 9.72450599e-02 2.14220017e-01 -7.73752749e-01 1.08586967e+00 2.50619888e-01 8.30174685e-01 -1.12752700e+00 -8.51474553e-02 1.37123600e-01 1.16507494e+00 -3.39809537e-01 2.86616027e-01 7.28565216e-01 5.52799523e-01 7.12184727e-01 3.14872235e-01 5.13813615e-01 3.06397051e-01 4.54219371e-01 -1.42386183e-02 2.97353983e-01 -1.45130470e-01 -4.37641323e-01 6.64586306e-01 9.81967151e-01 7.05562979e-02 -5.55313043e-02 -8.73671055e-01 8.26063812e-01 -1.55742705e+00 -1.71418822e+00 -3.01927447e-01 1.47940052e+00 1.19516695e+00 1.23664163e-01 2.64257878e-01 3.18044186e-01 7.37653077e-01 7.38114595e-01 2.01449513e-01 -8.87594044e-01 -2.08838016e-01 1.33178115e-01 2.21122056e-01 5.42874753e-01 -1.52227104e+00 7.57506192e-01 5.95798540e+00 1.23989713e+00 -1.01914310e+00 6.36890590e-01 5.68360329e-01 -2.01091081e-01 -4.76617962e-01 -2.04144746e-01 -7.18859136e-01 7.27741361e-01 8.71458411e-01 -3.84142578e-01 -1.00598624e-02 7.85123169e-01 2.52719373e-02 1.60240028e-02 -7.07407951e-01 5.75230181e-01 7.35977590e-01 -1.95779645e+00 1.76073015e-01 -1.24137953e-01 9.29210663e-01 -5.65125495e-02 1.20111234e-01 4.00670052e-01 4.71757591e-01 -1.03510582e+00 1.32247329e+00 1.14315756e-01 5.52843750e-01 -5.86776972e-01 7.48916209e-01 5.24628639e-01 -5.19050837e-01 -2.97174066e-01 1.64964236e-02 -4.32517529e-01 4.75752175e-01 1.23996842e+00 -4.69930738e-01 2.78254479e-01 4.06528920e-01 6.76936984e-01 -3.88623118e-01 4.49370474e-01 -3.92131984e-01 1.06115448e+00 -5.60407154e-02 -4.41422701e-01 2.72517502e-01 3.52734476e-01 6.35461807e-01 1.73961687e+00 1.22065708e-01 -1.85459182e-01 4.52854447e-02 6.86729670e-01 -1.77568108e-01 5.66323519e-01 -3.47655773e-01 -3.36823612e-01 4.83569533e-01 1.08941782e+00 -4.49738950e-01 -4.35598314e-01 -9.05870423e-02 5.14961958e-01 6.78948879e-01 -2.19335645e-01 -1.05763221e+00 -1.76719978e-01 4.48999405e-01 2.57221967e-01 2.46483702e-02 1.11475311e-01 -4.68509793e-01 -1.36044538e+00 -2.93345958e-01 -9.05364633e-01 5.79305410e-01 -1.03203267e-01 -1.71384907e+00 4.80191350e-01 -2.89691955e-01 -6.47773206e-01 1.76108763e-01 -3.01196069e-01 -7.35404730e-01 4.39822614e-01 -1.31454396e+00 -1.37544298e+00 1.80497736e-01 1.56979755e-01 6.66920722e-01 -3.84776324e-01 8.34131479e-01 1.68631911e-01 -3.75367135e-01 3.93848866e-01 -1.25379944e-02 4.98024791e-01 7.13167727e-01 -7.85671413e-01 -8.76627788e-02 6.45790517e-01 2.91756898e-01 4.97637093e-01 1.03934872e+00 -7.15323448e-01 -8.00102949e-01 -1.09237146e+00 1.54798722e+00 -5.79198003e-01 1.29287374e+00 -4.83767450e-01 -6.06815338e-01 3.70203197e-01 2.33093411e-01 -4.59449142e-01 1.16086686e+00 6.30713582e-01 -1.06554019e+00 2.85465658e-01 -1.05079341e+00 1.68026701e-01 8.66015375e-01 -7.31121242e-01 -1.32044482e+00 6.17974281e-01 4.63939041e-01 -1.08913243e-01 -6.43054426e-01 -2.15241879e-01 6.29830062e-01 -7.82190919e-01 6.58275783e-01 -9.75081742e-01 1.24726605e+00 1.20300859e-01 -3.80242974e-01 -1.30150938e+00 -5.60273588e-01 -1.98172972e-01 -2.34834746e-01 1.35478222e+00 3.82411182e-01 -4.40426022e-01 2.89604664e-01 -3.60667855e-01 1.56558245e-01 -4.69198465e-01 -1.24132299e+00 -5.84701002e-01 3.20602119e-01 -4.09777075e-01 1.97621614e-01 1.63665533e+00 3.32977355e-01 6.77005708e-01 -6.81971788e-01 -1.13043346e-01 6.97133422e-01 3.39266270e-01 4.47405696e-01 -9.93417740e-01 -9.40405875e-02 -6.13582492e-01 -4.19162571e-01 -4.96129543e-01 3.26723874e-01 -1.07079470e+00 -2.79709786e-01 -1.39115894e+00 5.59761405e-01 -2.78777272e-01 -8.79698694e-02 3.61144423e-01 2.04583928e-02 1.77532077e-01 1.03057940e-02 1.82624415e-01 -5.92202783e-01 5.23499906e-01 6.75177097e-01 -5.97755194e-01 5.15917279e-02 -3.48818094e-01 -1.00817752e+00 8.61838579e-01 7.42236912e-01 -8.31097662e-01 2.03118950e-01 -3.94382507e-01 7.52261817e-01 -3.02134514e-01 4.54816580e-01 -8.21963072e-01 -9.24680904e-02 -4.66920525e-01 1.73658833e-01 -4.85397667e-01 -1.61603615e-01 -3.28826874e-01 -7.70588741e-02 7.01097012e-01 -8.09497774e-01 -3.52467149e-01 -1.64611161e-01 4.14612979e-01 -1.30964562e-01 -1.70788065e-01 8.02897215e-01 2.50399202e-01 -2.63217330e-01 -2.54300922e-01 -8.14969718e-01 5.33151746e-01 9.17117000e-01 3.82011503e-01 -1.05773938e+00 -1.11097626e-01 -5.79828680e-01 -9.63149071e-02 1.45508572e-01 4.45482314e-01 1.99847281e-01 -1.37534845e+00 -1.38036942e+00 -5.63842729e-02 3.35859448e-01 -7.85583079e-01 2.74869382e-01 8.58389556e-01 -6.22600973e-01 -2.83019375e-02 -1.26159638e-01 2.31904779e-02 -1.30143631e+00 3.88443887e-01 -7.96608329e-02 -4.53924268e-01 -7.22068846e-01 9.11029518e-01 -1.87835783e-01 -1.33825839e-01 -1.66255593e-01 -2.50541642e-02 -4.56307441e-01 6.97150230e-01 1.01853549e+00 5.33069193e-01 -1.39227509e-01 -9.47695196e-01 -5.43856263e-01 2.22898558e-01 -2.46730760e-01 2.55659875e-02 1.57289374e+00 1.66150525e-01 -4.85892534e-01 6.08076930e-01 1.30294669e+00 6.24843299e-01 -2.56129056e-01 -3.50486398e-01 5.83787747e-02 -4.36696768e-01 2.65830368e-01 -1.03733182e+00 -3.46625179e-01 7.02181578e-01 -3.52333374e-02 6.64237499e-01 4.71136779e-01 3.32099319e-01 6.39560699e-01 -6.97225034e-02 1.87444255e-01 -1.50847054e+00 1.87320158e-01 7.25462437e-01 1.12416995e+00 -7.66278982e-01 3.45803320e-01 -3.23293388e-01 -6.79611802e-01 1.06695414e+00 5.61035424e-02 -1.66181982e-01 6.12222195e-01 2.28923187e-01 9.19356123e-02 -6.32346272e-01 -7.17621684e-01 2.42121816e-01 4.65413451e-01 3.07116330e-01 5.22089660e-01 4.88676578e-01 -8.95420432e-01 8.96116316e-01 -5.40005445e-01 -2.91273654e-01 3.40706766e-01 6.20437622e-01 -8.52902293e-01 -3.42337132e-01 -4.33903486e-01 7.38683701e-01 -8.08534920e-01 -4.22356278e-02 -5.65442026e-01 2.37138689e-01 3.98836434e-01 1.02585542e+00 1.72606990e-01 -5.40312529e-01 -8.58029053e-02 1.68804482e-01 3.98669422e-01 -5.53637862e-01 -9.93937075e-01 -3.29361796e-01 9.04777169e-01 -1.95474193e-01 -5.88247120e-01 -5.37146866e-01 -6.87670767e-01 -9.01705801e-01 -4.00762707e-01 3.44124347e-01 5.19735396e-01 1.11720324e+00 -5.35096489e-02 3.69704187e-01 7.50805557e-01 -2.53934413e-01 -7.97453105e-01 -1.12682223e+00 -4.03185248e-01 6.36448085e-01 2.56376475e-01 -6.68698609e-01 -5.89056671e-01 -5.95846698e-02]
[8.524947166442871, 10.530829429626465]
668dea0b-e83f-4825-a04f-9ecbab544051
local-global-fusion-network-for-video-super
null
null
https://ieeexplore.ieee.org/document/9203860/authors#authors
https://ieeexplore.ieee.org/document/9203860/authors#authors
Local-Global Fusion Network for Video Super-Resolution
The goal of video super-resolution technique is to address the problem of effectively restoring high-resolution (HR) videos from low-resolution (LR) ones. Previous methods commonly used optical flow to perform frame alignment and designed a framework from the perspective of space and time. However, inaccurate optical flow estimation may occur easily which leads to inferior restoration effects. In addition, how to effectively fuse the features of various video frames remains a challenging problem. In this paper, we propose a Local-Global Fusion Network (LGFN) to solve the above issues from a novel viewpoint. As an alternative to optical flow, deformable convolutions (DCs) with decreased multi-dilation convolution units (DMDCUs) are applied for efficient implicit alignment. Moreover, a structure with two branches, consisting of a Local Fusion Module (LFM) and a Global Fusion Module (GFM), is proposed to combine information from two different aspects. Specifically, LFM focuses on the relationship between adjacent frames and maintains the temporal consistency while GFM attempts to take advantage of all related features globally with a video shuffle strategy. Benefiting from our advanced network, experimental results on several datasets demonstrate that our LGFN can not only achieve comparative performance with state-of-the-art methods but also possess reliable ability on restoring a variety of video frames. The results on benchmark datasets of our LGFN are presented on https://github.com/BIOINSu/LGFN and the source code will be released as soon as the paper is accepted.
['Xinyi Peng', 'Xianfang Sun', 'Longcun Jin', 'Hua Wang', 'Dewei Su']
2020-09-22
null
null
null
ieee-access-2020-9
['video-super-resolution']
['computer-vision']
[ 1.16753690e-01 -5.78717411e-01 -3.24122980e-02 -1.56036854e-01 -4.07830447e-01 -1.68694407e-01 3.00507575e-01 -4.30670857e-01 -2.96677917e-01 9.45530057e-01 2.59700745e-01 2.25466508e-02 -1.28553569e-01 -7.24407077e-01 -5.63378274e-01 -8.31917226e-01 1.54233068e-01 -4.87316966e-01 4.46529150e-01 -2.78617024e-01 3.78674835e-01 4.53361958e-01 -1.46327269e+00 2.27950007e-01 1.07084286e+00 7.61218965e-01 5.40511966e-01 3.21442574e-01 -1.43690109e-01 9.03572381e-01 -2.39507228e-01 -7.69018084e-02 3.93273652e-01 -5.09613156e-01 -7.38882959e-01 6.43242849e-03 5.07863104e-01 -6.45130575e-01 -5.51839232e-01 1.12775731e+00 6.04526401e-01 4.60042685e-01 -2.02855170e-02 -1.01820064e+00 -7.41672099e-01 2.37729549e-01 -6.68486893e-01 7.63910651e-01 4.38620478e-01 1.69211641e-01 6.16691887e-01 -9.04237211e-01 6.64417148e-01 1.31658256e+00 3.50489765e-01 5.19159257e-01 -8.84845614e-01 -6.74595416e-01 2.81376928e-01 5.14558434e-01 -1.29345965e+00 -6.17238402e-01 7.14573443e-01 -3.20281327e-01 6.23087049e-01 2.13266641e-01 4.63814288e-01 6.95947587e-01 1.54685423e-01 3.47623229e-01 1.03286469e+00 -6.39498755e-02 -2.77132750e-01 -3.01100701e-01 -5.46934940e-02 5.54101706e-01 1.92022398e-01 1.23362377e-01 -5.34865320e-01 1.98986977e-01 1.23611057e+00 8.54420364e-02 -1.00075090e+00 4.49943542e-03 -1.32494032e+00 3.67284834e-01 4.57580417e-01 4.91551131e-01 -3.42143238e-01 -8.61727819e-02 3.56615335e-01 1.95859626e-01 3.44796538e-01 -5.44001460e-02 -1.94600716e-01 1.61929745e-02 -7.25927293e-01 1.02064442e-02 3.65849346e-01 6.86953485e-01 8.96888852e-01 1.86103925e-01 -2.07921341e-01 8.39429557e-01 1.29932314e-01 7.79979974e-02 6.97361350e-01 -1.17533398e+00 5.40432632e-01 3.04957658e-01 1.15046926e-01 -1.25319326e+00 -1.77003369e-01 -3.46131921e-01 -1.07517350e+00 1.62834436e-01 2.10128069e-01 1.74552687e-02 -5.78887880e-01 1.65926528e+00 5.01924038e-01 7.75761247e-01 1.66347343e-02 1.36864901e+00 9.14421797e-01 7.61466682e-01 -7.71283284e-02 -6.22093320e-01 1.08943057e+00 -9.45508659e-01 -9.86559391e-01 1.89252838e-01 2.31123492e-01 -1.04934597e+00 6.84939504e-01 2.12317184e-01 -1.27307081e+00 -9.91850615e-01 -1.05361342e+00 -2.83120900e-01 -4.45149206e-02 -2.00760216e-02 4.06590462e-01 7.99141452e-02 -1.13344324e+00 6.93447948e-01 -7.82059669e-01 -2.14684770e-01 2.78650284e-01 2.95481324e-01 -5.18397570e-01 -1.27576709e-01 -1.25297213e+00 6.65668607e-01 4.35381889e-01 5.23165107e-01 -5.69371223e-01 -6.13151312e-01 -7.68020570e-01 -8.19949135e-02 3.08074147e-01 -8.01022470e-01 9.11275208e-01 -1.03057718e+00 -1.55444968e+00 3.54214489e-01 -3.79401267e-01 -1.44845560e-01 4.57837999e-01 -2.81945825e-01 -5.23295403e-01 4.29299206e-01 -7.30169425e-03 4.34178740e-01 7.76614904e-01 -1.09051573e+00 -8.82881880e-01 -5.30648492e-02 1.83408499e-01 4.23829913e-01 -3.55120718e-01 1.15269020e-01 -4.26452577e-01 -7.55432248e-01 1.85636804e-01 -5.18249810e-01 4.93471175e-02 -3.31107490e-02 -5.01480065e-02 -2.52417848e-02 9.79346037e-01 -9.56710041e-01 1.56148934e+00 -2.04817176e+00 3.99657935e-01 -2.85714298e-01 3.03468287e-01 6.62651241e-01 -7.66094700e-02 1.56232461e-01 -2.85045236e-01 3.12717236e-03 -1.75090328e-01 -1.38400525e-01 -6.02039218e-01 1.06856510e-01 -1.35007307e-01 5.49327016e-01 1.13623559e-01 5.70873380e-01 -9.56802547e-01 -7.42982924e-01 5.29118836e-01 8.41317713e-01 -4.15837586e-01 1.98312968e-01 2.44548693e-01 9.33664501e-01 -4.74523962e-01 5.77118039e-01 9.98723626e-01 -1.65737376e-01 9.22466740e-02 -5.91998816e-01 -4.89576846e-01 -3.39535810e-02 -1.37662244e+00 1.69264781e+00 -2.14673027e-01 3.04613918e-01 2.21231133e-01 -8.88563454e-01 8.57326567e-01 3.66383791e-01 7.36963868e-01 -7.44738936e-01 1.20856412e-01 2.22163290e-01 -3.90411541e-02 -6.24404132e-01 4.56226110e-01 2.61702780e-02 6.02165341e-01 9.45733264e-02 -1.81675190e-03 6.21573925e-01 4.27746862e-01 3.23377438e-02 7.38555491e-01 5.78399956e-01 2.74023861e-01 -1.87634975e-01 1.26587236e+00 -4.46301401e-01 1.19325173e+00 2.24842995e-01 -4.78008866e-01 7.26469457e-01 1.24288857e-01 -6.57888114e-01 -8.39020848e-01 -8.65179896e-01 -1.11956373e-01 4.92351919e-01 6.25293732e-01 -4.09587562e-01 -6.04138136e-01 -3.29977661e-01 -3.31215382e-01 4.67986241e-02 -2.70755559e-01 4.25061248e-02 -1.09647620e+00 -7.37298489e-01 1.26147345e-01 3.60088050e-01 1.00917923e+00 -9.15615618e-01 -4.41203028e-01 2.04052642e-01 -5.87852359e-01 -1.26549232e+00 -7.67081499e-01 -7.25291193e-01 -9.22353864e-01 -1.09499431e+00 -6.74683690e-01 -7.99601376e-01 5.80856800e-01 9.42244470e-01 7.39351928e-01 4.60003555e-01 -1.37306482e-01 -4.06453274e-02 -4.04306471e-01 3.43230993e-01 -1.35894343e-01 -2.53414810e-01 5.56254946e-02 3.92170191e-01 5.62720783e-02 -8.16531599e-01 -8.95060241e-01 4.94722545e-01 -1.09987116e+00 3.21938306e-01 3.67509067e-01 7.68566549e-01 6.60370767e-01 1.59268826e-01 5.55397749e-01 -4.88805562e-01 3.83407235e-01 -3.82066280e-01 -5.81832588e-01 1.80055797e-01 -4.06105071e-01 -2.68361837e-01 8.17900836e-01 -3.34703088e-01 -1.40688121e+00 -1.52568594e-01 4.89885695e-02 -7.76347160e-01 -8.17162171e-03 2.27334902e-01 -2.26537660e-01 -2.72157699e-01 1.86909050e-01 3.45747739e-01 1.36725217e-01 -5.49673736e-01 1.10941365e-01 5.05339622e-01 7.00245619e-01 -5.66854119e-01 8.35501552e-01 6.21357679e-01 7.92241693e-02 -6.03472590e-01 -5.40136933e-01 -3.75129461e-01 -6.28641665e-01 -4.35240507e-01 8.71965885e-01 -9.80285466e-01 -7.68413305e-01 6.92289174e-01 -1.07065094e+00 4.26517799e-02 1.22494869e-01 7.62227416e-01 -4.47490931e-01 7.74429917e-01 -8.13460529e-01 -5.25504172e-01 -3.93741965e-01 -1.20307779e+00 6.50524795e-01 7.18125045e-01 3.92305881e-01 -9.07315314e-01 -9.23205689e-02 4.07142937e-01 4.92709279e-01 3.42081159e-01 3.61539900e-01 1.53161287e-01 -1.02775133e+00 3.13683063e-01 -4.07819033e-01 5.58887422e-01 5.62716424e-01 1.61796302e-01 -7.01739609e-01 -5.49017787e-01 1.71130747e-01 2.34297842e-01 8.27934504e-01 4.35381353e-01 1.10097826e+00 -1.87957942e-01 -1.22238308e-01 9.29681182e-01 1.72296834e+00 2.50618786e-01 1.00187278e+00 5.14053404e-01 9.78247881e-01 4.01108325e-01 8.42240989e-01 3.99465054e-01 3.71491909e-01 8.49502742e-01 3.79053533e-01 -2.32867450e-02 -4.38962072e-01 -1.03228167e-02 5.11106133e-01 8.98769438e-01 -5.86602211e-01 -1.25615850e-01 -4.37777936e-01 2.83946633e-01 -2.06286073e+00 -1.21085382e+00 -2.22927317e-01 2.29721236e+00 7.63257205e-01 -1.02543406e-01 -8.12037215e-02 6.74922019e-02 1.09430254e+00 3.41299713e-01 -3.37281018e-01 -7.54192621e-02 -3.42060626e-01 8.50690994e-03 3.00444603e-01 5.95585108e-01 -1.09617221e+00 7.42847562e-01 4.82497692e+00 8.15786660e-01 -1.24211502e+00 2.31167629e-01 4.37068999e-01 -5.43047115e-02 1.02792270e-02 1.16103336e-01 -7.54119873e-01 7.02449977e-01 4.97785389e-01 -1.64144263e-01 6.31549776e-01 1.51873976e-01 5.47084570e-01 6.70126872e-03 -4.88467366e-01 1.06285214e+00 -7.75287971e-02 -1.31502450e+00 8.92128050e-02 -4.96741720e-02 6.49717331e-01 -2.36915782e-01 -1.19508967e-01 -1.04925901e-01 -2.83315718e-01 -7.17703462e-01 4.51337099e-01 9.43401873e-01 7.71467626e-01 -7.05320954e-01 7.34963655e-01 1.44154042e-01 -1.63649893e+00 -9.67862606e-02 -3.51966232e-01 -1.25603572e-01 4.63670373e-01 5.78843296e-01 3.99868842e-03 1.19905877e+00 9.58557904e-01 1.27315795e+00 -3.47207397e-01 1.06828427e+00 -7.44238347e-02 5.06700203e-02 3.26397493e-02 7.01359928e-01 3.47448215e-02 -5.04799843e-01 7.45722353e-01 9.09727216e-01 4.32367682e-01 3.21492076e-01 1.84152424e-01 6.58360541e-01 1.96282476e-01 1.60501078e-01 -2.32951075e-01 3.04086030e-01 3.78368109e-01 1.58272243e+00 -4.13535327e-01 -2.24550664e-01 -6.84109688e-01 8.67076159e-01 2.06100792e-01 4.10628229e-01 -8.68195176e-01 -2.71192819e-01 7.63921440e-01 1.00673094e-01 2.66280085e-01 -3.43083471e-01 2.22493201e-01 -1.59421682e+00 2.40703627e-01 -7.41214097e-01 4.62175548e-01 -8.29065919e-01 -1.07697368e+00 7.51955569e-01 -1.24830917e-01 -1.78166199e+00 8.35933089e-02 -2.93414026e-01 -4.78926033e-01 1.11446583e+00 -2.02817607e+00 -9.35968876e-01 -7.31055021e-01 8.19660366e-01 5.56905746e-01 1.02559663e-01 3.29615772e-01 7.09254682e-01 -9.16437507e-01 3.23113590e-01 -8.89671743e-02 1.32509649e-01 9.26511288e-01 -8.38799000e-01 -1.49063379e-01 1.26356137e+00 -3.46306801e-01 6.78373396e-01 4.05952513e-01 -5.79889655e-01 -1.22590661e+00 -1.03933334e+00 6.74782991e-01 6.50326014e-02 3.96655679e-01 3.14226419e-01 -1.24597788e+00 5.77842474e-01 2.81455189e-01 3.41977268e-01 2.49174744e-01 -6.17549002e-01 -4.90451120e-02 -3.94842386e-01 -1.08255053e+00 4.29263920e-01 1.23106182e+00 -2.12391168e-01 -2.42688939e-01 9.68767852e-02 7.87356973e-01 -6.76859319e-01 -1.13941538e+00 5.96095681e-01 4.41481888e-01 -1.33786750e+00 1.07424021e+00 -2.18511358e-01 5.21466553e-01 -9.43087280e-01 -1.72044009e-01 -9.45310831e-01 -5.33372462e-01 -7.72909045e-01 -2.79702187e-01 1.47613168e+00 -4.40181971e-01 -8.86668503e-01 2.59114563e-01 1.84191212e-01 -2.23419040e-01 -7.29682803e-01 -8.89404297e-01 -6.18322372e-01 -2.91501045e-01 1.90892383e-01 5.07222056e-01 1.01871777e+00 -3.76022339e-01 1.58769354e-01 -5.92385948e-01 3.40717643e-01 6.60093069e-01 1.08203046e-01 4.94806081e-01 -9.70837057e-01 -1.43283635e-01 -3.97627980e-01 -4.24711734e-01 -1.02455711e+00 1.19283952e-01 -6.97949171e-01 -2.53549516e-01 -1.43717396e+00 1.50876552e-01 -2.67083794e-01 -5.44855118e-01 1.91179737e-01 -3.91177326e-01 2.21977711e-01 3.92865032e-01 5.25325239e-01 -4.62357759e-01 5.01736462e-01 1.71724653e+00 2.81813592e-01 -2.02640757e-01 -1.85671508e-01 -5.50939202e-01 5.08430898e-01 9.95557487e-01 3.93981561e-02 -4.14422780e-01 -6.29594564e-01 -3.38238746e-01 3.13660920e-01 4.98533309e-01 -1.02836454e+00 2.71155119e-01 -2.17100024e-01 3.55035365e-01 -4.79938507e-01 2.26349279e-01 -6.42908156e-01 4.69078869e-01 3.85372430e-01 6.96819201e-02 2.83210635e-01 3.74538973e-02 4.79449868e-01 -5.34961283e-01 1.55810229e-02 9.88943100e-01 -2.30643287e-01 -1.02297902e+00 5.64182639e-01 9.14210305e-02 -2.38391653e-01 9.77382123e-01 -2.97212809e-01 -6.98420942e-01 -1.63902551e-01 -5.74318349e-01 2.78263688e-01 5.06998301e-01 5.74289501e-01 8.45008731e-01 -1.46661162e+00 -8.15774083e-01 2.32900292e-01 -3.04340839e-01 2.87786238e-02 7.84016967e-01 1.20188868e+00 -7.27483153e-01 3.93842250e-01 -5.21659911e-01 -4.81466323e-01 -1.24647403e+00 5.84134936e-01 5.13930857e-01 -1.40752152e-01 -7.79067755e-01 5.70375741e-01 3.29683274e-01 -3.37555781e-02 -7.73331150e-02 -4.63454872e-02 -5.52065134e-01 -3.14845890e-02 8.76694202e-01 6.66560173e-01 -1.40998155e-01 -1.06164706e+00 -2.77644157e-01 9.84035075e-01 -9.73381400e-02 2.29042053e-01 1.22554159e+00 -6.50638342e-01 -4.46763337e-01 1.34669309e-02 1.15500534e+00 -1.48351103e-01 -1.40160251e+00 -3.95175368e-01 -4.61164206e-01 -9.05343115e-01 1.83475196e-01 -3.27243000e-01 -1.57607067e+00 7.05662131e-01 6.70281768e-01 1.70893210e-03 1.57539189e+00 -5.31865954e-01 1.04757416e+00 -2.54774034e-01 3.15122396e-01 -8.21643293e-01 -1.65185444e-02 2.57592767e-01 7.26826310e-01 -1.07591045e+00 5.19687533e-02 -6.67281926e-01 -3.16130728e-01 1.38675380e+00 8.90210032e-01 -1.96525946e-01 4.54168826e-01 -4.27899249e-02 -6.21870384e-02 2.10862845e-01 -6.16096616e-01 -1.72524989e-01 1.06551342e-01 3.41595799e-01 4.95034456e-01 -2.99712628e-01 -7.55087912e-01 2.75350094e-01 2.87284434e-01 4.58575159e-01 6.64898157e-01 9.19583380e-01 -4.81738597e-01 -1.18759465e+00 -4.02532697e-01 9.53086019e-02 -5.32932281e-01 -4.08849642e-02 4.41620767e-01 5.43639600e-01 3.65487367e-01 9.63893771e-01 -4.89967614e-02 -4.38972175e-01 1.81733623e-01 -3.72093707e-01 4.36420292e-01 -1.37364209e-01 -2.98089534e-01 3.42522651e-01 -1.88935921e-01 -8.39779615e-01 -1.06295455e+00 -6.44177616e-01 -1.30727589e+00 -5.55508494e-01 -1.38742924e-01 -7.75722563e-02 1.87344626e-01 7.04970062e-01 4.31189984e-01 7.31563151e-01 7.45676994e-01 -8.75457764e-01 -2.18423128e-01 -7.79189169e-01 -4.93003249e-01 3.96818042e-01 5.86045146e-01 -6.42683327e-01 -3.38472933e-01 3.04678828e-01]
