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
e1e9fd84-8bf2-495b-82da-510f03a1fa98
clip-models-are-few-shot-learners-empirical
2203.07190
null
https://arxiv.org/abs/2203.07190v1
https://arxiv.org/pdf/2203.07190v1.pdf
CLIP Models are Few-shot Learners: Empirical Studies on VQA and Visual Entailment
CLIP has shown a remarkable zero-shot capability on a wide range of vision tasks. Previously, CLIP is only regarded as a powerful visual encoder. However, after being pre-trained by language supervision from a large amount of image-caption pairs, CLIP itself should also have acquired some few-shot abilities for vision-language tasks. In this work, we empirically show that CLIP can be a strong vision-language few-shot learner by leveraging the power of language. We first evaluate CLIP's zero-shot performance on a typical visual question answering task and demonstrate a zero-shot cross-modality transfer capability of CLIP on the visual entailment task. Then we propose a parameter-efficient fine-tuning strategy to boost the few-shot performance on the vqa task. We achieve competitive zero/few-shot results on the visual question answering and visual entailment tasks without introducing any additional pre-training procedure.
['Furu Wei', 'Ting Liu', 'Wei-Nan Zhang', 'Li Dong', 'Haoyu Song']
2022-03-14
null
https://aclanthology.org/2022.acl-long.421
https://aclanthology.org/2022.acl-long.421.pdf
acl-2022-5
['visual-entailment']
['reasoning']
[ 1.76163111e-02 2.59146504e-02 -1.54573724e-01 -4.66140330e-01 -1.22594416e+00 -3.17622453e-01 9.10031080e-01 -1.75090700e-01 -6.01363063e-01 4.08598125e-01 2.02374816e-01 -4.66330141e-01 3.30598146e-01 -4.20883685e-01 -1.08790386e+00 -2.81938374e-01 3.81634116e-01 2.03494489e-01 2.63518274e-01 -3.09686154e-01 -1.59438118e-01 -1.24190442e-01 -1.50397110e+00 4.85977829e-01 8.06451857e-01 8.64115000e-01 4.59027976e-01 9.42777872e-01 -2.43507490e-01 1.34956527e+00 -3.12043935e-01 -6.86951816e-01 5.92455454e-02 -6.27388477e-01 -7.77608991e-01 1.39629170e-01 9.70551670e-01 -6.20431483e-01 -6.32019103e-01 9.42633212e-01 5.99788308e-01 2.16578677e-01 6.34623587e-01 -1.29273903e+00 -1.29106498e+00 2.45075881e-01 -6.30929530e-01 5.09672284e-01 4.24013644e-01 7.11403131e-01 1.14868963e+00 -1.35559809e+00 7.45545089e-01 1.24447489e+00 2.94243276e-01 7.54132330e-01 -1.07126701e+00 -4.95337337e-01 -8.52662511e-03 4.08996165e-01 -1.07148600e+00 -6.97570443e-01 5.81065416e-01 -4.82072562e-01 1.11716962e+00 -8.69311988e-02 4.30294156e-01 1.26932144e+00 -1.13865823e-01 1.02620327e+00 9.19025242e-01 -3.26307714e-01 1.69884101e-01 1.12358280e-01 1.67741597e-01 9.41837788e-01 -2.72755027e-01 5.83801977e-02 -7.30478108e-01 2.70686656e-01 3.79983395e-01 -1.83213018e-02 -4.21345830e-01 -3.73278499e-01 -1.04839706e+00 9.80353713e-01 8.10875356e-01 1.39556572e-01 -2.06389308e-01 5.54182053e-01 5.36757350e-01 4.08278525e-01 3.25525016e-01 3.69178534e-01 -1.27634034e-01 -1.62795484e-02 -7.92281628e-01 -1.27576783e-01 5.37904203e-01 1.18712544e+00 6.87642276e-01 1.98751390e-01 -9.53421295e-01 8.47605824e-01 1.50636822e-01 7.97310889e-01 1.33299917e-01 -9.84713972e-01 4.91686434e-01 2.38378048e-01 7.58933872e-02 -4.91006106e-01 1.37921184e-01 -1.57048807e-01 -7.82303512e-01 2.31243581e-01 4.74693865e-01 -3.32394801e-02 -1.25847459e+00 1.84504664e+00 1.75575726e-02 8.57213214e-02 1.86573312e-01 1.07609224e+00 1.18421626e+00 9.15336668e-01 3.16630453e-01 -1.05134673e-01 1.69434512e+00 -1.39213085e+00 -6.70759320e-01 -4.54380333e-01 3.97676885e-01 -4.86482352e-01 1.68427598e+00 -1.55200381e-02 -1.30860901e+00 -7.85812020e-01 -1.00889826e+00 -6.16216123e-01 -1.92057580e-01 -2.08937943e-01 5.35722792e-01 2.51794249e-01 -1.03194571e+00 -2.37818994e-02 -5.72893858e-01 -3.50309461e-01 8.52430046e-01 -1.90229133e-01 -4.38247114e-01 -6.99099123e-01 -1.11400115e+00 9.34333742e-01 1.25040421e-02 -3.09937894e-01 -1.46028316e+00 -9.92985904e-01 -1.07992172e+00 2.72445261e-01 5.80594242e-01 -1.20056510e+00 1.45194352e+00 -1.13972592e+00 -1.25478590e+00 1.18312538e+00 -3.34022433e-01 -6.67948544e-01 4.07599241e-01 -1.06491603e-01 -1.45015210e-01 6.24138236e-01 1.91088602e-01 9.89712059e-01 1.20600116e+00 -1.11550260e+00 -3.54560196e-01 -2.02386260e-01 5.38043797e-01 1.18644759e-01 -3.07157487e-01 1.40599951e-01 -8.45675945e-01 -3.24332863e-01 -7.51264691e-01 -6.26753628e-01 6.56485781e-02 5.10999084e-01 7.42448345e-02 -3.64602476e-01 4.98852074e-01 -5.73557019e-01 5.94129801e-01 -2.27055049e+00 4.05813605e-02 -6.20496869e-01 3.02263349e-01 5.66599131e-01 -6.39508963e-01 4.37404007e-01 1.52510554e-01 -2.30171308e-01 -5.03070354e-01 -4.79754627e-01 7.38663075e-04 2.73887604e-01 -4.84549195e-01 3.79574835e-01 5.03805578e-01 1.74085248e+00 -1.04445100e+00 -6.57834172e-01 2.48166978e-01 5.82937181e-01 -5.19526899e-01 5.21649837e-01 -5.49435019e-01 1.54014125e-01 -9.63720232e-02 6.29400373e-01 4.21171933e-01 -7.42104650e-01 -4.06684667e-01 -8.46253559e-02 1.95627317e-01 -3.73933613e-01 -1.05487198e-01 2.23963833e+00 -6.71443999e-01 1.00572205e+00 -6.03722706e-02 -8.72032225e-01 5.24479091e-01 3.49026382e-01 5.93089610e-02 -1.19057798e+00 1.55248150e-01 -2.35713542e-01 -1.83908679e-02 -9.71802890e-01 2.91706502e-01 -3.82031918e-01 -1.05849519e-01 3.22751969e-01 6.33420944e-01 -1.66496426e-01 1.28168806e-01 8.20604980e-01 9.36178565e-01 -1.40750706e-01 2.97834992e-01 6.69907779e-02 4.87866789e-01 4.16279286e-02 -4.38749567e-02 8.86840582e-01 -4.93159741e-01 8.15373421e-01 4.79998320e-01 2.05543879e-02 -1.25081778e+00 -1.15617335e+00 8.16242918e-02 1.56276727e+00 1.23539224e-01 -3.54802459e-01 -5.12176812e-01 -5.84881485e-01 -1.02452800e-01 8.46795678e-01 -7.49051094e-01 -3.70351315e-01 -1.14679020e-02 -1.77732795e-01 6.26698494e-01 6.93881631e-01 4.42622364e-01 -1.11604023e+00 -5.93948543e-01 -2.24388242e-01 -1.15310609e-01 -1.48323989e+00 -6.87631190e-01 -1.25479028e-01 -2.20183223e-01 -9.54183936e-01 -1.20968556e+00 -1.02297974e+00 4.71184671e-01 8.28294992e-01 1.41310287e+00 6.99447542e-02 -4.43299413e-01 9.45141196e-01 -5.17013609e-01 -4.02958930e-01 -3.17916214e-01 -3.74183476e-01 -4.00831580e-01 -4.26790379e-02 3.30861986e-01 -3.95490795e-01 -6.71145558e-01 -1.61372751e-01 -9.01305735e-01 8.02633017e-02 6.24137402e-01 1.01035452e+00 3.36829960e-01 -8.62460732e-01 7.01714933e-01 -5.41069329e-01 6.05013311e-01 -4.61351693e-01 -4.62220877e-01 5.52869499e-01 -1.09164916e-01 1.02305628e-01 5.02203405e-01 -4.65440601e-01 -1.04008543e+00 1.81668196e-02 -1.08802952e-01 -1.04171419e+00 7.24112615e-03 3.47625315e-01 -6.64894581e-02 -2.68423911e-02 6.90675318e-01 3.92615318e-01 9.53926817e-02 -1.54812396e-01 1.16169906e+00 5.54275274e-01 9.53484714e-01 -4.72455919e-01 8.45267355e-01 5.41768134e-01 -3.96041214e-01 -9.69241142e-01 -1.29800630e+00 -7.47969270e-01 -3.30653965e-01 -2.26757213e-01 1.40722620e+00 -1.46071076e+00 -7.39347994e-01 2.26241454e-01 -1.36421764e+00 -3.77609909e-01 -3.79095346e-01 1.34475097e-01 -6.87889099e-01 4.10466313e-01 -5.16038775e-01 -5.91627240e-01 -6.13640368e-01 -9.31398451e-01 1.11249900e+00 2.13986382e-01 3.70278805e-01 -7.76657581e-01 4.84202094e-02 6.23093903e-01 3.25801075e-01 -1.44007936e-01 7.29624927e-01 -2.96196312e-01 -6.33632123e-01 5.34505844e-02 -8.13183188e-01 4.60130721e-01 -4.14267927e-01 -4.16464120e-01 -1.22945869e+00 -4.95214403e-01 -6.67889928e-03 -1.08634937e+00 1.36571431e+00 1.43534288e-01 1.07852662e+00 -5.50472029e-02 1.81988195e-01 7.82880902e-01 1.61113775e+00 -2.32668787e-01 6.19322181e-01 -1.80252329e-01 8.52166593e-01 2.67461807e-01 4.18169528e-01 1.17927969e-01 6.48101449e-01 4.19255167e-01 5.90306580e-01 -2.61331916e-01 -5.34401476e-01 -4.32751209e-01 4.34141874e-01 6.58287644e-01 8.76149237e-02 -2.28246644e-01 -9.31097686e-01 8.32259417e-01 -1.77431977e+00 -1.12544990e+00 7.37996995e-02 1.78109443e+00 8.29963326e-01 -1.28971219e-01 1.19104356e-01 -4.79100049e-01 3.86356801e-01 4.26156312e-01 -7.21180499e-01 -3.13370615e-01 -1.94317356e-01 1.79988027e-01 2.44730219e-01 4.39069122e-01 -8.62018943e-01 1.14829302e+00 6.38982487e+00 6.39500618e-01 -8.74339044e-01 5.74458539e-01 4.01574254e-01 -3.25704545e-01 -3.60091627e-01 4.90731560e-03 -4.11286801e-01 3.32697541e-01 7.77313948e-01 -4.13573354e-01 2.64407903e-01 8.11547995e-01 -3.03871036e-01 -9.33398679e-02 -1.24587321e+00 1.38928306e+00 6.53611958e-01 -1.50395691e+00 2.29230657e-01 -1.52117759e-01 8.15162659e-01 4.66406703e-01 9.60974544e-02 6.74385190e-01 1.69174775e-01 -1.26396477e+00 5.25360584e-01 5.59400558e-01 1.15745986e+00 -5.22782922e-01 3.34606081e-01 3.34580988e-01 -8.43566418e-01 -1.14334643e-01 -6.63035631e-01 -4.66671623e-02 4.72438455e-01 2.61199862e-01 -3.55842590e-01 4.24153060e-01 5.42003751e-01 7.10168302e-01 -7.00321674e-01 8.83619070e-01 -3.66785020e-01 5.13826430e-01 2.91598111e-01 1.35284588e-01 4.65176165e-01 1.33723125e-01 4.12660420e-01 1.14347005e+00 -1.07479133e-02 3.56631547e-01 1.06046967e-01 1.00008917e+00 -3.67180765e-01 -6.82550371e-02 -8.42999518e-01 -2.16084719e-01 1.09775797e-01 1.14417064e+00 1.86479073e-02 -7.38720059e-01 -1.07155156e+00 1.31981647e+00 5.82911789e-01 6.38159037e-01 -9.96756792e-01 -3.79965752e-01 5.71895480e-01 -8.83461460e-02 6.27480090e-01 -2.35140890e-01 1.55200258e-01 -1.46585155e+00 -4.86308075e-02 -8.08374822e-01 2.96986163e-01 -1.47698319e+00 -1.49358833e+00 5.69503427e-01 -1.98387027e-01 -9.18056428e-01 -2.52122909e-01 -7.22241938e-01 -6.42285526e-01 7.05303550e-01 -1.77196014e+00 -1.50229526e+00 -4.99015301e-01 1.02823567e+00 8.76130521e-01 -3.11134875e-01 6.38073444e-01 1.84882566e-01 -2.40042537e-01 6.92675114e-01 -1.11380853e-01 2.80417085e-01 7.89088428e-01 -1.06281614e+00 3.66166830e-01 8.89613926e-01 5.20538330e-01 3.47694814e-01 6.82565391e-01 -2.05902413e-01 -1.67864287e+00 -1.07054305e+00 5.67850530e-01 -5.84338188e-01 8.54622543e-01 -5.15424013e-01 -1.11456954e+00 7.27145672e-01 8.19610119e-01 4.70270216e-01 4.63980138e-01 1.31029601e-03 -9.63226497e-01 4.55752611e-02 -6.79123759e-01 5.49851298e-01 9.89862919e-01 -1.27424252e+00 -1.08189511e+00 5.03230095e-01 1.20608890e+00 -4.67091799e-02 -6.28048003e-01 1.21574186e-01 3.16631287e-01 -7.60549307e-01 1.23926044e+00 -1.02268207e+00 9.96171713e-01 -3.52559872e-02 -3.29381049e-01 -1.07572818e+00 -2.89252013e-01 -4.26576108e-01 -3.42797428e-01 1.10443115e+00 1.74054757e-01 -1.45664215e-01 2.52658159e-01 3.33847523e-01 -1.57272011e-01 -5.43790638e-01 -7.19422400e-01 -8.73948216e-01 2.26305559e-01 -4.49274778e-01 3.82649414e-02 7.77894855e-01 -1.26035646e-01 1.01572049e+00 -8.09256136e-01 -1.83001280e-01 7.71510303e-01 1.54475451e-01 1.03057933e+00 -6.84498847e-01 -6.03679597e-01 -2.25398198e-01 -2.94926107e-01 -1.06082451e+00 3.43052864e-01 -1.15337229e+00 2.96106905e-01 -1.67100871e+00 7.21704125e-01 4.58073527e-01 -4.34670657e-01 4.00463402e-01 -4.64579910e-01 4.87836689e-01 6.72441840e-01 1.24361470e-01 -1.19054353e+00 7.76694000e-01 1.51909125e+00 -4.48968589e-01 2.25117996e-01 -4.75815117e-01 -8.00270796e-01 5.06785274e-01 3.65809739e-01 -1.04612291e-01 -7.02849030e-01 -8.03130090e-01 2.47108623e-01 2.35218793e-01 7.48339474e-01 -7.73739398e-01 2.84565419e-01 4.25561005e-03 2.09709167e-01 -4.26257432e-01 5.71788847e-01 -3.59013557e-01 -6.12768233e-01 2.45940194e-01 -4.42437440e-01 -1.72560245e-01 2.39939079e-01 8.72911513e-01 -3.23561996e-01 3.20770890e-02 9.34221745e-01 -2.79086202e-01 -1.13755941e+00 6.58504605e-01 -1.19267516e-01 6.77062929e-01 1.06906569e+00 7.94882476e-02 -6.26094282e-01 -5.79138875e-01 -5.51159084e-01 5.82339823e-01 5.74909031e-01 6.02679491e-01 8.76080155e-01 -1.21069336e+00 -8.39893818e-01 4.87384479e-03 7.73916125e-01 -3.11905801e-01 5.80624998e-01 8.40507925e-01 -2.19410375e-01 4.39574182e-01 -4.37987059e-01 -7.32688010e-01 -1.14927113e+00 1.20957386e+00 2.28826568e-01 1.74375519e-01 -9.98765588e-01 1.06983328e+00 7.51430690e-01 1.12443671e-01 3.11448246e-01 -3.94641235e-02 1.09961048e-01 -2.70099472e-03 9.79731679e-01 -1.11546084e-01 -4.85807121e-01 -5.32870650e-01 -2.79600888e-01 5.65597057e-01 -2.43038032e-02 -3.50203454e-01 1.22576523e+00 -2.75779575e-01 2.25016341e-01 6.07819617e-01 1.38370621e+00 -4.40613031e-01 -1.59718597e+00 -4.43670303e-01 -2.96255261e-01 -4.12779987e-01 -7.90142342e-02 -8.10083032e-01 -1.01232386e+00 1.52130306e+00 4.32707965e-01 -2.06148818e-01 1.02585804e+00 6.04921818e-01 8.69798243e-01 4.25472707e-01 8.34891126e-02 -7.74589300e-01 6.77825689e-01 5.41039586e-01 1.10720837e+00 -1.86667693e+00 -3.11831534e-01 -6.33734092e-02 -1.13737273e+00 5.13578296e-01 5.12449145e-01 -2.74534374e-01 4.13977474e-01 -1.08936280e-01 4.51298468e-02 -2.71692693e-01 -1.22600830e+00 -8.29164922e-01 5.63934267e-01 7.77271092e-01 1.72999829e-01 -1.00113958e-01 4.32876833e-02 4.10667002e-01 1.72142029e-01 2.13974550e-01 3.82690430e-01 4.81077373e-01 -6.80813670e-01 -4.60653394e-01 1.13126732e-01 1.66737899e-01 -2.92060018e-01 -3.53640169e-01 -3.63870025e-01 6.70720279e-01 -2.32208744e-01 8.78679574e-01 1.98995039e-01 -2.55611483e-02 1.22365452e-01 1.63847521e-01 8.51119995e-01 -5.67198873e-01 -3.36416483e-01 -2.46360943e-01 -9.19533893e-02 -7.44455099e-01 -3.67062718e-01 -9.05375332e-02 -1.02087879e+00 -9.66800228e-02 1.90671217e-02 -2.80985981e-01 5.41804731e-01 1.07090902e+00 3.20989788e-01 5.75186312e-01 1.92020833e-01 -3.35542351e-01 -6.52504981e-01 -8.31722379e-01 -3.00388157e-01 7.31656969e-01 5.97749710e-01 -4.75535989e-01 -1.63955346e-01 1.57675624e-01]
[10.769235610961914, 1.6640204191207886]
e16309be-c0be-4e82-baed-82fb272d6d01
diffpose-toward-more-reliable-3d-pose
2211.16940
null
https://arxiv.org/abs/2211.16940v3
https://arxiv.org/pdf/2211.16940v3.pdf
DiffPose: Toward More Reliable 3D Pose Estimation
Monocular 3D human pose estimation is quite challenging due to the inherent ambiguity and occlusion, which often lead to high uncertainty and indeterminacy. On the other hand, diffusion models have recently emerged as an effective tool for generating high-quality images from noise. Inspired by their capability, we explore a novel pose estimation framework (DiffPose) that formulates 3D pose estimation as a reverse diffusion process. We incorporate novel designs into our DiffPose to facilitate the diffusion process for 3D pose estimation: a pose-specific initialization of pose uncertainty distributions, a Gaussian Mixture Model-based forward diffusion process, and a context-conditioned reverse diffusion process. Our proposed DiffPose significantly outperforms existing methods on the widely used pose estimation benchmarks Human3.6M and MPI-INF-3DHP. Project page: https://gongjia0208.github.io/Diffpose/.
['Jun Liu', 'Hossein Rahmani', 'Qiuhong Ke', 'Zhipeng Fan', 'Lin Geng Foo', 'Jia Gong']
2022-11-30
null
http://openaccess.thecvf.com//content/CVPR2023/html/Gong_DiffPose_Toward_More_Reliable_3D_Pose_Estimation_CVPR_2023_paper.html
http://openaccess.thecvf.com//content/CVPR2023/papers/Gong_DiffPose_Toward_More_Reliable_3D_Pose_Estimation_CVPR_2023_paper.pdf
cvpr-2023-1
['3d-pose-estimation', '3d-human-pose-estimation', 'monocular-3d-human-pose-estimation']
['computer-vision', 'computer-vision', 'computer-vision']
[-3.93446714e-01 -1.21387690e-01 3.00602973e-01 -2.34331846e-01 -8.31042707e-01 -4.80461150e-01 7.11201251e-01 -5.13368189e-01 -4.02764827e-01 6.85291409e-01 4.04283792e-01 1.66139200e-01 9.79496613e-02 -4.73572940e-01 -5.96771240e-01 -6.34601712e-01 3.04750174e-01 9.53120470e-01 1.35292679e-01 -8.88389070e-03 4.74373773e-02 4.13326323e-01 -1.29864347e+00 -3.37293267e-01 8.87101889e-01 8.57605636e-01 2.80434787e-01 7.70331740e-01 8.32033455e-02 1.87919527e-01 -6.74817562e-01 -4.00172293e-01 5.23931265e-01 -3.74750704e-01 -5.55047929e-01 1.77818716e-01 4.48528349e-01 -5.01394928e-01 -2.68907696e-01 8.60599220e-01 9.91880238e-01 1.82070538e-01 9.08724785e-01 -1.30984879e+00 -2.29083896e-01 9.97589007e-02 -8.84484291e-01 6.22269399e-02 6.26901090e-01 2.94193834e-01 3.41793984e-01 -1.03954327e+00 9.32952821e-01 1.40725672e+00 5.51695466e-01 6.93735659e-01 -1.18263924e+00 -6.60406709e-01 8.90925080e-02 5.16175404e-02 -1.59855115e+00 -3.18377256e-01 8.01114261e-01 -5.44251502e-01 4.51539844e-01 2.66208887e-01 7.24034131e-01 1.55014741e+00 3.51135969e-01 8.55853498e-01 1.21579444e+00 -3.08352187e-02 3.45444471e-01 -3.35022956e-01 -3.92348170e-01 5.41456163e-01 1.93376422e-01 1.12111390e-01 -6.17204607e-01 -2.23575026e-01 1.22821581e+00 -3.44636232e-01 -2.75655866e-01 -8.00526083e-01 -1.25334537e+00 4.38515514e-01 3.51853848e-01 -2.02749208e-01 -5.08726537e-01 3.88882875e-01 3.76491919e-02 -1.09355152e-01 5.41659474e-01 2.13366777e-01 -3.10884953e-01 -3.60236317e-01 -7.76326358e-01 7.31678784e-01 7.12060809e-01 1.09929514e+00 3.71161669e-01 -2.15188116e-01 -2.66754091e-01 6.52346849e-01 6.40805542e-01 8.50612640e-01 1.10651636e-02 -1.33184266e+00 3.75305057e-01 1.27555579e-01 4.10934865e-01 -9.60891306e-01 -4.83164132e-01 -6.03640258e-01 -8.04314673e-01 2.99879521e-01 6.48644209e-01 -2.40558416e-01 -1.03644001e+00 1.78934097e+00 9.36493754e-01 6.99070245e-02 -5.10372519e-01 1.27028549e+00 7.70824075e-01 4.43112552e-01 -1.14152143e-02 1.76819995e-01 1.07631934e+00 -9.27841663e-01 -6.41984284e-01 -1.79630652e-01 2.18325958e-01 -8.92049491e-01 7.57676601e-01 5.29771745e-01 -1.18314445e+00 -4.86358821e-01 -8.31938922e-01 -1.54195249e-01 1.93511080e-02 1.04276702e-01 5.40468454e-01 6.61427617e-01 -9.12558317e-01 4.42507148e-01 -9.59476352e-01 -2.29493454e-01 3.18116277e-01 2.57240117e-01 -4.50852036e-01 -2.04241514e-01 -7.99030602e-01 9.17624235e-01 -1.38730016e-02 3.78338516e-01 -9.12207127e-01 -5.76406837e-01 -8.05195630e-01 -5.91351807e-01 5.75250864e-01 -1.18837535e+00 1.06818354e+00 -1.52482213e-02 -1.73964405e+00 8.92414033e-01 -5.47072962e-02 -1.13211937e-01 1.27499807e+00 -7.21257448e-01 2.11881772e-01 5.41140325e-02 6.27519190e-02 8.11602831e-01 9.85577166e-01 -1.41899037e+00 -1.09036058e-01 -6.12928331e-01 -2.03114480e-01 5.93717456e-01 2.35440984e-01 -2.27663219e-01 -1.05315506e+00 -7.37475872e-01 4.36646223e-01 -1.28258324e+00 -4.48837608e-01 3.06992739e-01 -6.55590653e-01 7.87434727e-02 4.90386248e-01 -8.41938138e-01 1.11001539e+00 -1.76779544e+00 7.68626451e-01 2.38150522e-01 3.91181499e-01 -6.14489727e-02 -2.24579573e-02 1.83286026e-01 3.08212161e-01 -1.33731931e-01 -1.59641802e-01 -9.26370978e-01 2.30830714e-01 1.76799193e-01 1.78519353e-01 7.27961659e-01 -4.38066535e-02 9.91463721e-01 -8.97215903e-01 -6.71095252e-01 4.86778051e-01 8.53922844e-01 -6.11201346e-01 2.33236924e-01 -2.34993607e-01 1.00863242e+00 -4.82059509e-01 7.02257335e-01 1.00601912e+00 -2.33838156e-01 -1.24452755e-01 -3.71329248e-01 8.36505517e-02 -1.93978101e-01 -1.53082061e+00 2.13187218e+00 -1.07448809e-01 1.42841190e-01 2.52819002e-01 -1.64249659e-01 6.42652214e-01 1.77213848e-01 6.18499994e-01 -1.55858248e-01 5.34733653e-01 1.14888214e-01 -2.23150298e-01 -2.17700630e-01 5.25704205e-01 5.20252362e-02 -9.82626006e-02 2.49652743e-01 5.09733818e-02 -6.65127218e-01 7.06670035e-05 1.99754849e-01 9.31216240e-01 7.60858834e-01 5.99878654e-02 -1.59733608e-01 2.96022773e-01 -2.90407330e-01 5.66214502e-01 5.15496612e-01 -4.50287193e-01 1.25321972e+00 1.87628657e-01 -2.19375074e-01 -9.71461713e-01 -1.54243183e+00 -1.33609511e-02 3.57135624e-01 3.17129552e-01 -4.72434044e-01 -8.59496593e-01 -7.34416664e-01 6.15835153e-02 3.47366512e-01 -6.62457049e-01 7.86476210e-02 -4.40341592e-01 -7.00698018e-01 4.19596523e-01 5.34227192e-01 5.62206089e-01 -5.79977393e-01 -4.12221342e-01 5.04382662e-02 -4.54664260e-01 -1.16862226e+00 -7.90326416e-01 -1.17947191e-01 -6.84788048e-01 -9.57246900e-01 -1.36438298e+00 -2.54485369e-01 7.84988225e-01 -1.08183898e-01 1.13227689e+00 -2.07070559e-01 -2.80665666e-01 5.53570271e-01 -8.84312317e-02 -3.09425533e-01 -2.28968963e-01 -7.26704448e-02 3.72374058e-01 -2.52654046e-01 1.62415393e-02 -5.49511254e-01 -1.00123131e+00 6.83408558e-01 -5.60829103e-01 1.59599215e-01 4.56946462e-01 5.45801222e-01 7.57233262e-01 -2.61434287e-01 1.05313502e-01 -5.14010072e-01 5.56639850e-01 -2.18852013e-01 -6.06847942e-01 -8.59519467e-02 -2.39845067e-01 -7.46003911e-03 -1.99191093e-01 -5.72085202e-01 -1.21916056e+00 2.72265285e-01 -3.95346850e-01 -5.32978415e-01 -2.13132426e-01 5.84473796e-02 -3.54563653e-01 -2.09766820e-01 7.45723009e-01 3.06174792e-02 2.61366665e-01 -5.56022286e-01 4.26111281e-01 3.67014825e-01 5.53834856e-01 -8.10370088e-01 1.06889820e+00 5.89605033e-01 7.87350908e-02 -7.70784080e-01 -5.97203255e-01 -3.71037394e-01 -8.71203482e-01 -4.67913926e-01 1.05100322e+00 -1.02369118e+00 -5.36230266e-01 8.64227355e-01 -1.18543732e+00 -4.30720925e-01 -7.89440945e-02 4.58327800e-01 -7.63436556e-01 5.65752864e-01 -7.11121917e-01 -7.37302184e-01 -2.32558787e-01 -1.33909166e+00 1.44783974e+00 1.19548798e-01 -5.67398489e-01 -7.63353169e-01 1.02574393e-01 5.47836542e-01 2.88218349e-01 6.17734671e-01 1.28435105e-01 1.06811419e-01 -6.40329719e-01 -1.17220975e-01 -1.72219742e-02 2.91046768e-01 -3.96313630e-02 -1.57166272e-01 -8.14747214e-01 -2.40677342e-01 -7.42270052e-02 -2.21926555e-01 4.43843901e-01 7.43477643e-01 8.56848240e-01 3.34808141e-01 -4.01046872e-01 7.29153335e-01 9.67399538e-01 -2.90514201e-01 7.57411957e-01 2.21088141e-01 8.06424737e-01 3.30309749e-01 6.53946280e-01 6.48685575e-01 5.43532670e-01 9.58163917e-01 3.33352268e-01 1.25171989e-01 -3.77092719e-01 -4.02734637e-01 9.26859900e-02 7.37676322e-01 -3.94558340e-01 -2.22509041e-01 -8.96046937e-01 1.38149276e-01 -1.53542209e+00 -5.19032598e-01 -1.56040251e-01 2.24528599e+00 8.12591195e-01 1.26386046e-01 2.44504914e-01 -7.93606415e-03 4.55022007e-01 5.24805114e-02 -4.73116338e-01 4.02737379e-01 4.32991683e-02 -7.05163479e-02 3.74583513e-01 5.84893584e-01 -1.07006383e+00 9.07196224e-01 5.83962440e+00 9.39466476e-01 -7.41615832e-01 3.73519361e-02 5.94948113e-01 -2.79869258e-01 -1.70772105e-01 -3.04969996e-01 -9.50729132e-01 4.64035034e-01 2.96935797e-01 9.21301320e-02 1.19485468e-01 7.07290888e-01 2.13469952e-01 -4.99527305e-01 -8.45910549e-01 1.30074322e+00 -1.11136734e-02 -8.70623052e-01 -1.73970953e-01 3.71418685e-01 7.31429040e-01 -1.43854529e-01 1.91849470e-01 8.84655714e-02 1.98437840e-01 -8.51495385e-01 9.11165178e-01 6.82725430e-01 6.10121906e-01 -5.92444241e-01 5.35297692e-01 4.59303975e-01 -1.04301560e+00 3.54969501e-01 -1.80018872e-01 1.46295264e-01 6.94374144e-01 7.98659921e-01 -5.85807085e-01 5.60734749e-01 6.85277045e-01 3.84731323e-01 -5.56121051e-01 1.19703805e+00 -4.86579776e-01 1.49288222e-01 -6.01609468e-01 8.39997679e-02 -2.35664666e-01 -2.84100413e-01 8.17008853e-01 7.87273228e-01 3.22730005e-01 3.72117199e-02 6.34088591e-02 9.85484660e-01 1.16491705e-01 -2.20334232e-01 -3.02492440e-01 3.38817209e-01 4.28176463e-01 1.13060260e+00 -8.31875920e-01 -1.07973784e-01 8.08513910e-02 1.39593673e+00 1.02712870e-01 2.75363445e-01 -1.15639079e+00 1.60890535e-01 7.41007566e-01 2.25349799e-01 7.38987029e-02 -7.41820633e-01 -1.82802841e-01 -1.23518991e+00 8.30218866e-02 -8.38574171e-01 9.07317400e-02 -9.30946231e-01 -1.20159709e+00 6.10919058e-01 3.78081828e-01 -1.25622058e+00 -3.55660111e-01 -4.79759246e-01 -1.44993812e-01 8.99766445e-01 -9.47304666e-01 -1.11209035e+00 -5.09788215e-01 3.81437570e-01 4.60281551e-01 3.65997553e-01 4.97961938e-01 4.32643503e-01 -3.38409066e-01 3.43443274e-01 -3.50954980e-01 -6.23628162e-02 8.24030459e-01 -1.30348885e+00 6.34965181e-01 6.08660281e-01 -1.08720278e-02 5.99267006e-01 9.98768985e-01 -1.03138447e+00 -1.47408009e+00 -7.40414977e-01 4.40761209e-01 -1.09680140e+00 2.68606961e-01 -5.59779048e-01 -4.81319666e-01 4.77832347e-01 -1.61204666e-01 8.73539373e-02 2.67344177e-01 -1.53369009e-01 5.98890707e-02 2.67788172e-01 -1.17725623e+00 8.05870175e-01 1.60567892e+00 -1.54246360e-01 -3.66672128e-01 2.30786115e-01 4.94150758e-01 -1.13595963e+00 -8.71505320e-01 4.76358205e-01 7.00884044e-01 -8.74766529e-01 1.34196627e+00 1.32708892e-01 2.22191274e-01 -4.44352120e-01 -9.65735093e-02 -1.39421856e+00 -2.61744499e-01 -8.06702077e-01 -3.64145279e-01 1.04978716e+00 7.92000890e-02 -3.65409553e-01 1.16079724e+00 7.07726300e-01 1.30190194e-01 -6.61359131e-01 -9.66798544e-01 -9.07480955e-01 3.77643667e-02 -5.55795968e-01 3.80031317e-01 3.94779384e-01 -5.77943623e-01 9.27154645e-02 -5.64056754e-01 1.93064019e-01 1.11092007e+00 -3.02273780e-01 1.27048922e+00 -1.00117147e+00 -4.90267426e-01 -2.28696853e-01 -3.77874196e-01 -1.61695468e+00 -1.90361291e-01 -2.54529357e-01 2.93684840e-01 -1.44697750e+00 3.67398635e-02 -4.36912507e-01 2.60110050e-01 -1.18555143e-01 -2.05287889e-01 2.82862902e-01 1.24829382e-01 5.04369326e-02 -5.22713482e-01 7.87213445e-01 1.69765615e+00 2.13604614e-01 -7.68860653e-02 2.49215066e-01 -2.86480039e-01 7.91304052e-01 6.47639811e-01 -2.94256508e-01 -4.27860975e-01 -5.04086792e-01 2.62147844e-01 9.62241739e-02 4.96390283e-01 -1.25764644e+00 1.20630987e-01 -8.59515890e-02 6.39247954e-01 -9.25085425e-01 8.19945812e-01 -6.72645926e-01 5.68083167e-01 4.67891008e-01 7.30995461e-02 1.40874982e-01 7.40700066e-02 6.83853030e-01 1.19663052e-01 2.10665807e-01 6.65667236e-01 -2.46549010e-01 -6.27534628e-01 5.64142346e-01 -1.06356286e-01 2.97695547e-01 8.81603122e-01 -2.17900261e-01 3.61635350e-02 -5.94415545e-01 -8.87054920e-01 2.95248181e-01 5.87658465e-01 5.38788617e-01 6.91374958e-01 -1.41619945e+00 -6.80917203e-01 3.54215838e-02 -5.10731302e-02 3.76694381e-01 3.55531901e-01 1.01038218e+00 -6.42905176e-01 -4.54655178e-02 -6.86933845e-02 -7.48184443e-01 -1.32420325e+00 2.59602815e-01 3.11145753e-01 -1.75761983e-01 -5.52235067e-01 1.23557115e+00 1.41755193e-01 -6.01753771e-01 2.71894664e-01 -3.49787980e-01 2.17834383e-01 -2.70931482e-01 3.40799898e-01 5.91539085e-01 -2.09220320e-01 -8.45750988e-01 -4.28430617e-01 7.94377983e-01 3.19335699e-01 -4.94916856e-01 1.11447847e+00 -4.60565776e-01 1.84379831e-01 1.78773105e-01 8.96049440e-01 1.82737693e-01 -1.68491268e+00 -1.43118920e-02 -2.28715032e-01 -6.39264703e-01 -2.16222797e-02 -8.02696288e-01 -9.08024371e-01 6.26983464e-01 5.23319125e-01 -5.44485390e-01 7.48611927e-01 1.03725269e-01 7.02176869e-01 -1.09854840e-01 8.45131636e-01 -1.12594509e+00 3.10812414e-01 4.69789773e-01 1.21520030e+00 -1.21776307e+00 2.13432327e-01 -8.59901607e-01 -4.83311296e-01 6.91182435e-01 7.63069510e-01 -4.96474840e-02 7.58075297e-01 4.69812036e-01 2.67278492e-01 -5.93021102e-02 -3.56402963e-01 -8.64372179e-02 6.42898560e-01 8.36059928e-01 3.04034114e-01 -8.82987529e-02 -3.51029515e-01 4.81646299e-01 -3.14931095e-01 1.07221596e-01 1.79854155e-01 1.03228605e+00 -5.59154786e-02 -1.16556215e+00 -7.90480137e-01 5.16174920e-02 -1.59490660e-01 2.45484024e-01 -2.58874267e-01 7.20086932e-01 4.78748046e-02 6.29137695e-01 -2.00273216e-01 -5.44349194e-01 3.29854459e-01 -1.18915975e-01 9.48398113e-01 -4.75037366e-01 -2.41207942e-01 3.52052271e-01 5.41103333e-02 -8.70809495e-01 -2.33087078e-01 -7.84462035e-01 -1.05346012e+00 -2.97403693e-01 -2.50624686e-01 -1.19447760e-01 7.83529639e-01 8.40055287e-01 4.73809510e-01 4.15134370e-01 -1.17302174e-02 -1.34570396e+00 -6.22045040e-01 -1.00056374e+00 -4.75386351e-01 4.80879188e-01 8.33711959e-03 -1.10534072e+00 -3.21439117e-01 -1.94667608e-01]
[6.991943359375, -1.06046724319458]
cbb1cb29-e9b6-47e3-a4dc-9e34392aaa4a
rakutens-participation-in-wat-2021-examining
null
null
https://aclanthology.org/2021.wat-1.9
https://aclanthology.org/2021.wat-1.9.pdf
Rakuten’s Participation in WAT 2021: Examining the Effectiveness of Pre-trained Models for Multilingual and Multimodal Machine Translation
This paper introduces our neural machine translation systems’ participation in the WAT 2021 shared translation tasks (team ID: sakura). We participated in the (i) NICT-SAP, (ii) Japanese-English multimodal translation, (iii) Multilingual Indic, and (iv) Myanmar-English translation tasks. Multilingual approaches such as mBART (Liu et al., 2020) are capable of pre-training a complete, multilingual sequence-to-sequence model through denoising objectives, making it a great starting point for building multilingual translation systems. Our main focus in this work is to investigate the effectiveness of multilingual finetuning on such a multilingual language model on various translation tasks, including low-resource, multimodal, and mixed-domain translation. We further explore a multimodal approach based on universal visual representation (Zhang et al., 2019) and compare its performance against a unimodal approach based on mBART alone.
['Ohnmar Htun', 'Mausam Jain', 'Sunil Yadav', 'Dongzhe Wang', 'Raymond Hendy Susanto']
null
null
null
null
acl-wat-2021-8
['multimodal-machine-translation']
['natural-language-processing']
[ 5.39927445e-02 -3.08025628e-01 -1.24327101e-01 -1.24963284e-01 -1.57668674e+00 -8.09642851e-01 8.94712806e-01 -3.06659371e-01 -6.12637341e-01 1.01397204e+00 1.68910056e-01 -9.54364181e-01 6.05990291e-01 -2.69879431e-01 -9.00324821e-01 -5.84277153e-01 4.52261627e-01 9.12191510e-01 -4.27959412e-01 -4.98038113e-01 -5.74807189e-02 1.92969695e-01 -7.19965756e-01 6.64329648e-01 1.19423652e+00 4.11054134e-01 6.79555178e-01 6.24221921e-01 -3.31820361e-02 2.90424824e-01 -3.28592032e-01 -8.45429003e-01 5.82013540e-02 -7.57264972e-01 -8.24845612e-01 -2.38248616e-01 4.92485285e-01 -2.87407916e-02 -2.71434989e-02 1.01829696e+00 8.87031496e-01 -8.14918354e-02 7.34895110e-01 -9.40952897e-01 -1.09488440e+00 8.67407620e-01 -5.01732945e-01 -1.42466694e-01 3.11521530e-01 2.57986605e-01 8.26304793e-01 -1.42151260e+00 1.09352136e+00 1.28320825e+00 5.99725485e-01 4.16673094e-01 -1.28598011e+00 -4.04499531e-01 -6.45249411e-02 3.26269269e-01 -1.19920731e+00 -6.82347298e-01 3.25829715e-01 -2.79719740e-01 1.28884101e+00 2.08527356e-01 2.45533437e-01 1.75597513e+00 3.22676659e-01 8.99463177e-01 1.38211858e+00 -8.04296851e-01 -2.74526447e-01 2.62582153e-01 -4.30436909e-01 7.14076281e-01 -3.28599900e-01 1.30914738e-02 -6.74996257e-01 -2.47801114e-02 4.15530622e-01 -6.73381150e-01 -2.10946739e-01 4.71927188e-02 -1.76946831e+00 8.64961267e-01 2.35047325e-01 5.80281019e-01 -3.77762556e-01 -9.59805623e-02 5.34730196e-01 7.49556184e-01 7.32722759e-01 3.10662717e-01 -5.60339689e-01 -3.03868651e-01 -8.95429432e-01 -2.77571887e-01 6.52978957e-01 9.72506464e-01 8.24831903e-01 3.52775842e-01 -1.28755704e-01 1.29475665e+00 2.46424273e-01 9.78684723e-01 5.99906802e-01 -4.42514181e-01 1.03203452e+00 -1.97201613e-02 -2.31910601e-01 -6.19134307e-01 -3.83822173e-01 -3.20697725e-01 -9.10822630e-01 -1.20465115e-01 2.45755360e-01 -4.31028157e-01 -8.55626285e-01 1.71575069e+00 -1.25696048e-01 -3.08205873e-01 3.85923654e-01 1.01580787e+00 8.05882156e-01 1.03897417e+00 -6.77773878e-02 -2.59914607e-01 1.06252217e+00 -1.29118741e+00 -7.64784276e-01 -1.68326288e-01 7.10688353e-01 -1.48842466e+00 1.15144253e+00 2.11291820e-01 -1.16033959e+00 -6.23207152e-01 -7.89108992e-01 -1.95816040e-01 -4.28072900e-01 7.53069520e-01 3.52893472e-01 4.80279595e-01 -1.30828178e+00 1.32723436e-01 -8.16007972e-01 -8.81525755e-01 -1.37189209e-01 2.52705127e-01 -5.74135244e-01 -2.80245423e-01 -1.41976368e+00 1.52982914e+00 6.10363632e-02 4.69042033e-01 -1.02018774e+00 -5.34095615e-02 -8.75104308e-01 -4.45101917e-01 1.23078693e-02 -7.54133701e-01 8.77687275e-01 -1.24769533e+00 -1.61271060e+00 1.10938311e+00 -2.56249696e-01 -1.73943952e-01 5.98980904e-01 -1.18279792e-01 -4.58592772e-01 -6.75641373e-02 1.68271348e-01 1.03549600e+00 6.70214295e-01 -1.23432446e+00 -2.68834621e-01 -4.44196492e-01 -4.52624321e-01 5.23218393e-01 -5.46780527e-02 5.19575477e-01 -6.14147723e-01 -7.96066105e-01 -3.03760976e-01 -1.21282768e+00 -3.77402343e-02 -4.86791015e-01 -4.78669167e-01 6.52004853e-02 4.34951395e-01 -1.43868053e+00 6.64098442e-01 -1.84719706e+00 8.79822552e-01 -2.00965881e-01 -4.32175785e-01 1.67139843e-01 -7.11035490e-01 7.72250056e-01 1.12928107e-01 1.44339859e-01 -3.45819235e-01 -7.53080845e-01 1.37607262e-01 1.48477212e-01 -2.16455311e-01 4.85298991e-01 3.61264974e-01 1.30212510e+00 -4.83492494e-01 -4.73956525e-01 -4.18894663e-02 7.13558853e-01 7.33105615e-02 4.60674129e-02 -1.12267405e-01 7.82180488e-01 8.74551982e-02 9.27720428e-01 5.31163633e-01 2.75241077e-01 4.01656866e-01 -3.56939703e-01 -4.72101837e-01 1.45669788e-01 -3.70518446e-01 2.14476252e+00 -6.50697887e-01 1.01482069e+00 3.80249113e-01 -6.90867901e-01 9.73008871e-01 5.13962865e-01 8.97045732e-02 -9.34035063e-01 3.19882296e-02 5.99019647e-01 4.47611846e-02 -3.37321818e-01 7.26025939e-01 -2.68479943e-01 -1.75580934e-01 6.39081657e-01 4.24843490e-01 8.15434661e-03 3.38888109e-01 1.46424159e-01 4.28243428e-01 7.54229605e-01 -1.27946168e-01 -2.81061679e-01 3.00133914e-01 3.11381698e-01 2.73988277e-01 4.92637843e-01 -2.43706957e-01 8.53740752e-01 2.80585289e-01 -3.61236453e-01 -1.18783867e+00 -9.74350810e-01 -2.18806509e-02 1.28854728e+00 -2.71129876e-01 -1.62436306e-01 -7.42023945e-01 -6.44206285e-01 -5.24104834e-01 5.26080072e-01 -4.39739287e-01 1.66364517e-02 -8.37651968e-01 -1.08101523e+00 1.02683830e+00 2.56544381e-01 3.54676455e-01 -1.13815653e+00 1.86753005e-01 2.54670411e-01 -1.06823516e+00 -1.27452409e+00 -7.53260672e-01 3.06776911e-01 -6.68251395e-01 -5.58486879e-01 -1.34489059e+00 -1.11103547e+00 2.98957735e-01 -1.66983262e-01 1.10709918e+00 -5.07576823e-01 -1.04110956e-01 3.11618656e-01 -3.57890517e-01 -1.51489988e-01 -7.27729380e-01 4.55839574e-01 2.21518919e-01 -3.20232995e-02 1.49518788e-01 -6.78481087e-02 1.99093260e-02 3.93669039e-01 -6.80832446e-01 2.98492670e-01 1.02373874e+00 9.26456332e-01 6.57340109e-01 -8.60684574e-01 6.34249628e-01 -4.83542770e-01 8.35683346e-01 -2.57465959e-01 -5.12920380e-01 7.80923545e-01 -3.55213702e-01 -1.70840457e-01 4.44485605e-01 -5.56774557e-01 -1.15651619e+00 -1.54923070e-02 -1.63107678e-01 -2.23595217e-01 8.12130496e-02 8.88719380e-01 -2.08730340e-01 -6.37653917e-02 6.29595935e-01 5.52157760e-01 -7.04560056e-02 -4.85610485e-01 5.86036026e-01 8.57485235e-01 4.48907793e-01 -7.36674786e-01 5.48949957e-01 -9.41213369e-02 -3.00367981e-01 -7.90124834e-01 -2.57218778e-02 -4.18580472e-02 -1.02938950e+00 -1.92895651e-01 1.16987073e+00 -1.15910149e+00 -1.23347394e-01 5.97801864e-01 -1.56577885e+00 -5.70254803e-01 5.23028791e-01 7.12316096e-01 -5.44116557e-01 3.47681016e-01 -1.01835299e+00 -4.09178346e-01 -4.59519476e-01 -1.57718384e+00 1.21958947e+00 -3.15869033e-01 -6.51333109e-02 -1.15957952e+00 6.04939103e-01 6.80292189e-01 6.18849397e-01 -1.19342715e-01 1.02674115e+00 -4.34021473e-01 -3.39196414e-01 2.13778079e-01 -3.78087044e-01 3.33339781e-01 -1.11748263e-01 8.58564451e-02 -7.45244443e-01 -4.40015197e-01 -3.81240517e-01 -6.66157782e-01 9.22944844e-01 1.70032084e-01 2.24203374e-02 2.12839562e-02 5.92692755e-02 6.38020277e-01 1.08515990e+00 1.04662627e-01 6.41364932e-01 3.91419947e-01 8.79509866e-01 6.01647556e-01 5.04411995e-01 -2.02561364e-01 6.49489522e-01 1.03831017e+00 7.41685331e-02 -5.10265768e-01 -2.66619295e-01 -1.78120863e-02 1.02541816e+00 1.49238193e+00 -4.17433649e-01 -3.10167164e-01 -1.05054879e+00 3.95901799e-01 -1.99600887e+00 -6.24524891e-01 -3.81001472e-01 1.98443139e+00 8.62411737e-01 -5.95133841e-01 9.08848271e-02 -7.66746163e-01 8.08909416e-01 -1.47758558e-01 -2.86935493e-02 -9.26927447e-01 -8.07610691e-01 1.20093077e-02 2.33094156e-01 6.75374806e-01 -9.83011961e-01 1.43195617e+00 5.76668787e+00 9.93341923e-01 -1.32174087e+00 5.15435815e-01 5.81188977e-01 1.44962534e-01 -2.75627941e-01 -1.02354012e-01 -5.76108813e-01 2.61619419e-01 1.17858732e+00 2.66961187e-01 8.26672435e-01 1.55598298e-01 8.76890589e-03 1.57433122e-01 -7.98724830e-01 8.17992866e-01 4.84127700e-01 -1.19574761e+00 1.83921516e-01 9.25552920e-02 9.61461127e-01 7.09352612e-01 9.89156067e-02 5.59517860e-01 1.53032914e-01 -1.17544186e+00 6.98385060e-01 4.45034593e-01 1.13398159e+00 -7.57260501e-01 8.30425739e-01 2.01821938e-01 -9.67779279e-01 5.60280561e-01 -4.07493055e-01 5.11148334e-01 4.21859741e-01 1.98559925e-01 -7.13377714e-01 1.07258844e+00 4.31553364e-01 5.78896284e-01 -5.32992542e-01 4.50963974e-01 -2.44193450e-01 6.39728904e-01 -5.93683869e-02 1.14828587e-01 4.92488950e-01 -6.11712158e-01 5.80245018e-01 1.61438835e+00 5.16655266e-01 -6.84200346e-01 1.66802704e-01 6.37577713e-01 -2.49507174e-01 4.82918680e-01 -5.87701976e-01 -4.22181576e-01 -1.81835741e-01 1.23744059e+00 -5.78110933e-01 -2.93269545e-01 -5.83884060e-01 1.52370358e+00 4.17254627e-01 8.97184551e-01 -7.90040314e-01 -1.60757437e-01 2.94110119e-01 -5.28125942e-01 1.13452844e-01 -4.26633716e-01 -2.28193060e-01 -1.49139166e+00 -1.29564017e-01 -1.46834528e+00 2.63505541e-02 -1.04350746e+00 -1.14056587e+00 1.26262558e+00 -4.37507540e-01 -1.18419123e+00 -3.95560235e-01 -7.79081821e-01 -2.53349692e-01 1.40049517e+00 -1.23397994e+00 -1.95885193e+00 3.45454067e-01 6.54870987e-01 5.54009557e-01 -6.00475907e-01 1.08708978e+00 5.85514784e-01 -6.13550484e-01 6.34350598e-01 2.79522747e-01 6.68520704e-02 1.37534916e+00 -9.84789073e-01 4.06637818e-01 8.09255064e-01 4.72375214e-01 5.74258685e-01 3.94698083e-01 -7.42087007e-01 -1.72610867e+00 -9.98365998e-01 1.38759208e+00 -4.91265446e-01 6.74969018e-01 -4.87182468e-01 -6.36156619e-01 9.63045299e-01 1.08177054e+00 -6.59025013e-01 6.12336636e-01 1.33618146e-01 -2.44273379e-01 1.91099465e-01 -7.79454172e-01 7.32210934e-01 5.55243254e-01 -8.05711150e-01 -2.51117796e-01 5.81346929e-01 5.39147377e-01 -2.95883298e-01 -7.12659836e-01 3.43417048e-01 6.43865645e-01 -5.30608654e-01 8.12956274e-01 -6.86499596e-01 6.44389510e-01 -1.69577792e-01 -4.22921449e-01 -1.76474142e+00 -1.60262035e-03 -7.31926382e-01 4.65626955e-01 1.04814434e+00 1.06325579e+00 -5.51713645e-01 -3.43514793e-02 -1.85399622e-01 -3.28713238e-01 -4.38946068e-01 -1.10076308e+00 -5.78374565e-01 4.43294466e-01 -3.14023018e-01 6.23912141e-02 1.26209486e+00 -1.02735892e-01 7.85092294e-01 -8.60134184e-01 -6.13308251e-02 2.90890783e-01 1.04283914e-01 7.19197035e-01 -5.58997750e-01 -4.43508595e-01 -6.30968273e-01 1.21326081e-01 -7.41624236e-01 3.58429581e-01 -1.32399058e+00 1.15232229e-01 -1.69282019e+00 4.10164237e-01 1.14089631e-01 -1.42427534e-01 5.92059851e-01 -9.71757919e-02 6.78546906e-01 3.82196099e-01 3.81207734e-01 -5.19450247e-01 6.38569593e-01 1.35140610e+00 -3.34931165e-01 4.19753633e-04 -3.39110941e-01 -3.80673140e-01 3.68440971e-02 6.45754755e-01 -2.75464386e-01 1.80338919e-01 -9.93353248e-01 2.66993463e-01 2.92566478e-01 9.55816433e-02 -2.92924196e-01 7.93397352e-02 1.32952249e-02 3.71528149e-01 -4.34804350e-01 3.55941236e-01 -4.29592669e-01 2.09897608e-01 4.01683778e-01 -1.98368818e-01 6.43463671e-01 4.82997060e-01 -7.53748640e-02 -5.22645235e-01 -1.82250142e-02 4.60599422e-01 -1.33009762e-01 -4.33058113e-01 -8.21103007e-02 -7.28571236e-01 -3.16257417e-01 5.73738456e-01 2.16784790e-01 -7.50120521e-01 -3.89634907e-01 -8.63626122e-01 1.58970445e-01 4.33771938e-01 5.87304235e-01 3.54219735e-01 -1.54446793e+00 -1.32777441e+00 9.16709080e-02 8.32523704e-02 -9.28689599e-01 4.85079996e-02 1.33941293e+00 -6.40794158e-01 6.71474457e-01 -4.43972647e-01 -6.64514184e-01 -1.30808532e+00 1.69805855e-01 2.19363555e-01 -4.25432324e-01 -7.37536773e-02 5.99166453e-01 6.30791858e-02 -1.26280797e+00 -1.20361932e-01 3.41205299e-01 1.74436420e-01 1.88660741e-01 1.90065429e-01 3.20799112e-01 3.41686398e-01 -1.08118856e+00 -4.15571660e-01 6.95115805e-01 8.40792954e-02 -8.64855349e-01 1.08219635e+00 -2.74530172e-01 -4.34109449e-01 5.64236820e-01 1.22701311e+00 8.69915709e-02 -5.08382618e-01 -1.11592904e-01 -1.93042070e-01 1.03598371e-01 -9.08673331e-02 -1.45170271e+00 -7.57308722e-01 1.29352713e+00 6.04469836e-01 -4.18086708e-01 1.02915728e+00 -1.93299294e-01 6.79279327e-01 5.50180614e-01 4.38748986e-01 -1.09374535e+00 -2.84339547e-01 9.54343021e-01 1.24667847e+00 -1.52356768e+00 -4.25945640e-01 1.49165988e-01 -1.07747471e+00 1.26317108e+00 3.59669715e-01 6.15485251e-01 -1.89696014e-01 2.16138065e-01 6.82077408e-01 2.71394819e-01 -8.81242573e-01 -3.84210259e-01 6.90762937e-01 6.38091981e-01 7.25156248e-01 2.69588828e-01 -4.08123463e-01 3.04174662e-01 1.43576756e-01 -3.29315275e-01 1.31902158e-01 6.68043375e-01 -7.31487898e-03 -1.44795263e+00 -5.33651710e-01 -2.36973852e-01 -3.15518141e-01 -6.68153107e-01 -8.11868966e-01 7.52771318e-01 -1.05169795e-01 9.90956604e-01 -2.21335739e-01 -3.99018914e-01 1.09370247e-01 5.27418375e-01 6.93800688e-01 -3.56879711e-01 -8.85821402e-01 6.14104927e-01 4.08900172e-01 -3.30335319e-01 -2.37738371e-01 -7.10181475e-01 -7.55553544e-01 -2.00050965e-01 -3.05084866e-02 3.35413218e-02 1.15053666e+00 8.99447799e-01 3.58630031e-01 3.70081812e-01 2.89550543e-01 -1.03471363e+00 -2.44527221e-01 -1.40650153e+00 9.90244672e-02 9.55119431e-02 4.59106341e-02 -8.05858597e-02 -1.44580051e-01 2.40271077e-01]
[11.510578155517578, 1.569311499595642]
5d874651-e8c6-4063-9ccf-a4a77d585000
exposure-correction-model-to-enhance-image
2204.10648
null
https://arxiv.org/abs/2204.10648v1
https://arxiv.org/pdf/2204.10648v1.pdf
Exposure Correction Model to Enhance Image Quality
Exposure errors in an image cause a degradation in the contrast and low visibility in the content. In this paper, we address this problem and propose an end-to-end exposure correction model in order to handle both under- and overexposure errors with a single model. Our model contains an image encoder, consecutive residual blocks, and image decoder to synthesize the corrected image. We utilize perceptual loss, feature matching loss, and multi-scale discriminator to increase the quality of the generated image as well as to make the training more stable. The experimental results indicate the effectiveness of proposed model. We achieve the state-of-the-art result on a large-scale exposure dataset. Besides, we investigate the effect of exposure setting of the image on the portrait matting task. We find that under- and overexposed images cause severe degradation in the performance of the portrait matting models. We show that after applying exposure correction with the proposed model, the portrait matting quality increases significantly. https://github.com/yamand16/ExposureCorrection
['Alexander Waibel', 'Hazim Kemal Ekenel', 'Dogucan Yaman', 'Fevziye Irem Eyiokur']
2022-04-22
null
null
null
null
['image-matting']
['computer-vision']
[ 5.67471862e-01 -3.58516991e-01 1.29054412e-01 -3.49650860e-01 -6.29169106e-01 -3.22540969e-01 2.85087079e-01 -3.18321317e-01 -8.92709941e-02 3.90033215e-01 2.02357277e-01 1.43480271e-01 2.60122389e-01 -8.10442865e-01 -9.26019311e-01 -6.86232507e-01 5.58359802e-01 -5.96194446e-01 2.68524826e-01 -1.04654901e-01 2.08975911e-01 1.22980170e-01 -1.48121357e+00 3.03082854e-01 1.37237608e+00 9.66173708e-01 5.05609512e-01 6.64161444e-01 3.06619823e-01 4.06857044e-01 -6.00606382e-01 -6.83481336e-01 5.98474443e-01 -5.57747543e-01 -1.90141737e-01 7.31516004e-01 1.06623435e+00 -7.02751637e-01 -6.88897610e-01 1.24862528e+00 5.64023376e-01 8.26030876e-03 5.07782698e-01 -1.19367242e+00 -1.00583267e+00 2.01648489e-01 -1.04892242e+00 3.98622267e-02 1.79133832e-01 4.95851636e-01 3.09429675e-01 -8.62907350e-01 5.13403654e-01 1.18859756e+00 7.17267871e-01 3.11990589e-01 -1.09718704e+00 -8.18175733e-01 -3.10869124e-02 2.34161958e-01 -1.24381554e+00 -6.47748947e-01 8.23791623e-01 -9.87392515e-02 1.41165495e-01 3.47689301e-01 4.92264688e-01 7.70965040e-01 5.00472426e-01 5.26069462e-01 1.40508676e+00 -5.89759052e-01 -1.91685632e-01 1.95065483e-01 -1.19665854e-01 6.25991464e-01 1.64586633e-01 1.51844725e-01 -5.18672049e-01 3.22441280e-01 5.28497279e-01 1.47567526e-01 -5.00452518e-01 2.67419636e-01 -8.47108722e-01 2.37407565e-01 6.90005064e-01 -1.24340780e-01 -2.55778670e-01 2.36699969e-01 9.46451426e-02 2.64356732e-01 8.53341579e-01 -4.49856967e-02 6.17836341e-02 3.81861418e-01 -1.25501871e+00 1.04044177e-01 7.73861408e-02 9.95126724e-01 5.70632577e-01 2.13011906e-01 -4.32527244e-01 9.68383253e-01 2.45006099e-01 9.83711541e-01 2.20006332e-01 -8.61616015e-01 6.74328029e-01 2.33000979e-01 1.49124889e-02 -1.09726858e+00 1.08913153e-01 -4.86505926e-01 -9.30850267e-01 4.09773618e-01 2.04274639e-01 -6.60039410e-02 -1.03775084e+00 1.73509085e+00 4.36889648e-01 2.60256767e-01 -5.04509322e-02 1.03961754e+00 6.82491899e-01 9.37404871e-01 -4.42324840e-02 -3.86291891e-01 1.27516770e+00 -1.27658284e+00 -1.19407368e+00 -3.11785460e-01 -1.36037478e-02 -1.27402627e+00 1.27711451e+00 3.97439808e-01 -1.46102870e+00 -9.02854800e-01 -1.30785382e+00 -1.75087839e-01 3.25988755e-02 3.37173879e-01 4.16277274e-02 7.59478033e-01 -8.14089835e-01 6.60001755e-01 -2.10960731e-01 -2.74676651e-01 4.43959445e-01 -1.00255042e-01 -1.02977894e-01 -5.38133383e-01 -1.07291496e+00 8.61969113e-01 1.02820024e-01 2.00897366e-01 -9.92711425e-01 -9.55918491e-01 -7.21215725e-01 -2.27392048e-01 2.57584840e-01 -5.31202376e-01 9.65902865e-01 -1.31111360e+00 -1.20210552e+00 7.58248210e-01 -1.13716066e-01 -1.25082552e-01 7.39793777e-01 -2.61593729e-01 -6.21331811e-01 1.52167588e-01 1.07232025e-02 6.33385718e-01 1.15520680e+00 -1.45962977e+00 -4.01773036e-01 -2.30812728e-01 -3.80996048e-01 4.44255680e-01 -3.76504600e-01 -1.36635140e-01 -7.04350173e-01 -9.40712214e-01 -1.36198159e-02 -9.04501259e-01 2.14529425e-01 5.05073726e-01 -5.35692573e-01 6.32874548e-01 9.34602499e-01 -1.24702251e+00 1.30402112e+00 -2.58209205e+00 -2.23774418e-01 -1.45338789e-01 8.37713107e-02 1.00323498e-01 -3.73243153e-01 3.11363190e-01 -2.71260575e-03 1.37241438e-01 -3.63020241e-01 -5.53150296e-01 -2.20004812e-01 -4.42067832e-02 -5.02507448e-01 7.45278537e-01 -8.29776190e-03 6.86128795e-01 -4.67498213e-01 -5.35028934e-01 2.99510151e-01 5.31617701e-01 -2.20862642e-01 4.41398829e-01 -1.10379368e-01 1.94473743e-01 1.06313236e-01 6.43132389e-01 1.37941277e+00 6.59908503e-02 -3.39520752e-01 -5.52340269e-01 -6.25671521e-02 -3.02518606e-01 -9.83907700e-01 1.54302597e+00 -4.43343222e-01 8.77443492e-01 1.67234347e-03 1.04008049e-01 8.19175482e-01 -3.89840379e-02 -4.63251607e-04 -1.00111938e+00 6.48960173e-02 1.88149557e-01 -2.00384945e-01 -6.71438277e-01 6.39841914e-01 -7.29921386e-02 1.90656021e-01 2.26575896e-01 -3.06424856e-01 -2.21392781e-01 3.33010852e-02 1.83437109e-01 4.61233288e-01 2.41026077e-02 -7.32413307e-02 -4.10074331e-02 3.49026620e-01 -3.99004072e-01 3.33303005e-01 3.48397374e-01 -1.52180851e-01 1.03816116e+00 9.96424183e-02 -8.74105915e-02 -1.54809940e+00 -1.15120935e+00 -2.63677575e-02 8.03997636e-01 6.08115613e-01 -1.10792920e-01 -1.04928231e+00 -1.25113338e-01 -2.93951184e-01 7.96281338e-01 -5.63086033e-01 -3.15563887e-01 -2.62188464e-01 -7.30700135e-01 4.96255368e-01 2.18059897e-01 1.39229381e+00 -7.19524443e-01 -2.01282516e-01 -1.73723668e-01 -4.12495136e-01 -1.12201560e+00 -1.11110914e+00 -5.29137254e-01 -6.65753484e-01 -8.78929853e-01 -8.24586451e-01 -6.53297246e-01 8.65588248e-01 6.74978554e-01 6.79667056e-01 1.95114717e-01 -2.79946625e-01 -1.57172848e-02 -2.76569873e-01 -2.33036071e-01 -5.24539292e-01 -2.89383501e-01 -1.63125962e-01 2.44159654e-01 -5.11162996e-01 -5.37254214e-01 -1.05309284e+00 5.53650498e-01 -1.31011891e+00 3.29264700e-01 6.78079188e-01 5.58404148e-01 4.53861415e-01 3.63671958e-01 3.33024591e-01 -5.33874810e-01 5.52636445e-01 -2.05658928e-01 -4.46107715e-01 2.82022178e-01 -8.83264601e-01 -3.27619672e-01 4.79045779e-01 -7.17850029e-01 -1.36905801e+00 -1.69992685e-01 -2.37175990e-02 -4.35968429e-01 1.56434998e-01 2.96807550e-02 -5.13734996e-01 -3.29339087e-01 5.27207375e-01 5.31045258e-01 1.21710226e-01 -3.75792325e-01 3.72804254e-01 6.79721892e-01 7.02116489e-01 -9.98344868e-02 1.12056375e+00 4.41095650e-01 3.33296619e-02 -7.02892482e-01 -8.01029682e-01 1.47325024e-01 -2.07506955e-01 -6.28478885e-01 7.11261034e-01 -1.10237610e+00 -3.17235678e-01 8.73639941e-01 -9.85804617e-01 -5.38492739e-01 -5.69357648e-02 4.06330936e-02 -1.88447252e-01 5.56528091e-01 -5.81742942e-01 -5.72935522e-01 -5.62436938e-01 -1.07055020e+00 9.80541945e-01 5.91360509e-01 3.15691769e-01 -5.13821483e-01 -8.27197284e-02 5.27758181e-01 6.25790298e-01 3.30120772e-01 5.70117176e-01 2.97336489e-01 -6.89781427e-01 -2.06756089e-02 -4.50333893e-01 5.73257744e-01 1.54355213e-01 8.25886205e-02 -1.12570310e+00 -3.44225794e-01 6.74539432e-02 -1.43704824e-02 9.39244151e-01 3.35937142e-01 1.15501761e+00 -4.70961958e-01 -2.60939002e-02 9.00680184e-01 1.58924043e+00 -1.47479782e-02 1.31853139e+00 2.62235552e-01 6.18853688e-01 5.08492887e-01 9.61307704e-01 2.64991015e-01 4.88048233e-02 7.48785913e-01 3.95555794e-01 -4.07594323e-01 -8.62451673e-01 -5.22588134e-01 5.47469199e-01 4.40662801e-01 2.92294532e-01 -7.39630401e-01 -3.57990742e-01 3.90827656e-01 -1.44213772e+00 -1.00070775e+00 -5.17196834e-01 2.27020311e+00 1.01816869e+00 -9.93157551e-02 -2.10689574e-01 1.21059800e-02 9.38739777e-01 3.78839910e-01 -5.46440482e-01 -2.46438786e-01 -5.01668394e-01 -3.82384539e-01 7.04101861e-01 6.95287228e-01 -8.18735003e-01 7.97038972e-01 6.16862106e+00 1.12504590e+00 -1.23616540e+00 1.85442358e-01 1.11584592e+00 -1.30064785e-01 -3.66207421e-01 -9.57328975e-02 -5.87412357e-01 9.77902949e-01 6.26526833e-01 -1.88089609e-01 4.97414798e-01 4.52409327e-01 5.22552371e-01 -4.30916488e-01 -5.78134477e-01 1.19088292e+00 5.94693244e-01 -9.79808271e-01 -1.64669544e-01 2.60144137e-02 9.80707347e-01 -5.55383921e-01 6.55786991e-01 -1.83439299e-01 -4.10397649e-01 -8.44460905e-01 9.49490249e-01 6.42575920e-01 1.21180105e+00 -6.69318020e-01 3.88434321e-01 -5.52823395e-02 -1.00863719e+00 1.02826081e-01 -2.72722870e-01 2.20398590e-01 2.72368371e-01 7.16035068e-01 -4.29060966e-01 3.98324192e-01 4.06258523e-01 4.58800435e-01 -1.09203827e+00 1.27058208e+00 -4.62717265e-01 5.24639308e-01 -8.11721757e-02 4.94749933e-01 -1.86696425e-01 -1.97375178e-01 5.04565477e-01 1.00900733e+00 5.04544616e-01 8.38049799e-02 -1.18949905e-01 7.99989283e-01 -3.58463764e-01 -9.25261751e-02 -3.61159414e-01 3.94379973e-01 4.57009703e-01 1.13390815e+00 -3.95366251e-01 -2.86486179e-01 -1.90250762e-02 1.50764704e+00 -2.90789783e-01 4.73230690e-01 -1.29196370e+00 -3.42365175e-01 3.89881670e-01 5.00506580e-01 1.82709843e-01 -3.50191928e-02 -4.63050604e-01 -9.39834416e-01 4.03947264e-01 -9.41022336e-01 -3.61551307e-02 -1.35657704e+00 -1.11472535e+00 4.79772508e-01 -5.21942861e-02 -1.27205575e+00 3.86857122e-01 -4.66002375e-02 -8.89562845e-01 8.98975611e-01 -1.50084436e+00 -1.37777531e+00 -9.67189848e-01 4.41621095e-01 8.48125041e-01 1.85903355e-01 1.51563790e-02 5.61208487e-01 -7.24808276e-01 7.67648220e-01 2.83304244e-01 -3.42997491e-01 1.25464618e+00 -6.67321026e-01 2.53119767e-01 1.39556563e+00 -3.56202543e-01 1.06470771e-01 1.00322139e+00 -9.89752054e-01 -1.30862427e+00 -1.49615622e+00 4.47567284e-01 -5.28344186e-03 7.69112334e-02 -3.26012611e-01 -9.03895617e-01 3.91993850e-01 7.44543493e-01 -2.16960996e-01 4.63048726e-01 -6.34562910e-01 -4.07300174e-01 -6.09556973e-01 -1.45304596e+00 7.77936280e-01 7.56785929e-01 -2.58731246e-01 -2.55940817e-02 1.84186965e-01 9.64875519e-01 -5.89944422e-01 -6.45854771e-01 1.48294687e-01 6.50349975e-01 -9.01421070e-01 9.44573700e-01 3.78760427e-01 9.02297556e-01 -5.27382791e-01 -1.90643027e-01 -1.34671831e+00 -2.88412303e-01 -7.00528264e-01 1.16658121e-01 1.61427248e+00 3.29435050e-01 -4.07747239e-01 3.00598770e-01 6.50501490e-01 -1.33641064e-01 -4.35013920e-01 -4.88733977e-01 -6.85811222e-01 -2.42719799e-01 -8.82724393e-03 6.13385320e-01 6.15291417e-01 -5.49694121e-01 1.47390947e-01 -1.02841461e+00 4.40469563e-01 1.09930456e+00 5.71814738e-02 8.08172703e-01 -2.97216713e-01 -1.93411365e-01 6.52988032e-02 -6.24374114e-02 -9.37385380e-01 -2.45445818e-01 -2.88587004e-01 1.84335396e-01 -1.29853451e+00 4.54663664e-01 -1.06981151e-01 1.26731068e-01 1.63883775e-01 -6.39571369e-01 6.93670392e-01 4.35382277e-01 3.10790062e-01 -3.91371608e-01 6.49910927e-01 1.55477452e+00 -2.19780385e-01 -4.14362475e-02 -3.35531652e-01 -7.45746732e-01 3.24892282e-01 9.84360158e-01 -4.89481032e-01 -4.18010443e-01 -7.74893761e-01 3.90320048e-02 -1.06987633e-01 5.95855415e-01 -1.09229505e+00 1.96362153e-01 -2.05957398e-01 9.23938036e-01 -5.38282216e-01 4.43901360e-01 -7.46516824e-01 4.15172249e-01 5.58435977e-01 -3.78162473e-01 -7.57446559e-03 3.98193091e-01 5.82719624e-01 1.61421038e-02 -1.79364383e-01 1.25933397e+00 3.13375831e-01 -5.39352238e-01 2.99659073e-01 3.48763727e-02 -2.72874951e-01 1.17415655e+00 -2.90774047e-01 -7.05256581e-01 -5.19400358e-01 -2.43649170e-01 1.31674066e-01 9.71050918e-01 3.02324951e-01 8.81569028e-01 -1.36387134e+00 -9.14982200e-01 1.43352151e-01 -5.75228743e-02 -3.11585844e-01 9.23377693e-01 8.50139380e-01 -6.36790752e-01 -4.12278086e-01 -3.68052691e-01 -2.69661516e-01 -1.67965436e+00 6.60650313e-01 2.05633521e-01 1.37990609e-01 -5.31493187e-01 5.57817578e-01 3.26631725e-01 1.92324907e-01 1.79081663e-01 -4.72706445e-02 3.61155480e-01 -3.58928323e-01 8.52415860e-01 3.65491301e-01 -2.26677641e-01 -4.53959078e-01 3.04572918e-02 6.40341341e-01 -2.07223058e-01 -1.70809105e-01 1.07449055e+00 -8.26559305e-01 8.14934298e-02 1.70098349e-01 1.03729475e+00 9.57828164e-02 -1.52445328e+00 -6.61896765e-02 -9.43268955e-01 -1.19348764e+00 2.30697393e-01 -9.90256488e-01 -1.12135780e+00 7.34984040e-01 1.08928573e+00 -1.15804926e-01 1.74897528e+00 -4.03052002e-01 1.06093144e+00 -3.55363309e-01 -1.88465342e-01 -9.84305143e-01 4.31755573e-01 -1.45865738e-01 1.19953334e+00 -1.24978888e+00 3.84552211e-01 -7.24587321e-01 -8.44004929e-01 8.52694750e-01 7.25517154e-01 -1.64758757e-01 1.83886945e-01 1.49225861e-01 -1.87479574e-02 1.22626208e-01 -5.69263697e-01 3.33122909e-02 2.70832002e-01 5.70231140e-01 8.03364366e-02 -2.15505108e-01 -2.30055004e-01 1.01122983e-01 7.34194694e-03 -6.40254766e-02 7.45939493e-01 5.45033872e-01 -5.29743016e-01 -7.92739272e-01 -6.75282001e-01 1.70188174e-01 -3.28080952e-01 -2.89420396e-01 -2.19251662e-01 4.93970037e-01 3.63738567e-01 1.13190353e+00 -4.80947308e-02 -4.48902547e-01 3.47847849e-01 -4.13681865e-01 6.47165716e-01 -8.38340893e-02 -4.53974664e-01 2.53070384e-01 -4.53045331e-02 -4.43943352e-01 -1.84068844e-01 -1.78532258e-01 -5.63749135e-01 -7.49514759e-01 -4.07640308e-01 -3.76668572e-01 6.82328761e-01 4.76231873e-01 4.02322680e-01 6.23658597e-01 1.14928639e+00 -5.50314665e-01 -3.45208377e-01 -1.10269392e+00 -7.00819552e-01 5.46249866e-01 4.02165949e-01 -2.35806391e-01 -4.84473526e-01 4.89003479e-01]
[10.899675369262695, -2.1522066593170166]
192732ec-4f2d-43a1-8617-1641a0074e2b
a-multi-modal-neural-geometric-solver-with
2302.11097
null
https://arxiv.org/abs/2302.11097v2
https://arxiv.org/pdf/2302.11097v2.pdf
A Multi-Modal Neural Geometric Solver with Textual Clauses Parsed from Diagram
Geometry problem solving (GPS) is a high-level mathematical reasoning requiring the capacities of multi-modal fusion and geometric knowledge application. Recently, neural solvers have shown great potential in GPS but still be short in diagram presentation and modal fusion. In this work, we convert diagrams into basic textual clauses to describe diagram features effectively, and propose a new neural solver called PGPSNet to fuse multi-modal information efficiently. Combining structural and semantic pre-training, data augmentation and self-limited decoding, PGPSNet is endowed with rich knowledge of geometry theorems and geometric representation, and therefore promotes geometric understanding and reasoning. In addition, to facilitate the research of GPS, we build a new large-scale and fine-annotated GPS dataset named PGPS9K, labeled with both fine-grained diagram annotation and interpretable solution program. Experiments on PGPS9K and an existing dataset Geometry3K validate the superiority of our method over the state-of-the-art neural solvers. Our code, dataset and appendix material are available at \url{https://github.com/mingliangzhang2018/PGPS}.
['Cheng-Lin Liu', 'Fei Yin', 'Ming-Liang Zhang']
2023-02-22
null
null
null
null
['mathematical-reasoning']
['natural-language-processing']
[-2.72210032e-01 3.22263598e-01 -9.17199254e-02 -4.13338482e-01 -7.42189765e-01 -4.96786565e-01 1.77769676e-01 -1.76306237e-02 4.20618832e-01 6.31892622e-01 2.39618674e-01 -4.81640100e-01 -3.77878696e-01 -1.38371229e+00 -1.14322317e+00 -3.52103561e-01 -2.42253728e-02 5.95090806e-01 -8.98401663e-02 -3.70762944e-01 2.04283874e-02 3.91711742e-01 -1.37988389e+00 6.23262405e-01 1.44555640e+00 1.24881601e+00 8.86578709e-02 2.65233904e-01 -3.41480196e-01 1.05818403e+00 -3.08380246e-01 -5.11600435e-01 7.33208135e-02 1.17403045e-01 -1.03073299e+00 -2.31026947e-01 5.54205835e-01 -2.95674682e-01 -9.22006547e-01 1.00752795e+00 2.76613832e-01 1.49514545e-02 3.64175379e-01 -1.55560601e+00 -1.18445349e+00 1.06300473e+00 -4.45532382e-01 -1.86518282e-01 6.87572598e-01 1.06399775e-01 8.93186927e-01 -8.74555528e-01 6.37426376e-01 1.33001292e+00 7.63785839e-01 2.42730439e-01 -6.59346342e-01 -8.88990998e-01 3.31608027e-01 6.31082177e-01 -1.77097833e+00 -6.10427111e-02 1.09605157e+00 -2.59054065e-01 9.99457121e-01 3.55993479e-01 8.72128963e-01 7.46928632e-01 -2.57179767e-01 1.13082218e+00 5.21156907e-01 6.30400255e-02 -4.44571227e-02 -4.32959735e-01 -5.19740433e-02 1.11658978e+00 5.38330227e-02 -1.91643238e-01 -4.28836107e-01 2.71768898e-01 1.19844866e+00 -3.57399940e-01 -2.07197756e-01 -4.25689161e-01 -1.30975413e+00 5.98366618e-01 9.09435391e-01 3.05704754e-02 -5.15950881e-02 6.95406437e-01 4.02697325e-01 -3.61394174e-02 2.94862092e-01 4.56276894e-01 -4.30774242e-01 -1.42931163e-01 -4.94572610e-01 6.50752306e-01 7.07049310e-01 1.55705214e+00 6.85279667e-01 5.92665784e-02 7.41237178e-02 6.04629040e-01 1.81546524e-01 6.64466262e-01 -1.17729478e-01 -1.11040020e+00 1.12397420e+00 1.13965440e+00 -3.18027377e-01 -1.51991355e+00 -5.17180383e-01 -2.48197868e-01 -1.15720677e+00 -3.98604751e-01 2.73880631e-01 6.38742521e-02 -7.04070389e-01 1.54629588e+00 3.59347343e-01 1.56899810e-01 3.13993469e-02 9.32074308e-01 1.68565667e+00 9.97249603e-01 -1.16582245e-01 4.10558403e-01 1.17379296e+00 -9.99409497e-01 -4.48262036e-01 -5.20197265e-02 9.19631362e-01 -2.98830092e-01 1.08376515e+00 5.52319229e-01 -1.33257473e+00 -5.41730821e-01 -1.13576055e+00 -6.19019389e-01 -5.58605731e-01 3.86274248e-01 1.20834374e+00 5.20542711e-02 -8.73090148e-01 5.49053729e-01 -5.95488667e-01 5.86683080e-02 8.87573659e-01 3.46485555e-01 -4.13113087e-01 -3.16072226e-01 -1.46938074e+00 8.50999117e-01 8.16902041e-01 4.54822093e-01 -5.19163966e-01 -1.03868103e+00 -1.33126283e+00 8.44940320e-02 7.14712620e-01 -7.98626602e-01 1.18129444e+00 -1.76309258e-01 -1.21429360e+00 7.17085361e-01 1.88060224e-01 -1.44300625e-01 4.56226051e-01 -1.31705731e-01 -4.07528758e-01 1.67179376e-01 4.09426950e-02 8.57446551e-01 1.57802388e-01 -1.03977978e+00 -4.34466153e-01 -9.32698101e-02 6.51477039e-01 2.34169126e-01 1.11174867e-01 -2.61494517e-01 -7.44817793e-01 -6.72642648e-01 3.27581853e-01 -4.71115291e-01 -5.71624152e-02 1.44771850e-02 -6.17334962e-01 -5.87428987e-01 5.81051111e-01 -8.80372941e-01 1.24176395e+00 -1.93985450e+00 4.45498049e-01 3.50601226e-01 5.44831038e-01 6.29315227e-02 2.60107890e-02 5.05517840e-01 -2.05141664e-01 1.46258458e-01 -2.21003100e-01 4.10505868e-02 4.31873918e-01 2.55987227e-01 -1.54402703e-01 2.53584802e-01 2.29315266e-01 1.42554748e+00 -8.22693408e-01 -7.20926642e-01 5.59694886e-01 2.77371347e-01 -6.18325353e-01 -2.24895254e-01 -5.54942667e-01 3.36401314e-01 -7.50741899e-01 9.92347062e-01 9.87560689e-01 -6.27111018e-01 -4.82706865e-03 -5.80046415e-01 1.28156886e-01 1.79470792e-01 -1.39526594e+00 2.34751201e+00 -4.51327026e-01 4.37348008e-01 -1.27047211e-01 -1.26834500e+00 9.99397516e-01 -8.15851912e-02 3.15979987e-01 -9.92933691e-01 3.08436960e-01 1.63888440e-01 -2.70568430e-01 -5.62549114e-01 5.15698671e-01 5.50647557e-01 -4.29200768e-01 -6.50593564e-02 2.32883040e-02 -5.25759339e-01 5.82131863e-01 6.80021405e-01 9.25566792e-01 3.25886905e-01 4.39104177e-02 -2.56280690e-01 7.83712327e-01 1.77413836e-01 5.32865584e-01 3.40115666e-01 3.67650419e-01 4.54998940e-01 7.29345381e-01 -6.16392493e-01 -1.00277817e+00 -1.14689219e+00 1.21004623e-03 5.35394669e-01 6.16379678e-01 -9.32741642e-01 -7.30442584e-01 -3.74371707e-01 1.42309666e-01 5.19484222e-01 -5.34242511e-01 -1.89161658e-01 -9.79567170e-01 -3.35752934e-01 8.35460663e-01 1.15354526e+00 9.39371347e-01 -8.85868311e-01 8.54439959e-02 1.62381772e-02 -3.87926877e-01 -1.27596152e+00 6.49032518e-02 -2.55142301e-01 -5.50275683e-01 -1.31716764e+00 -4.41027611e-01 -1.04969990e+00 8.78249586e-01 -1.63177028e-01 1.38463247e+00 3.46800506e-01 -2.00977221e-01 4.68129441e-02 -2.54314393e-01 -1.44979209e-01 5.88134862e-03 1.64405718e-01 -3.51010114e-01 -7.52739966e-01 -4.89491746e-02 -7.17227817e-01 -2.70596653e-01 1.86264604e-01 -6.21460080e-01 9.91240799e-01 6.18909299e-01 4.94596779e-01 5.50843358e-01 2.67490178e-01 3.36417288e-01 -6.00385725e-01 3.66370201e-01 -2.63399184e-01 -7.67268479e-01 4.11341459e-01 -9.51152965e-02 -1.03743933e-01 7.97794938e-01 -2.97063552e-02 -1.12772882e+00 -3.80619913e-02 -1.74365535e-01 -3.93670589e-01 -1.52633324e-01 9.08985019e-01 -8.77956390e-01 -2.02324092e-01 3.37922394e-01 1.04811043e-01 -3.93616855e-01 -3.09790283e-01 7.93968558e-01 1.50563806e-01 8.96537781e-01 -1.24987078e+00 8.47366095e-01 1.70586944e-01 1.81605801e-01 -3.74980003e-01 -8.36160600e-01 1.53330579e-01 -5.86097658e-01 -2.11209565e-01 6.90279067e-01 -9.48549688e-01 -1.22082722e+00 6.66619897e-01 -1.30595481e+00 -4.47051883e-01 -9.16163847e-02 1.20507866e-01 -6.66191041e-01 2.98007518e-01 -8.12108338e-01 -2.97766685e-01 -1.88438743e-01 -1.14763892e+00 1.09783387e+00 3.21747094e-01 4.04544100e-02 -8.50307822e-01 -5.63650489e-01 6.32666528e-01 3.50903943e-02 5.58683813e-01 1.07423151e+00 -1.23091146e-01 -1.11804569e+00 -4.20844257e-02 -8.81407380e-01 3.37265171e-02 -2.06437275e-01 1.30766436e-01 -4.66249585e-01 2.42632404e-01 -5.40215373e-01 -4.71035272e-01 6.87334120e-01 3.32520247e-01 1.99764013e+00 -3.37579459e-01 -4.42343205e-01 1.02185965e+00 1.41121244e+00 1.27580658e-01 7.22608387e-01 2.61423558e-01 1.40327954e+00 3.77995789e-01 5.67538142e-01 2.85701066e-01 1.15206409e+00 3.63441080e-01 5.94428658e-01 -1.66163445e-02 -4.69906591e-02 -5.05821347e-01 -3.31445783e-01 1.03411460e+00 -1.56117767e-01 -3.20507139e-01 -1.35652900e+00 4.31964606e-01 -2.17355323e+00 -6.41824841e-01 -4.76447314e-01 1.19090879e+00 7.19949782e-01 5.30412607e-02 -1.54692367e-01 3.09749335e-01 5.42120039e-01 1.93501592e-01 -4.59230423e-01 -3.53855826e-03 -3.19571435e-01 1.15871005e-01 4.81478184e-01 2.73495138e-01 -1.22167814e+00 1.02893186e+00 4.99563932e+00 1.42078221e+00 -6.88697278e-01 -3.44595939e-01 5.34055471e-01 -6.10447861e-03 -4.90989715e-01 -1.05048127e-01 -4.23853606e-01 4.36393827e-01 3.16982895e-01 -3.61249149e-01 6.22791946e-01 8.42249990e-01 -2.67450541e-01 4.72861603e-02 -1.34221661e+00 1.44921541e+00 5.07184826e-02 -1.99263680e+00 7.06897303e-02 -2.88794190e-01 6.21149004e-01 -3.54857326e-01 -1.09212533e-01 4.50297117e-01 4.61906672e-01 -1.42232668e+00 8.35084260e-01 6.50828123e-01 8.60079408e-01 -8.85247707e-01 6.55773699e-01 1.70272738e-01 -1.81855321e+00 1.14338644e-01 -1.51447520e-01 -2.89994150e-01 3.09507310e-01 3.89047235e-01 -1.59861401e-01 1.34586573e+00 7.42713153e-01 1.22181022e+00 -6.22077525e-01 8.01479340e-01 -5.44398904e-01 7.44937360e-02 -4.21059966e-01 1.43054929e-02 3.79550636e-01 -1.13208793e-01 8.44596475e-02 7.94450998e-01 2.65828311e-01 5.36409020e-01 2.78348476e-01 1.30265152e+00 -2.42892459e-01 -1.28243655e-01 -3.72289449e-01 -2.11506829e-01 7.17463970e-01 1.13449812e+00 -7.35745311e-01 -5.36762238e-01 -3.04881722e-01 4.13074821e-01 4.46505100e-01 2.71326810e-01 -1.31399012e+00 -4.93401498e-01 3.19245368e-01 -1.85596570e-01 9.94606838e-02 -2.93939620e-01 -8.58148932e-01 -1.17492414e+00 4.34378296e-01 -7.68138409e-01 4.09087896e-01 -1.31455541e+00 -9.93151724e-01 2.08815679e-01 3.72348696e-01 -9.39991772e-01 2.49495804e-01 -8.60156894e-01 -4.36504483e-01 7.65685976e-01 -9.46752310e-01 -1.78086758e+00 -8.41112316e-01 4.66925919e-01 3.03947419e-01 1.36169009e-02 4.04216439e-01 7.08804429e-01 -6.08865499e-01 4.70257342e-01 -2.46759191e-01 5.17505169e-01 4.49426519e-03 -1.29573321e+00 7.51895130e-01 5.50566494e-01 -2.10103989e-01 4.45242643e-01 3.93317252e-01 -7.49029338e-01 -1.78333259e+00 -1.06524169e+00 3.98180127e-01 -3.03608984e-01 7.28268564e-01 -4.58080232e-01 -8.69134665e-01 7.68086672e-01 -2.08369508e-01 -1.39742166e-01 1.11646041e-01 6.66634664e-02 -2.05444798e-01 1.22776628e-03 -9.41865444e-01 6.37099326e-01 1.69132912e+00 -5.16938746e-01 -6.26945853e-01 5.89519620e-01 9.17956293e-01 -1.32956648e+00 -1.11312175e+00 7.62783825e-01 4.18782055e-01 -5.57815671e-01 1.34506619e+00 -3.45549285e-01 9.96485770e-01 -4.04586494e-01 -1.85347840e-01 -1.09848654e+00 -2.72926033e-01 -1.14570796e-01 -4.12974238e-01 1.17799199e+00 2.09115520e-01 -3.08109045e-01 1.03372693e+00 6.64199233e-01 -7.01552153e-01 -1.21505022e+00 -4.97228295e-01 -7.05697000e-01 2.62836725e-01 -8.92190814e-01 1.26695037e+00 9.86327410e-01 1.84579834e-01 -2.47729546e-03 -1.97606519e-01 2.60800272e-01 3.59039277e-01 2.49075234e-01 9.36259449e-01 -1.10220838e+00 1.33955196e-01 -7.12386191e-01 -5.62102616e-01 -1.22382879e+00 4.61860448e-01 -1.14766586e+00 -3.07706803e-01 -2.11660457e+00 -2.03300819e-01 -5.53708613e-01 2.83812881e-01 6.58409059e-01 2.12786674e-01 -7.04277679e-02 5.87779805e-02 -2.98234940e-01 -8.74752700e-01 6.81989491e-01 1.74828935e+00 -4.96607602e-01 1.55486807e-01 -6.92134917e-01 -7.67028570e-01 8.42408121e-01 6.33372068e-01 1.02967128e-01 -5.18029928e-01 -9.25476432e-01 8.02465081e-01 1.98704138e-01 7.54512548e-01 -1.05419648e+00 4.60119367e-01 -3.36898029e-01 4.77401465e-01 -1.35043240e+00 4.79712188e-01 -5.65083086e-01 3.12745601e-01 1.72581851e-01 -1.28412724e-01 1.58113297e-02 3.96952242e-01 1.24731615e-01 -2.72069156e-01 1.20465748e-01 1.93667442e-01 -3.92643660e-01 -1.20350540e+00 6.12548947e-01 1.11930437e-01 2.55885214e-01 9.50567961e-01 -2.26256564e-01 -7.55495965e-01 -2.28043646e-01 -6.38670623e-01 9.17336881e-01 4.18897182e-01 3.75808388e-01 7.08428741e-01 -1.71304476e+00 -4.92356807e-01 8.94048437e-03 6.92958906e-02 1.06701946e+00 8.56287718e-01 5.86035788e-01 -1.22333682e+00 6.14006162e-01 -3.00280392e-01 -4.12411124e-01 -8.04928601e-01 5.57873011e-01 2.67154932e-01 -3.45683917e-02 -8.51677418e-01 1.20328522e+00 2.93077886e-01 -7.77028263e-01 2.98728496e-01 -8.36201131e-01 1.48568666e-02 -3.71425390e-01 2.52382278e-01 6.01053596e-01 3.39348391e-02 -5.16752303e-01 -5.44583440e-01 5.94641089e-01 1.32309735e-01 5.11828661e-01 1.30643237e+00 1.62933290e-01 -3.83155882e-01 5.41027635e-02 9.56305146e-01 -3.88796508e-01 -7.92716742e-01 -9.01420265e-02 -2.80459195e-01 -3.48438710e-01 -2.03853294e-01 -8.41772795e-01 -1.29613662e+00 8.05139422e-01 -2.28357762e-01 -2.36779861e-02 9.77283061e-01 2.29578272e-01 1.01816654e+00 6.34679437e-01 4.18197304e-01 -1.06391311e+00 -9.55345929e-02 7.23255157e-01 1.20430589e+00 -1.04026341e+00 9.71255824e-02 -1.22799730e+00 -3.72088343e-01 1.04796052e+00 1.17887700e+00 -4.38210309e-01 5.01381218e-01 2.94403166e-01 -3.19221497e-01 -5.76613784e-01 -4.92746472e-01 8.66132556e-04 4.94748294e-01 5.83691835e-01 1.58935204e-01 1.07654855e-01 1.76111124e-02 9.62215543e-01 -9.25324678e-01 1.24468781e-01 1.86427549e-01 8.18323135e-01 7.34948441e-02 -7.35759854e-01 -1.48642436e-01 5.20557880e-01 1.82035550e-01 -1.92791402e-01 -4.29891944e-01 1.06797957e+00 4.80526507e-01 5.11853099e-01 1.48968741e-01 -4.30499822e-01 2.80580789e-01 -4.14741427e-01 7.87643433e-01 -2.55892694e-01 -8.76301304e-02 -4.93039757e-01 4.08393145e-01 -7.46686876e-01 -2.02581972e-01 -4.14849252e-01 -1.63648689e+00 -8.13785315e-01 -6.01470880e-02 4.00261171e-02 3.52979898e-01 1.15622497e+00 2.00475201e-01 1.08675277e+00 -1.03942871e-01 -7.85770893e-01 7.71794245e-02 -6.35348678e-01 -3.96226168e-01 1.71250209e-01 -1.49965867e-01 -9.31683302e-01 1.82786882e-01 -2.24421695e-01]
[9.418240547180176, 7.4313201904296875]
412cad54-3b4b-4cc6-9b55-162c6a2cab87
gspmd-general-and-scalable-parallelization
2105.04663
null
https://arxiv.org/abs/2105.04663v2
https://arxiv.org/pdf/2105.04663v2.pdf
GSPMD: General and Scalable Parallelization for ML Computation Graphs
We present GSPMD, an automatic, compiler-based parallelization system for common machine learning computations. It allows users to write programs in the same way as for a single device, then give hints through a few annotations on how to distribute tensors, based on which GSPMD will parallelize the computation. Its representation of partitioning is simple yet general, allowing it to express different or mixed paradigms of parallelism on a wide variety of models. GSPMD infers the partitioning for every operator based on limited user annotations, making it convenient to scale existing single-device programs. It solves several technical challenges for production usage, allowing GSPMD to achieve 50% to 62% compute utilization on up to 2048 Cloud TPUv3 cores for models with up to one trillion parameters.
['Zhifeng Chen', 'Yonghui Wu', 'Tao Wang', 'Shibo Wang', 'Noam Shazeer', 'Ruoming Pang', 'Marcello Maggioni', 'Andy Ly', 'Dmitry Lepikhin', 'Maxim Krikun', 'Rahul Joshi', 'Yanping Huang', 'Blake Hechtman', 'Dehao Chen', 'HyoukJoong Lee', 'Yuanzhong Xu']
2021-05-10
null
null
null
null
['2048']
['playing-games']
[-5.12876332e-01 -2.23067731e-01 -4.52198893e-01 -3.17847550e-01 -5.76396704e-01 -5.97988725e-01 6.53103650e-01 1.98654041e-01 -1.03076838e-01 4.67490070e-02 -6.23515993e-03 -8.73555362e-01 9.09718103e-04 -6.53935969e-01 -4.57224905e-01 -5.37295461e-01 -7.94864073e-02 1.04120755e+00 1.87173173e-01 2.78207306e-02 -2.69853417e-02 7.11808264e-01 -1.57981002e+00 8.18559647e-01 1.81039140e-01 9.97844875e-01 3.09099078e-01 9.34352517e-01 -2.61342287e-01 1.02037084e+00 -3.40433270e-01 -1.78310558e-01 5.31475961e-01 3.59691232e-01 -1.29297304e+00 2.91903354e-02 3.73000622e-01 -2.34494016e-01 -1.02160975e-01 5.25004387e-01 9.74522382e-02 -1.50830254e-01 4.33185756e-01 -1.44880068e+00 7.78106004e-02 8.10191810e-01 -6.02799535e-01 4.68388945e-01 4.28057909e-02 2.99689591e-01 1.20149755e+00 -8.21010947e-01 4.43056673e-01 9.09997761e-01 6.99474156e-01 1.20411202e-01 -1.56459975e+00 -3.26577961e-01 -2.32574850e-01 -2.43785232e-01 -1.36266065e+00 -5.21338999e-01 2.47433782e-01 -9.64571118e-01 1.69225287e+00 6.00368857e-01 6.47506833e-01 3.70258659e-01 4.47400808e-01 6.69438064e-01 1.03618777e+00 -5.07268131e-01 1.58673733e-01 -5.51263466e-02 5.58562398e-01 1.00008655e+00 -1.09717369e-01 -2.81670630e-01 -6.16206288e-01 -6.88585877e-01 5.27212799e-01 -2.83600330e-01 1.91738576e-01 -3.62292290e-01 -1.54589903e+00 6.62913144e-01 1.41943023e-01 1.43711999e-01 -3.02195013e-01 1.90521926e-01 1.05228209e+00 1.67497516e-01 4.92321879e-01 5.44445634e-01 -1.14646375e+00 -4.25269306e-01 -9.11630630e-01 5.69176257e-01 9.46544826e-01 1.12547410e+00 1.00799155e+00 -2.07700580e-02 -1.75001711e-01 7.14714527e-01 1.26750261e-01 1.44967362e-01 2.58188248e-01 -1.39034104e+00 3.20509255e-01 3.75873536e-01 -1.66344583e-01 -7.56758988e-01 -7.40862966e-01 -2.78597921e-01 -8.98961067e-01 1.16410352e-01 7.98493624e-02 -2.03774303e-01 -4.68393415e-01 1.09970152e+00 4.67643440e-01 -1.12945162e-01 -3.15421104e-01 8.56066704e-01 5.37252307e-01 7.98414946e-01 7.17815608e-02 8.08610693e-02 1.58566248e+00 -1.33491933e+00 -2.53235940e-02 -1.45491436e-01 1.26329160e+00 -1.00579011e+00 1.35360491e+00 7.02619970e-01 -1.04773390e+00 -6.33270383e-01 -8.24803174e-01 -4.79653060e-01 -1.89681679e-01 3.04260664e-02 1.56213319e+00 5.92279971e-01 -1.51951194e+00 7.14396119e-01 -1.49052930e+00 -2.10071713e-01 5.45829516e-06 5.02124190e-01 -3.09775327e-03 1.23036645e-01 -3.90282065e-01 6.68903112e-01 6.51515663e-01 -2.98151731e-01 -4.64895338e-01 -1.36061227e+00 -4.41955090e-01 3.77070755e-02 -1.15464583e-01 -1.16503549e+00 1.57168424e+00 -5.31334281e-01 -1.37013090e+00 1.21218467e+00 -3.23024899e-01 -3.91788632e-01 2.16569901e-01 8.05716664e-02 -3.76186520e-01 -1.83580786e-01 1.93316653e-01 6.05761349e-01 6.16675973e-01 -6.34836316e-01 -5.70857406e-01 -2.77628243e-01 1.60174608e-01 3.13788116e-01 -3.02592814e-01 4.43816334e-01 -1.03330779e+00 -3.38598192e-01 1.22670187e-02 -1.14292598e+00 -2.03295559e-01 -3.08654934e-01 -4.94264930e-01 -2.85772711e-01 9.32806611e-01 -4.46845233e-01 1.02664852e+00 -1.96107292e+00 3.00470173e-01 8.90844315e-02 7.20407188e-01 -5.90189286e-02 3.04909982e-02 4.29574966e-01 -1.94593206e-01 3.02449521e-02 -3.63969663e-03 -4.37848181e-01 2.43833750e-01 4.75383133e-01 -4.65661377e-01 2.94842213e-01 -3.75387445e-02 3.79823059e-01 -6.20904684e-01 -3.99034023e-01 1.86595649e-01 1.04509577e-01 -8.86755168e-01 2.36039162e-01 -5.10030329e-01 4.09399986e-01 -3.69314492e-01 5.42729020e-01 9.27256405e-01 -7.70089328e-01 5.65437019e-01 -3.18537980e-01 -3.32754999e-01 6.05137765e-01 -1.09216285e+00 1.63091540e+00 -6.46826088e-01 6.05626762e-01 3.94829005e-01 -6.75253510e-01 2.91262299e-01 1.25812173e-01 6.91342175e-01 -1.69573158e-01 -2.19271615e-01 2.81333439e-02 -2.87485838e-01 -3.59192789e-01 6.52469993e-01 9.89910141e-02 -2.40489766e-01 9.51439083e-01 1.23563424e-01 -4.70686138e-01 4.67413425e-01 4.29678738e-01 1.37163138e+00 -7.12534413e-02 1.45549983e-01 -8.43358219e-01 9.61818919e-02 2.98565954e-01 4.69440401e-01 7.08919168e-01 3.09986115e-01 1.80834621e-01 8.01984787e-01 -9.34266865e-01 -1.25594270e+00 -8.82511020e-01 -4.01880175e-01 1.83247447e+00 -6.93486691e-01 -1.14471626e+00 -7.36046791e-01 -4.38861847e-01 3.66459563e-02 5.03797889e-01 -5.62235825e-02 6.20703638e-01 -4.18678135e-01 -1.16982424e+00 5.09136021e-01 4.94202375e-01 4.30198014e-01 -5.86103141e-01 -3.92778248e-01 -1.02261879e-01 -9.29208822e-04 -1.19141793e+00 -3.35624725e-01 4.34030831e-01 -1.12601995e+00 -8.31930995e-01 3.39630321e-02 -4.55096871e-01 4.15752083e-01 2.58785367e-01 1.83045042e+00 4.00013536e-01 -1.32695764e-01 6.96126446e-02 1.36380538e-01 -2.38945842e-01 -4.62724686e-01 6.76977217e-01 1.52738720e-01 -5.33586442e-01 4.60640728e-01 -6.96732223e-01 5.79174533e-02 -1.20886587e-01 -5.84745169e-01 8.56205761e-01 8.92542750e-02 6.71016812e-01 5.36878884e-01 1.77551210e-01 -5.77389359e-01 -1.30002999e+00 5.12030303e-01 -4.79181588e-01 -1.02551842e+00 -7.65483230e-02 -8.49923313e-01 2.94117510e-01 8.57022285e-01 -2.23999619e-02 -7.59436131e-01 1.65014863e-01 -2.28610083e-01 -5.49682081e-01 -2.55231917e-01 4.99946684e-01 4.16507311e-02 -1.00160599e-01 6.60222709e-01 -1.08892277e-01 -2.32895568e-01 -6.27297997e-01 4.61975485e-01 4.78002846e-01 2.82033592e-01 -1.33687818e+00 4.86489832e-01 2.00050324e-01 6.14761338e-02 -8.62300217e-01 -6.59311414e-01 -3.92208874e-01 -7.77272046e-01 2.11175293e-01 6.80882812e-01 -1.08414674e+00 -1.15728784e+00 5.75135112e-01 -1.15831017e+00 -8.32727253e-01 7.96513706e-02 2.42156655e-01 -4.16591078e-01 5.84312677e-02 -1.28143597e+00 -1.31883115e-01 -3.79649460e-01 -1.49020958e+00 1.27212548e+00 -3.30917388e-01 -3.57842624e-01 -1.12028384e+00 -7.50103127e-03 5.19864440e-01 4.77607340e-01 -3.10497195e-01 1.37429368e+00 -2.67640054e-01 -7.80773520e-01 2.22116202e-01 -3.93426597e-01 1.17611803e-01 -4.42442268e-01 3.58616292e-01 -1.03172553e+00 -3.81112814e-01 -1.00001931e-01 -1.18214227e-01 5.30265689e-01 4.04119566e-02 1.83882999e+00 -3.33720803e-01 -4.64755446e-01 1.02546382e+00 1.40024662e+00 -3.62396896e-01 3.35220039e-01 -6.49402896e-03 9.95818019e-01 -5.46327000e-03 1.15317836e-01 5.41302383e-01 5.22542894e-01 8.83757114e-01 2.70029549e-02 -1.71723172e-01 1.55404061e-01 1.31872103e-01 5.53037450e-02 1.59875703e+00 -3.18546593e-01 2.39683583e-01 -1.44239211e+00 6.36402741e-02 -1.71519542e+00 -5.00201762e-01 -6.62203133e-01 1.90678978e+00 7.09248662e-01 2.18635961e-01 1.13529183e-01 -1.63689584e-01 1.84728220e-01 2.73747265e-01 -1.41402751e-01 -8.66697133e-01 2.70566374e-01 5.06397367e-01 9.19055760e-01 5.65402865e-01 -1.08941591e+00 1.15357697e+00 8.03082085e+00 8.84099782e-01 -1.37786984e+00 6.29198372e-01 7.68431723e-01 -7.93339014e-02 -1.77155226e-01 1.72591954e-01 -8.14023793e-01 3.35675776e-01 1.21132874e+00 -6.71901926e-02 8.35769475e-01 1.10528791e+00 1.03528455e-01 -8.24106857e-02 -1.26482749e+00 9.42666173e-01 -3.30042452e-01 -1.77803922e+00 -1.92809135e-01 1.05348453e-01 7.35204995e-01 7.97808945e-01 -2.37193108e-01 4.48259652e-01 7.37274051e-01 -7.73482025e-01 5.73580921e-01 2.64164656e-01 7.58310795e-01 -6.95919394e-01 3.81359518e-01 5.94662666e-01 -1.28140640e+00 1.67588949e-01 -3.76838684e-01 -6.60024762e-01 -2.41364121e-01 9.14333045e-01 -9.74667728e-01 4.20826912e-01 9.20519233e-01 3.33510667e-01 -6.68103695e-01 3.47035825e-01 1.60382062e-01 7.65890718e-01 -4.86797243e-01 2.49244288e-01 -4.01444361e-02 -3.12210858e-01 1.47620831e-02 1.40082777e+00 1.47284031e-01 8.75073150e-02 4.78384018e-01 5.27790308e-01 -4.06872854e-02 1.00512788e-01 -4.76446241e-01 9.72396359e-02 4.14516628e-01 1.45634687e+00 -8.06641579e-01 -6.22653306e-01 -5.49951911e-01 9.39201057e-01 7.30530560e-01 2.13268131e-01 -8.66294086e-01 -1.22640608e-02 9.74603653e-01 2.63984412e-01 -9.34846513e-03 -7.67347932e-01 -5.85209072e-01 -1.44535506e+00 -1.16952933e-01 -1.09308076e+00 5.11065543e-01 -9.57238674e-01 -1.09163880e+00 5.60999334e-01 3.44687440e-02 -6.60645604e-01 -1.58774242e-01 -9.48500991e-01 -3.56534570e-01 1.04466081e+00 -7.21658587e-01 -1.01058984e+00 -9.51860771e-02 3.32338870e-01 1.24724910e-01 -3.21639985e-01 1.18238544e+00 3.98781806e-01 -7.57165551e-01 2.12088659e-01 -2.37369351e-02 3.21835168e-02 3.87292534e-01 -1.20489240e+00 7.72496581e-01 5.87948442e-01 2.78351575e-01 7.72287250e-01 6.99477911e-01 -3.19549024e-01 -1.90827370e+00 -9.90213990e-01 9.56756294e-01 -6.34961009e-01 8.63084614e-01 -5.10409057e-01 -7.10566163e-01 1.18519080e+00 4.45241779e-01 3.09707731e-01 6.75960124e-01 9.23734665e-01 -4.04503912e-01 -4.05235291e-01 -7.40628064e-01 5.04596174e-01 9.78583694e-01 -7.53039360e-01 4.81337029e-03 1.01387930e+00 6.42799258e-01 -1.02997339e+00 -1.29200661e+00 1.45582274e-01 -2.36350931e-02 -1.04572868e+00 7.53814995e-01 -7.98812866e-01 5.24266899e-01 -3.76820683e-01 -2.23483562e-01 -1.02684677e+00 -6.51745915e-01 -6.15524292e-01 -4.15364325e-01 8.59488130e-01 4.11574066e-01 -6.75110042e-01 7.48301387e-01 9.20485973e-01 -3.62105221e-01 -6.49797618e-01 -5.28420329e-01 -4.15469736e-01 3.16035859e-02 -1.09947562e+00 9.13971722e-01 1.26605344e+00 1.49853855e-01 5.24932504e-01 -9.47619751e-02 3.55865777e-01 4.12218422e-01 3.99497062e-01 1.13240647e+00 -1.11790359e+00 -9.57653999e-01 -6.00394011e-01 -1.11001968e-01 -1.29203498e+00 2.42609695e-01 -1.38990378e+00 -5.79504788e-01 -8.96681190e-01 5.05695343e-01 -9.92031693e-01 -1.18318703e-02 9.11568940e-01 3.61923039e-01 2.29738623e-01 1.30111292e-01 4.50747102e-01 -4.91817385e-01 -1.23417340e-01 7.42679954e-01 -3.52274254e-02 8.78020376e-02 -9.74528119e-02 -5.59749484e-01 8.28803897e-01 6.07591569e-01 -3.78575593e-01 -2.59036005e-01 -1.05736375e+00 4.24796194e-01 1.78915277e-01 2.88435698e-01 -1.15008891e+00 1.79568172e-01 -3.51817310e-01 3.63630712e-01 -5.01179397e-01 2.55068481e-01 -5.24760425e-01 5.91083407e-01 1.35753125e-01 -1.82136029e-01 6.32144094e-01 4.08609480e-01 -2.63960660e-01 -2.36185379e-02 8.95548537e-02 6.85256541e-01 -2.20689654e-01 -7.14980006e-01 4.41890895e-01 -5.18824697e-01 3.93143073e-02 7.49297380e-01 5.47344983e-01 -2.61991024e-01 1.46324858e-01 -6.36323571e-01 5.32991774e-02 7.83447266e-01 2.35108823e-01 -1.61077172e-01 -1.20816457e+00 -4.26547766e-01 4.75459725e-01 2.49712020e-02 -3.11041437e-02 3.52166235e-01 8.49131942e-01 -1.18960571e+00 6.90751910e-01 -5.79673052e-02 -8.85828853e-01 -1.30384469e+00 8.47283185e-01 2.47369856e-01 -5.84417939e-01 -7.35890269e-01 7.58443236e-01 -1.04070686e-01 -7.47924984e-01 -3.65429938e-01 -4.72368151e-01 6.27875984e-01 -3.56593788e-01 4.84092087e-01 1.81891486e-01 8.34945321e-01 -4.13937092e-01 -2.57970452e-01 1.06980145e-01 -2.69261748e-03 3.30111012e-02 1.14328384e+00 2.43013218e-01 -7.80043244e-01 5.40465474e-01 1.24159694e+00 7.96507150e-02 -8.92042935e-01 -1.42167419e-01 -7.88151026e-02 -3.37426007e-01 3.27831924e-01 -5.59455276e-01 -1.24263155e+00 8.52847099e-01 1.75645545e-01 2.50486612e-01 1.07465076e+00 7.02858418e-02 6.28293157e-01 5.06240547e-01 8.08526397e-01 -1.07867777e+00 -5.42365849e-01 8.32720399e-01 4.49110031e-01 -8.74726117e-01 4.18187618e-01 -5.85038483e-01 -4.08629328e-01 1.29818988e+00 5.11744857e-01 3.08345854e-01 9.86751735e-01 1.00592136e+00 -1.59517974e-02 -3.23709935e-01 -1.14600015e+00 2.54607975e-01 2.18714565e-01 3.26635987e-01 7.65675843e-01 6.51269734e-01 -3.48833427e-02 1.32536501e-01 -4.99705136e-01 -5.56455143e-02 3.12090605e-01 8.31652641e-01 1.47763148e-01 -1.38155258e+00 -3.76450002e-01 7.67905056e-01 -2.93216795e-01 -2.57346630e-01 3.99206340e-01 6.51153564e-01 2.60775864e-01 4.43698078e-01 5.57925165e-01 -6.03125751e-01 -1.72788113e-01 1.38872176e-01 4.57836539e-01 -8.50999177e-01 -8.14680874e-01 -5.61247543e-02 5.19689262e-01 -6.48905277e-01 8.81899595e-02 -6.15379453e-01 -1.22618389e+00 -1.12750888e+00 2.91988671e-01 9.60290506e-02 7.16314018e-01 9.53998744e-01 8.44619274e-01 4.03584808e-01 4.15504634e-01 -1.15819693e+00 -3.36846888e-01 -7.12739646e-01 -5.93477905e-01 1.29012316e-01 -2.58840859e-01 -2.74755001e-01 1.17284730e-02 6.84628859e-02]
[8.47344970703125, 3.4983651638031006]
e63951f1-2e76-44dd-af93-5da1eca53674
sigmorphon-2022-task-0-submission-description
null
null
https://aclanthology.org/2022.sigmorphon-1.20
https://aclanthology.org/2022.sigmorphon-1.20.pdf
SIGMORPHON 2022 Task 0 Submission Description: Modelling Morphological Inflection with Data-Driven and Rule-Based Approaches
This paper describes our participation in the 2022 SIGMORPHON-UniMorph Shared Task on Typologically Diverse and AcquisitionInspired Morphological Inflection Generation. We present two approaches: one being a modification of the neural baseline encoderdecoder model, the other being hand-coded morphological analyzers using finite-state tools (FST) and outside linguistic knowledge. While our proposed modification of the baseline encoder-decoder model underperforms the baseline for almost all languages, the FST methods outperform other systems in the respective languages by a large margin. This confirms that purely data-driven approaches have not yet reached the maturity to replace trained linguists for documentation and analysis especially considering low-resource and endangered languages.
['Ryan Soh-Eun Shim', 'Jingwen Li', 'Leander Girrbach', 'Nkonye Gbadegoye', 'Tatiana Merzhevich']
null
null
null
null
naacl-sigmorphon-2022-7
['morphological-inflection']
['natural-language-processing']
[ 1.48771495e-01 3.02669883e-01 -2.17240855e-01 -3.41019243e-01 -8.80913317e-01 -9.79320467e-01 6.15534782e-01 1.64986834e-01 -9.29288864e-01 9.13592041e-01 4.19455111e-01 -6.36277199e-01 9.63967890e-02 -4.80893642e-01 -6.06526732e-01 -3.13197792e-01 2.02052221e-01 8.71065140e-01 -6.56578541e-02 -4.99568671e-01 3.78291696e-01 2.42223099e-01 -1.26518381e+00 3.56666058e-01 9.77723897e-01 2.56164912e-02 2.00090781e-01 5.06546438e-01 -1.13380216e-01 4.71436471e-01 -6.47882462e-01 -1.15028572e+00 3.26772571e-01 -4.57551450e-01 -8.17851722e-01 -4.77769226e-01 8.22380006e-01 -3.53414118e-01 -1.81890633e-02 1.35549259e+00 5.45327723e-01 -1.70400307e-01 6.85391128e-01 -4.12040383e-01 -1.19944489e+00 1.69890237e+00 -2.12297916e-01 5.03607094e-01 3.84365529e-01 3.39632660e-01 1.11408007e+00 -7.85385966e-01 9.06817794e-01 1.46982968e+00 8.35000336e-01 5.88692784e-01 -1.43908489e+00 -3.49814713e-01 3.51996683e-02 1.62366912e-01 -1.32305288e+00 -6.59925461e-01 6.53976262e-01 -5.67092597e-01 1.58318114e+00 -1.58436149e-01 5.89879990e-01 1.11319983e+00 -1.27268001e-01 6.67910933e-01 1.12307084e+00 -7.15133131e-01 -5.68440855e-02 2.19680384e-01 1.32644832e-01 7.08071411e-01 6.77271903e-01 3.58063668e-01 -8.08833778e-01 5.56127876e-02 6.50984764e-01 -1.11930418e+00 4.45906725e-03 5.19833863e-01 -1.30226612e+00 7.28024542e-01 -1.19399220e-01 6.74838960e-01 -4.77487743e-01 3.51743365e-04 7.29214072e-01 6.84881747e-01 1.20649807e-01 6.25654757e-01 -6.07929885e-01 -3.01762074e-01 -1.23807311e+00 2.35133052e-01 6.92671001e-01 1.00928044e+00 4.02430475e-01 9.55395341e-01 2.76906252e-01 9.79159951e-01 1.87094614e-01 4.89701331e-01 8.98589849e-01 -4.66540605e-01 5.15365303e-01 5.09311974e-01 -5.47341049e-01 -3.24181795e-01 -2.27842301e-01 -1.46413445e-01 -4.35979754e-01 1.53651848e-01 7.71466732e-01 -1.52589232e-01 -7.52131045e-01 1.93998051e+00 -1.50495335e-01 -5.80216050e-01 8.70479364e-03 3.71632814e-01 5.04785001e-01 5.27987719e-01 3.88101339e-01 -1.01829499e-01 1.42049336e+00 -3.92726481e-01 -7.04815626e-01 -2.21827582e-01 5.93790710e-01 -7.37894058e-01 1.42051566e+00 7.10084140e-01 -1.62843764e+00 -6.98589385e-01 -1.46195292e+00 -1.86868280e-01 -6.72773123e-01 1.31922007e-01 7.03648984e-01 1.12279475e+00 -1.22141707e+00 8.10552537e-01 -6.78139627e-01 -4.27523762e-01 6.73317164e-02 3.02801043e-01 -4.68445688e-01 4.99604553e-01 -1.20819235e+00 1.28989244e+00 1.12342513e+00 -1.13713909e-02 -6.62008822e-01 -5.74902952e-01 -8.23966682e-01 -3.77033859e-01 -1.80036023e-01 -4.43653792e-01 1.25064576e+00 -8.87915313e-01 -1.70249116e+00 1.39139009e+00 3.24663460e-01 -7.33560920e-01 4.22254384e-01 -3.53880078e-01 -6.63221419e-01 -1.04904547e-01 -1.34841859e-01 7.47509003e-01 3.47176343e-01 -8.78179610e-01 -6.25647902e-01 -2.73512125e-01 -3.64401013e-01 7.88059756e-02 -2.94187963e-01 5.41773021e-01 2.11209998e-01 -8.44849050e-01 -3.19693834e-01 -7.02687144e-01 -1.13956472e-02 -4.04650509e-01 -4.98989791e-01 -1.91904843e-01 -7.19863549e-02 -1.34733343e+00 1.23679578e+00 -1.76715457e+00 1.58926710e-01 -7.89411068e-02 -4.65398967e-01 5.96714914e-01 -1.08088836e-01 7.19098389e-01 -1.08021840e-01 3.23442042e-01 -5.55326641e-01 -2.77814507e-01 3.50989610e-01 2.09496364e-01 -2.09827647e-01 4.75112587e-01 3.51600587e-01 9.77876663e-01 -9.00906563e-01 -5.11258721e-01 1.87464461e-01 1.60021514e-01 -6.16138697e-01 -1.66593686e-01 -3.39073867e-01 3.63870710e-01 4.44429100e-01 7.09018648e-01 4.94835228e-01 7.75089622e-01 4.20685887e-01 4.08909440e-01 -5.29240549e-01 9.21812475e-01 -1.10848165e+00 2.04007745e+00 -5.89228749e-01 6.00168765e-01 -7.32313246e-02 -4.88617063e-01 9.76080656e-01 3.81942034e-01 -2.26420715e-01 -4.86976296e-01 5.01619428e-02 9.20381784e-01 7.30683684e-01 -2.80771106e-01 8.08182538e-01 -6.31584346e-01 -3.00505459e-01 2.94701040e-01 5.47464073e-01 -1.71248510e-01 4.12506849e-01 -1.80771768e-01 8.66041660e-01 5.73168099e-01 7.45138407e-01 -7.37720191e-01 7.42255628e-01 1.06895804e-01 5.04927695e-01 3.97611767e-01 -3.95575836e-02 3.78118694e-01 2.53047019e-01 -2.86621004e-01 -1.27808237e+00 -1.37275457e+00 -4.02091771e-01 1.15937829e+00 -7.04192281e-01 -3.11513603e-01 -9.64926481e-01 -4.63486046e-01 -1.50598422e-01 1.36715138e+00 -3.22402209e-01 -1.51909322e-01 -1.10968971e+00 -7.27094948e-01 1.51059687e+00 5.39171934e-01 1.99100763e-01 -1.64237273e+00 -4.01339352e-01 5.38186967e-01 7.53220692e-02 -7.49654770e-01 -9.04886127e-02 2.75139451e-01 -6.25179708e-01 -5.86615503e-01 -6.22294009e-01 -1.11661482e+00 1.50897637e-01 -8.81427526e-01 9.63214815e-01 -2.66923755e-01 5.54648042e-02 -3.52153838e-01 -1.76831722e-01 -5.51126242e-01 -1.06970084e+00 4.94865566e-01 2.75080502e-01 -4.78367984e-01 7.59664655e-01 -7.34192014e-01 4.71594036e-02 -5.90003371e-01 -7.62062490e-01 -4.02873665e-01 7.00535715e-01 5.50884604e-01 4.15414900e-01 -4.81777787e-01 1.06857061e+00 -1.03820133e+00 6.34012818e-01 -3.47431928e-01 -5.99375725e-01 1.44613087e-01 -5.46792209e-01 1.93817213e-01 1.06786203e+00 -3.49680543e-01 -1.12580383e+00 1.02393299e-01 -8.31637979e-01 3.12160075e-01 -4.84236002e-01 6.32881999e-01 -6.17736638e-01 4.12921160e-01 7.26341128e-01 1.97639078e-01 -3.09661806e-01 -8.94599855e-01 6.69670105e-01 8.22951853e-01 1.23736525e+00 -6.66866064e-01 8.61424625e-01 -6.58855140e-02 -4.78285551e-01 -1.01068819e+00 -8.66819173e-02 3.12589668e-02 -1.20386016e+00 8.14204589e-02 7.78322220e-01 -6.80076301e-01 -2.18871996e-01 6.21135473e-01 -1.42719936e+00 -4.06443745e-01 -7.79435277e-01 5.09111762e-01 -6.50612414e-01 3.56244445e-01 -9.25012946e-01 -6.56590104e-01 -6.69437468e-01 -8.42160523e-01 8.18965614e-01 -8.78398642e-02 -6.59411788e-01 -1.11471045e+00 6.68438792e-01 -1.52396053e-01 3.32828373e-01 2.38521233e-01 1.69933891e+00 -1.09106505e+00 -3.32618766e-02 -8.85162652e-02 2.63560921e-01 6.52766705e-01 -9.56123918e-02 2.20795512e-01 -8.28319788e-01 -8.42390209e-02 -7.74945021e-02 -3.25571746e-01 6.93347812e-01 4.34612446e-02 3.03287923e-01 -3.07093561e-01 3.23640674e-01 7.57831097e-01 1.45313764e+00 1.95539832e-01 7.09784389e-01 4.15248185e-01 3.70919555e-01 6.42898023e-01 -1.47881910e-01 2.49554701e-02 4.65076506e-01 4.24976885e-01 -1.64246812e-01 2.22890198e-01 -6.81085408e-01 -7.37355173e-01 1.00448048e+00 1.58960557e+00 -5.97698912e-02 -2.82019496e-01 -1.19476664e+00 1.08126426e+00 -1.34000051e+00 -9.43075061e-01 -4.71346796e-01 2.37218595e+00 1.34925020e+00 3.04120183e-01 2.75568157e-01 2.04531580e-01 7.89009929e-01 -6.15633912e-02 -3.67334247e-01 -1.27218688e+00 -6.52418733e-01 8.13509464e-01 4.13265109e-01 6.09931707e-01 -5.97271740e-01 1.60989273e+00 6.94427919e+00 7.44695842e-01 -7.11313367e-01 1.55670553e-01 -8.46405551e-02 9.29223299e-02 -5.25815964e-01 2.13274479e-01 -9.31045294e-01 5.16857564e-01 1.59378779e+00 -4.67299998e-01 6.14719689e-01 6.54855192e-01 -1.89135760e-01 1.49280548e-01 -1.26699376e+00 6.38768911e-01 2.35295117e-01 -9.49793220e-01 3.56054813e-01 9.08163562e-02 4.22077298e-01 3.19298059e-01 -1.75601635e-02 3.20582867e-01 7.18219936e-01 -9.29851472e-01 1.20670497e+00 3.41168672e-01 1.31369424e+00 -8.11571360e-01 4.94505554e-01 2.31853917e-01 -8.76131952e-01 2.70048976e-01 -4.95350033e-01 -4.70671318e-02 5.08825719e-01 2.38071606e-01 -6.03674471e-01 4.17718768e-01 -2.87789628e-02 2.95549870e-01 -1.00860906e+00 9.21176016e-01 -7.66908526e-01 1.12761998e+00 -3.98794144e-01 4.10364643e-02 1.12299003e-01 -3.05751879e-02 9.08500969e-01 1.79627907e+00 1.55164808e-01 -5.00953257e-01 -2.01273829e-01 8.15813839e-01 3.12043037e-02 6.68166697e-01 -6.60363495e-01 -1.90217108e-01 5.00822783e-01 8.59556437e-01 -5.48843384e-01 -4.24147159e-01 -3.37498069e-01 9.85237241e-01 6.22388184e-01 -2.28670582e-01 -4.98298526e-01 -7.45886385e-01 4.50055301e-01 4.26811911e-02 1.86750025e-01 -4.76442575e-01 -4.73364055e-01 -9.90691185e-01 -1.22629747e-01 -1.11701584e+00 4.51332957e-01 -2.54531115e-01 -1.39148176e+00 9.22909379e-01 7.22947344e-02 -7.52011120e-01 -7.23598421e-01 -1.00230360e+00 -3.52577835e-01 8.98836792e-01 -1.22964394e+00 -1.44557142e+00 5.67215860e-01 4.29799914e-01 4.13779855e-01 -5.47574222e-01 1.05116975e+00 4.30183649e-01 -4.68614280e-01 8.72764885e-01 3.04430500e-02 2.29911059e-01 6.41870499e-01 -1.65097773e+00 8.82453620e-01 1.23355007e+00 5.40034592e-01 8.54108393e-01 4.53186363e-01 -8.95205379e-01 -9.56033826e-01 -7.62410760e-01 1.44229174e+00 -3.51384699e-01 9.10588145e-01 -7.12617815e-01 -9.52852488e-01 9.03594196e-01 9.48178351e-01 -8.07281256e-01 9.20457423e-01 5.19136377e-02 -3.30671877e-01 3.78413469e-01 -9.64650989e-01 8.41490388e-01 1.32300687e+00 -6.93085074e-01 -1.34588099e+00 2.22017556e-01 4.47778046e-01 1.05744712e-01 -8.51047993e-01 9.49014574e-02 4.53481048e-01 -6.19724095e-01 4.36792880e-01 -8.37790906e-01 4.71245348e-01 -1.15731440e-01 -2.26349950e-01 -1.54599440e+00 -3.81093025e-01 -1.04920471e+00 1.01615772e-01 1.73921978e+00 7.86930919e-01 -5.82287848e-01 4.56881553e-01 -3.69329304e-02 -5.99372566e-01 -2.46576779e-02 -1.17324436e+00 -1.08500445e+00 9.27415848e-01 -5.65464318e-01 6.42659724e-01 8.68732393e-01 2.13750362e-01 5.21263182e-01 2.00952385e-02 -1.87964335e-01 3.96728426e-01 -2.93525755e-01 3.11316639e-01 -1.09286976e+00 -4.43558246e-01 -7.89712906e-01 -6.52970374e-01 -4.33543742e-01 4.79691505e-01 -1.74300265e+00 4.59126197e-02 -1.17079854e+00 -1.95051625e-01 1.20078363e-02 -1.58611909e-01 5.45293272e-01 -1.02475230e-02 2.42559507e-01 9.88301560e-02 -1.10410996e-01 1.55089781e-01 2.94983357e-01 3.54162365e-01 3.04461032e-01 -1.90316439e-01 -3.73631746e-01 -8.63498271e-01 1.04174054e+00 9.06847537e-01 -7.19690979e-01 -1.59468744e-02 -7.73533344e-01 5.55739045e-01 -4.26262856e-01 4.24937978e-02 -1.07741368e+00 1.63801208e-01 1.30090371e-01 1.68921813e-01 -5.06967127e-01 -1.63210556e-01 -2.45791078e-01 1.81065407e-02 6.20769322e-01 -3.27811390e-01 6.69588864e-01 4.40917134e-01 -1.17362544e-01 -4.98824269e-02 -7.58355379e-01 7.95737326e-01 -3.43371391e-01 -6.87969029e-01 -1.08559892e-01 -9.38888013e-01 4.63701069e-01 6.72940969e-01 -3.98584038e-01 -3.76679689e-01 1.11821219e-01 -5.42777896e-01 -3.64604682e-01 6.67600870e-01 2.49225989e-01 2.50694335e-01 -1.11258161e+00 -1.34450197e+00 3.84116083e-01 7.16147944e-03 -5.60164213e-01 -1.73085734e-01 4.07228172e-01 -7.67414749e-01 4.20636505e-01 -6.90217197e-01 1.53207377e-01 -6.54363394e-01 5.18527508e-01 2.10554197e-01 -2.87204802e-01 -4.58840072e-01 9.59324181e-01 -3.72118622e-01 -8.44039977e-01 -6.28198311e-02 -1.28899410e-01 -1.84223637e-01 2.15613514e-01 4.70774531e-01 4.97552037e-01 3.64151537e-01 -9.44413960e-01 -3.75775784e-01 1.38403177e-01 -3.18942964e-01 -7.78548419e-01 1.28213656e+00 2.02095985e-01 -7.72224888e-02 8.08702350e-01 7.61547446e-01 6.14177525e-01 -6.10591650e-01 4.80494574e-02 6.38878167e-01 9.14638042e-02 -2.38845959e-01 -1.11137986e+00 -5.88538826e-01 9.93816555e-01 6.74420446e-02 -1.69276312e-01 7.29380965e-01 -3.64959657e-01 8.62720191e-01 4.38656032e-01 3.27767372e-01 -1.52228105e+00 -5.82025766e-01 8.79828155e-01 8.77114058e-01 -5.18128932e-01 -1.75621793e-01 -8.17442909e-02 -7.08651900e-01 1.13834214e+00 4.58347589e-01 -4.34083849e-01 3.23271781e-01 6.17607713e-01 7.30140731e-02 -2.00843941e-02 -7.12277472e-01 -4.71234649e-01 2.49310154e-02 8.59861970e-01 7.75326490e-01 1.11800842e-01 -8.45888197e-01 7.55689383e-01 -1.20367444e+00 -3.79682541e-01 4.88339603e-01 5.64303517e-01 -1.85810655e-01 -1.46339166e+00 -7.35677555e-02 2.93988943e-01 -7.24575162e-01 -7.04949260e-01 -7.73349404e-01 1.24918282e+00 4.76860851e-01 6.12920225e-01 9.36037153e-02 -2.96638429e-01 3.64165306e-01 7.47028291e-01 7.40046620e-01 -1.00056577e+00 -1.29399049e+00 -1.39374584e-01 5.08401752e-01 3.81054729e-02 1.02967530e-01 -1.29432607e+00 -1.10734785e+00 -2.83580691e-01 -3.13024014e-01 -1.84941024e-01 7.57817745e-01 6.77814305e-01 -1.66849390e-01 2.36581400e-01 -5.93914911e-02 -5.71910918e-01 -8.93399119e-01 -1.38897431e+00 -6.91384912e-01 1.60822153e-01 -1.92853391e-01 -1.10528558e-01 -1.78005010e-01 3.20943475e-01]
[10.727738380432129, 10.0040922164917]
837b349d-3245-412b-9c88-61668cb8503f
self-supervised-classification-network
2103.10994
null
https://arxiv.org/abs/2103.10994v3
https://arxiv.org/pdf/2103.10994v3.pdf
Self-Supervised Classification Network
We present Self-Classifier -- a novel self-supervised end-to-end classification learning approach. Self-Classifier learns labels and representations simultaneously in a single-stage end-to-end manner by optimizing for same-class prediction of two augmented views of the same sample. To guarantee non-degenerate solutions (i.e., solutions where all labels are assigned to the same class) we propose a mathematically motivated variant of the cross-entropy loss that has a uniform prior asserted on the predicted labels. In our theoretical analysis, we prove that degenerate solutions are not in the set of optimal solutions of our approach. Self-Classifier is simple to implement and scalable. Unlike other popular unsupervised classification and contrastive representation learning approaches, it does not require any form of pre-training, expectation-maximization, pseudo-labeling, external clustering, a second network, stop-gradient operation, or negative pairs. Despite its simplicity, our approach sets a new state of the art for unsupervised classification of ImageNet; and even achieves comparable to state-of-the-art results for unsupervised representation learning. Code is available at https://github.com/elad-amrani/self-classifier.
['Alex Bronstein', 'Leonid Karlinsky', 'Elad Amrani']
2021-03-19
null
null
null
null
['self-supervised-image-classification', 'unsupervised-image-classification']
['computer-vision', 'computer-vision']
[ 4.52994704e-01 4.32130158e-01 -4.51093346e-01 -8.26896787e-01 -8.59133184e-01 -6.16951942e-01 3.83364052e-01 2.49575496e-01 -2.68355280e-01 4.07643050e-01 6.00414276e-02 -1.16335526e-01 -6.36093318e-02 -5.30242443e-01 -5.67619145e-01 -7.62629867e-01 4.90524694e-02 7.37866998e-01 -1.10964663e-01 2.82672286e-01 1.69001862e-01 1.95680693e-01 -1.76372230e+00 4.35214698e-01 6.33275390e-01 1.36647236e+00 -6.90951720e-02 6.10047162e-01 1.48411915e-01 1.06873703e+00 -5.55259921e-02 -4.06867206e-01 4.59592909e-01 -4.99981284e-01 -9.61213827e-01 4.25109804e-01 7.91168928e-01 3.22487727e-02 -4.55163904e-02 1.10456884e+00 4.91831332e-01 2.05861688e-01 1.22712636e+00 -1.51765025e+00 -6.22487068e-01 2.63919115e-01 -6.16903603e-01 -1.02183573e-01 -2.24880818e-02 -1.27933413e-01 1.25086391e+00 -9.47397888e-01 6.95289969e-01 1.02479672e+00 9.28075135e-01 6.93566680e-01 -1.50443137e+00 -6.13954365e-01 -9.75289475e-03 1.56911716e-01 -1.31010747e+00 -4.37623888e-01 6.90699637e-01 -7.17025280e-01 7.24557757e-01 1.61968559e-01 3.70374829e-01 1.00085878e+00 -1.62838370e-01 7.97626793e-01 1.33460057e+00 -5.78758538e-01 3.57640982e-01 4.54626560e-01 3.43280345e-01 1.03611743e+00 -1.20322302e-01 1.08038403e-01 -3.16009700e-01 -1.92163587e-01 4.44639593e-01 1.91530734e-01 5.22118025e-02 -1.05166507e+00 -9.91448164e-01 1.04896903e+00 4.94711101e-01 -3.02964076e-02 -1.75621822e-01 2.56293952e-01 5.50206780e-01 4.27961826e-01 7.21374631e-01 2.61298418e-01 -4.28289711e-01 1.10844657e-01 -9.87153530e-01 -2.26859197e-01 7.37467766e-01 8.43039930e-01 9.77755010e-01 -1.87858060e-01 1.51697740e-01 8.44839394e-01 8.80852267e-02 9.77739319e-02 8.12818885e-01 -1.22533536e+00 8.96366164e-02 6.23496652e-01 -1.98134914e-01 -7.14905262e-01 -5.22805274e-01 -8.07709455e-01 -9.01384413e-01 2.54516423e-01 3.08638275e-01 -4.57886457e-02 -8.09235096e-01 1.83035338e+00 2.57085592e-01 2.83457071e-01 2.10139200e-01 8.79930258e-01 5.98355711e-01 2.81231999e-01 -2.41300873e-02 -2.27835238e-01 1.04318345e+00 -1.30081236e+00 -2.82508314e-01 -1.20655939e-01 7.83044875e-01 -5.40502727e-01 8.76889944e-01 2.30758607e-01 -8.06392908e-01 -4.50894564e-01 -9.38429832e-01 -3.13874776e-03 -4.16414320e-01 3.44768494e-01 6.85368478e-01 6.30184650e-01 -9.87262666e-01 7.65674591e-01 -7.70695627e-01 -3.99419516e-01 6.12338066e-01 3.81980926e-01 -6.76219761e-01 7.96926171e-02 -4.16010857e-01 9.44845319e-01 3.18310350e-01 -4.18398023e-01 -8.85018647e-01 -5.54822683e-01 -8.58464301e-01 -7.56470859e-02 2.77272582e-01 -4.78704005e-01 1.41189241e+00 -1.65251422e+00 -1.37878346e+00 1.48225260e+00 -3.08466971e-01 -4.96143758e-01 5.06253958e-01 -6.50426298e-02 -1.58358976e-01 2.41585046e-01 3.36537212e-01 8.70791018e-01 8.94490957e-01 -1.42264497e+00 -4.59838748e-01 -3.98584962e-01 -2.23861322e-01 2.80204505e-01 -3.59935105e-01 -1.52154371e-01 -7.50007033e-02 -5.06078482e-01 3.76777828e-01 -1.05704236e+00 -2.83455610e-01 2.69078255e-01 -5.12781858e-01 -2.58758873e-01 6.34570301e-01 -3.29022139e-01 8.11026394e-01 -2.31186891e+00 5.63122593e-02 3.09184700e-01 3.01161110e-01 -6.27996447e-03 -1.23471297e-01 3.09893221e-01 -5.45507848e-01 4.75900322e-02 -6.40762866e-01 -6.12511039e-01 3.75732011e-03 7.73230493e-02 -2.51433372e-01 9.69021320e-01 1.98537946e-01 4.85926002e-01 -1.08757782e+00 -6.15283251e-01 2.58862168e-01 3.31119001e-01 -6.52118862e-01 2.21916035e-01 -8.02623257e-02 4.19649988e-01 -6.03315569e-02 5.59272885e-01 5.67644000e-01 -6.03391647e-01 3.45665246e-01 -6.43941164e-02 1.92683727e-01 3.15577805e-01 -1.24397075e+00 1.72503459e+00 -5.29702663e-01 5.48093557e-01 -2.27399886e-01 -1.64824033e+00 8.79211664e-01 2.82014370e-01 5.89412987e-01 -4.17758346e-01 1.52551800e-01 3.44183803e-01 -2.71394312e-01 -2.63310134e-01 -1.09474167e-01 -2.69252568e-01 5.51794991e-02 5.35083711e-01 6.42167687e-01 9.57728624e-02 1.61665067e-01 7.30926469e-02 9.53761935e-01 3.30259144e-01 5.29178500e-01 -3.34599286e-01 3.53045046e-01 -1.27709523e-01 5.74904025e-01 5.94700217e-01 -6.27729669e-02 8.78181636e-01 3.99891287e-01 -3.12856764e-01 -1.07347596e+00 -1.09890020e+00 -2.39099890e-01 1.24612844e+00 -1.59995873e-02 -4.13964450e-01 -8.06183517e-01 -1.07904017e+00 -1.84819382e-02 6.26107275e-01 -6.49634242e-01 -1.78238466e-01 -1.20851502e-01 -3.76920015e-01 1.99922547e-01 4.48617369e-01 3.43384057e-01 -9.14899230e-01 -6.15234494e-01 9.67519581e-02 -7.61136189e-02 -8.62834871e-01 -3.19673657e-01 7.57122695e-01 -1.03556049e+00 -1.23691702e+00 -6.24179482e-01 -1.20634198e+00 1.11181736e+00 1.80468887e-01 1.16838515e+00 -1.17031269e-01 -2.38160983e-01 5.07887661e-01 -1.73034996e-01 -1.45041242e-01 -3.78097832e-01 1.90590322e-01 2.28204746e-02 2.91178197e-01 3.24588507e-01 -8.04465294e-01 -6.43281400e-01 4.82448012e-01 -4.45212245e-01 2.39350237e-02 3.87814492e-01 1.04726541e+00 9.04503047e-01 -1.02946557e-01 5.87840736e-01 -1.08469129e+00 1.36251852e-01 -6.46865785e-01 -5.20351112e-01 2.27354869e-01 -8.35978627e-01 1.98695600e-01 7.56837964e-01 -4.95852172e-01 -6.31285250e-01 6.20627582e-01 1.39365286e-01 -6.38534486e-01 -3.46323252e-01 3.64052534e-01 1.68508515e-01 1.34573162e-01 9.44072545e-01 9.98548567e-02 2.94275135e-01 -5.36692202e-01 5.59128761e-01 9.18514609e-01 5.57313383e-01 -3.55305582e-01 5.99291861e-01 5.86071312e-01 -7.69334659e-03 -4.92269337e-01 -1.36922181e+00 -9.48041022e-01 -7.42549360e-01 -1.88727096e-01 6.37417912e-01 -1.00539207e+00 -6.23374641e-01 2.61892706e-01 -6.79739058e-01 -3.76518935e-01 -6.43473148e-01 3.83848190e-01 -1.00448573e+00 3.34151983e-01 -3.79192472e-01 -8.77721310e-01 -4.32670683e-01 -8.55050862e-01 8.64315808e-01 1.06402993e-01 -2.82481343e-01 -9.09844041e-01 1.12008020e-01 4.63305235e-01 2.13283986e-01 1.33015409e-01 5.75577080e-01 -9.70781684e-01 -1.02488384e-01 -3.05313408e-01 -2.51027435e-01 8.32698286e-01 6.28128126e-02 -2.85886943e-01 -1.16654587e+00 -3.87601852e-01 -1.43080950e-01 -8.62500548e-01 1.16573811e+00 1.61220908e-01 1.36097062e+00 -5.38820446e-01 -2.17608735e-01 7.14837372e-01 1.54860497e+00 -8.39613155e-02 3.98639619e-01 1.94683313e-01 5.91942728e-01 8.15559566e-01 4.52511698e-01 3.88612479e-01 3.88998270e-01 5.08930862e-01 4.74584490e-01 -1.62412554e-01 -8.85627195e-02 -3.08336258e-01 3.13235462e-01 6.26830935e-01 1.20396294e-01 -6.16112240e-02 -8.17213297e-01 6.08103633e-01 -2.04365182e+00 -1.03044319e+00 6.43343851e-02 2.43594766e+00 7.54586995e-01 -1.20259672e-01 2.72241503e-01 2.07093120e-01 9.28180158e-01 -1.78501103e-02 -7.42477298e-01 -1.94278970e-01 -6.00208808e-03 3.54452163e-01 6.21039391e-01 4.34134245e-01 -1.58601260e+00 9.65980172e-01 6.05414200e+00 8.18022728e-01 -1.11917865e+00 4.33159024e-01 8.63962173e-01 1.64491832e-02 -7.33076595e-03 1.72854930e-01 -5.17255247e-01 3.49365115e-01 7.72010326e-01 1.90644294e-01 3.93546522e-01 1.30444562e+00 -4.08142835e-01 5.18464856e-02 -1.26090407e+00 1.08964157e+00 2.36839667e-01 -1.28016639e+00 -4.96739835e-01 -3.81575129e-03 8.43661547e-01 2.48075515e-01 6.52368814e-02 2.32351243e-01 5.14658213e-01 -8.78111541e-01 8.26446593e-01 2.83175647e-01 1.09058619e+00 -6.44021034e-01 6.28054380e-01 4.22597110e-01 -9.20858860e-01 -2.87646025e-01 -3.81035328e-01 -3.74212936e-02 -3.58500332e-01 5.75064242e-01 -6.56819105e-01 1.99792251e-01 4.59616363e-01 1.00816000e+00 -5.23107231e-01 9.26358700e-01 -2.88203716e-01 7.22134948e-01 -4.07826513e-01 2.08161086e-01 1.69988975e-01 -1.21639468e-01 2.62532085e-01 1.36947250e+00 8.11288208e-02 -1.98908359e-01 4.93135095e-01 4.89211589e-01 -1.24734409e-01 1.41662091e-01 -6.04061484e-01 1.23910546e-01 3.53264391e-01 1.28070712e+00 -7.82293379e-01 -4.37236100e-01 -2.52801657e-01 1.16463554e+00 7.66196907e-01 2.34505549e-01 -5.52307844e-01 -1.23476975e-01 3.29203874e-01 1.82948470e-01 4.27257568e-01 1.62635371e-01 -3.82070810e-01 -1.33169866e+00 -1.30291041e-02 -7.92362750e-01 7.15723217e-01 -4.25970256e-01 -1.68547535e+00 3.64654243e-01 -1.56250402e-01 -1.60375047e+00 -5.08481026e-01 -7.21076488e-01 -5.12951851e-01 3.47208470e-01 -1.42025530e+00 -1.11513722e+00 -2.20513314e-01 6.34703100e-01 2.83930302e-01 -4.56300676e-01 1.21528971e+00 1.04173519e-01 -4.39274102e-01 7.99319565e-01 4.21709746e-01 1.40145853e-01 7.16617584e-01 -1.37049580e+00 -9.84432995e-02 5.41048110e-01 2.98439234e-01 2.03127742e-01 4.36982751e-01 -2.48506412e-01 -8.32292438e-01 -1.10282993e+00 8.92011523e-01 -2.23803997e-01 5.68194568e-01 -3.53785723e-01 -5.29759109e-01 8.61414015e-01 1.22637004e-01 5.51258028e-01 9.82562363e-01 1.85126945e-01 -8.66479397e-01 -1.60366386e-01 -1.26940918e+00 3.35910469e-01 1.08773744e+00 -6.04286492e-01 -3.19346994e-01 7.32247651e-01 3.14916849e-01 -1.26465648e-01 -7.76920378e-01 4.48943257e-01 4.96379733e-01 -1.01201963e+00 8.78249168e-01 -6.24355435e-01 6.52520180e-01 -1.34591579e-01 -3.66906345e-01 -1.16643584e+00 -2.93923020e-01 -5.51102102e-01 -1.82682276e-01 1.05523288e+00 5.32348394e-01 -7.48758376e-01 9.83026922e-01 2.36924723e-01 -1.45945519e-01 -1.09931803e+00 -9.42563772e-01 -9.53976631e-01 -4.41420004e-02 -2.78815955e-01 -1.70874931e-02 1.22884226e+00 1.75592095e-01 4.41033751e-01 -3.17616254e-01 -8.18526074e-02 1.09159207e+00 3.08106601e-01 6.05702162e-01 -1.34749484e+00 -4.68482852e-01 -4.62605923e-01 -7.47928798e-01 -7.92167902e-01 5.45110524e-01 -1.42275333e+00 -1.29090816e-01 -1.49792480e+00 4.09795761e-01 -6.41517937e-01 -5.32857776e-01 8.23394537e-01 2.45740920e-01 4.47363764e-01 3.00306398e-02 4.36047584e-01 -8.53580832e-01 4.86757219e-01 7.30412662e-01 -1.94492936e-01 2.66522728e-02 1.00628316e-01 -6.70925736e-01 9.09930646e-01 1.05971253e+00 -8.67467463e-01 -4.52231288e-01 -1.38610927e-02 -1.91452559e-02 -5.38330860e-02 3.67794305e-01 -1.01156461e+00 1.43036753e-01 1.53859183e-01 3.03276658e-01 -3.62428278e-01 3.44689220e-01 -8.63451898e-01 -1.40019268e-01 4.52301919e-01 -7.21652687e-01 -1.95255518e-01 -2.67055571e-01 6.03323638e-01 -2.56087363e-01 -6.27584517e-01 1.07592380e+00 -2.76379973e-01 -5.03753483e-01 1.65605411e-01 -3.00657451e-01 1.83517307e-01 1.08145750e+00 -2.42033869e-01 -1.92829043e-01 -4.63911206e-01 -8.99308324e-01 1.28813818e-01 4.97719646e-01 1.54176190e-01 4.67650414e-01 -1.24972892e+00 -7.39557385e-01 1.25886664e-01 3.58113706e-01 -4.54932041e-02 1.00444764e-01 6.11136317e-01 -3.41372579e-01 -1.04002640e-01 -1.21891648e-01 -7.51413405e-01 -1.16791928e+00 5.93404651e-01 4.39558268e-01 -4.67487097e-01 -6.27127349e-01 8.67805243e-01 1.60216555e-01 -9.81737614e-01 4.81527239e-01 2.88330466e-01 -2.84712054e-02 2.28627250e-01 2.22450927e-01 2.53715068e-01 -6.32668063e-02 -6.44474149e-01 -3.73343587e-01 5.21013975e-01 -1.34945020e-01 -2.19934404e-01 1.37066901e+00 1.65049300e-01 2.71002986e-02 6.99321449e-01 1.72525477e+00 -3.74835223e-01 -1.32341790e+00 -3.85790318e-01 -6.50677085e-02 -1.87401354e-01 1.08916081e-01 -8.29634368e-01 -1.22182107e+00 8.71497333e-01 9.05944169e-01 -5.15674539e-02 1.04026616e+00 1.74580202e-01 3.38493973e-01 6.35156274e-01 2.75812089e-01 -1.23106575e+00 2.75968373e-01 3.89812708e-01 7.77417541e-01 -1.43963480e+00 3.30095962e-02 -4.19205904e-01 -8.01676810e-01 9.80957389e-01 5.79179525e-01 -4.69760954e-01 8.81249785e-01 5.36548942e-02 1.43324107e-01 -8.79136696e-02 -7.23412514e-01 -3.06509048e-01 2.61429191e-01 6.22216284e-01 4.57290471e-01 2.04521269e-01 -1.18310992e-02 2.66179413e-01 -2.38109395e-01 -1.75529316e-01 2.35824421e-01 9.94940519e-01 -2.14249149e-01 -8.84383023e-01 4.26364094e-02 6.14280045e-01 -3.69960308e-01 -1.01551831e-01 -3.55647147e-01 4.00681764e-01 5.28087467e-02 8.65235031e-01 3.11950743e-01 -4.33819294e-01 -2.42873896e-02 4.25169945e-01 4.60546702e-01 -7.58506358e-01 -5.09932518e-01 4.22819937e-03 6.34213462e-02 -3.71705502e-01 -5.44457674e-01 -6.84912622e-01 -1.19040716e+00 -2.95740012e-02 -5.87536633e-01 -2.17375048e-02 6.94947004e-01 7.01248288e-01 4.88205135e-01 4.02826406e-02 1.02473259e+00 -8.60695064e-01 -8.04987609e-01 -1.03422928e+00 -5.42139113e-01 6.79263294e-01 2.28046492e-01 -7.18091965e-01 -6.03416264e-01 2.18074918e-01]
[9.309185028076172, 2.9721572399139404]
d227c901-e97e-4ac8-83f3-7743bfdea82f
federated-nearest-neighbor-machine
2302.12211
null
https://arxiv.org/abs/2302.12211v1
https://arxiv.org/pdf/2302.12211v1.pdf
Federated Nearest Neighbor Machine Translation
To protect user privacy and meet legal regulations, federated learning (FL) is attracting significant attention. Training neural machine translation (NMT) models with traditional FL algorithm (e.g., FedAvg) typically relies on multi-round model-based interactions. However, it is impractical and inefficient for machine translation tasks due to the vast communication overheads and heavy synchronization. In this paper, we propose a novel federated nearest neighbor (FedNN) machine translation framework that, instead of multi-round model-based interactions, leverages one-round memorization-based interaction to share knowledge across different clients to build low-overhead privacy-preserving systems. The whole approach equips the public NMT model trained on large-scale accessible data with a $k$-nearest-neighbor ($$kNN) classifier and integrates the external datastore constructed by private text data in all clients to form the final FL model. A two-phase datastore encryption strategy is introduced to achieve privacy-preserving during this process. Extensive experiments show that FedNN significantly reduces computational and communication costs compared with FedAvg, while maintaining promising performance in different FL settings.