[11.042531967163086, -1.840067982673645]
c312a07c-c1ea-4676-8172-121302b423f0
highly-parallel-autoregressive-entity-linking
2109.03792
null
https://arxiv.org/abs/2109.03792v1
https://arxiv.org/pdf/2109.03792v1.pdf
Highly Parallel Autoregressive Entity Linking with Discriminative Correction
Generative approaches have been recently shown to be effective for both Entity Disambiguation and Entity Linking (i.e., joint mention detection and disambiguation). However, the previously proposed autoregressive formulation for EL suffers from i) high computational cost due to a complex (deep) decoder, ii) non-parallelizable decoding that scales with the source sequence length, and iii) the need for training on a large amount of data. In this work, we propose a very efficient approach that parallelizes autoregressive linking across all potential mentions and relies on a shallow and efficient decoder. Moreover, we augment the generative objective with an extra discriminative component, i.e., a correction term which lets us directly optimize the generator's ranking. When taken together, these techniques tackle all the above issues: our model is >70 times faster and more accurate than the previous generative method, outperforming state-of-the-art approaches on the standard English dataset AIDA-CoNLL. Source code available at https://github.com/nicola-decao/efficient-autoregressive-EL
['Ivan Titov', 'Wilker Aziz', 'Nicola De Cao']
2021-09-08
null
https://aclanthology.org/2021.emnlp-main.604
https://aclanthology.org/2021.emnlp-main.604.pdf
emnlp-2021-11
['entity-disambiguation']
['natural-language-processing']
[-2.01166272e-01 3.47623855e-01 6.12392873e-02 -2.28319690e-01 -1.48561323e+00 -6.80420816e-01 6.62674785e-01 2.43670732e-01 -5.91134250e-01 8.48324955e-01 2.39102080e-01 -3.09634924e-01 3.61210227e-01 -6.64866567e-01 -8.31762016e-01 -5.59818864e-01 7.41980746e-02 1.03823507e+00 7.12817386e-02 -2.65044183e-01 -2.04792097e-01 5.68212159e-02 -1.20022690e+00 -3.20336074e-01 1.20794415e+00 5.85029721e-01 4.52423126e-01 5.56297362e-01 -1.14260122e-01 5.87162673e-01 -5.88567555e-01 -9.67450559e-01 -1.61430433e-01 -3.63312960e-01 -6.57602012e-01 -4.10776526e-01 3.26518029e-01 -2.10557431e-01 -4.33032930e-01 1.06552553e+00 7.92072535e-01 1.17132939e-01 5.26127338e-01 -8.70317161e-01 -8.39392424e-01 1.19805372e+00 -5.33056021e-01 9.32200104e-02 2.08249778e-01 -1.42195582e-01 1.32624424e+00 -1.03590667e+00 6.65389001e-01 1.12900567e+00 6.45026982e-01 5.38514495e-01 -1.22680545e+00 -6.55148029e-01 2.41738647e-01 2.70245254e-01 -1.60296524e+00 -5.91482103e-01 5.05715251e-01 -3.35450530e-01 1.26709878e+00 2.68221889e-02 4.11536753e-01 1.26603878e+00 -1.19622335e-01 9.15273488e-01 6.43970668e-01 -2.76722759e-01 5.03168702e-02 -1.32800847e-01 1.14270948e-01 6.96548164e-01 3.74760240e-01 -3.12632143e-01 -2.77540177e-01 -2.23057479e-01 3.89151573e-01 -3.67500514e-01 -1.77775800e-01 -2.73400769e-02 -9.28630054e-01 8.23263168e-01 3.69205654e-01 4.66500580e-01 -5.56056082e-01 3.48806351e-01 3.33701134e-01 -5.68629801e-02 5.47874868e-01 3.88480455e-01 -5.31661510e-01 -3.19985002e-01 -1.07663882e+00 1.79192141e-01 9.75081086e-01 1.11150837e+00 6.93828225e-01 1.06053799e-01 4.80329469e-02 8.87598276e-01 5.58480978e-01 5.18052399e-01 3.08208913e-01 -3.61538380e-01 7.25301802e-01 2.45192736e-01 -9.73941386e-02 -7.63806045e-01 -3.12602937e-01 -6.42899573e-01 -8.14182937e-01 -5.10220408e-01 3.74336898e-01 -3.97746861e-01 -8.17642510e-01 2.04401445e+00 2.61049390e-01 3.45466793e-01 7.98567310e-02 7.30476916e-01 8.85500968e-01 6.25534594e-01 4.80866432e-01 7.81975873e-03 1.52900779e+00 -9.70996380e-01 -7.42814839e-01 -6.19271934e-01 5.91243625e-01 -8.79262388e-01 8.01492631e-01 -4.60351296e-02 -1.28644001e+00 -2.25595266e-01 -9.61157382e-01 -5.71715713e-01 -3.23724300e-01 6.05545580e-01 8.04634988e-01 4.39810187e-01 -1.14346862e+00 2.37093523e-01 -1.02955723e+00 -2.41562292e-01 1.36352303e-02 4.75612998e-01 -2.15252802e-01 3.48848477e-02 -1.48036587e+00 8.98897409e-01 6.92593992e-01 2.20449373e-01 -5.74189723e-01 -6.42788112e-01 -1.12641788e+00 3.77093613e-01 3.93175572e-01 -9.83564377e-01 1.25412941e+00 -3.94297898e-01 -1.56222963e+00 7.72361815e-01 -4.06893849e-01 -7.01157093e-01 5.84311306e-01 -6.72792435e-01 -2.14645863e-01 -1.88400537e-01 1.81164801e-01 6.31599903e-01 1.86370134e-01 -1.13979721e+00 -3.16500515e-01 -2.05851525e-01 -1.52604477e-02 1.48869365e-01 -1.50931537e-01 6.12917840e-02 -9.32889342e-01 -8.02112222e-01 -1.51181836e-02 -1.05398142e+00 -3.33036572e-01 -9.00145829e-01 -7.36409009e-01 -3.51289541e-01 2.07475245e-01 -1.17125416e+00 1.54433346e+00 -1.98435307e+00 3.37680459e-01 1.53649822e-01 1.32791221e-01 5.45520186e-01 8.13788828e-03 6.69514060e-01 2.67756302e-02 2.34333754e-01 -3.06015968e-01 -8.20174158e-01 3.47592950e-01 8.90013576e-02 -2.35968828e-01 2.53262162e-01 2.70595044e-01 1.20240653e+00 -1.06743670e+00 -5.99705994e-01 -2.91244872e-02 8.87753785e-01 -6.31945074e-01 1.96366206e-01 -2.36618847e-01 3.89764845e-01 -3.58712524e-01 4.70026374e-01 5.67900896e-01 -4.53750581e-01 5.30853212e-01 -7.60685205e-02 -5.49576357e-02 8.79894853e-01 -1.15790141e+00 1.79803574e+00 -4.92622077e-01 4.37781245e-01 -1.79183975e-01 -7.85986900e-01 8.85839403e-01 4.63710696e-01 2.31543005e-01 -3.75460267e-01 2.63804793e-01 6.31338000e-01 -2.36846939e-01 -5.28946444e-02 6.93352759e-01 1.40208691e-01 -4.08318907e-01 2.97763675e-01 4.70227480e-01 1.70757055e-01 5.14368892e-01 5.01293480e-01 9.78883445e-01 2.62657821e-01 3.91281068e-01 9.33935866e-03 3.40804636e-01 -3.06410521e-01 7.31032670e-01 6.94051027e-01 3.83877993e-01 4.90578234e-01 5.04018366e-01 5.16343936e-02 -1.03189456e+00 -8.57146263e-01 6.70310706e-02 1.04199553e+00 -8.52235220e-03 -8.14897656e-01 -8.49946439e-01 -5.36282241e-01 -1.26834825e-01 9.46012080e-01 -4.16337371e-01 4.39307429e-02 -9.42064524e-01 -9.93669808e-01 8.28583956e-01 6.42403007e-01 3.37618947e-01 -6.97583616e-01 5.10615483e-03 4.58073914e-01 -4.76286888e-01 -1.32433224e+00 -4.82566684e-01 2.29763463e-01 -5.79853415e-01 -7.02272058e-01 -7.58003354e-01 -6.95261180e-01 5.57897151e-01 -1.96422368e-01 1.44050205e+00 -6.03874028e-02 -3.10553201e-02 1.47741735e-01 -1.91445515e-01 -2.39393383e-01 -4.12722617e-01 6.59618735e-01 -3.07068467e-01 -3.87456656e-01 3.25085163e-01 -5.62309384e-01 -4.13767248e-01 -1.16922498e-01 -6.02647245e-01 1.90309640e-02 8.80192041e-01 9.88852739e-01 7.04801738e-01 -4.64434654e-01 6.00975573e-01 -1.22465408e+00 5.17546117e-01 -6.47707582e-01 -7.53803074e-01 3.47366095e-01 -5.70017159e-01 3.29096764e-01 4.35553282e-01 -2.75078416e-01 -1.08583736e+00 2.20893025e-02 -6.00539982e-01 -3.98988649e-02 1.60967365e-01 7.63328373e-01 -3.13426614e-01 3.23894203e-01 3.24801832e-01 7.44011998e-02 -5.49853802e-01 -7.37764895e-01 7.10136890e-01 4.26999837e-01 6.96737528e-01 -4.76011455e-01 8.72961640e-01 -5.49307391e-02 -2.28171960e-01 -4.32006568e-01 -7.84137607e-01 -4.17749643e-01 -6.77569151e-01 8.70030522e-02 7.99947262e-01 -1.33675814e+00 -5.67402780e-01 4.01808113e-01 -1.46160960e+00 -2.47932732e-01 7.56197050e-02 6.04959249e-01 -3.41707975e-01 3.92814785e-01 -9.73341703e-01 -6.41931295e-01 -7.14459717e-01 -9.49266732e-01 1.18838286e+00 1.36803746e-01 -2.61850148e-01 -9.94132280e-01 1.57119483e-01 3.17306370e-01 3.10536355e-01 9.55005735e-02 7.49287844e-01 -8.27667117e-01 -7.90434957e-01 -3.18123788e-01 -1.96119130e-01 8.31136554e-02 -3.88056427e-01 3.88935469e-02 -8.44072044e-01 -2.37756878e-01 -3.97611707e-01 -6.19251803e-02 9.43798959e-01 2.01694280e-01 6.30937159e-01 -4.73156273e-01 -3.90401155e-01 6.57222629e-01 1.43662345e+00 -1.12519681e-01 5.82555592e-01 3.67792308e-01 8.86190891e-01 9.29878354e-02 5.23080647e-01 3.68657202e-01 8.75635743e-01 8.13938379e-01 8.77678394e-02 -1.11820325e-01 -2.54159212e-01 -5.01762033e-01 3.59223217e-01 1.39190257e+00 -1.85645014e-01 -6.62801147e-01 -1.11980939e+00 6.50478244e-01 -2.03031778e+00 -8.08581889e-01 -3.46127599e-01 2.03419638e+00 1.12867916e+00 -9.91707668e-02 -1.06315821e-01 -2.90979177e-01 6.20551765e-01 6.97084814e-02 -3.37954223e-01 -1.15011618e-01 -2.58703083e-01 2.58526117e-01 6.54373348e-01 7.68090189e-01 -1.24393070e+00 1.28710771e+00 5.59069586e+00 8.47254813e-01 -9.10269022e-01 4.33638841e-01 4.31306899e-01 -5.49135767e-02 -4.47177410e-01 1.08561791e-01 -1.19625771e+00 5.91089547e-01 1.12168145e+00 -2.96361089e-01 2.75590420e-01 8.42726350e-01 -2.99647689e-01 5.09766862e-02 -8.17449570e-01 8.50651026e-01 1.99375451e-01 -9.47388053e-01 -1.52694523e-01 2.24093124e-02 6.73678458e-01 3.81234169e-01 -1.41534865e-01 4.84929591e-01 6.94522023e-01 -7.66817749e-01 7.87413418e-01 3.14073861e-01 6.81266308e-01 -7.78016508e-01 9.23560381e-01 1.98525026e-01 -1.23120868e+00 2.37422943e-01 -2.75193423e-01 4.91279393e-01 5.72783530e-01 8.95573437e-01 -8.50304365e-01 7.45617211e-01 3.18566322e-01 5.14099717e-01 -4.09427404e-01 9.02551711e-01 -9.12961185e-01 7.55376041e-01 -4.50760216e-01 3.32534574e-02 3.19241792e-01 -1.41246423e-01 6.04888439e-01 1.63676238e+00 5.42644620e-01 2.66449936e-02 -4.47433703e-02 7.91341484e-01 -3.98427308e-01 2.51956761e-01 -2.66045213e-01 -9.93404016e-02 7.37803161e-01 1.26708198e+00 -5.05898118e-01 -4.08909500e-01 -3.83531272e-01 1.00005043e+00 7.19627559e-01 3.42319161e-01 -1.07575655e+00 -3.47676277e-01 3.54720533e-01 -1.60077825e-01 4.91594106e-01 -5.44137239e-01 -1.12693407e-01 -1.32113147e+00 4.97515090e-02 -6.81842387e-01 4.65556920e-01 -5.10785520e-01 -1.03064847e+00 8.25124264e-01 -2.12689176e-01 -7.77782798e-01 -7.35275984e-01 -1.83851093e-01 -2.13798314e-01 1.04829144e+00 -1.73173618e+00 -1.23469305e+00 -6.43122867e-02 1.41823560e-01 2.22459078e-01 6.35063648e-02 8.60862315e-01 8.90430152e-01 -8.17910254e-01 8.51588726e-01 1.28960952e-01 3.54001552e-01 9.19242382e-01 -1.50185633e+00 8.96134794e-01 1.17360914e+00 3.92228395e-01 7.90709257e-01 7.33584404e-01 -6.68712020e-01 -1.39875865e+00 -1.03361654e+00 1.66803253e+00 -2.76587933e-01 8.40570033e-01 -6.23871922e-01 -9.91749704e-01 1.05544341e+00 3.19746107e-01 -5.58660030e-01 7.61848271e-01 3.97008181e-01 -3.57929766e-01 2.21636891e-01 -6.21406198e-01 5.38309753e-01 9.38171804e-01 -4.75506127e-01 -5.10056734e-01 3.53916824e-01 5.59759915e-01 -7.72641063e-01 -9.66084301e-01 1.82325855e-01 3.61409187e-01 -4.61281836e-01 9.06911194e-01 -2.84282207e-01 3.26990724e-01 -2.49714643e-01 1.20335452e-01 -1.23843491e+00 -3.47836345e-01 -9.23636973e-01 -4.80294645e-01 1.61829722e+00 7.33341694e-01 -7.96053767e-01 3.39917183e-01 6.00705445e-01 -2.32561156e-01 -8.13380837e-01 -8.05795312e-01 -7.34691024e-01 -8.33356082e-02 -2.18982056e-01 5.90849102e-01 7.94193029e-01 -5.66803217e-02 7.96727419e-01 -5.13071775e-01 4.20519263e-01 5.42343080e-01 6.40311763e-02 6.92813098e-01 -1.06803274e+00 -5.73917687e-01 -2.99015075e-01 -2.51966536e-01 -1.44818950e+00 3.46653819e-01 -1.01109993e+00 1.93006024e-01 -1.69575918e+00 1.95315331e-01 -6.01917505e-01 -7.89981261e-02 6.69606447e-01 -6.60970092e-01 1.41001314e-01 2.59736419e-01 3.25791150e-01 -7.03233838e-01 5.85024178e-01 6.35568917e-01 1.40314847e-01 -3.36149558e-02 -2.83799648e-01 -7.84460306e-01 5.70437431e-01 6.84613466e-01 -5.57318330e-01 -3.94286476e-02 -7.47070491e-01 5.38155258e-01 1.25115952e-02 2.76240259e-01 -6.98423088e-01 5.09019971e-01 4.83409643e-01 4.54977080e-02 -6.29640579e-01 4.70437855e-01 -2.14737266e-01 2.97704816e-01 2.38956317e-01 -2.03859210e-01 2.00465918e-01 1.20460175e-01 4.88085121e-01 -3.91540468e-01 -3.30021501e-01 4.79456604e-01 6.35074601e-02 -6.75792575e-01 2.54096836e-01 -9.78092253e-02 2.65250593e-01 7.32381999e-01 4.39053774e-01 -4.19686794e-01 -4.17866468e-01 -6.17166996e-01 2.94293106e-01 3.32654148e-01 4.01687473e-01 3.80479731e-02 -1.23356891e+00 -9.86647487e-01 -1.67779624e-01 -1.75897300e-01 5.42058013e-02 2.72317290e-01 9.44667697e-01 -3.41573924e-01 5.82526624e-01 4.12692100e-01 -1.59892380e-01 -1.20366573e+00 3.75381470e-01 -1.02276234e-02 -7.41715312e-01 -6.25571728e-01 9.51098442e-01 -6.06951118e-02 -2.63380200e-01 2.44413558e-02 -3.33009660e-02 -8.28217939e-02 2.82419026e-01 2.56579429e-01 2.92614400e-01 1.88834682e-01 -7.93229759e-01 -5.19238889e-01 3.80334884e-01 -2.16738582e-01 -1.26899987e-01 1.46785092e+00 -1.43313721e-01 -7.37274289e-02 2.55261987e-01 1.09024620e+00 2.23471582e-01 -9.17538285e-01 -4.04303014e-01 2.81228364e-01 -2.85128201e-03 -2.98975799e-02 -7.17036724e-01 -1.01517475e+00 8.23278368e-01 8.05150717e-02 -8.70365184e-03 8.70077193e-01 3.18232507e-01 1.10325289e+00 2.36704364e-01 3.14192712e-01 -9.57107842e-01 -4.99740064e-01 8.01252067e-01 6.26945436e-01 -1.04935300e+00 -9.57694352e-02 -6.78418338e-01 -7.14551330e-01 8.57753336e-01 2.24514306e-01 -6.80735037e-02 3.76601070e-01 4.98008728e-01 3.48798744e-02 -4.52662669e-02 -7.90793061e-01 -4.27554041e-01 4.20979202e-01 1.55667961e-01 8.43959868e-01 1.72266260e-01 -6.05131984e-01 8.35773170e-01 -4.40377712e-01 -1.49833068e-01 1.42031774e-01 5.17832398e-01 -8.78063589e-02 -1.34857452e+00 1.08504757e-01 1.45852014e-01 -7.31988549e-01 -6.46764696e-01 -1.89420149e-01 7.51948714e-01 -1.58470899e-01 8.39084506e-01 -4.56229933e-02 -3.92831713e-02 2.50004381e-01 2.36683905e-01 2.82764435e-01 -5.91756523e-01 -5.58761179e-01 3.05473119e-01 2.68497497e-01 -4.03556615e-01 -1.80982590e-01 -8.26133370e-01 -1.26033008e+00 -3.14349413e-01 -5.06710470e-01 2.90386885e-01 7.49021411e-01 8.36543560e-01 7.28639603e-01 5.99232912e-01 1.83570936e-01 -6.87712193e-01 -4.75881428e-01 -1.08592689e+00 -2.95361131e-01 2.79238045e-01 -1.00913189e-01 -5.74052632e-01 -9.82661769e-02 -7.00148791e-02]
[9.529861450195312, 8.946061134338379]
e346d4dd-a9e1-439f-b6d4-41c4fad2caa1
cyber-attack-detection-in-socio-technical
2103.11422
null
https://arxiv.org/abs/2103.11422v1
https://arxiv.org/pdf/2103.11422v1.pdf
Cyber-Attack Detection in Socio-Technical Transportation Systems Exploiting Redundancies Between Physical and Social Data
Cyber-physical-social connectivity is a key element in Intelligent Transportation Systems (ITSs) due to the ever-increasing interaction between human users and technological systems. Such connectivity translates the ITSs into dynamical systems of socio-technical nature. Exploiting this socio-technical feature to our advantage, we propose a cyber-attack detection scheme for ITSs that focuses on cyber-attacks on freeway traffic infrastructure. The proposed scheme combines two parallel macroscopic traffic model-based Partial Differential Equation (PDE) filters whose output residuals are compared to make decision on attack occurrences. One of the filters utilizes physical (vehicle/infrastructure) sensor data as feedback whereas the other utilizes social data from human users' mobile devices as feedback. The Social Data-based Filter is aided by a fake data isolator and a social signal processor that translates the social information into usable feedback signals. Mathematical convergence properties are analyzed for the filters using Lyapunov's stability theory. Lastly, we validate our proposed scheme by presenting simulation results.