['Enhong Chen', 'Tong Xu', 'Lemao Liu', 'Bingzhe Wu', 'Zhirui Zhang', 'Yichao Du']
2023-02-23
null
null
null
null
['nmt', 'memorization']
['computer-code', 'natural-language-processing']
[ 1.29458949e-01 -1.21115409e-01 -5.76632619e-01 -6.24813676e-01 -1.23965812e+00 -9.43246424e-01 5.22261858e-01 -6.33883178e-02 -5.44267833e-01 8.41680646e-01 -1.86991394e-01 -8.14242840e-01 9.58951861e-02 -8.00558805e-01 -1.05171382e+00 -7.85748541e-01 4.36812848e-01 2.55966306e-01 -3.81779730e-01 1.71031013e-01 1.55876726e-02 2.13104054e-01 -7.41855085e-01 5.40290773e-01 1.01257014e+00 1.10833991e+00 -1.06643163e-01 2.56457292e-02 -2.26834342e-01 4.35559064e-01 -1.91639334e-01 -1.32529664e+00 7.51762271e-01 -2.40743577e-01 -9.19876158e-01 -6.73257589e-01 3.19078743e-01 -6.49584353e-01 -4.63219970e-01 1.07241511e+00 3.40256780e-01 -3.25750202e-01 1.08971432e-01 -1.59577322e+00 -8.82253766e-01 8.21070015e-01 -3.69738430e-01 -4.53544945e-01 -6.09119870e-02 1.92117646e-01 7.82552779e-01 -7.92540908e-01 6.94582045e-01 8.89098346e-01 6.67777002e-01 7.37723708e-01 -1.38961935e+00 -1.13580620e+00 -2.10278824e-01 1.84038579e-01 -1.27941787e+00 -4.62300450e-01 4.51208264e-01 7.62315392e-02 8.00747097e-01 6.52589440e-01 3.92710388e-01 1.17617178e+00 4.46253359e-01 7.42503047e-01 1.32480812e+00 -2.25530118e-01 2.88997352e-01 6.25252306e-01 -3.50600109e-02 4.76298422e-01 4.86885875e-01 -9.11519155e-02 -6.48492992e-01 -8.48949492e-01 3.93504322e-01 5.54562151e-01 -4.73320812e-01 -7.60418117e-01 -1.11412036e+00 7.77188957e-01 3.00741941e-01 1.00792378e-01 -1.64381653e-01 1.08346730e-01 5.28711975e-01 7.68840075e-01 5.00034571e-01 -7.29587022e-03 -8.04840744e-01 -1.56568401e-02 -6.88864648e-01 1.39987499e-01 9.93463457e-01 1.25356650e+00 1.03760993e+00 -6.50857985e-01 3.43721211e-02 4.42895502e-01 -2.36365246e-03 5.58678329e-01 5.56697130e-01 -8.18275988e-01 1.00733697e+00 7.47840643e-01 4.13200259e-02 -8.66084993e-01 4.01538312e-01 -1.64599210e-01 -9.82657075e-01 -2.98143864e-01 1.06222428e-01 -4.49559897e-01 -3.13102603e-01 1.79631543e+00 5.45485020e-01 -1.25606760e-01 4.37344760e-01 6.15388215e-01 5.23538515e-02 6.32691205e-01 -2.26341575e-01 -3.03278595e-01 1.14468038e+00 -1.03134978e+00 -6.50232077e-01 1.29202992e-01 1.16622663e+00 -6.03775442e-01 7.85941660e-01 9.80943069e-02 -9.36605334e-01 2.15197876e-01 -8.05179775e-01 -3.34131539e-01 -5.45570612e-01 3.31175141e-02 7.64829993e-01 8.65639985e-01 -9.91995096e-01 4.59391117e-01 -7.87908852e-01 -2.17833579e-01 8.86832774e-01 9.04785097e-01 -9.25376296e-01 -3.09124827e-01 -1.05543435e+00 5.71199059e-01 9.85149443e-02 -1.52073503e-01 -7.08615065e-01 -6.23555481e-01 -5.44774055e-01 2.64566951e-02 2.48512626e-01 -6.97467327e-01 1.27794921e+00 -5.20911992e-01 -1.51438642e+00 7.20850885e-01 -1.84232548e-01 -8.52689385e-01 8.17117453e-01 -2.17878781e-02 -1.10574834e-01 4.26884368e-02 -2.66835332e-01 1.23669036e-01 7.89449930e-01 -1.05715990e+00 -6.22120738e-01 -9.66373384e-01 -7.41384551e-02 1.06980585e-01 -8.98394883e-01 1.27288789e-01 -2.64321119e-01 -4.45370942e-01 1.06672652e-01 -1.00591409e+00 -1.14455819e-01 1.96546435e-01 -4.48458016e-01 1.02669448e-02 1.41762030e+00 -6.62357926e-01 1.13265359e+00 -1.98978519e+00 -1.36114940e-01 2.88028508e-01 2.79687256e-01 5.08696735e-01 9.34848487e-02 7.19415963e-01 3.96170348e-01 2.20562786e-01 -2.97997177e-01 -7.88751245e-01 2.25805447e-01 1.49633184e-01 -5.27624786e-01 6.20384991e-01 -4.06857133e-01 1.21710050e+00 -4.04506236e-01 -5.86609840e-01 -5.82241900e-02 4.11744058e-01 -5.63467145e-01 1.83882400e-01 -2.13947847e-01 3.77077907e-01 -6.70875430e-01 6.86578810e-01 1.07176435e+00 -2.81964123e-01 5.08588195e-01 4.18588705e-02 -5.03923930e-02 1.75810158e-01 -6.17121398e-01 1.89168644e+00 -5.60782850e-01 1.99133635e-01 5.34490883e-01 -7.40790665e-01 7.89100766e-01 5.73645890e-01 2.89782226e-01 -4.92430657e-01 4.20288295e-01 4.86687213e-01 -6.34052575e-01 -2.32632235e-01 2.19835639e-01 2.58676887e-01 -2.17840761e-01 1.06445730e+00 -6.23072684e-02 3.38975251e-01 -7.96703458e-01 2.44363427e-01 1.15716088e+00 -1.72475446e-02 -1.16306506e-02 -2.89352564e-03 5.09284794e-01 -6.16043471e-02 7.48846233e-01 4.06015605e-01 -2.92750984e-01 1.88146651e-01 1.71604410e-01 -5.97492814e-01 -9.11081851e-01 -5.64451039e-01 2.45968744e-01 9.06664968e-01 1.28305063e-01 -6.02068067e-01 -1.18666387e+00 -1.08707666e+00 3.23048025e-01 5.05975008e-01 -3.93718630e-01 -3.91177267e-01 -6.65095389e-01 -3.83960396e-01 6.63141847e-01 -2.79916674e-02 8.82115364e-01 -4.55926061e-01 -2.90701807e-01 -1.97901968e-02 -4.15532589e-01 -8.33674431e-01 -9.35445249e-01 1.93243027e-01 -9.20917988e-01 -7.65247643e-01 -5.45095026e-01 -7.34553814e-01 7.19614983e-01 6.60626411e-01 2.71527886e-01 -3.24647456e-01 1.23383977e-01 2.82653811e-04 -8.71599838e-02 -2.90574193e-01 -3.76848757e-01 3.57349783e-01 2.17619300e-01 5.64345896e-01 6.89936936e-01 -6.23007298e-01 -8.15936208e-01 4.09051478e-01 -8.69491100e-01 5.34262508e-02 7.67667055e-01 8.49777758e-01 6.51398540e-01 -2.80019224e-01 2.63930857e-01 -1.18016565e+00 5.79666793e-01 -5.46589017e-01 -8.02043974e-01 7.23838151e-01 -1.02920496e+00 -4.40105051e-02 1.18692720e+00 -6.01705492e-01 -1.01021540e+00 5.91750108e-02 3.92092675e-01 -8.78974676e-01 3.06023240e-01 8.34154896e-03 -5.07242858e-01 -6.04866564e-01 3.22022885e-01 6.65588140e-01 2.43369222e-01 -6.34774923e-01 4.31527048e-01 1.27214086e+00 3.52750301e-01 -3.94203424e-01 9.60678637e-01 4.55400437e-01 -1.78230032e-01 1.58373207e-01 -2.11694911e-01 6.36551483e-03 -1.47319496e-01 4.29659426e-01 6.06266320e-01 -9.69618201e-01 -1.07789147e+00 6.48143351e-01 -1.21086967e+00 -4.26790789e-02 5.67722134e-03 5.23932159e-01 -5.24884641e-01 1.83386102e-01 -6.33593142e-01 -7.77749717e-01 -1.06098604e+00 -1.08211601e+00 7.62871385e-01 3.35945487e-02 2.57554740e-01 -4.58684057e-01 -2.93696336e-02 8.13359380e-01 7.34452307e-01 -2.47491021e-02 1.00641239e+00 -8.92327011e-01 -9.78311896e-01 -6.37313187e-01 -1.96059927e-01 3.39005709e-01 2.14905843e-01 -6.83334470e-01 -8.85929346e-01 -6.55579388e-01 4.26591039e-01 -5.49728632e-01 2.38343820e-01 -2.48856753e-01 1.38498640e+00 -1.07647836e+00 -5.39599478e-01 1.09085131e+00 1.42976868e+00 1.06509678e-01 2.33681023e-01 1.46555394e-01 6.66758716e-01 4.01776254e-01 4.65805441e-01 4.47018772e-01 5.39338648e-01 5.61744630e-01 3.76057506e-01 3.54112610e-02 5.91642976e-01 -5.99980354e-01 3.08796376e-01 6.90600753e-01 4.60252464e-01 -1.00436971e-01 -5.09175539e-01 1.87837824e-01 -1.95893037e+00 -7.75908232e-01 3.71988922e-01 2.34124041e+00 1.01608002e+00 -4.21352386e-01 -1.87712088e-01 -2.36429870e-01 7.08226562e-01 -1.97927821e-02 -8.74836564e-01 -6.41622484e-01 -7.54738152e-02 4.60617840e-02 1.05569172e+00 4.35971543e-02 -8.05615425e-01 8.82728398e-01 4.82329607e+00 8.91371369e-01 -1.26048005e+00 6.18276060e-01 8.53455126e-01 -3.74766439e-01 -4.47847694e-01 1.45457581e-01 -6.76072955e-01 6.12722039e-01 1.07474160e+00 -5.76639891e-01 9.49095488e-01 9.48570669e-01 -2.01361701e-01 5.95409453e-01 -1.21196270e+00 1.21648979e+00 -3.37476760e-01 -1.68185937e+00 1.13846488e-01 5.44105113e-01 6.60468519e-01 2.15670332e-01 3.28196794e-01 1.63484886e-01 4.35550213e-01 -6.96769834e-01 4.53694612e-01 1.79575354e-01 1.15659475e+00 -1.05438936e+00 5.25699914e-01 7.48050213e-01 -9.05823410e-01 -4.86806005e-01 -4.32182103e-01 4.11835819e-01 -2.71375149e-01 2.38763988e-01 -6.80389166e-01 8.97904813e-01 8.38545084e-01 1.61651015e-01 -6.39179274e-02 3.39879453e-01 4.01651889e-01 4.16871995e-01 -3.88134271e-01 -1.18203973e-02 2.40957618e-01 -5.84597588e-01 1.84519246e-01 7.60963202e-01 4.28523481e-01 1.96117014e-01 -2.21431211e-01 7.84929335e-01 -8.55823576e-01 3.24312359e-01 -9.66850817e-01 -1.48215979e-01 9.71104562e-01 1.15824008e+00 1.72372565e-01 -7.72763789e-02 -5.58962822e-01 1.37451899e+00 6.64403319e-01 1.42983750e-01 -5.51794767e-01 -4.81006742e-01 9.84191656e-01 -9.68921110e-02 2.56020993e-01 -1.47391438e-01 -2.11938709e-01 -1.42891860e+00 5.58604360e-01 -1.12762141e+00 4.76212412e-01 -1.57535821e-01 -1.22896290e+00 7.30176091e-01 -3.41442108e-01 -1.26748049e+00 7.37645850e-03 1.05336800e-01 -4.38703895e-01 1.06297255e+00 -1.39055490e+00 -1.69208205e+00 2.03611881e-01 1.18718004e+00 -1.49694413e-01 -1.30233169e-01 1.18434155e+00 2.72958606e-01 -7.87097752e-01 1.40266573e+00 7.67624080e-01 1.48297459e-01 8.08569491e-01 -6.66555941e-01 5.57212949e-01 7.56595790e-01 -9.19988751e-02 1.22442281e+00 6.65983036e-02 -6.01048410e-01 -2.27473617e+00 -1.60889649e+00 1.36707759e+00 -3.82780701e-01 2.11252257e-01 -1.02373445e+00 -7.15913415e-01 1.10784066e+00 1.31526530e-01 3.19995284e-01 9.50044692e-01 -4.10456449e-01 -8.03287446e-01 -5.99444747e-01 -1.92386770e+00 5.82032681e-01 1.01745319e+00 -9.29509163e-01 -3.48578170e-02 4.42348361e-01 1.26515162e+00 -2.40804538e-01 -8.69504631e-01 6.74307048e-02 7.04010546e-01 -6.65028811e-01 5.99937677e-01 -7.00389683e-01 -9.10664201e-02 4.42857146e-02 -4.34202820e-01 -8.52411389e-01 2.36254916e-01 -1.50067258e+00 -3.87412488e-01 1.29304600e+00 3.95639271e-01 -1.43230557e+00 1.18880522e+00 1.24653196e+00 4.47560072e-01 -8.68463039e-01 -1.41199172e+00 -7.37890363e-01 2.10735291e-01 -6.21286966e-02 1.41144228e+00 1.17020226e+00 2.85164148e-01 -1.87999502e-01 -6.56291068e-01 1.26827419e-01 9.15050626e-01 4.07026052e-01 1.15553856e+00 -7.58905172e-01 -2.45370999e-01 2.22711340e-01 -4.43274938e-02 -8.91230941e-01 1.92870602e-01 -1.26212871e+00 -4.77254182e-01 -9.19221520e-01 3.11693460e-01 -6.20566905e-01 -3.44921321e-01 7.95689881e-01 2.54037261e-01 1.36737645e-01 1.17533684e-01 5.00705600e-01 -4.75835115e-01 6.64270043e-01 9.84009326e-01 1.74965197e-03 3.28163393e-02 1.73541605e-01 -9.39439058e-01 1.84149016e-02 8.79173696e-01 -7.26183772e-01 -4.73794073e-01 -8.44533384e-01 -2.58498549e-01 2.28094488e-01 2.10915431e-01 -5.99836648e-01 6.03041828e-01 -2.44962186e-01 1.17320046e-01 -4.11402404e-01 2.14242995e-01 -1.38569117e+00 5.19081354e-01 5.52693188e-01 -4.16700274e-01 1.66852310e-01 -2.42217630e-01 6.83518112e-01 8.66535585e-03 4.71151680e-01 5.42000234e-01 -1.19155115e-02 2.94264108e-01 9.28236365e-01 1.22006759e-01 -5.30233741e-01 1.24121141e+00 -1.21271841e-01 -5.32403231e-01 -2.21194446e-01 -7.25074857e-02 3.20291579e-01 9.22193706e-01 2.36401737e-01 3.78566235e-01 -1.18319452e+00 -4.57231760e-01 4.75550443e-01 1.58297762e-01 -3.88422459e-02 1.66866407e-01 8.51755202e-01 -1.03410184e-01 7.15866268e-01 -4.90668267e-02 -2.37115175e-01 -1.44844365e+00 7.36596227e-01 1.50312692e-01 -3.53844941e-01 -5.97987175e-01 8.22199702e-01 -4.37796414e-02 -9.21766341e-01 4.14434105e-01 -1.01907812e-01 7.41167247e-01 -5.33556044e-01 5.50832510e-01 2.85191238e-01 2.50902772e-01 -4.79568034e-01 -2.31752530e-01 1.67752117e-01 -4.47792977e-01 -1.09692834e-01 1.22553885e+00 -4.67271447e-01 -4.00139958e-01 -2.70877540e-01 1.84913528e+00 -8.47292617e-02 -8.82511377e-01 -7.24596977e-01 -2.89651841e-01 -8.58081818e-01 1.67640820e-02 -7.23888338e-01 -1.16128206e+00 6.81403875e-01 5.98655283e-01 -3.44871193e-01 1.20687222e+00 -4.53873366e-01 1.65677965e+00 6.92148268e-01 1.14178288e+00 -8.31650019e-01 -7.73673773e-01 -7.65440390e-02 3.39849323e-01 -1.15478599e+00 -3.33711326e-01 -2.54931092e-01 -4.20079142e-01 7.45612204e-01 5.13629615e-01 5.02319634e-01 6.00072920e-01 1.15572333e-01 1.60359651e-01 2.36929163e-01 -1.00000131e+00 9.05836701e-01 -3.89764190e-01 4.99802232e-01 -3.29951584e-01 1.50328115e-01 -4.21214014e-01 8.92923772e-01 -7.17398450e-02 3.62203658e-01 -1.02249019e-01 1.39303195e+00 2.54839629e-01 -1.76599622e+00 -9.30971429e-02 5.41565239e-01 -7.40160525e-01 -5.84450131e-03 -5.41044533e-01 2.32086539e-01 8.67265165e-02 9.20243323e-01 -2.78983355e-01 -8.74181509e-01 -4.32383753e-02 2.33854607e-01 7.28601441e-02 -2.50841618e-01 -1.22113919e+00 -3.66195172e-01 -2.27130368e-01 -9.10068810e-01 1.32636532e-01 -6.06495619e-01 -9.42403853e-01 -9.98169363e-01 -5.17947257e-01 5.14017165e-01 1.01833987e+00 5.18867850e-01 1.02201641e+00 -6.75983131e-01 1.35529137e+00 -2.03924865e-01 -1.13111961e+00 -5.41118801e-01 -4.04660940e-01 1.55910909e-01 3.30638051e-01 1.21637069e-01 -2.33238786e-01 -2.30701774e-01]
[5.829962730407715, 6.505829334259033]
2684c411-996f-46f6-8e11-a5b6202654cf
efficiently-discovering-locally-exceptional
1709.07941
null
http://arxiv.org/abs/1709.07941v1
http://arxiv.org/pdf/1709.07941v1.pdf
Efficiently Discovering Locally Exceptional yet Globally Representative Subgroups
Subgroup discovery is a local pattern mining technique to find interpretable descriptions of sub-populations that stand out on a given target variable. That is, these sub-populations are exceptional with regard to the global distribution. In this paper we argue that in many applications, such as scientific discovery, subgroups are only useful if they are additionally representative of the global distribution with regard to a control variable. That is, when the distribution of this control variable is the same, or almost the same, as over the whole data. We formalise this objective function and give an efficient algorithm to compute its tight optimistic estimator for the case of a numeric target and a binary control variable. This enables us to use the branch-and-bound framework to efficiently discover the top-$k$ subgroups that are both exceptional as well as representative. Experimental evaluation on a wide range of datasets shows that with this algorithm we discover meaningful representative patterns and are up to orders of magnitude faster in terms of node evaluations as well as time.
['Janis Kalofolias', 'Mario Boley', 'Jilles Vreeken']
2017-09-22
null
null
null
null
['subgroup-discovery']
['methodology']
[ 1.45598143e-01 3.66296411e-01 -7.18020916e-01 -3.83836895e-01 -5.61931074e-01 -5.14935315e-01 3.76286983e-01 5.97028971e-01 4.30538319e-02 1.14289606e+00 4.45304215e-02 -3.81431609e-01 -6.65758073e-01 -9.33483899e-01 -7.49107122e-01 -9.58200157e-01 -4.79612440e-01 1.09545362e+00 2.41006881e-01 1.18242443e-01 3.54488164e-01 4.71202642e-01 -1.76411462e+00 1.62201971e-01 4.70181882e-01 1.02974892e+00 -3.59247327e-01 1.84255779e-01 -2.65640244e-02 4.53249723e-01 -7.52631366e-01 -8.69326591e-02 1.27631694e-01 -3.81341428e-01 -7.91653395e-01 9.16322544e-02 -6.00726008e-02 2.40535676e-01 4.88627762e-01 9.91172433e-01 1.47103414e-01 7.63324322e-03 7.18261421e-01 -1.48681343e+00 3.06583166e-01 9.88083780e-01 -1.09331012e+00 2.22364828e-01 3.48803699e-01 -2.56364524e-01 1.24839354e+00 -1.53474450e-01 9.87890303e-01 1.13875532e+00 4.43383485e-01 3.46077271e-02 -1.60010386e+00 -6.48385823e-01 3.95588964e-01 6.44011423e-02 -1.65927076e+00 -3.42903823e-01 5.15429378e-01 -1.71647444e-01 7.08586872e-01 8.72085392e-01 3.91451448e-01 6.04333460e-01 4.44366217e-01 3.91533345e-01 7.86887705e-01 -4.79893804e-01 7.22363234e-01 2.09376544e-01 2.73740411e-01 4.98000711e-01 7.56103039e-01 -2.61262745e-01 -7.73423851e-01 -8.80002677e-01 -9.64141488e-02 -2.53754944e-01 -1.97228357e-01 -5.23526371e-01 -1.05792105e+00 9.32598650e-01 -1.75223246e-01 2.56029874e-01 -3.99213791e-01 2.09481835e-01 3.27895582e-01 2.03094974e-01 5.82413912e-01 4.04488415e-01 -6.99387193e-01 -1.23289898e-01 -8.71703327e-01 4.68401223e-01 1.27074242e+00 9.25637007e-01 8.13783050e-01 -5.29877961e-01 6.12780824e-02 3.00542772e-01 2.28488836e-02 3.61755371e-01 2.53922939e-01 -7.67950714e-01 8.41158405e-02 9.26900685e-01 3.17890614e-01 -9.25479710e-01 -4.81471717e-01 -5.78771830e-01 -9.28114712e-01 1.45480216e-01 5.28730333e-01 -1.77503470e-02 -5.56390166e-01 1.95920944e+00 5.65563142e-01 -1.14889398e-01 -1.18210129e-01 3.53665411e-01 2.30328232e-01 6.22677445e-01 -4.20718864e-02 -8.94773245e-01 1.50340724e+00 -4.75962423e-02 -2.74207115e-01 -1.68517623e-02 6.90463305e-01 -2.80882925e-01 5.44447958e-01 5.67676425e-01 -9.54769015e-01 1.12211861e-01 -1.01400375e+00 6.16656899e-01 -4.58553135e-02 -3.61664504e-01 7.13873088e-01 6.74693823e-01 -7.11158276e-01 4.29554164e-01 -7.06186295e-01 -3.41368705e-01 1.87624156e-01 5.97022951e-01 -2.23866180e-01 2.23828867e-01 -7.10872293e-01 2.80717105e-01 5.43492794e-01 -4.94757056e-01 -5.37045300e-01 -1.06710088e+00 -2.69857943e-01 3.22964787e-01 6.05210543e-01 -6.82890713e-01 9.03723240e-01 -8.62526119e-01 -7.60849833e-01 7.61847198e-01 -6.42880440e-01 -5.07422447e-01 4.38764423e-01 5.96500993e-01 -2.44715944e-01 -6.03587218e-02 3.28790545e-01 -9.55004692e-02 4.31675911e-01 -1.06797707e+00 -1.10073388e+00 -7.11425424e-01 -1.83846787e-01 -1.27498150e-01 -4.56428863e-02 -6.35055155e-02 -2.15069190e-01 -1.15858085e-01 2.14609802e-01 -8.35601211e-01 -4.50085402e-01 -3.85403395e-01 -6.98386550e-01 -5.50235510e-01 5.81386209e-01 -4.82783206e-02 1.65557766e+00 -2.03299546e+00 -7.68174380e-02 1.19231689e+00 2.52577186e-01 -4.78202105e-01 6.46340311e-01 3.63598734e-01 -1.72941357e-01 5.02902269e-01 -3.71220499e-01 5.96809015e-02 8.90468210e-02 2.61413038e-01 -3.36669326e-01 8.63725424e-01 -2.38470927e-01 1.31605431e-01 -6.51963890e-01 -4.22639519e-01 -3.75880569e-01 -2.61209100e-01 -2.65587658e-01 -1.37769550e-01 -6.42696798e-01 2.84777451e-02 -7.21987963e-01 5.15006661e-01 5.50806046e-01 -2.86961943e-01 6.83676481e-01 2.40008801e-01 -1.26327172e-01 2.08724514e-02 -1.63283181e+00 7.80414581e-01 -1.75326143e-03 1.71274304e-01 5.06304093e-02 -1.10371649e+00 1.02344406e+00 8.36773142e-02 6.98370934e-01 -2.47249946e-01 2.60907877e-02 5.83618641e-01 -1.48090953e-02 2.62453109e-02 1.89648837e-01 -3.48459691e-01 -3.26213956e-01 8.49829316e-01 -5.64504087e-01 3.43878150e-01 4.19897884e-01 6.45055529e-03 1.38855577e+00 -5.32248199e-01 8.05604875e-01 -8.79064381e-01 5.12785017e-01 6.78272247e-02 9.31556880e-01 7.16369629e-01 1.55540243e-01 1.62355065e-01 1.35190320e+00 -5.31018913e-01 -8.48637760e-01 -7.36032844e-01 -1.80920750e-01 8.46269190e-01 1.96574450e-01 -5.98746479e-01 -3.64269435e-01 -4.92461145e-01 2.89684117e-01 7.55025327e-01 -8.22126865e-01 5.89265861e-02 -3.11920285e-01 -1.04115081e+00 1.45399556e-01 1.54512882e-01 1.60489231e-01 -7.04243839e-01 -9.47957456e-01 7.30245858e-02 9.21240672e-02 -5.62449992e-01 -9.09082815e-02 5.04982233e-01 -6.82010651e-01 -1.26007509e+00 -2.97139809e-02 -4.63473737e-01 6.96163118e-01 -2.19427541e-01 1.15560699e+00 -6.29318133e-02 -1.21748209e-01 -1.16245247e-01 -4.74857688e-02 -6.42227411e-01 -4.36379284e-01 1.51424423e-01 1.13238111e-01 6.06806353e-02 4.48150754e-01 -7.70563304e-01 -3.33368748e-01 3.70733887e-01 -5.94718337e-01 -3.53075355e-01 3.01512867e-01 4.71556455e-01 1.09321058e+00 7.83679545e-01 5.11311829e-01 -9.74179089e-01 5.84864438e-01 -9.52371180e-01 -8.23751271e-01 5.08995295e-01 -8.88009608e-01 4.37994272e-01 6.13262415e-01 -2.70301938e-01 -7.13748634e-01 1.97155684e-01 3.86860192e-01 -2.18403265e-01 1.49144894e-02 7.33017027e-01 -2.86800712e-01 4.17033106e-01 6.81279182e-01 1.48217678e-01 5.06437160e-02 -3.80107135e-01 -8.71269628e-02 6.28298044e-01 2.71281689e-01 -7.06126928e-01 4.47416127e-01 7.34395742e-01 7.27734983e-01 -8.46618652e-01 -3.21213901e-01 -6.05467379e-01 -4.70742024e-02 1.05811067e-01 1.72068015e-01 -3.99295270e-01 -1.13542807e+00 -1.33369133e-01 -6.28329992e-01 1.28012240e-01 -4.12589729e-01 1.80439278e-01 -6.32235587e-01 1.57041714e-01 2.47401923e-01 -1.00342917e+00 -4.72152412e-01 -1.01985693e+00 7.67640173e-01 8.26943517e-02 -7.79843628e-01 -7.03078687e-01 6.46524504e-02 -2.51221925e-01 -1.24170575e-02 7.48463333e-01 1.33685029e+00 -1.14676404e+00 -4.17779982e-01 -3.33331436e-01 2.59333551e-01 -4.11320060e-01 2.71829348e-02 2.62553900e-01 -4.57762867e-01 -3.25882792e-01 -5.63284643e-02 2.49102309e-01 5.48870564e-01 6.48080111e-01 1.32089341e+00 -6.06915593e-01 -1.03333068e+00 4.92514312e-01 1.20162678e+00 3.09102237e-01 2.38372430e-01 3.09743762e-01 -1.54943228e-01 7.08898008e-01 7.41479039e-01 8.42085361e-01 7.68805817e-02 8.64301860e-01 2.87639529e-01 3.90141129e-01 5.50917804e-01 1.99778844e-02 5.50073273e-02 -1.69178218e-01 1.39329061e-01 -2.57920295e-01 -9.39928472e-01 9.15208042e-01 -1.87901223e+00 -1.13595378e+00 -9.78295431e-02 2.47526503e+00 9.98208046e-01 3.95212203e-01 5.05332172e-01 2.60256976e-01 7.95124650e-01 1.62936803e-02 -6.92299783e-01 -8.55812252e-01 -2.25029066e-02 2.59065568e-01 7.83206463e-01 4.77096707e-01 -7.43580818e-01 2.96089351e-01 6.17153978e+00 1.05596268e+00 -9.97725368e-01 -3.25463861e-01 1.04092550e+00 -1.69210657e-01 -7.03104258e-01 9.91564989e-02 -9.28649962e-01 4.86548305e-01 1.14036989e+00 -8.31510723e-01 1.42374262e-01 1.07484639e+00 1.20313458e-01 -4.81713831e-01 -1.18802476e+00 5.06624758e-01 -4.34481949e-01 -1.36444998e+00 -1.85041264e-01 5.38351417e-01 6.19036555e-01 -3.22211266e-01 -2.31487185e-01 -3.02136898e-01 3.28201383e-01 -9.52609837e-01 6.02719665e-01 2.08258480e-01 6.98793113e-01 -1.37423611e+00 7.05864727e-01 4.92531955e-01 -9.41394687e-01 -8.06645676e-03 -2.90631622e-01 -3.91723514e-02 -1.67684719e-01 1.04868841e+00 -1.04504406e+00 5.33264756e-01 8.60771835e-01 2.49830008e-01 -5.85430004e-02 8.65018070e-01 1.27122462e-01 6.16885900e-01 -9.78173912e-01 -3.30978155e-01 9.49242711e-03 -6.93658218e-02 7.43573487e-01 9.58266735e-01 4.49890792e-01 3.21787335e-02 -4.33640331e-02 6.99120343e-01 -9.16342158e-03 1.79755449e-01 -4.11606133e-01 -2.81882733e-02 7.31808364e-01 8.82967234e-01 -9.51711118e-01 -2.46470436e-01 1.02130502e-01 8.65206867e-02 -3.74073312e-02 9.99518782e-02 -5.07474482e-01 -4.28587884e-01 1.10541487e+00 2.89794832e-01 5.36574543e-01 2.98489600e-01 -6.25494719e-01 -6.83368146e-01 1.80678189e-01 -8.38380814e-01 7.83421397e-01 7.15211630e-02 -9.50399041e-01 4.04871464e-01 3.94977957e-01 -8.94825637e-01 -7.50515282e-01 -2.43080258e-01 -6.06424868e-01 7.45513558e-01 -8.45872998e-01 -6.46257102e-01 9.94587541e-02 4.83894676e-01 -4.93759513e-02 -4.38623726e-02 7.57886350e-01 -3.63044858e-01 -3.76269132e-01 7.04460919e-01 2.92613357e-01 -5.95774233e-01 6.56110197e-02 -1.09685898e+00 -1.36136234e-01 7.94645727e-01 4.25002910e-02 7.52380133e-01 1.29595983e+00 -7.10336208e-01 -1.29451299e+00 -7.64608383e-01 1.09033585e+00 -1.71552137e-01 5.34103334e-01 -1.35992616e-01 -4.97149676e-01 5.73480844e-01 -1.78563967e-01 -1.00714453e-01 7.82050610e-01 5.42358875e-01 -1.74228400e-01 -5.08996665e-01 -1.38553023e+00 4.77321565e-01 8.37983251e-01 8.03878009e-02 -1.82320207e-01 2.71058291e-01 2.94121206e-01 -2.35259965e-01 -8.87175083e-01 5.31347930e-01 6.09534681e-01 -1.21923721e+00 6.70084953e-01 -5.71166515e-01 1.94376975e-01 -5.79332232e-01 -2.97064811e-01 -9.14558053e-01 -1.46063238e-01 -9.40941632e-01 -2.54148036e-01 1.18019676e+00 6.21454239e-01 -7.28649080e-01 9.52176273e-01 4.90273237e-01 2.72891521e-01 -1.22446239e+00 -1.42091453e+00 -7.75777340e-01 -1.97168916e-01 -2.21886843e-01 1.11510527e+00 8.15285504e-01 3.47280771e-01 1.66584760e-01 -1.37850121e-01 1.79561883e-01 7.30962515e-01 8.79452169e-01 7.31603801e-01 -1.45261383e+00 -2.55174518e-01 -4.53119904e-01 -5.92747986e-01 -4.46749866e-01 9.67123508e-02 -6.17866755e-01 -1.16690308e-01 -1.17648494e+00 4.37057167e-01 -6.13811493e-01 -1.18077181e-01 3.97233039e-01 1.71630993e-01 -9.42143425e-03 -5.36791146e-01 2.14856163e-01 -3.74670386e-01 -1.03418916e-01 4.83415067e-01 8.27881992e-02 -3.23716909e-01 4.40378487e-01 -1.21890819e+00 6.58914924e-01 9.57535028e-01 -6.28942251e-01 -4.44962621e-01 5.01188338e-01 5.84618211e-01 2.00080723e-02 -1.44455075e-01 -6.52537048e-01 7.84634352e-02 -6.70914054e-01 1.62126988e-01 -6.18605971e-01 -1.14228867e-01 -8.76684844e-01 8.41591477e-01 7.40788460e-01 -2.98976213e-01 -8.92414153e-02 -1.09584117e-02 6.64775491e-01 4.67216447e-02 -1.67711169e-01 7.25950956e-01 8.41798037e-02 -4.38189685e-01 2.46885996e-02 -1.86407045e-01 -7.54294768e-02 1.45335484e+00 -3.66887420e-01 -4.65466201e-01 -5.47610700e-01 -3.97600383e-01 3.54785681e-01 5.74092925e-01 -2.18216404e-01 2.68530369e-01 -8.31535399e-01 -6.98778152e-01 1.52816087e-01 3.10836911e-01 1.22776978e-01 -1.65377036e-01 9.46330786e-01 -4.90258902e-01 4.64192778e-01 4.28541116e-02 -6.51823759e-01 -1.49492013e+00 5.67324579e-01 2.70916373e-01 -5.07373929e-01 -5.29635847e-01 6.75859153e-01 1.69915155e-01 1.63490936e-01 1.12260632e-01 -4.23236877e-01 1.83187872e-01 2.83563435e-01 6.46259606e-01 5.35325825e-01 -1.35513857e-01 -3.81304055e-01 -7.87123621e-01 3.72467458e-01 1.55598566e-01 1.87095478e-01 1.44094563e+00 3.91546637e-03 -6.04115188e-01 3.42455626e-01 1.09632361e+00 4.88165170e-01 -7.21886039e-01 -3.37916426e-02 3.51950347e-01 -4.53227043e-01 -2.90862650e-01 -6.72533810e-01 -9.56620097e-01 1.14539266e-01 2.57995967e-02 6.36924088e-01 1.24425375e+00 5.31663597e-01 4.80964035e-02 1.95304185e-01 4.98122752e-01 -8.96836460e-01 -6.38206124e-01 4.06602025e-02 6.95001245e-01 -5.58871210e-01 5.65548480e-01 -4.70177442e-01 -3.96419823e-01 1.07174778e+00 1.95938110e-01 2.75589079e-02 7.20419943e-01 2.85062253e-01 -5.66112280e-01 -2.48076633e-01 -1.14897144e+00 1.28532395e-01 -8.02432746e-02 3.44263822e-01 2.29535978e-02 3.38403344e-01 -8.12223673e-01 6.91600204e-01 -3.49638611e-01 -1.39694363e-01 5.70837498e-01 8.19292665e-01 -6.18092179e-01 -1.05563414e+00 -3.60234529e-01 8.45359325e-01 -6.30686045e-01 2.90118307e-01 -5.19236028e-01 8.75000596e-01 1.54764012e-01 8.86842489e-01 2.51639456e-01 -2.67486662e-01 1.89809620e-01 6.94426671e-02 1.79265946e-01 -3.69318873e-01 -4.06529486e-01 5.31894714e-03 3.91428411e-01 -5.59299588e-01 -2.01480806e-01 -7.32095599e-01 -1.27835584e+00 -7.56858170e-01 -1.96991295e-01 5.65315247e-01 4.53793049e-01 8.61750662e-01 3.43493611e-01 7.50135481e-02 6.78647399e-01 -9.05570984e-02 -6.73602045e-01 -5.94080746e-01 -1.09367108e+00 6.68175612e-03 2.28548288e-01 -6.34048939e-01 -6.88688338e-01 -2.49288499e-01]
[7.679121971130371, 4.760691165924072]
fd974aa1-b129-4d93-a911-53ae2dcb46d8
clustering-aided-weakly-supervised-training
2203.13704
null
https://arxiv.org/abs/2203.13704v1
https://arxiv.org/pdf/2203.13704v1.pdf
Clustering Aided Weakly Supervised Training to Detect Anomalous Events in Surveillance Videos
Formulating learning systems for the detection of real-world anomalous events using only video-level labels is a challenging task mainly due to the presence of noisy labels as well as the rare occurrence of anomalous events in the training data. We propose a weakly supervised anomaly detection system which has multiple contributions including a random batch selection mechanism to reduce inter-batch correlation and a normalcy suppression block which learns to minimize anomaly scores over normal regions of a video by utilizing the overall information available in a training batch. In addition, a clustering loss block is proposed to mitigate the label noise and to improve the representation learning for the anomalous and normal regions. This block encourages the backbone network to produce two distinct feature clusters representing normal and anomalous events. Extensive analysis of the proposed approach is provided using three popular anomaly detection datasets including UCF-Crime, ShanghaiTech, and UCSD Ped2. The experiments demonstrate a superior anomaly detection capability of our approach.
['Seung-Ik Lee', 'Marcella Astrid', 'Arif Mahmood', 'Muhammad Zaigham Zaheer']
2022-03-25
null
null
null
null
['supervised-anomaly-detection']
['computer-vision']
[ 1.95384920e-01 -1.94568679e-01 6.62142038e-02 -5.62364221e-01 -5.66672385e-01 -1.67269737e-01 3.55863065e-01 5.29645741e-01 -2.75899619e-01 3.91449392e-01 1.21536113e-01 1.13881201e-01 -3.20140906e-02 -3.86128247e-01 -5.46756446e-01 -8.11827600e-01 -5.28431296e-01 7.99825415e-02 3.45230460e-01 -3.43036503e-02 2.96484679e-01 5.64994752e-01 -1.61812699e+00 4.60070342e-01 9.57650781e-01 1.46003449e+00 -3.71204942e-01 4.41982627e-01 -2.53602684e-01 1.34330976e+00 -6.59783483e-01 1.31457791e-01 5.07621884e-01 -4.33679402e-01 -4.68201458e-01 5.11035264e-01 4.56727624e-01 -4.07271683e-01 -4.62028861e-01 1.09352040e+00 2.66671389e-01 4.50716615e-01 5.63979626e-01 -1.67793155e+00 -6.89930841e-02 9.03887600e-02 -8.35497320e-01 8.24487448e-01 1.96596354e-01 9.34344679e-02 1.01253104e+00 -7.39497662e-01 3.34343582e-01 8.63938034e-01 4.48550075e-01 4.20703560e-01 -8.02120149e-01 -7.01100349e-01 6.26119733e-01 5.57957649e-01 -1.29523849e+00 -3.61478239e-01 9.35940146e-01 -4.24741030e-01 6.41542554e-01 1.12470686e-01 4.33004767e-01 1.19567668e+00 2.68899370e-03 9.27999675e-01 5.30005276e-01 -1.88554823e-01 3.54050905e-01 -1.99326426e-01 2.89964467e-01 7.17883110e-01 -2.28136741e-02 -2.22225100e-01 -5.91587663e-01 -5.00762343e-01 4.21025336e-01 4.33995605e-01 -1.22942671e-01 -3.85210603e-01 -8.49243879e-01 6.67702854e-01 1.04331434e-01 5.31557053e-02 -6.05738044e-01 -2.38198027e-01 9.97733235e-01 6.16649985e-01 8.77631068e-01 1.83480218e-01 -4.50442195e-01 -1.91156805e-01 -8.48665476e-01 1.62315339e-01 5.10058999e-01 7.40052938e-01 4.66851354e-01 5.20310700e-01 -2.44484231e-01 8.78552198e-01 2.10169435e-01 4.51015756e-02 5.22240043e-01 -9.23374474e-01 5.07094622e-01 8.60125124e-01 -2.41544265e-02 -1.30116093e+00 -2.96853781e-01 -3.83887082e-01 -7.92179227e-01 2.29459926e-01 4.09354180e-01 -1.15121454e-01 -8.80397975e-01 1.67602932e+00 2.60118306e-01 7.65778720e-01 -1.25118434e-01 7.43839741e-01 4.64283675e-01 6.74141824e-01 1.87765181e-01 -4.61074203e-01 7.13476121e-01 -7.51062870e-01 -7.52157569e-01 -3.42359953e-02 9.85074759e-01 -3.52013588e-01 8.99062276e-01 5.60662508e-01 -8.04374754e-01 -2.69857615e-01 -7.82790959e-01 4.59141344e-01 -3.04368019e-01 -7.76022719e-03 4.18428063e-01 1.75781429e-01 -5.72626412e-01 4.70540255e-01 -9.54321444e-01 -3.41228306e-01 7.61004031e-01 1.32611364e-01 -3.78314316e-01 -1.43902838e-01 -8.43439877e-01 2.24355549e-01 5.53574204e-01 5.85807562e-02 -8.79296362e-01 -4.57741559e-01 -9.08872366e-01 5.79452924e-02 6.26084089e-01 1.86029986e-01 8.47264290e-01 -1.40709543e+00 -8.25368106e-01 6.99655831e-01 -1.13639779e-01 -6.80085540e-01 5.77736080e-01 -3.95919621e-01 -8.87871027e-01 3.84394735e-01 1.50960445e-01 1.05464630e-01 9.55812097e-01 -8.54287446e-01 -9.47098136e-01 -3.39677662e-01 -4.15135711e-01 1.95903271e-01 -5.24991274e-01 8.74313414e-02 -3.36576104e-01 -1.08964813e+00 3.67652565e-01 -6.39138520e-01 -3.10703009e-01 -2.06318617e-01 -3.59135985e-01 -3.15308362e-01 1.26406372e+00 -6.78446293e-01 1.42578650e+00 -2.55592275e+00 -3.21668148e-01 7.31295943e-01 1.97899461e-01 1.54998019e-01 -2.21235782e-01 4.91293184e-02 -4.28930193e-01 -2.65054584e-01 -1.52185306e-01 -1.36594445e-01 -4.92431730e-01 4.10614014e-01 -4.54765320e-01 6.99835658e-01 4.07645762e-01 1.62593588e-01 -8.96682084e-01 -4.39506829e-01 9.35147181e-02 -1.32906303e-01 -5.29680073e-01 4.40834701e-01 -1.34413823e-01 6.03423238e-01 -4.15178359e-01 8.26865315e-01 4.03874397e-01 -6.02669418e-02 -3.07476282e-01 1.64491236e-01 2.00204253e-01 -3.25911343e-02 -1.35376370e+00 1.45688522e+00 2.27519572e-01 2.70750612e-01 -1.72042958e-02 -1.34480357e+00 8.82344246e-01 4.55593735e-01 9.71271574e-01 -5.74034154e-01 4.05779816e-02 1.14680529e-01 -1.31045684e-01 -6.59189880e-01 3.17836255e-01 2.69674182e-01 9.46521759e-02 4.29595947e-01 1.03413880e-01 7.71843016e-01 4.75968391e-01 5.31425774e-01 1.48326766e+00 -9.80589986e-02 5.82270063e-02 4.65542786e-02 7.48462558e-01 -1.42945811e-01 1.19772804e+00 8.62496257e-01 -5.38433075e-01 6.04096293e-01 7.19711781e-01 -8.01479042e-01 -9.07067001e-01 -1.01227295e+00 9.99577045e-02 1.23768735e+00 1.24428071e-01 -5.44094265e-01 -4.31546569e-01 -1.31655669e+00 -1.76528171e-01 5.80294073e-01 -5.39986968e-01 -5.13688326e-01 -5.36130250e-01 -8.37563455e-01 3.80058289e-01 5.26502848e-01 3.78514320e-01 -1.15650010e+00 -1.48303971e-01 -4.08433191e-02 -3.50119591e-01 -1.11708403e+00 -4.35761988e-01 2.89686441e-01 -8.95780981e-01 -1.14671075e+00 -1.98802147e-02 -6.04870319e-01 9.19121206e-01 1.07557788e-01 9.08591986e-01 1.57857418e-01 -2.37175137e-01 1.73309535e-01 -5.64519167e-01 -2.66458124e-01 -3.15353036e-01 -2.07726926e-01 3.39658171e-01 5.43714166e-01 7.40753531e-01 -4.70222056e-01 -3.76612067e-01 3.71350616e-01 -9.46381271e-01 -5.78071713e-01 3.63011092e-01 8.14098060e-01 6.91076159e-01 3.18611801e-01 6.71653807e-01 -9.12584066e-01 3.09134543e-01 -1.07922757e+00 -4.05903935e-01 -1.50449604e-01 -5.88531315e-01 -3.31997782e-01 8.15654933e-01 -3.94598275e-01 -1.09959733e+00 8.50476176e-02 -5.65623268e-02 -5.99419177e-01 -7.32828200e-01 1.35177433e-01 -3.11023910e-02 2.40761116e-01 5.20575762e-01 2.24622771e-01 -3.76510769e-02 -4.31328505e-01 -1.42509595e-01 4.56096321e-01 6.88857377e-01 -4.45085019e-01 6.57315433e-01 5.49555361e-01 -1.33549109e-01 -8.96642208e-01 -1.04114568e+00 -1.00104666e+00 -5.10295272e-01 -2.67756999e-01 6.27768576e-01 -1.02254450e+00 -2.08341420e-01 7.04749465e-01 -6.69203401e-01 1.10650420e-01 -6.51847839e-01 5.19985378e-01 -3.91069055e-01 6.23212934e-01 -7.05480933e-01 -7.88963675e-01 -1.04933441e-01 -7.65312374e-01 6.63640857e-01 1.28841549e-01 -2.29753554e-01 -7.62523174e-01 7.27212057e-02 1.13513410e-01 1.30871683e-01 4.73272264e-01 8.25454235e-01 -1.44288325e+00 -3.14123273e-01 -5.86114764e-01 -1.84862480e-01 7.09851146e-01 2.27550015e-01 3.02401464e-02 -8.13554823e-01 -3.95172238e-01 2.51734816e-02 -3.50052714e-01 8.78745794e-01 2.89831728e-01 1.61802530e+00 -2.98981249e-01 -7.59122074e-02 6.29829168e-01 1.09463429e+00 2.82721967e-01 4.68316108e-01 4.68877167e-01 9.18345690e-01 4.74492073e-01 8.13134789e-01 8.33095253e-01 -9.39708948e-02 3.52465719e-01 5.14755905e-01 -1.16987020e-01 3.78629744e-01 8.93625319e-02 5.32483757e-01 7.10903466e-01 4.62833792e-02 -2.64082372e-01 -8.66505921e-01 7.00028181e-01 -2.12792754e+00 -1.24650013e+00 -2.26433143e-01 2.22788167e+00 3.53924215e-01 3.20707053e-01 3.68799061e-01 4.22480434e-01 7.33467519e-01 2.68481970e-01 -5.13832808e-01 -3.30417275e-01 -3.79885197e-01 -2.57956177e-01 1.72103792e-01 -3.75889195e-03 -1.49562407e+00 6.41430676e-01 6.06502676e+00 7.56294847e-01 -8.78158152e-01 -4.54631783e-02 8.14479530e-01 -4.76612329e-01 4.18477446e-01 -3.43744695e-01 -3.93348336e-01 7.57435560e-01 8.31245184e-01 2.54482359e-01 -5.53870481e-03 1.00285280e+00 4.32702601e-01 -4.67928126e-03 -8.98380756e-01 8.85185599e-01 2.46483296e-01 -8.14270079e-01 1.91643208e-01 -1.92792952e-01 7.47402370e-01 2.62313217e-01 -1.18528739e-01 2.70557642e-01 9.92107987e-02 -6.40744865e-01 4.54411447e-01 5.03640532e-01 2.81671286e-01 -1.00726724e+00 8.77375782e-01 2.82609671e-01 -9.72242594e-01 -5.30696809e-01 -1.33810222e-01 -1.27307251e-01 -4.18705270e-02 8.14044118e-01 -5.80920398e-01 2.85169572e-01 1.05668616e+00 1.04388666e+00 -6.92107439e-01 1.00322497e+00 -2.11135927e-03 9.71792877e-01 -4.77850884e-01 7.22632349e-01 4.13603842e-01 -2.96372086e-01 8.33918095e-01 1.11204779e+00 1.63141847e-01 -1.10460907e-01 6.05030656e-01 2.80154675e-01 -2.88008302e-02 3.15978020e-01 -6.37386024e-01 2.01973394e-01 2.94070452e-01 1.18324840e+00 -6.97905242e-01 -2.81815320e-01 -7.21107543e-01 1.16372454e+00 2.39437222e-01 4.56547469e-01 -8.06481779e-01 -3.39523971e-01 8.11591864e-01 5.08551449e-02 1.61140725e-01 4.21899483e-02 -1.80915326e-01 -1.36503732e+00 2.99202979e-01 -8.84859979e-01 9.58910048e-01 -2.84776598e-01 -1.54455364e+00 4.94723946e-01 -1.76135913e-01 -1.56213212e+00 -2.82744467e-01 -2.47914165e-01 -1.01468050e+00 4.14998293e-01 -1.26981974e+00 -6.71908259e-01 -5.00156045e-01 9.86388981e-01 6.27929747e-01 -7.36225665e-01 5.34215868e-01 5.75838268e-01 -9.93078768e-01 5.93460858e-01 2.00055972e-01 4.24622744e-01 8.73656094e-01 -1.30149186e+00 1.17924929e-01 1.38429904e+00 3.60812694e-02 1.17340563e-02 5.14270067e-01 -8.21406841e-01 -5.30043304e-01 -1.44853830e+00 5.41352153e-01 -2.76906341e-01 6.63392186e-01 -1.25274435e-01 -1.42747033e+00 7.81434774e-01 -3.49128187e-01 5.86931646e-01 7.78832078e-01 -2.30293386e-02 -3.96146208e-01 -1.41388386e-01 -1.27986610e+00 3.82299602e-01 9.73813057e-01 -3.08511883e-01 -3.83810073e-01 4.11584347e-01 2.28333920e-01 -3.57560486e-01 -4.15239304e-01 3.94230515e-01 1.60240665e-01 -9.53165531e-01 6.38844967e-01 -9.46650088e-01 2.43565485e-01 -4.58048671e-01 1.09276911e-02 -1.23707855e+00 -2.67924964e-01 -4.01983351e-01 -6.34651363e-01 1.38011611e+00 1.28803834e-01 -4.03755754e-01 9.54553604e-01 4.09006596e-01 -2.12647900e-01 -7.25693583e-01 -9.60291028e-01 -7.68499494e-01 -5.89447916e-01 -5.98406732e-01 2.51681805e-01 1.15233862e+00 -1.36548221e-01 -4.32759859e-02 -6.60781741e-01 4.78315592e-01 7.03341424e-01 -3.77646506e-01 5.29804349e-01 -1.31326580e+00 -2.34402679e-02 -6.09086752e-02 -8.10858965e-01 -6.31502390e-01 2.32994080e-01 -6.69008017e-01 1.55840054e-01 -6.91464901e-01 7.38997981e-02 -3.98106456e-01 -9.06732380e-01 5.44027865e-01 -2.27877960e-01 3.44957083e-01 -2.98177451e-01 4.62137371e-01 -1.15005767e+00 4.98815298e-01 5.27136624e-01 1.47411898e-01 -2.82332867e-01 1.97952762e-01 -3.05595428e-01 1.20570219e+00 8.33656251e-01 -6.62848294e-01 -4.09652352e-01 -1.78350106e-01 -2.34017484e-02 -3.02606225e-01 2.46435940e-01 -1.17177284e+00 4.46376279e-02 -1.30978584e-01 5.36992133e-01 -5.81246436e-01 -9.34919715e-03 -1.02633119e+00 -4.82023388e-01 1.08136535e-01 -5.18034756e-01 2.62744874e-01 -7.77861401e-02 1.02021861e+00 -4.90849197e-01 6.58179894e-02 9.28150773e-01 4.29150015e-02 -1.04195511e+00 5.79909384e-01 -5.85089505e-01 2.92859226e-01 1.35206068e+00 -2.04315394e-01 -1.31798491e-01 -4.81665015e-01 -8.13641310e-01 5.91497242e-01 2.88833290e-01 5.88168085e-01 6.39219820e-01 -1.44438159e+00 -7.01364398e-01 5.60718238e-01 5.29562891e-01 -7.68972337e-02 3.74717265e-01 8.80899727e-01 -4.30289924e-01 -2.03371987e-01 -3.16279680e-01 -5.88681400e-01 -1.32009280e+00 4.59307283e-01 1.27964020e-01 -2.38462955e-01 -8.11403394e-01 7.11117387e-01 9.50564519e-02 -2.76039746e-02 7.17329264e-01 7.84444734e-02 -2.64408857e-01 -1.22410301e-02 7.32541800e-01 6.66732132e-01 8.48637074e-02 -8.61397922e-01 -3.68011057e-01 2.48908089e-03 -5.98519504e-01 5.45274794e-01 1.17935777e+00 -3.38454425e-01 -1.35499891e-02 5.25112271e-01 9.66807783e-01 -2.73164481e-01 -1.31576788e+00 -4.92346615e-01 2.92313397e-01 -5.94706655e-01 -1.07894996e-02 -4.90363002e-01 -1.25285506e+00 4.58278120e-01 8.15347672e-01 2.93099642e-01 1.40287614e+00 -9.25950482e-02 1.00422525e+00 4.40074474e-01 -2.88829684e-01 -1.46418560e+00 3.71258944e-01 6.35764897e-01 3.67757738e-01 -1.50476110e+00 -2.05921665e-01 -1.78764552e-01 -7.79463470e-01 8.04701686e-01 1.00306785e+00 -3.34286511e-01 5.17319381e-01 5.97369485e-02 1.96569756e-01 -2.11714178e-01 -7.01677740e-01 4.83837761e-02 2.30478898e-01 5.10576308e-01 1.71042427e-01 -4.33160961e-01 -4.22014222e-02 4.32840496e-01 5.37812769e-01 -2.93095857e-01 5.81838429e-01 1.10481322e+00 -4.32201833e-01 -6.25513852e-01 -3.25848758e-01 9.92355764e-01 -1.11533058e+00 1.70635611e-01 -2.07653165e-01 2.70937771e-01 3.40315759e-01 9.02838588e-01 5.00413895e-01 -3.08628768e-01 2.92807698e-01 4.12415057e-01 -2.72037864e-01 -5.93502879e-01 -4.01046962e-01 -2.11507399e-02 -1.34632558e-01 -1.03559363e+00 -2.63745993e-01 -7.11253405e-01 -1.20282888e+00 -7.34468549e-02 -1.91565588e-01 2.75421679e-01 1.69528291e-01 1.10252893e+00 4.40446258e-01 4.91101176e-01 8.71351421e-01 -4.73588675e-01 -3.89568329e-01 -7.80439317e-01 -9.85417604e-01 1.11083043e+00 5.76302767e-01 -6.18967831e-01 -7.01065004e-01 1.54384017e-01]
[7.831862926483154, 1.6581337451934814]
5e0b7369-07c7-4d3e-b3c0-ba9c66775693
outlier-robust-extreme-learning-machine-for
1905.09368
null
https://arxiv.org/abs/1905.09368v2
https://arxiv.org/pdf/1905.09368v2.pdf
Outlier Robust Extreme Learning Machine for Multi-Target Regression
The popularity of algorithms based on Extreme Learning Machine (ELM), which can be used to train Single Layer Feedforward Neural Networks (SLFN), has increased in the past years. They have been successfully applied to a wide range of classification and regression tasks. The most commonly used methods are the ones based on minimizing the $\ell_2$ norm of the error, which is not suitable to deal with outliers, essentially in regression tasks. The use of $\ell_1$ norm was proposed in Outlier Robust ELM (OR-ELM), which is defined to one-dimensional outputs. In this paper, we generalize OR-ELM to deal with multi-target regression problems, using the error $\ell_{2,1}$ norm and the Elastic Net theory, which can result in a more sparse network, resulting in our method, Generalized Outlier Robust ELM (GOR-ELM). We use Alternating Direction Method of Multipliers (ADMM) to solve the resulting optimization problem. An incremental version of GOR-ELM is also proposed. We chose 15 public real-world multi-target regression datasets to test our methods. Our conducted experiments show that they are statistically better than other ELM-based techniques, when considering data contaminated with outliers, and equivalent to them, otherwise.
['Bruno Légora Souza da Silva', 'Patrick Marques Ciarelli', 'Fernando Kentaro Inaba', 'Evandro Ottoni Teatini Salles']
2019-05-22
null
null
null
null
['multi-target-regression']
['miscellaneous']
[-2.29494244e-01 -1.28543496e-01 9.00839344e-02 -3.32730711e-01 -3.03703874e-01 2.05121294e-01 1.11259468e-01 1.23510838e-01 -6.99101269e-01 9.13275301e-01 -2.75353253e-01 -7.10563213e-02 -4.08156067e-01 -6.70315504e-01 -8.70445549e-01 -7.97504008e-01 -1.88563108e-01 3.16483676e-01 -1.03395857e-01 -2.03293219e-01 3.78624141e-01 5.08885145e-01 -1.57529843e+00 5.89369312e-02 1.11078167e+00 1.21996057e+00 -2.52250850e-01 9.03926119e-02 -3.52128714e-01 7.86473572e-01 -3.35767686e-01 -3.04381669e-01 4.48242098e-01 -2.11234421e-01 -1.50395721e-01 -3.28642726e-01 3.91793191e-01 1.42007634e-01 2.38779690e-02 1.11449075e+00 6.57417715e-01 4.63720202e-01 9.69993770e-01 -1.50115871e+00 -8.14615309e-01 5.54805815e-01 -8.83101165e-01 1.17791668e-01 -2.57273018e-02 -5.26971780e-02 7.62577772e-01 -1.36510289e+00 3.08364511e-01 1.25201869e+00 1.19992268e+00 2.82049268e-01 -1.02376556e+00 -7.35166967e-01 3.04943144e-01 3.21809351e-01 -1.50493538e+00 -1.92199387e-02 9.21119452e-01 -4.84655648e-01 9.45742488e-01 1.63952112e-01 2.56634265e-01 9.18731868e-01 4.65441704e-01 7.75511026e-01 1.12652242e+00 -4.95668560e-01 4.26596224e-01 1.06533490e-01 7.88551793e-02 5.88928759e-01 5.86805344e-01 -7.46858446e-03 -4.05188054e-01 -2.02237040e-01 2.86889344e-01 4.00564313e-01 -1.55830801e-01 -1.75296038e-01 -6.73111916e-01 1.06232309e+00 4.57859039e-01 4.98302162e-01 -5.72673142e-01 4.85384986e-02 4.02125180e-01 7.30753005e-01 8.14637125e-01 3.19843084e-01 -3.80323917e-01 2.63242811e-01 -1.16649616e+00 7.38316923e-02 6.48239136e-01 5.99767268e-01 5.67312539e-01 4.84381586e-01 -5.66353695e-03 1.14440572e+00 7.24727452e-01 1.23931684e-01 8.38793576e-01 -5.36581457e-01 5.38452864e-01 1.10759377e+00 -1.35292888e-01 -1.25488627e+00 -5.91593683e-01 -5.46749830e-01 -1.46725535e+00 6.17258012e-01 3.32023382e-01 -3.08097661e-01 -8.34901989e-01 1.66964006e+00 1.44108325e-01 4.87955660e-01 1.54057994e-01 7.55450428e-01 5.56231558e-01 4.87817645e-01 -1.27402738e-01 -4.29194033e-01 5.64297616e-01 -7.81550944e-01 -6.32069230e-01 -8.67229700e-02 6.33004606e-01 -6.58025682e-01 8.95278633e-01 6.23000085e-01 -8.82761300e-01 -4.79552239e-01 -1.01856911e+00 5.05294502e-01 -7.73515105e-01 -8.47438052e-02 3.21371764e-01 4.96693462e-01 -8.54201972e-01 1.03600097e+00 -6.08950734e-01 -2.01847460e-02 5.55196822e-01 4.67355102e-01 -3.34160864e-01 8.85922369e-03 -1.10069072e+00 8.82373750e-01 6.00141525e-01 5.64777851e-01 -2.73077846e-01 -4.99856085e-01 -6.18078530e-01 -1.68770954e-01 2.77327478e-01 -2.59818226e-01 5.32777965e-01 -1.10132039e+00 -1.37871122e+00 5.63436925e-01 2.04456925e-01 -4.37162250e-01 8.06247830e-01 -2.43218720e-01 -5.30564427e-01 -4.55060422e-01 -2.10269928e-01 3.13481420e-01 1.03756809e+00 -9.97527480e-01 -3.23998660e-01 -4.42448735e-01 -7.04526365e-01 -1.17480360e-01 -4.79929954e-01 -5.96439429e-02 1.45275772e-01 -8.65060925e-01 4.90560234e-01 -8.85333478e-01 -3.50835919e-01 -6.31949455e-02 -3.77465934e-01 -2.84715086e-01 9.94055331e-01 -6.60869718e-01 1.27732301e+00 -2.19079542e+00 1.72221377e-01 6.00039124e-01 1.22453779e-01 4.07418579e-01 1.13293193e-02 2.69107133e-01 -3.90224397e-01 7.75567293e-02 -5.21934986e-01 -4.90905195e-01 6.78911507e-02 1.77836940e-01 -1.67186916e-01 7.04437494e-01 9.58173200e-02 6.06768250e-01 -5.90379596e-01 -3.07075948e-01 8.88649076e-02 4.48002785e-01 -3.82517070e-01 4.89853360e-02 -2.55896658e-01 3.79211724e-01 -2.48358220e-01 8.06733787e-01 7.27789700e-01 1.80790320e-01 -5.19212008e-01 6.37682527e-02 -1.59952164e-01 -8.74994516e-01 -1.89008439e+00 1.25467575e+00 -3.38363886e-01 6.34189427e-01 -2.05059901e-01 -1.40345085e+00 1.48554158e+00 3.17944169e-01 7.50799477e-01 -4.70130086e-01 2.05442786e-01 7.38946795e-01 -8.26027393e-02 -5.60351968e-01 2.71291602e-02 -3.58715951e-02 1.28730983e-01 1.87700599e-01 -1.31770293e-03 4.69174147e-01 2.75571853e-01 -3.12674880e-01 8.55695188e-01 4.32981700e-02 2.00832620e-01 -2.66143590e-01 7.07561672e-01 -2.55174488e-01 8.09024274e-01 8.38989258e-01 -6.45077899e-02 5.38796246e-01 4.42273676e-01 -7.34893799e-01 -9.53766048e-01 -9.34319973e-01 -3.23783070e-01 7.58384645e-01 -2.38127112e-01 8.35648328e-02 -3.19057912e-01 -5.09861708e-01 2.95730472e-01 7.38681436e-01 -3.55952531e-01 -4.33628798e-01 -5.15724599e-01 -1.23589075e+00 4.77251023e-01 3.85121256e-01 6.42479777e-01 -1.21079254e+00 -3.38361919e-01 3.72058034e-01 1.53524056e-01 -4.90155160e-01 -2.39841156e-02 2.99481332e-01 -1.14734733e+00 -9.65073705e-01 -9.03515935e-01 -5.79034030e-01 7.18414664e-01 -6.23359501e-01 9.57588434e-01 2.11658344e-01 -4.02569503e-01 -9.29474533e-02 -3.84562641e-01 -7.03265846e-01 -2.18278378e-01 -1.30784854e-01 5.10304749e-01 4.45489734e-01 4.88969624e-01 -7.64678895e-01 -2.95255691e-01 2.52368808e-01 -9.94369447e-01 -4.65467751e-01 7.33038425e-01 9.61017728e-01 7.75238097e-01 1.62215009e-01 8.81726265e-01 -1.06386197e+00 7.93290615e-01 -8.41066003e-01 -6.60598874e-01 2.43743837e-01 -9.65708673e-01 2.30881080e-01 1.24438787e+00 -7.45574713e-01 -4.98889089e-01 -1.87741026e-01 -3.09335768e-01 -9.28131819e-01 -1.22778416e-01 6.28055751e-01 9.47719514e-02 -3.31605941e-01 6.45185709e-01 -2.48637255e-02 -2.60784358e-01 -6.41790926e-01 -3.88008244e-02 8.71480048e-01 1.66529715e-01 -3.01381081e-01 6.65740967e-01 1.81348965e-01 3.63817662e-01 -7.88742006e-01 -7.08888710e-01 -4.13062066e-01 -4.31061655e-01 -1.81304440e-01 4.09826219e-01 -4.70777661e-01 -6.37473941e-01 8.76197696e-01 -7.05887198e-01 -7.77121261e-02 -4.61773008e-01 7.67459691e-01 -5.91928065e-01 1.42954424e-01 -4.22797561e-01 -1.10117364e+00 -4.73052949e-01 -1.06026542e+00 3.07409942e-01 1.99275523e-01 1.02276519e-01 -1.10041118e+00 1.39176667e-01 -1.22799791e-01 7.60752857e-01 9.19835448e-01 8.96376848e-01 -9.71604943e-01 9.37383100e-02 -3.28305751e-01 8.16018656e-02 7.09565938e-01 -1.50511175e-01 2.23381609e-01 -8.22184026e-01 -5.26163399e-01 2.73903072e-01 -2.29726791e-01 1.06408548e+00 4.13775682e-01 1.23783684e+00 -3.98084372e-01 -1.93579420e-01 8.00638497e-01 1.80813026e+00 1.13178685e-01 6.99233294e-01 5.55320740e-01 5.27692914e-01 3.26501966e-01 5.49019337e-01 5.14914691e-01 -2.54431129e-01 3.56159449e-01 4.99619216e-01 -3.02436143e-01 3.49417001e-01 2.59520020e-02 4.84993756e-01 9.45946336e-01 -5.22926338e-02 -9.56443027e-02 -9.18778181e-01 3.18928629e-01 -2.16197896e+00 -7.97630072e-01 -2.65868455e-01 2.19546223e+00 4.76162791e-01 3.43636751e-01 -9.05430391e-02 4.05657411e-01 7.34945893e-01 -1.34070978e-01 -7.69892871e-01 -6.88556552e-01 -3.12031835e-01 3.52949917e-01 5.36402881e-01 2.25570902e-01 -9.07142580e-01 5.13550580e-01 5.41461897e+00 8.19590092e-01 -1.30809700e+00 3.43536437e-02 5.47807634e-01 -1.39287859e-01 1.70333207e-01 -2.58615375e-01 -7.26686001e-01 8.83814514e-01 9.84829664e-01 1.36839122e-01 2.14167625e-01 1.02677453e+00 2.95175999e-01 -3.86318099e-03 -9.73717749e-01 1.06811333e+00 1.57519624e-01 -7.60771453e-01 -9.95328128e-02 -3.04537177e-01 9.93730724e-01 1.88006401e-01 1.81901231e-01 5.39681494e-01 1.23679399e-01 -1.22626126e+00 3.74089509e-01 9.84428048e-01 5.07838249e-01 -9.50913131e-01 1.17339230e+00 5.78396201e-01 -1.06546354e+00 -5.56339622e-01 -7.79694617e-01 -3.65541615e-02 -1.59892350e-01 1.05652833e+00 -3.61771822e-01 7.18609214e-01 1.00474226e+00 6.74016356e-01 -4.94200617e-01 1.44210136e+00 1.11241147e-01 4.99528706e-01 -7.02410340e-01 -5.73350042e-02 2.67088473e-01 -6.44636512e-01 6.50235951e-01 8.32031250e-01 6.00801826e-01 -4.26962227e-01 1.53319821e-01 9.26523209e-01 -1.67313099e-01 6.94559336e-01 -6.91997707e-01 3.22836399e-01 3.19796562e-01 1.05489480e+00 -6.46010160e-01 -1.78445224e-02 -4.13921684e-01 7.39185333e-01 3.91285300e-01 3.85145664e-01 -8.19114983e-01 -6.97984219e-01 3.71013969e-01 -1.40040383e-01 1.60665140e-02 -4.69125733e-02 -1.52487844e-01 -1.05081809e+00 3.29653442e-01 -8.64281118e-01 6.26286626e-01 -5.35737753e-01 -1.75147378e+00 6.82719469e-01 -1.19740285e-01 -1.35512602e+00 -1.02174245e-01 -8.93596411e-01 -7.62026906e-01 8.05909574e-01 -1.65456450e+00 -6.25992775e-01 -3.74194294e-01 7.54837453e-01 3.31673294e-01 -4.53697711e-01 6.07870281e-01 6.70909882e-01 -9.76970553e-01 7.20246255e-01 7.53784716e-01 9.09232795e-02 7.90407419e-01 -1.12342453e+00 -1.83228418e-01 7.30982900e-01 1.59870595e-01 3.27097446e-01 6.25793576e-01 -5.38856208e-01 -1.06809652e+00 -1.27202976e+00 6.91660225e-01 -9.70197562e-03 3.39659333e-01 -2.50295214e-02 -1.05422068e+00 7.27446198e-01 -3.08450699e-01 3.65265399e-01 5.63209414e-01 -1.30336836e-01 -9.50876325e-02 -3.35844576e-01 -1.57021582e+00 5.85212171e-01 7.27781355e-01 1.58310682e-01 -6.48627520e-01 3.73694152e-01 2.92070419e-01 -2.80473292e-01 -1.09148169e+00 6.99849308e-01 2.16198385e-01 -1.04500961e+00 7.55496621e-01 -7.35181630e-01 2.00121969e-01 -3.33013922e-01 -2.60880530e-01 -1.48638558e+00 -1.43773794e-01 -2.58084506e-01 -4.13412422e-01 1.10570419e+00 4.49805886e-01 -1.01730490e+00 6.30813420e-01 3.92548233e-01 -1.17355630e-01 -1.27819109e+00 -1.12861311e+00 -9.60032940e-01 2.44862184e-01 -4.36871588e-01 3.30121726e-01 1.04494309e+00 -3.28774214e-01 -2.19459966e-01 -4.20972556e-01 -7.86968619e-02 7.48602569e-01 -2.58776307e-01 4.29802507e-01 -1.77882099e+00 -2.54165143e-01 -5.65155864e-01 -7.00175226e-01 -2.43612796e-01 2.58658499e-01 -1.02408028e+00 -1.56452715e-01 -1.29206073e+00 -3.62339407e-01 -5.60628593e-01 -7.34717846e-01 5.40697515e-01 6.36052489e-02 4.17927653e-01 2.16579549e-02 3.03244531e-01 -4.23723489e-01 6.82178855e-01 5.35204113e-01 -1.07717007e-01 -5.38657963e-01 2.41049856e-01 -3.57157171e-01 1.05613506e+00 9.05776024e-01 -8.10535312e-01 -4.83711883e-02 -1.26230240e-01 4.37169433e-01 -4.15528774e-01 -4.65727225e-02 -1.43771601e+00 3.24629605e-01 -1.41085565e-01 7.34104872e-01 -4.02606964e-01 1.15224302e-01 -1.15647662e+00 2.79805034e-01 6.39985502e-01 -1.71595022e-01 3.05475950e-01 -2.25867052e-02 4.21947807e-01 -4.65485007e-01 -6.66907728e-01 1.01694489e+00 -1.40422925e-01 -5.35265982e-01 3.73916179e-01 -3.44430394e-02 9.02702585e-02 1.32606125e+00 -3.66544962e-01 1.41908228e-01 -4.26129848e-02 -7.91287184e-01 2.55030185e-01 2.88645059e-01 2.40650341e-01 7.98784971e-01 -1.47976005e+00 -9.72997189e-01 3.35909754e-01 -3.06270476e-02 4.80255559e-02 2.86648255e-02 1.22605002e+00 -5.37356079e-01 1.00770816e-02 -2.07995057e-01 -6.15161538e-01 -9.38509583e-01 5.02219439e-01 5.41136324e-01 -3.52001846e-01 -6.83920503e-01 7.26757705e-01 -6.09744728e-01 -5.81705451e-01 5.31709373e-01 -1.82034120e-01 -2.93256849e-01 -8.04847106e-02 1.98102191e-01 7.38368332e-01 2.72416711e-01 -4.54357952e-01 -3.55707467e-01 7.36203074e-01 1.33449942e-01 2.61288792e-01 1.68431652e+00 2.37995282e-01 -3.39196265e-01 8.93803716e-01 1.30330706e+00 -1.79088041e-01 -1.08858550e+00 -1.77509829e-01 4.22420084e-01 -2.16426060e-01 1.04238130e-02 -4.17874545e-01 -1.38068926e+00 6.56680167e-01 9.38560367e-01 1.28963515e-01 1.15483773e+00 -6.10285938e-01 6.40388489e-01 6.52952433e-01 1.75903350e-01 -1.43115973e+00 8.91103223e-02 6.11863494e-01 8.17676425e-01 -1.26041365e+00 -4.83822897e-02 2.79932886e-01 -2.79078394e-01 1.34276617e+00 7.97883213e-01 -5.43217599e-01 8.62237036e-01 2.64618158e-01 -6.75075352e-02 1.12929113e-01 -4.76057082e-01 2.65193343e-01 7.23323971e-02 7.97337741e-02 3.59471112e-01 -2.78406411e-01 -5.34324467e-01 6.15312994e-01 1.40056372e-01 1.48545116e-01 2.16506973e-01 8.95622551e-01 -5.10074258e-01 -9.33028460e-01 -6.31443441e-01 8.81888747e-01 -5.63766956e-01 4.60987948e-02 -1.65376037e-01 7.18809605e-01 4.64978665e-01 7.69282043e-01 5.35375178e-02 -2.41170213e-01 4.75378662e-01 5.06140113e-01 1.03933401e-01 -1.40299991e-01 -4.71367329e-01 -5.37413061e-01 -5.97684383e-01 -4.29145783e-01 -2.58732170e-01 -4.06279892e-01 -1.25326121e+00 -3.13098073e-01 -4.54544455e-01 1.75284237e-01 7.37506866e-01 9.21049178e-01 2.65966177e-01 3.75338435e-01 8.66100907e-01 -6.14769757e-01 -8.32552016e-01 -1.12866092e+00 -7.58081257e-01 6.27239704e-01 2.42752671e-01 -8.90174806e-01 -7.92294025e-01 -4.07260865e-01]
[7.729569911956787, 2.849924087524414]
ba7d94be-b795-4586-825c-346078750b8b
the-role-of-global-and-local-context-in-named
2305.03132
null
https://arxiv.org/abs/2305.03132v2
https://arxiv.org/pdf/2305.03132v2.pdf
The Role of Global and Local Context in Named Entity Recognition
Pre-trained transformer-based models have recently shown great performance when applied to Named Entity Recognition (NER). As the complexity of their self-attention mechanism prevents them from processing long documents at once, these models are usually applied in a sequential fashion. Such an approach unfortunately only incorporates local context and prevents leveraging global document context in long documents such as novels, which might hinder performance. In this article, we explore the impact of global document context, and its relationships with local context. We find that correctly retrieving global document context has a greater impact on performance than only leveraging local context, prompting for further research on how to better retrieve that context.
['Richard Dufour', 'Vincent Labatut', 'Arthur Amalvy']
2023-05-04
null
null
null
null
['named-entity-recognition-ner']
['natural-language-processing']
[ 2.49515474e-02 -1.66835785e-01 -2.52123475e-01 -3.49099517e-01 -9.16444242e-01 -7.52160192e-01 8.90100718e-01 5.34124672e-01 -9.48852897e-01 5.18700421e-01 7.30399609e-01 -8.02648544e-01 4.18932177e-02 -6.18680894e-01 -4.53121483e-01 -2.63319343e-01 1.70986265e-01 1.84657127e-01 2.58958697e-01 -1.62543803e-01 3.89934450e-01 4.38670427e-01 -9.66973960e-01 3.04097116e-01 7.13026464e-01 2.24955916e-01 4.53179419e-01 6.26534104e-01 -6.95705891e-01 9.06045258e-01 -9.28866684e-01 -3.92531395e-01 -1.22496918e-01 -4.13264185e-01 -1.12210405e+00 -4.41708893e-01 5.56334972e-01 -3.92211288e-01 -3.69956166e-01 7.70370841e-01 4.96970505e-01 3.34681839e-01 4.51400995e-01 -6.03260100e-01 -9.32392657e-01 7.86708355e-01 -1.54735133e-01 9.19535935e-01 2.76571870e-01 -1.73410296e-01 1.20626831e+00 -6.99222386e-01 6.82707250e-01 9.40516949e-01 7.16144741e-01 4.17689115e-01 -1.05798090e+00 -5.33998311e-01 5.81135392e-01 1.47184700e-01 -1.11002409e+00 -6.66755795e-01 6.41654193e-01 -1.22109009e-02 1.62642038e+00 1.75836578e-01 3.13476235e-01 1.16919565e+00 2.41184637e-01 8.16040158e-01 8.71231735e-01 -6.78351700e-01 7.76439905e-02 -1.29871994e-01 2.64499873e-01 2.10800007e-01 3.19393843e-01 -7.16776177e-02 -3.64246964e-01 -2.41638929e-01 4.58650112e-01 -4.18240763e-02 -4.53643017e-02 2.51892984e-01 -1.07817400e+00 6.27737820e-01 4.27076489e-01 9.16523576e-01 -4.04232115e-01 1.69639647e-01 5.36736071e-01 4.73480225e-01 3.57287526e-01 8.50444496e-01 -6.32926285e-01 -3.60657811e-01 -1.14409196e+00 3.77071500e-02 7.32991099e-01 8.68302464e-01 4.87241596e-01 4.72685397e-02 -8.62353370e-02 8.29301298e-01 2.31584720e-02 3.22030932e-01 7.30589688e-01 -6.42512739e-01 6.40226245e-01 4.14730579e-01 1.70308769e-01 -9.96294498e-01 -4.21009511e-01 -3.76358896e-01 -1.75752461e-01 -4.84111965e-01 4.26458061e-01 -1.62493989e-01 -9.75190222e-01 1.76503110e+00 5.66169759e-03 -1.37058794e-02 1.91896826e-01 5.13921082e-01 4.56793934e-01 7.76901603e-01 6.38971329e-01 -1.70022652e-01 1.36804450e+00 -7.83528328e-01 -8.21103454e-01 -7.65696704e-01 8.91018510e-01 -1.02675116e+00 1.13407147e+00 -1.17369257e-01 -9.72738326e-01 -1.50083333e-01 -8.20395291e-01 -2.94498563e-01 -6.94359481e-01 -1.57003462e-01 5.52795768e-01 4.99782860e-01 -1.09570611e+00 5.11013865e-01 -8.96090984e-01 -8.44714820e-01 -6.48138449e-02 9.15575475e-02 -1.65298268e-01 -1.71800032e-01 -1.33363163e+00 1.37833846e+00 3.30777466e-01 2.00027347e-01 -2.62873441e-01 -4.38634902e-01 -6.94674671e-01 3.23887318e-01 3.99061352e-01 -4.95017678e-01 1.51886463e+00 -6.49618447e-01 -1.05347776e+00 4.49092120e-01 -6.57161355e-01 -3.31965536e-01 -7.03368261e-02 -4.02166873e-01 -7.83217967e-01 8.77258833e-03 1.30945534e-01 3.76278996e-01 4.76221114e-01 -8.75151932e-01 -5.90029716e-01 -4.06817466e-01 -4.82760277e-03 3.90342176e-01 -6.06860459e-01 5.33714652e-01 -4.54316825e-01 -8.31734359e-01 3.03266160e-02 -7.12169528e-01 -2.13763714e-01 -6.07941985e-01 -5.94489053e-02 -3.44853878e-01 7.84671128e-01 -7.61845946e-01 1.41971159e+00 -2.10327411e+00 -3.35088521e-01 4.63445447e-02 -1.36996433e-01 3.69763851e-01 -3.81084055e-01 8.51390779e-01 9.45224762e-02 4.99255627e-01 1.30008176e-01 -1.47374541e-01 -9.72928554e-02 1.58287719e-01 -4.85011548e-01 -1.55540686e-02 2.40605175e-01 1.08485103e+00 -1.06383824e+00 -5.30023217e-01 -8.09753984e-02 4.92874950e-01 -3.53970796e-01 -9.07722712e-02 -9.87676159e-02 -1.21131241e-01 -6.13377452e-01 5.30553222e-01 2.61146039e-01 -2.24553645e-01 4.75364506e-01 -2.90186945e-02 -1.01372987e-01 1.10381949e+00 -6.47537947e-01 1.32532477e+00 -9.11598742e-01 8.63838375e-01 -8.12939182e-03 -7.50014961e-01 7.87771046e-01 5.17088592e-01 8.54114145e-02 -9.84266818e-01 -9.66596305e-02 1.36462793e-01 7.86765963e-02 -3.84027630e-01 9.07403648e-01 -3.82695228e-01 -2.99879517e-02 6.72775030e-01 -9.85343978e-02 2.68993080e-01 1.87155470e-01 3.97430569e-01 1.37296832e+00 -1.92404777e-01 3.03848773e-01 -1.87902331e-01 -3.54089551e-02 1.99873850e-01 4.69540209e-01 9.22263622e-01 -1.62415519e-01 2.86650807e-01 2.02428717e-02 -1.73044950e-01 -1.01072288e+00 -6.42467678e-01 -1.99633390e-02 1.45463836e+00 -8.86534899e-02 -7.00383782e-01 -3.42949897e-01 -1.00123286e+00 -1.07020251e-01 1.02590311e+00 -3.60413939e-01 -2.91910708e-01 -8.42600465e-01 -6.40878141e-01 7.34686911e-01 1.09766197e+00 1.91211790e-01 -1.11593783e+00 -5.83824813e-01 6.46407962e-01 -2.61935413e-01 -1.13342416e+00 -6.87420368e-01 5.60970843e-01 -1.27205491e+00 -6.80186093e-01 -6.08317494e-01 -6.81446254e-01 4.47077990e-01 4.39822584e-01 1.22480798e+00 2.41464555e-01 2.78435111e-01 4.92858559e-01 -4.51168746e-01 -3.02667469e-01 -2.69392163e-01 4.78840381e-01 -3.25977147e-01 -6.18317306e-01 7.29993463e-01 -4.81930345e-01 -2.83960670e-01 2.35154912e-01 -8.92957568e-01 -3.19105983e-01 8.07525516e-01 1.05522919e+00 1.41972646e-01 4.23941128e-02 8.39681506e-01 -1.00588143e+00 9.06753659e-01 -5.97772598e-01 -1.76283032e-01 5.48878372e-01 -7.88944781e-01 1.43516600e-01 6.29450321e-01 -5.29786587e-01 -1.38167679e+00 -5.21938145e-01 -1.31138310e-01 2.73390487e-02 -1.52906477e-01 9.82400119e-01 -5.92464320e-02 2.72853732e-01 5.56439042e-01 1.73466682e-01 -6.24857664e-01 -8.13308120e-01 5.72070256e-02 7.25981712e-01 3.29161912e-01 -7.42980123e-01 4.94639486e-01 5.23465388e-02 -5.95177114e-01 -8.84964764e-01 -8.85202229e-01 -4.96507049e-01 -2.99639344e-01 1.21071160e-01 6.31031871e-01 -8.02499235e-01 -2.82788686e-02 4.35401872e-02 -1.11686277e+00 -3.29745620e-01 -1.20492607e-01 5.18905282e-01 4.43492904e-02 3.57164353e-01 -8.90296936e-01 -6.77196562e-01 -2.99437225e-01 -6.80956364e-01 8.03178847e-01 1.80584714e-01 -4.97231841e-01 -1.16744041e+00 7.68426135e-02 2.36504391e-01 8.56263578e-01 -3.74497414e-01 1.11203313e+00 -1.08418834e+00 -3.20558757e-01 -3.37798655e-01 -1.55315936e-01 -3.47048081e-02 1.73032448e-01 -1.68485373e-01 -9.28086162e-01 -2.25142181e-01 -9.70902070e-02 -1.49110043e-02 7.67464817e-01 -7.67129585e-02 7.40324438e-01 -3.61590594e-01 -5.29431045e-01 1.25288889e-01 1.29780030e+00 4.78671223e-01 3.70817691e-01 6.74277961e-01 6.62258446e-01 3.54908317e-01 4.38312411e-01 2.69743260e-02 5.37602544e-01 5.06389737e-01 -2.22482085e-01 9.09727663e-02 -1.43478334e-01 -4.21023995e-01 2.73920685e-01 1.03285921e+00 2.38736793e-01 -5.61380625e-01 -1.20682514e+00 9.29458976e-01 -1.48628926e+00 -1.01709306e+00 3.33291352e-01 1.88771808e+00 8.62712085e-01 3.17788959e-01 -2.60406166e-01 -1.51765078e-01 7.65122473e-01 3.52830052e-01 -4.75473344e-01 -8.18990707e-01 -9.73665416e-02 1.63965687e-01 6.31056249e-01 3.98832053e-01 -8.13113928e-01 1.02868998e+00 7.63388729e+00 4.77133125e-01 -1.30339050e+00 1.38235748e-01 3.82934391e-01 -1.71663433e-01 -6.98086202e-01 3.10968816e-01 -9.49782729e-01 4.24560219e-01 1.33121371e+00 -2.35686064e-01 2.06939608e-01 7.27819502e-01 -2.03567687e-02 -9.35133770e-02 -9.86224830e-01 3.56299758e-01 -5.35032526e-02 -1.01644933e+00 1.71015039e-01 1.58313587e-01 4.34564471e-01 3.75330865e-01 -2.09169865e-01 5.45376897e-01 3.26895803e-01 -7.41874218e-01 5.20359993e-01 1.99628219e-01 3.58337194e-01 -7.08344638e-01 6.18188262e-01 3.60022247e-01 -1.01474619e+00 -1.88001871e-01 -3.43462914e-01 -8.54893588e-03 2.81428993e-01 6.20334268e-01 -9.13585067e-01 3.55358601e-01 5.16293645e-01 2.80001640e-01 -6.98102117e-01 9.81812656e-01 -1.38103113e-01 9.45460618e-01 -5.55341899e-01 -1.91461846e-01 4.81971413e-01 2.55666047e-01 5.93144774e-01 1.60573196e+00 2.39570990e-01 1.43356830e-01 -1.16829075e-01 3.71008128e-01 -9.47706997e-02 2.84471661e-01 -7.02299118e-01 -5.25880814e-01 9.23626482e-01 7.99287379e-01 -8.48199189e-01 -5.68531990e-01 -7.81133056e-01 8.12410474e-01 4.95482445e-01 5.19892633e-01 -3.88179272e-01 -5.32258153e-01 4.41805214e-01 1.28551960e-01 5.87820947e-01 -5.34062445e-01 -3.39127630e-01 -1.26738381e+00 1.77944794e-01 -7.28515625e-01 7.10102797e-01 -6.48405373e-01 -1.15136468e+00 6.47930324e-01 -1.62642673e-02 -5.99612713e-01 -2.65484035e-01 -2.55899787e-01 -8.74820411e-01 8.13557386e-01 -1.51558435e+00 -9.61175025e-01 2.83508271e-01 4.07953620e-01 5.30754805e-01 1.51701048e-01 8.67922068e-01 3.75885189e-01 -5.72447598e-01 8.02070677e-01 3.39785963e-01 2.52827495e-01 9.84459281e-01 -1.26020634e+00 7.52828002e-01 1.16593206e+00 4.90734279e-01 1.18408048e+00 6.05719268e-01 -7.73492157e-01 -1.32846999e+00 -7.61539996e-01 1.74686515e+00 -6.08206570e-01 6.07543647e-01 -1.56120047e-01 -1.17367625e+00 1.00316811e+00 3.11818004e-01 -2.02293694e-01 7.23494649e-01 8.01514745e-01 -4.97614861e-01 8.36283267e-02 -9.14254427e-01 9.04394329e-01 9.98577774e-01 -9.84131515e-01 -1.11884606e+00 -9.40647945e-02 7.91543961e-01 -9.19490308e-02 -8.49165082e-01 2.39517048e-01 3.55863899e-01 -5.02667665e-01 8.09292138e-01 -5.75998902e-01 8.80202129e-02 -1.31019980e-01 -3.79932821e-02 -1.15214384e+00 -4.82613802e-01 -3.33188802e-01 -3.69074196e-01 1.56693804e+00 6.25601232e-01 -6.31873071e-01 5.20658970e-01 9.94711578e-01 -1.51766866e-01 -5.03792286e-01 -7.58988917e-01 -9.11653459e-01 3.37071598e-01 -4.29047257e-01 8.22321057e-01 1.33602667e+00 2.41215751e-01 5.36477089e-01 7.45054185e-02 7.27168936e-03 -3.07952259e-02 2.19351631e-02 1.49356619e-01 -8.60344470e-01 -9.24813226e-02 -5.82390666e-01 -2.81219244e-01 -1.21656358e+00 1.62675589e-01 -7.67063975e-01 7.17953816e-02 -1.56612158e+00 2.41832256e-01 -6.15081251e-01 -6.72995806e-01 7.34670639e-01 -4.89715248e-01 -7.06614852e-02 3.16706598e-01 2.08224460e-01 -6.19006455e-01 2.52249539e-01 8.61808896e-01 -4.47821431e-02 -2.00459659e-01 -3.16938132e-01 -1.17748582e+00 2.86999106e-01 8.66389155e-01 -6.09497547e-01 -2.59689003e-01 -8.23698342e-01 2.24881649e-01 -1.05768532e-01 -7.25533813e-02 -8.57000887e-01 4.23345983e-01 -1.61691517e-01 5.31682134e-01 -5.28015614e-01 7.99955875e-02 -6.72800124e-01 -2.77487844e-01 -1.65447574e-02 -5.44245958e-01 6.22157991e-01 4.61854458e-01 5.11241257e-01 -3.80316556e-01 -4.76763964e-01 2.53957212e-01 -1.93200380e-01 -7.76518047e-01 -1.86385915e-01 -9.20795977e-01 2.68287122e-01 5.02964020e-01 -1.54471979e-01 -2.74177313e-01 -3.14306736e-01 -6.00656867e-01 1.87035538e-02 5.78041136e-01 6.70419872e-01 2.32145593e-01 -1.12080586e+00 -3.83204222e-01 1.60858370e-02 1.17087185e-01 -2.79746622e-01 1.21984128e-02 5.05591571e-01 -1.03955995e-02 7.26415396e-01 1.86964437e-01 -1.01675749e-01 -1.11020780e+00 6.65872872e-01 1.89878508e-01 -3.84465069e-01 -7.81095386e-01 6.40832901e-01 -8.28603432e-02 -1.45891681e-01 1.67417243e-01 -4.65554297e-01 -2.04910755e-01 4.85897847e-02 6.32380307e-01 1.36566386e-01 3.81935000e-01 -5.03492653e-01 -4.34236586e-01 2.66964942e-01 -6.79409742e-01 -2.65864104e-01 1.15498447e+00 -3.20342481e-01 5.63942157e-02 3.85764152e-01 1.22547495e+00 1.89717144e-01 -7.39428818e-01 -3.72597307e-01 6.59622967e-01 -4.45572019e-01 4.30483758e-01 -9.39951062e-01 -8.39910269e-01 6.80599928e-01 1.94415852e-01 2.05046386e-01 1.04599571e+00 -4.19234671e-02 1.09100223e+00 8.27109933e-01 2.37445772e-01 -1.18835211e+00 -2.26596043e-01 9.90614951e-01 3.57808888e-01 -1.03154373e+00 -1.23276241e-01 1.62687302e-01 -5.07640123e-01 9.06141102e-01 5.77653229e-01 1.80947155e-01 5.62725484e-01 4.00151879e-01 2.15300009e-01 -2.06522107e-01 -9.28062201e-01 -1.59356758e-01 1.29806280e-01 3.59040380e-01 8.58847857e-01 -1.65324256e-01 -3.63105148e-01 5.44693232e-01 -1.51315972e-01 -1.77060544e-01 3.96023542e-01 1.55038691e+00 -3.87306601e-01 -1.34084880e+00 -4.54954684e-01 6.32459700e-01 -8.74520898e-01 -5.85114181e-01 -5.40749907e-01 5.95207810e-01 -2.94699311e-01 1.06592083e+00 1.59863994e-01 -1.22041762e-01 3.89090359e-01 5.98384202e-01 3.12687546e-01 -7.28652477e-01 -8.87344539e-01 3.46763991e-02 4.25312817e-01 -4.16883200e-01 -1.28712296e-01 -7.45768070e-01 -1.07688153e+00 -3.26088399e-01 -6.64934635e-01 4.29563701e-01 8.02060246e-01 1.16110373e+00 6.72049522e-01 2.79140860e-01 2.43906558e-01 -3.14982295e-01 -4.52193141e-01 -9.93229628e-01 -2.19182342e-01 1.05950244e-01 3.01343292e-01 -3.51990730e-01 -2.07972452e-01 -1.97209194e-01]
[10.795117378234863, 8.505051612854004]
f8339d2d-2e6a-43db-a178-31e298b7716f
vision-transformer-adapter-for-dense
2205.08534
null
https://arxiv.org/abs/2205.08534v4
https://arxiv.org/pdf/2205.08534v4.pdf
Vision Transformer Adapter for Dense Predictions
This work investigates a simple yet powerful dense prediction task adapter for Vision Transformer (ViT). Unlike recently advanced variants that incorporate vision-specific inductive biases into their architectures, the plain ViT suffers inferior performance on dense predictions due to weak prior assumptions. To address this issue, we propose the ViT-Adapter, which allows plain ViT to achieve comparable performance to vision-specific transformers. Specifically, the backbone in our framework is a plain ViT that can learn powerful representations from large-scale multi-modal data. When transferring to downstream tasks, a pre-training-free adapter is used to introduce the image-related inductive biases into the model, making it suitable for these tasks. We verify ViT-Adapter on multiple dense prediction tasks, including object detection, instance segmentation, and semantic segmentation. Notably, without using extra detection data, our ViT-Adapter-L yields state-of-the-art 60.9 box AP and 53.0 mask AP on COCO test-dev. We hope that the ViT-Adapter could serve as an alternative for vision-specific transformers and facilitate future research. The code and models will be released at https://github.com/czczup/ViT-Adapter.
['Yu Qiao', 'Jifeng Dai', 'Tong Lu', 'Junjun He', 'Wenhai Wang', 'Yuchen Duan', 'Zhe Chen']
2022-05-17
null
null
null
null
['real-time-object-detection']
['computer-vision']
[ 1.85610831e-01 3.42476070e-01 -9.40500107e-03 -4.45145160e-01 -8.55795622e-01 -3.74139488e-01 6.70905173e-01 -5.34753799e-01 -2.99194455e-01 3.03581685e-01 1.44427449e-01 -4.10572588e-01 4.74973917e-01 -6.55230105e-01 -1.04772449e+00 -6.18490040e-01 5.72917998e-01 5.14008105e-01 5.37365675e-01 -4.75700535e-02 -7.39327893e-02 1.46814361e-02 -1.35151434e+00 5.18745005e-01 7.17600942e-01 1.22083592e+00 7.30819046e-01 5.63481092e-01 5.85194975e-02 1.06001270e+00 -1.74250156e-01 -7.15102315e-01 4.29252416e-01 -1.58563003e-01 -8.22964191e-01 -5.39165037e-03 8.37667525e-01 -4.42376703e-01 -4.26768512e-01 9.07787383e-01 5.23777902e-01 -1.20806307e-01 6.10586822e-01 -1.19471371e+00 -8.07252765e-01 6.15384102e-01 -6.43661797e-01 7.41939023e-02 -1.14198022e-01 7.84510672e-01 1.21189713e+00 -1.14865303e+00 3.57040197e-01 1.25030029e+00 1.01042879e+00 6.34883702e-01 -1.25055754e+00 -6.90014005e-01 3.87188464e-01 3.62833560e-01 -1.08903527e+00 -6.33327663e-01 4.57644910e-01 -3.55691850e-01 1.05203211e+00 -8.35055206e-03 6.38871670e-01 1.38446343e+00 9.88024101e-02 1.35911536e+00 1.15364099e+00 -1.38770983e-01 -5.26274741e-02 6.94391504e-02 3.67549658e-02 8.17780077e-01 -5.77055812e-02 2.02759981e-01 -2.80252635e-01 4.15467560e-01 7.43925869e-01 -9.64543894e-02 -2.72221208e-01 -1.61854610e-01 -1.16685021e+00 7.98115909e-01 8.26105475e-01 -1.74511954e-01 -3.37567568e-01 3.07844877e-01 3.75944167e-01 3.25781517e-02 5.00271499e-01 4.44665968e-01 -6.00486398e-01 -2.45910473e-02 -9.44155693e-01 -1.51765812e-02 5.39256811e-01 1.11523592e+00 7.14496076e-01 1.33502558e-01 -5.78931153e-01 9.48412895e-01 4.40141320e-01 5.24237037e-01 4.20752943e-01 -1.06488132e+00 4.40907836e-01 4.64826286e-01 -3.14880222e-01 -2.87676692e-01 -2.93496370e-01 -7.36476243e-01 -7.52223670e-01 1.92067876e-01 4.84186709e-01 -1.37654424e-01 -1.46307850e+00 1.68257761e+00 2.33039975e-01 5.27226567e-01 -6.37208819e-02 1.07375228e+00 1.24385512e+00 6.84948742e-01 2.06103563e-01 4.44961399e-01 1.37615955e+00 -1.66464961e+00 -2.25077510e-01 -7.16043770e-01 4.98784214e-01 -8.09859276e-01 1.56417048e+00 2.42662966e-01 -1.20988619e+00 -7.24850476e-01 -6.92026794e-01 -4.96989846e-01 -5.45267649e-02 3.83022189e-01 7.29650140e-01 3.23637486e-01 -1.32352197e+00 2.07289934e-01 -8.38347614e-01 -3.46503109e-01 8.67114365e-01 9.08760726e-02 -1.02789335e-01 -2.85954356e-01 -7.80512810e-01 9.80022252e-01 -8.05452392e-02 2.26296648e-01 -1.49593782e+00 -9.84093487e-01 -9.67837811e-01 -6.73959125e-03 3.04867744e-01 -1.07231176e+00 1.69287813e+00 -9.73354101e-01 -1.52766991e+00 1.02419710e+00 -2.68047243e-01 -6.13122642e-01 6.74598813e-01 -2.94440806e-01 1.53707817e-01 -5.44555811e-03 2.84279794e-01 1.14866042e+00 1.00488043e+00 -1.12124097e+00 -6.31748021e-01 -5.80919087e-02 3.34106833e-01 1.52007148e-01 -1.11684866e-01 -2.54680783e-01 -9.70342934e-01 -5.92396736e-01 -3.79282415e-01 -9.51871276e-01 -2.74942964e-01 1.93138793e-01 -5.81153035e-01 -2.31606901e-01 6.98395431e-01 -5.10351777e-01 5.23064792e-01 -2.15148258e+00 1.43052816e-01 -3.31780136e-01 3.12498868e-01 4.32598442e-01 -6.88753188e-01 -3.99003699e-02 1.57577887e-01 -2.35026434e-01 -2.51128942e-01 -9.37061608e-01 -2.99619362e-02 2.34015286e-01 -2.73722500e-01 3.69014114e-01 4.34519827e-01 1.32325542e+00 -6.72017872e-01 -5.06542981e-01 4.46895868e-01 4.54153478e-01 -6.18575513e-01 2.86194772e-01 -5.28308928e-01 2.88760245e-01 -2.98273683e-01 9.19616640e-01 6.03386343e-01 -5.77790856e-01 -3.74840260e-01 -5.08717477e-01 6.88195974e-02 3.85022253e-01 -4.38556284e-01 1.72506320e+00 -7.04425275e-01 9.22564447e-01 1.65054515e-01 -1.00166619e+00 5.70199370e-01 5.39599024e-02 3.74962902e-03 -9.62098181e-01 1.08599134e-01 -1.66807428e-01 -5.81610799e-02 -3.14192116e-01 3.75689268e-01 -1.93617523e-01 2.65559763e-01 2.02673063e-01 3.46363872e-01 -3.01886529e-01 -2.67585013e-02 2.64820665e-01 1.03679597e+00 2.75302321e-01 -1.82036698e-01 -5.05380854e-02 3.11189294e-01 7.97237977e-02 5.99235952e-01 7.74324417e-01 -4.35607582e-01 1.05345702e+00 2.73990571e-01 -1.82950601e-01 -1.04568541e+00 -1.21460819e+00 -1.90601647e-01 1.19900918e+00 6.69464320e-02 -2.75898784e-01 -6.25815988e-01 -8.22635174e-01 2.31992453e-02 7.36698151e-01 -7.58693993e-01 -7.61407092e-02 -1.85683250e-01 -7.29608417e-01 6.93238556e-01 8.63087356e-01 6.57429457e-01 -1.13648045e+00 -2.81550705e-01 -3.89029421e-02 -3.54981214e-01 -1.28495586e+00 -4.63860512e-01 3.08168709e-01 -7.80495048e-01 -1.09144604e+00 -8.65863800e-01 -7.39194274e-01 4.89635497e-01 4.14257675e-01 1.29783785e+00 -1.11956336e-01 -1.38870075e-01 4.61113185e-01 -2.09477842e-01 -4.43822831e-01 -1.57724619e-01 2.43695959e-01 -3.90689343e-01 -2.42621034e-01 2.32303888e-01 -2.94717848e-01 -8.33572626e-01 4.13436681e-01 -5.59009314e-01 5.67993760e-01 8.56933951e-01 8.74241948e-01 7.65240371e-01 -5.30853152e-01 3.27330112e-01 -8.65803659e-01 1.60791516e-01 -4.60821033e-01 -4.50239539e-01 2.06884205e-01 -4.48515236e-01 -1.20718151e-01 5.57700992e-01 -5.26455462e-01 -1.11003399e+00 2.03264832e-01 -5.42297602e-01 -7.88992107e-01 -5.34589812e-02 3.21817070e-01 -1.46859840e-01 -4.46996056e-02 5.37900984e-01 4.03559715e-01 1.13617945e-02 -4.92386162e-01 5.78251839e-01 5.08069277e-01 6.53100610e-01 -5.09142041e-01 7.04261720e-01 3.94301146e-01 -3.50384712e-01 -6.42879367e-01 -1.44022691e+00 -4.33582455e-01 -2.36655623e-01 -1.38591807e-02 1.01272869e+00 -1.43531621e+00 -4.01503891e-01 6.83779240e-01 -1.01826549e+00 -1.15884554e+00 -4.44482297e-01 1.91123649e-01 -6.04202867e-01 1.71034962e-01 -8.66928756e-01 -4.37262028e-01 -5.53099513e-01 -1.31475532e+00 1.25142109e+00 2.42269814e-01 -7.99657835e-04 -9.39106464e-01 -1.21433549e-02 9.61610913e-01 5.08242905e-01 -2.32728586e-01 3.88944417e-01 -2.52066940e-01 -6.86502278e-01 8.82175863e-02 -6.61952078e-01 6.97557986e-01 -1.23367108e-01 4.49210815e-02 -1.50856590e+00 -2.59596169e-01 -2.61464238e-01 -8.01135123e-01 1.35492682e+00 6.24909103e-01 1.45031619e+00 -7.57828876e-02 -3.74705315e-01 1.08975554e+00 1.20532751e+00 -3.68706703e-01 7.57467866e-01 3.46140921e-01 1.08479881e+00 2.88514733e-01 5.23038268e-01 8.57219473e-02 9.46813881e-01 5.84948897e-01 6.88066542e-01 -5.23564577e-01 -6.19831324e-01 -3.14529032e-01 5.81120968e-01 6.17901862e-01 -4.00790088e-02 -1.73592016e-01 -9.07255352e-01 7.75840700e-01 -1.80810213e+00 -7.51838148e-01 -2.29561105e-01 1.70667505e+00 9.03049767e-01 2.45504394e-01 1.12181291e-01 -2.71322399e-01 3.72129709e-01 1.28115147e-01 -8.39208841e-01 -1.54477164e-01 -8.04073289e-02 3.78223717e-01 5.20850837e-01 4.66278166e-01 -1.18813276e+00 1.33495390e+00 6.30140781e+00 9.29478228e-01 -1.27622974e+00 4.69506264e-01 8.54564428e-01 -2.10647553e-01 -3.61497104e-01 -1.10895924e-01 -9.15560842e-01 4.52569336e-01 7.89547384e-01 2.36752257e-01 7.55655542e-02 9.98780131e-01 1.71752274e-01 -4.26936559e-02 -1.16259575e+00 8.24114382e-01 -5.20835817e-02 -1.38334990e+00 2.44551376e-01 -5.70558123e-02 5.93919694e-01 8.22893739e-01 3.48223329e-01 6.57315135e-01 4.92859811e-01 -1.21140528e+00 9.37458277e-01 3.66953194e-01 9.81070459e-01 -2.11679891e-01 7.14703023e-01 1.30753964e-01 -1.10479367e+00 1.09734582e-02 -6.26289070e-01 -1.26882240e-01 1.00973748e-01 6.84884250e-01 -7.87407696e-01 2.09615797e-01 8.39645207e-01 1.13097107e+00 -7.90979683e-01 1.15103257e+00 -4.73179281e-01 1.01927960e+00 -3.33804756e-01 3.77995640e-01 4.66750711e-01 1.94675326e-02 2.85715550e-01 1.33145940e+00 1.39854163e-01 -2.88454205e-01 7.16952756e-02 1.09186506e+00 -1.94257930e-01 -3.06059599e-01 -4.78553027e-01 1.82101443e-01 2.78653264e-01 1.41220570e+00 -4.55692023e-01 -4.42582250e-01 -7.51943767e-01 9.91965532e-01 5.22952855e-01 4.37468797e-01 -1.07563472e+00 1.96731374e-01 8.89638722e-01 1.86506182e-01 6.38680339e-01 6.89494163e-02 -5.50296366e-01 -1.22902620e+00 -2.29500815e-01 -9.07469451e-01 2.58911639e-01 -9.92239177e-01 -1.59614480e+00 5.67288458e-01 -2.57974863e-01 -9.83517945e-01 7.52686337e-02 -7.69514799e-01 -8.68476450e-01 7.64352500e-01 -1.85151935e+00 -1.73330462e+00 -5.19468963e-01 8.40561032e-01 8.72775257e-01 -1.10821813e-01 5.46433270e-01 2.13819787e-01 -8.53999913e-01 8.37630987e-01 -1.99616253e-01 1.84386075e-01 8.43005240e-01 -1.33466947e+00 6.45115077e-01 8.88254464e-01 2.29237616e-01 2.45809257e-01 4.81165767e-01 -3.29332054e-01 -1.21919322e+00 -1.53747284e+00 4.16994870e-01 -7.91172683e-01 6.66666806e-01 -3.91449124e-01 -7.73250520e-01 1.08464229e+00 3.37818414e-01 2.92126328e-01 4.45048362e-01 3.01159114e-01 -5.24929345e-01 -2.51987338e-01 -9.08065438e-01 6.19968951e-01 1.12786782e+00 -6.16416216e-01 -6.11623764e-01 3.62136960e-01 8.58188331e-01 -4.96761233e-01 -6.41019046e-01 3.96954179e-01 5.12664437e-01 -1.01085734e+00 1.14340329e+00 -1.69649079e-01 8.11784148e-01 -3.08729529e-01 -7.17546344e-02 -1.28049827e+00 -6.08815074e-01 -1.95618391e-01 -9.97106582e-02 1.19179058e+00 6.31542742e-01 -6.09243453e-01 8.25065374e-01 2.70363450e-01 -5.07866502e-01 -8.85013044e-01 -7.45665371e-01 -5.81984937e-01 3.04220647e-01 -7.20691800e-01 3.63679647e-01 7.25304782e-01 -5.55714488e-01 6.22478962e-01 -3.08402747e-01 8.48654434e-02 6.98506892e-01 8.79526287e-02 9.41145420e-01 -8.32307696e-01 -5.80470264e-01 -4.97659117e-01 -1.50130853e-01 -1.42501891e+00 1.75420120e-01 -8.16924810e-01 2.31551424e-01 -1.76039195e+00 4.12920892e-01 -6.66726351e-01 -3.42326969e-01 8.63074839e-01 -3.26971591e-01 7.23390937e-01 4.61241245e-01 1.84039593e-01 -7.98162699e-01 8.64620507e-01 1.55293190e+00 -4.63678777e-01 1.66699082e-01 1.15958653e-01 -8.81149232e-01 7.82942653e-01 9.04152453e-01 -1.79055035e-01 -6.32177472e-01 -8.07281554e-01 -9.31740180e-02 -4.25419390e-01 6.09682381e-01 -1.01192677e+00 1.12993985e-01 5.90927899e-02 4.13345575e-01 -4.40933943e-01 5.68525970e-01 -4.75410730e-01 -1.30916715e-01 2.29652554e-01 -1.20715037e-01 -2.82920718e-01 4.23759043e-01 3.94425929e-01 -2.05308735e-01 3.12497970e-02 9.19042826e-01 -2.01986060e-01 -9.37166274e-01 5.03906071e-01 -2.81853020e-01 2.46895030e-01 8.85180175e-01 -2.61847556e-01 -6.09669864e-01 -2.87476510e-01 -6.36869609e-01 4.86567110e-01 6.33061111e-01 4.22289044e-01 7.17245042e-01 -9.76993382e-01 -8.31709504e-01 7.70123005e-02 2.69077212e-01 4.62380379e-01 1.38970852e-01 1.02300727e+00 -3.06882322e-01 2.21222937e-01 -2.18894631e-01 -9.96088624e-01 -1.05791795e+00 3.70229393e-01 5.77840269e-01 -1.22535288e-01 -9.91973698e-01 1.34325159e+00 9.32224929e-01 -4.04698163e-01 2.70775139e-01 -6.22498453e-01 7.97585621e-02 -2.20563397e-01 4.41310853e-01 -2.03241613e-02 -1.34710595e-01 -3.86392325e-01 -3.70006144e-01 4.80823576e-01 -2.25611344e-01 5.38543016e-02 1.37923491e+00 -1.81523502e-01 1.71979725e-01 2.95602739e-01 8.08778167e-01 -3.67392153e-01 -1.90541470e+00 -2.93769151e-01 -3.94776642e-01 -2.73269713e-01 2.81452954e-01 -1.01768482e+00 -1.52524149e+00 1.09291446e+00 5.64113796e-01 -3.77809227e-01 1.18598485e+00 2.98154444e-01 7.93107986e-01 1.76091880e-01 1.71136200e-01 -7.26541758e-01 1.53423354e-01 7.31404841e-01 9.46793020e-01 -1.58076012e+00 -2.72141010e-01 -3.52802157e-01 -8.13695669e-01 5.65233648e-01 1.08282995e+00 -9.91476849e-02 3.79973412e-01 4.56394613e-01 2.36236900e-01 -1.40508905e-01 -1.02592206e+00 -6.20632112e-01 3.94269615e-01 8.77221048e-01 4.76391017e-01 -1.53823541e-02 3.76117200e-01 5.35533905e-01 -2.19614074e-01 1.02284305e-01 3.28535914e-01 3.98818344e-01 -1.47939444e-01 -8.68951678e-01 -1.64825231e-01 7.10997701e-01 -3.06253999e-01 -5.64657032e-01 -3.00673038e-01 5.79737902e-01 2.66161174e-01 8.74691844e-01 9.05811489e-02 -3.66052955e-01 2.44939744e-01 -1.84191540e-01 6.47948623e-01 -7.31857419e-01 -6.27986372e-01 -1.32282332e-01 1.12413570e-01 -8.71931911e-01 -2.58195817e-01 -5.00420272e-01 -9.68969703e-01 -3.59591097e-01 -2.16667056e-01 -3.31259459e-01 3.24506968e-01 9.02544379e-01 3.40733320e-01 6.42004132e-01 4.48306389e-02 -1.00769985e+00 -3.92361075e-01 -1.16621518e+00 -1.62947670e-01 2.13496178e-01 4.95475262e-01 -6.62266076e-01 -1.43039301e-01 6.73654974e-02]
[9.760509490966797, 1.3476005792617798]
2e8a1b0c-e54c-4bbf-8134-56c6fe491ddd
icta-system-combination-for-chinese-semantic
null
null
https://aclanthology.org/S12-1073
https://aclanthology.org/S12-1073.pdf
ICT:A System Combination for Chinese Semantic Dependency Parsing
null
['Qun Liu', 'Hao Xiong']
2012-07-01
null
null
null
semeval-2012-7
['semantic-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]
[-7.270680904388428, 3.7627005577087402]
6ec5d7ae-fbca-4b72-9551-ae9066958c4d
towards-multi-domain-single-image-dehazing
null
null
http://openaccess.thecvf.com//content/CVPR2022/html/Liu_Towards_Multi-Domain_Single_Image_Dehazing_via_Test-Time_Training_CVPR_2022_paper.html
http://openaccess.thecvf.com//content/CVPR2022/papers/Liu_Towards_Multi-Domain_Single_Image_Dehazing_via_Test-Time_Training_CVPR_2022_paper.pdf
Towards Multi-Domain Single Image Dehazing via Test-Time Training
Recent years have witnessed significant progress in the area of single image dehazing, thanks to the employment of deep neural networks and diverse datasets. Most of the existing methods perform well when the training and testing are conducted on a single dataset. However, they are not able to handle different types of hazy images using a dehazing model trained on a particular dataset. One possible remedy is to perform training on multiple datasets jointly. However, we observe that this training strategy tends to compromise the model performance on individual datasets. Motivated by this observation, we propose a test-time training method which leverages a helper network to assist the dehazing model in better adapting to a domain of interest. Specifically, during the test time, the helper network evaluates the quality of the dehazing results, then directs the dehazing network to improve the quality by adjusting its parameters via self-supervision. Nevertheless, the inclusion of the helper network does not automatically ensure the desired performance improvement. For this reason, a meta-learning approach is employed to make the objectives of the dehazing and helper networks consistent with each other. We demonstrate the effectiveness of the proposed method by providing extensive supporting experiments.
['Keyan Wang', 'Jun Chen', 'Sadaf Salehkalaibar', 'Liangyan Li', 'Zijun Wu', 'Huan Liu']
2022-01-01
null
null
null
cvpr-2022-1
['image-dehazing']
['computer-vision']
[ 3.03718150e-01 -1.76557451e-01 -7.91959986e-02 -3.34220350e-01 -4.08018798e-01 -2.08963856e-01 5.04743934e-01 1.17223173e-01 -3.01780313e-01 4.77518827e-01 -1.41862154e-01 -9.13123041e-02 -1.70404673e-01 -8.67312849e-01 -6.52387977e-01 -1.00140142e+00 3.46373618e-01 1.44304588e-01 4.24721271e-01 -3.45080316e-01 3.47011566e-01 3.42117876e-01 -1.63428223e+00 1.31409481e-01 1.30317414e+00 1.01901948e+00 2.44434446e-01 5.77108920e-01 3.51069272e-01 7.88070679e-01 -9.07402039e-01 -3.71770531e-01 3.68153572e-01 -3.86818886e-01 -3.78503293e-01 3.42301399e-01 4.54934120e-01 -5.39872766e-01 -4.84521240e-02 1.10350168e+00 4.55255210e-01 2.54757613e-01 4.84184861e-01 -1.26386917e+00 -6.46919906e-01 2.44121045e-01 -3.92286569e-01 2.34692380e-01 -2.81570643e-01 2.30624482e-01 7.96462774e-01 -7.03934073e-01 2.11199835e-01 8.81604314e-01 4.37161118e-01 4.56182152e-01 -9.56005275e-01 -6.54565811e-01 1.12057775e-01 1.44733921e-01 -1.40691352e+00 -4.21326756e-01 9.97941971e-01 -2.30232298e-01 4.02359903e-01 -1.53732687e-01 3.40726405e-01 8.37387741e-01 1.05944246e-01 6.52707160e-01 1.03412747e+00 -4.03708965e-01 2.86307007e-01 5.49677849e-01 -2.76174862e-02 4.52254951e-01 3.50353420e-01 1.16065532e-01 -3.83750170e-01 1.15725294e-01 3.96585792e-01 9.28578824e-02 -2.19735533e-01 -4.74596679e-01 -7.99893975e-01 7.20653772e-01 5.06212115e-01 4.07962233e-01 -5.01945734e-01 -2.32716233e-01 1.33170232e-01 4.46395189e-01 5.85482895e-01 5.67622066e-01 -2.05651090e-01 2.27845728e-01 -1.11063564e+00 2.12652445e-01 4.12248731e-01 6.13853335e-01 1.00166523e+00 1.30751044e-01 -7.52462670e-02 8.17651033e-01 1.85075447e-01 1.85562596e-01 4.14494127e-01 -6.61070228e-01 4.67037380e-01 8.22529614e-01 1.41857639e-01 -1.14404619e+00 -1.29455730e-01 -6.44842088e-01 -7.80829966e-01 6.16264641e-01 2.43943498e-01 -1.14587046e-01 -1.09414089e+00 1.47929347e+00 4.95908797e-01 3.15033704e-01 1.72854438e-01 8.34832311e-01 3.25765729e-01 6.73945069e-01 -4.77945283e-02 5.61570488e-02 7.89757788e-01 -1.16258359e+00 -6.02583468e-01 -2.37180620e-01 7.03876674e-01 -5.99834800e-01 1.01582050e+00 5.36942899e-01 -9.95569885e-01 -7.59765267e-01 -1.51797342e+00 2.40886346e-01 -4.53436226e-01 7.08769783e-02 -1.37036905e-01 5.87092221e-01 -9.20469701e-01 5.82800567e-01 -6.13165259e-01 -1.21659182e-01 4.12255198e-01 3.60091567e-01 -8.39754045e-02 -3.12129587e-01 -1.14389217e+00 8.66187096e-01 7.31348395e-01 1.83243513e-01 -9.96086538e-01 -6.53770983e-01 -4.23142821e-01 4.85171638e-02 4.54323113e-01 -5.42109907e-01 8.94098163e-01 -1.39073479e+00 -1.38186669e+00 5.47198474e-01 4.02127475e-01 -6.21301293e-01 6.06064737e-01 -3.48522186e-01 -5.37436903e-01 2.63615578e-01 -7.97250792e-02 5.97117364e-01 1.29792964e+00 -1.53972852e+00 -9.29264843e-01 -3.28730494e-01 3.03819597e-01 2.12671578e-01 -6.54189169e-01 -1.98175505e-01 -4.19621378e-01 -8.81111264e-01 -2.83700317e-01 -6.91776037e-01 -3.42575833e-02 -1.08796626e-01 -2.77899563e-01 4.43349145e-02 1.06985092e+00 -5.52161932e-01 1.41146159e+00 -2.66176963e+00 1.64068699e-01 2.43622586e-01 2.36881703e-01 7.62360990e-01 -2.32648775e-01 2.22577035e-01 -2.72209439e-02 1.70964152e-01 -3.04405123e-01 -3.28030586e-01 -3.38535041e-01 2.29283199e-01 -1.64164230e-01 3.74444544e-01 5.18720984e-01 4.95816737e-01 -7.16025233e-01 -3.93798351e-01 2.76062310e-01 4.62182224e-01 -6.48904562e-01 6.01609826e-01 -3.09930354e-01 5.00614822e-01 -3.56362164e-01 5.27751029e-01 6.81574106e-01 -2.79689729e-01 -1.08987011e-01 -1.85643598e-01 -7.00953826e-02 8.22889432e-03 -1.17416608e+00 1.13811219e+00 -5.21244049e-01 3.20224166e-01 1.18630663e-01 -1.21189606e+00 8.61875176e-01 3.02140772e-01 3.72828752e-01 -8.20837557e-01 1.40984461e-01 3.55651408e-01 1.03985183e-01 -5.60490608e-01 4.62039888e-01 -1.36339530e-01 4.41017181e-01 5.46373367e-01 -8.64702389e-02 5.35319932e-02 2.15522230e-01 -5.16227707e-02 7.65471399e-01 -8.70074555e-02 1.44413248e-01 6.98196515e-02 6.94399655e-01 3.67642976e-02 4.17792439e-01 8.44514072e-01 -2.76116997e-01 6.70988023e-01 4.84393239e-02 -4.91521508e-01 -1.14198351e+00 -6.69025421e-01 -3.14467773e-02 1.20154464e+00 3.27213973e-01 -6.07533790e-02 -8.41303825e-01 -8.72642875e-01 -1.57018244e-01 7.11523712e-01 -8.17843378e-01 -5.94062448e-01 -4.91208404e-01 -7.77490377e-01 3.54444385e-01 1.38468802e-01 1.00400579e+00 -9.18356657e-01 -7.48248160e-01 2.08259895e-01 -6.98834332e-03 -9.88671899e-01 -2.98755229e-01 1.66701317e-01 -8.69374990e-01 -1.00002038e+00 -6.63080633e-01 -7.36039817e-01 7.09936500e-01 7.43304193e-01 7.57519603e-01 5.68626642e-01 2.24488214e-01 1.16812654e-01 -5.67326665e-01 -4.11732346e-01 -7.65590429e-01 3.55102867e-01 -1.93184093e-01 4.00008321e-01 1.28291920e-01 -6.50951147e-01 -6.33361459e-01 4.83452052e-01 -1.50523257e+00 -4.73577762e-03 6.94226980e-01 8.95847499e-01 2.83774197e-01 6.65023625e-01 5.71911871e-01 -8.65491271e-01 4.74034250e-01 -6.87402844e-01 -5.84406137e-01 3.41344476e-01 -9.62169766e-01 5.41740246e-02 7.55935788e-01 -6.12334609e-01 -1.01293671e+00 -3.60595614e-01 2.40165796e-02 -5.74507773e-01 -2.37727866e-01 6.77288473e-01 -1.83568448e-01 -1.44625634e-01 7.50376999e-01 2.27493346e-01 -7.83284288e-03 -2.41346270e-01 -7.21532330e-02 7.08968878e-01 5.29181361e-01 -3.19601685e-01 1.13783121e+00 4.67209280e-01 -3.75775605e-01 -4.60700989e-01 -1.06665254e+00 -2.65577942e-01 -5.56610882e-01 -2.82394022e-01 8.32756937e-01 -7.98335314e-01 -1.24841817e-01 6.83126807e-01 -8.57655704e-01 -4.11611587e-01 -5.08638797e-03 2.89106727e-01 -1.03157587e-01 1.10171460e-01 -6.28369674e-02 -7.54405141e-01 -1.70474723e-01 -1.15891016e+00 8.47749650e-01 2.67411739e-01 1.89599827e-01 -1.12888086e+00 -5.43306628e-03 4.42190826e-01 6.15544736e-01 1.61927879e-01 1.00415492e+00 -7.13054001e-01 -6.15758002e-01 -3.04807007e-01 -2.62641698e-01 7.61479795e-01 4.37926710e-01 1.44137874e-01 -9.55348432e-01 -3.78145069e-01 2.41820395e-01 -2.52517045e-01 8.00674856e-01 1.48118556e-01 1.10082126e+00 -2.45778218e-01 1.19547687e-01 6.25931501e-01 1.41032839e+00 3.00573498e-01 5.59842229e-01 1.00151491e+00 7.11400867e-01 6.42366707e-01 7.41273522e-01 2.98373908e-01 3.60116512e-01 6.02507770e-01 8.83274853e-01 -2.83575922e-01 -3.52079282e-03 6.05170708e-03 3.20495665e-01 3.87693971e-01 -1.20810670e-04 -4.97683108e-01 -8.30279529e-01 5.27110398e-01 -1.75275803e+00 -6.23543441e-01 3.87852609e-01 2.23029304e+00 6.55585110e-01 2.93336302e-01 1.31070226e-01 2.80144870e-01 6.90163732e-01 2.68921405e-01 -5.78458786e-01 -1.52209014e-01 -3.47892046e-02 -1.74012348e-01 3.54688764e-01 2.06881672e-01 -1.08099568e+00 8.50619912e-01 5.73682117e+00 6.22263730e-01 -1.40272498e+00 -5.26210293e-02 6.82053387e-01 6.79190271e-03 -2.11126119e-01 -1.25557885e-01 -6.09289706e-01 6.06419027e-01 9.04993713e-01 -9.63203460e-02 5.86751521e-01 6.69776797e-01 2.71084338e-01 -1.72177389e-01 -8.36785018e-01 6.16770685e-01 1.82604313e-01 -1.10357153e+00 2.29028419e-01 -1.20341778e-02 8.92787039e-01 -1.97759762e-01 3.12803656e-01 2.79259712e-01 8.93811285e-02 -8.36920023e-01 7.33317375e-01 3.30583572e-01 2.04879820e-01 -8.66859853e-01 9.78041291e-01 5.74877322e-01 -6.83806360e-01 -2.66125768e-01 -1.83587834e-01 2.28354663e-01 -1.46741927e-01 5.44062197e-01 -9.02860403e-01 7.46100187e-01 7.83889771e-01 5.93059480e-01 -9.13111567e-01 9.60140884e-01 -1.58397824e-01 7.18777239e-01 2.29125787e-02 4.94248331e-01 2.61039078e-01 -1.94607258e-01 3.61643285e-01 8.57544959e-01 2.91037709e-01 -1.31999701e-01 2.29717031e-01 5.41422009e-01 -1.18313961e-01 4.04445566e-02 -5.62042296e-01 -1.65915146e-01 4.67348933e-01 9.68142152e-01 -3.62888068e-01 -3.93679887e-01 -4.72775966e-01 7.27458715e-01 3.73978347e-01 3.77517968e-01 -9.19998765e-01 -2.77077943e-01 4.63840097e-01 1.63068354e-01 4.83099163e-01 -1.25067756e-01 -2.94587880e-01 -9.61683929e-01 1.21217750e-01 -1.13694823e+00 3.81666154e-01 -7.85526931e-01 -1.21432519e+00 5.97602665e-01 -4.93421964e-02 -1.41262138e+00 -4.66074906e-02 -3.36437047e-01 -6.49323046e-01 5.60262561e-01 -2.05910945e+00 -9.95216489e-01 -5.60591280e-01 6.53145492e-01 5.02267599e-01 -1.62484720e-01 2.78973103e-01 5.02762616e-01 -9.00016606e-01 6.23885036e-01 2.10759044e-03 -7.76875094e-02 1.00747156e+00 -1.09700763e+00 -6.42652735e-02 1.23773193e+00 -9.30098295e-02 5.19196332e-01 9.05051351e-01 -4.90181178e-01 -1.01837933e+00 -1.34820807e+00 4.57792014e-01 -1.29920632e-01 5.91305315e-01 9.99085233e-02 -1.36542094e+00 3.97208631e-01 2.81772494e-01 1.62829965e-01 4.61266667e-01 -2.35970125e-01 -3.25819135e-01 -3.89088690e-01 -1.07581306e+00 6.44541800e-01 5.84203660e-01 -3.75030845e-01 -4.29426670e-01 8.77070129e-02 6.54285133e-01 -4.21697587e-01 -7.74210155e-01 5.57265341e-01 2.30287865e-01 -1.19645846e+00 8.12412381e-01 -3.50895464e-01 5.99356234e-01 -4.76043761e-01 -8.69088843e-02 -1.41214621e+00 -1.43438116e-01 -1.69991106e-01 -1.82200700e-01 1.26874077e+00 2.50630319e-01 -5.06759822e-01 6.79070711e-01 4.44969624e-01 -1.71120420e-01 -6.69700265e-01 -4.91164505e-01 -6.32552981e-01 -9.16836932e-02 -2.06402585e-01 8.29927027e-01 9.76059675e-01 -7.09468722e-01 3.15566696e-02 -6.70770645e-01 6.05747759e-01 5.01128852e-01 -1.98287740e-02 9.79703307e-01 -1.13343048e+00 -2.12867290e-01 -2.31144086e-01 -3.06730270e-01 -7.31600344e-01 2.11815294e-02 -4.18314159e-01 2.71335423e-01 -1.05900931e+00 -1.05927072e-01 -5.18870652e-01 -6.01389527e-01 5.20072520e-01 -5.94684184e-01 2.86427617e-01 2.09053993e-01 3.15243393e-01 -6.02652371e-01 5.41541040e-01 1.28994572e+00 -2.93269485e-01 -4.11251396e-01 2.10592434e-01 -8.92502189e-01 5.29585063e-01 9.80741858e-01 -6.36313975e-01 -6.65095031e-01 -6.19513392e-01 1.39669225e-01 -3.05317461e-01 4.47893232e-01 -1.20103800e+00 3.09374809e-01 -2.60988981e-01 1.91750079e-01 -3.36239129e-01 2.29231231e-02 -1.00202119e+00 -4.90628816e-02 1.77732497e-01 -3.06080312e-01 5.88586070e-02 1.53147662e-02 5.39859414e-01 -5.23670316e-01 -3.52661639e-01 1.10322452e+00 4.81640771e-02 -7.28571057e-01 2.82076716e-01 -2.79461473e-01 -8.77060965e-02 1.01239526e+00 -5.30501485e-01 -2.50968754e-01 -2.72159517e-01 -3.75206769e-01 2.05550671e-01 6.96018040e-01 3.59100491e-01 5.23939908e-01 -1.00490069e+00 -5.86410284e-01 2.74924278e-01 2.40981728e-01 2.47715473e-01 2.08338082e-01 8.41199040e-01 -2.67139822e-01 -5.37321456e-02 -2.44733989e-01 -5.93637824e-01 -1.13299930e+00 6.64393842e-01 5.62129617e-01 -3.36609274e-01 -4.48605657e-01 4.44790721e-01 2.15848729e-01 -2.10012227e-01 2.62706161e-01 4.15998921e-02 -3.21601987e-01 4.92722075e-03 6.47847354e-01 2.52377003e-01 3.86846274e-01 -3.52936834e-01 -2.71324757e-02 2.39336655e-01 -4.07357901e-01 3.13848071e-02 1.58299613e+00 -3.66944432e-01 9.18170959e-02 2.63855666e-01 1.00381398e+00 -7.75234252e-02 -1.55646908e+00 -3.78690392e-01 -3.12905423e-02 -6.20495737e-01 2.83495158e-01 -6.54647589e-01 -1.33984017e+00 7.64567614e-01 5.87639630e-01 3.80300134e-01 1.64261556e+00 -4.36210692e-01 6.00915253e-01 3.50762814e-01 1.01743445e-01 -1.13051867e+00 4.41902936e-01 2.96897948e-01 5.84335327e-01 -1.38187802e+00 -6.80052936e-02 -1.23307668e-01 -7.60480106e-01 1.03626680e+00 8.40832412e-01 -1.64169073e-01 5.03089666e-01 -3.36828157e-02 2.85559028e-01 -1.35608032e-01 -7.17077374e-01 -4.41190116e-02 3.04843694e-01 3.93515766e-01 2.40125936e-02 -4.12116051e-01 2.19159082e-01 4.51662987e-02 4.01635394e-02 1.22898601e-01 5.13950408e-01 1.16538429e+00 -5.74544370e-01 -1.11808157e+00 -4.68280047e-01 2.28483781e-01 -3.07193130e-01 1.03237800e-01 -2.67486840e-01 8.41324389e-01 2.80219942e-01 1.13568711e+00 -1.54049948e-01 -4.50070888e-01 3.65591645e-01 -3.80705521e-02 1.69824749e-01 -6.56025052e-01 -7.65992463e-01 -5.42276911e-02 -2.68987149e-01 -2.58673877e-01 -6.66202366e-01 -3.47229421e-01 -8.09041679e-01 -3.09297264e-01 -4.44677025e-01 2.02048063e-01 4.09692138e-01 1.27974498e+00 3.62685770e-01 7.40250528e-01 1.09355879e+00 -6.44803345e-01 -5.90665936e-01 -8.02782118e-01 -5.76653779e-01 4.57981110e-01 6.97954834e-01 -7.54038692e-01 -5.18050253e-01 6.29828647e-02]
[10.944283485412598, -3.0267539024353027]
c7889580-743e-45ea-b39a-075cf42dad98
enabling-weakly-supervised-temporal-action
2208.12673
null
https://arxiv.org/abs/2208.12673v1
https://arxiv.org/pdf/2208.12673v1.pdf
Enabling Weakly-Supervised Temporal Action Localization from On-Device Learning of the Video Stream
Detecting actions in videos have been widely applied in on-device applications. Practical on-device videos are always untrimmed with both action and background. It is desirable for a model to both recognize the class of action and localize the temporal position where the action happens. Such a task is called temporal action location (TAL), which is always trained on the cloud where multiple untrimmed videos are collected and labeled. It is desirable for a TAL model to continuously and locally learn from new data, which can directly improve the action detection precision while protecting customers' privacy. However, it is non-trivial to train a TAL model, since tremendous video samples with temporal annotations are required. However, annotating videos frame by frame is exorbitantly time-consuming and expensive. Although weakly-supervised TAL (W-TAL) has been proposed to learn from untrimmed videos with only video-level labels, such an approach is also not suitable for on-device learning scenarios. In practical on-device learning applications, data are collected in streaming. Dividing such a long video stream into multiple video segments requires lots of human effort, which hinders the exploration of applying the TAL tasks to realistic on-device learning applications. To enable W-TAL models to learn from a long, untrimmed streaming video, we propose an efficient video learning approach that can directly adapt to new environments. We first propose a self-adaptive video dividing approach with a contrast score-based segment merging approach to convert the video stream into multiple segments. Then, we explore different sampling strategies on the TAL tasks to request as few labels as possible. To the best of our knowledge, we are the first attempt to directly learn from the on-device, long video stream.
['Jingtong Hu', 'Peipei Zhou', 'Yawen Wu', 'Yue Tang']
2022-08-25
null
null
null
null
['weakly-supervised-temporal-action', 'action-localization']
['computer-vision', 'computer-vision']
[ 0.37980536 -0.19723947 -0.5666264 -0.28603348 -0.7064961 -0.60501426 0.02747823 -0.17954212 -0.29059818 0.3299041 -0.19973788 -0.24200776 0.12039312 -0.414344 -0.8807798 -0.658428 -0.22102194 0.04155475 0.5060434 0.35870153 -0.06121735 0.32716718 -1.4703604 0.6325117 0.55480736 1.3480175 0.3300662 0.7160167 0.01542797 1.0231411 -0.55862105 -0.30413523 0.48131436 -0.45634648 -0.5663152 0.6956697 0.47857258 -0.6038932 -0.42967752 0.85487545 0.24668328 0.07615774 0.26568168 -1.4106659 -0.20687288 0.34253332 -0.65429056 0.27426127 0.45417944 0.14629334 0.6538646 -0.51128066 0.47988665 0.8597268 0.52225494 0.49480152 -0.8435602 -0.7614538 0.49133512 0.40020344 -1.189147 -0.45199957 0.9400043 -0.44154584 0.36662567 0.2867079 0.6925522 1.1079625 -0.21775353 1.4178264 0.75925004 -0.21726757 0.43820676 0.24030258 -0.3453583 0.5575806 -0.23244871 -0.37550583 -0.6807272 0.0848946 0.7552052 0.60040486 -0.37818816 -0.5354705 -1.1014603 0.40958482 0.1600881 0.23961379 -0.37821513 0.03712492 0.6640266 0.4598328 0.41733944 0.20883715 -0.62672466 -0.5566256 -1.2242969 -0.13709883 0.594847 1.1648852 0.6119353 -0.02563436 -0.03455029 0.5469336 -0.23847304 0.304234 0.3530151 -1.0144418 0.58202255 0.3914788 0.1646772 -0.9964111 0.045689 -0.18449067 -0.6956956 -0.08544098 0.50924665 -0.25525218 -0.6836701 1.4596645 0.36993748 0.871404 -0.2775383 0.8211364 0.3002616 0.7618323 -0.05607104 -0.7150899 1.0382205 -1.092916 -0.627334 -0.02956529 0.8017479 -0.51994765 1.2328067 0.7737078 -0.8911249 -0.70024216 -0.9295063 0.2002556 -0.22423773 0.23255304 0.4888177 0.598728 -0.4885338 0.43689516 -0.93662566 -0.25006494 0.7939266 0.44515604 -0.41427416 -0.4104567 -0.8303024 0.23457041 0.29902104 -0.04450204 -1.0525349 -0.4502972 -0.72730994 -0.09282483 1.0668583 -0.10236188 1.3614353 -1.6983265 -1.6384646 0.5294021 -0.20558137 -0.3334907 0.5764681 -0.21967196 -0.5452756 0.41346866 -0.1555018 0.35034102 1.4228613 -1.211102 -0.96711683 -0.241908 0.40976864 0.11674 -0.79836863 0.09981265 -0.85148984 -0.6375714 -0.19418177 -0.93517476 0.10165901 0.19138303 -0.02668576 -0.14910088 1.4077479 -0.74644154 1.2853411 -2.4823477 -0.14235812 -0.02330445 0.12456258 0.6197207 -0.10019404 0.06859857 0.18266724 -0.06063356 0.2608683 -0.36702818 -0.29579365 0.23042867 -0.42100915 0.3202434 0.12198037 0.6075674 -1.103855 -0.48843268 0.24201974 0.32958102 -0.7021443 0.40800342 -0.30686688 0.55679744 -0.60266006 0.9064157 0.61435425 -0.27141258 0.17511557 -0.3239145 0.09419858 -0.07616012 -1.0959088 1.6621097 -0.5898504 0.5780364 0.11345241 -1.2553086 0.65601784 0.5250231 0.933342 -0.6064302 0.00738182 0.11421215 -0.27477965 -0.9478982 0.2721069 -0.07725034 -0.02645826 0.40443793 0.00602144 0.43266705 0.1625024 0.04900429 1.3322045 0.33646262 0.08417419 0.33926708 0.5314597 -0.2920968 0.8632341 0.5704077 -0.36762092 0.61993295 0.45463192 -0.5470208 -0.72106117 -0.8611647 0.17655057 1.2887592 0.26460397 -0.53648615 -0.79738677 -1.1940287 -0.2709916 0.30683205 -0.16599923 -0.14180736 -0.6290346 -0.24632116 0.26505136 0.58466417 0.49091905 -0.95941764 -0.7539895 0.25945103 -0.25035363 -1.3179927 -0.8105154 0.01179174 -0.75217146 -1.1198282 -0.56256783 -0.7904409 0.7567721 0.63287884 0.71076685 -0.10457116 -0.01939931 0.5324937 -0.66784745 -0.28596985 -0.12089156 0.05527213 0.18339534 0.7589892 0.54764724 -0.5514862 -0.607841 0.6520324 -1.0924932 -0.01819305 0.5852411 0.56158984 0.71978587 0.55646694 0.6320774 -0.7533933 -0.00917719 -0.48518977 -0.5788516 0.40755117 -0.07628706 -0.44301647 0.9803677 -1.0653521 -0.85331076 0.38713592 0.18511832 -1.0959969 -0.18473455 0.23409237 -0.47147125 0.01732316 0.24125011 0.23709595 -0.12050385 -0.36209708 0.13804051 1.0383505 0.35074914 -0.40953574 0.66815895 0.48815084 -0.29677022 -0.8865704 -0.79266584 -0.8080071 -0.7617447 -0.45864922 0.85158545 -1.0425124 -0.79988396 0.43002042 -0.9137275 -0.42431566 -0.34218165 0.56923693 -0.69504285 0.49583575 -0.33961788 -0.7354283 0.00873508 -1.1665297 1.1055919 0.0787368 -0.04064006 -0.85802937 -0.27381927 0.6105037 0.13596104 0.00703414 0.31514755 -0.49502873 -0.85596883 -0.55602324 -0.03279745 0.50635594 0.48032326 -0.10839944 -0.9194345 -0.49998567 0.24920374 -0.55892247 0.58939654 0.19358443 1.7840463 -0.42221564 -0.34683552 0.6040846 1.0374143 0.4712908 0.5001355 0.12712063 0.90714747 0.37182036 1.155998 0.56248754 0.19213963 0.88242775 0.37713253 -0.01479967 0.12937902 -0.38868532 0.88707316 0.7326372 -0.05104638 -0.46476045 -0.5779163 0.51031214 -2.0266128 -1.112226 0.46855628 2.4461849 0.65106624 0.11311958 0.36259705 0.13736321 0.679714 0.15073746 -0.8259233 0.01237218 0.2903678 -0.11310944 0.40713197 0.07830032 -1.4437661 0.64133716 4.96357 1.0560054 -1.3773261 0.30972788 0.72456276 -0.49042344 0.09577568 -0.04278705 -0.7077991 0.82710046 0.87726945 0.17733124 0.44448406 1.017216 0.44430733 0.03328385 -1.3647557 1.472964 0.2660217 -1.0826734 -0.18602572 0.05437936 0.72544146 -0.25504574 -0.04010833 0.43179473 -0.2535399 -0.63972175 0.65741795 0.22230853 0.88586867 -0.560763 0.673398 0.6190983 -1.4096318 -0.47662053 -0.17717685 0.09259284 0.3625402 0.49325502 -0.63887715 0.26423657 0.864984 1.1048987 -0.3814179 0.984046 0.12597865 0.7314972 -0.23929572 0.38697568 0.09945364 -0.07615692 0.24541791 1.0059232 0.6351356 0.16174322 0.70802075 0.12638038 -0.27479982 0.1174994 -0.7477269 -0.12298799 0.32184207 1.152627 -0.6554113 -0.34347987 -0.88222826 1.1961831 0.04445664 0.2901643 -1.0093462 0.00852759 0.527635 0.55732214 0.47417986 -0.2213149 0.26687732 -1.4223863 0.33345047 -1.0840704 0.4998274 -0.61850953 -1.1216166 0.38124087 -0.15261316 -1.8733933 -0.17293249 -0.46012196 -0.43702412 0.08074299 -1.2752523 -1.0861751 -0.36636364 0.9541892 0.9296433 -0.2221492 0.4207616 0.66882026 -0.6643193 0.8035436 -0.02451786 0.12016314 0.8470724 -0.8249256 -0.20970927 0.864025 0.3291421 0.2456819 0.38630733 -0.5381582 -1.4638005 -1.3918198 0.40321258 -0.26234666 0.6633698 -0.41429195 -1.0139476 0.8626229 0.03009195 0.44907013 0.72091573 -0.22221164 -0.19672944 -0.6063272 -1.01598 0.49346656 1.1685568 -0.69279504 -0.12807539 0.6029538 0.5828617 -0.15558755 -0.76908576 0.34803075 0.40421984 -0.78689027 0.72434473 -0.5318506 0.1140067 -0.5307703 -0.14167953 -0.8532499 0.04731464 -0.93720883 -0.5412131 1.3444302 0.01593636 -0.28626698 1.0737717 0.6382881 0.01067052 -0.71082073 -0.89456934 -0.9837648 -0.55091244 -0.58612794 0.34276557 0.9054343 0.02478457 0.01512522 -0.9029405 0.24978964 0.40312204 0.06300148 0.9478161 -0.8076671 -0.6650328 0.01259984 -0.56651545 -1.5427712 0.10510187 -0.48110983 0.10696468 -0.8855502 0.11750641 -0.5053907 -0.52344227 0.4905023 0.13467708 0.21325797 0.23450862 0.2795023 -1.1477467 0.45522568 0.9444834 -0.25708148 -0.3602386 0.5625841 -0.2895603 0.80675226 0.8673355 -0.44273216 -0.9067647 -0.37440357 -0.01791081 0.19468258 0.11597336 -1.0287886 0.18165232 -0.39778742 0.02871392 -0.5365503 0.34979725 -1.281258 0.14894968 0.19115679 -0.18720265 -0.18511066 -0.17507194 0.7446216 -0.48577484 -0.17632657 0.5621994 -0.08465233 -0.9034695 0.73082626 -0.28814563 -0.2311909 1.5006444 -0.43214387 0.05121065 -0.501991 -0.63447976 0.24741772 0.5819667 0.57267827 0.640853 -1.3042912 -0.15288411 0.28552425 0.16075727 0.09251025 0.40865 0.8792496 -0.2554395 0.03167832 -0.02985363 -0.7026132 -1.511266 1.1403333 0.03980708 -0.17557204 -0.6014429 0.6110582 0.2888977 0.06564227 0.5257585 -0.24702345 -0.04912958 0.10795079 0.7291699 0.28440353 -0.10366056 -0.5222689 -0.09664522 0.63204014 -0.21442135 0.34228528 1.1463457 -0.28210473 0.22264996 0.5536522 1.3759134 0.02197494 -1.8935698 -0.36646682 -0.14513946 -0.854725 -0.1154063 -0.24680756 -1.4217914 0.9964852 0.7125418 0.17263651 1.5887864 -0.09096958 1.282446 0.36135986 0.6706551 -1.3093431 0.64291346 -0.02800733 0.43009079 -1.3009037 -0.13699503 -0.40700668 -0.7320432 0.9831482 0.62659276 0.32949063 0.8237399 0.22272867 0.05177324 0.15516496 -0.5769182 -0.03087192 0.13584316 0.5858357 -0.11800069 -0.01274161 0.26739407 0.3864806 0.39133516 0.36720207 0.47904485 1.0664967 -0.26759797 -1.222402 -0.23888691 0.47893584 -0.5530566 0.35762885 -0.01828217 0.3796351 0.5299685 1.0107199 0.07123385 -0.6127183 0.20115218 -0.0351773 0.19771191 -0.6082237 -0.12624139 0.18583605 -0.13988727 -0.72314566 -0.727211 -0.9753018 -0.98974854 -0.20227848 -0.22847725 -0.10694307 0.36174715 0.91614515 0.44323415 0.3486062 1.0049257 -0.966098 -0.24336256 -0.56678367 -0.5747604 0.5696366 0.44322097 -0.53846544 -0.2098919 0.44926387]
[8.552701950073242, 0.6540181040763855]
7d3a9468-2f6b-4ec8-9d31-79178ca24317
enhancing-transformer-for-end-to-end-speech
null
null
https://aclanthology.org/W19-6603
https://aclanthology.org/W19-6603.pdf
Enhancing Transformer for End-to-end Speech-to-Text Translation
null
['Roldano Cattoni', 'Roberto Dessi', 'Mattia Antonino Di Gangi', 'Marco Turchi', 'Matteo Negri']
2019-08-01
null
null
null
ws-2019-8
['speech-to-text-translation']
['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]
[-7.384832859039307, 3.748279333114624]
c91ed3ee-3c89-4a13-99c3-d857999cb8fe
evolving-domain-generalization
2206.00047
null
https://arxiv.org/abs/2206.00047v2
https://arxiv.org/pdf/2206.00047v2.pdf
Evolving Domain Generalization
Domain generalization aims to learn a predictive model from multiple different but related source tasks that can generalize well to a target task without the need of accessing any target data. Existing domain generalization methods ignore the relationship between tasks, implicitly assuming that all the tasks are sampled from a stationary environment. Therefore, they can fail when deployed in an evolving environment. To this end, we formulate and study the \emph{evolving domain generalization} (EDG) scenario, which exploits not only the source data but also their evolving pattern to generate a model for the unseen task. Our theoretical result reveals the benefits of modeling the relation between two consecutive tasks by learning a globally consistent directional mapping function. In practice, our analysis also suggests solving the DDG problem in a meta-learning manner, which leads to \emph{directional prototypical network}, the first method for the DDG problem. Empirical evaluation of both synthetic and real-world data sets validates the effectiveness of our approach.
['William Wei Wang', 'Boyu Wang', 'Christian Gagné', 'Charles Ling', 'Changjian Shui', 'Fan Zhou', 'Jiaqi Li', 'Ruizhi Pu', 'Gezheng Xu']
2022-05-31
null
null
null
null
['evolving-domain-generalization']
['computer-vision']
[ 5.46667099e-01 7.12455958e-02 -3.18618976e-02 -3.59995633e-01 -9.17003751e-02 -7.00491130e-01 8.38555098e-01 -3.27251386e-04 -1.74674913e-01 9.40823197e-01 -5.68479896e-02 -9.39924270e-02 -5.57875991e-01 -5.78284979e-01 -7.05580354e-01 -7.87506282e-01 -2.21200392e-01 6.38383687e-01 2.91240782e-01 -4.14833069e-01 1.49195075e-01 2.21923366e-01 -1.60118747e+00 1.47983387e-01 1.47166765e+00 7.31976390e-01 6.19001567e-01 3.04494381e-01 -2.67004110e-02 3.20126474e-01 -8.88750911e-01 -3.26109469e-01 4.12614912e-01 -5.79146028e-01 -8.50028038e-01 5.43550067e-02 2.05839097e-01 1.63876638e-02 -6.10866435e-02 1.01140451e+00 3.79119366e-01 5.19452095e-01 7.47469366e-01 -1.53804469e+00 -7.48503685e-01 5.52983105e-01 -2.03221500e-01 3.63157153e-01 2.12389082e-02 -7.97900371e-03 6.46528184e-01 -4.84190971e-01 7.96050072e-01 1.00557923e+00 5.42852342e-01 1.02672172e+00 -1.46660411e+00 -7.53873765e-01 6.22411668e-01 1.26758695e-01 -1.16689825e+00 -1.01894185e-01 1.02903914e+00 -4.72294092e-01 6.42553389e-01 -1.26733584e-02 4.66155857e-01 1.73820245e+00 1.39498949e-01 5.70545256e-01 1.23609400e+00 -2.29076341e-01 5.01413584e-01 3.78729850e-01 1.94153890e-01 4.01880443e-02 5.03066659e-01 1.98252037e-01 -8.09834242e-01 -2.41612002e-01 3.74085724e-01 -9.92934182e-02 -3.63082767e-01 -5.03746331e-01 -1.03343725e+00 4.50302273e-01 2.26969644e-01 3.81071448e-01 -5.33811450e-01 -2.12739885e-01 4.32977945e-01 7.81477451e-01 7.95728564e-01 5.65229595e-01 -6.25464797e-01 1.80373807e-02 -6.29875779e-01 3.40457201e-01 8.28146636e-01 1.16637886e+00 8.20855260e-01 1.38098836e-01 5.12532257e-02 6.64359093e-01 -2.33730897e-01 1.22909397e-01 7.52857447e-01 -6.99784100e-01 5.25878489e-01 5.69545269e-01 1.62370294e-01 -5.92230260e-01 -4.03115630e-01 -8.18309844e-01 -7.16984510e-01 -1.30180214e-02 4.28147376e-01 -4.45501417e-01 -7.07414746e-01 2.31764030e+00 3.34130436e-01 5.03992140e-01 4.53942895e-01 5.66111088e-01 1.85348392e-01 4.72759634e-01 2.04288110e-01 -3.53483468e-01 7.34221876e-01 -7.53474176e-01 -4.41133231e-01 -6.38229132e-01 5.23522496e-01 -4.48381640e-02 1.23481989e+00 4.50998455e-01 -6.53794289e-01 -7.76631713e-01 -1.15217865e+00 4.11803544e-01 -3.82240206e-01 -4.24938351e-01 2.71666169e-01 5.79736531e-01 -1.05437636e+00 6.70307219e-01 -5.97993910e-01 -6.52722359e-01 2.87323207e-01 2.89167494e-01 -1.61252096e-01 -1.24062963e-01 -1.21842110e+00 8.14805508e-01 9.15787697e-01 -2.08270252e-01 -1.04553998e+00 -8.87999713e-01 -2.97221750e-01 5.30381203e-02 4.72408950e-01 -7.83077180e-01 1.02971971e+00 -1.33782089e+00 -1.12901056e+00 6.33811116e-01 -2.06743807e-01 -4.79093522e-01 5.74753881e-01 -1.78170219e-01 -5.87143719e-01 -3.46617848e-01 5.12911789e-02 4.20639336e-01 9.86783326e-01 -1.63706708e+00 -9.03731227e-01 -6.10911369e-01 1.00260697e-01 4.21080500e-01 -7.90099025e-01 -6.20983422e-01 1.59217149e-01 -7.03448534e-01 1.97887644e-02 -9.90233302e-01 1.10644236e-01 -5.64778328e-01 -1.72099352e-01 -5.44790983e-01 1.02666938e+00 -5.34428716e-01 1.37470818e+00 -2.28515244e+00 5.64607441e-01 1.46263167e-01 2.97412902e-01 7.77183250e-02 -2.52613872e-01 4.25586462e-01 -2.80696452e-01 1.40991345e-01 -2.50886023e-01 -3.03569764e-01 -2.31184497e-01 4.38504189e-01 -3.64928782e-01 5.36928233e-03 -7.07271025e-02 6.16959631e-01 -1.15308130e+00 -2.81228900e-01 -4.88671064e-01 9.19294953e-02 -2.53306627e-01 1.32911637e-01 -4.34722990e-01 7.59229481e-01 -5.99196732e-01 1.39264956e-01 5.87761223e-01 -2.02749610e-01 6.45655990e-01 1.84403673e-01 1.84221819e-01 1.43078253e-01 -1.03002775e+00 1.83687687e+00 -6.32647872e-01 5.35845280e-01 -8.99533257e-02 -1.35942793e+00 1.30497265e+00 1.91316590e-01 3.07817250e-01 -5.76709092e-01 -3.05257708e-01 4.86693859e-01 2.65889585e-01 -3.64382595e-01 3.20720762e-01 -3.43832910e-01 -8.06343257e-02 6.37784660e-01 2.51990467e-01 2.09072158e-01 1.61065727e-01 2.60749832e-03 9.40595567e-01 3.25684905e-01 1.62517011e-01 -5.72496355e-01 3.20267618e-01 9.64721143e-02 6.61132216e-01 9.12957191e-01 -6.68636858e-02 1.35072470e-01 4.66458887e-01 -4.62303698e-01 -9.17002320e-01 -1.20214415e+00 -5.61741330e-02 1.35124362e+00 3.49549502e-01 -8.00586045e-02 -6.80402577e-01 -9.64484215e-01 2.58364100e-02 1.10881543e+00 -8.68369997e-01 -7.03673601e-01 -7.32289851e-01 -5.77696919e-01 2.69325465e-01 4.07368958e-01 5.72075367e-01 -9.80650783e-01 -8.16035390e-01 1.95564777e-01 -3.00722152e-01 -7.53136098e-01 -1.83106557e-01 4.80827630e-01 -1.07910419e+00 -9.92111206e-01 -6.43633723e-01 -8.37556005e-01 4.75990593e-01 2.09039807e-01 1.04913855e+00 -5.62794507e-02 2.59571314e-01 3.23098153e-01 -1.96722031e-01 -4.94442999e-01 -7.03768611e-01 5.64857662e-01 3.86202455e-01 -1.37395076e-02 5.32424986e-01 -1.09394372e+00 -3.58250767e-01 4.70846444e-01 -9.80876505e-01 -7.94860050e-02 3.99627090e-01 9.52072144e-01 4.16787088e-01 5.17975867e-01 1.28731036e+00 -1.04307711e+00 1.21877015e+00 -8.67136598e-01 -3.32992166e-01 6.79191768e-01 -9.96835053e-01 1.71184927e-01 8.98334563e-01 -8.94417763e-01 -1.51083899e+00 -6.34042174e-02 6.18853569e-01 -3.39952856e-01 -2.50318915e-01 5.08866847e-01 -2.59203643e-01 4.19492364e-01 1.15254915e+00 5.79871893e-01 -2.28346400e-02 -4.39034611e-01 1.48305684e-01 5.27009904e-01 5.80995381e-01 -9.50206161e-01 8.86265874e-01 4.60083306e-01 5.33332974e-02 -5.91329634e-01 -8.70255947e-01 -5.15332073e-02 -9.42898810e-01 -1.92036271e-01 3.88866305e-01 -5.82284868e-01 -3.46722573e-01 4.69258070e-01 -1.00228620e+00 -6.26151264e-01 -4.15951878e-01 2.94463247e-01 -6.00816786e-01 1.11704029e-01 1.29215062e-01 -7.78183520e-01 -6.82184771e-02 -7.31887996e-01 6.72143638e-01 2.48736754e-01 -3.10039312e-01 -1.44389582e+00 5.31901792e-02 -6.91130757e-02 5.32333612e-01 3.39811832e-01 1.42200267e+00 -1.17028570e+00 -1.81571022e-01 1.72750026e-01 2.19096825e-01 4.16326731e-01 4.97094095e-01 -5.75837195e-01 -9.17467177e-01 -4.04005975e-01 3.82525235e-01 -2.06132144e-01 6.51504159e-01 1.41315103e-01 1.03849113e+00 -2.66560704e-01 -6.39080226e-01 4.73004401e-01 1.27491164e+00 5.62476337e-01 2.20659912e-01 5.43454230e-01 3.30508977e-01 8.99390578e-01 6.86892152e-01 3.34476888e-01 1.67913988e-01 5.02718568e-01 3.93318832e-01 2.97891915e-01 -3.25582922e-02 -3.88122052e-01 2.25016296e-01 3.18382621e-01 -1.05470337e-01 -4.26129937e-01 -1.13933849e+00 7.34072208e-01 -1.88157189e+00 -9.71364558e-01 2.57917672e-01 2.39540839e+00 8.90724182e-01 1.89500660e-01 2.58263022e-01 -1.65828347e-01 9.49989140e-01 -1.16292849e-01 -1.13900471e+00 -1.12778388e-01 -1.62267670e-01 1.25764340e-01 -6.82768822e-02 1.84381723e-01 -7.25748956e-01 8.07824969e-01 5.86558819e+00 5.93563616e-01 -1.25346613e+00 1.30085573e-01 4.21844125e-01 -1.11022599e-01 -5.08619606e-01 2.32131965e-02 -6.58286333e-01 5.52817166e-01 7.78157592e-01 -9.45963502e-01 6.09710217e-01 9.18946326e-01 -1.61313072e-01 1.47957385e-01 -1.47128177e+00 5.33993542e-01 -2.43198052e-02 -8.84231746e-01 1.43943071e-01 7.51290889e-03 8.32332373e-01 -1.94912508e-01 3.33006054e-01 3.43891174e-01 4.70039129e-01 -7.73569405e-01 6.55227959e-01 5.00384867e-01 6.20487273e-01 -6.40734375e-01 4.04827595e-01 8.91249359e-01 -9.67066169e-01 -5.29986739e-01 -3.52068931e-01 8.50631595e-02 -2.60466076e-02 1.89389050e-01 -1.15550995e+00 9.07501519e-01 8.06582212e-01 7.05535650e-01 -6.66116238e-01 7.50386775e-01 -8.14264566e-02 4.95791912e-01 -2.00219721e-01 9.35144499e-02 -8.52196198e-03 -1.84548959e-01 7.70534098e-01 7.96052575e-01 6.84594810e-01 -1.47983074e-01 1.12685330e-01 8.76574934e-01 1.47571877e-01 -2.38519609e-01 -9.28391039e-01 7.64532536e-02 8.30998421e-01 7.16552556e-01 -5.48154414e-01 -1.13478735e-01 -5.71559779e-02 9.56208825e-01 4.92120206e-01 8.95842791e-01 -7.40354240e-01 -2.23397374e-01 5.85291207e-01 1.57333016e-01 7.68609270e-02 -1.91866502e-01 -4.00100499e-01 -9.01212573e-01 2.88779616e-01 -9.37959850e-01 5.23495674e-01 -5.49333692e-01 -1.54570103e+00 8.78182709e-01 4.77258682e-01 -1.20761311e+00 -5.20576775e-01 -2.59033799e-01 -6.03683889e-01 1.00462079e+00 -1.49474585e+00 -8.75408649e-01 -6.58330739e-01 9.41196203e-01 6.88904107e-01 -4.22936916e-01 5.75879097e-01 -8.20414647e-02 -4.52522010e-01 4.16879416e-01 2.06273332e-01 -4.75526214e-01 7.55688727e-01 -1.22433972e+00 3.13748837e-01 8.09874713e-01 -1.94831699e-01 6.82600856e-01 7.79989898e-01 -9.39279199e-01 -8.55695784e-01 -1.19990456e+00 6.93445921e-01 -4.94263977e-01 6.38297379e-01 -5.93597412e-01 -1.26790261e+00 7.92326927e-01 8.43549445e-02 -4.08311695e-01 4.54350680e-01 2.55219877e-01 -2.98592180e-01 -3.31348717e-01 -1.02949250e+00 5.65503359e-01 1.75698709e+00 -3.03223014e-01 -6.73576653e-01 3.53215009e-01 8.40533972e-01 -1.43541887e-01 -7.32436359e-01 2.22386107e-01 1.57318011e-01 -9.43387270e-01 7.43252635e-01 -9.90225792e-01 3.55240911e-01 -1.07840542e-02 -1.47054538e-01 -1.96640611e+00 -3.06796819e-01 -6.03564262e-01 -2.72922873e-01 1.26914680e+00 5.05654693e-01 -9.16088045e-01 5.76128066e-01 5.08376718e-01 -1.65972501e-01 -3.69070470e-01 -1.01052797e+00 -1.44672871e+00 4.17306632e-01 -1.63515940e-01 9.20611501e-01 1.15947950e+00 -9.10846964e-02 5.79501569e-01 -2.89803445e-01 2.12190419e-01 7.02004731e-01 1.23471767e-01 6.86735213e-01 -1.86869860e+00 -2.43646577e-01 -3.27699035e-01 -3.21055669e-03 -9.55688059e-01 6.86012864e-01 -9.80456233e-01 -1.58587322e-01 -1.22363031e+00 -4.88155968e-02 -8.83475304e-01 -2.91203707e-01 2.54205793e-01 -2.87616193e-01 -5.70654869e-01 -2.18139477e-02 4.42722440e-01 -2.59988219e-01 4.89699274e-01 1.16543210e+00 -7.98602328e-02 -5.09277523e-01 3.19600642e-01 -1.00086427e+00 4.31020647e-01 1.03254306e+00 -6.96747839e-01 -1.27805495e+00 -7.21342146e-01 1.41938061e-01 -2.91027278e-01 2.90971905e-01 -9.24318135e-01 2.74202824e-01 -3.72536212e-01 1.97483957e-01 -1.54080167e-02 2.60987282e-01 -9.59475279e-01 3.53338093e-01 4.32034582e-01 -3.54875714e-01 1.54101327e-01 1.19011901e-01 9.31707084e-01 -4.89337929e-02 -3.31524432e-01 6.05559468e-01 -6.13367371e-02 -9.37739849e-01 1.90684333e-01 8.22663233e-02 3.42489451e-01 1.40559232e+00 -4.91017908e-01 -6.53317690e-01 -3.56413484e-01 -9.35132384e-01 3.53593230e-01 4.25259888e-01 7.03138292e-01 3.49801123e-01 -1.06135976e+00 -7.90935755e-01 2.01663226e-01 1.57158867e-01 1.29323825e-01 3.46664906e-01 4.99021262e-01 2.25333154e-01 1.84269592e-01 -3.80978495e-01 -6.89248860e-01 -8.54758203e-01 7.81215847e-01 4.93936837e-01 -2.50944793e-01 -4.97885942e-01 6.59022868e-01 4.62895304e-01 -3.67001057e-01 1.46360829e-01 9.96449143e-02 -1.65169120e-01 1.08107761e-01 9.11038145e-02 5.11195958e-01 3.38381641e-02 -9.91482809e-02 -2.13721141e-01 1.21768482e-01 -2.51819134e-01 -8.88543054e-02 1.38507605e+00 -2.67217755e-01 1.46032423e-01 5.85957170e-01 8.09569120e-01 -4.10014182e-01 -1.53768218e+00 -5.47024488e-01 4.17028427e-01 -4.25882131e-01 -4.48949844e-01 -7.87521660e-01 -7.59999633e-01 6.67088866e-01 4.07570273e-01 5.96449971e-01 1.50163484e+00 1.33832663e-01 3.21051806e-01 5.02008557e-01 5.14401913e-01 -1.32733727e+00 4.68971163e-01 3.39187592e-01 8.82992864e-01 -8.74417186e-01 -3.12790632e-01 -2.22056493e-01 -8.43380451e-01 9.54538524e-01 1.09260297e+00 8.98843855e-02 5.31615496e-01 -1.46121323e-01 -2.15258241e-01 6.86909929e-02 -1.01160109e+00 1.56409740e-01 9.07072239e-03 1.25538468e+00 6.75103068e-02 -1.63764298e-01 -1.33031279e-01 7.03644276e-01 -2.90331304e-01 1.88336316e-02 4.38550413e-01 9.80473042e-01 -2.68875390e-01 -1.24376750e+00 -2.95516476e-02 2.47804761e-01 7.01210424e-02 1.48843527e-02 -6.21654272e-01 1.08453989e+00 1.57019407e-01 6.82704508e-01 1.12320356e-01 -3.88513952e-01 5.25173306e-01 5.78455269e-01 3.75701696e-01 -6.50220394e-01 -4.93511200e-01 -2.57035851e-01 -1.23192497e-01 -1.06563412e-01 -3.72970670e-01 -7.14172661e-01 -9.12485242e-01 -1.50856614e-01 6.69904724e-02 2.05875590e-01 4.00901645e-01 8.08284819e-01 6.79821730e-01 6.67893291e-01 8.09734404e-01 -4.44108844e-01 -1.01790273e+00 -8.70865583e-01 -7.12257683e-01 6.50449574e-01 2.70685852e-01 -9.50176239e-01 -3.39721292e-01 1.92861483e-01]
[10.17762565612793, 3.1112685203552246]
13be319b-b53d-4697-9e9f-5be726bba964
using-set-covering-to-generate-databases-for
2211.03447
null
https://arxiv.org/abs/2211.03447v1
https://arxiv.org/pdf/2211.03447v1.pdf
Using Set Covering to Generate Databases for Holistic Steganalysis
Within an operational framework, covers used by a steganographer are likely to come from different sensors and different processing pipelines than the ones used by researchers for training their steganalysis models. Thus, a performance gap is unavoidable when it comes to out-of-distributions covers, an extremely frequent scenario called Cover Source Mismatch (CSM). Here, we explore a grid of processing pipelines to study the origins of CSM, to better understand it, and to better tackle it. A set-covering greedy algorithm is used to select representative pipelines minimizing the maximum regret between the representative and the pipelines within the set. Our main contribution is a methodology for generating relevant bases able to tackle operational CSM. Experimental validation highlights that, for a given number of training samples, our set covering selection is a better strategy than selecting random pipelines or using all the available pipelines. Our analysis also shows that parameters as denoising, sharpening, and downsampling are very important to foster diversity. Finally, different benchmarks for classical and wild databases show the good generalization property of the extracted databases. Additional resources are available at github.com/RonyAbecidan/HolisticSteganalysisWithSetCovering.
['Tomáš Pevný', 'Patrick Bas', 'Jérémie Boulanger', 'Vincent Itier', 'Rony Abecidan']
2022-11-07
null
null
null
null
['steganalysis']
['computer-vision']
[ 5.37851632e-01 5.84865138e-02 5.17031774e-02 1.91416815e-02 -9.02427316e-01 -5.91471732e-01 4.98179287e-01 2.22070083e-01 -9.27819312e-02 3.49912465e-01 -6.20080903e-02 8.17100611e-03 -8.99559110e-02 -1.01371133e+00 -1.01314402e+00 -9.24127758e-01 -1.85789779e-01 3.91570181e-01 3.24980408e-01 -3.42895687e-01 4.10684168e-01 2.69802868e-01 -1.81671858e+00 3.39162171e-01 8.38400066e-01 6.86875582e-01 3.52774352e-01 4.93727505e-01 8.14419091e-02 3.96357328e-01 -7.83158600e-01 -5.18233776e-01 6.45535648e-01 -8.09126019e-01 -4.16150331e-01 2.59352941e-02 6.02933951e-03 -7.73762316e-02 1.31844372e-01 1.34542990e+00 5.39747536e-01 -4.33677226e-01 4.84018594e-01 -1.06824911e+00 3.05309206e-01 1.17102611e+00 -4.48238999e-01 -1.20145902e-01 3.16954345e-01 4.22106892e-01 7.53977120e-01 -5.62308669e-01 8.11232746e-01 1.04970050e+00 6.62932932e-01 2.35179260e-01 -9.64338779e-01 -7.22708404e-01 -2.97601938e-01 1.30426332e-01 -1.40268481e+00 -5.01029730e-01 6.59910619e-01 -3.71155947e-01 5.13053477e-01 6.52116001e-01 1.05283356e+00 1.29705858e+00 2.69045115e-01 3.94646376e-01 1.21825147e+00 -6.60012841e-01 3.03420514e-01 4.60473269e-01 -2.41479993e-01 2.46750057e-01 9.53596652e-01 2.28133857e-01 -6.30046964e-01 -1.05903134e-01 1.07464746e-01 -2.47972131e-01 -5.33883333e-01 -1.36472508e-01 -1.12528133e+00 6.91673875e-01 1.32571170e-02 6.23175859e-01 -4.20329481e-01 1.99701130e-01 4.42415625e-01 5.63380957e-01 4.21790481e-01 7.56536245e-01 -2.41428375e-01 1.12795182e-01 -1.21756852e+00 1.67189136e-01 1.06573284e+00 1.09039843e+00 9.44506347e-01 -1.03338994e-01 -5.68452924e-02 -3.51534644e-03 3.24347287e-01 6.72667742e-01 6.06916025e-02 -4.11162347e-01 4.52800393e-01 5.94448090e-01 -2.03838691e-01 -1.16823626e+00 -7.58294165e-02 -7.25600123e-01 -7.47196198e-01 3.04766893e-02 3.89965862e-01 -8.09724256e-02 -5.74562550e-01 1.06690454e+00 3.07475865e-01 2.12197825e-01 1.10133462e-01 6.66384041e-01 5.35096467e-01 3.83084983e-01 -2.48065740e-01 -2.83040047e-01 1.60933483e+00 -3.42381120e-01 -6.06551528e-01 -1.60629392e-01 5.89239299e-01 -8.01566482e-01 6.05852067e-01 7.38937497e-01 -7.29442358e-01 -7.08950013e-02 -1.17819488e+00 6.45607829e-01 -3.86848122e-01 -2.01766267e-01 2.46163040e-01 1.09294331e+00 -7.18951166e-01 8.22903931e-01 -6.68394387e-01 -5.37674546e-01 3.83779168e-01 -4.76385839e-02 1.85906596e-03 -1.91617802e-01 -1.26491368e+00 7.87064672e-01 8.89025867e-01 -5.62559515e-02 -1.31275690e+00 -5.82787573e-01 -4.66749310e-01 -7.87136778e-02 5.09293258e-01 -3.78098279e-01 4.10914332e-01 -9.54643369e-01 -1.01038849e+00 9.58794355e-01 2.77867109e-01 -7.87508845e-01 8.29800129e-01 9.13208500e-02 -3.06086332e-01 6.99409395e-02 -2.35060051e-01 1.06041245e-01 1.22506821e+00 -1.44075108e+00 -4.23688084e-01 -1.89426482e-01 -2.35386878e-01 -1.86623260e-01 -3.31469446e-01 8.73904079e-02 -3.91915530e-01 -5.04775345e-01 1.06678732e-01 -9.27611053e-01 -1.81902111e-01 -6.92562342e-01 -7.83531189e-01 3.79041314e-01 5.53010046e-01 -7.22636938e-01 1.26664209e+00 -2.13591599e+00 2.41948038e-01 5.84211111e-01 -2.79014912e-02 2.31247053e-01 1.10271454e-01 9.68731046e-01 1.80206951e-02 4.28807914e-01 -3.98144096e-01 -4.47838753e-01 -7.25483596e-02 -5.51944301e-02 -2.67238736e-01 8.01029623e-01 -3.27397473e-02 4.63624209e-01 -8.72931898e-01 -4.50897366e-01 2.00564265e-01 3.64077896e-01 -1.97892830e-01 5.16451262e-02 -5.54743111e-01 6.17242157e-01 -2.15766445e-01 9.30103242e-01 9.11379099e-01 8.94708280e-03 3.43962997e-01 -2.32522786e-01 -1.20744400e-01 -8.45630020e-02 -1.36492085e+00 1.51378238e+00 -2.80340910e-01 5.80147624e-01 -4.45398968e-03 -7.69588530e-01 1.10903847e+00 2.76023567e-01 3.63613814e-01 -4.66289520e-01 5.06607711e-01 7.42077768e-01 -1.25903919e-01 -5.11695683e-01 4.70170230e-01 1.29290998e-01 -1.60827726e-01 4.11652744e-01 -3.86183592e-03 -3.62641871e-01 2.81428069e-01 -9.15351361e-02 1.19041586e+00 -1.36767536e-01 4.24110562e-01 -5.33554912e-01 3.78248841e-01 3.40523779e-01 5.46171844e-01 8.57379258e-01 7.41867274e-02 5.71902633e-01 7.42573202e-01 -2.64787167e-01 -1.25195849e+00 -1.09754257e-01 7.07776248e-02 4.89350885e-01 3.12201411e-01 -4.46366966e-01 -1.02536392e+00 -3.85222822e-01 -3.12519997e-01 7.85357416e-01 -5.43918252e-01 -7.10673677e-03 -4.96567070e-01 -7.67133653e-01 8.44670832e-01 -4.93453413e-01 4.70064104e-01 -9.67132092e-01 -1.16308272e+00 -4.90572453e-02 -2.15774417e-01 -9.56392229e-01 1.39901461e-02 1.15006991e-01 -7.19112813e-01 -1.46353447e+00 -4.26304489e-01 -1.64680228e-01 5.58857203e-01 2.31181264e-01 1.19180572e+00 4.08844918e-01 -1.42432436e-01 9.55224633e-02 -9.31411982e-01 -8.02761197e-01 -1.06268799e+00 1.86620250e-01 -3.57535571e-01 1.84910953e-01 2.56246418e-01 -3.05845469e-01 -6.66375816e-01 2.49453098e-01 -1.23291123e+00 -1.02741413e-01 7.93624580e-01 3.20046723e-01 5.91961563e-01 5.47544360e-01 -5.81544600e-02 -9.80727434e-01 2.17266068e-01 -8.10768187e-01 -8.71511400e-01 3.23103070e-01 -7.23056078e-01 -3.43982801e-02 3.60916257e-01 -1.26364857e-01 -4.57811922e-01 8.54097828e-02 4.76228260e-02 -5.80799222e-01 -1.80803701e-01 3.66281807e-01 -3.91109824e-01 -8.88554566e-03 7.73781717e-01 3.18073362e-01 9.40502435e-02 -4.13606942e-01 5.86306956e-03 4.48837370e-01 -1.82935193e-01 -1.46942839e-01 1.05520082e+00 7.36786962e-01 7.86898062e-02 -8.72531056e-01 -1.61870494e-01 -4.77347702e-01 -2.35136911e-01 -2.93207228e-01 5.61429441e-01 -9.38608587e-01 -3.26658815e-01 7.40922987e-01 -1.23054159e+00 -1.24590658e-01 -4.14085567e-01 1.87766939e-01 -7.97682554e-02 3.39467376e-01 1.80029310e-02 -8.89413178e-01 -5.80301881e-01 -1.42585230e+00 1.01849115e+00 1.70152932e-02 -2.64642574e-02 -8.76591802e-01 -1.63267225e-01 9.31208357e-02 5.11702955e-01 9.16834772e-01 3.52120668e-01 -1.04479277e+00 -7.74901986e-01 -3.50678831e-01 2.68154114e-01 2.99850315e-01 -8.91741291e-02 2.68200427e-01 -1.05856419e+00 -5.75088263e-01 2.19947919e-01 3.00614029e-01 1.17275989e+00 2.46339098e-01 9.02565002e-01 -4.96384859e-01 -4.41070855e-01 6.44090474e-01 1.80210578e+00 -1.42125174e-01 9.40058172e-01 5.28519332e-01 4.06378508e-01 1.11631727e+00 8.67150187e-01 5.58586121e-01 -1.29565686e-01 5.00912130e-01 1.05733871e+00 1.44780532e-01 1.73941195e-01 -1.08187653e-01 6.08200073e-01 4.51508492e-01 6.05104826e-02 -5.75153172e-01 -9.21265483e-01 5.84229887e-01 -1.48015928e+00 -9.12918568e-01 -4.18619782e-01 2.42618084e+00 4.15505201e-01 2.82294720e-01 1.98290989e-01 5.10197461e-01 1.01038098e+00 3.46476018e-01 -1.30942598e-01 -7.80770257e-02 -3.10661674e-01 -5.98136485e-02 1.15015721e+00 2.24387527e-01 -1.04131341e+00 5.31972051e-01 5.31213379e+00 1.27337754e+00 -1.10715497e+00 2.10181504e-01 3.81359905e-01 -6.74669966e-02 -7.58703291e-01 3.95506889e-01 -7.69078195e-01 7.97815502e-01 1.15551174e+00 3.17710452e-02 3.75454247e-01 5.62933862e-01 2.05856398e-01 -1.33806124e-01 -8.56997609e-01 8.26702774e-01 1.77719325e-01 -1.40395308e+00 -2.25508109e-01 4.01029229e-01 7.91033089e-01 1.64726406e-01 -1.21752687e-01 -3.43059480e-01 2.42327437e-01 -9.61036205e-01 1.07525742e+00 5.36468625e-01 8.46262038e-01 -6.23349726e-01 9.70889807e-01 4.33911502e-01 -9.52728331e-01 6.12312183e-03 -2.66820639e-01 4.20170069e-01 -5.78651391e-02 1.28431261e+00 -1.00159371e+00 1.01920986e+00 7.01421857e-01 3.47695529e-01 -4.89312023e-01 1.10519361e+00 -2.98508495e-01 6.38068438e-01 -4.19256091e-01 -3.73654477e-02 -1.24968089e-01 5.24830893e-02 1.11887801e+00 1.41751564e+00 7.20881999e-01 -4.19414848e-01 -1.37202233e-01 7.36183286e-01 9.53664407e-02 6.49306476e-02 -8.16348791e-01 -9.82022136e-02 9.15166378e-01 1.10855401e+00 -9.86678183e-01 -3.72717460e-03 1.09301664e-01 6.44151986e-01 -6.06519461e-01 -1.42655000e-01 -8.96260977e-01 -2.42134765e-01 4.02652800e-01 5.17421901e-01 5.04364431e-01 1.31581277e-01 -3.90018910e-01 -9.29409266e-01 -1.21819995e-01 -1.31307518e+00 2.06144750e-01 -3.09208725e-02 -7.25288928e-01 4.91127759e-01 1.60629943e-01 -1.71280789e+00 -2.59741750e-02 -2.22566634e-01 -5.98102570e-01 6.12382472e-01 -1.43483567e+00 -1.18570626e+00 -5.66450894e-01 1.52125791e-01 4.13556814e-01 -2.71141171e-01 5.63185394e-01 2.12617874e-01 -5.83368480e-01 4.46298838e-01 1.35404438e-01 -2.37466842e-01 4.49551135e-01 -8.98800015e-01 3.81892771e-01 1.36550999e+00 -4.81657088e-02 2.75040179e-01 1.24287939e+00 -8.22855592e-01 -1.75942707e+00 -1.12945092e+00 5.17847121e-01 -2.15896890e-01 4.76043522e-01 -3.84859502e-01 -6.26073360e-01 3.05183619e-01 3.94269139e-01 -3.20118994e-01 3.73822361e-01 -4.30231452e-01 -2.70311594e-01 -1.54905796e-01 -1.41879356e+00 1.08166017e-01 8.22049320e-01 -8.45927298e-02 1.59322560e-01 3.67637932e-01 5.25706887e-01 -4.93308425e-01 -8.37143719e-01 2.10582554e-01 3.05098027e-01 -1.40682280e+00 5.11102855e-01 2.58914053e-01 4.50190157e-01 -2.87237972e-01 -1.92651361e-01 -1.18926466e+00 1.99063078e-01 -9.40266669e-01 -1.01833761e-01 1.34442866e+00 2.83947289e-01 -7.59864569e-01 5.99809170e-01 -2.17763692e-01 9.64203924e-02 -7.12324858e-01 -9.05356348e-01 -7.83805609e-01 -1.64841801e-01 -5.01953602e-01 1.14698493e+00 8.16652477e-01 -5.81089616e-01 -4.27803874e-01 -5.57248175e-01 3.61489505e-01 1.07320571e+00 -3.76844108e-02 1.05340815e+00 -1.11810470e+00 -4.44404036e-01 -4.00053263e-01 -3.39230388e-01 -2.85761833e-01 -4.11155611e-01 -6.34426177e-01 1.48554534e-01 -1.12439477e+00 -5.65942265e-02 -5.94885170e-01 -1.68736260e-02 8.70005935e-02 1.92680657e-01 3.10847253e-01 1.78980857e-01 5.21821678e-01 -3.78349990e-01 -4.52043079e-02 1.01251101e+00 -9.85717550e-02 -1.61214061e-02 1.13879062e-01 -7.63349950e-01 5.98443568e-01 8.19962263e-01 -1.07599270e+00 -6.47583082e-02 -1.20203845e-01 6.99024141e-01 3.27073317e-03 4.74970847e-01 -1.26785815e+00 2.17408523e-01 -7.25713521e-02 -1.49899498e-01 -6.40257001e-01 1.06635071e-01 -1.02620029e+00 1.03455508e+00 1.04496229e+00 2.16118321e-01 -4.92971271e-01 -1.23638175e-01 5.51099896e-01 3.15293409e-02 -6.25091076e-01 8.32494676e-01 -4.52859640e-01 -6.72751129e-01 3.34489197e-02 -7.50351623e-02 -1.30564794e-02 1.33475399e+00 -4.04613703e-01 -1.78607523e-01 -1.19464412e-01 -2.44541109e-01 -5.55842044e-03 9.42839563e-01 1.01444140e-01 3.69457573e-01 -5.95622659e-01 -1.09019244e+00 1.36561707e-01 1.98645771e-01 9.59188715e-02 2.89226711e-01 9.75737691e-01 -8.74055326e-01 1.36387646e-01 7.16874599e-02 -5.95535219e-01 -1.31106293e+00 2.86228329e-01 3.26499015e-01 -5.38405299e-01 -5.28233051e-01 9.10279214e-01 -6.02156818e-01 -9.14635211e-02 9.34233665e-02 -1.13275051e-01 -1.03176512e-01 5.75207949e-01 4.98900145e-01 6.59091473e-01 4.30449635e-01 -5.21962047e-01 -4.82003540e-01 3.25326473e-01 3.31564695e-01 1.74074784e-01 1.40379500e+00 -1.55518740e-01 -3.25530201e-01 2.86507010e-01 8.11030269e-01 2.02966586e-01 -9.38307703e-01 -3.20388116e-02 7.75706396e-02 -6.44755900e-01 -2.35237226e-01 -2.93951720e-01 -1.08404768e+00 4.54402864e-01 5.80127001e-01 7.81138539e-01 1.32235110e+00 2.41249725e-02 5.11799753e-01 -1.10637419e-01 7.99276650e-01 -9.41374302e-01 -2.15703413e-01 9.49368626e-02 9.12552476e-01 -1.07593977e+00 2.68498749e-01 -5.47118902e-01 -5.44304490e-01 9.37576056e-01 -2.87270378e-02 -2.11134627e-01 5.52405596e-01 2.85862058e-01 -3.41828138e-01 -3.37649524e-01 -3.86445969e-01 -2.11635992e-01 -3.42211992e-01 5.41696250e-01 -8.32524300e-02 1.06180526e-01 -3.93202186e-01 1.66364819e-01 -2.63463795e-01 -2.71930039e-01 6.67268932e-01 6.92295253e-01 -4.73842472e-01 -1.11139226e+00 -9.30658042e-01 3.52960080e-01 -5.74022949e-01 -1.46453558e-02 -1.78319037e-01 6.56590819e-01 5.73167145e-01 1.02388990e+00 -2.03927010e-01 -7.98819065e-01 2.67895848e-01 -2.80757725e-01 3.01130801e-01 -5.28448224e-01 -8.66639793e-01 9.14576501e-02 1.35373667e-01 -4.66125071e-01 -5.23235202e-01 -8.32060754e-01 -7.24832833e-01 -6.13259017e-01 -6.74169838e-01 1.82903722e-01 9.34826434e-01 8.31183016e-01 3.25232357e-01 5.71628213e-01 7.49907196e-01 -8.31719875e-01 -6.71101213e-01 -1.05083442e+00 -5.65063119e-01 1.36481389e-01 1.32088616e-01 -3.56700450e-01 -7.64057577e-01 7.98577268e-04]
[4.3997578620910645, 7.982824325561523]
3bdb12f9-d00e-454b-8ccd-3f88d40ebff4
sst-reversiblenet-reversible-prior-based
2305.04054
null
https://arxiv.org/abs/2305.04054v1
https://arxiv.org/pdf/2305.04054v1.pdf
SST-ReversibleNet: Reversible-prior-based Spectral-Spatial Transformer for Efficient Hyperspectral Image Reconstruction
Spectral image reconstruction is an important task in snapshot compressed imaging. This paper aims to propose a new end-to-end framework with iterative capabilities similar to a deep unfolding network to improve reconstruction accuracy, independent of optimization conditions, and to reduce the number of parameters. A novel framework called the reversible-prior-based method is proposed. Inspired by the reversibility of the optical path, the reversible-prior-based framework projects the reconstructions back into the measurement space, and then the residuals between the projected data and the real measurements are fed into the network for iteration. The reconstruction subnet in the network then learns the mapping of the residuals to the true values to improve reconstruction accuracy. In addition, a novel spectral-spatial transformer is proposed to account for the global correlation of spectral data in both spatial and spectral dimensions while balancing network depth and computational complexity, in response to the shortcomings of existing transformer-based denoising modules that ignore spatial texture features or learn local spatial features at the expense of global spatial features. Extensive experiments show that our SST-ReversibleNet significantly outperforms state-of-the-art methods on simulated and real HSI datasets, while requiring lower computational and storage costs. https://github.com/caizeyu1992/SST
['Feipeng Da', 'Chengqian Jin', 'Ziyu Zhang', 'Jian Yu', 'Zeyu Cai']
2023-05-06
null
null
null
null
['image-reconstruction']
['computer-vision']
[ 4.51465279e-01 -3.96642417e-01 2.99332947e-01 -4.45760101e-01 -5.93578696e-01 -1.81378409e-01 8.34704712e-02 -5.15094817e-01 -3.81637007e-01 5.39882839e-01 3.30653310e-01 1.98940616e-02 -4.82101619e-01 -9.08196211e-01 -5.65922976e-01 -1.08587039e+00 1.43810421e-01 -1.54965445e-01 1.28450215e-01 -1.63911939e-01 1.37218043e-01 4.77862895e-01 -1.21012223e+00 1.42365722e-02 9.21279311e-01 1.14939547e+00 5.55692017e-01 3.55080634e-01 3.33301544e-01 7.38646746e-01 4.68959175e-02 2.04720385e-02 6.63863897e-01 -5.66049337e-01 -4.13456202e-01 1.61843836e-01 3.04967731e-01 -7.37814665e-01 -9.86867249e-01 1.32204509e+00 6.14236355e-01 2.85799950e-01 2.44296081e-02 -5.34443200e-01 -6.63238525e-01 2.38776505e-01 -4.93476689e-01 -2.75944434e-02 -6.58132881e-02 2.84950256e-01 5.74899077e-01 -1.12365794e+00 4.73320574e-01 8.03068221e-01 1.07631910e+00 4.69220467e-02 -1.30984175e+00 -5.87328017e-01 -1.07063130e-01 4.88863528e-01 -1.51169431e+00 -6.14161789e-01 1.14915442e+00 -1.43542066e-01 6.53429270e-01 2.60693103e-01 8.31721604e-01 7.78152227e-01 1.15102038e-01 2.00460091e-01 1.16665161e+00 -3.15602809e-01 -6.19521551e-02 -4.69306409e-01 -8.79110321e-02 7.56077826e-01 -5.67028262e-02 3.59772563e-01 -8.68986845e-01 5.48251905e-02 9.57046270e-01 4.75312918e-01 -9.40056801e-01 -3.22809666e-01 -1.14695585e+00 6.32233739e-01 9.13453341e-01 1.97536305e-01 -5.59380412e-01 8.56587365e-02 8.33234042e-02 2.69535571e-01 6.27181292e-01 2.52719700e-01 -2.35178843e-01 3.17747921e-01 -1.13108718e+00 -1.43988982e-01 4.69831437e-01 5.03518701e-01 1.10008991e+00 3.14741433e-01 8.78155828e-02 9.24181163e-01 3.26681435e-01 7.05586016e-01 3.73340845e-01 -1.14947772e+00 1.79919749e-01 2.99981833e-01 -1.08040005e-01 -1.31708872e+00 -3.52155983e-01 -7.85095811e-01 -1.39423823e+00 5.51397800e-02 1.46470424e-02 1.94248423e-01 -8.34535003e-01 1.60181320e+00 3.02706391e-01 6.27550364e-01 -1.11105286e-01 1.32485604e+00 6.28753901e-01 8.86200845e-01 -6.20394051e-01 -3.38987023e-01 9.28157449e-01 -8.25489581e-01 -7.01974154e-01 -1.78288743e-01 2.29985248e-02 -8.38346183e-01 1.00778341e+00 2.74305910e-01 -1.00957692e+00 -6.05431080e-01 -1.08076680e+00 -2.81489819e-01 6.35168031e-02 3.33056748e-01 2.78282821e-01 1.94312111e-01 -1.14530003e+00 9.87291276e-01 -9.17452693e-01 -1.77839905e-01 1.30580842e-01 -4.63641025e-02 -1.04207247e-01 -4.47291106e-01 -9.30787921e-01 5.90442836e-01 3.65594417e-01 8.21790278e-01 -8.59460771e-01 -7.96771467e-01 -6.93038762e-01 1.36215210e-01 4.52011824e-01 -5.79760611e-01 6.09397829e-01 -8.60432446e-01 -1.71599567e+00 3.10700089e-01 -2.52449453e-01 -2.00116232e-01 4.43454564e-01 -1.49359722e-02 -3.34066540e-01 3.35708141e-01 -5.10288551e-02 2.65206963e-01 1.06041539e+00 -1.14943540e+00 -1.88875005e-01 -3.00184876e-01 -3.17683697e-01 2.98552126e-01 -4.19912636e-01 -5.72181702e-01 -4.78053749e-01 -6.94401920e-01 7.52016842e-01 -8.91993880e-01 -2.47087464e-01 3.69550377e-01 -3.28671396e-01 5.74242234e-01 6.60738885e-01 -1.08970881e+00 9.06684756e-01 -2.50530815e+00 3.38279724e-01 3.80115867e-01 3.40667963e-01 1.23978011e-01 -4.17538285e-01 4.39895868e-01 -2.48978108e-01 -4.08145040e-01 -6.93153381e-01 -2.60934800e-01 -3.46635550e-01 1.84614703e-01 -4.24326837e-01 7.76653230e-01 -1.77871987e-01 6.70603931e-01 -7.97206759e-01 -2.30904743e-02 3.55173141e-01 7.56852210e-01 -4.03304070e-01 2.61895835e-01 2.84186840e-01 7.01722264e-01 -1.49754480e-01 5.34246862e-01 1.11000752e+00 -1.91866979e-01 2.46346563e-01 -8.45797122e-01 -4.71596807e-01 3.77546221e-01 -1.20943785e+00 1.97415841e+00 -4.82506573e-01 5.89680970e-01 5.40060878e-01 -1.17498875e+00 8.82003427e-01 1.45901456e-01 6.50842309e-01 -9.72633719e-01 -1.21442959e-01 4.21756715e-01 -2.60597467e-01 -4.85622168e-01 4.10255194e-01 -1.56960309e-01 5.37657022e-01 8.41483623e-02 -3.25825065e-02 -1.82720810e-01 -1.39300842e-02 -5.34684844e-02 8.31485927e-01 3.30099344e-01 2.44218986e-02 -3.09330791e-01 5.04148901e-01 -8.68487880e-02 9.04131770e-01 5.10982692e-01 9.77763310e-02 8.97376299e-01 5.24145626e-02 -6.04910374e-01 -1.08606720e+00 -9.96673346e-01 -1.16201662e-01 6.14189923e-01 3.18377256e-01 -2.46837810e-01 -4.38007414e-01 -6.57298863e-02 -2.51113206e-01 4.38581944e-01 -3.69150519e-01 -1.38768122e-01 -5.58479548e-01 -8.47988009e-01 1.07961990e-01 1.35691136e-01 1.18406844e+00 -6.58706069e-01 -5.43655872e-01 3.18801314e-01 -6.79380655e-01 -1.08112776e+00 -3.69217724e-01 5.61212376e-02 -1.12690568e+00 -9.81947064e-01 -4.98234093e-01 -4.73269880e-01 6.45183384e-01 6.86671078e-01 7.49551177e-01 1.24345839e-01 -2.09993422e-01 1.80234089e-01 -2.71988600e-01 2.13267952e-01 4.05014269e-02 -1.93564191e-01 -2.92108446e-01 3.89941812e-01 -3.87068063e-01 -1.15386331e+00 -9.92102146e-01 2.61341780e-01 -1.08735216e+00 4.70926225e-01 5.57362139e-01 1.21267986e+00 8.31158936e-01 1.20037816e-01 -1.84701011e-01 -5.54365396e-01 2.19934627e-01 -2.67185390e-01 -6.09560788e-01 1.63479552e-01 -6.96359515e-01 -5.26359789e-02 6.49404585e-01 -2.80940741e-01 -1.09056914e+00 2.52099782e-01 -2.03860983e-01 -5.74153483e-01 2.94118315e-01 9.48724151e-01 9.99817550e-02 -4.10387486e-01 4.83940184e-01 7.95216382e-01 3.07111025e-01 -6.61404729e-01 2.89818225e-03 1.67565450e-01 7.61866331e-01 -1.29048571e-01 1.07279265e+00 8.41151476e-01 2.97622204e-01 -9.48450863e-01 -1.03779435e+00 -3.76458466e-01 -6.64109945e-01 -4.25516576e-01 5.09057522e-01 -1.00420713e+00 -5.51768005e-01 8.01378310e-01 -9.33688462e-01 -4.74006772e-01 -4.63185072e-01 5.20329416e-01 -2.91054934e-01 6.15478814e-01 -7.51762152e-01 -2.08469599e-01 -4.85923171e-01 -1.08442545e+00 7.95369565e-01 1.29253805e-01 5.59807897e-01 -7.41479278e-01 2.89058257e-02 1.14793077e-01 8.09588909e-01 8.22302997e-02 6.09868467e-01 3.56105864e-01 -9.45721805e-01 -6.77220300e-02 -4.75845635e-01 7.04851031e-01 -2.91190818e-02 -2.65066266e-01 -8.20771813e-01 -5.59608638e-01 6.25540197e-01 -1.89921595e-02 1.05775189e+00 5.96875608e-01 1.32466745e+00 -3.27272624e-01 2.07332999e-01 1.48796618e+00 1.87146032e+00 -2.70061672e-01 9.78751481e-01 2.71331459e-01 7.98019886e-01 4.48348969e-01 1.64006636e-01 3.82425815e-01 3.12753648e-01 4.39546317e-01 7.15162873e-01 -3.38164985e-01 -4.31290030e-01 -1.58336878e-01 3.19631308e-01 1.06743157e+00 -2.38776095e-02 -1.05786763e-01 -7.63050139e-01 4.17646110e-01 -1.81814337e+00 -8.75045896e-01 -3.09550136e-01 2.12924075e+00 6.21963799e-01 -3.73924077e-01 -5.76558113e-01 4.90316823e-02 4.96445000e-01 6.45183146e-01 -7.39421248e-01 3.81421179e-01 -2.66102970e-01 1.54726550e-01 8.36244881e-01 6.59711778e-01 -7.30683446e-01 5.15537262e-01 5.36363697e+00 7.13225484e-01 -1.45469093e+00 3.16346943e-01 4.15243477e-01 -9.35664475e-02 -1.78283244e-01 2.34430432e-01 -1.38183549e-01 2.77494103e-01 6.41154110e-01 1.26331791e-01 1.05861926e+00 3.33780795e-01 3.22651327e-01 -1.41852766e-01 -5.59517622e-01 1.07857740e+00 4.66488749e-02 -1.48217666e+00 -2.04666466e-01 -2.66045276e-02 6.05330586e-01 4.02699351e-01 6.05829060e-02 -2.37088367e-01 -1.16432808e-01 -7.24679053e-01 7.81691432e-01 9.93958831e-01 8.73808682e-01 -5.26691735e-01 7.18125165e-01 2.87965953e-01 -1.28253591e+00 -1.43721402e-01 -5.78473985e-01 -1.16186053e-01 2.93516725e-01 1.15918851e+00 -3.38670045e-01 8.68244708e-01 9.98606741e-01 1.22257102e+00 -3.31718922e-01 9.29074109e-01 -3.82190824e-01 6.93557858e-01 -4.10862446e-01 5.99017859e-01 1.59429193e-01 -7.73642302e-01 6.27297223e-01 7.82160997e-01 7.76482165e-01 4.17927951e-01 2.45478824e-01 9.35978413e-01 4.77328822e-02 -3.31712037e-01 -3.72062445e-01 3.41730803e-01 2.64038026e-01 1.44766331e+00 -4.79577720e-01 -1.43695548e-01 -2.27461308e-01 1.00352752e+00 1.31010702e-02 6.58001602e-01 -4.95764345e-01 -1.83593601e-01 4.57035005e-01 1.94975108e-01 3.25658739e-01 -4.50311750e-01 -2.22295165e-01 -1.28416538e+00 3.17453295e-01 -7.28652179e-01 7.05578178e-02 -1.08988225e+00 -1.18310237e+00 5.07625639e-01 -3.30869049e-01 -1.48252678e+00 2.33026117e-01 -3.81469131e-01 -5.02497733e-01 1.05089021e+00 -1.91916227e+00 -1.11986363e+00 -9.10800397e-01 7.68960595e-01 2.36839384e-01 2.08961338e-01 5.87147057e-01 3.92390072e-01 -5.77185154e-01 -5.01241162e-02 5.85073590e-01 4.27164324e-02 7.04129338e-01 -7.61006534e-01 1.02524728e-01 1.32052851e+00 -2.49383152e-01 5.42020261e-01 5.27466118e-01 -5.29517651e-01 -1.57074118e+00 -1.10888255e+00 4.40201163e-01 4.57334459e-01 5.59780777e-01 -1.20668776e-01 -1.10758078e+00 4.25908238e-01 1.51176393e-01 4.43913609e-01 2.82197326e-01 -3.73361498e-01 -3.44226867e-01 -7.63545692e-01 -1.03349376e+00 4.26106960e-01 1.12045908e+00 -7.33434498e-01 7.87297040e-02 4.53377306e-01 5.34921169e-01 -6.11764312e-01 -7.72760987e-01 4.44691479e-01 4.80648577e-01 -1.31587994e+00 1.12213421e+00 3.82036835e-01 5.80314755e-01 -5.71676075e-01 -2.86177576e-01 -1.44489551e+00 -7.08154082e-01 -5.30172110e-01 1.17685134e-02 8.76228452e-01 4.58175652e-02 -1.03991878e+00 4.67094898e-01 1.23596162e-01 -4.57100719e-01 -5.89189410e-01 -1.08108675e+00 -5.89197397e-01 -3.93914163e-01 -2.66047746e-01 5.08772373e-01 9.06864703e-01 -7.13242233e-01 -1.88608584e-03 -5.51022530e-01 6.05529904e-01 1.03199410e+00 3.28286678e-01 3.98232996e-01 -8.89316916e-01 -3.12990099e-01 -1.69036075e-01 -1.22488841e-01 -1.27153790e+00 -1.03970431e-01 -8.88105869e-01 2.90533006e-01 -1.42608857e+00 1.87240750e-01 -2.51535833e-01 -3.29946786e-01 3.71381611e-01 -1.15710665e-02 5.06655335e-01 3.41173053e-01 5.91642439e-01 -4.06411430e-03 9.54411387e-01 1.32063925e+00 -7.03449398e-02 3.35930325e-02 -2.85542130e-01 -2.54422486e-01 5.69043458e-01 6.10224783e-01 -5.60390115e-01 -2.95872808e-01 -9.94018018e-01 2.20819354e-01 3.65728587e-01 8.16875756e-01 -1.15834272e+00 4.95379359e-01 -1.67026296e-01 3.99690598e-01 -5.79260707e-01 4.58771855e-01 -1.04518700e+00 4.77285951e-01 4.79805619e-01 -1.67465404e-01 -1.62016153e-01 -1.33269057e-01 6.28688633e-01 -3.43303651e-01 -1.49849728e-01 9.70374405e-01 -8.65550786e-02 -6.17169678e-01 4.04330164e-01 -4.80829924e-03 -4.82055008e-01 5.57884932e-01 -2.37797603e-01 -3.14270437e-01 -3.63782376e-01 -5.45393646e-01 8.22079480e-02 5.11560082e-01 -2.13683754e-01 9.08400416e-01 -1.15187001e+00 -8.00558269e-01 5.12763917e-01 -2.43261084e-01 1.67265877e-01 7.98203468e-01 1.24809361e+00 -9.05831158e-01 -1.98865339e-01 -2.06279188e-01 -7.47562468e-01 -9.22936559e-01 1.55544400e-01 6.69110298e-01 -3.32301259e-01 -1.01081276e+00 6.80418670e-01 -9.62160379e-02 -6.97794259e-01 -6.74416032e-03 -1.59852102e-01 3.60468119e-01 -1.26478776e-01 4.07617152e-01 5.19759178e-01 1.67672053e-01 -4.59368885e-01 -3.32655013e-02 7.32337832e-01 3.29938710e-01 -1.64074168e-01 1.97801459e+00 -4.02209520e-01 -6.62334323e-01 1.32782504e-01 1.34314549e+00 -8.43496397e-02 -1.67365038e+00 -6.97062671e-01 -5.44410050e-01 -8.17246795e-01 6.28919363e-01 -6.69484735e-01 -1.72099316e+00 8.16862285e-01 9.24780130e-01 4.03292701e-02 1.80325210e+00 -6.54665053e-01 9.49399829e-01 3.28564167e-01 1.88622251e-01 -9.18908179e-01 2.03164835e-02 6.27100468e-01 1.00979292e+00 -1.02008414e+00 4.80158299e-01 -4.62486923e-01 -1.94413140e-01 1.37371337e+00 1.85935780e-01 -2.30869815e-01 7.47431397e-01 5.42148575e-02 -3.47221904e-02 -1.71683416e-01 -2.48485893e-01 9.52064469e-02 1.86908603e-01 3.87926012e-01 4.48672138e-02 -1.09909840e-01 -9.30353701e-02 -1.45317987e-01 7.16421679e-02 6.19402751e-02 4.99660313e-01 8.35369289e-01 -2.25552469e-01 -7.45073617e-01 -3.83437395e-01 3.59424412e-01 7.02666417e-02 -2.67761022e-01 2.23863915e-01 3.30417156e-01 5.98009191e-02 7.01010585e-01 -8.33597928e-02 -4.09732908e-01 3.56654644e-01 -2.97444135e-01 2.53595412e-01 -2.42856607e-01 -2.99826264e-01 2.22240940e-01 -2.70616919e-01 -9.17762160e-01 -5.63250542e-01 -5.62550962e-01 -9.10775006e-01 -4.41606879e-01 -1.73042059e-01 -1.77054837e-01 7.73516238e-01 7.09603786e-01 4.97631639e-01 4.99983758e-01 9.81042027e-01 -1.22835839e+00 -6.62623525e-01 -9.16787744e-01 -5.91285944e-01 2.37902865e-01 6.02314055e-01 -2.42033347e-01 -4.50177759e-01 8.62031505e-02]
[10.799715995788574, -2.288935661315918]
c33fbf66-54c1-4c46-a0fe-b754bf0248eb
unsupervised-audio-source-separation-using-1
2201.09592
null
https://arxiv.org/abs/2201.09592v2
https://arxiv.org/pdf/2201.09592v2.pdf
Unsupervised Music Source Separation Using Differentiable Parametric Source Models
Supervised deep learning approaches to underdetermined audio source separation achieve state-of-the-art performance but require a dataset of mixtures along with their corresponding isolated source signals. Such datasets can be extremely costly to obtain for musical mixtures. This raises a need for unsupervised methods. We propose a novel unsupervised model-based deep learning approach to musical source separation. Each source is modelled with a differentiable parametric source-filter model. A neural network is trained to reconstruct the observed mixture as a sum of the sources by estimating the source models' parameters given their fundamental frequencies. At test time, soft masks are obtained from the synthesized source signals. The experimental evaluation on a vocal ensemble separation task shows that the proposed method outperforms learning-free methods based on nonnegative matrix factorization and a supervised deep learning baseline. Integrating domain knowledge in the form of source models into a data-driven method leads to high data efficiency: the proposed approach achieves good separation quality even when trained on less than three minutes of audio. This work makes powerful deep learning based separation usable in scenarios where training data with ground truth is expensive or nonexistent.
['Liam Kelley', 'Gaël Richard', 'Roland Badeau', 'Clement S. J. Doire', 'Kilian Schulze-Forster']
2022-01-24
null
null
null
null
['audio-source-separation', 'music-source-separation']
['audio', 'music']
[ 3.07619333e-01 -1.29942670e-01 4.64019775e-02 -2.39876106e-01 -1.41359508e+00 -6.87053800e-01 4.36941147e-01 -2.71285236e-01 -8.66791680e-02 5.71002603e-01 3.30340445e-01 1.22118860e-01 -3.91800135e-01 -2.24178493e-01 -7.60315061e-01 -9.38371003e-01 6.75970037e-03 5.69840193e-01 -4.52545434e-01 -5.15914857e-02 -1.27840787e-01 3.11058104e-01 -1.71627510e+00 4.89054948e-01 9.75814819e-01 1.06196964e+00 2.90109217e-01 1.02948856e+00 -1.76726386e-01 8.43046784e-01 -9.59070683e-01 -5.46633825e-02 4.60245162e-01 -6.52921855e-01 -2.91850805e-01 3.13611627e-01 4.69753236e-01 -3.70958328e-01 -1.33598819e-01 9.69908416e-01 7.43472457e-01 2.85741776e-01 8.42029214e-01 -1.26095176e+00 -2.41025910e-01 9.90207076e-01 -4.56499040e-01 2.94418987e-02 6.57593682e-02 -1.87447920e-01 1.01963735e+00 -1.02894580e+00 -7.63715282e-02 1.00541902e+00 7.21186101e-01 2.07195401e-01 -1.52364206e+00 -7.88047254e-01 -1.23066001e-01 -7.10901077e-05 -1.25426018e+00 -1.22061145e+00 1.24703431e+00 -5.48996866e-01 7.38769889e-01 3.10481519e-01 2.40097865e-01 1.11112821e+00 -3.83822113e-01 8.88461709e-01 8.44814301e-01 -5.99097192e-01 3.04286063e-01 1.87051430e-01 -2.00934350e-01 2.70163685e-01 -4.61638533e-02 -4.67605591e-02 -9.26411808e-01 -2.24971160e-01 8.51863623e-01 -2.04050034e-01 -3.08518767e-01 -3.31453830e-01 -1.21477461e+00 5.97820103e-01 1.57058418e-01 4.76068377e-01 -5.60336053e-01 2.22286973e-02 2.48110518e-01 5.15860319e-01 4.90932167e-01 4.39193934e-01 -3.25403422e-01 -4.21012640e-02 -1.69432020e+00 2.07937643e-01 9.85281944e-01 6.95653200e-01 4.36970264e-01 8.39604616e-01 2.69432604e-01 1.27258694e+00 3.58809143e-01 5.01065373e-01 7.54617929e-01 -1.18747294e+00 4.22215313e-01 1.98878303e-01 3.06683064e-01 -8.96355569e-01 -1.67752832e-01 -1.01427662e+00 -1.04789555e+00 3.97861838e-01 6.15290046e-01 -3.10630381e-01 -7.75756180e-01 1.71925223e+00 1.97405636e-01 7.18639970e-01 3.79854441e-01 1.05303061e+00 7.85638332e-01 7.28138030e-01 -5.88662326e-01 -4.13743645e-01 8.09572816e-01 -1.00231981e+00 -8.53066623e-01 -3.14983189e-01 3.83554995e-02 -1.06670308e+00 6.51256144e-01 1.03298199e+00 -1.24998963e+00 -8.76223087e-01 -1.13508379e+00 2.83870310e-01 1.18101113e-01 7.09211946e-01 3.04440647e-01 7.67486036e-01 -9.05056655e-01 5.93024671e-01 -8.24504375e-01 3.01275104e-01 2.20186204e-01 4.96656299e-01 -3.08337569e-01 3.18843365e-01 -8.22171271e-01 3.21341157e-01 1.96890131e-01 3.31392884e-01 -1.41654027e+00 -5.91610491e-01 -7.72061467e-01 2.87752539e-01 1.83975190e-01 -5.57084143e-01 1.43061864e+00 -1.59567535e+00 -1.91654932e+00 3.34653258e-01 -1.94189474e-01 -6.85807884e-01 3.82105649e-01 -5.52205026e-01 -6.29820943e-01 2.77412027e-01 1.11181876e-02 2.58853972e-01 1.62220180e+00 -1.56998718e+00 -4.50500607e-01 -1.19657014e-02 -3.80987495e-01 1.35214299e-01 -3.93066943e-01 -3.51425335e-02 -9.10372380e-03 -9.82797146e-01 2.97572225e-01 -8.23451579e-01 -7.71398246e-02 -4.32512611e-01 -5.69611371e-01 2.13623434e-01 3.95482838e-01 -9.49898243e-01 8.80976856e-01 -2.41406798e+00 4.74532694e-01 1.87335014e-01 1.32909879e-01 2.06062868e-01 -2.72430241e-01 3.54629666e-01 -4.21623200e-01 -4.34765130e-01 -3.98600399e-01 -9.06398118e-01 2.18312383e-01 -1.40554592e-01 -7.36636043e-01 4.38391268e-01 1.02164172e-01 3.50969732e-01 -7.24652112e-01 -1.67556182e-01 2.23913118e-01 7.61676967e-01 -4.73728240e-01 5.78803837e-01 -2.20997885e-01 6.43343568e-01 2.84596443e-01 5.03449202e-01 6.14192486e-01 2.17701457e-02 2.47004360e-01 -1.64836451e-01 1.69482335e-01 4.39093679e-01 -1.70469570e+00 2.10741663e+00 -4.82595205e-01 6.91864431e-01 6.87422812e-01 -1.14456892e+00 1.07437861e+00 9.06946898e-01 6.14902675e-01 1.03140756e-01 2.81431496e-01 5.26314974e-01 2.77581245e-01 -3.07268739e-01 2.80908167e-01 -5.20494998e-01 1.13622911e-01 4.72491413e-01 7.06336021e-01 -2.03235999e-01 4.25971858e-02 8.44531413e-03 7.71774232e-01 2.33676568e-01 -8.26301202e-02 -1.14502192e-01 4.05845493e-01 -4.43623364e-01 5.50010622e-01 6.56855464e-01 2.81941950e-01 7.51885414e-01 3.44668739e-02 -1.12218335e-02 -5.75812936e-01 -1.21396029e+00 2.79471546e-01 1.18232536e+00 -4.11466688e-01 -3.75814110e-01 -8.16361248e-01 -2.15212122e-01 -2.69577980e-01 6.34419203e-01 -2.53544122e-01 -3.64954211e-02 -4.51233059e-01 -6.11198366e-01 6.89637363e-01 3.86887699e-01 3.62827145e-02 -7.85507262e-01 -2.09710494e-01 4.86747593e-01 -3.70858639e-01 -9.08761919e-01 -2.03720212e-01 5.31227291e-01 -9.55089390e-01 -5.88537633e-01 -9.37655330e-01 -8.62265825e-01 3.31514686e-01 2.37333283e-01 9.37197626e-01 -6.20153785e-01 8.98626819e-02 1.65208057e-01 -1.71244442e-01 -5.96327126e-01 -5.58476865e-01 -6.07892163e-02 5.44599354e-01 7.65453756e-01 4.01201919e-02 -1.32375515e+00 -3.41059268e-01 9.13084000e-02 -9.67858791e-01 -2.09522754e-01 5.95643342e-01 6.28550470e-01 4.59087491e-01 5.01958072e-01 1.02464414e+00 -4.00840998e-01 7.98290670e-01 -5.08569419e-01 -3.82127017e-01 -9.65729877e-02 -4.13199738e-02 5.07502109e-02 8.15882504e-01 -8.46315145e-01 -1.24101543e+00 3.28102708e-01 8.28535706e-02 -9.02654946e-01 -4.41819072e-01 4.85150874e-01 -4.93086934e-01 4.53674138e-01 8.74626100e-01 3.03027093e-01 -7.82677382e-02 -1.04934955e+00 4.42083418e-01 7.87072659e-01 9.80221391e-01 -5.33112645e-01 8.49223852e-01 3.77347201e-01 -3.74844879e-01 -9.62937474e-01 -8.12646568e-01 -5.31324804e-01 -8.98262322e-01 -1.89235985e-01 4.25274491e-01 -1.32256246e+00 -2.14856237e-01 5.47218442e-01 -1.08163512e+00 -3.02729219e-01 -3.99847716e-01 8.74310315e-01 -5.77543974e-01 2.08080307e-01 -4.65843320e-01 -1.15600777e+00 -1.78047419e-01 -8.92993867e-01 1.07503784e+00 1.97201297e-02 -3.65849257e-01 -7.77908206e-01 2.78417200e-01 4.67328519e-01 2.78493196e-01 3.34155858e-02 4.53884512e-01 -8.76131296e-01 -4.02037948e-01 -2.79788613e-01 3.97440314e-01 8.30874741e-01 5.47392488e-01 -1.18272014e-01 -1.59962475e+00 -2.62491643e-01 4.64136332e-01 -3.21464062e-01 9.88686621e-01 5.32062709e-01 8.00507784e-01 -3.91261488e-01 6.57166094e-02 6.22332871e-01 1.06359494e+00 4.74526137e-02 7.52852708e-02 -2.26244241e-01 7.45535016e-01 5.16345322e-01 1.55912802e-01 4.89323109e-01 -1.33129150e-01 3.37785274e-01 2.73411512e-01 -8.78746659e-02 -1.98578089e-01 -8.33313316e-02 6.21161401e-01 1.28812766e+00 -9.47274715e-02 -1.53421327e-01 -6.20447457e-01 7.59011865e-01 -1.75919640e+00 -1.02338696e+00 5.15182950e-02 2.29162002e+00 9.02797937e-01 1.42955244e-01 4.61434931e-01 9.56711173e-01 4.86431509e-01 8.73162672e-02 -4.96171623e-01 1.39192510e-02 -9.14658010e-02 2.60463804e-01 -8.87631774e-02 6.54599011e-01 -1.21584773e+00 4.73663092e-01 5.90153217e+00 8.83268476e-01 -1.23871756e+00 2.47617781e-01 8.99729803e-02 -6.16791606e-01 -2.31997207e-01 -3.74647349e-01 -3.03754479e-01 2.69616008e-01 1.23837388e+00 1.14805892e-01 6.96596980e-01 6.32642210e-01 2.12438092e-01 2.00296104e-01 -1.26137948e+00 1.40633655e+00 2.01606765e-01 -1.08325589e+00 -1.94367215e-01 -6.31561801e-02 7.02173114e-01 -8.97154883e-02 2.12259993e-01 1.57948196e-01 2.82240093e-01 -1.08469558e+00 1.00882518e+00 6.52797461e-01 5.25477290e-01 -9.40428555e-01 3.62332731e-01 8.56242537e-01 -9.68367338e-01 -1.51562080e-01 -3.05878878e-01 -2.68774152e-01 8.03403109e-02 8.64489436e-01 -9.77434397e-01 6.38995707e-01 4.21490878e-01 7.01408625e-01 -1.06499434e-01 9.15326655e-01 -9.99127552e-02 1.06617570e+00 -4.53115940e-01 6.08851731e-01 -5.45701683e-02 -2.32806236e-01 9.36069965e-01 1.29303312e+00 4.50104386e-01 -3.54754895e-01 5.01406230e-02 8.81510019e-01 -1.16642401e-01 5.97059913e-02 -4.91655320e-01 -2.34686300e-01 2.62767434e-01 1.34162140e+00 -7.98032045e-01 -2.89323896e-01 5.56775322e-03 9.11557972e-01 1.66559711e-01 4.51459080e-01 -6.60709023e-01 -2.61324495e-01 5.13899505e-01 -3.15242559e-02 3.76575381e-01 -3.32965136e-01 -1.85984984e-01 -1.32770514e+00 -5.17334715e-02 -1.27508438e+00 2.00133789e-02 -6.45936191e-01 -1.35520065e+00 8.63692820e-01 -1.55824631e-01 -1.52212000e+00 -8.46391261e-01 -2.99250692e-01 -6.17270231e-01 9.46803689e-01 -1.21846974e+00 -8.34313750e-01 2.48944685e-02 6.11779332e-01 6.83985293e-01 -6.26105368e-01 1.11480582e+00 4.39035535e-01 -4.49334800e-01 4.26076442e-01 4.61235642e-01 1.64713129e-01 8.28424573e-01 -1.28983951e+00 7.78006390e-02 9.07622576e-01 1.04659402e+00 5.58419287e-01 7.94343591e-01 -2.51985550e-01 -1.19084954e+00 -8.34667802e-01 5.43608725e-01 -4.10587490e-01 4.63659614e-01 -5.68431616e-01 -9.52337384e-01 4.63665128e-01 3.80240381e-01 -1.01222292e-01 1.34530163e+00 9.00863409e-02 -3.49119931e-01 -3.98644060e-01 -8.05060983e-01 1.06586486e-01 6.44312918e-01 -7.24015057e-01 -7.83038020e-01 2.04272479e-01 5.21652579e-01 -1.55447096e-01 -5.48745573e-01 1.06592961e-01 5.33408642e-01 -1.10229456e+00 1.02938569e+00 -4.74182278e-01 3.85979027e-01 -4.43413317e-01 -3.21399033e-01 -1.61129439e+00 -2.15490162e-01 -1.03767180e+00 -5.51524520e-01 1.49140000e+00 5.08046031e-01 -2.65189439e-01 4.75420743e-01 2.60273188e-01 -4.81256992e-02 -2.38790393e-01 -8.52702439e-01 -8.44580948e-01 -1.32512480e-01 -7.18728244e-01 5.76653600e-01 9.93474901e-01 -1.65065661e-01 4.93530542e-01 -6.23440206e-01 5.11410177e-01 7.98286915e-01 4.83699650e-01 9.49681520e-01 -1.31703341e+00 -9.41788971e-01 -3.65037262e-01 -4.34784554e-02 -9.65650916e-01 4.50523973e-01 -9.09092009e-01 1.62101239e-01 -1.35277581e+00 -3.71997952e-01 -2.86653012e-01 -6.28686488e-01 1.87531412e-01 3.10096089e-02 2.52382070e-01 1.92719743e-01 2.64456123e-01 -2.68688560e-01 5.41327417e-01 6.14370763e-01 -3.37797403e-01 -4.44176257e-01 3.00562054e-01 -7.86593854e-01 1.07909799e+00 8.31848025e-01 -6.48333967e-01 -6.81045890e-01 -6.08088553e-01 -1.33492732e-02 4.07397419e-01 2.73259997e-01 -1.36810052e+00 2.66972333e-01 1.22864544e-01 5.01409590e-01 -4.02365744e-01 9.00340378e-01 -9.63321745e-01 4.05023307e-01 2.36508828e-02 -5.15428722e-01 -6.28662407e-01 2.13740557e-01 6.74161434e-01 -5.78069150e-01 -2.75309414e-01 5.36868572e-01 3.70549075e-02 1.09622329e-02 -1.67888552e-01 -3.60857606e-01 -1.40586615e-01 3.60717297e-01 -5.59615120e-02 4.35402900e-01 -8.22196782e-01 -1.06849456e+00 -4.48635072e-01 -1.20426647e-01 3.39692265e-01 5.74744880e-01 -1.35840535e+00 -9.81604159e-01 5.75811446e-01 -4.32895273e-01 7.90547207e-02 2.40680993e-01 6.55387998e-01 7.78728127e-02 1.93218991e-01 -7.69506618e-02 -5.74197054e-01 -1.24163067e+00 5.21489084e-01 3.80952120e-01 -7.69165298e-03 -3.06365401e-01 1.21828067e+00 2.26086304e-01 -4.76246059e-01 5.42804480e-01 -4.55515265e-01 -1.37729332e-01 3.55625600e-01 6.40950143e-01 4.66280013e-01 6.44223019e-02 -8.04530025e-01 -1.53501183e-01 2.26894647e-01 3.50856453e-01 -6.82074726e-01 1.57928705e+00 1.44680336e-01 -8.83261487e-02 9.92071450e-01 1.07168388e+00 6.14341140e-01 -1.37969732e+00 -4.30235714e-01 -8.99374112e-02 -4.36865956e-01 2.03499153e-01 -7.31433392e-01 -1.04332411e+00 1.19589186e+00 3.75276148e-01 3.85325938e-01 1.33424056e+00 -2.36650184e-01 6.30733311e-01 2.84314334e-01 1.08052760e-01 -7.59913266e-01 1.03115864e-01 1.45998240e-01 1.07648349e+00 -1.10442221e+00 -3.14335048e-01 -1.63104162e-01 -3.48384172e-01 1.03620684e+00 1.04445629e-01 -4.82120544e-01 7.63459086e-01 5.69708705e-01 3.56185466e-01 1.93117365e-01 -5.59991419e-01 -9.60346535e-02 7.67395854e-01 5.42480350e-01 4.01355445e-01 1.48773223e-01 5.87966263e-01 1.26651132e+00 -5.91341257e-01 -9.41364914e-02 2.22984791e-01 5.52201033e-01 -3.97963643e-01 -1.20710218e+00 -8.47770095e-01 3.68132651e-01 -6.58965111e-01 -2.14067206e-01 -4.91636127e-01 5.19746505e-02 1.44899398e-01 1.37498748e+00 -6.29823133e-02 -2.93549687e-01 -1.94714125e-02 4.78717476e-01 4.92626250e-01 -6.43091083e-01 -4.45456654e-01 9.64812279e-01 2.31018700e-02 -6.81449920e-02 -6.76705539e-01 -6.27901733e-01 -1.01405394e+00 2.58387834e-01 -4.22148883e-01 3.99407744e-01 6.64776266e-01 8.23761523e-01 3.23211253e-01 8.04169536e-01 6.40160084e-01 -1.35307753e+00 -5.49407721e-01 -1.17129195e+00 -8.50232601e-01 1.42470255e-01 9.29142833e-01 -4.49072540e-01 -4.69504982e-01 6.73193276e-01]
[15.397066116333008, 5.582434177398682]
21f3c233-3b5e-4722-9529-0c03b34a14dc
webcpm-interactive-web-search-for-chinese
2305.06849
null
https://arxiv.org/abs/2305.06849v2
https://arxiv.org/pdf/2305.06849v2.pdf
WebCPM: Interactive Web Search for Chinese Long-form Question Answering
Long-form question answering (LFQA) aims at answering complex, open-ended questions with detailed, paragraph-length responses. The de facto paradigm of LFQA necessitates two procedures: information retrieval, which searches for relevant supporting facts, and information synthesis, which integrates these facts into a coherent answer. In this paper, we introduce WebCPM, the first Chinese LFQA dataset. One unique feature of WebCPM is that its information retrieval is based on interactive web search, which engages with a search engine in real time. Following WebGPT, we develop a web search interface. We recruit annotators to search for relevant information using our interface and then answer questions. Meanwhile, the web search behaviors of our annotators would be recorded. In total, we collect 5,500 high-quality question-answer pairs, together with 14,315 supporting facts and 121,330 web search actions. We fine-tune pre-trained language models to imitate human behaviors for web search and to generate answers based on the collected facts. Our LFQA pipeline, built on these fine-tuned models, generates answers that are no worse than human-written ones in 32.5% and 47.5% of the cases on our dataset and DuReader, respectively.
['Jie zhou', 'Maosong Sun', 'Zhiyuan Liu', 'Fanchao Qi', 'Ruobing Xie', 'Huadong Wang', 'Ning Ding', 'Xu Han', 'Yankai Lin', 'Kunlun Zhu', 'Shihao Liang', 'Lan Yan', 'Dian Jin', 'Zihan Cai', 'Yujia Qin']
2023-05-11
null
null
null
null
['long-form-question-answering']
['natural-language-processing']
[-2.00268716e-01 2.54242241e-01 9.79649555e-03 -3.87268364e-01 -1.87430346e+00 -1.18949056e+00 4.01056647e-01 1.22326948e-01 -5.64637482e-01 7.73679852e-01 5.41433096e-01 -4.97347444e-01 -3.25175710e-02 -8.64662230e-01 -6.04314506e-01 1.62666589e-01 5.65593302e-01 1.02405059e+00 5.82064807e-01 -5.90863824e-01 3.27508181e-01 -3.20526361e-01 -1.09129846e+00 9.51765180e-01 1.45808733e+00 1.16173005e+00 3.58894229e-01 1.07406723e+00 -7.23875284e-01 9.16003466e-01 -8.90458822e-01 -1.25841486e+00 -1.49725422e-01 -2.98820525e-01 -1.51092625e+00 -4.35257375e-01 3.82134140e-01 -4.33173597e-01 -1.22202344e-01 6.64445281e-01 4.52188432e-01 -4.41946127e-02 5.03991663e-01 -7.20169067e-01 -1.02388442e+00 7.19635904e-01 1.37278408e-01 1.87253252e-01 1.28519499e+00 3.91871363e-01 1.48162842e+00 -9.37557340e-01 4.84493583e-01 1.13227177e+00 2.14504138e-01 6.22077048e-01 -7.69664884e-01 -4.02185440e-01 -7.93278366e-02 1.80529922e-01 -9.61859286e-01 -1.94922403e-01 4.93254036e-01 -2.03705311e-01 9.33532476e-01 6.99722052e-01 3.62857610e-01 1.17615664e+00 9.12613049e-02 8.05157006e-01 7.84505486e-01 -4.99032676e-01 3.62603404e-02 2.03439534e-01 4.62588042e-01 4.84607190e-01 -4.00145911e-02 -5.48527598e-01 -5.94248712e-01 -4.97341007e-01 2.41161406e-01 -3.50845098e-01 -3.75839770e-01 4.71576601e-01 -1.05899465e+00 9.50604022e-01 2.97682524e-01 2.64992058e-01 -5.79493105e-01 -3.01691145e-01 1.60507277e-01 4.20369178e-01 1.00315154e-01 1.01978528e+00 -7.89312005e-01 -3.48102957e-01 -3.20437223e-01 5.69616258e-01 1.49542642e+00 1.15772319e+00 6.51801288e-01 -6.23064637e-01 -6.27795815e-01 1.07085001e+00 2.74948180e-01 9.10274327e-01 7.01155603e-01 -1.25830555e+00 9.72000957e-01 1.05650914e+00 5.30122876e-01 -1.08694267e+00 -2.05162555e-01 -1.95698112e-01 -2.68696457e-01 -9.42906439e-01 5.13205886e-01 -3.68119389e-01 -3.94337744e-01 1.46865845e+00 9.02498588e-02 -7.77040839e-01 1.72822252e-01 9.25288856e-01 1.41133475e+00 9.05352473e-01 1.35280699e-01 -6.14523180e-02 1.89849794e+00 -1.11394083e+00 -9.78459060e-01 -3.94303203e-01 4.42684084e-01 -1.00628722e+00 1.91330755e+00 3.05271238e-01 -1.32801926e+00 -4.08933580e-01 -4.77478623e-01 -4.25948083e-01 -3.42858076e-01 9.26589817e-02 2.60977328e-01 4.18597370e-01 -8.48743737e-01 -2.40013897e-01 -3.02164722e-02 -1.74079895e-01 -1.18526943e-01 -2.31248364e-02 1.51801378e-01 -1.82340920e-01 -1.69668055e+00 7.60866404e-01 1.15443677e-01 -2.65232533e-01 -5.52576184e-01 -6.33880615e-01 -6.21445835e-01 1.82622492e-01 8.56624722e-01 -8.44191253e-01 2.03430200e+00 -3.04725945e-01 -1.45547569e+00 7.73602426e-01 -4.14170682e-01 -1.80502728e-01 1.29288599e-01 -5.36347449e-01 -5.99807262e-01 4.12480086e-01 5.22369981e-01 5.69195867e-01 3.10195982e-01 -1.02398491e+00 -6.11129642e-01 -5.33768423e-02 5.61731040e-01 2.26552650e-01 -2.30584875e-01 1.12884149e-01 -9.76956367e-01 -3.55936974e-01 -2.08216384e-01 -5.26763320e-01 -7.42467418e-02 -5.26532173e-01 -3.73674959e-01 -7.73447990e-01 2.02922806e-01 -1.06829381e+00 1.69585860e+00 -1.66311324e+00 -2.28424236e-01 6.81801140e-02 2.05676287e-01 2.45446756e-01 -5.52548409e-01 8.89425933e-01 5.71260095e-01 3.43281627e-01 2.34095827e-02 1.33864611e-01 3.61986428e-01 1.92182839e-01 -6.88174367e-01 -7.60417283e-01 1.41321287e-01 1.38363266e+00 -1.08570802e+00 -7.57720351e-01 -4.89954501e-01 -2.92141169e-01 -8.63988936e-01 7.34497547e-01 -9.39571500e-01 2.97382146e-01 -7.88359880e-01 6.48531854e-01 2.53921390e-01 -6.63345993e-01 -9.78910178e-02 6.52612671e-02 1.34067744e-01 9.03726518e-01 -5.69966257e-01 1.74538565e+00 -8.53807509e-01 2.45221481e-01 -1.32345095e-01 -1.40117466e-01 8.38855922e-01 4.24292207e-01 -6.37180684e-03 -1.33106840e+00 -6.59047812e-02 5.84458113e-01 -3.50422710e-01 -1.10968494e+00 6.00755036e-01 4.23444420e-01 -5.51477611e-01 6.79003179e-01 1.14263989e-01 -6.07384622e-01 5.65991759e-01 4.97696280e-01 1.35435832e+00 -2.97362477e-01 6.79124668e-02 -1.22759137e-02 9.16354835e-01 3.51689786e-01 1.06784321e-01 1.20323551e+00 3.95636819e-02 4.21262980e-01 4.95300651e-01 -3.99992943e-01 -6.35580719e-01 -1.03028452e+00 3.41604590e-01 1.21252477e+00 -1.02386057e-01 -6.41420782e-01 -9.82439876e-01 -9.98526394e-01 -3.67358297e-01 9.21213090e-01 -1.74003869e-01 3.80485058e-02 -6.65807307e-01 1.23628294e-02 4.81238216e-01 6.30748048e-02 7.51368463e-01 -1.48030198e+00 -4.31926817e-01 4.79465544e-01 -1.22232664e+00 -1.13676441e+00 -9.74497974e-01 -3.05953741e-01 -3.42339516e-01 -1.35056484e+00 -6.61131620e-01 -8.48530054e-01 1.15107886e-01 8.64463523e-02 1.95247042e+00 4.12056208e-01 2.07804084e-01 7.36021757e-01 -7.68399417e-01 -2.81743377e-01 -4.37422097e-01 5.10945499e-01 -5.62114418e-01 -2.59551048e-01 6.45565987e-01 -1.61402702e-01 -8.14872921e-01 5.14897048e-01 -1.02122927e+00 -3.13681886e-02 6.86346471e-01 6.54036760e-01 3.07661414e-01 -3.76264364e-01 9.18938875e-01 -8.82708609e-01 1.40909195e+00 -6.41191661e-01 -4.77192521e-01 8.82571697e-01 -2.59924173e-01 1.30391046e-01 6.63413823e-01 -3.81342620e-01 -1.21369207e+00 -3.58661592e-01 -7.96676993e-01 3.65783721e-01 -2.87985429e-02 8.63076389e-01 -1.16972968e-01 2.77259737e-01 1.02306628e+00 2.64556170e-01 -4.60841060e-01 -4.47992831e-01 5.65778434e-01 9.72181857e-01 6.00516438e-01 -9.54645693e-01 6.50502622e-01 -1.48422629e-01 -9.43120778e-01 -3.84842187e-01 -1.46414053e+00 -6.48600876e-01 -8.83973464e-02 -3.10581714e-01 9.36421275e-01 -6.25976205e-01 -1.18860257e+00 1.26426727e-01 -1.50141418e+00 -4.85405356e-01 -9.31214318e-02 4.77334894e-02 -4.63639766e-01 1.93515092e-01 -7.09431767e-01 -7.48291135e-01 -7.30337262e-01 -9.94914711e-01 1.03757751e+00 3.71925831e-01 -6.03760600e-01 -7.24910975e-01 1.65697083e-01 1.16685212e+00 4.97541398e-01 -4.65920776e-01 1.20079601e+00 -9.88040090e-01 -4.77193564e-01 -3.34721446e-01 -1.15521230e-01 6.81797415e-02 -1.76399741e-02 -2.99000382e-01 -5.69128931e-01 3.14263552e-01 7.38791823e-02 -7.79026091e-01 3.13128233e-01 -2.25596637e-01 1.17714489e+00 -9.64823604e-01 1.79449558e-01 -1.80553690e-01 1.01844728e+00 3.52001280e-01 5.10311425e-01 1.92287400e-01 1.45672202e-01 9.36637461e-01 7.32637167e-01 1.34305149e-01 9.22532558e-01 6.50324166e-01 1.74183249e-01 4.88770187e-01 3.74635984e-03 -6.45802557e-01 2.05879927e-01 1.24276376e+00 4.08574849e-01 -5.70975542e-01 -1.08868217e+00 5.96449733e-01 -1.59864771e+00 -6.61491811e-01 -2.95770288e-01 1.68966281e+00 1.47664511e+00 8.02846029e-02 -3.01767942e-02 -3.65973383e-01 2.21776083e-01 -2.24347368e-01 -4.29598778e-01 -3.23740095e-01 2.61110980e-02 4.43531066e-01 -2.02629089e-01 7.37773240e-01 -5.42781174e-01 8.90421510e-01 5.95276260e+00 8.02806497e-01 -4.12343323e-01 4.94396910e-02 5.78990519e-01 2.06964314e-01 -8.59452248e-01 -1.71741307e-01 -7.78796911e-01 5.73129117e-01 1.04481137e+00 -2.61370152e-01 6.30386949e-01 7.47759163e-01 3.71161923e-02 -1.23405859e-01 -8.91078234e-01 7.56306112e-01 9.54815447e-02 -1.32939982e+00 5.48457086e-01 -4.40493137e-01 6.68494642e-01 -3.24160725e-01 -6.64542168e-02 9.30068016e-01 6.44815028e-01 -7.64606476e-01 6.23702526e-01 6.68856680e-01 4.05043572e-01 -4.54347819e-01 9.52466607e-01 8.19594860e-01 -9.97627914e-01 -2.02623174e-01 -7.14807734e-02 -5.51077817e-03 3.34089965e-01 3.95174474e-01 -7.06932724e-01 5.73658526e-01 9.35104430e-01 -1.89824238e-01 -8.14298213e-01 8.25501919e-01 -6.11413658e-01 7.22995877e-01 -3.09605598e-01 -7.92078853e-01 3.08267146e-01 -2.00883865e-01 2.19600230e-01 1.03539574e+00 1.43526763e-01 6.43860161e-01 1.94817781e-02 8.51096213e-01 -4.55116570e-01 4.06589717e-01 -2.28063568e-01 -1.71157330e-01 7.47800469e-01 1.19683146e+00 -6.98323324e-02 -5.29205322e-01 -5.80561340e-01 9.02212083e-01 5.15661657e-01 5.66906929e-01 -6.31876647e-01 -5.44341445e-01 3.73456366e-02 -2.18459629e-02 -1.35439530e-01 2.24579405e-02 -9.70910415e-02 -1.38547051e+00 5.27305782e-01 -1.66371143e+00 8.27172220e-01 -1.20546830e+00 -1.57479525e+00 7.83950269e-01 -2.24817216e-01 -6.69966459e-01 -6.81524396e-01 -3.07842642e-01 -5.62248826e-01 1.01949859e+00 -1.42232561e+00 -9.06669796e-01 -4.79280293e-01 7.49236703e-01 1.01957691e+00 2.27803677e-01 8.56289744e-01 5.13704300e-01 -1.76321834e-01 5.07608712e-01 -5.89143276e-01 1.45632863e-01 7.89440572e-01 -1.26185334e+00 4.56942171e-01 5.00267386e-01 2.32962996e-01 9.64724898e-01 4.23804462e-01 -7.24654675e-01 -1.42698514e+00 -7.16377079e-01 1.67929339e+00 -9.88068342e-01 7.96975613e-01 -4.35176343e-01 -1.21887898e+00 4.72933590e-01 5.50690889e-01 -6.89712763e-01 8.26031029e-01 3.25462446e-02 -1.95640683e-01 1.27288536e-03 -8.60555947e-01 8.75549376e-01 7.13859499e-01 -7.27455199e-01 -1.23415363e+00 5.67595541e-01 1.32946134e+00 -5.40782452e-01 -7.11572886e-01 1.70433015e-01 3.25230747e-01 -5.34386456e-01 8.15462112e-01 -7.85773396e-01 5.29434800e-01 -9.66259837e-02 -1.72270760e-01 -9.86892521e-01 -1.02901332e-01 -7.05933452e-01 -1.27893969e-01 1.28895128e+00 1.06050265e+00 -5.85156620e-01 4.20467526e-01 1.05784583e+00 -1.09728880e-01 -8.31266522e-01 -7.45796025e-01 -3.52342099e-01 5.47522157e-02 -3.89806479e-01 7.56323934e-01 4.78376001e-01 9.45519805e-02 5.97968757e-01 1.21775545e-01 3.33818123e-02 1.10785831e-02 2.03067869e-01 8.63269448e-01 -8.51338923e-01 -3.80299419e-01 -3.27254593e-01 8.46793115e-01 -1.62646401e+00 5.57484198e-03 -5.90215027e-01 8.83921906e-02 -1.77806580e+00 1.55651331e-01 -2.54678994e-01 3.47315729e-01 3.74243796e-01 -6.01053655e-01 -1.65866949e-02 -3.30288671e-02 2.05441162e-01 -1.00678313e+00 5.32450199e-01 1.33546257e+00 -1.23874813e-01 -9.89113748e-02 1.79199338e-01 -9.47486103e-01 5.34979641e-01 6.87864780e-01 -2.70806015e-01 -4.00147796e-01 -7.46103704e-01 9.17334676e-01 5.01053274e-01 1.39282316e-01 -6.64782643e-01 4.51424897e-01 -3.02603334e-01 -5.87891862e-02 -7.86391079e-01 1.93197057e-01 -7.11221337e-01 -3.56934667e-01 3.42942387e-01 -6.97083592e-01 3.36705744e-01 6.32096231e-02 1.95291236e-01 -4.42629904e-01 -7.09797561e-01 -2.48209629e-02 -3.93511027e-01 -4.20398951e-01 1.59898862e-01 -6.57146156e-01 7.66167104e-01 4.46016252e-01 3.36520702e-01 -6.15663052e-01 -1.00312042e+00 -3.13941807e-01 7.84291387e-01 -1.33532077e-01 5.58947921e-01 6.01039350e-01 -1.23011243e+00 -7.53423393e-01 -2.47054696e-01 3.83968771e-01 5.53233624e-02 2.24283412e-01 1.83418557e-01 -4.20688003e-01 1.01828957e+00 3.97696048e-01 -1.96691841e-01 -6.49262071e-01 4.42686200e-01 1.01361528e-01 -6.80502713e-01 -3.06378268e-02 7.43767440e-01 -5.89060299e-02 -7.99249589e-01 1.61035001e-01 -3.56629729e-01 -5.87780476e-01 3.56876925e-02 7.05771625e-01 2.58315075e-02 1.48952991e-01 6.86775446e-02 -7.15024024e-02 5.26504040e-01 -1.31858075e-02 -4.71312732e-01 8.28833044e-01 -2.55065888e-01 -2.04521835e-01 2.13797092e-01 1.02425790e+00 4.61832881e-01 -6.62006378e-01 -4.65624481e-01 3.71774048e-01 -2.04922512e-01 -5.95283091e-01 -1.43906796e+00 -6.14919484e-01 5.62156737e-01 -2.29852468e-01 5.51119566e-01 1.16192293e+00 4.53839034e-01 1.36035740e+00 1.02258301e+00 4.76728559e-01 -9.64622796e-01 6.84178889e-01 8.96266580e-01 1.36895287e+00 -1.27680182e+00 -6.73253715e-01 -3.69336754e-01 -6.24938726e-01 1.03836942e+00 9.98347342e-01 2.99236089e-01 6.60053343e-02 -2.70936430e-01 4.52928245e-01 -4.55516845e-01 -1.18083954e+00 -1.01830050e-01 6.36770189e-01 3.06676179e-01 3.49461764e-01 -2.17216581e-01 -5.39170146e-01 1.24422383e+00 -6.74196839e-01 -1.53518558e-01 2.53928870e-01 7.21754372e-01 -5.15862644e-01 -1.01025498e+00 -2.34455347e-01 5.54340363e-01 -5.66851139e-01 -3.65107983e-01 -7.16932654e-01 3.70239973e-01 -5.32782257e-01 1.69047201e+00 -2.05282941e-01 -2.39900857e-01 7.86319196e-01 2.36026406e-01 -1.97834089e-01 -6.76913738e-01 -1.07429206e+00 -5.49934089e-01 5.27646780e-01 -6.09187901e-01 1.56007424e-01 -1.79537430e-01 -1.08282304e+00 -6.76333979e-02 -2.64496982e-01 9.15993035e-01 4.81685787e-01 9.04293180e-01 4.78136271e-01 1.50483236e-01 5.44122577e-01 1.78302541e-01 -7.38594711e-01 -1.12051034e+00 3.11823875e-01 4.58838433e-01 9.47151855e-02 -4.44338731e-02 -3.39519024e-01 3.44306091e-03]
[11.503085136413574, 8.017892837524414]
895e2d70-d29a-45a3-ad7d-390556b3fceb
hybrid-and-collaborative-passage-reranking
2305.09313
null
https://arxiv.org/abs/2305.09313v1
https://arxiv.org/pdf/2305.09313v1.pdf
Hybrid and Collaborative Passage Reranking
In passage retrieval system, the initial passage retrieval results may be unsatisfactory, which can be refined by a reranking scheme. Existing solutions to passage reranking focus on enriching the interaction between query and each passage separately, neglecting the context among the top-ranked passages in the initial retrieval list. To tackle this problem, we propose a Hybrid and Collaborative Passage Reranking (HybRank) method, which leverages the substantial similarity measurements of upstream retrievers for passage collaboration and incorporates the lexical and semantic properties of sparse and dense retrievers for reranking. Besides, built on off-the-shelf retriever features, HybRank is a plug-in reranker capable of enhancing arbitrary passage lists including previously reranked ones. Extensive experiments demonstrate the stable improvements of performance over prevalent retrieval and reranking methods, and verify the effectiveness of the core components of HybRank.
['Houqiang Li', 'Jiaxin Shi', 'Wengang Zhou', 'Zongmeng Zhang']
2023-05-16
null
null
null
null
['passage-retrieval']
['natural-language-processing']
[-3.55116218e-01 -5.20635068e-01 -2.62638032e-01 1.36666685e-01 -1.26854312e+00 -7.29502618e-01 7.42926419e-01 8.23185980e-01 -4.13961977e-01 9.15873468e-01 1.00062680e+00 1.07693616e-02 -8.19820583e-01 -8.54188740e-01 -2.60200381e-01 -4.07554388e-01 -3.87475431e-01 4.36399043e-01 8.04955542e-01 -7.16214895e-01 5.84018111e-01 3.81797343e-01 -1.55622995e+00 6.45133913e-01 1.03766513e+00 5.74207067e-01 2.07922727e-01 7.09064960e-01 -4.41953987e-01 5.36207676e-01 -4.90605235e-01 1.86703522e-02 1.19598925e-01 -4.03973877e-01 -1.13003421e+00 -6.51943207e-01 2.86700428e-01 -4.55452919e-01 -8.13467205e-01 6.80380523e-01 4.90195781e-01 7.57986009e-01 4.86654162e-01 -7.86732197e-01 -6.17745161e-01 8.66658568e-01 -1.50720432e-01 8.43983114e-01 1.06774986e+00 -5.14550030e-01 1.79157758e+00 -1.06713271e+00 6.31956041e-01 8.69599819e-01 5.85944712e-01 -1.08288899e-01 -8.89462948e-01 -2.40755305e-01 2.42355064e-01 3.52456212e-01 -1.66362095e+00 -2.05307588e-01 4.21718895e-01 -1.05122708e-01 9.77269709e-01 7.68175900e-01 7.50510037e-01 5.51149786e-01 -3.85281742e-01 1.06503832e+00 3.95135224e-01 -2.35662922e-01 2.69427188e-02 -1.31174028e-01 6.10221684e-01 3.64142418e-01 1.98642090e-01 -2.43872434e-01 -5.60228348e-01 -7.77760386e-01 6.08936846e-01 3.85379225e-01 -6.74401462e-01 -1.50330171e-01 -1.27124667e+00 4.86923546e-01 5.30670702e-01 6.73466146e-01 -3.10712516e-01 -5.44272244e-01 3.32609683e-01 5.72884142e-01 4.32786226e-01 9.76674616e-01 -4.26619500e-01 -6.39225915e-02 -1.47265971e+00 3.99492502e-01 7.82823384e-01 9.48107839e-01 9.72178757e-01 -9.58379507e-01 -1.11407983e+00 1.02874172e+00 3.34972322e-01 1.84794307e-01 5.95482051e-01 -4.42903966e-01 3.37406993e-01 9.27086234e-01 4.35207605e-01 -1.07074571e+00 -3.16836745e-01 -5.17237723e-01 -5.76065660e-01 -8.55336964e-01 -1.51227951e-01 4.91089076e-01 -6.25151753e-01 1.03940916e+00 3.46000850e-01 2.31752187e-01 -2.72635873e-02 9.26019311e-01 1.18511736e+00 8.54944050e-01 -1.24030504e-02 -3.67794782e-01 9.40149903e-01 -1.34085381e+00 -2.58899987e-01 5.20982265e-01 8.54097188e-01 -9.56406891e-01 1.16908932e+00 -1.96412072e-01 -1.11039996e+00 -4.74454433e-01 -8.29215825e-01 -1.94861770e-01 -2.11362466e-01 -5.18960990e-02 3.89115959e-01 1.75090749e-02 -1.32142556e+00 7.30755448e-01 -2.20740750e-01 -4.74295735e-01 -1.93959981e-01 1.56515956e-01 -2.59117156e-01 -2.36149430e-01 -1.73379993e+00 4.72780615e-01 4.27991629e-01 -1.91590905e-01 -6.16172731e-01 -1.25653172e+00 -3.57340783e-01 5.44223189e-01 2.12567657e-01 -1.02656841e+00 9.25228775e-01 -4.28586394e-01 -1.15574467e+00 4.29532290e-01 -2.77429700e-01 -5.12031801e-02 3.29002619e-01 -4.99383003e-01 -5.90586782e-01 5.53245902e-01 2.07597315e-01 4.60259207e-02 4.26586479e-01 -1.14227128e+00 -9.17995632e-01 1.59825638e-01 4.13920730e-01 7.79068887e-01 -5.15459061e-01 -3.48723843e-03 -9.89556611e-01 -5.40726960e-01 1.82145074e-01 -4.10973430e-01 -3.14169765e-01 -6.48592055e-01 -3.06518137e-01 -5.22716522e-01 3.70233029e-01 -4.50702995e-01 1.91613841e+00 -2.13028741e+00 -4.46562283e-02 6.28348410e-01 4.36460167e-01 3.78398061e-01 -5.73825061e-01 1.16701126e+00 3.64027679e-01 3.46816868e-01 4.05962944e-01 1.51765929e-03 -4.36871545e-03 -1.84959829e-01 -5.38761616e-01 1.94247097e-01 -4.81566548e-01 1.08583105e+00 -1.36716199e+00 -7.86067545e-01 -4.27183732e-02 1.45860404e-01 -6.73672199e-01 4.10088152e-01 -1.45368371e-02 3.72646332e-01 -9.88485396e-01 5.00680387e-01 2.78164744e-01 -5.01896203e-01 -1.53566644e-01 -1.48400560e-01 -7.91438445e-02 7.25893080e-01 -9.96535122e-01 1.77916980e+00 -1.94727570e-01 1.91791162e-01 -3.34376246e-01 -4.80549842e-01 5.93738735e-01 3.22991133e-01 7.86959767e-01 -8.68090451e-01 -3.08187783e-01 3.21979105e-01 -6.32699728e-01 -5.41674137e-01 1.24888682e+00 3.75854671e-01 1.23313377e-02 6.27472937e-01 -1.25870332e-01 3.16015869e-01 6.50173783e-01 9.63015854e-01 1.41207242e+00 -1.10598966e-01 3.62421036e-01 -1.82490066e-01 8.81027460e-01 8.96264985e-02 1.03088856e-01 1.51401675e+00 1.15096942e-01 8.15999746e-01 -1.68773279e-01 -4.25412655e-02 -9.40322638e-01 -1.07682431e+00 -2.03471221e-02 1.48881757e+00 6.63409591e-01 -8.91681015e-01 -2.00574070e-01 -7.34874547e-01 7.52598047e-02 1.29722208e-01 -3.36448640e-01 -1.96015343e-01 -5.89642882e-01 -5.87383926e-01 5.59591293e-01 2.86478341e-01 3.02187771e-01 -8.37153912e-01 4.04373169e-01 2.95039386e-01 -6.49891376e-01 -5.78994572e-01 -7.93462694e-01 -5.43780386e-01 -8.09064627e-01 -1.03329813e+00 -1.00580227e+00 -7.38788188e-01 5.59801340e-01 8.92718434e-01 1.43245566e+00 9.91316259e-01 6.97636530e-02 7.35693574e-01 -1.11925840e+00 5.83783746e-01 5.47825471e-02 6.43259823e-01 2.06449572e-02 -2.86516905e-01 1.48035750e-01 -6.05023861e-01 -1.09676301e+00 3.78860444e-01 -9.92168903e-01 -4.22532529e-01 4.93630707e-01 7.98791409e-01 5.60859859e-01 4.60744426e-02 7.55926967e-01 -6.03071928e-01 9.74280000e-01 -7.44567633e-01 -4.76970188e-02 6.46046042e-01 -8.13954234e-01 2.01850776e-02 4.20435041e-01 -2.42189094e-01 -1.02069533e+00 -7.40191162e-01 -3.09318751e-01 1.18000463e-01 1.47381678e-01 8.99583638e-01 1.62113696e-01 -8.49265158e-02 3.72326553e-01 4.88674611e-01 -6.95882380e-01 -9.20960128e-01 5.63740253e-01 6.41043782e-01 1.79690421e-01 -6.00385547e-01 8.67571473e-01 2.97005355e-01 -3.20143282e-01 -6.97885454e-01 -1.00802350e+00 -1.60666120e+00 -5.48361540e-01 -1.71016335e-01 1.76363856e-01 -1.03124487e+00 -4.87067640e-01 2.47444622e-02 -6.93971634e-01 2.57210821e-01 -5.51667154e-01 5.37561715e-01 -3.00177820e-02 9.39005315e-01 -8.24841678e-01 -3.36407036e-01 -4.70840544e-01 -5.43088198e-01 9.89757895e-01 3.03406000e-01 -2.16282532e-01 -9.92082417e-01 6.06723964e-01 2.87005991e-01 4.79157567e-01 -6.25665307e-01 1.04944849e+00 -1.38368857e+00 -7.61976480e-01 -5.01397669e-01 -3.71860176e-01 -3.47864121e-01 8.28373432e-02 -1.63523078e-01 -4.85741079e-01 -6.04144692e-01 -5.77620327e-01 -1.20849930e-01 1.15302336e+00 6.27518445e-02 5.81580520e-01 -4.52291578e-01 -6.60202444e-01 1.79528296e-01 1.27905226e+00 -2.76277333e-01 6.72764957e-01 6.37462795e-01 5.96545815e-01 1.20387815e-01 6.54961050e-01 5.86764634e-01 4.75566655e-01 5.05250990e-01 -2.97713727e-02 -1.45277828e-01 -9.20105278e-02 -6.76816285e-01 1.22508474e-01 1.40643942e+00 2.14238353e-02 -4.65550542e-01 -5.02493083e-01 7.58478820e-01 -1.83194435e+00 -1.08823276e+00 -1.69862941e-01 2.31353521e+00 9.76935685e-01 -2.60743707e-01 1.86688676e-01 2.60777827e-02 6.55870438e-01 3.03292453e-01 -1.21334031e-01 3.22147608e-01 -1.38083205e-01 -5.03587211e-03 8.01963508e-02 6.83391631e-01 -9.02725399e-01 9.14383769e-01 6.99163437e+00 1.06078649e+00 -4.43076968e-01 -2.13289753e-01 -6.13151900e-02 -1.39200777e-01 -7.74047017e-01 2.20169023e-01 -1.04974449e+00 1.65900514e-01 5.15318394e-01 -6.58747137e-01 3.33160460e-01 4.67909604e-01 1.90404430e-01 -1.03598058e-01 -1.13622105e+00 5.16037643e-01 2.17534557e-01 -1.45394862e+00 6.33536518e-01 -4.19168353e-01 8.70239735e-01 1.63170353e-01 -2.84928828e-01 7.09345281e-01 4.01733726e-01 -4.67106938e-01 2.05535427e-01 7.23852575e-01 1.59496352e-01 -5.13984501e-01 7.86025941e-01 3.13892752e-01 -1.42534971e+00 5.66406883e-02 -3.61668199e-01 2.49117408e-02 2.17098027e-01 6.15785599e-01 -6.38212621e-01 1.12915444e+00 8.27900767e-01 9.35648024e-01 -6.53929651e-01 1.91118693e+00 7.56228715e-02 6.85904980e-01 -3.94045711e-01 -2.50588447e-01 3.66886556e-01 8.04076120e-02 1.00530636e+00 1.34689558e+00 3.86816025e-01 4.34363365e-01 4.80094492e-01 1.83983341e-01 -2.14376912e-01 7.77794540e-01 -3.33982483e-02 7.80972019e-02 6.86715543e-01 1.10463166e+00 -5.68390667e-01 -5.83454311e-01 -3.97759885e-01 9.89614308e-01 3.12560529e-01 6.97289407e-01 -3.78663838e-01 -4.91306722e-01 5.67979157e-01 1.21205226e-01 3.34323972e-01 1.29549831e-01 4.14852113e-01 -1.49835885e+00 1.17151678e-01 -4.89261717e-01 8.72213960e-01 -5.85937202e-01 -1.63650393e+00 6.43483400e-01 3.56233940e-02 -1.48190320e+00 1.74204886e-01 1.06573246e-01 -4.10178155e-01 7.30631769e-01 -1.90302873e+00 -8.13761353e-01 -7.84443691e-02 6.48077667e-01 4.48352724e-01 6.84392527e-02 7.07502961e-01 6.80903196e-01 -1.64835125e-01 5.86857021e-01 6.27304673e-01 8.42918754e-02 7.59453058e-01 -1.08285451e+00 -2.31332835e-02 6.62660837e-01 4.25314456e-01 1.07797575e+00 5.10485590e-01 -5.66745341e-01 -1.05547214e+00 -8.87329936e-01 1.32614720e+00 -3.13244522e-01 8.31438959e-01 3.12068969e-01 -1.18616188e+00 2.92199582e-01 9.35333893e-02 -2.36458927e-01 9.84004140e-01 5.56332409e-01 -4.08868879e-01 -6.07870780e-02 -6.87357843e-01 7.21963406e-01 1.16123486e+00 -6.79262757e-01 -1.11630797e+00 3.35039556e-01 7.77666926e-01 -1.60022721e-01 -8.65431905e-01 4.88754600e-01 6.08208001e-01 -7.33263254e-01 1.48838580e+00 -8.02416265e-01 8.71237740e-02 -5.39108455e-01 -2.37451196e-02 -1.18234074e+00 -7.76945353e-01 -5.84913969e-01 -1.67996287e-01 1.36844492e+00 5.31167507e-01 -2.79013723e-01 6.29704356e-01 3.25366884e-01 -9.14326534e-02 -5.46921849e-01 -6.60844326e-01 -5.92226863e-01 -7.49146864e-02 6.49857074e-02 8.58850241e-01 8.31492901e-01 6.21364832e-01 3.52632523e-01 -2.51048386e-01 3.48153740e-01 1.39993995e-01 6.38499618e-01 4.64471579e-01 -1.29597843e+00 -1.92098737e-01 -5.41881979e-01 -2.11883292e-01 -1.77152979e+00 -9.63059515e-02 -1.25818872e+00 1.51703820e-01 -2.02067542e+00 7.11797655e-01 -7.84570456e-01 -1.06405365e+00 1.93031952e-01 -7.00316906e-01 3.01052004e-01 -2.70670671e-02 9.61821496e-01 -1.58798611e+00 6.40420675e-01 1.20045054e+00 -2.19273239e-01 -5.71424186e-01 1.70148425e-02 -6.28904760e-01 3.67880046e-01 2.99716651e-01 -5.09254813e-01 -5.00605822e-01 -3.88903320e-01 7.05090404e-01 8.42396915e-02 1.17138168e-02 -6.62727177e-01 7.38194346e-01 -9.41217244e-02 6.16626479e-02 -8.02393973e-01 -1.00795582e-01 -5.95239222e-01 -6.38045594e-02 -2.26213083e-01 -8.18240225e-01 7.52069652e-02 -1.16825789e-01 7.73340285e-01 -5.78796208e-01 -3.49931389e-01 1.22741804e-01 -2.05430493e-01 -8.33393931e-01 5.30809283e-01 -4.75932896e-01 3.61322910e-01 4.06511337e-01 1.82870831e-02 -2.88526148e-01 -3.24953049e-01 -6.91517651e-01 5.54706395e-01 3.80624264e-01 4.96677637e-01 6.44669652e-01 -1.48072481e+00 -8.89435947e-01 -7.15803802e-02 5.13003349e-01 -3.36251408e-01 4.07064617e-01 7.62966096e-01 -3.00412089e-01 6.43143296e-01 4.97272998e-01 -1.58718705e-01 -1.36628306e+00 4.96901900e-01 -3.85264531e-02 -8.96904647e-01 -8.61393034e-01 8.99945617e-01 -1.76802408e-02 -2.28591502e-01 1.68819293e-01 -1.58256382e-01 -7.84084678e-01 2.51019448e-01 8.25330496e-01 5.23981273e-01 2.57163763e-01 -5.35437644e-01 -2.70944774e-01 3.17136258e-01 -5.49586654e-01 -1.74394362e-02 1.01820385e+00 -8.38606775e-01 -4.65716690e-01 2.26576433e-01 1.16622031e+00 4.31067586e-01 -3.88106853e-01 -6.54591203e-01 1.52529761e-01 -5.14562607e-01 -6.03004657e-02 -7.29316890e-01 -5.79380631e-01 4.34785187e-02 -1.11065075e-01 3.82398844e-01 1.09044838e+00 2.93808095e-02 1.12064135e+00 8.79406393e-01 4.80086297e-01 -8.79299223e-01 4.71781082e-02 6.93956673e-01 7.13975966e-01 -7.77407110e-01 1.96910232e-01 -2.89799929e-01 -1.82255700e-01 6.74736023e-01 1.80853039e-01 -1.92085300e-02 8.73408198e-01 -3.14465940e-01 -1.85957197e-02 -1.37781054e-01 -6.34702921e-01 -4.07545179e-01 6.87477946e-01 -6.37907013e-02 3.93188477e-01 -3.56842697e-01 -7.92199850e-01 5.34175932e-01 1.20817132e-01 1.50947673e-02 5.43934777e-02 7.94784725e-01 -7.58846641e-01 -1.14936662e+00 -1.59065530e-01 6.07895911e-01 -5.01770616e-01 -6.84206426e-01 -1.93896607e-01 3.51697683e-01 -3.44954073e-01 1.05480719e+00 -1.47977158e-01 -2.99399644e-01 1.69616684e-01 -1.15287058e-01 5.32276742e-02 -6.14966571e-01 -9.58783329e-01 1.66653600e-02 8.90488848e-02 -5.31914771e-01 -4.25384730e-01 -4.97176945e-01 -1.08098972e+00 -1.38665900e-01 -6.99743330e-01 1.17512178e+00 -2.31761098e-01 1.07828391e+00 5.80326438e-01 2.84455627e-01 9.09658670e-01 -5.21105409e-01 -3.52009803e-01 -8.54631007e-01 -8.49646866e-01 6.98279738e-01 5.97438633e-01 -3.13238710e-01 -5.88140547e-01 -3.17182779e-01]
[11.46517276763916, 7.668643951416016]
d2558e97-44f9-44d5-a3c8-a1a05292ef9a
boundary-enhanced-co-training-for-weakly
null
null
http://openaccess.thecvf.com//content/CVPR2023/html/Rong_Boundary-Enhanced_Co-Training_for_Weakly_Supervised_Semantic_Segmentation_CVPR_2023_paper.html
http://openaccess.thecvf.com//content/CVPR2023/papers/Rong_Boundary-Enhanced_Co-Training_for_Weakly_Supervised_Semantic_Segmentation_CVPR_2023_paper.pdf
Boundary-Enhanced Co-Training for Weakly Supervised Semantic Segmentation
The existing weakly supervised semantic segmentation (WSSS) methods pay much attention to generating accurate and complete class activation maps (CAMs) as pseudo-labels, while ignoring the importance of training the segmentation networks. In this work, we observe that there is an inconsistency between the quality of the pseudo-labels in CAMs and the performance of the final segmentation model, and the mislabeled pixels mainly lie on the boundary areas. Inspired by these findings, we argue that the focus of WSSS should be shifted to robust learning given the noisy pseudo-labels, and further propose a boundary-enhanced co-training (BECO) method for training the segmentation model. To be specific, we first propose to use a co-training paradigm with two interactive networks to improve the learning of uncertain pixels. Then we propose a boundary-enhanced strategy to boost the prediction of difficult boundary areas, which utilizes reliable predictions to construct artificial boundaries. Benefiting from the design of co-training and boundary enhancement, our method can achieve promising segmentation performance for different CAMs. Extensive experiments on PASCAL VOC 2012 and MS COCO 2014 validate the superiority of our BECO over other state-of-the-art methods.
['Junjie Li', 'Zilei Wang', 'Bohai Tu', 'Shenghai Rong']
2023-01-01
null
null
null
cvpr-2023-1
['weakly-supervised-semantic-segmentation']
['computer-vision']
[ 4.49000865e-01 4.08251941e-01 -1.06270820e-01 -6.66284442e-01 -8.44511807e-01 -4.06489640e-01 3.48562628e-01 -5.53929880e-02 -4.31137979e-01 4.89630878e-01 -2.04375744e-01 -1.90843105e-01 1.46030501e-01 -6.33060515e-01 -9.08347011e-01 -6.97878063e-01 4.11518604e-01 3.16049695e-01 8.32747757e-01 -7.29756877e-02 1.97850168e-01 9.72300544e-02 -1.35704887e+00 3.26050103e-01 1.48143053e+00 1.16760743e+00 3.62311244e-01 3.30982029e-01 -2.51465559e-01 5.05231678e-01 -5.84668934e-01 -2.73678929e-01 1.73115253e-01 -3.47921491e-01 -8.88210654e-01 2.56199658e-01 2.45222583e-01 -1.44923246e-02 2.68096983e-01 1.25802028e+00 1.66163594e-01 2.85082217e-02 5.17773449e-01 -9.71862555e-01 -5.41084528e-01 7.40489662e-01 -6.27198935e-01 -1.85918227e-01 -1.76654503e-01 -3.11845331e-03 9.29004312e-01 -9.97527242e-01 4.85492259e-01 9.62530673e-01 8.21372867e-01 6.22506142e-01 -9.63254035e-01 -4.97518152e-01 6.58975601e-01 2.50943035e-01 -1.40619874e+00 -2.04527766e-01 8.82054269e-01 -3.46621841e-01 1.37166724e-01 1.35926411e-01 5.31970143e-01 9.42221463e-01 -4.04046237e-01 1.30831027e+00 1.23250818e+00 -4.93697345e-01 3.23278487e-01 2.94921100e-01 2.33368054e-01 5.93791842e-01 1.14344813e-01 -1.36422608e-02 -1.46120682e-01 2.84357995e-01 6.08158231e-01 -2.22272232e-01 -3.02260369e-01 -3.25763077e-01 -9.29977655e-01 4.92780060e-01 7.37176001e-01 4.17028606e-01 -2.33321637e-01 -1.44734547e-01 1.95409968e-01 -3.11970830e-01 7.33791292e-01 5.28299034e-01 -5.54308832e-01 2.07568467e-01 -1.13667333e+00 2.02259514e-02 4.52846736e-01 8.75142157e-01 8.75401437e-01 -1.12540178e-01 -2.79372215e-01 1.07245100e+00 3.75411123e-01 2.36193463e-01 2.85912931e-01 -9.61092114e-01 4.79258925e-01 7.88137615e-01 6.23322316e-02 -5.41710436e-01 -3.21178436e-01 -7.33892441e-01 -4.99186844e-01 9.98834893e-02 6.30008042e-01 -2.27055863e-01 -1.35246730e+00 1.55038881e+00 5.50890625e-01 3.86821300e-01 -4.97557707e-02 1.07214308e+00 6.36138022e-01 5.14146030e-01 1.68124586e-01 1.95671022e-01 9.82626915e-01 -1.38582003e+00 -5.77978790e-01 -5.05010962e-01 7.13614583e-01 -4.98782635e-01 1.14382613e+00 2.39307359e-01 -7.53261626e-01 -8.39058042e-01 -1.07698309e+00 1.76233009e-01 -3.87685925e-01 3.85150522e-01 5.67102671e-01 3.41374665e-01 -9.17310894e-01 7.30435550e-01 -8.97377074e-01 -1.91767141e-02 7.87018180e-01 1.42058641e-01 3.27173211e-02 -3.02383285e-02 -1.10055673e+00 4.96639222e-01 7.36567080e-01 6.97797656e-01 -7.66891122e-01 -5.32595098e-01 -7.46207952e-01 3.34999748e-02 8.30363989e-01 -3.12191129e-01 1.04822576e+00 -1.42298710e+00 -1.38218272e+00 6.88143671e-01 -2.70734001e-02 -3.29415262e-01 6.22385800e-01 -1.40837297e-01 -5.70037812e-02 2.56816059e-01 2.16386542e-01 1.18166709e+00 7.25362420e-01 -1.79251301e+00 -8.65695596e-01 -3.24146926e-01 -1.26319109e-02 1.84057191e-01 -1.32161334e-01 -4.12235439e-01 -7.83151388e-01 -6.52522385e-01 5.13915241e-01 -9.55093503e-01 -4.48295146e-01 -1.48431003e-01 -6.93553209e-01 -3.61354202e-01 8.09648573e-01 -6.06550932e-01 1.06315243e+00 -2.07854438e+00 -6.58969656e-02 2.94202298e-01 6.73281727e-03 5.67302585e-01 -1.41029984e-01 -3.35833013e-01 1.48057684e-01 2.98445255e-01 -7.62858629e-01 -5.78167140e-01 -9.25190374e-02 3.43935072e-01 -6.83099106e-02 1.16534032e-01 5.09184182e-01 1.08140779e+00 -8.64260912e-01 -7.74572492e-01 2.51183331e-01 3.12565893e-01 -3.62783402e-01 3.74418288e-01 -4.86008525e-01 7.27763414e-01 -6.19013488e-01 9.25927103e-01 8.45322549e-01 -3.37986946e-01 -3.30247134e-02 -1.16862901e-01 -1.06746249e-01 1.26142905e-03 -1.14807487e+00 1.54646051e+00 -9.98909101e-02 1.57404944e-01 1.91342875e-01 -1.09684157e+00 9.16919708e-01 3.10816448e-02 1.88597113e-01 -6.92668855e-01 2.52072424e-01 4.18808639e-01 -9.29003209e-02 -4.31305945e-01 3.15142363e-01 2.94730160e-03 2.31567606e-01 -5.23111932e-02 9.84679675e-04 -3.05171132e-01 1.56134203e-01 3.25675979e-02 4.39286083e-01 6.83473349e-01 -3.45576793e-01 -3.00639570e-01 7.11695492e-01 7.99833983e-02 1.02343071e+00 6.93830132e-01 -5.37743866e-01 1.01752460e+00 5.38900316e-01 -3.59566845e-02 -7.75226891e-01 -7.51404703e-01 -2.72919655e-01 9.87573087e-01 5.45941532e-01 -6.84699863e-02 -1.41407037e+00 -1.17016196e+00 -3.42194200e-01 6.36640549e-01 -6.82383955e-01 -3.18351910e-02 -6.08014047e-01 -9.68730807e-01 3.60318512e-01 9.42587197e-01 7.15220630e-01 -1.28341866e+00 -1.59491062e-01 1.37623817e-01 -2.54165202e-01 -1.30189574e+00 -4.33174282e-01 3.16881061e-01 -8.62900555e-01 -1.09270072e+00 -9.16983545e-01 -1.02876616e+00 1.02118516e+00 4.20649275e-02 7.33021736e-01 2.77822495e-01 5.99418059e-02 -1.12285443e-01 -5.10553181e-01 -3.09634656e-01 -3.69529307e-01 2.63973445e-01 -4.37954277e-01 2.19792277e-01 8.01170021e-02 -2.26786077e-01 -6.48216605e-01 6.02276385e-01 -8.74600947e-01 4.32675689e-01 5.61263800e-01 9.34149921e-01 9.72126126e-01 -5.78781776e-02 7.42238879e-01 -1.37674940e+00 1.43734798e-01 -2.55705863e-01 -6.45162463e-01 3.78385991e-01 -9.46210623e-01 -1.10278219e-01 6.19838417e-01 -3.95668209e-01 -1.46007860e+00 2.76235163e-01 -5.45133352e-01 -2.68657953e-01 -4.03406650e-01 3.95443469e-01 -3.03973883e-01 -4.54640724e-02 4.84257251e-01 6.60256529e-03 -2.79488742e-01 -4.98293549e-01 2.83985049e-01 6.25273883e-01 5.28740883e-01 -6.93191409e-01 4.66471314e-01 5.62426805e-01 -3.46451074e-01 -4.60955024e-01 -1.34068429e+00 -5.79398751e-01 -7.59888232e-01 -2.70748764e-01 9.01590884e-01 -8.07598233e-01 -1.64309084e-01 7.54545867e-01 -8.06429625e-01 -8.68719935e-01 -3.37608397e-01 1.78841457e-01 -4.12361473e-01 4.05674100e-01 -6.32631421e-01 -7.20215619e-01 -1.78763643e-01 -1.29174888e+00 1.22286475e+00 5.53243279e-01 1.53836936e-01 -9.31789637e-01 -3.28842849e-01 8.49779725e-01 2.65144736e-01 1.51655897e-01 6.21249497e-01 -6.02361441e-01 -5.39687753e-01 -1.30591720e-01 -5.98733544e-01 8.06570947e-01 -1.33552447e-01 1.42629659e-02 -1.24793875e+00 1.03954785e-01 -1.60245687e-01 -4.35537666e-01 1.05399394e+00 5.96123517e-01 1.59015930e+00 1.62859425e-01 -4.77598816e-01 7.23410606e-01 1.27996016e+00 1.37269780e-01 5.67151725e-01 2.61375219e-01 9.13517535e-01 8.78472865e-01 9.43233132e-01 -2.62117721e-02 5.19075394e-01 4.75430429e-01 5.04526138e-01 -5.29733300e-01 -3.21591228e-01 -3.67608637e-01 3.90426628e-02 6.24661088e-01 1.31463051e-01 -2.03665286e-01 -7.85015643e-01 7.56919086e-01 -2.02438092e+00 -2.02373952e-01 -3.92727703e-01 1.85769546e+00 9.62743759e-01 5.29554784e-01 -9.82465893e-02 8.70998576e-02 8.94407630e-01 5.19257896e-02 -6.70100510e-01 4.68051359e-02 -1.57842427e-01 6.97748810e-02 4.66058999e-01 5.66330969e-01 -1.25897086e+00 1.34840930e+00 5.46646643e+00 1.10704315e+00 -1.10244286e+00 1.31036609e-01 1.22319591e+00 4.54692066e-01 -1.72177181e-01 1.27180189e-01 -8.13999295e-01 5.28398097e-01 5.63558698e-01 6.91972494e-01 8.24642107e-02 9.61958468e-01 1.76088199e-01 -3.44742656e-01 -7.99390793e-01 4.78562474e-01 -7.44229630e-02 -1.05504715e+00 -2.10493088e-01 -3.42487335e-01 1.19125247e+00 7.66273355e-04 -8.26693848e-02 3.42919648e-01 8.05700123e-02 -7.78647125e-01 9.56031144e-01 4.49986726e-01 4.61361498e-01 -4.12594855e-01 1.06191552e+00 3.94178241e-01 -1.07784891e+00 -8.21181610e-02 -3.13042134e-01 3.23176652e-01 1.12079233e-01 7.32133746e-01 -7.07919538e-01 4.97380346e-01 8.39427054e-01 5.95006108e-01 -7.93186784e-01 1.11778879e+00 -7.87522852e-01 9.99162614e-01 -2.97669858e-01 1.02094091e-01 4.60011989e-01 -2.96726614e-01 9.93577465e-02 1.16136348e+00 -8.64260346e-02 5.69196157e-02 2.62156695e-01 1.28496861e+00 1.45274205e-02 -3.06381378e-03 9.50422138e-02 8.86794701e-02 3.67563307e-01 1.32608545e+00 -1.26210988e+00 -5.16674936e-01 -1.95707962e-01 8.37336481e-01 3.15775663e-01 5.48676908e-01 -8.15403402e-01 -2.53078878e-01 4.28373665e-02 6.92953095e-02 3.88067782e-01 9.90725905e-02 -9.06759202e-01 -1.03988349e+00 2.24142805e-01 -5.78311503e-01 2.11226016e-01 -7.21361279e-01 -1.12153697e+00 6.06053293e-01 -9.99640226e-02 -8.87345195e-01 2.75704384e-01 -4.46466267e-01 -6.83430910e-01 6.03876710e-01 -1.93615317e+00 -1.29953837e+00 -3.96749318e-01 1.02739267e-01 6.83984339e-01 4.40325797e-01 2.48651639e-01 2.50446796e-01 -7.36177862e-01 4.64771807e-01 -1.50825754e-01 1.41351029e-01 5.89478314e-01 -1.46085632e+00 3.04094046e-01 9.34565485e-01 -3.24253030e-02 2.37704754e-01 3.91060054e-01 -7.26312220e-01 -4.64119971e-01 -1.26417434e+00 5.11773825e-01 -3.08619320e-01 3.15291524e-01 -4.68188018e-01 -1.24566495e+00 3.42639029e-01 -1.03558511e-01 2.63697147e-01 2.44628698e-01 1.59545038e-02 -1.93522568e-03 -1.33724749e-01 -1.06947458e+00 5.26506484e-01 1.04784310e+00 -1.99404076e-01 -5.35956681e-01 2.50123709e-01 8.93659472e-01 -5.07763267e-01 -4.61974651e-01 8.79060805e-01 2.51172185e-01 -9.09948647e-01 8.47897410e-01 -7.75509551e-02 2.29024947e-01 -4.56006616e-01 1.90596357e-01 -1.19584191e+00 4.00545709e-02 -3.00960559e-02 2.70989835e-01 1.54492271e+00 7.18495667e-01 -3.95116419e-01 1.11723816e+00 6.61220193e-01 -4.89464790e-01 -1.02083099e+00 -6.91693485e-01 -5.10833502e-01 8.16085748e-03 -5.41614592e-01 4.81677920e-01 8.79887998e-01 -3.52045566e-01 2.59520598e-02 1.14922803e-02 2.94525504e-01 4.68858600e-01 1.47877917e-01 4.44806099e-01 -1.17574322e+00 -2.13665724e-01 -3.53714257e-01 1.50409758e-01 -1.28970897e+00 1.95946798e-01 -7.08964288e-01 7.42052078e-01 -1.57035661e+00 -1.89672578e-02 -1.00217736e+00 -4.07496631e-01 6.87189996e-01 -6.95018530e-01 4.47702736e-01 1.53485075e-01 1.82512313e-01 -9.34280336e-01 7.29533911e-01 1.69818473e+00 -1.08273782e-01 -2.82882690e-01 2.21603811e-01 -7.13777781e-01 9.79733884e-01 6.07965589e-01 -4.23768789e-01 -3.27142835e-01 -3.86628300e-01 -8.08201507e-02 -2.30697423e-01 3.67617339e-01 -9.51058090e-01 1.20482162e-01 -7.07425401e-02 2.82167107e-01 -5.94312072e-01 1.12014964e-01 -8.49194407e-01 -5.01283348e-01 2.13876992e-01 -3.22192699e-01 -7.28819668e-01 7.75850490e-02 5.86257815e-01 -3.68047923e-01 -5.33437192e-01 9.41164434e-01 -2.32143342e-01 -8.76790404e-01 1.65696889e-01 -6.68625813e-03 2.56155133e-01 9.96545613e-01 -3.96803111e-01 -1.33259267e-01 8.27207714e-02 -8.52719069e-01 7.04149187e-01 4.24965173e-01 3.24649334e-01 4.79108632e-01 -8.80724967e-01 -3.58031631e-01 3.55147421e-01 1.16881870e-01 7.65718639e-01 2.69611478e-01 8.22328925e-01 -4.30386871e-01 1.33179620e-01 2.51818776e-01 -8.32744300e-01 -9.19416785e-01 2.62322247e-01 3.99487227e-01 -1.78024322e-01 -3.87852222e-01 1.14119899e+00 3.79344732e-01 -6.56303644e-01 5.32032549e-01 -5.00267446e-01 -3.14685881e-01 -6.55703023e-02 2.03448623e-01 1.78048894e-01 9.54255986e-04 -4.22897071e-01 -2.70028621e-01 5.87622583e-01 -1.09824851e-01 1.96911022e-01 1.18215382e+00 -2.97112018e-01 -5.86904259e-03 3.24054539e-01 8.01201284e-01 -2.54083782e-01 -1.90227401e+00 -1.54867634e-01 2.68368661e-01 -2.25296080e-01 6.52989671e-02 -1.11500776e+00 -1.31205964e+00 9.48221564e-01 6.36401236e-01 3.53821740e-02 1.10188067e+00 1.20039776e-01 9.33085501e-01 1.91889722e-02 2.67318100e-01 -1.51103055e+00 7.35928863e-02 3.14774573e-01 4.03989017e-01 -1.50047851e+00 -4.06090289e-01 -9.51134264e-01 -8.81236494e-01 8.06820154e-01 1.01822281e+00 -9.30800438e-02 5.27395487e-01 1.09028324e-01 3.04135203e-01 -3.73708978e-02 -2.09255531e-01 -3.76159191e-01 4.62315887e-01 4.93759453e-01 1.44968942e-01 1.16067603e-01 -3.64448547e-01 8.03428531e-01 2.63369739e-01 -1.53996749e-02 2.18027011e-01 7.88028419e-01 -5.33597410e-01 -1.13782001e+00 -3.15491736e-01 2.99165636e-01 -3.34736526e-01 -8.95826444e-02 -4.52906609e-01 4.97429579e-01 6.30111754e-01 9.57641840e-01 -1.85723215e-01 -2.97089010e-01 2.17172682e-01 2.18357399e-01 1.41002208e-01 -7.62431324e-01 -5.30605674e-01 3.40167910e-01 -1.30918354e-01 -4.69827712e-01 -5.42565644e-01 -4.31483388e-01 -1.54020333e+00 3.91903371e-01 -8.99717391e-01 2.29765043e-01 6.72468960e-01 1.22871935e+00 2.39421293e-01 7.90786505e-01 4.26353633e-01 -7.56103754e-01 -4.33225811e-01 -1.01133251e+00 -4.70458716e-01 6.35154426e-01 1.11206941e-01 -6.36573672e-01 -3.36290300e-01 3.36396322e-02]
[9.576639175415039, 0.6250138878822327]
72965288-aeb4-4677-a0e9-71a318c92a2d
spectral-aware-softmax-for-visible-infrared
2302.01512
null
https://arxiv.org/abs/2302.01512v1
https://arxiv.org/pdf/2302.01512v1.pdf
Spectral Aware Softmax for Visible-Infrared Person Re-Identification
Visible-infrared person re-identification (VI-ReID) aims to match specific pedestrian images from different modalities. Although suffering an extra modality discrepancy, existing methods still follow the softmax loss training paradigm, which is widely used in single-modality classification tasks. The softmax loss lacks an explicit penalty for the apparent modality gap, which adversely limits the performance upper bound of the VI-ReID task. In this paper, we propose the spectral-aware softmax (SA-Softmax) loss, which can fully explore the embedding space with the modality information and has clear interpretability. Specifically, SA-Softmax loss utilizes an asynchronous optimization strategy based on the modality prototype instead of the synchronous optimization based on the identity prototype in the original softmax loss. To encourage a high overlapping between two modalities, SA-Softmax optimizes each sample by the prototype from another spectrum. Based on the observation and analysis of SA-Softmax, we modify the SA-Softmax with the Feature Mask and Absolute-Similarity Term to alleviate the ambiguous optimization during model training. Extensive experimental evaluations conducted on RegDB and SYSU-MM01 demonstrate the superior performance of the SA-Softmax over the state-of-the-art methods in such a cross-modality condition.
['Rongrong Ji', 'Yongjian Wu', 'Mingliang Xu', 'Qixiang Ye', 'Pingyang Dai', 'Lei Tan']
2023-02-03
null
null
null
null
['person-re-identification']
['computer-vision']
[ 2.83987612e-01 -2.85933614e-01 -1.07990876e-01 -5.84186435e-01 -6.50216639e-01 -1.51612163e-01 5.18008351e-01 -3.14133823e-01 -7.03313231e-01 6.50507510e-01 3.39355975e-01 3.34000662e-02 2.76047084e-02 -4.78943676e-01 -6.23967826e-01 -9.60500062e-01 4.60669190e-01 -3.51467043e-01 -2.49057338e-01 -1.62133768e-01 -3.01261485e-01 -7.36664012e-02 -1.50183320e+00 2.81480014e-01 8.46494913e-01 1.27338982e+00 1.31579101e-01 2.67371386e-01 4.75427089e-03 3.38605106e-01 -4.32429284e-01 -7.74410367e-01 5.29427588e-01 -3.72555345e-01 -3.68760437e-01 -1.99757099e-01 7.13033617e-01 -2.07067490e-01 -8.56499732e-01 1.23451889e+00 8.87696862e-01 3.22028697e-01 3.83889139e-01 -1.46911681e+00 -9.51144755e-01 4.19486940e-01 -6.90558672e-01 1.14604846e-01 5.58735311e-01 1.89046040e-01 7.57319033e-01 -9.49462116e-01 2.27221891e-01 1.39615750e+00 9.61159110e-01 6.20935261e-01 -1.13957107e+00 -8.55556190e-01 3.52735072e-01 6.59285843e-01 -1.74120414e+00 -4.70110774e-01 8.75000000e-01 -3.13702881e-01 6.00658476e-01 3.53423238e-01 2.58333027e-01 1.40800476e+00 -4.11693484e-01 8.36577296e-01 1.04426789e+00 -5.08287728e-01 -2.51610518e-01 4.76461560e-01 2.00450420e-01 5.53369164e-01 6.66679535e-03 5.02980351e-01 -6.67842090e-01 2.59010978e-02 4.13445622e-01 -6.60847202e-02 -4.24952149e-01 -9.94071141e-02 -1.26036406e+00 5.10125697e-01 5.50944984e-01 -1.08661696e-01 -1.23732006e-02 -1.40947863e-01 5.55121362e-01 3.55360508e-01 3.85671496e-01 -2.13666543e-01 -2.76126027e-01 1.32382825e-01 -5.59864223e-01 -1.29322454e-01 2.27116928e-01 9.11469221e-01 4.61773366e-01 3.52914371e-02 -4.87885475e-01 1.24967754e+00 5.95485032e-01 6.65962040e-01 3.76718402e-01 -4.55247045e-01 7.61626124e-01 3.65664333e-01 -6.18199110e-02 -5.03508389e-01 -3.89365584e-01 -6.97318137e-01 -1.03432429e+00 3.92743945e-01 5.24281621e-01 -2.56232858e-01 -8.50415051e-01 2.15318036e+00 4.26955879e-01 3.70996147e-01 1.23184413e-01 1.31376827e+00 1.14587665e+00 4.61487949e-01 5.09879529e-01 -9.24074724e-02 1.63062871e+00 -8.77901435e-01 -5.34366071e-01 -2.18267694e-01 2.23198161e-01 -8.05018485e-01 1.00151432e+00 -5.46452701e-02 -5.66756248e-01 -1.11352742e+00 -1.26084912e+00 2.59987218e-03 -4.13507700e-01 4.60372210e-01 4.46940839e-01 9.61026251e-01 -6.68029904e-01 1.54843926e-01 -2.66204506e-01 -4.31487262e-01 1.99013680e-01 1.94345638e-01 -6.05697751e-01 -6.80428669e-02 -1.54687214e+00 8.13831985e-01 6.37231171e-01 4.31079090e-01 -1.13049366e-01 -7.60899544e-01 -1.05005896e+00 -5.61551005e-02 4.11943167e-01 -8.79641771e-01 7.50771105e-01 -1.06976819e+00 -1.68993616e+00 1.07038653e+00 -1.88871041e-01 -1.78702831e-01 8.03539157e-01 -5.65410852e-02 -9.37835276e-01 -1.59284994e-01 -1.52115226e-01 5.49314201e-01 9.40519869e-01 -1.41754270e+00 -7.59792566e-01 -3.25877458e-01 -9.14339721e-03 4.68819141e-01 -5.52006245e-01 6.16410933e-02 -4.98706698e-01 -8.92149329e-01 -1.48913031e-02 -8.40374053e-01 3.72356743e-01 1.62301734e-01 -4.74481165e-01 -2.75497437e-01 5.46190083e-01 -8.56677949e-01 1.02769077e+00 -2.44245172e+00 -1.16396442e-01 2.52957791e-01 -4.60267924e-02 3.33594114e-01 -2.86020190e-01 8.40255991e-02 -3.68371338e-01 -2.43832812e-01 -1.58129930e-01 -6.86349094e-01 3.15981001e-01 -8.06369931e-02 -5.21882735e-02 6.63195610e-01 -1.30304202e-01 6.61557555e-01 -6.98153019e-01 -5.85135758e-01 3.62757057e-01 6.40565991e-01 -1.06544435e-01 1.27889156e-01 3.95112664e-01 3.94351602e-01 -1.79337755e-01 7.58514166e-01 1.01777840e+00 -7.93879107e-02 -1.49947807e-01 -8.30379069e-01 -2.84933560e-02 -2.16862500e-01 -1.35404348e+00 1.78245485e+00 -1.35978729e-01 5.44284225e-01 3.86652611e-02 -1.09430003e+00 8.78128409e-01 2.53172666e-01 5.77886403e-01 -1.00982213e+00 1.19664282e-01 8.02383423e-02 -2.22070768e-01 -5.82505345e-01 4.01199907e-01 -2.04273850e-01 6.62932321e-02 -3.60198319e-02 8.25361907e-02 9.02381539e-01 -9.81310382e-02 -2.31509849e-01 3.11095119e-01 3.07653725e-01 -1.11221559e-01 5.62693067e-02 9.41904962e-01 -6.19193316e-01 7.31205523e-01 8.07976186e-01 -5.56629717e-01 7.02383280e-01 -2.46944308e-01 -1.11157350e-01 -8.31177354e-01 -1.45615935e+00 -5.21168172e-01 1.40846932e+00 7.37970054e-01 -1.51853383e-01 -4.61039335e-01 -6.46338403e-01 1.46735221e-01 4.26667660e-01 -5.09092450e-01 -4.51223344e-01 -4.51656461e-01 -1.09307694e+00 7.67915368e-01 5.85738182e-01 1.03033495e+00 -5.22000134e-01 4.15218733e-02 -1.08797587e-01 -4.91386235e-01 -1.30166519e+00 -8.90765429e-01 -3.79937172e-01 -1.94720492e-01 -8.91543269e-01 -1.06426597e+00 -1.03691852e+00 5.98193347e-01 2.70007938e-01 5.39102256e-01 -1.60642669e-01 -1.11898862e-01 6.24400318e-01 -3.17263752e-01 -2.74126321e-01 -9.55862030e-02 -3.07314005e-02 2.14088023e-01 6.18476212e-01 5.92875063e-01 -2.76791722e-01 -6.39854729e-01 3.76521319e-01 -5.50714195e-01 7.59971216e-02 4.99519110e-01 1.19359493e+00 5.04392922e-01 -1.24776475e-02 7.85915494e-01 -1.44075543e-01 4.85360563e-01 -4.21489924e-01 -3.43801707e-01 7.06381500e-01 -7.76266694e-01 -1.17821582e-02 4.64785874e-01 -7.82712460e-01 -1.52026772e+00 6.65346980e-02 -2.64809549e-01 -3.49572718e-01 -8.59188959e-02 4.92049783e-01 -5.91879308e-01 -2.87470609e-01 4.72225279e-01 4.56906706e-01 5.41050360e-02 -5.41260302e-01 3.09620619e-01 8.27782393e-01 1.03861892e+00 -5.12608528e-01 9.37398255e-01 3.07691634e-01 -6.38751090e-02 -7.61568904e-01 -5.87606192e-01 -5.38905442e-01 -1.44318432e-01 -3.92443210e-01 9.12593782e-01 -1.24812090e+00 -1.17739820e+00 9.49391305e-01 -8.76134634e-01 1.38502821e-01 -2.27596760e-01 9.42998707e-01 -1.29042685e-01 7.10768163e-01 -4.28513706e-01 -8.42353046e-01 -3.50421876e-01 -9.13786888e-01 8.42131019e-01 8.08787644e-01 3.03179353e-01 -8.28870356e-01 -2.14886069e-01 6.33472860e-01 4.24846023e-01 5.17239757e-02 4.91049826e-01 -4.54160988e-01 -2.05892429e-01 -2.56344706e-01 -6.94972515e-01 4.43950087e-01 1.63583890e-01 -4.56705302e-01 -1.41831493e+00 -4.78004605e-01 -3.86770666e-01 -1.57342240e-01 1.06998301e+00 2.47294158e-01 1.16988778e+00 -7.83393160e-02 -2.57028908e-01 9.91131186e-01 1.23121798e+00 1.95768569e-02 5.88586271e-01 5.76360285e-01 9.18517530e-01 4.20993149e-01 3.50762665e-01 3.93615782e-01 6.84073150e-01 9.02210474e-01 3.01931292e-01 -3.51938128e-01 -2.38457561e-01 -3.63737047e-01 3.85213941e-01 2.78596282e-01 -1.97293311e-01 -2.27125183e-01 -3.30083281e-01 1.64143622e-01 -1.90070081e+00 -1.30624485e+00 1.70995340e-01 2.57811069e+00 8.66321087e-01 1.30934520e-02 3.09176952e-01 8.24664608e-02 1.11307228e+00 1.99613675e-01 -6.63368523e-01 4.75317203e-02 -6.53863370e-01 -2.06171229e-01 7.70784855e-01 4.64505315e-01 -1.38961446e+00 4.65340972e-01 5.40446138e+00 1.15010631e+00 -1.17017651e+00 1.87037200e-01 3.27044070e-01 -1.16067119e-01 -4.52214032e-02 -3.16530704e-01 -7.80296028e-01 9.01096523e-01 4.49496984e-01 -1.97564676e-01 6.02811337e-01 5.05458474e-01 1.03156336e-01 5.78060858e-02 -1.05297494e+00 1.61684537e+00 2.10053697e-01 -8.07876945e-01 -2.80942976e-01 -9.09669548e-02 4.34457064e-01 -1.32549554e-01 2.36709774e-01 3.96917462e-01 -3.15524191e-01 -1.01989925e+00 1.02261472e+00 7.30076432e-01 9.44827855e-01 -5.88342667e-01 8.06561530e-01 1.04562476e-01 -1.55676162e+00 -2.67124981e-01 -2.07245678e-01 2.71169811e-01 3.62250924e-01 2.38214731e-01 -2.89689958e-01 9.76896524e-01 7.96675086e-01 6.70034349e-01 -5.82822084e-01 1.27471375e+00 5.71002960e-02 3.86360586e-01 -4.18894738e-01 3.14795673e-01 -2.94323653e-01 -3.27943206e-01 8.17711830e-01 1.41894341e+00 9.96022671e-02 -1.39934108e-01 3.82964671e-01 7.92678535e-01 -1.67321265e-01 -5.11579439e-02 4.23103981e-02 2.25190654e-01 5.63763201e-01 8.66680622e-01 7.02528134e-02 -1.86015680e-01 -6.41949117e-01 1.17041647e+00 -5.32069318e-02 7.26434410e-01 -9.71250057e-01 -5.26519835e-01 9.26832557e-01 -2.41443485e-01 4.54572402e-02 5.79911731e-02 -4.26879197e-01 -1.19138944e+00 3.93683761e-01 -5.28405428e-01 6.75993025e-01 -6.75387621e-01 -1.92031240e+00 4.45642143e-01 1.67630896e-01 -1.51203763e+00 1.57474980e-01 -5.60112000e-01 -4.55626488e-01 1.40030968e+00 -1.75021553e+00 -1.65284050e+00 -5.10037422e-01 9.03863609e-01 2.75704920e-01 -3.62043858e-01 4.65014845e-01 8.41310263e-01 -8.31406593e-01 1.62384093e+00 2.09899649e-01 2.69864947e-01 1.18555689e+00 -9.93410885e-01 -9.98990908e-02 9.57397461e-01 -3.77015531e-01 4.79365498e-01 6.08822107e-01 -3.71072710e-01 -1.09381068e+00 -8.64414632e-01 7.00922966e-01 3.43081392e-02 2.13401973e-01 -9.57184434e-02 -9.55700397e-01 3.47930074e-01 1.27533376e-01 1.07776634e-01 8.26933622e-01 1.53792739e-01 -6.76325381e-01 -5.65923393e-01 -1.23380697e+00 4.28634226e-01 1.32015455e+00 -9.37810838e-01 -3.25659394e-01 1.68923721e-01 6.01639926e-01 -3.98692220e-01 -9.88854825e-01 5.50718546e-01 9.63168740e-01 -6.01394296e-01 1.54344583e+00 -4.56728905e-01 -1.34116918e-01 -5.07246315e-01 -2.20905006e-01 -1.11546636e+00 -4.37087059e-01 -3.31254452e-01 -1.18054815e-01 1.51199317e+00 1.68967649e-01 -1.04356730e+00 4.48164523e-01 6.97852969e-01 -1.09195523e-02 -3.30807179e-01 -1.17445397e+00 -9.84770238e-01 -3.10977012e-01 -2.69050688e-01 6.67118371e-01 9.82031167e-01 -1.45914793e-01 4.18681800e-02 -7.38102257e-01 4.36706036e-01 1.15082967e+00 -2.60001302e-01 5.37155330e-01 -1.06374300e+00 -4.57794905e-01 -6.34340882e-01 -4.06448215e-01 -1.13465273e+00 2.63563722e-01 -8.03072214e-01 -1.06279773e-03 -1.29227543e+00 2.68580884e-01 -3.90324354e-01 -9.05061960e-01 4.25407678e-01 -6.37796223e-01 2.64199853e-01 3.67197126e-01 2.12049827e-01 -4.74561304e-01 8.23370695e-01 1.01157415e+00 -5.45448482e-01 -2.46367171e-01 7.32131451e-02 -8.08103800e-01 4.07915920e-01 5.42622507e-01 -8.94979239e-02 -4.28473622e-01 -3.70623440e-01 -2.90259682e-02 -3.96061152e-01 7.09456623e-01 -8.94162536e-01 3.81904304e-01 -1.79544792e-01 6.67086542e-01 -3.03389013e-01 5.84587872e-01 -7.85038829e-01 2.25199938e-01 1.42842144e-01 -4.39428091e-01 -4.22127068e-01 1.13757730e-01 7.23346591e-01 -1.60193652e-01 3.43318991e-02 7.18326628e-01 3.46704245e-01 -1.01393163e+00 2.94140786e-01 1.30354688e-01 -2.19815686e-01 7.55703390e-01 -5.60496509e-01 -6.29016519e-01 -2.18218952e-01 -6.36516035e-01 4.55961883e-01 3.14573497e-01 7.37748504e-01 5.25717080e-01 -1.81770504e+00 -9.29926217e-01 2.23935008e-01 5.34356534e-01 -5.25512815e-01 7.49153495e-01 8.67541671e-01 1.17476091e-01 -3.05004977e-02 -2.78863996e-01 -5.12135088e-01 -1.43605137e+00 4.87932056e-01 6.68145120e-01 1.04661241e-01 -4.73105192e-01 7.66604245e-01 3.00406665e-01 -3.85366499e-01 4.84477460e-01 3.01736891e-01 -2.84774274e-01 -1.00775205e-01 6.47794545e-01 5.22743225e-01 -1.40146449e-01 -9.77314770e-01 -5.45891881e-01 5.04987121e-01 1.43209204e-01 -6.41181245e-02 8.23763609e-01 -5.45035660e-01 2.85693794e-01 2.10215285e-01 1.37046444e+00 -1.07318841e-01 -1.42289913e+00 -7.71363735e-01 -5.82860112e-01 -5.49251318e-01 3.70638072e-02 -9.91757751e-01 -1.01177526e+00 5.08399069e-01 1.32474327e+00 2.04541758e-02 1.30132174e+00 -2.57244527e-01 9.93868530e-01 1.39105052e-01 7.27685029e-03 -1.33551943e+00 -1.79503039e-01 1.51162878e-01 6.86803997e-01 -1.63381207e+00 -1.75720990e-01 -4.09388304e-01 -5.25181830e-01 1.03306508e+00 7.42634237e-01 5.05861282e-01 6.78246915e-01 -2.18560681e-01 1.48964778e-01 3.76596957e-01 7.25724697e-02 -3.97296637e-01 6.85849011e-01 7.78184295e-01 6.58610389e-02 7.51716271e-02 -2.73125440e-01 8.95269454e-01 1.00107327e-01 -7.23800436e-02 -1.84608027e-01 6.18235469e-01 -1.36173815e-01 -9.24172103e-01 -6.95681095e-01 8.09872970e-02 -1.66479379e-01 4.19376083e-02 -3.46635580e-02 3.80418748e-01 5.61009169e-01 1.21260202e+00 -1.80734783e-01 -7.63220727e-01 4.95989352e-01 9.34000835e-02 5.72211206e-01 1.98541671e-01 -3.77604932e-01 -1.76262558e-01 3.86668369e-02 -1.74033105e-01 -6.11117363e-01 -5.56556761e-01 -9.24765766e-01 -5.17334998e-01 -4.18256640e-01 -3.26332927e-01 7.17456460e-01 1.03080499e+00 3.08404088e-01 4.16895151e-01 7.49337792e-01 -6.73649013e-01 -8.43353629e-01 -9.30345654e-01 -4.49706644e-01 7.79027462e-01 4.60086912e-01 -7.82220542e-01 -2.23820373e-01 1.66154936e-01]
[14.69723129272461, 0.945119321346283]
bc8e47e9-75b8-4a5b-af6c-2c77d6e1ee35
a-real-time-automated-point-process-method
null
null
https://ieeexplore.ieee.org/document/6257443/authors#authors
https://ieeexplore.ieee.org/document/6257443/authors#authors
A Real-Time Automated Point-Process Method for the Detection and Correction of Erroneous and Ectopic Heartbeats
The presence of recurring arrhythmic events (also known as cardiac dysrhythmia or irregular heartbeats), as well as erroneous beat detection due to low signal quality, significantly affects estimation of both time and frequency domain indices of heart rate variability (HRV). A reliable, real-time classification and correction of ECG-derived heartbeats is a necessary prerequisite for an accurate online monitoring of HRV and cardiovascular control. We have developed a novel point-process-based method for real-time R-R interval error detection and correction. Given an R-wave event, we assume that the length of the next R-R interval follows a physiologically motivated, time-varying inverse Gaussian probability distribution. We then devise an instantaneous automated detection and correction procedure for erroneous and arrhythmic beats by using the information on the probability of occurrence of the observed beat provided by the model. We test our algorithm over two datasets from the PhysioNet archive. The Fantasia normal rhythm database is artificially corrupted with known erroneous beats to test both the detection procedure and correction procedure. The benchmark MIT-BIH Arrhythmia database is further considered to test the detection procedure of real arrhythmic events and compare it with results from previously published algorithms. Our automated algorithm represents an improvement over previous procedures, with best specificity for the detection of correct beats, as well as highest sensitivity to missed and extra beats, artificially misplaced beats, and for real arrhythmic events. A near-optimal heartbeat classification and correction, together with the ability to adapt to time-varying changes of heartbeat dynamics in an online fashion, may provide a solid base for building a more reliable real-time HRV monitoring device.