['Satadru Dey', 'Sara Sattarzadeh', 'Tanushree Roy']
2021-03-21
null
null
null
null
['cyber-attack-detection']
['miscellaneous']
[ 5.03110047e-03 3.49601865e-01 1.53136045e-01 3.54242176e-01 -2.44049802e-01 -6.17011309e-01 6.73080385e-01 -2.81429626e-02 -3.39383066e-01 8.86222541e-01 -7.48535916e-02 -7.62762487e-01 -1.86391622e-01 -9.40751255e-01 -6.49826407e-01 -7.54290998e-01 -7.83963799e-02 -3.22599530e-01 7.87442029e-01 -4.74208891e-01 2.49735355e-01 5.74890673e-01 -1.42697799e+00 -2.74611205e-01 9.86083388e-01 1.11390305e+00 -2.85352319e-01 1.14285016e+00 3.72316092e-01 4.91724372e-01 -5.86330414e-01 4.29135151e-02 4.37509388e-01 -1.41705438e-01 4.24335375e-02 -2.31141552e-01 -2.90962040e-01 -1.32379189e-01 -5.20784140e-01 1.24569166e+00 1.81313097e-01 3.60488109e-02 7.28278756e-01 -1.97388279e+00 -2.44223058e-01 6.85104309e-03 -3.12031716e-01 1.57096684e-01 3.92667770e-01 6.65975392e-01 6.52295798e-02 -4.59641188e-01 5.34933567e-01 1.31793261e+00 6.98870242e-01 3.01539570e-01 -1.00769699e+00 -9.03171718e-01 -1.14447556e-01 1.29917890e-01 -1.27429855e+00 -2.52979249e-01 8.68275344e-01 -7.11317360e-01 5.92834115e-01 5.53005338e-01 8.18268538e-01 1.01254380e+00 9.98181999e-01 2.74182349e-01 1.05186462e+00 3.61449458e-02 4.39302117e-01 6.20146811e-01 4.34368908e-01 5.86162388e-01 6.84639454e-01 3.80505323e-01 -1.67187657e-02 -4.48236585e-01 3.97058755e-01 -5.99800274e-02 -1.08059764e-01 2.60509282e-01 -6.31116688e-01 3.13628584e-01 3.47148955e-01 2.28234753e-01 -7.94555068e-01 2.50978976e-01 3.43138427e-01 7.64572442e-01 3.72858733e-01 -6.96072876e-02 -1.25100911e-01 -1.70010939e-01 -2.30361685e-01 -3.36177759e-02 8.34399819e-01 8.55179012e-01 5.10593653e-01 1.18817791e-01 2.41702184e-01 -8.89092386e-02 6.48490012e-01 1.15487826e+00 -2.43488874e-04 -6.84515119e-01 3.89769971e-01 6.44410253e-01 4.17531163e-01 -1.69503617e+00 -4.51589733e-01 -2.67273545e-01 -9.02665019e-01 3.45786393e-01 2.74298936e-01 -5.77458918e-01 -2.35971406e-01 1.52116144e+00 6.84707165e-01 6.23039126e-01 1.45044908e-01 7.33700991e-01 7.91575089e-02 7.00786471e-01 -2.64512189e-02 -4.30158913e-01 1.13738894e+00 -1.16976023e-01 -9.35551465e-01 5.95110595e-01 5.37953138e-01 -5.52489877e-01 4.31012869e-01 2.46951193e-01 -8.33458841e-01 -2.54257351e-01 -1.20603704e+00 5.72439969e-01 -9.03863788e-01 -1.99625522e-01 -2.90733337e-01 1.08224165e+00 -8.77919436e-01 3.15647364e-01 -6.44018412e-01 -5.96479952e-01 9.19406042e-02 4.40794587e-01 -4.60587107e-02 7.83657908e-01 -1.63829565e+00 9.69287634e-01 -2.82717794e-01 2.02082530e-01 -4.51467156e-01 -7.55379975e-01 -4.25511569e-01 -2.01140672e-01 1.78758994e-01 -6.50144160e-01 5.96815109e-01 -6.60304070e-01 -1.79073811e+00 6.86512589e-02 2.59925038e-01 -6.53897107e-01 9.89834130e-01 1.85403466e-01 -9.55417812e-01 3.73445362e-01 -2.16938913e-01 -1.73093781e-01 9.31333899e-01 -1.29418099e+00 -6.60336912e-01 -1.48238719e-01 -2.89290935e-01 -3.32472712e-01 -4.05446351e-01 -3.25733662e-01 4.57087606e-01 -7.29604736e-02 -2.31368288e-01 -1.16520810e+00 -1.84048772e-01 4.40901406e-02 -7.96417475e-01 9.21588987e-02 1.61380923e+00 -4.33399826e-01 1.21121538e+00 -1.96235847e+00 -3.47915560e-01 7.57150948e-01 2.55709708e-01 7.50046074e-01 -3.60441506e-02 1.01765954e+00 2.58657336e-01 1.23326153e-01 1.84261799e-01 -3.15750800e-02 -2.16158956e-01 -2.55053520e-01 -5.74722707e-01 9.26066935e-01 2.35695407e-01 4.47984904e-01 -1.00009000e+00 -1.39310881e-01 6.11040652e-01 4.81582135e-01 -4.44794416e-01 -1.35196432e-01 8.97583961e-02 7.36458957e-01 -9.29435909e-01 1.99985951e-01 9.29866612e-01 4.13498104e-01 -1.63299158e-01 2.74043158e-02 -6.52240396e-01 -6.30197451e-02 -1.11104596e+00 5.43698430e-01 -3.80410939e-01 7.44759619e-01 4.85426098e-01 -7.92090237e-01 7.09267020e-01 4.24906522e-01 6.85410082e-01 -6.01198614e-01 6.74384475e-01 2.92283267e-01 -3.06785628e-02 -8.32652330e-01 3.93459678e-01 2.41268903e-01 -1.76966727e-01 5.12023270e-01 -4.56082642e-01 5.73536195e-02 -2.62883216e-01 3.33351493e-01 1.61561680e+00 -5.53200483e-01 -9.30857658e-02 -5.44247806e-01 1.04595888e+00 1.27853770e-02 2.90680617e-01 4.53005314e-01 -6.21626437e-01 -3.99213314e-01 4.65519726e-01 -4.73073348e-02 -7.81641543e-01 -8.67472887e-01 1.10442713e-01 -1.31991450e-02 5.35348892e-01 4.96052252e-03 -8.43930662e-01 -4.81492132e-01 4.05467898e-01 6.09342158e-01 -6.42997503e-01 -7.22481906e-01 -1.76871374e-01 -4.80938815e-02 7.23906875e-01 -1.44390255e-01 4.65484232e-01 -3.77961367e-01 -5.05752802e-01 4.17087495e-01 2.45852947e-01 -1.06652236e+00 -2.89874554e-01 -6.01161599e-01 -3.62837374e-01 -1.30911887e+00 -5.30968904e-02 9.56425816e-03 7.96264350e-01 7.15283096e-01 5.28336726e-02 1.30742058e-01 -4.18641418e-02 8.54005575e-01 -2.38524508e-02 -6.30395770e-01 -9.65678394e-01 -2.26154774e-01 6.13528550e-01 8.90449882e-01 9.76117402e-02 -7.48396158e-01 -7.91730821e-01 5.88542998e-01 -8.10731351e-01 -1.76591992e-01 2.14395642e-01 2.92215019e-01 -2.45334223e-01 3.98669988e-01 9.17625070e-01 -4.94657308e-01 8.23596358e-01 -1.07108140e+00 -1.01221883e+00 -1.43217891e-01 -3.80572379e-01 -2.37505570e-01 8.85117114e-01 -5.06015539e-01 -8.11628699e-01 -3.54886144e-01 3.92250478e-01 -1.82672918e-01 -3.76364514e-02 3.47705662e-01 -2.33871415e-01 -3.73383820e-01 6.01709247e-01 4.27318225e-03 4.36548173e-01 4.51242924e-02 2.44803846e-01 1.21394813e+00 2.10707664e-01 -4.80606854e-02 1.22965300e+00 8.73923898e-01 4.80304658e-01 -1.67847848e+00 2.34467313e-01 -2.93374330e-01 -1.40692547e-01 -1.09462023e+00 5.64651370e-01 -4.95610058e-01 -1.59370184e+00 7.66169429e-01 -1.37864292e+00 5.66588566e-02 -4.93406318e-02 5.40464699e-01 -3.12288523e-01 2.55463235e-02 -4.85924691e-01 -1.61239994e+00 1.03138395e-01 -6.87597990e-01 7.53889918e-01 1.67834684e-01 8.74111056e-03 -1.06763113e+00 2.63243794e-01 1.78682148e-01 5.90499878e-01 7.60359347e-01 3.28302115e-01 -2.34102562e-01 -7.84024477e-01 -7.67589450e-01 -1.90677315e-01 3.00408751e-01 7.95950741e-02 3.98906380e-01 -7.26564944e-01 -2.75440812e-01 1.92378283e-01 4.32567120e-01 2.79956669e-01 2.61983186e-01 1.65599406e-01 -4.90448534e-01 -5.63217103e-01 -1.04228377e-01 1.38954329e+00 4.20998275e-01 6.72894835e-01 -1.64334953e-01 4.94786561e-01 8.89962256e-01 5.80445409e-01 3.39258254e-01 3.72678936e-01 2.95073330e-01 5.89110494e-01 -8.10734928e-02 1.69809550e-01 -2.55237728e-01 8.67195249e-01 8.48280549e-01 1.60900667e-01 -4.76351410e-01 -6.31895602e-01 2.94866621e-01 -1.92585504e+00 -8.81353319e-01 -9.88145828e-01 2.33797956e+00 1.32718179e-02 3.26385409e-01 2.74790287e-01 4.37574923e-01 1.11902344e+00 -2.44003788e-01 -5.14517903e-01 -4.30511057e-01 2.33020574e-01 -4.86618966e-01 1.10684490e+00 7.89073110e-01 -5.44311523e-01 4.32572573e-01 5.09658051e+00 5.03257573e-01 -1.16362226e+00 1.98269069e-01 -3.33128795e-02 3.60747308e-01 -1.54519185e-01 -7.86052123e-02 -3.69928867e-01 8.44688714e-01 1.45927417e+00 -3.74981910e-01 8.42643604e-02 2.30299219e-01 1.25443077e+00 -5.44580102e-01 -3.07204753e-01 3.18404853e-01 -5.68930626e-01 -1.11831701e+00 -3.28179061e-01 5.41499317e-01 5.69531024e-01 -1.66475326e-01 1.60137638e-01 -2.47639641e-01 2.06059754e-01 -2.09555432e-01 6.45613074e-01 9.66551840e-01 4.17597055e-01 -7.57540762e-01 4.53306943e-01 5.47876179e-01 -1.41411889e+00 -3.18248034e-01 5.19001894e-02 -2.95028865e-01 4.99262869e-01 6.31731033e-01 -4.81268346e-01 5.99699795e-01 -1.04239635e-01 6.38862967e-01 -1.78594768e-01 7.53912389e-01 1.66496202e-01 8.30865622e-01 -6.49360836e-01 -3.53447795e-01 1.22853808e-01 -4.44741488e-01 1.34906209e+00 9.14737165e-01 2.46976465e-01 2.90865690e-01 -1.01416700e-01 7.70408332e-01 5.29569089e-01 -2.85713822e-01 -1.26719654e+00 -7.98224509e-02 5.95387518e-01 9.69804108e-01 -5.54148853e-01 -3.35096121e-01 -2.05736265e-01 4.42034841e-01 -4.20756787e-01 5.38762033e-01 -1.00975585e+00 -6.31568968e-01 1.12363768e+00 6.12189054e-01 -3.09461743e-01 -5.91989160e-01 -2.94544756e-01 -8.34317327e-01 -2.86161285e-02 -7.54040554e-02 -1.11745574e-01 -4.81023937e-01 -1.00595045e+00 1.74066514e-01 -2.96420842e-01 -1.73575425e+00 2.32212722e-01 -2.95260668e-01 -6.64752603e-01 6.25922143e-01 -1.15250993e+00 -9.07659233e-01 -1.38370067e-01 7.02368498e-01 -1.19926564e-01 -1.36748299e-01 2.63857961e-01 4.70470726e-01 -7.67914534e-01 1.82471693e-01 1.22774117e-01 -1.72252253e-01 1.89808294e-01 -5.97396553e-01 3.20153981e-01 1.10581422e+00 -9.11858976e-01 5.50434411e-01 1.01716149e+00 -1.09585428e+00 -1.90998852e+00 -1.07235289e+00 7.58994102e-01 -2.98116416e-01 1.19128156e+00 -3.94463837e-01 -5.68480551e-01 1.86642945e-01 2.73345381e-01 -8.52358639e-02 3.09518486e-01 -7.77045190e-01 -1.06542893e-02 -2.39536092e-01 -1.58963990e+00 7.51663268e-01 8.42678845e-01 -5.08360267e-01 -1.03999667e-01 1.61727697e-01 5.87884545e-01 1.42020598e-01 -6.12045109e-01 4.89623100e-02 6.40033126e-01 -6.99665189e-01 5.33091247e-01 -5.86727440e-01 -3.15865725e-01 -7.00308502e-01 1.48905724e-01 -1.36582386e+00 1.31212577e-01 -1.40566123e+00 1.44898042e-01 1.04696620e+00 1.42286524e-01 -1.66981888e+00 4.10609722e-01 7.78372347e-01 1.14377238e-01 -1.37821808e-01 -1.19473445e+00 -1.11402512e+00 -2.88474947e-01 -5.65265417e-01 3.01438272e-01 7.91281521e-01 4.01720822e-01 2.99397800e-02 -3.45290244e-01 7.57176697e-01 8.08150232e-01 -8.45408618e-01 9.93259013e-01 -1.27367413e+00 2.36041814e-01 -3.20679456e-01 -8.46497059e-01 -6.06268942e-01 -9.07324348e-03 -4.80596811e-01 -1.08166121e-01 -8.81665111e-01 -6.07856393e-01 -2.85340130e-01 -3.00768971e-01 -2.85706818e-01 3.97546202e-01 7.67996758e-02 6.75842613e-02 -4.25740518e-02 -1.30455405e-01 5.15674829e-01 9.87483501e-01 -2.68108565e-02 -5.15063822e-01 2.31800899e-01 -1.60774305e-01 3.93922120e-01 7.92500317e-01 -6.77687168e-01 -5.58550298e-01 3.43192905e-01 5.26298508e-02 4.70476002e-01 7.15032935e-01 -1.35553586e+00 6.68437719e-01 -2.45269969e-01 -5.59327662e-01 -5.01028240e-01 1.08031519e-01 -1.33639383e+00 4.10320878e-01 1.21965873e+00 5.53370900e-02 -9.41393375e-02 1.07877903e-01 1.19319236e+00 8.97086337e-02 7.38902271e-01 9.18450654e-01 6.84704185e-01 2.18980804e-01 4.28652987e-02 -1.43032074e+00 -4.15440083e-01 1.52642035e+00 -4.38838333e-01 -7.09676445e-01 -5.98822534e-01 -5.13377845e-01 4.13209856e-01 1.91024020e-01 5.04943252e-01 3.59974861e-01 -1.15666187e+00 -3.25115800e-01 3.07233006e-01 -3.00504178e-01 -7.80618012e-01 4.09811288e-01 1.30591691e+00 -3.41532350e-01 5.79443634e-01 -2.69154429e-01 -5.64410567e-01 -1.10506058e+00 3.43971133e-01 3.13400894e-01 2.82280952e-01 -2.18912512e-01 1.24714710e-01 -4.75083351e-01 -6.78366050e-02 -3.55675220e-02 -5.09673715e-01 -1.41084613e-02 1.17171578e-01 4.20924038e-01 1.05964708e+00 -1.68509692e-01 -8.76976848e-01 -2.95112580e-01 4.64195430e-01 5.60642481e-01 -4.84388411e-01 8.81041408e-01 -5.84717691e-01 2.62433384e-02 4.98494685e-01 1.31032634e+00 1.32873073e-01 -1.20329225e+00 2.06121076e-02 -6.34477064e-02 -3.68916243e-01 1.06054239e-01 -4.18123662e-01 -8.87982547e-01 4.75483924e-01 5.11300087e-01 1.14412487e+00 8.22559118e-01 -7.55860448e-01 1.15968609e+00 2.01556355e-01 7.02955306e-01 -1.26984537e+00 -2.25093648e-01 3.23934495e-01 7.35214591e-01 -8.39649320e-01 -3.87617499e-01 -8.10155749e-01 -4.99700978e-02 8.80966187e-01 2.14381903e-01 -7.29903460e-01 1.48434341e+00 3.32078546e-01 -6.95941001e-02 -1.25057802e-01 -8.77808332e-01 -2.49675572e-01 -1.35130510e-01 7.07637250e-01 -5.87210834e-01 2.03130960e-01 -6.34754241e-01 4.45778966e-01 1.32424265e-01 7.96291381e-02 9.46195066e-01 9.28339303e-01 -5.09466946e-01 -8.02541912e-01 -5.76770782e-01 1.73141941e-01 1.67603225e-01 6.05151653e-01 -4.98239875e-01 6.64465129e-01 4.10510320e-03 1.58378828e+00 -1.05499730e-01 -9.98356938e-01 6.40046597e-01 -1.93190187e-01 -1.47388041e-01 2.39978954e-02 -7.60420144e-01 -4.26426500e-01 5.27478606e-02 -7.71857202e-01 -1.63886711e-01 -7.89834976e-01 -1.14791596e+00 -7.95406163e-01 -4.70149666e-01 3.11384887e-01 1.05108500e+00 9.81904328e-01 8.82477701e-01 4.78147358e-01 1.35734844e+00 -6.76066339e-01 -2.13902771e-01 -6.14921093e-01 -5.33160508e-01 8.17128941e-02 7.05036998e-01 -8.54307413e-01 -8.37258518e-01 -2.86474437e-01]
[5.551424980163574, 1.6717194318771362]
bbca93b5-56ef-46a1-8c31-f1f5b750721f
a-new-target-specific-object-proposal
1803.10098
null
http://arxiv.org/abs/1803.10098v1
http://arxiv.org/pdf/1803.10098v1.pdf
A New Target-specific Object Proposal Generation Method for Visual Tracking
Object proposal generation methods have been widely applied to many computer vision tasks. However, existing object proposal generation methods often suffer from the problems of motion blur, low contrast, deformation, etc., when they are applied to video related tasks. In this paper, we propose an effective and highly accurate target-specific object proposal generation (TOPG) method, which takes full advantage of the context information of a video to alleviate these problems. Specifically, we propose to generate target-specific object proposals by integrating the information of two important objectness cues: colors and edges, which are complementary to each other for different challenging environments in the process of generating object proposals. As a result, the recall of the proposed TOPG method is significantly increased. Furthermore, we propose an object proposal ranking strategy to increase the rank accuracy of the generated object proposals. The proposed TOPG method has yielded significant recall gain (about 20%-60% higher) compared with several state-of-the-art object proposal methods on several challenging visual tracking datasets. Then, we apply the proposed TOPG method to the task of visual tracking and propose a TOPG-based tracker (called as TOPGT), where TOPG is used as a sample selection strategy to select a small number of high-quality target candidates from the generated object proposals. Since the object proposals generated by the proposed TOPG cover many hard negative samples and positive samples, these object proposals can not only be used for training an effective classifier, but also be used as target candidates for visual tracking. Experimental results show the superior performance of TOPGT for visual tracking compared with several other state-of-the-art visual trackers (about 3%-11% higher than the winner of the VOT2015 challenge in term of distance precision).