['Riccardo Barbieri', 'Emery N. Brown', 'Luca Citi']
2012-08-02
null
null
null
null
['heart-rate-variability', 'heartbeat-classification']
['medical', 'medical']
[ 4.88780260e-01 -2.64330208e-01 2.84845531e-01 -2.04200000e-01 -4.14584249e-01 -7.19700575e-01 7.24020973e-02 4.79783595e-01 -2.00467706e-01 9.02554095e-01 -8.98546502e-02 -2.50413477e-01 -2.83435792e-01 -6.02254152e-01 -2.22929910e-01 -5.39145291e-01 -5.99484921e-01 5.14638364e-01 -1.71978865e-02 2.12359235e-01 8.44743252e-02 5.96191704e-01 -1.26691282e+00 -1.36582568e-01 7.88286626e-01 7.38071084e-01 -7.65182748e-02 1.07435775e+00 6.05430961e-01 1.94541678e-01 -9.68785524e-01 6.56144619e-02 1.33641317e-01 -9.88448083e-01 -3.01218063e-01 -1.28463119e-01 -1.25154689e-01 -9.37040225e-02 1.99120089e-01 4.53397065e-01 9.73836541e-01 -1.29773214e-01 5.89320123e-01 -7.27285266e-01 2.43466929e-01 3.11993867e-01 -1.66536197e-01 6.00523651e-01 2.81129062e-01 2.64600098e-01 5.02614439e-01 -4.82633561e-01 5.38647056e-01 3.96859020e-01 1.26563728e+00 1.96328819e-01 -1.60805345e+00 -2.34871745e-01 -7.39995480e-01 -6.88165287e-03 -1.61486506e+00 -2.86408454e-01 7.53541291e-01 -5.22437274e-01 5.26003778e-01 6.49182677e-01 1.04685843e+00 7.99517453e-01 5.31933248e-01 -1.80446759e-01 8.90473783e-01 -4.54185009e-01 2.59475291e-01 -1.34169281e-01 1.23227648e-01 2.35199973e-01 5.30571103e-01 4.77663934e-01 -1.58918142e-01 -5.86549163e-01 8.80071700e-01 -2.02327278e-02 -5.67610085e-01 5.97066917e-02 -1.54583347e+00 3.21159869e-01 -9.47736204e-03 6.14965558e-01 -6.57468438e-01 9.33691561e-02 6.11776650e-01 2.91099429e-01 3.13076496e-01 6.22755945e-01 -5.80473423e-01 -4.33435917e-01 -1.17200470e+00 2.49024302e-01 8.62909138e-01 2.00224862e-01 4.50587690e-01 2.48183712e-01 -5.50453365e-01 7.12888956e-01 5.15473709e-02 5.14467597e-01 6.76265121e-01 -1.10450637e+00 -1.81100845e-01 2.82192886e-01 3.26393455e-01 -1.12036979e+00 -8.25414479e-01 -9.13417757e-01 -1.18660152e+00 -1.25502571e-01 6.96250737e-01 -7.04428777e-02 -3.67937773e-01 1.59825575e+00 2.79498398e-01 3.64479572e-01 5.91162825e-03 7.71097243e-01 2.57293642e-01 3.08407307e-01 -1.48033664e-01 -1.18533337e+00 1.37436008e+00 2.68845201e-01 -9.06569302e-01 3.82052958e-01 5.81562817e-01 -6.51995361e-01 6.72766030e-01 4.94816989e-01 -8.26192319e-01 -7.60562360e-01 -8.35960388e-01 4.96493816e-01 1.80616856e-01 1.33518994e-01 1.62714481e-01 7.82887638e-01 -4.73624021e-01 1.11352825e+00 -8.20417643e-01 -2.62217522e-01 1.18323423e-01 5.61211705e-02 6.41792733e-03 5.05365849e-01 -1.27498949e+00 7.83806682e-01 2.43960187e-01 3.13035995e-01 -5.24407566e-01 -9.86361563e-01 -5.49428225e-01 -1.08271196e-01 -6.82389215e-02 -7.94777870e-01 8.32004428e-01 -6.65511429e-01 -1.15874016e+00 9.02461112e-01 -1.85593247e-01 -8.16823900e-01 8.52841675e-01 9.51499939e-02 -4.95296389e-01 9.43241492e-02 -1.34179309e-01 -3.39338720e-01 8.57854366e-01 -8.49746644e-01 2.80790478e-02 -4.11752135e-01 -8.87543380e-01 -7.49785528e-02 4.21979278e-01 -2.88646758e-01 1.80829450e-01 -8.15277815e-01 3.10285538e-01 -8.84001970e-01 -4.40773778e-02 -3.90170097e-01 -2.61198878e-01 5.00161290e-01 2.32239410e-01 -8.40583801e-01 1.58876252e+00 -2.28511095e+00 -6.49232045e-02 5.29137433e-01 2.22798064e-01 3.78982663e-01 4.53458935e-01 4.04338926e-01 -1.55650347e-01 1.71973929e-02 -4.72317010e-01 1.00623019e-01 -3.56695503e-01 2.90049821e-01 -3.87956798e-01 7.55124629e-01 -1.36551648e-01 6.93355441e-01 -8.11667979e-01 -2.21172825e-01 4.21587855e-01 6.81666553e-01 -1.43580571e-01 2.32189685e-01 1.57990903e-01 1.02617753e+00 2.55706280e-01 2.34541699e-01 2.04625249e-01 9.90651175e-02 2.80039817e-01 -3.42274964e-01 -1.67975366e-01 7.87069276e-02 -1.19365180e+00 1.22105026e+00 -1.43207684e-01 6.48243368e-01 -4.41575289e-01 -7.88081527e-01 1.33203256e+00 8.46334040e-01 7.54409194e-01 -4.27103013e-01 2.12440461e-01 4.64290172e-01 3.72951508e-01 -3.94350737e-01 -1.64861888e-01 -3.29941690e-01 2.70385057e-01 3.64986807e-01 -1.36278346e-01 -1.92018732e-01 1.47631794e-01 -3.26759756e-01 1.17627478e+00 8.32094252e-02 9.17534411e-01 -4.02278781e-01 6.52903855e-01 -3.75657946e-01 7.99809933e-01 8.16817403e-01 -2.80340850e-01 1.05451214e+00 5.46510339e-01 -9.39541399e-01 -8.62791359e-01 -1.07710660e+00 -4.70088154e-01 1.53718829e-01 -2.44452909e-01 -4.28528190e-01 -4.43412393e-01 -6.40512258e-02 9.42037534e-03 5.96514583e-01 -4.56914246e-01 -3.91608000e-01 -6.67512178e-01 -1.25437844e+00 6.67055786e-01 1.33255169e-01 3.67809553e-03 -1.17522120e+00 -1.15708804e+00 8.18875730e-01 -4.74572122e-01 -8.40634227e-01 -2.03168422e-01 1.76917195e-01 -1.21312714e+00 -1.31395543e+00 -6.23047173e-01 -7.18264747e-03 2.97924221e-01 -5.79426467e-01 1.35572648e+00 2.76215047e-01 -1.13280249e+00 2.37258375e-01 -1.50030702e-01 -6.79113507e-01 -8.44530046e-01 -3.49606335e-01 2.08304837e-01 2.11581379e-01 -8.74034762e-02 -8.31903458e-01 -8.50492477e-01 4.19049829e-01 -4.18116271e-01 -3.28871042e-01 1.26736760e-01 7.24757910e-01 9.42659378e-01 -7.75423944e-02 9.76488948e-01 -8.00460339e-01 6.13741994e-01 -1.95794031e-01 -6.31106257e-01 -8.37365240e-02 -9.75002408e-01 -2.95806378e-01 5.53610146e-01 -3.53598624e-01 -3.76058787e-01 4.35812445e-03 -1.10530570e-01 -2.36864448e-01 -3.01676393e-01 3.02401215e-01 2.92909443e-01 2.75039822e-01 1.12017143e+00 4.61068273e-01 2.51130700e-01 -1.77088335e-01 -1.87772691e-01 3.21026415e-01 9.05353487e-01 -4.15823102e-01 4.98465776e-01 3.23744148e-01 3.48431140e-01 -9.90926206e-01 -3.67838234e-01 -6.86032176e-01 -7.52919436e-01 -2.44876325e-01 7.35785484e-01 -6.51936531e-01 -8.04915369e-01 4.91551161e-01 -1.04762590e+00 -1.95106581e-01 -7.78779089e-01 6.13778830e-01 -9.40482736e-01 5.55041969e-01 -4.20946240e-01 -1.17506754e+00 -7.97844470e-01 -5.25744140e-01 5.86085498e-01 -6.09138273e-02 -8.07993174e-01 -9.04019594e-01 6.18508518e-01 -1.21707290e-01 3.99041325e-01 1.01145935e+00 6.38643265e-01 -5.93172789e-01 -6.24053106e-02 -5.25656164e-01 3.86199683e-01 4.56399262e-01 4.68172342e-01 2.46300280e-01 -8.79356503e-01 -9.77802351e-02 3.13439637e-01 5.18208206e-01 4.23411280e-01 6.99628234e-01 7.37563610e-01 -1.81975305e-01 -1.95444152e-01 5.03194273e-01 1.37001038e+00 2.93836147e-01 8.33518267e-01 -1.35617226e-01 2.57505178e-01 2.91579008e-01 5.28548241e-01 7.08730400e-01 -8.57463181e-02 6.95508003e-01 6.15860224e-02 -1.83343753e-01 6.26488850e-02 1.90976262e-01 5.28632589e-02 7.38987565e-01 -7.01854050e-01 1.10905871e-01 -1.01429927e+00 5.50880849e-01 -1.66342783e+00 -1.35643375e+00 -7.10717082e-01 3.08010864e+00 9.83891308e-01 1.78356498e-01 3.88296694e-01 8.34573150e-01 7.59298384e-01 -2.28288785e-01 -3.94209236e-01 -4.21661556e-01 1.95380133e-02 2.77484536e-01 2.18315050e-01 3.61036837e-01 -7.88401604e-01 -9.39692631e-02 6.69852924e+00 -8.02350491e-02 -1.22936773e+00 -5.44954874e-02 5.92028916e-01 5.40467538e-02 2.24015951e-01 -1.65943205e-01 -4.06746894e-01 5.88120520e-01 1.41236556e+00 -3.17703247e-01 3.80956441e-01 3.94045770e-01 8.17163944e-01 -2.38203816e-02 -1.02923274e+00 1.21570218e+00 -1.91084892e-01 -1.25481510e+00 -5.17585278e-01 -6.28777146e-02 2.68736035e-01 -1.05119258e-01 -5.78995228e-01 -3.08180284e-02 -8.04360211e-01 -7.38992453e-01 4.26277488e-01 1.05340934e+00 9.46075737e-01 -6.92889273e-01 9.37438786e-01 3.26876730e-01 -1.13797081e+00 8.49598795e-02 -1.24907661e-02 -1.44690216e-01 1.32085323e-01 1.35565853e+00 -8.26968968e-01 6.40693665e-01 1.66151986e-01 7.37729549e-01 -3.69769663e-01 1.25529850e+00 5.14242891e-03 9.80716228e-01 -4.87790108e-01 4.54682022e-01 -8.88851464e-01 -1.36533096e-01 1.02638733e+00 9.68291223e-01 3.89044702e-01 1.74038246e-01 -8.60877931e-02 9.84440267e-01 5.04933894e-01 1.11372620e-01 -6.01012111e-01 1.38989896e-01 4.96075511e-01 1.12478232e+00 -8.13554287e-01 -3.29551488e-01 -5.68535477e-02 6.14142537e-01 -3.22682083e-01 2.35868216e-01 -7.30391860e-01 -3.49971652e-01 3.07827562e-01 5.43781638e-01 -2.51747310e-01 -3.25212181e-02 -5.60976207e-01 -9.16116595e-01 1.21676616e-01 -8.71919096e-01 4.96858627e-01 -3.81781876e-01 -8.80498886e-01 6.62957370e-01 -1.96239263e-01 -1.42208648e+00 -6.54190481e-01 1.73979849e-02 -9.44491506e-01 1.10437512e+00 -9.98025775e-01 -3.19018215e-01 -2.88094521e-01 2.77313828e-01 -7.17509585e-03 1.87788486e-01 1.32299650e+00 2.97041774e-01 -1.66470677e-01 2.13321850e-01 -3.03763419e-01 -8.70725587e-02 6.82873368e-01 -1.40137494e+00 3.08040250e-02 6.83745921e-01 1.93443209e-01 3.67159486e-01 1.11164343e+00 -6.21020615e-01 -7.53275871e-01 -1.09917629e+00 1.01951408e+00 -6.73056781e-01 2.40649983e-01 -1.94383599e-02 -1.06407583e+00 3.15001637e-01 -5.42110384e-01 2.53544539e-01 7.62050390e-01 -1.20238490e-01 2.82787800e-01 -4.93316561e-01 -1.00315201e+00 7.33536482e-02 5.51325977e-01 -2.62892038e-01 -8.70038927e-01 1.67797998e-01 -4.17021709e-03 -3.65141183e-01 -1.29649770e+00 7.68754482e-01 9.20532227e-01 -1.23625350e+00 9.45736945e-01 2.62570139e-02 -2.47640774e-01 -2.95185208e-01 3.50542754e-01 -1.06951022e+00 -1.77653342e-01 -1.36601698e+00 -1.93122983e-01 1.05307996e+00 3.66462730e-02 -7.76510537e-01 2.53133446e-01 2.21845627e-01 1.09459683e-01 -4.97446567e-01 -1.01993132e+00 -7.87636876e-01 -5.32960415e-01 -4.66234982e-01 1.05560690e-01 7.51499772e-01 -4.01772223e-02 8.95755216e-02 -3.50899637e-01 9.95132700e-02 8.55082095e-01 1.85695529e-01 5.30142128e-01 -1.71735108e+00 -4.60656911e-01 -2.61833310e-01 -6.37661636e-01 2.39087924e-01 -5.78127384e-01 -5.89830160e-01 1.04796864e-01 -1.03828645e+00 -1.76956177e-01 -3.82825583e-01 -5.71080506e-01 -1.43010111e-04 -4.04365033e-01 4.99740273e-01 -9.99757126e-02 2.83903509e-01 1.70504615e-01 6.34344742e-02 8.16269755e-01 5.44209838e-01 -8.84946585e-01 6.12690866e-01 2.61409152e-02 6.43967450e-01 6.73126757e-01 -7.80477762e-01 -3.19315910e-01 6.76274180e-01 3.17002416e-01 6.22385085e-01 6.31317437e-01 -1.32400119e+00 -4.34566677e-01 4.00750041e-01 5.09549975e-01 -5.23914397e-01 -4.47836928e-02 -6.20174885e-01 8.60925853e-01 9.50737357e-01 -1.78067550e-01 9.79522318e-02 1.23421416e-01 4.95064974e-01 -1.20252267e-01 -2.60129496e-02 9.54551637e-01 -2.17219979e-01 2.75636017e-01 1.78701326e-01 -5.72840750e-01 2.64683902e-01 9.55511510e-01 -3.60102952e-01 4.10481766e-02 -2.03725815e-01 -1.27083635e+00 -3.94311368e-01 2.14091048e-01 1.41418830e-01 3.85837078e-01 -1.04123056e+00 -9.98213470e-01 4.38645899e-01 6.54852539e-02 -3.86013210e-01 2.98966914e-01 1.37089360e+00 -6.90169513e-01 1.83590770e-01 -2.50092059e-01 -8.96113813e-01 -1.33625829e+00 2.34326422e-01 7.50563800e-01 -2.03710794e-01 -8.84531796e-01 2.03541249e-01 -4.11146581e-01 2.17680007e-01 5.93221076e-02 -4.84642595e-01 -2.39576638e-01 1.59256354e-01 5.69797039e-01 5.38901806e-01 2.40595803e-01 -3.67248535e-01 -3.96145254e-01 5.58532834e-01 6.16742074e-01 9.69602987e-02 8.24902833e-01 -2.98473835e-01 -2.13372841e-01 1.17607045e+00 5.77397168e-01 -3.33479047e-02 -7.73093402e-01 2.43330464e-01 -8.78376290e-02 -1.55034870e-01 -1.87179312e-01 -8.82976770e-01 -6.81237042e-01 6.57621861e-01 9.84970331e-01 5.51820099e-01 1.19863451e+00 -6.49000406e-01 4.75132704e-01 -1.58835977e-01 3.87613863e-01 -7.46891975e-01 -3.03795844e-01 -1.03116155e-01 9.03621256e-01 -7.31837153e-01 2.19756216e-01 -2.22859234e-01 -3.75651568e-01 1.27143073e+00 -1.11477382e-01 -1.76491827e-01 7.66366303e-01 1.05707593e-01 3.49602640e-01 1.45507921e-02 -7.04507053e-01 7.56604373e-02 2.03678578e-01 7.52873063e-01 6.20555341e-01 2.09460929e-01 -8.90048146e-01 4.47506219e-01 -8.86757150e-02 5.07901609e-01 6.20323122e-01 6.06626391e-01 -1.63879231e-01 -8.46391916e-01 -5.05082190e-01 5.26351154e-01 -6.98524117e-01 1.15136519e-01 4.55589816e-02 5.00332534e-01 1.42661139e-01 6.75863743e-01 7.51272589e-02 1.76390201e-01 3.63707602e-01 5.57202399e-01 4.04418230e-01 -4.93177831e-01 -9.06855106e-01 2.36027017e-01 1.34072751e-01 -5.40076733e-01 -2.91751206e-01 -7.36207783e-01 -1.32533026e+00 9.34780166e-02 -2.98771471e-01 6.38215244e-02 4.89078522e-01 8.77890944e-01 4.70866859e-01 7.06475616e-01 5.78810632e-01 -6.51742280e-01 -6.38967514e-01 -1.03321445e+00 -8.61095011e-01 7.48689532e-01 5.59707046e-01 -2.15033382e-01 -6.42729700e-01 4.80773062e-01]
[14.18603515625, 3.175640106201172]
a4920eba-cf8a-4186-b9a9-489db5e6f3c9
divergence-regulated-encoder-network-for
2012.15764
null
https://arxiv.org/abs/2012.15764v6
https://arxiv.org/pdf/2012.15764v6.pdf
Divergence Regulated Encoder Network for Joint Dimensionality Reduction and Classification
Feature representation is an important aspect of remote-sensing based image classification. While deep convolutional neural networks are able to effectively amalgamate information, large numbers of parameters often make learned features inscrutable and difficult to transfer to alternative models. In order to better represent statistical texture information for remote-sensing image classification, in this paper, we investigate performing joint dimensionality reduction and classification using a novel histogram neural network. Motivated by a popular dimensionality reduction approach, t-Distributed Stochastic Neighbor Embedding (t-SNE), our proposed method incorporates a classification loss computed on samples in a low-dimensional embedding space. We compare the learned sample embeddings against coordinates found by t-SNE in terms of classification accuracy and qualitative assessment. We also explore use of various divergence measures in the t-SNE objective. The proposed method has several advantages such as readily embedding out-of-sample points and reducing feature dimensionality while retaining class discriminability. Our results show that the proposed approach maintains and/or improves classification performance and reveals characteristics of features produced by neural networks that may be helpful for other applications.
['Weihuang Xu', 'James Keller', 'Alina Zare', 'Connor McCurley', 'Sarah Walker', 'Joshua Peeples']
2020-12-31
null
null
null
null
['remote-sensing-image-classification']
['miscellaneous']
[ 3.22287977e-01 -2.86514848e-01 2.04076841e-02 -6.38038993e-01 -8.12944353e-01 -3.29616517e-01 7.59403765e-01 3.41630071e-01 -5.13240457e-01 6.60668373e-01 1.83130741e-01 -5.11236042e-02 -7.72224844e-01 -1.22305059e+00 -3.88820767e-01 -1.19610035e+00 -4.01453465e-01 9.10235047e-02 -1.67092100e-01 2.33156681e-02 2.59944111e-01 1.10158277e+00 -1.89161921e+00 4.17641439e-02 7.69830704e-01 1.31757462e+00 3.09378225e-02 3.90797794e-01 6.19546548e-02 2.94242442e-01 -2.95478970e-01 6.29225746e-02 5.07447183e-01 -2.47835174e-01 -4.57724541e-01 -1.75695330e-01 5.91123223e-01 -1.92940146e-01 -1.17724724e-01 1.09928799e+00 6.18381858e-01 4.69587028e-01 1.19950914e+00 -1.10925865e+00 -1.03006661e+00 -5.28229065e-02 -3.11801463e-01 1.50783360e-01 -3.26461755e-02 -1.88623890e-01 9.83991206e-01 -9.67588246e-01 5.38429976e-01 1.06816316e+00 9.55127954e-01 5.54987788e-02 -1.49573147e+00 -4.50791389e-01 -2.99943179e-01 1.60549730e-01 -1.62846756e+00 -1.56611070e-01 1.02019799e+00 -6.23741806e-01 5.99127829e-01 5.11756003e-01 5.98457694e-01 6.23268008e-01 5.38472235e-01 5.03757119e-01 1.43421507e+00 -3.93109411e-01 4.40418065e-01 3.32635224e-01 1.88566837e-02 5.02012491e-01 3.08543980e-01 5.49920127e-02 -1.58280745e-01 -2.85459965e-01 6.83419883e-01 4.90376890e-01 -1.82572484e-01 -6.17965579e-01 -1.00643992e+00 1.32908928e+00 7.84592211e-01 3.15788299e-01 -5.28049171e-01 3.50621566e-02 2.56681442e-01 2.84741729e-01 9.86791611e-01 4.34042513e-01 -2.32349873e-01 2.01397523e-01 -8.96445453e-01 3.46465290e-01 3.86408895e-01 4.82675582e-02 1.21925282e+00 -3.96323167e-02 -1.22361049e-01 1.01452243e+00 1.04189470e-01 7.27283597e-01 5.08760095e-01 -7.08151698e-01 -6.90304339e-02 5.52573919e-01 -1.72147483e-01 -1.60112381e+00 -5.12668073e-01 -3.16261500e-01 -1.21016645e+00 4.51522589e-01 1.14766583e-01 2.02820957e-01 -5.29114246e-01 1.41050124e+00 4.06480849e-01 -3.01508427e-01 -2.13547833e-02 8.15143645e-01 4.59735900e-01 7.11366117e-01 -1.06072187e-01 7.84789026e-02 1.00274396e+00 -4.93835419e-01 -4.20335829e-01 3.18181217e-01 5.84336281e-01 -4.43606138e-01 1.04193163e+00 1.93385094e-01 -4.88768786e-01 -4.50791210e-01 -8.89462948e-01 1.57882855e-03 -7.80199468e-01 2.27130160e-01 5.11390746e-01 5.73557913e-01 -1.00189579e+00 8.82751048e-01 -9.26933229e-01 -4.98325795e-01 7.44064748e-01 1.67547494e-01 -6.27224028e-01 1.30160581e-02 -8.45416725e-01 6.83700264e-01 2.49548763e-01 2.07151622e-01 -3.97983313e-01 -6.61764562e-01 -1.00834167e+00 2.01629907e-01 -2.50304401e-01 -1.69929922e-01 4.56690311e-01 -5.86024642e-01 -1.27767313e+00 4.75930870e-01 3.40040959e-02 -2.08267316e-01 2.42600977e-01 -9.61272717e-02 -3.37595731e-01 3.38787109e-01 1.80584148e-01 6.46072447e-01 8.57191741e-01 -1.00463450e+00 -4.09686804e-01 -7.11391628e-01 -2.59365588e-01 -4.13449667e-02 -7.94138908e-01 -3.89274657e-01 6.71569586e-01 -8.34609747e-01 4.27701980e-01 -8.86056304e-01 -8.70799944e-02 7.18349755e-01 -2.03898266e-01 -1.80361882e-01 1.19860268e+00 -5.55065393e-01 9.25048709e-01 -1.99190474e+00 -1.02668509e-01 4.48634923e-01 2.76285410e-01 1.80459470e-01 -2.89264321e-01 6.21202946e-01 -8.00999478e-02 2.23299801e-01 -5.37605584e-01 1.08499005e-01 7.91338384e-02 1.47120118e-01 -1.12740979e-01 8.66228759e-01 5.75711012e-01 8.13746572e-01 -6.88163519e-01 -2.21376404e-01 4.86979663e-01 1.06207573e+00 -5.63942313e-01 -5.64040914e-02 1.66312054e-01 2.59666413e-01 -4.72319901e-01 6.19345367e-01 8.38205993e-01 -3.51819545e-02 -2.72239327e-01 -2.94540763e-01 -5.01485839e-02 -4.98985760e-02 -9.89663005e-01 1.28570807e+00 -5.42905509e-01 9.41180825e-01 -3.35064858e-01 -1.42260122e+00 1.32885647e+00 -2.40807146e-01 5.14924467e-01 -8.09864879e-01 -1.76126063e-01 5.74843548e-02 -2.53705084e-01 -4.99211311e-01 5.48684299e-01 -2.39530042e-01 1.03214756e-01 4.13561553e-01 -2.03191832e-01 6.48197532e-02 -1.80009648e-01 -4.01202828e-01 7.42293417e-01 -1.39014572e-02 4.80105191e-01 -5.39724290e-01 3.66727710e-01 -1.40448719e-01 2.28642136e-01 6.48805976e-01 -2.67334044e-01 5.51929653e-01 2.82487839e-01 -8.26716125e-01 -1.13459587e+00 -1.04952264e+00 -7.46119499e-01 9.19216633e-01 -9.00463313e-02 1.02859505e-01 -3.58670920e-01 -5.37651420e-01 3.55322450e-01 4.87385660e-01 -9.69119430e-01 -2.70894945e-01 -1.13019988e-01 -9.55885231e-01 4.87845033e-01 4.39435840e-01 5.26366115e-01 -7.41697490e-01 -7.90276527e-01 1.33648857e-01 9.79176909e-02 -4.49356258e-01 2.48571746e-02 2.83038855e-01 -1.04433954e+00 -7.87655056e-01 -7.74719238e-01 -4.93474036e-01 6.41261399e-01 4.69722539e-01 6.41907454e-01 -2.86279023e-01 -6.22769117e-01 2.03865349e-01 -5.86785734e-01 -2.50624478e-01 -1.17022924e-01 1.80429965e-01 -6.66562468e-02 2.33221114e-01 7.42497563e-01 -6.79506958e-01 -8.14343333e-01 1.72726303e-01 -1.21588051e+00 -5.24244905e-01 6.81619823e-01 1.10515523e+00 5.99244833e-01 3.99616420e-01 4.37462538e-01 -4.84701633e-01 6.11406147e-01 -3.62864703e-01 -3.49014699e-01 2.49697566e-02 -7.93059766e-01 2.91019768e-01 5.69799066e-01 -2.58497536e-01 -6.91066921e-01 -1.37313426e-01 -7.21373186e-02 -3.21938545e-01 -3.11123222e-01 6.60506725e-01 1.74131125e-01 -3.90637279e-01 8.33885670e-01 4.52408612e-01 5.11508882e-01 -6.52276814e-01 2.98710525e-01 8.83195460e-01 -1.15389369e-01 -2.82055676e-01 7.93524444e-01 7.25277841e-01 2.64389038e-01 -1.41337276e+00 -5.78683615e-01 -4.26110804e-01 -7.50778079e-01 -1.88053455e-02 9.27223146e-01 -6.47316039e-01 -6.38215363e-01 3.70127976e-01 -6.77080929e-01 8.59001204e-02 -4.56990033e-01 7.33265579e-01 -3.43228608e-01 2.66408563e-01 -3.04522395e-01 -1.02389276e+00 -2.91381806e-01 -1.09636045e+00 1.33109820e+00 1.74410474e-02 7.75291175e-02 -1.27731550e+00 3.52650166e-01 -6.64202198e-02 8.46128285e-01 6.51404262e-01 1.05562329e+00 -6.76605225e-01 -1.31023020e-01 -3.60043496e-01 -4.28074628e-01 6.23412430e-01 2.83050209e-01 -7.16267005e-02 -1.15605772e+00 -4.63598073e-01 -9.60100517e-02 -4.45625514e-01 1.12480474e+00 6.12665296e-01 1.39951181e+00 -3.98928791e-01 -8.52830410e-02 6.89111710e-01 1.82251608e+00 1.67647246e-02 6.91682279e-01 6.23386621e-01 5.48579931e-01 8.37627888e-01 2.65883714e-01 5.99786997e-01 2.37309873e-01 4.76235420e-01 3.74668509e-01 -1.74456641e-01 2.09278345e-01 -3.11571155e-02 1.50648244e-02 6.72004640e-01 -1.39922559e-01 1.28041893e-01 -8.33610773e-01 4.77567047e-01 -1.46486032e+00 -1.04197645e+00 9.07365605e-02 2.05750608e+00 4.26411748e-01 -3.77964139e-01 -4.76588160e-02 4.05407369e-01 6.31094992e-01 3.59753460e-01 -4.39604282e-01 -3.24976414e-01 -2.04763949e-01 3.02018136e-01 4.02663201e-01 3.86868477e-01 -1.33244336e+00 4.35604483e-01 5.74149323e+00 8.95698428e-01 -1.45124245e+00 -7.35026821e-02 6.95559561e-01 1.30138442e-01 -2.96326309e-01 -3.28042537e-01 -4.33275491e-01 1.86873987e-01 6.71975791e-01 1.15845844e-01 1.80726171e-01 8.90825629e-01 2.69387126e-01 -1.58725768e-01 -8.94427121e-01 9.46021974e-01 1.15088679e-01 -1.33755517e+00 4.77714151e-01 3.64921361e-01 6.56341136e-01 -2.99962685e-02 4.32885468e-01 8.99375454e-02 -3.27900082e-01 -1.18404734e+00 4.68727857e-01 6.80848539e-01 8.40398967e-01 -8.43187153e-01 8.19371283e-01 6.42604679e-02 -1.23609233e+00 -1.07911311e-01 -7.78852463e-01 -2.35934719e-01 -4.11550581e-01 8.37998390e-01 -7.52678692e-01 4.93747741e-01 7.11356759e-01 9.11085010e-01 -7.26694167e-01 8.95440042e-01 4.07051504e-01 2.92528868e-01 -3.35805774e-01 -3.32203776e-01 3.47873032e-01 -6.07218683e-01 3.99963647e-01 1.03573883e+00 5.89110613e-01 -2.21607909e-01 1.26620606e-01 1.00010860e+00 2.40027949e-01 1.79501489e-01 -9.82464910e-01 -2.85744786e-01 5.23759425e-01 1.28349710e+00 -8.49598110e-01 2.23025884e-02 -9.35704634e-02 7.48348832e-01 1.84586853e-01 2.92315841e-01 -3.21450740e-01 -7.83850551e-01 8.52954566e-01 6.83824196e-02 3.92300487e-01 -2.41873622e-01 -1.90716788e-01 -1.04126072e+00 1.94747657e-01 -4.51970100e-01 3.43335718e-02 -5.11634111e-01 -1.42439425e+00 6.74790740e-01 6.89125434e-02 -1.55842292e+00 -1.14839599e-01 -8.13923895e-01 -3.87311399e-01 7.61666656e-01 -1.75760841e+00 -1.21166992e+00 -3.62444431e-01 2.91464984e-01 2.18276262e-01 -2.11172581e-01 1.10357940e+00 1.90207630e-01 -2.82029986e-01 4.99458343e-01 8.72076452e-01 1.22059725e-01 3.38540107e-01 -1.16885912e+00 -8.92242789e-02 4.51537460e-01 4.63410988e-02 4.93333310e-01 5.05350411e-01 -2.89889276e-01 -1.21661973e+00 -1.31961679e+00 6.11130476e-01 -6.36335537e-02 5.36957026e-01 -3.28902215e-01 -1.01555276e+00 2.52979338e-01 -3.83276641e-01 2.50646293e-01 9.13205862e-01 -1.17072411e-01 -4.98793304e-01 -4.67900753e-01 -1.41049945e+00 2.76189059e-01 5.20655990e-01 -8.54772329e-01 -4.24957007e-01 2.55796313e-01 2.35845938e-01 4.16628778e-01 -1.16508734e+00 3.32627565e-01 8.66284728e-01 -1.10040724e+00 8.77555192e-01 -4.18480575e-01 6.92472219e-01 -1.34472987e-02 -6.25329018e-01 -1.46665776e+00 -4.79049534e-01 7.74180666e-02 4.61483568e-01 8.78077805e-01 2.44517997e-01 -9.05710578e-01 8.96385312e-01 1.50727242e-01 2.29912966e-01 -9.10200059e-01 -1.11824906e+00 -9.43985522e-01 4.77118522e-01 -2.97698230e-01 5.36378622e-01 1.16028142e+00 -3.31225663e-01 -2.10659832e-01 -1.66078165e-01 8.50324631e-02 8.59073162e-01 6.91320300e-02 5.72595835e-01 -1.60021758e+00 1.76429689e-01 -4.95540231e-01 -1.00952554e+00 -4.38527793e-01 1.57530770e-01 -1.16659427e+00 -1.29850924e-01 -1.43134403e+00 1.75385550e-01 -5.48224926e-01 -3.91421199e-01 4.41969395e-01 2.03990445e-01 5.07048011e-01 -1.65057946e-02 3.31063986e-01 -1.18440911e-01 1.04438126e+00 1.06816053e+00 -2.68036306e-01 -1.89001009e-01 -2.98148632e-01 -4.65687662e-01 5.86587310e-01 7.83011079e-01 -5.53021610e-01 -2.04374418e-01 -2.49134466e-01 6.23840503e-02 -2.71091402e-01 6.49809897e-01 -1.06357288e+00 -2.51263659e-02 -3.64086106e-02 7.96707273e-01 -4.28518534e-01 3.44554871e-01 -8.01155031e-01 1.46724522e-01 5.87148666e-01 -4.33746159e-01 -1.05438717e-01 -2.71168607e-03 7.17967808e-01 -4.26058799e-01 -8.81462991e-02 9.08413649e-01 1.38653308e-01 -5.54574013e-01 1.63913816e-01 -3.11733425e-01 -4.54693288e-01 9.47259486e-01 -6.47035658e-01 -3.09068382e-01 -2.22363889e-01 -6.32327437e-01 -3.50523025e-01 3.63761902e-01 2.19775051e-01 8.34795594e-01 -1.65699983e+00 -7.51462996e-01 4.96334672e-01 4.82306987e-01 -3.53477895e-01 2.59840757e-01 7.41913378e-01 -8.82796407e-01 3.27927172e-01 -5.71424901e-01 -9.14402187e-01 -9.40162122e-01 1.52074635e-01 2.45473668e-01 4.15926725e-02 -7.68043876e-01 6.60946429e-01 9.06417742e-02 -7.83768356e-01 -1.13873146e-01 -3.48535240e-01 -1.48332790e-01 4.45195735e-01 5.24320066e-01 7.12851286e-01 1.69296339e-01 -6.98848844e-01 -3.56035292e-01 7.33950555e-01 5.27260900e-02 2.96339532e-03 1.86369324e+00 -6.04457445e-02 -3.31407696e-01 5.46943724e-01 1.90414178e+00 -3.94344091e-01 -1.14645982e+00 -2.64593184e-01 -7.22226277e-02 -7.28726804e-01 5.45625687e-01 -3.71927559e-01 -9.76734817e-01 1.03840566e+00 1.24412644e+00 4.70334470e-01 1.10826683e+00 -2.36445189e-01 5.62649012e-01 8.37478757e-01 1.17442332e-01 -1.08984756e+00 -4.34622280e-02 3.07592750e-01 8.73296916e-01 -1.43009627e+00 3.65870148e-02 -3.69874714e-03 -2.00603411e-01 1.26461732e+00 6.07918156e-03 -3.78474504e-01 1.05760324e+00 -1.42328337e-01 1.09583057e-01 -3.50060940e-01 -2.41114736e-01 -6.37926757e-02 4.33405250e-01 8.92839611e-01 3.90426993e-01 1.37671798e-01 -2.92951763e-01 -3.47553454e-02 -1.91157028e-01 -3.09988052e-01 4.00828153e-01 9.32743311e-01 -5.67000270e-01 -6.98969007e-01 -4.37785536e-01 8.97678614e-01 -2.37934604e-01 3.94647531e-02 -2.30174690e-01 8.80798221e-01 1.94272064e-02 5.64973116e-01 3.34261030e-01 -5.53172112e-01 4.54961658e-02 4.80890758e-02 1.74685687e-01 -2.26977468e-01 -1.30824149e-01 -1.08961038e-01 -3.69740546e-01 -4.36728805e-01 -7.94406176e-01 -6.37602627e-01 -5.89810312e-01 -1.51447862e-01 -3.71073663e-01 1.09987304e-01 9.95850444e-01 7.71659791e-01 3.00382912e-01 1.34630889e-01 9.75307763e-01 -1.21338761e+00 -8.05576921e-01 -9.29551542e-01 -1.20079052e+00 4.10812199e-01 5.26835918e-01 -7.67606199e-01 -6.54030979e-01 -8.28299075e-02]
[9.387486457824707, 2.304417133331299]
772c9e97-1bfe-4c83-b716-dae7760d74b2
an-adversarial-multi-task-learning-method-for
2306.16313
null
https://arxiv.org/abs/2306.16313v1
https://arxiv.org/pdf/2306.16313v1.pdf
An Adversarial Multi-Task Learning Method for Chinese Text Correction with Semantic Detection
Text correction, especially the semantic correction of more widely used scenes, is strongly required to improve, for the fluency and writing efficiency of the text. An adversarial multi-task learning method is proposed to enhance the modeling and detection ability of character polysemy in Chinese sentence context. Wherein, two models, the masked language model and scoring language model, are introduced as a pair of not only coupled but also adversarial learning tasks. Moreover, the Monte Carlo tree search strategy and a policy network are introduced to accomplish the efficient Chinese text correction task with semantic detection. The experiments are executed on three datasets and five comparable methods, and the experimental results show that our method can obtain good performance in Chinese text correction task for better semantic rationality.
['Zhenping Xie', 'Fanyu Wang']
2023-06-28
null
null
null
null
['multi-task-learning']
['methodology']
[ 3.72136474e-01 -4.91576940e-01 1.42788678e-01 -1.97273433e-01 -8.51258993e-01 -7.29335621e-02 6.04023039e-01 -1.14406973e-01 -6.69986904e-01 9.50672984e-01 5.68668902e-01 -6.31930232e-02 3.24506283e-01 -5.38662910e-01 -4.49794292e-01 -6.04332268e-01 9.82065380e-01 2.73047984e-01 4.45048153e-01 -2.74185002e-01 4.52752858e-01 1.80000588e-02 -6.81176424e-01 4.11893934e-01 1.67759049e+00 3.44678313e-01 5.54958224e-01 7.23889709e-01 -3.60092282e-01 1.17001319e+00 -7.35925972e-01 -1.04762328e+00 -2.52739191e-01 -6.09084904e-01 -5.31092644e-01 -1.89023837e-01 1.75030425e-01 -3.19070995e-01 -3.87255788e-01 1.66268814e+00 9.25778568e-01 1.25227362e-01 6.69100225e-01 -5.63407063e-01 -1.17551458e+00 9.05855775e-01 -6.21729374e-01 2.59375006e-01 1.10851802e-01 -4.72776452e-03 7.85434723e-01 -9.50871229e-01 1.99508891e-01 1.34238362e+00 6.90560699e-01 9.20286000e-01 -4.45380658e-01 -8.27887118e-01 5.83339967e-02 4.68864739e-01 -1.21116972e+00 -2.57204711e-01 6.16789818e-01 -1.74617067e-01 7.94513047e-01 1.44190162e-01 2.39939660e-01 1.40648282e+00 6.35250211e-01 8.70675683e-01 1.08677030e+00 -4.64596778e-01 -1.49603516e-01 3.16991329e-01 -2.73580521e-01 5.66506088e-01 1.60145760e-01 -7.57847056e-02 -3.58571231e-01 1.65339291e-01 4.09480989e-01 7.43956119e-02 4.68086973e-02 3.35572422e-01 -9.42493737e-01 8.08818102e-01 5.77927157e-02 4.47496116e-01 5.32629639e-02 2.49533117e-01 7.85463035e-01 8.18468332e-02 8.92476082e-01 2.57655054e-01 -3.80227387e-01 -1.23059954e-02 -7.88040757e-01 1.77436590e-01 3.10189158e-01 1.26707566e+00 2.69024998e-01 3.20500851e-01 -7.20884919e-01 9.51117873e-01 1.72686338e-01 9.22869325e-01 9.25977528e-01 -4.32680488e-01 9.03564870e-01 3.51365626e-01 -2.68106125e-02 -1.07014787e+00 -1.61167026e-01 -5.30149221e-01 -1.00530410e+00 -4.03402269e-01 -9.28316414e-02 -2.85839081e-01 -6.51229501e-01 1.52422929e+00 -7.24145621e-02 1.64764538e-01 1.63770132e-02 8.02465081e-01 6.93137527e-01 5.95651865e-01 4.49719191e-01 -4.35186297e-01 1.36581874e+00 -1.37220383e+00 -1.37436843e+00 -6.37913465e-01 8.13620389e-01 -1.29968131e+00 1.31841254e+00 3.02499563e-01 -1.20771813e+00 -6.14953876e-01 -1.03631043e+00 -3.16751778e-01 -1.85886726e-01 6.43052220e-01 4.64702308e-01 7.80913770e-01 -5.99071145e-01 5.49185693e-01 -4.60461736e-01 -1.42968938e-01 5.36975145e-01 -1.02619529e-01 2.29054272e-01 -1.07308403e-01 -1.64563107e+00 1.34049332e+00 4.96844262e-01 2.19901621e-01 -8.68581831e-01 -3.40095311e-01 -4.74126101e-01 1.07842639e-01 3.58223379e-01 -6.20973945e-01 1.31809258e+00 -1.24627566e+00 -1.47363579e+00 9.30155754e-01 -1.00871511e-01 -3.26600194e-01 9.84457970e-01 -5.57997823e-01 -8.29132617e-01 4.55318904e-03 2.01077148e-01 9.47509985e-03 1.07622743e+00 -8.01443756e-01 -6.18058264e-01 -2.12075338e-01 -5.46561599e-01 5.80683470e-01 -7.32926309e-01 7.31919527e-01 -4.60306734e-01 -1.24705744e+00 -4.44644094e-01 -4.93864417e-01 -2.79832184e-01 -2.71363914e-01 -3.02061141e-01 -1.37998059e-01 6.50840878e-01 -1.26781118e+00 1.49865460e+00 -1.90319848e+00 1.99097276e-01 -2.20177516e-01 1.32463887e-01 7.17720568e-01 -1.21878482e-01 9.78558287e-02 2.31199642e-03 2.49076977e-01 -4.45049345e-01 -7.23246396e-01 -2.01550916e-01 -8.00256506e-02 -3.90337378e-01 3.22823614e-01 -6.63081333e-02 1.04507506e+00 -9.33772743e-01 -7.80021429e-01 6.65163249e-02 -6.77947924e-02 -5.12760997e-01 2.71729141e-01 -2.97935724e-01 4.38365430e-01 -6.62697434e-01 2.36751482e-01 9.55447733e-01 -1.33568436e-01 -1.04191655e-03 1.14785358e-01 1.75132275e-01 -5.44059277e-02 -7.90738821e-01 1.77002323e+00 -3.52866709e-01 1.23487316e-01 -3.05173457e-01 -7.02642918e-01 9.06140506e-01 1.57990128e-01 -5.22293933e-02 -1.12467480e+00 4.82609212e-01 2.28251204e-01 -1.48272753e-01 -9.04548049e-01 8.62156212e-01 -2.87414491e-01 -1.82725698e-01 3.43610287e-01 -2.08505109e-01 -3.74562263e-01 -1.69300094e-01 4.11518604e-01 6.42237365e-01 1.21160187e-01 1.16140082e-01 -2.08290488e-01 9.74522769e-01 -1.37545511e-01 8.85754764e-01 6.60142362e-01 -4.56295013e-01 4.35924023e-01 3.18509400e-01 8.53605792e-02 -1.31260884e+00 -8.13879073e-01 1.87006354e-01 1.16246295e+00 4.69428182e-01 -3.48679483e-01 -1.01407528e+00 -8.79935503e-01 -3.76220554e-01 1.02117586e+00 -3.15732062e-01 -8.10610771e-01 -7.67814934e-01 -1.08704722e+00 7.15665340e-01 7.44468629e-01 8.47744763e-01 -1.10130668e+00 3.33223552e-01 8.85479227e-02 -5.02639771e-01 -1.32876563e+00 -1.07692778e+00 -3.13017935e-01 -7.39213347e-01 -6.65938377e-01 -6.90707743e-01 -1.00184369e+00 5.26574612e-01 1.69815525e-01 4.97846335e-01 4.71298009e-01 2.85177324e-02 -3.19622941e-02 -6.10204041e-01 -5.52291453e-01 -8.67824018e-01 -9.19497907e-02 1.05703093e-01 -8.32807049e-02 4.47859734e-01 -1.86456114e-01 -3.07173997e-01 4.31410335e-02 -6.99849844e-01 3.91826153e-01 7.30085731e-01 1.11296546e+00 2.92642802e-01 -1.18702821e-01 5.62080860e-01 -1.18120968e+00 8.26423585e-01 -1.84025973e-01 -3.91460985e-01 5.35241306e-01 -8.65680575e-01 -1.89290106e-01 1.02902329e+00 -3.42813104e-01 -1.67760670e+00 -2.67116845e-01 -3.41185957e-01 -2.69145161e-01 2.18485579e-01 3.58866632e-01 -4.16533768e-01 -1.02850832e-01 5.88645637e-01 5.69413543e-01 -3.82449538e-01 -4.36678559e-01 2.78212368e-01 7.72454262e-01 6.13623798e-01 -4.25176114e-01 8.25786173e-01 9.68885273e-02 -3.62632662e-01 -2.93703675e-01 -1.15112960e+00 -2.12712381e-02 -6.74864054e-01 -1.43271387e-01 1.16818583e+00 -1.04648936e+00 -4.40476090e-01 9.98470008e-01 -1.51973844e+00 -9.70863029e-02 4.43119317e-01 5.86591363e-01 -4.28761244e-01 8.89517069e-01 -9.37039554e-01 -6.12257540e-01 -7.25991964e-01 -9.43308115e-01 7.65353143e-01 3.91233414e-01 3.62158328e-01 -1.19567537e+00 -1.99636787e-01 7.05450654e-01 3.04295570e-01 -3.07921857e-01 9.78579640e-01 -7.66065657e-01 -4.33153301e-01 -8.26500915e-03 -4.30272102e-01 7.94525325e-01 -1.22839078e-01 -1.31919011e-01 -8.71888876e-01 -2.86879361e-01 2.73255020e-01 -3.67047012e-01 8.70960236e-01 9.32413191e-02 1.63036513e+00 -2.72791743e-01 -1.79158911e-01 6.22419715e-01 1.24213755e+00 1.23345375e-01 9.13677514e-01 1.00457810e-01 1.00481701e+00 1.87753186e-01 7.98652649e-01 3.61777246e-01 3.19868803e-01 4.68974143e-01 1.60722986e-01 4.00158279e-02 -3.69563401e-01 -6.36350214e-01 3.27423930e-01 1.55800748e+00 1.08052768e-01 -4.58269686e-01 -6.84642255e-01 2.68358886e-01 -1.82876778e+00 -9.86820042e-01 -6.62391841e-01 1.79403126e+00 1.10144019e+00 2.64960378e-01 -4.31102484e-01 -2.65303582e-01 1.31915927e+00 1.29342318e-01 -5.02150357e-01 -3.83864105e-01 -5.37005305e-01 5.82762137e-02 6.41412675e-01 5.30979753e-01 -1.12187731e+00 1.52837336e+00 6.42475748e+00 1.42598307e+00 -8.26930404e-01 7.70324528e-01 5.59818149e-01 2.09035978e-01 -4.42798048e-01 -8.94352570e-02 -8.22859824e-01 8.67884457e-01 5.77342391e-01 -3.86087328e-01 5.31182051e-01 5.38102984e-01 5.36725760e-01 5.00782542e-02 -3.59272361e-01 1.01581800e+00 5.54726481e-01 -1.08840251e+00 3.51199836e-01 -5.67977369e-01 9.53478456e-01 -1.84566155e-01 6.45560995e-02 3.43529999e-01 3.84408295e-01 -8.82825911e-01 9.07637119e-01 7.82100677e-01 9.07391906e-01 -8.46199155e-01 7.98318267e-01 5.83129466e-01 -7.94500291e-01 -1.22903399e-01 -5.99378765e-01 -1.36653975e-01 9.38597023e-02 3.69324923e-01 -1.52690455e-01 5.61064720e-01 4.03729260e-01 1.03248787e+00 -1.00207269e+00 8.05613220e-01 -8.51055205e-01 6.16795838e-01 5.49206257e-01 -6.03749275e-01 2.96578207e-03 -2.63318717e-01 4.28845495e-01 1.25409794e+00 1.29484892e-01 1.32068023e-01 -1.00901246e-01 7.68300414e-01 -2.49167860e-01 4.82513517e-01 -9.18815434e-02 6.04762323e-02 5.16041160e-01 1.20836091e+00 -3.91586244e-01 -4.44315314e-01 -4.34668243e-01 1.59361148e+00 4.16008711e-01 2.10408553e-01 -1.37107503e+00 -6.87956929e-01 9.96017605e-02 -5.54762602e-01 5.30587323e-02 -7.50065828e-03 -9.61963356e-01 -1.60694468e+00 1.10045627e-01 -8.80179763e-01 2.69601911e-01 -1.04694068e+00 -1.36767578e+00 1.03407629e-01 -5.80982208e-01 -1.13742208e+00 4.25356388e-01 -4.95128006e-01 -9.40774143e-01 1.17202342e+00 -1.62355232e+00 -1.34199178e+00 -2.61983007e-01 7.99382210e-01 7.07949460e-01 -4.67033863e-01 5.75467944e-01 6.47896349e-01 -8.90777707e-01 1.09904075e+00 5.38092613e-01 1.53047517e-01 1.01280308e+00 -9.72080767e-01 2.40864441e-01 1.36038375e+00 -4.65566278e-01 2.50091821e-01 6.05597734e-01 -1.18441391e+00 -8.63704383e-01 -1.37637663e+00 1.31030250e+00 -5.14345169e-01 6.85092747e-01 -4.93237019e-01 -9.43782091e-01 6.56447291e-01 2.43462592e-01 -5.57428002e-01 2.91483432e-01 -3.06039155e-01 6.58250600e-02 -1.24751016e-01 -1.10815811e+00 8.18164825e-01 1.12346017e+00 -6.34644091e-01 -9.67188656e-01 8.06921661e-01 1.17157042e+00 -6.83890343e-01 -5.31770945e-01 3.68745551e-02 -8.31806938e-06 -4.60157245e-01 7.37199187e-01 -6.85670197e-01 1.03164732e+00 -1.62932917e-01 1.50041625e-01 -1.43483019e+00 -6.29165590e-01 -4.79253352e-01 2.33610258e-01 1.52899683e+00 3.46920878e-01 -4.49966699e-01 5.04664540e-01 2.91230589e-01 -5.00309944e-01 -2.09799051e-01 -8.85252893e-01 -5.10419190e-01 5.32460988e-01 -2.78498411e-01 5.08494675e-01 9.89445448e-01 -2.68878520e-01 6.07082486e-01 -9.77257788e-01 -2.68978775e-02 3.92333627e-01 -4.11031455e-01 3.09446424e-01 -7.81589091e-01 -7.35874921e-02 -4.75742340e-01 2.00476244e-01 -9.16063905e-01 4.40108627e-01 -1.12382448e+00 -5.15498929e-02 -1.21302021e+00 6.86603725e-01 -1.25890881e-01 -8.12279508e-02 1.24007724e-01 -9.15356636e-01 -2.64356919e-02 1.52563065e-01 2.14946896e-01 -7.39598155e-01 1.10568357e+00 1.69473124e+00 -2.25711286e-01 3.35074246e-01 -8.60169623e-03 -7.76363790e-01 7.92943716e-01 6.07983589e-01 -7.47444272e-01 -1.28497645e-01 -9.92732704e-01 3.88062418e-01 7.57407248e-02 2.39161626e-01 -7.56840706e-01 4.40392941e-01 -2.67065406e-01 5.61463714e-01 -5.50835669e-01 -4.26616669e-02 -5.66459477e-01 -5.12372077e-01 7.98059285e-01 -5.65441310e-01 2.00201035e-01 -4.97224443e-02 6.67290926e-01 7.42470548e-02 -5.87619781e-01 1.20502245e+00 -2.17122599e-01 -7.01176345e-01 4.18168664e-01 -3.47973019e-01 3.06091607e-01 9.70390737e-01 1.87508076e-01 -5.09883583e-01 -2.24158883e-01 -5.12008846e-01 4.41794306e-01 1.45846784e-01 6.59540355e-01 6.75633252e-01 -1.51533878e+00 -1.05960858e+00 -1.43141104e-02 1.32330611e-01 -4.74233150e-01 5.69014907e-01 7.61254966e-01 -7.87746906e-01 1.67129427e-01 2.15212815e-02 -1.35507984e-02 -1.10660589e+00 7.74779856e-01 5.54112434e-01 -4.09752816e-01 -3.88865173e-01 8.10095429e-01 -8.62126127e-02 -5.17043710e-01 9.99927670e-02 1.04248174e-01 -3.60454381e-01 -3.23996097e-01 7.18742549e-01 4.81210053e-01 5.83814271e-02 -4.67110395e-01 -1.35015950e-01 5.80268323e-01 -3.06573570e-01 1.40246630e-01 8.82505476e-01 -6.40506268e-01 -5.13322055e-01 9.97091532e-02 8.26807618e-01 3.73633623e-01 -8.29903126e-01 -5.34797490e-01 -8.69033113e-02 -5.26824594e-01 -1.85636394e-02 -1.19477117e+00 -7.72067308e-01 1.08039606e+00 5.27840257e-01 -3.47646415e-01 1.15962005e+00 -3.70472282e-01 1.16783810e+00 1.84031636e-01 1.00526540e-02 -1.65062141e+00 4.84983385e-01 9.49067831e-01 7.22706378e-01 -1.19487870e+00 -1.69801805e-02 -3.38158637e-01 -1.01610231e+00 1.15408659e+00 1.16176236e+00 -1.87135831e-01 2.31540814e-01 1.48349345e-01 2.62321346e-02 2.25906968e-01 -4.89529490e-01 2.15571731e-01 -6.82552010e-02 4.97973263e-01 3.59879941e-01 1.18195377e-01 -9.92808461e-01 1.14560497e+00 -7.67918304e-02 -1.52107492e-01 6.58075154e-01 3.25087160e-01 -4.52217966e-01 -9.50474203e-01 -2.88204014e-01 4.67750967e-01 -7.20154822e-01 -6.67743206e-01 -2.75273353e-01 1.19483806e-01 2.10476667e-01 9.23789263e-01 -1.21401049e-01 -5.19776702e-01 2.32239693e-01 -1.22756831e-01 2.70033985e-01 -5.70413530e-01 -7.94022024e-01 5.92579087e-03 -5.19177914e-02 -1.68970227e-01 -2.27576956e-01 -5.56742787e-01 -1.22971141e+00 -6.63078487e-01 -6.30356014e-01 -4.81905509e-03 3.56261849e-01 1.26825821e+00 -6.63846079e-03 7.77942657e-01 9.35342133e-01 -4.38570976e-02 -9.54815626e-01 -1.25644541e+00 -6.36670887e-01 5.46121538e-01 -1.66366652e-01 -1.88988268e-01 -1.96118027e-01 -1.49204396e-02]
[10.963138580322266, 10.82551383972168]
b7598349-2c23-41e9-abfe-7a3eabf3dede
contrastively-enforcing-distinctiveness-for
null
null
https://openreview.net/forum?id=jNsynsmDkl
https://openreview.net/pdf?id=jNsynsmDkl
Contrastively Enforcing Distinctiveness for Multi-Label Classification
Recently, as an effective way of learning latent representations, contrastive learning has been increasingly popular and successful in various domains. The success of constrastive learning in single-label classifications motivates us to leverage this learning framework to enhance distinctiveness for better performance in multi-label image classification. In this paper, we show that a direct application of contrastive learning can hardly improve in multi-label cases. Accordingly, we propose a novel framework for multi-label classification with contrastive learning in a fully supervised setting, which learns multiple representations of an image under the context of different labels. This facilities a simple yet intuitive adaption of contrastive learning into our model to boost its performance in multi-label image classification. Extensive experiments on two benchmark datasets show that the proposed framework achieves state-of-the-art performance in the comparison with the advanced methods in multi-label classification.
['Jianfei Cai', 'Dinh Phung', 'He Zhao', 'Son Duy Dao']
2021-09-29
null
null
null
null
['multi-label-image-classification']
['computer-vision']
[ 6.54421687e-01 -3.36272180e-01 -6.12248898e-01 -4.89669442e-01 -1.06872880e+00 -5.86192012e-01 6.62459970e-01 3.19570869e-01 -3.07282239e-01 4.60632831e-01 -2.42992446e-01 -4.03521359e-02 -1.99198455e-01 -3.03182989e-01 -3.53947461e-01 -1.03432870e+00 2.83149719e-01 1.32657677e-01 -1.40681639e-01 6.20165318e-02 3.11296582e-01 3.30833644e-01 -1.63527346e+00 6.06087565e-01 4.98325706e-01 1.05472434e+00 8.22808072e-02 3.42934072e-01 -7.52777159e-02 1.03204966e+00 -3.00866038e-01 -3.82402658e-01 1.71631411e-01 -4.63357478e-01 -1.07561910e+00 4.53364462e-01 9.31988001e-01 1.29928738e-01 3.11202351e-02 1.14413679e+00 2.76254147e-01 -9.18178633e-02 1.08620703e+00 -1.38009965e+00 -7.34189153e-01 2.70437688e-01 -1.03698778e+00 7.56020918e-02 1.09227061e-01 -4.34927166e-01 1.33000302e+00 -1.05529404e+00 3.64636958e-01 1.25219131e+00 7.61726618e-01 5.04519761e-01 -1.36212015e+00 -8.26290369e-01 3.52683008e-01 1.52135924e-01 -1.32615805e+00 -2.46359229e-01 8.44828308e-01 -3.28015774e-01 2.77176738e-01 3.37881923e-01 5.88386096e-02 1.13366461e+00 2.58569449e-01 1.14935291e+00 1.83348823e+00 -7.13940203e-01 -1.94413751e-01 3.20887059e-01 3.50069582e-01 9.14608240e-01 -6.50518686e-02 -2.33080462e-02 -3.46220523e-01 -3.14249843e-01 3.83694261e-01 4.77340668e-01 1.30849071e-02 -6.09742939e-01 -1.00413716e+00 9.95981693e-01 4.57211256e-01 3.50097150e-01 1.65715180e-02 1.63204804e-01 6.81431472e-01 4.84372169e-01 9.15063560e-01 3.78337771e-01 -3.49476039e-01 4.76651400e-01 -9.22968984e-01 9.84116569e-02 1.74711704e-01 8.12311828e-01 8.07599545e-01 -2.22138271e-01 -1.80914462e-01 1.24663568e+00 3.06772113e-01 2.29764879e-01 7.54048407e-01 -6.33134723e-01 2.53931224e-01 7.43640184e-01 -3.16158503e-01 -7.88621962e-01 -6.96204841e-01 -7.50656188e-01 -1.00022793e+00 2.62749910e-01 1.84574515e-01 2.93363899e-01 -7.30358064e-01 1.53887928e+00 2.07290590e-01 6.26936436e-01 1.26968935e-01 3.75569463e-01 7.34916747e-01 3.67506057e-01 3.36550862e-01 -4.94229257e-01 1.39437032e+00 -1.49904299e+00 -6.04589939e-01 -9.48068425e-02 1.01351035e+00 -8.13948274e-01 8.67946684e-01 3.67803842e-01 -6.00167632e-01 -6.91166937e-01 -1.06378806e+00 1.01912871e-01 -3.93516630e-01 1.82679698e-01 6.86730564e-01 6.38118207e-01 -6.11841500e-01 5.24461269e-01 -2.22219363e-01 -3.07342112e-01 6.70980811e-01 1.70331433e-01 -6.30881131e-01 -2.55193144e-01 -7.96308339e-01 7.96355665e-01 4.47500885e-01 -2.71458358e-01 -6.93875551e-01 -4.16181147e-01 -7.90566802e-01 -1.50925606e-01 5.26956856e-01 -3.03813517e-01 1.23151743e+00 -1.05656588e+00 -1.12031996e+00 1.45537031e+00 1.61296993e-01 -8.80056173e-02 3.76236141e-01 -4.12705131e-02 -4.87852037e-01 2.88458377e-01 4.34573174e-01 6.17461741e-01 1.09521222e+00 -1.62114346e+00 -9.78762686e-01 -3.62152994e-01 1.18301392e-01 2.21445337e-01 -5.14378488e-01 7.20488727e-02 8.64734575e-02 -8.43527019e-01 2.53677458e-01 -1.30191922e+00 -8.14236552e-02 -2.42743760e-01 -2.86092877e-01 -6.28421009e-01 8.16439331e-01 2.39463337e-02 1.06887817e+00 -2.19658899e+00 9.79001671e-02 -1.29802870e-02 3.34698975e-01 1.04733348e-01 -2.65020400e-01 3.82742643e-01 -4.33837622e-01 1.16452850e-01 -1.12291850e-01 -6.10963523e-01 -3.73940468e-02 7.90166184e-02 -2.22977564e-01 7.03269362e-01 1.50056943e-01 1.11039507e+00 -1.03977132e+00 -8.36272359e-01 7.23669678e-02 8.67080614e-02 -1.15130790e-01 2.03731745e-01 2.54792392e-01 5.66739738e-01 -3.19661409e-01 9.04468775e-01 6.87208474e-01 -8.02082956e-01 4.75289166e-01 -2.26475239e-01 1.09305926e-01 -3.96095872e-01 -1.02837694e+00 1.67962027e+00 -6.55908048e-01 2.99172938e-01 -2.20790222e-01 -1.24768484e+00 6.62526190e-01 4.07395273e-01 6.01245165e-01 -5.52052736e-01 1.30675092e-01 3.65832567e-01 -4.82575655e-01 -3.49535644e-01 3.97326767e-01 -4.70205933e-01 -3.16086531e-01 7.08768964e-01 2.86495000e-01 1.07625574e-01 5.92015982e-02 3.35790440e-02 5.20695448e-01 7.27447420e-02 8.62270296e-01 -3.80088717e-01 6.28621340e-01 -4.07324195e-01 3.55981231e-01 7.12867379e-01 -3.16730291e-01 4.55234021e-01 1.11631438e-01 -4.84374136e-01 -6.88795686e-01 -7.41106689e-01 -4.61532176e-01 1.78573954e+00 2.86257058e-01 -3.88866544e-01 -3.49678904e-01 -1.47107315e+00 9.45544150e-03 1.52312964e-01 -7.73106337e-01 3.57636041e-03 -5.04469395e-01 -9.65494156e-01 6.46596432e-01 4.32901025e-01 5.06555259e-01 -8.60409856e-01 -3.11680496e-01 3.62703167e-02 -5.45134321e-02 -1.01149189e+00 -3.56725454e-01 6.35069072e-01 -8.70340407e-01 -1.18133307e+00 -8.53712261e-01 -1.29926550e+00 7.58512199e-01 8.43072355e-01 9.42948818e-01 1.55922562e-01 -1.96519181e-01 4.03190196e-01 -4.51252013e-01 -7.71521255e-02 -5.93351483e-01 2.35139057e-01 -5.48372418e-02 4.27137524e-01 1.09352134e-01 -3.39182824e-01 -3.00089210e-01 5.18351793e-01 -1.20351160e+00 2.97062527e-02 6.97735667e-01 1.26750326e+00 6.16563201e-01 4.10626382e-02 9.13494170e-01 -1.49910164e+00 4.43553716e-01 -6.83175027e-01 -2.75234342e-01 6.53102398e-01 -1.10473382e+00 1.47975340e-01 6.40540302e-01 -5.41101754e-01 -8.68729472e-01 1.82335868e-01 1.10306963e-01 -3.37406963e-01 -2.23859504e-01 3.80847275e-01 1.62459195e-01 -4.64654475e-01 4.63890672e-01 1.78378224e-01 -1.81029156e-01 -6.67899072e-01 6.16751015e-01 8.80643964e-01 2.15740070e-01 -4.36085403e-01 6.19802952e-01 5.25928855e-01 4.55796242e-01 -2.28441566e-01 -1.49263871e+00 -9.55914795e-01 -9.38713670e-01 -1.65131256e-01 6.61493003e-01 -1.04944074e+00 -5.80639005e-01 6.44702256e-01 -6.59002483e-01 6.79970682e-02 1.44691318e-01 1.24383688e-01 -5.95934927e-01 5.56833446e-01 -6.60942554e-01 -4.12282407e-01 -1.32117063e-01 -1.21340084e+00 1.23548079e+00 2.56810337e-01 3.64452638e-02 -1.38594759e+00 9.02396515e-02 6.04767144e-01 6.87282309e-02 6.82640001e-02 1.04052329e+00 -9.10828829e-01 -1.20917253e-01 -2.68363208e-02 -4.27114516e-01 3.83447707e-01 4.55994934e-01 -5.14038026e-01 -1.39490747e+00 -6.98435307e-01 1.86886378e-02 -9.58764851e-01 1.12745535e+00 -9.91608053e-02 1.18361127e+00 -9.50433612e-02 -5.84909618e-01 4.80261475e-01 1.76898301e+00 -1.87086120e-01 1.02916718e-01 4.00514662e-01 1.00038195e+00 5.37601113e-01 9.40169930e-01 3.14683557e-01 3.29058409e-01 9.68443811e-01 5.22226870e-01 -3.47602427e-01 -4.87340800e-02 5.97293451e-02 -1.02155045e-01 8.53683293e-01 8.47258791e-02 4.90905829e-02 -6.53714657e-01 2.28114590e-01 -1.93809462e+00 -9.47814107e-01 4.38474258e-03 1.80899560e+00 8.59144330e-01 -9.78374258e-02 1.71444118e-01 1.41149044e-01 8.52536142e-01 4.67625916e-01 -3.41961294e-01 -2.53933758e-01 -1.55252695e-01 9.05813873e-02 4.84373659e-01 1.37018204e-01 -1.80953872e+00 8.65927041e-01 6.74037981e+00 1.24221325e+00 -1.20070755e+00 4.94905829e-01 8.07053447e-01 2.14539006e-01 5.72225973e-02 -3.35554093e-01 -1.03200030e+00 2.70757675e-01 8.28842700e-01 1.75178498e-01 -8.09197724e-02 9.32802081e-01 -4.12326902e-01 1.94825798e-01 -1.13938916e+00 1.24346805e+00 5.08458436e-01 -1.22295189e+00 1.40296612e-02 1.00356713e-01 1.22037709e+00 -3.97017032e-01 6.28806770e-01 4.94501650e-01 -3.72842886e-02 -9.94498789e-01 5.27027488e-01 7.66726583e-02 9.32258368e-01 -7.99089909e-01 6.47339046e-01 3.17860931e-01 -1.34057379e+00 -4.91535068e-01 -2.62016088e-01 2.19472367e-02 -1.27186596e-01 3.15573603e-01 -7.64486492e-01 8.39697361e-01 1.50524929e-01 1.02079082e+00 -1.09016454e+00 7.85622954e-01 -1.71988364e-02 4.05649692e-01 4.59408075e-01 2.78393745e-01 5.28479099e-01 8.27024505e-03 -1.51121775e-02 1.41972351e+00 -6.31279871e-02 -3.39492291e-01 6.71661198e-01 1.17406458e-01 -3.58915031e-01 5.76893568e-01 -6.78890526e-01 2.52633572e-01 2.39429563e-01 1.64908504e+00 -9.69940722e-01 -4.04668570e-01 -8.00247908e-01 1.04379117e+00 6.95375741e-01 1.81726992e-01 -8.25924277e-01 9.09787714e-02 2.47356117e-01 -3.39784741e-01 2.30292886e-01 2.15272114e-01 -8.59109759e-02 -1.06947017e+00 -2.58322507e-01 -1.05340815e+00 8.23893070e-01 -3.15413773e-01 -1.64824176e+00 5.39800406e-01 3.63022052e-02 -1.63028729e+00 -2.51192540e-01 -6.98410988e-01 -6.70305565e-02 3.56209815e-01 -2.04878902e+00 -1.78504133e+00 -1.15452960e-01 5.29942214e-01 7.36643314e-01 -4.84812617e-01 1.16235399e+00 4.52567786e-01 -3.39395374e-01 9.41989243e-01 6.13374949e-01 -6.34447113e-02 1.18906581e+00 -1.40838492e+00 -1.40716642e-01 3.91476572e-01 6.05816841e-01 3.23286176e-01 1.71334550e-01 -2.13607147e-01 -7.88481355e-01 -1.30068278e+00 7.17167258e-01 -3.64420384e-01 5.05714417e-01 -2.24740833e-01 -7.51690745e-01 7.23442018e-01 2.59804338e-01 3.53990674e-01 1.24455559e+00 2.30866864e-01 -9.03324783e-01 -1.72431380e-01 -1.08652020e+00 3.12112629e-01 7.58320212e-01 -8.81549001e-01 -3.33624005e-01 5.76705098e-01 5.14577568e-01 -6.63531721e-02 -9.98297691e-01 5.33037245e-01 6.47389472e-01 -7.95984447e-01 1.05101287e+00 -5.90962112e-01 3.53370756e-01 -8.28447118e-02 -3.28843266e-01 -1.32033408e+00 -6.32874846e-01 -1.96059495e-01 8.67550150e-02 1.23651743e+00 7.58230090e-02 -5.39255023e-01 4.58000779e-01 -1.71090260e-01 -1.51583506e-02 -1.01883137e+00 -7.45590627e-01 -7.88689017e-01 3.66484910e-01 1.18144728e-01 2.32344463e-01 1.43522441e+00 -4.53959107e-02 5.91735959e-01 -7.75393367e-01 3.24396715e-02 7.93084502e-01 5.83821058e-01 3.55161667e-01 -1.54847431e+00 -2.42618755e-01 -3.92819613e-01 -4.48046237e-01 -1.10547996e+00 7.82646000e-01 -1.32328212e+00 -1.66498974e-01 -1.02835941e+00 8.55450153e-01 -7.12498367e-01 -1.01588905e+00 4.96698529e-01 -6.15807116e-01 9.28724110e-01 3.57637137e-01 5.49698114e-01 -1.15228438e+00 1.40904441e-01 1.26340199e+00 -4.93853837e-01 3.45752537e-01 1.04765721e-01 -8.42236638e-01 6.44083619e-01 6.14034057e-01 -7.78078258e-01 -3.68931085e-01 7.40489662e-02 8.84023309e-02 -1.74214840e-01 1.40381828e-01 -7.86329448e-01 -8.49231780e-02 -1.81543201e-01 2.02194855e-01 -3.74890774e-01 2.89026201e-01 -7.98462510e-01 -6.48214668e-02 3.47762108e-01 -7.42944777e-01 8.27362482e-03 -1.52531385e-01 7.13681698e-01 -4.23601091e-01 -5.79115510e-01 1.06397033e+00 -2.24886522e-01 -9.03511107e-01 2.70431399e-01 -7.26954117e-02 -5.16643673e-02 1.06017625e+00 4.53423150e-02 -3.21931481e-01 4.01734971e-02 -6.43740535e-01 -1.39203072e-02 4.32799011e-01 5.11735380e-01 3.88140827e-01 -1.67337227e+00 -6.93703830e-01 1.37727305e-01 6.94286883e-01 -6.84453011e-01 2.52796352e-01 7.65182495e-01 3.02941240e-02 4.17754769e-01 -2.73451626e-01 -7.51996338e-01 -1.82488465e+00 8.40802312e-01 1.42310604e-01 -7.44550228e-01 -3.67402405e-01 8.14344347e-01 6.63803220e-01 -4.88397539e-01 8.09723139e-02 1.27331182e-01 -6.44156992e-01 5.72379351e-01 5.72002947e-01 3.31620336e-01 -2.20312048e-02 -9.03023899e-01 -3.10048431e-01 1.02097499e+00 -6.15049362e-01 2.78548151e-01 1.09930348e+00 -4.42081243e-01 -1.00871131e-01 9.39657986e-01 1.77537847e+00 -9.61973295e-02 -1.03778303e+00 -5.20570040e-01 2.97096074e-01 -4.19798255e-01 -6.77302405e-02 -5.67349851e-01 -9.09076631e-01 7.60478437e-01 8.71469200e-01 3.94507587e-01 1.11286139e+00 5.80372848e-02 4.92472202e-01 2.99629599e-01 3.82759362e-01 -1.00445390e+00 7.04680741e-01 3.02494437e-01 4.81683791e-01 -1.81957221e+00 2.18771040e-01 -4.91039246e-01 -6.46064520e-01 1.20266449e+00 3.63512844e-01 3.06928214e-02 7.16793656e-01 -2.39774033e-01 4.38812137e-01 -2.56037176e-01 -6.19713366e-01 -1.78539336e-01 4.15264249e-01 1.99320883e-01 5.62224209e-01 1.59517303e-01 -3.21726263e-01 1.09122843e-01 5.96698046e-01 -4.29475069e-01 2.79407471e-01 1.02573204e+00 -4.69845116e-01 -1.49087095e+00 -2.97895789e-01 2.54960507e-01 -8.78892899e-01 -1.79888792e-02 -8.32785293e-02 6.69253707e-01 2.14505062e-01 1.04875445e+00 -3.26640218e-01 -2.66366333e-01 -6.73903972e-02 3.51127297e-01 7.20892072e-01 -8.59273493e-01 -6.97238088e-01 1.57648399e-01 -1.16455548e-01 -2.75176704e-01 -1.01842356e+00 -5.51646352e-01 -9.60697234e-01 1.81381136e-01 -6.03485107e-01 -4.73606773e-03 4.72311467e-01 1.07234728e+00 5.10646217e-02 4.75741565e-01 1.09712255e+00 -6.15568936e-01 -7.70934105e-01 -9.07715559e-01 -9.02429461e-01 8.31414819e-01 4.02524889e-01 -9.90446270e-01 -3.96468610e-01 -3.03886011e-02]
[9.660860061645508, 4.185973167419434]
5ed7011d-778a-4840-bafa-0dfb8bad4901
full-duplex-strategy-for-video-object
2108.03151
null
https://arxiv.org/abs/2108.03151v3
https://arxiv.org/pdf/2108.03151v3.pdf
Full-Duplex Strategy for Video Object Segmentation
Previous video object segmentation approaches mainly focus on using simplex solutions between appearance and motion, limiting feature collaboration efficiency among and across these two cues. In this work, we study a novel and efficient full-duplex strategy network (FSNet) to address this issue, by considering a better mutual restraint scheme between motion and appearance in exploiting the cross-modal features from the fusion and decoding stage. Specifically, we introduce the relational cross-attention module (RCAM) to achieve bidirectional message propagation across embedding sub-spaces. To improve the model's robustness and update the inconsistent features from the spatial-temporal embeddings, we adopt the bidirectional purification module (BPM) after the RCAM. Extensive experiments on five popular benchmarks show that our FSNet is robust to various challenging scenarios (e.g., motion blur, occlusion) and achieves favourable performance against existing cutting-edges both in the video object segmentation and video salient object detection tasks. The project is publicly available at: https://dpfan.net/FSNet.
['Ling Shao', 'Jianbing Shen', 'Deng-Ping Fan', 'Zhe Wu', 'Keren Fu', 'Ge-Peng Ji']
2021-08-06
null
http://openaccess.thecvf.com//content/ICCV2021/html/Ji_Full-Duplex_Strategy_for_Video_Object_Segmentation_ICCV_2021_paper.html
http://openaccess.thecvf.com//content/ICCV2021/papers/Ji_Full-Duplex_Strategy_for_Video_Object_Segmentation_ICCV_2021_paper.pdf
iccv-2021-1
['video-salient-object-detection', 'unsupervised-video-object-segmentation', 'video-polyp-segmentation']
['computer-vision', 'computer-vision', 'computer-vision']
[ 1.71829298e-01 -1.39288232e-01 -3.15938264e-01 -2.54936278e-01 -4.62291837e-01 -4.06394929e-01 6.16470873e-01 -4.07210886e-01 -4.71525311e-01 4.82318670e-01 3.60586703e-01 -6.96643814e-02 -2.82580964e-04 -4.26493496e-01 -7.47594774e-01 -8.22541177e-01 -8.10031444e-02 -2.31764480e-01 5.92231035e-01 1.06368981e-01 2.77466536e-01 2.87990481e-01 -1.42140245e+00 1.71952665e-01 7.93339491e-01 9.64949727e-01 4.15945172e-01 5.86227894e-01 -2.26468936e-01 8.18443179e-01 -1.10193998e-01 -4.86848325e-01 4.32031929e-01 -2.86187410e-01 -9.58914757e-01 2.61088014e-01 3.32311153e-01 -5.36047161e-01 -7.15430439e-01 1.22826195e+00 4.30663049e-01 7.13438168e-02 3.26593786e-01 -1.60420477e+00 -7.04647899e-01 6.13943756e-01 -9.92582142e-01 3.52249593e-01 1.10397711e-01 3.20870250e-01 1.02775824e+00 -1.07262278e+00 7.53235161e-01 1.32912087e+00 5.26405394e-01 5.94605446e-01 -1.31561017e+00 -6.76159322e-01 5.11091053e-01 6.50513947e-01 -1.42852128e+00 -4.97071207e-01 8.84758949e-01 -4.80963737e-01 5.89446187e-01 3.19077253e-01 6.00661814e-01 1.10187685e+00 1.47029117e-01 1.11272335e+00 8.14662457e-01 9.57722962e-03 -4.84674536e-02 1.63100079e-01 2.79598832e-01 6.96363926e-01 1.06253088e-01 -3.41886282e-02 -6.48919761e-01 -7.77011588e-02 8.07842433e-01 -6.25151768e-02 -5.57580769e-01 -6.16271555e-01 -1.11385727e+00 6.27279341e-01 4.57105428e-01 2.53529280e-01 -2.15928242e-01 2.09999472e-01 3.31885546e-01 1.23082951e-01 4.42474335e-01 1.24276131e-01 -4.25700277e-01 -5.56129143e-02 -8.65930855e-01 1.87455844e-02 6.72706842e-01 1.09938169e+00 8.64818454e-01 -2.46185198e-01 -4.73141193e-01 8.25292170e-01 6.72591925e-01 1.45025387e-01 2.31895998e-01 -1.01229298e+00 4.29570794e-01 4.21284378e-01 -4.63822260e-02 -1.26762247e+00 -4.08124745e-01 -2.73149699e-01 -8.49730551e-01 -1.17534898e-01 2.77427465e-01 -5.89776300e-02 -9.75660980e-01 1.84676719e+00 6.86418653e-01 7.93084562e-01 -1.42929927e-01 1.40330827e+00 1.08568418e+00 7.83420742e-01 1.97810844e-01 -1.04487091e-01 1.28302348e+00 -1.21315408e+00 -7.37297416e-01 -2.36091793e-01 5.79466343e-01 -8.94909084e-01 4.60213929e-01 -2.71746013e-02 -1.13102591e+00 -5.62725246e-01 -1.02213347e+00 -4.18787628e-01 -2.48531252e-01 2.20200941e-01 5.87931871e-01 3.77562732e-01 -1.09489536e+00 4.21256810e-01 -8.60429764e-01 -3.48217517e-01 5.32066107e-01 5.34639180e-01 -3.43031615e-01 -2.69007206e-01 -1.14595008e+00 6.24248624e-01 1.89957321e-01 4.75132048e-01 -6.50836647e-01 -7.31221318e-01 -8.44509542e-01 -1.15941264e-01 3.97521913e-01 -7.42501318e-01 7.17492759e-01 -1.13773823e+00 -1.44889307e+00 6.45510614e-01 -2.02889815e-01 -2.81357080e-01 6.52705371e-01 -2.98479140e-01 -1.50614992e-01 1.11560188e-01 3.00636832e-02 1.21125495e+00 9.97973502e-01 -1.30410945e+00 -6.36561096e-01 -1.40294299e-01 1.50558025e-01 2.85181284e-01 -3.32476735e-01 1.03941403e-01 -1.15800142e+00 -6.95815682e-01 1.95199862e-01 -1.03518391e+00 -1.66558236e-01 1.99555740e-01 -5.95146060e-01 -1.81594625e-01 1.02572739e+00 -7.27466881e-01 1.31006598e+00 -2.32080603e+00 5.95227063e-01 2.04936847e-01 3.01121831e-01 3.61909062e-01 -3.92165989e-01 1.33812636e-01 -7.72978645e-04 -2.00799620e-03 -4.13600683e-01 -5.68177342e-01 -4.15465087e-02 2.57097632e-01 3.46688032e-02 8.22722197e-01 5.40060341e-01 9.66602981e-01 -7.06591785e-01 -7.58556247e-01 2.79482841e-01 6.45555139e-01 -7.48282731e-01 1.33795798e-01 7.20486790e-02 4.11213189e-01 -4.44187135e-01 5.83324254e-01 1.18294799e+00 -3.41802508e-01 6.49274811e-02 -5.73435724e-01 -2.14677632e-01 6.38581142e-02 -1.26815987e+00 1.87565565e+00 -3.06069031e-02 7.54661977e-01 3.40450257e-01 -8.88636589e-01 5.17117023e-01 5.91435544e-02 6.57010853e-01 -5.63348293e-01 3.75955582e-01 -1.12132609e-01 -4.36895750e-02 -6.35355115e-01 5.38452864e-01 4.58564043e-01 3.06936085e-01 1.22688241e-01 2.63461590e-01 3.80878121e-01 2.82121569e-01 4.51017082e-01 9.89857733e-01 3.16050380e-01 -1.30216643e-01 -4.11070228e-01 7.83499360e-01 -2.69578278e-01 9.15351987e-01 6.17667019e-01 -5.06485343e-01 7.81388938e-01 5.59851646e-01 -1.06760986e-01 -8.80547643e-01 -8.50874543e-01 -1.53579950e-01 1.01771462e+00 8.09775472e-01 -5.28820693e-01 -6.62581801e-01 -6.71585739e-01 -1.27423741e-03 2.47520193e-01 -6.57311141e-01 -1.17353357e-01 -6.10114932e-01 -9.36059117e-01 2.95943469e-01 4.47934568e-01 7.09780335e-01 -7.05120385e-01 -3.04995537e-01 2.52531111e-01 -3.01592261e-01 -1.25993502e+00 -7.40486860e-01 -1.21378139e-01 -4.23057050e-01 -1.05761397e+00 -7.76059806e-01 -7.81333447e-01 5.96515656e-01 4.87638712e-01 5.64501166e-01 1.86073989e-01 -4.00715232e-01 5.06510913e-01 -4.11800712e-01 2.33723000e-01 1.49227679e-02 8.98809955e-02 -2.59207308e-01 4.35217768e-01 3.13486874e-01 -3.39542359e-01 -8.04381013e-01 3.90716553e-01 -9.87485647e-01 4.45562184e-01 4.98594105e-01 8.18318725e-01 2.75063127e-01 -3.88273746e-01 2.66453266e-01 -5.21242023e-01 2.47932330e-01 -7.75834322e-01 -6.61544740e-01 2.31884554e-01 -3.53174210e-01 -1.39856815e-01 3.29303816e-02 -4.05021042e-01 -1.16615844e+00 1.33671284e-01 -1.46950141e-01 -4.63599831e-01 5.33892177e-02 9.69912186e-02 -3.16036940e-01 -2.72540599e-01 -3.29892151e-02 2.32078925e-01 -1.55336961e-01 -4.06524211e-01 3.77156407e-01 4.77916926e-01 3.35599959e-01 -2.58690298e-01 8.87706339e-01 6.64828539e-01 -1.05125725e-01 -8.34326148e-01 -5.44848740e-01 -7.40707338e-01 -5.92749476e-01 -3.30133885e-01 1.17480028e+00 -9.79347944e-01 -5.04574955e-01 6.72210276e-01 -1.43998277e+00 6.16311058e-02 1.06899552e-01 5.93775034e-01 -3.58316720e-01 6.07528985e-01 -8.28767300e-01 -5.63513756e-01 -2.69670263e-02 -1.44792664e+00 1.06383634e+00 4.44812447e-01 4.88512069e-02 -7.22168863e-01 -1.22239433e-01 3.02355409e-01 3.77193511e-01 -1.06522888e-01 5.23620248e-01 -1.96886122e-01 -1.07681572e+00 3.40071589e-01 -8.14355791e-01 3.42144340e-01 1.24191061e-01 1.81737930e-01 -1.00020599e+00 -2.90771455e-01 -2.13989034e-01 7.11734518e-02 1.27197182e+00 4.21847433e-01 1.00885558e+00 -1.84375748e-01 -4.63204205e-01 8.72470438e-01 1.32455456e+00 -7.77548626e-02 6.35529816e-01 4.05055910e-01 1.07548606e+00 7.16742516e-01 5.71360648e-01 4.59991574e-01 5.65345943e-01 6.77186668e-01 3.30981880e-01 -2.19841525e-01 -4.13631618e-01 9.62999016e-02 4.75446224e-01 8.70691061e-01 8.82848203e-02 -2.95300663e-01 -5.70861697e-01 5.48645258e-01 -2.14821267e+00 -8.15813482e-01 -2.59076476e-01 1.72570908e+00 7.71220267e-01 -7.24201500e-02 -3.45020145e-02 -3.22257131e-01 9.38346922e-01 4.39393908e-01 -5.02445936e-01 -8.76062140e-02 -3.38161945e-01 -3.86190206e-01 6.89711571e-01 5.85841417e-01 -1.38652074e+00 1.04672599e+00 5.57672691e+00 8.38079631e-01 -1.01178527e+00 3.03080142e-01 6.43399477e-01 -2.22651601e-01 -3.52886826e-01 -1.61319077e-02 -8.86224508e-01 6.58607721e-01 3.29649538e-01 1.44020244e-01 2.77246028e-01 3.56773227e-01 1.25799209e-01 -1.26747102e-01 -8.71706486e-01 9.17361379e-01 4.19790596e-02 -1.42820668e+00 -2.99164299e-02 -7.60821849e-02 6.66125834e-01 2.66842991e-01 1.40498430e-01 4.89942357e-02 1.12821460e-01 -5.66054702e-01 7.26310790e-01 4.91845697e-01 3.48127306e-01 -5.49792826e-01 6.40707195e-01 -7.93122277e-02 -1.35555208e+00 -6.34137094e-02 -1.08351164e-01 4.12233204e-01 4.05614793e-01 5.13303936e-01 -2.26650387e-01 6.93021536e-01 8.77677977e-01 9.87203777e-01 -5.08401990e-01 1.27415669e+00 1.27990395e-01 3.52883875e-01 -4.42471772e-01 2.94609785e-01 4.86190230e-01 -3.79869759e-01 8.33093882e-01 1.48736167e+00 1.17014386e-01 2.22298093e-02 6.84825704e-02 9.90504265e-01 -6.03146516e-02 -2.77974680e-02 -2.73466796e-01 7.48345852e-02 1.64189786e-01 1.35784924e+00 -7.77155578e-01 -1.78939819e-01 -7.67742574e-01 1.21167886e+00 2.51349986e-01 6.40124559e-01 -1.23377919e+00 -7.04816729e-02 1.00321615e+00 -3.22411895e-01 6.36672616e-01 -3.25401723e-01 3.59574705e-02 -1.21343076e+00 8.75012949e-02 -5.45333505e-01 2.41500273e-01 -5.11958897e-01 -1.21547735e+00 3.56355160e-01 -7.12709650e-02 -1.09399331e+00 3.91675025e-01 -3.98337454e-01 -4.28230643e-01 5.74012578e-01 -1.87706125e+00 -1.18947279e+00 -1.88979402e-01 5.94627738e-01 5.57113290e-01 2.69654095e-01 9.58291590e-02 7.42514908e-01 -9.59612489e-01 5.31364739e-01 -3.89116481e-02 7.92653561e-02 6.93391085e-01 -8.58234882e-01 2.29294121e-01 1.03640461e+00 2.36776005e-02 3.47234815e-01 6.58119917e-01 -5.93712270e-01 -1.61456978e+00 -1.27023983e+00 4.73616660e-01 -8.02808180e-02 7.48029411e-01 -4.93642449e-01 -1.09044087e+00 4.81895417e-01 3.14697325e-01 1.89918384e-01 4.41066921e-01 -2.64585793e-01 -2.92633682e-01 -5.27990051e-02 -8.29620659e-01 7.14352787e-01 1.44791293e+00 -3.97832423e-01 -2.42610008e-01 3.68087925e-02 7.89962828e-01 -2.75258124e-01 -6.47933125e-01 6.43647552e-01 4.55185562e-01 -8.46729100e-01 9.86715496e-01 -4.80679691e-01 3.39956880e-01 -5.96252859e-01 -1.39021620e-01 -8.78785014e-01 -6.14534080e-01 -8.30320656e-01 -2.54193902e-01 1.51135695e+00 3.11031371e-01 -5.31378925e-01 6.72611296e-01 6.15720570e-01 5.04189683e-03 -7.78919995e-01 -1.15630221e+00 -5.24508119e-01 -3.59348804e-01 -3.99532557e-01 3.81358534e-01 9.20324683e-01 -2.98174560e-01 1.48193941e-01 -6.61051214e-01 3.85281622e-01 5.74898779e-01 7.21284226e-02 6.10148549e-01 -7.63808072e-01 -2.55259454e-01 -6.33401573e-01 -5.27531981e-01 -1.52865911e+00 2.62391657e-01 -8.38467836e-01 6.96256012e-02 -1.34354842e+00 3.87502193e-01 -3.06249976e-01 -4.90218669e-01 1.96167246e-01 -2.34865978e-01 4.09619898e-01 4.09295201e-01 1.68164954e-01 -1.00846648e+00 8.36722612e-01 1.34386885e+00 -2.23245889e-01 -1.16453163e-01 -3.68160903e-01 -4.97506291e-01 7.10632026e-01 6.54383659e-01 -4.27800447e-01 -2.29899704e-01 -6.76347554e-01 -4.50083911e-02 -3.97635043e-01 5.17354429e-01 -7.13741601e-01 5.24345517e-01 7.11404858e-03 3.35891247e-01 -6.08003736e-01 5.17212570e-01 -9.34407890e-01 9.70858410e-02 1.97415009e-01 -1.39869213e-01 -9.04090628e-02 3.59910578e-01 7.95063615e-01 -1.52953893e-01 -1.67572200e-02 7.83713877e-01 3.44193250e-01 -1.00697362e+00 4.58566278e-01 -3.14264089e-01 -1.72401175e-01 1.21024036e+00 -3.23219329e-01 -4.76783782e-01 -1.11698315e-01 -6.90483868e-01 6.51098430e-01 3.67967308e-01 8.08330834e-01 7.06343710e-01 -1.36173630e+00 -6.11144304e-01 3.22133273e-01 -5.39869033e-02 -2.76636124e-01 5.07177651e-01 1.40856051e+00 -3.91879350e-01 2.92739779e-01 -1.04358226e-01 -7.55110502e-01 -1.40506542e+00 4.53944147e-01 3.26441467e-01 1.20438583e-01 -5.79596460e-01 1.12840652e+00 5.29559672e-01 -8.30040872e-02 4.30192947e-01 -2.45412022e-01 -2.50084192e-01 1.28723651e-01 4.04497534e-01 5.25879741e-01 -3.96075428e-01 -1.04335535e+00 -6.06896937e-01 7.27331102e-01 -2.86243469e-01 1.09242119e-01 1.15820479e+00 -6.22693062e-01 -2.63987094e-01 1.79024816e-01 1.50414979e+00 -2.78725177e-01 -1.63623405e+00 -3.43384266e-01 -8.10388010e-03 -6.46897197e-01 2.67670423e-01 -2.25941375e-01 -1.40826249e+00 5.17491460e-01 6.82195902e-01 1.81380332e-01 1.07689607e+00 1.25696242e-01 8.18224192e-01 -1.17880687e-01 -8.21592659e-02 -1.14747965e+00 -1.82754084e-01 3.74207407e-01 8.12359571e-01 -1.32951546e+00 -4.38991375e-02 -7.92820036e-01 -5.27557611e-01 8.43549311e-01 7.78688014e-01 -1.31266549e-01 9.07961190e-01 2.30962485e-01 -1.18211783e-01 3.20132203e-05 -7.81333864e-01 -2.91686177e-01 4.69946772e-01 1.84662968e-01 2.55640090e-01 -2.86988169e-01 -4.70847785e-01 3.33558470e-01 5.36870003e-01 -1.03164561e-01 2.24214211e-01 8.00417542e-01 -1.88119560e-01 -1.02573657e+00 -1.75628588e-01 2.36634523e-01 -4.70450610e-01 -1.02827251e-01 -3.04774076e-01 7.17827916e-01 4.22423065e-01 8.95523131e-01 1.53962761e-01 -4.05204296e-01 -9.77114309e-03 -2.16706142e-01 5.20705760e-01 -1.33812159e-01 -3.98120552e-01 2.89435714e-01 8.37634727e-02 -1.00104427e+00 -8.13669622e-01 -8.05622280e-01 -1.06863487e+00 -3.06282103e-01 -4.70405310e-01 -2.13715598e-01 5.70320010e-01 8.47935915e-01 5.93453169e-01 5.41194320e-01 5.70241690e-01 -9.84187961e-01 -1.06220856e-01 -6.77343309e-01 -5.13348401e-01 3.18074524e-01 5.46221852e-01 -7.31410027e-01 -4.57855076e-01 1.44530181e-02]
[9.275411605834961, -0.15825045108795166]
afd1a5b5-9868-4b76-a73c-52e6bc3e54e7
equiangular-basis-vectors
2303.11637
null
https://arxiv.org/abs/2303.11637v2
https://arxiv.org/pdf/2303.11637v2.pdf
Equiangular Basis Vectors
We propose Equiangular Basis Vectors (EBVs) for classification tasks. In deep neural networks, models usually end with a k-way fully connected layer with softmax to handle different classification tasks. The learning objective of these methods can be summarized as mapping the learned feature representations to the samples' label space. While in metric learning approaches, the main objective is to learn a transformation function that maps training data points from the original space to a new space where similar points are closer while dissimilar points become farther apart. Different from previous methods, our EBVs generate normalized vector embeddings as "predefined classifiers" which are required to not only be with the equal status between each other, but also be as orthogonal as possible. By minimizing the spherical distance of the embedding of an input between its categorical EBV in training, the predictions can be obtained by identifying the categorical EBV with the smallest distance during inference. Various experiments on the ImageNet-1K dataset and other downstream tasks demonstrate that our method outperforms the general fully connected classifier while it does not introduce huge additional computation compared with classical metric learning methods. Our EBVs won the first place in the 2022 DIGIX Global AI Challenge, and our code is open-source and available at https://github.com/NJUST-VIPGroup/Equiangular-Basis-Vectors.
['Xiu-Shen Wei', 'Xuhao Sun', 'Yang shen']
2023-03-21
null
http://openaccess.thecvf.com//content/CVPR2023/html/Shen_Equiangular_Basis_Vectors_CVPR_2023_paper.html
http://openaccess.thecvf.com//content/CVPR2023/papers/Shen_Equiangular_Basis_Vectors_CVPR_2023_paper.pdf
cvpr-2023-1
['metric-learning', 'metric-learning']
['computer-vision', 'methodology']
[-1.31238863e-01 1.07185863e-01 -4.32017028e-01 -8.39903891e-01 -3.00552368e-01 -4.68517244e-01 5.62337160e-01 -4.05747369e-02 -5.56225777e-01 6.98359132e-01 4.84368168e-02 -9.61357355e-02 -3.47595334e-01 -8.62079442e-01 -7.28452861e-01 -8.07488263e-01 3.36872190e-02 6.36508644e-01 -1.99351996e-01 1.36350915e-02 1.93674117e-01 4.34515953e-01 -1.41738427e+00 1.64995506e-01 5.18496037e-01 1.33920884e+00 6.21743724e-02 3.36587876e-01 1.59232959e-01 3.70568305e-01 -4.53108102e-01 -3.18121195e-01 5.14147520e-01 -1.30077258e-01 -8.16596568e-01 -3.40666980e-01 5.43635428e-01 -1.95755512e-01 -6.21127188e-01 1.23105752e+00 3.17782342e-01 3.25396895e-01 1.13524020e+00 -1.76355278e+00 -1.11935437e+00 5.24188280e-01 -3.13681185e-01 -5.32838218e-02 -1.31743580e-01 -2.21261621e-01 1.31916165e+00 -1.05294847e+00 5.07495284e-01 1.02608478e+00 5.70476890e-01 6.69459343e-01 -1.27932370e+00 -7.81085193e-01 1.04074934e-02 6.44819736e-01 -1.68403244e+00 -2.33246610e-01 7.44909823e-01 -5.72222471e-01 6.94696248e-01 3.37406576e-01 4.76185262e-01 1.04850066e+00 -1.11697754e-02 5.99791825e-01 6.37280405e-01 -1.89710885e-01 3.47083420e-01 2.18185395e-01 2.54895896e-01 6.66172743e-01 5.90666868e-02 2.49777585e-02 -4.64745492e-01 -5.50493747e-02 4.79655027e-01 1.92580596e-01 -3.36737424e-01 -7.58291185e-01 -1.49304676e+00 1.14758956e+00 8.60517561e-01 3.15896243e-01 -1.48742959e-01 2.72088766e-01 2.13552073e-01 3.81199479e-01 3.82346064e-01 4.73715872e-01 -3.97907495e-01 9.46606845e-02 -7.99215496e-01 1.35125816e-01 6.30825281e-01 9.47376013e-01 8.82157087e-01 -3.56800497e-01 -6.47439137e-02 9.17676091e-01 3.53904575e-01 1.37405694e-01 7.55661428e-01 -1.03344107e+00 3.83909434e-01 7.82964408e-01 -2.03528702e-01 -1.15865767e+00 -3.54092181e-01 -3.82546514e-01 -1.14475477e+00 1.78344116e-01 4.11457866e-01 5.29912906e-03 -7.62894988e-01 1.84699655e+00 2.28796333e-01 3.71364594e-01 -8.18452090e-02 9.72975135e-01 8.20760965e-01 5.80912173e-01 -3.86494070e-01 3.06977719e-01 1.07361341e+00 -8.75202060e-01 -3.95508170e-01 -3.46634090e-02 8.57806027e-01 -3.38297248e-01 1.07415009e+00 3.96370143e-01 -6.42368913e-01 -5.16879022e-01 -1.42314470e+00 -2.93650895e-01 -7.31182337e-01 2.04667449e-01 4.91029859e-01 3.86399865e-01 -1.16499424e+00 7.18831182e-01 -7.32231140e-01 -1.64426386e-01 5.72655976e-01 3.66167963e-01 -6.80325627e-01 -5.98004051e-02 -1.27731562e+00 1.09634233e+00 4.90814209e-01 1.24275774e-01 -7.60377884e-01 -8.16830933e-01 -9.79934990e-01 8.54240544e-03 -8.32936317e-02 -4.40506876e-01 7.76467860e-01 -7.13111162e-01 -1.15108693e+00 9.58686829e-01 2.95726974e-02 -4.07108516e-01 4.21208799e-01 9.91171375e-02 -1.36452034e-01 -1.99514508e-01 6.22346103e-02 1.09672177e+00 5.07007897e-01 -8.50921094e-01 -4.78072673e-01 -4.17698085e-01 -1.20689929e-01 2.91323394e-01 -9.44312096e-01 -3.01240027e-01 -2.07676679e-01 -4.40341949e-01 2.59645164e-01 -8.66626024e-01 -1.30933523e-01 4.76925671e-01 -4.74648684e-01 -6.58577859e-01 8.66693377e-01 -3.55221838e-01 9.69832420e-01 -2.15369368e+00 4.68987882e-01 3.70247036e-01 4.38150495e-01 3.39673981e-02 -1.99018180e-01 5.42420074e-02 -1.71169490e-01 -1.86673608e-02 -3.72940570e-01 -3.10490012e-01 2.45997518e-01 3.71797174e-01 -1.74715325e-01 8.24374676e-01 3.30755636e-02 8.85941386e-01 -9.81723487e-01 -2.85747200e-01 2.65744776e-01 5.51272452e-01 -5.19645452e-01 8.85293186e-02 2.00063512e-01 1.28438100e-01 -2.03743447e-02 3.96174669e-01 5.76927304e-01 -1.98511079e-01 -1.09761059e-01 -3.58987898e-01 2.00145543e-01 2.12746114e-01 -1.08911741e+00 1.90271628e+00 -2.33649611e-01 9.12989438e-01 -4.12348539e-01 -1.59206223e+00 1.24432397e+00 9.60578248e-02 4.73896325e-01 -4.97362882e-01 1.98511139e-01 6.58714920e-02 7.42668360e-02 -1.90945089e-01 1.43554091e-01 8.86976495e-02 -1.45009369e-01 3.48877341e-01 3.01584780e-01 5.09944977e-03 3.40762958e-02 5.13329841e-02 7.77148068e-01 -1.05251960e-01 1.85792059e-01 -3.50298733e-01 4.39534754e-01 -2.72526175e-01 7.78622448e-01 3.05628926e-01 -3.10461551e-01 7.11290598e-01 6.44820631e-01 -5.55265844e-01 -8.60942066e-01 -1.35002244e+00 -5.50646663e-01 8.50865901e-01 2.29235590e-01 -2.11490735e-01 -6.79760516e-01 -9.05873060e-01 2.58786440e-01 8.15642357e-01 -1.02810466e+00 -5.30332685e-01 -2.28073999e-01 -4.76283342e-01 7.02690303e-01 6.23860836e-01 2.90412635e-01 -9.25161719e-01 -2.19303265e-01 -1.18878342e-01 -1.17561825e-01 -7.33204544e-01 -6.40583813e-01 4.60627764e-01 -5.92849016e-01 -1.12170768e+00 -6.60820007e-01 -1.06695807e+00 1.03530049e+00 2.77703186e-03 7.39205599e-01 -2.14708462e-01 -3.18289101e-01 -8.74968842e-02 -1.59199029e-01 -2.85787672e-01 7.28736669e-02 6.06953092e-02 3.57307732e-01 2.51520067e-01 8.63649487e-01 -5.77982664e-01 -7.29662359e-01 5.62924445e-01 -6.95974886e-01 -3.40594612e-02 1.11439593e-01 1.02512825e+00 7.70212770e-01 -2.68422753e-01 6.47879124e-01 -5.10587633e-01 4.90167528e-01 -8.42090070e-01 -4.21581835e-01 3.81757110e-01 -6.84012890e-01 1.23240873e-01 7.37426698e-01 -6.01506174e-01 -3.07968974e-01 -3.89778279e-02 1.41391039e-01 -6.94671750e-01 -1.55651331e-01 4.16046679e-01 -3.25505406e-01 6.27716035e-02 7.08231628e-01 1.06426582e-01 1.36986986e-01 -3.85621041e-01 6.30212903e-01 8.67146969e-01 5.50352216e-01 -5.58834612e-01 6.99362099e-01 4.14498299e-01 -5.11970259e-02 -5.66632569e-01 -7.28726387e-01 -3.36379558e-01 -1.04726410e+00 -1.27922982e-01 9.27455008e-01 -5.57237625e-01 -6.26046062e-01 3.31571400e-01 -8.84552777e-01 -3.35639715e-01 -3.46349746e-01 7.00207114e-01 -5.53551376e-01 7.29164332e-02 -2.67469525e-01 -2.40102317e-02 -2.64888048e-01 -1.04512596e+00 5.51122785e-01 1.58743992e-01 -3.94694716e-01 -1.18953395e+00 1.74150363e-01 1.86154127e-01 3.62544239e-01 1.26742542e-01 9.80381787e-01 -9.91476357e-01 -2.93753538e-02 -4.19983745e-01 -3.34562868e-01 8.67964983e-01 3.46354067e-01 -9.20762792e-02 -8.29207659e-01 -3.15016627e-01 -7.53865764e-02 -4.28420275e-01 9.01401222e-01 3.50057513e-01 1.62255871e+00 -2.84936547e-01 -3.70765746e-01 1.16447246e+00 1.15751374e+00 1.51272327e-01 4.42209750e-01 3.52081031e-01 8.02640617e-01 5.21532118e-01 3.94426525e-01 1.40893012e-01 5.15138507e-01 6.16557419e-01 6.55135810e-01 -3.84592116e-02 -1.30491033e-02 -1.87168211e-01 1.68348029e-01 7.78523803e-01 2.11592659e-01 -3.41101252e-02 -8.58795226e-01 6.48818433e-01 -1.87183046e+00 -8.01476419e-01 1.15278758e-01 2.32895923e+00 8.27059269e-01 -5.83317988e-02 -2.46865585e-01 1.36520058e-01 6.65652573e-01 2.81484723e-02 -8.75010669e-01 -4.70935553e-01 -9.07454044e-02 1.30013883e-01 4.00700629e-01 4.75869507e-01 -1.14935577e+00 6.99666262e-01 5.24700499e+00 6.68003917e-01 -1.18799186e+00 1.06540821e-01 7.52552688e-01 -5.58850348e-01 -1.89387664e-01 -1.37290969e-01 -6.57242060e-01 5.91317713e-01 8.15743506e-01 -3.67454976e-01 4.52790797e-01 8.64781797e-01 -1.81867853e-01 3.63176554e-01 -1.68572259e+00 1.13543701e+00 1.12766519e-01 -1.32013404e+00 3.58526073e-02 1.00993112e-01 6.95779800e-01 3.26869488e-01 3.42607886e-01 3.26279581e-01 1.65083691e-01 -1.41071463e+00 5.62378824e-01 5.85682213e-01 1.09524703e+00 -8.23583484e-01 5.86051226e-01 3.22120696e-01 -1.03350806e+00 8.32993835e-02 -8.33268523e-01 -6.31660447e-02 -1.97186276e-01 4.75841463e-01 -7.51214147e-01 2.18637139e-01 6.57937467e-01 9.90210056e-01 -3.94407213e-01 9.72622454e-01 -2.29530811e-01 3.35014611e-01 -2.58803189e-01 -4.54221144e-02 3.06544960e-01 -5.07511497e-01 1.94583267e-01 1.00621235e+00 4.62325871e-01 -1.97034746e-01 -1.42908730e-02 9.52393055e-01 -4.76077437e-01 2.19928678e-02 -8.12932074e-01 1.99881434e-01 6.23647451e-01 1.24575424e+00 -4.31573123e-01 -2.08163008e-01 -3.88751209e-01 1.01704884e+00 6.70226514e-01 2.66239285e-01 -8.55957448e-01 -7.88997889e-01 1.16293132e+00 -1.53362513e-01 1.53844923e-01 -4.96755056e-02 -3.74413818e-01 -1.02050602e+00 6.02023005e-02 -4.53297585e-01 4.07889098e-01 -5.41637480e-01 -1.42587674e+00 5.71293712e-01 -7.20785782e-02 -1.29531038e+00 -1.03966177e-01 -9.46567893e-01 -5.99688947e-01 9.32693779e-01 -1.30940652e+00 -9.28520620e-01 -2.05946267e-01 8.22138131e-01 2.07751662e-01 -2.84491450e-01 1.04013240e+00 4.73041505e-01 -4.00119901e-01 1.06281233e+00 5.17156482e-01 4.94398773e-01 8.17353427e-01 -1.32289314e+00 2.69311816e-01 4.01876807e-01 3.86844605e-01 5.89310527e-01 2.01094136e-01 -2.19802991e-01 -9.22216237e-01 -1.20327652e+00 9.69130754e-01 -4.86244351e-01 6.82411134e-01 -6.24329150e-01 -9.63799417e-01 7.77638495e-01 -6.82339743e-02 4.80736136e-01 9.14481699e-01 1.32554114e-01 -7.10581303e-01 -2.99001217e-01 -1.16909242e+00 4.59084362e-01 1.09750640e+00 -5.32036841e-01 -6.06616497e-01 4.30704147e-01 5.09025991e-01 -1.02623373e-01 -1.10684049e+00 3.72901112e-01 6.27883375e-01 -6.94140136e-01 9.95497942e-01 -9.04366314e-01 4.50212568e-01 -2.96288878e-01 -5.15927196e-01 -1.64430630e+00 -4.21435267e-01 2.30047293e-02 -1.45628765e-01 9.73740160e-01 4.11183536e-01 -8.22911859e-01 8.42584729e-01 3.93329620e-01 -1.49345860e-01 -1.23660970e+00 -1.27421486e+00 -9.13608730e-01 3.48730952e-01 -3.40902954e-01 5.86226463e-01 1.36431730e+00 1.41607493e-01 2.65733719e-01 -2.11888060e-01 9.74952355e-02 7.58536160e-01 -7.36212805e-02 3.41969222e-01 -1.44318140e+00 3.44150178e-02 -7.10703075e-01 -1.02766287e+00 -8.73067617e-01 4.21737283e-01 -1.69797230e+00 -1.47962704e-01 -1.51436496e+00 1.64318696e-01 -5.35972059e-01 -7.95099318e-01 7.55584657e-01 8.97079408e-02 4.34082776e-01 -5.24198003e-02 1.05828285e-01 -4.12405312e-01 7.45916486e-01 8.65640461e-01 -3.88878673e-01 -2.18835417e-02 -8.87640193e-02 -7.74238050e-01 7.31496334e-01 8.99896204e-01 -4.53024596e-01 -5.40141642e-01 -6.60633922e-01 4.36520483e-03 -4.67058569e-01 3.18502903e-01 -8.09003830e-01 4.03268963e-01 -1.56568602e-01 6.15823627e-01 -4.84345257e-01 4.48883951e-01 -7.71161497e-01 -3.82503420e-02 3.64752859e-01 -6.25081480e-01 -5.64923994e-02 -2.10894585e-01 2.81427890e-01 -1.76200673e-01 -2.73965657e-01 7.54236758e-01 1.84137151e-01 -5.05698383e-01 6.58685386e-01 7.31705800e-02 6.92002401e-02 1.25137520e+00 -2.58838832e-01 -3.65128011e-01 -9.73911956e-02 -6.58916652e-01 4.70762193e-01 3.51615846e-01 7.24189043e-01 8.65319192e-01 -1.73856997e+00 -8.56156111e-01 3.26443613e-01 3.01399291e-01 1.58561260e-01 -3.08045391e-02 6.25092208e-01 -3.52000535e-01 4.52700794e-01 -4.24550027e-01 -6.47648931e-01 -1.02588296e+00 5.54324806e-01 5.06685853e-01 1.32560700e-01 -5.13266087e-01 1.15843022e+00 3.36833030e-01 -1.02535677e+00 5.94862282e-01 -2.81065702e-01 -1.76938519e-01 2.68833071e-01 5.28598607e-01 4.14243698e-01 1.36275098e-01 -6.14880323e-01 -4.74728525e-01 3.17704827e-01 -2.45157555e-01 1.25349775e-01 1.48928940e+00 2.41714284e-01 -1.78201094e-01 6.15690470e-01 1.93027854e+00 -6.81210577e-01 -1.22042155e+00 -5.00062585e-01 -7.00424835e-02 -4.25844043e-01 5.53111061e-02 -4.73309219e-01 -1.24933648e+00 1.06248367e+00 7.47342408e-01 -2.18254700e-02 7.80641615e-01 2.12532938e-01 5.19861341e-01 5.19647121e-01 8.03686753e-02 -1.05318892e+00 3.74102145e-02 3.83403271e-01 1.08561814e+00 -1.24167919e+00 -1.96466699e-01 2.64842123e-01 -4.98647541e-01 1.06450808e+00 7.68936992e-01 -2.97593832e-01 9.55290616e-01 -6.18794374e-03 3.29146422e-02 -1.94063276e-01 -7.27289379e-01 2.70935833e-01 5.93893707e-01 7.01347768e-01 3.45936984e-01 3.23177457e-01 -1.55851737e-01 6.14034295e-01 -5.67178369e-01 -3.00126404e-01 1.89539328e-01 4.77106988e-01 -3.11816782e-01 -9.76591527e-01 -5.72801679e-02 7.84226418e-01 5.50612994e-02 -1.07546188e-01 -3.37453187e-01 5.96718848e-01 2.95392990e-01 7.48636067e-01 5.04659951e-01 -6.03532732e-01 1.95030659e-01 8.65749717e-02 4.08412784e-01 -6.00259483e-01 -7.11771995e-02 -5.91243982e-01 -4.24682438e-01 -5.84424913e-01 6.67643845e-02 -6.30159736e-01 -1.34447849e+00 -3.07421029e-01 -1.73309550e-01 9.26823467e-02 9.01692271e-01 7.60469556e-01 2.13094622e-01 3.41856778e-01 8.00047815e-01 -7.77030349e-01 -8.04909766e-01 -9.76968348e-01 -5.43133259e-01 4.29921210e-01 2.55964279e-01 -7.37053752e-01 -5.22695541e-01 -9.36226919e-02]
[9.451921463012695, 3.074852228164673]
2a841b38-d3a2-46bf-98f4-a7b234003780
dan-deep-attention-neural-network-for-news
null
null
https://ojs.aaai.org/index.php/AAAI/article/view/4549
https://ojs.aaai.org/index.php/AAAI/article/view/4549/4427
Dan: Deep attention neural network for news recommendation
With the rapid information explosion of news, making personalized news recommendation for users becomes an increasingly challenging problem. Many existing recommendation methods that regard the recommendation procedure as the static process, have achieved better recommendation performance. However, they usually fail with the dynamic diversity of news and user’s interests, or ignore the importance of sequential information of user’s clicking selection. In this paper, taking full advantages of convolution neural network (CNN), recurrent neural network (RNN) and attention mechanism, we propose a deep attention neural network DAN for news recommendation. Our DAN model presents to use attention-based parallel CNN for aggregating user’s interest features and attention-based RNN for capturing richer hidden sequential features of user’s clicks, and combines these features for new recommendation. We conduct experiment on real-world news data sets, and the experimental results demonstrate the superiority and effectiveness of our proposed DAN model.
['Qiannan Zhu']
2019-07-17
null
null
null
aaai-2019-7
['deep-attention', 'deep-attention']
['computer-vision', 'natural-language-processing']
[-3.47320676e-01 -5.44076979e-01 -4.27997023e-01 -5.13380408e-01 -8.12647119e-02 -8.65601227e-02 5.64504385e-01 -3.35786730e-01 -3.38496596e-01 4.93647128e-01 1.05712438e+00 -3.68216008e-01 -2.39524513e-01 -8.99975061e-01 -4.72596884e-01 -3.39856386e-01 3.82043235e-02 1.58095852e-01 2.52098590e-01 -4.46531832e-01 3.60082209e-01 6.79160133e-02 -1.48090458e+00 5.07385731e-01 6.45727992e-01 1.23565042e+00 3.18456739e-01 4.91255790e-01 -3.24807405e-01 9.31865335e-01 -3.62776369e-01 -1.89098641e-01 1.29299283e-01 -3.47530246e-01 -6.20846570e-01 -4.04847592e-01 -7.04376772e-02 -8.42728972e-01 -9.38297510e-01 7.26187229e-01 6.47525132e-01 7.77823985e-01 1.95602328e-01 -5.87180853e-01 -1.52747440e+00 1.21677268e+00 -2.94432521e-01 7.99046576e-01 2.10231096e-01 -3.23353171e-01 1.27359593e+00 -7.20997334e-01 2.80891299e-01 9.21442807e-01 5.63997269e-01 5.97391129e-01 -5.33728778e-01 -4.99137640e-01 8.45419705e-01 2.99149185e-01 -6.92513108e-01 -8.36427361e-02 6.53004050e-01 -1.94262043e-01 1.00277209e+00 3.06094468e-01 8.90841782e-01 1.44297540e+00 1.59076408e-01 1.26114476e+00 4.36888665e-01 2.38450542e-01 -1.52329192e-01 -1.35048136e-01 6.35901392e-01 1.18397269e-02 1.87664494e-01 8.48962590e-02 -1.17997542e-01 -3.05794198e-02 8.92953515e-01 1.10651326e+00 -3.63346756e-01 3.63898307e-01 -1.16739261e+00 7.74035752e-01 7.10791409e-01 3.53438586e-01 -6.22158647e-01 2.75747925e-02 4.32406127e-01 4.74244863e-01 7.96806276e-01 4.11095798e-01 -6.62588120e-01 -1.17702030e-01 -7.16713190e-01 2.74993032e-01 7.18167484e-01 8.01611185e-01 4.17907983e-02 3.24910432e-01 -4.97766674e-01 8.59065771e-01 2.73693532e-01 2.51832187e-01 1.16116869e+00 -1.33697301e-01 2.05780402e-01 4.41570789e-01 1.70269012e-01 -1.06638718e+00 -5.18400669e-01 -1.19030797e+00 -9.69354391e-01 -5.88489890e-01 -1.88251987e-01 -5.30390561e-01 -7.98002183e-01 1.30926514e+00 -2.08059642e-02 3.92727822e-01 -4.55908338e-03 1.08761120e+00 1.23422253e+00 1.10569477e+00 -5.33672683e-02 -3.41464341e-01 1.09552181e+00 -1.42972863e+00 -7.62161076e-01 1.62160590e-01 2.50761747e-01 -5.39857447e-01 1.06366956e+00 3.29954118e-01 -9.80181754e-01 -8.36667418e-01 -7.53253460e-01 -2.75689036e-01 -2.57168084e-01 3.20642114e-01 9.33211148e-01 4.30748202e-02 -5.34141541e-01 6.82110846e-01 -5.04356205e-01 -2.46228322e-01 3.49960566e-01 6.37484372e-01 3.21539611e-01 1.60728782e-01 -1.44477808e+00 3.27559173e-01 5.85526377e-02 5.11613727e-01 -6.20165348e-01 -5.04308522e-01 -1.39809236e-01 6.06472313e-01 3.54298145e-01 -6.80177391e-01 1.52384984e+00 -1.08976531e+00 -1.74966502e+00 -2.93359786e-01 8.60701874e-02 -3.79272044e-01 1.76547095e-01 -7.47646391e-01 -9.45671916e-01 -7.20765889e-01 -4.08348739e-01 -1.77473694e-01 4.55193818e-01 -7.68053949e-01 -1.17422259e+00 -1.69050202e-01 2.51106024e-01 2.06728190e-01 -5.80782712e-01 1.44741580e-01 -3.54835540e-01 -8.71387243e-01 -1.36265352e-01 -5.52059948e-01 -4.08141464e-01 -8.01031470e-01 -2.41738573e-01 -3.54768097e-01 7.65864015e-01 -7.39315927e-01 1.57437861e+00 -2.13346934e+00 2.99034826e-02 1.25409111e-01 2.43466020e-01 5.62115192e-01 -2.50157386e-01 3.79625827e-01 2.81202346e-01 4.41117398e-03 6.73502445e-01 3.46533023e-02 -1.29305527e-01 5.39512672e-02 -7.31205642e-01 1.45037500e-02 -5.32935560e-01 1.08448410e+00 -9.13768768e-01 2.21612930e-01 -1.15527011e-01 5.01129210e-01 -7.82692134e-01 3.74692798e-01 -4.53580230e-01 4.02353942e-01 -1.16728222e+00 8.75593275e-02 3.86075139e-01 -5.48725605e-01 -2.25097179e-01 -1.58108398e-01 -2.38454297e-01 8.01110327e-01 -8.65042090e-01 1.21862745e+00 -5.09108245e-01 2.66548812e-01 -3.20526898e-01 -6.97457671e-01 8.99139643e-01 3.59378994e-01 1.54274076e-01 -9.82815623e-01 4.14198488e-01 -1.80645883e-01 3.19417983e-01 -6.11767888e-01 7.41508543e-01 2.97708690e-01 2.80438453e-01 1.08511055e+00 -1.85615476e-02 1.04925931e+00 -2.23405167e-01 1.39231801e-01 8.45529616e-01 9.94623825e-02 1.04836049e-02 -2.80072931e-02 3.59520674e-01 -5.40251434e-01 5.51301003e-01 7.26379037e-01 2.84343004e-01 4.25713003e-01 2.19382748e-01 -7.37340331e-01 -9.86535370e-01 -3.12157750e-01 2.11528108e-01 1.82698703e+00 5.71843944e-02 -2.79236585e-01 -2.06476644e-01 -7.24523127e-01 -2.58050054e-01 6.97959125e-01 -8.51241231e-01 -2.63654828e-01 -5.07685840e-01 -8.84233296e-01 -1.22135088e-01 7.85675287e-01 5.93831658e-01 -1.26599932e+00 1.21108152e-01 5.32356620e-01 -1.07765384e-01 -4.19828653e-01 -9.17114735e-01 -2.74617314e-01 -8.38722587e-01 -5.61928689e-01 -1.01185954e+00 -8.04672599e-01 5.08304954e-01 7.44818151e-01 8.32469761e-01 3.25742930e-01 3.42314422e-01 1.34859294e-01 -8.85027766e-01 -7.10206777e-02 4.10529345e-01 5.45315623e-01 4.81342897e-02 3.01154137e-01 3.14245969e-01 -7.09487736e-01 -1.11089540e+00 4.24895406e-01 -8.27016413e-01 1.37836218e-01 6.48078144e-01 8.92577529e-01 3.08387339e-01 -1.75986826e-01 8.43593538e-01 -1.36530602e+00 1.21366167e+00 -1.10550308e+00 -1.63160935e-01 2.20353216e-01 -7.87872076e-01 -2.11580366e-01 1.08948302e+00 -6.44621074e-01 -1.44838130e+00 -6.51427150e-01 -5.33197641e-01 4.30096872e-02 1.35003567e-01 1.04123843e+00 7.68765807e-02 3.64271820e-01 3.73950452e-01 4.36877787e-01 -5.31747401e-01 -1.10207224e+00 3.92746478e-01 9.25112247e-01 2.47230917e-01 -2.13601440e-01 1.67533696e-01 1.64250880e-01 -8.94717336e-01 -1.85951278e-01 -1.22353947e+00 -4.74640131e-01 -7.24002570e-02 -5.96562997e-02 4.98677820e-01 -7.33864307e-01 -8.10273111e-01 3.32373440e-01 -9.63989615e-01 -4.05693352e-02 -1.53332502e-01 1.06447589e+00 2.20640693e-02 1.16417140e-01 -1.10774112e+00 -5.67444861e-01 -8.34566772e-01 -8.55449617e-01 3.28531176e-01 5.99547863e-01 2.00406715e-01 -9.53435481e-01 1.01190194e-01 8.97908732e-02 1.05321550e+00 -3.96555454e-01 5.30886173e-01 -1.22130597e+00 -4.90937889e-01 -5.93479097e-01 -4.60733742e-01 2.04314753e-01 1.61549285e-01 -1.03264309e-01 -6.82757556e-01 -8.92323703e-02 -5.96107394e-02 1.21917270e-01 1.01634419e+00 4.32423055e-01 1.52721786e+00 -7.94037640e-01 -1.74537539e-01 6.78644001e-01 9.89610493e-01 4.60113347e-01 6.38155520e-01 2.86557734e-01 8.67553711e-01 7.60038123e-02 3.49780619e-01 5.94329655e-01 4.74200219e-01 4.65989739e-01 3.07837725e-01 -5.91944940e-02 9.87507328e-02 -4.49422926e-01 2.50448465e-01 1.42025459e+00 -8.44137192e-01 -6.55408919e-01 -1.05277494e-01 2.53078431e-01 -2.17793584e+00 -1.36389661e+00 5.00829443e-02 1.93846035e+00 3.39869499e-01 2.13620231e-01 2.40081117e-01 -3.86632174e-01 7.82548130e-01 1.74320444e-01 -7.43813038e-01 -3.09606999e-01 6.64787069e-02 -1.46890491e-01 3.63778830e-01 1.81202501e-01 -9.27240133e-01 7.75491297e-01 5.86843920e+00 7.58561492e-01 -1.38017106e+00 1.55845970e-01 6.27064407e-01 -4.40186888e-01 -6.69819534e-01 -4.06739116e-01 -8.32767546e-01 7.96596169e-01 7.81306624e-01 -3.54272425e-01 7.20512807e-01 1.01384509e+00 3.27694327e-01 5.78279614e-01 -7.02244878e-01 7.08313465e-01 1.10385001e-01 -1.47798622e+00 3.45305264e-01 -8.19100626e-03 1.00706756e+00 4.48823750e-01 3.71006310e-01 6.30939722e-01 6.95674419e-01 -7.80840993e-01 3.36529136e-01 9.35927987e-01 1.28373027e-01 -7.89336324e-01 1.14904618e+00 3.10414732e-01 -9.80476558e-01 -3.55048180e-01 -6.13097370e-01 -5.19788802e-01 2.21090376e-01 5.66270769e-01 -2.68560976e-01 4.66674834e-01 7.28612006e-01 1.35955858e+00 -2.46130094e-01 1.17336619e+00 -5.60348853e-02 1.09005690e+00 -3.52460407e-02 -6.35730386e-01 4.24839199e-01 -3.49492490e-01 3.36539567e-01 1.17034280e+00 5.36454976e-01 5.04292309e-01 3.65200043e-02 3.40805620e-01 -3.09600860e-01 4.63461250e-01 -5.02705835e-02 -1.77459359e-01 6.33879453e-02 1.39979386e+00 -5.85735261e-01 -4.04282272e-01 -6.30371571e-01 8.10012817e-01 2.96313941e-01 5.87926805e-01 -8.86553884e-01 -4.67812508e-01 4.31104481e-01 8.88631418e-02 7.15306818e-01 1.84647664e-02 5.14533632e-02 -1.37214971e+00 -1.06162354e-01 -6.35448933e-01 4.14511144e-01 -7.75808394e-01 -1.51835895e+00 8.60759676e-01 -6.12609506e-01 -1.14453208e+00 1.42082259e-01 -1.89724967e-01 -1.17223561e+00 6.43677890e-01 -1.22033477e+00 -1.18478048e+00 -1.51256800e-01 6.03808045e-01 6.64410233e-01 -3.83109123e-01 4.90687072e-01 8.41062605e-01 -7.82389760e-01 7.04686701e-01 7.88578391e-01 1.12006791e-01 4.73914146e-01 -8.88788164e-01 6.25077307e-01 4.07407939e-01 -8.43859240e-02 1.05505359e+00 4.12293613e-01 -5.63012481e-01 -1.35224330e+00 -1.22438562e+00 7.32995570e-01 -1.61246538e-01 6.34620488e-01 -2.93770079e-02 -8.06728184e-01 1.24939823e+00 1.63943484e-01 -1.85800165e-01 9.94089186e-01 6.89374983e-01 -2.30416819e-01 -2.24585533e-01 -5.58394074e-01 6.98792398e-01 1.20947480e+00 -1.37496144e-01 -7.53295958e-01 3.45101476e-01 1.07801986e+00 -2.54154801e-01 -6.92771375e-01 1.80977821e-01 7.54356384e-01 -5.92144251e-01 8.57571661e-01 -1.29901540e+00 6.39052987e-01 -1.26030564e-01 1.50282577e-01 -1.54410255e+00 -1.16012621e+00 -4.07685667e-01 -2.19044730e-01 1.11592281e+00 5.54771721e-01 -7.75312603e-01 4.83444631e-01 4.48030651e-01 -5.20390987e-01 -9.09894347e-01 -3.46862823e-01 -1.15888260e-01 -4.90603060e-01 -1.48539767e-01 1.05305231e+00 9.58959997e-01 1.65163726e-01 6.95951879e-01 -1.21342885e+00 -2.31948942e-01 -3.44342381e-01 4.47478503e-01 5.92379749e-01 -1.29480433e+00 -5.67726672e-01 -5.80164909e-01 -2.59099882e-02 -1.72380447e+00 -2.81277299e-01 -7.14305103e-01 -1.91152662e-01 -1.88138211e+00 3.58398438e-01 -2.89779156e-01 -9.48038995e-01 2.70228297e-01 -1.44538686e-01 -7.03802854e-02 -1.13724865e-01 3.74318838e-01 -1.17225003e+00 6.87530875e-01 1.56598485e+00 4.19717208e-02 -2.50716537e-01 5.45376122e-01 -1.06855309e+00 6.10571086e-01 6.27061903e-01 -2.47663870e-01 -4.34280306e-01 -7.71758378e-01 7.39781320e-01 -1.16003431e-01 -1.77692726e-01 -7.07827866e-01 4.65598315e-01 -1.91148371e-01 5.42995930e-01 -5.94182491e-01 -4.22306433e-02 -6.61584258e-01 5.11318482e-02 2.33012289e-01 -8.35080028e-01 -2.10380908e-02 -2.04853058e-01 8.94977391e-01 -1.21306017e-01 1.25063896e-01 1.31640390e-01 -7.50289038e-02 -6.81588054e-01 1.05717564e+00 -4.05262709e-01 -2.26230204e-01 5.07088482e-01 1.28762037e-01 -4.22119766e-01 -6.42466426e-01 -9.31942463e-01 3.78295332e-01 -1.65319607e-01 7.66623616e-01 5.74888587e-01 -1.40938067e+00 -6.68525577e-01 6.70737624e-02 -1.87759757e-01 -2.44732589e-01 7.04153895e-01 5.59566557e-01 -2.88572729e-01 3.32380742e-01 -2.48292193e-01 3.47994179e-01 -8.69474649e-01 7.97951221e-01 1.99124560e-01 -2.12880820e-01 -7.26448417e-01 1.00624192e+00 2.25117326e-01 -2.94305533e-01 2.98998237e-01 -5.21767139e-01 -9.45767224e-01 1.70657164e-04 9.55719471e-01 4.31908190e-01 -2.11289123e-01 -2.86159694e-01 3.59602481e-01 3.36077034e-01 -7.70957947e-01 3.72470766e-01 1.66052377e+00 -3.60048383e-01 2.24481486e-02 4.30034846e-01 9.25345719e-01 7.62586445e-02 -8.24483454e-01 -6.07790649e-01 -4.37374026e-01 -4.49596703e-01 4.58568603e-01 -7.91684628e-01 -1.56540680e+00 6.56844497e-01 2.16049150e-01 5.78920424e-01 7.31761277e-01 -2.03056082e-01 1.65210545e+00 6.32734895e-01 -8.19548368e-02 -1.08979475e+00 -1.64125517e-01 8.42391133e-01 7.71286249e-01 -1.07666552e+00 -3.36253159e-02 1.81222439e-01 -6.44132257e-01 1.17111874e+00 8.29084158e-01 -4.67295766e-01 1.14837337e+00 -3.00704360e-01 -1.00846790e-01 -1.56978350e-02 -8.53624523e-01 -5.93329631e-02 3.77067685e-01 -4.34115753e-02 6.62066221e-01 -4.12889980e-02 -6.06712759e-01 1.66551602e+00 -1.19156443e-01 4.94420499e-01 2.75853604e-01 5.54493964e-01 -8.44257712e-01 -6.77616000e-01 2.08099306e-01 1.25887990e+00 -7.59213507e-01 -3.04311484e-01 1.30683854e-01 2.39489615e-01 9.67530608e-02 6.51156902e-01 3.16870600e-01 -8.53101790e-01 2.97051042e-01 -3.35195035e-01 -2.32098624e-01 -6.05740309e-01 -7.97874093e-01 9.96822715e-02 -5.49132712e-02 -3.14906150e-01 -1.32916033e-01 -3.69846851e-01 -1.07630491e+00 -5.51986992e-01 -5.37985206e-01 5.94565094e-01 4.41036463e-01 1.09278238e+00 7.20232904e-01 9.20842528e-01 7.70610452e-01 -7.34595895e-01 -5.99565506e-01 -1.13834298e+00 -7.65277624e-01 2.56064236e-01 3.81341368e-01 -5.17664790e-01 -1.86928421e-01 -2.77568161e-01]
[10.156679153442383, 5.598781585693359]
8a0e19b5-4694-4521-be41-bbe2636e321c
multi-target-multiplicity-flexibility-and
2306.13738
null
https://arxiv.org/abs/2306.13738v1
https://arxiv.org/pdf/2306.13738v1.pdf
Multi-Target Multiplicity: Flexibility and Fairness in Target Specification under Resource Constraints
Prediction models have been widely adopted as the basis for decision-making in domains as diverse as employment, education, lending, and health. Yet, few real world problems readily present themselves as precisely formulated prediction tasks. In particular, there are often many reasonable target variable options. Prior work has argued that this is an important and sometimes underappreciated choice, and has also shown that target choice can have a significant impact on the fairness of the resulting model. However, the existing literature does not offer a formal framework for characterizing the extent to which target choice matters in a particular task. Our work fills this gap by drawing connections between the problem of target choice and recent work on predictive multiplicity. Specifically, we introduce a conceptual and computational framework for assessing how the choice of target affects individuals' outcomes and selection rate disparities across groups. We call this multi-target multiplicity. Along the way, we refine the study of single-target multiplicity by introducing notions of multiplicity that respect resource constraints -- a feature of many real-world tasks that is not captured by existing notions of predictive multiplicity. We apply our methods on a healthcare dataset, and show that the level of multiplicity that stems from target variable choice can be greater than that stemming from nearly-optimal models of a single target.
['Alexandra Chouldechova', 'Jake M. Hofman', 'Solon Barocas', 'Jamelle Watson-Daniels']
2023-06-23
null
null
null
null
['fairness', 'fairness', 'decision-making']
['computer-vision', 'miscellaneous', 'reasoning']
[ 4.54602540e-01 1.26271188e-01 -9.06019330e-01 -2.85089642e-01 -3.97131294e-01 -2.37677351e-01 5.78619361e-01 4.14362311e-01 -4.72341210e-01 8.65696073e-01 3.88301671e-01 -5.63418448e-01 -7.13025868e-01 -7.25811899e-01 -5.03993630e-01 -5.80421031e-01 2.95040399e-01 5.25145411e-01 -1.77549869e-01 -2.41336212e-01 3.56518507e-01 3.55328321e-01 -1.37414181e+00 5.83666563e-02 9.59074318e-01 7.20752716e-01 -9.61411446e-02 9.61481184e-02 1.39086843e-01 4.94809926e-01 -4.12918031e-01 -6.80697560e-01 2.96565145e-01 -3.16182017e-01 -7.78447986e-01 -1.07083842e-02 2.34450534e-01 1.79236084e-01 1.11647785e-01 8.27113986e-01 4.83797580e-01 8.04410055e-02 9.46281672e-01 -1.54789793e+00 -4.84822333e-01 7.26229370e-01 -3.06693137e-01 2.42971659e-01 6.27858639e-02 1.45943910e-01 1.31226766e+00 -4.35842395e-01 4.42359507e-01 1.21562934e+00 6.08414292e-01 3.51939440e-01 -1.57936954e+00 -7.04602599e-01 3.21658909e-01 -1.19860899e-02 -1.30351973e+00 -5.13724446e-01 5.17199874e-01 -9.20356870e-01 5.89121938e-01 7.09441602e-01 4.07437921e-01 9.32585299e-01 3.71831536e-01 2.97955185e-01 1.12146521e+00 -5.65732121e-01 8.32188576e-02 1.90851063e-01 2.10080981e-01 2.79258311e-01 5.30776501e-01 1.03721112e-01 -4.17003393e-01 -4.25830960e-01 5.98364890e-01 1.62366614e-01 -2.42151916e-01 -4.04752493e-01 -1.16700888e+00 1.07405150e+00 -2.08423734e-02 2.66280800e-01 -2.49509811e-01 -1.73596799e-01 2.61480480e-01 3.21465015e-01 6.46167517e-01 8.33469868e-01 -5.39744258e-01 1.18464559e-01 -7.75023699e-01 5.33472896e-01 7.61807263e-01 5.56949317e-01 5.03978252e-01 -3.70973289e-01 -3.91811371e-01 7.07283020e-01 -1.44705856e-02 1.67438492e-01 4.26169038e-01 -1.01898611e+00 3.90281945e-01 5.50984442e-01 4.45858926e-01 -1.17833161e+00 -5.71442008e-01 -6.14564598e-01 -7.66267300e-01 6.72402754e-02 6.99389040e-01 -1.30767733e-01 -3.26440632e-01 2.26978397e+00 2.22932369e-01 1.36194289e-01 -3.92856926e-01 7.50260472e-01 1.91393420e-01 2.71022320e-03 3.76739234e-01 -5.20751238e-01 1.24957037e+00 -7.09310055e-01 -4.09043610e-01 -3.99449706e-01 6.81439817e-01 -5.69470644e-01 8.87638569e-01 1.36203051e-01 -9.43693101e-01 -8.36109072e-02 -5.31749725e-01 1.20852098e-01 -6.62314817e-02 -3.91448081e-01 7.49367356e-01 8.71854961e-01 -7.43184268e-01 6.52495682e-01 -3.47589791e-01 -2.45230436e-01 3.11445475e-01 3.64647985e-01 -1.86071008e-01 -3.69565226e-02 -1.09458613e+00 1.26688695e+00 9.80442166e-02 -3.22774649e-01 -1.29006088e-01 -1.09917212e+00 -5.30188918e-01 2.76670277e-01 7.11364627e-01 -1.15950406e+00 1.25239110e+00 -1.23943019e+00 -9.94400620e-01 6.02510035e-01 -2.07757682e-01 -2.33804494e-01 6.55692101e-01 7.77188838e-02 -3.58372450e-01 -5.46053767e-01 3.92705441e-01 1.07610554e-01 6.20779335e-01 -7.97706068e-01 -6.61871493e-01 -5.30744195e-01 3.20464045e-01 4.36947823e-01 -2.06689954e-01 2.54030585e-01 4.68944535e-02 -6.89583898e-01 -2.53083467e-01 -8.64009142e-01 -6.70279860e-01 -1.66258276e-01 -4.64430571e-01 -3.83358479e-01 -1.01034805e-01 -7.63767809e-02 1.60763991e+00 -2.03919196e+00 3.07678908e-01 2.00028360e-01 3.56027186e-01 -1.55593053e-01 -3.44325276e-03 3.43873620e-01 -1.48301840e-01 5.99202096e-01 -2.30670311e-02 -2.87312716e-01 2.73776114e-01 7.03706592e-02 -2.93270707e-01 5.50162435e-01 7.28805959e-02 6.61322594e-01 -6.52382314e-01 -3.54159355e-01 1.97406746e-02 2.65741259e-01 -7.44571149e-01 -2.35404164e-01 -1.44741476e-01 2.42077872e-01 -5.41533411e-01 4.46846426e-01 2.33211696e-01 -4.60018635e-01 4.14475113e-01 2.33860970e-01 -1.40345037e-01 4.90080237e-01 -1.16401303e+00 9.57594216e-01 -1.08173840e-01 1.73161656e-01 -1.30272493e-01 -1.24528277e+00 3.91619802e-01 4.98776585e-02 7.02606678e-01 -4.55196142e-01 8.01158324e-03 1.96970612e-01 4.59379971e-01 -3.01678985e-01 4.91215348e-01 -5.99375725e-01 -1.27468497e-01 4.20093834e-01 -6.42824769e-01 1.56923279e-01 3.51760648e-02 -1.81201607e-01 8.63825798e-01 -4.70512629e-01 7.40124047e-01 -5.78349710e-01 1.40114397e-01 1.40422434e-01 8.48180532e-01 1.00912678e+00 -9.59654674e-02 3.95415008e-01 7.94544280e-01 -2.62405723e-01 -1.03633606e+00 -7.09417105e-01 -5.19100368e-01 1.24893904e+00 3.65435556e-02 -2.03972384e-01 -3.55356753e-01 -3.79425615e-01 3.53498161e-01 6.94946051e-01 -7.72050023e-01 4.66662943e-02 -2.49776140e-01 -1.16780794e+00 2.47255936e-01 2.08290979e-01 -2.35032842e-01 -6.66320324e-01 -7.86788404e-01 1.94868326e-01 -1.39834151e-01 -8.10438335e-01 -5.27155399e-01 1.31525174e-02 -7.92286754e-01 -1.16644144e+00 -4.98389065e-01 -1.43667340e-01 4.34460491e-01 1.24204382e-01 1.40121448e+00 2.34508634e-01 -1.56882286e-01 3.81471395e-01 -1.14022776e-01 -6.71091318e-01 -2.96082854e-01 1.75657630e-01 2.18090281e-01 -1.23799421e-01 3.75045180e-01 -4.53442484e-01 -6.18301272e-01 2.67754614e-01 -7.94964790e-01 2.81877935e-01 5.22979796e-01 8.53646636e-01 3.71458203e-01 5.06874323e-02 8.76414597e-01 -1.25454676e+00 7.46117651e-01 -8.68718922e-01 -3.33310843e-01 3.21129203e-01 -1.10772395e+00 -9.58392397e-04 3.89645338e-01 -6.30809307e-01 -5.87641954e-01 -2.71598220e-01 1.98775187e-01 -6.89541921e-02 -5.40410504e-02 8.31258297e-01 -8.19557458e-02 1.53943926e-01 5.16413689e-01 -2.48794556e-01 1.94265723e-01 -3.07543755e-01 9.93693173e-02 4.16664273e-01 -2.79753655e-02 -7.97556341e-01 4.43541497e-01 1.40384108e-01 3.18014085e-01 -5.94130695e-01 -9.56609249e-01 -2.58777440e-01 -3.48706394e-01 2.26241693e-01 6.68794632e-01 -7.22700000e-01 -8.46229196e-01 -2.35854238e-02 -6.32201195e-01 -4.39016551e-01 -2.06282854e-01 4.52069014e-01 -5.51832557e-01 1.60879716e-01 -1.78994253e-01 -8.94989014e-01 7.08709657e-02 -1.37344086e+00 6.10549986e-01 -7.84156621e-02 -5.24215221e-01 -1.08607590e+00 -1.56396598e-01 4.52027023e-01 5.59044242e-01 2.87308186e-01 1.47590899e+00 -7.35276937e-01 -4.34378624e-01 1.76520452e-01 2.01117527e-02 -3.23487192e-01 5.54178581e-02 -2.11119145e-01 -5.48106134e-01 -2.58832332e-03 -9.10713002e-02 4.44363765e-02 6.43108845e-01 7.24741280e-01 1.04251754e+00 -5.72082400e-01 -4.52955902e-01 2.50878692e-01 1.47314763e+00 9.85087007e-02 2.85171628e-01 4.35267061e-01 5.23472786e-01 7.85281003e-01 5.46716034e-01 6.11290216e-01 5.96235573e-01 1.08753335e+00 1.59626693e-01 -7.37047866e-02 4.61508304e-01 6.81571066e-02 -7.54541755e-02 1.30722255e-01 -9.26864147e-02 -1.22356817e-01 -1.01191401e+00 4.04917926e-01 -2.01830053e+00 -1.09344494e+00 2.14754865e-02 2.47950053e+00 1.03137374e+00 1.44639015e-01 5.96753776e-01 1.88734710e-01 5.84780037e-01 -1.01612709e-01 -5.02166629e-01 -4.41683859e-01 1.02577984e-01 3.91095951e-02 6.40777886e-01 6.29823089e-01 -8.81520987e-01 3.96842271e-01 7.07171249e+00 7.90571451e-01 -1.05529737e+00 5.13203442e-02 1.13590240e+00 -2.36670703e-01 -5.64075708e-01 -1.26398623e-01 -7.04596281e-01 5.63735723e-01 8.40824962e-01 -7.28911519e-01 3.09710532e-01 7.67478108e-01 4.15223598e-01 5.63409552e-02 -1.55051851e+00 6.18602872e-01 -1.35022312e-01 -1.14771557e+00 -9.84486192e-02 4.66742605e-01 5.97126484e-01 -3.06186527e-01 3.83121103e-01 2.07452342e-01 4.54473138e-01 -1.54809165e+00 7.13690162e-01 3.82237911e-01 7.40260124e-01 -7.08265901e-01 4.04048830e-01 5.70293307e-01 -6.31706595e-01 -5.22876978e-01 -1.42610535e-01 -5.56603611e-01 -1.22403890e-01 7.67289519e-01 -6.99439764e-01 3.97237480e-01 3.05518359e-01 2.30725005e-01 -3.17771137e-01 1.03822744e+00 3.75398219e-01 5.73204756e-01 -1.57753468e-01 1.22142248e-01 -6.28894791e-02 -1.61891520e-01 4.58770096e-01 9.56065178e-01 3.74063432e-01 2.21213326e-01 2.92808622e-01 6.92748189e-01 -5.26899658e-03 3.86844873e-01 -6.93233609e-01 -1.28554902e-03 6.93083584e-01 6.38408780e-01 -5.76420426e-01 -6.72087818e-02 -6.27319336e-01 2.67787427e-01 1.91418111e-01 2.58323431e-01 -6.56935573e-01 3.21970582e-01 1.10780764e+00 4.13591027e-01 -1.45805240e-01 1.75541546e-02 -7.84815192e-01 -1.11210954e+00 -1.74744055e-01 -1.16675949e+00 5.06611168e-01 -1.86420932e-01 -1.50109553e+00 -8.50567371e-02 2.40318850e-01 -7.72616625e-01 -2.57121503e-01 -5.18360615e-01 -3.93163323e-01 1.11814606e+00 -1.36400604e+00 -8.19218755e-01 2.84307837e-01 2.92450130e-01 4.27428305e-01 7.93311149e-02 8.43905330e-01 2.48102441e-01 -8.12736273e-01 6.36675954e-01 3.19800042e-02 -2.95672894e-01 6.71009183e-01 -1.12074149e+00 -1.21142417e-01 6.83313608e-01 -1.56944498e-01 6.54146254e-01 8.80146027e-01 -4.85313773e-01 -1.06179547e+00 -8.27881515e-01 1.24342012e+00 -7.87386358e-01 6.10971630e-01 1.54850632e-02 -5.01402915e-01 8.03397417e-01 -3.04560721e-01 -3.63415182e-01 9.86904502e-01 7.42552936e-01 -3.18235993e-01 2.46394519e-02 -1.08520627e+00 7.78526306e-01 1.09794831e+00 -2.12165862e-01 -3.37461323e-01 2.64538884e-01 7.81616271e-01 -2.58109748e-01 -1.00280929e+00 5.08375883e-01 6.98588789e-01 -1.01230812e+00 1.18198395e+00 -9.92079377e-01 6.76397502e-01 2.98211157e-01 -2.72477180e-01 -1.14629531e+00 -7.12856412e-01 -2.86266059e-01 3.03652644e-01 9.43609416e-01 6.26260161e-01 -7.95754671e-01 6.35027051e-01 1.44030452e+00 4.33936566e-02 -1.05702078e+00 -9.81850863e-01 -6.17520332e-01 6.80001915e-01 -3.74889135e-01 8.14886451e-01 1.36153817e+00 -4.68340069e-02 3.37162137e-01 -3.74702454e-01 -4.83737420e-03 5.05823374e-01 3.76710951e-01 6.19598806e-01 -1.66522193e+00 -4.41545486e-01 -9.82349157e-01 -3.48736167e-01 -5.05076647e-01 2.08883673e-01 -7.65647888e-01 -4.11406636e-01 -1.32361960e+00 5.84406853e-01 -1.02889156e+00 -4.42296207e-01 4.73099709e-01 -3.71095359e-01 -8.46519917e-02 3.72277737e-01 3.57912660e-01 -2.45291770e-01 2.97371596e-02 1.08347607e+00 9.46385562e-02 -1.67947143e-01 2.43914574e-01 -1.47535086e+00 6.12693489e-01 7.91839957e-01 -4.91271973e-01 -4.63388503e-01 -1.27932787e-01 5.19310236e-01 2.03440726e-01 3.78698349e-01 -3.47175300e-01 -5.92081398e-02 -1.13591325e+00 1.67367607e-01 2.78595477e-01 7.81323165e-02 -8.38926911e-01 4.76909757e-01 4.64896888e-01 -6.38521791e-01 1.05523169e-01 -1.92164883e-01 4.98481482e-01 3.21676522e-01 -2.41569728e-01 5.50132275e-01 -1.11831605e-01 -3.43660653e-01 2.28691161e-01 -1.23212323e-01 3.78487051e-01 9.88367677e-01 -2.45841015e-02 -3.96819413e-01 -2.16817990e-01 -6.05663240e-01 6.15834221e-02 6.24050558e-01 3.99507701e-01 4.84964326e-02 -1.12247002e+00 -7.34151483e-01 -1.01139784e-01 3.00577916e-02 -2.83696115e-01 -3.56702060e-02 1.12240577e+00 1.41138941e-01 3.85277212e-01 -7.65786618e-02 -2.09482640e-01 -1.23621500e+00 6.26858830e-01 3.47448528e-01 -4.04046893e-01 -4.93973911e-01 5.13811469e-01 4.15459961e-01 -2.07525849e-01 6.85925633e-02 -1.68026209e-01 -1.75322473e-01 6.99676275e-02 2.93570340e-01 4.05786395e-01 -2.31662706e-01 -6.71526909e-01 -4.07298326e-01 2.95981050e-01 2.14838713e-01 -5.71351200e-02 1.32989752e+00 -2.36940622e-01 -1.20376535e-01 5.33857822e-01 6.03987098e-01 1.42576545e-01 -7.18945503e-01 -1.80537865e-01 1.11508466e-01 -5.25047183e-01 -2.77690232e-01 -8.24203968e-01 -6.44210935e-01 3.99198562e-01 2.69895613e-01 3.68762434e-01 1.06635559e+00 -1.41733974e-01 8.35397616e-02 1.10262498e-01 3.08817208e-01 -8.13545942e-01 -2.60720700e-01 2.72227019e-01 6.91342533e-01 -1.40968633e+00 7.45255351e-02 -5.49805522e-01 -6.21090174e-01 5.71203172e-01 4.25645173e-01 2.48709023e-01 8.27456355e-01 7.75898769e-02 -2.77420938e-01 -2.52778810e-02 -8.93656850e-01 -1.13891870e-01 4.01296139e-01 3.10716391e-01 5.94260573e-01 6.01223946e-01 -8.61951292e-01 9.46929753e-01 -3.33877057e-01 7.47313276e-02 5.86833000e-01 6.49057329e-01 -2.81981558e-01 -1.31144142e+00 -4.41858768e-01 1.00246572e+00 -9.59537625e-01 -2.38811880e-01 -1.83397949e-01 7.31989801e-01 3.24364841e-01 9.33210671e-01 1.64771900e-01 -2.08078533e-01 9.91594121e-02 2.08979249e-01 3.60359877e-01 -8.62584352e-01 -4.82090533e-01 -8.79662186e-02 1.86263919e-01 -2.49033645e-01 -4.61827815e-01 -7.41514683e-01 -5.74807882e-01 -6.68436825e-01 -2.30981767e-01 1.84468836e-01 2.62468904e-01 1.09181356e+00 3.99941683e-01 3.29196513e-01 3.79242837e-01 -4.94673103e-01 -8.85866940e-01 -6.29099607e-01 -5.60831428e-01 4.52828854e-01 3.31313848e-01 -9.27061558e-01 -3.93517435e-01 -2.79691786e-01]
[8.486918449401855, 5.370142459869385]
53d39896-c604-4192-b545-dfde9e861ebf
pento-diaref-a-diagnostic-dataset-for
2305.15087
null
https://arxiv.org/abs/2305.15087v1
https://arxiv.org/pdf/2305.15087v1.pdf
Pento-DIARef: A Diagnostic Dataset for Learning the Incremental Algorithm for Referring Expression Generation from Examples
NLP tasks are typically defined extensionally through datasets containing example instantiations (e.g., pairs of image i and text t), but motivated intensionally through capabilities invoked in verbal descriptions of the task (e.g., "t is a description of i, for which the content of i needs to be recognised and understood"). We present Pento-DIARef, a diagnostic dataset in a visual domain of puzzle pieces where referring expressions are generated by a well-known symbolic algorithm (the "Incremental Algorithm"), which itself is motivated by appeal to a hypothesised capability (eliminating distractors through application of Gricean maxims). Our question then is whether the extensional description (the dataset) is sufficient for a neural model to pick up the underlying regularity and exhibit this capability given the simple task definition of producing expressions from visual inputs. We find that a model supported by a vision detection step and a targeted data generation scheme achieves an almost perfect BLEU@1 score and sentence accuracy, whereas simpler baselines do not.