['Hong-Yuan Mark Liao', 'Yan Yan', 'Hanzi Wang', 'Guanjun Guo', 'Bo Li']
2018-03-27
null
null
null
null
['object-proposal-generation']
['computer-vision']
[-7.09012896e-03 -4.88700390e-01 -2.60392457e-01 -1.33775726e-01 -6.05270863e-01 -2.87965477e-01 6.16274953e-01 -3.55311893e-02 -4.52831686e-01 5.93250096e-01 -4.26105484e-02 1.10383511e-01 8.48094076e-02 -5.02421856e-01 -5.58688939e-01 -8.19328785e-01 2.44119659e-01 3.20973426e-01 1.02919745e+00 -5.00791445e-02 2.80020297e-01 3.58903348e-01 -1.72225368e+00 8.48347247e-02 9.49016571e-01 1.09593105e+00 4.60031509e-01 3.39423604e-02 -1.99023172e-01 2.53568053e-01 -6.36670232e-01 -1.50813147e-01 3.56576860e-01 -2.08326325e-01 -7.18885213e-02 1.29083306e-01 7.44475663e-01 -4.11001354e-01 -1.06823415e-01 1.21100175e+00 5.11748433e-01 1.14822067e-01 5.58672786e-01 -1.43708897e+00 -2.79719293e-01 3.26388478e-01 -7.40305960e-01 2.58732885e-01 3.37690972e-02 2.86749393e-01 7.09569752e-01 -1.27138877e+00 6.22232020e-01 1.40832388e+00 4.37664509e-01 6.72867119e-01 -9.08725262e-01 -1.14103901e+00 4.77402449e-01 3.18776250e-01 -1.45837831e+00 -3.78985941e-01 7.71214604e-01 -6.18383884e-01 2.15197682e-01 2.71427274e-01 8.43215764e-01 7.89422512e-01 6.23876862e-02 1.03358305e+00 9.22932744e-01 -1.79348648e-01 1.30897127e-02 3.37606013e-01 1.91541597e-01 6.44744217e-01 7.01159656e-01 3.29110771e-01 -3.59630525e-01 -2.39875704e-01 6.19072318e-01 9.07156840e-02 -5.34777224e-01 -6.05907738e-01 -1.36365318e+00 6.03611112e-01 7.87760615e-01 2.49081403e-01 -2.68433601e-01 9.28168893e-02 2.62100041e-01 -3.73917729e-01 2.64033437e-01 1.40472949e-01 -7.00800121e-02 4.05884564e-01 -8.00011814e-01 4.15285856e-01 1.31553337e-01 1.00223005e+00 4.67339635e-01 1.25559017e-01 -1.04561853e+00 6.58216536e-01 7.40560174e-01 7.08897948e-01 4.72586125e-01 -4.34532434e-01 5.58116019e-01 7.04643488e-01 5.93656421e-01 -1.11297560e+00 -2.06580907e-01 -6.06379867e-01 -5.43594778e-01 1.86242774e-01 3.46928507e-01 1.49161980e-01 -1.14648056e+00 1.66197574e+00 7.75908768e-01 3.58528614e-01 -1.83017090e-01 1.22603428e+00 1.16557503e+00 7.65715301e-01 2.33307019e-01 -4.07579303e-01 1.52534783e+00 -9.58666742e-01 -6.15575612e-01 -2.37554044e-01 2.90466309e-01 -9.30227339e-01 5.93083739e-01 1.66486070e-01 -5.93858838e-01 -9.49985445e-01 -9.29329395e-01 6.07943892e-01 6.39789701e-02 7.14821994e-01 4.14473563e-01 5.64614832e-01 -7.27505326e-01 1.29383400e-01 -5.93185663e-01 -3.87639701e-01 6.24688268e-01 2.61058599e-01 -4.66161221e-02 -2.65245676e-01 -7.92850792e-01 7.88478255e-01 7.88390875e-01 2.81782895e-01 -1.03921819e+00 -5.23310244e-01 -5.48619986e-01 -1.49501547e-01 6.18944824e-01 -5.23652077e-01 9.97524142e-01 -6.54079139e-01 -1.03359449e+00 3.87414098e-01 -1.35532960e-01 -1.76111907e-01 5.59129775e-01 -1.96277961e-01 -4.33486670e-01 1.10564874e-02 1.86350137e-01 9.13580358e-01 1.02482831e+00 -1.27842486e+00 -1.04597843e+00 -1.26564145e-01 -2.97677368e-01 1.38961151e-01 -2.74833500e-01 1.33461982e-01 -8.93951595e-01 -8.39284062e-01 2.33234599e-01 -1.07116473e+00 -1.43584982e-01 3.90022248e-01 -2.39681229e-01 -6.61597490e-01 1.15582156e+00 -5.17137527e-01 1.09688234e+00 -2.05186105e+00 -9.44325775e-02 6.11980855e-02 2.10448727e-01 8.55123699e-01 -1.93573385e-01 -5.64389564e-02 3.47893387e-01 -9.90015194e-02 1.45471439e-01 -1.89574212e-02 -1.57934949e-01 -2.52797097e-01 -5.46074025e-02 3.22607100e-01 2.80649871e-01 7.97690809e-01 -1.07786214e+00 -7.85361528e-01 4.08087671e-01 3.44930261e-01 -2.06239074e-01 2.08936766e-01 -3.15580517e-01 4.17855412e-01 -8.89838696e-01 7.73348868e-01 8.47518742e-01 -6.30006194e-02 -2.66365390e-02 -6.45910680e-01 -8.54808018e-02 -1.77809760e-01 -1.20959246e+00 1.04206944e+00 1.63803637e-01 4.43896919e-01 -2.49183655e-01 -5.82801342e-01 1.19383109e+00 2.51761168e-01 4.32113737e-01 -4.86528665e-01 2.36642912e-01 2.87442863e-01 2.48582959e-01 -3.43160391e-01 5.49978316e-01 4.59227636e-02 2.81149417e-01 -1.82886759e-03 -1.23275973e-01 3.38318497e-01 3.51329803e-01 2.57708281e-01 6.46262944e-01 3.16617817e-01 2.03721792e-01 -9.21027958e-02 7.26494312e-01 9.19534937e-02 1.10917377e+00 6.80900156e-01 -5.94688773e-01 4.31091309e-01 -1.56879023e-01 -3.61256719e-01 -8.64203095e-01 -8.10909867e-01 -2.73732189e-02 5.99260032e-01 6.61959767e-01 -2.07809642e-01 -3.98232728e-01 -8.69359851e-01 3.46309394e-02 3.48981470e-01 -3.72372001e-01 -2.22684637e-01 -6.61884844e-01 -6.80777192e-01 1.91532329e-01 5.89661181e-01 6.62623465e-01 -1.23948014e+00 -5.28166711e-01 2.61158526e-01 -2.68777728e-01 -1.17356884e+00 -7.47418940e-01 -4.95412558e-01 -9.39829409e-01 -1.10669267e+00 -9.69877124e-01 -7.19037950e-01 7.78325617e-01 7.86367238e-01 7.02433169e-01 3.73251468e-01 -2.00827658e-01 1.09352484e-01 -4.36951011e-01 -4.87463415e-01 -2.41824344e-01 -2.28241906e-01 2.66926557e-01 3.89898151e-01 2.19265670e-01 2.80055732e-01 -7.51847923e-01 6.84001327e-01 -6.09283805e-01 1.11598223e-01 9.03877556e-01 8.14466834e-01 5.98621845e-01 -1.34615526e-01 6.34948075e-01 -4.60467070e-01 1.70427263e-01 -1.24679461e-01 -1.11976910e+00 4.02846426e-01 -4.29095477e-01 -7.75872469e-02 3.40232134e-01 -8.74753714e-01 -1.08437920e+00 2.02376559e-01 2.28192031e-01 -7.37284899e-01 1.94820240e-01 1.04475833e-01 -2.68232465e-01 -4.11157697e-01 3.57216567e-01 2.77065784e-01 -7.49902427e-02 -4.82015342e-01 1.94966272e-01 4.56572801e-01 3.10197264e-01 -3.51023108e-01 1.15648437e+00 2.98983008e-01 2.20231358e-02 -5.10192692e-01 -6.54539287e-01 -6.93509400e-01 -2.27057680e-01 -6.20797753e-01 7.71282792e-01 -9.75981712e-01 -6.65309250e-01 4.45206493e-01 -1.11534822e+00 2.70498723e-01 1.22438781e-01 8.91202807e-01 -7.54397213e-02 4.15620565e-01 -1.37914240e-01 -1.07488084e+00 -6.04063630e-01 -1.30763018e+00 1.06198204e+00 6.84650242e-01 4.34645623e-01 -4.60255206e-01 -4.13497508e-01 2.91688740e-01 3.46438736e-01 1.94199502e-01 2.86257058e-01 -5.23612857e-01 -1.19053626e+00 -2.87483901e-01 -4.66086715e-01 1.26880094e-01 4.86966297e-02 -2.43590660e-02 -7.07238078e-01 -7.09385216e-01 -3.60930115e-01 -4.65204827e-02 1.01423872e+00 5.30242324e-01 6.95768535e-01 -1.61082223e-01 -9.56063628e-01 3.15831274e-01 1.46058345e+00 3.41995150e-01 3.43149662e-01 3.27483058e-01 8.17165196e-01 3.48270148e-01 1.53642094e+00 3.74484181e-01 2.15399191e-02 1.37076604e+00 4.55087572e-01 9.63719934e-02 -3.50680530e-01 -1.95114121e-01 5.53318918e-01 4.05383199e-01 -3.81359346e-02 -1.23769991e-01 -5.55288851e-01 6.50475442e-01 -2.13848376e+00 -9.39096451e-01 -3.63793135e-01 2.34970760e+00 5.07760048e-01 2.99032778e-01 3.98517251e-01 -2.19027147e-01 1.17345524e+00 -4.33940925e-02 -4.28666860e-01 4.68097717e-01 1.78292379e-01 -1.82043090e-01 4.98482138e-01 -8.48509073e-02 -1.26157725e+00 9.40508187e-01 4.89965677e+00 1.12286389e+00 -1.08050847e+00 1.08537816e-01 7.97296017e-02 1.48700848e-01 1.60421729e-01 3.76421846e-02 -1.49716032e+00 7.60041416e-01 2.03138113e-01 -1.79136515e-01 -1.16737939e-01 9.23827291e-01 2.76537776e-01 -6.41251802e-02 -7.59020388e-01 1.00801313e+00 5.65680712e-02 -1.11218357e+00 1.97582394e-01 -8.46569091e-02 6.25881255e-01 -1.67885140e-01 -6.11280501e-02 2.98203140e-01 1.52465990e-02 -4.66330379e-01 9.41882908e-01 4.36147153e-01 3.75035286e-01 -4.50285673e-01 9.06325519e-01 2.51396179e-01 -1.78338373e+00 -1.95733815e-01 -6.93795562e-01 5.38883746e-01 2.97860563e-01 4.87322628e-01 -9.83892202e-01 7.68153548e-01 7.30127573e-01 6.47402048e-01 -8.48881245e-01 2.01117039e+00 -1.17583483e-01 5.99522531e-01 -3.64471704e-01 -2.45758191e-01 1.59141943e-01 1.78504307e-02 8.33229601e-01 8.24095666e-01 5.76554120e-01 1.51241839e-01 7.08353579e-01 7.43545592e-01 9.39859226e-02 2.56859392e-01 -3.30315053e-01 2.09480971e-01 7.80329287e-01 1.48106790e+00 -8.59394431e-01 -5.73551834e-01 -2.16996223e-01 3.52333397e-01 2.93106679e-02 1.74630910e-01 -1.03116596e+00 -2.84565449e-01 3.86963904e-01 1.29442990e-01 7.85537720e-01 7.98809081e-02 3.25988024e-01 -1.08550215e+00 2.12030470e-01 -7.17999876e-01 3.45234782e-01 -8.40564430e-01 -1.11996150e+00 6.02439582e-01 1.19251914e-01 -1.96580565e+00 1.49804249e-01 -5.60384572e-01 -6.86460972e-01 6.92020893e-01 -1.30088377e+00 -1.32232296e+00 -7.00356841e-01 2.83497185e-01 5.08531809e-01 -2.58949906e-01 4.68159579e-02 5.17415762e-01 -5.28869689e-01 4.97602969e-01 -1.03068240e-01 2.28681043e-01 7.85639942e-01 -6.88772023e-01 1.99581042e-01 1.06801963e+00 1.09918363e-01 5.12337804e-01 4.76797819e-01 -9.54821885e-01 -1.25806272e+00 -1.55286360e+00 4.70463157e-01 -3.70616138e-01 2.27205098e-01 -1.24413431e-01 -9.18589354e-01 3.16529483e-01 -2.87656605e-01 3.87956113e-01 3.13530453e-02 -3.67925376e-01 -1.02269314e-01 -4.23885316e-01 -9.95745003e-01 5.02033234e-01 9.69988286e-01 2.81426996e-01 -5.71215689e-01 1.47215798e-01 6.47820055e-01 -4.35881704e-01 -4.78068441e-01 7.68252492e-01 6.49200261e-01 -6.16712093e-01 1.07843888e+00 -2.25300580e-01 -3.38006556e-01 -1.14737284e+00 1.47110775e-01 -9.90029275e-01 -5.56835115e-01 -1.75300404e-01 -1.44806787e-01 1.49753881e+00 -4.84783351e-02 -4.71944720e-01 6.28814638e-01 1.11938879e-01 -1.98192984e-01 -6.14981472e-01 -8.08734894e-01 -1.06939638e+00 -5.93341172e-01 1.40045034e-02 4.23825920e-01 5.36192715e-01 -5.55527985e-01 2.34048039e-01 -5.91557145e-01 1.77865863e-01 1.00040734e+00 3.12150925e-01 9.93104219e-01 -1.37977660e+00 2.31104400e-02 -3.27159733e-01 -6.96511328e-01 -1.13501179e+00 -3.72482210e-01 -7.63508737e-01 4.67819721e-01 -1.60026979e+00 4.52350289e-01 -7.31483519e-01 -4.72368419e-01 3.30969453e-01 -7.28987634e-01 5.47366202e-01 7.76669621e-01 4.13788348e-01 -8.22774053e-01 7.65428960e-01 1.38747561e+00 -3.62451822e-01 -2.21567675e-01 3.76479924e-01 -4.70309645e-01 4.95769650e-01 4.97106642e-01 -6.55201733e-01 -1.94251612e-01 -3.77090573e-02 -5.00868618e-01 -2.00556606e-01 5.43646276e-01 -1.20606649e+00 1.58464074e-01 -3.18294793e-01 5.17286241e-01 -1.08297420e+00 4.20661241e-01 -7.73616850e-01 2.51900047e-01 7.90237606e-01 2.02650994e-01 -2.96734512e-01 2.54884779e-01 7.67801702e-01 -1.35530874e-01 -2.26066351e-01 9.31502342e-01 5.97170815e-02 -1.08416963e+00 5.06330431e-01 6.84535354e-02 -5.56353778e-02 1.23450315e+00 -2.39823461e-01 -5.07273316e-01 2.07180586e-02 -2.41731077e-01 4.08398002e-01 2.65593231e-01 7.94073343e-01 8.02316487e-01 -1.72933638e+00 -7.85357773e-01 -3.40167806e-02 4.82413769e-01 -2.17082620e-01 9.52666178e-02 1.00807416e+00 -7.05112964e-02 4.80714083e-01 -3.20317954e-01 -1.01681197e+00 -1.64756966e+00 7.44474471e-01 8.81652161e-02 -1.08713374e-01 -7.86722302e-01 6.79338396e-01 5.93584955e-01 6.84843212e-02 1.67052910e-01 -3.85789275e-01 -4.46768820e-01 -1.09119222e-01 6.18241608e-01 1.44448087e-01 -1.90581679e-01 -8.55233908e-01 -7.38162160e-01 9.63031590e-01 -1.90915540e-01 2.31980354e-01 9.48125839e-01 1.30973101e-01 3.83634597e-01 3.27431820e-02 5.53043902e-01 -1.10673510e-01 -1.28827012e+00 -3.97275090e-01 8.30678493e-02 -8.25610042e-01 -2.02838019e-01 -5.61939180e-01 -1.30333841e+00 6.16270185e-01 9.78497565e-01 -1.45984009e-01 9.22750652e-01 -9.69087705e-03 7.56273985e-01 7.54750669e-02 6.90123379e-01 -7.79441595e-01 3.13000798e-01 4.52862568e-02 7.56490886e-01 -1.41290152e+00 3.07450503e-01 -6.14741981e-01 -5.43152988e-01 8.71547937e-01 1.06582725e+00 5.89814745e-02 2.89582014e-01 -2.86982954e-01 2.53457725e-02 1.83125446e-03 -5.69312572e-01 -4.96195883e-01 8.66060078e-01 6.47593141e-01 2.00224549e-01 -1.86171085e-02 -7.16444671e-01 3.97863388e-01 4.33363944e-01 1.14494473e-01 2.05265790e-01 8.18861663e-01 -8.45965683e-01 -1.12305880e+00 -8.27254236e-01 5.65536678e-01 -1.05308905e-01 2.42520213e-01 -8.08513612e-02 8.46392751e-01 3.45585793e-01 7.98018038e-01 -2.86124974e-01 -4.43436742e-01 2.32155830e-01 -3.00455987e-01 5.02241313e-01 -5.07411242e-01 -3.79021585e-01 3.67147833e-01 -5.08175530e-02 -4.14874703e-01 -6.55053914e-01 -6.11206234e-01 -1.09239972e+00 2.25223184e-01 -9.03254569e-01 2.41859868e-01 5.37138462e-01 8.36764693e-01 2.43216336e-01 5.28298914e-01 3.36873710e-01 -1.04343295e+00 -5.07444501e-01 -1.11093605e+00 -1.97435349e-01 5.34804165e-01 4.52490449e-02 -1.39260256e+00 1.52442642e-02 -1.33028835e-01]
[6.327364921569824, -2.128077507019043]
26e46a11-a6b2-4f3f-bb7b-c43342b6b9f6
diffuscene-scene-graph-denoising-diffusion
2303.14207
null
https://arxiv.org/abs/2303.14207v1
https://arxiv.org/pdf/2303.14207v1.pdf
DiffuScene: Scene Graph Denoising Diffusion Probabilistic Model for Generative Indoor Scene Synthesis
We present DiffuScene for indoor 3D scene synthesis based on a novel scene graph denoising diffusion probabilistic model, which generates 3D instance properties stored in a fully-connected scene graph and then retrieves the most similar object geometry for each graph node i.e. object instance which is characterized as a concatenation of different attributes, including location, size, orientation, semantic, and geometry features. Based on this scene graph, we designed a diffusion model to determine the placements and types of 3D instances. Our method can facilitate many downstream applications, including scene completion, scene arrangement, and text-conditioned scene synthesis. Experiments on the 3D-FRONT dataset show that our method can synthesize more physically plausible and diverse indoor scenes than state-of-the-art methods. Extensive ablation studies verify the effectiveness of our design choice in scene diffusion models.
['Matthias Nießner', 'Justus Thies', 'Angela Dai', 'Lev Markhasin', 'Yinyu Nie', 'Jiapeng Tang']
2023-03-24
null
null
null
null
['indoor-scene-synthesis']
['computer-vision']
[ 2.43240401e-01 -3.53271291e-02 4.21899587e-01 -2.60214508e-01 -3.73845577e-01 -7.98873007e-01 6.41011715e-01 2.70050883e-01 2.83450693e-01 2.91393012e-01 4.26118553e-01 -2.48286560e-01 -3.06797594e-01 -1.14756823e+00 -8.70196879e-01 -4.80137289e-01 1.46035552e-01 7.00527132e-01 3.59212279e-01 8.48715827e-02 2.19026312e-01 1.06608701e+00 -1.46556520e+00 7.31015801e-02 7.68974245e-01 7.49353588e-01 7.85993278e-01 7.75652528e-01 -2.54158497e-01 6.41451001e-01 -4.95567322e-01 -2.30658408e-02 3.37246776e-01 -1.31163090e-01 -4.25145775e-01 5.36766827e-01 5.84016919e-01 -3.38114858e-01 -7.52884090e-01 6.31971836e-01 4.85891014e-01 3.83134663e-01 9.66182590e-01 -1.00105846e+00 -6.76379561e-01 1.50022447e-01 -3.78940344e-01 -3.16335976e-01 6.37539983e-01 1.99079677e-01 5.17471075e-01 -9.96297836e-01 9.35210228e-01 1.30006635e+00 4.32487458e-01 2.19021127e-01 -1.36661851e+00 -2.69897193e-01 4.50274736e-01 -3.67584378e-01 -1.54579711e+00 -3.72924328e-01 1.11926460e+00 -5.58298469e-01 9.42455292e-01 3.52516472e-01 8.90994787e-01 1.00518239e+00 1.84580863e-01 6.10911250e-01 8.56262386e-01 -2.97627896e-01 6.94774270e-01 -4.15963382e-02 -2.04739794e-01 9.75536525e-01 2.47888148e-01 -1.88963100e-01 -4.49850649e-01 -2.65349358e-01 1.09263122e+00 -8.54719579e-02 -9.25497562e-02 -9.90559280e-01 -1.35154772e+00 3.75720918e-01 6.04488015e-01 -3.16971391e-01 -3.70204270e-01 3.88904691e-01 -3.64768505e-01 -4.20888364e-01 6.96323335e-01 5.17623961e-01 -3.71310234e-01 2.50656843e-01 -6.27752364e-01 6.36928916e-01 7.22612321e-01 1.68321395e+00 8.30341518e-01 7.07351975e-03 -3.55041862e-01 7.56698847e-01 4.51897651e-01 1.03094149e+00 -6.38476014e-01 -1.48014653e+00 3.82695526e-01 6.51580930e-01 -5.24975024e-02 -1.01301444e+00 -3.76115441e-01 -8.69289860e-02 -7.42920518e-01 1.15396241e-02 9.15516354e-03 3.23035941e-02 -1.33460963e+00 1.28326631e+00 6.44558668e-01 2.70904750e-01 -1.29985586e-01 7.49624312e-01 1.16970706e+00 7.31977582e-01 1.52229760e-02 3.01367640e-01 8.62921894e-01 -6.39068961e-01 -2.18655452e-01 -1.91329822e-01 3.43062341e-01 -9.31713641e-01 8.79178166e-01 9.56226289e-02 -9.44027126e-01 -4.54372108e-01 -6.99386239e-01 -2.55692869e-01 -2.80123770e-01 -7.53585547e-02 1.03206241e+00 7.27139652e-01 -9.37923849e-01 3.10142398e-01 -6.79158151e-01 -3.78065914e-01 4.53475058e-01 1.52582854e-01 -1.15902580e-01 -5.85277438e-01 -2.97673017e-01 2.84876794e-01 1.45939767e-01 -8.63226429e-02 -1.18661582e+00 -1.09351492e+00 -1.22361100e+00 -1.52188182e-01 3.80537868e-01 -1.54922962e+00 8.81340146e-01 7.23210871e-02 -1.33564961e+00 7.03140974e-01 -3.24117869e-01 8.81824717e-02 2.56126672e-01 2.45280638e-01 -7.68098757e-02 -3.21236365e-02 2.22837910e-01 8.79198551e-01 7.12415934e-01 -1.84995306e+00 -3.81747961e-01 -2.17916414e-01 3.03824902e-01 4.94612932e-01 3.63581717e-01 -5.32317281e-01 -9.00717735e-01 -5.29464483e-01 5.69304705e-01 -9.62610602e-01 -7.09419549e-01 2.99020439e-01 -1.00356984e+00 1.65652677e-01 7.13854611e-01 -2.96725810e-01 6.72967017e-01 -2.37319112e+00 2.51951814e-01 5.77151179e-01 3.25351864e-01 -6.39977515e-01 -1.45777181e-01 4.14422989e-01 1.77320719e-01 2.59767622e-01 -4.17172432e-01 -7.43196607e-01 -9.19590238e-03 8.22650790e-02 -2.58798063e-01 5.06888747e-01 -3.31378914e-02 6.40613675e-01 -1.07882750e+00 -4.96138781e-01 9.31472361e-01 6.09510839e-01 -1.00036955e+00 2.57207096e-01 -6.92856312e-01 5.95949769e-01 -8.28105748e-01 7.66881585e-01 9.37553883e-01 -2.84303904e-01 -1.03870742e-02 -3.23781252e-01 -1.74387306e-01 1.32220224e-01 -1.32885122e+00 2.38049555e+00 -5.30987382e-01 3.39114815e-01 -1.64184153e-01 -4.71687876e-02 9.72115338e-01 -2.41994977e-01 5.33392429e-01 -2.61472791e-01 -1.09871946e-01 -2.00400367e-01 -7.03899324e-01 -1.94133610e-01 8.85766804e-01 4.39414412e-01 -1.69462681e-01 1.90490037e-01 -3.03287655e-01 -1.30396903e+00 -2.94347331e-02 8.65252376e-01 1.18901229e+00 4.76927280e-01 -2.28637010e-01 -4.38647091e-01 2.28205342e-02 2.73409277e-01 2.17492461e-01 9.32737291e-01 4.50035751e-01 1.20707870e+00 1.73990533e-01 -7.98135251e-02 -1.10352826e+00 -1.46330965e+00 -3.68218236e-02 3.37575138e-01 6.52962863e-01 -6.26177847e-01 -5.14004350e-01 -4.78283525e-01 1.96363553e-01 1.14640439e+00 -4.03747022e-01 -4.05358262e-02 -2.83501565e-01 -5.46549320e-01 -1.18129104e-01 1.90151811e-01 3.84193450e-01 -8.01088035e-01 -2.94059575e-01 -1.57476719e-02 -1.04592601e-02 -1.29596460e+00 -5.01525819e-01 -1.04125299e-01 -6.69918239e-01 -9.27588940e-01 -4.80042219e-01 -7.28428483e-01 1.14796722e+00 8.04804683e-01 1.24098372e+00 5.79911843e-02 -3.78132284e-01 8.50321472e-01 -2.08970904e-01 -1.82663441e-01 -4.45913672e-01 4.11278307e-02 3.86128463e-02 -3.85543182e-02 -4.79574829e-01 -6.73010230e-01 -7.37627625e-01 1.40283808e-01 -7.93511629e-01 7.27779210e-01 1.20528564e-01 -5.46211787e-02 1.11318648e+00 4.67478603e-01 -3.57637495e-01 -9.64624941e-01 3.38237107e-01 -4.53187287e-01 -7.32263863e-01 2.56474376e-01 -1.30687386e-01 1.44912779e-01 3.30635339e-01 -1.99159056e-01 -1.26047194e+00 5.27137697e-01 1.72176689e-01 -4.49422777e-01 -4.55019772e-01 2.44836524e-01 -5.82671404e-01 -8.78629386e-02 5.13817549e-01 2.02606335e-01 -7.35684931e-01 -5.11955559e-01 8.54556799e-01 1.06289633e-01 4.03391957e-01 -1.04271364e+00 9.78701830e-01 7.04977334e-01 4.09142911e-01 -1.33112895e+00 -5.54155231e-01 -2.92246342e-01 -7.18260586e-01 -3.65277618e-01 8.93965781e-01 -1.04840314e+00 -4.65642601e-01 5.71937084e-01 -1.34824228e+00 -7.23115802e-01 -4.25792903e-01 2.59625584e-01 -5.64663470e-01 1.11444399e-01 -2.66616911e-01 -8.89242947e-01 3.79072726e-01 -1.09576643e+00 1.65008008e+00 5.08827064e-03 -1.29233778e-01 -9.32169437e-01 -1.43119991e-01 2.39131987e-01 1.63984597e-02 3.73569548e-01 1.33070886e+00 3.51074040e-01 -1.48458707e+00 1.32680848e-01 -3.17387670e-01 -2.89285511e-01 3.03924024e-01 4.50886518e-01 -8.95478785e-01 1.29309624e-01 -4.38121229e-01 1.95630029e-01 6.31926954e-01 8.15126657e-01 1.53163624e+00 1.03445388e-01 -6.45080209e-01 1.13991010e+00 1.41062868e+00 -5.63650727e-02 6.17603481e-01 -1.45924032e-01 1.29762936e+00 5.42780280e-01 2.19551802e-01 5.25100946e-01 6.27803683e-01 6.19402766e-01 5.98660290e-01 -1.99414194e-01 -8.75998974e-01 -8.25501263e-01 -1.30255565e-01 6.58039927e-01 1.48761094e-01 -7.64508903e-01 -8.74248326e-01 4.30908322e-01 -1.55949318e+00 -6.13920212e-01 -4.99302745e-01 2.08666492e+00 2.15416580e-01 9.59602371e-02 -3.32360059e-01 -2.99437642e-01 4.52326357e-01 4.22029376e-01 -6.74298227e-01 9.39698219e-02 -4.36697721e-01 2.59221792e-02 8.94509852e-01 8.17910194e-01 -7.48291254e-01 1.25534034e+00 6.40511942e+00 7.65663028e-01 -4.93639976e-01 -3.56533885e-01 6.78097486e-01 -3.23996842e-02 -1.09075618e+00 2.10458517e-01 -7.61414468e-01 3.82107310e-02 1.81891710e-01 -3.82593647e-02 7.71320820e-01 8.45023274e-01 4.19424325e-01 -3.69765759e-01 -9.79775071e-01 1.13326907e+00 8.12591314e-02 -1.84874368e+00 4.95689124e-01 1.88056275e-01 1.12748635e+00 -1.55265450e-01 1.14956349e-02 -2.32760966e-01 8.92495215e-01 -6.43442094e-01 1.11325073e+00 6.41411841e-01 8.85035396e-01 -5.61829388e-01 -2.29519844e-01 3.14248502e-01 -1.32326365e+00 3.12592119e-01 -3.01713228e-01 2.11416721e-01 3.26937735e-01 9.06828344e-01 -9.07790959e-01 7.82526910e-01 7.10928082e-01 8.62317741e-01 -6.80795252e-01 1.12329352e+00 -4.30727631e-01 4.14427102e-01 -6.90970600e-01 6.72788871e-03 -1.67017370e-01 -3.25328171e-01 6.99178934e-01 9.53580797e-01 4.79918182e-01 2.97451854e-01 3.36533248e-01 1.37283373e+00 -1.17675290e-01 -1.56657815e-01 -9.53754544e-01 3.37066442e-01 7.27518380e-01 9.65792120e-01 -1.11596501e+00 -2.00350240e-01 3.51536497e-02 1.05065715e+00 9.68582854e-02 6.71106994e-01 -7.72168696e-01 4.94814627e-02 6.89903200e-01 3.94775361e-01 3.04538161e-01 -9.48852718e-01 -6.28530025e-01 -9.54259157e-01 -1.02495268e-01 -2.77657241e-01 -9.42349657e-02 -1.38649380e+00 -1.22540677e+00 3.67134839e-01 3.89142394e-01 -8.94055307e-01 2.44601697e-01 -3.81202728e-01 -3.43045592e-01 7.68089414e-01 -1.15222788e+00 -1.25084782e+00 -5.53890467e-01 7.71818459e-01 7.13331461e-01 3.78727168e-01 7.63689458e-01 5.35387844e-02 -6.14651978e-01 -2.74539798e-01 -1.18317679e-01 -2.59207994e-01 1.78746164e-01 -1.19656873e+00 9.16933894e-01 9.11201239e-01 2.00254545e-01 4.03166920e-01 5.94621897e-01 -1.12836862e+00 -1.81378901e+00 -1.38346231e+00 4.62196648e-01 -7.34100342e-01 2.13819921e-01 -8.48085642e-01 -2.60600239e-01 3.92506570e-01 -2.49657124e-01 -2.11810976e-01 1.76104352e-01 -8.24805275e-02 -5.83460853e-02 1.10860392e-01 -1.12664235e+00 1.08334601e+00 1.94050574e+00 -3.86249065e-01 4.87833284e-02 8.22194874e-01 1.16802704e+00 -8.01635027e-01 -6.12552106e-01 2.99376786e-01 -1.06957923e-04 -7.06145406e-01 1.44325423e+00 -1.64251849e-01 4.60683376e-01 -5.95231354e-01 -5.95728040e-01 -1.25516427e+00 -6.70846760e-01 -6.27495646e-01 -8.60357434e-02 1.11722064e+00 4.43877816e-01 -1.68396622e-01 8.38783324e-01 9.81161892e-01 -5.06701887e-01 -3.67330432e-01 -5.89525878e-01 -5.90321124e-01 -3.62322032e-01 -7.98023582e-01 9.95518029e-01 5.83286524e-01 -9.61594939e-01 3.62129182e-01 -1.16844423e-01 7.08168030e-01 9.27665055e-01 5.06636858e-01 1.27511358e+00 -9.65824485e-01 -3.32887843e-02 -3.23617578e-01 -1.82533175e-01 -1.67979026e+00 -2.51343604e-02 -9.09981191e-01 1.02866270e-01 -2.14441562e+00 -7.47802258e-02 -1.22306192e+00 4.22105938e-01 1.09140120e-01 2.94198524e-02 -7.27618560e-02 5.41433021e-02 -1.28767654e-01 -4.50496376e-01 7.52420783e-01 1.47409391e+00 -3.27858150e-01 -6.46284282e-01 -1.52003735e-01 -5.55243492e-01 5.83652794e-01 5.04918575e-01 -4.56786126e-01 -6.23373032e-01 -8.26844752e-01 2.50460744e-01 -6.49351031e-02 3.90662491e-01 -1.04539645e+00 1.58484146e-01 -7.24683106e-01 5.64557910e-01 -9.33493197e-01 8.91128540e-01 -9.46039379e-01 6.20704293e-01 -7.06527904e-02 -5.90095744e-02 -2.05408111e-01 1.64395198e-01 7.26957262e-01 6.70966983e-01 2.45161265e-01 2.62118012e-01 -2.36833006e-01 -8.35885584e-01 8.56395543e-01 -1.05535664e-01 -1.86154842e-01 9.09320652e-01 -4.32870150e-01 -2.11614043e-01 -3.81599873e-01 -4.65504348e-01 2.47046426e-02 9.91814137e-01 4.79961455e-01 1.03785884e+00 -1.31351268e+00 -7.56940186e-01 3.75069588e-01 2.37072274e-01 7.90529311e-01 4.67233032e-01 -1.02911025e-01 -8.42368841e-01 -1.88736394e-02 4.05158073e-01 -8.39663327e-01 -9.79587853e-01 6.44266129e-01 9.42872912e-02 2.91611284e-01 -9.48810756e-01 8.78375769e-01 6.22050285e-01 -6.99102581e-01 -4.20226529e-02 -6.30186677e-01 5.10224700e-01 -6.59183621e-01 -1.19842373e-01 2.37025052e-01 -6.03970252e-02 -6.11538053e-01 -4.27966118e-01 8.07121396e-01 4.95234519e-01 -2.00743333e-01 1.28044271e+00 -2.00005695e-01 3.52890454e-02 2.48638168e-01 8.06395054e-01 5.19108713e-01 -1.51207387e+00 4.44930010e-02 -5.94516814e-01 -9.12345290e-01 1.94287568e-01 -5.98069191e-01 -1.05653346e+00 3.85964364e-01 5.87832704e-02 -7.01832352e-03 7.21049190e-01 3.23135585e-01 4.96844649e-01 1.99053004e-01 7.51940072e-01 -6.22619271e-01 -1.14707202e-01 4.78826255e-01 1.01437223e+00 -7.48319864e-01 4.25547719e-01 -1.27764702e+00 -3.63210380e-01 7.78958499e-01 4.56463337e-01 -1.05065241e-01 1.04037011e+00 2.12312713e-01 -3.15113962e-01 -6.08433783e-01 -3.04655910e-01 -9.68906134e-02 4.52279389e-01 9.08452272e-01 -1.22500896e-01 3.90555531e-01 7.09923744e-01 2.85252631e-02 -4.76932496e-01 -4.57960129e-01 3.94653141e-01 9.01028633e-01 -1.36872739e-01 -8.19215000e-01 -3.78019989e-01 3.19525421e-01 5.72870731e-01 -1.12050287e-01 -3.99132580e-01 5.15653014e-01 1.68624401e-01 9.86565590e-01 4.84993905e-02 -4.21240926e-01 7.42201269e-01 -5.44634759e-01 8.67698967e-01 -1.04097211e+00 3.73486951e-02 -1.38859794e-01 1.93652689e-01 -6.32649958e-01 -1.65960833e-01 -6.04824245e-01 -1.13890171e+00 -4.39742327e-01 -1.19626351e-01 -3.77602339e-01 9.21547890e-01 5.86044848e-01 6.68716669e-01 6.74228370e-01 6.33875906e-01 -1.18396568e+00 4.28291768e-01 -3.63226414e-01 -5.91790199e-01 4.58387733e-01 1.02042414e-01 -6.86725914e-01 -2.31875286e-01 1.79469526e-01]
[9.226799964904785, -3.137993097305298]
fc35441a-6e2b-4dde-b154-6aa1af4bd74d
content-determination-for-chess-as-a-source
null
null
https://aclanthology.org/W18-6605
https://aclanthology.org/W18-6605.pdf
Content Determination for Chess as a Source for Suspenseful Narratives
null
["Pablo Gerv{\\'a}s", 'Richard Doust']
2018-11-01
null
null
null
ws-2018-11
['game-of-chess']
['playing-games']
[-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]
[-7.269283771514893, 3.769989252090454]
2a3582f9-130f-45cd-ba5e-092f047ddfd8
contrastive-lift-3d-object-instance
2306.04633
null
https://arxiv.org/abs/2306.04633v1
https://arxiv.org/pdf/2306.04633v1.pdf
Contrastive Lift: 3D Object Instance Segmentation by Slow-Fast Contrastive Fusion
Instance segmentation in 3D is a challenging task due to the lack of large-scale annotated datasets. In this paper, we show that this task can be addressed effectively by leveraging instead 2D pre-trained models for instance segmentation. We propose a novel approach to lift 2D segments to 3D and fuse them by means of a neural field representation, which encourages multi-view consistency across frames. The core of our approach is a slow-fast clustering objective function, which is scalable and well-suited for scenes with a large number of objects. Unlike previous approaches, our method does not require an upper bound on the number of objects or object tracking across frames. To demonstrate the scalability of the slow-fast clustering, we create a new semi-realistic dataset called the Messy Rooms dataset, which features scenes with up to 500 objects per scene. Our approach outperforms the state-of-the-art on challenging scenes from the ScanNet, Hypersim, and Replica datasets, as well as on our newly created Messy Rooms dataset, demonstrating the effectiveness and scalability of our slow-fast clustering method.