['David Schlangen', 'Philipp Sadler']
2023-05-24
null
null
null
null
['referring-expression-generation', 'referring-expression']
['computer-vision', 'computer-vision']
[ 7.86878169e-01 6.54970109e-01 2.58614700e-02 -5.41166365e-01 -6.95939660e-01 -1.02874207e+00 1.19125795e+00 -6.03680797e-02 -2.04656333e-01 5.25724113e-01 4.90705818e-01 -3.77073556e-01 -3.31912220e-01 -5.06502450e-01 -9.08399999e-01 -6.05694115e-01 8.47693905e-02 7.27275372e-01 -1.64250553e-01 -3.18332523e-01 5.42295039e-01 3.06594640e-01 -1.37383354e+00 9.30389404e-01 4.22256052e-01 8.75456691e-01 5.02154410e-01 8.01249325e-01 -1.76294222e-01 1.15941262e+00 -8.34320366e-01 -7.21480191e-01 6.63510486e-02 -7.99995661e-01 -1.21015632e+00 5.06505251e-01 4.80644494e-01 -3.00371170e-01 -2.28105322e-01 9.28850949e-01 1.90631330e-01 -4.37261127e-02 8.78153503e-01 -1.32125521e+00 -1.04229093e+00 9.29144800e-01 -1.88585624e-01 4.15014513e-02 7.81064212e-01 7.16670632e-01 1.30110943e+00 -6.07408583e-01 1.09124911e+00 1.32806563e+00 3.63660395e-01 8.42305005e-01 -1.58053303e+00 -4.87249345e-02 -5.51324524e-03 -7.89089575e-02 -1.03640699e+00 -4.70138192e-01 7.18161523e-01 -4.51522112e-01 1.23527682e+00 5.45704842e-01 6.46359265e-01 1.52064443e+00 -4.84398156e-02 9.26130235e-01 1.22710705e+00 -4.47617799e-01 2.83594877e-01 1.48486614e-01 -9.77524146e-02 7.43211508e-01 -6.33971347e-03 9.15481970e-02 -7.92985082e-01 -7.14909378e-03 1.07223642e+00 -6.78513944e-01 -4.45202827e-01 -3.78335953e-01 -1.51290441e+00 7.43648291e-01 4.51106101e-01 2.40285963e-01 -5.27917564e-01 2.99993515e-01 5.63182592e-01 2.95277953e-01 -2.09143817e-01 1.04699147e+00 -3.61716598e-01 -8.88191685e-02 -9.84317362e-01 4.24621850e-01 7.69271553e-01 1.09974778e+00 3.08942854e-01 1.23438768e-01 -3.37479562e-01 3.33310485e-01 7.58418674e-03 2.03082666e-01 6.09907687e-01 -1.50368047e+00 3.68079454e-01 5.21246195e-01 1.73582897e-01 -7.25616574e-01 -3.91729385e-01 -2.29429692e-01 -3.71807426e-01 3.07139069e-01 8.30113888e-01 5.23214005e-02 -6.89816952e-01 2.02999091e+00 -1.89070508e-01 -3.21030080e-01 4.15832937e-01 1.07811069e+00 9.75460291e-01 5.38401902e-01 2.02563956e-01 -2.15859726e-01 1.59791124e+00 -6.39551818e-01 -4.33424026e-01 -4.27620322e-01 6.72351062e-01 -3.07546437e-01 1.58447134e+00 5.13416171e-01 -1.42135239e+00 -4.53433663e-01 -1.04089737e+00 -4.23766375e-01 -1.65665731e-01 8.83359239e-02 9.69654500e-01 2.52551138e-01 -1.12271285e+00 3.78271699e-01 -2.31666282e-01 -5.90635657e-01 5.09565353e-01 1.39798820e-01 -6.03286862e-01 1.40809827e-02 -8.30964684e-01 9.16005373e-01 5.56791604e-01 -1.03128947e-01 -1.23415756e+00 -1.92805186e-01 -1.02396405e+00 -3.16243842e-02 4.07334566e-01 -1.03901958e+00 1.34638834e+00 -1.76146483e+00 -1.02886450e+00 1.55391514e+00 -1.84586421e-02 -6.27462864e-01 5.21964967e-01 5.29134907e-02 1.14189431e-01 5.89480221e-01 2.02976480e-01 1.28329194e+00 1.00182199e+00 -1.47013533e+00 -2.52509296e-01 -3.25567663e-01 8.07134509e-01 3.94481272e-01 3.33345205e-01 2.20245466e-01 -2.13851050e-01 -6.23052359e-01 1.11734957e-01 -8.16864252e-01 1.22677766e-01 8.07165205e-02 -7.28758514e-01 -3.52817953e-01 3.25047672e-01 -7.42050886e-01 6.13219142e-01 -2.23581100e+00 3.39310050e-01 -8.77448544e-02 2.06704080e-01 -3.16434801e-01 -2.46826142e-01 4.21019554e-01 -3.14868212e-01 4.08645749e-01 -2.35640109e-01 -2.11565588e-02 4.47771460e-01 4.41425502e-01 -5.47923148e-01 2.93261141e-01 5.70595980e-01 1.30131149e+00 -8.61195147e-01 -5.79992414e-01 -3.41273025e-02 2.05405161e-01 -2.86713958e-01 1.63644731e-01 -7.30035305e-01 2.00975999e-01 -2.05345154e-01 6.28310323e-01 1.19313277e-01 -4.68679607e-01 2.20489994e-01 -2.08578214e-01 8.71591121e-02 4.14038718e-01 -7.70523608e-01 1.88297212e+00 -1.52074903e-01 7.92589545e-01 1.02632483e-02 -9.32863474e-01 6.21844471e-01 3.25973094e-01 -2.64990717e-01 -5.76145768e-01 1.33311585e-01 4.97722626e-02 3.29581439e-01 -9.13110852e-01 3.18386495e-01 -3.79721671e-01 -3.11405391e-01 5.51770985e-01 -8.03754181e-02 -5.23957193e-01 2.32276335e-01 6.99518323e-01 1.22597957e+00 6.80159092e-01 4.76265162e-01 -1.78382769e-01 2.42962658e-01 5.70253909e-01 2.28448689e-01 1.09015357e+00 -2.17732284e-02 6.83579266e-01 9.95070875e-01 -4.09327179e-01 -1.34965968e+00 -1.17499924e+00 6.81096967e-03 1.02848077e+00 -3.49767171e-02 -4.23317879e-01 -8.07931960e-01 -7.46067941e-01 -3.55406165e-01 1.13542223e+00 -8.76043737e-01 1.53760770e-02 -4.97943312e-01 -3.08813304e-01 8.12668383e-01 6.14061117e-01 3.78667295e-01 -1.66434371e+00 -1.30864191e+00 6.60061985e-02 -2.47432128e-01 -1.05863452e+00 -3.54963727e-02 3.85006875e-01 -6.84547782e-01 -8.96288693e-01 -4.73715901e-01 -8.27310801e-01 8.79892588e-01 -1.79476187e-01 1.46725368e+00 2.07658693e-01 -3.26448977e-01 7.37241507e-01 -3.94272178e-01 -2.64859349e-01 -4.05494541e-01 -5.55022418e-01 -2.88757771e-01 -3.53063792e-01 2.24403635e-01 -3.30544591e-01 -4.22992647e-01 -8.31299648e-02 -9.48444366e-01 5.51868737e-01 8.21955144e-01 8.76925588e-01 4.80621397e-01 -2.88223207e-01 3.36353689e-01 -7.52216339e-01 8.01247835e-01 -2.86297321e-01 -1.42882541e-01 3.38187248e-01 1.12482391e-01 2.79346406e-01 4.66016501e-01 -6.73841834e-01 -1.04419804e+00 2.33607471e-01 2.01306731e-01 -9.52805057e-02 -5.89004159e-01 2.57245004e-01 -3.28107804e-01 3.32017124e-01 9.13731635e-01 5.72131693e-01 -9.52931792e-02 6.99808076e-02 6.42195225e-01 2.88622081e-01 1.10053849e+00 -1.10830688e+00 4.61607337e-01 4.69330817e-01 6.85260296e-02 -6.46280169e-01 -8.60044003e-01 3.56572680e-02 -5.21399081e-01 -2.45849207e-01 1.03520203e+00 -6.96522176e-01 -8.01015317e-01 -1.64455473e-01 -1.63334572e+00 -4.96454656e-01 -6.14078283e-01 1.91882342e-01 -1.25605094e+00 2.02926606e-01 -5.88929892e-01 -8.81279528e-01 -6.07034117e-02 -1.00794899e+00 1.16361570e+00 -1.41893486e-02 -8.60229492e-01 -6.72287881e-01 -5.45652688e-01 3.76570731e-01 1.68401673e-02 5.48401833e-01 1.35812473e+00 -8.23911607e-01 -4.58811313e-01 1.01879545e-01 -5.05552173e-01 1.62133291e-01 -2.46466175e-01 -7.25956112e-02 -8.85475814e-01 1.79846540e-01 2.70745724e-01 -9.80929673e-01 5.56472659e-01 1.76608309e-01 1.18373024e+00 -7.01785803e-01 -7.98738524e-02 3.81242305e-01 1.39638734e+00 1.28012136e-01 8.38676453e-01 2.44366720e-01 2.96130598e-01 1.02844524e+00 3.56828839e-01 2.75311828e-01 2.96573997e-01 6.48738325e-01 5.56827128e-01 1.26257613e-01 -2.33233377e-01 -2.63175607e-01 4.72110301e-01 -1.41133413e-01 -1.64119899e-01 -1.19759031e-01 -8.81873548e-01 5.94722211e-01 -1.74792051e+00 -1.30160701e+00 -3.50225806e-01 1.74469304e+00 1.01366818e+00 3.50118458e-01 2.18437575e-02 2.96952277e-02 4.61714566e-01 1.27141699e-01 -5.56992948e-01 -8.23641300e-01 -4.80965376e-01 1.85405195e-01 1.26670916e-02 3.39356750e-01 -7.42322862e-01 8.45989347e-01 6.61057186e+00 6.67111158e-01 -7.00063944e-01 -7.54303038e-02 6.07071459e-01 -1.19089983e-01 -3.80171180e-01 -4.58292430e-03 -3.21350038e-01 2.45288908e-01 7.19090760e-01 3.21701616e-02 5.12255132e-01 6.89083815e-01 1.02455050e-01 -3.52295935e-01 -1.69624221e+00 9.00888681e-01 6.06382728e-01 -1.38085306e+00 3.21995109e-01 -1.25315547e-01 2.96594232e-01 -5.08070648e-01 2.60307997e-01 2.74867088e-01 3.84543240e-01 -1.40587378e+00 1.14250493e+00 3.87268007e-01 8.06797326e-01 -3.32565904e-01 3.45880300e-01 6.22862935e-01 -5.24906993e-01 -7.16036707e-02 -2.10945249e-01 -3.07802618e-01 1.58287019e-01 -6.09887615e-02 -7.88476527e-01 2.49953628e-01 2.27503315e-01 1.59683123e-01 -5.37293136e-01 4.81670052e-01 -7.10341871e-01 3.24848175e-01 -3.40529121e-02 -2.46154711e-01 5.49995899e-01 1.32341787e-01 7.60807872e-01 1.37589204e+00 1.86508447e-02 2.54622728e-01 -6.44261986e-02 1.41747200e+00 4.05290574e-02 -7.13033304e-02 -8.89827728e-01 -5.32666743e-02 2.37731770e-01 1.22324955e+00 -6.63975477e-01 -5.24018109e-01 -1.89508155e-01 1.32803679e+00 3.24317932e-01 4.20205235e-01 -8.45950544e-01 -1.41621128e-01 6.26778305e-02 2.31665924e-01 3.31372857e-01 -1.38191745e-01 -4.54193830e-01 -8.34693074e-01 6.26657829e-02 -1.03411186e+00 2.25629047e-01 -1.85941041e+00 -1.21593702e+00 5.59763432e-01 7.40970597e-02 -7.58841634e-01 -6.40601218e-01 -1.02029169e+00 -5.44209063e-01 8.04837704e-01 -9.03195560e-01 -1.46368301e+00 -2.36557260e-01 6.40835345e-01 7.96349943e-01 7.70050883e-02 1.01230824e+00 -6.70040667e-01 2.03933883e-02 2.87968427e-01 -8.19282830e-01 2.20773101e-01 3.66887927e-01 -1.57661140e+00 2.12608799e-01 6.53717875e-01 6.74403727e-01 8.02009642e-01 1.16354024e+00 -5.09054720e-01 -1.48877692e+00 -6.14907682e-01 1.00476551e+00 -9.50638950e-01 7.68302083e-01 -4.24192071e-01 -7.42170095e-01 9.48922276e-01 5.45416951e-01 -2.77875334e-01 4.91019368e-01 -3.19487900e-01 -7.09764421e-01 5.29352248e-01 -1.29388988e+00 6.94699347e-01 1.18631291e+00 -5.47044218e-01 -1.41067445e+00 6.90652788e-01 6.51455045e-01 -3.92038912e-01 -3.14713508e-01 1.20238349e-01 4.63683456e-01 -9.71156418e-01 7.98754871e-01 -1.20678532e+00 1.03326809e+00 -3.22419554e-01 -4.89993900e-01 -1.09166610e+00 -2.02653781e-01 -5.86042762e-01 1.38211027e-02 1.10245585e+00 5.18731475e-01 -7.68987834e-02 4.87761110e-01 7.29290724e-01 -3.84344086e-02 -5.23557305e-01 -9.33857322e-01 -8.16207826e-01 -1.51701823e-01 -6.64697409e-01 2.81642377e-01 1.00799048e+00 4.17491704e-01 5.71783245e-01 -6.09731339e-02 -1.29594475e-01 4.23601717e-01 1.23209298e-01 6.31199718e-01 -8.27605069e-01 -6.40529215e-01 -4.08266991e-01 -4.25686270e-01 -1.09165907e+00 3.43980223e-01 -1.01630628e+00 1.39433652e-01 -1.71049726e+00 3.82300317e-01 2.13367026e-02 3.07686806e-01 5.56559205e-01 3.87239248e-01 2.88217008e-01 3.10263664e-01 2.27442637e-01 -7.53706634e-01 5.45114763e-02 1.22265875e+00 -1.73595265e-01 9.78634804e-02 -6.43500388e-01 -1.16235137e+00 1.00729704e+00 6.25579119e-01 -1.56927645e-01 -4.34853017e-01 -6.00142300e-01 8.13557923e-01 2.34184131e-01 1.01394665e+00 -5.88757396e-01 1.10803850e-01 -2.26930246e-01 5.70183873e-01 -5.05059808e-02 4.85672057e-01 -6.15161002e-01 4.48790425e-03 2.91014940e-01 -8.09077144e-01 2.41141111e-01 6.97262362e-02 3.42906386e-01 3.34891304e-02 -4.87292975e-01 4.53905642e-01 -6.02061689e-01 -9.57916379e-01 -4.27725226e-01 -3.33272785e-01 1.53662756e-01 9.63181376e-01 -4.14188415e-01 -7.20345378e-01 -4.30987865e-01 -9.54664946e-01 -9.78642181e-02 6.58932149e-01 6.45216107e-02 8.65167558e-01 -1.15944898e+00 -8.30625892e-01 -1.41602889e-01 4.28333282e-01 -6.98711816e-03 -3.98468338e-02 7.61776865e-01 -4.56228465e-01 1.09331571e-01 -3.23874295e-01 -4.40245241e-01 -1.03431010e+00 7.72636294e-01 2.14520156e-01 1.27599522e-01 -7.48818934e-01 1.11607599e+00 5.62426627e-01 -7.57945562e-03 1.26150146e-01 -3.29584509e-01 -8.18607509e-02 -1.08110130e-01 6.96422875e-01 -3.56159210e-01 -5.17556906e-01 -8.26429605e-01 -2.68291324e-01 2.77214289e-01 1.55221643e-02 -4.90104824e-01 1.11666536e+00 1.32425651e-02 -1.18185811e-01 5.22251844e-01 5.93826592e-01 -1.34505183e-01 -1.18960261e+00 1.25635177e-01 -1.65192261e-02 -2.06948921e-01 -5.33515513e-01 -1.36595511e+00 -4.11221832e-01 8.72791648e-01 -1.46989807e-01 3.44688416e-01 9.67276990e-01 6.86531961e-01 2.15976432e-01 4.77287889e-01 2.81303614e-01 -8.13674808e-01 4.55255657e-01 3.91168654e-01 1.44053793e+00 -1.03563571e+00 -1.16543591e-01 -1.71752721e-01 -1.18775654e+00 1.07929158e+00 6.90740526e-01 -1.52181029e-01 -4.73846853e-01 3.90818000e-01 -2.62784392e-01 -6.37200117e-01 -1.09350669e+00 -2.30283737e-01 9.95618999e-02 9.21234429e-01 2.58632779e-01 8.76167119e-02 -2.45199263e-01 8.71585071e-01 -4.57626343e-01 -3.02377760e-01 5.68698525e-01 6.87164843e-01 -4.43126619e-01 -4.46655989e-01 -2.85001636e-01 3.57511878e-01 -3.29509288e-01 -3.08776230e-01 -1.01292181e+00 1.11143863e+00 3.36857975e-01 7.16202140e-01 2.47082442e-01 4.37027887e-02 9.98260602e-02 3.24969649e-01 9.48699057e-01 -8.15986931e-01 -6.17661417e-01 -1.18316025e-01 4.33169901e-01 -4.26081777e-01 -5.44314504e-01 -7.14267313e-01 -1.44189823e+00 1.27393782e-01 1.74428821e-01 -1.93778008e-01 6.48154080e-01 1.12777793e+00 -1.90474302e-01 3.73304397e-01 -7.02385306e-02 -7.37055480e-01 -5.59548378e-01 -8.75934184e-01 -4.81459856e-01 8.21080625e-01 9.39511210e-02 -3.37967247e-01 -4.72803950e-01 5.78566492e-01]
[10.715112686157227, 1.8249149322509766]
1f5481e5-1c08-40cc-bb5f-beb47ecdc3b2
localization-distillation-for-object
2102.12252
null
https://arxiv.org/abs/2102.12252v4
https://arxiv.org/pdf/2102.12252v4.pdf
Localization Distillation for Dense Object Detection
Knowledge distillation (KD) has witnessed its powerful capability in learning compact models in object detection. Previous KD methods for object detection mostly focus on imitating deep features within the imitation regions instead of mimicking classification logit due to its inefficiency in distilling localization information and trivial improvement. In this paper, by reformulating the knowledge distillation process on localization, we present a novel localization distillation (LD) method which can efficiently transfer the localization knowledge from the teacher to the student. Moreover, we also heuristically introduce the concept of valuable localization region that can aid to selectively distill the semantic and localization knowledge for a certain region. Combining these two new components, for the first time, we show that logit mimicking can outperform feature imitation and localization knowledge distillation is more important and efficient than semantic knowledge for distilling object detectors. Our distillation scheme is simple as well as effective and can be easily applied to different dense object detectors. Experiments show that our LD can boost the AP score of GFocal-ResNet-50 with a single-scale 1x training schedule from 40.1 to 42.1 on the COCO benchmark without any sacrifice on the inference speed. Our source code and trained models are publicly available at https://github.com/HikariTJU/LD
['Ming-Ming Cheng', 'Qibin Hou', 'Dongwei Ren', 'WangMeng Zuo', 'Ping Wang', 'Rongguang Ye', 'Zhaohui Zheng']
2021-02-24
null
http://openaccess.thecvf.com//content/CVPR2022/html/Zheng_Localization_Distillation_for_Dense_Object_Detection_CVPR_2022_paper.html
http://openaccess.thecvf.com//content/CVPR2022/papers/Zheng_Localization_Distillation_for_Dense_Object_Detection_CVPR_2022_paper.pdf
cvpr-2022-1
['dense-object-detection']
['computer-vision']
[-5.50165951e-01 1.95270523e-01 -2.16063425e-01 -9.68451947e-02 -7.44827688e-01 -7.16801643e-01 4.63553280e-01 -1.68369170e-02 -7.29949176e-01 6.04906619e-01 -1.93159640e-01 -2.12120429e-01 -4.80578579e-02 -7.02251673e-01 -1.10562396e+00 -6.90837979e-01 8.74259546e-02 4.07011896e-01 6.14678442e-01 1.50430068e-01 -4.01583351e-02 6.24791384e-01 -1.21852922e+00 2.30309833e-02 8.25771630e-01 9.85914111e-01 7.28120327e-01 5.85404754e-01 1.32928744e-01 9.27702248e-01 -4.94408071e-01 -4.30230647e-01 2.92085916e-01 -1.99540392e-01 -8.29845428e-01 -5.02070189e-01 4.29559857e-01 -6.61709666e-01 -6.98030710e-01 1.24791527e+00 7.57280827e-01 9.40632373e-02 6.52135611e-01 -1.26908171e+00 -9.38633025e-01 8.64251316e-01 -7.87307441e-01 4.91124541e-01 2.54322421e-02 2.10641176e-01 8.67364109e-01 -1.25165153e+00 4.12527353e-01 1.25743639e+00 6.94253504e-01 4.97048944e-01 -8.62719715e-01 -8.83746028e-01 2.94150174e-01 5.26649714e-01 -1.71336639e+00 6.62998110e-02 5.41583240e-01 -3.34924579e-01 7.56418347e-01 -2.28673562e-01 5.26962578e-01 9.61127937e-01 -3.27102929e-01 1.42838538e+00 9.23972845e-01 -4.29230034e-01 -2.96542712e-04 4.77995366e-01 5.09911142e-02 1.09137499e+00 3.11341524e-01 1.62123978e-01 -2.12956339e-01 2.78006613e-01 9.95565355e-01 2.60581851e-01 -1.70836881e-01 -4.24944997e-01 -1.17585123e+00 9.33829248e-01 1.24212182e+00 2.21764818e-01 -1.51670367e-01 4.83524054e-01 2.73878664e-01 1.94184870e-01 2.15754449e-01 5.14246583e-01 -5.42402327e-01 -3.94001305e-02 -7.22182274e-01 5.49949706e-02 6.42060578e-01 1.16909754e+00 7.99266756e-01 -9.55993310e-02 -3.75573814e-01 6.64335310e-01 1.77217886e-01 7.98286498e-01 6.19044244e-01 -8.37054551e-01 1.53588578e-01 5.24107873e-01 3.20218690e-02 -6.28737688e-01 -2.46207327e-01 -7.33947396e-01 -5.29528618e-01 2.00144015e-02 3.98024261e-01 -2.31344298e-01 -1.02720952e+00 1.78302801e+00 4.18754280e-01 6.54563785e-01 -4.15399931e-02 8.29931140e-01 9.90352333e-01 4.35920447e-01 6.40987903e-02 4.49045688e-01 1.52308238e+00 -1.36492598e+00 -1.61951497e-01 -8.88772160e-02 8.63784850e-01 -5.66578746e-01 1.05159044e+00 1.40808433e-01 -9.41153347e-01 -6.93725348e-01 -9.45547998e-01 -2.12053105e-01 -6.26288116e-01 4.85397160e-01 7.33240247e-01 3.48805428e-01 -1.11592197e+00 4.06967819e-01 -8.98153722e-01 -2.44763687e-01 8.64042103e-01 3.79918665e-01 -1.92142487e-01 -1.79134324e-01 -1.11845767e+00 1.05652606e+00 7.79713571e-01 -1.76507339e-01 -1.21709824e+00 -9.66219127e-01 -7.81256199e-01 2.75361061e-01 6.39733851e-01 -6.33395970e-01 1.52494597e+00 -8.01879168e-01 -1.42759156e+00 5.95654607e-01 2.14851439e-01 -7.23262727e-01 6.77978992e-01 -4.59042847e-01 8.20471942e-02 3.77077878e-01 1.66083321e-01 1.27791154e+00 6.96360111e-01 -8.67428303e-01 -1.03492558e+00 -9.92387719e-03 3.06589127e-01 2.13131085e-01 -3.44995171e-01 -2.44616389e-01 -7.62374103e-01 -5.75807273e-01 -1.46544978e-01 -7.49556422e-01 -1.48481160e-01 3.16378444e-01 -2.89104551e-01 -4.99665231e-01 8.92695606e-01 -3.92860442e-01 7.91094661e-01 -2.20044208e+00 2.79993210e-02 -8.30726326e-03 5.12184143e-01 6.42826140e-01 -2.56858140e-01 -7.08818212e-02 1.63236901e-01 -1.86109081e-01 3.06433979e-02 -2.73793012e-01 2.28471339e-01 1.70599997e-01 -4.23009366e-01 4.82699394e-01 2.49590263e-01 1.48109651e+00 -1.20557714e+00 -4.67507273e-01 3.49833727e-01 6.68288648e-01 -7.26309121e-01 4.53888029e-02 -2.38229468e-01 3.18608314e-01 -4.93621141e-01 6.12582326e-01 7.60374486e-01 -4.22123492e-01 -3.19910586e-01 -3.40581015e-02 6.13986030e-02 9.19584632e-02 -1.11792827e+00 1.70221090e+00 -5.15629828e-01 6.93847179e-01 -8.06952938e-02 -1.01412773e+00 5.88239193e-01 -6.87541142e-02 9.91256386e-02 -6.29877985e-01 1.68142587e-01 2.68012106e-01 -1.17729411e-01 -2.37229124e-01 3.42512429e-01 1.93926886e-01 -4.24079970e-02 2.87930161e-01 5.02564371e-01 1.75400674e-02 -3.25798010e-03 4.03770804e-01 8.91979337e-01 1.54814184e-01 2.10162103e-01 -1.35752723e-01 3.51628959e-01 -1.50444940e-01 3.19703460e-01 1.14442384e+00 -4.21450198e-01 2.83363879e-01 3.34729463e-01 -5.86194396e-02 -7.78203964e-01 -1.28545177e+00 -1.72388360e-01 1.26589811e+00 4.82323021e-01 -1.48594543e-01 -6.94549620e-01 -1.07131648e+00 3.16828430e-01 5.55766642e-01 -6.76452279e-01 -3.23968768e-01 -4.72576410e-01 -4.24599230e-01 8.26064944e-01 1.03858113e+00 9.31171060e-01 -9.76288199e-01 -4.91377711e-01 -1.00200534e-01 6.97689727e-02 -1.11885345e+00 -4.09155697e-01 2.18879178e-01 -5.97452700e-01 -1.00463235e+00 -1.06939864e+00 -1.10469890e+00 7.41282225e-01 5.88936448e-01 8.62920046e-01 -7.23083019e-02 -5.06283879e-01 5.25787413e-01 -3.60766590e-01 -6.62118196e-01 -4.49480489e-02 3.36314112e-01 1.40587240e-01 -5.07376790e-01 5.38655460e-01 -3.36547613e-01 -8.31849158e-01 2.95449674e-01 -6.24142289e-01 -8.23799055e-03 9.54223275e-01 7.81861365e-01 5.57322621e-01 -1.18053421e-01 5.90878725e-01 -5.67729354e-01 2.03104943e-01 -6.16889536e-01 -7.67929435e-01 2.19339028e-01 -4.43976998e-01 2.79684484e-01 4.84064579e-01 -7.03247190e-01 -8.89369190e-01 1.74221441e-01 -1.50429606e-01 -8.36130083e-01 -4.07495573e-02 2.08403736e-01 1.65729120e-01 -4.47263688e-01 6.36281908e-01 3.13620806e-01 -1.83585703e-01 -6.90255344e-01 8.58739614e-01 3.14985067e-01 6.54887795e-01 -6.40528560e-01 9.32727754e-01 4.77394134e-01 -2.79193729e-01 -3.79020840e-01 -9.84457552e-01 -6.22961164e-01 -6.26321375e-01 1.70203716e-01 6.47214651e-01 -1.29739070e+00 -1.00083458e+00 3.46897811e-01 -9.36048269e-01 -4.57050174e-01 -5.46962798e-01 6.61270559e-01 -3.41152191e-01 2.87256658e-01 -6.52846634e-01 -3.49463403e-01 -1.56669617e-01 -1.17056966e+00 9.88776624e-01 3.97020757e-01 4.15484101e-01 -9.21355426e-01 -9.79146659e-02 6.68717772e-02 4.73285049e-01 -2.28463367e-01 3.79663199e-01 -6.68320119e-01 -8.27140450e-01 -2.78754532e-01 -7.69477010e-01 3.15376520e-01 -1.12961151e-01 -4.20675665e-01 -1.11703503e+00 -4.24606681e-01 -2.94244796e-01 -6.23684764e-01 1.41879988e+00 2.65459687e-01 1.34989977e+00 -2.71076888e-01 -5.70528507e-01 6.71607137e-01 1.31745410e+00 -1.83900580e-01 2.33591393e-01 2.06293106e-01 8.79381835e-01 8.33009407e-02 5.80195904e-01 2.32118979e-01 4.69385654e-01 5.99354923e-01 3.91342819e-01 -3.03582728e-01 -6.14123285e-01 -6.48282528e-01 4.34492111e-01 5.13667524e-01 1.05923489e-01 4.76874933e-02 -7.59729266e-01 7.67881036e-01 -1.86480451e+00 -7.78450727e-01 4.28752422e-01 1.88936031e+00 1.00968540e+00 9.30719078e-03 2.79031307e-01 -3.24674964e-01 7.14858353e-01 -2.59840727e-01 -6.48256302e-01 1.96812227e-02 1.24329031e-01 1.81680501e-01 7.69063771e-01 6.59582973e-01 -1.18299377e+00 1.37439132e+00 5.32878876e+00 1.27178538e+00 -1.04524565e+00 5.63572705e-01 3.21674347e-01 -1.20918818e-01 1.53124690e-01 -2.95400381e-01 -1.18017948e+00 3.95043433e-01 6.61137819e-01 -1.30547717e-01 1.55273259e-01 1.40627730e+00 -2.98925400e-01 -9.63188484e-02 -1.13170171e+00 9.86172020e-01 -2.03828156e-01 -1.26801538e+00 1.16283804e-01 -2.25969553e-01 7.55231142e-01 3.76939118e-01 4.00911242e-01 8.87438834e-01 5.80639005e-01 -9.19419527e-01 8.98194253e-01 1.79649070e-01 7.46210456e-01 -6.69899583e-01 8.27885151e-01 5.05474627e-01 -1.27399194e+00 -1.87295973e-01 -6.44869208e-01 -2.92513799e-02 -3.00177127e-01 3.56981874e-01 -1.41440582e+00 3.64332378e-01 7.56922901e-01 5.65435767e-01 -7.51460612e-01 1.32637751e+00 -7.52062917e-01 5.37174642e-01 -5.30434906e-01 -1.95543662e-01 5.76453745e-01 3.19748610e-01 3.47722262e-01 1.47003627e+00 2.02351317e-01 -1.15087353e-01 2.63779372e-01 1.11222005e+00 -4.16692227e-01 -2.22559810e-01 -6.13921940e-01 2.42654756e-01 7.04441071e-01 1.14193106e+00 -8.08517277e-01 -5.64884722e-01 -3.62369448e-01 1.07354438e+00 7.11373568e-01 3.22417974e-01 -1.18510997e+00 -5.83039343e-01 5.10808825e-01 -1.87680945e-01 8.98679554e-01 -9.77222696e-02 1.14252880e-01 -1.17274868e+00 -1.44159570e-01 -2.71182895e-01 3.86728466e-01 -6.60763502e-01 -1.05048108e+00 2.70253897e-01 1.51498377e-01 -9.61333632e-01 4.68422659e-02 -8.32217455e-01 -4.28720385e-01 7.06259370e-01 -1.91672361e+00 -1.18104923e+00 -3.94699842e-01 6.88021600e-01 4.91809964e-01 2.20568534e-02 4.70166832e-01 4.94356543e-01 -5.66800654e-01 1.02520919e+00 2.16928661e-01 4.96514469e-01 6.88186288e-01 -1.44635856e+00 2.17282474e-01 8.39649796e-01 3.59638125e-01 6.47631466e-01 3.08862895e-01 -2.36995384e-01 -1.18040407e+00 -1.38216627e+00 4.29985821e-01 -6.70267403e-01 7.54234850e-01 -3.68420005e-01 -7.20528364e-01 8.77446890e-01 -5.81048466e-02 3.57438236e-01 1.21069364e-01 -3.04381475e-02 -6.88062787e-01 -7.64698237e-02 -1.21528172e+00 5.06403685e-01 9.91309941e-01 -5.74875534e-01 -6.67472720e-01 5.22160113e-01 9.67389941e-01 -5.65136075e-01 -5.61591029e-01 3.19213420e-01 3.23538154e-01 -5.67101717e-01 1.27512610e+00 -3.86503994e-01 9.78948176e-02 -3.98177743e-01 -2.96839923e-02 -1.02475584e+00 -3.78236264e-01 -3.26244354e-01 -4.73362625e-01 1.04178512e+00 3.86267871e-01 -7.30481803e-01 7.87301004e-01 4.17234711e-02 -4.96338345e-02 -8.49106014e-01 -8.42621088e-01 -1.11169839e+00 2.54179358e-01 -3.91269803e-01 4.28972900e-01 8.74243379e-01 -1.55254304e-01 2.10522771e-01 -7.61377960e-02 4.25642014e-01 5.47795355e-01 1.74611405e-01 6.64610684e-01 -8.72695506e-01 -6.30130708e-01 -6.06153071e-01 -6.09193027e-01 -1.65022540e+00 7.03402534e-02 -1.27924955e+00 3.07232793e-02 -1.33802605e+00 5.36322653e-01 -5.53344011e-01 -6.89691603e-01 8.60140741e-01 -1.98717386e-01 4.25868541e-01 3.78553182e-01 1.34704411e-01 -1.01660061e+00 5.56315124e-01 1.35868967e+00 -2.20269367e-01 -1.22023180e-01 -7.16521144e-02 -8.12287688e-01 7.82659173e-01 6.71508491e-01 -6.09871805e-01 -4.79042858e-01 -4.65320319e-01 -1.61503837e-01 -3.88447434e-01 7.73344934e-01 -9.41435575e-01 4.17295992e-01 4.35145915e-01 4.56050903e-01 -4.33889717e-01 2.15962440e-01 -6.93850338e-01 -5.54435372e-01 8.17953646e-01 -2.18059435e-01 -3.58547479e-01 4.08816308e-01 7.11854339e-01 -5.66860288e-02 -3.03002924e-01 9.16840255e-01 -1.42700970e-01 -1.28173161e+00 3.75031948e-01 1.42569065e-01 1.13458082e-01 1.26025093e+00 5.87674715e-02 -4.67946261e-01 -1.97989941e-01 -7.39509821e-01 5.04548013e-01 2.26481155e-01 3.54037225e-01 5.49043238e-01 -1.26216638e+00 -5.64862072e-01 1.37293682e-01 -5.68094524e-03 1.71993881e-01 1.03516012e-01 1.04488969e+00 -5.73385239e-01 6.66669846e-01 -5.99175654e-02 -7.82462716e-01 -9.35524940e-01 8.05682778e-01 4.33875829e-01 -1.19753793e-01 -7.76113093e-01 1.50672579e+00 6.82318211e-01 -3.43519866e-01 6.97749078e-01 -6.42101645e-01 1.18024565e-01 -1.42502323e-01 6.49460554e-01 4.17567134e-01 -2.12393656e-01 -2.11330682e-01 -4.99937236e-01 5.29637396e-01 -4.65988904e-01 1.91728249e-01 9.93304968e-01 3.88285257e-02 3.38598073e-01 1.64598655e-02 1.18535960e+00 -2.84362853e-01 -1.65807486e+00 -6.55804753e-01 -1.36952505e-01 -2.77841747e-01 1.43872172e-01 -9.30199146e-01 -1.19379902e+00 9.54434574e-01 7.18677402e-01 -1.76593959e-01 8.83836150e-01 5.71407914e-01 6.35162294e-01 7.91031361e-01 3.82432342e-01 -7.81839311e-01 4.09569144e-01 4.72173780e-01 6.47448361e-01 -1.40497029e+00 -1.64309189e-01 -2.58458436e-01 -5.24331212e-01 7.63436913e-01 8.18247259e-01 -3.09019774e-01 6.55636430e-01 3.21359992e-01 -2.52134621e-01 -2.03605160e-01 -3.85633111e-01 -5.47199786e-01 2.31016323e-01 4.54286665e-01 3.48031484e-02 1.46411896e-01 1.76276669e-01 6.63542747e-01 -3.81041109e-03 1.92347616e-02 1.62406608e-01 7.90215194e-01 -8.39235187e-01 -6.46279573e-01 -1.66891918e-01 1.55825377e-01 -4.35988128e-01 -2.37866297e-01 -1.83459267e-01 1.07197821e+00 3.14748794e-01 5.20937741e-01 -4.61991802e-02 -1.89342618e-01 1.77945554e-01 -2.16640443e-01 6.60099447e-01 -6.22277081e-01 -2.51036644e-01 -1.47370145e-01 -4.65082616e-01 -6.20928109e-01 -8.78765956e-02 -2.91487724e-01 -1.48593724e+00 -2.47379974e-01 -5.74317217e-01 1.73113674e-01 4.91323173e-01 8.51971328e-01 4.80274796e-01 5.74532568e-01 3.46634865e-01 -8.89136255e-01 -8.06211948e-01 -9.58068192e-01 -4.66119826e-01 -2.43504662e-02 3.65761817e-01 -1.01397383e+00 -3.11618537e-01 -3.19414377e-01]
[9.325974464416504, 1.3807191848754883]
88076794-0d2a-4ea9-ac6b-ee512881a9d0
me-gcn-multi-dimensional-edge-enhanced-graph
null
null
https://openreview.net/forum?id=pY5QUHPRhNN
https://openreview.net/pdf?id=pY5QUHPRhNN
ME-GCN: Multi-dimensional Edge-Enhanced Graph Convolutional Networks for Semi-supervised Text Classification
Compared to sequential learning models, graph-based neural networks exhibit excellent ability in capturing global information and have been used for semi-supervised learning tasks, including citation network analysis or text classification. Most Graph Convolutional Networks are designed with the single-dimensional edge feature and failed to utilise the rich edge information about graphs. In this paper, we introduce the ME-GCN (Multi-dimensional Edge-enhanced Graph Convolutional Networks) for semi-supervised text classification. A text graph for an entire corpus is firstly constructed to describe the undirected and multi-dimensional relationship of word-to-word, document-document, and word-to-document. The graph is initialised with corpus-trained multi-dimensional word and document node representation, and the relations are represented according to the distance of those words/documents nodes. Then, the generated graph is trained with ME-GCN, which considers the edge features as multi-stream signals, and each stream performs a separate graph convolutional operation. Our ME-GCN can integrate a rich source of graph edge information of the entire text corpus. The results have demonstrated that our proposed model has significantly outperformed the state-of-the-art methods across eight benchmark datasets.
['Anonymous']
2021-12-17
null
null
null
acl-arr-december-2022-12
['semi-supervised-text-classification-1']
['natural-language-processing']
[-1.95251429e-03 2.66746506e-02 -1.74359024e-01 -8.94868895e-02 6.22959137e-02 -3.21816742e-01 8.42051268e-01 5.24275959e-01 -1.39175102e-01 1.43213466e-01 2.43318439e-01 -5.13297141e-01 -1.99158773e-01 -1.20658970e+00 -2.81827569e-01 -4.79395509e-01 -4.44836527e-01 4.26612735e-01 2.01966465e-01 -2.20311448e-01 3.48176807e-01 3.41213107e-01 -1.15924394e+00 7.28236511e-02 4.73935604e-01 9.80780065e-01 1.63389608e-01 1.00395393e+00 -9.89305615e-01 8.03671062e-01 -4.89741266e-01 -3.24635357e-01 -2.62163371e-01 -5.71027637e-01 -7.06520319e-01 3.73797631e-03 2.14629978e-01 -1.25554511e-02 -9.06141162e-01 1.22722518e+00 4.11915034e-01 1.57335520e-01 7.24033535e-01 -1.42466986e+00 -9.38023627e-01 8.35004330e-01 -5.02902627e-01 2.51312494e-01 4.39097434e-01 -3.77750814e-01 1.37458920e+00 -8.29387128e-01 8.38121235e-01 1.31095004e+00 5.28931081e-01 9.13463011e-02 -6.53439164e-01 -5.52093804e-01 3.53513300e-01 1.35751948e-01 -1.27095151e+00 1.64373070e-01 1.11731052e+00 -3.20684284e-01 1.30916440e+00 -4.33189869e-02 9.83280301e-01 9.60578799e-01 7.27148831e-01 6.89973056e-01 1.86234742e-01 -4.04831618e-01 -6.97081983e-02 -4.48489070e-01 6.78803205e-01 1.20446444e+00 4.20852482e-01 -4.35341179e-01 -5.56714296e-01 1.09164966e-02 7.15970278e-01 1.96248606e-01 -3.10271680e-01 -8.29397514e-02 -1.11777377e+00 9.10169899e-01 7.46817529e-01 7.19584882e-01 -2.99552888e-01 1.75851509e-01 6.92538381e-01 4.02233243e-01 7.91392744e-01 1.06032401e-01 -8.81947726e-02 2.57877856e-02 -7.09063947e-01 -1.44405574e-01 1.04768038e+00 1.15227318e+00 6.36906922e-01 6.86682090e-02 -2.47252598e-01 4.78522718e-01 6.36995018e-01 1.52572691e-01 7.48921752e-01 1.35604337e-01 7.66534150e-01 1.35190737e+00 -8.26713920e-01 -1.72652555e+00 -8.31426919e-01 -6.48837209e-01 -1.45269370e+00 -4.08909231e-01 -7.69791678e-02 -1.49148449e-01 -8.55551779e-01 1.15961564e+00 5.62684983e-02 1.85476840e-01 -4.59145149e-03 5.67716420e-01 1.66397035e+00 8.59240770e-01 -2.22690836e-01 -2.13064879e-01 1.25011849e+00 -1.08484542e+00 -1.00525212e+00 -2.03621253e-01 9.27283108e-01 -3.09842795e-01 7.99714804e-01 7.47220367e-02 -7.70459175e-01 -4.98895615e-01 -9.84321356e-01 -1.19424976e-01 -9.18157339e-01 -8.88233930e-02 7.74122894e-01 1.64367482e-01 -1.23667300e+00 4.73820180e-01 -4.70238417e-01 -4.14942652e-01 5.09674788e-01 2.01830193e-01 -4.50792223e-01 -5.18308803e-02 -1.42379439e+00 3.11726063e-01 6.54316545e-01 2.75113821e-01 -3.90657127e-01 -7.15546533e-02 -1.16522133e+00 5.45786142e-01 1.16293333e-01 -5.66837132e-01 5.09279191e-01 -5.93278587e-01 -1.19631612e+00 5.51147223e-01 5.54349683e-02 -2.02410057e-01 2.85445284e-02 3.70334625e-01 -8.42961192e-01 2.89273530e-01 -3.40831429e-02 1.99386492e-01 7.12568045e-01 -8.76561821e-01 -3.68787795e-01 -5.11924803e-01 -3.02956313e-01 4.15013969e-01 -8.73822212e-01 -1.35125205e-01 -9.37308013e-01 -6.90922737e-01 2.76631504e-01 -5.31332910e-01 4.64409254e-02 -3.29817206e-01 -7.24947453e-01 -6.46121442e-01 1.08432055e+00 -3.34385931e-01 1.68217468e+00 -2.08447552e+00 2.27998391e-01 3.98454517e-01 8.87358129e-01 2.02904671e-01 -4.27102268e-01 7.37605810e-01 -3.00320871e-02 1.71916112e-01 3.89964245e-02 -4.21200901e-01 -1.08729959e-01 7.53736794e-02 9.77690369e-02 4.60775435e-01 -6.37991130e-02 1.40153825e+00 -1.07955098e+00 -7.84287632e-01 6.26681279e-03 5.72500885e-01 6.23800084e-02 1.66896567e-01 -1.91900074e-01 -2.09449887e-01 -6.68602288e-01 3.84345800e-01 5.10734081e-01 -7.89490104e-01 2.65311927e-01 -9.01706740e-02 3.58865589e-01 7.14212358e-02 -1.08529389e+00 1.74733329e+00 -1.41450241e-01 1.04162705e+00 -2.13734642e-01 -1.16894650e+00 1.33561134e+00 1.13848634e-01 3.85928988e-01 -6.02304518e-01 5.10184169e-01 -1.27952561e-01 -3.23206768e-03 -4.68734443e-01 5.84462404e-01 5.09082675e-01 5.12263132e-03 5.97110629e-01 4.90522683e-01 -1.32771164e-01 3.79087538e-01 8.53613913e-01 1.13467979e+00 -3.97022009e-01 1.26474336e-01 -2.29515582e-01 7.01073945e-01 -2.74758548e-01 -2.25459337e-01 3.99348468e-01 2.75226444e-01 4.03043032e-01 9.17515039e-01 -3.58217597e-01 -5.12943804e-01 -5.51125526e-01 2.29277685e-01 9.91566658e-01 1.82660595e-01 -8.95278692e-01 -4.73822415e-01 -8.69865417e-01 6.24078810e-02 4.05036718e-01 -7.12340415e-01 -3.84213179e-01 -2.51974165e-01 -5.90772986e-01 3.80836427e-01 4.31482583e-01 5.90976894e-01 -1.09013247e+00 1.99015439e-01 3.13887388e-01 2.34174758e-01 -9.60110426e-01 -7.83551097e-01 2.04727799e-01 -7.83180714e-01 -1.33230615e+00 -6.83299601e-01 -1.40514374e+00 8.55975688e-01 4.16533738e-01 1.16175723e+00 4.78247672e-01 -2.81956851e-01 4.18409616e-01 -6.17128253e-01 -2.10430339e-01 -1.52476877e-01 3.69320720e-01 -4.94032264e-01 2.12634221e-01 4.38944459e-01 -3.76579940e-01 -4.48477417e-01 -9.74418148e-02 -9.10840034e-01 -8.53412375e-02 2.52100259e-01 7.83283830e-01 3.70958298e-01 3.49655539e-01 5.98991871e-01 -1.16532063e+00 1.51026261e+00 -6.38744056e-01 -3.68013233e-01 3.07651401e-01 -8.68466198e-01 2.98911668e-02 8.90068531e-01 -3.89228016e-01 -5.26873767e-01 -2.32781425e-01 1.64611593e-01 -2.46307626e-01 2.58204907e-01 1.21108735e+00 -1.63483303e-02 -3.12083792e-02 4.02142584e-01 3.23700815e-01 1.75797585e-02 -1.60208583e-01 7.19212711e-01 1.05049348e+00 3.36672962e-01 -6.69074357e-02 4.38476175e-01 1.49169683e-01 3.91344547e-01 -8.15658808e-01 -4.31779653e-01 -6.01398408e-01 -6.70218289e-01 -4.24975663e-01 9.41421509e-01 -5.26430070e-01 -6.10761166e-01 8.25693488e-01 -1.23855245e+00 -2.37142280e-01 1.15851119e-01 5.22940159e-01 9.13427621e-02 6.09157741e-01 -7.72943318e-01 -4.91862148e-01 -8.11421275e-01 -7.15595603e-01 1.05549896e+00 2.29152754e-01 2.29392037e-01 -1.74954975e+00 -2.89622396e-02 -2.99299389e-01 4.05025393e-01 2.22122505e-01 1.06911993e+00 -1.16674793e+00 4.46285168e-03 -6.32095337e-01 -5.08507073e-01 5.63007034e-03 2.93516457e-01 1.11044273e-01 -4.22382444e-01 -4.17579710e-01 -4.82619494e-01 -8.14780816e-02 9.92813468e-01 3.96974534e-01 1.21317101e+00 -2.31708914e-01 -6.80840552e-01 5.94032526e-01 1.35903776e+00 5.39323920e-03 2.72066623e-01 1.75270826e-01 1.27688146e+00 3.57920110e-01 -9.78078544e-02 2.72094905e-01 6.10219300e-01 -1.10074906e-02 6.68117225e-01 -2.54328609e-01 -2.54493594e-01 -1.77926272e-01 -5.99439070e-02 1.43102956e+00 1.24205865e-01 -1.12740242e+00 -1.01288199e+00 4.77265716e-01 -1.99837279e+00 -5.82112789e-01 -7.89139032e-01 1.52813327e+00 2.33595401e-01 4.94616151e-01 -6.83386847e-02 1.52431026e-01 1.02670312e+00 6.14183903e-01 -4.37472165e-01 -2.66024023e-01 -2.15496376e-01 9.66355503e-02 3.48621100e-01 3.14707607e-01 -8.34904134e-01 9.88135457e-01 5.82813787e+00 5.41290760e-01 -1.12027669e+00 -3.28406423e-01 3.30998242e-01 3.24912906e-01 -3.65640670e-01 -3.25110525e-01 -4.65690225e-01 3.68292630e-01 9.17732477e-01 -5.11670470e-01 2.93736428e-01 5.36977053e-01 -3.50371629e-01 3.71602744e-01 -9.50364411e-01 1.07799590e+00 3.66260648e-01 -1.45444202e+00 4.66884792e-01 8.11981224e-03 5.61048687e-01 1.43058345e-01 -2.85344154e-01 2.46206790e-01 4.14483875e-01 -1.05283344e+00 1.83210939e-01 6.54574394e-01 9.44419742e-01 -7.71976292e-01 9.40118909e-01 3.55132252e-01 -1.98537338e+00 1.18728645e-01 -3.86378795e-01 -5.92730492e-02 -6.48091808e-02 6.81767821e-01 -6.77359283e-01 1.06249726e+00 3.47630531e-01 1.37394571e+00 -7.73837209e-01 5.87747514e-01 -1.26918018e-01 5.35097659e-01 -1.19848344e-02 -7.96865225e-01 5.76243639e-01 -3.92852962e-01 3.51359665e-01 1.48745620e+00 3.58273298e-01 -5.71323186e-02 1.06754676e-01 5.50072253e-01 -5.97991824e-01 3.79851192e-01 -8.06332171e-01 -7.05327392e-01 3.43096882e-01 1.55753434e+00 -1.15800905e+00 -5.86921215e-01 -6.88093245e-01 9.79490340e-01 5.61339855e-01 4.96983796e-01 -3.54233801e-01 -1.11864758e+00 -5.19411899e-02 -2.70365655e-01 2.46758670e-01 -4.06159461e-01 2.53098644e-02 -1.04128742e+00 -1.59471452e-01 -3.23440522e-01 8.01673412e-01 -8.93067300e-01 -1.47689843e+00 1.04105949e+00 -3.56924564e-01 -9.67638135e-01 -2.30967388e-01 -7.82400191e-01 -1.06854856e+00 9.26980376e-01 -1.34010720e+00 -1.20791340e+00 -6.63071454e-01 9.57312763e-01 3.25511932e-01 -6.12873018e-01 8.56371045e-01 -7.91759863e-02 -5.89060187e-01 4.86107171e-01 1.37395635e-01 6.74124479e-01 2.61840761e-01 -1.34158552e+00 8.89901340e-01 6.59421444e-01 4.47892427e-01 4.04645920e-01 1.79464594e-02 -9.41637397e-01 -1.83559680e+00 -1.27546692e+00 9.06521142e-01 6.08887374e-02 1.01732302e+00 -5.31866729e-01 -9.67009306e-01 7.42393017e-01 4.04758543e-01 4.67828721e-01 4.27994847e-01 2.88547799e-02 -2.85913050e-01 1.61166638e-01 -7.23049939e-01 5.01343548e-01 1.27008808e+00 -6.66425288e-01 -4.12883937e-01 5.55465817e-01 1.10371912e+00 -3.95815104e-01 -9.50873196e-01 -2.11050138e-01 1.96283460e-01 -4.08935785e-01 6.35531604e-01 -7.84787655e-01 5.38726687e-01 8.79955590e-02 1.89135179e-01 -1.54974771e+00 -6.90400720e-01 -5.08848906e-01 -4.53416049e-01 1.18150222e+00 3.64749670e-01 -8.49800825e-01 6.42112613e-01 -1.18649177e-01 -1.96915403e-01 -8.18273783e-01 -6.42057955e-01 -5.05032182e-01 -1.18503682e-01 -3.18806320e-01 7.61462629e-01 1.23243034e+00 4.31767076e-01 9.74703968e-01 1.40338093e-01 -8.32050219e-02 5.20955443e-01 1.52706206e-01 4.71964091e-01 -1.67267799e+00 1.80414453e-01 -9.58268523e-01 -8.01222146e-01 -1.15676153e+00 4.81474817e-01 -1.58920538e+00 -3.95802021e-01 -2.42734790e+00 1.33019043e-02 9.82925072e-02 -2.86453933e-01 3.03082585e-01 -3.01328808e-01 -2.40363792e-01 -1.79156333e-01 1.10517405e-01 -7.10220456e-01 5.28244793e-01 1.65743971e+00 -5.59122682e-01 -2.67479330e-01 -2.26183757e-01 -5.23096442e-01 2.22399592e-01 6.45301282e-01 -3.07771772e-01 -6.12385750e-01 -4.76104438e-01 5.93938410e-01 1.14384450e-01 3.02547892e-03 -6.37206495e-01 9.10236776e-01 2.51532108e-01 3.59991491e-01 -7.38873303e-01 -1.92231759e-01 -8.34480047e-01 -2.61524230e-01 3.93685043e-01 -3.99524182e-01 1.57735229e-01 -7.92044401e-02 9.87180352e-01 -3.42140138e-01 -6.66537210e-02 8.72285739e-02 -2.05858037e-01 -4.67623740e-01 8.07675481e-01 -1.73350707e-01 3.43529582e-02 7.35046983e-01 -1.07665271e-01 -5.89808881e-01 -4.70872849e-01 -4.59074885e-01 4.95506465e-01 -1.38631970e-01 6.08351350e-01 8.53281379e-01 -1.41039634e+00 -7.06503093e-01 3.52620780e-01 2.55414635e-01 3.63945574e-01 -3.83770429e-02 2.97087997e-01 -5.12096882e-01 4.84083951e-01 8.25195387e-02 -5.19810498e-01 -1.04899848e+00 7.13772535e-01 1.94607437e-01 -4.62261707e-01 -9.84524071e-01 9.12826180e-01 -2.04961434e-01 -3.41943145e-01 4.22339052e-01 -4.01202440e-01 -7.89107680e-01 3.73105526e-01 1.39777705e-01 2.40286782e-01 2.10154489e-01 -6.71993613e-01 -3.41497958e-01 5.93086481e-01 -7.69060925e-02 2.54495054e-01 1.43731332e+00 -1.04484865e-02 -4.86512482e-01 5.13311684e-01 1.67802835e+00 -3.63225907e-01 -5.98165095e-01 -5.65824151e-01 -6.64169490e-02 -1.34345712e-02 5.69884479e-01 -2.00481072e-01 -1.45369935e+00 9.57618177e-01 1.32208496e-01 9.42168295e-01 9.22400475e-01 1.06476218e-01 8.19027126e-01 5.34217954e-01 -1.12033173e-01 -9.46953416e-01 3.71073872e-01 7.92534888e-01 7.79840946e-01 -9.90814626e-01 1.64704621e-01 -3.11651915e-01 -4.40938860e-01 1.77549410e+00 4.94959980e-01 -1.76146016e-01 1.22275138e+00 1.34702742e-01 -2.66299307e-01 -1.05076778e+00 -6.27915025e-01 -2.01967180e-01 6.65527344e-01 4.75416452e-01 4.54051733e-01 1.53499395e-02 -1.72336221e-01 5.30661047e-01 -2.12877050e-01 -2.40718275e-01 3.05470645e-01 7.07374811e-01 -3.56643707e-01 -6.94984436e-01 1.48735955e-01 1.04161870e+00 -2.73399078e-03 -3.77319694e-01 -7.58341670e-01 6.73349738e-01 -5.60826957e-01 9.55759645e-01 5.15572548e-01 -6.18267477e-01 1.62485719e-01 4.71659936e-02 7.82467797e-02 -6.76989079e-01 -5.69227993e-01 -6.03519939e-02 4.72586602e-02 -2.65076756e-01 -2.57402271e-01 -1.27361029e-01 -1.62999308e+00 -2.67269999e-01 -6.42684758e-01 1.96880832e-01 7.65494764e-01 8.36343229e-01 4.68543917e-01 1.14089262e+00 7.46115148e-01 -6.85137630e-01 1.27712771e-01 -1.20970225e+00 -9.76143658e-01 3.61064792e-01 2.96546191e-01 -2.78881907e-01 -5.62612295e-01 -2.97537625e-01]
[9.945577621459961, 6.757925033569336]
baf77286-6720-4d20-b974-aaf90d348664
schema-inference-for-interpretable-image
2303.06635
null
https://arxiv.org/abs/2303.06635v1
https://arxiv.org/pdf/2303.06635v1.pdf
Schema Inference for Interpretable Image Classification
In this paper, we study a novel inference paradigm, termed as schema inference, that learns to deductively infer the explainable predictions by rebuilding the prior deep neural network (DNN) forwarding scheme, guided by the prevalent philosophical cognitive concept of schema. We strive to reformulate the conventional model inference pipeline into a graph matching policy that associates the extracted visual concepts of an image with the pre-computed scene impression, by analogy with human reasoning mechanism via impression matching. To this end, we devise an elaborated architecture, termed as SchemaNet, as a dedicated instantiation of the proposed schema inference concept, that models both the visual semantics of input instances and the learned abstract imaginations of target categories as topological relational graphs. Meanwhile, to capture and leverage the compositional contributions of visual semantics in a global view, we also introduce a universal Feat2Graph scheme in SchemaNet to establish the relational graphs that contain abundant interaction information. Both the theoretical analysis and the experimental results on several benchmarks demonstrate that the proposed schema inference achieves encouraging performance and meanwhile yields a clear picture of the deductive process leading to the predictions. Our code is available at https://github.com/zhfeing/SchemaNet-PyTorch.
['Mingli Song', 'Jie Song', 'KaiXuan Chen', 'Xiaokang Liu', 'Mengqi Xue', 'Haofei Zhang']
2023-03-12
null
null
null
null
['graph-matching']
['graphs']
[ 2.66238302e-01 7.28911459e-01 -1.45382240e-01 -7.06862867e-01 9.18891504e-02 -4.96859550e-01 1.06032336e+00 9.93496552e-02 2.94796795e-01 3.15375149e-01 4.56572741e-01 -4.26855624e-01 -4.04244542e-01 -1.21071863e+00 -1.12566185e+00 -3.47041994e-01 2.54236400e-01 4.56544250e-01 7.89457094e-03 -3.97054166e-01 3.91325921e-01 2.05194712e-01 -1.54715908e+00 7.13647306e-01 7.83812881e-01 1.05419099e+00 2.57645428e-01 2.82564908e-01 -3.12541604e-01 1.26166558e+00 1.25854090e-01 -1.05667031e+00 7.39719272e-02 -4.54471886e-01 -1.03979492e+00 1.75342653e-02 7.70045638e-01 -1.50066733e-01 -6.03687584e-01 1.33630526e+00 -2.35154822e-01 -5.67382202e-02 5.42723477e-01 -1.40046763e+00 -1.27649903e+00 8.71489286e-01 -2.11879686e-01 5.51016182e-02 3.02454352e-01 9.64364037e-02 1.42098033e+00 -8.67137134e-01 8.55975866e-01 1.48300290e+00 4.52919483e-01 4.47581887e-01 -1.20720470e+00 -5.46298921e-01 3.88298810e-01 5.35685718e-01 -1.27588236e+00 -1.60341993e-01 1.20707357e+00 -4.53133285e-01 5.11471629e-01 3.23478431e-01 9.10597980e-01 1.26809561e+00 1.20741509e-01 7.85265923e-01 1.21638906e+00 -4.78679612e-02 1.17197014e-01 2.02101529e-01 3.28462303e-01 1.04964089e+00 3.44848394e-01 3.96547526e-01 -6.84193373e-01 2.01402590e-01 9.75573659e-01 3.16123933e-01 -3.15775782e-01 -7.11370289e-01 -1.10304916e+00 8.09971869e-01 1.18936098e+00 9.61135626e-02 -3.96808416e-01 3.79552126e-01 2.78872013e-01 1.45181462e-01 2.47833818e-01 2.08763644e-01 -9.94546413e-02 6.97308242e-01 -5.99793732e-01 3.25910866e-01 7.03007162e-01 1.25394893e+00 9.44847524e-01 -2.40601912e-01 -5.76043315e-02 2.34218776e-01 5.86585224e-01 3.03207099e-01 4.23912816e-02 -1.12012017e+00 2.99562812e-01 1.11745656e+00 -3.14729810e-01 -1.37317848e+00 -1.79847330e-01 -5.47858298e-01 -1.02281284e+00 -3.84036601e-02 3.13179381e-02 5.08191347e-01 -7.83056676e-01 1.92912209e+00 3.15769434e-01 4.17948425e-01 1.68119133e-01 9.34208751e-01 1.08516550e+00 4.60572511e-01 3.36098582e-01 8.07244629e-02 1.66048551e+00 -7.96804190e-01 -6.03801787e-01 -1.26781970e-01 1.81814238e-01 -2.61968732e-01 1.21823931e+00 8.85501802e-02 -1.18886173e+00 -7.85484970e-01 -1.10140884e+00 -4.93233174e-01 -6.58537686e-01 -2.50649631e-01 9.25430775e-01 1.07804239e-01 -1.15337622e+00 6.23801470e-01 -4.26664650e-01 -5.61451733e-01 8.51528823e-01 1.23047102e-02 -2.62484878e-01 -1.05116002e-01 -1.31655335e+00 7.35051513e-01 7.94899225e-01 1.54947922e-01 -9.19825137e-01 -9.12545443e-01 -8.11897695e-01 1.99090913e-01 6.37121618e-01 -1.43979645e+00 9.01125669e-01 -1.16816020e+00 -9.99358237e-01 1.25078487e+00 -1.00113861e-01 -5.79301536e-01 4.29006159e-01 -2.20730938e-02 -1.91648364e-01 1.44030586e-01 -1.18290931e-01 7.98156500e-01 5.64346492e-01 -1.74182498e+00 -3.71646971e-01 -4.73147631e-01 7.26460457e-01 3.13952714e-01 2.12439485e-02 -4.10243899e-01 -3.43885750e-01 -5.25286555e-01 2.86159158e-01 -5.12746632e-01 1.63519345e-02 2.88489699e-01 -6.50236070e-01 -2.75871128e-01 4.86167520e-01 -3.69252890e-01 1.02514303e+00 -1.94826233e+00 4.41868156e-01 2.97756225e-01 8.81148994e-01 -1.44648075e-01 1.33355960e-01 5.94121277e-01 -1.48546726e-01 3.94855142e-02 -2.97788084e-01 -1.56088203e-01 3.32387477e-01 3.47497225e-01 -9.02688265e-01 1.13980241e-01 -4.10464033e-02 1.40023720e+00 -9.63336825e-01 -5.09981394e-01 1.51723042e-01 3.49724174e-01 -7.14423537e-01 3.48608673e-01 -5.70805430e-01 2.50373304e-01 -5.37459552e-01 2.91381776e-01 7.72616625e-01 -7.45117188e-01 5.00365555e-01 -7.41499662e-01 1.65278107e-01 3.68885219e-01 -8.07258546e-01 2.06104994e+00 -8.62684920e-02 2.57247537e-01 -2.01509818e-01 -1.26353049e+00 9.90660250e-01 -4.89844754e-02 -1.23813860e-01 -7.65404284e-01 8.40317011e-02 -2.20564038e-01 -2.25650981e-01 -4.70003426e-01 3.02583694e-01 -3.77973080e-01 -2.52770670e-02 3.97359997e-01 2.43065834e-01 -6.46762252e-02 -2.26500124e-01 8.51563096e-01 6.32434666e-01 4.13770020e-01 3.86118829e-01 -6.31348073e-01 5.30346632e-01 4.41685766e-02 2.48880982e-01 7.32089996e-01 2.71198936e-02 9.07777101e-02 6.74254894e-01 -8.05998147e-01 -1.12187803e+00 -1.63662255e+00 1.36187479e-01 8.99271429e-01 7.30842412e-01 -5.81583023e-01 -8.19923937e-01 -4.59277242e-01 5.11067063e-02 8.17834139e-01 -9.93730545e-01 -2.85236269e-01 -2.91035473e-01 -1.66424423e-01 3.78472298e-01 6.22780085e-01 7.11365879e-01 -1.19285882e+00 -3.47324640e-01 -2.39924058e-01 -1.73224807e-01 -9.68852162e-01 1.99687341e-03 -2.55873710e-01 -6.54355705e-01 -1.17885804e+00 1.18315488e-01 -7.15262651e-01 7.16343641e-01 1.63288265e-01 1.37782896e+00 5.95990181e-01 -5.66266440e-02 4.09474075e-01 -8.77174363e-02 -3.06866288e-01 -2.45175645e-01 -2.72297025e-01 -2.69381225e-01 2.05113724e-01 3.81778628e-01 -9.23000216e-01 -8.43206346e-01 -7.99742937e-02 -9.56497371e-01 1.06978726e+00 5.62575281e-01 6.20631933e-01 6.74827158e-01 -5.73087074e-02 2.75212467e-01 -1.35545087e+00 5.13280988e-01 -7.97710240e-01 -4.99622494e-01 4.96474117e-01 -7.00847387e-01 3.03769082e-01 6.85092449e-01 4.06602509e-02 -1.47078955e+00 -3.52102220e-01 1.34170413e-01 -5.75873494e-01 -3.13564569e-01 6.62357032e-01 -3.76027703e-01 2.75948822e-01 3.90681148e-01 4.35356468e-01 -8.04505870e-02 -3.32756937e-01 9.59451020e-01 5.66961318e-02 8.51183772e-01 -7.74632394e-01 9.78161037e-01 9.75589395e-01 2.48747543e-01 -1.04930729e-01 -1.34628701e+00 6.58247545e-02 -7.35464573e-01 -2.00207502e-01 9.78694737e-01 -8.37759256e-01 -1.22634602e+00 -1.37164793e-03 -1.25862110e+00 -2.01915905e-01 -1.27905309e-01 -2.41904780e-02 -9.25271988e-01 3.21166039e-01 -5.72183609e-01 -4.46369678e-01 -3.80765498e-01 -7.89489925e-01 8.94613385e-01 2.62536883e-01 -2.19606422e-02 -1.24455953e+00 -1.51927933e-01 4.77324635e-01 3.33042622e-01 4.30153698e-01 1.40904975e+00 -5.60096145e-01 -1.24455905e+00 5.07734835e-01 -8.53381336e-01 -1.21016935e-01 -2.95922428e-01 -3.10333878e-01 -1.05407715e+00 -9.48050898e-03 4.44815354e-03 -2.06760973e-01 9.83937383e-01 6.56969193e-03 1.43420637e+00 -4.38366175e-01 -3.29790831e-01 8.57384920e-01 1.77757299e+00 -9.72746089e-02 7.58903265e-01 7.45844990e-02 9.31362152e-01 8.63062799e-01 2.60804296e-01 2.32866690e-01 8.88822973e-01 3.21199417e-01 7.09239304e-01 -5.89921959e-02 -2.44340554e-01 -1.05038404e+00 -3.47970836e-02 7.69819677e-01 -2.47375950e-01 -7.19570890e-02 -8.31112385e-01 1.76090375e-01 -2.11636829e+00 -1.28668928e+00 -1.27722666e-01 1.65542400e+00 6.34520888e-01 1.92030028e-01 -3.10792685e-01 -3.82992268e-01 7.26537108e-01 2.48352781e-01 -5.71768641e-01 -3.13122034e-01 -5.85987717e-02 2.18233868e-01 2.60238331e-02 5.40815175e-01 -6.47957921e-01 1.20345330e+00 5.53008223e+00 6.19479477e-01 -7.32181668e-01 8.79058689e-02 5.24101436e-01 1.33930460e-01 -9.51249897e-01 4.53529805e-01 -4.74043459e-01 1.72275379e-01 7.11853623e-01 -4.32294965e-01 7.58806229e-01 5.96442640e-01 1.07693458e-02 3.57975721e-01 -1.31754029e+00 8.50237787e-01 7.58843124e-02 -1.91203940e+00 7.94440925e-01 5.00500575e-03 3.81817073e-01 -3.26774836e-01 1.08507194e-01 2.48535916e-01 4.60401624e-01 -1.08267963e+00 1.10574937e+00 1.13671756e+00 5.84609568e-01 -5.84875107e-01 2.51184732e-01 2.92733520e-01 -1.29824746e+00 -5.84867708e-02 -6.10156655e-01 -2.93533683e-01 6.24242760e-02 2.74932802e-01 -4.54668671e-01 9.01547492e-01 6.04511797e-01 1.02944028e+00 -6.16032481e-01 4.66599524e-01 -5.71391284e-01 3.57666761e-01 2.80539691e-01 1.37880787e-01 1.90516487e-01 -3.40499848e-01 2.69057870e-01 8.45922709e-01 8.08684751e-02 4.29933608e-01 -1.59523726e-01 1.83975899e+00 -3.60982925e-01 -1.08074844e-01 -7.93210328e-01 1.50669396e-01 5.41986942e-01 1.27030265e+00 -6.02465212e-01 -6.48961544e-01 -3.86046082e-01 7.32658803e-01 8.91394138e-01 4.26924884e-01 -1.04865670e+00 9.29213837e-02 5.95189035e-01 1.01159334e-01 2.37618268e-01 5.81506863e-02 -4.43362623e-01 -1.29277968e+00 -1.30776867e-01 -5.57223201e-01 4.16616648e-01 -1.48393214e+00 -1.45441186e+00 5.68145275e-01 2.14363858e-01 -8.40121865e-01 2.12477386e-01 -8.22499573e-01 -9.08237398e-01 7.52789378e-01 -1.54317069e+00 -1.74118102e+00 -7.46000350e-01 8.57366383e-01 2.22761199e-01 1.31834909e-01 6.78617954e-01 -1.39322579e-01 -2.46180534e-01 2.74057895e-01 -4.52313960e-01 5.81319891e-02 1.89077809e-01 -1.28921127e+00 5.75431228e-01 5.49026787e-01 2.75630593e-01 1.06686985e+00 6.80496991e-01 -5.59300900e-01 -1.56529927e+00 -1.08829534e+00 6.95508063e-01 -6.95795774e-01 9.95580912e-01 -6.72962844e-01 -1.14339244e+00 1.08018017e+00 4.60937023e-01 -2.18086943e-01 5.57097912e-01 9.82260630e-02 -8.63912642e-01 -4.47579920e-02 -8.60518992e-01 8.12670350e-01 1.85976112e+00 -8.79755974e-01 -1.08200610e+00 1.84284255e-01 1.08071876e+00 -2.47193128e-01 -9.73367274e-01 2.94944406e-01 5.01168966e-01 -1.23652518e+00 1.16019845e+00 -9.35078800e-01 8.48071456e-01 -3.63754004e-01 -3.79028559e-01 -9.01725352e-01 -6.28285527e-01 -1.35286212e-01 -9.57107097e-02 1.20471704e+00 2.12339625e-01 -6.01173639e-01 6.32984400e-01 5.00235021e-01 -1.44411936e-01 -8.39190960e-01 -5.23508489e-01 -3.01431745e-01 2.34021991e-02 -3.87629479e-01 9.68713105e-01 1.19828677e+00 7.76260346e-02 6.07565165e-01 -2.29446471e-01 3.44329715e-01 9.92900848e-01 6.98571086e-01 5.48805594e-01 -1.35015726e+00 -3.27893287e-01 -4.76637512e-01 -4.21102941e-01 -9.67830658e-01 4.75129396e-01 -1.47700155e+00 -4.74620908e-01 -1.68357563e+00 7.28117764e-01 -1.65654227e-01 -5.12445092e-01 3.29587877e-01 -2.40040999e-02 9.18441713e-02 3.51731062e-01 2.87812948e-01 -6.77049041e-01 4.61309850e-01 1.49793088e+00 -1.17750347e-01 4.97999281e-01 -5.17795384e-01 -1.26579785e+00 6.77569747e-01 6.09100103e-01 -7.02185258e-02 -8.51318479e-01 -5.02799571e-01 7.14352071e-01 9.97306257e-02 1.19184363e+00 -5.65389931e-01 3.20721269e-01 -4.31314617e-01 3.03113550e-01 -5.44599414e-01 2.38038242e-01 -8.26137066e-01 5.03563344e-01 2.76598573e-01 -7.12825656e-01 1.01178564e-01 -2.57669557e-02 7.53888011e-01 -1.02574363e-01 2.06784829e-01 5.54958761e-01 -3.12832087e-01 -1.11542010e+00 5.30294299e-01 4.02623922e-01 1.70458364e-03 8.07845533e-01 -9.23417583e-02 -8.02437782e-01 -2.32541561e-01 -8.23257864e-01 1.36566654e-01 5.36513388e-01 4.13625777e-01 7.87657857e-01 -1.56070828e+00 -3.95503640e-01 1.50598153e-01 4.36639875e-01 -2.04682872e-01 4.50139612e-01 7.52958238e-01 -3.40809733e-01 4.09860253e-01 -4.33206409e-01 -4.91349488e-01 -8.36904168e-01 1.11794221e+00 3.96635294e-01 -1.45304380e-02 -8.12430143e-01 8.34513068e-01 1.07574832e+00 -3.84250700e-01 -1.60891011e-01 -3.99496496e-01 -1.34197742e-01 -3.27732116e-01 3.18964213e-01 -4.73767482e-02 -4.75524932e-01 -6.96397543e-01 -1.10543452e-01 3.77664268e-01 3.52847576e-02 2.23406166e-01 1.17970073e+00 -2.65827924e-01 -5.54786801e-01 3.64056081e-01 1.06388664e+00 -4.40258592e-01 -1.18476439e+00 -4.57592696e-01 -2.40644477e-02 -3.89520705e-01 -7.68979564e-02 -7.81654954e-01 -1.04718113e+00 1.00985384e+00 1.57905266e-01 2.67059892e-01 1.08497608e+00 4.56458330e-01 4.98172611e-01 2.29275629e-01 4.04072702e-01 -4.67727065e-01 1.35972381e-01 1.76329926e-01 1.04998791e+00 -1.00494981e+00 -1.14392690e-01 -6.38141215e-01 -6.60403907e-01 1.05001557e+00 7.25413978e-01 -3.69438857e-01 6.60586655e-01 -2.06409261e-01 -2.89606601e-01 -9.45221186e-01 -1.05781674e+00 -1.94802791e-01 5.34852445e-01 4.15648669e-01 1.39758095e-01 3.73471737e-01 -9.75204408e-02 8.15386176e-01 -4.39976513e-01 5.15799187e-02 2.19824895e-01 4.37156171e-01 -4.50325847e-01 -6.63042188e-01 1.89835310e-01 3.09055984e-01 1.52116850e-01 -3.26323181e-01 -5.91880620e-01 8.45263362e-01 3.60864431e-01 3.65423739e-01 2.41169393e-01 -4.48151499e-01 3.05424124e-01 -3.96209396e-02 5.77326775e-01 -4.53640908e-01 -2.35300422e-01 -5.31446636e-01 -1.48862138e-01 -8.55048776e-01 -5.32680035e-01 -1.19941816e-01 -1.45097280e+00 -7.41989255e-01 1.26599580e-01 -5.16637228e-02 2.71440178e-01 1.08388019e+00 4.14420933e-01 4.36038703e-01 1.43234894e-01 -4.77360934e-01 -1.90286890e-01 -5.27973115e-01 -4.95592654e-01 1.05717659e+00 4.17350605e-02 -8.74726474e-01 -1.85194612e-01 2.76007771e-01]
[10.55648422241211, 1.7776635885238647]
b4e212e5-9f29-4ec4-a4e8-c2aaf25890c9
toward-automatic-misinformation-detection
null
null
https://openreview.net/forum?id=bzqte0FZQH
https://openreview.net/pdf?id=bzqte0FZQH
Toward Automatic Misinformation Detection Utilizing Fact-checked Information
We proposed a new task FCCKB: Fact-checking by Claim Knowledge Base. The goal was to fact-check a sentence utilizing verified claims stored in the database. To retrieve relevant claims from the large database, we proposed applying Semantic Role Labeling(SRL) on the input sentence having rich semantics and then encoding the results to get fine-grained sentence embeddings. That improved semantic matching between the input sentence and the relevant claims. We used three sentence encoders for sentence encoding. In FEVER dataset, precision and recall was improved by more than 5 percent after SRL was applied.
['Anonymous']
2022-01-16
null
null
null
acl-arr-january-2022-1
['semantic-role-labeling']
['natural-language-processing']
[ 4.98711795e-01 6.01493061e-01 -5.88614643e-01 -5.52110851e-01 -1.42142117e+00 -4.90421683e-01 4.49412197e-01 9.28363979e-01 -6.92072570e-01 1.15267789e+00 9.46847856e-01 -8.81731585e-02 -3.48599941e-01 -9.76292312e-01 -1.05613613e+00 2.19316617e-01 2.50577539e-01 3.62160951e-01 7.79224634e-01 -6.22921705e-01 4.55447286e-01 1.13801420e-01 -1.50043154e+00 9.50557172e-01 5.95312595e-01 1.40300226e+00 2.15372279e-01 4.74032372e-01 -2.07963645e-01 1.77781427e+00 -9.64423120e-01 -1.13243902e+00 -1.51920095e-01 -1.23065256e-01 -1.41424239e+00 -2.74009585e-01 3.01976919e-01 2.23188430e-01 -5.84537685e-01 1.34671152e+00 2.07073539e-01 1.00718006e-01 2.62837738e-01 -8.15296471e-01 -1.73741007e+00 1.05604768e+00 8.86050537e-02 7.08195925e-01 7.33797908e-01 -6.81205034e-01 1.20648634e+00 -6.58084929e-01 1.02547061e+00 1.29279542e+00 4.89857435e-01 6.22066200e-01 -6.25327170e-01 -3.94794077e-01 -3.46766382e-01 6.93546772e-01 -1.08889282e+00 -4.67173040e-01 8.63106966e-01 -2.30372548e-01 1.35688448e+00 2.94266433e-01 1.95326611e-01 1.04458785e+00 2.36124009e-01 7.16626823e-01 7.57880926e-01 -4.07532901e-01 1.05279408e-01 3.43778521e-01 7.63633072e-01 8.39940131e-01 5.59932888e-01 -2.96025425e-01 -6.54833853e-01 -6.06681705e-01 2.42981642e-01 -1.27433047e-01 -9.64989811e-02 3.70456815e-01 -9.64252293e-01 1.19679475e+00 5.76827526e-01 3.84135574e-01 -4.69303071e-01 1.78321347e-01 1.16151989e+00 4.47502792e-01 6.06696188e-01 7.86158562e-01 -8.09158266e-01 1.25030145e-01 -5.37485540e-01 2.82723516e-01 5.85025847e-01 6.81920767e-01 2.55540431e-01 -4.60683852e-01 -7.62051105e-01 8.79859507e-01 -9.49041173e-02 6.67238772e-01 6.32283568e-01 -8.76497924e-01 9.69293892e-01 1.24104106e+00 1.46083936e-01 -1.40104401e+00 -1.34486139e-01 -6.12755343e-02 -4.56317037e-01 -4.49591339e-01 -3.15608919e-01 2.63815165e-01 -8.51512849e-01 1.42693341e+00 5.02892174e-02 6.14328571e-02 6.75679445e-01 6.82149053e-01 9.32632983e-01 2.79337227e-01 2.48786479e-01 -2.29801033e-02 1.84547889e+00 -7.06444204e-01 -1.06558311e+00 -2.05352921e-02 7.17185080e-01 -4.89967018e-01 1.21714389e+00 -3.46187830e-01 -8.20729017e-01 -4.09593821e-01 -1.36629736e+00 -4.48832661e-01 -6.46739066e-01 7.63281137e-02 5.55234671e-01 4.49196517e-01 -6.38617456e-01 3.92157614e-01 -3.51228714e-01 1.66213959e-01 8.33024561e-01 2.23563146e-02 -6.25119328e-01 -1.39290825e-01 -2.07775235e+00 1.00661302e+00 8.70011687e-01 -4.55462873e-01 -5.27212679e-01 -8.05640221e-01 -1.47312748e+00 1.69362575e-01 8.05104733e-01 -6.36435449e-01 1.21776581e+00 -3.95707607e-01 -8.89029801e-01 9.78589177e-01 -3.28663766e-01 -1.17491484e+00 -2.58554816e-01 -1.74720109e-01 -1.25165856e+00 6.53439641e-01 8.06326807e-01 -1.59525305e-01 5.28328955e-01 -8.97143483e-01 -2.42556036e-01 -3.98308307e-01 3.74617726e-01 -1.44927487e-01 -8.62553194e-02 6.04400516e-01 7.01136962e-02 -8.63286853e-01 -2.29368791e-01 -2.27626145e-01 2.16831446e-01 -5.39660454e-01 -3.27753454e-01 -6.61136329e-01 8.65388691e-01 -1.04598296e+00 1.32265317e+00 -2.12736034e+00 -6.60554945e-01 -2.04487279e-01 -7.28265643e-02 2.98035026e-01 1.49934432e-02 3.11576933e-01 4.07351777e-02 3.06134760e-01 -5.47887266e-01 3.79823335e-02 -1.02790110e-01 3.54413480e-01 -8.19621742e-01 -6.10658061e-03 6.38955355e-01 1.25913537e+00 -1.07055974e+00 -7.65464485e-01 -3.55288237e-01 2.22211063e-01 -4.82859761e-01 -1.47140846e-01 -3.96028429e-01 -3.38331163e-01 -5.38176835e-01 6.36829555e-01 3.28582704e-01 -4.77974743e-01 -1.70470566e-01 -3.41344982e-01 7.70501375e-01 8.80705118e-01 -6.03257179e-01 1.78201783e+00 -5.33083677e-01 3.57349128e-01 -4.38745260e-01 -8.24585378e-01 9.90107238e-01 5.22004247e-01 2.77481407e-01 -1.21389306e+00 -5.48576415e-02 1.50395468e-01 -5.12500465e-01 -7.63718307e-01 8.25445890e-01 -4.40666795e-01 -7.35906839e-01 3.89999747e-01 4.68756966e-02 2.34902909e-05 1.23285942e-01 5.82545042e-01 1.34308958e+00 -5.79517543e-01 1.83373794e-01 -6.45532981e-02 9.35601473e-01 7.06088066e-01 3.10590446e-01 5.47163725e-01 5.49612418e-02 2.27212504e-01 5.83052337e-01 -3.97893131e-01 -8.88072371e-01 -8.71851265e-01 -1.69764161e-01 8.24509442e-01 1.59541398e-01 -4.40693706e-01 -4.16651189e-01 -1.18022311e+00 4.23596799e-01 8.45005274e-01 -8.35832059e-01 -4.70126003e-01 -5.06102622e-01 -4.04354930e-01 1.05330265e+00 6.60371065e-01 7.73587346e-01 -1.21684313e+00 -5.00245810e-01 2.84886003e-01 -2.54123241e-01 -1.38847995e+00 -4.68964010e-01 -1.97508596e-02 -3.94219995e-01 -1.78587258e+00 1.75026238e-01 -8.57688427e-01 3.60347509e-01 -1.89997062e-01 1.16858089e+00 9.20074731e-02 4.16983180e-02 -2.33475253e-01 -5.90223014e-01 -2.75888324e-01 -5.63276649e-01 -3.64191085e-01 1.22176567e-02 -1.18631534e-01 3.50276947e-01 2.69492656e-01 -2.13575691e-01 -3.66892278e-01 -1.10318851e+00 -4.88203377e-01 2.07529396e-01 9.72143650e-01 5.40371239e-01 3.53400916e-01 1.04950857e+00 -1.25701404e+00 1.30680490e+00 -4.42398846e-01 -1.68047786e-01 6.68118417e-01 -5.62410235e-01 4.85024691e-01 5.13546407e-01 1.80996656e-02 -1.14563119e+00 -4.94381219e-01 -3.79257530e-01 -2.88214117e-01 3.99842203e-01 7.16087878e-01 3.09167616e-02 5.13931096e-01 6.92728996e-01 3.58913019e-02 -1.60319179e-01 -3.97829264e-01 4.26842123e-01 8.51447225e-01 8.38091075e-01 -5.58816552e-01 1.60896048e-01 3.42881441e-01 -2.20200300e-01 -6.47662133e-02 -1.78481567e+00 -2.96450406e-01 -2.65336484e-01 4.55189854e-01 1.07339883e+00 -8.45598280e-01 -7.82568038e-01 -3.01884562e-01 -1.41605020e+00 3.12720448e-01 -7.50982702e-01 2.19759881e-01 -2.63302237e-01 2.90231794e-01 -7.93021083e-01 -7.00539827e-01 -6.33391857e-01 -6.92243516e-01 1.21351480e+00 -3.45580846e-01 -7.38156065e-02 -8.40114176e-01 1.02684766e-01 8.29691112e-01 3.04619130e-02 1.29269645e-01 1.18743157e+00 -1.08876598e+00 1.29427537e-01 -5.96250355e-01 -5.98496854e-01 5.61986744e-01 2.62096196e-01 -8.19891334e-01 -8.26404691e-01 2.28638634e-01 2.67817348e-01 -4.82847363e-01 1.11844301e+00 -1.87056422e-01 1.27026820e+00 -7.69586444e-01 -5.00740930e-02 -1.86077178e-01 1.53413475e+00 1.67113945e-01 4.60020363e-01 4.05210733e-01 5.02954066e-01 2.09885910e-01 7.83537090e-01 -5.25969714e-02 4.42881525e-01 6.66465640e-01 6.60585091e-02 5.06348193e-01 -3.56755316e-01 -7.38870800e-01 2.72755563e-01 5.98032832e-01 3.83000791e-01 5.97596653e-02 -7.23378718e-01 7.23537326e-01 -1.87999427e+00 -1.36553133e+00 -5.80720901e-02 1.54373252e+00 1.23410666e+00 5.00506938e-01 -2.56942779e-01 4.09339577e-01 7.41443336e-01 1.43117949e-01 -3.96429688e-01 -4.70723093e-01 -5.59370160e-01 5.10018229e-01 7.62327671e-01 7.60549128e-01 -1.08338022e+00 1.03303552e+00 6.43452692e+00 9.77552772e-01 -6.22428238e-01 9.24893379e-01 3.63469273e-01 2.21506506e-02 -7.34511554e-01 -1.12860091e-01 -7.01927602e-01 6.85974002e-01 1.14210808e+00 -3.30041856e-01 -1.28156334e-01 8.48176241e-01 -1.28227651e-01 -7.87636787e-02 -4.39939260e-01 8.74823868e-01 5.75533152e-01 -2.20913482e+00 4.03620750e-01 -5.19138753e-01 5.63939333e-01 -1.33830622e-01 -3.99802923e-01 6.21696174e-01 3.53134930e-01 -9.25510108e-01 9.51028883e-01 5.92508852e-01 9.24072206e-01 -7.41200864e-01 1.31833816e+00 9.65954363e-02 -8.77569020e-01 -3.53409469e-01 -5.90081215e-01 -7.86330625e-02 3.27023655e-01 9.40768242e-01 -9.22562540e-01 8.25984240e-01 4.54480648e-01 6.43003345e-01 -8.73738050e-01 2.82023400e-01 -5.24819970e-01 4.46880788e-01 2.63785988e-01 -1.64522342e-02 1.15956903e-01 5.35888851e-01 1.54026955e-01 1.10031629e+00 -4.27516326e-02 1.25991210e-01 1.55735359e-01 6.97419107e-01 -6.61419094e-01 -4.49146964e-02 -6.36565745e-01 -2.83270806e-01 5.36205411e-01 4.92741376e-01 -4.36794728e-01 -8.33262384e-01 -1.63846716e-01 1.13307452e+00 4.84772176e-01 -1.64857149e-01 -8.66559148e-01 -4.94808942e-01 5.30756652e-01 7.66264573e-02 1.44236833e-01 2.27279976e-01 -4.71756727e-01 -1.20296848e+00 3.92880291e-01 -4.09496754e-01 9.41383779e-01 -1.00481093e+00 -1.41580760e+00 8.22606087e-01 -8.97830650e-02 -5.48706651e-01 1.00874916e-01 -5.12317061e-01 1.23387046e-01 8.30946267e-01 -1.52567530e+00 -1.17592120e+00 5.26545703e-01 7.80999899e-01 6.11357927e-01 -3.52308303e-01 9.85487521e-01 5.08269846e-01 -1.21178031e-01 4.51359510e-01 -6.33883715e-01 5.90449393e-01 4.23092663e-01 -1.12020874e+00 4.46242809e-01 7.78512239e-01 3.59327972e-01 7.72082329e-01 3.31366956e-01 -1.20199418e+00 -1.05599225e+00 -1.31765246e+00 1.72833741e+00 -1.09417403e+00 1.11526740e+00 -1.94805309e-01 -1.00382531e+00 4.06656682e-01 1.57634616e-01 1.19361430e-01 6.26336396e-01 -1.50981084e-01 -8.31207931e-01 1.91110827e-03 -1.56605804e+00 -2.28948835e-02 1.02483714e+00 -1.49925101e+00 -1.50448918e+00 3.90840948e-01 1.76596308e+00 -3.08772504e-01 -8.82194936e-01 3.75239998e-01 -1.36145085e-01 -4.73757274e-02 1.02436996e+00 -1.23193479e+00 6.29243374e-01 -4.47602361e-01 -6.24098539e-01 -9.23325837e-01 -1.85093224e-01 1.88844800e-01 -6.16552472e-01 9.11167741e-01 8.57580662e-01 -3.85233730e-01 3.59644175e-01 4.24169332e-01 -2.65928954e-01 -7.55228639e-01 -1.05578732e+00 -8.79357338e-01 -3.48144799e-01 -7.26908326e-01 1.01617837e+00 9.08425033e-01 3.78540695e-01 6.64030790e-01 -6.54331222e-02 4.19200182e-01 3.08154583e-01 2.74461359e-01 -3.35388303e-01 -9.96518016e-01 -1.06578097e-01 7.27201551e-02 -5.95876634e-01 -3.51693720e-01 6.48423433e-01 -1.50777030e+00 -3.78912717e-01 -1.75897956e+00 4.84895855e-01 -1.88700989e-01 -5.20022035e-01 7.46702671e-01 -4.62196022e-01 3.78371298e-01 3.73608693e-02 -2.95066517e-02 -7.16205776e-01 4.23782706e-01 1.16543889e+00 -5.91132641e-01 3.37006390e-01 -2.39653274e-01 -1.31255484e+00 2.34481364e-01 8.99518013e-01 -8.83794606e-01 -5.32103121e-01 -4.97970104e-01 8.46927464e-01 1.94247544e-01 7.48413861e-01 -8.06438625e-01 2.23437756e-01 5.98075055e-02 3.26994732e-02 -4.75239664e-01 3.28254044e-01 -8.14432025e-01 -2.48775646e-01 4.05886948e-01 -7.61993170e-01 1.38194874e-01 1.41261399e-01 8.04588437e-01 -6.52557433e-01 -4.37978119e-01 3.16272348e-01 -9.43871066e-02 -1.01603615e+00 -1.28466547e-01 2.31788814e-01 4.46669132e-01 7.76209712e-01 2.10866511e-01 -8.54986489e-01 1.06394172e-01 -8.68303120e-01 -2.81453338e-02 1.25168666e-01 6.71333432e-01 9.91579235e-01 -1.57066786e+00 -6.49115503e-01 2.19535753e-01 5.18527269e-01 -3.52463365e-01 1.36740133e-01 4.79078561e-01 -2.81320214e-01 6.20719671e-01 1.47738263e-01 2.57740945e-01 -1.19640386e+00 1.08619690e+00 -6.08288832e-02 -3.92774165e-01 -5.38727880e-01 9.67854023e-01 -8.72274935e-01 -3.37184638e-01 -2.23087952e-01 -5.91173351e-01 -4.39387679e-01 -1.07085984e-02 8.47989976e-01 9.37975198e-02 5.62880516e-01 -6.99403644e-01 -6.66338980e-01 8.96263868e-02 1.15856722e-01 -5.33740111e-02 1.33454585e+00 2.25699767e-01 -3.90143842e-01 2.07914352e-01 1.29197383e+00 3.00437957e-01 -3.57849061e-01 -2.76079208e-01 7.30088174e-01 -5.07027268e-01 9.75214019e-02 -1.02666736e+00 -7.54771113e-01 6.45052940e-02 9.79627296e-02 3.99818450e-01 8.02289009e-01 6.61963642e-01 1.15922785e+00 6.04729354e-01 6.51816905e-01 -1.24555695e+00 -1.95490406e-03 8.72263253e-01 1.16895843e+00 -1.21773744e+00 -2.18754888e-01 -5.93340993e-01 -1.14738572e+00 6.35344803e-01 7.97374845e-02 -2.97721893e-01 7.37724423e-01 7.60042444e-02 -1.55079812e-01 -1.08397114e+00 -6.93171442e-01 -3.68994921e-01 4.59557414e-01 3.21708173e-01 1.17323264e-01 2.19874725e-01 -4.76589143e-01 1.21497047e+00 -4.46018279e-01 7.32484683e-02 3.73179406e-01 9.33658183e-01 -4.73441184e-01 -9.65405047e-01 -9.95269418e-02 8.48082244e-01 -8.10836792e-01 -3.40597093e-01 -5.56636155e-01 2.72970259e-01 3.37853462e-01 1.20461774e+00 -1.21138453e-01 -5.58938682e-01 7.29633629e-01 4.37247872e-01 2.83066571e-01 -1.29930604e+00 -6.82141900e-01 -1.11582708e+00 8.04155648e-01 -4.97344971e-01 -6.85993254e-01 -2.76558965e-01 -1.56430531e+00 4.48316708e-02 -6.13671653e-02 5.86451411e-01 3.83712441e-01 1.03848052e+00 5.83320320e-01 6.18948996e-01 3.70984554e-01 4.76436377e-01 -5.65463305e-01 -9.32983994e-01 -5.48026800e-01 9.17047679e-01 3.95419776e-01 -7.39886045e-01 9.04380064e-03 9.73063260e-02]
[9.741408348083496, 8.47488784790039]
3a78a030-58b0-4a1e-b895-0eb23550e178
predicting-classification-accuracy-when-1
2010.15011
null
https://arxiv.org/abs/2010.15011v3
https://arxiv.org/pdf/2010.15011v3.pdf
Predicting Classification Accuracy When Adding New Unobserved Classes
Multiclass classifiers are often designed and evaluated only on a sample from the classes on which they will eventually be applied. Hence, their final accuracy remains unknown. In this work we study how a classifier's performance over the initial class sample can be used to extrapolate its expected accuracy on a larger, unobserved set of classes. For this, we define a measure of separation between correct and incorrect classes that is independent of the number of classes: the "reversed ROC" (rROC), which is obtained by replacing the roles of classes and data-points in the common ROC. We show that the classification accuracy is a function of the rROC in multiclass classifiers, for which the learned representation of data from the initial class sample remains unchanged when new classes are added. Using these results we formulate a robust neural-network-based algorithm, "CleaneX", which learns to estimate the accuracy of such classifiers on arbitrarily large sets of classes. Unlike previous methods, our method uses both the observed accuracies of the classifier and densities of classification scores, and therefore achieves remarkably better predictions than current state-of-the-art methods on both simulations and real datasets of object detection, face recognition, and brain decoding.
['Yuval Benjamini', 'Yuli Slavutsky']
2020-10-28
predicting-classification-accuracy-when
https://openreview.net/forum?id=Y9McSeEaqUh
https://openreview.net/pdf?id=Y9McSeEaqUh
iclr-2021-1
['brain-decoding', 'brain-decoding']
['medical', 'miscellaneous']
[ 5.99579036e-01 4.84630093e-02 -3.32754791e-01 -8.95335138e-01 -6.36330903e-01 -4.99902517e-01 5.75149715e-01 2.01828182e-01 -4.85753238e-01 9.16374147e-01 -3.89568716e-01 -1.47401065e-01 -1.62481919e-01 -5.28105021e-01 -8.91815662e-01 -8.82298827e-01 2.10623555e-02 5.47366679e-01 3.64122659e-01 3.18922520e-01 1.45745710e-01 4.50565875e-01 -1.93603039e+00 7.69866586e-01 6.16686642e-01 1.20621681e+00 -1.21306479e-01 7.91235626e-01 2.40412891e-01 5.42497694e-01 -8.31856489e-01 -4.37129050e-01 1.72232054e-02 -6.50676310e-01 -5.14799356e-01 -1.76817194e-01 5.43965220e-01 -2.65755355e-01 -2.46086344e-01 1.19825017e+00 8.33612159e-02 -1.16296887e-01 1.19653213e+00 -1.39311457e+00 -4.13180053e-01 5.66736758e-01 -1.32249236e-01 9.70744342e-02 -3.68655892e-03 -1.85463458e-01 8.50302219e-01 -8.68004203e-01 1.77040979e-01 1.08726752e+00 8.43958020e-01 7.65694261e-01 -1.52071571e+00 -9.33235765e-01 1.17703237e-01 7.74798170e-02 -1.44577491e+00 -5.08254707e-01 4.63801950e-01 -6.51103020e-01 4.86378580e-01 2.72350550e-01 5.15853107e-01 1.20088542e+00 4.59702611e-01 6.31603181e-01 1.20038998e+00 -5.86845756e-01 5.55300653e-01 5.89670479e-01 5.23699522e-01 3.57264459e-01 4.98576581e-01 4.18118089e-01 -3.10984254e-01 -3.52671206e-01 3.37302089e-01 8.79204720e-02 -2.61535406e-01 -4.74587113e-01 -7.43941665e-01 7.56359518e-01 1.86011136e-01 1.93454236e-01 -1.93559706e-01 2.95275543e-02 2.54701208e-02 2.40792096e-01 4.73402560e-01 2.72973124e-02 -6.58940971e-01 2.07299218e-01 -9.85350788e-01 1.68650210e-01 7.70738721e-01 8.34285796e-01 5.43839753e-01 -2.35996798e-01 -1.28766522e-01 9.47696805e-01 -6.46832632e-03 4.56019133e-01 7.06833303e-01 -6.69816792e-01 3.33237089e-02 5.43162346e-01 1.11445725e-01 -3.38825047e-01 -5.02601027e-01 -8.21405232e-01 -6.08476460e-01 2.13176534e-01 6.55073225e-01 -9.02471691e-02 -1.07517219e+00 2.04063487e+00 6.21037930e-02 1.50541008e-01 2.05527216e-01 3.69129330e-01 3.74158710e-01 3.65242392e-01 1.24031091e-02 -2.00174645e-01 8.75750124e-01 -3.34296614e-01 -3.61265719e-01 -2.96306789e-01 6.24675512e-01 -2.28864208e-01 6.30063713e-01 6.38672888e-01 -8.07375610e-01 -5.04810572e-01 -1.20243371e+00 6.06996655e-01 -4.23947811e-01 3.89330685e-01 4.63484854e-01 9.99058723e-01 -8.23321462e-01 8.84566009e-01 -6.89384937e-01 -1.52671948e-01 6.23720765e-01 4.89567012e-01 -2.47203290e-01 -1.14443369e-01 -8.26070964e-01 8.91213238e-01 4.57883477e-01 7.87334070e-02 -9.80208039e-01 -4.55640435e-01 -3.80322069e-01 1.48432299e-01 -6.68045208e-02 -1.93588033e-01 1.33944046e+00 -1.48899329e+00 -1.05156922e+00 6.74682260e-01 -3.60280812e-01 -4.17809278e-01 3.96448582e-01 9.30376723e-02 -4.42464173e-01 -7.65737221e-02 3.18136578e-03 6.52891576e-01 1.01064920e+00 -1.36411989e+00 -9.17655945e-01 -6.27518237e-01 -2.91016489e-01 -2.03829497e-01 -1.60916120e-01 -2.35889837e-01 1.07971713e-01 -3.29418838e-01 3.05547863e-01 -9.80905354e-01 1.67533860e-01 9.91584882e-02 -3.30611944e-01 -2.10085809e-01 5.00353634e-01 -1.99471250e-01 9.56167340e-01 -2.15378547e+00 2.88682301e-02 3.30951005e-01 1.32180348e-01 9.85624716e-02 -1.50563478e-01 -4.23221663e-02 -4.11631405e-01 -1.30107120e-01 -4.50121850e-01 -4.87900451e-02 -3.31429422e-01 2.04725593e-01 -3.45671326e-01 6.24470353e-01 4.97628003e-01 4.16367322e-01 -6.28316760e-01 -6.06660917e-02 -1.76461950e-01 2.96109647e-01 -6.29461586e-01 2.64003556e-02 1.98342308e-01 2.62850463e-01 -1.59105852e-01 5.04995048e-01 7.58429229e-01 -3.40699673e-01 3.58157754e-01 -2.97937300e-02 2.58078367e-01 1.04256952e-02 -1.12247789e+00 7.75969565e-01 -1.86368793e-01 6.59998775e-01 -4.63127255e-01 -1.38743341e+00 9.42087352e-01 2.27388173e-01 1.18855402e-01 -2.79049546e-01 3.39612186e-01 2.48214945e-01 5.36420941e-01 -1.21087059e-01 -9.22602508e-03 -2.99091458e-01 2.79600602e-02 2.35278532e-01 4.85779494e-01 1.74520254e-01 -1.21032208e-01 -1.66087821e-01 1.06801581e+00 -1.05791554e-01 3.41326118e-01 -3.21161360e-01 2.93085098e-01 -2.31909081e-01 6.94105685e-01 1.33153319e+00 -3.23123544e-01 4.67799515e-01 7.23380089e-01 -4.31701422e-01 -1.00942087e+00 -1.31392658e+00 -1.02464211e+00 1.04720855e+00 -1.31629318e-01 2.12361366e-01 -9.39066410e-01 -8.67848694e-01 1.46570429e-01 9.67090607e-01 -9.66534078e-01 -7.23711133e-01 -1.53506860e-01 -1.20839107e+00 4.71534133e-01 5.82080543e-01 6.99594384e-04 -8.24096203e-01 -3.92299712e-01 1.35989949e-01 1.79058760e-01 -9.68788385e-01 1.71549141e-01 6.34065568e-01 -9.19511616e-01 -1.37122166e+00 -4.69420373e-01 -6.44545436e-01 7.75029302e-01 -1.53642505e-01 7.67604291e-01 3.85558717e-02 -7.12518543e-02 3.33290100e-01 -2.25016177e-01 -7.02156305e-01 -6.39125645e-01 -1.00631870e-01 3.18399012e-01 4.60899234e-01 5.58480978e-01 -4.37113941e-01 -2.97025472e-01 4.37913388e-01 -7.42660522e-01 -1.59254014e-01 5.36961555e-01 1.12672758e+00 3.43804032e-01 1.36503652e-01 1.18219650e+00 -1.08373463e+00 3.00364137e-01 -5.51538050e-01 -5.99029839e-01 5.79069257e-01 -6.56641126e-01 1.77117079e-01 7.44555593e-01 -9.74519491e-01 -7.52665997e-01 2.14855164e-01 1.34217456e-01 -1.79818690e-01 -1.75747171e-01 1.34666875e-01 3.76473032e-02 2.27034241e-02 8.28202307e-01 2.81732410e-01 -1.53415352e-01 -3.71508300e-01 -2.95225605e-02 1.14497721e+00 5.53981960e-01 -5.06284475e-01 3.82278681e-01 3.10617834e-01 1.38258353e-01 -7.48432636e-01 -1.11817777e+00 -1.20913327e-01 -9.81194139e-01 -3.41187745e-01 5.04851758e-01 -7.04682171e-01 -6.24107778e-01 5.48795819e-01 -1.16020751e+00 -3.08384091e-01 -3.32007170e-01 8.11791062e-01 -4.98120844e-01 -1.13310166e-01 -7.07239211e-02 -1.20557225e+00 1.98764428e-01 -1.18031490e+00 9.61775959e-01 2.77413696e-01 -1.71885222e-01 -7.98842132e-01 -1.94317281e-01 -8.49968344e-02 -1.91344395e-02 -8.32466111e-02 1.21172619e+00 -1.16813290e+00 -2.68739283e-01 -7.82042205e-01 -1.89806581e-01 7.87301004e-01 -7.82853141e-02 -5.87940477e-02 -1.42053819e+00 -4.53295231e-01 1.58195961e-02 -3.03264737e-01 9.78417516e-01 5.45699716e-01 1.44546831e+00 -2.09438711e-01 -6.40578330e-01 3.23779881e-01 1.16303515e+00 5.15377223e-01 4.22937751e-01 -1.16331145e-01 1.09892882e-01 6.75207675e-01 2.64473736e-01 2.14715838e-01 -6.92143440e-02 4.92372245e-01 3.03396136e-01 3.51512015e-01 1.40047416e-01 -1.24346420e-01 2.04481587e-01 3.55325043e-01 1.82984635e-01 -2.94680864e-01 -7.66398191e-01 4.77512091e-01 -1.62475550e+00 -8.79135370e-01 1.12950020e-01 2.82234120e+00 7.10057616e-01 4.70196009e-01 8.58613104e-02 3.82135153e-01 1.06906307e+00 -5.55902302e-01 -1.01064312e+00 -1.86964408e-01 -1.32215485e-01 4.29662645e-01 4.18224305e-01 3.86869580e-01 -9.90561426e-01 3.43134761e-01 7.49473190e+00 6.86814964e-01 -8.57551694e-01 1.06156364e-01 9.74341989e-01 -8.34522173e-02 2.37333059e-01 -9.32936370e-03 -1.06000912e+00 5.21994770e-01 1.39966726e+00 -2.13736277e-02 5.24946988e-01 8.90530705e-01 -3.54179293e-01 -2.19911233e-01 -1.70958424e+00 6.20645106e-01 2.52452463e-01 -9.65080500e-01 1.05102479e-01 -5.22366948e-02 5.57399452e-01 1.63761117e-02 1.45980716e-01 4.47057456e-01 2.14567944e-01 -8.56963277e-01 8.89836609e-01 6.62675202e-01 9.86596763e-01 -5.52910566e-01 7.72724211e-01 6.64105296e-01 -6.59243226e-01 -6.81398869e-01 -4.84324843e-01 2.86780354e-02 -6.25030160e-01 4.26534384e-01 -7.66174793e-01 -9.54036191e-02 5.48527598e-01 4.37069565e-01 -7.01351762e-01 1.11398244e+00 5.21553606e-02 7.83202648e-01 -2.60426193e-01 -2.31015548e-01 -1.94291249e-01 2.03627095e-01 2.29701430e-01 9.62637961e-01 3.77330035e-01 9.57010910e-02 -9.00496170e-02 7.40940988e-01 -1.10381007e-01 -2.90540665e-01 -4.90545899e-01 2.10070103e-01 5.68444192e-01 9.51028943e-01 -7.37738311e-01 -5.88201225e-01 -2.95079410e-01 5.16688883e-01 5.04506588e-01 3.99745822e-01 -6.59003794e-01 -4.08332318e-01 4.10897523e-01 4.18793038e-02 3.14601153e-01 3.09992582e-01 -2.43608534e-01 -1.08704877e+00 1.94210594e-03 -4.78036255e-01 3.27074021e-01 -8.15875173e-01 -1.38472342e+00 7.10641921e-01 2.15296030e-01 -1.26707256e+00 -2.18144834e-01 -1.18350816e+00 -3.49759579e-01 6.68012083e-01 -1.15667295e+00 -6.15482509e-01 1.62374657e-02 9.13789123e-02 2.37218276e-01 -4.18772906e-01 8.95572960e-01 5.45786060e-02 -6.07537806e-01 9.42422271e-01 7.10216403e-01 2.67306209e-01 4.85681474e-01 -1.05283296e+00 -7.10028931e-02 4.43745464e-01 1.45385131e-01 2.86112696e-01 5.88396788e-01 -4.53528523e-01 -8.07975769e-01 -1.10653138e+00 6.97727382e-01 -8.28916728e-01 4.65957820e-01 -4.39686835e-01 -1.03899395e+00 8.30788732e-01 -7.09693611e-01 3.21775377e-01 6.97059691e-01 1.54575601e-01 -4.71610159e-01 -3.75153840e-01 -1.29135251e+00 2.80677855e-01 8.27263772e-01 -5.01124561e-01 -6.02845848e-01 3.85828316e-01 3.69089037e-01 -4.96567339e-02 -7.06980884e-01 6.28027976e-01 1.11136341e+00 -1.07659042e+00 8.14090312e-01 -1.13026083e+00 3.37932020e-01 7.77681693e-02 -6.60024524e-01 -1.25448108e+00 -2.98708498e-01 1.47001296e-01 1.20857805e-01 8.33924532e-01 7.10458338e-01 -8.85289550e-01 6.86807215e-01 5.13162494e-01 2.11804003e-01 -8.94346178e-01 -1.31014192e+00 -8.93808007e-01 3.62218261e-01 -5.89729011e-01 6.69027209e-01 7.26920247e-01 -1.57244086e-01 2.46071100e-01 -8.61490890e-03 2.87099391e-01 8.38753760e-01 -1.72235165e-02 3.26081365e-01 -1.78288317e+00 -1.71501845e-01 -3.14983130e-01 -8.31997812e-01 -7.42057681e-01 2.94233501e-01 -1.01363301e+00 1.29717380e-01 -6.79389775e-01 5.85533321e-01 -6.96469367e-01 -6.06400549e-01 4.01767790e-01 -2.38112621e-02 1.89697057e-01 -4.48960625e-02 2.42353633e-01 -3.80875140e-01 3.65787357e-01 7.70199358e-01 -5.42562120e-02 -2.06343234e-01 3.99476141e-01 -6.44323945e-01 1.03539217e+00 6.20333493e-01 -9.91360128e-01 -1.14085346e-01 8.96076187e-02 -7.38667995e-02 2.33919397e-01 5.70398867e-01 -1.40198028e+00 3.83046977e-02 -9.76856425e-02 1.00778425e+00 -4.08657372e-01 5.15829206e-01 -7.57869303e-01 4.02317680e-02 5.65366387e-01 -8.66693258e-01 -2.19776616e-01 1.30728588e-01 8.12465966e-01 2.68275857e-01 -6.91745043e-01 9.89940047e-01 2.13751093e-01 5.37937656e-02 1.06436843e-02 -4.62626547e-01 7.67954066e-02 9.74858403e-01 -2.45507464e-01 -3.64982158e-01 -3.21588755e-01 -1.00065434e+00 -3.38624269e-01 1.34209543e-01 2.23749831e-01 5.45073569e-01 -1.36803126e+00 -7.11228728e-01 5.56535304e-01 2.98917055e-01 -5.01468599e-01 -7.77467936e-02 5.42103648e-01 1.82768837e-01 3.41274053e-01 -1.04025856e-01 -8.55721295e-01 -1.08463395e+00 4.29580659e-01 6.66715443e-01 1.13145024e-01 -7.27810413e-02 5.05411088e-01 4.65142965e-01 -5.53174734e-01 2.05710918e-01 -3.52244675e-01 -1.08819552e-01 -5.95227703e-02 5.53153515e-01 9.60936397e-02 6.23375401e-02 -6.43204629e-01 -1.46701813e-01 3.19697171e-01 -3.38995963e-01 -3.02327201e-02 1.21006954e+00 2.78516054e-01 2.17708498e-01 9.66379642e-01 1.12187541e+00 -4.82756048e-01 -1.23909247e+00 -2.26202354e-01 -1.18407600e-01 -5.12696207e-01 -1.15407266e-01 -1.03837013e+00 -6.00795209e-01 1.04256535e+00 9.84924197e-01 3.07719409e-01 9.56000388e-01 1.28053606e-01 -1.13730229e-01 5.41353643e-01 4.17718470e-01 -9.33741271e-01 -2.42171869e-01 2.21600115e-01 7.59302020e-01 -1.20118928e+00 -7.82902632e-03 -2.33142257e-01 -3.40732992e-01 1.20061469e+00 5.22022545e-01 -6.00745380e-02 8.22051525e-01 6.01054579e-02 -3.89484525e-01 2.29543030e-01 -8.10003459e-01 2.17415035e-01 4.16191578e-01 7.74691999e-01 1.61488071e-01 4.12392408e-01 1.67298496e-01 1.00749254e+00 -1.69658467e-01 1.61814183e-01 6.45774066e-01 6.27835631e-01 -6.75209045e-01 -6.65216267e-01 -3.06620181e-01 1.07915330e+00 -3.62623930e-01 -8.50890856e-03 -2.38248765e-01 7.07378387e-01 2.39372030e-01 7.62845874e-01 3.60374212e-01 -4.96234149e-01 2.01500997e-01 4.51435775e-01 7.19616115e-01 -6.43060386e-01 2.54754741e-02 -4.66085166e-01 -6.55387416e-02 -1.54975384e-01 -2.80078590e-01 -8.76647472e-01 -8.03549469e-01 1.33017004e-01 -7.72695005e-01 9.62914154e-02 7.54910350e-01 1.20494437e+00 7.03965798e-02 3.37586492e-01 7.25054324e-01 -7.21115232e-01 -9.87246871e-01 -9.74038899e-01 -7.53289878e-01 1.49868593e-01 5.58587015e-01 -9.32167411e-01 -6.83055639e-01 -2.16680571e-01]
[8.7008638381958, 4.049533843994141]
98cc1661-b49c-4464-aaaf-ebd235e90cee
a-policy-guided-imitation-approach-for
2210.08323
null
https://arxiv.org/abs/2210.08323v3
https://arxiv.org/pdf/2210.08323v3.pdf
A Policy-Guided Imitation Approach for Offline Reinforcement Learning
Offline reinforcement learning (RL) methods can generally be categorized into two types: RL-based and Imitation-based. RL-based methods could in principle enjoy out-of-distribution generalization but suffer from erroneous off-policy evaluation. Imitation-based methods avoid off-policy evaluation but are too conservative to surpass the dataset. In this study, we propose an alternative approach, inheriting the training stability of imitation-style methods while still allowing logical out-of-distribution generalization. We decompose the conventional reward-maximizing policy in offline RL into a guide-policy and an execute-policy. During training, the guide-poicy and execute-policy are learned using only data from the dataset, in a supervised and decoupled manner. During evaluation, the guide-policy guides the execute-policy by telling where it should go so that the reward can be maximized, serving as the \textit{Prophet}. By doing so, our algorithm allows \textit{state-compositionality} from the dataset, rather than \textit{action-compositionality} conducted in prior imitation-style methods. We dumb this new approach Policy-guided Offline RL (\texttt{POR}). \texttt{POR} demonstrates the state-of-the-art performance on D4RL, a standard benchmark for offline RL. We also highlight the benefits of \texttt{POR} in terms of improving with supplementary suboptimal data and easily adapting to new tasks by only changing the guide-poicy.
['Xianyuan Zhan', 'Jianxiong Li', 'Li Jiang', 'Haoran Xu']
2022-10-15
null
null
null
null
['d4rl']
['robots']
[-9.63488147e-02 2.00529158e-01 -5.70911705e-01 -1.25972643e-01 -8.50518405e-01 -1.09864640e+00 8.01302433e-01 -1.16878174e-01 -8.99369538e-01 1.13042259e+00 -1.34824663e-01 -4.43419784e-01 -1.23709276e-01 -5.04051566e-01 -8.85399818e-01 -9.03545141e-01 -1.57761499e-01 6.71566606e-01 8.21656585e-02 -4.20913309e-01 2.06517175e-01 4.78178561e-01 -1.40902841e+00 1.12729646e-01 7.88647175e-01 8.77079248e-01 4.12451774e-01 5.30834258e-01 8.17924961e-02 1.01761854e+00 -7.33471692e-01 -1.26507515e-02 4.87643272e-01 -7.56153524e-01 -6.48124337e-01 -2.81916261e-01 -2.59174556e-01 -9.91169572e-01 -3.44037801e-01 8.11140180e-01 6.62700593e-01 3.87641400e-01 5.13878703e-01 -1.23329866e+00 -4.10315305e-01 6.75063431e-01 -1.39452443e-01 -3.82564701e-02 4.39093351e-01 6.69342935e-01 1.03023326e+00 -4.09046501e-01 8.16752017e-01 1.14907980e+00 2.33431935e-01 9.53750193e-01 -1.28846383e+00 -5.92444718e-01 4.72774684e-01 -7.02572241e-02 -7.90915549e-01 -2.91008651e-01 6.45426035e-01 -1.97615609e-01 9.24585581e-01 7.53861740e-02 7.29517281e-01 1.46878171e+00 -1.05021950e-02 1.38384283e+00 1.53963327e+00 -1.16292544e-01 7.19595432e-01 -1.27619088e-01 -5.57277143e-01 4.18652654e-01 -2.90856987e-01 8.08451355e-01 -5.29243231e-01 -3.41634937e-02 1.07923579e+00 -2.10504994e-01 -3.89936939e-02 -5.36875486e-01 -1.47653079e+00 7.69078135e-01 2.69832671e-01 1.15942702e-01 -5.09138703e-01 5.08105934e-01 6.11192048e-01 7.56404221e-01 9.66298357e-02 5.84517241e-01 -6.80017650e-01 -7.17778087e-01 -8.46958101e-01 6.25060976e-01 6.81026816e-01 9.47330594e-01 6.85880303e-01 3.09208572e-01 -4.80141431e-01 6.76368952e-01 4.19322141e-02 5.73703468e-01 7.78578997e-01 -1.52762747e+00 4.87885147e-01 2.80276299e-01 5.72558582e-01 -2.20032156e-01 -3.87297362e-01 -3.45496297e-01 -3.58485758e-01 6.27346575e-01 6.53817177e-01 -4.94083345e-01 -6.25854611e-01 2.03778958e+00 4.14158732e-01 -2.59767205e-01 1.62466958e-01 8.48324299e-01 2.68347740e-01 4.32985961e-01 -2.62396839e-02 -5.02614081e-01 7.86286891e-01 -1.04613924e+00 -3.75075281e-01 -4.26555574e-02 8.42194021e-01 -3.01567852e-01 1.36506772e+00 6.82839274e-01 -1.01946485e+00 -3.79731268e-01 -8.32592547e-01 3.37685674e-01 -1.13137595e-01 3.00310999e-01 3.69208515e-01 3.05170268e-01 -1.10471344e+00 1.15373361e+00 -8.62769544e-01 -6.01114333e-02 2.66602069e-01 5.38538456e-01 -2.29666710e-01 3.38744223e-01 -1.11504650e+00 9.28675115e-01 4.77515042e-01 -1.16985835e-01 -1.64634442e+00 -3.94674301e-01 -6.29540741e-01 -3.26575905e-01 7.40321577e-01 -2.62418091e-01 1.87725043e+00 -1.25638115e+00 -2.39271522e+00 5.95499158e-01 1.32109165e-01 -5.09727895e-01 1.05121779e+00 -1.45991251e-01 2.21849345e-02 1.88264191e-01 1.08277842e-01 8.68488848e-01 1.15375483e+00 -1.25200570e+00 -7.04074919e-01 -6.19707294e-02 4.73706216e-01 5.39141893e-01 8.33906382e-02 -2.93163955e-01 -1.54468119e-01 -7.11876035e-01 -3.46223563e-01 -1.25716698e+00 6.40265411e-03 -7.41420761e-02 -3.27380270e-01 -4.68274415e-01 6.99022233e-01 -3.14962745e-01 1.02628541e+00 -1.97644913e+00 2.99881339e-01 -1.71471998e-01 2.04760581e-02 3.28146309e-01 -4.16191310e-01 6.78389907e-01 1.96770355e-01 -9.98442993e-02 -1.33992344e-01 -3.61457974e-01 2.74845749e-01 6.10652626e-01 -4.96151596e-01 4.27952915e-01 -3.52158733e-02 9.98199582e-01 -1.47805870e+00 -2.22131178e-01 2.16925472e-01 -1.19898498e-01 -7.36886322e-01 3.71406049e-01 -8.54301393e-01 1.15622425e+00 -7.64488459e-01 4.83837336e-01 1.41235560e-01 1.98374137e-01 4.09416735e-01 4.06140715e-01 -2.59367287e-01 4.40766573e-01 -9.68787014e-01 1.71979856e+00 -4.58213866e-01 2.59292156e-01 4.70237341e-03 -1.05199885e+00 9.08521175e-01 2.79458553e-01 8.25287223e-01 -9.18873250e-01 8.49077776e-02 4.16404963e-01 2.41829038e-01 -2.49757096e-01 4.06215727e-01 -4.76210332e-03 -1.28884703e-01 8.10767829e-01 3.71246636e-02 -3.27093363e-01 3.29750150e-01 -2.44384725e-03 1.05000865e+00 1.19817829e+00 2.44552225e-01 -1.91969752e-01 3.89324844e-01 2.29758061e-02 4.58376378e-01 9.96401370e-01 -4.07720864e-01 3.50135639e-02 7.80895650e-01 -2.12443843e-01 -1.01784921e+00 -1.15106869e+00 3.10182244e-01 1.44804513e+00 -6.24974333e-02 -1.04009219e-01 -6.75272942e-01 -1.25889719e+00 2.12366581e-01 1.03144670e+00 -7.14582145e-01 -1.39259711e-01 -6.65426016e-01 -1.05908483e-01 8.01095307e-01 3.95553440e-01 5.57998896e-01 -1.63090205e+00 -9.60885346e-01 4.41481352e-01 -6.33575097e-02 -7.83337116e-01 -5.70994794e-01 4.25894201e-01 -9.06788886e-01 -8.37878227e-01 -9.13343191e-01 -3.72117370e-01 5.39376020e-01 -1.70078799e-01 8.61727238e-01 -3.48510258e-02 3.67520034e-01 4.50397819e-01 -4.83000219e-01 -3.77697319e-01 -7.01096296e-01 4.10981774e-02 3.04951787e-01 -2.31922746e-01 -3.45772468e-02 -6.48134708e-01 -6.81664288e-01 6.11162484e-01 -6.29827142e-01 -2.50917286e-01 5.55388629e-01 8.95046532e-01 7.59803534e-01 -4.85890776e-01 8.61688912e-01 -5.61895609e-01 8.09895396e-01 -3.71220648e-01 -7.21092939e-01 -6.32633418e-02 -7.60138512e-01 4.53502536e-01 1.16749597e+00 -9.66999650e-01 -8.34643483e-01 -1.06697883e-02 -1.07671097e-01 -6.53171480e-01 3.26862093e-03 6.43498003e-02 1.65348291e-01 1.63888305e-01 6.89491630e-01 7.06103861e-01 3.11809510e-01 -3.35292429e-01 5.31552613e-01 4.36550945e-01 2.82005817e-01 -1.20010281e+00 5.34059167e-01 3.48144829e-01 -1.66964844e-01 -4.57411051e-01 -6.88211143e-01 -9.48125198e-02 -3.89821678e-01 -4.60290879e-01 4.38713312e-01 -5.89684606e-01 -1.21297669e+00 4.12185669e-01 -7.40771711e-01 -1.27248478e+00 -8.64020348e-01 4.58817869e-01 -1.26133931e+00 2.39504322e-01 -4.73708034e-01 -9.36396956e-01 -5.98227233e-02 -1.42438602e+00 9.92829382e-01 1.41263261e-01 -1.37112916e-01 -8.75657618e-01 1.73762962e-01 -1.47730693e-01 3.40712607e-01 1.63129330e-01 6.06540322e-01 -8.53867352e-01 -2.18191728e-01 1.73699528e-01 1.81774169e-01 5.04198253e-01 1.98708326e-02 -2.45786205e-01 -7.68867254e-01 -6.71343923e-01 -1.51453942e-01 -9.23734605e-01 4.20384616e-01 1.77590117e-01 1.12794459e+00 -4.91710842e-01 1.09037384e-01 4.74859118e-01 1.13674712e+00 4.19540793e-01 4.90739793e-01 7.49697566e-01 2.04392597e-01 3.74653637e-01 9.14069176e-01 7.44140863e-01 3.20297778e-01 7.43426681e-01 5.22916853e-01 3.79763842e-01 4.39171195e-02 -8.09939682e-01 1.02609551e+00 2.49481946e-01 -2.00026497e-01 -9.51999351e-02 -2.82469034e-01 2.82295376e-01 -2.00162148e+00 -9.79497910e-01 3.52588832e-01 2.29033899e+00 1.33186781e+00 4.92563210e-02 6.13032997e-01 -1.17292441e-01 1.57022223e-01 1.94281101e-01 -1.11905992e+00 -7.08657146e-01 1.32624805e-01 1.07572421e-01 5.12012780e-01 4.07077432e-01 -6.55542314e-01 1.25354636e+00 5.65879679e+00 1.09158921e+00 -1.26759028e+00 1.03019729e-01 3.15233797e-01 -2.92858630e-01 -3.29487413e-01 1.79097235e-01 -8.03982556e-01 5.54008663e-01 6.43247962e-01 6.76818267e-02 1.12460244e+00 9.47326243e-01 5.26533008e-01 -4.71141875e-01 -1.45494366e+00 6.75768018e-01 -3.93276721e-01 -9.34331119e-01 -2.09243178e-01 1.91381916e-01 7.11979210e-01 1.09709829e-01 3.57819088e-02 9.38593924e-01 6.81086063e-01 -7.23208010e-01 1.19739640e+00 4.26519334e-01 1.07367933e+00 -5.48959076e-01 2.06996799e-01 8.58839691e-01 -9.27544951e-01 -1.47528246e-01 -1.11588039e-01 -1.53959736e-01 -1.11483566e-01 1.24461748e-01 -6.31024599e-01 5.47587454e-01 3.38968456e-01 6.46186531e-01 -1.13925477e-03 3.90830100e-01 -8.95165205e-01 5.51974714e-01 -3.56488526e-01 -4.33299899e-01 6.05486274e-01 -3.92012656e-01 7.47084379e-01 8.06582630e-01 2.44976189e-02 -1.13600627e-01 5.14660716e-01 7.81180441e-01 -2.64030825e-02 -6.72633871e-02 -6.68369412e-01 -2.23853573e-01 4.90352750e-01 9.22763526e-01 -3.76005739e-01 -2.55784839e-01 2.37123802e-01 1.12937951e+00 6.07086062e-01 5.61530948e-01 -8.66413534e-01 3.94863747e-02 4.66341853e-01 -1.75774381e-01 4.09449995e-01 -5.29950321e-01 2.09854722e-01 -9.81202960e-01 -3.90738174e-02 -1.34662592e+00 2.44225413e-01 -5.57561934e-01 -8.84505033e-01 2.75583744e-01 3.66410285e-01 -1.40322518e+00 -7.49312937e-01 -4.62871522e-01 -2.72673130e-01 4.54991400e-01 -1.52026939e+00 -8.68863821e-01 2.38100216e-01 6.97105825e-01 7.05042124e-01 -2.30592072e-01 6.71980798e-01 -1.58443034e-01 -3.34861755e-01 6.66462660e-01 5.15735567e-01 -9.72459763e-02 6.97394371e-01 -1.36515486e+00 -5.18952943e-02 2.53392041e-01 -1.58850983e-01 4.50471461e-01 6.39784753e-01 -6.36946142e-01 -1.50514591e+00 -8.45982134e-01 2.85187215e-01 -2.49730289e-01 8.19116712e-01 -2.26093411e-01 -4.27254558e-01 7.36079633e-01 2.94453494e-04 -1.63913831e-01 5.13176760e-03 -1.90522462e-01 -1.36464983e-01 -1.32067904e-01 -1.24714780e+00 8.82566988e-01 1.08906305e+00 -3.67484987e-01 -4.59835976e-01 3.27208847e-01 6.61185801e-01 -5.65735161e-01 -6.79489255e-01 7.57037997e-02 6.36066556e-01 -1.11006391e+00 7.13587940e-01 -5.25127411e-01 3.25434834e-01 -2.31280640e-01 5.47765903e-02 -1.51644552e+00 1.12264574e-01 -1.22925651e+00 -3.40137690e-01 8.75226438e-01 2.67321944e-01 -1.02224624e+00 6.62322700e-01 8.03843811e-02 -4.77190204e-02 -1.05280232e+00 -1.02775884e+00 -1.41286600e+00 4.34958488e-01 -4.22790319e-01 4.87449557e-01 6.24802768e-01 3.09183002e-02 -6.03911988e-02 -4.97082323e-01 -4.22758430e-01 5.29082716e-01 -1.56296454e-02 9.73338842e-01 -5.71705103e-01 -7.56053746e-01 -6.02623403e-01 4.28692222e-01 -1.67197108e+00 3.15692335e-01 -8.50064099e-01 2.86484212e-01 -1.32084632e+00 -2.39907831e-01 -6.80653095e-01 -1.11247078e-01 8.34617674e-01 2.35934004e-01 -3.04060310e-01 4.39660549e-01 6.51988804e-01 -5.91013730e-01 8.90985727e-01 1.78590703e+00 1.06537417e-01 -4.88733888e-01 2.27587596e-02 -2.24348322e-01 6.21333778e-01 9.78323638e-01 -7.02620625e-01 -7.09923625e-01 -1.94615498e-01 7.41285458e-02 3.36170405e-01 3.02946419e-01 -6.05684996e-01 -7.63909742e-02 -4.62775975e-01 8.78354684e-02 -3.15864980e-01 1.83745563e-01 -4.81241733e-01 -3.02204996e-01 7.95575976e-01 -5.61483383e-01 1.55093640e-01 1.36348158e-01 6.00540876e-01 2.81301975e-01 -2.64211357e-01 6.66492701e-01 -3.07114482e-01 -6.08856678e-01 1.92542776e-01 -4.94582593e-01 3.33502084e-01 1.13611543e+00 -1.59048676e-01 -3.12078208e-01 -5.36716461e-01 -7.55179644e-01 4.44058627e-01 5.63916504e-01 2.12676242e-01 4.21287626e-01 -1.09650600e+00 -5.33096433e-01 3.68086398e-02 -1.02682211e-01 -6.18056133e-02 -8.78991187e-03 1.08807862e+00 -1.83008194e-01 2.74235487e-01 -1.79344401e-01 -4.74646181e-01 -5.22678077e-01 5.70565104e-01 5.04388452e-01 -7.45451689e-01 -6.80671811e-01 5.00917673e-01 5.02562430e-03 -8.51721764e-01 3.30755115e-01 -3.91769350e-01 -1.54783614e-02 -3.24057378e-02 1.63004950e-01 3.55976462e-01 -3.23438793e-01 -5.66626377e-02 -2.19822794e-01 2.01722220e-01 1.57815143e-01 -7.40590453e-01 1.15233171e+00 9.89067182e-02 1.98614612e-01 6.38356090e-01 8.37484658e-01 -4.32453267e-02 -1.94200563e+00 3.78020667e-02 -3.93189937e-02 -2.36705810e-01 -2.43274242e-01 -1.28722143e+00 -6.53318226e-01 6.00874364e-01 3.26027989e-01 1.27413347e-01 9.05093312e-01 -2.33345285e-01 6.03809357e-01 7.16850936e-01 8.67560387e-01 -1.56111801e+00 3.70629042e-01 5.73067546e-01 1.10377443e+00 -1.11004937e+00 -2.37630662e-02 5.84126532e-01 -1.06568205e+00 9.99753475e-01 5.59192955e-01 -5.06746352e-01 9.64485034e-02 3.67601775e-02 -1.13588542e-01 1.91820920e-01 -8.31159234e-01 -4.45403665e-01 -8.89127478e-02 6.74255669e-01 5.98819740e-02 1.51783228e-01 -4.45722222e-01 3.53802443e-01 -9.76687893e-02 1.97087988e-01 2.34621793e-01 1.21678746e+00 -4.29487646e-01 -1.46194482e+00 -1.08342327e-01 2.96607971e-01 -1.38193831e-01 1.45654932e-01 -4.20567900e-01 1.12961221e+00 -1.71788856e-01 9.13125038e-01 -2.11780578e-01 -3.07334572e-01 3.10283005e-01 -1.18660592e-01 7.61182904e-01 -4.75833535e-01 -8.17122281e-01 1.51078194e-01 2.57122219e-01 -9.35003340e-01 -1.64337620e-01 -7.68942833e-01 -1.69725263e+00 -1.81178331e-01 3.48421894e-02 1.35414869e-01 5.42036653e-01 1.23237157e+00 2.85258740e-01 4.03329760e-01 9.36796606e-01 -1.06234765e+00 -1.42069089e+00 -8.48255396e-01 -5.01497149e-01 2.58540690e-01 4.36334431e-01 -8.11744034e-01 -3.63498062e-01 -3.64010185e-01]
[4.091348648071289, 2.0806221961975098]
25c3a3a3-b888-486d-b78a-5a7991e3eeec
generalization-challenges-for-neural
1803.08629
null
http://arxiv.org/abs/1803.08629v2
http://arxiv.org/pdf/1803.08629v2.pdf
Generalization Challenges for Neural Architectures in Audio Source Separation
Recent work has shown that recurrent neural networks can be trained to separate individual speakers in a sound mixture with high fidelity. Here we explore convolutional neural network models as an alternative and show that they achieve state-of-the-art results with an order of magnitude fewer parameters. We also characterize and compare the robustness and ability of these different approaches to generalize under three different test conditions: longer time sequences, the addition of intermittent noise, and different datasets not seen during training. For the last condition, we create a new dataset, RealTalkLibri, to test source separation in real-world environments. We show that the acoustics of the environment have significant impact on the structure of the waveform and the overall performance of neural network models, with the convolutional model showing superior ability to generalize to new environments. The code for our study is available at https://github.com/ShariqM/source_separation.
['Bruno Olshausen', 'Brian Cheung', 'Shariq Mobin']
2018-03-23
null
null
null
null
['audio-source-separation']
['audio']
[ 1.57623395e-01 -4.13914531e-01 3.17141980e-01 -3.64811569e-01 -1.12311995e+00 -7.60887325e-01 3.39805841e-01 -2.38851294e-01 -2.63218254e-01 4.84072357e-01 3.44573945e-01 -4.09642428e-01 -1.76644206e-01 -2.84996092e-01 -7.22171307e-01 -5.63775122e-01 -3.06503296e-01 1.42905265e-01 1.66212648e-01 -1.03843503e-01 -2.80373812e-01 6.01879120e-01 -1.68556333e+00 2.87705034e-01 4.71512407e-01 8.07771385e-01 3.53627019e-02 1.17894638e+00 2.99388200e-01 4.22365397e-01 -1.04458165e+00 3.88551839e-02 3.58752966e-01 -4.72661257e-01 -4.00246292e-01 -2.52419949e-01 6.71610892e-01 -1.32160231e-01 -5.46966910e-01 7.94370055e-01 1.05073082e+00 4.27377701e-01 2.93480724e-01 -1.03542817e+00 -2.84238756e-01 8.69955957e-01 2.99140979e-02 6.41736567e-01 1.86686590e-01 2.82804400e-01 6.98560655e-01 -7.23048568e-01 9.60528255e-02 1.01581240e+00 9.65364337e-01 5.39222538e-01 -1.24547541e+00 -1.03794563e+00 -6.95864763e-03 -8.66869185e-03 -1.17708099e+00 -1.19929957e+00 6.86931729e-01 -1.87053904e-01 9.07439470e-01 3.34155291e-01 2.42540479e-01 1.47541070e+00 -2.33667061e-01 4.59293604e-01 8.54658484e-01 -2.94515610e-01 1.50738463e-01 -1.44398771e-02 1.89183772e-01 7.02282637e-02 6.64789602e-03 3.06485534e-01 -5.85675895e-01 -5.79896718e-02 2.66917467e-01 -4.14121747e-01 -5.44858336e-01 2.02946216e-01 -8.55432749e-01 3.93244386e-01 2.26742879e-01 4.17251438e-01 -1.38303772e-01 3.48020792e-01 3.22111398e-01 3.37536722e-01 4.67165112e-01 4.21829134e-01 -5.13355792e-01 -4.24076587e-01 -1.09099364e+00 2.31383741e-01 9.95922565e-01 7.07521915e-01 2.93132961e-01 8.26491058e-01 7.38579631e-02 1.09078491e+00 1.00547010e-02 5.87926924e-01 7.79363215e-01 -1.07175624e+00 3.22504759e-01 -3.46401751e-01 -8.79591256e-02 -4.86046463e-01 -5.19071639e-01 -1.05430567e+00 -4.58532482e-01 1.60213545e-01 5.28297246e-01 -6.87395573e-01 -1.21539116e+00 1.96812224e+00 -1.38692439e-01 6.66435778e-01 2.31989980e-01 5.53285301e-01 1.06638992e+00 6.59227610e-01 -2.66303748e-01 9.80900675e-02 7.99749374e-01 -7.70615458e-01 -6.39791250e-01 -4.12310749e-01 2.24107176e-01 -9.61329877e-01 7.88672864e-01 4.65071827e-01 -1.23961258e+00 -6.98538542e-01 -1.21826267e+00 4.22686905e-01 -4.26904112e-01 2.06211153e-02 2.61905313e-01 9.89934444e-01 -1.14170551e+00 8.12532485e-01 -7.91633964e-01 -1.35983363e-01 2.05109566e-01 3.44479024e-01 -1.56937703e-01 1.26658186e-01 -1.26928389e+00 4.86179203e-01 6.18642196e-02 1.89258471e-01 -1.16334999e+00 -9.40912902e-01 -7.33038008e-01 2.59744555e-01 7.89522529e-02 -1.03941523e-01 1.74869406e+00 -7.41383255e-01 -1.42914140e+00 2.32395723e-01 -1.37731403e-01 -5.30920088e-01 3.78256887e-01 -4.08173293e-01 -9.28150177e-01 -1.14176929e-01 -2.26615399e-01 3.69412273e-01 6.85118496e-01 -1.26847553e+00 -4.84648854e-01 5.81118912e-02 -2.80227005e-01 -5.77145591e-02 -6.35297224e-02 1.26084477e-01 -1.64591834e-01 -7.25582004e-01 -1.05436824e-01 -9.65778053e-01 6.31255433e-02 -5.99834919e-01 -3.16624105e-01 1.37818590e-01 6.56305909e-01 -7.76666164e-01 8.35400760e-01 -2.55364895e+00 -2.09525526e-01 5.75507991e-02 -1.02751903e-01 5.70307016e-01 -4.63334113e-01 4.67656106e-01 -4.10832465e-01 2.98418552e-01 -2.25701362e-01 -5.74541569e-01 -9.20382387e-04 -1.06686331e-01 -3.93652856e-01 2.99612075e-01 2.52498567e-01 3.82948607e-01 -5.41127205e-01 2.93323785e-01 -1.78758189e-01 8.23507071e-01 -3.57635617e-01 2.31669486e-01 1.04820803e-01 4.87930924e-01 3.18925112e-01 3.26038033e-01 6.87671244e-01 2.77447909e-01 -3.71910892e-02 -1.24962041e-02 -8.85702744e-02 7.80091584e-01 -1.37443745e+00 1.41245008e+00 -6.78140819e-01 1.19140363e+00 3.63998264e-01 -5.52117348e-01 7.64250398e-01 6.28926754e-01 1.54328048e-01 -5.60153008e-01 2.48239815e-01 4.52119321e-01 6.62830114e-01 -2.55634665e-01 3.25080931e-01 -1.28020525e-01 2.05475777e-01 6.29171312e-01 2.97589809e-01 -1.57551467e-01 1.41959205e-01 -1.37887094e-02 1.14229167e+00 -2.87177891e-01 -4.25185680e-01 -1.24766327e-01 -2.04228535e-02 -6.27238691e-01 5.75078964e-01 9.71439481e-01 -1.13206282e-01 9.01912987e-01 1.83489062e-02 -2.35055983e-02 -6.85861766e-01 -1.27386618e+00 -2.43712217e-01 1.08076358e+00 -3.07989120e-01 -1.31965339e-01 -6.50355756e-01 -1.30673796e-02 -3.65466997e-02 8.88798356e-01 -4.74714071e-01 -2.73713678e-01 -7.29969203e-01 -8.29471529e-01 1.21532202e+00 6.48651004e-01 3.01196516e-01 -1.06555748e+00 -5.13579845e-01 2.00916886e-01 -1.07797273e-01 -1.27557027e+00 -2.56247669e-01 7.59787560e-01 -4.44700688e-01 -5.85880637e-01 -5.77350199e-01 -6.03740573e-01 -3.48201171e-02 1.22349456e-01 1.08982575e+00 -1.27941947e-02 -1.41102716e-01 4.14053768e-01 -2.69298464e-01 -7.64685750e-01 -8.19564581e-01 2.50282586e-01 3.10437977e-01 -8.85957330e-02 3.89791414e-04 -9.60785151e-01 -3.19925576e-01 1.87422559e-01 -1.00929189e+00 -4.82540876e-01 2.99956203e-01 4.24067944e-01 -2.99814250e-02 2.96379894e-01 9.85998809e-01 -5.60788393e-01 9.52700436e-01 -6.70710742e-01 -2.59575874e-01 -1.66941941e-01 -2.41633937e-01 -8.05958137e-02 5.64267457e-01 -6.33228838e-01 -9.42163289e-01 -3.27550501e-01 -5.98145604e-01 -3.70522171e-01 -5.18650949e-01 2.91944206e-01 -2.06731260e-01 1.50200129e-01 7.08664775e-01 3.26178819e-02 -2.16575012e-01 -8.38822246e-01 1.42858595e-01 8.51614892e-01 6.61833346e-01 -4.28549021e-01 7.86665738e-01 1.89668298e-01 -3.72585267e-01 -8.76190126e-01 -6.71193004e-01 -2.33230963e-01 -5.04839718e-01 -8.53825361e-02 5.05081654e-01 -9.59336400e-01 -3.10527205e-01 7.97245800e-01 -9.30675805e-01 -8.20142925e-01 -2.12010965e-01 7.43786335e-01 -2.48435095e-01 -8.18305984e-02 -6.53561890e-01 -8.78384352e-01 -1.88966632e-01 -1.08772993e+00 6.07590318e-01 3.69585454e-01 -1.01819210e-01 -9.26859319e-01 3.05213481e-01 -2.60402411e-02 7.68273830e-01 5.11920452e-02 6.42401934e-01 -9.63644326e-01 -1.69679403e-01 -7.59626105e-02 1.89146549e-01 6.63440704e-01 3.79796386e-01 1.99014783e-01 -1.73155820e+00 -3.91735494e-01 1.33454964e-01 -1.27620161e-01 1.07296360e+00 3.19408506e-01 7.62713253e-01 -6.07660525e-02 -5.62136024e-02 8.36481452e-01 9.68961179e-01 4.89567220e-01 5.91269076e-01 9.66622978e-02 5.11230528e-01 4.97291803e-01 -9.61397141e-02 9.26276520e-02 -2.02312246e-01 5.40764391e-01 2.00698689e-01 -2.03602716e-01 -5.67144811e-01 4.38821763e-02 5.43725550e-01 9.78719532e-01 2.50046313e-01 -5.61598301e-01 -9.22173619e-01 7.19455004e-01 -1.19783890e+00 -1.00110078e+00 1.61826000e-01 2.10136724e+00 7.99890876e-01 2.55636215e-01 2.92037666e-01 4.91239935e-01 6.12872064e-01 2.30290994e-01 -4.58550960e-01 -5.95345914e-01 -1.68022588e-01 5.06771207e-01 2.92087913e-01 5.82210541e-01 -1.00909567e+00 5.43055415e-01 7.04078913e+00 5.03046632e-01 -1.51745558e+00 9.74370390e-02 3.86683583e-01 -4.88678128e-01 -7.93053433e-02 -4.80741233e-01 -6.79210186e-01 3.70409548e-01 1.67924094e+00 4.56356630e-02 4.58836526e-01 2.80676574e-01 6.34042621e-02 2.83278078e-01 -1.03478634e+00 5.35734236e-01 1.24794416e-01 -9.35235023e-01 -4.03572321e-01 -1.21153928e-01 5.46679318e-01 5.79078734e-01 4.75242168e-01 3.55875611e-01 3.88109833e-01 -1.14952362e+00 8.25027883e-01 3.27711999e-01 6.80103958e-01 -7.28532732e-01 6.77235961e-01 1.87989593e-01 -1.07019973e+00 -1.38734892e-01 -5.82083012e-04 -3.98927629e-02 1.40810817e-01 5.01673937e-01 -1.03350890e+00 4.59589392e-01 8.81774962e-01 3.77706766e-01 -7.53834009e-01 1.29917455e+00 -2.15261281e-02 1.26767147e+00 -5.04083335e-01 3.01490158e-01 8.31468999e-02 3.78672689e-01 8.83222103e-01 1.49687433e+00 5.40402114e-01 -3.10545862e-01 -1.65662006e-01 6.46238267e-01 -1.26384616e-01 -3.19286734e-01 -4.47513431e-01 -3.59770693e-02 5.91558576e-01 9.36070740e-01 -5.06116748e-01 -5.10288738e-02 -1.59989193e-01 4.69668865e-01 9.87021104e-02 7.96134412e-01 -7.65009403e-01 -5.65516889e-01 9.12026942e-01 -2.64525749e-02 5.91706038e-01 -2.66869485e-01 -3.24687846e-02 -8.23045313e-01 -9.66872275e-02 -1.09009278e+00 1.51696190e-01 -7.84204662e-01 -1.16693616e+00 9.77720678e-01 1.19442925e-01 -1.00114429e+00 -5.26633859e-01 -5.63010156e-01 -9.74493921e-01 9.76409912e-01 -1.36724710e+00 -6.22794747e-01 -2.80407369e-01 2.53463268e-01 5.39383113e-01 -3.38081837e-01 8.07049811e-01 5.26942015e-01 -7.72072792e-01 8.47206652e-01 3.27289999e-01 2.69681066e-01 8.36804450e-01 -1.15528297e+00 7.42624998e-01 1.09551549e+00 4.92976546e-01 6.01095080e-01 9.58366990e-01 -3.10111940e-01 -8.60853970e-01 -1.22794354e+00 3.53729784e-01 -3.88289511e-01 5.77141106e-01 -6.44387662e-01 -1.11951315e+00 7.57761717e-01 4.38860148e-01 -1.23302214e-01 9.36476052e-01 2.86878467e-01 -6.13944709e-01 -2.45724812e-01 -7.83122659e-01 5.37651658e-01 8.49331439e-01 -5.12740910e-01 -3.33504736e-01 4.55231778e-02 9.49989498e-01 -5.21340549e-01 -5.83514631e-01 5.00449657e-01 7.33752668e-01 -1.17333007e+00 9.19464171e-01 -4.58269984e-01 5.90644218e-02 -1.78880811e-01 -3.30934584e-01 -1.68558896e+00 -2.55248100e-01 -7.54553020e-01 3.31795886e-02 1.45196366e+00 8.56923461e-01 -8.76410902e-01 2.43408799e-01 1.64924383e-01 -4.00291473e-01 -5.11227906e-01 -8.79853129e-01 -1.06944370e+00 3.57678622e-01 -8.07631850e-01 8.81364286e-01 5.60929418e-01 -4.62873906e-01 2.03298032e-01 -2.31417984e-01 4.85551894e-01 1.96812630e-01 -1.25253871e-01 6.05126143e-01 -1.14062023e+00 -5.46762526e-01 -4.85306352e-01 -3.29539359e-01 -7.27581263e-01 2.97016561e-01 -7.79418230e-01 3.82987648e-01 -1.31671023e+00 -4.03555185e-01 -4.28269327e-01 -5.90919435e-01 3.63365263e-01 -4.21292856e-02 4.94792998e-01 3.80645514e-01 -1.36231016e-02 -8.70053917e-02 3.10448170e-01 5.87762296e-01 -9.93272290e-02 -3.03914905e-01 4.03503090e-01 -7.67781854e-01 6.00611389e-01 1.23478770e+00 -5.90643466e-01 -4.86415774e-01 -7.40124404e-01 -1.08525813e-01 -1.99834809e-01 2.58470148e-01 -1.70467639e+00 -8.62224773e-03 3.39799553e-01 5.12078404e-01 -2.40733266e-01 6.56854868e-01 -5.86220086e-01 2.83038944e-01 2.23804161e-01 -5.73339760e-01 1.12797722e-01 9.36082959e-01 2.88209647e-01 -9.15359184e-02 -3.66974533e-01 7.00448155e-01 1.37117177e-01 -2.02504873e-01 -3.89838517e-02 -5.61381578e-01 2.54103869e-01 2.53427923e-01 1.31022617e-01 -3.96248728e-01 -8.26045871e-01 -7.31398463e-01 -3.00699353e-01 3.92663926e-02 5.30261695e-01 2.45843366e-01 -1.09760606e+00 -9.44810987e-01 3.64612550e-01 -3.24637055e-01 -3.40303928e-01 1.97442859e-01 5.19477546e-01 -1.60736814e-01 1.17698707e-01 -2.73901150e-02 -4.76626068e-01 -1.28235054e+00 6.88603669e-02 8.17993641e-01 1.79666534e-01 -3.55009794e-01 9.41434324e-01 4.19524051e-02 -5.30620456e-01 4.87914234e-01 -4.38406765e-01 -5.07189706e-02 -5.57017811e-02 5.82054973e-01 3.93458903e-01 3.77808601e-01 -6.55924678e-01 -3.03312421e-01 2.67750323e-01 8.15847740e-02 -5.41162431e-01 1.32420921e+00 7.94692710e-02 3.72613907e-01 9.12095964e-01 1.21496129e+00 5.30606747e-01 -1.18229270e+00 -1.46707833e-01 -2.97129542e-01 -1.88647628e-01 2.72406749e-02 -9.36688423e-01 -1.06142509e+00 1.04201496e+00 9.41212237e-01 5.50444067e-01 1.07799315e+00 -1.38876617e-01 7.75990963e-01 2.69016325e-01 -2.70677179e-01 -7.12527096e-01 -7.39392266e-02 7.17258215e-01 8.72107506e-01 -8.59703422e-01 -2.59190738e-01 -6.25647306e-02 -2.68270254e-01 9.81188536e-01 3.93998742e-01 1.26013374e-02 9.13664818e-01 7.93832362e-01 6.38725817e-01 9.07633230e-02 -7.78556287e-01 -2.70168841e-01 2.56390363e-01 7.99766839e-01 5.47685325e-01 1.34314010e-02 6.74854517e-01 5.63883007e-01 -7.64937758e-01 -4.54579592e-01 6.10701203e-01 6.90606892e-01 -2.68037736e-01 -9.36991215e-01 -4.35964853e-01 3.09572995e-01 -7.09663570e-01 -1.97795331e-01 -3.56580079e-01 5.92752874e-01 -4.15498726e-02 1.34015298e+00 2.49416620e-01 -4.49818194e-01 4.26863223e-01 5.13621390e-01 2.00625092e-01 -6.60156190e-01 -7.38344967e-01 2.16387421e-01 3.30725551e-01 -1.32199705e-01 -3.01357001e-01 -7.13521719e-01 -1.09500039e+00 -4.81152646e-02 -2.91885495e-01 3.00687432e-01 7.78275132e-01 7.28541017e-01 3.73897135e-01 1.20095110e+00 5.35464704e-01 -1.10218954e+00 -4.97747332e-01 -1.33232307e+00 -4.66222435e-01 2.36042902e-01 8.96537423e-01 -4.23793316e-01 -7.99363971e-01 4.79292758e-02]
[15.044066429138184, 5.758535385131836]
0dc0fac5-c88e-4fb4-8f31-c7bdab0c1f50
face-presentation-attack-detection-using
2111.11046
null
https://arxiv.org/abs/2111.11046v2
https://arxiv.org/pdf/2111.11046v2.pdf
FRT-PAD: Effective Presentation Attack Detection Driven by Face Related Task
The robustness and generalization ability of Presentation Attack Detection (PAD) methods is critical to ensure the security of Face Recognition Systems (FRSs). However, in a real scenario, Presentation Attacks (PAs) are various and it is hard to predict the Presentation Attack Instrument (PAI) species that will be used by the attacker. Existing PAD methods are highly dependent on the limited training set and cannot generalize well to unknown PAI species. Unlike this specific PAD task, other face related tasks trained by huge amount of real faces (e.g. face recognition and attribute editing) can be effectively adopted into different application scenarios. Inspired by this, we propose to trade position of PAD and face related work in a face system and apply the free acquired prior knowledge from face related tasks to solve face PAD, so as to improve the generalization ability in detecting PAs. The proposed method, first introduces task specific features from other face related task, then, we design a Cross-Modal Adapter using a Graph Attention Network (GAT) to re-map such features to adapt to PAD task. Finally, face PAD is achieved by using the hierarchical features from a CNN-based PA detector and the re-mapped features. The experimental results show that the proposed method can achieve significant improvements in the complicated and hybrid datasets, when compared with the state-of-the-art methods. In particular, when training on the datasets OULU-NPU, CASIA-FASD, and Idiap Replay-Attack, we obtain HTER (Half Total Error Rate) of 5.48% for the testing dataset MSU-MFSD, outperforming the baseline by 7.39%.
['Raghavendra Ramachandra', 'Feng Liu', 'Christoph Busch', 'Haozhe Liu', 'Wentian Zhang']
2021-11-22
null
null
null
null
['face-presentation-attack-detection']
['computer-vision']
[ 2.32804537e-01 -3.40831548e-01 2.40659371e-01 -2.67544180e-01 -2.94921726e-01 -5.74337900e-01 4.75283831e-01 -4.38853592e-01 1.09076845e-02 2.17206612e-01 -3.34245384e-01 -1.40084639e-01 -2.82252319e-02 -7.87633955e-01 -6.02496743e-01 -6.14456296e-01 -2.83246368e-01 1.33187711e-01 1.03670791e-01 -3.73835951e-01 2.18344882e-01 1.09058785e+00 -1.69894660e+00 4.35918272e-01 5.11815131e-01 1.57328415e+00 -2.52523855e-03 2.70415932e-01 -5.95599264e-02 3.05053174e-01 -9.45877671e-01 -9.08074200e-01 4.56842065e-01 -2.97461331e-01 -3.97226691e-01 -9.49704051e-02 5.69506288e-01 -5.23999214e-01 -5.12874246e-01 1.12144160e+00 7.76199400e-01 -1.01072811e-01 6.90813541e-01 -1.65523422e+00 -7.48502851e-01 3.23652804e-01 -8.13853681e-01 2.39483759e-01 4.61880744e-01 1.33477405e-01 1.66482776e-01 -9.61243391e-01 2.51910061e-01 1.68710887e+00 6.31112456e-01 9.52105284e-01 -7.23437548e-01 -1.52674985e+00 -1.05530433e-02 5.58468103e-01 -1.61880922e+00 -4.82821941e-01 1.05866218e+00 -1.89950094e-01 4.45215940e-01 1.25840709e-01 2.64481418e-02 1.64161956e+00 -4.04598676e-02 3.11681032e-01 1.02206326e+00 -2.32036740e-01 -1.10766783e-01 3.25168461e-01 9.33655351e-02 7.84210742e-01 1.46760389e-01 7.52255768e-02 -4.16865766e-01 -2.53915161e-01 5.19837022e-01 -6.78279027e-02 -4.95167971e-01 2.20043898e-01 -4.05532241e-01 6.57455683e-01 4.68711972e-01 2.56553411e-01 -2.29060650e-01 -3.64601731e-01 4.92311358e-01 1.73274115e-01 2.27071755e-02 1.18640624e-01 -3.34777921e-01 3.34591478e-01 -6.40400767e-01 -2.25050151e-02 7.99210489e-01 8.80099833e-01 7.12960005e-01 3.14302862e-01 -3.05601984e-01 7.89654255e-01 2.35179037e-01 6.65363014e-01 4.00795698e-01 -2.64523953e-01 4.49468672e-01 5.01805723e-01 -2.57467359e-01 -1.39943266e+00 -3.12658370e-01 -2.11285219e-01 -8.68826449e-01 1.95499822e-01 1.93764120e-01 -1.50422499e-01 -1.09009290e+00 1.75182891e+00 2.10693881e-01 6.31624997e-01 -2.35961713e-02 6.03090346e-01 9.24260437e-01 7.23941028e-01 2.54108787e-01 -3.78866076e-01 1.62920189e+00 -4.85245705e-01 -7.80163646e-01 1.02860555e-02 1.21089146e-01 -9.08173621e-01 1.05389738e+00 5.97907126e-01 -4.90280300e-01 -9.34723377e-01 -1.29442394e+00 5.50836444e-01 -6.23135805e-01 3.44772935e-01 2.96015143e-01 1.35882199e+00 -7.74485350e-01 4.05572563e-01 -1.46851569e-01 -3.67452919e-01 8.38727176e-01 6.49681091e-01 -6.58818066e-01 -1.48031726e-01 -1.33964717e+00 6.92677915e-01 3.59461188e-01 3.79630595e-01 -1.30294585e+00 -6.59822822e-01 -5.89413226e-01 2.47673705e-01 5.38607597e-01 5.92678860e-02 7.58575141e-01 -8.98235679e-01 -1.62761772e+00 5.38419962e-01 2.63610423e-01 -2.37575129e-01 2.40998283e-01 -1.54933751e-01 -9.59491789e-01 2.55038232e-01 -3.40329051e-01 3.08418214e-01 1.38811433e+00 -1.24668682e+00 -2.71697551e-01 -4.55370784e-01 3.10707092e-02 -2.86314219e-01 -9.41361725e-01 5.59113920e-01 -3.33536953e-01 -4.29242253e-01 -5.46586335e-01 -6.86559618e-01 5.46111524e-01 8.90010521e-02 -3.20412129e-01 -4.77387130e-01 1.55657983e+00 -8.76109242e-01 1.14073789e+00 -2.38405013e+00 -1.95922121e-01 3.87714922e-01 -1.21635124e-01 1.00564432e+00 -4.15672690e-01 2.75414884e-01 -3.91392797e-01 1.40357807e-01 -6.77209208e-03 -1.64528400e-01 -1.09574787e-01 -5.43362685e-02 -4.87139463e-01 3.60024244e-01 6.42556667e-01 4.12015259e-01 -3.32484603e-01 -3.69567901e-01 2.09938418e-02 5.10627091e-01 -4.05752748e-01 6.43624425e-01 6.97864965e-02 3.97910267e-01 -4.66010571e-01 7.89747357e-01 1.39747822e+00 1.68700397e-01 -4.09199074e-02 -6.14953756e-01 4.02499318e-01 -4.66071129e-01 -1.32178938e+00 1.19249630e+00 -4.76633430e-01 2.84760386e-01 1.52636990e-01 -9.79235113e-01 1.24743092e+00 3.55081558e-01 6.57836273e-02 -4.62550163e-01 4.68526989e-01 8.95924345e-02 1.39185801e-01 -6.13343954e-01 -2.34990433e-01 2.14553714e-01 2.14122653e-01 1.53462335e-01 4.25892651e-01 3.42773378e-01 -1.55079007e-01 5.75637771e-03 8.94579470e-01 -1.11744970e-01 -2.04937141e-02 -8.26192424e-02 1.22195089e+00 -7.27189541e-01 6.27559423e-01 4.30481493e-01 -2.09129751e-01 3.46415490e-01 5.63892007e-01 -4.64006931e-01 -3.96357208e-01 -9.67890561e-01 -2.19520733e-01 9.37628031e-01 1.28656790e-01 -4.59530115e-01 -1.11426902e+00 -1.21994865e+00 -6.90739229e-02 4.15168971e-01 -6.38990283e-01 -4.13775325e-01 -7.35791743e-01 -6.99147642e-01 1.00589097e+00 3.90655965e-01 1.02198970e+00 -1.18614721e+00 -4.37336043e-02 -1.40530705e-01 7.03668222e-02 -1.33286607e+00 -3.47869009e-01 -3.90992403e-01 -1.99300870e-02 -1.23832929e+00 -5.26584864e-01 -8.41295004e-01 5.16418993e-01 7.41849393e-02 5.10353148e-01 3.58591139e-01 -5.30331731e-01 3.32229048e-01 -2.76133835e-01 -6.22539639e-01 -3.03828388e-01 -1.99958459e-01 3.46387029e-01 7.54663050e-01 5.02332509e-01 -4.02942359e-01 -5.24115324e-01 6.03648782e-01 -8.41910779e-01 -6.78133667e-01 5.93557656e-01 8.41358840e-01 -5.76325133e-03 7.76911527e-02 9.10673499e-01 -7.24451542e-01 6.28012180e-01 -4.90122646e-01 -6.69753075e-01 4.67828155e-01 -2.25146711e-01 -4.49309051e-01 8.96583378e-01 -1.05127335e+00 -1.17767394e+00 8.99805874e-03 -2.38063186e-01 -8.32249522e-01 -3.43522817e-01 1.95141733e-02 -1.00469267e+00 -6.27008319e-01 7.51664281e-01 1.73418254e-01 1.88377708e-01 -3.47983271e-01 4.69364747e-02 9.80138838e-01 4.46306705e-01 -5.85938871e-01 1.13631940e+00 9.96336043e-02 3.07851821e-01 -8.64688933e-01 -4.03903067e-01 1.85869113e-01 -3.16303283e-01 -3.83141309e-01 6.71586037e-01 -7.40989089e-01 -1.15518379e+00 8.40149939e-01 -1.25582790e+00 2.43826240e-01 5.04116178e-01 1.87871918e-01 -1.65224522e-01 8.14844072e-01 -5.82598686e-01 -9.14925158e-01 -4.90295172e-01 -1.41786075e+00 9.89702940e-01 4.33139294e-01 2.93968946e-01 -4.61083949e-01 -5.23269475e-01 2.42014244e-01 4.70430851e-01 1.57564476e-01 9.74825263e-01 -9.83488083e-01 -3.06770951e-01 -3.63925010e-01 -4.26159918e-01 6.88909411e-01 2.46710375e-01 7.42436647e-02 -1.42436612e+00 -3.82486969e-01 2.62639582e-01 -3.00440609e-01 4.71948743e-01 -2.58820534e-01 1.64686549e+00 -4.71477002e-01 -3.88262719e-01 7.21506596e-01 1.14096653e+00 4.34656143e-01 1.00388336e+00 -5.21228500e-02 6.50636077e-01 8.12458694e-01 6.02270186e-01 3.07074428e-01 -2.04169542e-01 8.00404668e-01 5.34593105e-01 1.83449909e-01 -2.19906330e-01 -1.88437209e-01 5.41631520e-01 1.55022874e-01 1.35983333e-01 -4.22615707e-01 -7.10222781e-01 4.77824658e-02 -1.33063769e+00 -8.91473174e-01 2.04804659e-01 1.97905672e+00 4.27114576e-01 -1.26593903e-01 7.98011422e-02 3.29994529e-01 1.07815731e+00 -5.85166775e-02 -3.60933334e-01 -3.78730595e-01 2.03951332e-03 6.47748351e-01 1.94496408e-01 -9.76006314e-02 -1.16406465e+00 9.45820570e-01 4.94293022e+00 1.31897354e+00 -1.42621469e+00 1.41246229e-01 6.19596660e-01 4.34393078e-01 4.12188739e-01 -5.16202748e-01 -1.01105320e+00 6.72703624e-01 9.23979223e-01 1.51651591e-01 6.40101075e-01 8.53214681e-01 -3.17550927e-01 6.05017781e-01 -1.07481229e+00 1.43005848e+00 6.87061012e-01 -7.42525876e-01 4.18987364e-01 -8.20660293e-02 1.37057334e-01 -8.39067161e-01 5.18794537e-01 3.67662579e-01 -2.63517678e-01 -1.25639820e+00 2.51731426e-01 1.62793428e-01 9.49938238e-01 -9.44320202e-01 8.60022902e-01 1.79480880e-01 -1.24680328e+00 -5.69834292e-01 -6.81638420e-01 4.26088065e-01 -2.74525493e-01 6.67273924e-02 -8.92974973e-01 7.73130298e-01 8.73937309e-01 2.93081194e-01 -7.61620760e-01 7.22583175e-01 -2.14709580e-01 6.32753551e-01 -3.22458446e-01 6.27388433e-02 -1.94276601e-01 1.26913741e-01 4.72935766e-01 9.73047912e-01 6.83452606e-01 1.01617612e-01 -1.10976890e-01 6.61102891e-01 -3.46985966e-01 3.07785302e-01 -6.47620797e-01 -5.35070561e-02 6.27117872e-01 1.68277597e+00 -2.85878360e-01 2.67403368e-02 -3.79230201e-01 8.81383777e-01 2.80362159e-01 3.03628832e-01 -1.02819955e+00 -5.22102773e-01 6.23324990e-01 3.89766693e-02 4.61277038e-01 2.50754863e-01 3.40853959e-01 -8.76873732e-01 9.89296511e-02 -1.15267098e+00 6.15283430e-01 -6.80188060e-01 -1.63859963e+00 1.13523936e+00 -1.01832002e-01 -1.30199182e+00 -3.94174606e-02 -1.12688911e+00 -9.40401435e-01 9.82160807e-01 -1.39168167e+00 -1.66416585e+00 -4.80717570e-01 1.03621745e+00 3.94622475e-01 -7.50207961e-01 8.41254711e-01 7.03143239e-01 -8.96147311e-01 1.31114709e+00 -6.63004398e-01 4.48543161e-01 9.21395361e-01 -4.99216706e-01 1.24694891e-01 9.48292434e-01 -5.91719747e-02 6.14693522e-01 2.28587016e-01 -5.34193754e-01 -1.50318336e+00 -1.28349054e+00 1.15410902e-01 -2.67428488e-01 4.67799097e-01 -6.41548693e-01 -1.13776135e+00 4.04290199e-01 1.49172023e-01 2.36826167e-01 6.20036244e-01 -2.97306538e-01 -8.83093655e-01 -5.68527699e-01 -1.62923563e+00 5.51087618e-01 1.11285603e+00 -5.06632686e-01 -2.57110894e-01 2.96898335e-01 4.58298415e-01 -8.71557519e-02 -6.90049231e-01 6.79051220e-01 4.55922633e-01 -6.81784809e-01 1.19333720e+00 -6.69870913e-01 -4.54508103e-02 -3.01163912e-01 -1.65697798e-01 -1.03495371e+00 -2.33354717e-01 -6.37671888e-01 -1.30452827e-01 1.78261459e+00 1.88690927e-02 -7.57052362e-01 4.72293168e-01 2.46002421e-01 5.77374995e-02 -5.01398504e-01 -9.41786587e-01 -8.49290550e-01 -1.55393958e-01 -1.09730326e-01 1.16970515e+00 8.85471642e-01 -4.09776568e-01 8.23872909e-02 -5.71062982e-01 8.31393957e-01 6.75022840e-01 -3.75418663e-01 8.80013704e-01 -1.39311206e+00 -9.20268819e-02 -1.96338281e-01 -8.01488817e-01 -5.22836506e-01 5.02895713e-01 -6.94606483e-01 -2.51299322e-01 -6.08968616e-01 -2.58090310e-02 -2.22342089e-01 -4.53956038e-01 5.97272873e-01 -2.31861487e-01 3.80856097e-01 4.00860995e-01 -8.73738229e-02 -8.42575207e-02 4.93342221e-01 9.26650226e-01 -2.84266800e-01 1.02251165e-01 -7.83316568e-02 -5.84996223e-01 6.10776305e-01 6.64171934e-01 -2.80696005e-01 -3.84145677e-01 -5.14292903e-02 -4.58410591e-01 -2.96182223e-02 3.77754331e-01 -1.25661123e+00 1.57037571e-01 -1.60708204e-02 7.55631149e-01 -4.10273999e-01 4.96379405e-01 -1.01460421e+00 1.09126464e-01 3.15022975e-01 1.90549195e-01 -1.68980900e-02 5.49908400e-01 3.97480786e-01 -4.62636463e-02 -3.12582016e-01 8.76667917e-01 2.12712452e-01 -7.79816329e-01 7.72436202e-01 -8.94679278e-02 -3.23416829e-01 1.29713202e+00 -2.22590044e-01 -6.93331897e-01 -2.57921547e-01 -4.10081208e-01 -1.47861466e-01 7.65380915e-04 7.22989380e-01 8.29449177e-01 -1.38737929e+00 -7.74252772e-01 7.09533215e-01 1.49646774e-01 -5.13801515e-01 5.58729768e-01 3.35224688e-01 -2.83888519e-01 -6.39583766e-02 -6.93047941e-01 -4.23999131e-01 -1.52784646e+00 1.02075922e+00 3.20019037e-01 1.29983574e-01 -1.09026223e-01 7.09764302e-01 5.07382393e-01 -2.37254471e-01 2.68441379e-01 5.36020458e-01 -6.18418396e-01 5.10187522e-02 9.63679671e-01 2.04039127e-01 2.61439770e-01 -7.85617471e-01 -4.82985914e-01 6.95777595e-01 -2.44417027e-01 3.60246778e-01 1.02221179e+00 3.66425246e-01 -1.81285039e-01 -4.74788547e-01 1.36740243e+00 8.87925625e-02 -8.72980893e-01 -8.86800792e-03 -2.51397818e-01 -7.05533326e-01 -3.53762686e-01 -6.99063897e-01 -1.38473845e+00 1.25759244e+00 1.22076392e+00 1.02597371e-01 1.41853476e+00 -2.67279834e-01 6.83866382e-01 2.43822590e-01 4.29624736e-01 -5.54585457e-01 4.16954428e-01 1.44941926e-01 1.22684216e+00 -9.59320903e-01 -2.90457010e-01 -8.63829076e-01 -4.87731099e-01 1.31091678e+00 1.15406597e+00 -1.71547160e-02 8.96036148e-01 1.87447459e-01 -2.38017552e-02 -6.43144995e-02 -3.36271912e-01 1.37010247e-01 3.49121630e-01 1.14742041e+00 -1.09542057e-01 -2.90692747e-01 6.29564971e-02 8.94362450e-01 -8.55396464e-02 -2.52650946e-01 2.36429870e-01 5.48200369e-01 -1.11280113e-01 -1.15807450e+00 -5.40219247e-01 4.59657818e-01 -6.59160376e-01 2.14068070e-01 -4.66123253e-01 7.24355936e-01 3.40788424e-01 1.09928215e+00 -5.02723642e-02 -6.67016149e-01 3.72327089e-01 1.15759842e-01 4.33774978e-01 -3.61718416e-01 -6.86289132e-01 -3.35444301e-01 -2.17178002e-01 -4.03271466e-01 -6.64346740e-02 -2.96704501e-01 -7.42231011e-01 -3.74037176e-01 -5.77446282e-01 2.45605763e-02 5.24449348e-01 6.48858070e-01 4.69656467e-01 3.47595870e-01 1.15180254e+00 -6.96486950e-01 -7.21515834e-01 -1.03111625e+00 -4.42630112e-01 5.72042286e-01 1.89576414e-04 -1.05696189e+00 -4.82764214e-01 -2.16985181e-01]
[13.061765670776367, 1.1314411163330078]
3ed50582-2c24-467c-a687-b61d34b6e1aa
representation-and-synthesis-of-geometric
null
null
https://aclanthology.org/2022.signlang-1.9
https://aclanthology.org/2022.signlang-1.9.pdf
Representation and Synthesis of Geometric Relocations
One of the key features of signed discourse is the geometric placements of gestural units in signing space. Signers use the geometry of signing space to describe the placements and forms of objects and also use it to contrast participants or locales in a story. Depending on the specific functions of the placement in the discourse, features such as geometric precision, gaze redirection and timing will all differ. A signing avatar must capture these differences to sign such discourse naturally. This paper builds on prior work that animated geometric depictions to enable a signing avatar to more naturally use signing space for opposing participants and concepts in discourse. Building from a structured linguistic description of a signed newscast, they system automatically synthesizes animation that correctly utilizes signing space to lay out the opposing locales in the report. The efficacy of the approach is demonstrated through comparisons of the avatar’s motion with the source signing.
['John McDonald', 'Michael Filhol']
null
null
null
null
signlang-lrec-2022-6
['gaze-redirection']
['computer-vision']
[ 2.74083674e-01 2.64429718e-01 -5.16336691e-03 -4.75309342e-01 -3.17295432e-01 -1.19323838e+00 1.23241103e+00 1.42302826e-01 9.14342254e-02 2.79224664e-01 9.38301444e-01 -3.21532309e-01 -4.89414372e-02 -4.46186155e-01 -3.62822920e-01 -1.66689873e-01 -5.73836081e-02 2.36850977e-01 2.15555459e-01 -6.48531735e-01 8.69123697e-01 6.42418325e-01 -1.47436476e+00 7.33237147e-01 1.31779984e-01 3.51746321e-01 -1.41173273e-01 1.17324781e+00 -1.79403439e-01 1.02431798e+00 -1.28323209e+00 6.11400651e-03 1.01151623e-01 -8.88252199e-01 -6.00177467e-01 1.21565096e-01 6.61894262e-01 -4.13631856e-01 -7.22143799e-02 5.32563567e-01 1.42996043e-01 3.66605036e-02 8.76820385e-01 -1.22735190e+00 -4.41143304e-01 8.85908186e-01 -4.81554776e-01 -3.33677620e-01 9.51715052e-01 -4.31172550e-02 1.19788039e+00 -4.25782293e-01 1.26226699e+00 1.31348574e+00 6.47111952e-01 5.53988934e-01 -9.18107927e-01 -4.51537192e-01 -8.08108423e-04 -2.04854697e-01 -9.28375721e-01 -5.81563950e-01 1.04632235e+00 -7.32060909e-01 3.22205693e-01 7.56624520e-01 1.15757024e+00 7.57268906e-01 -1.12591654e-01 5.15143752e-01 8.93725514e-01 -7.88873911e-01 2.37769678e-01 -1.63503289e-01 2.59356737e-01 4.75686789e-01 -1.54684007e-01 -7.55333155e-02 -8.11476052e-01 -4.91167068e-01 9.66611743e-01 -6.06426001e-01 -4.52724874e-01 -4.39337522e-01 -1.55592418e+00 7.12570429e-01 6.48689196e-02 5.67642450e-01 1.65418386e-02 6.45734191e-01 5.21666765e-01 -3.53729501e-02 3.37229744e-02 7.60282338e-01 8.73827562e-02 -6.64467931e-01 -9.00798917e-01 8.83481920e-01 1.02693260e+00 7.32586265e-01 1.51448041e-01 -1.93592772e-01 -8.07029605e-02 1.56853706e-01 6.18873477e-01 9.46749970e-02 -7.60806352e-02 -1.41548860e+00 3.57228547e-01 7.60339618e-01 3.00856709e-01 -1.43169904e+00 -3.29910129e-01 5.05716681e-01 1.41388074e-01 7.99774170e-01 7.54505932e-01 -1.89084008e-01 -3.82434756e-01 1.48667967e+00 5.10030866e-01 -1.04312606e-01 -8.15665200e-02 8.98330629e-01 8.87402236e-01 4.64742064e-01 4.47266102e-02 1.15460344e-01 1.67813027e+00 -4.11959708e-01 -9.86751914e-01 9.16800201e-02 1.02130365e+00 -8.09014082e-01 1.05524957e+00 6.50300179e-03 -1.22431135e+00 3.73021290e-02 -1.45502746e+00 -2.82730877e-01 3.30999121e-02 -1.30036250e-01 3.23932737e-01 6.86119795e-01 -6.62271082e-01 6.52358830e-01 -6.99865460e-01 -3.00681293e-01 4.15329216e-03 -2.22552158e-02 5.32916039e-02 6.36860013e-01 -7.82377958e-01 9.88975227e-01 1.69513315e-01 -7.26605579e-02 -9.78373140e-02 -8.01555157e-01 -1.00924802e+00 -3.70530397e-01 1.31738067e-01 -3.03370237e-01 1.68528306e+00 -1.20168734e+00 -1.71175110e+00 9.96685207e-01 2.21619010e-01 -8.84127915e-02 9.82454181e-01 -1.60082579e-01 -1.99150473e-01 3.51138473e-01 -6.91301897e-02 4.84754413e-01 5.88139713e-01 -1.70048642e+00 -3.98417652e-01 -1.76461056e-01 5.94245136e-01 3.39740604e-01 3.28634948e-01 6.38836503e-01 1.65349945e-01 -9.66957510e-01 2.48244107e-01 -7.33241558e-01 4.85842884e-01 3.85506123e-01 -4.73359853e-01 1.07953392e-01 1.67364895e+00 -8.24855685e-01 1.46415377e+00 -2.10752058e+00 -1.71273071e-02 3.80061358e-01 4.01196927e-01 -7.05863386e-02 2.07104430e-01 7.93602824e-01 5.00638559e-02 1.66061357e-01 -2.57130533e-01 -1.10032611e-01 3.10812950e-01 5.62040545e-02 -1.78636342e-01 8.82119656e-01 -3.01265776e-01 9.12899137e-01 -9.63086009e-01 -5.31802237e-01 2.64549032e-02 7.40598023e-01 -4.09350604e-01 -6.70808777e-02 -3.95397812e-01 3.95972669e-01 -5.28680086e-01 3.30810249e-01 1.45724148e-01 1.53427392e-01 6.23166025e-01 -2.61698633e-01 -4.79023576e-01 5.54000199e-01 -1.39135599e+00 1.50920188e+00 -7.82409608e-02 1.20851517e+00 5.26258409e-01 -2.04265475e-01 9.46029127e-01 4.02033091e-01 5.59797809e-02 -6.98471963e-02 3.94014925e-01 2.76368022e-01 8.65758676e-03 -6.11763716e-01 9.41430151e-01 -1.43963426e-01 -3.78058821e-01 1.13158619e+00 -8.43388617e-01 -6.73794806e-01 -1.18776353e-03 3.47981542e-01 7.97960579e-01 6.90313578e-01 6.12183452e-01 -2.31514305e-01 1.06254749e-01 1.16545096e-01 -5.81668057e-02 1.35042638e-01 1.03617974e-01 9.42476690e-01 8.28829169e-01 -4.62010533e-01 -1.24346423e+00 -6.17955863e-01 1.81585521e-01 1.03793752e+00 2.25116551e-01 -7.09152937e-01 -8.70890439e-01 -4.81628001e-01 -2.06579447e-01 1.12523532e+00 -8.41487527e-01 3.16666067e-01 -1.32440937e+00 9.70766991e-02 8.25185061e-01 4.01201755e-01 3.55544798e-02 -8.92874599e-01 -1.86608338e+00 6.28984869e-02 2.77134404e-02 -7.23314583e-01 -6.68610275e-01 -4.69171017e-01 -1.77220061e-01 -1.44109774e+00 -5.97050667e-01 -6.06829822e-01 6.86647713e-01 -5.45819178e-02 7.93897331e-01 4.66106355e-01 -5.74429408e-02 4.03489292e-01 -6.72178745e-01 -4.87755597e-01 -1.11563993e+00 -5.42115331e-01 -2.85735279e-01 -1.66073740e-01 -3.75128239e-01 -3.81206453e-01 -3.56695950e-01 2.41510943e-01 -9.46243048e-01 6.60484731e-01 -6.56246662e-01 3.69318396e-01 -1.58590615e-01 -6.32257998e-01 -1.80855930e-01 -8.61711860e-01 8.54575992e-01 6.43522069e-02 -2.35122472e-01 3.39140385e-01 2.32394218e-01 8.00832957e-02 2.64998693e-02 -6.43656254e-01 -6.43752933e-01 -4.20734823e-01 6.55458391e-01 3.21181446e-01 1.56108197e-02 2.37054810e-01 -1.61506444e-01 9.53737274e-02 6.58292413e-01 -5.87907195e-01 -9.53513160e-02 -1.89743087e-01 5.64050734e-01 4.72573102e-01 5.54246187e-01 -7.62368798e-01 7.86328733e-01 5.37378311e-01 1.19641840e-01 -6.62560940e-01 -1.29337758e-01 2.21701846e-01 -8.91721308e-01 -5.78694463e-01 8.74447942e-01 -2.13926494e-01 -1.05435216e+00 3.24940920e-01 -1.63297057e+00 -4.91598368e-01 -6.56716406e-01 3.14654619e-01 -8.43871117e-01 3.81849587e-01 -2.98614532e-01 -8.83861721e-01 2.20694974e-01 -1.02338386e+00 1.16452742e+00 6.06349371e-02 -1.36072230e+00 -1.04041266e+00 9.04248133e-02 1.09144464e-01 9.73016471e-02 1.30358517e+00 1.05715942e+00 -6.35503948e-01 -1.70479283e-01 -1.94146231e-01 1.11563951e-01 -4.83372092e-01 2.11872548e-01 7.99886703e-01 -7.02987134e-01 2.50296921e-01 -2.90709049e-01 9.87877953e-04 -7.14336932e-02 2.72771448e-01 1.24530509e-01 -7.49276459e-01 -2.55317152e-01 2.80055434e-01 7.66692817e-01 4.72858340e-01 4.58552331e-01 6.68584466e-01 7.19331622e-01 1.16838324e+00 4.44859833e-01 4.98239547e-01 3.93343091e-01 9.46642041e-01 1.98885158e-01 6.60147965e-02 -3.17120463e-01 -4.17781115e-01 4.26088244e-01 3.20539236e-01 -2.39042535e-01 -1.22822233e-01 -9.60467458e-01 2.65776694e-01 -1.80255389e+00 -1.19662011e+00 -5.79632699e-01 1.80906296e+00 6.74455285e-01 -1.10011563e-01 2.43845120e-01 3.99984062e-01 8.01948488e-01 7.15676129e-01 3.91167551e-02 -8.70741248e-01 -9.80596393e-02 -2.30758175e-01 3.33269626e-01 8.57092261e-01 -6.67532206e-01 5.22254229e-01 6.46655130e+00 4.31600571e-01 -1.00942993e+00 -1.91069975e-01 2.39004999e-01 -1.42554324e-02 -7.02592790e-01 3.69351596e-01 -2.85118341e-01 8.17003474e-02 4.35358852e-01 -2.31023118e-01 1.92411795e-01 1.69377282e-01 6.02631927e-01 -3.22210997e-01 -1.54290915e+00 6.78640544e-01 3.76866996e-01 -1.75823915e+00 -1.65131629e-01 -2.47821704e-01 4.60943341e-01 -1.04572678e+00 -3.00349742e-01 -5.33158422e-01 2.44106114e-01 -1.12629592e+00 1.72645366e+00 4.12353754e-01 9.31970358e-01 -3.94070417e-01 1.23921305e-01 5.56612723e-02 -1.19338405e+00 3.66189212e-01 8.48266184e-01 -5.01030982e-01 7.21823871e-01 -5.30064583e-01 -1.30073798e+00 1.94967374e-01 8.44734721e-03 2.29797661e-01 -1.87122688e-01 6.77000761e-01 -5.57392597e-01 4.88182724e-01 -3.18401814e-01 -4.11422372e-01 2.03500330e-01 -2.01280639e-01 8.99257302e-01 1.33942437e+00 3.46133709e-01 3.28949660e-01 -7.43906349e-02 8.40929449e-01 2.75328875e-01 1.51303232e-01 -5.38540363e-01 -1.96163446e-01 6.65806293e-01 7.64943779e-01 -9.64986742e-01 -3.54057670e-01 1.00159355e-01 7.54572153e-01 -5.43823183e-01 3.95996124e-01 -8.32537949e-01 -3.72367054e-01 6.84802234e-01 8.49662960e-01 3.02657522e-02 -6.46582603e-01 -6.60514295e-01 -4.81028646e-01 1.36854708e-01 -9.34428871e-01 3.66886444e-02 -1.23797441e+00 -3.05014402e-01 2.75910825e-01 5.45525253e-01 -1.21574438e+00 -4.29205120e-01 -2.67602623e-01 -9.34958398e-01 6.50161505e-01 -7.14736223e-01 -1.59565175e+00 -2.13468120e-01 -1.41996769e-02 4.31498587e-01 3.36680114e-01 7.68737614e-01 -3.90692264e-01 1.68615669e-01 2.35192984e-01 -3.85738343e-01 5.01711488e-01 4.65906292e-01 -1.18335867e+00 6.52300775e-01 5.40710866e-01 2.48721182e-01 6.18658364e-01 1.31243539e+00 -6.66286945e-01 -1.11579490e+00 -9.36809108e-02 1.07085013e+00 -7.18217969e-01 9.09570277e-01 -5.73124409e-01 -5.90272665e-01 8.64944875e-01 3.04965734e-01 -3.43000621e-01 5.83575785e-01 -3.66230637e-01 -6.79308355e-01 6.26480520e-01 -1.21138752e+00 8.95195305e-01 8.91880572e-01 -5.65843523e-01 -1.10097277e+00 1.11207962e-01 4.36082333e-01 -9.87508714e-01 -3.27738553e-01 -5.23825660e-02 1.04991591e+00 -9.74226058e-01 7.38660693e-01 -5.93175232e-01 3.94518167e-01 -5.16298473e-01 -1.26317009e-01 -9.71767902e-01 3.58938456e-01 -1.23995006e+00 5.27468026e-02 1.35679054e+00 9.14291516e-02 -2.14978248e-01 6.73818767e-01 9.82368290e-01 -3.85005325e-02 -1.48561075e-01 -8.10710907e-01 -2.60651112e-01 -7.93349594e-02 -5.09406984e-01 7.35850573e-01 1.14475560e+00 6.62744224e-01 -1.32228285e-01 -9.13669020e-02 -1.25636980e-01 1.58039734e-01 1.47856921e-02 9.19941425e-01 -1.32931018e+00 -2.27262989e-01 -6.21132433e-01 -5.12223363e-01 -9.05033231e-01 -1.61946103e-01 -5.43498993e-01 -1.28625050e-01 -1.64574909e+00 -3.90355617e-01 -5.29946566e-01 9.13178802e-01 3.97679269e-01 2.49507040e-01 1.80677369e-01 8.24641466e-01 2.90785491e-01 -2.41758481e-01 6.71406835e-02 1.36811996e+00 9.58115682e-02 -7.53534734e-01 -4.33242083e-01 -3.79906386e-01 1.17890966e+00 5.44506252e-01 -3.23143423e-01 -1.22251347e-01 -3.55569035e-01 3.94980013e-01 3.00581306e-01 5.20374417e-01 -6.22858524e-01 2.02027619e-01 -4.21714664e-01 -4.44798023e-02 -4.55677241e-01 5.39927185e-01 -6.83644474e-01 4.51276481e-01 5.01126409e-01 -7.65385568e-01 4.03873980e-01 -3.60032022e-02 -6.68639466e-02 2.37098083e-01 -2.44953722e-01 3.76999468e-01 9.32278484e-02 -3.72132063e-01 -6.81815326e-01 -4.42645848e-01 4.31900434e-02 1.29108274e+00 -1.02753544e+00 -3.02773774e-01 -7.34800041e-01 -5.51218510e-01 -1.33077249e-01 9.78767693e-01 3.41381818e-01 4.07007784e-01 -1.23950636e+00 -7.76581228e-01 -8.20406675e-02 -1.41953185e-01 5.76034077e-02 -1.44467548e-01 4.56025541e-01 -1.24810815e+00 -3.78050506e-01 -4.88378368e-02 -2.08336458e-01 -1.84388757e+00 -1.49924651e-01 2.77882606e-01 2.30265662e-01 -8.55033755e-01 6.19952142e-01 4.37525138e-02 -7.71698654e-02 3.92470099e-02 -6.66448355e-01 -2.20558122e-01 4.75888848e-01 9.20526385e-01 5.35000205e-01 -4.17456418e-01 -1.19289160e+00 -3.39112341e-01 8.65750194e-01 4.29691702e-01 -8.44741285e-01 1.21528280e+00 -1.14875853e-01 7.92430341e-02 8.27512681e-01 1.06478584e+00 8.17468226e-01 -1.45500195e+00 3.23604554e-01 2.60035425e-01 -5.83803952e-01 -5.06390870e-01 -7.72421360e-01 -3.43048453e-01 5.19329190e-01 -2.33744100e-01 2.97925323e-01 2.78845996e-01 2.37235382e-01 5.27870119e-01 -8.97882283e-02 2.54311502e-01 -1.01522708e+00 4.72831336e-04 4.90157425e-01 1.46403468e+00 -4.96895432e-01 2.68858880e-01 -4.30370182e-01 -8.12333703e-01 1.66226161e+00 4.54570130e-02 -1.83446854e-01 1.42048389e-01 8.25511158e-01 6.73561275e-01 -6.52146816e-01 -1.02448255e-01 6.10810757e-01 2.41543308e-01 7.72248626e-01 7.67374992e-01 2.55598992e-01 -8.43951583e-01 2.62109756e-01 -8.83344233e-01 -4.20118272e-01 9.65306342e-01 1.49978101e+00 -3.36858749e-01 -9.35381532e-01 -9.54846203e-01 -2.88282335e-01 -1.14553258e-01 3.10313016e-01 -9.31689739e-01 1.20488083e+00 8.19840282e-02 8.79962564e-01 2.96208084e-01 -1.06892362e-01 5.10752261e-01 -5.49759343e-02 6.22678876e-01 -6.47286177e-01 -1.21726108e+00 -3.68555859e-02 6.47617340e-01 -2.69753784e-01 -7.48531759e-01 -9.75577891e-01 -2.01232171e+00 -3.39949965e-01 -1.36202589e-01 1.17013983e-01 9.07446802e-01 8.97844732e-01 -3.89749855e-02 3.79559964e-01 7.14842603e-02 -1.20972776e+00 -2.46616706e-01 -6.87061906e-01 -5.16802013e-01 7.19956100e-01 5.25984824e-01 -4.61052746e-01 -2.76761889e-01 3.59732419e-01]
[11.200165748596191, 0.8048818707466125]
4f7665b1-915d-4672-b661-257ec1f8b4e0
wave-simulation-in-non-smooth-media-by-pinn
2208.08276
null
https://arxiv.org/abs/2208.08276v2
https://arxiv.org/pdf/2208.08276v2.pdf
Wave simulation in non-smooth media by PINN with quadratic neural network and PML condition
Frequency-domain simulation of seismic waves plays an important role in seismic inversion, but it remains challenging in large models. The recently proposed physics-informed neural network (PINN), as an effective deep learning method, has achieved successful applications in solving a wide range of partial differential equations (PDEs), and there is still room for improvement on this front. For example, PINN can lead to inaccurate solutions when PDE coefficients are non-smooth and describe structurally-complex media. In this paper, we solve the acoustic and visco-acoustic scattered-field wave equation in the frequency domain with PINN instead of the wave equation to remove source singularity. We first illustrate that non-smooth velocity models lead to inaccurate wavefields when no boundary conditions are implemented in the loss function. Then, we add the perfectly matched layer (PML) conditions in the loss function of PINN and design a quadratic neural network to overcome the detrimental effects of non-smooth models in PINN. We show that PML and quadratic neurons improve the results as well as attenuation and discuss the reason for this improvement. We also illustrate that a network trained during a wavefield simulation can be used to pre-train the neural network of another wavefield simulation after PDE-coefficient alteration and improve the convergence speed accordingly. This pre-training strategy should find application in iterative full waveform inversion (FWI) and time-lag target-oriented imaging when the model perturbation between two consecutive iterations or two consecutive experiments can be small.
['Jianwei Ma', 'Stephane Operto', 'Hossein S. Aghamiry', 'Yanqi Wu']
2022-08-16
null
null
null
null
['seismic-inversion']
['miscellaneous']
[ 2.79504657e-01 -1.40255019e-01 7.04041362e-01 1.18731298e-01 -7.33703911e-01 2.21194960e-02 1.04215592e-01 -2.54863858e-01 -6.01197183e-01 9.52121079e-01 -3.03349085e-03 -2.32969016e-01 -5.96748888e-01 -8.56466651e-01 -7.33653128e-01 -1.04699898e+00 -3.35589856e-01 3.97225618e-01 4.63817000e-01 -5.17537713e-01 -4.76836506e-03 7.31914937e-01 -1.29902875e+00 1.51390314e-01 8.68415594e-01 8.87245357e-01 6.70218319e-02 5.24650812e-01 9.42908823e-02 7.84489393e-01 -4.12541628e-01 8.74390602e-02 1.10631146e-01 -4.36876595e-01 -6.35640502e-01 -3.96051645e-01 1.17645063e-01 -5.37902951e-01 -2.46000156e-01 9.89314854e-01 1.09530103e+00 4.59998518e-01 8.16536725e-01 -8.91903639e-01 -2.89935470e-01 3.99484664e-01 -7.33072996e-01 2.27599382e-01 -3.04540657e-02 2.28404433e-01 3.50457072e-01 -1.06936586e+00 3.15049320e-01 9.83021557e-01 1.50791061e+00 3.59074533e-01 -1.04580033e+00 -5.32311738e-01 -6.11239851e-01 2.26688311e-01 -1.34815478e+00 -1.40478730e-01 1.13114512e+00 -6.03858054e-01 8.49682510e-01 1.93813160e-01 6.59894943e-01 6.83931768e-01 4.06806886e-01 1.31288469e-01 7.75443733e-01 -4.95137513e-01 2.01862112e-01 -3.50954145e-01 6.08893530e-03 4.56960857e-01 3.06091160e-01 4.25181419e-01 -3.31862390e-01 -2.99083441e-01 1.14471781e+00 -3.72421473e-01 -6.86193168e-01 6.06785007e-02 -8.68163645e-01 9.08000052e-01 4.26867276e-01 6.22207046e-01 -6.35991573e-01 3.67024273e-01 2.19958469e-01 1.83507830e-01 5.99017620e-01 9.41498220e-01 -3.21330935e-01 6.48430660e-02 -1.30486715e+00 5.49108863e-01 7.12414861e-01 1.11300126e-01 7.73694932e-01 7.68744290e-01 5.31747676e-02 9.39929247e-01 3.10399294e-01 6.95749044e-01 3.37596297e-01 -1.23973596e+00 -5.21594621e-02 -2.87249591e-02 2.77840108e-01 -1.10033453e+00 -7.14313924e-01 -6.88916326e-01 -1.22795713e+00 6.40383065e-01 6.26361847e-01 -6.06357753e-01 -8.76731098e-01 1.37422144e+00 3.10057253e-01 7.08720148e-01 1.61551431e-01 1.23079944e+00 9.27599430e-01 1.06887782e+00 -1.26988456e-01 -2.26443380e-01 9.35343683e-01 -5.91347992e-01 -6.75955415e-01 -2.16436908e-01 7.09849536e-01 -7.29037106e-01 6.73427880e-01 2.96251595e-01 -1.38441908e+00 -2.92903662e-01 -9.86382782e-01 1.74534738e-01 4.16956432e-02 -4.99573737e-01 2.61018395e-01 1.26334399e-01 -1.18286467e+00 1.10967445e+00 -8.72906864e-01 3.97644937e-01 2.35483035e-01 3.30550879e-01 -1.73320659e-02 2.65002958e-02 -1.67166173e+00 8.26926947e-01 -2.13653445e-02 5.00465751e-01 -5.82958221e-01 -1.41784823e+00 -6.39363110e-01 3.37401003e-01 9.97466035e-03 -6.38135135e-01 9.79358673e-01 -8.83768797e-01 -1.56653571e+00 3.16358775e-01 3.71761262e-01 -2.36281887e-01 5.66839039e-01 3.48744774e-03 -2.41721541e-01 2.01762825e-01 -2.07550645e-01 1.72110647e-01 7.68442750e-01 -1.44205117e+00 -1.06790930e-01 3.31032544e-01 -1.14163242e-01 -4.48872037e-02 3.63987386e-02 -9.06567648e-02 9.71289128e-02 -7.41889179e-01 1.28432900e-01 -6.67039990e-01 -5.03093898e-01 1.48158982e-01 -1.32316843e-01 3.86329234e-01 6.70727849e-01 -1.04474902e+00 8.89698625e-01 -1.97068596e+00 -2.17781384e-02 3.87790382e-01 1.75899327e-01 5.07442832e-01 -2.32459977e-01 5.99111438e-01 -3.21328998e-01 -1.81179255e-01 -9.39621866e-01 7.65079781e-02 -4.58577514e-01 1.24583535e-01 -1.85159341e-01 5.96448421e-01 3.11299205e-01 5.16424716e-01 -5.59046745e-01 -2.28542000e-01 -3.41789126e-02 9.08353567e-01 -8.62865448e-01 6.30475506e-02 -5.16017377e-02 7.61777878e-01 -4.35670286e-01 1.33970395e-01 1.29629111e+00 -7.13027194e-02 -4.01317328e-01 -3.95442992e-01 -5.27854741e-01 -1.80046871e-01 -1.34491563e+00 1.05790949e+00 -7.22544730e-01 7.78100848e-01 6.37500346e-01 -1.41566169e+00 7.15438128e-01 5.98244071e-01 5.71708143e-01 -7.62322366e-01 1.80910304e-01 6.50998056e-01 4.17143583e-01 -8.86031210e-01 2.39560343e-02 -7.37962782e-01 5.53619266e-01 2.22238585e-01 -2.17185691e-01 -4.15115505e-01 -1.05148792e-01 -3.41731995e-01 1.06566978e+00 -1.77702904e-01 -3.10305744e-01 -6.15928292e-01 6.44744694e-01 1.25732496e-01 5.77078164e-01 6.90739036e-01 1.61563829e-01 9.40244317e-01 3.04690272e-01 -5.84827185e-01 -1.12794685e+00 -6.27538502e-01 -3.35069805e-01 6.56973541e-01 4.09884341e-02 3.59496802e-01 -5.76682270e-01 3.23753297e-01 -1.51089311e-01 7.71713555e-01 -3.04486752e-01 -2.30553284e-01 -1.22367537e+00 -1.11355329e+00 7.51427770e-01 4.47488964e-01 6.31852210e-01 -1.18360710e+00 -6.16131961e-01 5.26752651e-01 -2.65204668e-01 -9.36842978e-01 1.98991045e-01 2.89590389e-01 -7.62843013e-01 -8.14197421e-01 -1.23155308e+00 -9.16460097e-01 3.51058900e-01 -2.12687477e-01 8.92086565e-01 3.71658653e-01 5.08025140e-02 1.61823377e-01 -1.85322315e-01 -2.97856003e-01 -6.04365706e-01 -1.24353714e-01 -1.40171736e-01 -1.60840899e-02 -5.66902518e-01 -1.04390037e+00 -6.82280362e-01 1.48348793e-01 -1.10459769e+00 9.83576924e-02 1.67198330e-01 9.44523156e-01 2.14271083e-01 2.62029737e-01 6.86924994e-01 -5.40874839e-01 6.84278011e-01 -2.85998493e-01 -6.79879785e-01 -2.24974364e-01 -1.76526442e-01 4.18545455e-02 6.89783752e-01 -6.31676435e-01 -1.22871721e+00 -4.72549349e-01 -1.00529301e+00 -3.68391514e-01 1.30144641e-01 1.07216442e+00 2.45077580e-01 -6.68844879e-01 7.68542588e-01 1.09216265e-01 5.59841394e-02 -5.17601192e-01 -5.12428522e-01 1.26745537e-01 3.32998395e-01 -4.92984742e-01 1.05806804e+00 4.99709666e-01 5.80271959e-01 -1.26789963e+00 -5.87091923e-01 -1.76235020e-01 -1.60890415e-01 -3.19578886e-01 6.84043586e-01 -5.13219297e-01 -6.31892145e-01 9.41164374e-01 -1.42730367e+00 -8.38793755e-01 -3.52172643e-01 8.81429791e-01 -3.86385620e-01 2.61983216e-01 -8.00373077e-01 -7.40437269e-01 -3.54733407e-01 -1.11097729e+00 5.31228721e-01 2.62796849e-01 4.79918607e-02 -1.31059194e+00 1.24415018e-01 -1.85752049e-01 9.69041169e-01 5.82197785e-01 1.11343622e+00 -8.14339295e-02 -3.10901195e-01 2.13343594e-02 -2.76108950e-01 3.36462170e-01 -3.79174292e-01 -6.24306442e-04 -9.94920790e-01 -1.94258347e-01 6.00598216e-01 -1.01641610e-01 9.81838584e-01 9.58871961e-01 8.67877305e-01 -1.42431095e-01 -1.58856869e-01 1.05806530e+00 1.85746264e+00 2.47510433e-01 7.63341784e-01 2.22421274e-01 7.45920658e-01 6.34207964e-01 -5.36127090e-02 4.44967896e-01 -3.05210859e-01 4.08856452e-01 3.96933883e-01 -6.53011441e-01 -2.83783793e-01 4.60998744e-01 5.40422602e-03 8.61790419e-01 -5.87102592e-01 -2.53658026e-01 -1.08137846e+00 5.03356338e-01 -1.41687131e+00 -9.07091498e-01 -6.91032648e-01 1.80841553e+00 8.32217038e-01 1.20058395e-02 -4.29763377e-01 4.89930212e-01 5.09944916e-01 -2.32504285e-03 -3.45942110e-01 -3.33975881e-01 -2.37554044e-01 6.37616456e-01 5.22402227e-01 9.76343453e-01 -7.93733716e-01 4.03192252e-01 5.77328730e+00 7.98312008e-01 -1.79176891e+00 2.97923028e-01 2.67569780e-01 3.23914587e-01 -3.62970859e-01 -2.42096305e-01 -2.46803895e-01 3.19025218e-01 8.70111883e-01 7.76387006e-02 1.68766707e-01 2.28365391e-01 5.38027644e-01 -1.14644617e-01 -5.44532120e-01 6.85207009e-01 -3.42098117e-01 -1.56804478e+00 -2.36585885e-01 -2.53122717e-01 6.78924263e-01 7.80868679e-02 -1.93433717e-01 7.85532445e-02 -2.79529482e-01 -1.02947688e+00 5.29691875e-01 7.51498222e-01 8.27879429e-01 -7.01611102e-01 9.56248045e-01 5.57856083e-01 -8.60778093e-01 2.53332973e-01 -3.08102638e-01 -2.88351476e-01 7.40087032e-01 8.92426550e-01 -3.55386436e-01 4.52957243e-01 6.72546983e-01 3.18490416e-01 1.72745317e-01 1.47870708e+00 2.09666491e-01 1.02121997e+00 -5.46032965e-01 2.10797176e-01 5.31833947e-01 -3.28786701e-01 9.03037488e-01 1.14887750e+00 7.05171108e-01 4.61861879e-01 -1.88550293e-01 9.39243793e-01 3.11905086e-01 -1.44689053e-01 -2.39388153e-01 4.61620301e-01 -2.27305233e-01 7.76669323e-01 -6.83257222e-01 3.27018760e-02 -3.70699525e-01 3.06736976e-01 -1.88729689e-01 9.62401628e-01 -8.93738627e-01 -3.78652275e-01 5.15030801e-01 6.88098490e-01 3.19282830e-01 -2.75599629e-01 -2.36866638e-01 -6.72694683e-01 -2.34433830e-01 -3.65971267e-01 5.13960645e-02 -9.72577691e-01 -1.24811232e+00 6.11569583e-01 1.58930689e-01 -1.07267344e+00 -4.89505157e-02 -7.41892576e-01 -8.66665423e-01 1.09940159e+00 -2.02013874e+00 -6.93530202e-01 -1.80384949e-01 4.77938443e-01 3.52333114e-02 1.92548722e-01 5.80794573e-01 7.64912367e-01 -1.98263064e-01 2.33011484e-01 3.94297689e-01 2.82274634e-01 4.38291013e-01 -7.29034543e-01 -8.53789449e-02 7.26365566e-01 -7.91218817e-01 1.72600195e-01 1.22034967e+00 -6.31918252e-01 -1.09122241e+00 -1.01991141e+00 6.43034160e-01 5.71963072e-01 6.01858616e-01 8.08150619e-02 -1.71000457e+00 2.09826857e-01 2.03043818e-01 1.74220875e-01 3.11509103e-01 -4.58924890e-01 2.72743285e-01 -1.21615887e-01 -1.16957998e+00 4.04898882e-01 5.56212783e-01 5.59278913e-02 -3.07728976e-01 2.44770437e-01 5.42161286e-01 -5.92042327e-01 -6.97280645e-01 8.29188824e-01 3.52833420e-01 -1.02364397e+00 1.19185829e+00 -1.89534590e-01 7.70581424e-01 -1.50487810e-01 2.10799351e-01 -1.59683442e+00 -4.38408107e-01 -7.19226360e-01 1.17963865e-01 8.08832765e-01 3.54045659e-01 -1.03077948e+00 6.58572912e-01 3.84395838e-01 -6.29715979e-01 -9.01778877e-01 -1.20518208e+00 -7.12067366e-01 5.02598822e-01 -5.53007782e-01 2.81683147e-01 1.04814792e+00 -4.95474190e-01 -1.30110428e-01 -2.91359752e-01 5.60249686e-01 6.12613142e-01 -1.85991839e-01 4.05181982e-02 -1.43284285e+00 -2.74142087e-01 -4.35567260e-01 5.77644892e-02 -8.73015106e-01 6.39169067e-02 -5.63333392e-01 3.96422058e-01 -1.60436332e+00 -4.56692934e-01 -8.17476511e-01 1.58006642e-02 1.35628149e-01 -1.89599451e-02 3.83422911e-01 -1.75258830e-01 1.47782281e-01 4.29335058e-01 5.62678814e-01 1.68902469e+00 -2.03259904e-02 -2.41308227e-01 1.44790068e-01 -4.95051406e-02 1.04049575e+00 6.82446659e-01 -7.15011418e-01 -2.10836709e-01 -7.20884800e-01 4.77453500e-01 3.71122122e-01 5.88585734e-01 -1.34617400e+00 4.09060448e-01 -5.30940816e-02 2.54141331e-01 -3.64136755e-01 3.96180987e-01 -8.59867275e-01 2.47646973e-01 5.81190467e-01 2.13325815e-03 -2.20693037e-01 5.66595614e-01 -5.21790944e-02 -3.54231060e-01 -6.65180266e-01 1.35055912e+00 -6.47791401e-02 -4.02475864e-01 6.65466264e-02 -9.16797400e-01 2.43988186e-01 3.34966242e-01 -2.74008721e-01 1.05420180e-01 -4.66009676e-01 -5.32663047e-01 -1.05287544e-01 3.24100144e-02 -5.01563489e-01 7.36258984e-01 -1.06163216e+00 -1.05687809e+00 3.83221388e-01 -5.77107549e-01 3.21982950e-01 7.62233019e-01 1.12096345e+00 -1.35427177e+00 -2.05132529e-01 -2.08579563e-02 -5.81014216e-01 -6.36977136e-01 9.58327949e-03 1.07755017e+00 -4.91224974e-01 -6.87622607e-01 1.14748061e+00 2.70374030e-01 -3.65036994e-01 -1.80085972e-02 -4.14698064e-01 -3.16758573e-01 -7.55819008e-02 4.92869437e-01 5.82038343e-01 2.34907344e-01 -6.19141698e-01 -2.14351982e-01 9.84716237e-01 5.71181238e-01 -2.37068966e-01 1.70203865e+00 2.26530194e-01 -1.31372616e-01 1.47005916e-01 1.21417773e+00 -2.49356255e-02 -1.39302826e+00 -1.91348642e-02 -5.47267199e-01 6.15842314e-03 6.08168304e-01 -5.86901844e-01 -1.51369321e+00 9.54131246e-01 4.02288705e-01 2.54546463e-01 1.15650344e+00 -2.97959030e-01 1.16493452e+00 2.11789563e-01 -9.52835232e-02 -7.19320297e-01 -9.32840556e-02 9.49055791e-01 1.11695540e+00 -8.63715231e-01 3.12595107e-02 -5.23432732e-01 -8.00442882e-03 1.27846038e+00 3.53667766e-01 -4.23950374e-01 1.20832217e+00 8.83891165e-01 2.51696259e-01 -2.03874633e-01 -1.31451637e-01 1.48945242e-01 2.48940229e-01 5.58866382e-01 4.73885030e-01 -4.50538933e-01 -3.05508405e-01 4.22176093e-01 2.30345782e-02 1.60637692e-01 6.30554616e-01 8.56860697e-01 -3.85381281e-01 -6.47918582e-01 -7.05216646e-01 2.20804587e-01 -4.27990258e-01 -2.98800290e-01 4.11593586e-01 7.93729842e-01 2.96543717e-01 6.38640523e-01 8.49849060e-02 2.17845485e-01 3.76035333e-01 7.09460909e-03 3.89775544e-01 -1.82954654e-01 -6.31342053e-01 3.00601125e-01 1.56829704e-03 -1.04224302e-01 -5.89048684e-01 -3.94973963e-01 -1.70140982e+00 -3.06793004e-01 -3.30369622e-01 3.52799654e-01 2.82307327e-01 1.04014122e+00 2.12850329e-02 8.05913389e-01 2.21482292e-01 -1.46684301e+00 -1.08528979e-01 -9.49754357e-01 -7.87938058e-01 1.27862871e-01 7.42587924e-01 -9.37337875e-01 -7.87728846e-01 2.07956191e-02]
[6.795405387878418, 2.5864317417144775]
4442ae7b-94ca-44e5-a3bf-9949a55ab34f
neural-collaborative-graph-machines-for-table
2111.13359
null
https://arxiv.org/abs/2111.13359v2
https://arxiv.org/pdf/2111.13359v2.pdf
Neural Collaborative Graph Machines for Table Structure Recognition
Recently, table structure recognition has achieved impressive progress with the help of deep graph models. Most of them exploit single visual cues of tabular elements or simply combine visual cues with other modalities via early fusion to reason their graph relationships. However, neither early fusion nor individually reasoning in terms of multiple modalities can be appropriate for all varieties of table structures with great diversity. Instead, different modalities are expected to collaborate with each other in different patterns for different table cases. In the community, the importance of intra-inter modality interactions for table structure reasoning is still unexplored. In this paper, we define it as heterogeneous table structure recognition (Hetero-TSR) problem. With the aim of filling this gap, we present a novel Neural Collaborative Graph Machines (NCGM) equipped with stacked collaborative blocks, which alternatively extracts intra-modality context and models inter-modality interactions in a hierarchical way. It can represent the intra-inter modality relationships of tabular elements more robustly, which significantly improves the recognition performance. We also show that the proposed NCGM can modulate collaborative pattern of different modalities conditioned on the context of intra-modality cues, which is vital for diversified table cases. Experimental results on benchmarks demonstrate our proposed NCGM achieves state-of-the-art performance and beats other contemporary methods by a large margin especially under challenging scenarios.
['Bo Ren', 'Yinsong Liu', 'Deqiang Jiang', 'Bing Liu', 'Xin Li', 'Hao liu']
2021-11-26
null
http://openaccess.thecvf.com//content/CVPR2022/html/Liu_Neural_Collaborative_Graph_Machines_for_Table_Structure_Recognition_CVPR_2022_paper.html
http://openaccess.thecvf.com//content/CVPR2022/papers/Liu_Neural_Collaborative_Graph_Machines_for_Table_Structure_Recognition_CVPR_2022_paper.pdf
cvpr-2022-1
['table-recognition']
['computer-vision']
[ 2.94678450e-01 -2.24696714e-02 -3.66822958e-01 -1.43959031e-01 -5.97251713e-01 -7.14262307e-01 5.52427113e-01 3.35567981e-01 2.13271290e-01 3.53662372e-01 4.13234591e-01 -1.06106274e-01 -2.73891062e-01 -8.31504107e-01 -7.44803965e-01 -5.15950501e-01 2.60668784e-01 6.46546602e-01 2.10077643e-01 -4.91113216e-01 8.73847231e-02 2.71240413e-01 -1.41853750e+00 1.07625985e+00 7.60609627e-01 1.11688924e+00 -6.77998737e-02 3.83829743e-01 -5.71308911e-01 1.45724380e+00 -3.80478591e-01 -7.87700415e-01 2.18182743e-01 -4.75574434e-01 -4.89209354e-01 3.73413771e-01 7.45315254e-01 1.19131826e-01 -8.37247133e-01 1.00092995e+00 1.66798174e-01 4.69072312e-02 5.30064404e-01 -1.29032123e+00 -1.00577438e+00 1.22440386e+00 -8.81024361e-01 4.43791002e-02 6.34069681e-01 -4.79275063e-02 1.38156211e+00 -9.19668257e-01 7.65831411e-01 1.41593993e+00 3.64511371e-01 1.80900499e-01 -1.13745141e+00 -5.98836839e-01 6.75652742e-01 5.97238779e-01 -1.32879162e+00 -2.26320192e-01 1.20238972e+00 -1.86398193e-01 8.14053357e-01 3.92196953e-01 3.96176845e-01 1.14970064e+00 1.49918854e-01 1.14727831e+00 1.06655550e+00 -2.22304806e-01 -1.83755755e-01 -3.79031040e-02 1.18689775e-01 8.86484444e-01 4.09951329e-01 -6.51950538e-01 -1.11292744e+00 1.85797811e-01 5.35077751e-01 3.34950298e-01 -3.29589784e-01 -7.65667737e-01 -1.60887671e+00 4.47895914e-01 9.07906294e-01 4.77751464e-01 -1.18271977e-01 -2.22338006e-01 2.50615090e-01 3.55644315e-01 -2.06563026e-01 8.90380219e-02 -1.41375542e-01 3.08065355e-01 -5.23144126e-01 -2.65253544e-01 8.65415990e-01 1.10422194e+00 6.13770545e-01 2.65009999e-02 -4.71946478e-01 8.01326215e-01 4.25750852e-01 5.35614252e-01 -3.61584350e-02 -4.61351573e-01 1.21474373e+00 1.30523741e+00 -4.08249080e-01 -1.46676004e+00 -4.27070826e-01 -5.19504249e-01 -1.47852695e+00 -1.84225261e-01 5.20077348e-01 3.42717588e-01 -9.92665172e-01 1.42793751e+00 2.17602745e-01 -1.94058567e-01 1.85182869e-01 7.23454416e-01 1.19324708e+00 6.24088585e-01 -1.71490058e-01 -8.79838914e-02 1.53762996e+00 -9.85649168e-01 -9.96183336e-01 -3.13886315e-01 3.18576485e-01 -7.38050580e-01 8.12114000e-01 4.69782948e-01 -9.38656151e-01 -7.04447985e-01 -1.14177239e+00 -2.19099239e-01 -6.85758233e-01 9.80880558e-02 6.62468851e-01 5.74750841e-01 -8.69599640e-01 1.08257368e-01 -4.60261047e-01 -2.56738663e-01 5.21327198e-01 2.69208133e-01 -6.07457221e-01 -5.11740267e-01 -1.10000741e+00 4.57399279e-01 3.89933228e-01 5.31577826e-01 -3.15814972e-01 -4.08955365e-01 -1.01118863e+00 2.95428902e-01 1.26101553e+00 -7.09546387e-01 4.77959007e-01 -4.95122552e-01 -1.09094477e+00 6.61472976e-01 3.07787918e-02 -7.36306757e-02 4.22049522e-01 1.52432784e-01 -6.94097579e-01 -2.87399013e-02 -1.67934895e-01 3.43264014e-01 7.40258276e-01 -1.62442756e+00 -3.92094135e-01 -7.07211971e-01 1.83312863e-01 4.62378174e-01 -2.74654269e-01 -3.11475992e-01 -1.24580204e+00 -6.13360882e-01 4.31931853e-01 -8.03920984e-01 5.17724156e-02 -1.24179997e-01 -6.32260263e-01 -1.16810307e-01 6.60241783e-01 -4.41616863e-01 1.40720642e+00 -2.03539300e+00 4.96806979e-01 4.10574526e-01 6.78981960e-01 -4.60738242e-02 -5.16377250e-03 7.53529012e-01 1.51752114e-01 -8.84665251e-02 -1.75115049e-01 -1.47770196e-01 2.75096953e-01 2.72538036e-01 -2.89647698e-01 1.41604796e-01 1.19263612e-01 1.06631410e+00 -6.74167275e-01 -5.96077323e-01 1.50587931e-01 5.21468937e-01 -3.53313029e-01 -1.71698794e-01 -2.28329286e-01 1.93482354e-01 -3.22712243e-01 1.04171515e+00 7.28732347e-01 -8.18550944e-01 8.86794388e-01 -7.31594563e-01 4.53853428e-01 -2.53159702e-01 -1.64122021e+00 1.72329783e+00 -1.53359044e-02 3.42927217e-01 -9.76503715e-02 -8.90846014e-01 1.01506865e+00 -1.68062910e-01 2.62760520e-01 -8.75773370e-01 1.07095562e-01 7.84257054e-02 2.62483835e-01 -8.89647454e-02 4.95699644e-01 3.62534076e-01 -2.87414849e-01 -5.52149676e-03 1.45906238e-02 2.86937237e-01 4.68341142e-01 6.95379734e-01 1.05307889e+00 -3.89544107e-02 2.87209868e-01 2.30510727e-01 8.36185813e-01 -2.23261997e-01 3.98367584e-01 9.39466834e-01 7.57162040e-03 5.04483998e-01 7.00153232e-01 -2.17117965e-01 -4.22102541e-01 -1.06163025e+00 3.31680596e-01 1.07551837e+00 6.94913328e-01 -7.80915737e-01 -8.66556913e-02 -8.47529590e-01 2.35769272e-01 1.74764350e-01 -9.18115079e-01 3.52926296e-03 -6.03568196e-01 -5.93997598e-01 2.84261197e-01 8.41486216e-01 6.42876804e-01 -9.73997056e-01 1.95044830e-01 6.90216124e-02 -1.62667140e-01 -1.47998571e+00 -4.03487295e-01 2.03789592e-01 -5.14190972e-01 -1.21642375e+00 -4.85100120e-01 -6.92732692e-01 5.84983289e-01 5.32436132e-01 1.26484978e+00 2.79668689e-01 1.71239883e-01 5.79755485e-01 -3.48741621e-01 -7.47518148e-04 -2.38645256e-01 2.35881843e-02 -4.18041527e-01 6.06155574e-01 1.83806852e-01 -5.17421722e-01 -3.86768073e-01 5.63633978e-01 -1.02826726e+00 3.28374356e-01 8.75359416e-01 7.16611266e-01 6.68385386e-01 1.48591995e-01 6.82724565e-02 -1.27231264e+00 4.11314875e-01 -4.89876956e-01 -3.75440150e-01 1.00108171e+00 -4.49220747e-01 3.06014329e-01 7.64997602e-01 -2.39748672e-01 -1.07484472e+00 -1.55425966e-01 5.24084091e-01 -7.16619253e-01 1.02513850e-01 9.22995150e-01 -7.09018230e-01 5.83185703e-02 2.62018889e-01 3.11787516e-01 -4.56506521e-01 -4.50135529e-01 4.86349165e-01 1.92537472e-01 6.86354876e-01 -7.65196621e-01 9.93162930e-01 5.12464345e-01 4.76240933e-01 -3.03029925e-01 -7.48323381e-01 -3.69096577e-01 -6.60476029e-01 -4.92996603e-01 7.63248384e-01 -9.90869880e-01 -8.40716660e-01 4.94755656e-01 -6.71944976e-01 -1.41290883e-02 2.86512375e-01 4.63128388e-02 3.79690826e-02 6.70763314e-01 -5.44699788e-01 -6.04151607e-01 2.54724734e-02 -1.20141947e+00 9.63145733e-01 4.01417971e-01 3.41955513e-01 -9.14735854e-01 -3.62650096e-01 8.77954602e-01 1.97662249e-01 2.15357721e-01 9.69258189e-01 -7.12016940e-01 -1.12464678e+00 -1.94790304e-01 -5.22530079e-01 -3.53442222e-01 1.95255354e-01 1.49392143e-01 -6.96655929e-01 -7.00155795e-02 -5.49109280e-01 -2.77386844e-01 1.13660574e+00 -1.51192144e-01 1.11636961e+00 1.14743978e-01 -2.95355499e-01 4.97756779e-01 1.36457717e+00 3.31802726e-01 6.97849095e-01 1.82846472e-01 1.50217962e+00 5.24765253e-01 3.68272424e-01 2.76876032e-01 8.43908966e-01 6.12615764e-01 4.80097294e-01 1.25332743e-01 -2.57886499e-01 -3.77241731e-01 2.52133697e-01 1.02082455e+00 -1.27826393e-01 -8.28256011e-01 -1.01618814e+00 1.66142583e-01 -2.20455289e+00 -9.28846002e-01 -3.31641257e-01 1.85708094e+00 3.62901062e-01 4.60395873e-01 -6.52232915e-02 2.54224569e-01 7.93404162e-01 4.53726053e-01 -5.88249087e-01 -1.93810519e-02 -6.47680640e-01 -2.14486852e-01 2.49789461e-01 1.40007362e-01 -9.17465150e-01 7.06832945e-01 4.70028591e+00 1.00615871e+00 -6.75221324e-01 -4.05866921e-01 6.67865396e-01 3.13033462e-01 -4.83721435e-01 -4.83811200e-02 -7.74459541e-01 1.68509915e-01 2.81777084e-01 7.32395649e-02 6.12714887e-01 4.34838116e-01 -7.38861740e-01 -9.73681454e-03 -1.10299897e+00 1.38005567e+00 4.48136657e-01 -1.49203765e+00 5.42469919e-01 -2.46912725e-02 7.78910220e-01 -4.67637718e-01 1.47359177e-01 5.25499344e-01 1.59378111e-01 -8.69815648e-01 5.70494711e-01 6.28191411e-01 5.65405905e-01 -6.46848440e-01 6.32291317e-01 1.00909988e-03 -1.82700217e+00 -3.69268842e-02 -2.54166424e-02 1.97890714e-01 -2.62429535e-01 2.63848066e-01 -3.60503107e-01 1.34414387e+00 5.86549401e-01 1.02168691e+00 -1.06493032e+00 7.14296222e-01 -1.40871942e-01 2.36895472e-01 -2.38805935e-01 -1.78424921e-02 1.59325853e-01 -2.19026104e-01 4.05668855e-01 8.82594407e-01 9.31145698e-02 1.10331355e-02 3.81632954e-01 6.03807271e-01 -4.80274796e-01 1.43006340e-01 -5.65056741e-01 -2.30043858e-01 3.44179273e-01 1.30017138e+00 -1.12047815e+00 -4.61821377e-01 -6.27029896e-01 7.89683521e-01 4.90478277e-01 4.11804438e-01 -7.69421875e-01 -1.35709923e-02 7.12365881e-02 -2.68132716e-01 6.34886980e-01 -1.22178867e-01 -5.56396186e-01 -1.63014030e+00 4.13249850e-01 -1.18232548e+00 1.15280163e+00 -9.99093831e-01 -1.44223344e+00 6.52745247e-01 -5.26977144e-02 -1.35478663e+00 1.93131864e-01 -8.06594610e-01 -2.07336068e-01 4.64295357e-01 -1.25863004e+00 -1.55108094e+00 -5.38998723e-01 9.88215327e-01 4.37087834e-01 -2.96628624e-01 3.01176369e-01 1.80265740e-01 -8.55312109e-01 8.08257997e-01 -4.65012453e-02 3.54337960e-01 8.25844407e-01 -1.49195075e+00 9.85629261e-02 9.94423747e-01 7.09155500e-01 5.42098343e-01 3.36828113e-01 -8.27969313e-01 -2.20932102e+00 -7.34140456e-01 6.70325458e-01 -6.14727676e-01 9.08473611e-01 -6.00541949e-01 -1.19824791e+00 6.24774158e-01 4.67563868e-01 7.02726990e-02 6.49989069e-01 4.25259531e-01 -9.72109795e-01 -4.36434001e-01 -7.25243092e-01 8.36564064e-01 1.38432205e+00 -6.92655087e-01 -3.29330206e-01 1.46362379e-01 3.30623269e-01 -6.02725029e-01 -9.93933082e-01 6.46992207e-01 4.81467158e-01 -1.12554049e+00 9.30375755e-01 -3.29272002e-01 4.84691381e-01 -6.24325573e-01 -5.10231614e-01 -8.85835469e-01 -3.93217564e-01 -5.40643156e-01 -6.35099053e-01 1.36264169e+00 4.47010964e-01 -4.08914238e-01 7.25585222e-01 4.50501978e-01 -1.95610803e-03 -6.92408562e-01 -6.04303300e-01 -4.13752079e-01 -4.85328257e-01 -3.58141601e-01 6.51427507e-01 1.16622007e+00 8.41277763e-02 6.39553428e-01 -5.93811750e-01 3.76400322e-01 5.31159580e-01 6.08615816e-01 9.10320878e-01 -9.93963242e-01 -4.89774197e-01 -6.40973926e-01 -5.01879871e-01 -9.82659519e-01 -1.13769136e-01 -1.03439653e+00 -3.64377886e-01 -1.79888999e+00 4.40811574e-01 1.61188822e-02 -7.47832954e-01 6.38642192e-01 -5.45845687e-01 3.08762640e-01 6.72427177e-01 1.94983259e-01 -1.13569105e+00 4.20924693e-01 1.49321687e+00 -7.78381288e-01 -4.24868651e-02 -4.99469548e-01 -9.35274303e-01 2.67646760e-01 3.76389682e-01 -8.89889598e-02 -5.77067554e-01 -3.17039698e-01 6.59200132e-01 4.51713204e-01 5.00080734e-02 -1.09241664e+00 5.85433900e-01 -1.15953654e-01 7.11961925e-01 -9.10288334e-01 2.53852308e-01 -9.97975528e-01 4.39413637e-01 -9.52443667e-03 -2.26529524e-01 1.46755189e-01 2.28425637e-01 1.06687939e+00 -4.37064767e-01 5.32726884e-01 1.11038864e-01 2.39508990e-02 -9.46156919e-01 2.66330630e-01 -1.66641116e-01 -7.36300498e-02 6.13144398e-01 -3.58361095e-01 -6.11662269e-01 -3.22333932e-01 -7.97044933e-01 4.30482775e-01 3.45054835e-01 6.62746489e-01 5.83537400e-01 -1.57544267e+00 -5.98421574e-01 1.94226339e-01 6.90620482e-01 -1.98082358e-01 7.48653233e-01 9.73587334e-01 -1.75945461e-01 2.56829083e-01 -1.67851388e-01 -7.18083024e-01 -1.27133441e+00 9.97777939e-01 1.12721659e-01 -8.09274077e-01 -5.02926230e-01 6.74786985e-01 4.33656394e-01 -1.94596693e-01 4.01195049e-01 -3.61727208e-01 -5.18620551e-01 3.34442586e-01 3.60719979e-01 -4.60607223e-02 1.77151084e-01 -7.81658411e-01 -5.19937575e-01 6.73367143e-01 -3.10595840e-01 4.56487238e-01 9.70979393e-01 -1.23191118e-01 5.40344380e-02 6.69139266e-01 6.68263972e-01 3.27504545e-01 -8.97727728e-01 -6.52648985e-01 -1.10277653e-01 -4.96632427e-01 -2.25663677e-01 -1.14593685e+00 -1.43611324e+00 5.76582253e-01 1.35292247e-01 3.22846383e-01 1.32425582e+00 -3.49608175e-02 4.97088701e-01 5.94965816e-01 3.36171806e-01 -7.65593290e-01 2.43539989e-01 3.85787785e-01 9.56908047e-01 -1.43725336e+00 7.80554190e-02 -8.22340846e-01 -9.50367212e-01 1.35288918e+00 6.83679163e-01 2.31303751e-01 5.52113593e-01 1.43861905e-01 5.74349388e-02 -5.08657277e-01 -9.06629741e-01 -4.69299525e-01 9.81842935e-01 4.13668990e-01 3.33626151e-01 1.00691892e-01 1.29570842e-01 4.92548108e-01 3.78626823e-01 -4.08038825e-01 2.72988111e-01 9.52752054e-01 -1.69475209e-02 -1.17720294e+00 -5.85572064e-01 4.26632434e-01 -3.81159410e-02 -7.00046644e-02 -9.22466695e-01 9.19746816e-01 -4.47636023e-02 1.11834955e+00 -2.39660427e-01 -6.91100895e-01 3.53304625e-01 -2.64417729e-03 6.79830253e-01 -2.18930304e-01 -7.40552008e-01 3.00777644e-01 1.98686182e-01 -4.72164214e-01 -5.27025342e-01 -4.98716086e-01 -9.94346261e-01 -3.66219431e-01 3.68065462e-02 -1.66465715e-01 3.45322862e-02 8.52305830e-01 4.40181166e-01 7.89860308e-01 4.14554954e-01 -4.49386328e-01 -6.62895590e-02 -6.36980355e-01 -5.52910268e-01 6.34757638e-01 2.42533833e-02 -8.10919523e-01 -1.91427827e-01 -1.23190880e-01]
[11.098712921142578, 2.4537389278411865]
b760117d-13eb-46ec-9038-b415d40f2b37
decoupling-the-skeleton-parsing-and-schema
2302.05965
null
https://arxiv.org/abs/2302.05965v3
https://arxiv.org/pdf/2302.05965v3.pdf
RESDSQL: Decoupling Schema Linking and Skeleton Parsing for Text-to-SQL
One of the recent best attempts at Text-to-SQL is the pre-trained language model. Due to the structural property of the SQL queries, the seq2seq model takes the responsibility of parsing both the schema items (i.e., tables and columns) and the skeleton (i.e., SQL keywords). Such coupled targets increase the difficulty of parsing the correct SQL queries especially when they involve many schema items and logic operators. This paper proposes a ranking-enhanced encoding and skeleton-aware decoding framework to decouple the schema linking and the skeleton parsing. Specifically, for a seq2seq encoder-decode model, its encoder is injected by the most relevant schema items instead of the whole unordered ones, which could alleviate the schema linking effort during SQL parsing, and its decoder first generates the skeleton and then the actual SQL query, which could implicitly constrain the SQL parsing. We evaluate our proposed framework on Spider and its three robustness variants: Spider-DK, Spider-Syn, and Spider-Realistic. The experimental results show that our framework delivers promising performance and robustness. Our code is available at https://github.com/RUCKBReasoning/RESDSQL.
['Hong Chen', 'Cuiping Li', 'Jing Zhang', 'Haoyang Li']
2023-02-12
null
null
null
null
['text-to-sql', 'semantic-parsing']
['computer-code', 'natural-language-processing']
[ 5.30557968e-02 1.50475368e-01 -1.45721167e-01 -5.50293207e-01 -1.00533116e+00 -8.39774489e-01 1.58795491e-01 3.08060080e-01 -3.31804492e-02 3.03984851e-01 4.19477403e-01 -5.54720938e-01 2.34153152e-01 -1.10630357e+00 -1.08289623e+00 -1.33550838e-01 1.74500883e-01 3.40814024e-01 4.79239792e-01 -2.93646634e-01 3.43962014e-01 5.44805303e-02 -1.42238355e+00 6.14870191e-01 9.62997019e-01 1.06807625e+00 2.53026366e-01 5.82889795e-01 -7.85934031e-01 9.69197750e-01 -5.07448852e-01 -9.19624448e-01 1.60942897e-01 -1.43500656e-01 -5.77225745e-01 -3.21457922e-01 2.80541003e-01 -7.00367928e-01 -3.13435793e-01 1.05548096e+00 4.42804784e-01 -5.38837850e-01 -1.20609701e-01 -9.58516181e-01 -3.32056463e-01 1.10088241e+00 -4.21214670e-01 -2.11202860e-01 3.72057736e-01 9.62179899e-02 1.32415760e+00 -7.79838860e-01 5.09735823e-01 1.20315230e+00 2.87701726e-01 2.66065389e-01 -8.93651009e-01 -4.77051795e-01 1.19795755e-01 -8.28716680e-02 -1.26827466e+00 -6.37652934e-01 5.99043906e-01 -1.59559816e-01 1.01818085e+00 4.84239429e-01 8.61192495e-02 9.76747215e-01 2.40120828e-01 9.01470006e-01 5.74521661e-01 6.76066801e-02 7.11947083e-02 7.28356689e-02 3.43059093e-01 7.71721363e-01 2.91097611e-01 -6.28540590e-02 -5.33850670e-01 -2.65406579e-01 5.44798493e-01 -4.46637496e-02 1.11188866e-01 -3.71601343e-01 -1.10612452e+00 5.92114687e-01 1.21991910e-01 -2.36348733e-01 -1.70989364e-01 1.27347872e-01 9.25812960e-01 1.02221265e-01 -2.21748143e-01 4.55935746e-02 -7.40387797e-01 -3.63113165e-01 -7.53697932e-01 1.83442265e-01 1.13039505e+00 1.64451945e+00 7.39016235e-01 -1.90670237e-01 -1.63913220e-01 7.93965101e-01 3.46360058e-01 5.70783734e-01 3.15585472e-02 -6.41852319e-01 1.12108362e+00 9.84090984e-01 -8.97139981e-02 -7.92526960e-01 -5.90904281e-02 -3.96892935e-01 -5.91019392e-01 -4.94933486e-01 5.24472222e-02 -3.16277072e-02 -6.25033438e-01 1.61258960e+00 2.58466125e-01 -4.24949259e-01 1.76134020e-01 5.89512050e-01 8.24653149e-01 7.48523533e-01 -1.40789106e-01 -9.69110206e-02 1.49863219e+00 -8.64523590e-01 -5.94935656e-01 -3.64185214e-01 6.57730997e-01 -9.39088762e-01 1.27228403e+00 2.27358967e-01 -1.30423951e+00 -5.75627863e-01 -9.10294890e-01 -3.71166408e-01 -1.86195165e-01 3.04058254e-01 4.51155216e-01 3.59710246e-01 -4.85072166e-01 1.61378562e-01 -8.17342937e-01 -2.81165272e-01 1.24720678e-01 1.08325966e-01 -1.57347247e-01 7.67868385e-02 -1.13601828e+00 2.70825118e-01 8.21012974e-01 1.76101312e-01 -7.21611023e-01 -4.95524794e-01 -7.15051770e-01 2.55936235e-01 9.95742202e-01 -4.55951065e-01 1.28603971e+00 -3.25840294e-01 -1.35315490e+00 5.05513430e-01 -2.59692609e-01 -1.66324750e-01 4.95542228e-01 -4.77232009e-01 -3.46672982e-01 -1.65418342e-01 2.63075769e-01 3.03717762e-01 4.36676830e-01 -9.92261529e-01 -6.24336123e-01 -4.30239379e-01 1.67663798e-01 3.56027484e-02 -1.05607823e-01 1.35180935e-01 -1.17971897e+00 -5.96760213e-01 1.27176523e-01 -6.12192929e-01 2.16429666e-01 -2.73970664e-01 -1.04983747e+00 1.87783953e-05 3.42703283e-01 -8.51942956e-01 1.83908606e+00 -2.48761415e+00 7.46978074e-02 2.42847875e-01 -2.84267031e-02 5.73807731e-02 -5.69699854e-02 9.76286054e-01 -8.75990167e-02 3.71730924e-01 -2.52921432e-01 3.02594036e-01 1.58152938e-01 4.10499275e-01 -7.26647973e-01 -2.25033030e-01 3.80187541e-01 9.74970222e-01 -5.41472614e-01 -6.52571797e-01 -2.42562070e-01 -8.07700828e-02 -8.77583086e-01 6.49185061e-01 -7.13790417e-01 -1.02979392e-01 -6.47481084e-01 8.58794570e-01 8.83161008e-01 -1.46802381e-01 4.78889644e-01 -5.26928008e-01 -2.84103692e-01 8.99772167e-01 -1.44904864e+00 1.71998775e+00 -4.41076070e-01 -2.26789773e-01 -1.69103257e-02 -5.92823744e-01 1.03749621e+00 -1.25131514e-02 1.19411692e-01 -9.33457315e-01 -3.46668601e-01 3.87066066e-01 -2.70893335e-01 -7.42979825e-01 3.38366449e-01 4.20128793e-01 -5.82325280e-01 2.89033115e-01 -2.03449145e-01 5.80217727e-02 3.54705632e-01 4.94905055e-01 1.09339750e+00 4.12037939e-01 3.25736433e-01 -1.30930087e-02 8.58340263e-01 -9.11740400e-03 7.66134024e-01 6.24813616e-01 4.53485578e-01 3.26328069e-01 1.28758359e+00 -1.89305261e-01 -1.18254745e+00 -1.44239712e+00 5.38847595e-02 1.14547980e+00 9.18338522e-02 -9.60920513e-01 -6.66200757e-01 -9.04521525e-01 -6.23255316e-03 9.94706750e-01 -1.57376140e-01 -1.54154420e-01 -9.84072685e-01 -2.92764693e-01 8.63404751e-01 5.56335628e-01 5.29169798e-01 -7.77899384e-01 -3.90035003e-01 3.20706069e-01 -3.00857037e-01 -9.91618276e-01 -7.09763050e-01 4.80360121e-01 -8.31388593e-01 -1.15645361e+00 1.07411653e-01 -6.19844496e-01 5.22687018e-01 -2.90060192e-01 1.03668654e+00 9.89098996e-02 7.01537207e-02 -2.59722471e-01 -2.42247224e-01 -5.57830520e-02 -4.81822997e-01 2.33041197e-01 -7.50754058e-01 -1.20686643e-01 2.23768443e-01 -3.04785252e-01 -3.27351004e-01 3.35310698e-01 -1.18527997e+00 3.07118535e-01 8.64117384e-01 8.62128854e-01 7.24776328e-01 -7.84631297e-02 2.30738759e-01 -1.30222833e+00 4.38763142e-01 -4.95226830e-01 -8.96340668e-01 7.26206303e-01 -4.96745080e-01 5.20810783e-01 9.91854548e-01 1.11097634e-01 -1.10232937e+00 5.29168658e-02 -4.70213085e-01 3.09355929e-03 2.90835798e-02 7.16576517e-01 -7.98415184e-01 6.02273583e-01 2.53667057e-01 5.75055242e-01 -1.89266518e-01 -7.10798800e-01 3.62312555e-01 6.99171245e-01 6.20176613e-01 -8.96483660e-01 8.34983647e-01 1.74132232e-02 -2.68322408e-01 -3.13463449e-01 -5.06707311e-01 -9.07020569e-02 -2.99139559e-01 1.92620203e-01 6.13235235e-01 -8.10701430e-01 -7.23153114e-01 3.40630263e-01 -1.48485792e+00 8.88030082e-02 -1.43716842e-01 -1.23546146e-01 -4.12514627e-01 5.53106189e-01 -7.39030659e-01 -4.56886679e-01 -4.40313309e-01 -1.39527106e+00 1.26941979e+00 2.52957232e-02 -8.54610354e-02 -4.45407659e-01 -1.57588676e-01 3.63732636e-01 2.03684121e-01 -8.76841322e-02 1.71521842e+00 -9.44201827e-01 -1.11708248e+00 -3.81797515e-02 -5.08110106e-01 3.36458266e-01 -1.62163824e-01 2.15396196e-01 -5.02114058e-01 9.24125463e-02 -8.75103381e-03 -4.03601915e-01 4.51191872e-01 -5.03168643e-01 1.29331851e+00 -5.61299562e-01 -8.03556144e-02 7.80457735e-01 1.59240782e+00 2.98431516e-01 7.86075413e-01 7.84608349e-02 5.99104106e-01 5.56197107e-01 6.96750104e-01 5.46760559e-01 6.15300119e-01 5.63240349e-01 6.43526316e-01 4.33894575e-01 7.62960538e-02 -9.27795768e-01 5.87703526e-01 1.27443182e+00 6.27072453e-01 -4.97531921e-01 -1.00533009e+00 1.23701520e-01 -1.77604604e+00 -4.55904692e-01 -3.22776377e-01 2.26155329e+00 9.60114539e-01 3.28923941e-01 -3.21040362e-01 -8.60671848e-02 6.01089954e-01 2.16534808e-01 -6.80972099e-01 -6.73510373e-01 -8.18094891e-03 2.55167216e-01 5.01955271e-01 3.85149658e-01 -7.25865662e-01 9.59429264e-01 5.21535301e+00 7.82014549e-01 -8.49073291e-01 -1.62448272e-01 2.35005364e-01 -7.92132393e-02 -6.65162206e-01 2.58979887e-01 -1.17195714e+00 7.04626858e-01 8.93095672e-01 -1.31535575e-01 4.62664753e-01 8.02037954e-01 9.43036303e-02 -5.71741350e-02 -1.36098826e+00 5.28166354e-01 -2.61707991e-01 -1.15015042e+00 3.59319896e-01 -2.39123553e-01 7.50396103e-02 -1.53077528e-01 -2.52860725e-01 4.11938339e-01 -2.97074951e-02 -6.28523052e-01 9.33791935e-01 6.59372449e-01 8.41468036e-01 -7.25901842e-01 7.11690605e-01 2.68684417e-01 -1.27957308e+00 -1.19270079e-01 -2.56316304e-01 4.92771059e-01 5.07948138e-02 6.41388416e-01 -4.83923584e-01 8.56365502e-01 6.68713570e-01 5.39129555e-01 -7.38697767e-01 4.89080846e-01 -3.51346076e-01 4.77657050e-01 -3.01367074e-01 -1.62141874e-01 -4.25787419e-02 -1.43221214e-01 3.63082021e-01 1.19265842e+00 3.72329593e-01 -9.91419181e-02 1.44497067e-01 9.87125099e-01 -1.25843808e-01 2.61457890e-01 -3.09578359e-01 -3.37223202e-01 7.59453714e-01 8.67133796e-01 -2.86767572e-01 -4.98192191e-01 -6.84614897e-01 7.98144460e-01 3.53992462e-01 2.43116438e-01 -1.15201879e+00 -7.57353902e-01 5.76017141e-01 2.02226639e-01 5.76314270e-01 -1.29797965e-01 -5.59924006e-01 -1.35768986e+00 6.82854950e-01 -1.34220648e+00 4.04020756e-01 -7.81461000e-01 -8.58558953e-01 4.36395675e-01 -1.36305660e-01 -8.51599276e-01 -1.50186613e-01 -2.33550757e-01 -4.51114625e-01 9.33787942e-01 -1.32534993e+00 -9.72033024e-01 -1.84315950e-01 4.63173985e-01 3.77569377e-01 9.87238600e-04 6.16715550e-01 5.87971330e-01 -9.88577902e-01 8.01635206e-01 -9.55464393e-02 2.34767556e-01 7.07632840e-01 -9.64636266e-01 6.69312656e-01 9.60739255e-01 -2.58697897e-01 1.04646754e+00 3.59283835e-01 -9.48534489e-01 -2.09490442e+00 -1.09977293e+00 1.12505412e+00 -1.61451101e-01 7.15592563e-01 -7.86353171e-01 -1.05248070e+00 7.03882158e-01 4.75768596e-02 -4.47053879e-01 4.75938201e-01 -3.31208467e-01 -6.57221019e-01 -5.75826287e-01 -6.32391810e-01 6.98023081e-01 1.08914733e+00 -7.23743141e-01 -6.20783150e-01 1.86424509e-01 1.08277941e+00 -6.50306404e-01 -9.48010325e-01 4.86627996e-01 6.04712069e-01 -1.14532506e+00 8.11602414e-01 -4.74184871e-01 8.35124552e-01 -3.82114679e-01 -4.54346091e-01 -4.81647104e-01 1.58784121e-01 -4.30046082e-01 -2.66845644e-01 1.67053068e+00 5.41885495e-01 -4.64690864e-01 7.59259462e-01 3.90210032e-01 -2.84115970e-01 -8.89813662e-01 -4.86882508e-01 -5.66560328e-01 -3.42588693e-01 -4.93302464e-01 8.85156810e-01 3.68718117e-01 -2.46728376e-01 3.49952608e-01 -2.21971661e-01 3.95370066e-01 4.34374332e-01 4.61261094e-01 9.22728121e-01 -6.56865597e-01 -5.57631850e-01 -1.65644780e-01 1.47533208e-01 -1.44391215e+00 -1.96667939e-01 -9.42891836e-01 4.24317606e-02 -1.35430253e+00 1.18867785e-01 -4.35655504e-01 3.12177464e-02 5.11527956e-01 -1.53711557e-01 -7.45745778e-01 1.20505050e-01 1.63986713e-01 -4.72934633e-01 2.65323877e-01 8.92910779e-01 -9.67454538e-02 -5.13519607e-02 -3.59936617e-03 -8.14466357e-01 2.37130210e-01 9.40788805e-01 -6.30087018e-01 -5.81048906e-01 -8.40388358e-01 6.94125295e-01 5.96479595e-01 2.51391858e-01 -6.70858264e-01 4.45981205e-01 -1.62423640e-01 -7.00092018e-02 -9.11348939e-01 -9.98373628e-02 -7.61413097e-01 2.34666258e-01 5.43567479e-01 -5.45917869e-01 4.86309290e-01 2.01261565e-01 4.42537576e-01 -3.86008531e-01 -2.83251673e-01 4.34462875e-01 -1.24114685e-01 -4.98617113e-01 2.81913459e-01 6.48124889e-03 6.01370037e-01 7.55668044e-01 1.88093379e-01 -5.98382473e-01 -8.02224874e-02 -2.80707568e-01 4.13293988e-01 3.93907517e-01 5.55420339e-01 5.99753499e-01 -1.07976747e+00 -4.78881150e-01 6.22660220e-01 3.57032597e-01 2.65100718e-01 9.57840867e-03 6.14034295e-01 -7.96238244e-01 5.93264461e-01 -4.36272696e-02 -4.29894626e-01 -9.71914351e-01 6.70270503e-01 1.15374595e-01 -4.99663532e-01 -1.88395053e-01 6.85438812e-01 3.49342227e-01 -6.57960534e-01 4.86541092e-01 -4.63205695e-01 1.96318954e-01 -1.53954774e-01 1.31942093e-01 2.15692118e-01 1.30123436e-01 -2.06313767e-02 -5.30853271e-01 4.77878749e-01 -3.67389113e-01 1.68213263e-01 1.02618313e+00 8.62690955e-02 -7.11054206e-01 2.39057317e-01 1.26152468e+00 4.30380881e-01 -8.07496965e-01 -1.62187248e-01 3.30568582e-01 -2.49814183e-01 -4.78821367e-01 -9.37786162e-01 -9.73180234e-01 1.04739332e+00 -6.87990040e-02 1.52986720e-01 1.25282693e+00 -8.30183700e-02 1.08738554e+00 4.80103672e-01 3.13225657e-01 -9.79923368e-01 -9.18447897e-02 6.91954494e-01 7.70908892e-01 -8.28392327e-01 -1.53169230e-01 -7.43118346e-01 -5.63066125e-01 1.38987517e+00 7.41417527e-01 1.87977433e-01 3.10925007e-01 6.94458842e-01 -1.24856435e-01 -1.08802952e-02 -1.04479384e+00 9.48665813e-02 -2.15655230e-02 9.28589627e-02 5.07312953e-01 -1.19916461e-01 -4.63388532e-01 9.82591093e-01 -1.63010508e-01 -7.75474384e-02 3.60985279e-01 1.06363368e+00 -3.23825598e-01 -1.62547112e+00 -1.50760219e-01 6.00914121e-01 -5.00607550e-01 -4.79919136e-01 -3.25654328e-01 6.59027219e-01 1.76629752e-01 7.21141517e-01 8.37743189e-03 -6.12064004e-01 7.46088564e-01 3.89172286e-01 -2.14167181e-02 -7.10040867e-01 -8.64686906e-01 1.84386835e-01 4.01057631e-01 -9.49405253e-01 4.19606566e-01 -5.39465606e-01 -1.48121917e+00 -3.91832054e-01 -9.76298526e-02 1.73076570e-01 6.08445823e-01 2.81927228e-01 7.37170041e-01 5.79431236e-01 7.73489833e-01 4.03038025e-01 -9.42994893e-01 -4.34470952e-01 -1.43764034e-01 1.63413808e-01 4.58108336e-02 1.75634697e-02 8.23142380e-02 -6.79056048e-02]
[9.919173240661621, 7.826427459716797]
5b3fcafc-2797-4648-97ed-4f41c6a12810
epg-mgcn-ego-planning-guided-multi-graph
2303.17027
null
https://arxiv.org/abs/2303.17027v1
https://arxiv.org/pdf/2303.17027v1.pdf
EPG-MGCN: Ego-Planning Guided Multi-Graph Convolutional Network for Heterogeneous Agent Trajectory Prediction
To drive safely in complex traffic environments, autonomous vehicles need to make an accurate prediction of the future trajectories of nearby heterogeneous traffic agents (i.e., vehicles, pedestrians, bicyclists, etc). Due to the interactive nature, human drivers are accustomed to infer what the future situations will become if they are going to execute different maneuvers. To fully exploit the impacts of interactions, this paper proposes a ego-planning guided multi-graph convolutional network (EPG-MGCN) to predict the trajectories of heterogeneous agents using both historical trajectory information and ego vehicle's future planning information. The EPG-MGCN first models the social interactions by employing four graph topologies, i.e., distance graphs, visibility graphs, planning graphs and category graphs. Then, the planning information of the ego vehicle is encoded by both the planning graph and the subsequent planning-guided prediction module to reduce uncertainty in the trajectory prediction. Finally, a category-specific gated recurrent unit (CS-GRU) encoder-decoder is designed to generate future trajectories for each specific type of agents. Our network is evaluated on two real-world trajectory datasets: ApolloScape and NGSIM. The experimental results show that the proposed EPG-MGCN achieves state-of-the-art performance compared to existing methods.
['Sikai Chen', 'Zilin Huang', 'Zihao Sheng']
2023-03-29
null
null
null
null
['trajectory-prediction']
['computer-vision']
[-4.35927480e-01 3.61612618e-01 -6.91521317e-02 -4.20971364e-01 -2.40305245e-01 -2.08116934e-01 8.76103461e-01 -1.63427144e-01 -1.72398929e-02 6.16004407e-01 3.66357863e-01 -5.73915720e-01 7.40990639e-02 -1.22374892e+00 -8.76639962e-01 -5.93990624e-01 -4.84809101e-01 7.18406439e-01 6.21040404e-01 -5.30198812e-01 -2.60878690e-02 5.00792980e-01 -1.64004147e+00 6.92438930e-02 9.01056588e-01 5.59555531e-01 3.51766914e-01 6.68461382e-01 6.34186566e-02 1.01993263e+00 -8.67433697e-02 -4.20591563e-01 1.25317752e-01 -1.72386825e-01 -2.04000860e-01 -5.37837185e-02 -3.25817108e-01 -4.68929112e-01 -1.38509583e+00 8.57384741e-01 1.19709931e-01 6.01983249e-01 7.18119204e-01 -1.70365167e+00 -3.43599409e-01 7.47911751e-01 -7.40886852e-02 3.94716322e-01 -4.70092706e-02 6.55853510e-01 5.25009334e-01 -4.86260921e-01 7.29045391e-01 1.53810132e+00 2.29442149e-01 4.13212478e-01 -4.67822790e-01 -7.72395790e-01 6.82293832e-01 9.79733348e-01 -1.42402363e+00 -3.45685452e-01 7.73937762e-01 -4.49367106e-01 9.94292080e-01 -7.57850781e-02 7.13085532e-01 1.13555479e+00 7.49161422e-01 7.47899234e-01 2.28779003e-01 5.52489638e-01 1.75905496e-01 -6.76486641e-02 -4.03063968e-02 6.04220867e-01 2.57347450e-02 6.41393900e-01 -4.15428691e-02 1.58687428e-01 1.69568747e-01 6.85076565e-02 2.03434184e-01 -5.81806637e-02 -1.01276541e+00 7.83945262e-01 6.83714449e-01 -3.14053297e-01 -6.92461133e-01 3.90068501e-01 3.97973508e-01 1.65616408e-01 4.96687591e-01 -4.09191161e-01 2.30840296e-01 -2.47453213e-01 -3.99244189e-01 6.07322335e-01 6.52937293e-01 1.59588873e+00 8.60555947e-01 2.06881151e-01 -3.70524704e-01 2.48576328e-01 5.69199562e-01 6.64576173e-01 1.03677474e-01 -7.60176480e-01 8.69651973e-01 6.33097231e-01 1.22361630e-01 -1.25586581e+00 -5.79667032e-01 -4.04058456e-01 -7.16089964e-01 -4.95941304e-02 -2.43639145e-02 -5.39669394e-01 -9.41093743e-01 1.66812956e+00 3.87223095e-01 8.00077915e-01 3.36200446e-01 9.48756635e-01 8.37513328e-01 1.07737172e+00 1.51467144e-01 5.50824888e-02 7.39847958e-01 -1.22591114e+00 -6.30817652e-01 -3.88310164e-01 7.11013675e-01 -1.29108265e-01 1.15292683e-01 -3.35467517e-01 -8.21430206e-01 -6.33296132e-01 -8.07762682e-01 3.35025638e-01 -5.01560211e-01 -8.62217620e-02 3.24028283e-01 1.72334090e-01 -9.54927027e-01 3.89466137e-01 -9.21236634e-01 -2.34873995e-01 4.48832959e-01 2.32468769e-01 1.49166044e-02 -2.42591515e-01 -1.43185008e+00 8.23215544e-01 5.24188638e-01 3.61855060e-01 -1.45983160e+00 -3.99435669e-01 -1.00179827e+00 7.21875876e-02 3.63850057e-01 -5.93103647e-01 9.01143849e-01 -3.97140473e-01 -1.26565695e+00 8.25002119e-02 -3.36469561e-01 -7.39364326e-01 7.77840555e-01 4.06080723e-01 -9.55073237e-01 -1.67838737e-01 1.77196145e-01 6.15941882e-01 5.03505528e-01 -1.04759085e+00 -1.12577152e+00 -2.59274662e-01 9.85861048e-02 5.85535944e-01 4.52888936e-01 -3.70771706e-01 -6.74518049e-01 -1.60406739e-01 -3.38678598e-01 -1.32643616e+00 -5.33542275e-01 -5.73918223e-01 -5.83967447e-01 -6.92631662e-01 1.13310254e+00 -4.85420018e-01 1.16395783e+00 -2.30395508e+00 2.01969758e-01 3.24243903e-01 3.45673174e-01 2.37073034e-01 -2.85810173e-01 7.78982341e-01 3.93250853e-01 -2.14812949e-01 1.80434301e-01 -3.18887889e-01 1.07795469e-01 4.00438637e-01 -2.57942170e-01 5.06337166e-01 -7.08711594e-02 1.10793769e+00 -1.11211455e+00 -2.03556001e-01 6.16357923e-01 5.45694172e-01 -3.37261647e-01 2.17224300e-01 -3.74800146e-01 4.94783431e-01 -9.95558858e-01 1.83589146e-01 6.45936549e-01 -3.63355242e-02 -2.41822116e-02 3.81719351e-01 -3.62553805e-01 2.16217577e-01 -8.08911026e-01 9.67250049e-01 -3.24543327e-01 9.28851724e-01 -3.22511822e-01 -6.94383264e-01 9.35140252e-01 1.61338598e-01 2.24021226e-01 -7.67718256e-01 3.10016423e-01 7.25930333e-02 1.90305576e-01 -4.75936413e-01 5.57327390e-01 4.54698831e-01 -9.97654051e-02 1.99848354e-01 -4.01265949e-01 4.86920655e-01 3.46977890e-01 4.69676465e-01 1.07636714e+00 -1.50983438e-01 -2.68543154e-01 -3.63873467e-02 5.86038411e-01 1.77068144e-01 8.07862222e-01 5.15021026e-01 -4.32593107e-01 2.69780010e-01 4.52155113e-01 -6.90740645e-01 -9.95350361e-01 -1.04159033e+00 4.25269127e-01 7.09781706e-01 7.38593340e-01 -1.96183681e-01 -5.34981728e-01 -4.83704299e-01 6.35970607e-02 1.22103822e+00 -6.14090025e-01 -5.96080542e-01 -8.02387476e-01 -4.93180990e-01 2.09267527e-01 3.78051996e-01 6.33373499e-01 -1.07144487e+00 -2.49488980e-01 5.08159518e-01 -7.24649727e-02 -1.29374015e+00 -6.64068758e-01 -7.49293089e-01 -1.64496362e-01 -1.03888702e+00 -1.43130869e-01 -5.37285507e-01 5.15098095e-01 6.59670353e-01 5.56942940e-01 7.06326813e-02 4.08248782e-01 9.21737850e-02 -3.51791799e-01 -3.37469846e-01 -6.63015783e-01 1.34946287e-01 8.53572041e-02 5.36497712e-01 5.10710061e-01 -6.59778059e-01 -8.32566440e-01 5.61468005e-01 -2.25473046e-01 4.80642021e-01 4.54329133e-01 2.91543156e-01 3.98633033e-01 3.62317652e-01 7.64626741e-01 -7.21821606e-01 4.48714584e-01 -1.26409364e+00 -5.62751472e-01 -1.21641334e-03 -4.80686605e-01 -2.70534039e-01 9.17756677e-01 -2.42365152e-01 -1.10209537e+00 -1.39307126e-01 -1.77470699e-01 -6.33604765e-01 -1.76598176e-01 5.36400974e-01 -2.05939531e-01 2.66866416e-01 1.40785009e-01 4.19423789e-01 -7.30030462e-02 2.49381647e-01 4.99137402e-01 5.73651254e-01 4.34906870e-01 1.30074114e-01 8.45545292e-01 4.69498873e-01 5.64935729e-02 -8.22948694e-01 -1.76928908e-01 -1.76689759e-01 -2.58017629e-01 -9.56123948e-01 8.96734893e-01 -1.13981307e+00 -9.13636148e-01 5.32247126e-01 -1.20438874e+00 -5.67089319e-01 5.94459698e-02 6.63316011e-01 -6.78052783e-01 1.18554467e-02 -4.47056502e-01 -8.17438841e-01 1.12666517e-01 -1.43432546e+00 8.02241087e-01 5.14054656e-01 4.18590009e-01 -1.17629945e+00 -9.73916873e-02 1.50719181e-01 3.69459450e-01 3.87033969e-01 7.68794358e-01 -5.55629075e-01 -1.20172691e+00 -2.55442083e-01 -3.34065527e-01 -3.24302375e-01 -8.89987871e-02 -1.03283897e-01 -4.28063124e-01 -2.80517429e-01 -6.84656322e-01 4.38712776e-01 7.25560606e-01 4.71091866e-01 8.56191218e-01 -5.01039326e-01 -1.00377464e+00 6.44605994e-01 9.64209497e-01 5.85794628e-01 8.04279089e-01 -4.15747426e-02 1.02145743e+00 7.58422136e-01 6.98392987e-01 2.64943033e-01 1.33716011e+00 7.33759463e-01 8.95569980e-01 4.38024104e-01 -2.94223040e-01 -6.89520419e-01 5.43436110e-01 7.90112674e-01 -1.01835541e-01 -8.21290076e-01 -8.99101555e-01 7.61895359e-01 -2.37774539e+00 -1.23207557e+00 -5.13664663e-01 1.80429673e+00 -3.45476598e-01 1.53932407e-01 1.94601372e-01 -5.90572894e-01 1.00618017e+00 4.06003624e-01 -1.05062079e+00 -3.57457310e-01 5.85794896e-02 -9.40608799e-01 8.34514380e-01 6.50766790e-01 -8.64618778e-01 1.27397990e+00 4.80274820e+00 8.82848620e-01 -7.92757750e-01 7.13180676e-02 5.88062584e-01 -5.72011918e-02 -4.04589713e-01 -9.43527147e-02 -1.06851912e+00 9.19427216e-01 1.40953851e+00 -6.90546036e-01 7.21672714e-01 7.16843605e-01 7.28525162e-01 1.95020124e-01 -7.84038782e-01 6.39813602e-01 -1.82778627e-01 -1.63453770e+00 1.74509317e-01 2.65927732e-01 6.00025833e-01 7.67304182e-01 9.00013372e-02 6.29620492e-01 7.10819066e-01 -9.03068781e-01 7.29143798e-01 8.80795002e-01 3.82418692e-01 -1.28886676e+00 6.52267873e-01 6.80599272e-01 -1.60554731e+00 -3.79922271e-01 -3.26818049e-01 8.16024169e-02 8.13953042e-01 1.06790014e-01 -9.37282860e-01 8.22433054e-01 2.95034915e-01 1.19012034e+00 -2.96053231e-01 8.38617384e-01 -2.10936666e-01 5.23519695e-01 -1.40563026e-01 -2.82853246e-01 7.87187397e-01 -5.08579552e-01 1.09437430e+00 9.86484051e-01 5.59409142e-01 3.61943394e-01 2.42493585e-01 9.06258464e-01 2.15474054e-01 -3.43579590e-01 -1.10712504e+00 1.64624639e-02 6.71051443e-01 1.04136026e+00 -4.36644077e-01 -3.38698953e-01 -4.27413821e-01 4.93395448e-01 3.98903161e-01 7.44778454e-01 -1.02640486e+00 -2.17178598e-01 1.01746190e+00 1.63204029e-01 2.93198436e-01 -4.01764363e-01 3.07658970e-01 -7.87758589e-01 -1.83699027e-01 -1.34000227e-01 9.47109014e-02 -7.51648068e-01 -8.65292668e-01 9.32887733e-01 -3.81186828e-02 -1.38295877e+00 -6.41298771e-01 -1.84263960e-01 -1.21841764e+00 7.79642522e-01 -1.62707722e+00 -1.30260432e+00 -2.77131647e-01 5.38163841e-01 6.82864547e-01 -5.58495760e-01 -2.59342827e-02 2.99762040e-01 -7.77661622e-01 2.32819721e-01 8.29371288e-02 8.64607543e-02 -1.10078387e-01 -6.87338710e-01 9.62213516e-01 9.09067810e-01 -4.24780816e-01 -6.08004890e-02 6.85343504e-01 -1.04773033e+00 -1.59990764e+00 -2.06241775e+00 9.14410532e-01 -1.84090704e-01 6.04832828e-01 -2.22668260e-01 -6.65335894e-01 1.02223206e+00 2.44569220e-02 -3.04402020e-02 1.55659154e-01 -3.46734256e-01 2.97757983e-01 -2.09066737e-02 -7.21716464e-01 1.03970098e+00 1.37841368e+00 -2.41007000e-01 -1.27115741e-01 4.82508630e-01 9.07584906e-01 -5.25278866e-01 -3.86301786e-01 2.57004768e-01 3.22664708e-01 -8.27328146e-01 6.92440093e-01 -6.04879320e-01 1.06374294e-01 -4.02943343e-01 2.02479935e-03 -1.69097304e+00 -5.64312875e-01 -5.52177906e-01 -1.91313148e-01 8.28705966e-01 4.83610928e-01 -8.79601538e-01 6.78213656e-01 6.71889782e-01 -7.48897195e-01 -7.59223819e-01 -1.11291015e+00 -5.94813108e-01 -1.14898622e-01 -7.02307105e-01 1.13428593e+00 3.57961774e-01 -1.29656270e-01 2.71397054e-01 -5.06673455e-01 7.22330272e-01 6.73537433e-01 4.50475104e-02 9.43916321e-01 -8.50396276e-01 3.01086068e-01 -4.94577646e-01 -7.33791828e-01 -1.17984366e+00 5.64017892e-01 -1.13861704e+00 7.11209979e-03 -1.67030263e+00 -1.56309187e-01 -4.47878361e-01 8.28720406e-02 -1.57995045e-01 -3.59886251e-02 -3.02726775e-01 2.08411992e-01 -9.04262764e-04 -8.04498196e-01 9.93120372e-01 1.43899548e+00 -2.49545947e-01 -3.48294705e-01 4.05632108e-01 -2.42283195e-01 4.73717391e-01 8.21422219e-01 -2.81109184e-01 -8.45017135e-01 -3.91289353e-01 1.58035502e-01 5.46983480e-01 5.07874191e-01 -9.90156054e-01 7.71726310e-01 -3.89815420e-01 -2.69304067e-01 -1.16695428e+00 3.95554155e-01 -6.08294368e-01 5.79346299e-01 4.25187260e-01 -1.52179956e-01 9.67558846e-02 5.92546239e-02 1.21308339e+00 -4.59800195e-03 3.95353913e-01 2.70382583e-01 1.31480470e-01 -1.12497139e+00 1.10222089e+00 -9.76674080e-01 -1.38629183e-01 1.50928700e+00 -1.77441567e-01 -5.61213911e-01 -8.22534084e-01 -6.83590829e-01 1.26595938e+00 6.24603517e-02 9.11932409e-01 9.05774593e-01 -1.55270815e+00 -9.47366178e-01 3.39581102e-01 1.88744009e-01 -3.24361771e-02 8.78925920e-01 7.18540072e-01 -3.45488697e-01 5.74479461e-01 -9.83096510e-02 -4.46496129e-01 -7.83317089e-01 6.46167040e-01 4.15565461e-01 -9.00717974e-02 -1.05253005e+00 3.83309931e-01 4.32375669e-01 -4.30780143e-01 -6.91886097e-02 -6.40883595e-02 -5.95681727e-01 -2.53645182e-01 5.44702172e-01 7.74456620e-01 -3.35977346e-01 -1.35215414e+00 -1.58014759e-01 1.66278318e-01 -6.71245381e-02 7.88253024e-02 1.03037047e+00 -5.44260919e-01 3.47123593e-01 2.96251565e-01 1.10090685e+00 -6.74123824e-01 -1.74132431e+00 -1.39934450e-01 -4.29078788e-01 -3.05245429e-01 7.82110319e-02 -3.02212209e-01 -1.33651233e+00 8.36162567e-01 2.21431866e-01 2.18636319e-01 6.13143384e-01 2.37800162e-02 1.16705871e+00 1.87208280e-01 7.26411641e-01 -1.08027041e+00 -5.71507573e-01 7.59338617e-01 8.12995672e-01 -1.02726460e+00 -5.30067921e-01 -4.68146265e-01 -8.54149044e-01 9.23963487e-01 7.28566051e-01 -2.80797780e-01 9.62638617e-01 -2.99475729e-01 -2.56633610e-01 -2.57480621e-01 -1.30216646e+00 -4.22586083e-01 2.67291665e-01 7.19492853e-01 -5.07773161e-01 6.36982858e-01 3.03503349e-02 3.61067861e-01 -3.09533566e-01 -2.63021588e-01 6.76857114e-01 2.46468425e-01 -4.66272473e-01 -6.17736042e-01 2.46611372e-01 6.11119270e-01 3.62687528e-01 3.99764597e-01 -2.51431130e-02 6.69264674e-01 3.92062217e-02 1.34532440e+00 4.40285057e-01 -8.76961410e-01 2.95283258e-01 -4.46028113e-01 -2.09326044e-01 -4.37142700e-01 -1.46823093e-01 -4.19463605e-01 3.84277642e-01 -7.20313966e-01 9.29150078e-03 -1.01793063e+00 -1.56002390e+00 -8.96114409e-01 2.37576857e-01 1.75054282e-01 4.16292846e-01 1.11533606e+00 8.62613499e-01 8.65975380e-01 9.89256442e-01 -1.00763261e+00 8.32891986e-02 -7.55356908e-01 -4.17616487e-01 1.79966912e-01 4.27461326e-01 -7.86140680e-01 -1.75717786e-01 -5.19673347e-01]
[5.948138236999512, 0.9262908697128296]
2506af5c-f9b7-428f-8ca7-055f80c2611f
a-closer-look-at-novel-class-discovery-from
2209.09120
null
https://arxiv.org/abs/2209.09120v4
https://arxiv.org/pdf/2209.09120v4.pdf
A Closer Look at Novel Class Discovery from the Labeled Set
Novel class discovery (NCD) aims to infer novel categories in an unlabeled dataset leveraging prior knowledge of a labeled set comprising disjoint but related classes. Existing research focuses primarily on utilizing the labeled set at the methodological level, with less emphasis on the analysis of the labeled set itself. Thus, in this paper, we rethink novel class discovery from the labeled set and focus on two core questions: (i) Given a specific unlabeled set, what kind of labeled set can best support novel class discovery? (ii) A fundamental premise of NCD is that the labeled set must be related to the unlabeled set, but how can we measure this relation? For (i), we propose and substantiate the hypothesis that NCD could benefit more from a labeled set with a large degree of semantic similarity to the unlabeled set. Specifically, we establish an extensive and large-scale benchmark with varying degrees of semantic similarity between labeled/unlabeled datasets on ImageNet by leveraging its hierarchical class structure. As a sharp contrast, the existing NCD benchmarks are developed based on labeled sets with different number of categories and images, and completely ignore the semantic relation. For (ii), we introduce a mathematical definition for quantifying the semantic similarity between labeled and unlabeled sets. In addition, we use this metric to confirm the validity of our proposed benchmark and demonstrate that it highly correlates with NCD performance. Furthermore, without quantitative analysis, previous works commonly believe that label information is always beneficial. However, counterintuitively, our experimental results show that using labels may lead to sub-optimal outcomes in low-similarity settings.
['Haojin Yang', 'Christoph Meinel', 'Di Hu', 'Ben Dai', 'Jona Otholt', 'Ziyun Li']
2022-09-19
null
null
null
null
['novel-class-discovery', 'novel-class-discovery']
['computer-vision', 'methodology']
[ 2.98000127e-01 1.93846852e-01 -3.85058075e-01 -6.73269749e-01 -2.95671076e-01 -8.95499945e-01 6.29848182e-01 2.99164802e-01 -3.04168701e-01 6.60114467e-01 3.55108157e-02 -1.45787865e-01 -2.87841707e-01 -8.24701905e-01 -5.38882554e-01 -5.91181397e-01 1.94817737e-01 2.14814842e-01 1.76144555e-01 -1.49822803e-02 2.30804488e-01 5.23324072e-01 -1.81943464e+00 3.40681598e-02 7.26674139e-01 1.05322683e+00 -4.46539856e-02 -4.06226143e-02 -1.06107935e-01 6.00542784e-01 -3.87698084e-01 -3.40465844e-01 5.07112741e-01 -6.58711195e-01 -1.14906132e+00 2.82937735e-01 6.37653947e-01 9.60426480e-02 -6.46537021e-02 1.43069577e+00 1.41428366e-01 4.28752229e-02 8.83222044e-01 -1.40197623e+00 -6.80874407e-01 6.81372523e-01 -4.56699461e-01 2.39644527e-01 1.39048383e-01 4.43253741e-02 1.22564161e+00 -9.09700036e-01 7.95517981e-01 1.00126421e+00 5.54199517e-01 5.28594553e-01 -1.26382804e+00 -9.43463385e-01 4.58361655e-01 2.15493977e-01 -1.57876861e+00 -3.10776412e-01 7.95399964e-01 -5.39683342e-01 2.43411556e-01 2.77869970e-01 4.15074557e-01 9.30030823e-01 -4.37105983e-01 5.77553988e-01 1.41278219e+00 -5.29447258e-01 3.63385886e-01 4.38151985e-01 6.85320973e-01 4.34786737e-01 4.64430660e-01 2.85022140e-01 -2.65887380e-01 2.42399737e-01 3.63860369e-01 8.64852369e-02 -2.80358702e-01 -5.97462296e-01 -1.29341638e+00 8.30865026e-01 4.60851759e-01 5.57607651e-01 -1.43085420e-01 -3.06460768e-01 3.58415216e-01 4.28716868e-01 3.41579914e-01 7.82517076e-01 -5.06462514e-01 3.42566639e-01 -7.80319929e-01 -3.14059965e-02 6.58486128e-01 1.07166946e+00 1.17216551e+00 -2.53751606e-01 -1.14156954e-01 5.89599609e-01 1.89740434e-01 3.03454220e-01 6.29903674e-01 -7.71641552e-01 -9.12148599e-03 8.48153770e-01 -1.91273883e-01 -1.01309967e+00 -3.20568860e-01 -5.57203889e-01 -6.71103477e-01 2.61170715e-02 3.84918392e-01 1.37802362e-01 -8.53886783e-01 1.96615005e+00 4.22769189e-01 4.28010017e-01 1.67505279e-01 8.37314010e-01 1.00991285e+00 -3.11284941e-02 6.54092105e-03 -4.34854418e-01 1.27950966e+00 -8.47560823e-01 -4.80623752e-01 -7.31522068e-02 7.08209693e-01 -5.59484303e-01 1.22056448e+00 5.66477552e-02 -4.38776523e-01 -6.56864464e-01 -1.09622169e+00 1.75310537e-01 -5.60895801e-01 -7.79036358e-02 6.57569170e-01 7.81718254e-01 -8.57211292e-01 3.56147707e-01 -3.18810582e-01 -7.37433612e-01 3.69438857e-01 8.32408667e-02 -4.23923969e-01 -2.77823836e-01 -1.10061097e+00 7.64167488e-01 8.49626780e-01 -2.45668828e-01 -8.42509151e-01 -7.01018035e-01 -5.98789275e-01 9.53519940e-02 6.65384829e-01 -6.07098758e-01 9.65521038e-01 -1.19633687e+00 -9.86396790e-01 1.20093477e+00 -8.55381116e-02 -3.99942666e-01 2.17576265e-01 2.02469021e-01 -6.27842009e-01 2.98566371e-01 3.81983429e-01 8.28030288e-01 6.87478542e-01 -1.67804599e+00 -8.51115882e-01 -4.85440314e-01 4.16715294e-01 1.39757112e-01 -5.44970155e-01 -2.91239887e-01 -1.27666295e-01 -6.46237195e-01 4.38115776e-01 -9.02627468e-01 1.73575617e-02 -9.62726921e-02 -5.30354559e-01 -3.10209960e-01 9.87733901e-01 9.58869606e-02 1.06640875e+00 -2.34161854e+00 -2.75458962e-01 2.61994869e-01 7.42452264e-01 2.29285926e-01 2.80091688e-02 5.14095463e-02 -4.05966371e-01 3.51252675e-01 -2.08049268e-01 1.42181695e-01 -1.88910477e-02 2.19686687e-01 -4.34808850e-01 3.43969941e-01 1.07050411e-01 9.19656515e-01 -1.14981282e+00 -5.88739693e-01 1.49385467e-01 -1.27062142e-01 -4.32178766e-01 5.21516167e-02 4.87461872e-02 5.01874030e-01 -4.77413386e-01 7.50588238e-01 7.06600785e-01 -5.48041880e-01 3.47872406e-01 -5.74485421e-01 7.38078542e-03 -1.01121128e-01 -1.05091405e+00 1.33506167e+00 9.75403190e-03 5.06535649e-01 -2.92985290e-01 -1.50042355e+00 9.75141406e-01 4.44356240e-02 4.74165916e-01 -5.45986772e-01 2.84812719e-01 2.19893366e-01 9.68317613e-02 -5.10369241e-01 2.48242512e-01 -5.72835684e-01 8.42069238e-02 4.50938672e-01 1.81606159e-01 1.34617999e-01 1.83339775e-01 2.56278813e-01 1.05020678e+00 -2.04622269e-01 5.61158717e-01 -5.57058275e-01 3.71317744e-01 7.27585182e-02 7.30938494e-01 8.77715468e-01 -5.68921924e-01 4.85795051e-01 2.96631575e-01 -1.04400076e-01 -7.64643312e-01 -1.22517991e+00 -4.02578801e-01 9.18133676e-01 7.22250879e-01 -1.64526477e-01 -4.33193207e-01 -1.20036519e+00 -3.74093652e-02 7.61614740e-01 -8.96430254e-01 -1.95491329e-01 -1.45757645e-01 -5.91057122e-01 5.13701022e-01 3.67344916e-01 6.53003454e-01 -7.89986551e-01 -3.62300932e-01 -2.25450456e-01 -1.46161094e-01 -1.22092915e+00 -3.78678143e-01 2.05008626e-01 -8.56357157e-01 -1.58407199e+00 -2.71576703e-01 -9.26721752e-01 8.91372740e-01 7.28337467e-01 1.24396598e+00 9.11210105e-02 7.57075555e-04 6.11069441e-01 -7.61051655e-01 -3.87186885e-01 -2.75781959e-01 3.95864435e-02 8.67117494e-02 1.28828764e-01 6.79689109e-01 -6.59272611e-01 -5.06562769e-01 6.28778279e-01 -1.06734538e+00 -4.34083305e-03 5.88845015e-01 7.01398492e-01 6.42799497e-01 2.73649931e-01 8.79722118e-01 -1.13417697e+00 2.74443686e-01 -8.33662629e-01 -2.84784704e-01 4.93956178e-01 -1.05637789e+00 -4.01266031e-02 5.32215834e-01 -4.54664767e-01 -9.18262422e-01 -5.69071025e-02 3.26673150e-01 -5.14342487e-01 -4.36926574e-01 4.99038994e-01 -4.09224987e-01 2.73735840e-02 7.77789891e-01 1.23148799e-01 -9.27686840e-02 -3.48453522e-01 5.63499391e-01 4.67369676e-01 4.87094432e-01 -8.49121094e-01 9.18290854e-01 8.50051939e-01 -6.39964361e-03 -5.53081095e-01 -1.35688567e+00 -6.62681818e-01 -9.82847393e-01 -1.75977737e-01 7.12229609e-01 -7.72461176e-01 -6.20625436e-01 -4.48672846e-02 -5.18579781e-01 9.63420793e-02 -6.55172348e-01 6.80536389e-01 -3.04065377e-01 5.30372798e-01 -5.89703396e-02 -4.75228012e-01 3.02405488e-02 -9.71245050e-01 7.01907575e-01 2.03022942e-01 -2.11657986e-01 -1.15049469e+00 -1.93323284e-01 3.68070215e-01 -3.09289945e-03 2.79842615e-01 1.16015875e+00 -1.22351384e+00 -3.16932499e-01 -7.73291215e-02 -5.27625263e-01 4.63738978e-01 4.71078604e-01 -2.25741684e-01 -1.11485541e+00 -3.05614322e-01 1.50588900e-01 -3.57635915e-01 7.81501472e-01 1.53107479e-01 1.16244745e+00 -1.35983303e-01 -4.98921156e-01 3.53466749e-01 1.50687432e+00 2.27126747e-01 2.26717517e-01 2.65333712e-01 6.99474514e-01 8.04056525e-01 7.98541784e-01 1.57902315e-01 3.82653624e-01 4.11950231e-01 4.11710829e-01 -6.15669191e-02 -2.60138720e-01 -2.63960510e-01 -3.98444533e-02 7.81536460e-01 1.76516443e-01 -1.36241406e-01 -7.84324884e-01 5.89509606e-01 -1.74590552e+00 -8.23395729e-01 2.36798301e-02 2.26586938e+00 8.54314804e-01 1.27134457e-01 2.94375047e-02 2.26371095e-01 1.05671728e+00 -1.60498381e-01 -7.45859385e-01 1.62220985e-01 -3.84552926e-01 8.78307745e-02 2.57926971e-01 6.09482192e-02 -1.03090227e+00 8.77119899e-01 6.70209837e+00 1.01534677e+00 -1.10340142e+00 2.60460395e-02 6.39362276e-01 4.17751037e-02 -4.57540363e-01 2.88640618e-01 -9.12614763e-01 3.06427062e-01 6.13599718e-01 -5.49918652e-01 -5.93017414e-03 8.16655934e-01 -1.71368681e-02 6.77329898e-02 -1.60112190e+00 8.93746614e-01 4.58530821e-02 -1.18224740e+00 2.17275411e-01 1.78786978e-01 9.19282317e-01 -2.59987056e-01 8.52560624e-02 3.51699293e-01 4.08302128e-01 -9.47346032e-01 7.08076656e-01 2.24715859e-01 7.45588422e-01 -4.97354776e-01 6.96033716e-01 2.92317241e-01 -1.11594546e+00 -1.29411355e-01 -1.77578643e-01 -7.47107118e-02 -4.03697461e-01 7.65934646e-01 -8.39899242e-01 7.16483057e-01 6.55280292e-01 1.04134738e+00 -8.14925313e-01 8.35245192e-01 -3.13028723e-01 7.66234398e-01 1.13037657e-02 3.87053579e-01 3.43919784e-01 -1.48934558e-01 3.35873932e-01 8.36341918e-01 1.33535683e-01 2.37029120e-01 4.39183831e-01 9.17144418e-01 -3.02213818e-01 1.18609883e-01 -6.51841700e-01 -8.77918154e-02 7.55706489e-01 1.15606999e+00 -1.20109200e+00 -6.15066528e-01 -5.23118079e-01 4.03802365e-01 1.93853855e-01 3.88965487e-01 -7.60640144e-01 -8.34511444e-02 5.08441031e-01 4.40248214e-02 1.96549088e-01 1.68285072e-01 -5.37532508e-01 -1.29171288e+00 -1.30974382e-01 -6.58931434e-01 6.92959964e-01 -5.57157338e-01 -1.71509552e+00 4.16859716e-01 1.71660930e-01 -1.45659554e+00 7.26977587e-02 -4.45454717e-01 -1.86726704e-01 4.98867899e-01 -1.54326403e+00 -1.02602470e+00 -4.16517556e-01 6.41792953e-01 4.91818815e-01 -8.74196589e-02 7.69640684e-01 2.37533376e-01 -3.70247662e-01 6.94631815e-01 1.28187045e-01 2.59530336e-01 7.80878484e-01 -1.12354565e+00 -2.03292415e-01 9.03528512e-01 2.39169061e-01 7.52991378e-01 6.08020902e-01 -4.82665926e-01 -7.70069897e-01 -1.29266667e+00 5.92554569e-01 -4.12671626e-01 6.07459247e-01 -7.54587203e-02 -9.84326243e-01 7.41618335e-01 -1.65263638e-01 2.31996015e-01 9.69532609e-01 2.81218439e-01 -8.02288711e-01 3.87732908e-02 -1.40176225e+00 4.45483387e-01 1.34681606e+00 -5.21097064e-01 -8.17450285e-01 2.25537479e-01 9.48541939e-01 2.25272343e-01 -8.97053003e-01 7.73930788e-01 4.67593849e-01 -1.01249599e+00 9.43431079e-01 -7.37129390e-01 5.00787914e-01 -5.50178111e-01 -4.64211732e-01 -1.16456366e+00 -3.32803547e-01 1.68610349e-01 1.56451285e-01 1.24598765e+00 2.47883499e-01 -7.44026661e-01 8.40579569e-01 4.75911260e-01 -1.59763008e-01 -5.53954720e-01 -6.85030282e-01 -1.25380862e+00 1.82416379e-01 -3.13437253e-01 6.09353781e-01 1.68083155e+00 -1.82536423e-01 5.42036176e-01 1.99418906e-02 2.31714129e-01 6.58755362e-01 5.23369431e-01 5.64109921e-01 -1.53939474e+00 8.73182043e-02 -5.47575831e-01 -6.12538755e-01 -8.42907310e-01 4.58283335e-01 -1.11813235e+00 -1.92629658e-02 -1.07795596e+00 5.81113458e-01 -6.85404360e-01 -5.92587769e-01 6.59416437e-01 -3.33627760e-01 4.66826975e-01 1.48380890e-01 5.85539997e-01 -7.79578090e-01 5.29288292e-01 1.25869644e+00 -3.31623465e-01 1.77315678e-02 -2.22367242e-01 -1.17306340e+00 8.08596015e-01 6.10390186e-01 -4.79217917e-01 -7.08297372e-01 -5.49210533e-02 2.74226144e-02 -4.64007974e-01 4.14416015e-01 -7.77981877e-01 8.28746930e-02 -4.18387711e-01 3.21502462e-02 -3.09488744e-01 -8.73973966e-02 -9.65064108e-01 -7.69196253e-04 3.74827743e-01 -4.85534608e-01 -5.00822186e-01 -8.06343555e-02 7.55820215e-01 -4.12065178e-01 -3.29381824e-01 9.07109737e-01 -1.32622287e-01 -1.18298507e+00 3.68278742e-01 -1.32729616e-02 4.10311788e-01 1.21811473e+00 -5.53471565e-01 -3.66294891e-01 -8.44166502e-02 -8.00086379e-01 9.06803459e-02 5.77769816e-01 4.70400244e-01 4.18746173e-01 -1.46760273e+00 -5.52962840e-01 -8.19534287e-02 7.33628035e-01 -9.82078686e-02 4.43392470e-02 6.41267657e-01 -3.04566268e-02 3.19680750e-01 -9.71021280e-02 -6.88221574e-01 -1.25329745e+00 9.16776597e-01 1.81924671e-01 2.56744605e-02 -5.08241177e-01 7.27873683e-01 6.98644757e-01 -5.31028569e-01 1.38158677e-02 -7.86947235e-02 -3.47576886e-01 2.61418670e-01 2.77173400e-01 2.87300140e-01 -2.32977644e-02 -6.99048162e-01 -3.90130907e-01 4.15104955e-01 -1.88386172e-01 3.52109671e-01 1.06161857e+00 -3.65612507e-01 -5.57367541e-02 6.22359693e-01 1.31900167e+00 -1.77853152e-01 -9.22247648e-01 -6.49545372e-01 2.60449141e-01 -6.07755005e-01 -1.61085576e-01 -6.86189532e-01 -1.13523257e+00 6.24928415e-01 7.88603127e-01 2.32982054e-01 1.28999448e+00 2.83328652e-01 3.86584282e-01 3.40939909e-01 3.99314433e-01 -9.75704789e-01 2.34274849e-01 3.46056014e-01 4.80280876e-01 -1.60208619e+00 2.56830472e-02 -8.08774590e-01 -4.52260047e-01 9.11464810e-01 7.32911587e-01 2.02010542e-01 8.74112427e-01 -2.38871545e-01 5.68334870e-02 -4.74279821e-01 -3.69795710e-01 -5.64494967e-01 3.34399968e-01 6.66337073e-01 3.68910700e-01 2.09135294e-01 -4.12007183e-01 5.73896348e-01 -2.31145799e-01 -1.04522899e-01 4.37453806e-01 8.36773455e-01 -5.49145401e-01 -9.02794003e-01 -2.33755186e-01 5.52188277e-01 -1.59999877e-01 -1.16332315e-01 -4.38217103e-01 8.62233937e-01 4.22906309e-01 1.11416101e+00 4.28568088e-02 -5.87125957e-01 1.51164308e-01 -5.06312065e-02 3.62637669e-01 -1.02980077e+00 -3.38825971e-01 -3.06643575e-01 -2.81724423e-01 -2.94939369e-01 -8.48770738e-01 -3.75948250e-01 -1.18856728e+00 -2.07120851e-01 -5.04646540e-01 3.23692381e-01 3.64140451e-01 1.20174956e+00 3.63784730e-01 3.21278334e-01 8.01498175e-01 -1.72422007e-01 -4.41548944e-01 -8.13998520e-01 -8.76923859e-01 8.49320531e-01 -2.97885668e-02 -9.62394476e-01 -6.39043391e-01 2.41671383e-01]
[9.656194686889648, 2.9676802158355713]
303e9950-43f0-4e33-b474-457478f2861d
enhancing-general-face-forgery-detection-via
2303.00917
null
https://arxiv.org/abs/2303.00917v2
https://arxiv.org/pdf/2303.00917v2.pdf
Enhancing General Face Forgery Detection via Vision Transformer with Low-Rank Adaptation
Nowadays, forgery faces pose pressing security concerns over fake news, fraud, impersonation, etc. Despite the demonstrated success in intra-domain face forgery detection, existing detection methods lack generalization capability and tend to suffer from dramatic performance drops when deployed to unforeseen domains. To mitigate this issue, this paper designs a more general fake face detection model based on the vision transformer(ViT) architecture. In the training phase, the pretrained ViT weights are freezed, and only the Low-Rank Adaptation(LoRA) modules are updated. Additionally, the Single Center Loss(SCL) is applied to supervise the training process, further improving the generalization capability of the model. The proposed method achieves state-of-the-arts detection performances in both cross-manipulation and cross-dataset evaluations.
['Shiqi Wang', 'Haoliang Li', 'Chenqi Kong']
2023-03-02
null
null
null
null
['face-detection']
['computer-vision']
[ 2.31488109e-01 -2.37540424e-01 -1.49628073e-01 -5.31322777e-01 -4.38251644e-01 -2.03482106e-01 7.66553760e-01 -1.31798968e-01 -3.23513359e-01 4.86978412e-01 -1.08033895e-01 -6.11710958e-02 6.69681802e-02 -5.46624959e-01 -4.26662683e-01 -6.23057425e-01 1.37213498e-01 -9.97485872e-03 1.34323463e-01 -2.15254903e-01 2.15886101e-01 3.89370203e-01 -1.37455928e+00 4.32028383e-01 9.36384439e-01 1.11345315e+00 -3.64787936e-01 -3.36689106e-03 1.29028693e-01 6.46455884e-01 -6.99438274e-01 -1.05842328e+00 1.75355896e-01 -2.41844818e-01 -3.33434135e-01 1.98122799e-01 6.89915478e-01 -4.60036129e-01 -7.01211572e-01 1.40287614e+00 5.00914693e-01 -1.92441180e-01 2.69893438e-01 -1.57565486e+00 -7.06502259e-01 7.38109946e-02 -8.46430004e-01 2.73814738e-01 1.92257226e-01 1.30951956e-01 5.75938761e-01 -1.33140171e+00 3.97485584e-01 1.39745653e+00 6.49817765e-01 8.49056542e-01 -1.06050134e+00 -1.09007645e+00 1.71329781e-01 3.37093949e-01 -1.32085025e+00 -7.09722459e-01 9.25826490e-01 -1.24917641e-01 4.15964246e-01 8.89417678e-02 3.30798090e-01 1.47727108e+00 1.17089532e-01 7.21549928e-01 1.03516448e+00 -1.40912414e-01 -1.31432265e-01 4.61899966e-01 -1.35544538e-01 7.67677009e-01 4.92212236e-01 2.61737257e-01 -8.22282910e-01 -3.18272710e-01 5.32071590e-01 1.06787421e-01 -2.40687549e-01 -2.02126786e-01 -4.15979415e-01 9.05619144e-01 4.82081175e-01 2.99086660e-01 -2.13024467e-01 -7.85326585e-02 6.11976087e-01 4.06183183e-01 6.27539694e-01 9.26039368e-02 1.96384471e-02 3.18015188e-01 -1.11582744e+00 1.26027837e-01 3.34566325e-01 5.48443079e-01 4.77330297e-01 1.38487846e-01 -1.84432983e-01 1.06434155e+00 2.77267098e-01 1.99786499e-01 4.60765600e-01 -3.88910323e-01 6.18621171e-01 5.17330647e-01 -5.82546368e-03 -1.46253765e+00 -1.37952715e-01 -8.19254339e-01 -8.28882694e-01 -8.76681283e-02 1.90435529e-01 1.70044497e-01 -8.89565170e-01 1.55081880e+00 4.91568089e-01 4.78430152e-01 -1.73905298e-01 1.07615328e+00 5.11764705e-01 4.09383833e-01 2.82833666e-01 -2.46584326e-01 1.30596256e+00 -8.62847030e-01 -7.59842813e-01 -5.38254261e-01 3.25861722e-01 -8.65685880e-01 9.05633390e-01 6.09935939e-01 -7.00007975e-01 -3.68427455e-01 -1.00333190e+00 4.81256023e-02 -2.03693986e-01 4.18755352e-01 7.29624391e-01 1.03836882e+00 -6.53731704e-01 4.51369882e-01 -5.16370356e-01 -1.80467233e-01 9.95761573e-01 2.15656281e-01 -4.99286890e-01 -5.31942904e-01 -1.22413063e+00 8.00876319e-01 1.03285387e-01 4.77806687e-01 -1.11982524e+00 -6.11867189e-01 -4.71099585e-01 -3.68403085e-02 3.32762718e-01 -3.18033755e-01 8.12578857e-01 -1.19117618e+00 -1.26419532e+00 1.03796363e+00 1.30288348e-01 -4.15144771e-01 8.00764680e-01 -3.58029127e-01 -8.84351909e-01 1.64634749e-01 -1.30851297e-02 3.87668461e-01 1.68162668e+00 -1.23152053e+00 -4.68630791e-01 -6.75014496e-01 -1.10486388e-01 -4.55932431e-02 -8.68043661e-01 2.40466610e-01 -4.89568770e-01 -9.79476213e-01 1.66348755e-01 -7.19920635e-01 2.42424533e-01 5.49168363e-02 -1.84994414e-01 -9.99983102e-02 1.46459889e+00 -8.01740646e-01 1.34181094e+00 -2.45653367e+00 -1.22737318e-01 1.24470823e-01 1.55087069e-01 7.75730848e-01 -1.81447744e-01 6.43824413e-02 -1.21915154e-01 -2.84049630e-01 -2.17206359e-01 -6.37132049e-01 -2.87911415e-01 1.75128020e-02 -4.07872856e-01 8.72051954e-01 4.82553333e-01 5.01874506e-01 -8.00816536e-01 -3.50446492e-01 6.66793138e-02 6.05203807e-01 -6.11214459e-01 8.14992376e-03 5.83692342e-02 4.20301139e-01 -4.22611415e-01 8.25320959e-01 1.23786652e+00 -8.93926993e-02 3.06785643e-01 -2.08382338e-01 4.22718227e-01 1.46554768e-01 -1.07019043e+00 1.22478545e+00 -3.50849181e-01 4.30047542e-01 3.57612163e-01 -1.09042943e+00 7.78661132e-01 1.73403263e-01 2.56496161e-01 -5.96672595e-01 2.37880751e-01 2.86341280e-01 -1.86713994e-01 -3.27780932e-01 4.09887016e-01 -1.91899493e-01 1.91820696e-01 1.03408240e-01 7.37634674e-02 4.63020325e-01 -3.12973350e-01 1.89159825e-01 8.82832885e-01 -1.01349384e-01 -2.31637120e-01 -5.64814322e-02 7.96556294e-01 -3.40694696e-01 6.80480838e-01 2.58915037e-01 -3.93315315e-01 2.81669170e-01 5.09724855e-01 -4.07206714e-01 -7.52116263e-01 -8.51925313e-01 -2.60895103e-01 9.97153759e-01 2.73055792e-01 -2.89788067e-01 -6.89494312e-01 -1.18509007e+00 3.07896197e-01 3.86897892e-01 -4.82932001e-01 -6.57035828e-01 -5.00517309e-01 -9.67857659e-01 9.02784586e-01 9.92956981e-02 8.53315711e-01 -6.87952697e-01 -1.97253898e-01 1.87183306e-01 -1.01334609e-01 -1.39217281e+00 -4.02508736e-01 -5.26587427e-01 -8.71778190e-01 -1.10557520e+00 -5.51699460e-01 -7.22471595e-01 7.99467862e-01 5.18510103e-01 5.47865987e-01 3.78439903e-01 -3.57632011e-01 -7.19501674e-02 -2.36230463e-01 -1.07765108e-01 -3.16689491e-01 -1.98459372e-01 2.74267644e-01 6.39400244e-01 4.01564091e-01 -3.84399742e-01 -6.61053598e-01 3.92540723e-01 -8.91640604e-01 -5.05085707e-01 5.26963174e-01 1.27906609e+00 5.70621714e-02 2.30473414e-01 7.81963646e-01 -9.22683358e-01 6.35488391e-01 -3.69155139e-01 -6.97298050e-01 1.46532491e-01 -8.93189967e-01 -3.58067065e-01 6.71175182e-01 -6.16426051e-01 -1.28708625e+00 -2.24368826e-01 1.30536899e-01 -8.02650928e-01 1.94879070e-01 8.55118632e-02 -2.64981151e-01 -4.56292808e-01 5.23750842e-01 2.42869526e-01 -8.42951983e-02 -4.97397453e-01 5.11166789e-02 6.90689087e-01 4.37602937e-01 -1.25986025e-01 1.18636918e+00 6.33928955e-01 -4.38507609e-02 -8.33203316e-01 -7.56524444e-01 -3.30941111e-01 -1.61605515e-02 -3.06210339e-01 2.31847435e-01 -1.10479236e+00 -5.59760213e-01 8.75611663e-01 -1.08408654e+00 3.72344524e-01 3.56222898e-01 3.28054309e-01 9.67917144e-02 7.62363255e-01 -6.93230927e-01 -9.92673397e-01 -2.66787142e-01 -9.44414079e-01 9.56205726e-01 1.21424124e-01 3.58053356e-01 -6.06406510e-01 -4.22235876e-01 8.11419010e-01 4.57594573e-01 5.21828309e-02 5.65012515e-01 -3.07844460e-01 -3.82206231e-01 -6.00326598e-01 -5.45470297e-01 7.69493461e-01 6.74068704e-02 -2.20186010e-01 -1.34920037e+00 -7.30044663e-01 2.04240873e-01 -3.44973862e-01 1.13992989e+00 -2.08531410e-01 1.20406103e+00 -5.26320755e-01 -4.97334719e-01 5.21814048e-01 1.17738080e+00 -1.61934525e-01 7.42161274e-01 7.20742792e-02 7.15462863e-01 7.49848545e-01 6.28549755e-01 5.50086141e-01 2.12358505e-01 7.85431027e-01 6.58156693e-01 1.04214586e-01 -1.23174749e-01 -3.66793990e-01 6.03011191e-01 2.85321236e-01 3.05908859e-01 -1.72985151e-01 -6.12734616e-01 4.36731189e-01 -1.56874728e+00 -1.02677417e+00 -1.39973685e-03 2.25818825e+00 5.52646160e-01 2.61757582e-01 4.89135087e-02 2.05507100e-01 8.89163077e-01 3.38225394e-01 -5.04296482e-01 -1.98054969e-01 -2.44408533e-01 8.99775773e-02 4.97145921e-01 3.29752952e-01 -1.37713039e+00 1.32598126e+00 5.60953760e+00 1.17938745e+00 -1.39800453e+00 5.35285830e-01 6.53946221e-01 -8.49802718e-02 4.67690974e-02 4.49795334e-04 -6.67181611e-01 6.61966622e-01 5.56853533e-01 3.99204612e-01 4.82004374e-01 9.16207373e-01 1.58269674e-01 2.57199928e-02 -5.33763766e-01 1.10096192e+00 3.49768132e-01 -1.00944126e+00 1.43004969e-01 9.49866995e-02 5.00664711e-01 -2.66199470e-01 5.16192734e-01 2.62704194e-01 -2.41240129e-01 -9.41039503e-01 4.91780072e-01 1.90716654e-01 9.34700847e-01 -9.32373345e-01 6.43284738e-01 2.00479090e-01 -7.48745561e-01 -4.69494015e-01 -5.89887977e-01 1.43637434e-01 -1.01709731e-01 8.82569551e-01 -6.65382147e-01 3.82400364e-01 5.65704226e-01 6.05330288e-01 -5.80738425e-01 8.37273479e-01 -2.71281302e-01 6.39726222e-01 -2.06771925e-01 3.99152517e-01 2.04992086e-01 -1.08686976e-01 6.51568890e-01 9.49088037e-01 1.08977832e-01 -2.40744829e-01 -1.75902247e-02 6.40505612e-01 -4.07683045e-01 -1.23602994e-01 -4.31216359e-01 -7.93311074e-02 3.95723820e-01 1.26358676e+00 -3.12401175e-01 -1.14223741e-01 -3.20967436e-01 1.24321663e+00 2.82516569e-01 2.50368297e-01 -9.87373352e-01 -2.39871249e-01 8.37826848e-01 3.12821865e-01 3.21115464e-01 2.09671017e-02 -9.36727375e-02 -1.43766212e+00 2.95231104e-01 -1.07994843e+00 4.71107841e-01 -1.64818734e-01 -1.57024610e+00 4.69734848e-01 -4.02529120e-01 -1.25941133e+00 2.40341648e-01 -5.61000705e-01 -4.64652926e-01 5.39458752e-01 -1.80900025e+00 -1.27229583e+00 -2.59521157e-01 6.67698681e-01 3.21685433e-01 -4.46576327e-01 4.53947514e-01 9.23416495e-01 -8.86186898e-01 1.16367149e+00 -1.52291385e-02 2.33827546e-01 8.90316308e-01 -4.88198996e-01 1.70490384e-01 9.20149386e-01 -9.33424234e-02 5.74735641e-01 6.10087872e-01 -6.63608611e-01 -1.35527825e+00 -1.23770678e+00 7.21785367e-01 -1.71019912e-01 5.40707767e-01 -4.83912557e-01 -9.71090019e-01 3.21031779e-01 -3.49176705e-01 4.79227602e-01 3.48611206e-01 -1.60880104e-01 -9.44609880e-01 -3.96609575e-01 -1.60753548e+00 2.88393259e-01 9.40101147e-01 -8.50929558e-01 -2.63429463e-01 4.78745759e-01 2.14038044e-01 -1.94139272e-01 -4.89978194e-01 3.26265335e-01 7.57156312e-01 -9.67017770e-01 8.89085531e-01 -6.78332210e-01 2.11265087e-01 -3.06643575e-01 6.54536784e-02 -9.07383561e-01 -2.17873588e-01 -6.47247136e-01 -3.82635832e-01 1.24004245e+00 1.06784783e-01 -6.98634207e-01 1.00553942e+00 8.02784860e-02 2.24617735e-01 -5.97772002e-01 -1.30174828e+00 -8.46471488e-01 -2.27914050e-01 -1.46225989e-01 4.60962921e-01 1.24410534e+00 -1.23423152e-01 3.39982837e-01 -7.93916106e-01 5.15501857e-01 1.02136207e+00 -2.49558374e-01 5.75075388e-01 -1.15500259e+00 -1.62116244e-01 -1.63848802e-01 -5.42819321e-01 -8.89693737e-01 1.46670327e-01 -6.48597002e-01 -3.26796532e-01 -6.82117641e-01 8.45023915e-02 -4.34940398e-01 -3.22848767e-01 3.76673281e-01 -2.63314962e-01 6.22971117e-01 2.27586865e-01 2.72738725e-01 -4.57622349e-01 7.51457095e-01 1.11163986e+00 -2.42478013e-01 2.00293168e-01 1.18856221e-01 -4.99130249e-01 6.63347423e-01 7.72343397e-01 -7.53264129e-01 -3.24888617e-01 -4.60830420e-01 -1.05496213e-01 -7.79763311e-02 6.50558650e-01 -9.90814924e-01 1.95525661e-01 6.76110294e-03 4.54382181e-01 -2.20549479e-01 5.48856795e-01 -7.32703269e-01 -2.44337752e-01 5.97492635e-01 8.37876499e-02 -1.39262751e-01 1.68867186e-01 8.34290445e-01 -2.69984603e-01 4.31437306e-02 1.22107995e+00 1.44717693e-01 -3.23074907e-01 4.21220243e-01 1.71659533e-02 -1.29119992e-01 9.29080486e-01 -3.59350950e-01 -4.23215300e-01 -1.90705270e-01 -3.63099426e-01 6.26164898e-02 4.12826270e-01 6.16195142e-01 8.97902906e-01 -1.29414833e+00 -6.92594409e-01 4.98629600e-01 2.76391625e-01 -4.78792667e-01 4.25948650e-01 8.02248836e-01 -1.77640006e-01 1.15224101e-01 -1.25256062e-01 -3.16265106e-01 -1.21579623e+00 6.10262513e-01 3.78771871e-01 -2.99247503e-01 -3.71027946e-01 1.02294981e+00 1.50378337e-02 -1.80128306e-01 2.47092724e-01 4.39100385e-01 -8.26599076e-02 2.12004095e-01 5.72742939e-01 5.07293046e-01 2.12291628e-01 -9.51471806e-01 -6.38946831e-01 1.77460402e-01 -4.16206628e-01 2.96692699e-01 1.14213228e+00 -4.43300046e-02 -2.34876335e-01 -2.45385185e-01 1.17799127e+00 -3.32746655e-03 -1.16347003e+00 -3.25519234e-01 1.01500832e-01 -8.82690072e-01 3.70914936e-01 -6.60472453e-01 -1.36098528e+00 7.79488981e-01 8.97889018e-01 -9.03703198e-02 1.20044291e+00 -3.95974398e-01 9.85632479e-01 9.84383225e-02 4.87973601e-01 -1.14271641e+00 3.87852758e-01 2.89608359e-01 7.35281646e-01 -1.28027868e+00 1.56924501e-01 -6.07693255e-01 -5.28861105e-01 7.96494842e-01 7.16997862e-01 -2.06291154e-01 6.18350208e-01 -1.28791943e-01 -1.62080213e-01 -2.05870852e-01 -5.05208910e-01 3.60222280e-01 1.17638804e-01 4.92161483e-01 5.33648841e-02 -4.72020097e-02 -4.17602122e-01 3.97834033e-01 2.23310143e-01 1.90350004e-02 3.19794118e-01 7.62450159e-01 -2.03770459e-01 -1.13344514e+00 -5.05047977e-01 4.64708358e-01 -6.51859879e-01 6.57056319e-03 -4.07629371e-01 4.51858699e-01 3.69329572e-01 1.03446102e+00 -3.09896559e-01 -4.85741764e-01 2.44256809e-01 -1.27970248e-01 4.23436582e-01 -4.00767416e-01 -7.27318525e-01 -1.05040573e-01 -9.54907164e-02 -7.39254236e-01 -2.07710654e-01 -4.88646507e-01 -6.32249832e-01 -5.24167717e-01 -7.96905816e-01 1.82161108e-01 4.83884454e-01 7.45955527e-01 5.00228167e-01 1.62915200e-01 9.49455559e-01 -4.66728896e-01 -9.55991149e-01 -1.00527370e+00 -5.74120939e-01 4.49845374e-01 5.08272350e-01 -9.39813435e-01 -4.91558760e-01 -3.08537871e-01]
[12.75704288482666, 1.0083798170089722]
3a29cdcc-0151-4434-8ac4-7c00ddfb81b6
noisy-boundaries-lemon-or-lemonade-for-semi
2203.13427
null
https://arxiv.org/abs/2203.13427v1
https://arxiv.org/pdf/2203.13427v1.pdf
Noisy Boundaries: Lemon or Lemonade for Semi-supervised Instance Segmentation?
Current instance segmentation methods rely heavily on pixel-level annotated images. The huge cost to obtain such fully-annotated images restricts the dataset scale and limits the performance. In this paper, we formally address semi-supervised instance segmentation, where unlabeled images are employed to boost the performance. We construct a framework for semi-supervised instance segmentation by assigning pixel-level pseudo labels. Under this framework, we point out that noisy boundaries associated with pseudo labels are double-edged. We propose to exploit and resist them in a unified manner simultaneously: 1) To combat the negative effects of noisy boundaries, we propose a noise-tolerant mask head by leveraging low-resolution features. 2) To enhance the positive impacts, we introduce a boundary-preserving map for learning detailed information within boundary-relevant regions. We evaluate our approach by extensive experiments. It behaves extraordinarily, outperforming the supervised baseline by a large margin, more than 6% on Cityscapes, 7% on COCO and 4.5% on BDD100k. On Cityscapes, our method achieves comparable performance by utilizing only 30% labeled images.