['Andrea Vedaldi', 'Andrew Zisserman', 'João F. Henriques', 'Iro Laina', 'Yash Bhalgat']
2023-06-07
null
null
null
null
['object-tracking']
['computer-vision']
[ 6.90689981e-02 -7.42483214e-02 2.30904251e-01 -5.06046772e-01 -8.40660870e-01 -6.70771003e-01 6.06184542e-01 6.69250041e-02 -4.11136240e-01 2.49629721e-01 2.02134307e-02 -1.34381698e-02 -1.33536592e-01 -4.42534655e-01 -9.20915008e-01 -5.51776946e-01 -1.86719999e-01 8.43392134e-01 7.31281936e-01 6.07441105e-02 -2.88496409e-02 5.42658269e-01 -1.63301885e+00 3.60723101e-02 7.45011330e-01 9.96737123e-01 4.58765060e-01 6.72521234e-01 9.31949355e-03 3.57554674e-01 -2.29653910e-01 8.97400528e-02 6.23359323e-01 -2.46955857e-01 -9.42946494e-01 6.85295165e-01 8.86423588e-01 -1.91006839e-01 4.01963405e-02 7.08980083e-01 3.13863397e-01 4.16627556e-01 4.68151838e-01 -1.08892262e+00 6.89544901e-03 2.55939603e-01 -6.58753633e-01 -8.87190178e-03 9.41746458e-02 -2.88754832e-02 8.65525603e-01 -7.27077007e-01 7.88431704e-01 1.17528152e+00 8.05578649e-01 2.65899509e-01 -1.38843858e+00 -2.66992718e-01 5.92318118e-01 -8.82106051e-02 -1.32929850e+00 -3.82327378e-01 4.06783819e-01 -6.55176461e-01 9.58951533e-01 1.84517696e-01 7.84162700e-01 7.57037103e-01 -4.46048677e-01 8.77788365e-01 9.34481859e-01 -2.17860520e-01 3.94305885e-01 -1.24907412e-01 4.36587259e-02 6.32299125e-01 7.54357576e-02 -1.69298321e-01 -7.68502429e-02 7.89242238e-02 6.94424987e-01 2.11059541e-01 -3.35219860e-01 -9.24260974e-01 -1.57193542e+00 6.97005808e-01 4.36730772e-01 9.23266336e-02 -1.92973286e-01 2.22608775e-01 4.04666573e-01 -2.04322815e-01 6.19021177e-01 4.28687602e-01 -7.30251253e-01 -3.45504843e-02 -1.23488462e+00 1.99735045e-01 8.04771841e-01 1.20728064e+00 8.22252512e-01 -3.47564906e-01 3.40610780e-02 7.38828659e-01 2.55211264e-01 3.69061112e-01 -5.61431795e-03 -1.31596780e+00 2.37370387e-01 5.65021932e-01 2.11416677e-01 -6.20405793e-01 -6.47803009e-01 -2.00781986e-01 -6.06839538e-01 3.95352058e-02 4.70451266e-01 1.10984147e-01 -1.18338931e+00 1.65223360e+00 8.26945424e-01 3.19396973e-01 -2.82938421e-01 8.96583021e-01 5.68693697e-01 5.40352166e-01 -2.34106630e-01 1.41392723e-02 1.07850122e+00 -1.44013619e+00 -3.24026167e-01 -2.75849551e-01 4.61849421e-01 -6.49735928e-01 1.02990019e+00 4.45643067e-01 -1.04523134e+00 -5.39872766e-01 -7.75704503e-01 -6.37850165e-02 -3.09137374e-01 -3.32140401e-02 8.92784655e-01 4.69147444e-01 -9.95884836e-01 5.80263197e-01 -1.20556748e+00 -7.87734091e-01 5.53962350e-01 3.02840084e-01 -4.40112799e-01 -3.05725247e-01 -2.94880450e-01 5.03151476e-01 5.39146841e-01 -1.07507899e-01 -1.01978302e+00 -7.10344136e-01 -9.55871880e-01 -7.95914829e-02 6.45013213e-01 -6.88424706e-01 1.07868147e+00 -7.33092725e-01 -1.18592310e+00 9.74321783e-01 -7.79011101e-02 -3.19088310e-01 6.51185155e-01 -4.64012742e-01 1.25547305e-01 3.61617774e-01 2.23305687e-01 1.03840351e+00 5.94454765e-01 -1.52656353e+00 -7.14395821e-01 -4.49123472e-01 9.37978476e-02 2.45282814e-01 -4.54287753e-02 -2.20836103e-01 -1.02555668e+00 -2.79673070e-01 2.46102914e-01 -1.17451370e+00 -6.04650259e-01 3.84248197e-02 -5.14105618e-01 -2.73511803e-04 8.95688772e-01 -3.10914993e-01 6.98715687e-01 -2.35892677e+00 2.51643628e-01 1.73384190e-01 1.02394849e-01 1.33697465e-01 -1.58281341e-01 2.09010750e-01 2.00273722e-01 1.43961469e-02 -5.79798877e-01 -6.95711017e-01 1.03087164e-01 5.24640799e-01 4.25019711e-02 6.66737735e-01 6.23460524e-02 5.97981930e-01 -9.47495759e-01 -3.68578225e-01 3.59145373e-01 3.86472553e-01 -7.28397846e-01 2.39500031e-01 -3.98000598e-01 5.99319100e-01 -1.28673315e-01 5.12801826e-01 8.55869055e-01 -6.11601055e-01 1.12089748e-02 -1.09847166e-01 -2.10861102e-01 2.61740863e-01 -1.40625405e+00 2.23870897e+00 -4.66693133e-01 4.07337368e-01 4.24368680e-01 -9.99120891e-01 5.65018237e-01 -2.51967851e-02 6.47504568e-01 -2.39147559e-01 -8.86102021e-02 1.49565920e-01 -3.62935543e-01 -5.02529085e-01 3.71471733e-01 1.73512787e-01 2.11769305e-02 4.55085099e-01 1.06800236e-01 -5.85679531e-01 5.84425688e-01 2.91684270e-01 1.16647959e+00 3.86545211e-01 -5.16707227e-02 -4.73728329e-01 1.80925161e-01 2.33409688e-01 4.76561844e-01 8.04629505e-01 -3.77537981e-02 9.40180421e-01 1.50911227e-01 -4.93685305e-01 -1.13186955e+00 -8.84185195e-01 -2.17125326e-01 9.50400591e-01 3.71922523e-01 -5.47305167e-01 -8.10876846e-01 -8.53420019e-01 -1.20252492e-02 3.01400691e-01 -6.23567641e-01 3.75853062e-01 -4.20944005e-01 -7.68872619e-01 2.01382712e-01 6.49706900e-01 5.84308147e-01 -5.98676085e-01 -8.31168473e-01 9.31889862e-02 -2.24050239e-01 -1.72336829e+00 -4.00517613e-01 4.12632138e-01 -8.48473370e-01 -1.20778036e+00 -5.21757126e-01 -6.76954627e-01 7.33154118e-01 6.34679139e-01 1.33507681e+00 9.75444093e-02 -3.82426023e-01 6.78439856e-01 -3.95836741e-01 -1.31486326e-01 -1.20916560e-01 2.08102345e-01 -1.32682040e-01 -1.83307350e-01 -5.12972334e-03 -5.26663005e-01 -6.67475700e-01 5.04546106e-01 -9.72623467e-01 2.56644696e-01 3.26558620e-01 5.87202072e-01 9.35905635e-01 -5.89733385e-02 8.87929499e-02 -9.00715590e-01 -1.31439835e-01 -2.97220558e-01 -8.47665071e-01 1.02399766e-01 -2.40823641e-01 -1.08083166e-01 4.18418914e-01 -3.18281174e-01 -7.85777152e-01 6.02074742e-01 -9.97767900e-04 -4.82169569e-01 -4.94721204e-01 2.00908422e-01 -8.91368315e-02 1.50749041e-02 3.62219632e-01 -7.47159347e-02 -1.98245406e-01 -5.76219797e-01 6.26376510e-01 3.67003202e-01 4.82428551e-01 -5.73377192e-01 7.86381602e-01 1.01020896e+00 -2.70027742e-02 -7.89950907e-01 -1.11199522e+00 -1.00671828e+00 -1.05530500e+00 -8.46671686e-02 1.15653491e+00 -1.14842618e+00 -3.58607620e-01 3.21861535e-01 -9.80519772e-01 -7.96796083e-01 -3.93108219e-01 5.71227968e-01 -8.19345236e-01 3.20962369e-01 -5.38612068e-01 -6.22941494e-01 4.50007580e-02 -1.16139746e+00 1.43795347e+00 -1.41841173e-01 -6.79929601e-03 -8.77980471e-01 6.67465702e-02 4.03225273e-01 1.73707426e-01 5.19671679e-01 5.28945327e-01 -5.61247170e-01 -8.46508741e-01 -1.77358184e-02 -3.09142590e-01 2.81617463e-01 -6.02481775e-02 1.64202169e-01 -8.85782957e-01 -3.51116925e-01 -8.39659944e-02 -4.74072754e-01 1.15661597e+00 3.73725712e-01 1.20490623e+00 1.19290769e-01 -4.90905493e-01 9.13423538e-01 1.36313415e+00 -1.52183577e-01 3.57460737e-01 3.03240269e-01 9.12454903e-01 6.27242029e-01 5.98712623e-01 3.79473299e-01 7.06338108e-01 7.58241773e-01 6.95240200e-01 -3.28885019e-01 -1.21100917e-01 1.43919826e-01 2.63049621e-02 7.80532777e-01 -8.88201222e-02 -2.23562717e-01 -9.67583895e-01 7.87750065e-01 -2.06947827e+00 -8.30354691e-01 -3.78107995e-01 2.15851545e+00 3.73788595e-01 -7.09267482e-02 1.82462275e-01 7.22424313e-02 5.42817116e-01 1.42378181e-01 -3.98937553e-01 8.82256180e-02 7.10065290e-02 -2.99365842e-03 4.07477826e-01 3.36633146e-01 -1.50950110e+00 8.53492498e-01 6.61582899e+00 5.97251534e-01 -7.93944955e-01 1.92395136e-01 6.22566462e-01 -2.88361251e-01 -4.65568416e-02 7.78872892e-02 -7.51714051e-01 2.98178077e-01 6.84317827e-01 4.45370167e-01 5.45579731e-01 9.49217558e-01 1.34316772e-01 -3.86452377e-01 -1.27748394e+00 8.87331843e-01 1.57538831e-01 -1.31347775e+00 -4.56407338e-01 1.17998704e-01 9.47274685e-01 5.93737245e-01 -3.25429082e-01 6.59547225e-02 4.06800777e-01 -8.28954041e-01 8.55289757e-01 1.51110321e-01 4.99457121e-01 -6.17710471e-01 3.98322105e-01 4.79592949e-01 -1.30293882e+00 1.17910326e-01 -3.51492494e-01 1.74991116e-01 2.63700128e-01 8.51301670e-01 -6.99890792e-01 5.47452211e-01 1.21289718e+00 8.20423007e-01 -7.48870015e-01 1.21963537e+00 -1.11582048e-01 4.74560350e-01 -9.53848660e-01 3.70900840e-01 4.37633783e-01 -2.60509998e-01 2.18819618e-01 1.43109429e+00 3.55116785e-01 -3.81462984e-02 7.40267038e-01 7.88700163e-01 -5.81915267e-02 -2.05678523e-01 -5.03720641e-01 2.57079780e-01 3.74084085e-01 1.67086792e+00 -1.24447048e+00 -4.74561602e-01 -3.94849062e-01 9.88817573e-01 3.75869811e-01 2.67649293e-01 -8.28349292e-01 3.54971066e-02 3.79873365e-01 2.06083521e-01 8.50302637e-01 -5.83330989e-01 -4.19294387e-02 -1.11642611e+00 -2.64245700e-02 -6.86836123e-01 2.70841211e-01 -6.55088007e-01 -1.27855372e+00 6.27030969e-01 1.25341073e-01 -1.13433397e+00 -6.81919083e-02 -4.70773131e-01 -4.27958548e-01 3.61065030e-01 -1.50288892e+00 -1.33483410e+00 -7.63693213e-01 6.58389926e-01 6.42237544e-01 3.29515815e-01 6.92317784e-01 4.13788319e-01 -5.67977786e-01 -3.43060456e-02 2.13863254e-01 -2.06334982e-02 6.85360014e-01 -1.35165250e+00 4.90063250e-01 8.86315465e-01 2.41814166e-01 3.19330603e-01 4.80895579e-01 -3.50176424e-01 -1.20766664e+00 -1.33309436e+00 3.33422124e-01 -6.23023629e-01 4.59632695e-01 -7.70303011e-01 -8.68103981e-01 8.33947361e-01 1.65811688e-01 4.96749729e-01 6.67742372e-01 1.06752731e-01 -3.12474757e-01 1.44810259e-01 -1.08160293e+00 2.06428424e-01 1.39578366e+00 -1.02186963e-01 -2.85830379e-01 6.92749321e-01 8.48472476e-01 -6.43708587e-01 -8.22665215e-01 5.42633116e-01 1.67777970e-01 -1.22962987e+00 9.58587646e-01 -2.43873507e-01 2.89025337e-01 -5.84771514e-01 -4.37503844e-01 -1.19200647e+00 -2.02774808e-01 -5.36009550e-01 -5.06822988e-02 1.15528107e+00 2.56723851e-01 -1.70507073e-01 7.11658716e-01 4.64373767e-01 -4.51726437e-01 -6.44126058e-01 -8.75726700e-01 -9.25806522e-01 -2.33514249e-01 -5.55678129e-01 5.27460158e-01 8.87809038e-01 -4.73831594e-01 2.65926957e-01 -1.40072182e-01 2.56723553e-01 6.66930556e-01 4.24184322e-01 1.17894650e+00 -1.31219816e+00 -4.56194758e-01 -2.22885638e-01 -3.05316597e-01 -1.24896955e+00 4.41042595e-02 -7.69820571e-01 3.15656394e-01 -1.65150428e+00 3.39450866e-01 -7.32731998e-01 1.95844471e-01 3.88515115e-01 -1.32683843e-01 4.64503825e-01 3.32259923e-01 2.19014958e-01 -1.29506862e+00 5.35549879e-01 1.10801566e+00 -5.13346978e-02 -1.84176937e-01 -1.30342960e-01 -3.16094518e-01 1.02711916e+00 5.27832985e-01 -3.97942245e-01 -4.31556076e-01 -6.52312696e-01 -1.85775962e-02 -2.49366790e-01 4.72155660e-01 -1.16850662e+00 9.46666971e-02 -4.67739208e-03 4.02272642e-01 -9.05483961e-01 5.98378420e-01 -9.56903338e-01 9.83752608e-02 1.46546990e-01 -5.40112108e-02 6.18650056e-02 2.73469627e-01 8.28525484e-01 -7.69124106e-02 1.10058514e-02 8.44156146e-01 -3.77172381e-01 -7.48844087e-01 3.80319238e-01 -1.02544345e-01 2.13997021e-01 1.26360047e+00 -1.85853347e-01 -5.71269654e-02 -1.09010532e-01 -7.24181175e-01 5.14917731e-01 8.51065397e-01 3.40442955e-01 3.57981324e-01 -1.09518325e+00 -4.34030205e-01 1.69094041e-01 1.31949216e-01 8.64673555e-01 2.16217890e-01 1.04707634e+00 -6.22101784e-01 8.93218517e-02 3.84250060e-02 -1.26555681e+00 -1.14212310e+00 6.63585305e-01 2.82769829e-01 -2.87136644e-01 -8.43701541e-01 8.47638369e-01 3.78023922e-01 -7.11907506e-01 2.56207108e-01 -4.59110141e-01 1.96582213e-01 -6.68647215e-02 3.94573748e-01 2.86096871e-01 1.63942233e-01 -5.01355648e-01 -5.23497939e-01 7.07175076e-01 -1.76864467e-03 -1.13544449e-01 1.73911083e+00 -3.62505376e-01 -1.48625433e-01 4.47721243e-01 1.04377961e+00 -1.82011917e-01 -1.76528692e+00 -2.88428012e-02 -4.77098301e-02 -6.11891210e-01 -5.74955046e-02 -5.90472162e-01 -1.18339419e+00 8.55989575e-01 5.59864104e-01 2.25900024e-01 1.01090419e+00 4.37341928e-01 8.29477370e-01 4.75307524e-01 5.29698908e-01 -1.13415778e+00 1.97740108e-01 6.46508753e-01 5.50968289e-01 -1.48171961e+00 1.08593062e-01 -5.51859915e-01 -5.02287507e-01 9.02449906e-01 5.78548372e-01 -1.80348799e-01 5.18207967e-01 3.67347211e-01 5.00852317e-02 -3.63117188e-01 -6.23538554e-01 -4.93283808e-01 1.11787178e-01 6.17688537e-01 1.48328394e-01 -1.90083683e-01 2.89071679e-01 1.23439498e-01 4.33980227e-02 -2.28497013e-01 4.84652519e-01 9.98664916e-01 -4.67375994e-01 -9.41772044e-01 -4.74481195e-01 2.41144553e-01 -2.19538689e-01 1.01418279e-01 -2.65741974e-01 9.73016024e-01 1.97802246e-01 9.20628488e-01 2.87804902e-01 -1.57816693e-01 2.42329419e-01 -9.67935473e-02 5.78841150e-01 -7.51969278e-01 -3.73100191e-01 4.25079763e-01 -1.52414724e-01 -9.55472946e-01 -1.05804360e+00 -7.82029450e-01 -1.28360713e+00 -9.78844985e-02 -4.15115207e-01 -4.45098802e-02 7.90959537e-01 9.49318945e-01 4.96652156e-01 5.10137439e-01 4.69607711e-01 -1.50364017e+00 -1.61787361e-01 -7.17125297e-01 -4.82303351e-01 6.99463367e-01 1.87584713e-01 -6.51383758e-01 -3.35870087e-01 2.71674454e-01]
[8.282027244567871, -2.6273043155670166]
996a3500-638a-4227-afe3-e7913c21f09b
putting-3d-spatially-sparse-networks-on-a
2112.01316
null
https://arxiv.org/abs/2112.01316v2
https://arxiv.org/pdf/2112.01316v2.pdf
Putting 3D Spatially Sparse Networks on a Diet
3D neural networks have become prevalent for many 3D vision tasks including object detection, segmentation, registration, and various perception tasks for 3D inputs. However, due to the sparsity and irregularity of 3D data, custom 3D operators or network designs have been the primary focus of research, while the size of networks or efficacy of parameters has been overlooked. In this work, we perform the first comprehensive study on the weight sparsity of spatially sparse 3D convolutional networks and propose a compact weight-sparse and spatially sparse 3D convnet (WS^3-Convnet) for semantic and instance segmentation on the real-world indoor and outdoor datasets. We employ various network pruning strategies to find compact networks and show our WS^3-Convnet achieves minimal loss in performance (2.15\% drop) with orders-of-magnitude smaller number of parameters (99\% compression rate) and computational cost (95\% reduction). Finally, we systematically analyze the compression patterns of WS^3-Convnet and show interesting emerging sparsity patterns common in our compressed networks to further speed up inference (45\% faster). \keywords{Efficient network architecture, Network pruning, 3D scene segmentation, Spatially sparse convolution}
['Jaesik Park', 'Christopher Choy', 'Junha Lee']
2021-12-02
null
null
null
null
['scene-segmentation']
['computer-vision']
[ 4.54366982e-01 1.30502298e-01 -6.63824603e-02 -5.16230822e-01 3.67712140e-01 -1.99945018e-01 1.05691738e-01 -1.37652025e-01 -5.73112130e-01 2.95411259e-01 -1.70429990e-01 -5.52357793e-01 -4.90630805e-01 -7.33111739e-01 -7.97852993e-01 -4.74236995e-01 -3.95812392e-01 3.74516957e-02 5.16937792e-01 1.06441244e-01 2.90397722e-02 1.14863265e+00 -1.74835861e+00 -5.35185263e-02 6.28941357e-01 1.64622331e+00 3.42836499e-01 3.18676621e-01 -9.90289971e-02 4.91175115e-01 -6.44263864e-01 -1.82044983e-01 8.73822033e-01 -7.73474947e-02 -5.93780518e-01 2.73613960e-01 7.91935503e-01 -2.42406279e-01 -7.76807368e-01 1.16101122e+00 4.69872028e-01 2.56307632e-01 4.00290698e-01 -9.51546073e-01 -3.64755005e-01 5.45884788e-01 -4.76771295e-01 5.35714805e-01 -3.31569076e-01 2.66342703e-02 5.49769104e-01 -7.37563729e-01 6.12146437e-01 1.19652808e+00 9.43303764e-01 4.24334407e-01 -8.33116353e-01 -5.31441987e-01 2.62583882e-01 1.33228861e-02 -1.56105399e+00 -3.35778177e-01 7.31448114e-01 -9.86072794e-02 1.47308207e+00 2.42441088e-01 9.79882181e-01 7.58192599e-01 -1.07314281e-01 6.01834774e-01 8.88293862e-01 -2.24804699e-01 2.21005827e-01 -2.59379745e-01 2.49785274e-01 1.08821642e+00 5.60429215e-01 -1.10120267e-01 -6.87038958e-01 4.01323706e-01 1.11684334e+00 2.41738752e-01 -4.19984711e-03 -1.33902758e-01 -7.47273088e-01 5.85354865e-01 7.81925321e-01 1.98230594e-01 -2.94051379e-01 3.31014693e-01 2.68231779e-01 2.79781908e-01 5.16512096e-01 2.43436262e-01 -5.22436917e-01 -6.36398271e-02 -1.07411289e+00 3.56251657e-01 6.81536853e-01 1.35123217e+00 5.48702002e-01 6.01707101e-01 1.37210414e-01 1.01515031e+00 5.81011251e-02 6.15849853e-01 8.87042508e-02 -1.07503533e+00 4.72703427e-01 1.00419700e+00 -5.56273341e-01 -1.18838930e+00 -6.07065856e-01 -7.17497170e-01 -1.44854188e+00 6.73619583e-02 1.62821114e-01 -9.90800858e-02 -1.25522554e+00 1.34965742e+00 3.28336477e-01 1.23066798e-01 -1.82799041e-01 9.56772447e-01 1.14525735e+00 2.45446980e-01 -1.95307940e-01 1.42317191e-01 1.02908194e+00 -5.78294575e-01 -1.83407754e-01 -2.17890531e-01 4.03585851e-01 -4.63485390e-01 6.35810435e-01 1.60440937e-01 -1.35369229e+00 -6.30687356e-01 -1.13115954e+00 -1.92316264e-01 -3.32619160e-01 6.82253437e-03 8.57472181e-01 7.45405972e-01 -9.92688000e-01 7.45563984e-01 -9.94229198e-01 -3.24692607e-01 1.36090922e+00 8.18837345e-01 -2.03530908e-01 -4.40205425e-01 -6.61129832e-01 6.15465939e-01 5.49869657e-01 2.59596229e-01 -7.22750068e-01 -8.14609945e-01 -9.19317544e-01 6.14786707e-02 2.50508219e-01 -5.77945352e-01 8.01411033e-01 -3.36568892e-01 -1.12970662e+00 1.03159106e+00 1.69563293e-02 -9.00272012e-01 7.67125487e-02 -2.33334690e-01 -1.11888677e-01 2.97590375e-01 -1.13412149e-01 1.05879223e+00 5.87666929e-01 -7.93272614e-01 -5.51868796e-01 -5.91661632e-01 1.38399631e-01 2.06932619e-01 -3.33565950e-01 -2.04821870e-01 -5.45414209e-01 -6.39870405e-01 8.04315507e-01 -7.73218215e-01 -6.91970825e-01 3.94884109e-01 -5.02977133e-01 -1.90438062e-01 1.23043132e+00 -1.38895676e-01 9.04157341e-01 -2.28154898e+00 -1.24963760e-01 4.50304896e-01 6.53891921e-01 5.56943059e-01 7.22821476e-03 -3.02330613e-01 1.76277846e-01 1.73331171e-01 -6.01472974e-01 -4.46515977e-01 -1.06977277e-01 5.01269937e-01 -2.31377613e-02 5.34711182e-01 3.01333427e-01 8.91152024e-01 -3.63753557e-01 -5.21352828e-01 3.60304743e-01 6.97325408e-01 -8.78261566e-01 -2.25238308e-01 -7.21853375e-02 -8.52926150e-02 -5.80648482e-01 1.07882833e+00 7.23579168e-01 -4.13518071e-01 -1.79283425e-01 -5.18369019e-01 -1.49993330e-01 5.77308424e-02 -1.20021701e+00 1.95123589e+00 -4.09105718e-02 7.30200529e-01 2.70652026e-01 -1.44360602e+00 1.20963347e+00 -1.09245978e-01 6.52029455e-01 -8.79134953e-01 4.51418966e-01 2.10242867e-01 -2.47611608e-02 -3.58763993e-01 4.51006740e-01 1.96074024e-01 -1.96507201e-02 2.64144331e-01 2.31720433e-01 -4.52157646e-01 1.51105508e-01 -4.27262532e-03 1.24824381e+00 -2.79152483e-01 -1.04183465e-01 -4.47000444e-01 1.26523059e-02 1.63051665e-01 5.25982916e-01 9.43991005e-01 -2.95678586e-01 3.91548723e-01 3.93205017e-01 -7.06186235e-01 -9.61944461e-01 -7.29711354e-01 -3.01636606e-01 6.38409376e-01 3.71474355e-01 -5.03565557e-02 -5.84815264e-01 -4.31828260e-01 1.02054484e-01 1.00013241e-01 -2.55545914e-01 -9.21920314e-02 -9.34762716e-01 -6.99671805e-01 1.01020885e+00 6.42850339e-01 1.14967024e+00 -8.87713611e-01 -1.06692648e+00 -5.04811257e-02 5.51172674e-01 -1.53304720e+00 -5.59416562e-02 7.94813097e-01 -1.45603180e+00 -1.07353437e+00 -3.78766358e-01 -1.11150157e+00 8.57484162e-01 4.61564511e-01 8.71862292e-01 9.63083953e-02 -5.91465354e-01 1.24311022e-01 -1.26939431e-01 -5.14954448e-01 3.29704791e-01 1.96004853e-01 1.62600264e-01 -5.78605890e-01 2.93696910e-01 -1.00365698e+00 -7.13677585e-01 1.37092501e-01 -9.34937000e-01 -1.60276726e-01 5.86220145e-01 5.06564260e-01 9.88406479e-01 3.61074060e-01 2.16654941e-01 -8.23170066e-01 5.46403766e-01 -1.95934907e-01 -7.60160148e-01 -2.95780927e-01 -2.81877756e-01 -1.48256630e-01 4.68339950e-01 -3.74354035e-01 -6.37944877e-01 2.66450763e-01 -4.52862889e-01 -9.58538413e-01 -3.92525792e-01 2.11498722e-01 6.43546730e-02 -4.52838063e-01 5.63243270e-01 1.84121996e-01 -1.60918962e-02 -6.31904185e-01 2.49883980e-01 2.30795532e-01 4.76757854e-01 -3.57184976e-01 6.62917376e-01 6.42746985e-01 5.13178527e-01 -1.16506398e+00 -1.01833498e+00 -2.04362303e-01 -7.97462523e-01 5.57658114e-02 8.20115507e-01 -9.81150329e-01 -7.31016338e-01 5.12694597e-01 -1.12362731e+00 -4.09616530e-01 -7.77484894e-01 4.39104646e-01 -1.81481570e-01 1.03780432e-02 -5.15625238e-01 -5.42900980e-01 -4.05957818e-01 -1.25896430e+00 8.14390242e-01 2.77351916e-01 -6.87860651e-04 -7.67287612e-01 -7.04914451e-01 1.55750945e-01 5.73231876e-01 4.12696213e-01 1.12973022e+00 -5.08556306e-01 -9.14278030e-01 -7.56249279e-02 -6.68584049e-01 5.31874001e-01 -1.69925496e-01 -3.63471568e-01 -7.75360823e-01 1.05486825e-01 7.61191770e-02 -1.91527858e-01 9.69882905e-01 8.79675210e-01 1.84978223e+00 -3.22081596e-01 -2.41766930e-01 1.25606990e+00 1.35577440e+00 1.57822534e-01 2.44297236e-01 -2.52158791e-01 8.77416790e-01 3.20572197e-01 5.85409515e-02 4.73226756e-01 -1.15950271e-01 6.09744415e-02 8.18075299e-01 -1.60904676e-01 -4.77762401e-01 -4.51479405e-02 -3.22963655e-01 1.05931735e+00 -2.79044181e-01 -2.12532505e-01 -6.87650442e-01 4.57957566e-01 -1.34050262e+00 -7.62402892e-01 1.08921051e-01 1.66724360e+00 4.93377626e-01 5.82893848e-01 -1.82441980e-01 3.23372632e-01 5.02197146e-01 4.20751065e-01 -9.26400542e-01 -3.28224987e-01 -4.97488767e-01 8.68558943e-01 8.69285583e-01 2.09589265e-02 -1.10241795e+00 9.73255277e-01 6.09914970e+00 9.44616735e-01 -1.11241567e+00 -1.71997711e-01 7.64523268e-01 -5.07748902e-01 6.74089119e-02 -4.72859204e-01 -1.21583307e+00 2.52412225e-04 4.04744565e-01 3.36019725e-01 1.07389502e-01 1.07353628e+00 1.47064716e-01 -7.64742270e-02 -8.90848935e-01 1.35260940e+00 -5.41155376e-02 -1.93043983e+00 1.89904824e-01 2.03062311e-01 9.37617779e-01 4.03012305e-01 2.24894751e-02 -2.88211135e-03 7.80672133e-02 -1.29495347e+00 7.28130698e-01 1.63433805e-01 9.82219994e-01 -8.20417821e-01 5.07034779e-01 3.90715420e-01 -1.18866491e+00 -8.34401473e-02 -6.88719094e-01 -4.92223911e-02 1.78487524e-01 1.06842661e+00 -5.34609258e-01 -1.74672566e-02 1.23399472e+00 9.67902422e-01 -2.98698008e-01 1.15882421e+00 3.23290110e-01 5.71245492e-01 -8.92698526e-01 -2.58987725e-01 6.43360853e-01 -4.96798418e-02 5.66515386e-01 1.16266978e+00 3.53246570e-01 4.57584620e-01 1.24598183e-01 1.02164876e+00 -6.39288783e-01 -3.87014031e-01 -8.36871088e-01 1.32592350e-01 6.56019330e-01 7.68236339e-01 -1.11609149e+00 -7.11393654e-02 -2.48459890e-01 7.92793751e-01 1.57026574e-01 2.21689120e-01 -4.68867213e-01 -6.44687831e-01 1.01214695e+00 1.19318627e-01 7.18056202e-01 -7.31066644e-01 -6.93409860e-01 -5.69957197e-01 3.79420407e-02 -4.96058255e-01 8.49083625e-03 -3.37913185e-01 -9.98747051e-01 6.03972256e-01 1.56099036e-01 -8.17354560e-01 4.43484455e-01 -8.63337457e-01 -2.46371686e-01 4.35909480e-01 -1.48916745e+00 -6.26319706e-01 -4.40610796e-01 8.96576285e-01 7.13774383e-01 -2.36451596e-01 3.51367831e-01 5.32846272e-01 -6.24326468e-01 4.48249072e-01 -2.38085806e-01 2.81408250e-01 -2.58776814e-01 -7.08972096e-01 4.20234501e-01 6.96855307e-01 6.95983395e-02 5.55490315e-01 1.73975958e-03 -5.36451042e-01 -1.63090467e+00 -1.40483451e+00 5.71194172e-01 2.21201405e-01 2.33618796e-01 -3.89223486e-01 -6.12643838e-01 2.61509240e-01 -1.60847351e-01 4.07354116e-01 4.40548122e-01 -1.79046959e-01 -1.29989713e-01 -2.10638329e-01 -1.42728078e+00 6.54987395e-01 2.00358295e+00 -2.14291364e-01 -1.91898182e-01 3.21981192e-01 1.38493943e+00 -7.97548056e-01 -9.98023748e-01 6.73459291e-01 3.64890188e-01 -1.06451678e+00 1.35728216e+00 -2.76348859e-01 3.23830783e-01 -1.30741149e-01 -4.33815181e-01 -5.69092929e-01 -4.29121554e-02 -3.93330395e-01 -3.70332301e-01 4.89363641e-01 1.70199260e-01 -2.94512719e-01 1.38109684e+00 3.80945385e-01 -7.56342649e-01 -1.25732875e+00 -1.23683584e+00 -5.80636322e-01 -3.56206775e-01 -9.40376759e-01 5.48365712e-01 6.86642766e-01 -7.72803724e-01 2.55212281e-03 1.71409883e-02 1.79716855e-01 7.69237995e-01 4.64522652e-02 3.15742493e-01 -1.20351899e+00 1.81762695e-01 -7.48731554e-01 -6.24304354e-01 -1.88064611e+00 -1.22719981e-01 -9.97362912e-01 -2.01077163e-01 -1.41184759e+00 -3.78358841e-01 -9.76320505e-01 -1.37491003e-02 6.27025247e-01 5.84308028e-01 4.95985776e-01 5.77722751e-02 1.11431535e-02 -7.09919393e-01 6.55850589e-01 1.50104189e+00 -1.34984717e-01 -4.42328870e-01 9.09726173e-02 -5.28017104e-01 8.82463098e-01 7.33853400e-01 -4.38314766e-01 -5.90824008e-01 -9.20996428e-01 2.82031531e-03 -3.28451872e-01 5.83050430e-01 -1.30167806e+00 4.26195741e-01 -3.28577943e-02 6.10274017e-01 -1.07566559e+00 5.80408216e-01 -9.55746293e-01 -2.06547812e-01 4.39254761e-01 -5.90561368e-02 1.23902790e-01 4.47660685e-01 3.74205410e-01 -1.53529748e-01 -1.71513006e-01 6.77086174e-01 -4.39290196e-01 -8.84442687e-01 8.31653833e-01 -1.18289769e-01 1.91922799e-01 7.92232037e-01 -9.22249436e-01 -1.47165637e-02 3.97256762e-01 -6.40354753e-01 1.45045361e-02 -1.10903099e-01 3.82044129e-02 1.12607920e+00 -1.09390855e+00 -2.75680989e-01 4.92456645e-01 -5.22655666e-01 1.07469869e+00 3.40155780e-01 5.64288855e-01 -9.26804602e-01 5.80571711e-01 -1.73098296e-01 -1.00310862e+00 -1.10105205e+00 -9.36903656e-02 4.48354363e-01 1.30921334e-01 -9.44244564e-01 1.32472610e+00 -3.06560785e-01 -3.06279600e-01 8.42455924e-01 -7.61744082e-01 1.20085165e-01 -3.32987934e-01 -5.60639054e-02 4.68629152e-01 1.78220123e-01 -2.53572375e-01 -2.81626374e-01 6.48773789e-01 1.43886745e-01 6.19700134e-01 1.90294671e+00 1.62335813e-01 -7.43876100e-02 -9.09858495e-02 1.23702240e+00 -7.13590384e-01 -1.45296180e+00 -4.19505328e-01 -2.34563723e-01 -3.76362920e-01 2.49475986e-01 -1.91292971e-01 -1.69138193e+00 8.24526906e-01 5.41458130e-01 1.73103824e-01 1.21562541e+00 2.81006575e-01 1.01681197e+00 9.07858074e-01 2.36023456e-01 -9.77460563e-01 -2.17133854e-02 8.90082836e-01 5.91974199e-01 -9.11165178e-01 3.79262567e-01 -5.67438424e-01 -1.02628216e-01 8.08148146e-01 7.64075696e-01 -4.32458699e-01 1.36143982e+00 4.72089142e-01 -4.47312236e-01 -8.21977615e-01 -1.77767500e-01 -3.17527026e-01 1.62944213e-01 7.02723265e-01 -8.81356969e-02 -2.45736375e-01 3.04949611e-01 4.20523047e-01 -4.78947967e-01 -2.36643702e-01 -1.59511849e-01 1.08366406e+00 -5.32621980e-01 -6.61466718e-01 9.54623669e-02 9.26764369e-01 -2.37813890e-01 -1.97184071e-01 -6.07713163e-02 8.58408570e-01 5.70808768e-01 4.77501214e-01 3.42513978e-01 -3.69524539e-01 6.00716114e-01 -2.43244722e-01 7.06756830e-01 -4.95730191e-01 -5.96997440e-01 -2.22751096e-01 -1.47536427e-01 -7.05057681e-01 -9.07018721e-01 -3.52274984e-01 -1.22446144e+00 -5.05831182e-01 -1.61532775e-01 -4.65565890e-01 8.31792653e-01 8.46658647e-01 6.15964830e-01 5.31196833e-01 3.41089129e-01 -9.69578266e-01 -1.34310901e-01 -6.35011375e-01 -6.53616011e-01 1.14056572e-01 6.73830807e-02 -5.42950392e-01 -1.37878135e-01 5.46126142e-02]
[8.003463745117188, -3.565056800842285]
a1ffb3e6-32ae-48c4-9317-dfaf08437a8c
a-multi-task-deep-learning-framework-for
2104.09375
null
https://arxiv.org/abs/2104.09375v1
https://arxiv.org/pdf/2104.09375v1.pdf
A Multi-Task Deep Learning Framework for Building Footprint Segmentation
The task of building footprint segmentation has been well-studied in the context of remote sensing (RS) as it provides valuable information in many aspects, however, difficulties brought by the nature of RS images such as variations in the spatial arrangements and in-consistent constructional patterns require studying further, since it often causes poorly classified segmentation maps. We address this need by designing a joint optimization scheme for the task of building footprint delineation and introducing two auxiliary tasks; image reconstruction and building footprint boundary segmentation with the intent to reveal the common underlying structure to advance the classification accuracy of a single task model under the favor of auxiliary tasks. In particular, we propose a deep multi-task learning (MTL) based unified fully convolutional framework which operates in an end-to-end manner by making use of joint loss function with learnable loss weights considering the homoscedastic uncertainty of each task loss. Experimental results conducted on the SpaceNet6 dataset demonstrate the potential of the proposed MTL framework as it improves the classification accuracy considerably compared to single-task and lesser compounded tasks.