['Shengjin Wang', 'YaLi Li', 'Zhenyu Wang']
2022-03-25
null
http://openaccess.thecvf.com//content/CVPR2022/html/Wang_Noisy_Boundaries_Lemon_or_Lemonade_for_Semi-Supervised_Instance_Segmentation_CVPR_2022_paper.html
http://openaccess.thecvf.com//content/CVPR2022/papers/Wang_Noisy_Boundaries_Lemon_or_Lemonade_for_Semi-Supervised_Instance_Segmentation_CVPR_2022_paper.pdf
cvpr-2022-1
['semi-supervised-instance-segmentation']
['computer-vision']
[ 4.72653508e-01 5.12704313e-01 -3.50680977e-01 -4.43855345e-01 -1.21658826e+00 -7.51948059e-01 4.37650681e-01 -4.99664433e-02 -6.94173753e-01 5.90572000e-01 -1.21181887e-02 -1.49532303e-01 2.62763292e-01 -5.18774152e-01 -8.24595869e-01 -6.70220256e-01 1.31826937e-01 2.82554775e-01 6.55157268e-01 1.32148668e-01 1.20615751e-01 2.11405084e-01 -1.29770279e+00 2.55114347e-01 1.18276370e+00 9.07749116e-01 1.55195057e-01 3.42803210e-01 -7.22671524e-02 5.04800260e-01 -4.96575147e-01 -3.01154673e-01 5.48844099e-01 -2.68165946e-01 -1.06831634e+00 7.53365695e-01 5.76286137e-01 -2.36324996e-01 -6.85738027e-02 1.12281656e+00 2.05671296e-01 -3.66928279e-02 6.76718295e-01 -9.34995532e-01 -3.81529003e-01 5.98944664e-01 -1.05016685e+00 1.16008550e-01 -6.53453097e-02 2.78785229e-01 1.01455295e+00 -8.39900255e-01 7.55037546e-01 9.22423959e-01 6.29997432e-01 5.22421658e-01 -1.42699552e+00 -3.18424791e-01 5.15611708e-01 -2.89850324e-01 -1.45384586e+00 -3.54749650e-01 7.80332029e-01 -3.13444376e-01 5.45080364e-01 1.71114817e-01 2.82815456e-01 8.76257420e-01 -3.26916814e-01 1.15874577e+00 1.39187407e+00 -4.72952425e-01 2.11918876e-01 1.40625149e-01 1.20341323e-01 7.44067609e-01 1.71879530e-01 -8.94663706e-02 -1.37709469e-01 1.04099914e-01 7.12000966e-01 -1.37624964e-01 -2.88493007e-01 -4.52324480e-01 -1.15702868e+00 5.99798977e-01 5.30799448e-01 2.62126774e-01 -9.35829356e-02 -1.28401965e-01 3.73991847e-01 -3.33203673e-02 7.94780433e-01 3.36236000e-01 -6.87736869e-01 1.78158626e-01 -1.28492153e+00 -6.69819787e-02 5.13361275e-01 1.10033667e+00 9.97170091e-01 -2.54947484e-01 -2.38189355e-01 1.02312732e+00 2.33611297e-02 2.87944764e-01 3.03283215e-01 -1.20610821e+00 5.88634789e-01 6.66919589e-01 2.12671325e-01 -6.02495492e-01 -3.55779022e-01 -4.73390698e-01 -7.84865618e-01 2.32570749e-02 6.85561955e-01 -2.07913905e-01 -1.51316535e+00 1.48336935e+00 4.57717896e-01 3.38006884e-01 -1.05551802e-01 9.19771314e-01 6.05394423e-01 4.55597162e-01 1.24291040e-01 -9.85830203e-02 1.19041717e+00 -1.49793148e+00 -5.63353002e-01 -4.33786154e-01 7.76414514e-01 -5.56907237e-01 1.38177824e+00 3.79365772e-01 -9.94243205e-01 -4.65140998e-01 -8.77833545e-01 -2.68026590e-02 -2.39561409e-01 2.78516322e-01 3.75425249e-01 6.59847975e-01 -1.00094855e+00 5.34707069e-01 -1.04757071e+00 -7.62454271e-02 6.52405202e-01 3.09990168e-01 -3.75391841e-01 -1.74040243e-01 -7.90922344e-01 3.71432275e-01 5.58121324e-01 4.23627049e-02 -6.28768921e-01 -5.70207059e-01 -1.07074523e+00 -1.79926127e-01 7.35155821e-01 -2.25376755e-01 1.08218849e+00 -1.09704363e+00 -1.31557250e+00 1.19287634e+00 -6.63478076e-02 -4.83937144e-01 9.09246564e-01 -2.65494168e-01 -1.12311542e-01 4.03225332e-01 4.02194053e-01 1.12493777e+00 6.77186847e-01 -1.66276312e+00 -8.34327757e-01 -3.39994133e-01 -7.85928816e-02 1.72500595e-01 -2.34742999e-01 -2.80774176e-01 -1.25278723e+00 -8.34584594e-01 3.55181217e-01 -1.09181404e+00 -6.99942946e-01 -1.72617391e-01 -8.82997751e-01 5.78018129e-02 8.31036091e-01 -4.41914350e-01 9.73102272e-01 -2.18264008e+00 -2.96406269e-01 1.99583650e-01 1.68309972e-01 4.07805115e-01 -2.14297935e-01 -1.98734194e-01 2.83664852e-01 3.84788692e-01 -9.19307470e-01 -6.47555470e-01 -1.84777185e-01 4.17321235e-01 -9.93575826e-02 4.54056621e-01 4.61438835e-01 1.00499737e+00 -8.26669276e-01 -7.98505902e-01 2.72462606e-01 3.14351857e-01 -4.81497288e-01 -7.44585022e-02 -2.83684611e-01 5.13134420e-01 -5.60110688e-01 7.32173860e-01 9.25492585e-01 -3.31335723e-01 5.54117002e-02 -2.53160112e-02 -5.02390675e-02 1.28947999e-02 -1.28015780e+00 1.79992104e+00 -2.42836371e-01 2.94937998e-01 3.21736276e-01 -1.00255144e+00 7.91495979e-01 -7.35048251e-03 3.89966667e-01 -6.84965968e-01 -7.55625069e-02 2.20740929e-01 -4.37252492e-01 -2.87290871e-01 3.84295225e-01 1.34199440e-01 -1.34003639e-01 2.34933391e-01 -1.62813425e-01 -2.49280617e-01 3.27632219e-01 2.06289753e-01 1.03804743e+00 3.46089780e-01 6.65140226e-02 -4.26601738e-01 4.50739354e-01 1.31908521e-01 8.50975394e-01 8.23737323e-01 -5.98281801e-01 1.18388546e+00 4.55025047e-01 -2.70081013e-01 -8.62410367e-01 -9.70520616e-01 -3.24514985e-01 8.55831802e-01 4.52810526e-01 -2.04873532e-01 -1.14320028e+00 -1.23678041e+00 -1.89577281e-01 3.97426188e-01 -7.02402771e-01 1.72485366e-01 -6.70016944e-01 -9.58351731e-01 5.56199789e-01 6.87563598e-01 8.31784070e-01 -8.74145687e-01 -3.61513525e-01 1.32435665e-01 -2.83451557e-01 -1.55454421e+00 -6.48759365e-01 3.91714275e-01 -9.82756615e-01 -8.71179283e-01 -1.04519248e+00 -9.14315104e-01 8.67440104e-01 4.02733624e-01 1.20091379e+00 5.72249033e-02 -2.56798327e-01 5.49496897e-02 -3.60112220e-01 1.45423552e-02 -7.93097541e-02 3.75528187e-01 -4.03682292e-01 5.46483137e-02 2.20158733e-02 -3.06822598e-01 -7.65480101e-01 5.85158050e-01 -9.92575943e-01 7.81652108e-02 7.39900708e-01 8.00465345e-01 1.14983261e+00 6.27006665e-02 5.24959147e-01 -1.44877505e+00 -4.17532399e-02 -1.98453069e-01 -5.96383631e-01 1.77085578e-01 -5.27332962e-01 2.75811739e-03 5.15031099e-01 -3.57797563e-01 -1.18125153e+00 6.28842771e-01 -1.46783367e-01 -2.02170119e-01 -5.85791409e-01 -4.88327071e-02 -3.62058997e-01 -5.42185456e-02 5.39715290e-01 1.01650842e-01 -4.36236948e-01 -5.96581399e-01 6.43942535e-01 7.33509600e-01 7.08173096e-01 -5.75701296e-01 7.42058039e-01 9.06037927e-01 -4.43919778e-01 -7.70581484e-01 -1.19773102e+00 -6.37441278e-01 -8.54020596e-01 1.32037103e-01 9.11464632e-01 -1.07816958e+00 -2.86068972e-02 3.90859872e-01 -7.19780147e-01 -7.75151849e-01 -3.43353897e-01 1.84722424e-01 -4.91793424e-01 4.69901860e-01 -8.25238347e-01 -5.80468953e-01 -1.59985855e-01 -1.26158106e+00 1.37781465e+00 2.36108869e-01 1.73507571e-01 -7.86423147e-01 -1.23815596e-01 7.07873344e-01 3.70139666e-02 3.69788915e-01 4.19558257e-01 -6.36280656e-01 -5.48870802e-01 -6.24150448e-02 -5.90034783e-01 4.91618127e-01 1.38239622e-01 -4.36852798e-02 -1.11719906e+00 -4.85981628e-02 -3.07811409e-01 -4.21264976e-01 1.32493103e+00 3.79987597e-01 1.46130884e+00 3.07792495e-03 -4.01807547e-01 6.22369051e-01 1.33450651e+00 -2.59191215e-01 4.49423462e-01 3.03869784e-01 8.88706386e-01 7.88640559e-01 8.17953646e-01 1.67077333e-01 4.18620706e-01 5.18857598e-01 3.20622385e-01 -6.34917915e-01 -3.34128350e-01 -1.95731238e-01 9.56030488e-02 3.73032898e-01 1.16631076e-01 -2.71254092e-01 -9.49710369e-01 8.40560257e-01 -1.77915752e+00 -3.41181725e-01 -2.73042053e-01 1.96950376e+00 9.76598263e-01 5.91923475e-01 2.53465474e-01 1.62886336e-01 8.51345658e-01 2.45180041e-01 -6.65835619e-01 1.06493175e-01 -2.54963338e-01 4.42101061e-02 7.98392296e-01 5.87853909e-01 -1.57345617e+00 1.41104579e+00 6.17417002e+00 1.05269778e+00 -9.01633859e-01 7.76381046e-02 1.33362138e+00 5.40050417e-02 -1.45588666e-01 -1.74152553e-01 -7.29914308e-01 4.48487014e-01 4.72220331e-01 5.60175717e-01 -7.86694065e-02 8.80915701e-01 2.61817306e-01 -2.96247005e-01 -7.24844933e-01 7.19711363e-01 -7.84570351e-02 -1.10934591e+00 -1.88592926e-01 7.73669183e-02 1.30803359e+00 2.20798045e-01 3.20517384e-02 1.91856682e-01 4.38887000e-01 -8.64952266e-01 7.10323811e-01 -1.15931980e-01 7.95675933e-01 -7.31558561e-01 6.31488681e-01 3.94011617e-01 -1.14577413e+00 2.13074952e-01 -1.61517188e-01 1.69786796e-01 1.58567727e-01 9.89937067e-01 -4.87309664e-01 5.66537976e-01 7.43553519e-01 5.71698010e-01 -7.13759661e-01 9.70785320e-01 -5.34185708e-01 9.05814409e-01 -5.08898675e-01 6.23726726e-01 6.80982172e-01 -2.86747217e-01 1.05104528e-01 1.42404723e+00 -1.80327415e-01 3.93878967e-02 5.64161658e-01 8.28199029e-01 -2.46453434e-01 1.22660227e-01 -4.29310322e-01 3.23472410e-01 2.32503146e-01 1.37506819e+00 -1.44771373e+00 -4.99095768e-01 -4.54003215e-01 1.31648231e+00 3.63871157e-01 5.43248177e-01 -8.34554911e-01 -2.16180965e-01 2.10173875e-01 1.05753519e-01 4.39209670e-01 -8.57125819e-02 -6.40053093e-01 -1.15117872e+00 2.64984459e-01 -5.42615116e-01 3.97236973e-01 -2.22077921e-01 -1.08298349e+00 6.34879887e-01 -1.07272819e-01 -9.69050884e-01 1.22358344e-01 -4.11794156e-01 -5.20347536e-01 4.52797890e-01 -1.95692098e+00 -1.13401818e+00 -2.86941111e-01 3.63994569e-01 7.16907978e-01 5.76248288e-01 3.36818933e-01 4.33596581e-01 -9.05835986e-01 6.45169973e-01 3.09721194e-02 4.14527565e-01 7.39506483e-01 -1.59192455e+00 5.72035074e-01 9.92788970e-01 4.03605044e-01 2.71766365e-01 4.33727682e-01 -5.28108060e-01 -6.59211695e-01 -1.40820253e+00 6.45654380e-01 -3.52432758e-01 3.77972156e-01 -4.78817105e-01 -1.07141030e+00 6.06522739e-01 4.00268473e-03 6.34244263e-01 4.10972238e-01 -1.52752176e-01 -3.72818768e-01 9.52395797e-02 -1.26263189e+00 6.49397433e-01 1.22147930e+00 -2.60838002e-01 -4.49274957e-01 4.15269703e-01 8.97466242e-01 -3.76647472e-01 -6.32256210e-01 6.02688968e-01 7.55858496e-02 -1.04076815e+00 9.08435702e-01 -2.24499404e-01 3.57052654e-01 -5.32184422e-01 1.06334254e-01 -9.39413488e-01 -4.72016283e-04 -6.51066720e-01 2.05205068e-01 1.41603601e+00 6.74981415e-01 -4.34009314e-01 1.24894178e+00 6.02216005e-01 -1.60362989e-01 -7.96664178e-01 -7.94228554e-01 -8.82138848e-01 2.21869543e-01 -4.25792336e-01 2.58827537e-01 9.07779515e-01 -3.03248286e-01 1.88607767e-01 -1.87541947e-01 2.42391765e-01 6.34570420e-01 2.03245565e-01 6.47776186e-01 -8.39545012e-01 -2.49479279e-01 -3.51423770e-01 -8.66857097e-02 -1.41493499e+00 1.84636846e-01 -6.60268724e-01 3.64093661e-01 -1.28696704e+00 2.65724003e-01 -7.16790557e-01 -3.55966657e-01 5.89254677e-01 -5.26647568e-01 9.31329012e-01 6.23314492e-02 2.60593325e-01 -1.05842650e+00 3.41141075e-01 1.32472372e+00 -1.16077796e-01 -3.62993181e-01 -1.17882304e-01 -6.40064955e-01 1.00389636e+00 8.62007618e-01 -4.25863057e-01 -3.40940028e-01 -5.07159650e-01 -3.24668586e-01 -2.88595110e-01 2.11688951e-01 -8.73629153e-01 -1.97357442e-02 -1.76966731e-02 2.23706499e-01 -5.86337805e-01 9.24386755e-02 -6.79912567e-01 -4.00928885e-01 3.09044093e-01 -4.02369201e-01 -4.21596229e-01 7.57006705e-02 7.08165526e-01 -3.35268855e-01 -1.56267866e-01 1.04085922e+00 -1.22989304e-01 -8.34332228e-01 3.27670246e-01 -1.74817532e-01 4.87503409e-01 1.16532838e+00 -2.09907755e-01 -7.36275837e-02 -5.60213663e-02 -7.35566795e-01 4.98872519e-01 7.11741149e-01 7.16202110e-02 2.59722769e-01 -8.14401984e-01 -4.57130998e-01 1.47614360e-01 3.29937100e-01 5.55582523e-01 2.24410430e-01 7.59997189e-01 -4.97711867e-01 1.55553371e-01 2.62249619e-01 -7.77822733e-01 -1.00598073e+00 5.11706471e-01 1.76989853e-01 -4.15781379e-01 -8.94016206e-01 8.78373861e-01 5.25114417e-01 -5.59038162e-01 3.30438882e-01 -5.26951551e-01 -3.83685790e-02 -6.73079714e-02 4.21736121e-01 1.08342089e-01 5.06745614e-02 -6.74640536e-01 -3.35360527e-01 6.82573318e-01 -3.37767869e-01 -2.33551636e-01 1.18600070e+00 -3.85523200e-01 2.71789402e-01 1.51350111e-01 1.16115189e+00 6.94494992e-02 -1.83912802e+00 -3.86900604e-01 3.58267933e-01 -3.47562402e-01 2.19766125e-01 -8.65892828e-01 -1.36041701e+00 6.68834150e-01 4.11607832e-01 6.69536218e-02 1.12643600e+00 1.63094565e-01 9.41819072e-01 1.04145721e-01 1.63105026e-01 -1.24725318e+00 4.89100814e-03 2.80501306e-01 2.55996495e-01 -1.69058883e+00 -7.90052563e-02 -1.00654411e+00 -8.03517520e-01 6.64367974e-01 5.86003184e-01 -2.01435000e-01 4.26399678e-01 4.17155862e-01 3.53516847e-01 -3.85640049e-03 -3.31364036e-01 -6.39566004e-01 3.39485615e-01 5.52329421e-01 2.82009482e-01 7.74169341e-02 -2.14160562e-01 4.68281895e-01 2.26243868e-01 -2.12648571e-01 4.28453892e-01 9.22651768e-01 -5.16665578e-01 -1.12919211e+00 -3.93300653e-01 3.24775994e-01 -6.27124012e-01 -1.67225465e-01 -4.77301240e-01 9.09538329e-01 2.31824294e-01 1.00141919e+00 7.88569674e-02 -7.33158067e-02 3.15336525e-01 -1.01470225e-01 1.15570307e-01 -6.87866330e-01 -3.86050820e-01 4.95679557e-01 -2.72177719e-02 -5.82017720e-01 -5.75181186e-01 -5.37970424e-01 -1.44821906e+00 2.63155013e-01 -4.07807112e-01 2.24219752e-03 3.70135874e-01 8.83632958e-01 3.69457901e-01 3.49775285e-01 6.16455376e-01 -8.27061713e-01 -3.02916586e-01 -7.68182158e-01 -6.47127271e-01 4.53005284e-01 2.37907439e-01 -4.25020397e-01 -4.87063140e-01 2.45209500e-01]
[9.535964965820312, 0.5929101705551147]
a6c04aa7-21eb-485a-8984-87a0d696626f
physiological-and-affective-computing-through
1908.10307
null
https://arxiv.org/abs/1908.10307v1
https://arxiv.org/pdf/1908.10307v1.pdf
Physiological and Affective Computing through Thermal Imaging: A Survey
Thermal imaging-based physiological and affective computing is an emerging research area enabling technologies to monitor our bodily functions and understand psychological and affective needs in a contactless manner. However, up to recently, research has been mainly carried out in very controlled lab settings. As small size and even low-cost versions of thermal video cameras have started to appear on the market, mobile thermal imaging is opening its door to ubiquitous and real-world applications. Here we review the literature on the use of thermal imaging to track changes in physiological cues relevant to affective computing and the technological requirements set so far. In doing so, we aim to establish computational and methodological pipelines from thermal images of the human skin to affective states and outline the research opportunities and challenges to be tackled to make ubiquitous real-life thermal imaging-based affect monitoring a possibility.
['Nadia Bianchi-Berthouze', 'Youngjun Cho']
2019-08-27
null
null
null
null
['physiological-computing']
['computer-vision']
[ 5.89341342e-01 -3.71407390e-01 1.31563663e-01 -4.65373397e-01 -1.58231527e-01 -5.54355681e-01 2.91150689e-01 -7.93932229e-02 -5.52106082e-01 4.18789178e-01 -1.65667683e-01 4.63032097e-01 4.13145512e-01 -2.43741661e-01 5.22414185e-02 -8.49610507e-01 3.29789892e-02 -4.06120449e-01 -3.05619240e-01 1.38668492e-01 1.42077267e-01 2.01081425e-01 -1.78497481e+00 4.74108905e-01 4.93031144e-01 1.31198788e+00 -3.82559933e-02 8.43933761e-01 2.90107965e-01 3.03055465e-01 -4.39925194e-01 -3.06135684e-01 3.00366729e-02 -3.48401666e-01 -5.38814425e-01 -1.65318072e-01 1.89309314e-01 -3.72360319e-01 8.02744552e-03 8.48604143e-01 9.82828557e-01 9.60878134e-02 1.44055724e-01 -1.04307580e+00 -5.66772103e-01 -2.19599709e-01 -6.05411649e-01 2.08469346e-01 9.41142082e-01 2.62429297e-01 1.47249043e-01 -7.41189957e-01 2.74007589e-01 7.46297419e-01 6.12888932e-01 9.70521867e-01 -1.09263098e+00 -4.27090734e-01 -8.40026736e-02 3.73044640e-01 -1.06482375e+00 -8.56794178e-01 7.97248542e-01 -2.09042877e-01 1.08358431e+00 6.75954878e-01 1.27488458e+00 1.70108807e+00 4.76705939e-01 5.38899839e-01 1.85352576e+00 -3.31141889e-01 5.06091833e-01 4.55119222e-01 -2.01896489e-01 2.95767874e-01 3.43738683e-03 -2.01437071e-01 -8.00036311e-01 -9.48666111e-02 6.87195957e-01 -2.37092942e-01 -2.28831097e-01 5.22886515e-02 -9.05914426e-01 2.51438469e-03 -1.46512046e-01 7.92521298e-01 -6.74762130e-01 5.00893354e-01 6.84563041e-01 1.18381582e-01 7.18787432e-01 4.72320467e-01 -2.60702938e-01 -8.55256021e-01 -7.58418679e-01 -3.36081356e-01 4.97379065e-01 4.20070916e-01 2.53586024e-01 -1.31582007e-01 1.21364452e-01 8.47342610e-01 5.00400253e-02 5.75839102e-01 3.73644233e-01 -1.16066802e+00 -4.73343343e-01 3.88588697e-01 1.33471400e-01 -8.77612770e-01 -5.91970563e-01 4.51264083e-01 -4.98703361e-01 1.00082912e-01 1.42662168e-01 -5.42241275e-01 -4.59593266e-01 1.48584557e+00 6.49506748e-01 2.69158691e-01 -3.16904098e-01 1.16140246e+00 7.29222894e-01 4.07159269e-01 3.17202359e-01 -7.66883552e-01 1.61535811e+00 -1.74412921e-01 -1.12345386e+00 -2.01240152e-01 8.24719965e-02 -6.49137020e-01 1.09548509e+00 6.22978568e-01 -1.11063850e+00 -3.68723631e-01 -7.42155254e-01 8.18758737e-03 -4.10858244e-01 -2.35646218e-01 1.06539834e+00 1.55601919e+00 -1.30306101e+00 2.64384389e-01 -1.12669909e+00 -9.96690035e-01 9.49928463e-02 4.14841622e-01 -3.78275305e-01 1.80252627e-01 -1.13107049e+00 8.78476083e-01 -1.28446206e-01 6.28185868e-01 -4.85746354e-01 -6.12261236e-01 -4.82696712e-01 -4.00019735e-01 1.29756242e-01 -7.44314730e-01 1.03662860e+00 -1.31771004e+00 -2.24539137e+00 1.19625247e+00 -2.44682565e-01 2.87160307e-01 1.34766316e-02 -3.23181570e-01 -5.25065243e-01 6.88174486e-01 -4.92102623e-01 5.05967975e-01 7.74621785e-01 -9.40881729e-01 2.18567446e-01 -6.02954209e-01 -8.19962621e-02 4.31855500e-01 -7.81372964e-01 4.74991500e-01 -3.11023384e-01 6.05690405e-02 -1.48306102e-01 -9.76213932e-01 -8.54686052e-02 1.59298524e-01 1.90301672e-01 3.59994590e-01 8.24261844e-01 -3.81977767e-01 8.91636670e-01 -1.95885122e+00 -8.09299722e-02 9.06057879e-02 3.35597917e-02 4.16464210e-01 2.95599818e-01 4.19813544e-01 -1.06605776e-02 5.33735193e-03 1.92566793e-02 -2.13220820e-01 -1.25231713e-01 -9.89874750e-02 2.54856735e-01 7.10132480e-01 -3.18310350e-01 1.05667818e+00 -1.06501794e+00 -5.24848819e-01 8.95394206e-01 7.01468468e-01 -6.87699839e-02 2.59057403e-01 3.56513083e-01 8.60915959e-01 -4.30887759e-01 9.31459188e-01 4.98372138e-01 5.30461848e-01 2.52746761e-01 -1.60365582e-01 -3.33297014e-01 -2.54510343e-01 -7.29761124e-01 1.78029454e+00 -6.85904562e-01 6.81470037e-01 6.07774317e-01 -6.66791260e-01 7.35454142e-01 7.35730469e-01 1.05030262e+00 -1.23787630e+00 3.47989559e-01 -1.18389197e-01 -4.48618412e-01 -1.20104396e+00 4.96031970e-01 -6.28576279e-01 -1.54517844e-01 3.07631105e-01 -1.48623824e-01 -2.36156628e-01 -5.11959314e-01 -3.47840101e-01 1.02029157e+00 5.80135584e-01 2.71286845e-01 -1.25728533e-01 4.06938106e-01 -3.80910426e-01 1.68420017e-01 2.97318608e-01 -9.93682504e-01 3.66453469e-01 -2.01873027e-05 -3.48942101e-01 -5.24857283e-01 -8.04417610e-01 -3.34662944e-02 1.28842723e+00 2.15084046e-01 -2.79451221e-01 -1.07534039e+00 7.95655549e-02 -4.89570171e-01 2.98619062e-01 -8.82519841e-01 -1.70721933e-01 3.81915942e-02 -8.95761788e-01 5.55735826e-01 2.45720088e-01 4.69106138e-01 -1.32527852e+00 -1.32671881e+00 1.57260478e-01 -2.90813535e-01 -1.35174465e+00 1.65944085e-01 1.06669039e-01 -7.53731489e-01 -5.64050972e-01 -5.69456935e-01 -7.45087713e-02 4.75951493e-01 2.37185881e-01 6.71717346e-01 -3.67613703e-01 -8.46907854e-01 1.23248422e+00 -3.36920440e-01 -5.15907884e-01 2.79405594e-01 -3.67336512e-01 3.61351371e-01 2.25548938e-01 5.10024190e-01 -8.23908806e-01 -1.16161776e+00 9.40137208e-02 -8.81312549e-01 -2.33348563e-01 2.03090668e-01 8.67786705e-02 3.90471727e-01 -4.41283345e-01 5.28633833e-01 -5.62187910e-01 7.85054505e-01 -3.68678302e-01 1.61710948e-01 1.16179273e-01 -2.97096610e-01 -6.58761621e-01 5.01023173e-01 -5.07222354e-01 -1.51116967e+00 9.29711834e-02 -1.98519096e-01 -2.41460130e-01 -6.33023620e-01 3.75306040e-01 1.33490162e-02 -3.62733692e-01 7.04926074e-01 -1.08407423e-01 -1.16320826e-01 2.36145064e-01 4.10992712e-01 9.46020007e-01 5.64728856e-01 -6.44193351e-01 1.91775218e-01 7.48235703e-01 -1.58641651e-01 -1.38659394e+00 -4.34536219e-01 -5.19911587e-01 -7.90884376e-01 -1.23433197e+00 1.01997232e+00 -6.11982763e-01 -1.12485015e+00 6.48001254e-01 -5.31776845e-01 -3.89681250e-01 4.69002724e-02 2.60748565e-01 -5.25555670e-01 5.61756015e-01 -7.28219450e-01 -1.57841754e+00 -7.29843855e-01 -6.81129575e-01 1.14413834e+00 4.78509277e-01 -7.93730319e-01 -1.00931931e+00 5.13418019e-01 7.01437652e-01 6.37610257e-01 7.02944160e-01 1.30298719e-01 5.06215930e-01 1.50077894e-01 -2.35435069e-01 3.02470237e-01 -2.17595268e-02 2.75205672e-01 2.31046334e-01 -1.89548755e+00 -2.78183129e-02 5.07869422e-01 -6.69720888e-01 2.36556083e-01 5.95523715e-01 1.14473915e+00 2.81555116e-01 -2.63162076e-01 4.58418697e-01 1.31812167e+00 3.81471038e-01 9.13693726e-01 -1.35270447e-01 3.13353568e-01 8.85341942e-01 8.47755909e-01 6.01492107e-01 -5.81485219e-02 5.05842626e-01 3.78561109e-01 -1.52248666e-01 4.82274979e-01 2.23783359e-01 5.15374660e-01 9.01586354e-01 -4.30575669e-01 1.88030317e-01 -5.34826696e-01 3.06840152e-01 -1.46968377e+00 -1.12772560e+00 -3.52754951e-01 2.48263550e+00 6.68845832e-01 -3.59128803e-01 -1.39568010e-02 -9.46503729e-02 5.85277617e-01 3.70090641e-02 -5.43210685e-01 -1.20258415e+00 2.96877861e-01 4.92389083e-01 -1.22582138e-01 5.57005890e-02 -9.69250441e-01 6.92219257e-01 7.03546810e+00 1.32661670e-01 -1.63117516e+00 2.53081828e-01 7.77058244e-01 -5.64372420e-01 -8.35430324e-02 -3.12332928e-01 2.42989361e-01 4.18145239e-01 1.41452849e+00 3.44247937e-01 5.05732894e-01 5.74248791e-01 7.66392708e-01 -9.71544445e-01 -9.99905050e-01 1.30801535e+00 9.78257274e-04 -3.74309182e-01 -8.39963257e-01 -1.67925321e-02 4.18170571e-01 -3.56765181e-01 2.18057603e-01 -1.00858562e-01 -7.82436430e-01 -1.01596653e+00 2.39657804e-01 7.21245527e-01 1.22293389e+00 -5.77700794e-01 5.69158435e-01 6.41774908e-02 -9.61620808e-01 2.42948785e-01 -2.07863063e-01 -5.76422334e-01 8.64888504e-02 7.47946084e-01 -3.99324715e-01 2.15771601e-01 1.00449681e+00 1.93192720e-01 -1.97695717e-01 6.78912520e-01 7.84034505e-02 6.22653842e-01 -4.00664002e-01 -4.63976860e-01 -1.48685783e-01 -4.08813030e-01 1.17187567e-01 1.31664538e+00 1.72311455e-01 5.83470166e-01 -4.32991087e-01 7.67057240e-01 4.27755326e-01 8.00087824e-02 -7.95906961e-01 -1.74098790e-01 2.73219973e-01 2.11882782e+00 -1.04880190e+00 -7.15972260e-02 -3.29061389e-01 1.62987804e+00 -1.36669159e-01 1.24614120e-01 -1.01937032e+00 -1.06124133e-01 9.87629533e-01 -1.31706566e-01 -5.22514164e-01 -4.22695607e-01 -4.45321977e-01 -1.11208367e+00 -5.28571308e-02 -5.44773161e-01 -5.12114391e-02 -1.10357666e+00 -9.78291512e-01 7.18974173e-02 -6.60729781e-02 -7.93997169e-01 1.27783179e-01 -4.00058329e-01 -7.72936463e-01 5.98019242e-01 -9.71011996e-01 -7.74822176e-01 -6.81998670e-01 5.67601919e-01 2.17302963e-01 9.13135231e-01 1.36418843e+00 2.88300842e-01 -7.80011952e-01 3.22209656e-01 -1.70694981e-02 -5.16361475e-01 1.09351051e+00 -1.00180626e+00 -3.32651325e-02 5.41762233e-01 -3.83190155e-01 8.31313908e-01 7.20852137e-01 -2.69900382e-01 -2.03852868e+00 -3.11799794e-01 1.62373647e-01 -5.56183577e-01 3.62927049e-01 -7.03562558e-01 -5.17498732e-01 9.35323238e-02 6.27833009e-01 -1.89391915e-02 1.20249760e+00 1.04658403e-01 1.27165794e-01 -2.99587756e-01 -1.56942129e+00 6.92127168e-01 7.26240695e-01 -8.04153323e-01 -1.55630708e-01 -1.18844612e-02 -1.13543749e-01 -2.50967860e-01 -1.13780141e+00 1.06350519e-01 1.24415362e+00 -1.17123032e+00 6.51547670e-01 2.60183662e-01 1.52317420e-01 -9.47762951e-02 4.04683501e-01 -1.10144341e+00 -2.89152581e-02 -1.01169801e+00 1.63618416e-01 1.29921401e+00 -2.64402956e-01 -6.00922525e-01 8.22305679e-01 1.33204412e+00 4.98847030e-02 -6.14639997e-01 -9.78694856e-01 -7.95572102e-02 -3.20377916e-01 -9.25172389e-01 -1.32198691e-01 1.06741023e+00 9.98937190e-01 7.44235143e-02 -3.64897907e-01 -2.35887781e-01 1.93234131e-01 -7.83390850e-02 2.75604248e-01 -8.76245320e-01 2.11967513e-01 -1.81224778e-01 -4.46265131e-01 -2.24247575e-01 -1.26156196e-01 -1.81655958e-01 1.64535776e-01 -1.25174081e+00 4.49796379e-01 1.52491122e-01 -3.40697974e-01 2.74380565e-01 -2.53393918e-01 7.36235023e-01 -1.34642228e-01 -3.44805658e-01 -5.77032864e-01 2.83888310e-01 1.08980906e+00 3.66791993e-01 -2.52533585e-01 -4.07733917e-01 -6.41245306e-01 5.56090117e-01 7.98319995e-01 1.18669651e-01 -4.98386800e-01 1.42892048e-01 4.45200682e-01 1.72630399e-01 2.04459041e-01 -1.17494714e+00 2.14724660e-01 -3.05942208e-01 6.63862169e-01 -1.11049023e-02 8.25418890e-01 -1.02366543e+00 3.35531116e-01 1.95628956e-01 -1.50291443e-01 -2.46403843e-01 1.81273550e-01 1.74768999e-01 1.88439220e-01 7.88354278e-02 8.29501987e-01 -5.07758975e-01 -8.21477890e-01 -1.03028484e-01 -1.06539524e+00 -2.95497030e-01 1.39196265e+00 -7.78338790e-01 -2.44552240e-01 -5.17729223e-01 -7.10809827e-01 -5.62307276e-02 6.35662079e-01 1.91008806e-01 7.03493416e-01 -9.94491220e-01 1.42491609e-01 -4.03852724e-02 1.46645650e-01 -7.30076730e-01 7.72690475e-01 1.31359720e+00 -3.21291715e-01 3.40216070e-01 -6.23614490e-01 -5.51431835e-01 -1.65120542e+00 4.75630015e-01 4.84952062e-01 3.26470286e-01 -2.75906265e-01 7.16322720e-01 -9.14741084e-02 -9.63698477e-02 -2.07272172e-01 -1.29816577e-01 -8.39916170e-02 2.19053194e-01 4.71294165e-01 4.92742598e-01 1.93218902e-01 -2.99692988e-01 -6.04075551e-01 6.03800893e-01 7.22814202e-01 -4.74449307e-01 1.04921603e+00 -6.83601022e-01 -4.02696490e-01 1.02282774e+00 9.12250221e-01 -2.31225580e-01 -9.28185225e-01 5.21778524e-01 -2.73401618e-01 -4.32599127e-01 9.43296477e-02 -9.26478446e-01 -8.02222013e-01 1.02753031e+00 9.98554468e-01 1.96218878e-01 1.65298390e+00 -3.36797476e-01 7.35324860e-01 1.70038894e-01 6.14424169e-01 -1.69573641e+00 2.60424048e-01 5.98020740e-02 3.93077791e-01 -8.67786646e-01 1.76472589e-01 -4.99828845e-01 -8.10454607e-01 1.12932575e+00 6.97099507e-01 1.78458363e-01 4.54267144e-01 6.56475186e-01 3.11134994e-01 -2.73516804e-01 -8.69810343e-01 -1.19620577e-01 -9.16389376e-02 8.74031365e-01 1.00030684e+00 1.36279836e-01 -3.35407943e-01 1.19221173e-01 1.66079849e-01 3.39924872e-01 5.36113143e-01 1.03798449e+00 -3.36312592e-01 -8.23066711e-01 -6.03542387e-01 4.22866255e-01 -9.68790710e-01 1.58657849e-01 -9.44935799e-01 3.99797827e-01 1.28441527e-01 1.23295689e+00 -2.49426320e-01 -6.20389521e-01 1.10628061e-01 5.47842741e-01 9.16573942e-01 -4.40014809e-01 -9.33218539e-01 1.66475371e-01 1.53605744e-01 -1.06775129e+00 -1.02185476e+00 -8.22541177e-01 -1.10194755e+00 -2.28650033e-01 -1.54768392e-01 -3.52811366e-01 1.11880720e+00 7.77827084e-01 4.66092676e-01 3.03162634e-01 5.40752172e-01 -1.19250584e+00 4.71147478e-01 -9.78805721e-01 -6.11979723e-01 1.44345909e-01 -8.35883841e-02 -3.19191515e-01 2.40337458e-02 1.03520965e-02]
[13.586858749389648, 2.524019241333008]
e3ac909e-3cb0-4dc7-995a-5a1134a31d2c
cross-diffusion-model-makes-controllable
2305.16936
null
https://arxiv.org/abs/2305.16936v1
https://arxiv.org/pdf/2305.16936v1.pdf
CRoSS: Diffusion Model Makes Controllable, Robust and Secure Image Steganography
Current image steganography techniques are mainly focused on cover-based methods, which commonly have the risk of leaking secret images and poor robustness against degraded container images. Inspired by recent developments in diffusion models, we discovered that two properties of diffusion models, the ability to achieve translation between two images without training, and robustness to noisy data, can be used to improve security and natural robustness in image steganography tasks. For the choice of diffusion model, we selected Stable Diffusion, a type of conditional diffusion model, and fully utilized the latest tools from open-source communities, such as LoRAs and ControlNets, to improve the controllability and diversity of container images. In summary, we propose a novel image steganography framework, named Controllable, Robust and Secure Image Steganography (CRoSS), which has significant advantages in controllability, robustness, and security compared to cover-based image steganography methods. These benefits are obtained without additional training. To our knowledge, this is the first work to introduce diffusion models to the field of image steganography. In the experimental section, we conducted detailed experiments to demonstrate the advantages of our proposed CRoSS framework in controllability, robustness, and security.
['Jian Zhang', 'Youmin Xu', 'Xuanyu Zhang', 'Jiwen Yu']
2023-05-26
null
null
null
null
['image-steganography']
['computer-vision']
[ 6.60961747e-01 5.55153796e-03 3.80728841e-02 3.14989984e-01 1.80804223e-01 -5.67393541e-01 6.76172435e-01 -2.48768926e-01 -1.60753772e-01 5.49209237e-01 -1.20319039e-01 -6.62512243e-01 -8.24289545e-02 -1.04636264e+00 -4.82852906e-01 -9.42820549e-01 -5.14914989e-01 -3.99283350e-01 4.07188237e-01 -5.91357172e-01 3.85889977e-01 2.34308481e-01 -1.16873145e+00 -1.27451926e-01 8.60562027e-01 6.23879552e-01 5.61818331e-02 7.14392245e-01 3.23276460e-01 9.28179681e-01 -5.39873779e-01 -1.85766190e-01 3.07097107e-01 -8.48303080e-01 -4.93123144e-01 5.03933311e-01 -5.90620220e-01 -2.74455130e-01 -6.29516125e-01 1.26924670e+00 4.17349458e-01 -5.07731915e-01 3.10555816e-01 -1.33319926e+00 -9.65567470e-01 8.24133277e-01 -5.86596549e-01 9.47969034e-03 2.13554174e-01 4.40635532e-01 2.42527932e-01 -1.91352695e-01 8.70881796e-01 1.09441555e+00 4.78579521e-01 6.05644941e-01 -8.20168972e-01 -9.05202091e-01 -6.71395212e-02 2.54980549e-02 -1.22174084e+00 -4.44543391e-01 7.30357409e-01 -1.18421987e-01 5.57007849e-01 5.38932502e-01 7.90418029e-01 9.13384497e-01 4.95685875e-01 4.78068531e-01 1.72693491e+00 -6.13260984e-01 -7.77121782e-02 4.16751266e-01 -5.43545663e-01 6.55474186e-01 3.97274554e-01 4.79442775e-01 5.40882312e-02 -7.60417208e-02 8.61463785e-01 1.14031672e-01 -8.41711104e-01 -3.41938049e-01 -1.60603595e+00 9.15386379e-01 5.33433795e-01 6.15637064e-01 -5.97424246e-03 2.72872239e-01 3.48785035e-02 9.40977216e-01 2.53733367e-01 2.58233935e-01 1.04179665e-01 3.52357119e-01 -5.55467606e-01 -1.70779705e-01 9.89928424e-01 1.01469123e+00 4.54564363e-01 2.18897253e-01 2.24336565e-01 -6.54974580e-02 6.80312455e-01 1.02131844e+00 4.44826335e-01 -5.29987216e-01 2.82874107e-01 4.48742121e-01 -2.32097194e-01 -1.61774814e+00 -7.12664612e-03 -3.58777314e-01 -1.24686027e+00 2.60784507e-01 -2.02096090e-01 -1.29376411e-01 -9.27510500e-01 1.47905064e+00 1.58128500e-01 2.68822968e-01 5.27252614e-01 5.11798084e-01 6.00614905e-01 7.40473807e-01 -3.90400708e-01 -5.67958117e-01 9.09582973e-01 -7.37494826e-01 -8.70215058e-01 1.60635084e-01 5.78364730e-01 -1.01407111e+00 3.79008710e-01 1.50670568e-02 -9.22378659e-01 -2.09489182e-01 -1.42506373e+00 7.75469184e-01 -3.96792650e-01 -6.39544487e-01 4.70897824e-01 1.30664158e+00 -1.16405416e+00 4.54170674e-01 -6.26336455e-01 -3.70440751e-01 3.48047465e-01 3.68144900e-01 -3.74752373e-01 -2.11425528e-01 -1.53963280e+00 6.19760692e-01 4.78628963e-01 -1.05103627e-01 -1.13899720e+00 1.25207633e-01 -8.74991417e-01 -1.34981543e-01 3.39240700e-01 -4.12608802e-01 4.53033447e-01 -1.01369846e+00 -1.57242322e+00 6.32716358e-01 4.50146377e-01 -6.38076484e-01 6.53304458e-01 6.75070643e-01 -6.87835455e-01 2.46159151e-01 -2.61749923e-01 3.37736964e-01 1.26731217e+00 -1.40115654e+00 -3.38280052e-01 -8.71968493e-02 2.34745946e-02 -2.27587298e-01 -4.87736821e-01 7.16890395e-02 -2.53697455e-01 -5.69170058e-01 1.75111666e-01 -1.24445736e+00 -3.86810869e-01 -1.93162203e-01 -4.49989885e-01 5.21355450e-01 1.33586836e+00 -3.96596700e-01 1.14854312e+00 -2.14582968e+00 1.01272032e-01 5.65932453e-01 3.59794497e-01 6.73435926e-01 -1.87370613e-01 7.32374966e-01 1.42688304e-01 6.34776592e-01 -4.03954744e-01 1.57110110e-01 -3.61229002e-01 1.90857813e-01 -2.21755922e-01 8.52460146e-01 -1.09633349e-01 8.52443337e-01 -8.65543246e-01 -4.45719749e-01 2.89262861e-01 6.71730757e-01 -2.47658446e-01 -1.20075308e-01 7.61509910e-02 9.11157250e-01 -4.29577172e-01 5.17230272e-01 9.49842870e-01 -2.70827174e-01 3.59194726e-01 2.75606632e-01 -1.61420867e-01 -3.78431976e-01 -1.06676865e+00 1.03037405e+00 -2.62484014e-01 4.64951456e-01 7.32101649e-02 -7.83281505e-01 9.36357856e-01 6.35358453e-01 3.40893120e-01 -4.82898057e-01 4.44743067e-01 5.09166241e-01 7.29000345e-02 -8.35742652e-01 2.08015889e-01 1.71196178e-01 1.31321043e-01 8.08911741e-01 -3.33617926e-01 -3.13605815e-01 4.86061163e-02 1.82977736e-01 9.65254426e-01 -1.39493972e-01 1.51660040e-01 -1.76100418e-01 7.78200030e-01 -2.27531180e-01 1.94573879e-01 8.28019381e-01 -1.06248751e-01 3.11900675e-01 3.79800081e-01 -1.75963230e-02 -1.17171311e+00 -4.39058602e-01 1.91100195e-01 1.72137141e-01 8.22870672e-01 -3.58994156e-01 -8.10930014e-01 -5.30489206e-01 -4.79881287e-01 1.23843499e-01 -2.97033161e-01 -3.15282881e-01 -6.28505647e-01 -8.93444300e-01 8.96997392e-01 -3.88236940e-01 1.28592777e+00 -1.00083041e+00 -4.01784182e-01 5.42973541e-02 -2.51445383e-01 -1.15974641e+00 -3.58017713e-01 -5.07630825e-01 -7.89063156e-01 -1.27127230e+00 -9.54233170e-01 -1.01214290e+00 9.15620625e-01 7.94983387e-01 5.03423810e-01 7.58162975e-01 -6.00984283e-02 2.04299539e-01 -7.05717206e-01 -2.98956305e-01 -1.13351190e+00 2.08862548e-05 -1.30481645e-01 2.84584731e-01 -4.10945594e-01 -5.10478318e-01 -7.88722336e-01 7.09121644e-01 -1.54060459e+00 7.63294399e-02 7.73754001e-01 7.02576578e-01 9.83484462e-02 8.14758182e-01 1.65291876e-01 -8.81761789e-01 6.61976755e-01 -5.78222632e-01 -4.78620470e-01 2.10865647e-01 -1.02274108e+00 -3.97452600e-02 4.12582040e-01 -5.61320424e-01 -7.07214475e-01 -3.04049730e-01 -2.24839859e-02 -2.48707621e-03 2.44306996e-01 5.93754590e-01 3.06211412e-02 -9.54218447e-01 4.96924192e-01 7.28929996e-01 5.41466117e-01 -8.95727426e-03 2.47851208e-01 8.94107699e-01 2.53798943e-02 1.59241855e-01 1.14217186e+00 9.04321194e-01 1.69063240e-01 -7.67242789e-01 1.41359344e-01 -6.01866357e-02 -5.88373654e-02 -9.48022380e-02 5.54523289e-01 -7.38284171e-01 -6.69559717e-01 1.12338006e+00 -8.69306505e-01 -7.50138834e-02 1.24888003e-01 3.87766004e-01 -3.15509200e-01 9.33235765e-01 -5.17193854e-01 -7.35411346e-01 -3.16660404e-01 -1.37173057e+00 3.45974982e-01 4.69688997e-02 4.90004927e-01 -1.15973592e+00 -2.13229060e-01 7.74427876e-02 1.03994310e+00 7.14488089e-01 6.82177305e-01 -2.19010249e-01 -9.77028131e-01 -3.94092768e-01 -8.93125758e-02 4.64265376e-01 2.61294097e-01 -1.63173392e-01 -3.79946947e-01 -8.88942122e-01 5.16371846e-01 -7.66990036e-02 9.43768442e-01 -4.62404974e-02 5.30391514e-01 -6.92733407e-01 -5.21026075e-01 7.40685582e-01 1.70366931e+00 3.16876441e-01 1.25554228e+00 5.55476129e-01 3.80595624e-01 3.59257162e-01 2.23873645e-01 3.57405990e-01 3.29818606e-01 2.54273862e-01 7.30530798e-01 -2.73309320e-01 -1.09168582e-01 -2.25304633e-01 4.54764634e-01 1.06352723e+00 -3.12645942e-01 -7.83242464e-01 -4.88357961e-01 4.70951557e-01 -1.62639749e+00 -1.01008117e+00 -1.98533401e-01 2.03740168e+00 4.90442514e-01 1.94987193e-01 -1.77199438e-01 5.17774582e-01 1.12354851e+00 4.71801966e-01 -1.49538279e-01 -1.54256925e-01 -4.17737603e-01 -1.19228169e-01 1.04156840e+00 5.22806227e-01 -9.51850712e-01 9.38141763e-01 6.20528793e+00 8.81761730e-01 -1.34399498e+00 3.14769268e-01 3.93586725e-01 5.28281868e-01 -5.83589673e-01 4.48966771e-01 -3.00700754e-01 5.04725873e-01 6.60018802e-01 -3.09677515e-02 5.82262635e-01 2.66832918e-01 1.48574961e-02 1.70716688e-01 -2.12112963e-01 7.38653898e-01 2.59622693e-01 -1.31535947e+00 4.18328978e-02 3.81515205e-01 1.14447200e+00 -2.74146199e-01 4.72819984e-01 -2.23615542e-01 1.88721418e-01 -1.06751728e+00 3.72064531e-01 1.22076854e-01 7.09351957e-01 -6.03011429e-01 8.13045621e-01 3.60857844e-01 -8.08903456e-01 -6.84634149e-02 -3.99850398e-01 7.49735981e-02 1.33100405e-01 3.49382222e-01 -4.19813991e-01 7.46846259e-01 3.76061708e-01 7.37834513e-01 -2.53694385e-01 8.89678776e-01 -4.52704191e-01 6.58203602e-01 -1.61166012e-01 -1.01663888e-01 4.07752246e-01 5.82070723e-02 9.46131766e-01 9.00631487e-01 7.13199019e-01 -5.19804209e-02 1.02210872e-01 6.62564933e-01 9.14740339e-02 -3.54069583e-02 -1.01833355e+00 -6.99921623e-02 2.69482642e-01 8.48206639e-01 -8.53722453e-01 -1.89697102e-01 -1.76556289e-01 1.06762540e+00 -6.94001973e-01 3.14379990e-01 -7.50580430e-01 -7.18543828e-01 3.07297297e-02 2.16070458e-01 5.47563851e-01 -4.77232575e-01 4.74158600e-02 -1.36037719e+00 -3.10729593e-01 -1.27975082e+00 -7.06836954e-02 -2.92921126e-01 -6.14295900e-01 8.28369915e-01 -1.11624017e-01 -1.48081434e+00 2.13222608e-01 -8.03058073e-02 -4.35299844e-01 4.21701163e-01 -1.82881129e+00 -1.53921127e+00 -1.66291669e-01 1.00847721e+00 7.34445080e-02 -4.40877497e-01 7.41016090e-01 -3.12450022e-04 -2.02450648e-01 4.29033935e-01 4.64429617e-01 5.28906249e-02 3.34926367e-01 -5.86397648e-01 5.62895238e-01 1.17832601e+00 -1.94792330e-01 5.43160975e-01 8.50253880e-01 -6.94629788e-01 -1.54645789e+00 -1.05867374e+00 7.38254249e-01 7.49321058e-02 5.21215916e-01 -1.62830010e-01 -4.32063401e-01 5.52523732e-01 4.83243287e-01 -1.49349406e-01 3.09411943e-01 -8.86715829e-01 -2.21919566e-01 1.54018015e-01 -1.46345484e+00 6.73350811e-01 1.00049102e+00 -2.19118863e-01 6.94564357e-02 3.85967046e-01 9.10694957e-01 -2.96264201e-01 -7.92840719e-01 2.96923667e-01 4.47830766e-01 -1.10045671e+00 9.12245154e-01 2.09835112e-01 1.26828209e-01 -3.93433243e-01 7.54913464e-02 -1.21324122e+00 -1.65224314e-01 -1.42308438e+00 -4.02311534e-02 1.01588166e+00 2.04650939e-01 -1.15144992e+00 3.84850949e-01 -1.96278527e-01 4.46625322e-01 -2.95438677e-01 -6.48265302e-01 -9.18504477e-01 -1.07997194e-01 -9.38032195e-02 9.72986758e-01 9.54144180e-01 -7.56438300e-02 -2.90010750e-01 -1.02704978e+00 3.53371799e-01 8.48500252e-01 -9.57929417e-02 6.95952535e-01 -7.63605237e-01 -2.13204563e-01 -1.93729028e-01 -7.75505662e-01 -9.73846376e-01 -2.67526925e-01 -6.12644017e-01 -2.41782933e-01 -1.15253592e+00 -5.72588108e-02 -7.17494726e-01 -1.59632981e-01 8.56433585e-02 1.16119020e-01 7.62163758e-01 4.38959569e-01 7.49111831e-01 -3.51212382e-01 2.51783013e-01 1.86055458e+00 -3.36304933e-01 -2.88596123e-01 1.41241267e-01 -8.09804320e-01 3.54204357e-01 9.88035858e-01 -8.18768442e-01 -5.47050774e-01 -3.63905221e-01 3.55706751e-01 1.82111382e-01 3.04695636e-01 -1.05597663e+00 2.32719243e-01 -2.77469516e-01 -2.72452950e-01 2.09850073e-01 -4.51283082e-02 -9.82272685e-01 4.73895639e-01 1.39223611e+00 -1.16108321e-01 -2.42764577e-01 -3.52468193e-01 8.39438438e-01 -2.26124227e-01 -1.19510442e-01 8.93248618e-01 -4.74493235e-01 -6.25922441e-01 3.64221781e-01 -6.12980187e-01 -5.99433839e-01 1.39648914e+00 -4.03160781e-01 -4.64502364e-01 -7.57233918e-01 -5.70915639e-01 -8.80331397e-02 7.61430621e-01 3.28118354e-01 9.55486178e-01 -1.13543665e+00 -9.81721699e-01 6.60804570e-01 2.67730430e-02 -4.63267624e-01 6.55515194e-02 1.05714679e+00 -9.38446820e-01 3.15329045e-01 -2.62217402e-01 -5.24612427e-01 -1.33729327e+00 8.72082293e-01 1.54939070e-01 -3.48825067e-01 -5.76896369e-01 4.78984624e-01 -9.51071531e-02 -2.16187939e-01 -1.58966407e-02 -7.69773945e-02 -2.57920891e-01 -5.98175406e-01 6.33247495e-01 2.61788726e-01 -3.58004481e-01 -8.74749959e-01 -1.91105396e-01 7.69321263e-01 -2.40496844e-02 -2.05683082e-01 9.52567399e-01 -7.38370180e-01 -5.90560853e-01 -2.98344135e-01 1.09179354e+00 4.47244495e-02 -7.35807478e-01 -1.94966003e-01 -4.80802417e-01 -7.85719573e-01 1.44777775e-01 -4.51564372e-01 -1.30433309e+00 5.52262127e-01 7.48746276e-01 8.55836093e-01 1.11953712e+00 -4.92353201e-01 1.16507924e+00 1.02944694e-01 6.30928755e-01 -5.10858774e-01 8.84763524e-02 3.07490021e-01 6.01705730e-01 -1.15046895e+00 -1.83422148e-01 -6.08128428e-01 -5.29411376e-01 1.07351482e+00 8.18666890e-02 -2.68197149e-01 9.65393186e-01 1.22121215e-01 8.05683210e-02 -1.83933377e-01 -3.77231240e-01 -1.54352993e-01 -3.77695322e-01 9.91006851e-01 -9.34814587e-02 -1.17231525e-01 -4.51251686e-01 -4.21905518e-01 1.56232223e-01 2.06760734e-01 8.88166010e-01 1.21157503e+00 -6.39521420e-01 -1.26429212e+00 -6.89161718e-01 -3.08370799e-01 -7.08086193e-01 -9.66672599e-02 -2.23488733e-01 7.01942205e-01 2.27943897e-01 1.48723173e+00 -5.22116065e-01 -6.86303735e-01 -1.71626255e-01 -7.30244219e-01 4.36968297e-01 -4.28887643e-02 -3.63978475e-01 -6.09656014e-02 -2.96022177e-01 -3.37351002e-02 -1.02476799e+00 -4.02122922e-02 -6.93488479e-01 -9.34045315e-01 -8.12962770e-01 2.33181223e-01 8.51621807e-01 8.35887432e-01 4.36479181e-01 3.88519228e-01 1.23701608e+00 -4.19717610e-01 -4.05283779e-01 -7.46071696e-01 -6.34110332e-01 3.22433352e-01 5.04027426e-01 -6.40182942e-02 -6.10871732e-01 5.82513064e-02]
[4.316466331481934, 8.053154945373535]
688e2665-0c68-4cce-8685-7a983a173843
beir-pl-zero-shot-information-retrieval
2305.19840
null
https://arxiv.org/abs/2305.19840v1
https://arxiv.org/pdf/2305.19840v1.pdf
BEIR-PL: Zero Shot Information Retrieval Benchmark for the Polish Language
The BEIR dataset is a large, heterogeneous benchmark for Information Retrieval (IR) in zero-shot settings, garnering considerable attention within the research community. However, BEIR and analogous datasets are predominantly restricted to the English language. Our objective is to establish extensive large-scale resources for IR in the Polish language, thereby advancing the research in this NLP area. In this work, inspired by mMARCO and Mr.~TyDi datasets, we translated all accessible open IR datasets into Polish, and we introduced the BEIR-PL benchmark -- a new benchmark which comprises 13 datasets, facilitating further development, training and evaluation of modern Polish language models for IR tasks. We executed an evaluation and comparison of numerous IR models on the newly introduced BEIR-PL benchmark. Furthermore, we publish pre-trained open IR models for Polish language,d marking a pioneering development in this field. Additionally, the evaluation revealed that BM25 achieved significantly lower scores for Polish than for English, which can be attributed to high inflection and intricate morphological structure of the Polish language. Finally, we trained various re-ranking models to enhance the BM25 retrieval, and we compared their performance to identify their unique characteristic features. To ensure accurate model comparisons, it is necessary to scrutinise individual results rather than to average across the entire benchmark. Thus, we thoroughly analysed the outcomes of IR models in relation to each individual data subset encompassed by the BEIR benchmark. The benchmark data is available at URL {\bf https://huggingface.co/clarin-knext}.
['Maciej Piasecki', 'Arkadiusz Janz', 'Kacper Wołowiec', 'Vadim Shishkin', 'Konrad Wojtasik']
2023-05-31
null
null
null
null
['information-retrieval']
['natural-language-processing']
[ 7.63585344e-02 -5.47565036e-02 -4.44527566e-01 1.15224749e-01 -1.34330714e+00 -7.91372538e-01 9.30107236e-01 3.90377641e-01 -7.82771170e-01 5.38451791e-01 4.83879209e-01 -2.92917401e-01 -6.16154194e-01 -6.56573534e-01 -1.75920501e-01 -3.44890058e-01 2.85722405e-01 7.77525902e-01 3.21411192e-02 -7.62248695e-01 3.28565687e-01 4.21567529e-01 -1.62210679e+00 4.60898191e-01 6.95073187e-01 5.68183482e-01 3.38191897e-01 4.15672749e-01 -2.33824402e-01 3.70278865e-01 -5.39684534e-01 -6.44296646e-01 1.15558945e-01 1.26240164e-01 -1.17945099e+00 -5.38400948e-01 3.04675668e-01 2.85337210e-01 -4.41849291e-01 8.35087955e-01 8.23044837e-01 3.92227143e-01 8.03667367e-01 -5.11439025e-01 -1.07078552e+00 8.26366365e-01 -3.15518618e-01 1.42591625e-01 6.83617830e-01 -2.13631883e-01 1.68321574e+00 -8.88333023e-01 1.15823102e+00 1.07549727e+00 2.19731838e-01 4.10125107e-01 -9.54308987e-01 -4.65971768e-01 -2.71419168e-01 1.34706587e-01 -1.62110531e+00 -4.10292923e-01 5.08482575e-01 -1.15803540e-01 9.42246318e-01 5.75580835e-01 6.76943883e-02 1.16014707e+00 -1.61320508e-01 8.14404368e-01 1.02112544e+00 -9.23302174e-01 -1.90129638e-01 2.25166321e-01 5.60974002e-01 2.28232950e-01 1.84238285e-01 -1.84217826e-01 -2.03191593e-01 -1.57787979e-01 2.82017171e-01 -1.77862525e-01 -3.89974654e-01 9.67135057e-02 -1.27516925e+00 7.82911658e-01 3.80271018e-01 8.14736247e-01 -1.85836062e-01 -4.25233632e-01 4.78563637e-01 6.35473251e-01 7.69090533e-01 8.05490255e-01 -5.45758963e-01 -2.32934505e-02 -6.84644043e-01 8.71252418e-02 7.49580562e-01 6.70864999e-01 5.23614943e-01 -6.69397950e-01 -5.47851861e-01 1.60533893e+00 1.88879177e-01 4.32577163e-01 6.74551606e-01 -6.36507928e-01 6.47632480e-01 6.68350756e-01 -5.71151404e-03 -8.81046832e-01 -3.17315042e-01 -2.31610075e-01 -6.64001167e-01 -3.48510832e-01 2.90850401e-01 3.10885191e-01 -5.74539304e-01 1.39850068e+00 -4.72493609e-03 -5.34271836e-01 4.89183456e-01 6.16001964e-01 1.21303546e+00 6.95570350e-01 1.64946839e-01 -1.74198985e-01 1.60670185e+00 -8.29286098e-01 -5.39775968e-01 -6.94005415e-02 9.44435596e-01 -1.19596672e+00 1.41510689e+00 2.47206032e-01 -9.02865231e-01 -3.10933650e-01 -8.67961764e-01 -3.77786785e-01 -8.21202457e-01 3.56264114e-01 3.49417031e-01 3.54866058e-01 -1.07435703e+00 4.14468497e-01 1.78215988e-02 -6.88256741e-01 -7.59567022e-02 1.10352799e-01 -6.05296969e-01 -4.95174408e-01 -1.71490085e+00 1.18964672e+00 4.92881328e-01 -1.21281147e-01 -2.41706416e-01 -6.37847900e-01 -6.47775114e-01 4.04907167e-02 2.48538464e-01 -4.74620402e-01 1.18783557e+00 -4.42691416e-01 -1.48191500e+00 1.40238380e+00 1.55887336e-01 -1.17939733e-01 2.45919555e-01 -1.77956358e-01 -7.69633234e-01 2.15115651e-01 -1.43777907e-01 4.73695219e-01 4.09323156e-01 -1.17271578e+00 -2.06991509e-01 -4.62462991e-01 7.31178522e-02 3.03224504e-01 -3.87610793e-01 7.31278479e-01 -6.47371888e-01 -7.12144494e-01 -2.32860312e-01 -7.54841924e-01 3.37824188e-02 -6.78880870e-01 -1.37341678e-01 -7.35376000e-01 1.14438087e-01 -6.07617378e-01 1.66092706e+00 -2.12281990e+00 1.77466813e-02 2.63998777e-01 1.83142554e-02 5.66301644e-01 -5.26712179e-01 8.29503357e-01 -2.78107315e-01 3.84380072e-01 1.22577772e-01 -1.91055819e-01 1.35893226e-01 -9.39999968e-02 -5.57550490e-01 2.94713795e-01 -1.11754937e-02 9.29698408e-01 -8.50127339e-01 -5.98144650e-01 1.40134752e-01 5.42875767e-01 -1.61162049e-01 -9.75425243e-02 -2.63871849e-01 3.96092273e-02 -4.58751559e-01 6.15121841e-01 3.35710704e-01 -4.76606749e-03 -1.11580277e-02 -2.18515955e-02 -2.85198510e-01 2.51848251e-01 -7.97585249e-01 1.75799918e+00 -6.97210968e-01 4.41812098e-01 -3.89275789e-01 -6.17083311e-01 1.14666343e+00 6.08737469e-01 4.30560231e-01 -8.77130628e-01 2.61613905e-01 4.16602314e-01 -2.52483517e-01 -9.34879258e-02 1.01152766e+00 1.58552766e-01 -3.11078042e-01 5.48633754e-01 1.87008664e-01 -2.29550257e-01 5.04019678e-01 1.94818437e-01 9.45813119e-01 -1.42399669e-01 4.45917040e-01 -3.36268485e-01 7.37803042e-01 1.32911354e-01 -1.98921748e-02 7.87096381e-01 4.18267399e-02 7.03306198e-01 1.64874539e-01 -2.79023349e-01 -7.14344680e-01 -9.85534012e-01 -8.82286310e-01 1.27398527e+00 -1.75798178e-01 -5.87468684e-01 -3.09014767e-01 -4.34320092e-01 -2.99137235e-01 7.66100883e-01 -4.59423512e-01 -1.87507868e-01 -2.65040457e-01 -8.68784666e-01 8.86558294e-01 -3.18885922e-01 1.43500060e-01 -1.35227406e+00 -1.75567009e-02 -4.66070734e-02 -4.95300382e-01 -8.99146080e-01 -3.08256656e-01 -4.57695685e-02 -5.60917258e-01 -1.13150215e+00 -1.19271100e+00 -9.96629953e-01 1.98444217e-01 3.27939481e-01 1.39756775e+00 8.63487273e-02 -5.29385448e-01 6.91315889e-01 -7.15876281e-01 -5.16132295e-01 -5.21588743e-01 4.85003948e-01 6.96039665e-03 -3.48671079e-01 6.90368891e-01 9.71935838e-02 -4.06378776e-01 4.08399194e-01 -1.16344953e+00 -5.15455723e-01 5.95087051e-01 6.47756219e-01 4.72364962e-01 -1.73767477e-01 5.59985399e-01 -7.03020394e-01 9.87973690e-01 -4.56143379e-01 -4.94649768e-01 7.80961096e-01 -6.07558966e-01 1.07136354e-01 1.27424315e-01 -3.58815283e-01 -1.23988152e+00 -6.39653981e-01 -3.28204632e-01 2.02835444e-02 -7.01467842e-02 7.36513138e-01 1.10467337e-01 8.80493596e-02 7.56321967e-01 -1.39687881e-01 -2.50389040e-01 -7.95418322e-01 4.54434365e-01 1.28913665e+00 2.20003337e-01 -8.50927234e-01 5.49789846e-01 1.26140818e-01 -3.72751653e-01 -1.20756066e+00 -8.71146202e-01 -1.06808531e+00 -4.17568892e-01 -2.11550426e-02 7.20641315e-01 -8.81271303e-01 -4.39951390e-01 2.09577307e-01 -1.14854324e+00 -6.84136748e-02 -2.62112617e-01 4.99578834e-01 -3.12064767e-01 4.57155794e-01 -7.86830544e-01 -6.20724320e-01 -8.70708525e-01 -9.46881354e-01 1.04011297e+00 -6.83069006e-02 -5.48845351e-01 -1.13200080e+00 8.27813566e-01 5.76624691e-01 2.08619371e-01 -4.49620813e-01 1.10372305e+00 -1.13453650e+00 1.90002769e-01 -4.51180488e-01 -2.60837883e-01 2.14565858e-01 -1.68476596e-01 1.98537976e-01 -1.01692295e+00 -1.56312600e-01 -3.44780028e-01 -5.36427259e-01 8.68330598e-01 -9.39718857e-02 7.83811390e-01 1.71762243e-01 -1.40332341e-01 3.59709896e-02 1.40037942e+00 -5.69023788e-02 8.90196025e-01 7.78719664e-01 1.83775872e-01 7.90916085e-01 7.12070107e-01 1.61985099e-01 5.44924401e-02 8.45940232e-01 -1.44956112e-01 4.26760539e-02 -1.12357430e-01 -1.42474025e-01 3.56028795e-01 1.05209637e+00 -3.51952940e-01 -2.54620552e-01 -9.89998698e-01 5.27621448e-01 -1.51383030e+00 -8.56637478e-01 2.14616433e-02 2.48073316e+00 1.03532779e+00 -1.10536724e-01 -2.45465994e-01 -4.24988158e-02 6.56139731e-01 3.20487559e-01 3.11614841e-01 -5.36684334e-01 -4.11479443e-01 6.42959833e-01 1.55817345e-01 5.44155538e-01 -8.18399131e-01 1.13585567e+00 6.53257179e+00 1.18279815e+00 -7.15851426e-01 -6.82398304e-02 2.99352854e-01 -9.13200155e-03 -4.15530801e-01 -1.85844883e-01 -9.91347849e-01 -9.58492793e-03 1.05181098e+00 -4.30555969e-01 4.24494177e-01 5.56737959e-01 -2.71593109e-02 1.36695012e-01 -8.55526567e-01 7.38761187e-01 1.80574000e-01 -9.53247547e-01 2.97374368e-01 -8.28349590e-02 5.57793140e-01 5.18307507e-01 1.02333643e-01 6.95998192e-01 5.76595128e-01 -8.98722589e-01 2.46040225e-01 3.05532485e-01 9.52477396e-01 -6.51238084e-01 9.24007833e-01 8.23142305e-02 -8.88708293e-01 9.37469900e-02 -7.65210927e-01 3.19443166e-01 -9.79054794e-02 4.76809770e-01 -4.61133927e-01 9.48172033e-01 6.78252578e-01 4.60084677e-01 -6.99907601e-01 7.54751027e-01 -1.42762423e-01 1.65390849e-01 -1.13898329e-01 5.45719527e-02 2.30574846e-01 -3.86769950e-01 5.84353089e-01 1.30883014e+00 2.29614928e-01 -1.29455626e-01 -1.60247639e-01 5.59167206e-01 -2.60095179e-01 7.91455209e-01 -6.96039855e-01 -3.29751313e-01 2.57763296e-01 1.51633859e+00 -3.14570606e-01 -1.27883449e-01 -6.34512961e-01 6.44718528e-01 5.51005244e-01 3.07710677e-01 -4.25180793e-01 -6.71331584e-01 5.66136658e-01 -1.83152527e-01 -1.38147563e-01 1.55216515e-01 1.96014732e-01 -1.29189575e+00 -1.87755138e-01 -9.20681179e-01 1.01342988e+00 -7.30884850e-01 -1.50766015e+00 9.37596321e-01 1.60195783e-01 -1.11902118e+00 -9.44208167e-03 -6.50007010e-01 -2.67990679e-01 1.06186831e+00 -1.66118813e+00 -1.03386748e+00 1.87737465e-01 6.48583651e-01 5.45418501e-01 -1.94151789e-01 1.24486434e+00 4.36243564e-01 -6.35773003e-01 5.97539842e-01 4.80424345e-01 1.04879752e-01 1.34515548e+00 -8.21684361e-01 9.01513323e-02 3.49490672e-01 5.84320188e-01 9.76721346e-01 4.27215487e-01 -4.59798515e-01 -1.12271762e+00 -7.31654704e-01 1.49230337e+00 -7.38295913e-01 1.14082527e+00 6.31723031e-02 -8.76477003e-01 5.85096419e-01 4.32049006e-01 -3.36975396e-01 7.91171670e-01 4.55453306e-01 -5.94866812e-01 2.74033565e-03 -9.76009548e-01 7.52293408e-01 7.45712399e-01 -6.90830827e-01 -1.20388615e+00 3.47408921e-01 8.47616136e-01 -3.60857695e-02 -1.32828414e+00 3.54529232e-01 4.45284545e-01 -4.59007859e-01 1.31223285e+00 -6.68721437e-01 4.40618068e-01 1.20555632e-01 -1.70423970e-01 -1.30615866e+00 -2.91687429e-01 -3.66692364e-01 3.28764975e-01 1.49519622e+00 5.98503828e-01 -7.06879437e-01 7.06280544e-02 4.84050035e-01 -1.01653956e-01 -4.06598389e-01 -7.98703611e-01 -5.82586944e-01 4.34673756e-01 -5.21746695e-01 3.66535902e-01 9.79311645e-01 5.08906603e-01 4.35764968e-01 1.03149414e-01 -2.61740685e-01 3.27171594e-01 5.56816198e-02 3.75379473e-01 -1.40081906e+00 -7.73211718e-02 -7.13815749e-01 -2.15864833e-02 -4.62616354e-01 4.73281235e-01 -1.37466502e+00 -1.05822496e-01 -1.41980243e+00 3.94808203e-01 -5.07730663e-01 -6.78830564e-01 4.22774851e-01 -2.16575354e-01 5.08400917e-01 2.47245178e-01 4.85145360e-01 -7.23128796e-01 2.72244871e-01 1.13836539e+00 -2.56772190e-01 -2.25365177e-01 -2.79436018e-02 -7.60124505e-01 4.30967301e-01 8.81667733e-01 -2.96004355e-01 -2.00364828e-01 -4.62728053e-01 4.28989440e-01 -1.63593233e-01 7.39825051e-03 -6.35188997e-01 5.66430129e-02 1.22192048e-01 -8.79990160e-02 -4.34908092e-01 9.26039666e-02 -6.96832538e-01 -8.72766525e-02 4.59838472e-02 -4.97182637e-01 1.42630637e-01 2.08671972e-01 1.71082333e-01 -5.16292572e-01 -5.24093211e-01 4.92693096e-01 -2.68547565e-01 -8.60764444e-01 2.52320319e-01 -3.24443221e-01 3.78774971e-01 6.15617394e-01 3.25689763e-01 -5.62587261e-01 -7.73039013e-02 -4.03822899e-01 5.30426092e-02 4.19346571e-01 8.89725447e-01 3.05291682e-01 -1.15326703e+00 -9.01493251e-01 -8.45691189e-02 8.26416016e-01 -2.64559627e-01 2.67653793e-01 6.54686213e-01 -5.33823490e-01 1.08826768e+00 6.22500479e-02 -8.86610746e-02 -1.33472455e+00 7.01852918e-01 -2.08732769e-01 -9.30809796e-01 -5.02651393e-01 2.50913292e-01 -7.53092393e-02 -8.53961885e-01 2.03589350e-01 4.13197242e-02 -7.65746057e-01 4.04832512e-01 7.66844034e-01 4.09633845e-01 4.09554183e-01 -7.46666849e-01 -9.48476493e-02 6.34999573e-01 -2.67875791e-01 -2.78446048e-01 1.28920186e+00 -1.85753033e-01 -5.33977509e-01 4.74885166e-01 1.31990099e+00 1.34678036e-01 -3.35507095e-02 -5.07180214e-01 3.68365288e-01 -1.29846811e-01 1.13349974e-01 -8.54970574e-01 -5.04730105e-01 6.17565572e-01 2.55175859e-01 2.18034405e-02 8.79619360e-01 2.21302480e-01 4.45296615e-01 5.96972942e-01 5.71963608e-01 -1.12225533e+00 -2.42350072e-01 9.14388597e-01 1.04074013e+00 -1.14896333e+00 -5.20331524e-02 -1.18518986e-01 -6.31092072e-01 9.92676735e-01 1.06858701e-01 2.68519372e-01 5.23926139e-01 -2.42998689e-01 3.12865645e-01 -2.90194511e-01 -6.05715334e-01 -3.90755355e-01 7.38804936e-01 2.56047338e-01 7.73060083e-01 -1.29580468e-01 -9.14538026e-01 7.02512503e-01 -1.93392381e-01 1.13036945e-01 9.07592922e-02 6.30964041e-01 -2.41880059e-01 -1.52223766e+00 -4.03616637e-01 2.99104154e-01 -7.51232326e-01 -4.56580371e-01 -5.73396206e-01 1.00775635e+00 -5.43385148e-01 1.00662303e+00 -9.50228944e-02 -1.91429332e-01 2.84744561e-01 3.33771318e-01 2.69259214e-01 -6.56865537e-01 -7.66816258e-01 -2.27963597e-01 4.03446376e-01 -2.19356790e-01 -3.01117748e-01 -4.22752559e-01 -6.10994875e-01 -2.58823991e-01 -3.53731006e-01 5.05130351e-01 5.57734132e-01 6.67355001e-01 1.38840020e-01 2.12476835e-01 4.84689206e-01 -5.57768464e-01 -7.77262390e-01 -1.25148237e+00 -6.95070863e-01 4.98391926e-01 -2.60379106e-01 -4.31212276e-01 -6.85378969e-01 -4.12216902e-01]
[11.357308387756348, 9.83283805847168]
664edaa2-06ac-4261-853e-ca0babf665b1
signal-enhancement-for-two-dimensional-cryo
2212.01421
null
https://arxiv.org/abs/2212.01421v1
https://arxiv.org/pdf/2212.01421v1.pdf
Signal enhancement for two-dimensional cryo-EM data processing
Different tasks in the computational pipeline of single-particle cryo-electron microscopy (cryo-EM) require enhancing the quality of the highly noisy raw images. To this end, we develop an efficient algorithm for signal enhancement of cryo-EM images. The enhanced images can be used for a variety of downstream tasks, such as 2-D classification, removing uninformative images, constructing {ab initio} models, generating templates for particle picking, providing a quick assessment of the data set, dimensionality reduction, and symmetry detection. The algorithm includes built-in quality measures to assess its performance and alleviate the risk of model bias. We demonstrate the effectiveness of the proposed algorithm on several experimental data sets. In particular, we show that the quality of the resulting images is high enough to produce ab initio models of $\sim 10$ \AA resolution. The algorithm is accompanied by a publicly available, documented and easy-to-use code.
['Tamir Bendory', 'Yoel Shkolnisky', 'Guy Sharon']
2022-12-02
null
null
null
null
['symmetry-detection']
['computer-vision']
[ 5.49745798e-01 -3.76582354e-01 7.12926865e-01 -4.98983115e-01 -1.17639458e+00 -5.14535308e-01 2.23909557e-01 3.76397759e-01 -6.64274573e-01 7.39263058e-01 -2.75784165e-01 -1.83934659e-01 -1.98491421e-02 -3.76212716e-01 -5.21008909e-01 -1.14958835e+00 3.74542177e-02 5.92031658e-01 2.07070187e-01 1.60691649e-01 6.17588222e-01 8.60327601e-01 -1.38111889e+00 2.33186245e-01 6.20356441e-01 8.26209843e-01 8.59899580e-01 5.57856679e-01 2.98892796e-01 1.91670179e-01 -4.21947986e-01 -5.88347502e-02 5.82561130e-03 -4.44695204e-01 -7.30152130e-01 8.04203227e-02 7.91171044e-02 -3.42228785e-02 1.88426822e-01 1.28293693e+00 7.73297369e-01 1.60689503e-01 5.16429663e-01 -5.29511213e-01 2.20223349e-02 -2.32325345e-01 -1.92210838e-01 5.12588441e-01 1.13407057e-02 5.28481722e-01 7.46188819e-01 -1.18649828e+00 1.18028486e+00 9.10051644e-01 2.73201317e-01 4.26152110e-01 -1.67952466e+00 -4.58870620e-01 -4.14369375e-01 1.85509995e-01 -8.63835454e-01 -5.59364378e-01 6.50961459e-01 -5.29786468e-01 8.33841085e-01 2.36223862e-01 3.36747885e-01 6.50702596e-01 2.94277549e-01 7.58540705e-02 1.49308300e+00 -4.59726065e-01 4.83536512e-01 -1.74478054e-01 3.62676501e-01 6.99088931e-01 9.13167968e-02 -9.33453143e-02 -4.17056412e-01 -4.65859562e-01 4.02866870e-01 -1.03157826e-01 -3.84237200e-01 -2.62660503e-01 -1.14270496e+00 4.75168228e-01 1.57020435e-01 8.41385126e-02 -5.19840598e-01 -5.12627184e-01 2.96197355e-01 -4.44027334e-02 3.47000867e-01 8.05087149e-01 -3.23587358e-01 1.09036518e-02 -8.70964527e-01 3.79089504e-01 2.18790658e-02 4.75983083e-01 8.05455148e-01 -4.77003574e-01 1.12457573e-01 6.45107031e-01 3.48055102e-02 3.39488208e-01 3.26492667e-01 -1.12947094e+00 7.91775584e-02 5.09608328e-01 2.40202427e-01 -6.27314031e-01 -6.54408693e-01 1.15484156e-01 -7.72704482e-01 5.25402248e-01 3.14478368e-01 7.49602169e-02 -8.75895262e-01 1.45999825e+00 5.67205191e-01 -3.25852454e-01 -1.29309133e-01 9.02359843e-01 6.93642855e-01 6.89960480e-01 -4.86181192e-02 -6.65981412e-01 1.41575778e+00 -3.64506692e-01 -5.33678710e-01 8.48183781e-02 3.79442513e-01 -8.28136861e-01 8.53244245e-01 4.21026796e-01 -1.16938269e+00 -5.15288591e-01 -1.04579127e+00 -1.83380738e-01 -9.18112770e-02 4.13271219e-01 2.32727751e-01 1.84398592e-01 -6.34864509e-01 1.11978769e+00 -1.11861706e+00 2.08273400e-02 4.47443634e-01 5.49439311e-01 -7.29459941e-01 5.56087494e-02 -4.91712242e-01 8.03281426e-01 3.13882500e-01 -1.38661101e-01 -8.30255866e-01 -7.20987439e-01 -5.43381333e-01 -4.46075723e-02 -3.21382843e-02 -4.73264694e-01 8.52138579e-01 -3.49579632e-01 -1.08855283e+00 1.21273768e+00 -5.27346909e-01 -3.18151265e-01 5.93859926e-02 2.93862224e-01 -1.07570313e-01 5.18697500e-01 1.74905822e-01 4.46659267e-01 7.67970145e-01 -1.08912325e+00 -4.60535824e-01 -8.52621257e-01 -5.67648232e-01 1.41978702e-02 2.62636691e-01 4.33037251e-01 -1.72621593e-01 -3.23966831e-01 4.90779936e-01 -8.22869718e-01 -3.82179469e-01 -1.65184096e-01 -3.24968636e-01 4.32075143e-01 8.65376294e-01 -7.56071031e-01 7.33001888e-01 -2.11601830e+00 1.97720453e-01 2.47676477e-01 3.71707410e-01 2.69431323e-01 8.88387114e-02 3.01176995e-01 -2.14776799e-01 -9.37985703e-02 -2.06174195e-01 -4.07248467e-01 -3.50056529e-01 -2.29547068e-01 9.03746262e-02 7.33217776e-01 3.74197662e-01 4.94495809e-01 -7.02147543e-01 -3.33995223e-01 4.49546963e-01 5.56607068e-01 -2.71490216e-01 3.47707331e-01 -6.43848721e-03 9.46074367e-01 -1.72588751e-01 4.11782712e-01 7.59700358e-01 -4.57741320e-01 4.50065464e-01 -4.01919305e-01 -3.63352269e-01 5.58542073e-01 -1.06926644e+00 1.28131402e+00 4.59081084e-02 5.02029240e-01 5.86655498e-01 -7.44170904e-01 8.51252198e-01 -8.60464852e-03 4.75860804e-01 -6.20184481e-01 1.10342212e-01 2.01354384e-01 1.98371336e-01 -3.51200938e-01 4.19172496e-01 -3.21141660e-01 4.17994738e-01 5.90664029e-01 1.60970986e-01 -2.36338973e-01 4.64168459e-01 1.02045260e-01 1.07178056e+00 -6.67568445e-02 1.15964085e-01 -5.98637521e-01 4.75713730e-01 2.35221580e-01 5.47188103e-01 4.33608145e-01 -1.87860966e-01 7.93737471e-01 2.41727620e-01 -4.47907716e-01 -1.61643815e+00 -8.00100327e-01 -4.30605233e-01 6.50279760e-01 8.68318751e-02 -4.06995803e-01 -9.04447436e-01 -1.25533491e-01 -3.61004949e-01 8.95838439e-02 -2.50994235e-01 1.16508864e-02 -4.83699620e-01 -1.28840780e+00 -5.75861298e-02 2.16837630e-01 -2.01529227e-02 -1.23514688e+00 -6.78680718e-01 2.11474985e-01 -1.70172364e-01 -1.03733683e+00 -2.22205997e-01 6.25471056e-01 -1.02254856e+00 -1.15764105e+00 -2.43677720e-01 -5.71342111e-01 1.16555345e+00 3.69924337e-01 8.94035161e-01 1.78548157e-01 -6.98544741e-01 -2.06772581e-01 -5.51989041e-02 -5.25738448e-02 -4.75146055e-01 -3.31772715e-01 3.21668386e-01 -2.23211035e-01 2.24803656e-01 -5.20170093e-01 -6.30731821e-01 2.49610573e-01 -8.15855145e-01 1.28218055e-01 3.23677868e-01 1.21376848e+00 1.25186300e+00 1.76517636e-01 2.67722815e-01 -7.29289591e-01 6.48761690e-01 1.41176656e-01 -8.70317280e-01 -9.32548046e-02 -5.44652581e-01 1.96654618e-01 8.40087295e-01 -3.46731208e-02 -9.67194319e-01 3.41335267e-01 -2.92351872e-01 -1.72769815e-01 -3.40800315e-01 1.78276122e-01 -2.54424095e-01 -2.72256434e-01 5.83900988e-01 2.90953338e-01 2.84344971e-01 -6.38897955e-01 -7.93711320e-02 5.29878378e-01 7.45296538e-01 -5.05410016e-01 5.09483278e-01 7.84916699e-01 2.36108184e-01 -9.64548826e-01 -4.30577606e-01 -7.19901919e-01 -7.94039965e-01 -4.02631648e-02 7.97082245e-01 -6.00864649e-01 -8.59771430e-01 3.84027898e-01 -1.01151061e+00 1.87204536e-02 1.97179273e-01 4.11127537e-01 -6.22734606e-01 7.28463829e-01 -7.43073642e-01 -7.95831203e-01 -6.16248906e-01 -1.72832406e+00 1.20726299e+00 1.90165207e-01 -1.62164539e-01 -5.47109842e-01 7.99650997e-02 5.88753402e-01 -4.09705788e-02 2.11137794e-02 1.18244994e+00 -4.52262104e-01 -5.15068650e-01 -7.75531903e-02 -8.21561813e-02 2.28329539e-01 -1.94912776e-02 2.26016209e-01 -9.74455833e-01 -5.10109484e-01 2.25729883e-01 -2.92223424e-01 8.53296280e-01 4.91766334e-01 1.15586364e+00 1.88831151e-01 -4.21311080e-01 6.25233591e-01 1.26259911e+00 2.19897136e-01 7.39486516e-01 3.41908306e-01 3.01515967e-01 6.09376073e-01 8.19589198e-01 4.33414936e-01 -2.78133512e-01 7.44506657e-01 3.73610020e-01 -1.53565601e-01 2.19145223e-01 2.34598577e-01 1.49644002e-01 5.21523297e-01 -1.53175414e-01 1.00146972e-01 -7.08173275e-01 3.07404310e-01 -1.63062811e+00 -1.17651939e+00 -3.68686140e-01 2.24573159e+00 7.30364859e-01 9.09861401e-02 9.38441455e-02 2.50761151e-01 8.13938558e-01 -2.47398823e-01 -4.75233197e-01 -5.69427833e-02 -6.45791441e-02 3.75555098e-01 2.48719811e-01 3.96137744e-01 -1.08920622e+00 6.29868031e-01 6.70142365e+00 5.73689997e-01 -1.10297525e+00 -1.66739359e-01 6.81018174e-01 -1.17394209e-01 1.21954903e-01 -9.97559167e-03 -9.21388209e-01 7.33098984e-01 8.75763118e-01 -4.40320931e-02 2.80157864e-01 7.34304428e-01 8.88566613e-01 -3.73216510e-01 -8.28269899e-01 9.74607766e-01 -3.94515425e-01 -1.78246319e+00 -1.96462706e-01 1.00142390e-01 3.08207721e-01 1.21776991e-01 -3.07077020e-01 -4.56287742e-01 -5.65097891e-02 -8.81345868e-01 3.47215205e-01 3.86867017e-01 7.76127458e-01 -8.93634796e-01 6.69674218e-01 3.39474887e-01 -7.61233270e-01 2.31988147e-01 -6.71071291e-01 1.63122877e-01 3.86607647e-01 1.00729311e+00 -6.88300908e-01 4.23505336e-01 6.92402422e-01 1.89233318e-01 -3.96721750e-01 7.89583266e-01 1.53786033e-01 2.19544515e-01 -3.82336795e-01 1.53293595e-01 -1.28059492e-01 -5.27925432e-01 5.20139635e-01 1.08791792e+00 6.96666315e-02 3.61172199e-01 1.01395100e-01 9.34783101e-01 -9.43368822e-02 1.85967106e-02 -2.55776227e-01 -7.70531967e-02 5.07200181e-01 1.62720644e+00 -1.02626514e+00 -1.98679477e-01 -5.67952059e-02 7.33984888e-01 3.69178474e-01 -5.62205017e-02 -2.87861288e-01 -3.56532812e-01 7.88211048e-01 2.96571821e-01 3.67292225e-01 -3.03694218e-01 -1.82398751e-01 -9.97307777e-01 1.55860633e-01 -8.70832562e-01 1.03062093e-01 -8.97238076e-01 -1.16024172e+00 4.97266799e-01 -2.62320489e-01 -8.44601989e-01 -9.96507630e-02 -9.76750016e-01 -6.29743755e-01 9.19590116e-01 -1.16747808e+00 -4.58758622e-01 -3.21520835e-01 1.94494605e-01 1.56338140e-01 1.39075536e-02 8.47463667e-01 2.01312006e-01 -5.69249630e-01 -1.53192788e-01 5.03383458e-01 -2.51923144e-01 6.51306868e-01 -1.11443865e+00 3.34781617e-01 9.72842932e-01 -2.12565050e-01 7.04221606e-01 9.35849011e-01 -6.94607496e-01 -1.38390994e+00 -8.11856925e-01 6.19480908e-01 -3.77663732e-01 3.98413211e-01 -3.31245214e-01 -1.05289054e+00 1.66071147e-01 -2.61363059e-01 1.23833925e-01 8.00873756e-01 -4.38398570e-01 1.84210032e-01 2.22351417e-01 -1.50587964e+00 2.17856795e-01 6.36043847e-01 -4.12611902e-01 -3.61658990e-01 4.23300356e-01 1.58112720e-02 -2.94182688e-01 -9.79973614e-01 2.03464389e-01 3.03295285e-01 -1.16745102e+00 9.83913839e-01 -5.90087950e-01 3.66071612e-01 -5.39627075e-01 9.56418067e-02 -1.06435180e+00 -5.19728243e-01 -5.15321016e-01 3.88947845e-01 7.39361525e-01 3.54561508e-01 -1.99653059e-01 7.87082434e-01 5.79665065e-01 -3.01800579e-01 -5.81376076e-01 -9.75464106e-01 -3.39775711e-01 -2.84870058e-01 -6.64772466e-02 2.31200233e-01 5.80866754e-01 2.24071339e-01 3.39912415e-01 -1.27835184e-01 3.23618561e-01 9.60584581e-01 5.14019012e-01 6.18866622e-01 -1.34146309e+00 -1.17215306e-01 1.24772526e-02 -4.84869689e-01 -7.20047951e-01 -1.47076976e-03 -6.71358705e-01 2.24025548e-01 -1.14604521e+00 6.72287226e-01 -2.02697024e-01 5.05241454e-02 2.20709428e-01 -3.16707313e-01 3.80947590e-01 2.85347868e-02 4.61957753e-01 -6.20571375e-01 3.76080573e-01 1.07724190e+00 2.68921927e-02 -2.68683434e-02 -2.39464909e-01 -3.78446460e-01 5.20648658e-01 7.99523652e-01 -5.58882833e-01 1.00013651e-01 1.34164020e-01 -1.42750099e-01 -2.22269520e-01 5.05867481e-01 -9.97652769e-01 -2.70023979e-02 -6.92525646e-03 5.28787673e-01 -8.65116537e-01 5.23752034e-01 -5.42858005e-01 3.17986608e-01 3.63703102e-01 -3.85883846e-03 1.10354848e-01 1.01399444e-01 1.71157271e-01 -1.46317735e-01 -5.81757247e-01 1.42024171e+00 -4.71942037e-01 -3.68697792e-01 1.85890079e-01 -3.82609159e-01 -1.77069768e-01 8.27750146e-01 -4.25160155e-02 -3.58624965e-01 -1.40912803e-02 -9.35985863e-01 9.62941572e-02 1.12118936e+00 -3.96999389e-01 6.63088143e-01 -8.69885445e-01 -3.82581919e-01 4.21979636e-01 5.19777834e-02 -3.56672220e-02 3.53357166e-01 1.00296521e+00 -6.74099565e-01 2.91393310e-01 -4.90918845e-01 -6.75924838e-01 -1.75623345e+00 4.59970772e-01 3.41459721e-01 -3.26136559e-01 -6.91929102e-01 4.71563160e-01 7.68267661e-02 -2.88971335e-01 -1.96538970e-01 -1.91925496e-01 5.22175990e-03 -2.97887355e-01 9.30194974e-01 3.55686128e-01 5.54010272e-01 -6.49217308e-01 -3.85730594e-01 2.90916860e-01 -2.33584568e-01 5.86050116e-02 1.86404586e+00 -1.93415552e-01 -2.92744130e-01 5.83014973e-02 1.00772929e+00 -2.28880882e-01 -1.32295120e+00 2.24829406e-01 -1.78167775e-01 -3.93476754e-01 2.03238592e-01 -3.55626643e-01 -6.22110963e-01 9.24767911e-01 6.44419611e-01 -6.79970533e-02 1.00098336e+00 -2.77196299e-02 5.01743734e-01 4.95591462e-01 4.12496597e-01 -1.25519586e+00 -2.24960878e-01 3.16818446e-01 3.64347905e-01 -1.25935996e+00 3.31537455e-01 -3.52359861e-01 -3.31853002e-01 1.08502781e+00 1.76584989e-01 1.68166980e-01 2.11683795e-01 4.32327747e-01 5.70054352e-02 -6.07650280e-01 -8.67609739e-01 -1.06365144e-01 -1.28108487e-01 7.83790767e-01 2.59924769e-01 -2.18209192e-01 -2.93919891e-01 4.06526953e-01 3.74439657e-02 -1.26592830e-01 6.06920481e-01 1.00989580e+00 -6.71019971e-01 -1.12057102e+00 -5.28165996e-01 4.86556113e-01 -7.82761991e-01 -1.06550321e-01 -4.37330902e-01 2.31567100e-01 -1.74062192e-01 6.59633696e-01 -1.21499233e-01 -8.33704919e-02 1.03004932e-01 2.34564558e-01 5.55883586e-01 -5.04219592e-01 -3.68915290e-01 2.03405574e-01 -6.01897249e-04 -3.72435123e-01 -5.35437047e-01 -5.94429970e-01 -1.52450562e+00 -3.03281397e-01 -5.68603396e-01 2.83426493e-01 7.77693748e-01 8.09787452e-01 8.01416337e-01 1.94237366e-01 5.64480603e-01 -1.38056529e+00 -3.29237431e-01 -8.50210905e-01 -5.60361385e-01 7.77954578e-01 1.10440895e-01 -7.40496635e-01 -4.37758595e-01 3.00526142e-01]
[13.393728256225586, -3.069051742553711]
87df80cd-2459-4309-a7c3-15fd1ffdb4b3
pmp-net-point-cloud-completion-by-learning
2012.03408
null
https://arxiv.org/abs/2012.03408v3
https://arxiv.org/pdf/2012.03408v3.pdf
PMP-Net: Point Cloud Completion by Learning Multi-step Point Moving Paths
The task of point cloud completion aims to predict the missing part for an incomplete 3D shape. A widely used strategy is to generate a complete point cloud from the incomplete one. However, the unordered nature of point clouds will degrade the generation of high-quality 3D shapes, as the detailed topology and structure of discrete points are hard to be captured by the generative process only using a latent code. In this paper, we address the above problem by reconsidering the completion task from a new perspective, where we formulate the prediction as a point cloud deformation process. Specifically, we design a novel neural network, named PMP-Net, to mimic the behavior of an earth mover. It moves each point of the incomplete input to complete the point cloud, where the total distance of point moving paths (PMP) should be shortest. Therefore, PMP-Net predicts a unique point moving path for each point according to the constraint of total point moving distances. As a result, the network learns a strict and unique correspondence on point-level, which can capture the detailed topology and structure relationships between the incomplete shape and the complete target, and thus improves the quality of the predicted complete shape. We conduct comprehensive experiments on Completion3D and PCN datasets, which demonstrate our advantages over the state-of-the-art point cloud completion methods.
['Yu-Shen Liu', 'Wen Zheng', 'Pengfei Wan', 'Yan-Pei Cao', 'Zhizhong Han', 'Peng Xiang', 'Xin Wen']
2020-12-07
null
http://openaccess.thecvf.com//content/CVPR2021/html/Wen_PMP-Net_Point_Cloud_Completion_by_Learning_Multi-Step_Point_Moving_Paths_CVPR_2021_paper.html
http://openaccess.thecvf.com//content/CVPR2021/papers/Wen_PMP-Net_Point_Cloud_Completion_by_Learning_Multi-Step_Point_Moving_Paths_CVPR_2021_paper.pdf
cvpr-2021-1
['point-cloud-completion']
['computer-vision']
[-1.47029892e-01 -2.95195859e-02 1.57611415e-01 -1.25760198e-01 -3.66010994e-01 -5.12588978e-01 3.75429720e-01 -2.81341970e-01 2.11615562e-01 3.22415382e-01 -1.75603017e-01 -1.59593374e-01 -7.88348019e-02 -1.13275468e+00 -1.01286042e+00 -5.97581744e-01 1.62048057e-01 1.08668303e+00 1.33380219e-01 -2.00253233e-01 6.88515827e-02 9.86771643e-01 -1.24878740e+00 -3.24874073e-02 1.01101303e+00 7.07529545e-01 5.26774764e-01 1.11020528e-01 -3.39480519e-01 -2.34218001e-01 -2.70482332e-01 -1.65075138e-01 4.24922943e-01 1.41550064e-01 -4.22653705e-01 1.95096970e-01 2.42755726e-01 -4.22663897e-01 -5.58033645e-01 9.03884649e-01 1.19449921e-01 -4.45001852e-03 7.77311265e-01 -1.36017799e+00 -8.02146673e-01 1.55240178e-01 -5.36753833e-01 -6.98029101e-01 1.30747408e-01 3.14860679e-02 7.11265028e-01 -1.08579266e+00 7.49629617e-01 1.11197281e+00 6.82695508e-01 4.58851695e-01 -1.23817945e+00 -7.48626173e-01 2.16462791e-01 -2.75899738e-01 -1.59842694e+00 1.40258844e-03 1.24263608e+00 -5.08230567e-01 5.36436677e-01 1.98338762e-01 9.42609429e-01 7.68161297e-01 1.98615238e-01 5.34020126e-01 3.79245311e-01 4.63878289e-02 1.65705323e-01 -4.54766989e-01 -4.43542331e-01 4.40684170e-01 1.57863960e-01 1.68889299e-01 -3.67479511e-02 -3.61143827e-01 1.31183290e+00 6.16080761e-01 -3.21095824e-01 -7.99671173e-01 -1.37058663e+00 4.87504452e-01 7.51540124e-01 1.10430561e-01 -4.99034435e-01 1.21027388e-01 -2.38376081e-01 -1.34359360e-01 6.07314169e-01 -1.92352254e-02 -5.17181218e-01 5.04683442e-02 -8.78926516e-01 5.33208430e-01 6.50344312e-01 1.42642450e+00 1.09506392e+00 -3.08624338e-02 1.59724548e-01 4.77922559e-01 5.47255695e-01 6.85921490e-01 -8.32248181e-02 -1.00216043e+00 7.19871938e-01 1.09819698e+00 2.18893141e-01 -1.47406232e+00 -2.45773226e-01 -5.16632199e-01 -1.33080089e+00 2.13945046e-01 -8.78752151e-04 6.51260838e-02 -9.13451195e-01 1.43180645e+00 4.54538614e-01 5.05376935e-01 -1.82681277e-01 8.95895779e-01 5.76769412e-01 1.05947745e+00 -4.08387750e-01 -7.87156373e-02 8.52364063e-01 -6.13248408e-01 -1.85921013e-01 4.12446372e-02 2.89957106e-01 -6.38707757e-01 8.45540643e-01 9.15519819e-02 -1.03258109e+00 -7.00696826e-01 -8.96229565e-01 -2.45561055e-03 2.11426705e-01 1.09551750e-01 3.99966270e-01 -2.02110991e-01 -7.39881754e-01 8.59941542e-01 -1.06441700e+00 2.49461174e-01 4.75293785e-01 2.71095037e-01 -3.37488562e-01 -2.94887662e-01 -6.04806006e-01 3.63846213e-01 1.94129482e-01 3.54600370e-01 -7.02840567e-01 -1.02645814e+00 -6.02361202e-01 2.70446211e-01 1.46407127e-01 -1.07895911e+00 9.16575909e-01 -4.05953526e-01 -1.10626853e+00 6.39431119e-01 -2.21974373e-01 1.06476210e-01 5.97425163e-01 7.10878745e-02 -1.03117913e-01 -1.75894231e-01 1.61594346e-01 8.00155401e-01 6.90385103e-01 -1.74542999e+00 -3.79167467e-01 -4.82663542e-01 -2.21822143e-01 2.67509550e-01 1.86028510e-01 -6.39961779e-01 -7.73761868e-01 -6.44548893e-01 8.24598551e-01 -1.12962794e+00 -3.22005153e-01 2.34546781e-01 -4.89175290e-01 -3.77189875e-01 1.12834322e+00 -5.63423514e-01 7.79657602e-01 -2.19784284e+00 3.03295523e-01 3.18749160e-01 3.76149118e-01 -6.27249554e-02 -1.59879550e-01 5.47396660e-01 -7.09232837e-02 2.61550963e-01 -6.36254013e-01 -6.55337512e-01 -3.24426144e-02 5.38031816e-01 -5.68951428e-01 2.76612043e-01 4.41894457e-02 9.99339521e-01 -8.48587215e-01 -7.60773346e-02 1.49546757e-01 5.75453103e-01 -6.20632410e-01 2.00640932e-01 -5.38047612e-01 5.87965906e-01 -8.15428019e-01 6.77566886e-01 1.18485188e+00 -2.01026797e-01 -2.64318585e-01 -2.64520258e-01 -2.38946468e-01 -1.53178334e-01 -1.22826242e+00 2.14303017e+00 -2.82479197e-01 9.51319486e-02 8.78810696e-03 -6.56384349e-01 1.46801579e+00 2.82032341e-01 8.27218175e-01 -3.20347220e-01 -2.87689596e-01 2.88025439e-01 -2.75786728e-01 -5.58285639e-02 4.77784872e-01 -2.62156725e-01 6.26959726e-02 2.29003906e-01 -5.39692938e-01 -4.83285457e-01 -4.46464002e-01 1.14864781e-01 7.18028367e-01 5.10622382e-01 -2.71044165e-01 1.30599052e-01 2.59181976e-01 2.53716409e-01 8.47411156e-01 2.21999824e-01 3.73641849e-01 1.13377917e+00 3.02481532e-01 -6.48378313e-01 -1.48569214e+00 -1.26768136e+00 -8.81038792e-03 6.39213175e-02 5.31117857e-01 -2.23077178e-01 -4.45429534e-01 -4.25726175e-01 1.83877051e-01 5.81979156e-01 -3.11595589e-01 -1.99039847e-01 -8.32195282e-01 -3.24480414e-01 -2.74449084e-02 4.52148020e-01 3.34631503e-01 -1.07334542e+00 -1.45307139e-01 3.20026726e-01 -2.94354558e-01 -8.89487207e-01 -6.19575322e-01 -3.89809340e-01 -1.31280613e+00 -9.55855668e-01 -7.21171796e-01 -8.78607213e-01 1.10754907e+00 3.60258698e-01 1.05729568e+00 4.96359706e-01 2.02532724e-01 -4.24028002e-02 -2.15118229e-01 -2.09801659e-01 -3.05377424e-01 1.25939965e-01 7.24068284e-02 -4.29111421e-02 6.64197654e-02 -1.10880792e+00 -6.36974216e-01 4.83748883e-01 -9.60516095e-01 5.49999714e-01 6.33086026e-01 5.56702793e-01 1.18079066e+00 2.82646894e-01 1.50645375e-01 -3.50512117e-01 3.13877016e-01 -4.60256666e-01 -6.42240286e-01 -1.91801041e-02 -1.73839346e-01 -1.39866695e-01 6.49145246e-01 -3.86790067e-01 -7.64274895e-01 4.87577319e-01 -2.42237911e-01 -1.12797499e+00 -1.72492191e-01 4.88247931e-01 -5.13855159e-01 -6.76544383e-02 2.37864152e-01 6.34982407e-01 3.29228714e-02 -9.04784560e-01 1.92243144e-01 2.51340449e-01 6.03053451e-01 -6.76415384e-01 1.37756860e+00 6.68255866e-01 3.53209496e-01 -5.92600346e-01 -4.02605325e-01 -2.53336906e-01 -1.08775389e+00 -6.59200102e-02 5.83334029e-01 -8.18103611e-01 -7.88617492e-01 4.49437559e-01 -1.64467669e+00 -1.21975407e-01 -2.77654976e-01 2.62796164e-01 -6.29356623e-01 4.31240708e-01 -2.91857392e-01 -3.65284652e-01 -2.91227072e-01 -1.07848477e+00 1.26889956e+00 9.98077635e-03 1.49547786e-01 -7.29647636e-01 2.42622510e-01 -1.07849799e-01 -9.29408744e-02 5.66546500e-01 1.04233527e+00 -4.54224013e-02 -1.18289030e+00 -4.26768124e-01 -3.13143730e-02 1.36630803e-01 1.70509592e-01 2.59666771e-01 -3.46674651e-01 -2.20945239e-01 1.66742906e-01 4.29072678e-01 3.75051051e-01 3.85328621e-01 1.38405919e+00 -3.28889072e-01 -6.40464783e-01 1.02871621e+00 1.47187257e+00 7.02912211e-02 6.61451221e-01 9.90673900e-02 1.02577996e+00 4.11687285e-01 4.63414252e-01 4.04438317e-01 5.23406506e-01 7.22897112e-01 9.59570467e-01 4.11603078e-02 1.77838206e-01 -9.35504436e-01 -1.85372949e-01 1.07606530e+00 -2.33164594e-01 -5.25400750e-02 -1.18219471e+00 4.15305197e-01 -1.92655969e+00 -6.90911591e-01 -3.22897196e-01 2.20825911e+00 3.33845139e-01 -2.39340458e-02 -2.54551500e-01 -1.11503914e-01 8.43053222e-01 1.32573741e-02 -7.20744073e-01 2.51946986e-01 1.52202785e-01 -1.68755308e-01 1.17910780e-01 2.74391949e-01 -6.01296723e-01 7.46900201e-01 5.37704182e+00 8.30013275e-01 -9.78297651e-01 -2.62610435e-01 2.96241939e-01 8.60832706e-02 -6.11332774e-01 3.41191977e-01 -6.90583646e-01 6.97544992e-01 1.35505930e-01 -1.95155531e-01 4.15729135e-01 9.92150903e-01 2.87380010e-01 6.08451426e-01 -1.10245550e+00 1.02302980e+00 -2.07623765e-01 -1.58914661e+00 5.38595319e-01 4.13015097e-01 8.51860404e-01 1.17504358e-01 -1.60965905e-01 8.14297795e-02 2.04472467e-02 -8.94975662e-01 6.68088436e-01 9.11808431e-01 8.90819192e-01 -9.00253415e-01 2.99954295e-01 1.12180293e+00 -1.36037421e+00 3.05223018e-01 -7.94035971e-01 -9.34840832e-03 4.25507605e-01 7.21707821e-01 -5.53205371e-01 9.40259457e-01 5.57652771e-01 9.49300230e-01 -1.01537384e-01 1.21749318e+00 -1.80794284e-01 1.70161843e-01 -5.28436184e-01 2.99583673e-01 8.92367363e-02 -7.84125388e-01 8.97133112e-01 3.66646945e-01 8.32657397e-01 3.84745061e-01 4.12757933e-01 1.36981893e+00 -2.19057634e-01 -1.51345342e-01 -7.04320371e-01 3.10423464e-01 7.55638361e-01 1.23928249e+00 -5.26012897e-01 -6.42501116e-02 -1.09916151e-01 6.80532813e-01 3.31749439e-01 3.91309500e-01 -5.93815207e-01 -1.40180858e-02 6.15877807e-01 3.87636691e-01 2.70632565e-01 -6.92427695e-01 -5.50837815e-01 -1.13462985e+00 4.22799289e-01 -3.03044319e-01 -1.45626381e-01 -1.18680084e+00 -1.23626602e+00 5.75815380e-01 -6.79459125e-02 -1.86691868e+00 1.29312545e-01 -2.47022629e-01 -1.04334807e+00 1.09749830e+00 -1.18037093e+00 -1.36670041e+00 -5.86045980e-01 4.02450383e-01 4.08575505e-01 1.04407869e-01 6.29064202e-01 1.23582333e-01 -2.40127146e-01 2.97657936e-03 1.31986648e-01 4.34272364e-02 1.28283143e-01 -8.83694470e-01 8.43424082e-01 6.07506037e-01 7.95983896e-02 6.01026058e-01 2.39030614e-01 -1.12051606e+00 -1.51517642e+00 -1.32960892e+00 6.70052588e-01 -4.80572820e-01 1.67658210e-01 -3.34890008e-01 -1.20008647e+00 7.07202673e-01 -5.72531641e-01 3.96713912e-02 9.22744628e-03 -1.46203727e-01 4.42929976e-02 -5.99391460e-02 -9.32143569e-01 7.05690861e-01 1.30300426e+00 -1.48519933e-01 -6.88954473e-01 3.08018088e-01 1.08915639e+00 -7.45621622e-01 -8.80341709e-01 6.11405015e-01 2.48302430e-01 -6.31370068e-01 1.22826517e+00 -3.66978526e-01 7.49410450e-01 -6.32313669e-01 -8.00975971e-03 -1.47019958e+00 -4.73981172e-01 -4.81459498e-01 -2.49611497e-01 1.14068139e+00 5.94491474e-02 -4.96425867e-01 1.24725735e+00 6.08627558e-01 -5.92077017e-01 -1.08331668e+00 -1.00474381e+00 -6.88377380e-01 3.23455215e-01 -4.17338014e-01 1.24283636e+00 8.31022382e-01 -6.48945808e-01 -3.86363715e-02 -3.54642421e-01 5.29865623e-01 6.92718863e-01 6.24121189e-01 1.04848325e+00 -1.53230178e+00 6.04184531e-02 -1.95482060e-01 -2.53166229e-01 -1.64314651e+00 1.33406341e-01 -9.21246946e-01 2.42569260e-02 -1.69526684e+00 -1.03293307e-01 -1.01360559e+00 3.54900032e-01 3.71390402e-01 5.61931543e-02 -2.64446158e-02 3.38773698e-01 8.38856101e-01 8.49380791e-02 1.08834589e+00 1.78561544e+00 -1.14095189e-01 -4.00865853e-01 3.34066927e-01 -4.02195930e-01 8.55602503e-01 5.17881155e-01 -6.08578920e-01 -2.63743132e-01 -8.26582432e-01 3.68410289e-01 4.17441130e-01 5.60920119e-01 -9.35414791e-01 4.82357204e-01 -4.38964903e-01 5.22709429e-01 -1.45323181e+00 5.95864356e-01 -1.29311967e+00 7.24535108e-01 3.37422371e-01 2.87704587e-01 2.31464222e-01 3.01331393e-02 7.55917430e-01 -1.12994120e-01 -4.34846580e-02 3.18056345e-01 -2.70006180e-01 -2.76587605e-01 1.26576746e+00 4.01530892e-01 -3.68108451e-01 9.53660429e-01 -4.56612706e-01 -9.99531075e-02 -6.04926683e-02 -6.44897938e-01 4.71630335e-01 9.63059366e-01 5.90099335e-01 1.04247046e+00 -1.73645198e+00 -7.50499189e-01 4.75630373e-01 1.76201649e-02 1.20662558e+00 5.00609994e-01 3.06490064e-01 -6.92420602e-01 1.39865145e-01 -5.38753830e-02 -9.44247663e-01 -7.48921871e-01 5.80616236e-01 3.16530257e-01 1.23165525e-01 -1.16581166e+00 2.55585909e-01 3.87489259e-01 -9.44763303e-01 -1.65226042e-01 -3.41204822e-01 1.02058306e-01 -4.63915944e-01 1.32106626e-02 2.65115827e-01 -6.45620599e-02 -7.03949153e-01 -3.17886323e-02 9.71427560e-01 2.34758988e-01 2.35369712e-01 1.63635206e+00 3.43107991e-02 -2.46515885e-01 2.33511046e-01 1.09409440e+00 -4.57929373e-02 -1.51247406e+00 -1.90377846e-01 -5.59762418e-01 -7.01257706e-01 -1.74604148e-01 -3.09673101e-01 -1.15076065e+00 8.02169919e-01 4.66508372e-03 3.34358425e-03 7.42552578e-01 5.31875528e-02 1.05375445e+00 2.85341263e-01 5.68922520e-01 -5.77731013e-01 -1.77938521e-01 5.44031739e-01 1.22810066e+00 -6.34051919e-01 -3.64624970e-02 -7.14391351e-01 -2.92391896e-01 1.14672518e+00 7.54609942e-01 -4.19220299e-01 7.40182400e-01 -2.06867576e-01 -3.57455641e-01 -3.63740593e-01 -4.82796937e-01 4.71732318e-01 4.14379269e-01 5.84360301e-01 -3.39639604e-01 2.98278004e-01 1.74446538e-01 4.62506741e-01 -4.79884207e-01 -1.24264704e-02 2.94671744e-01 7.05376625e-01 -2.67871350e-01 -1.17601931e+00 -5.43505192e-01 3.92894506e-01 3.49860221e-01 1.88923433e-01 -1.47870541e-01 6.22216463e-01 2.43537486e-01 2.92711347e-01 4.03335899e-01 -4.71061707e-01 5.70334554e-01 -2.77176291e-01 1.72870010e-02 -7.34753430e-01 9.49737281e-02 2.30737757e-02 -6.36948466e-01 -3.47040415e-01 -9.81901363e-02 -5.85468590e-01 -1.42469537e+00 -4.53460813e-01 -6.11093342e-02 1.65923879e-01 7.68683910e-01 7.12769628e-01 7.01926053e-01 2.64508784e-01 9.16960001e-01 -1.28150737e+00 -4.90342021e-01 -6.59031570e-01 -6.27463102e-01 5.31644404e-01 1.42533183e-01 -6.47669971e-01 -2.76728451e-01 -9.58854184e-02]
[8.39932632446289, -3.5729141235351562]
cc5550f3-9b43-4aaf-b9e3-9bcbfb4dbab2
near-optimal-heteroscedastic-regression-with
2306.14288
null
https://arxiv.org/abs/2306.14288v2
https://arxiv.org/pdf/2306.14288v2.pdf
Near Optimal Heteroscedastic Regression with Symbiotic Learning
We consider the problem of heteroscedastic linear regression, where, given $n$ samples $(\mathbf{x}_i, y_i)$ from $y_i = \langle \mathbf{w}^{*}, \mathbf{x}_i \rangle + \epsilon_i \cdot \langle \mathbf{f}^{*}, \mathbf{x}_i \rangle$ with $\mathbf{x}_i \sim N(0,\mathbf{I})$, $\epsilon_i \sim N(0,1)$, we aim to estimate $\mathbf{w}^{*}$. Beyond classical applications of such models in statistics, econometrics, time series analysis etc., it is also particularly relevant in machine learning when data is collected from multiple sources of varying but apriori unknown quality. Our work shows that we can estimate $\mathbf{w}^{*}$ in squared norm up to an error of $\tilde{O}\left(\|\mathbf{f}^{*}\|^2 \cdot \left(\frac{1}{n} + \left(\frac{d}{n}\right)^2\right)\right)$ and prove a matching lower bound (upto log factors). This represents a substantial improvement upon the previous best known upper bound of $\tilde{O}\left(\|\mathbf{f}^{*}\|^2\cdot \frac{d}{n}\right)$. Our algorithm is an alternating minimization procedure with two key subroutines 1. An adaptation of the classical weighted least squares heuristic to estimate $\mathbf{w}^{*}$, for which we provide the first non-asymptotic guarantee. 2. A nonconvex pseudogradient descent procedure for estimating $\mathbf{f}^{*}$ inspired by phase retrieval. As corollaries, we obtain fast non-asymptotic rates for two important problems, linear regression with multiplicative noise and phase retrieval with multiplicative noise, both of which are of independent interest. Beyond this, the proof of our lower bound, which involves a novel adaptation of LeCam's method for handling infinite mutual information quantities (thereby preventing a direct application of standard techniques like Fano's method), could also be of broader interest for establishing lower bounds for other heteroscedastic or heavy-tailed statistical problems.
['Praneeth Netrapalli', 'Dheeraj Nagaraj', 'Aniket Das', 'Dheeraj Baby']
2023-06-25
null
null
null
null
['retrieval', 'econometrics', 'time-series']
['methodology', 'miscellaneous', 'time-series']
[ 3.22550982e-01 1.77633658e-01 9.27419215e-02 -2.35861093e-01 -1.31848598e+00 -6.08715594e-01 -6.09100536e-02 3.59129049e-02 -8.59190702e-01 1.11440122e+00 -5.48945725e-01 -4.72237051e-01 -1.00204456e+00 -9.60211515e-01 -9.70053852e-01 -1.22632897e+00 -7.69202232e-01 3.71347576e-01 -1.31525531e-01 -3.33972871e-01 1.90681025e-01 3.76452178e-01 -1.82776964e+00 -4.73067164e-01 9.64052439e-01 1.63501608e+00 -6.50450736e-02 7.24064469e-01 9.72732976e-02 7.03634441e-01 -5.35380363e-01 -3.55422378e-01 4.35201943e-01 -7.34536946e-01 -6.08161569e-01 -2.70223349e-01 2.09891021e-01 1.46775573e-01 2.07809592e-03 1.54406798e+00 3.09720039e-01 3.97030234e-01 1.02902138e+00 -8.88256967e-01 -2.88762242e-01 3.98646027e-01 -1.01338756e+00 1.27247617e-01 8.51197243e-02 -2.20300794e-01 8.27717900e-01 -6.90771103e-01 2.39827916e-01 7.20020413e-01 7.85605669e-01 1.71824515e-01 -1.19382453e+00 -1.20822740e+00 3.66712026e-02 -4.24699903e-01 -1.69154513e+00 -1.28979072e-01 4.00111914e-01 -4.95165318e-01 5.49787283e-01 6.30206168e-01 1.85587481e-01 8.22172910e-02 3.93570513e-02 2.68296272e-01 1.23713779e+00 -7.74756312e-01 8.54182765e-02 1.91781715e-01 3.04462552e-01 9.35939252e-01 1.59074336e-01 1.82448044e-01 -3.75940859e-01 2.14006528e-02 9.17232633e-01 -1.43862860e-02 -3.62010151e-01 2.52393007e-01 -9.34839785e-01 1.01675701e+00 1.61262676e-02 5.20324647e-01 -1.33130297e-01 4.90717739e-01 -1.35019183e-01 4.79808718e-01 6.05573773e-01 2.21774548e-01 -6.08702064e-01 -3.41411680e-03 -1.16201341e+00 2.64197648e-01 7.84523368e-01 1.29746044e+00 1.02800846e+00 1.20728590e-01 3.43128666e-02 8.65976691e-01 2.70086616e-01 1.29480922e+00 -9.12937373e-02 -1.23058188e+00 5.80514669e-01 1.57200739e-01 2.13141263e-01 -6.89664423e-01 -4.45664346e-01 -3.66588205e-01 -1.26454461e+00 2.59968966e-01 9.21892405e-01 -3.67186785e-01 -3.74930173e-01 2.02538776e+00 -7.46248960e-02 -3.81358415e-01 -3.08828026e-01 5.87173581e-01 3.15133244e-01 7.62291193e-01 -1.69024020e-01 -1.02707100e+00 1.23145032e+00 -1.94568738e-01 -3.68101895e-01 1.34622604e-01 4.24127966e-01 -1.10104144e+00 8.09443057e-01 6.71586990e-01 -1.68884242e+00 -3.95292491e-01 -6.93114698e-01 4.08109248e-01 -2.08765909e-01 -1.70089960e-01 6.48696661e-01 9.94198442e-01 -9.78996515e-01 7.80121803e-01 -5.39287508e-01 3.34184408e-01 7.67481476e-02 7.17249155e-01 -4.12557691e-01 4.46218848e-02 -9.58542407e-01 3.53533894e-01 -8.86572748e-02 3.14327776e-01 -2.90043592e-01 -9.33967471e-01 -8.82902265e-01 -6.34531230e-02 5.42363226e-01 -1.15925387e-01 6.67169273e-01 -7.51401424e-01 -1.02459526e+00 9.31642056e-01 -1.83498010e-01 -1.30704297e-02 4.87902135e-01 2.37323552e-01 -3.51336151e-01 -2.46748514e-02 2.74278820e-01 9.08261538e-02 7.38722086e-01 -1.06646621e+00 -6.54559851e-01 -9.51779783e-01 -2.77914882e-01 -2.68169463e-01 -1.47441313e-01 3.79581094e-01 -1.77366540e-01 -8.44266653e-01 3.97799164e-01 -8.15659344e-01 -2.13138610e-01 -2.97262847e-01 -3.07685256e-01 3.26382220e-02 5.14839329e-02 -5.02984643e-01 1.45279908e+00 -1.94636846e+00 1.48298174e-01 9.64095414e-01 2.70544171e-01 -9.78886709e-02 4.51854765e-01 3.33070457e-01 -3.68579715e-01 1.69342011e-01 -6.91133857e-01 4.85368036e-02 1.32231504e-01 -6.87294602e-02 7.63195157e-02 8.28696907e-01 -2.69637555e-01 1.76005870e-01 -6.12988830e-01 -1.04876094e-01 -5.06018326e-02 3.22361350e-01 -3.82404506e-01 -5.93452975e-02 -4.88832360e-03 3.81458491e-01 -4.06815380e-01 4.29142177e-01 1.08910167e+00 -3.51160586e-01 -5.71212061e-02 5.31710982e-02 -5.17293513e-01 -3.24096441e-01 -2.10472345e+00 1.06688511e+00 -3.05203021e-01 1.13647439e-01 7.40115404e-01 -1.72161055e+00 1.01055300e+00 1.36835396e-01 1.09231591e+00 -8.75016809e-01 3.03167224e-01 4.57667112e-01 -3.74907255e-01 -2.39931494e-01 1.04983494e-01 -8.21546555e-01 -2.76116490e-01 6.10299826e-01 -9.56597179e-02 -2.86839336e-01 2.78495729e-01 -1.91060171e-01 1.26564884e+00 -1.01546034e-01 -6.01885840e-04 -7.77963042e-01 8.07950914e-01 -2.98375189e-01 3.18800747e-01 9.19412196e-01 -6.07760297e-03 3.51528734e-01 8.11594903e-01 -5.34291528e-02 -8.51367354e-01 -1.08955562e+00 -6.73965931e-01 1.28511691e+00 1.18692562e-01 -1.36003718e-01 -7.35140443e-01 -2.39371940e-01 -1.98008969e-01 4.65655029e-01 -6.09524012e-01 1.54656589e-01 -7.26107180e-01 -1.33736730e+00 4.82648909e-01 4.88276541e-01 4.63379920e-01 -9.02719021e-01 -3.03111583e-01 -2.32759658e-02 5.25376983e-02 -5.78437626e-01 -4.93173063e-01 8.21171880e-01 -6.00386739e-01 -1.06127894e+00 -1.03457808e+00 -4.49866533e-01 7.95286119e-01 -1.91373646e-01 1.13050890e+00 1.05486572e-01 -4.72845733e-01 4.61433858e-01 -1.54374927e-01 -6.55714452e-01 1.26318876e-02 -4.60881591e-01 -4.82481197e-02 -1.52952611e-01 1.70125067e-01 -6.48230970e-01 -7.66305268e-01 4.01726425e-01 -1.14408398e+00 -6.08028889e-01 2.00244233e-01 9.70449984e-01 1.01286268e+00 3.40113282e-01 2.95310527e-01 -8.81404102e-01 1.94555968e-01 -1.73716903e-01 -1.35061336e+00 3.16307135e-03 -8.25578630e-01 4.42962684e-02 8.20461273e-01 -1.63930148e-01 -6.01680100e-01 -3.33556950e-01 -1.96038380e-01 -2.79084086e-01 1.59300387e-01 3.86433303e-01 3.32794748e-02 -4.41784635e-02 7.02132225e-01 1.09283164e-01 -1.26410514e-01 -4.50131744e-01 1.39106065e-01 3.65527362e-01 6.19159281e-01 -7.11221874e-01 8.23659539e-01 4.54607546e-01 5.44690251e-01 -7.80771971e-01 -7.58037448e-01 -3.01071137e-01 -2.54632950e-01 3.58853154e-02 8.05869639e-01 -7.38470554e-01 -1.67224228e+00 4.84298766e-02 -4.84103471e-01 -2.06360102e-01 -6.17052019e-01 8.24879289e-01 -8.80593836e-01 2.66120642e-01 -3.72501791e-01 -1.32700419e+00 -2.73414195e-01 -9.94389057e-01 9.64974880e-01 1.47086695e-01 -6.40738606e-02 -9.74991083e-01 -2.06572726e-01 4.33920383e-01 3.55256855e-01 1.82835445e-01 1.01256728e+00 -4.94912595e-01 -5.39791703e-01 -4.21530932e-01 -4.44947511e-01 7.89921165e-01 -3.10824454e-01 -3.77510577e-01 -6.26489758e-01 -2.30093896e-01 3.58283937e-01 2.98686214e-02 6.83260798e-01 6.48014605e-01 1.36561644e+00 -6.46320105e-01 -1.11113124e-01 5.72276056e-01 1.62307048e+00 1.86177373e-01 8.24650943e-01 1.77656990e-02 1.08359717e-01 4.73716319e-01 3.49710196e-01 8.20763111e-01 -2.86919046e-02 4.86003697e-01 2.75390506e-01 1.03839129e-01 4.26534444e-01 4.66092557e-01 2.51595616e-01 6.58525467e-01 -4.97093916e-01 -1.31261736e-01 -5.92892945e-01 4.54104215e-01 -1.36553097e+00 -1.02576506e+00 -3.51396054e-01 2.77132773e+00 9.02619004e-01 6.55792430e-02 9.43891704e-02 4.31789458e-01 5.95461667e-01 -5.58166355e-02 -1.30891412e-01 -3.62853825e-01 -2.68784493e-01 1.50232196e+00 9.88260150e-01 6.26049876e-01 -8.17303896e-01 3.04137439e-01 3.80050945e+00 1.44110906e+00 -7.56363153e-01 -4.55535688e-02 6.32100701e-01 -3.17368329e-01 -4.48245585e-01 -1.92610249e-02 -8.24912071e-01 7.94331670e-01 8.05342913e-01 6.96261823e-02 6.36474609e-01 6.25070095e-01 4.07777093e-02 -6.54988706e-01 -9.28997934e-01 9.66288567e-01 1.03050217e-01 -9.92924690e-01 -9.00277257e-01 2.39101216e-01 7.62712002e-01 -5.05586982e-01 3.12582701e-01 2.53904462e-01 5.28271437e-01 -1.35362446e+00 4.62606132e-01 4.58629280e-01 1.17413855e+00 -8.82276297e-01 5.22836328e-01 4.47379768e-01 -1.22784686e+00 7.14612007e-02 -1.55995935e-01 -1.23916622e-02 6.02933355e-02 1.13383174e+00 5.20095713e-02 7.10862756e-01 1.01422453e+00 -2.49896981e-02 3.57785881e-01 6.16395473e-01 1.79110527e-01 3.97356004e-01 -6.30010843e-01 3.37587260e-02 1.43786579e-01 -6.02975368e-01 3.41959834e-01 9.39977467e-01 7.59899378e-01 7.14123368e-01 1.78020149e-01 8.64247501e-01 -1.93425596e-01 4.06834960e-01 -2.30496451e-01 2.57257968e-01 1.66651368e-01 8.26239347e-01 -8.15581441e-01 1.40864119e-01 -3.09135258e-01 3.27350229e-01 -1.04030520e-01 4.00455892e-01 -9.09520209e-01 -9.26604211e-01 5.28591096e-01 4.85809237e-01 3.50843072e-01 -1.88358128e-01 -4.38044339e-01 -8.06929469e-01 3.13531518e-01 -6.54315233e-01 5.74247479e-01 -3.36187422e-01 -1.25696445e+00 3.09623808e-01 2.49095142e-01 -1.14280272e+00 -2.72445261e-01 -7.51082242e-01 2.72757448e-02 1.40728331e+00 -1.24263549e+00 -4.84213829e-01 1.45732999e-01 8.83185565e-01 -2.37389788e-01 -2.59205461e-01 1.03671002e+00 5.63641191e-01 -2.08224475e-01 8.16658735e-01 5.75184047e-01 8.79532248e-02 4.56685424e-01 -9.89414930e-01 -7.01878607e-01 5.63858330e-01 -2.34195694e-01 5.09570956e-01 7.97067761e-01 -4.26731855e-01 -1.43395483e+00 -5.45470834e-01 9.30769682e-01 -2.27657378e-01 6.52224123e-01 -1.89564914e-01 -6.98295057e-01 7.88340986e-01 -6.11818656e-02 1.98342681e-01 7.92263031e-01 -4.81464230e-02 -1.16114020e-01 -3.69369596e-01 -1.40694082e+00 2.60556161e-01 7.62438595e-01 -1.67840049e-01 1.28915071e-01 5.30774176e-01 1.56957924e-01 -3.41165781e-01 -1.39217114e+00 7.70045161e-01 4.95132297e-01 -1.24319124e+00 1.00014162e+00 -3.52183342e-01 1.81577802e-01 -7.60240778e-02 -6.32453263e-01 -5.07910132e-01 1.65659845e-01 -1.08280182e+00 3.35766852e-01 1.05568719e+00 6.38436317e-01 -6.57648802e-01 8.00005794e-01 8.54708314e-01 -1.95383467e-03 -5.72400570e-01 -1.43671191e+00 -7.75902867e-01 6.17231846e-01 -7.36574054e-01 6.12580217e-03 6.85009718e-01 -6.48395866e-02 -3.19339871e-01 -4.12637532e-01 -4.74631488e-02 6.86355948e-01 -5.93957603e-02 3.85721505e-01 -1.21971583e+00 -7.00522423e-01 -6.00825310e-01 -4.78783995e-02 -8.91502142e-01 -7.87774026e-02 -7.48397589e-01 3.67110185e-02 -8.82969499e-01 1.88143641e-01 -1.04009771e+00 -3.05920184e-01 2.02507883e-01 2.50004649e-01 2.11871818e-01 2.74892058e-02 -5.03273271e-02 -2.93381989e-01 1.09021835e-01 1.19021916e+00 -1.04213186e-01 -9.23516154e-02 3.19110632e-01 -6.97903872e-01 8.62978578e-01 1.81715637e-01 -5.86538076e-01 -1.02086440e-01 -2.71311164e-01 9.04318511e-01 7.37443566e-01 1.90529853e-01 -7.33181357e-01 2.71620274e-01 -1.32211432e-01 4.03740346e-01 -3.46146345e-01 4.68745112e-01 -7.16953039e-01 1.72360912e-01 8.14188868e-02 -1.85603231e-01 1.40626609e-01 3.93391959e-02 2.62517542e-01 -8.78158361e-02 -6.74625874e-01 9.64233458e-01 -3.27965319e-01 1.23833761e-01 2.62589842e-01 -1.85560197e-01 2.89993137e-01 1.02124679e+00 3.91503088e-02 -1.59975186e-01 -6.69539750e-01 -9.33471203e-01 1.15735985e-01 -1.39405811e-02 -4.41645741e-01 1.10830650e-01 -1.11043525e+00 -7.37673700e-01 4.85089988e-01 -1.72180146e-01 3.54738176e-01 5.93619823e-01 1.31156576e+00 -4.04188931e-01 2.52616972e-01 4.57259327e-01 -5.34240663e-01 -7.30992913e-01 4.18677002e-01 1.87288687e-01 -4.77081090e-01 1.10042974e-01 1.39851928e+00 -1.06572621e-01 -2.29556113e-01 2.32342005e-01 -2.51030356e-01 2.95604169e-01 1.25027850e-01 4.21352386e-01 9.25610781e-01 2.47485429e-01 -5.52712679e-01 -2.28683800e-01 8.12264144e-01 2.15768978e-01 -3.47631216e-01 1.31468403e+00 -8.52839649e-02 -6.81715608e-01 2.90359914e-01 1.54244173e+00 3.60493362e-01 -8.80838156e-01 -3.66107598e-02 -1.79711565e-01 -2.12705910e-01 -5.48155546e-01 -6.10609353e-01 -1.04892111e+00 9.35073256e-01 7.33771563e-01 4.19763952e-01 1.45237982e+00 1.26888067e-01 4.90115821e-01 1.20848417e-01 4.92147803e-01 -1.46632457e+00 -6.19488992e-02 4.72098231e-01 4.93874580e-01 -1.24848628e+00 9.94390398e-02 -2.98229486e-01 -2.11802691e-01 1.09550190e+00 -2.14765109e-02 -2.02045232e-01 1.26395321e+00 4.49804038e-01 -1.29776299e-01 -2.50966549e-01 -8.86808708e-03 -2.49355391e-01 3.03017020e-01 -6.74567670e-02 8.05647373e-01 -7.61666102e-03 -6.28656209e-01 9.65352118e-01 -4.54390734e-01 -6.24385513e-02 2.21963763e-01 8.98852587e-01 -5.59255123e-01 -1.18672442e+00 -5.87238729e-01 6.32902145e-01 -8.69722247e-01 -6.17605448e-02 4.68562305e-01 9.77677941e-01 4.46864307e-01 9.55143571e-01 -5.66986650e-02 2.20359147e-01 3.74393344e-01 1.74587458e-01 7.80606687e-01 -2.02754542e-01 -4.75096405e-01 7.23713160e-01 -2.14695752e-01 -3.14872622e-01 -5.23353457e-01 -7.61486292e-01 -1.53150415e+00 -4.32761639e-01 -2.23892450e-01 4.35811728e-01 6.74512982e-01 1.06800163e+00 -4.54972833e-01 2.52256572e-01 7.55680084e-01 -3.86926770e-01 -6.70631230e-01 -7.34988451e-01 -1.33306432e+00 3.50194037e-01 5.57404272e-02 -5.37397563e-01 -6.06971979e-01 -3.82815627e-03]
[6.427547931671143, 4.557592868804932]
870464f9-fc31-445e-b592-540688d9f7aa
potato-the-portable-text-annotation-tool
2212.08620
null
https://arxiv.org/abs/2212.08620v2
https://arxiv.org/pdf/2212.08620v2.pdf
POTATO: The Portable Text Annotation Tool
We present POTATO, the Portable text annotation tool, a free, fully open-sourced annotation system that 1) supports labeling many types of text and multimodal data; 2) offers easy-to-configure features to maximize the productivity of both deployers and annotators (convenient templates for common ML/NLP tasks, active learning, keypress shortcuts, keyword highlights, tooltips); and 3) supports a high degree of customization (editable UI, inserting pre-screening questions, attention and qualification tests). Experiments over two annotation tasks suggest that POTATO improves labeling speed through its specially-designed productivity features, especially for long documents and complex tasks. POTATO is available at https://github.com/davidjurgens/potato and will continue to be updated.
['David Jurgens', 'Apostolos Dedeloudis', 'Jackson Sargent', 'Naitian Zhou', 'Xingyao Wang', 'Aparna Ananthasubramaniam', 'Jiaxin Pei']
2022-12-16
null
null
null
null
['text-annotation']
['natural-language-processing']
[-4.19743359e-02 4.02864456e-01 -3.34728658e-01 -4.20553386e-01 -1.10077477e+00 -1.27365172e+00 2.94331610e-01 6.45251274e-01 -4.99486655e-01 6.92346931e-01 3.53207350e-01 -5.47827601e-01 -2.72938788e-01 9.63286385e-02 9.73392352e-02 -1.20661549e-01 3.38802151e-02 1.01544726e+00 3.50145191e-01 -1.91903040e-01 3.06663543e-01 3.64539415e-01 -1.25875068e+00 2.64010578e-01 9.48669076e-01 4.80947644e-01 5.70819497e-01 9.59940851e-01 -5.42818189e-01 7.19546497e-01 -5.94632328e-01 -6.24063790e-01 -3.16344597e-03 1.21500984e-01 -1.28906512e+00 -1.35953248e-01 2.49089062e-01 -1.15159497e-01 1.35254577e-01 5.32882214e-01 8.55437279e-01 1.15144894e-01 2.18319386e-01 -1.36651778e+00 -8.23640227e-01 7.21527576e-01 -3.03528726e-01 2.41795450e-01 8.14896166e-01 3.41613114e-01 1.16349530e+00 -1.21234000e+00 8.23689342e-01 1.04381514e+00 8.01585674e-01 4.59328175e-01 -9.74068820e-01 -6.49393976e-01 -1.88437015e-01 -2.25441158e-01 -1.25574851e+00 -7.96456337e-01 7.46020228e-02 -5.23122549e-01 1.33877051e+00 5.70030928e-01 2.96302557e-01 1.05893958e+00 -8.40535015e-02 8.98968279e-01 6.75374091e-01 -8.69136155e-01 -8.95463675e-02 4.98903096e-01 4.67593819e-01 1.06682396e+00 -3.75056006e-02 -8.49246323e-01 -5.85369408e-01 -6.03825688e-01 2.28323936e-01 -3.06504369e-01 -2.79447466e-01 -5.34439944e-02 -1.18867600e+00 3.77740502e-01 -4.83632118e-01 2.83294022e-01 -8.06460977e-02 -2.47798547e-01 7.27005959e-01 2.01568425e-01 5.96572578e-01 8.29046488e-01 -1.08985877e+00 -6.67687535e-01 -6.02516353e-01 1.45705312e-01 1.15430343e+00 1.47798097e+00 7.43620574e-01 -5.91658831e-01 -4.86618400e-01 1.26752901e+00 4.33691442e-01 1.82257205e-01 3.46275032e-01 -1.12573647e+00 5.17161369e-01 9.46935117e-01 5.17358661e-01 -5.30606568e-01 -7.38193214e-01 -1.40057802e-01 -1.05398178e-01 -1.18490823e-01 4.76084977e-01 -4.03366178e-01 -5.42722583e-01 1.26971114e+00 3.54092300e-01 -7.35742033e-01 -1.64137125e-01 4.93330240e-01 1.27129769e+00 3.07627887e-01 5.88084280e-01 -2.85053432e-01 1.53938568e+00 -8.37095141e-01 -1.12163746e+00 -3.90254319e-01 1.12162781e+00 -1.20389223e+00 1.70689189e+00 3.00476193e-01 -1.08930445e+00 -3.18097442e-01 -6.03468478e-01 -6.29918098e-01 -8.73856068e-01 5.01298010e-01 7.29293406e-01 6.05937958e-01 -1.04439998e+00 3.43598843e-01 -9.51666296e-01 -7.82656968e-01 1.74550474e-01 3.75536501e-01 -3.83471757e-01 -8.97541293e-04 -9.98522341e-01 9.20707703e-01 6.14430189e-01 -1.96174711e-01 -1.88966930e-01 -6.61209702e-01 -6.90635681e-01 4.38252762e-02 7.55751908e-01 -5.02036035e-01 1.64892542e+00 -7.69616485e-01 -1.35361850e+00 1.01150084e+00 -6.42819181e-02 5.60169041e-01 3.96290988e-01 -5.09332955e-01 -3.30488414e-01 8.86626914e-02 2.33779952e-01 8.63718331e-01 2.06213415e-01 -8.32401395e-01 -6.07395172e-01 -2.69828856e-01 2.30900217e-02 6.25392437e-01 -6.47524118e-01 7.88912714e-01 -8.71943116e-01 -5.24617970e-01 -4.05999392e-01 -9.20260370e-01 3.28546800e-02 3.26541662e-02 -4.92435515e-01 -6.49212718e-01 1.05913138e+00 -1.14332998e+00 1.34972441e+00 -2.14497709e+00 -1.33628637e-01 -1.82231933e-01 3.33300978e-01 2.94477969e-01 6.15313575e-02 1.08591902e+00 6.23038821e-02 5.43232620e-01 1.33394878e-02 -5.30981839e-01 4.11459655e-01 2.21439660e-01 2.90423751e-01 2.06121728e-02 8.14444572e-02 9.79577780e-01 -1.07392085e+00 -9.76866066e-01 -3.16179954e-02 1.26363203e-01 -8.55716690e-02 1.74019158e-01 -4.28477019e-01 2.91998774e-01 -2.11667359e-01 9.72005129e-01 2.12659389e-01 -4.52591598e-01 2.97414929e-01 8.60242620e-02 -4.79730695e-01 1.90242574e-01 -1.24909317e+00 2.00965929e+00 -4.16524947e-01 7.34519958e-01 3.64980012e-01 -8.27506259e-02 8.00332725e-01 6.57095313e-01 3.96326959e-01 -2.43530199e-02 -2.93160006e-02 2.52943397e-01 -5.33700645e-01 -1.03660703e+00 8.76308978e-01 8.58202517e-01 -2.56289691e-01 6.79281533e-01 4.11466539e-01 -2.88034175e-02 7.12574303e-01 7.01016009e-01 1.18281734e+00 2.32591122e-01 4.06776667e-01 -2.48898998e-01 1.27310827e-01 2.75257021e-01 3.24235767e-01 8.25354815e-01 -6.84177056e-02 4.32393998e-02 5.06273210e-01 3.70869972e-02 -1.04222596e+00 -4.88563538e-01 -6.48997501e-02 1.93633807e+00 -5.18487036e-01 -8.15087676e-01 -5.50072074e-01 -8.77126396e-01 -1.64037883e-01 7.99294829e-01 -2.41999120e-01 4.09288585e-01 -4.47495608e-03 -3.11126560e-01 8.48357797e-01 5.66911161e-01 2.21230909e-01 -1.11619842e+00 -3.68462265e-01 5.39889000e-03 -3.07169616e-01 -8.17537725e-01 -6.26705527e-01 5.87178171e-01 -5.49491107e-01 -9.77702796e-01 -4.32354987e-01 -8.99878919e-01 6.07127607e-01 -8.88391435e-02 1.17058885e+00 2.36653507e-01 -4.22356725e-01 8.56774390e-01 -8.25111508e-01 -5.86457253e-01 -2.78249264e-01 3.69722515e-01 -3.55470300e-01 -7.28479683e-01 3.39323968e-01 -3.85879874e-02 -1.27524719e-01 4.66069579e-01 -8.21573794e-01 1.47615999e-01 2.93331951e-01 6.20791912e-01 3.98617804e-01 -2.63962954e-01 3.86065364e-01 -1.21602404e+00 1.00368834e+00 -4.10038650e-01 -4.72756416e-01 6.53931379e-01 -8.02794278e-01 -2.89402366e-01 1.50342584e-01 -3.00908476e-01 -1.20525217e+00 4.76894945e-01 -3.18753779e-01 1.45175084e-01 -4.43581194e-01 6.63127363e-01 -2.41941586e-01 6.04129257e-03 6.97306514e-01 -4.72055912e-01 -2.30907857e-01 -7.73088515e-01 4.82829452e-01 1.17045736e+00 4.58792150e-01 -5.84156394e-01 4.76924300e-01 -1.69851258e-01 -6.91135287e-01 -7.00536788e-01 -7.10864305e-01 -8.42780948e-01 -9.79450345e-01 -2.61607230e-01 6.50761485e-01 -6.58845484e-01 -7.90049374e-01 6.51506484e-02 -1.00024796e+00 -8.00722361e-01 -2.01682582e-01 1.72918186e-01 -1.98625907e-01 4.82231408e-01 -6.76511824e-01 -9.58746493e-01 -7.25886822e-01 -8.79855096e-01 1.05681336e+00 4.31717068e-01 -9.32770491e-01 -1.11845636e+00 7.33486935e-02 5.32912552e-01 2.20592454e-01 -1.16845295e-01 9.15377259e-01 -1.27172494e+00 2.92484872e-02 -3.05483043e-01 -3.65381353e-02 -3.70201953e-02 -1.45652562e-01 6.37106895e-01 -7.75960088e-01 -1.49258912e-01 -8.29260170e-01 -7.01480150e-01 2.54693508e-01 -2.49014888e-02 1.20793295e+00 -5.07843018e-01 -5.11748075e-01 1.22188881e-01 9.33483660e-01 2.88750708e-01 4.08244997e-01 2.10857391e-01 4.56669450e-01 6.23388231e-01 7.68821359e-01 5.21026909e-01 3.34068805e-01 6.92859232e-01 8.79878998e-02 -2.51437664e-01 3.02797437e-01 -6.78605139e-02 -6.33462816e-02 8.63354266e-01 2.05901101e-01 -6.21111631e-01 -1.43886864e+00 3.05599153e-01 -2.10111189e+00 -6.93205774e-01 -2.95477897e-01 1.97271895e+00 1.05093002e+00 5.32280132e-02 1.39527455e-01 3.90807875e-02 4.93211925e-01 -2.98032165e-01 -4.28099424e-01 -4.81361985e-01 -5.51301204e-02 1.72443911e-01 3.73834133e-01 6.00765526e-01 -1.03913689e+00 1.07137227e+00 6.43892765e+00 8.61118317e-01 -6.92642748e-01 6.54883265e-01 4.08102185e-01 6.29835799e-02 -5.49388342e-02 -4.93215583e-02 -1.04859233e+00 3.49798173e-01 7.00102746e-01 -1.96350917e-01 1.42751291e-01 9.02079403e-01 2.33058557e-01 -2.47545704e-01 -8.81588936e-01 5.07340491e-01 -2.19475120e-01 -1.20389783e+00 -5.49248934e-01 -1.96752205e-01 8.03657696e-02 1.98493838e-01 -2.51474380e-01 4.82442677e-01 6.26795411e-01 -6.46514475e-01 7.04207480e-01 4.41264689e-01 1.02637458e+00 -6.01571083e-01 6.58406317e-01 4.16835517e-01 -8.96932662e-01 -1.26128808e-01 -6.19069152e-02 8.02066401e-02 2.10119292e-01 3.93326610e-01 -1.32461262e+00 4.00203884e-01 8.29219997e-01 3.29903662e-01 -1.11130953e+00 1.08478677e+00 -4.64715697e-02 5.59782207e-01 -2.59768546e-01 -2.61743307e-01 -1.85550183e-01 2.02055633e-01 4.66267645e-01 1.97668326e+00 -2.73599364e-02 1.97899565e-01 6.17192268e-01 3.81645203e-01 1.18522123e-01 7.26490140e-01 -2.58907318e-01 -5.63928485e-01 9.50280726e-01 1.88801718e+00 -1.17391717e+00 -1.78685635e-01 -2.69342363e-01 9.03682590e-01 3.32475811e-01 3.72311890e-01 -4.86564696e-01 -8.78276289e-01 8.03939402e-02 -3.05465721e-02 -9.42428410e-02 -4.43475306e-01 -1.24302201e-01 -8.42816889e-01 -2.86684930e-01 -1.01046991e+00 7.32415318e-01 -1.32539678e+00 -9.44526136e-01 4.81183201e-01 6.35932386e-02 -6.88727498e-01 -7.99784660e-02 -6.40084505e-01 -2.05626890e-01 7.56662488e-01 -5.44722557e-01 -1.23827255e+00 -4.52964693e-01 6.14441931e-01 6.76208675e-01 3.16102728e-02 1.32396257e+00 5.95881402e-01 -8.61032069e-01 6.39014661e-01 6.22835681e-02 4.20646846e-01 1.08794379e+00 -1.56991816e+00 3.23673815e-01 4.32151049e-01 1.42269880e-01 8.41945350e-01 5.37512422e-01 -8.09482098e-01 -1.42441368e+00 -6.85315549e-01 1.23372304e+00 -8.48802507e-01 6.68652892e-01 -7.21515477e-01 -8.33906472e-01 1.01039529e+00 5.69622695e-01 -4.55803275e-01 1.22916937e+00 4.86523032e-01 -1.71380252e-01 2.76422232e-01 -1.11611629e+00 4.97743458e-01 8.54638100e-01 -4.42599326e-01 -3.36921126e-01 1.12088060e+00 6.29899859e-01 -7.33272195e-01 -1.01621020e+00 1.40690401e-01 5.68575084e-01 -4.16550428e-01 5.50477862e-01 -4.25385743e-01 -4.39851321e-02 7.93415587e-03 2.58881181e-01 -9.64084923e-01 -4.69504297e-01 -1.06449652e+00 -5.94161563e-02 1.77681041e+00 1.12021279e+00 -6.03984296e-01 3.08139473e-01 9.38185036e-01 -5.37239194e-01 -6.72784269e-01 -4.99523252e-01 -4.02710617e-01 -6.49442971e-01 -5.38734734e-01 2.93263465e-01 1.45196235e+00 7.12550581e-01 5.34289777e-01 -6.42190054e-02 -1.32546164e-02 -6.44662753e-02 -6.55357838e-01 6.63950443e-01 -1.29490972e+00 -3.61347407e-01 -3.94899011e-01 2.37504095e-01 -6.14175856e-01 -9.74191353e-02 -8.52722168e-01 -1.04229867e-01 -1.76724434e+00 3.68073620e-02 -7.58121848e-01 2.57790446e-01 1.49386489e+00 -1.56928435e-01 2.15787098e-01 -6.52150065e-03 3.72976243e-01 -1.04782629e+00 -2.58321822e-01 7.26928592e-01 1.11794353e-01 -2.88711935e-01 -1.44186497e-01 -7.70995796e-01 5.27532279e-01 7.56301343e-01 -4.67560887e-01 -2.04798654e-01 -4.38051701e-01 7.10492790e-01 -1.59564525e-01 -1.03439927e-01 -6.14981234e-01 3.90621185e-01 -1.32702217e-01 4.73306715e-01 -4.40527320e-01 1.74106061e-01 -6.20638609e-01 2.45469332e-01 1.36951962e-02 -6.32928967e-01 5.01523018e-01 4.82473135e-01 -4.39750664e-02 3.19441348e-01 -7.26253390e-01 3.30991596e-01 -3.36509138e-01 -7.21779227e-01 1.22371651e-01 -6.89531624e-01 2.60539293e-01 8.31357002e-01 -1.55286998e-01 -6.94250345e-01 -3.02700460e-01 -9.04875040e-01 5.87813139e-01 5.04991293e-01 4.94373679e-01 -7.22351894e-02 -7.66946018e-01 -4.35791939e-01 -3.04490238e-01 4.41616803e-01 -8.73816386e-02 -6.08932227e-02 8.73247743e-01 -7.07013488e-01 3.52223098e-01 9.00333226e-02 -3.99107218e-01 -1.71151972e+00 3.17320883e-01 -7.94695541e-02 -1.13995597e-01 -5.30167460e-01 1.03649926e+00 -6.90936565e-01 -7.11720824e-01 6.27269089e-01 1.20009609e-01 -3.94646913e-01 4.07233268e-01 5.48089743e-01 5.45479000e-01 4.16044801e-01 -2.17692077e-01 -2.92105407e-01 -2.10074037e-02 -1.32162601e-01 -2.77328640e-01 1.21890724e+00 -3.61273289e-01 -1.76530302e-01 7.21722722e-01 6.65574968e-01 4.04233694e-01 -7.80879319e-01 -2.03972146e-01 5.18908799e-01 -2.84118146e-01 -1.16100892e-01 -1.57058549e+00 -4.83582854e-01 3.23366642e-01 3.93433094e-01 4.89379406e-01 8.35054338e-01 2.36011222e-01 2.07336470e-01 7.35078692e-01 7.42498338e-02 -1.38683379e+00 -8.40498193e-04 5.48771739e-01 8.29691470e-01 -9.46104109e-01 2.09922239e-01 -4.27552372e-01 -9.50622678e-01 1.23980784e+00 6.59286082e-01 9.29623127e-01 3.50538820e-01 7.44924068e-01 4.60510999e-01 -4.88880992e-01 -1.09813464e+00 -1.22668311e-01 1.31310910e-01 6.81529403e-01 1.02248585e+00 -2.25213543e-01 -5.56120336e-01 4.79300529e-01 -1.33035295e-02 1.57451294e-02 3.17753732e-01 1.18451071e+00 -3.14405352e-01 -1.18705130e+00 -2.75667280e-01 5.18325388e-01 -4.89348203e-01 -1.88988507e-01 -8.78824532e-01 7.63969004e-01 1.72501013e-01 9.84635353e-01 4.06094454e-02 -4.38223630e-02 2.25896627e-01 6.10893726e-01 1.57178581e-01 -1.18186522e+00 -1.02921534e+00 1.26427993e-01 8.48159611e-01 -2.82445759e-01 -4.55529764e-02 -7.42465317e-01 -1.17477489e+00 -1.64938688e-01 -6.87591791e-01 4.66259480e-01 8.09851468e-01 7.29103327e-01 7.42722392e-01 1.25745222e-01 -1.19819351e-01 -7.64844179e-01 -1.25707567e-01 -1.47709298e+00 -3.44141573e-01 -2.09071413e-02 -4.30134892e-01 -6.24869108e-01 -5.99652790e-02 2.60448337e-01]
[9.246037483215332, 8.724334716796875]
c9eeebca-5ba5-4077-8856-29bc45ea81ed
cross-domain-face-presentation-attack
2004.01959
null
https://arxiv.org/abs/2004.01959v1
https://arxiv.org/pdf/2004.01959v1.pdf
Cross-domain Face Presentation Attack Detection via Multi-domain Disentangled Representation Learning
Face presentation attack detection (PAD) has been an urgent problem to be solved in the face recognition systems. Conventional approaches usually assume the testing and training are within the same domain; as a result, they may not generalize well into unseen scenarios because the representations learned for PAD may overfit to the subjects in the training set. In light of this, we propose an efficient disentangled representation learning for cross-domain face PAD. Our approach consists of disentangled representation learning (DR-Net) and multi-domain learning (MD-Net). DR-Net learns a pair of encoders via generative models that can disentangle PAD informative features from subject discriminative features. The disentangled features from different domains are fed to MD-Net which learns domain-independent features for the final cross-domain face PAD task. Extensive experiments on several public datasets validate the effectiveness of the proposed approach for cross-domain PAD.