['Elif Sertel', 'Burak Ekim']
2021-04-19
null
null
null
null
['remote-sensing-image-classification']
['miscellaneous']
[ 5.97193539e-01 -4.11573537e-02 2.42239445e-01 -6.07866645e-01 -9.65670943e-01 -3.06359440e-01 4.65046883e-01 1.44046038e-01 -5.27800500e-01 6.05563879e-01 2.08831653e-02 -3.32004339e-01 -4.72573668e-01 -8.47235620e-01 -8.69987905e-01 -7.88504720e-01 -1.17966942e-01 2.99312890e-01 4.29663025e-02 1.58015452e-02 7.17547014e-02 4.74777222e-01 -1.43044329e+00 -1.44374624e-01 1.18126070e+00 1.06976330e+00 4.58310783e-01 2.61974037e-01 2.08788738e-02 4.28556085e-01 -1.84707846e-02 -6.13448769e-02 5.19867003e-01 1.36688367e-01 -7.76069343e-01 3.86075646e-01 4.05888051e-01 -1.15831241e-01 1.71447411e-01 8.18778813e-01 5.17312407e-01 3.85532409e-01 5.66996336e-01 -7.86090851e-01 -2.23201454e-01 2.87941486e-01 -6.75730526e-01 1.92036033e-02 -2.57766813e-01 -1.40629103e-02 1.31985235e+00 -6.82441890e-01 9.76006240e-02 1.08442950e+00 8.31564903e-01 -9.09593105e-02 -1.37234545e+00 -4.56777751e-01 4.32092130e-01 -1.13309957e-01 -1.51220250e+00 -1.83847651e-01 9.23526287e-01 -6.74032331e-01 3.78063828e-01 1.97326586e-01 1.75274655e-01 5.95493078e-01 -3.24202508e-01 8.23265016e-01 1.15230560e+00 -1.62241697e-01 1.40711412e-01 -1.53890938e-01 6.88554272e-02 5.38949013e-01 -1.41824614e-02 -7.92237073e-02 -8.09363946e-02 2.09826335e-01 6.10708416e-01 -3.85145061e-02 -3.09527159e-01 -4.43762124e-01 -1.05561042e+00 6.65483773e-01 8.26138437e-01 2.19466358e-01 -3.74166876e-01 2.53868818e-01 1.95751354e-01 -9.21265781e-03 8.87031674e-01 3.20268631e-01 -4.21240866e-01 4.08633709e-01 -1.12254989e+00 3.47471058e-01 2.87686855e-01 4.18475121e-01 9.93447661e-01 -2.21536644e-02 -1.60524026e-01 1.03999889e+00 3.62326682e-01 2.89379060e-01 -1.63809270e-01 -7.89304852e-01 6.83010817e-01 5.89434981e-01 2.03929678e-01 -8.71910632e-01 -5.25937259e-01 -8.37119162e-01 -8.16361964e-01 1.50080428e-01 1.96334526e-01 -6.18273281e-02 -7.99258053e-01 1.76863348e+00 3.51436198e-01 -3.19603160e-02 -3.04286808e-01 8.59965146e-01 4.32936192e-01 5.47044098e-01 3.19189310e-01 3.67865771e-01 1.00492382e+00 -8.43959749e-01 -2.92134583e-01 -4.85922962e-01 3.40159118e-01 -3.89833003e-01 1.22027135e+00 2.32039653e-02 -5.44116259e-01 -6.60122037e-01 -1.04607403e+00 -2.01952830e-02 -3.91751468e-01 6.06039524e-01 7.16779351e-01 4.35861230e-01 -8.20328116e-01 5.16220629e-01 -8.73287201e-01 -1.27414539e-01 7.86421537e-01 2.22091511e-01 -8.87795761e-02 1.16207406e-01 -8.77886832e-01 7.76824653e-01 3.79014194e-01 7.89832592e-01 -7.88470149e-01 -9.07842040e-01 -8.73726547e-01 1.75562054e-01 5.67802548e-01 -5.06060421e-01 8.97701383e-01 -8.07183921e-01 -1.04714453e+00 7.85180390e-01 2.91326106e-01 -2.71690637e-01 8.59989047e-01 -3.44090670e-01 -3.44657525e-02 -3.02779108e-01 4.24387544e-01 5.87901711e-01 6.16190791e-01 -1.31554127e+00 -6.53264225e-01 -6.64125144e-01 8.95284861e-02 2.18801260e-01 -6.59281462e-02 -3.66925180e-01 -2.63586432e-01 -6.95938170e-01 2.46446580e-01 -9.10418868e-01 -3.92428011e-01 2.76253313e-01 -3.92614275e-01 3.18126939e-02 7.28684247e-01 -9.32218611e-01 1.01878905e+00 -2.07082319e+00 1.35663837e-01 1.15822136e-01 -2.01944411e-01 1.94618434e-01 -3.59669887e-02 2.37758473e-01 1.18437715e-01 1.22565493e-01 -9.82698798e-01 -4.72821414e-01 2.58158930e-02 1.74105227e-01 -9.23593193e-02 5.08033097e-01 4.36567396e-01 7.09683955e-01 -6.98491573e-01 -1.90303341e-01 5.27076125e-01 5.88498294e-01 -2.03616902e-01 2.73179442e-01 -3.98181379e-01 7.25226820e-01 -8.11705470e-01 5.43356955e-01 7.59315073e-01 -1.80731058e-01 -5.04167639e-02 -2.62087137e-01 -3.92157137e-01 1.42020822e-01 -1.23205125e+00 1.73605502e+00 -9.48094010e-01 5.06784737e-01 2.98332900e-01 -1.23193133e+00 9.06098306e-01 5.22566773e-02 4.12409693e-01 -5.27254879e-01 -1.37348518e-01 1.74220204e-01 -4.18918341e-01 -4.56687868e-01 3.82662326e-01 -2.19875425e-01 -1.51348591e-01 3.23484153e-01 -3.94332528e-01 -2.55939335e-01 -1.49346322e-01 -2.96290308e-01 5.69778025e-01 3.76657456e-01 -6.90909699e-02 -3.83427173e-01 6.04709983e-01 -1.55942321e-01 3.91351342e-01 5.31088650e-01 -3.79485302e-02 6.93899989e-01 2.56120209e-02 -4.61267382e-01 -8.45512927e-01 -9.18039441e-01 -4.79652852e-01 1.06615472e+00 1.37800351e-01 3.34810644e-01 -5.48544228e-01 -5.48510194e-01 1.14313088e-01 7.59412944e-01 -6.95245862e-01 1.91259786e-01 -4.63035077e-01 -1.19245827e+00 5.01624942e-01 5.30039072e-01 9.78258014e-01 -8.91819417e-01 -8.28570902e-01 2.90152609e-01 -3.15769702e-01 -1.25789142e+00 -1.28517166e-01 3.01802963e-01 -7.19112933e-01 -1.02414536e+00 -7.39137113e-01 -5.15134096e-01 5.57089210e-01 3.01719785e-01 8.03773880e-01 -1.46531120e-01 -2.88740247e-01 1.54277042e-01 -1.77104756e-01 -1.47421271e-01 -2.34909374e-02 4.14887905e-01 -7.04253912e-01 3.42588902e-01 -1.54392317e-01 -7.52394259e-01 -6.40787899e-01 3.03125650e-01 -1.01272309e+00 3.47845346e-01 5.92839956e-01 5.39366961e-01 5.18376350e-01 2.63633251e-01 5.38579106e-01 -8.48249018e-01 3.23483407e-01 -5.01288235e-01 -6.49324477e-01 4.87693846e-01 -5.89353085e-01 1.25091195e-01 3.54070902e-01 3.96296054e-01 -1.25697732e+00 1.42118260e-01 -1.86945215e-01 1.12634599e-01 -2.65864938e-01 6.89840257e-01 -3.39051366e-01 -1.40154377e-01 2.92450458e-01 2.54828304e-01 -1.11077040e-01 -7.89864540e-01 2.72063345e-01 5.94999075e-01 5.02439559e-01 -6.97366774e-01 7.74071336e-01 5.84142387e-01 2.24224538e-01 -9.45957124e-01 -1.14753449e+00 -6.86732113e-01 -7.25844264e-01 -2.27378398e-01 9.21259701e-01 -1.11010957e+00 -2.19457582e-01 7.21235991e-01 -8.19638073e-01 -3.77332538e-01 -3.10507733e-02 1.17548347e-01 -3.65279734e-01 3.95956576e-01 3.47760431e-02 -9.26517427e-01 -4.49274957e-01 -1.26713693e+00 1.24558210e+00 -8.33364427e-02 3.28999072e-01 -9.93673444e-01 -2.72667501e-02 7.06915140e-01 4.96991992e-01 6.75833642e-01 9.85359490e-01 -1.10384263e-01 -7.54570961e-01 -8.81749615e-02 -4.84554768e-01 7.22290516e-01 1.99466079e-01 -1.85717151e-01 -1.17255354e+00 -1.37935176e-01 -5.08653671e-02 -2.95542538e-01 1.29105341e+00 5.15637755e-01 1.36607635e+00 7.49326721e-02 -1.03979126e-01 7.75308013e-01 1.74828982e+00 -2.60719627e-01 4.16241318e-01 5.46062231e-01 8.77042115e-01 9.13639665e-01 5.25831223e-01 3.82509559e-01 6.43950582e-01 7.79927373e-01 7.68443406e-01 -3.07698816e-01 4.64784838e-02 -6.22923225e-02 -1.98396191e-01 6.39710054e-02 -1.12705462e-01 -8.08637589e-02 -1.00760353e+00 6.62487566e-01 -1.94406116e+00 -6.41316175e-01 -9.20674577e-02 2.13246918e+00 5.15066147e-01 -1.45187452e-01 -1.67351767e-01 2.36973539e-01 6.33558393e-01 7.44654477e-01 -5.96702039e-01 8.25974122e-02 -2.08361581e-01 1.36357397e-02 7.96376526e-01 4.60237235e-01 -1.56217325e+00 8.08680475e-01 4.76420879e+00 8.35598350e-01 -1.20191693e+00 9.04311240e-02 7.68710732e-01 3.96964848e-01 -4.25496370e-01 -4.72698212e-02 -5.40854156e-01 4.23751533e-01 4.09566551e-01 4.06149358e-01 3.28842223e-01 6.93797290e-01 5.50441325e-01 -2.27586403e-01 -7.25841105e-01 6.28836215e-01 -3.70709866e-01 -1.07284760e+00 -4.56219092e-02 2.22105091e-03 8.63035858e-01 3.65341753e-01 2.01628536e-01 4.42055091e-02 5.78027517e-02 -8.69188905e-01 1.01080680e+00 3.14422101e-01 6.26909137e-01 -6.15070403e-01 4.57893342e-01 3.97010982e-01 -1.39193487e+00 -1.43072993e-01 -7.36951753e-02 -7.88152292e-02 2.55917042e-01 7.83932686e-01 -7.31013596e-01 8.53413582e-01 4.80117202e-01 5.91210067e-01 -5.78932703e-01 1.00733614e+00 -2.78624356e-01 5.45132339e-01 -4.30770665e-01 4.11343813e-01 6.13374591e-01 -4.92526323e-01 3.86496603e-01 1.02737057e+00 1.60836041e-01 -1.93075702e-01 4.28460896e-01 1.16261470e+00 4.94881868e-02 -1.90725289e-02 -2.68269837e-01 1.16783068e-01 1.38167754e-01 1.21286035e+00 -7.26428807e-01 3.80800724e-01 -2.03346223e-01 7.77728140e-01 3.91575843e-01 3.59134555e-01 -7.18559921e-01 1.25516742e-03 3.58075887e-01 2.72531897e-01 5.50531268e-01 -3.27681363e-01 -6.39858305e-01 -8.17182362e-01 3.31546724e-01 -2.77961522e-01 2.27056712e-01 -4.33311582e-01 -1.02715182e+00 3.76995921e-01 -1.21741489e-01 -8.88670325e-01 1.55384690e-01 -3.01016897e-01 -7.22672582e-01 9.16892111e-01 -2.07858539e+00 -1.72763705e+00 -5.44201016e-01 2.09776431e-01 6.92915499e-01 2.99637377e-01 4.60174620e-01 5.00224113e-01 -7.06977010e-01 2.08209306e-01 2.99641281e-01 -5.69349416e-02 1.16064973e-01 -1.26515973e+00 3.39734167e-01 9.45744336e-01 -2.03050882e-01 -2.95486883e-03 4.31153446e-01 -2.52219051e-01 -8.65024745e-01 -1.70902669e+00 6.93338275e-01 1.36315366e-02 4.24238116e-01 -2.35726178e-01 -7.88527787e-01 4.34169680e-01 -5.22705019e-01 1.19123398e-03 4.74624217e-01 2.60562152e-02 -2.12871179e-01 -3.21684539e-01 -1.07433426e+00 2.25527018e-01 7.97441185e-01 -6.67491376e-01 -2.12220818e-01 3.67547184e-01 4.71750408e-01 -4.26656157e-02 -5.51504254e-01 6.95271492e-01 4.35799718e-01 -8.79938602e-01 1.11147296e+00 -3.86062264e-02 4.47484076e-01 -4.64032948e-01 -4.44830745e-01 -9.26754534e-01 -1.11683369e-01 -1.39470533e-01 5.90083480e-01 1.23184252e+00 4.20976818e-01 -4.04154301e-01 5.96870720e-01 6.63033009e-01 -3.32834363e-01 -7.52386332e-01 -1.02694416e+00 -6.62704110e-01 9.82736237e-03 -6.52888238e-01 4.70833987e-01 6.82211220e-01 -9.34344172e-01 1.79295778e-01 -3.39188546e-01 5.18131137e-01 8.00098360e-01 3.83255184e-01 6.74925327e-01 -1.32854402e+00 -4.02421802e-02 -4.07344371e-01 -7.01976940e-02 -8.87761712e-01 2.78025061e-01 -8.62615407e-01 3.18034738e-01 -1.54983377e+00 2.25384682e-02 -1.00103152e+00 -2.50193924e-01 3.92745048e-01 -2.89546520e-01 5.84538952e-02 4.98002674e-03 2.23116189e-01 -4.34163392e-01 9.22854304e-01 9.64051306e-01 -3.05603385e-01 -2.49485373e-02 4.02139843e-01 -5.14752865e-01 5.98134816e-01 6.23754144e-01 -3.15567821e-01 -3.92861962e-01 -8.47995818e-01 1.80994853e-01 -1.62678599e-01 7.59111881e-01 -1.13953519e+00 -9.54100862e-02 3.57510261e-02 -6.62057474e-02 -6.95031047e-01 1.44622982e-01 -1.01646852e+00 3.34085047e-01 1.85840711e-01 -3.53370905e-01 -5.47921419e-01 1.15508132e-01 7.83466399e-01 -1.72703624e-01 -6.34949729e-02 8.93185973e-01 -9.20185819e-02 -8.26503038e-01 3.62640321e-01 1.62062004e-01 4.24749143e-02 7.54078865e-01 -3.20301443e-01 1.33414686e-01 7.69558772e-02 -5.99076092e-01 4.12733346e-01 3.42919350e-01 2.75565773e-01 2.07047895e-01 -8.09688032e-01 -7.57899225e-01 9.08801854e-02 1.17163412e-01 6.63731039e-01 4.76888388e-01 7.05219984e-01 -6.13543570e-01 3.17645758e-01 -4.44987603e-02 -6.86845422e-01 -7.66611099e-01 -7.31067732e-02 4.70825255e-01 -4.34064209e-01 -6.32682562e-01 8.75784814e-01 3.81139427e-01 -8.49818587e-01 3.19397956e-01 -3.58043283e-01 -1.86649308e-01 1.96451440e-01 8.92890897e-03 5.72243214e-01 3.03588748e-01 -6.54614627e-01 -2.21923515e-01 6.42203093e-01 1.76073015e-01 8.28247964e-02 1.70941412e+00 -2.55903274e-01 -1.54287023e-02 4.30081308e-01 1.05461633e+00 -6.46622062e-01 -1.65842402e+00 -3.58221591e-01 2.44943351e-01 -4.73652333e-01 6.07448280e-01 -9.07697678e-01 -1.16581774e+00 9.69541490e-01 6.43709660e-01 -1.21413700e-01 1.01342130e+00 -1.36472642e-01 7.73141026e-01 2.70940691e-01 2.84837693e-01 -1.13528943e+00 -2.69109577e-01 1.92187577e-01 8.82984102e-01 -1.53715026e+00 -9.17153656e-02 -4.31866109e-01 -3.64853829e-01 8.40399444e-01 3.27355474e-01 -4.59181480e-02 7.33752429e-01 -1.38943508e-01 -5.50705567e-02 -3.52672786e-01 -7.81812370e-02 -4.69752818e-01 4.51989532e-01 3.50331366e-01 1.93327263e-01 2.95033425e-01 -1.90344259e-01 4.60143447e-01 1.92310497e-01 2.41158139e-02 -2.53451038e-02 8.77522647e-01 -4.41327512e-01 -7.51796544e-01 -1.23258576e-01 2.58349329e-01 -4.28878158e-01 1.86775327e-02 1.50891058e-02 6.28666639e-01 2.34027460e-01 7.64530897e-01 -4.29757833e-02 -1.34267822e-01 2.52447754e-01 -2.82262713e-01 1.58108205e-01 -6.19928241e-01 -5.64434826e-01 -5.68589978e-02 3.26159820e-02 -2.89332896e-01 -7.05709934e-01 -6.47235870e-01 -8.86263132e-01 1.58415347e-01 -4.07762229e-01 -2.24158429e-02 9.27136242e-01 1.18563128e+00 1.13821000e-01 5.46575367e-01 7.49717236e-01 -1.09114611e+00 -7.70178199e-01 -7.59979010e-01 -6.64093256e-01 2.83111781e-01 4.53369141e-01 -6.71995938e-01 -3.87049198e-01 -1.62330568e-01]
[9.637982368469238, -1.3832008838653564]
fa4a0a23-13b7-4b9f-b36d-ebf3795ea7e3
190600041
1906.00041
null
https://arxiv.org/abs/1906.00041v1
https://arxiv.org/pdf/1906.00041v1.pdf
Table2Vec: Neural Word and Entity Embeddings for Table Population and Retrieval
Tables contain valuable knowledge in a structured form. We employ neural language modeling approaches to embed tabular data into vector spaces. Specifically, we consider different table elements, such caption, column headings, and cells, for training word and entity embeddings. These embeddings are then utilized in three particular table-related tasks, row population, column population, and table retrieval, by incorporating them into existing retrieval models as additional semantic similarity signals. Evaluation results show that table embeddings can significantly improve upon the performance of state-of-the-art baselines.
['Li Deng', 'Shuo Zhang', 'Krisztian Balog']
2019-05-31
null
null
null
null
['table-retrieval']
['natural-language-processing']
[-2.70584494e-01 6.00423254e-02 -8.22262466e-01 -1.77429572e-01 -1.19192171e+00 -6.72843039e-01 6.97913527e-01 1.05452204e+00 -3.72041196e-01 8.18255186e-01 9.63568330e-01 -2.61294067e-01 2.99151745e-02 -1.07511950e+00 -7.29189098e-01 -1.59584731e-01 1.73885718e-01 6.20209157e-01 -1.90423980e-01 -5.86015880e-01 4.74197924e-01 1.64124012e-01 -1.06604254e+00 6.57276988e-01 7.66756833e-01 1.26813483e+00 -3.90495360e-01 1.88634649e-01 -8.62135887e-01 1.09339941e+00 -6.89195514e-01 -8.17631185e-01 -1.55996129e-01 1.14232332e-01 -4.91923988e-01 -4.38788950e-01 4.86418486e-01 -2.77528733e-01 -9.49529946e-01 8.52017641e-01 3.52438688e-01 3.95282298e-01 1.06035948e+00 -9.31688488e-01 -1.75618410e+00 1.09511554e+00 -3.69837880e-01 3.60789210e-01 4.97914344e-01 -5.28262377e-01 1.54002774e+00 -1.17446077e+00 1.02567232e+00 1.53640580e+00 4.05971289e-01 4.08133358e-01 -1.14041865e+00 -4.78940755e-01 1.14502892e-01 1.56647757e-01 -1.27111614e+00 -3.51317883e-01 7.03173161e-01 -3.24774295e-01 1.27491283e+00 1.19197682e-01 1.54904991e-01 1.18001306e+00 4.71418768e-01 9.73083913e-01 2.35496551e-01 -3.11667562e-01 1.13196597e-01 8.05844009e-01 4.66558814e-01 4.98533994e-01 9.98529136e-01 -6.03552759e-01 -6.58051014e-01 -5.00482060e-02 3.65478426e-01 1.35670647e-01 -1.64149448e-01 -7.42109716e-01 -1.32013237e+00 1.00195014e+00 6.71458364e-01 1.37838021e-01 -2.86180139e-01 2.64083713e-01 7.75219262e-01 8.08347166e-02 4.32716429e-01 9.61965621e-01 -4.72946376e-01 -5.28579839e-02 -3.57811719e-01 3.11328441e-01 8.00220132e-01 1.36287832e+00 5.63703895e-01 -1.27002284e-01 -7.53102779e-01 9.86110628e-01 2.62998760e-01 5.89647710e-01 5.59878051e-01 -5.14584541e-01 1.15044188e+00 9.37176704e-01 8.62705410e-02 -1.36793673e+00 -2.46263936e-01 -1.90381408e-01 -5.94779611e-01 -9.29066420e-01 -1.85756356e-01 1.03808314e-01 -7.88146853e-01 1.27998257e+00 -9.26095545e-02 -7.27429390e-01 5.53702235e-01 3.53530198e-01 1.27936876e+00 9.64829981e-01 1.19024344e-01 1.96669817e-01 1.67524898e+00 -1.07333398e+00 -1.33474874e+00 -3.05197954e-01 9.44177151e-01 -5.94510674e-01 1.04660690e+00 -2.10042864e-01 -9.81076241e-01 -5.37260294e-01 -1.17049491e+00 -5.57381511e-01 -1.43622482e+00 1.12528555e-01 8.99432778e-01 5.87107062e-01 -7.27929473e-01 3.65788847e-01 -5.70632219e-01 -3.96392137e-01 6.80748999e-01 -1.87723935e-01 -4.55013245e-01 -2.98813820e-01 -1.55760038e+00 9.71844494e-01 4.89424229e-01 -9.01113898e-02 -2.15092838e-01 -1.06234992e+00 -1.61642599e+00 3.97357285e-01 2.34769329e-01 -7.12630332e-01 8.42132866e-01 3.19523841e-01 -6.36776149e-01 7.94548333e-01 -2.80901760e-01 -4.76208121e-01 -4.57267821e-01 -3.15407008e-01 -8.14504802e-01 6.32861704e-02 2.57928133e-01 7.26968050e-01 2.12092683e-01 -1.28238082e+00 -2.02281371e-01 -2.97631413e-01 1.11346971e-02 2.83896685e-01 -8.98309708e-01 -1.79678455e-01 -8.68787885e-01 -9.40625846e-01 3.48710008e-02 -3.49639684e-01 -4.12279293e-02 -1.88280959e-02 -5.75526655e-01 -5.68385482e-01 1.76648244e-01 -7.52149582e-01 1.73155630e+00 -2.20706487e+00 9.24987197e-02 1.32899925e-01 3.65150064e-01 -2.16796398e-01 -1.26051903e-01 8.52524340e-01 1.80264860e-01 5.09862244e-01 9.72741246e-02 -2.14954078e-01 5.21266758e-01 -1.27809029e-02 -5.56982160e-01 -1.41349390e-01 4.66228247e-01 1.48876953e+00 -7.44573832e-01 -5.98119676e-01 -4.89471108e-02 5.70012093e-01 -6.95528388e-01 6.83495477e-02 -1.27996922e-01 -9.02581871e-01 -5.01050889e-01 1.08091009e+00 2.68330693e-01 -3.98958027e-01 1.54038534e-01 -4.25208330e-01 5.11638761e-01 7.84612417e-01 -8.47382069e-01 1.66609538e+00 -4.41263139e-01 7.41840541e-01 -8.05212677e-01 -5.87800324e-01 1.09591782e+00 -1.10084891e-01 1.01931229e-01 -9.41246033e-01 2.93692112e-01 1.55724093e-01 -2.26809114e-01 -1.50350586e-01 1.08134139e+00 2.93710500e-01 -5.34974575e-01 2.08997980e-01 2.31987029e-01 -1.45904478e-02 5.29959917e-01 7.43196785e-01 9.12289739e-01 -2.46852472e-01 3.48471552e-01 -2.49691352e-01 4.86233205e-01 6.52678758e-02 -7.34579787e-02 6.90149903e-01 5.40381409e-02 4.68390077e-01 6.97861910e-01 -2.98017025e-01 -9.11175847e-01 -1.40107524e+00 -3.07546139e-01 9.86399174e-01 1.56871215e-01 -8.43035460e-01 -4.14152116e-01 -6.24857187e-01 8.80529165e-01 9.47661936e-01 -1.14891708e+00 -6.19269729e-01 -4.58358854e-01 -3.55338991e-01 1.98007926e-01 1.28379667e+00 4.40425016e-02 -1.17416465e+00 1.15335822e-01 4.47234958e-01 7.31439590e-02 -1.14245820e+00 -8.01111460e-01 7.05669641e-01 -8.41281235e-01 -7.08667636e-01 -4.72500235e-01 -9.49747384e-01 4.93993491e-01 2.38656148e-01 1.58812416e+00 -1.77297011e-01 -1.70092389e-01 5.74970782e-01 -2.61943340e-01 -3.60749602e-01 1.45141631e-01 2.83111781e-01 2.55021043e-02 -5.01953542e-01 9.26238596e-01 5.37470952e-02 -2.59932160e-01 -2.30514377e-01 -9.33972359e-01 -6.41714036e-01 3.96154881e-01 1.00460124e+00 6.61490798e-01 -4.26878035e-01 5.96102655e-01 -1.16439044e+00 1.23914397e+00 -6.93122864e-01 -3.99795711e-01 6.75374806e-01 -8.14799666e-01 6.61069453e-01 5.93878746e-01 -1.47986993e-01 -6.33155107e-01 -7.79877007e-01 2.66458035e-01 -4.63024586e-01 3.97648305e-01 9.12139058e-01 -2.98733264e-01 3.80409211e-01 6.00509167e-01 2.93476105e-01 -3.09748918e-01 -4.19747025e-01 8.34846973e-01 6.58307135e-01 3.78759652e-01 -5.37657380e-01 9.12436545e-01 2.52580583e-01 -4.01584178e-01 -3.69824052e-01 -8.11659694e-01 -4.95510399e-01 -5.65752625e-01 2.61176288e-01 9.43012416e-01 -1.06084752e+00 -3.21975440e-01 -3.20982248e-01 -1.12188804e+00 4.05702442e-01 -3.35554242e-01 2.73544699e-01 -1.56184092e-01 -1.13873303e-01 -6.79048300e-01 -4.84762549e-01 -2.74500698e-01 -9.13982570e-01 1.10154808e+00 2.43408859e-01 -3.43222976e-01 -1.38228393e+00 5.05332127e-02 3.73773873e-01 4.18853998e-01 9.67767183e-03 1.61872244e+00 -1.13375330e+00 -6.28918290e-01 -6.11735582e-01 -5.49391270e-01 -5.71035109e-02 1.97452411e-01 -1.56619310e-01 -6.22450888e-01 6.16152063e-02 -7.94889569e-01 -6.34338975e-01 1.18865240e+00 1.02623262e-01 1.33678067e+00 -2.96930343e-01 -6.77343845e-01 7.43190706e-01 1.47490084e+00 6.33682668e-01 6.66585445e-01 7.51643598e-01 8.19223404e-01 6.43158555e-01 5.74026942e-01 4.64770466e-01 7.59139895e-01 4.73857075e-01 -8.77070501e-02 1.33256093e-01 4.52148393e-02 -8.91381919e-01 5.55508249e-02 1.12801790e+00 8.12592924e-01 -7.32290864e-01 -8.26197267e-01 6.12586796e-01 -1.50238228e+00 -6.77702487e-01 5.07852077e-01 1.68191087e+00 9.73647535e-01 4.25573498e-01 -4.36304599e-01 -7.01480135e-02 3.05605710e-01 6.01721048e-01 -5.13125300e-01 -7.27247000e-01 -2.92583019e-01 1.34983987e-01 5.80343485e-01 1.30472168e-01 -1.21249247e+00 1.17497396e+00 6.71447849e+00 6.28320396e-01 -4.69847351e-01 -4.51902956e-01 6.91676497e-01 2.79773306e-02 -9.53011572e-01 -4.19281989e-01 -1.13071275e+00 1.63824722e-01 1.01043272e+00 -6.14431322e-01 4.00841743e-01 9.56818998e-01 -6.87635422e-01 2.93723434e-01 -1.47688556e+00 1.29191148e+00 6.46960676e-01 -1.85991585e+00 9.11469221e-01 -1.77035093e-01 5.62188387e-01 -6.60430193e-01 4.27151144e-01 9.21895623e-01 2.02200696e-01 -1.28454876e+00 2.59771168e-01 4.12276626e-01 7.34824955e-01 -8.95411611e-01 9.09247339e-01 -6.15229726e-01 -1.27013743e+00 2.21405178e-01 -7.61633754e-01 4.66170520e-01 -2.03034207e-01 -1.03847273e-02 -5.25959015e-01 5.32938957e-01 7.44674265e-01 1.18597269e+00 -9.13372397e-01 4.24235702e-01 2.26167813e-02 1.20880708e-01 1.58450976e-01 -7.08074331e-01 3.38509053e-01 7.50268484e-03 2.23948769e-02 1.19103765e+00 -5.11054359e-02 -1.25572979e-01 -2.58749962e-01 1.06761336e+00 -9.28300023e-01 3.11856866e-01 -1.03293324e+00 -7.68952906e-01 9.05855656e-01 9.82829034e-01 -6.22642577e-01 -5.53254604e-01 -6.33510947e-01 6.02447808e-01 5.50071955e-01 4.64536697e-01 -5.42821229e-01 -1.08113062e+00 1.12791109e+00 -3.20770323e-01 5.60098171e-01 5.24478965e-02 -5.67848504e-01 -1.28475058e+00 3.14476758e-01 -8.72064412e-01 5.75810194e-01 -9.13021863e-01 -1.48000634e+00 6.20384157e-01 3.90237309e-02 -9.86662924e-01 -3.24512899e-01 -9.05230463e-01 -6.02253564e-02 7.60625780e-01 -1.48939967e+00 -6.26652777e-01 1.28129711e-02 3.00990254e-01 5.55348098e-01 -7.13136375e-01 9.15574908e-01 3.37915570e-01 -6.83895707e-01 1.22079194e+00 7.22618878e-01 7.55963027e-01 8.55563879e-01 -1.50260830e+00 6.78125441e-01 4.99016285e-01 5.55833876e-01 1.27262247e+00 2.78285593e-01 -6.48053765e-01 -1.97620785e+00 -1.00609803e+00 1.23611152e+00 -1.16043329e+00 1.03951061e+00 -8.91724110e-01 -9.90801036e-01 8.48309100e-01 5.31049728e-01 -9.34386402e-02 1.00370669e+00 4.74331975e-01 -8.90570402e-01 -3.20349067e-01 -8.97137225e-01 9.56151128e-01 6.50230587e-01 -8.32657933e-01 -9.24721658e-01 3.11027318e-01 1.41970658e+00 -3.47517967e-01 -1.16981113e+00 -3.78702627e-03 6.77229464e-01 -6.04041666e-02 1.37852681e+00 -1.12334120e+00 9.85298634e-01 8.54362100e-02 -5.93978524e-01 -1.42860067e+00 -3.95971924e-01 -1.41897172e-01 -7.66283453e-01 1.31191933e+00 9.35487211e-01 -5.36937118e-01 7.10727155e-01 6.28185213e-01 5.76475561e-02 -9.38548744e-01 -6.46977603e-01 -8.09633195e-01 4.49569702e-01 6.95026070e-02 8.37465465e-01 9.40552592e-01 4.66697752e-01 5.84540844e-01 1.68555245e-01 -2.20058396e-01 3.81821990e-01 6.02693595e-02 5.57094872e-01 -1.05481184e+00 4.15811628e-01 -5.52398980e-01 -5.02841294e-01 -1.04948413e+00 2.58284122e-01 -1.21319127e+00 -1.93151325e-01 -2.06981707e+00 1.92451596e-01 -7.05180764e-02 -9.26771343e-01 2.42707521e-01 -6.01828098e-01 -1.41941026e-01 3.12174618e-01 -1.27604023e-01 -7.97620356e-01 9.04226124e-01 1.04540467e+00 -7.31048703e-01 9.19536352e-02 -8.86804283e-01 -1.29732585e+00 -5.54881655e-02 5.11645496e-01 -6.21252775e-01 -3.82014126e-01 -6.07396424e-01 3.88352841e-01 1.51750669e-01 -5.65250039e-01 -9.32997465e-01 2.68727511e-01 6.66080862e-02 8.17570031e-01 -8.24768543e-01 3.10401291e-01 -8.47062111e-01 -7.39945292e-01 6.86280727e-02 -1.06310308e+00 8.08899343e-01 7.04925120e-01 7.47666299e-01 -6.37241006e-01 -5.22298738e-02 1.17904939e-01 1.66866958e-01 -5.78977823e-01 2.12192133e-01 -3.18640321e-01 6.50242388e-01 5.26511550e-01 -1.39694172e-03 -6.10678852e-01 -2.40458027e-01 -1.26117900e-01 5.38649917e-01 1.52285770e-01 1.07555580e+00 9.46337104e-01 -1.79972088e+00 -4.44272846e-01 9.75188613e-02 9.74738896e-01 -4.05782640e-01 -2.26492837e-01 1.42498329e-01 -5.11281133e-01 1.12458599e+00 -1.94032863e-01 -6.01722263e-02 -7.41054595e-01 1.06746483e+00 -1.85597435e-01 -5.05615413e-01 -3.85180533e-01 9.21715140e-01 4.82321419e-02 -5.07753134e-01 7.66502023e-01 -8.03058803e-01 -6.70464635e-01 5.51247060e-01 6.66592836e-01 -4.12546881e-02 1.72691032e-01 9.64418128e-02 -5.69797039e-01 3.59061450e-01 -5.35885930e-01 1.49355367e-01 1.06678200e+00 4.89703566e-02 -2.25970931e-02 8.39613855e-01 1.60920548e+00 2.00131729e-01 -4.46452796e-01 -3.56609643e-01 4.06656981e-01 -3.45340639e-01 2.31863279e-02 -9.19807136e-01 -9.34713483e-01 9.15761173e-01 2.27425277e-01 -9.41917002e-02 5.73087990e-01 -5.64607494e-02 8.89845848e-01 1.16478562e+00 1.32302254e-01 -1.03572524e+00 2.41060685e-02 6.13491297e-01 7.80846000e-01 -1.44226432e+00 5.22412881e-02 -1.91246748e-01 -7.84350395e-01 1.25562727e+00 7.93692052e-01 -1.00900248e-01 7.25144267e-01 1.83439553e-01 2.20243096e-01 -2.74526000e-01 -1.23939621e+00 -1.07181460e-01 7.35123754e-01 7.35059440e-01 8.18527937e-01 -1.36182606e-01 -1.83638841e-01 8.16711724e-01 -8.64463300e-02 -5.25717258e-01 4.34086561e-01 1.05939960e+00 -2.79559195e-01 -8.38583052e-01 -5.63263521e-02 9.21091437e-01 -2.35271379e-01 -8.41455400e-01 -6.02317631e-01 1.04868805e+00 -6.38514698e-01 6.91356242e-01 4.07656074e-01 -4.12784755e-01 6.38979018e-01 3.75816107e-01 9.94019210e-02 -7.57740796e-01 -7.57815540e-01 -5.22473216e-01 1.46173701e-01 -4.96670127e-01 3.20058405e-01 -4.14224088e-01 -1.06387889e+00 -3.69434446e-01 2.43848041e-02 5.42173862e-01 3.68245095e-01 3.60123038e-01 4.61147070e-01 9.29165184e-01 3.65687430e-01 -5.71786193e-03 -6.84540331e-01 -1.05151749e+00 -5.06559670e-01 4.19329971e-01 2.02638775e-01 -8.77748847e-01 -3.59156758e-01 -2.66541123e-01]
[9.749168395996094, 7.888470649719238]
e56c6bce-2e96-4670-8b06-752e3350387c
dynamic-steerable-blocks-in-deep-residual
1706.00598
null
http://arxiv.org/abs/1706.00598v2
http://arxiv.org/pdf/1706.00598v2.pdf
Dynamic Steerable Blocks in Deep Residual Networks
Filters in convolutional networks are typically parameterized in a pixel basis, that does not take prior knowledge about the visual world into account. We investigate the generalized notion of frames designed with image properties in mind, as alternatives to this parametrization. We show that frame-based ResNets and Densenets can improve performance on Cifar-10+ consistently, while having additional pleasant properties like steerability. By exploiting these transformation properties explicitly, we arrive at dynamic steerable blocks. They are an extension of residual blocks, that are able to seamlessly transform filters under pre-defined transformations, conditioned on the input at training and inference time. Dynamic steerable blocks learn the degree of invariance from data and locally adapt filters, allowing them to apply a different geometrical variant of the same filter to each location of the feature map. When evaluated on the Berkeley Segmentation contour detection dataset, our approach outperforms all competing approaches that do not utilize pre-training. Our results highlight the benefits of image-based regularization to deep networks.
['Bert de Brabandere', 'Jörn-Henrik Jacobsen', 'Arnold W. M. Smeulders']
2017-06-02
null
null
null
null
['contour-detection']
['computer-vision']
[ 1.18812345e-01 1.61159530e-01 -1.11809904e-02 -5.25873542e-01 -6.73494563e-02 -7.68664718e-01 9.18351173e-01 -1.57698050e-01 -8.57531309e-01 5.46250880e-01 1.98184893e-01 -2.40206835e-03 -8.73666555e-02 -9.43875492e-01 -1.33826017e+00 -7.31564701e-01 1.49177462e-01 1.96640328e-01 5.50651312e-01 -3.52013946e-01 5.15191408e-04 8.78876090e-01 -1.26906121e+00 3.07421833e-01 5.69423556e-01 8.89275134e-01 2.84922272e-01 5.94317675e-01 1.65850073e-01 8.26597333e-01 -8.00222829e-02 -3.45377535e-01 4.76542085e-01 -3.15639339e-02 -7.52005875e-01 2.17202485e-01 9.66039658e-01 -2.15310916e-01 -5.02269149e-01 1.19580019e+00 1.45151332e-01 3.26608598e-01 5.29321253e-01 -6.13908529e-01 -8.43405604e-01 6.32991612e-01 -2.98357606e-01 5.49914479e-01 -2.33820099e-02 1.83342397e-01 1.08254063e+00 -9.39353406e-01 9.92047608e-01 1.30484915e+00 7.75775194e-01 5.74966729e-01 -1.66522443e+00 -2.65066832e-01 4.77997243e-01 8.25520754e-02 -1.20948696e+00 -4.53739196e-01 6.40670240e-01 -4.91814613e-01 9.06737089e-01 2.25236490e-01 7.26349175e-01 9.64869380e-01 1.32606372e-01 6.61434531e-01 1.01299739e+00 -4.08316016e-01 2.38807186e-01 3.88335660e-02 3.54990989e-01 7.54288375e-01 1.62540421e-01 1.70619234e-01 -4.96774763e-01 3.08028698e-01 1.19448698e+00 8.87325406e-03 -5.82501709e-01 -8.81107748e-01 -1.09785485e+00 7.54057646e-01 9.50349867e-01 2.76442587e-01 -4.35984105e-01 4.08612847e-01 1.50842398e-01 2.67944872e-01 5.14041781e-01 1.91414252e-01 -5.22272289e-01 3.84542108e-01 -8.60347271e-01 1.58022195e-01 5.87513506e-01 8.49060655e-01 1.18915677e+00 1.66083381e-01 -2.86035389e-01 3.77358258e-01 1.82999969e-01 3.09825987e-01 3.40598106e-01 -8.64218354e-01 1.06404588e-01 4.38952148e-01 -5.65676950e-02 -8.82184386e-01 -6.15118206e-01 -7.58737326e-01 -7.67019987e-01 2.34796748e-01 5.93009889e-01 -6.93525225e-02 -1.36920345e+00 1.93802106e+00 2.14807525e-01 4.38206851e-01 -9.94217098e-02 9.06434834e-01 6.08163893e-01 3.99180293e-01 7.04854280e-02 1.66825280e-01 1.28102267e+00 -7.57479191e-01 -2.62263298e-01 -3.65338683e-01 3.24964672e-01 -5.10781586e-01 1.07205510e+00 2.24233583e-01 -1.32461786e+00 -7.81466007e-01 -9.88565683e-01 -2.30867088e-01 -4.14540529e-01 4.42613214e-02 6.48262918e-01 8.35002005e-01 -1.58204567e+00 7.99111187e-01 -1.16015279e+00 -2.28166550e-01 5.84541678e-01 4.38117951e-01 -3.72574449e-01 2.65066475e-01 -8.40450168e-01 9.60561037e-01 4.56891149e-01 3.03251863e-01 -9.20624375e-01 -8.73511553e-01 -7.76335299e-01 2.56632805e-01 2.76951522e-01 -8.08340967e-01 9.40946579e-01 -1.45534658e+00 -1.43966782e+00 9.26331162e-01 -1.26411960e-01 -1.03601825e+00 7.97798276e-01 -3.84059817e-01 -6.61194548e-02 3.12520027e-01 -1.63181737e-01 9.97974277e-01 1.11928761e+00 -1.07566214e+00 -5.74948370e-01 -6.63116798e-02 5.89559793e-01 1.29000610e-02 -3.01427931e-01 -2.90128410e-01 -5.94032824e-01 -7.17931449e-01 5.97864240e-02 -9.68204618e-01 -4.44558203e-01 5.49806878e-02 -3.24277967e-01 3.98570485e-02 8.26181114e-01 -3.31136018e-01 5.84707201e-01 -2.05786276e+00 4.33573902e-01 3.31226081e-01 2.14485168e-01 4.44515675e-01 -3.04565966e-01 -1.02988705e-01 -2.34844834e-01 1.71174482e-01 -9.73759070e-02 -2.05396906e-01 -1.48393959e-01 3.24029267e-01 -2.73145407e-01 7.38148630e-01 4.31925029e-01 1.10159659e+00 -7.01240599e-01 -9.82128233e-02 6.57923520e-01 7.56645858e-01 -8.43039453e-01 -2.67141640e-01 -3.59627396e-01 5.76656520e-01 -3.86645824e-01 -9.08874348e-02 8.77909541e-01 -1.67867050e-01 -1.15833394e-01 -4.52170730e-01 -2.71175027e-01 1.22123607e-01 -1.12489104e+00 1.79677153e+00 -5.90172827e-01 8.16089749e-01 1.08325675e-01 -1.23047054e+00 7.59518921e-01 3.64344418e-02 3.40211213e-01 -5.84507823e-01 1.93931535e-01 -8.49780962e-02 7.85820559e-02 -1.24809168e-01 2.11529940e-01 9.67370793e-02 4.31039035e-01 2.14708615e-02 5.24562657e-01 1.62516728e-01 2.47398034e-01 9.96077284e-02 9.96348560e-01 2.48435587e-01 -2.16760803e-02 -7.85469115e-01 7.19837070e-01 -4.63373482e-01 3.85643899e-01 9.82249022e-01 1.26662657e-01 7.55066872e-01 3.73196244e-01 -8.90400350e-01 -9.89216566e-01 -1.12810564e+00 -3.75910193e-01 1.14337885e+00 1.84258997e-01 -1.88954055e-01 -8.32373381e-01 -5.18352807e-01 -5.59047721e-02 4.59048033e-01 -1.01485646e+00 -1.85151964e-01 -8.57823253e-01 -6.50108397e-01 3.16802830e-01 6.13775194e-01 5.66228092e-01 -9.41685498e-01 -8.72090280e-01 3.45480770e-01 1.94187149e-01 -1.24560940e+00 -3.23700100e-01 4.59903657e-01 -7.64911056e-01 -9.76112843e-01 -7.64743805e-01 -7.71396875e-01 8.50943804e-01 1.99107841e-01 1.08094871e+00 1.56444356e-01 -2.55822718e-01 5.68936408e-01 -2.61365116e-01 6.07891791e-02 -1.48770303e-01 2.09999576e-01 -2.56611377e-01 3.96913201e-01 -4.92239697e-03 -6.08209372e-01 -6.88821852e-01 3.60462815e-01 -1.10745287e+00 2.14955181e-01 3.68679106e-01 7.81651556e-01 5.33923507e-01 -2.72581041e-01 1.37257606e-01 -1.09399259e+00 1.46744594e-01 3.59977446e-02 -7.74409950e-01 2.88584292e-01 -1.72145933e-01 4.46336389e-01 7.23277271e-01 -5.23371816e-01 -1.16682196e+00 5.04681408e-01 -1.43833473e-01 -3.91736835e-01 -1.76852971e-01 2.67811030e-01 -1.05565377e-01 -6.07573986e-01 8.16063404e-01 8.54558349e-02 -3.58703017e-01 -3.77240121e-01 7.78769076e-01 -2.25055069e-01 7.18177855e-01 -6.17807150e-01 1.02923107e+00 9.72995162e-01 1.88093930e-02 -8.37703884e-01 -7.49010980e-01 -2.90073156e-01 -1.04155874e+00 -2.30053142e-01 9.91510332e-01 -9.03691649e-01 -5.31803906e-01 3.93488526e-01 -1.11547792e+00 -4.51397032e-01 -5.44713259e-01 3.99483353e-01 -5.88649273e-01 1.01618744e-01 -4.48750287e-01 -2.47207865e-01 3.84770408e-02 -1.19234955e+00 7.99524605e-01 3.10820132e-01 2.67843664e-01 -1.19548094e+00 -2.57016420e-01 -2.21617848e-01 5.53967416e-01 2.66820520e-01 5.89012206e-01 -4.68084097e-01 -8.70627463e-01 -3.58133949e-02 -3.21530491e-01 5.25024176e-01 2.28898805e-02 2.38979429e-01 -1.17253840e+00 -3.57023150e-01 1.15293181e-02 -9.25377980e-02 1.49421620e+00 7.38906026e-01 1.13887179e+00 -1.68218479e-01 -1.05216473e-01 1.17319179e+00 1.42429757e+00 -1.91551015e-01 6.95594132e-01 3.88110042e-01 8.56427550e-01 2.34361351e-01 -7.17623951e-03 2.14644611e-01 6.51262105e-02 6.12040937e-01 5.89187980e-01 -3.41265529e-01 -2.75598347e-01 -2.29858868e-02 1.88879669e-01 1.63350012e-02 -5.60255945e-01 -8.15945491e-02 -6.56004369e-01 3.58532459e-01 -1.74571884e+00 -7.89726675e-01 -4.50942889e-02 2.06650186e+00 6.97423756e-01 3.78801703e-01 -8.69109631e-02 -2.11206332e-01 7.09256947e-01 3.79613429e-01 -4.28878099e-01 -2.38640100e-01 -3.19657534e-01 7.86053002e-01 1.01890290e+00 7.59296358e-01 -1.37259269e+00 1.32623446e+00 5.87286043e+00 6.77006781e-01 -1.45463192e+00 1.17393985e-01 5.61305106e-01 1.76170394e-01 -4.03584510e-01 2.46219248e-01 -8.10774267e-01 -4.89098660e-04 6.33700311e-01 9.45717394e-02 5.18046618e-01 6.07339144e-01 1.18594319e-01 2.21572652e-01 -1.25440097e+00 5.15636623e-01 -2.20214993e-01 -1.79630685e+00 6.77803382e-02 -1.01502374e-01 7.58029580e-01 2.67027050e-01 3.39629978e-01 -2.92785335e-02 4.88875955e-01 -1.01715136e+00 1.09375608e+00 7.13156462e-01 3.11747104e-01 -5.93606412e-01 2.63979346e-01 1.06726401e-01 -1.11618805e+00 1.04366668e-01 -7.91547775e-01 -4.55784518e-03 1.69058770e-01 4.83762681e-01 -6.64290130e-01 3.76695544e-01 7.46585548e-01 9.69417274e-01 -7.37821758e-01 1.00695908e+00 -2.58574963e-01 6.41661346e-01 -4.40764844e-01 2.52693862e-01 4.45631295e-01 -1.52133986e-01 4.14666265e-01 1.27912545e+00 -1.45760663e-02 -1.50428906e-01 4.04330380e-02 1.03034616e+00 -1.67100579e-01 -6.43371046e-02 -2.80262589e-01 4.55519766e-01 7.23075413e-04 1.25048769e+00 -1.12541449e+00 -2.20417380e-01 -5.31601846e-01 8.57556880e-01 4.17358160e-01 6.57986224e-01 -8.21041465e-01 1.04034781e-01 5.67016125e-01 1.47609830e-01 8.66435707e-01 -3.12147856e-01 -1.57379538e-01 -1.29538441e+00 -1.11033767e-03 -6.24406397e-01 2.06260588e-02 -6.03465557e-01 -8.63745928e-01 5.97494245e-01 -1.15669109e-01 -8.97010267e-01 1.25349700e-01 -8.33182037e-01 -5.22366285e-01 6.55530035e-01 -1.72047627e+00 -1.28566802e+00 -2.78900534e-01 7.95673370e-01 4.05043811e-01 1.10861413e-01 5.70418239e-01 1.40627280e-01 -3.49841565e-01 4.03200865e-01 8.26675538e-03 2.80587852e-01 4.08604294e-01 -1.29244637e+00 7.37510204e-01 1.16876006e+00 6.26966894e-01 7.03814149e-01 6.94948554e-01 -2.59819299e-01 -1.15528023e+00 -1.21236908e+00 2.47127756e-01 -3.54056060e-01 7.46201336e-01 -5.63337028e-01 -9.97767389e-01 9.57353890e-01 2.11403266e-01 6.32487714e-01 -2.32102140e-03 5.96510395e-02 -4.87502068e-01 -2.18101874e-01 -8.01038444e-01 4.43464458e-01 1.13143194e+00 -4.59877640e-01 -4.75997388e-01 3.10965091e-01 7.00247169e-01 -5.57900012e-01 -5.66489637e-01 3.25617522e-01 3.85583937e-01 -1.17695308e+00 1.43905771e+00 -6.97755635e-01 3.44290845e-02 -1.66578546e-01 5.93998618e-02 -1.11691177e+00 -5.35243154e-01 -7.08254218e-01 4.07396674e-01 1.00846612e+00 3.11393887e-01 -6.74268007e-01 9.06882763e-01 4.68137026e-01 -2.69128740e-01 -3.44198376e-01 -9.92973864e-01 -6.76344693e-01 2.55733579e-01 -3.51919830e-01 3.62301618e-01 6.02019966e-01 -6.44628823e-01 -8.61719809e-03 -3.57007325e-01 3.76232326e-01 4.36770469e-01 -1.84083685e-01 6.46910429e-01 -1.23570120e+00 -3.18715751e-01 -5.82652450e-01 -7.18330324e-01 -1.25028741e+00 4.64904964e-01 -9.15652215e-01 -1.52466178e-01 -1.10372353e+00 -1.32789254e-01 -3.73488039e-01 -3.04176122e-01 5.76996207e-01 6.24060184e-02 4.56925184e-01 3.04332256e-01 3.53196114e-02 -3.59423548e-01 3.62429649e-01 1.27063262e+00 -9.94647741e-02 -2.55678445e-01 1.12331115e-01 -4.54032481e-01 1.09853888e+00 5.81860185e-01 -2.18422875e-01 -4.51869041e-01 -6.99302197e-01 1.84558064e-01 -3.18868250e-01 8.20156932e-01 -1.13164127e+00 2.95591831e-01 -1.47898763e-01 4.84794319e-01 -1.86632037e-01 2.38736272e-01 -9.10512090e-01 1.45069286e-01 3.95802885e-01 -5.50042868e-01 -1.72559962e-01 2.57552654e-01 6.40889227e-01 3.96280661e-02 -2.71435410e-01 1.10892415e+00 -1.07288510e-01 -9.67562914e-01 3.39322388e-01 -3.00622374e-01 5.41209839e-02 9.24543500e-01 -2.21622556e-01 -1.73267201e-01 -2.05339953e-01 -1.09833491e+00 -7.82029107e-02 5.26651561e-01 3.45614791e-01 5.06438613e-01 -1.00804031e+00 -5.45931041e-01 4.16604221e-01 -3.88222426e-01 1.04721136e-01 1.37840077e-01 7.46480227e-01 -8.25758994e-01 2.86548167e-01 -5.40327311e-01 -7.61744738e-01 -9.11928236e-01 5.66498339e-01 7.22028553e-01 8.45056586e-03 -1.01359117e+00 1.06463671e+00 6.15473330e-01 -1.57788485e-01 2.81999737e-01 -7.81306505e-01 -1.97018549e-01 3.02772708e-02 2.62809306e-01 -2.04412073e-01 3.64126652e-01 -6.79063320e-01 -3.72320384e-01 7.36281097e-01 -2.32401699e-01 -7.40852803e-02 1.51470029e+00 -8.32420588e-03 -5.70874065e-02 2.53848229e-02 1.10782611e+00 7.70669337e-03 -1.86752510e+00 -4.66795385e-01 6.70847893e-02 -4.73849267e-01 2.57108033e-01 -4.75184321e-01 -1.71277392e+00 8.54195952e-01 5.73721647e-01 1.38339445e-01 1.01183903e+00 -5.05194627e-02 2.50854492e-01 4.65403676e-01 1.38117358e-01 -8.97559106e-01 4.39868085e-02 6.55551732e-01 7.95112073e-01 -8.28361869e-01 -1.69918835e-01 -5.04998624e-01 -3.32736701e-01 1.36614954e+00 4.20718908e-01 -8.92362416e-01 8.52898836e-01 2.80732840e-01 -1.80228669e-02 -3.24100582e-03 -5.54960608e-01 -4.75653350e-01 5.21221638e-01 6.66872203e-01 2.82096624e-01 -2.26741403e-01 -8.70876759e-02 1.78123489e-01 -2.16880322e-01 -2.28333876e-01 4.90282267e-01 6.15818381e-01 -6.36431158e-01 -9.86756861e-01 -1.48864821e-01 2.22849593e-01 -3.71815771e-01 -1.95766374e-01 -1.28478870e-01 1.01668835e+00 1.65447503e-01 4.46341634e-01 2.99220055e-01 -5.13021052e-02 2.86024302e-01 -1.58401564e-01 5.87504745e-01 -5.80248475e-01 -5.58161974e-01 2.65883356e-01 -2.50128150e-01 -6.71786010e-01 -7.31103301e-01 -6.77265108e-01 -1.14853811e+00 -9.98890325e-02 -3.26857239e-01 -2.52238214e-01 4.37213868e-01 1.06062174e+00 1.06833875e-01 5.97517431e-01 3.03489357e-01 -1.22342765e+00 -3.82820100e-01 -5.33732653e-01 -3.82651955e-01 5.30404568e-01 6.52388692e-01 -6.83108747e-01 -2.52723068e-01 4.04428631e-01]
[9.095501899719238, 2.229712963104248]