['Xilin Chen', 'Shiguang Shan', 'Guoqing Wang', 'Hu Han']
2020-04-04
cross-domain-face-presentation-attack-1
http://openaccess.thecvf.com/content_CVPR_2020/html/Wang_Cross-Domain_Face_Presentation_Attack_Detection_via_Multi-Domain_Disentangled_Representation_Learning_CVPR_2020_paper.html
http://openaccess.thecvf.com/content_CVPR_2020/papers/Wang_Cross-Domain_Face_Presentation_Attack_Detection_via_Multi-Domain_Disentangled_Representation_Learning_CVPR_2020_paper.pdf
cvpr-2020-6
['face-presentation-attack-detection']
['computer-vision']
[ 1.33777842e-01 -2.66423430e-02 -2.90952742e-01 -5.53622544e-01 -1.09441960e+00 -6.24558568e-01 5.13153911e-01 -4.71432358e-01 2.46783033e-01 8.65225971e-01 2.22094819e-01 2.98148394e-02 -2.43270993e-01 -7.93764353e-01 -5.58566868e-01 -8.90583038e-01 -1.33868799e-01 5.31732976e-01 -2.58734703e-01 -4.05982025e-02 -2.52584070e-01 6.46670520e-01 -1.24692523e+00 5.73532343e-01 5.98413885e-01 9.09815669e-01 -4.64800477e-01 3.63019198e-01 1.90760642e-01 4.34268236e-01 -6.79726422e-01 -8.88547659e-01 4.36375678e-01 -3.86215359e-01 -4.71673369e-01 6.69101700e-02 4.19198334e-01 -5.70734680e-01 -7.30407715e-01 7.78999269e-01 6.99818492e-01 -1.88305572e-01 8.84649694e-01 -1.70058727e+00 -9.62280512e-01 7.32500181e-02 -8.51844013e-01 1.96868941e-01 3.48490983e-01 -4.44468344e-04 7.95734107e-01 -1.00066113e+00 3.58997256e-01 1.47205305e+00 3.79133284e-01 9.33218539e-01 -1.33014548e+00 -1.39559567e+00 -8.10148492e-02 1.56966582e-01 -1.32839286e+00 -5.62491179e-01 1.16715944e+00 -3.88133049e-01 3.06750029e-01 -4.49020565e-02 -1.14465103e-01 1.83041418e+00 -9.96706039e-02 7.85053611e-01 1.16383326e+00 1.97725371e-02 -7.51157776e-02 4.86354887e-01 3.26027088e-02 6.39543056e-01 6.80100679e-01 4.21534628e-01 -7.56144583e-01 -5.65836668e-01 5.37365913e-01 5.84506467e-02 -2.98101753e-01 -6.92586482e-01 -5.14147162e-01 1.05381930e+00 8.41036513e-02 -6.19654432e-02 -2.42032155e-01 -4.55331475e-01 5.71027160e-01 3.34576279e-01 5.67836642e-01 9.01212245e-02 -5.00898719e-01 3.93092483e-01 -7.02532589e-01 3.29430193e-01 8.04952204e-01 6.82228804e-01 6.50168777e-01 2.19053671e-01 -1.27660427e-02 8.13854873e-01 4.50209886e-01 5.00152767e-01 4.36959296e-01 -1.86719760e-01 6.78030729e-01 5.36794662e-01 -1.22179069e-01 -1.09555292e+00 2.97842830e-01 -2.56379098e-01 -8.80551398e-01 3.30625683e-01 3.18310618e-01 -4.19164568e-01 -6.12350702e-01 2.01941729e+00 2.69604534e-01 7.08063483e-01 4.48782951e-01 5.98031342e-01 8.64242852e-01 6.04184985e-01 1.88350469e-01 -6.20549470e-02 1.47685587e+00 -4.55093831e-01 -7.01621711e-01 -3.73124629e-01 2.43074760e-01 -6.53541505e-01 5.69814861e-01 5.94603419e-01 -7.84532666e-01 -5.23825169e-01 -1.32988536e+00 2.54744172e-01 -1.55355245e-01 2.84706414e-01 3.66383076e-01 1.01971531e+00 -5.38010299e-01 2.86968321e-01 -2.58183330e-01 2.04285279e-01 1.16751277e+00 6.11771107e-01 -9.32925284e-01 -5.22630811e-01 -1.37608302e+00 6.25641584e-01 2.12195247e-01 -1.76769942e-02 -1.50176728e+00 -7.11351871e-01 -8.60951304e-01 1.96379527e-01 3.32802743e-01 -3.32154453e-01 8.61384749e-01 -9.40655708e-01 -1.30984247e+00 1.16924489e+00 2.09408440e-02 -1.72584459e-01 2.18980834e-01 -4.05237824e-01 -9.04964745e-01 1.68226108e-01 -4.40388434e-02 3.72612812e-02 1.44978487e+00 -1.41067553e+00 1.54676791e-02 -8.26373518e-01 -3.04199636e-01 -4.05473560e-02 -3.31896693e-01 4.27450798e-02 9.28230509e-02 -6.42732024e-01 -3.88832301e-01 -6.20501995e-01 3.13253611e-01 1.70339853e-01 -3.91675651e-01 -3.25610191e-01 1.28581440e+00 -8.41060042e-01 7.30686724e-01 -2.52712464e+00 1.43826738e-01 -1.88310333e-02 4.75207269e-01 5.80366492e-01 -4.69999045e-01 2.27328762e-01 -6.54788733e-01 -1.08698227e-01 1.28365770e-01 -4.02185827e-01 -9.85328201e-03 2.60666579e-01 -7.30807960e-01 6.85967267e-01 8.32527518e-01 4.69044000e-01 -6.44039690e-01 -4.30833697e-01 -1.28466904e-01 6.56712711e-01 -5.61914027e-01 7.70877540e-01 -9.77066457e-02 5.06211698e-01 -6.99795306e-01 5.02376676e-01 1.36978555e+00 -5.76256327e-02 3.85744154e-01 -1.59920946e-01 5.55122435e-01 5.60939908e-02 -1.05608726e+00 1.44422400e+00 -3.23421091e-01 3.81682605e-01 -1.04716010e-01 -1.24103999e+00 1.13134575e+00 7.29221344e-01 3.86323929e-01 -4.86141115e-01 1.35292485e-01 1.77553637e-04 -1.44207567e-01 -5.53331494e-01 -1.72715291e-01 -4.39889431e-01 -1.12512045e-01 4.88766342e-01 6.10320807e-01 5.20353377e-01 -5.44682324e-01 -8.35928600e-03 6.98851705e-01 4.13395502e-02 4.01568979e-01 5.26525341e-02 6.85938895e-01 -5.82239866e-01 7.97989964e-01 1.12815604e-01 -3.73491794e-01 4.13439989e-01 9.60917771e-01 -4.20500964e-01 -6.49264634e-01 -1.62918234e+00 -8.04081559e-02 9.54314053e-01 -1.09801114e-01 -1.03145577e-01 -5.43620169e-01 -1.23323035e+00 1.91086739e-01 4.04615104e-01 -7.58169472e-01 -5.79999506e-01 -5.07088542e-01 -7.03027964e-01 8.18794310e-01 4.43589121e-01 3.90661061e-01 -7.29344845e-01 1.01474270e-01 -3.37947428e-01 1.50875330e-01 -1.24028397e+00 -2.35381186e-01 -5.59026422e-03 -6.29911840e-01 -1.32951725e+00 -6.40177131e-01 -8.14907789e-01 6.39972985e-01 3.11863780e-01 8.99048328e-01 -4.85296458e-01 -4.42274988e-01 1.32307574e-01 -7.56151602e-02 -1.81643829e-01 -2.84909338e-01 -3.53306890e-01 2.97565311e-01 5.05863190e-01 7.64832199e-01 -7.20377207e-01 -6.22336566e-01 3.89982104e-01 -6.94583237e-01 -5.28352797e-01 8.04036081e-01 1.08072531e+00 1.50404513e-01 -2.68534254e-02 1.00052404e+00 -1.01288319e+00 7.66588032e-01 -9.99033093e-01 -5.25297225e-01 3.04471105e-01 -2.16219246e-01 2.27978185e-01 5.27648032e-01 -8.23663414e-01 -1.28903627e+00 6.61323294e-02 1.09574102e-01 -7.91314900e-01 -3.04749936e-01 -3.42446379e-02 -1.05658960e+00 -8.29728553e-04 5.95326245e-01 2.91326761e-01 2.08238974e-01 -4.51736540e-01 3.14352632e-01 7.46586919e-01 3.23653072e-01 -7.38194287e-01 1.08965981e+00 1.80612534e-01 -1.17598251e-01 -5.43544769e-01 -7.60857403e-01 -1.35667756e-01 -4.74204659e-01 1.99908495e-01 8.56932759e-01 -1.26092720e+00 -7.77409852e-01 3.19370449e-01 -1.18583512e+00 4.77271587e-01 -1.78855136e-02 4.07512903e-01 -4.44099098e-01 4.47020948e-01 -1.87463164e-01 -7.91899920e-01 -1.12219878e-01 -1.08196771e+00 1.17348325e+00 2.68327117e-01 -6.28722310e-02 -1.03424037e+00 3.30728292e-01 5.03778577e-01 -4.30864617e-02 4.34178412e-01 1.15848398e+00 -1.26142561e+00 -3.83984923e-01 -4.79337394e-01 -3.91165584e-01 6.66224420e-01 3.28081846e-01 -4.91053849e-01 -1.52250707e+00 -4.63513017e-01 2.29148760e-01 -7.44634032e-01 5.69297731e-01 -9.37354937e-02 1.38821554e+00 -6.14248872e-01 -2.91701376e-01 5.48146963e-01 1.29311347e+00 8.62968154e-04 6.59330904e-01 -3.87121588e-01 3.76068801e-01 6.07763290e-01 3.10848594e-01 6.06594503e-01 -6.19565696e-02 6.34857118e-01 3.76866579e-01 9.46837664e-02 -5.00424430e-02 -6.10276759e-01 6.01619661e-01 2.35471874e-01 4.13264155e-01 -2.43399724e-01 -5.36055982e-01 3.70398223e-01 -1.21963000e+00 -1.10029745e+00 3.78792942e-01 2.15380883e+00 7.31136203e-01 -1.07384093e-01 2.82158643e-01 1.81761265e-01 6.99288368e-01 2.18182117e-01 -6.82576060e-01 -4.25530255e-01 3.65145551e-03 7.19324589e-01 -7.46040866e-02 1.37028173e-01 -9.60271418e-01 5.29906154e-01 5.46565056e+00 6.96564019e-01 -9.39353585e-01 3.47214311e-01 5.52098513e-01 1.13112569e-01 -1.06295332e-01 -2.73902655e-01 -8.53320837e-01 3.67958367e-01 7.18784332e-01 -2.58543521e-01 1.45630553e-01 1.05978107e+00 -3.97867650e-01 4.84730721e-01 -1.44658446e+00 1.31078529e+00 3.58999074e-01 -8.42337847e-01 2.10220024e-01 3.20014328e-01 4.76098299e-01 -4.76274759e-01 8.52395952e-01 4.57746744e-01 3.05636734e-01 -1.36859357e+00 -1.58308342e-01 1.55554876e-01 1.03373194e+00 -8.83450031e-01 5.01670659e-01 2.46141076e-01 -8.00242722e-01 -2.84637306e-02 -4.20442671e-01 4.03799981e-01 -3.94638747e-01 3.55529875e-01 -1.13689375e+00 6.73409283e-01 1.41639620e-01 5.86005330e-01 -3.70513588e-01 4.59150702e-01 -1.80354193e-01 5.30868113e-01 1.50219142e-01 3.58262986e-01 -2.45983437e-01 2.70292815e-02 4.80111182e-01 7.90522933e-01 2.18794391e-01 2.68024625e-03 -8.32974166e-02 9.23629463e-01 -3.29409957e-01 -1.74515203e-01 -1.06031966e+00 -3.62138718e-01 3.90304267e-01 1.15420449e+00 1.98358402e-01 1.36682138e-01 -5.06085336e-01 1.19480157e+00 4.47803915e-01 3.90140176e-01 -7.92847633e-01 -2.45192587e-01 1.26011002e+00 -8.37555528e-02 2.66165584e-01 5.42755239e-02 9.36999638e-03 -1.36299515e+00 -1.75292715e-02 -1.06652927e+00 6.33544862e-01 -4.03498799e-01 -1.94722509e+00 9.09833312e-01 -1.66290440e-02 -1.41918099e+00 -2.41111740e-01 -9.67319727e-01 -6.15158558e-01 1.23243725e+00 -1.64971983e+00 -1.30468249e+00 -1.52778868e-02 1.02653408e+00 4.10141438e-01 -7.78922915e-01 1.28997755e+00 5.20246029e-01 -7.34190524e-01 1.21421766e+00 1.06016718e-01 5.48656940e-01 8.59688878e-01 -7.57015467e-01 6.54703230e-02 5.69076002e-01 8.94616172e-02 6.14663839e-01 3.82478327e-01 -5.00423729e-01 -1.41232646e+00 -1.04842293e+00 5.81935048e-01 -4.68483835e-01 3.42935890e-01 -7.66902685e-01 -9.56980824e-01 6.91346526e-01 2.69896895e-01 4.82510924e-01 1.56307054e+00 2.03803629e-01 -1.20806086e+00 -3.36051315e-01 -1.51314151e+00 2.82116145e-01 6.72241390e-01 -8.63165379e-01 -7.20950544e-01 3.67443442e-01 4.77208436e-01 1.02075012e-02 -7.06112325e-01 1.72956005e-01 5.84135771e-01 -7.72019029e-01 1.26601589e+00 -1.33168280e+00 4.48657066e-01 1.91073254e-01 -4.11972076e-01 -1.24998903e+00 -1.35336876e-01 -4.72765833e-01 -3.53987187e-01 1.39479756e+00 2.68290415e-02 -5.91686308e-01 8.82903218e-01 2.38381997e-01 5.11568964e-01 -6.44409478e-01 -1.06071436e+00 -7.65299559e-01 3.08697224e-01 -8.19415599e-02 8.46940577e-01 1.08261645e+00 -5.09302728e-02 5.21275222e-01 -8.66807282e-01 7.74107635e-01 9.81797159e-01 1.01788484e-01 6.52584314e-01 -1.52032530e+00 -5.09836316e-01 1.44809008e-01 -5.84106505e-01 -6.99051142e-01 7.09857643e-01 -9.11772847e-01 -3.70287210e-01 -6.10730767e-01 4.03199792e-01 -2.42531717e-01 -3.37819993e-01 3.14197451e-01 -1.29688412e-01 7.00641945e-02 -7.03600347e-02 -2.41483271e-01 -2.77689278e-01 8.02415073e-01 1.15757608e+00 -6.65215999e-02 2.47484088e-01 1.95760250e-01 -1.16003942e+00 4.45533067e-01 7.64360368e-01 -8.19028616e-01 -7.01171875e-01 -3.03787798e-01 -3.42257380e-01 4.23108608e-01 5.50193965e-01 -7.26565659e-01 -1.84318215e-01 -1.96391836e-01 8.36079538e-01 -2.42686242e-01 6.95211828e-01 -9.71760392e-01 -1.62321836e-01 3.00708592e-01 -1.41587526e-01 -1.91874743e-01 3.41427028e-01 7.55284846e-01 -1.73644647e-01 3.16190869e-02 9.53921795e-01 2.57005572e-01 -3.43829811e-01 7.02492476e-01 3.18412445e-02 2.46181473e-01 1.30422413e+00 -1.60015702e-01 -3.26859087e-01 -1.69916019e-01 -8.29862535e-01 -2.60008331e-02 -1.81923419e-01 6.59340918e-01 1.07011104e+00 -1.53810155e+00 -9.12401140e-01 9.19294775e-01 1.80883884e-01 -3.65379304e-01 4.90669370e-01 1.61395624e-01 6.36815056e-02 2.10823342e-01 -6.17011547e-01 -2.96906769e-01 -1.47798395e+00 8.43560219e-01 2.52302289e-01 -3.27010632e-01 -1.72281116e-01 8.61097753e-01 6.81826174e-01 -3.68770838e-01 5.41931093e-02 5.59683502e-01 -1.66448072e-01 4.19436283e-02 7.07430184e-01 6.65409341e-02 -1.15983240e-01 -7.16667712e-01 -5.01478195e-01 2.71795094e-01 -5.17395377e-01 -4.03600186e-02 1.52757609e+00 2.78833896e-01 1.21971317e-01 -1.11113332e-01 1.74069297e+00 7.93437362e-02 -1.14799619e+00 -3.86736363e-01 -2.26027817e-01 -9.03496265e-01 -3.40113729e-01 -7.34751999e-01 -1.01921403e+00 1.27640760e+00 9.23626006e-01 -3.57665211e-01 1.24404907e+00 1.88709293e-02 5.84980428e-01 -5.96406683e-03 1.70279026e-01 -3.87555480e-01 6.77758634e-01 -1.07481502e-01 8.89913321e-01 -1.14414060e+00 -1.16210006e-01 -6.44575536e-01 -8.16996694e-01 1.08942974e+00 9.29791808e-01 -4.89514112e-01 8.32354546e-01 2.23126009e-01 -2.65868068e-01 -9.34434086e-02 -8.18832159e-01 2.61452049e-01 4.03110355e-01 1.06623590e+00 2.42317572e-01 -6.15809038e-02 2.22353399e-01 1.19121754e+00 1.67102799e-01 -1.45217851e-02 1.18201785e-01 5.75068891e-01 7.08636642e-02 -1.50838602e+00 -2.12217212e-01 1.77405909e-01 -4.52423662e-01 4.04602915e-01 -5.17294765e-01 6.12731874e-01 3.23541433e-01 7.55510271e-01 -2.30273187e-01 -4.76807624e-01 1.79717973e-01 3.20626616e-01 7.29816139e-01 -6.06015563e-01 -2.21612081e-01 -4.21201825e-01 -2.15467244e-01 -3.35476398e-01 5.15719429e-02 -6.48272097e-01 -7.36007631e-01 -1.92588106e-01 -1.58541188e-01 2.56132394e-01 1.62347794e-01 7.88286507e-01 5.52899718e-01 2.85577387e-01 1.15634739e+00 -2.47194171e-01 -1.09087300e+00 -6.50906324e-01 -7.06186533e-01 4.60299075e-01 6.70303643e-01 -9.61546361e-01 -2.02345297e-01 -7.73950666e-02]
[13.087055206298828, 1.183016061782837]
9d469bf8-3d1c-4dde-9082-7e55d4981bca
multimodal-gpt-a-vision-and-language-model
2305.04790
null
https://arxiv.org/abs/2305.04790v3
https://arxiv.org/pdf/2305.04790v3.pdf
MultiModal-GPT: A Vision and Language Model for Dialogue with Humans
We present a vision and language model named MultiModal-GPT to conduct multi-round dialogue with humans. MultiModal-GPT can follow various instructions from humans, such as generating a detailed caption, counting the number of interested objects, and answering general questions from users. MultiModal-GPT is parameter-efficiently fine-tuned from OpenFlamingo, with Low-rank Adapter (LoRA) added both in the cross-attention part and the self-attention part of the language model. We first construct instruction templates with vision and language data for multi-modality instruction tuning to make the model understand and follow human instructions. We find the quality of training data is vital for the dialogue performance, where few data containing short answers can lead the model to respond shortly to any instructions. To further enhance the ability to chat with humans of the MultiModal-GPT, we utilize language-only instruction-following data to train the MultiModal-GPT jointly. The joint training of language-only and visual-language instructions with the \emph{same} instruction template effectively improves dialogue performance. Various demos show the ability of continuous dialogue of MultiModal-GPT with humans. Code, dataset, and demo are at https://github.com/open-mmlab/Multimodal-GPT
['Kai Chen', 'Ping Luo', 'Wenwei Zhang', 'Kuikun Liu', 'Qian Zhao', 'Miao Zheng', 'Yudong Wang', 'Shilong Zhang', 'Chengqi Lyu', 'Tao Gong']
2023-05-08
null
null
null
null
['instruction-following']
['natural-language-processing']
[-3.59468430e-01 6.63854554e-02 3.37804973e-01 -5.36485910e-01 -8.16137612e-01 -7.14156449e-01 5.81722140e-01 -2.44724751e-01 -6.33522272e-01 4.02058423e-01 4.91995692e-01 -3.20425868e-01 3.94653082e-01 -4.33978736e-01 -3.99520487e-01 -3.36696208e-01 4.83882129e-01 9.41252112e-01 1.40340418e-01 -7.61667490e-01 2.04460129e-01 -1.13257147e-01 -1.17729998e+00 8.97739232e-01 9.90295053e-01 6.65984452e-01 7.18290269e-01 1.24047530e+00 -4.48791265e-01 9.24901664e-01 -6.34188592e-01 -4.47688103e-01 -1.82414129e-01 -4.75871801e-01 -1.04548371e+00 3.37315276e-02 3.32379371e-01 -8.59175980e-01 -7.11110532e-01 6.14546299e-01 8.20815265e-01 6.01778865e-01 6.11026883e-01 -1.25674963e+00 -9.23573792e-01 3.45672369e-01 -2.19775185e-01 -1.32135123e-01 1.06253898e+00 9.85692680e-01 7.85633326e-01 -1.05105793e+00 2.14373633e-01 1.96927524e+00 1.59194306e-01 1.11591756e+00 -7.62735307e-01 -3.03656220e-01 3.70793082e-02 1.86698765e-01 -7.88660228e-01 -6.00985765e-01 2.99565643e-01 -2.38123566e-01 1.05952537e+00 4.41991538e-01 2.52405256e-01 1.40586197e+00 -2.36914475e-02 1.12120402e+00 9.10752654e-01 -2.72757947e-01 -4.67791706e-01 1.92820340e-01 1.90198302e-01 1.21408308e+00 -8.10409367e-01 -3.22357528e-02 -7.02431798e-01 8.10761750e-02 8.67196560e-01 -1.16573691e-01 -5.16369224e-01 2.25596830e-01 -1.41453314e+00 9.39022958e-01 2.83895493e-01 -1.52699780e-02 -2.14481816e-01 -9.10066292e-02 5.32214582e-01 3.31303895e-01 -6.67976812e-02 3.87859285e-01 -5.22880077e-01 -4.77587879e-01 -1.88228101e-01 2.89785694e-02 9.93563235e-01 1.11053181e+00 6.67377293e-01 -1.70949847e-01 -1.11017680e+00 1.27872157e+00 4.68024999e-01 7.91491568e-01 4.95473713e-01 -1.36976194e+00 9.57614541e-01 7.11997867e-01 6.91653416e-02 -6.60211921e-01 -4.48487997e-01 3.36720318e-01 -7.14828491e-01 -2.42344931e-01 7.99419880e-01 -6.36776507e-01 -5.47714531e-01 1.59306002e+00 3.43736827e-01 -4.35084820e-01 2.77065188e-01 1.00876129e+00 1.75603008e+00 1.00096941e+00 8.46605450e-02 1.08052418e-01 1.75928760e+00 -1.48748124e+00 -8.81788433e-01 -2.72270858e-01 7.09107280e-01 -9.26211476e-01 1.83202362e+00 1.26556724e-01 -1.07788479e+00 -8.95597816e-01 -2.29519486e-01 -5.90125620e-01 -2.40825474e-01 3.30111206e-01 2.60077059e-01 -3.58886644e-02 -9.76993203e-01 -1.36405677e-01 -3.51146847e-01 -5.22074103e-01 1.65907126e-02 2.23079935e-01 -3.91896158e-01 -2.44765192e-01 -1.39563000e+00 1.05999625e+00 1.44418716e-01 1.67868182e-01 -9.04079974e-01 -2.26591215e-01 -1.14466083e+00 9.09410939e-02 4.23832983e-01 -8.34024668e-01 1.69311988e+00 -5.04624784e-01 -2.01567554e+00 7.85980463e-01 -3.01388144e-01 -5.79775758e-02 5.88201225e-01 -2.98929006e-01 -6.82949722e-02 4.44142282e-01 -2.29711849e-02 1.24136603e+00 5.45633197e-01 -1.16206753e+00 -5.26878178e-01 -2.04197958e-01 4.32205141e-01 7.09386528e-01 -1.61999524e-01 -4.57309037e-02 -1.13017428e+00 -1.75069235e-02 -5.66214204e-01 -9.46625710e-01 8.99746791e-02 -1.59756035e-01 -5.60029447e-01 -7.59052455e-01 7.37965047e-01 -7.50842333e-01 1.08004105e+00 -2.05590701e+00 3.74077521e-02 -4.63687688e-01 1.80500746e-01 3.13890338e-01 -7.01505125e-01 5.75640500e-01 4.56809461e-01 -2.53024369e-01 1.16384625e-01 -5.50587833e-01 2.62697160e-01 4.12666947e-01 6.48877099e-02 2.27082223e-02 1.87048808e-01 1.40720379e+00 -9.26878810e-01 -7.15568542e-01 5.09840131e-01 4.69571829e-01 -5.89468837e-01 9.95589554e-01 -4.37192112e-01 8.56602550e-01 -5.51349699e-01 1.65766031e-01 1.42856613e-01 -4.93273020e-01 -3.01551551e-01 -3.77566844e-01 -2.12143641e-02 2.28554755e-01 -6.24917686e-01 1.75221658e+00 -7.39686131e-01 6.02601290e-01 4.59111243e-01 -3.70356977e-01 6.74377322e-01 3.75382751e-01 -5.59949689e-02 -9.37373757e-01 3.33161712e-01 -3.66571754e-01 -7.61303231e-02 -1.28442168e+00 5.72334826e-01 4.06297684e-01 -2.31841862e-01 5.45552909e-01 2.61480421e-01 -1.33708194e-01 3.33940864e-01 5.11207819e-01 7.10893095e-01 6.99099675e-02 -2.87952088e-02 2.64358938e-01 7.57750213e-01 -5.25379181e-02 -3.90107483e-02 8.15679371e-01 -2.15142101e-01 5.56040466e-01 4.19283152e-01 -4.05115150e-02 -5.02685905e-01 -8.05065215e-01 4.24605221e-01 2.09366083e+00 -5.83508313e-02 -2.64821321e-01 -8.80237401e-01 -8.43833208e-01 -1.99958220e-01 8.87260616e-01 -4.39823508e-01 -1.06378179e-02 -5.68201482e-01 -2.40564033e-01 3.89594793e-01 1.72062173e-01 9.24549639e-01 -1.51584733e+00 -4.62403923e-01 3.38758081e-02 -9.09018219e-01 -1.27748811e+00 -1.22842252e+00 -1.93216503e-01 -2.41759345e-01 -1.15062392e+00 -1.00196099e+00 -9.30091858e-01 6.73397183e-01 2.74176359e-01 1.26164746e+00 3.42770666e-01 6.61242232e-02 1.09609389e+00 -4.76539940e-01 -5.10887690e-02 -6.52097523e-01 -1.04942814e-01 -1.83895573e-01 3.29812132e-02 3.24274808e-01 1.29655868e-01 -6.40322864e-01 6.72231257e-01 -4.51176345e-01 4.29682225e-01 4.14521366e-01 9.58285034e-01 1.16506768e-02 -7.30885327e-01 3.45751226e-01 -5.87698698e-01 1.19767892e+00 -3.52788717e-01 -2.61778325e-01 4.12172765e-01 1.37361184e-01 6.05560951e-02 5.04868984e-01 -7.16508567e-01 -1.14481413e+00 -1.46403462e-01 -4.01790887e-01 -4.06885117e-01 -4.15687710e-01 2.61392742e-01 -2.72534341e-01 9.06670541e-02 4.60180491e-01 4.13406074e-01 2.18360752e-01 -2.40785584e-01 7.82021582e-01 9.61823642e-01 8.37023735e-01 -8.30608904e-01 3.87233764e-01 -2.14147791e-01 -5.89937389e-01 -9.81684804e-01 -7.17351198e-01 -4.22255546e-01 -3.52054983e-01 -5.20047486e-01 1.10245323e+00 -7.85117745e-01 -1.60835409e+00 3.77824724e-01 -1.57535696e+00 -7.83700407e-01 2.24795401e-01 3.36589068e-01 -5.05040705e-01 4.95190859e-01 -9.20741081e-01 -8.28192651e-01 -5.28887630e-01 -1.43091035e+00 1.07576060e+00 6.15840793e-01 -3.05520624e-01 -1.01998007e+00 -1.46545872e-01 1.05964017e+00 3.70824039e-01 -3.63017678e-01 6.78265750e-01 -9.27375913e-01 -3.03894311e-01 5.34936339e-02 -3.76183420e-01 2.36168727e-01 5.91944680e-02 -1.99650899e-01 -9.06633317e-01 -2.54233658e-01 -1.21726513e-01 -8.47598135e-01 5.63265979e-01 1.77700594e-01 9.50395346e-01 -6.17291152e-01 1.17574781e-02 2.34940901e-01 5.28258383e-01 1.04807936e-01 3.04103434e-01 -1.94961518e-01 9.68256295e-01 1.06277430e+00 7.68869042e-01 3.36367190e-01 1.13722444e+00 6.06609046e-01 3.23502392e-01 -1.99119046e-01 -1.01902954e-01 -3.02334726e-01 5.94259739e-01 1.04162133e+00 2.46934846e-01 -5.71004272e-01 -7.87022412e-01 3.24382186e-01 -1.94599736e+00 -8.48443568e-01 -2.21038520e-01 1.86347926e+00 1.01416433e+00 -2.81788111e-01 1.60863683e-01 -6.78088129e-01 7.38345504e-01 -1.77942477e-02 -4.37215656e-01 -4.52829689e-01 1.62365776e-03 -4.43362981e-01 -1.15696989e-01 1.01285970e+00 -8.16375136e-01 1.09869635e+00 5.46446037e+00 7.13859200e-01 -8.86161685e-01 4.21982855e-02 5.26884377e-01 8.22646543e-02 -8.92857462e-02 -4.73816425e-01 -8.67657542e-01 5.61130881e-01 7.96440542e-01 3.18037756e-02 5.39878249e-01 3.18560451e-01 5.26513994e-01 -2.38930911e-01 -1.17930281e+00 1.26123118e+00 1.81611791e-01 -8.63304973e-01 2.26763979e-01 -1.49655506e-01 2.28602618e-01 3.58824581e-02 1.38822824e-01 8.71416688e-01 5.27130008e-01 -1.09100854e+00 1.51877597e-01 7.47239232e-01 8.39324653e-01 -4.00407791e-01 5.82144201e-01 5.91993153e-01 -1.05583608e+00 2.67329831e-02 -6.10512942e-02 1.92869514e-01 2.27051303e-01 -3.02212685e-01 -1.01901317e+00 3.42884451e-01 4.99413550e-01 2.33670041e-01 -4.99290317e-01 6.62769258e-01 -3.74307722e-01 2.52015799e-01 -2.77237535e-01 -3.17458183e-01 3.76009256e-01 -2.28183687e-01 3.87192786e-01 1.44763541e+00 1.24555901e-01 3.56675774e-01 4.61867660e-01 6.04653299e-01 -1.45231411e-01 2.43900552e-01 -5.65757453e-01 -7.18090683e-02 3.68115932e-01 1.38029373e+00 -1.77975316e-02 -2.84668326e-01 -6.96283400e-01 1.22762001e+00 3.08493376e-01 6.41570687e-01 -7.57614195e-01 -3.23077410e-01 4.28563327e-01 -2.33582333e-01 -2.06951201e-02 -2.37285718e-01 2.50369519e-01 -1.10925174e+00 -3.26806873e-01 -1.20204794e+00 4.95155782e-01 -1.12165320e+00 -1.35653639e+00 7.03944445e-01 -1.77750036e-01 -9.36810434e-01 -3.59544516e-01 -7.39825845e-01 -7.42070079e-01 1.04994786e+00 -1.20968008e+00 -1.15917110e+00 -5.12300789e-01 1.02030778e+00 1.16550839e+00 -3.34792793e-01 9.03939247e-01 5.52289933e-03 -6.26059532e-01 7.53313065e-01 -4.82479721e-01 4.84323591e-01 1.29615140e+00 -1.22395384e+00 4.99361567e-02 2.54577458e-01 -1.29803613e-01 6.70965374e-01 7.18185723e-01 -5.60136080e-01 -1.48665285e+00 -9.22338605e-01 6.67806387e-01 -7.85791755e-01 4.16192591e-01 -2.72249043e-01 -1.10282969e+00 4.57877517e-01 1.01644802e+00 -3.73689801e-01 7.19189405e-01 -4.14647907e-03 -7.98784420e-02 2.47032449e-01 -9.32236850e-01 8.29143822e-01 5.17019153e-01 -7.48614669e-01 -8.65830541e-01 5.87025702e-01 9.01314974e-01 -7.29931831e-01 -7.29764760e-01 3.28460671e-02 4.64831561e-01 -7.65300572e-01 8.54629457e-01 -6.24033213e-01 4.07176733e-01 -9.06133726e-02 6.08119592e-02 -1.37930846e+00 -9.97078642e-02 -9.75777268e-01 -8.88601393e-02 1.26220214e+00 4.30638105e-01 -4.50275600e-01 3.02886158e-01 7.18861759e-01 -1.83148503e-01 -6.48365617e-01 -5.47551811e-01 -6.07209094e-02 1.71860605e-02 -3.07099335e-02 3.96313697e-01 7.21860945e-01 1.37555584e-01 1.08865392e+00 -5.49960554e-01 7.67733380e-02 2.13868022e-01 -1.35964215e-01 1.21821749e+00 -8.86192739e-01 -2.80910552e-01 -4.21904176e-01 6.79935753e-01 -1.95729256e+00 1.01304784e-01 -6.32906020e-01 2.78474361e-01 -1.76457334e+00 8.26455131e-02 2.02707499e-01 3.15167964e-01 6.49813712e-01 -5.33885002e-01 -2.66272187e-01 4.45341527e-01 6.63846657e-02 -1.13459826e+00 8.87398005e-01 1.99560964e+00 -2.43412554e-01 -2.76001632e-01 -4.54854630e-02 -4.77256566e-01 5.92702627e-01 5.90557277e-01 6.58902451e-02 -4.20510679e-01 -6.79880142e-01 -2.00022057e-01 7.09785461e-01 3.13163579e-01 -6.50687575e-01 6.54740870e-01 -1.89374894e-01 3.09101820e-01 -6.53832138e-01 5.87032974e-01 -5.13712049e-01 -6.86861217e-01 1.56638548e-01 -5.62538564e-01 2.55418956e-01 3.11127216e-01 2.52508640e-01 -4.53814603e-02 -1.00172378e-01 6.06779039e-01 -4.72932249e-01 -7.11561620e-01 3.34598333e-01 -6.13177180e-01 2.53195822e-01 5.40763021e-01 1.97325066e-01 -7.74917245e-01 -1.00672317e+00 -7.87325203e-01 1.29398894e+00 1.45992115e-01 6.89016521e-01 8.55599642e-01 -1.12728989e+00 -9.77045238e-01 -6.97294474e-02 2.70145744e-01 1.06223650e-01 6.43368244e-01 8.42093289e-01 -4.21191305e-01 6.44802451e-01 -1.11789107e-01 -6.70667291e-01 -1.51379836e+00 3.69187474e-01 5.01516938e-01 -1.28437757e-01 -3.91718298e-01 1.00779021e+00 6.01562500e-01 -1.10195279e+00 5.64164758e-01 -3.24379057e-02 -5.95381677e-01 1.11836633e-02 6.81205034e-01 1.50972724e-01 -5.17241597e-01 -7.36475170e-01 -7.53483549e-02 5.00169635e-01 -1.09388024e-01 -2.60168344e-01 7.47873724e-01 -6.42607987e-01 -2.62490120e-02 5.19641161e-01 1.08260870e+00 -1.60022214e-01 -1.42075479e+00 -2.70993054e-01 -6.77635908e-01 -6.18614890e-02 -4.17338431e-01 -1.12722933e+00 -7.87423193e-01 1.33107233e+00 4.40572262e-01 1.83045641e-01 8.16609442e-01 7.49312267e-02 8.17534208e-01 9.01440859e-01 1.24113016e-01 -1.05438137e+00 6.33674204e-01 1.08530998e+00 1.36068916e+00 -1.68911541e+00 -6.08984411e-01 -6.25025183e-02 -1.24995732e+00 1.11308444e+00 1.28889704e+00 3.96801502e-01 -1.10841468e-02 -3.44674103e-02 7.05722034e-01 -2.67516613e-01 -1.10817218e+00 -4.05971378e-01 4.48409647e-01 5.90199292e-01 5.63484669e-01 -1.40755087e-01 1.57838628e-01 5.72421491e-01 -2.38038734e-01 -3.47029179e-01 2.27932230e-01 4.60653752e-01 -4.58822340e-01 -8.90926003e-01 -4.85616952e-01 3.06705207e-01 -2.15658937e-02 -2.02900440e-01 -4.30828869e-01 5.64821601e-01 -3.63723516e-01 1.58692908e+00 -1.18571356e-01 -3.24735731e-01 6.61106467e-01 2.09655941e-01 3.29236567e-01 -7.84522295e-01 -1.03488958e+00 1.11928061e-01 4.07426327e-01 -6.83649719e-01 -9.73219518e-03 -2.12553740e-01 -1.34101856e+00 -3.46252739e-01 -8.75470042e-02 3.87067318e-01 1.94583446e-01 1.04622328e+00 1.98511124e-01 6.26086414e-01 5.99048078e-01 -8.61350775e-01 -4.13018316e-01 -1.46153855e+00 2.01579556e-01 3.25043350e-01 4.90203291e-01 -3.28295529e-01 -3.15771580e-01 -6.28228709e-02]
[10.936434745788574, 1.412665605545044]
a4fa52c3-c2f8-4b30-b217-c2fe2c050ff2
discovering-optimal-scoring-mechanisms-in
2302.06804
null
https://arxiv.org/abs/2302.06804v2
https://arxiv.org/pdf/2302.06804v2.pdf
Discovering Optimal Scoring Mechanisms in Causal Strategic Prediction
Faced with data-driven policies, individuals will manipulate their features to obtain favorable decisions. While earlier works cast these manipulations as undesirable gaming, recent works have adopted a more nuanced causal framing in which manipulations can improve outcomes of interest, and setting coherent mechanisms requires accounting for both predictive accuracy and improvement of the outcome. Typically, these works focus on known causal graphs, consisting only of an outcome and its parents. In this paper, we introduce a general framework in which an outcome and n observed features are related by an arbitrary unknown graph and manipulations are restricted by a fixed budget and cost structure. We develop algorithms that leverage strategic responses to discover the causal graph in a finite number of steps. Given this graph structure, we can then derive mechanisms that trade off between accuracy and improvement. Altogether, our work deepens links between causal discovery and incentive design and provides a more nuanced view of learning under causal strategic prediction.
['Zachary Lipton', 'Shantanu Gupta', 'Tom Yan']
2023-02-14
null
null
null
null
['causal-discovery']
['knowledge-base']
[ 3.53793174e-01 6.94253385e-01 -9.95562613e-01 -3.94369572e-01 -1.49004117e-01 -6.81720018e-01 5.56292892e-01 1.26300499e-01 -3.41957271e-01 9.04131174e-01 6.23936594e-01 -4.12628680e-01 -7.00161815e-01 -1.08627641e+00 -6.97391808e-01 -3.20307851e-01 -3.76446456e-01 1.66179538e-01 -2.88837880e-01 4.59018238e-02 3.20195973e-01 -5.50095923e-02 -1.30669045e+00 -1.19038329e-01 7.87675858e-01 4.90510046e-01 -1.77205969e-02 7.86775827e-01 5.17059386e-01 1.05257344e+00 -3.12337041e-01 -7.58932233e-01 5.59873581e-01 -4.53123540e-01 -5.83140373e-01 1.33951798e-01 1.68200985e-01 -7.30506837e-01 -4.94470477e-01 9.63714182e-01 3.51929307e-01 -1.25996724e-01 5.17872155e-01 -1.44792616e+00 -1.19509566e+00 1.40594411e+00 -6.18724942e-01 1.83081284e-01 3.48957807e-01 4.38287288e-01 1.73206580e+00 2.67002285e-02 7.63442814e-01 1.41034448e+00 3.11326712e-01 5.55650353e-01 -1.66037059e+00 -7.91146874e-01 6.65083706e-01 -6.61945269e-02 -4.54383194e-01 -2.81636298e-01 7.39124238e-01 -6.40860140e-01 3.60030264e-01 3.65113646e-01 9.80188310e-01 1.41797507e+00 -2.58795805e-02 6.24537885e-01 1.39013898e+00 -3.25195581e-01 2.25728706e-01 -2.14741483e-01 5.68786800e-01 4.37145531e-01 8.93557966e-01 8.80320907e-01 -7.79707372e-01 -3.12150419e-01 8.06522548e-01 1.85494218e-02 -2.82701820e-01 -4.91586745e-01 -6.74598396e-01 1.17231131e+00 2.79651940e-01 -1.29654005e-01 -5.93504012e-01 5.50402761e-01 -7.27973208e-02 5.51547945e-01 1.12115182e-01 1.01581693e+00 -4.61012721e-01 -2.42850021e-01 -3.07739317e-01 6.43997431e-01 7.52883255e-01 6.87827468e-01 7.51531243e-01 -2.33665764e-01 -4.68571901e-01 1.46248579e-01 2.13895246e-01 2.78899699e-01 -7.94147775e-02 -1.51229155e+00 4.14743602e-01 4.92944598e-01 4.76426333e-01 -1.15217435e+00 -4.27799135e-01 -5.57466745e-01 -3.58639419e-01 1.48423165e-02 6.94803536e-01 -7.10330248e-01 -6.22860312e-01 2.47756004e+00 2.34172985e-01 2.01220438e-01 -3.75031799e-01 9.17360842e-01 9.09861177e-02 1.00547925e-01 1.81452870e-01 -3.49730611e-01 9.77367938e-01 -3.85534972e-01 -5.60850620e-01 -3.13235015e-01 5.52037835e-01 -1.38104424e-01 9.84846592e-01 7.27620497e-02 -1.23728311e+00 9.05457363e-02 -6.14083230e-01 1.11639105e-01 3.14408571e-01 -4.72201169e-01 1.21495330e+00 8.73129904e-01 -1.09573245e+00 7.16349959e-01 -4.88465518e-01 -1.98756129e-01 5.91964304e-01 5.18525302e-01 2.86858469e-01 1.12187259e-01 -1.26228356e+00 5.92984617e-01 -1.01789586e-01 -2.04949051e-01 -1.19690454e+00 -1.20962703e+00 -3.60920191e-01 3.08160484e-01 1.17934859e+00 -1.27174425e+00 1.43667340e+00 -1.02377868e+00 -1.15763402e+00 4.11317170e-01 3.38159710e-01 -5.03803432e-01 6.17599130e-01 -2.27264985e-01 1.23054326e-01 -3.93040895e-01 3.36563408e-01 2.64780849e-01 8.32877159e-01 -1.13176703e+00 -7.60558069e-01 -5.09998977e-01 8.76853049e-01 4.50971246e-01 -4.37229216e-01 -1.07323527e-01 2.71857738e-01 -3.89756918e-01 -5.46068788e-01 -9.17089105e-01 -6.78277493e-01 -4.46047753e-01 -7.94246972e-01 2.63191629e-02 2.86292702e-01 -8.65764916e-02 1.33309186e+00 -1.67470729e+00 1.16685212e-01 2.68211663e-01 8.59541953e-01 -4.25634235e-01 -3.13752860e-01 2.71168083e-01 3.95031460e-02 8.31194162e-01 1.23978473e-01 -9.36694145e-02 1.96196839e-01 -8.47322680e-03 -4.70462233e-01 4.22730923e-01 -8.98536220e-02 9.85780060e-01 -9.43750620e-01 -2.11036336e-02 -1.84493542e-01 -7.07016587e-02 -1.14965379e+00 1.28668159e-01 -2.20184177e-01 2.76748031e-01 -8.79791617e-01 4.00836647e-01 1.42864183e-01 -3.10227096e-01 7.23355174e-01 4.54964221e-01 -2.06805244e-01 5.88284910e-01 -1.09801459e+00 8.68505776e-01 -1.98710803e-03 2.81474322e-01 2.46417567e-01 -7.96565175e-01 2.51652569e-01 -8.77290145e-02 4.09350365e-01 -4.20825899e-01 1.68965027e-01 -2.22904518e-01 3.58238459e-01 -4.50134903e-01 4.97167617e-01 -2.23809183e-02 -1.95061475e-01 7.25546241e-01 -2.99036801e-01 2.34367952e-01 -9.85591561e-02 4.28451002e-01 1.35675132e+00 -6.05406314e-02 5.43456078e-01 -3.94530654e-01 -3.83958906e-01 1.56537578e-01 8.55066836e-01 1.51371956e+00 -9.34029892e-02 5.39785437e-02 1.26430392e+00 -1.10050470e-01 -1.20131993e+00 -7.27067888e-01 1.95646271e-01 1.25079191e+00 1.55869275e-01 -2.63763607e-01 -4.68844146e-01 -6.09128833e-01 6.52307332e-01 7.56348133e-01 -1.15399170e+00 -1.96190715e-01 -2.40141451e-01 -8.29049528e-01 2.36393392e-01 4.09253925e-01 2.70230323e-01 -5.86980462e-01 -5.67670643e-01 -2.69695129e-02 1.28333896e-01 -3.46886545e-01 -6.42755568e-01 -2.12299779e-01 -7.89356530e-01 -1.24223363e+00 -8.11408609e-02 1.47426326e-03 4.10549968e-01 4.46849465e-01 1.02379107e+00 2.74014205e-01 -3.91317345e-02 6.28134787e-01 -6.31814152e-02 -4.84614611e-01 -4.03264612e-02 5.17333187e-02 1.24366917e-01 -8.76906738e-02 1.51340276e-01 -8.72499704e-01 -7.79069662e-01 5.79320788e-02 -4.23055261e-01 7.42140487e-02 5.93208671e-01 7.73585141e-01 6.67164251e-02 -2.56334037e-01 6.87254190e-01 -1.37448621e+00 1.02579439e+00 -9.44842041e-01 -9.16763902e-01 2.33450890e-01 -1.09749663e+00 1.72152191e-01 4.62548673e-01 -5.89636326e-01 -1.33961332e+00 -2.26958662e-01 7.39536822e-01 -1.21799313e-01 1.08049236e-01 7.22047150e-01 -2.86282182e-01 1.10163763e-01 8.60544145e-01 -4.56022531e-01 -1.80151835e-02 -3.89654696e-01 7.28139400e-01 1.85869440e-01 1.36190400e-01 -8.83819342e-01 8.26119006e-01 2.82084912e-01 2.21336946e-01 -1.31949723e-01 -1.03434539e+00 1.89632148e-01 -2.95278549e-01 -2.17673168e-01 5.88778913e-01 -8.06637526e-01 -1.41922152e+00 7.80421942e-02 -6.71053410e-01 -5.36501348e-01 -4.02524799e-01 4.88060296e-01 -6.30770802e-01 -9.97482464e-02 -4.57095116e-01 -1.20906246e+00 4.68920350e-01 -8.97276163e-01 5.43305635e-01 2.32350975e-01 -3.07931244e-01 -9.37313080e-01 4.64920290e-02 3.35573375e-01 4.77255315e-01 3.44638973e-02 9.81213331e-01 -5.24315178e-01 -1.30737579e+00 1.97279230e-01 -1.02281809e-01 -7.36955762e-01 6.48763180e-02 1.35386765e-01 -7.22707450e-01 -1.72080263e-01 -8.84446949e-02 -3.76203269e-01 7.74046183e-01 9.10665154e-01 1.06155205e+00 -9.62759376e-01 -5.84030628e-01 4.60460484e-01 1.16594005e+00 3.75046939e-01 1.48124292e-01 2.53455520e-01 5.81645191e-01 8.84692907e-01 4.07343626e-01 7.40894318e-01 6.63412273e-01 6.81215703e-01 6.01391673e-01 1.64928213e-01 2.43998960e-01 -8.08073282e-01 1.65462643e-01 -2.19375655e-01 -4.00270015e-01 -2.47666910e-01 -5.22047460e-01 5.88355064e-01 -2.03639984e+00 -1.34709322e+00 -3.22607905e-02 2.36034012e+00 9.25941825e-01 1.69418439e-01 5.87757647e-01 -1.05961807e-01 8.04607272e-01 2.27524519e-01 -8.63727033e-01 -1.36888832e-01 -3.43849249e-02 -1.80010647e-01 9.44258392e-01 8.47227514e-01 -6.50792181e-01 5.98074675e-01 7.29247522e+00 2.76184529e-01 -7.75942683e-01 -3.00078541e-02 9.62399244e-01 -4.49363202e-01 -1.06041241e+00 4.15741801e-01 -8.42656314e-01 2.53514826e-01 8.55586886e-01 -8.34689558e-01 7.36751556e-01 8.41396809e-01 6.11713529e-01 -3.95558365e-02 -1.36003757e+00 1.93857864e-01 -5.55999339e-01 -1.50796163e+00 -1.85331836e-01 6.24947071e-01 1.01517963e+00 -5.85158646e-01 4.86297190e-01 1.92810938e-01 1.59200370e+00 -1.31725895e+00 7.16270506e-01 4.03410107e-01 5.97412527e-01 -6.45821393e-01 7.21956491e-02 3.37678134e-01 -5.18280804e-01 -9.07699764e-01 -1.61370695e-01 -9.15374577e-01 -9.04690251e-02 6.36805117e-01 -6.76579773e-01 2.80203521e-01 3.54922742e-01 7.21876323e-01 -2.51136392e-01 7.33530760e-01 -4.61843729e-01 1.02908170e+00 1.99092224e-01 3.54810506e-02 -1.02600246e-03 -3.03349614e-01 6.32526278e-01 4.83854353e-01 6.00428768e-02 5.22957444e-01 3.16392064e-01 1.20207644e+00 -4.22601968e-01 -1.72697693e-01 -7.90401936e-01 -2.14384824e-01 8.07055950e-01 9.34517264e-01 -1.82182834e-01 -8.89135078e-02 -3.76270801e-01 9.94752422e-02 4.83004004e-01 4.18278128e-01 -6.81284487e-01 4.37087506e-01 1.20370471e+00 1.60365239e-01 -1.08876355e-01 1.05544984e-01 -9.13914919e-01 -1.23278582e+00 -2.39451513e-01 -7.30044246e-01 6.23413682e-01 -4.42005128e-01 -1.32899463e+00 -4.89490688e-01 1.63209587e-01 -5.06544352e-01 -2.13462919e-01 -1.23290390e-01 -5.71580708e-01 6.99090064e-01 -1.45940328e+00 -7.14146852e-01 1.48645088e-01 2.73222983e-01 2.94671804e-01 3.49699706e-01 5.40891945e-01 -2.02890739e-01 -8.48363400e-01 4.84188884e-01 -1.88582912e-01 -1.40812963e-01 3.71353477e-01 -1.41124237e+00 1.79168761e-01 8.83849859e-01 -2.47665524e-01 8.29387665e-01 8.75275493e-01 -9.90553916e-01 -1.49168456e+00 -9.67562795e-01 6.39349699e-01 -5.84101677e-01 1.01153982e+00 -2.54816025e-01 -4.14855957e-01 1.14139962e+00 1.07088082e-01 -3.48381490e-01 4.73195225e-01 7.99159467e-01 -2.50845462e-01 5.60087198e-03 -1.08588040e+00 1.06932366e+00 1.82532358e+00 -2.58495808e-01 -3.68138880e-01 9.31209102e-02 1.00325274e+00 -1.07814558e-01 -9.09975708e-01 9.74387601e-02 4.94951963e-01 -1.06230867e+00 7.88611591e-01 -1.26690185e+00 8.67904246e-01 1.32918641e-01 -2.76505630e-02 -1.54167068e+00 -6.91073895e-01 -1.06259120e+00 2.17787981e-01 1.14103472e+00 6.73893392e-01 -6.05692029e-01 8.98382306e-01 1.10394657e+00 2.36612573e-01 -4.19643790e-01 -4.09281611e-01 -4.94462967e-01 1.68659091e-01 -2.95235515e-01 9.04830873e-01 1.12054598e+00 2.36976117e-01 4.06744421e-01 -8.37610185e-01 2.17845440e-01 1.03475022e+00 3.83623064e-01 7.05497861e-01 -1.50572979e+00 -5.88202000e-01 -6.54259980e-01 -2.06975602e-02 -9.58033800e-01 2.23708183e-01 -5.96313715e-01 -3.41961145e-01 -1.03926408e+00 8.17617595e-01 -8.10383618e-01 -6.41060993e-02 5.42645156e-01 -5.38073003e-01 -3.59149128e-01 3.27613115e-01 2.04864696e-01 -3.27341318e-01 2.33648449e-01 1.17777205e+00 1.77725866e-01 -4.23795521e-01 1.38925850e-01 -1.64083350e+00 5.81231058e-01 6.38755322e-01 -3.98855478e-01 -7.42177486e-01 -3.09512705e-01 5.66907823e-01 7.18093276e-01 6.09597266e-01 1.07620992e-01 1.06323719e-01 -1.13200355e+00 1.63927540e-01 1.77715242e-01 -5.00667244e-02 -4.92426693e-01 3.87170017e-01 4.71948177e-01 -1.29525375e+00 -1.08459286e-01 -6.18423104e-01 9.20677722e-01 4.82765496e-01 1.32826298e-01 4.92196053e-01 -2.34391138e-01 -2.45402515e-01 4.32067961e-01 -2.39703074e-01 8.05251449e-02 1.13889420e+00 2.06993963e-03 -7.70652890e-01 -9.13552701e-01 -7.07511961e-01 3.23493928e-01 7.72405922e-01 2.52260178e-01 3.76154006e-01 -1.16003931e+00 -6.40868187e-01 -2.43728802e-01 -1.41373813e-01 -4.48262453e-01 7.01475888e-02 6.94206178e-01 5.71167111e-01 2.14099064e-01 -2.38277867e-01 -6.23801984e-02 -1.01883173e+00 7.31857717e-01 3.97149801e-01 -2.71137178e-01 -1.62133858e-01 7.87567973e-01 6.37063920e-01 -1.15161188e-01 -1.42928713e-03 4.50824481e-03 -2.74256557e-01 1.56392127e-01 3.00029516e-01 4.53336149e-01 -6.84526980e-01 -1.64958350e-02 -3.45384590e-02 4.49185073e-02 -1.39748733e-02 -2.94417530e-01 1.33815682e+00 -4.82692987e-01 -3.79419960e-02 2.46571124e-01 4.44226116e-01 3.65363270e-01 -1.72350240e+00 -1.96884155e-01 2.61762977e-01 -1.09540832e+00 -2.25516651e-02 -8.71405959e-01 -1.22497940e+00 1.55765519e-01 -6.98759332e-02 6.08454823e-01 9.31259990e-01 -6.46781325e-02 -1.24110863e-01 -2.55109463e-02 4.33567107e-01 -6.49986088e-01 -8.33154656e-03 1.03913313e-02 4.83966231e-01 -1.10385501e+00 9.83978156e-03 -6.57880187e-01 -6.02742076e-01 5.73485136e-01 6.43669724e-01 -3.61866727e-02 5.21430850e-01 2.67237544e-01 -5.22165835e-01 -2.02338815e-01 -1.30700696e+00 -4.03265834e-01 -6.54080883e-02 7.81849027e-01 2.60827541e-01 6.15623713e-01 -7.24229157e-01 8.90889883e-01 -4.53155279e-01 6.67300075e-02 1.23557758e+00 6.59290433e-01 -4.26227480e-01 -1.22828853e+00 -4.00141060e-01 1.07425857e+00 -7.44012058e-01 -2.19560012e-01 -6.77366853e-01 8.63328934e-01 -1.86868846e-01 1.11822772e+00 4.79046293e-02 -3.17162216e-01 3.59191388e-01 -3.79017383e-01 3.47737670e-01 -7.22076654e-01 -4.80955452e-01 -8.19651037e-02 4.01431262e-01 -8.21314931e-01 -4.64635879e-01 -1.04724491e+00 -5.34728408e-01 -7.05788195e-01 -2.07035169e-01 -5.46122640e-02 1.61475632e-02 7.78993726e-01 4.42154109e-01 3.43961686e-01 9.80699897e-01 -3.22255611e-01 -9.60586607e-01 -4.69844043e-01 -7.90187418e-01 4.37536687e-01 4.62404519e-01 -9.22921777e-01 -3.92656386e-01 -2.26926878e-01]
[8.136791229248047, 5.393988132476807]
f6b42750-10c5-4acf-b4b3-3deb8bb7b46a
longt5-efficient-text-to-text-transformer-for
2112.07916
null
https://arxiv.org/abs/2112.07916v2
https://arxiv.org/pdf/2112.07916v2.pdf
LongT5: Efficient Text-To-Text Transformer for Long Sequences
Recent work has shown that either (1) increasing the input length or (2) increasing model size can improve the performance of Transformer-based neural models. In this paper, we present a new model, called LongT5, with which we explore the effects of scaling both the input length and model size at the same time. Specifically, we integrated attention ideas from long-input transformers (ETC), and adopted pre-training strategies from summarization pre-training (PEGASUS) into the scalable T5 architecture. The result is a new attention mechanism we call {\em Transient Global} (TGlobal), which mimics ETC's local/global attention mechanism, but without requiring additional side-inputs. We are able to achieve state-of-the-art results on several summarization tasks and outperform the original T5 models on question answering tasks.
['Yinfei Yang', 'Yun-Hsuan Sung', 'Jianmo Ni', 'Santiago Ontanon', 'David Uthus', 'Joshua Ainslie', 'Mandy Guo']
2021-12-15
null
https://aclanthology.org/2022.findings-naacl.55
https://aclanthology.org/2022.findings-naacl.55.pdf
findings-naacl-2022-7
['long-range-modeling']
['natural-language-processing']
[ 2.95052469e-01 5.82307696e-01 1.49781823e-01 -2.10061267e-01 -8.25324774e-01 -4.56110299e-01 7.11224198e-01 1.39666572e-01 -5.04682064e-01 3.79982501e-01 5.77249706e-01 -6.50802314e-01 1.10330749e-02 -7.23367274e-01 -9.21067059e-01 -1.86751485e-01 2.83699989e-01 5.98220766e-01 5.09970427e-01 -3.04145396e-01 3.33361417e-01 5.14527000e-02 -1.30556381e+00 7.10862100e-01 1.19162822e+00 9.94159102e-01 3.61203760e-01 9.15873051e-01 -4.25373048e-01 1.14041543e+00 -7.34844327e-01 -6.91668570e-01 -8.10976774e-02 -4.88087028e-01 -1.42587280e+00 -2.05363005e-01 7.01435268e-01 -3.57435733e-01 -4.16406244e-01 4.90938574e-01 5.10805905e-01 2.69384414e-01 3.17411512e-01 -8.13409865e-01 -1.03511906e+00 1.14891768e+00 -1.82937577e-01 3.90100598e-01 5.16238250e-02 1.23539016e-01 1.39258862e+00 -7.73680329e-01 3.22188377e-01 1.39621675e+00 8.52733552e-01 4.84621644e-01 -1.27217174e+00 -2.52310693e-01 4.67694551e-01 4.43921804e-01 -8.52693379e-01 -5.91758192e-01 6.25051618e-01 -5.70736360e-03 1.72790289e+00 4.35960889e-01 5.00250697e-01 9.73148942e-01 2.57493883e-01 1.05103850e+00 6.78534508e-01 -5.15180111e-01 1.31400302e-02 -1.34553924e-01 3.70708436e-01 5.39985776e-01 -1.27126902e-01 -4.35365289e-01 -4.31782395e-01 -6.24984279e-02 4.89945740e-01 -3.34969938e-01 -2.49945208e-01 -1.25092678e-02 -9.80168045e-01 9.17039454e-01 6.27707243e-01 4.07279849e-01 -3.40434313e-01 5.03353715e-01 5.52982867e-01 5.45378983e-01 5.70800424e-01 9.22274828e-01 -7.85085797e-01 -5.66900857e-02 -8.49791288e-01 2.28905842e-01 6.80800319e-01 7.74726808e-01 4.71224606e-01 1.57259881e-01 -7.65395880e-01 9.44610655e-01 -1.18330657e-01 9.52511355e-02 7.52482414e-01 -1.10976160e+00 7.86991298e-01 7.24722147e-01 -2.59800069e-02 -4.69997346e-01 -4.50784385e-01 -7.95879960e-01 -5.88805735e-01 -5.80605209e-01 2.79483616e-01 -1.36716276e-01 -1.05488253e+00 1.99063647e+00 -1.86124086e-01 -1.00248173e-01 -3.26219248e-04 3.29880536e-01 1.05331302e+00 7.40610600e-01 1.07576028e-02 1.28167853e-01 1.22577024e+00 -1.47377276e+00 -6.97700858e-01 -4.58490074e-01 9.01214361e-01 -6.72309637e-01 1.44024980e+00 1.96614414e-01 -1.61488020e+00 -6.87846839e-01 -8.52597713e-01 -5.79544127e-01 -4.06132668e-01 7.14909509e-02 5.24591923e-01 4.38853115e-01 -1.53700304e+00 8.26829791e-01 -8.63900244e-01 -7.04749823e-01 2.07584918e-01 3.92034769e-01 7.00777546e-02 1.19421780e-01 -1.11066556e+00 1.24066114e+00 2.94624031e-01 -6.32073265e-03 -6.80895209e-01 -7.16683269e-01 -7.58738518e-01 4.59417045e-01 4.43538904e-01 -1.36093783e+00 1.76530564e+00 -7.56105125e-01 -1.67199230e+00 5.95103264e-01 -3.47724944e-01 -8.69213223e-01 7.41526261e-02 -5.62835336e-01 2.21466646e-01 1.66112483e-01 -2.42178038e-01 8.94778728e-01 7.02275753e-01 -8.89896274e-01 -2.47189671e-01 -2.24620134e-01 2.85270482e-01 2.32602224e-01 -4.85735178e-01 9.12754908e-02 -5.26166499e-01 -6.54376268e-01 -1.99455127e-01 -6.17918432e-01 -1.16452038e-01 -5.22875667e-01 -4.93664145e-01 -5.55265725e-01 7.57909119e-01 -9.66010988e-01 1.47251630e+00 -1.64006805e+00 5.74161947e-01 -3.64667892e-01 1.23316638e-01 5.91376424e-01 -6.69020891e-01 6.23850703e-01 -1.52257770e-01 2.92820185e-01 -3.01666230e-01 -7.32355773e-01 7.93519467e-02 3.57645422e-01 -4.70441431e-01 -2.77360141e-01 3.60103071e-01 1.53507853e+00 -6.42690897e-01 -3.79268318e-01 2.90316157e-02 9.85134244e-02 -1.06410587e+00 2.15969458e-01 -5.84757447e-01 8.08661282e-02 -2.37328440e-01 1.49965540e-01 2.73490429e-01 -3.95708084e-01 -3.49442586e-02 -1.82445571e-01 8.35209191e-02 1.01199317e+00 -4.46040273e-01 1.74801898e+00 -6.22201204e-01 5.73774457e-01 -6.67453408e-02 -1.08340907e+00 6.81987941e-01 3.63309473e-01 1.22331887e-01 -8.71424854e-01 3.16849262e-01 2.32608989e-03 1.72878653e-01 -3.92310709e-01 8.63643408e-01 7.79659897e-02 1.17259875e-01 6.80005550e-01 4.24071223e-01 -1.60629645e-01 2.68017352e-01 5.52016318e-01 1.23292804e+00 5.08600771e-02 4.64432873e-02 -3.40926528e-01 3.28403950e-01 -2.69472837e-01 1.28922403e-01 9.90601599e-01 2.25243345e-01 6.50013149e-01 8.40912640e-01 -5.65155707e-02 -1.18263900e+00 -7.78869271e-01 2.61590809e-01 1.52863431e+00 -4.19669420e-01 -8.55913937e-01 -1.02431929e+00 -5.91777146e-01 -1.05972916e-01 1.01316392e+00 -8.04553449e-01 -3.04290414e-01 -9.82043266e-01 -5.82118392e-01 8.24224770e-01 9.88695204e-01 5.52694380e-01 -1.23896682e+00 -6.53471947e-01 4.85433966e-01 -4.05751377e-01 -1.08586025e+00 -6.24292552e-01 3.46830070e-01 -1.16755593e+00 -5.44329464e-01 -6.53019965e-01 -5.98423958e-01 2.82022178e-01 1.17073186e-01 1.42002285e+00 2.65977889e-01 2.81366646e-01 2.36304551e-01 -4.45453018e-01 -3.91220778e-01 -2.29157493e-01 7.71033704e-01 -4.56380427e-01 -2.49335125e-01 -1.69287756e-01 -6.81285083e-01 -3.89217228e-01 1.58887550e-01 -9.47794616e-01 2.81282365e-01 7.04558313e-01 8.85922730e-01 1.82944819e-01 -5.00128508e-01 1.03389740e+00 -8.63130748e-01 9.76987064e-01 -6.08241670e-02 -1.29256234e-01 3.76537114e-01 -3.69761318e-01 3.62604052e-01 7.33548522e-01 -2.99470484e-01 -9.39958811e-01 -6.99829042e-01 -5.44604480e-01 -3.82776737e-01 3.61279517e-01 7.76648164e-01 -1.04496457e-01 2.35224530e-01 4.88377750e-01 2.12428242e-01 -8.37619230e-02 -8.38380635e-01 5.44809699e-01 4.32160199e-01 5.19864202e-01 -5.63620508e-01 3.01435173e-01 -7.26579204e-02 -5.56893945e-01 -6.47622585e-01 -1.02065969e+00 -2.04647079e-01 -4.24085796e-01 3.36539388e-01 7.48843193e-01 -4.92370188e-01 -3.92093629e-01 6.58560514e-01 -1.38356411e+00 -7.19000876e-01 -4.66416359e-01 -1.00898623e-01 -4.09466058e-01 4.08570379e-01 -1.11298609e+00 -4.64589536e-01 -7.94802547e-01 -9.89587486e-01 8.81771863e-01 4.40897755e-02 -3.60813349e-01 -1.06415534e+00 -1.79319859e-01 5.58812380e-01 1.02669013e+00 -3.63097161e-01 1.38757634e+00 -1.05495477e+00 -4.70205754e-01 1.56927913e-01 -3.17745030e-01 5.71802318e-01 -2.37938136e-01 -2.97889560e-01 -1.06298470e+00 -1.39336586e-01 1.19741447e-01 -2.89348960e-01 1.21877539e+00 4.74256486e-01 1.51156044e+00 -6.13234699e-01 -6.75640628e-02 4.57166076e-01 9.79445994e-01 -2.00508848e-01 9.19986188e-01 3.72494727e-01 6.60432875e-01 5.17183900e-01 5.18463925e-02 4.76553850e-02 6.51506305e-01 7.21559584e-01 3.85375977e-01 -1.43007413e-02 -4.17769104e-01 -4.58679885e-01 3.85602832e-01 1.23018610e+00 -1.28540453e-02 -5.68447232e-01 -6.89062238e-01 8.23943734e-01 -1.90669966e+00 -8.68486583e-01 4.01261821e-02 1.99741340e+00 9.31410909e-01 2.44988874e-01 4.86362800e-02 7.02132136e-02 3.67016554e-01 2.49687016e-01 -5.49988389e-01 -9.04100418e-01 -8.84220675e-02 6.39890194e-01 3.01186204e-01 6.03322923e-01 -7.58507192e-01 1.22608387e+00 7.10812855e+00 7.92411268e-01 -9.14250970e-01 2.24976152e-01 5.35983205e-01 -1.64481401e-01 -4.62604254e-01 -8.05829689e-02 -6.52963698e-01 1.90749317e-01 1.29339838e+00 -2.33220562e-01 2.87817538e-01 4.46601689e-01 -5.16959466e-02 9.83804613e-02 -1.20125210e+00 3.12988669e-01 2.75274634e-01 -1.32202852e+00 5.06307244e-01 -8.71267468e-02 5.11782885e-01 3.72281879e-01 -6.23909011e-03 7.68423796e-01 3.74552071e-01 -9.88987803e-01 5.94364047e-01 2.89795220e-01 4.71822441e-01 -4.87794429e-01 8.81393671e-01 4.30201501e-01 -1.04959464e+00 -3.36885154e-01 -3.41350317e-01 -1.24877691e-01 2.56404370e-01 9.19834152e-02 -7.54817307e-01 7.19664156e-01 6.23000503e-01 5.88331521e-01 -9.83305275e-01 1.02542257e+00 -3.10984462e-01 9.08194840e-01 -3.72369707e-01 5.59805110e-02 4.85987604e-01 1.11448452e-01 5.05825222e-01 1.23604643e+00 2.28671938e-01 -1.07587025e-01 -2.08940536e-01 8.03072453e-01 -3.71042728e-01 -7.31711015e-02 -2.05071419e-01 5.27234152e-02 3.70187968e-01 9.26371098e-01 -2.97457516e-01 -5.57856441e-01 -3.07741553e-01 8.65536630e-01 6.40478432e-01 4.45013970e-01 -8.73616993e-01 -5.14122725e-01 2.19090819e-01 1.59741476e-01 7.13619530e-01 1.06017906e-02 -3.82592767e-01 -1.18210590e+00 -3.14164604e-03 -9.65694904e-01 5.06755590e-01 -1.08196282e+00 -9.72888649e-01 7.49459147e-01 6.91876411e-02 -3.30042213e-01 -1.88672721e-01 -4.26045746e-01 -9.62303400e-01 8.43214333e-01 -1.43431616e+00 -1.28869593e+00 -2.25433931e-02 4.30621445e-01 7.47856259e-01 1.72979519e-01 7.99054503e-01 1.73515901e-01 -7.51406074e-01 9.51440513e-01 -1.95779484e-02 -1.44766197e-01 5.28115571e-01 -1.40855312e+00 8.39350998e-01 7.28938043e-01 8.20371956e-02 8.10426354e-01 7.01302409e-01 -2.26773843e-01 -1.18900740e+00 -1.10050511e+00 1.36959136e+00 -7.14537740e-01 8.16480219e-01 -3.79811436e-01 -1.20935631e+00 1.20710313e+00 8.42746913e-01 -5.80734968e-01 4.00404036e-01 4.36941445e-01 -3.80959064e-01 4.78190742e-02 -7.95946538e-01 5.60456634e-01 1.01049411e+00 -5.69154322e-01 -1.18874359e+00 1.31662428e-01 1.30650008e+00 -2.74607390e-01 -8.45074058e-01 4.75219548e-01 2.25980729e-01 -8.29629302e-01 9.63946164e-01 -7.95547903e-01 5.20868659e-01 2.06465095e-01 -8.85606781e-02 -1.56933856e+00 -5.42133570e-01 -6.03676260e-01 -4.57584977e-01 1.21830714e+00 6.00802541e-01 -8.11977088e-01 5.70633650e-01 4.56151575e-01 -7.68907189e-01 -1.12231290e+00 -9.51885700e-01 -6.91777766e-01 6.34759009e-01 -2.03284815e-01 6.36570573e-01 4.51231658e-01 9.27243009e-02 9.15118694e-01 -2.38731086e-01 -2.83693850e-01 6.78626448e-02 8.92844647e-02 6.03292108e-01 -9.57459271e-01 -4.88061756e-01 -7.53565669e-01 2.17492700e-01 -1.63929737e+00 1.85620207e-02 -9.25054789e-01 -2.29836389e-01 -1.98999333e+00 2.26163805e-01 5.97267784e-03 -4.30513710e-01 6.79373085e-01 -4.77614939e-01 -2.90346202e-02 4.28821921e-01 1.29496800e-02 -7.73572922e-01 1.01005518e+00 1.29475272e+00 -3.20756920e-02 -4.22884412e-02 -1.38897464e-01 -1.15074492e+00 5.00807881e-01 8.78399372e-01 -2.61893392e-01 -4.71988618e-01 -1.12891197e+00 1.55639917e-01 -8.38177130e-02 2.58287132e-01 -8.20075095e-01 2.28332236e-01 4.32323247e-01 1.95266027e-02 -7.38533139e-01 4.61110145e-01 -2.59911001e-01 -4.45830405e-01 4.27695721e-01 -7.31594265e-01 4.33764011e-01 5.35897255e-01 1.89348280e-01 -1.92707434e-01 -3.08890045e-01 5.75855851e-01 -1.89441919e-01 -2.42958322e-01 -7.09551480e-03 -4.26459372e-01 2.57264763e-01 2.97769666e-01 3.98623832e-02 -7.66123891e-01 -5.00083506e-01 -4.57309008e-01 4.93954360e-01 2.54479259e-01 4.15311992e-01 3.22854191e-01 -9.43979979e-01 -8.29030693e-01 -4.19769734e-02 -3.87385674e-02 -2.48756148e-02 1.72812372e-01 8.36423516e-01 -3.01658332e-01 9.29970443e-01 7.68127516e-02 -4.13229376e-01 -1.10194480e+00 4.88213927e-01 5.03602386e-01 -8.77657950e-01 -5.59186697e-01 1.04899120e+00 1.16777085e-01 -7.79970944e-01 2.31742099e-01 -7.11801469e-01 -2.51597226e-01 5.98601326e-02 3.31483603e-01 2.02991858e-01 3.39689791e-01 -1.31931528e-01 -5.66863567e-02 4.26187932e-01 -5.39186835e-01 -5.98059818e-02 1.40494907e+00 -4.28404361e-02 -1.44652739e-01 1.45750880e-01 1.02942908e+00 -2.66528189e-01 -8.38578403e-01 -4.49615866e-01 3.05278860e-02 -2.10105516e-02 3.51612382e-02 -9.73313153e-01 -1.00288105e+00 1.08144128e+00 -3.35549191e-02 4.05073613e-01 1.13758123e+00 1.44512102e-01 1.21091044e+00 5.76438069e-01 -1.29184062e-02 -8.65848422e-01 3.16410869e-01 9.17776942e-01 1.16070199e+00 -6.23112977e-01 -1.87583297e-01 1.27695836e-02 -6.63434029e-01 6.51340723e-01 7.80814648e-01 -1.01213381e-01 1.18759349e-01 7.87992217e-03 -2.41577968e-01 -1.18596479e-01 -1.51166320e+00 -1.66461885e-01 2.93170303e-01 2.75048912e-01 5.65631509e-01 -2.46898755e-01 -2.55183250e-01 8.01641107e-01 -4.62130219e-01 -1.17562257e-01 4.38577563e-01 8.19698691e-01 -5.88577867e-01 -1.25065351e+00 -1.00596264e-01 7.46057093e-01 -5.71493447e-01 -5.06721973e-01 -6.25281274e-01 6.91793144e-01 -1.39337391e-01 1.00369990e+00 1.01254612e-01 -4.33484524e-01 5.71725726e-01 3.57498050e-01 6.47734404e-01 -6.55099750e-01 -1.20723534e+00 -6.53658807e-02 4.48696256e-01 -4.41650391e-01 -6.80214763e-02 -4.19692904e-01 -8.48957300e-01 -3.58851403e-01 -5.55111825e-01 2.81965613e-01 2.00190619e-01 9.86407638e-01 7.46869683e-01 1.00524330e+00 5.46578765e-02 -8.52685452e-01 -7.21224368e-01 -1.56594837e+00 -9.91540030e-03 1.94554299e-01 2.29222223e-01 -2.11014166e-01 -2.37732902e-01 -9.98820066e-02]
[11.254128456115723, 8.428555488586426]
e70255c2-39a5-4a18-ac96-1c52e7a2eca6
scalable-matching-and-clustering-of-entities
null
null
https://csimq-journals.rtu.lv/article/view/csimq.2018-16.04/0
https://csimq-journals.rtu.lv/article/view/csimq.2018-16.04/1276
Scalable Matching and Clustering of Entities with FAMER
Entity resolution identifies semantically equivalent entities, e.g. describing the same product or customer. It is especially challenging for Big Data applications where large volumes of data from many sources have to be matched and integrated. We therefore introduce a scalable entity resolution framework called FAMER (FAst Multi-source Entity Resolution system) that is based on Apache Flink for distributed execution and that can holistically match entities from multiple sources. For the latter purpose, FAMER includes multiple clustering schemes that group matching entities from different sources within clusters. In addition to previously known clustering schemes FAMER includes new approaches tailored to multi-source entity resolution. We perform a detailed comparative evaluation of eight clustering schemes for different real-life and synthetically generated datasets. The evaluation considers both the match quality as well as the scalability for different numbers of machines and data sizes.
['Erhard Rahm', 'Eric Peukert', 'Markus Nentwig', 'Alieh Saeedi']
2018-10-01
null
null
null
complex-systems-informatics-and-modeling
['clustering-algorithms-evaluation', 'entity-resolution']
['methodology', 'natural-language-processing']
[-6.08926296e-01 -9.59115401e-02 9.80784837e-03 -1.41030565e-01 -9.72328961e-01 -8.20550084e-01 4.68564779e-01 1.14858735e+00 -3.45293641e-01 7.43032694e-01 1.00306801e-01 4.38957840e-01 -5.08289516e-01 -1.34471667e+00 -3.41597080e-01 -4.09452617e-02 -2.22430587e-01 1.13515282e+00 7.21120119e-01 -1.37249827e-01 -1.32308174e-02 6.07494593e-01 -1.69607270e+00 5.15847683e-01 9.05248940e-01 6.54798567e-01 2.09279791e-01 4.52808052e-01 -5.31192601e-01 4.96251673e-01 -7.56629646e-01 -6.35111988e-01 3.29866111e-01 1.71592250e-01 -8.56962860e-01 -3.43550503e-01 9.57889929e-02 3.75534266e-01 3.95855665e-01 9.39842522e-01 5.12261212e-01 3.55338752e-02 1.61766276e-01 -1.67792261e+00 -4.48227286e-01 8.11678886e-01 -4.41599399e-01 -1.29596200e-02 6.16940975e-01 -4.52458739e-01 8.22242677e-01 -9.81046915e-01 1.35873008e+00 9.68324840e-01 9.03384566e-01 2.69325674e-02 -1.10145152e+00 -5.60172558e-01 -2.85578161e-01 3.73725057e-01 -1.90051270e+00 -3.32341850e-01 3.30546722e-02 -4.25811350e-01 9.43651378e-01 4.23254788e-01 5.92903327e-03 2.15283051e-01 -1.94895625e-01 2.41726920e-01 6.25271022e-01 -1.21532008e-01 5.30725002e-01 5.30082941e-01 2.97820836e-01 4.74329479e-03 9.98235643e-01 -5.46167791e-01 -4.25416142e-01 -5.02557516e-01 2.04953998e-01 8.06749538e-02 2.08057582e-01 -4.72190350e-01 -1.43509209e+00 4.95704055e-01 4.23220009e-01 6.54440582e-01 -8.13971996e-01 -2.73124963e-01 3.72814924e-01 1.29033044e-01 -3.84004861e-02 6.41581893e-01 -6.07809722e-01 3.04936349e-01 -1.00715864e+00 4.80959326e-01 1.18664730e+00 1.49155998e+00 9.55022395e-01 -4.80193645e-01 1.51656820e-02 5.30687869e-01 1.30383790e-01 1.58163756e-01 3.07771832e-01 -8.64997745e-01 5.14670193e-01 1.18359959e+00 5.58567226e-01 -1.34624934e+00 -6.65135801e-01 -2.04096034e-01 -9.55760479e-01 -1.04719035e-01 1.60922959e-01 -5.39040864e-02 -3.48684222e-01 1.47782302e+00 8.40347648e-01 9.64241698e-02 4.53913629e-01 6.45783961e-01 1.28410912e+00 3.81647080e-01 4.67204899e-01 -3.43358010e-01 1.61658204e+00 -6.36799991e-01 -8.18955541e-01 3.31948161e-01 5.31690896e-01 -9.34069812e-01 6.43747225e-02 8.60264078e-02 -1.08020711e+00 -4.68069851e-01 -7.58033454e-01 5.13603762e-02 -1.26890576e+00 -1.64953068e-01 4.49041158e-01 2.84116060e-01 -8.87111902e-01 5.58314502e-01 -5.88818491e-01 -6.86925769e-01 -5.72237596e-02 4.40913528e-01 -7.11646676e-01 5.78875169e-02 -1.33751094e+00 8.15026641e-01 1.00940764e+00 -4.32608604e-01 1.17736109e-01 -9.70250845e-01 -6.45819366e-01 3.96409780e-01 4.67337430e-01 -8.05158615e-01 7.88901210e-01 -5.70486844e-01 -4.48557466e-01 6.70609772e-01 9.77203622e-03 -4.60911870e-01 1.60481945e-01 9.97539312e-02 -1.18685400e+00 -1.19139142e-01 4.72794026e-01 3.92952740e-01 -1.09102480e-01 -1.31769419e+00 -1.14682138e+00 -3.93936068e-01 -1.55648082e-01 -7.02000409e-02 -1.73450768e-01 5.40578902e-01 -5.22944212e-01 -5.26823223e-01 -1.51416942e-01 -6.12844706e-01 -3.61003220e-01 -7.00009286e-01 -3.02112192e-01 -3.25152546e-01 4.08602238e-01 -5.07770956e-01 1.65697360e+00 -1.64295673e+00 1.54332211e-02 4.77960885e-01 5.86997271e-01 2.74090648e-01 8.37645307e-02 1.02920318e+00 -6.81223795e-02 3.42120439e-01 1.85449002e-03 -6.96786046e-02 1.91797405e-01 -1.62138492e-01 1.53868437e-01 -2.39349566e-02 1.30404823e-03 8.03166747e-01 -7.36654878e-01 -7.91098237e-01 8.67119059e-03 4.54370588e-01 -2.31049314e-01 -3.24090272e-02 -5.79875335e-02 -5.53358719e-02 -3.18509579e-01 7.68681169e-01 9.64268565e-01 -5.46168506e-01 4.09784198e-01 -2.39960238e-01 -2.45318711e-01 -2.26153031e-01 -2.06017923e+00 1.36697853e+00 -9.49715748e-02 2.18105949e-02 3.63821656e-01 -3.56736422e-01 8.54970574e-01 5.42178869e-01 8.20347786e-01 -5.27785242e-01 -2.86965519e-01 5.29422700e-01 -3.80352259e-01 -1.83744773e-01 1.08671701e+00 2.49675959e-01 -4.56092238e-01 2.71354049e-01 5.31850616e-03 6.61620915e-01 8.45118105e-01 6.55572295e-01 1.33910048e+00 -3.77338290e-01 8.57458889e-01 -2.23576114e-01 5.40934324e-01 5.67486644e-01 8.45676184e-01 3.86155188e-01 1.71790943e-01 3.04538220e-01 9.98650044e-02 -4.27352816e-01 -1.11986876e+00 -9.72849131e-01 -6.55714795e-02 8.15679669e-01 5.54860115e-01 -1.07602644e+00 -5.79848170e-01 -3.30866426e-01 3.89453143e-01 4.43713248e-01 -3.08957994e-01 4.16618407e-01 -3.52291554e-01 -6.64800227e-01 4.11737233e-01 3.18498611e-01 2.00771585e-01 -1.13994300e+00 -4.67649162e-01 5.78080058e-01 -2.52803355e-01 -1.35646045e+00 1.01614445e-01 -3.92636091e-01 -2.05351457e-01 -1.44445765e+00 -2.29838163e-01 -6.75076962e-01 2.62403846e-01 1.21206239e-01 1.63866341e+00 1.24482810e-01 -3.34438801e-01 2.48847783e-01 -4.52804208e-01 -3.87487262e-01 -5.40600300e-01 5.16179919e-01 2.28377879e-01 -9.28165615e-02 8.01137865e-01 -4.22002554e-01 -3.81453604e-01 7.11387575e-01 -1.00374174e+00 -5.42287827e-02 3.90158504e-01 6.20221794e-02 8.37011278e-01 4.93252695e-01 1.10801339e+00 -1.24159038e+00 5.77623308e-01 -1.25087881e+00 -7.93154120e-01 8.32256377e-01 -9.50100362e-01 -2.23049968e-01 8.37742150e-01 -1.20937347e-01 -8.48381341e-01 1.77047208e-01 2.69972235e-02 -1.75280809e-01 -4.01472807e-01 5.53815722e-01 -2.52267569e-01 3.68694216e-01 5.92257082e-01 -3.78754586e-01 -5.37117958e-01 -8.13098311e-01 6.01390421e-01 8.18248272e-01 7.45212734e-01 -4.17975694e-01 9.19303238e-01 3.62608105e-01 -6.91299662e-02 -2.97492325e-01 -4.53839675e-02 -9.80057955e-01 -8.68389547e-01 1.02337420e-01 6.74958944e-01 -1.04267597e+00 -9.69713569e-01 1.46925405e-01 -1.07926917e+00 3.27371746e-01 -3.92041773e-01 2.09359407e-01 -1.66786611e-01 6.81629032e-02 -4.77064043e-01 -2.23476112e-01 -5.93121052e-01 -5.92604578e-01 8.71456563e-01 4.64332849e-01 -4.28588927e-01 -7.42318869e-01 5.08482516e-01 6.94007352e-02 5.96873462e-01 7.37958431e-01 5.85486948e-01 -1.26872313e+00 -7.30013430e-01 -3.67047161e-01 -3.11651379e-01 -5.86284637e-01 1.94081649e-01 1.43216863e-01 -3.60164732e-01 -3.01261991e-01 -6.85124159e-01 3.69334608e-01 2.74988592e-01 -3.58124673e-01 2.31045380e-01 -4.28237319e-01 -8.15767407e-01 2.53779352e-01 1.90240550e+00 1.91607140e-02 7.03182280e-01 8.38759303e-01 7.40058780e-01 7.76094615e-01 9.55329955e-01 5.72406411e-01 6.70853913e-01 8.59880805e-01 2.90264059e-02 -2.21015900e-01 1.23698786e-01 5.85116260e-02 -4.01454538e-01 7.75314629e-01 5.55984713e-02 -3.68218571e-01 -1.04081571e+00 8.31474960e-01 -2.16308951e+00 -1.24267614e+00 -5.75131416e-01 2.22798467e+00 6.45555496e-01 -3.70274723e-01 4.45314556e-01 6.26777438e-03 1.29134154e+00 -4.73038673e-01 -4.09171999e-01 -5.49590625e-02 -3.17765951e-01 5.50658070e-02 5.20020783e-01 1.61180526e-01 -9.43193555e-01 8.23204339e-01 5.98942423e+00 7.13662148e-01 -5.35890877e-01 3.30799252e-01 -3.02636577e-03 -6.05863929e-02 -1.44523218e-01 1.11435302e-01 -1.18113017e+00 5.97362578e-01 1.27449059e+00 -8.59594107e-01 1.70282364e-01 9.03690279e-01 -8.48084316e-02 -3.05926781e-02 -8.09793830e-01 7.23418117e-01 -4.71088886e-01 -1.63311088e+00 -2.29353458e-01 1.42423555e-01 7.96762824e-01 1.96026921e-01 -6.15491152e-01 1.95341468e-01 8.24013412e-01 -7.14002132e-01 5.35310090e-01 5.63875735e-01 6.18049622e-01 -8.56270969e-01 8.04578662e-01 2.98276424e-01 -1.83057559e+00 1.12008400e-01 -3.76456708e-01 4.85611051e-01 2.70651847e-01 7.00731397e-01 -7.48777032e-01 1.17527092e+00 9.26667929e-01 3.65524560e-01 -8.11497509e-01 1.10372794e+00 4.19087946e-01 -9.60559249e-02 -6.64198279e-01 4.70150709e-01 -3.46759051e-01 5.11817709e-02 3.58212620e-01 1.46619570e+00 5.21994650e-01 1.48098454e-01 1.24233112e-01 6.30491495e-01 -3.99998724e-01 4.84483272e-01 -2.89519519e-01 2.42919654e-01 1.28739166e+00 1.92740035e+00 -8.48998010e-01 -5.54422915e-01 -4.90297019e-01 4.18887168e-01 4.23184097e-01 5.03987633e-02 -7.02883959e-01 -6.36083305e-01 8.51530612e-01 4.32685405e-01 4.82931167e-01 -9.14251134e-02 2.89602447e-02 -1.07461762e+00 -2.80789565e-02 -6.31989896e-01 9.28802133e-01 -6.28437042e-01 -1.29882705e+00 7.33697832e-01 3.58406380e-02 -1.14419186e+00 -3.57725143e-01 7.55977258e-02 -4.09484327e-01 6.83321595e-01 -1.36963499e+00 -1.04870117e+00 -6.18618906e-01 7.93976486e-01 -3.79007459e-02 -2.40805596e-02 1.02773905e+00 8.09281945e-01 -6.36435866e-01 4.51275885e-01 4.74558204e-01 7.35622123e-02 8.98469210e-01 -1.41704869e+00 6.54044628e-01 7.58095145e-01 3.20232362e-02 9.82739091e-01 5.30538082e-01 -9.20343995e-01 -1.16159976e+00 -1.48876441e+00 1.23467290e+00 -6.00982368e-01 6.45134866e-01 -3.50201964e-01 -1.22142005e+00 6.04985893e-01 2.61771023e-01 1.46476716e-01 8.26777756e-01 1.06824592e-01 -4.38326180e-01 -2.26035252e-01 -1.60892820e+00 1.26337543e-01 7.94324696e-01 -5.42106964e-02 -3.96951079e-01 2.63911277e-01 5.56949914e-01 -1.67698711e-01 -2.03761744e+00 4.15844351e-01 2.83360749e-01 -8.68037343e-01 1.11539924e+00 -4.78750825e-01 -4.76473980e-02 -7.69376099e-01 -2.47314319e-01 -9.79639649e-01 -5.35508990e-01 -6.22367322e-01 -2.83921450e-01 2.06837702e+00 4.01120842e-01 -5.63553512e-01 3.97257626e-01 9.48573649e-01 4.96234745e-01 -1.94772705e-01 -7.58311927e-01 -9.41227376e-01 -3.32255661e-01 1.60251930e-01 1.39867938e+00 1.41536927e+00 1.46557972e-01 4.24911499e-01 -1.80578381e-02 5.79323053e-01 6.12177730e-01 5.64969480e-01 9.93548751e-01 -1.75424528e+00 7.44195953e-02 -2.08173051e-01 -5.97280502e-01 2.58307785e-01 -3.31947386e-01 -9.01666045e-01 -3.76689196e-01 -1.89278138e+00 2.00263798e-01 -8.90738308e-01 -3.30732942e-01 5.86040676e-01 -9.61843133e-02 1.21838436e-01 2.77803928e-01 7.92041361e-01 -1.21617711e+00 -1.74474552e-01 1.29867986e-01 3.57476324e-01 -3.38274986e-01 -1.49516448e-01 -7.02257693e-01 4.25993592e-01 6.74902797e-01 -7.13110507e-01 1.47282645e-01 -8.23962539e-02 7.32455373e-01 -3.50680901e-04 -8.32249224e-02 -1.23825097e+00 8.36956620e-01 -2.59658247e-01 2.22423390e-01 -8.28620791e-01 -2.84030169e-01 -1.07875967e+00 1.36080849e+00 2.66127046e-02 -3.15712728e-02 4.29154903e-01 2.17213929e-01 4.83848035e-01 -4.66355681e-01 1.03645436e-01 6.72585785e-01 -2.00558513e-01 -1.11839116e+00 2.25461900e-01 1.23337612e-01 1.15595058e-01 1.49736619e+00 1.16772363e-02 -7.24145710e-01 1.95394292e-01 -8.85786951e-01 6.79138005e-01 9.68483388e-01 6.04673386e-01 2.76105609e-02 -1.55826700e+00 -9.99041915e-01 -3.85102957e-01 5.25201917e-01 6.39891177e-02 1.04236387e-01 5.13718426e-01 -5.07011056e-01 3.83156240e-01 -2.44644582e-01 -1.82309389e-01 -1.55380428e+00 1.20382619e+00 4.73586880e-02 -6.21628284e-01 -3.70518029e-01 -3.53848301e-02 -2.83679068e-01 -5.73912323e-01 -1.97644442e-01 2.98950940e-01 -4.51505452e-01 4.47795719e-01 7.51454115e-01 8.59656036e-01 5.24774671e-01 -1.00727665e+00 -7.66184688e-01 3.89805406e-01 1.51022524e-02 3.52508813e-01 1.40319574e+00 -3.79804194e-01 -4.04321283e-01 1.72174424e-01 8.12453926e-01 4.34109867e-01 -3.60320270e-01 -4.00309771e-01 6.78825259e-01 -2.30448872e-01 -5.65311849e-01 -8.69527578e-01 -1.02922153e+00 -9.05378908e-02 4.15974796e-01 7.31995523e-01 1.13385880e+00 2.02162027e-01 7.29133308e-01 2.01120049e-01 7.82977164e-01 -1.12669039e+00 -7.72923112e-01 3.31275701e-03 4.73164260e-01 -1.05078161e+00 2.06417337e-01 -6.98879600e-01 -5.06923795e-01 7.53475726e-01 4.72751170e-01 1.37130301e-02 5.49299538e-01 4.93366122e-01 -1.37109950e-01 -4.49113220e-01 -8.23140204e-01 -5.06527781e-01 9.99620780e-02 4.68505859e-01 1.53794780e-01 2.39240140e-01 -5.66806018e-01 1.08114290e+00 -2.26015095e-02 1.84598804e-01 4.47987258e-01 9.03596997e-01 -4.53538358e-01 -1.43105197e+00 -5.83543181e-01 2.83741832e-01 -6.30867660e-01 5.84232360e-02 -6.06588662e-01 8.21548939e-01 4.19175923e-01 9.89889920e-01 2.44341001e-01 -2.93516994e-01 6.23237908e-01 -1.20795652e-01 -2.16681793e-01 -6.01771891e-01 -1.21384943e+00 -2.16290370e-01 1.45634592e-01 -6.62716985e-01 -5.41415095e-01 -6.37186766e-01 -1.57078028e+00 -7.46604800e-01 -3.64519775e-01 4.90500182e-01 8.91350448e-01 3.82704616e-01 1.21516967e+00 2.67798930e-01 4.65984046e-01 -3.23709190e-01 2.71710247e-01 -7.28996158e-01 -7.02403009e-01 8.16972196e-01 -2.24261582e-01 -4.69791263e-01 6.62172884e-02 7.00524170e-03]
[9.206306457519531, 7.99027156829834]
a02d0ef3-b7b6-454b-bf52-579400808e4c
few-shot-adversarial-domain-adaptation
1711.02536
null
http://arxiv.org/abs/1711.02536v1
http://arxiv.org/pdf/1711.02536v1.pdf
Few-Shot Adversarial Domain Adaptation
This work provides a framework for addressing the problem of supervised domain adaptation with deep models. The main idea is to exploit adversarial learning to learn an embedded subspace that simultaneously maximizes the confusion between two domains while semantically aligning their embedding. The supervised setting becomes attractive especially when there are only a few target data samples that need to be labeled. In this few-shot learning scenario, alignment and separation of semantic probability distributions is difficult because of the lack of data. We found that by carefully designing a training scheme whereby the typical binary adversarial discriminator is augmented to distinguish between four different classes, it is possible to effectively address the supervised adaptation problem. In addition, the approach has a high speed of adaptation, i.e. it requires an extremely low number of labeled target training samples, even one per category can be effective. We then extensively compare this approach to the state of the art in domain adaptation in two experiments: one using datasets for handwritten digit recognition, and one using datasets for visual object recognition.
['Seyed Mehdi Iranmanesh', 'Saeid Motiian', 'Gianfranco Doretto', 'Quinn Jones']
2017-11-05
few-shot-adversarial-domain-adaptation-1
http://papers.nips.cc/paper/7244-few-shot-adversarial-domain-adaptation
http://papers.nips.cc/paper/7244-few-shot-adversarial-domain-adaptation.pdf
neurips-2017-12
['handwritten-digit-recognition']
['computer-vision']
[ 4.28193003e-01 -2.37977393e-02 -2.31919065e-01 -3.71603668e-01 -8.07384253e-01 -7.67158628e-01 7.79178858e-01 4.44300286e-03 -6.35791957e-01 7.97539473e-01 -1.74686298e-01 -1.12116067e-02 -5.58809098e-03 -6.55205667e-01 -6.27116323e-01 -8.72975767e-01 2.86966056e-01 7.89647341e-01 2.79749781e-01 -7.90917054e-02 1.29345834e-01 6.61883891e-01 -1.41950691e+00 -1.43215863e-03 7.22050846e-01 1.01268101e+00 2.31566891e-01 6.35940611e-01 -2.80101240e-01 5.41534245e-01 -6.92196548e-01 -2.54550606e-01 5.28769672e-01 -5.73086023e-01 -7.94370472e-01 3.84659976e-01 6.44080639e-01 -3.35751355e-01 -2.53034174e-01 1.20282924e+00 5.36690831e-01 3.92013937e-01 1.16906035e+00 -1.37333095e+00 -6.37067080e-01 1.14392132e-01 -3.75019729e-01 1.43197089e-01 -1.18703038e-01 5.82120493e-02 6.61839664e-01 -7.63021469e-01 6.60881519e-01 9.66149032e-01 4.67820555e-01 9.13436592e-01 -1.60882211e+00 -4.65716124e-01 -4.82638516e-02 3.35959226e-01 -1.09728181e+00 -4.53960627e-01 9.42102194e-01 -5.84665656e-01 7.41252422e-01 -1.33669987e-01 2.88939297e-01 1.38955677e+00 -1.52784541e-01 5.68639636e-01 1.04838467e+00 -7.31292248e-01 7.86628842e-01 5.17277539e-01 8.77210125e-02 1.75991282e-01 8.72395486e-02 2.42536530e-01 -2.16641545e-01 -1.90708071e-01 7.13387787e-01 -9.96134244e-03 6.03560694e-02 -1.16240287e+00 -1.01206648e+00 1.02216268e+00 3.77482593e-01 3.82655442e-01 -2.30667621e-01 -2.55220860e-01 5.25479019e-01 5.41961014e-01 3.30087215e-01 6.68924391e-01 -3.23839694e-01 7.33048320e-02 -8.65151525e-01 1.43617868e-01 9.53158677e-01 8.29675674e-01 7.61001647e-01 1.17777094e-01 3.87536697e-02 1.15645432e+00 -1.53513148e-03 4.82651889e-01 6.27658248e-01 -1.00216424e+00 3.40398550e-01 2.17992306e-01 2.23643303e-01 -5.56193292e-01 1.18396498e-01 -1.40900880e-01 -8.56131732e-01 8.20931554e-01 8.93619716e-01 -1.72128260e-01 -1.21280527e+00 1.75113165e+00 1.53136656e-01 1.41956070e-02 4.50410694e-01 7.73416817e-01 2.10065529e-01 4.88923103e-01 2.50449836e-01 2.20293887e-02 1.15685093e+00 -8.24785948e-01 -5.08131266e-01 -7.33512819e-01 2.54839987e-01 -7.63299823e-01 1.10033107e+00 2.60661751e-01 -7.37588167e-01 -5.76006353e-01 -1.32837296e+00 2.24521652e-01 -6.71194613e-01 -4.67548054e-03 4.21878725e-01 6.10102892e-01 -5.82842827e-01 6.79060102e-01 -7.47152567e-01 -7.07915962e-01 6.53627276e-01 4.20325905e-01 -4.92745221e-01 -2.46694744e-01 -1.05449069e+00 1.11139905e+00 7.02080965e-01 -6.24501467e-01 -1.05464613e+00 -4.33384478e-01 -8.98114383e-01 9.46878791e-02 1.45499751e-01 -3.76013517e-01 1.26871657e+00 -1.59500182e+00 -1.57409966e+00 1.04489362e+00 9.87269580e-02 -4.32548493e-01 5.54182887e-01 2.52079149e-03 -2.79113680e-01 2.38588274e-01 2.54132040e-02 6.22109532e-01 1.19951606e+00 -1.15689933e+00 -4.08439666e-01 -5.04024923e-01 -7.02047944e-02 1.96949989e-01 -7.86645889e-01 -4.91414033e-02 -4.70958613e-02 -8.10970306e-01 -1.42047316e-01 -8.93949926e-01 -2.51330107e-01 2.25470066e-01 1.48259550e-02 8.35904572e-03 9.53835309e-01 -5.27299106e-01 5.06372333e-01 -2.37397861e+00 3.56964350e-01 1.61286026e-01 -1.10288084e-01 5.09667814e-01 -2.98156261e-01 2.37248272e-01 -2.85823882e-01 -1.69180572e-01 -6.16286337e-01 -2.33997419e-01 -6.76131248e-03 4.39858168e-01 -4.98837918e-01 4.58559632e-01 4.68400657e-01 6.26823366e-01 -9.21194732e-01 -3.37594509e-01 3.13706517e-01 2.22680375e-01 -3.25626910e-01 4.96859998e-01 -1.63863137e-01 4.65653747e-01 -2.18513831e-01 4.61841375e-01 6.99250102e-01 -1.47881791e-01 2.25457698e-01 1.87699661e-01 3.93172383e-01 -9.49784815e-02 -1.43347466e+00 1.63328373e+00 -5.25479317e-01 8.04461777e-01 -1.11189276e-01 -1.50378931e+00 1.07934868e+00 2.41580158e-01 3.48694324e-01 -3.63800168e-01 -1.73111744e-02 2.74401516e-01 -9.67053920e-02 -1.15123779e-01 6.96421508e-03 -5.70948720e-01 -1.37126967e-01 4.12726402e-01 6.62540734e-01 -2.60420442e-01 -7.91354105e-03 -1.49027720e-01 1.05310738e+00 1.13905318e-01 6.12031758e-01 -1.83784828e-01 4.30094600e-01 2.63598710e-01 4.84320998e-01 6.80374980e-01 -3.94308567e-01 6.80159688e-01 4.23703581e-01 -3.36672127e-01 -1.44764841e+00 -1.39605081e+00 -1.60108790e-01 1.00151861e+00 5.95940761e-02 3.87846708e-01 -7.91141331e-01 -8.76452208e-01 9.33498796e-03 7.79966354e-01 -7.56713271e-01 -4.01679158e-01 -4.28711891e-01 -4.82675344e-01 3.72482985e-01 7.69820869e-01 4.30597305e-01 -1.03730631e+00 -5.45173824e-01 1.31056368e-01 1.26967177e-01 -1.06832421e+00 -2.33295307e-01 6.29683197e-01 -1.01413774e+00 -9.83884454e-01 -1.08154416e+00 -1.04559839e+00 8.15916777e-01 1.21839587e-02 1.04878592e+00 -6.65741265e-01 -2.55705357e-01 4.39084738e-01 -2.81591445e-01 -4.97911960e-01 -5.75770438e-01 -1.69807505e-02 1.38438717e-01 2.06372123e-02 7.90500641e-01 -6.64232314e-01 -2.42314383e-01 3.60850602e-01 -9.60829735e-01 -4.78497207e-01 5.30604422e-01 1.35114372e+00 4.15917635e-01 -1.68376923e-01 6.98914051e-01 -7.05338001e-01 3.10852528e-01 -4.50706869e-01 -5.94408572e-01 1.80887640e-01 -3.26923162e-01 1.76280841e-01 1.10119987e+00 -8.78896475e-01 -1.10909033e+00 3.87203604e-01 1.28148004e-01 -6.04031742e-01 -6.20803595e-01 -6.90943301e-02 -2.80112416e-01 -2.14991286e-01 1.03864515e+00 1.96200296e-01 2.76168257e-01 -5.16293585e-01 4.08397645e-01 8.29192340e-01 6.20945334e-01 -4.11900342e-01 8.72447193e-01 4.58635300e-01 -1.76948592e-01 -9.30712640e-01 -5.45548081e-01 -4.89523679e-01 -9.95766103e-01 1.33699685e-01 7.81713307e-01 -7.07315087e-01 -7.90352076e-02 4.55584496e-01 -9.73767042e-01 -4.55950290e-01 -8.18655312e-01 4.97330546e-01 -9.84200001e-01 4.54521805e-01 -3.21162850e-01 -5.67526996e-01 1.67001784e-01 -9.19451058e-01 6.57042801e-01 1.45290077e-01 -2.58854002e-01 -1.21955335e+00 1.88065484e-01 -6.12702854e-02 3.41371655e-01 2.95682698e-01 9.85100210e-01 -1.26980436e+00 -1.39961481e-01 -2.58873731e-01 4.58559766e-02 8.31089437e-01 4.39286023e-01 -4.00016785e-01 -1.06953323e+00 -4.38889325e-01 1.10904470e-01 -7.30489194e-01 6.89902008e-01 1.33766562e-01 1.11885047e+00 1.91556644e-02 -2.39300191e-01 4.51785773e-01 1.41672134e+00 3.47430289e-01 5.52908182e-01 4.58540380e-01 4.81145024e-01 6.04451835e-01 6.44719481e-01 2.54494041e-01 -2.81911582e-01 8.42443466e-01 1.92261547e-01 -8.16901587e-03 -1.62015647e-01 -1.46700770e-01 1.79985911e-01 3.04092884e-01 2.64365554e-01 -6.08336739e-02 -1.07590878e+00 6.69329882e-01 -1.71849847e+00 -9.60471153e-01 6.46280527e-01 2.36352110e+00 8.02695751e-01 1.43266784e-03 2.07047895e-01 1.99645147e-01 8.10634375e-01 1.38133122e-02 -7.26773143e-01 -4.02930170e-01 -4.39228788e-02 4.79967475e-01 4.60710257e-01 3.38079065e-01 -1.42916596e+00 8.25392008e-01 6.57168531e+00 8.28633249e-01 -1.10054326e+00 1.15647227e-01 4.72195596e-01 1.03271388e-01 1.19646080e-01 -1.70152903e-01 -6.14494145e-01 5.03303587e-01 8.74686956e-01 -9.06746015e-02 4.15530115e-01 1.21759653e+00 -3.71270329e-01 8.16557836e-03 -1.39940095e+00 9.54908848e-01 1.18497625e-01 -9.49646235e-01 7.95017481e-02 -7.83897713e-02 6.92767501e-01 -1.80018455e-01 1.15273103e-01 2.79589683e-01 3.66761804e-01 -1.01419568e+00 2.92840064e-01 1.94569558e-01 8.96167219e-01 -8.21145594e-01 6.32310748e-01 4.73922282e-01 -6.65592611e-01 -3.01266074e-01 -7.03706324e-01 9.05263349e-02 -1.81344241e-01 1.80447012e-01 -7.88820148e-01 2.01609269e-01 5.43214083e-01 7.10346520e-01 -3.18446636e-01 1.02660251e+00 -8.41230750e-02 4.24243987e-01 -1.71792805e-01 1.01801164e-01 1.84395060e-01 -1.55226916e-01 5.84411800e-01 1.06612277e+00 2.99332917e-01 -5.56258038e-02 1.33794874e-01 6.92719460e-01 4.86219628e-03 -6.76206797e-02 -9.80817616e-01 1.32096596e-02 3.48327875e-01 8.79884481e-01 -6.20660365e-01 -4.58534420e-01 -4.30050105e-01 1.48093617e+00 5.09572685e-01 4.68540251e-01 -5.97949147e-01 -5.87833345e-01 8.25681686e-01 -1.63394719e-01 6.39182866e-01 -2.33715415e-01 -4.92805392e-01 -1.21976650e+00 -1.34294882e-01 -9.95222330e-01 3.60257804e-01 -3.92252266e-01 -1.71285093e+00 4.23384786e-01 6.52841553e-02 -1.51090240e+00 -5.09075880e-01 -9.67853725e-01 -5.37672222e-01 1.00735521e+00 -1.32150400e+00 -9.02256846e-01 -2.40304023e-01 7.48302341e-01 6.99972987e-01 -6.79347575e-01 1.21875894e+00 2.73289859e-01 -2.94804454e-01 7.78190136e-01 6.09650314e-01 3.11626464e-01 1.00098908e+00 -1.42390883e+00 2.61507124e-01 9.18474019e-01 1.87589481e-01 3.09458435e-01 7.30360389e-01 -2.93797612e-01 -9.50040936e-01 -1.06134963e+00 6.23166263e-01 -5.32012939e-01 5.61729908e-01 -4.30263996e-01 -1.09691954e+00 5.77248633e-01 1.44059390e-01 2.49687940e-01 9.14696932e-01 -1.06350020e-01 -4.53215241e-01 -1.95087597e-01 -1.45191646e+00 4.59220558e-01 7.48214424e-01 -5.52706957e-01 -1.06187427e+00 3.68606687e-01 2.09583148e-01 -1.07687980e-01 -7.58331954e-01 1.18178867e-01 3.12686801e-01 -7.24481881e-01 1.14322293e+00 -9.49654281e-01 4.40422952e-01 -1.56861916e-01 -3.31667513e-01 -1.69239712e+00 -3.36131662e-01 -1.99611187e-01 -2.11003870e-01 1.20312881e+00 2.17853427e-01 -7.22280800e-01 8.32744420e-01 6.10712945e-01 3.95737022e-01 -3.13122422e-01 -1.00289726e+00 -1.23984611e+00 4.11207676e-01 5.14018722e-02 2.43864909e-01 1.07685995e+00 -2.33124077e-01 3.03436249e-01 -3.01300645e-01 6.03961945e-02 8.33182216e-01 -4.50145826e-02 8.10601175e-01 -1.26033258e+00 -5.83001137e-01 -1.46698773e-01 -8.35583150e-01 -8.46338987e-01 4.12483782e-01 -7.46939600e-01 1.28311664e-01 -1.03355491e+00 1.34917781e-01 -3.93951118e-01 -5.20584583e-01 5.29158294e-01 -6.57041371e-02 1.88702077e-01 3.15304071e-01 1.74083337e-01 -1.68361887e-01 4.94106501e-01 7.67187715e-01 -4.20000434e-01 -2.03174174e-01 9.23278704e-02 -4.90903080e-01 7.39252865e-01 8.89314711e-01 -5.70827842e-01 -4.41992700e-01 -2.53575355e-01 -6.85145319e-01 -2.80800939e-01 5.00230491e-01 -1.10376716e+00 1.10430874e-01 -1.77838758e-01 8.02616656e-01 2.51284726e-02 4.57468480e-01 -1.14552796e+00 -1.66327581e-01 4.69305485e-01 -3.62972736e-01 -4.45915550e-01 2.83043861e-01 6.97664738e-01 -4.54888076e-01 -5.77648580e-01 1.53331935e+00 -2.48430133e-01 -1.14696646e+00 1.04916804e-01 -2.23582625e-01 1.60172164e-01 1.43601561e+00 -2.66584843e-01 -1.95160403e-03 -1.78983212e-01 -9.00369346e-01 -9.26098377e-02 6.90368474e-01 4.25641596e-01 5.13949931e-01 -1.59198236e+00 -4.98788923e-01 3.02993596e-01 3.18193734e-01 -3.43233943e-01 2.84644892e-03 2.08580658e-01 -3.31876695e-01 1.17630228e-01 -8.27661097e-01 -6.40315771e-01 -1.16971982e+00 9.11933482e-01 4.21434104e-01 -1.60620615e-01 -4.56508398e-01 7.51679003e-01 2.04508528e-01 -4.42270279e-01 3.07121515e-01 2.20598713e-01 -1.27918914e-01 1.51175678e-01 6.16642892e-01 3.22680026e-01 -6.16813339e-02 -3.85128289e-01 -3.47721040e-01 5.37967801e-01 -1.74242929e-01 -1.89479887e-01 1.27477169e+00 1.08421907e-01 2.28030488e-01 4.86877203e-01 1.37429643e+00 -4.11500424e-01 -1.65014410e+00 -4.28433537e-01 7.15584904e-02 -6.33387804e-01 -2.64331639e-01 -7.18741596e-01 -6.27620280e-01 1.17080986e+00 9.67950702e-01 1.15141757e-01 1.26224220e+00 -7.54294768e-02 4.39448982e-01 5.76874554e-01 2.54226714e-01 -1.33776200e+00 3.38752359e-01 5.08366406e-01 6.88410699e-01 -1.55611026e+00 -2.41835728e-01 -1.27758846e-01 -7.97885299e-01 1.26584172e+00 6.77563310e-01 -4.21478420e-01 5.26235580e-01 9.08484459e-02 1.39732480e-01 3.08735579e-01 -2.24676162e-01 -1.97310388e-01 1.92616433e-01 1.10308337e+00 5.03129587e-02 3.82226380e-03 2.65714545e-02 2.78676182e-01 2.15755701e-01 -2.19941318e-01 3.48539293e-01 9.41500187e-01 -4.40557182e-01 -1.49687159e+00 -4.30571645e-01 2.95945704e-01 -4.33388054e-01 1.05943941e-01 -4.66062695e-01 7.89008200e-01 -2.64540426e-02 7.44568408e-01 2.04543948e-01 -3.67624015e-02 3.68993968e-01 4.34211761e-01 6.05916023e-01 -8.08311224e-01 -1.09870441e-01 -3.31581503e-01 -2.63278574e-01 -2.87111044e-01 -2.97855198e-01 -6.31649375e-01 -8.03113639e-01 1.29511401e-01 4.81068008e-02 3.72361355e-02 5.37599862e-01 8.52889717e-01 2.01402858e-01 2.30553225e-01 7.70086884e-01 -8.95066559e-01 -1.18535101e+00 -9.13162470e-01 -7.69076407e-01 7.34501600e-01 4.80671376e-01 -8.92982125e-01 -3.89813602e-01 3.60141307e-01]
[9.904926300048828, 2.7128396034240723]
95fa74ab-eeb7-4226-bc82-403ba925b13e
dualfair-fair-representation-learning-at-both
2303.08403
null
https://arxiv.org/abs/2303.08403v1
https://arxiv.org/pdf/2303.08403v1.pdf
DualFair: Fair Representation Learning at Both Group and Individual Levels via Contrastive Self-supervision
Algorithmic fairness has become an important machine learning problem, especially for mission-critical Web applications. This work presents a self-supervised model, called DualFair, that can debias sensitive attributes like gender and race from learned representations. Unlike existing models that target a single type of fairness, our model jointly optimizes for two fairness criteria - group fairness and counterfactual fairness - and hence makes fairer predictions at both the group and individual levels. Our model uses contrastive loss to generate embeddings that are indistinguishable for each protected group, while forcing the embeddings of counterfactual pairs to be similar. It then uses a self-knowledge distillation method to maintain the quality of representation for the downstream tasks. Extensive analysis over multiple datasets confirms the model's validity and further shows the synergy of jointly addressing two fairness criteria, suggesting the model's potential value in fair intelligent Web applications.
['Meeyoung Cha', 'Xing Xie', 'Xiting Wang', 'Chuhan Wu', 'Sundong Kim', 'Fangzhao Wu', 'Seungeon Lee', 'Sungwon Han']
2023-03-15
null
null
null
null
['self-knowledge-distillation']
['computer-vision']
[-2.44982451e-01 5.40983140e-01 -7.28173673e-01 -8.33187759e-01 -3.64646822e-01 -3.95037681e-01 8.75136077e-01 3.66935223e-01 -6.19895339e-01 1.02526426e+00 5.80099463e-01 -4.06509399e-01 -1.67542547e-01 -8.73257339e-01 -2.17449203e-01 -1.59853593e-01 4.63948660e-02 3.07648093e-01 -4.25797701e-01 -2.93057859e-01 3.70140254e-01 4.68409806e-02 -1.46939862e+00 1.13992698e-01 1.28622329e+00 9.43196356e-01 -7.66572058e-01 3.72841895e-01 -6.36290060e-03 1.22685575e+00 -4.03334439e-01 -1.47997689e+00 7.08042145e-01 -2.23486796e-01 -8.52956653e-01 -7.20014036e-01 6.50405824e-01 -7.08039522e-01 -3.17990452e-01 1.11758995e+00 8.07772815e-01 2.80469298e-01 8.44557106e-01 -1.80592680e+00 -1.20479822e+00 8.53793919e-01 -5.26709914e-01 -2.81667978e-01 4.11394984e-02 2.96195388e-01 1.54941750e+00 -1.65901721e-01 2.57496089e-01 1.53847051e+00 8.71467590e-01 9.71021533e-01 -1.32737577e+00 -1.14761138e+00 -4.12774682e-02 1.25813931e-01 -9.15887117e-01 -5.82623303e-01 6.59013689e-01 -3.56212199e-01 4.79420215e-01 4.79371488e-01 3.48524392e-01 1.05345237e+00 2.14158967e-01 4.78412330e-01 1.41631520e+00 -2.72975743e-01 4.86475825e-01 1.95478737e-01 1.96155906e-01 4.26143736e-01 7.64355659e-01 5.58950245e-01 -4.44754362e-01 -7.18535781e-01 3.05081010e-01 9.31450445e-03 1.77912876e-01 -4.59932953e-01 -8.28250825e-01 1.32335865e+00 3.85789484e-01 -2.53931552e-01 -3.68587464e-01 4.63951647e-01 5.33089936e-01 3.57385606e-01 6.94634020e-01 7.32179403e-01 -2.44423300e-01 1.69292614e-02 -1.06202912e+00 8.63800585e-01 9.15319383e-01 4.46549892e-01 6.01287603e-01 -5.58959395e-02 -7.63536811e-01 8.40516806e-01 1.87869236e-01 2.92270124e-01 4.46549475e-01 -1.42204118e+00 3.51066679e-01 5.18928289e-01 3.40415239e-01 -1.00444484e+00 -1.23936750e-01 -4.11925204e-02 -6.60090148e-01 7.44618714e-01 5.08965790e-01 -2.37596110e-01 -4.96163905e-01 2.24254274e+00 3.77277881e-01 -2.58791476e-01 -1.15365729e-01 9.93576467e-01 6.44219279e-01 5.70379347e-02 8.04301977e-01 1.00981288e-01 1.10634851e+00 -7.76829422e-01 -6.91036582e-01 -1.72789425e-01 4.74496305e-01 -4.15468246e-01 9.60106850e-01 -1.08334839e-01 -1.03070509e+00 -2.01686054e-01 -7.86207080e-01 -4.30419147e-01 -3.08005780e-01 -3.92025679e-01 1.03965759e+00 1.15559947e+00 -9.17480826e-01 9.42103744e-01 3.17355022e-02 3.58333327e-02 9.15960670e-01 2.78178036e-01 -2.85651177e-01 -5.45971990e-02 -1.67759597e+00 1.09951758e+00 8.15961808e-02 -5.64176083e-01 -6.16720796e-01 -1.09964418e+00 -7.89635122e-01 4.39624757e-01 1.61166385e-01 -1.09680808e+00 9.83081877e-01 -1.28659546e+00 -1.42224121e+00 1.17809546e+00 2.78088272e-01 -5.97085536e-01 1.11959720e+00 -7.64695704e-02 -3.18667173e-01 -3.18839788e-01 3.63356173e-01 7.09735274e-01 7.88890898e-01 -1.13677764e+00 -4.62465763e-01 -4.67890382e-01 3.58129084e-01 4.40191448e-01 -5.27195215e-01 7.38217160e-02 4.77371305e-01 -6.10614836e-01 -9.08125162e-01 -4.59617913e-01 -3.84237260e-01 4.82757300e-01 -3.13665926e-01 -3.06012243e-01 1.16850026e-01 -6.40761554e-01 1.36196804e+00 -1.92028677e+00 -3.79245937e-01 1.54065087e-01 5.81953466e-01 2.32062697e-01 -2.71328270e-01 -9.11590382e-02 -3.18545163e-01 6.17803514e-01 -1.02205433e-01 -3.51379395e-01 5.64812660e-01 -1.70119032e-01 -1.07596576e-01 5.28944552e-01 1.02476239e-01 9.33553219e-01 -9.40674365e-01 -5.12080729e-01 -2.00999305e-02 -4.49813269e-02 -9.90023911e-01 3.59467477e-01 -3.13039236e-02 -2.29515016e-01 -2.16525018e-01 2.14210346e-01 1.04187906e+00 3.34272265e-01 4.08913553e-01 -1.26331121e-01 -4.06422503e-02 5.18237054e-01 -9.02482152e-01 1.05345845e+00 -3.77078235e-01 1.17450543e-01 2.62402240e-02 -7.12167621e-01 8.24783981e-01 -6.13374673e-02 3.07017773e-01 -1.00215542e+00 3.29162598e-01 1.62246004e-02 7.56151304e-02 -1.14766493e-01 6.96970344e-01 -6.83726251e-01 -4.17250216e-01 9.96063769e-01 -1.14222355e-01 -9.17980224e-02 -2.00435847e-01 3.32134128e-01 7.33091891e-01 -6.63767308e-02 5.90771973e-01 -6.55604243e-01 1.77222639e-01 -2.49399573e-01 8.01382959e-01 8.72711718e-01 -9.80857193e-01 4.27173704e-01 9.10695910e-01 -5.97553849e-01 -1.10288596e+00 -1.21675622e+00 -9.10844281e-02 1.52756751e+00 1.40584528e-01 -5.09669818e-02 -5.49469173e-01 -1.24528813e+00 1.00742030e+00 1.16168523e+00 -1.02359140e+00 -6.85188532e-01 4.03000079e-02 -7.95638084e-01 8.25137019e-01 1.95284650e-01 4.60881382e-01 -5.80524206e-01 -4.78680432e-01 -3.10632467e-01 -1.29095733e-01 -4.12739217e-01 -5.57435155e-01 -1.67429760e-01 -5.07686675e-01 -1.26744306e+00 -4.15072203e-01 -4.41814326e-02 1.95269346e-01 -7.73434639e-02 1.33963895e+00 1.45778567e-01 -2.80497670e-01 2.57381499e-01 2.46736892e-02 -5.99194169e-01 -3.14075798e-01 -4.56463695e-01 2.83175647e-01 -7.42568895e-02 5.47841072e-01 -5.00597656e-01 -8.12371969e-01 3.51558216e-02 -7.01309860e-01 -2.47492850e-01 -2.73558069e-02 9.54631746e-01 -1.04673855e-01 -3.59622121e-01 1.04240227e+00 -1.20545149e+00 1.28604019e+00 -9.33600843e-01 -1.82008401e-01 1.80261567e-01 -1.15889001e+00 2.81047244e-02 7.71701872e-01 -1.82639390e-01 -1.31041038e+00 -7.55995512e-01 3.05350453e-01 -1.07207432e-01 1.65308058e-01 -1.67323619e-01 -2.62340009e-01 1.87159628e-01 9.40467179e-01 -6.57373250e-01 4.90371972e-01 -1.90581918e-01 9.61753011e-01 8.42757344e-01 3.13850433e-01 -8.79474878e-01 7.01948524e-01 4.41176087e-01 -1.71197772e-01 -6.49437914e-03 -8.76219034e-01 1.10112689e-01 -6.33794665e-02 -8.55224803e-02 6.03037715e-01 -8.23756397e-01 -1.10171783e+00 2.05137014e-01 -7.86920011e-01 -3.19543660e-01 -6.75370812e-01 2.63308793e-01 -4.80372012e-01 3.08505863e-01 -3.43625039e-01 -1.14864898e+00 -4.79471236e-01 -4.99798417e-01 4.60838228e-01 2.78861374e-01 -6.08272672e-01 -1.10886014e+00 1.38730258e-01 5.40412664e-01 6.01173759e-01 2.37683699e-01 1.08140767e+00 -9.23436582e-01 9.67586413e-02 -1.20282874e-01 -6.38074040e-01 4.29794282e-01 1.58922046e-01 -1.00644313e-01 -1.13318443e+00 -3.19530927e-02 -3.62971127e-01 -7.15771675e-01 9.87007141e-01 2.60525584e-01 1.33425641e+00 -7.72090852e-01 1.91336140e-01 3.95815909e-01 1.25888181e+00 -3.80662888e-01 6.67558908e-01 3.35707277e-01 3.69619161e-01 8.43456447e-01 4.78455275e-01 8.67155015e-01 8.81093919e-01 3.14272940e-01 5.85391641e-01 -2.35181600e-01 -6.62903860e-02 -3.53408933e-01 2.51834512e-01 -2.78985232e-01 -5.31374924e-02 1.74071521e-01 -7.04996407e-01 5.94354868e-01 -1.87922120e+00 -1.38151431e+00 1.73055783e-01 2.49355865e+00 1.13890040e+00 -1.53705105e-01 4.16096151e-01 -1.44213066e-01 8.18524659e-01 4.42825824e-01 -7.11324751e-01 -1.23259890e+00 3.98581512e-02 3.64134938e-01 6.65313601e-01 5.40192306e-01 -1.03523040e+00 8.26465547e-01 6.84708261e+00 6.98795676e-01 -5.91379106e-01 1.78579345e-01 1.19446456e+00 -5.34769297e-01 -1.04054904e+00 1.38743455e-02 -1.25395119e-01 6.69433236e-01 7.12775946e-01 -9.14486170e-01 5.60962856e-01 9.31507170e-01 2.26786926e-01 -1.04928903e-01 -1.22131360e+00 6.11951768e-01 -2.63706893e-02 -1.18919313e+00 1.92112371e-01 2.76876688e-02 9.00948107e-01 -3.65544796e-01 4.01452988e-01 5.22939026e-01 1.29855108e+00 -1.39229417e+00 8.66024196e-01 5.12461543e-01 1.04377675e+00 -1.02442503e+00 7.84924328e-01 6.95700943e-02 -2.47014284e-01 -4.83923078e-01 -4.52883899e-01 -5.08633077e-01 -1.08812205e-01 8.74622762e-01 -3.49253833e-01 4.75178689e-01 5.92323363e-01 4.20353264e-01 -5.63315690e-01 6.36498630e-01 -9.95389298e-02 2.73721397e-01 1.74464568e-01 1.75264627e-01 -7.15051666e-02 -1.53676331e-01 1.58487558e-01 9.71287310e-01 -4.21179123e-02 -1.49249643e-01 -1.90939888e-01 1.26913118e+00 -6.42509103e-01 1.49405941e-01 -6.15921497e-01 -2.43795738e-02 7.60029078e-01 1.17265046e+00 2.19156653e-01 -3.55044037e-01 -8.00124928e-02 9.45285380e-01 6.34153724e-01 1.60358965e-01 -8.96122634e-01 -4.42231745e-01 1.53001046e+00 6.78298771e-02 -3.35405111e-01 3.21934700e-01 -9.48172510e-01 -1.12243104e+00 -3.15832675e-01 -9.76756215e-01 7.24647343e-01 -3.35904598e-01 -1.90260708e+00 -1.39271468e-01 -3.17940801e-01 -7.10740805e-01 -9.87718403e-02 -4.90042031e-01 -9.99215007e-01 1.22243011e+00 -1.95500624e+00 -1.31856287e+00 3.60190719e-02 4.22211707e-01 -3.77130479e-01 -2.00242534e-01 9.53863442e-01 1.21819921e-01 -4.60409045e-01 1.18974102e+00 1.11270204e-01 4.06456143e-02 1.41307354e+00 -1.31333876e+00 2.75697321e-01 4.23410833e-01 -6.40340030e-01 9.17361557e-01 6.05137229e-01 -5.80076933e-01 -6.53158128e-01 -1.12070417e+00 1.29606020e+00 -6.65691912e-01 5.63459158e-01 -1.92736775e-01 -4.41935301e-01 3.63335550e-01 2.26552203e-01 -2.10897345e-02 1.31889641e+00 5.36891699e-01 -1.11125660e+00 -2.82287180e-01 -1.89865112e+00 8.09617460e-01 1.14809656e+00 -6.45024002e-01 -6.23093843e-01 -4.46022907e-03 5.86027145e-01 2.11291507e-01 -6.87081039e-01 5.67431450e-02 9.60595012e-01 -1.36018884e+00 9.72753048e-01 -1.39160621e+00 1.12740409e+00 2.31637508e-01 -2.46884838e-01 -1.55486953e+00 -7.27905273e-01 -5.06885052e-01 1.12898294e-02 1.44923162e+00 1.43400177e-01 -8.06720793e-01 6.13807738e-01 1.47989810e+00 5.28234482e-01 -4.60945666e-01 -1.04514503e+00 -7.08483577e-01 7.61105001e-01 -2.13979155e-01 1.13667572e+00 1.38570869e+00 2.01883569e-01 7.50965327e-02 -6.72026098e-01 -2.52239347e-01 1.12044919e+00 1.70818344e-01 6.85612261e-01 -1.45539725e+00 -2.26673055e-02 -8.98100376e-01 -2.44970411e-01 -1.92287654e-01 7.46600509e-01 -1.15350616e+00 -4.26160604e-01 -1.17763507e+00 6.89011574e-01 -7.26429939e-01 -4.36906606e-01 5.52098513e-01 -6.22444630e-01 1.40590653e-01 3.82188767e-01 -3.00223023e-01 -5.15190363e-01 7.78063118e-01 8.71410131e-01 -5.03474027e-02 1.95065737e-01 -1.89508185e-01 -1.79134810e+00 4.87890273e-01 8.24437261e-01 -3.92290086e-01 -2.03161135e-01 -2.49414235e-01 2.75126278e-01 -4.53453034e-01 5.98173201e-01 -3.02928954e-01 -1.61845550e-01 -8.27990651e-01 4.45748687e-01 6.12506747e-01 -1.17958412e-02 -6.31861389e-01 -2.96247900e-01 3.01905125e-01 -8.75044525e-01 -1.90354288e-01 2.28053201e-02 5.13269424e-01 2.66267955e-01 -7.25621954e-02 1.07981038e+00 -1.27783015e-01 -3.34865987e-01 3.56050104e-01 2.77841780e-02 4.02139038e-01 9.64597642e-01 6.75822571e-02 -6.32449150e-01 -6.40627444e-01 -2.21608922e-01 5.92140794e-01 7.31150031e-01 4.19567108e-01 2.20740631e-01 -1.73420703e+00 -1.01720250e+00 9.30625293e-03 2.35042483e-01 -7.80392289e-01 3.02852005e-01 2.49506414e-01 -7.63454139e-02 -1.36283234e-01 -6.17092311e-01 3.64049017e-01 -1.02380860e+00 4.33383137e-01 4.30788875e-01 -3.08792472e-01 1.35337368e-01 7.14965284e-01 2.06232458e-01 -9.68396306e-01 -2.44897939e-02 5.08888841e-01 -5.02712503e-02 1.00937866e-01 5.65135419e-01 7.82959342e-01 -3.99216115e-01 -3.45678121e-01 -3.32090914e-01 -7.68693313e-02 8.10753256e-02 -6.78957179e-02 1.25516891e+00 -2.19392613e-01 -3.54366899e-01 -2.13702433e-02 1.08957386e+00 1.12459131e-01 -1.15499747e+00 -3.59676871e-03 -6.72349185e-02 -1.17564356e+00 6.69071302e-02 -1.22404265e+00 -7.89297640e-01 8.10794055e-01 4.84659314e-01 -6.01824559e-02 7.99355567e-01 -4.65785176e-01 2.92990118e-01 -1.21703997e-01 1.86893806e-01 -1.52785134e+00 -1.01246506e-01 2.12178499e-01 6.35390043e-01 -1.44259858e+00 2.81873405e-01 -4.67795208e-02 -9.94060278e-01 7.99095333e-01 8.09801757e-01 -7.18909204e-02 5.62191427e-01 2.23148731e-03 1.04662172e-01 2.05067322e-01 -8.01503658e-01 -6.99772909e-02 5.59036061e-02 8.10861528e-01 7.44241416e-01 6.84479952e-01 -8.47912490e-01 1.17610085e+00 -3.69232953e-01 4.73047681e-02 5.85535347e-01 4.44413662e-01 -1.15375802e-01 -1.25790620e+00 -1.86862096e-01 9.37422097e-01 -6.55232131e-01 -2.96623677e-01 -6.53777897e-01 4.63138759e-01 3.28041345e-01 9.05467212e-01 1.80444613e-01 -4.84094411e-01 2.02039719e-01 1.88094601e-01 2.23677740e-01 -4.62678432e-01 -9.41845179e-01 -8.75371099e-01 3.71698976e-01 -9.26368654e-01 -8.70300531e-02 -5.10627806e-01 -7.69827127e-01 -1.21389198e+00 5.60257435e-02 1.89039320e-01 2.65544444e-01 5.06788492e-01 5.46924591e-01 1.45218089e-01 1.01602101e+00 -4.00698602e-01 -1.26817596e+00 -4.45950389e-01 -8.34029317e-01 1.10708058e+00 2.36764699e-01 -5.97897649e-01 -4.81395513e-01 -4.21972871e-01]
[8.952082633972168, 5.317002773284912]
e93c1622-77da-4cd1-90c2-487faf235f94
ctrl-connect-tabular-and-language-model-for
2306.02841
null
https://arxiv.org/abs/2306.02841v2
https://arxiv.org/pdf/2306.02841v2.pdf
CTRL: Connect Tabular and Language Model for CTR Prediction
Traditional click-through rate (CTR) prediction models convert the tabular data into one-hot vectors and leverage the collaborative relations among features for inferring user's preference over items. This modeling paradigm discards the essential semantic information. Though some recent works like P5 and M6-Rec have explored the potential of using Pre-trained Language Models (PLMs) to extract semantic signals for CTR prediction, they are computationally expensive and suffer from low efficiency. Besides, the beneficial collaborative relations are not considered, hindering the recommendation performance. To solve these problems, in this paper, we propose a novel framework \textbf{CTRL}, which is industrial friendly and model-agnostic with high training and inference efficiency. Specifically, the original tabular data is first converted into textual data. Both tabular data and converted textual data are regarded as two different modalities and are separately fed into the collaborative CTR model and pre-trained language model. A cross-modal knowledge alignment procedure is performed to fine-grained align and integrate the collaborative and semantic signals, and the lightweight collaborative model can be deployed online for efficient serving after fine-tuned with supervised signals. Experimental results on three public datasets show that CTRL outperforms the SOTA CTR models significantly. Moreover, we further verify its effectiveness on a large-scale industrial recommender system.
['Ruiming Tang', 'Lu Hou', 'Bo Chen', 'Xiangyang Li']
2023-06-05
null
null
null
null
['click-through-rate-prediction']
['miscellaneous']
[ 2.68070903e-02 -5.57923019e-01 -6.73412085e-01 -5.13985813e-01 -7.58348942e-01 -5.22348225e-01 3.21503758e-01 -1.97750121e-01 -2.32072741e-01 3.35784286e-01 4.07040626e-01 -1.94709122e-01 -5.12615025e-01 -7.80389786e-01 -4.67626721e-01 -4.22415107e-01 2.51421988e-01 4.01595950e-01 -7.41085187e-02 -3.98211718e-01 1.35245040e-01 -3.16100627e-01 -1.45554459e+00 7.16429174e-01 1.08927512e+00 1.38118160e+00 5.32449841e-01 7.39812553e-02 -5.10764182e-01 6.10314667e-01 -1.13881633e-01 -7.24062204e-01 2.62435764e-01 -9.92342085e-02 -3.07210505e-01 -3.38541806e-01 -3.60447094e-02 -1.09002449e-01 -3.64070714e-01 8.55494797e-01 4.32466060e-01 3.35346967e-01 4.15193766e-01 -1.05927610e+00 -9.10484850e-01 1.22076404e+00 -5.46463132e-01 -1.98039696e-01 4.20409352e-01 -3.93421203e-01 1.63971603e+00 -1.19352722e+00 4.33272511e-01 1.06506479e+00 5.95649362e-01 3.07977647e-01 -1.11997890e+00 -1.04803169e+00 5.79362988e-01 3.82761091e-01 -1.27033103e+00 -4.06099856e-02 8.12298119e-01 -3.97316158e-01 6.45790815e-01 4.74607080e-01 4.46543217e-01 1.51643097e+00 -7.01346695e-02 1.04043305e+00 8.99730742e-01 -5.19900881e-02 -1.00480975e-04 4.72478449e-01 1.97809562e-01 4.42163885e-01 5.47211766e-02 -5.88268526e-02 -9.05540109e-01 7.40450714e-03 5.15351057e-01 5.91669083e-01 -5.32501936e-02 -2.93008983e-01 -1.28608251e+00 9.02325511e-01 3.77953202e-01 4.07942146e-01 -1.67865410e-01 -3.43245387e-01 4.30574387e-01 4.69149292e-01 5.43952942e-01 4.40739810e-01 -8.90527844e-01 -5.33397356e-03 -7.90627897e-01 -1.24438189e-01 5.83923101e-01 1.17826807e+00 6.00453198e-01 -2.14800015e-01 -4.12814289e-01 1.27327251e+00 6.31165385e-01 6.86844528e-01 8.16204011e-01 -4.16694969e-01 7.85520077e-01 6.43460333e-01 -3.44096944e-02 -1.46847296e+00 -2.71982551e-01 -8.49818230e-01 -7.80581713e-01 -9.17953730e-01 6.55345172e-02 -6.12947419e-02 -4.72400486e-01 1.51364517e+00 -9.59465001e-03 2.09301770e-01 -5.86822741e-02 1.06171846e+00 7.98445642e-01 6.86764300e-01 2.01275483e-01 -1.38104871e-01 1.29668844e+00 -1.13492346e+00 -7.73697197e-01 -2.58193582e-01 7.18180358e-01 -1.01370633e+00 1.17816138e+00 6.27817094e-01 -6.15078151e-01 -8.16766322e-01 -9.99154508e-01 1.74868610e-02 -4.28727955e-01 7.29214370e-01 9.96276975e-01 5.88754952e-01 -1.75625160e-01 4.37745005e-01 -3.43031287e-01 -3.94904427e-02 2.07427353e-01 2.03924745e-01 -3.83449569e-02 -3.20246339e-01 -1.54672360e+00 4.70581949e-01 3.27490628e-01 3.18098277e-01 -2.96287924e-01 -7.58754015e-01 -5.27635932e-01 9.67125148e-02 6.60775125e-01 -6.26860440e-01 9.54871953e-01 -8.34636033e-01 -1.68927348e+00 5.69202378e-03 -7.20866919e-02 -1.70445889e-01 2.22017124e-01 -3.99226725e-01 -1.09793520e+00 -5.65385938e-01 1.26860884e-03 5.98684438e-02 6.84580564e-01 -1.07758212e+00 -9.25386250e-01 -2.92288154e-01 9.18186456e-02 2.75452465e-01 -8.98369253e-01 -3.68217081e-02 -1.08199632e+00 -1.00380445e+00 1.46241382e-01 -7.56354153e-01 -3.47433448e-01 -5.80230236e-01 -4.08379614e-01 -1.22467868e-01 3.36940914e-01 -5.64374804e-01 1.61893332e+00 -2.13797140e+00 7.88312629e-02 6.11462533e-01 1.18186381e-02 1.45387888e-01 -3.98040622e-01 4.31052804e-01 1.88742116e-01 -1.21738791e-01 4.14116263e-01 -3.17553520e-01 3.21266502e-01 1.10147536e-01 -7.12449074e-01 3.92871760e-02 -4.02265877e-01 1.00010407e+00 -8.36074531e-01 -4.33153689e-01 2.89464742e-01 4.62566674e-01 -1.00352859e+00 2.24664643e-01 -2.97219038e-01 4.12065595e-01 -9.28007662e-01 5.76964617e-01 5.21944880e-01 -5.65163672e-01 5.14990926e-01 -8.07366788e-01 3.77080925e-02 4.62538570e-01 -1.17033792e+00 1.96827245e+00 -9.36464906e-01 -8.12396333e-02 -3.89402360e-01 -9.65955377e-01 1.07788980e+00 3.35775428e-02 8.28766227e-01 -1.17426944e+00 2.03237340e-01 2.04745650e-01 -3.64309520e-01 -4.78832692e-01 6.40020370e-01 1.21349409e-01 -3.44080627e-01 4.46810186e-01 1.98063821e-01 5.15294552e-01 5.11519276e-02 1.72768652e-01 5.95217466e-01 3.35960329e-01 -1.13012455e-01 1.53903365e-01 6.51519656e-01 -7.86454380e-02 6.95323706e-01 6.49568319e-01 5.66583574e-01 3.74380320e-01 -1.82985052e-01 -6.74758330e-02 -5.07381737e-01 -8.86847138e-01 2.45870184e-02 1.56980920e+00 5.54933548e-01 -1.03440297e+00 -1.22882888e-01 -8.06345284e-01 5.50603233e-02 8.62547696e-01 -3.81825686e-01 -2.59091973e-01 -3.65749985e-01 -8.47483933e-01 1.22152828e-01 6.00895166e-01 1.49945647e-01 -5.80475092e-01 3.07371497e-01 3.91488671e-01 -5.91664076e-01 -1.07915235e+00 -7.19252944e-01 5.32851694e-03 -6.70734107e-01 -7.36104608e-01 -3.18761587e-01 -4.04385418e-01 4.46065933e-01 6.36163771e-01 9.16429400e-01 -2.32952029e-01 1.20168217e-01 1.33749440e-01 -7.74293482e-01 -1.01151571e-01 2.86269873e-01 1.79066643e-01 1.02120265e-01 4.44321245e-01 6.19619787e-01 -3.97831470e-01 -6.66650891e-01 8.51714671e-01 -5.18509030e-01 2.01755121e-01 7.25160599e-01 9.19821560e-01 6.72531545e-01 1.01845235e-01 5.67140162e-01 -1.15289426e+00 6.41963422e-01 -8.34037364e-01 -4.11029190e-01 4.35509980e-01 -9.87162352e-01 -2.10258737e-02 1.00053060e+00 -5.79619944e-01 -1.33174002e+00 -1.64301351e-01 -1.06870227e-01 -4.39401418e-01 3.14044148e-01 1.07329607e+00 -1.68268070e-01 3.47606421e-01 3.64993602e-01 2.95028210e-01 -4.53003466e-01 -9.92320836e-01 6.89483285e-01 7.90998220e-01 2.65936166e-01 -6.77572250e-01 1.04453444e+00 2.54150182e-01 -5.40782571e-01 -1.47340208e-01 -1.50888145e+00 -8.58138442e-01 -3.87313902e-01 -9.65408757e-02 4.04449224e-01 -1.13845646e+00 -5.72857618e-01 -1.45896956e-01 -6.09482884e-01 2.01021656e-01 -8.63413066e-02 9.47018206e-01 -3.85159522e-01 3.20048928e-01 -4.18061882e-01 -5.91290355e-01 -4.24329966e-01 -9.05143261e-01 9.03328896e-01 1.90478340e-01 -8.54515284e-02 -9.58635449e-01 -1.09800644e-01 9.39032733e-01 6.47360981e-01 -7.96211243e-01 8.53953481e-01 -9.10819829e-01 -4.48205292e-01 -2.84407705e-01 -3.96474689e-01 1.61395982e-01 1.98594958e-01 -2.87962943e-01 -8.29117298e-01 -1.86723396e-01 -2.84630060e-01 -2.21199140e-01 7.20184386e-01 -7.24252984e-02 1.65356636e+00 -3.08382392e-01 -4.04272676e-01 6.25893891e-01 1.20359087e+00 1.49993286e-01 2.54242867e-01 2.18185872e-01 8.94498467e-01 4.54232126e-01 1.27754664e+00 5.44520497e-01 6.25356734e-01 9.73106027e-01 6.20982908e-02 6.66604377e-03 1.14200609e-02 -7.21510589e-01 5.83717465e-01 1.62114859e+00 -1.33763671e-01 -2.29990929e-01 -2.74834454e-01 6.71100020e-02 -2.12614012e+00 -9.72533822e-01 -3.81091014e-02 2.12966895e+00 9.47547257e-01 7.01730838e-03 -7.00981095e-02 -8.29044282e-02 3.72984082e-01 -5.36331125e-02 -4.45324898e-01 6.81809336e-02 -9.93161798e-02 1.47229761e-01 5.21001279e-01 5.55856824e-02 -8.91848624e-01 8.76026690e-01 5.02180958e+00 1.43117464e+00 -1.13867593e+00 3.35429341e-01 2.06052169e-01 -3.37811828e-01 -6.52472794e-01 -9.99183655e-02 -9.89586651e-01 6.68951094e-01 8.80679846e-01 -3.40319276e-01 6.67482376e-01 8.72811019e-01 1.98116466e-01 5.38649261e-01 -9.79828000e-01 1.16318476e+00 1.28109351e-01 -1.20376897e+00 2.11490810e-01 -1.11977622e-01 6.54172182e-01 4.03106436e-02 3.12087089e-01 9.03239369e-01 4.67769325e-01 -8.16003501e-01 7.25124776e-01 7.04766631e-01 6.42718196e-01 -6.09123707e-01 7.77482212e-01 2.38812208e-01 -1.41708720e+00 -4.13539082e-01 -5.21347702e-01 3.50014687e-01 7.67177716e-02 7.37334609e-01 -3.74648482e-01 1.27973807e+00 7.62064219e-01 1.28403747e+00 -4.57492054e-01 8.57968390e-01 3.80751751e-02 6.42015576e-01 -8.27536210e-02 -7.98351914e-02 5.46485409e-02 -4.89143819e-01 1.59282133e-01 1.08004332e+00 6.57985628e-01 -1.44642711e-01 4.21377271e-01 7.49721944e-01 -1.07771792e-01 6.70003295e-01 -1.50742978e-01 -1.80477023e-01 4.56777245e-01 1.42686188e+00 -2.40474954e-01 -1.03898928e-01 -9.16713297e-01 7.79933512e-01 6.89417645e-02 3.82455617e-01 -1.14179862e+00 -9.77158248e-02 5.71501553e-01 -2.70673037e-01 4.81232256e-01 -1.03211239e-01 -2.77074486e-01 -1.70350027e+00 -2.35066060e-02 -9.43107188e-01 5.41455865e-01 -5.57923377e-01 -1.81284821e+00 4.05950606e-01 -2.89229721e-01 -1.71882164e+00 6.73087388e-02 -5.15645742e-01 -8.56220648e-02 7.34014511e-01 -1.51723325e+00 -1.30035889e+00 -1.87554851e-01 8.90245497e-01 7.42825866e-01 -4.09713119e-01 7.79630303e-01 9.59978223e-01 -8.04723144e-01 9.93757904e-01 3.48138720e-01 -3.01459432e-02 9.24300671e-01 -8.74792516e-01 -9.03479680e-02 4.65745568e-01 5.26024401e-01 1.12722421e+00 4.53418434e-01 -5.44264734e-01 -1.91514683e+00 -1.39901257e+00 7.98018813e-01 -4.38430160e-01 9.34737504e-01 -5.16384661e-01 -6.02347970e-01 4.97336775e-01 -2.78551131e-01 -2.14587882e-01 1.14555526e+00 7.22144365e-01 -7.68063784e-01 -4.22401845e-01 -5.88846743e-01 6.69449866e-01 1.26698267e+00 -7.02657700e-01 -5.11752248e-01 3.31131041e-01 6.53661788e-01 -4.05100659e-02 -1.25327170e+00 3.01075131e-01 8.67482185e-01 -5.63209116e-01 1.15876544e+00 -6.27471030e-01 2.94663548e-01 -3.60495836e-01 -6.25075638e-01 -1.31760430e+00 -3.61280859e-01 -3.09240133e-01 -1.00940451e-01 1.36237621e+00 7.28128254e-01 -5.89942157e-01 5.07708848e-01 3.34822595e-01 5.06695546e-02 -6.22914076e-01 -4.85095799e-01 -7.06471145e-01 -3.80755275e-01 -7.95444071e-01 7.53410995e-01 1.15977550e+00 2.75257319e-01 6.25474632e-01 -8.25660408e-01 -7.76977092e-02 3.59855115e-01 7.88082480e-01 6.93997741e-01 -1.32521653e+00 -5.34369946e-01 -3.42084527e-01 2.42547393e-01 -1.49565029e+00 -6.11883700e-02 -1.26450086e+00 -2.52909988e-01 -1.12749672e+00 3.13328713e-01 -1.02064693e+00 -9.37694609e-01 5.59066474e-01 3.28058340e-02 1.82073936e-01 9.88957286e-02 3.86523634e-01 -9.40769851e-01 7.24618435e-01 1.16381526e+00 -2.97953129e-01 -1.80499867e-01 1.77530393e-01 -1.01086187e+00 4.88660157e-01 5.24553359e-01 -4.35920388e-01 -7.97477067e-01 -3.81925464e-01 6.74785256e-01 6.90131783e-02 -1.51693791e-01 -6.04373455e-01 3.82029682e-01 -3.85619193e-01 3.50926757e-01 -6.30437315e-01 4.17018652e-01 -1.13686502e+00 2.72053778e-01 -1.18642308e-01 -7.13721991e-01 -3.05940986e-01 -2.59399027e-01 8.85531366e-01 -1.95827201e-01 1.60790846e-01 8.66574794e-02 2.60191858e-01 -5.89122713e-01 6.30772710e-01 5.28593548e-02 -3.08549404e-01 6.41371548e-01 2.01850496e-02 -2.52983421e-01 -1.82472438e-01 -6.49746180e-01 3.48242253e-01 -1.25494853e-01 1.17157662e+00 4.17546034e-01 -1.50673199e+00 -4.00091052e-01 2.46859714e-01 4.44394469e-01 -5.59553444e-01 5.88911235e-01 1.05577278e+00 4.76266980e-01 7.86993802e-01 1.12014348e-02 -2.39877701e-01 -9.03915048e-01 7.39705443e-01 -1.13555416e-01 -4.40154999e-01 -4.57581282e-01 6.23922884e-01 1.56308651e-01 -6.85217321e-01 1.56668484e-01 -1.11841559e-01 -6.25287771e-01 3.94966573e-01 4.03109819e-01 8.88787583e-02 1.69398412e-01 -4.50178087e-01 -2.03935340e-01 6.16862535e-01 -3.50153565e-01 2.34257221e-01 1.45963836e+00 -5.42185664e-01 1.68348223e-01 5.37962496e-01 1.05968916e+00 4.40292090e-01 -6.05436981e-01 -8.55118275e-01 -1.51921958e-02 -6.49257958e-01 2.32924715e-01 -1.02091110e+00 -1.54782057e+00 5.84553719e-01 3.50425392e-01 -6.47092052e-03 1.10368621e+00 -1.83664471e-01 1.04329026e+00 5.02598882e-01 6.53325796e-01 -1.31638360e+00 4.32917029e-02 3.49470437e-01 6.08270943e-01 -1.05691981e+00 -4.52749208e-02 -5.86254835e-01 -7.63373673e-01 8.97721171e-01 5.87381661e-01 1.45938426e-01 8.50780427e-01 -6.82705641e-02 -4.21807691e-02 1.92191992e-02 -9.72081184e-01 -1.28311470e-01 7.84291089e-01 2.90342599e-01 6.28719866e-01 2.46788874e-01 -4.06052977e-01 1.63922942e+00 -2.18436152e-01 1.10728651e-01 -7.03961402e-02 4.14456218e-01 -1.79005399e-01 -1.46958363e+00 1.28790960e-01 7.91595459e-01 -2.47737542e-01 -3.18900019e-01 2.51338873e-02 2.41259992e-01 1.58817038e-01 1.18407357e+00 -4.33303535e-01 -1.13918316e+00 3.30587953e-01 -1.26602829e-01 1.13151282e-01 -5.32319367e-01 -6.00074649e-01 3.14068824e-01 3.45450610e-01 -9.15942430e-01 -3.22868437e-01 -5.77664256e-01 -9.44417238e-01 -1.67977348e-01 -6.03281975e-01 3.43039006e-01 6.95082724e-01 9.45049405e-01 5.29892862e-01 5.94207227e-01 1.04781294e+00 -4.17556196e-01 -7.44077742e-01 -7.44591117e-01 -5.65391004e-01 4.26905632e-01 -2.92527825e-01 -8.68812203e-01 -1.13600671e-01 -1.47682920e-01]
[10.176714897155762, 5.529623031616211]
bcd64d7d-bd27-4c28-9d1c-7b24c133452e
few-shot-representation-learning-for-out-of
1907.00505
null
https://arxiv.org/abs/1907.00505v1
https://arxiv.org/pdf/1907.00505v1.pdf
Few-Shot Representation Learning for Out-Of-Vocabulary Words
Existing approaches for learning word embeddings often assume there are sufficient occurrences for each word in the corpus, such that the representation of words can be accurately estimated from their contexts. However, in real-world scenarios, out-of-vocabulary (a.k.a. OOV) words that do not appear in training corpus emerge frequently. It is challenging to learn accurate representations of these words with only a few observations. In this paper, we formulate the learning of OOV embeddings as a few-shot regression problem, and address it by training a representation function to predict the oracle embedding vector (defined as embedding trained with abundant observations) based on limited observations. Specifically, we propose a novel hierarchical attention-based architecture to serve as the neural regression function, with which the context information of a word is encoded and aggregated from K observations. Furthermore, our approach can leverage Model-Agnostic Meta-Learning (MAML) for adapting the learned model to the new corpus fast and robustly. Experiments show that the proposed approach significantly outperforms existing methods in constructing accurate embeddings for OOV words, and improves downstream tasks where these embeddings are utilized.
['Kai-Wei Chang', 'Ziniu Hu', 'Ting Chen', 'Yizhou Sun']
2019-07-01
few-shot-representation-learning-for-out-of-1
https://aclanthology.org/P19-1402
https://aclanthology.org/P19-1402.pdf
acl-2019-7
['learning-word-embeddings']
['methodology']
[-4.10813978e-03 -1.44233434e-02 -5.23514271e-01 -4.86700326e-01 -8.71064961e-01 -3.85773003e-01 6.08051300e-01 4.49684471e-01 -6.12730145e-01 3.50541711e-01 4.55124378e-01 -2.35622510e-01 2.09521964e-01 -6.46010280e-01 -6.69339895e-01 -4.88166362e-01 1.67284027e-01 3.74068618e-01 -8.98385197e-02 -1.86832145e-01 1.64607719e-01 -1.83008593e-02 -1.42863035e+00 9.60281938e-02 8.00360560e-01 8.26578617e-01 6.91272378e-01 6.66828752e-01 -7.11935222e-01 5.05090594e-01 -5.87785125e-01 -2.81744301e-01 -8.60143453e-02 -6.26444519e-02 -6.24426544e-01 -2.72055715e-02 1.87584549e-01 -4.56350297e-01 -5.97505152e-01 6.99563920e-01 2.84374893e-01 5.90340555e-01 8.92638206e-01 -8.90300930e-01 -1.54608583e+00 6.58403635e-01 -1.87685475e-01 4.25023258e-01 -2.64244489e-02 1.54603422e-01 1.53207386e+00 -1.42935109e+00 3.24355423e-01 1.15398180e+00 2.25832134e-01 6.99443877e-01 -1.12293017e+00 -6.26939714e-01 7.67750919e-01 3.60579252e-01 -1.41702878e+00 -5.32241583e-01 6.27775848e-01 -3.36730093e-01 1.44851673e+00 -1.23988248e-01 5.72358370e-01 1.44495463e+00 4.07908224e-02 8.92468810e-01 3.02745908e-01 -6.03963912e-01 3.74176115e-01 2.42513120e-01 5.50897419e-01 2.83793300e-01 3.48788559e-01 -7.15625510e-02 -4.16765451e-01 -1.97186276e-01 2.27887303e-01 5.43782175e-01 -1.47914618e-01 -1.61915287e-01 -8.39064360e-01 1.17168903e+00 4.09635782e-01 4.11863983e-01 -4.25497681e-01 4.53945488e-01 4.67328191e-01 2.35120833e-01 7.04665482e-01 4.48806167e-01 -7.23667622e-01 -2.08107457e-01 -6.19822323e-01 1.47604086e-02 4.56520349e-01 1.13284564e+00 1.05151355e+00 2.14461952e-01 -1.04515597e-01 1.15559256e+00 5.07711887e-01 3.19543004e-01 1.26167715e+00 -2.14351788e-01 4.61169451e-01 4.88975257e-01 2.95287557e-02 -7.05902517e-01 -2.01584790e-02 -1.80036575e-01 -4.53290701e-01 -4.75446343e-01 -1.56889945e-01 7.56367967e-02 -1.18691111e+00 1.75887775e+00 3.27565074e-01 7.38561511e-01 3.48730326e-01 4.46212023e-01 7.03909516e-01 1.28156388e+00 2.34658957e-01 -1.57084331e-01 1.29671013e+00 -1.14684296e+00 -9.03686404e-01 -8.78434420e-01 9.31460857e-01 -4.10315990e-01 1.42463136e+00 1.06965065e-01 -5.39783597e-01 -5.88778198e-01 -1.05866909e+00 -3.74550492e-01 -7.46064842e-01 -3.51991802e-02 3.41527492e-01 2.89619714e-01 -6.76198363e-01 2.46736720e-01 -6.79268003e-01 -3.00669163e-01 3.34519356e-01 1.48340091e-01 -2.46605963e-01 -4.51935679e-01 -1.37886453e+00 8.41881037e-01 7.36102283e-01 4.13923673e-02 -1.08296895e+00 -6.75614178e-01 -1.29048693e+00 1.41088292e-01 2.92235404e-01 -2.46557653e-01 1.23336172e+00 -7.98330069e-01 -1.19631422e+00 4.05837744e-01 -5.90555668e-01 -5.03149509e-01 -4.38356847e-01 -4.57143813e-01 -5.22249520e-01 -2.23298386e-01 5.76734133e-02 2.90327847e-01 1.12541115e+00 -1.17627633e+00 -7.00823069e-01 -2.38217846e-01 2.39044890e-01 -1.09102353e-02 -1.08658779e+00 -3.69706273e-01 -4.05217409e-01 -6.50893569e-01 -2.92139143e-01 -5.58200896e-01 -2.21557856e-01 -1.84491515e-01 -1.77919008e-02 -7.61666417e-01 7.09356368e-01 -5.89830995e-01 1.75182331e+00 -2.34310436e+00 1.29212681e-02 -1.22067206e-01 2.12658718e-01 4.02821094e-01 -5.91288269e-01 6.93045080e-01 1.41147971e-01 1.15685411e-01 -4.09938991e-02 -6.70200109e-01 8.79622325e-02 6.93051517e-01 -8.17601323e-01 1.53493360e-01 3.73680979e-01 1.03858483e+00 -1.22667956e+00 -1.68751404e-01 9.83240083e-02 5.08910477e-01 -6.12464488e-01 6.52018368e-01 -3.40475261e-01 -1.62966534e-01 -5.37843287e-01 3.63324404e-01 2.61584908e-01 -3.62227261e-01 1.94147885e-01 1.79977641e-02 2.03987181e-01 3.28308225e-01 -8.66594672e-01 1.72456312e+00 -1.28516829e+00 5.21776617e-01 -4.54901546e-01 -1.28134799e+00 1.05973101e+00 4.82010275e-01 -2.44833082e-02 -4.27941263e-01 3.27190422e-02 1.89923570e-01 -1.41507149e-01 -6.70917511e-01 6.49315298e-01 -2.02243894e-01 -2.09224239e-01 4.79939759e-01 5.66525102e-01 2.22501904e-01 1.48483797e-03 2.00128749e-01 1.05680037e+00 -2.97213703e-01 7.42499590e-01 1.75082356e-01 3.66499126e-01 -2.21202895e-01 5.23871660e-01 6.91051424e-01 -1.23848870e-01 3.28950226e-01 1.77220732e-01 -5.19054055e-01 -1.08948839e+00 -8.15783560e-01 -7.85953924e-02 1.52340436e+00 1.24177799e-01 -7.23180413e-01 -3.63461196e-01 -6.34993196e-01 1.57644734e-01 1.15773654e+00 -9.60205615e-01 -6.42285049e-01 -4.79399115e-01 -4.79184806e-01 -1.71470251e-02 7.80793726e-01 -4.65161711e-01 -1.07627714e+00 -1.78903952e-01 6.42612278e-01 5.61235994e-02 -1.09882796e+00 -7.39904583e-01 4.07320470e-01 -8.08010817e-01 -7.65580952e-01 -4.43606347e-01 -1.03335297e+00 6.97422028e-01 6.35003448e-01 1.00331748e+00 1.84041142e-01 -2.07436308e-01 3.73878986e-01 -8.08883667e-01 -3.97395343e-01 -2.05223307e-01 2.08584234e-01 2.83759326e-01 2.39857391e-01 1.04618812e+00 -4.89071965e-01 -3.96006823e-01 -2.09703490e-01 -1.07900965e+00 -5.06090820e-01 5.73326588e-01 1.25474012e+00 7.34220564e-01 -2.73976982e-01 9.03856337e-01 -1.01721501e+00 7.39263296e-01 -9.79617715e-01 -2.61795104e-01 1.77542061e-01 -7.34711826e-01 2.60868818e-01 1.04374397e+00 -9.56508160e-01 -6.87977970e-01 -1.66852847e-01 -4.99752685e-02 -7.80230045e-01 6.52197525e-02 5.93289554e-01 -1.24375723e-01 5.26966989e-01 4.21622157e-01 5.20271420e-01 -3.61587167e-01 -6.95393264e-01 8.13614964e-01 1.08923602e+00 5.98767921e-02 -5.68626940e-01 8.29551101e-01 2.04690158e-01 -9.41668451e-01 -1.03029943e+00 -1.22532642e+00 -9.04181123e-01 -6.10516608e-01 3.68251592e-01 8.07040989e-01 -1.14331162e+00 7.16698989e-02 -1.72845662e-01 -1.26789832e+00 -2.05474824e-01 -4.19367820e-01 4.91163552e-01 -2.03263730e-01 6.85494244e-02 -2.00365469e-01 -1.07276750e+00 -2.83018857e-01 -1.07904553e+00 1.12341702e+00 8.99428204e-02 -2.82079041e-01 -1.33289218e+00 3.90296727e-01 2.11669859e-02 4.07375216e-01 -3.38542342e-01 1.09046972e+00 -1.18000400e+00 -2.64110237e-01 -6.43263817e-01 1.02297157e-01 5.78812838e-01 3.87092531e-01 -2.51940191e-01 -1.19771338e+00 -4.57391858e-01 -1.20851509e-01 -5.29208004e-01 9.44182396e-01 2.20066234e-01 1.44050789e+00 -6.07043147e-01 -4.17184114e-01 4.38207179e-01 1.45856738e+00 8.52175280e-02 1.81811795e-01 6.08773828e-02 7.76112497e-01 4.35515225e-01 6.60135329e-01 5.07895231e-01 5.27715385e-01 5.07765114e-01 4.36318725e-01 4.13066387e-01 7.16239959e-02 -6.06352270e-01 5.69353998e-01 1.58214521e+00 4.76030976e-01 -3.78572822e-01 -7.02647924e-01 1.16623437e+00 -1.67459130e+00 -5.70244312e-01 4.00515169e-01 1.97198749e+00 7.66367614e-01 1.11447655e-01 -3.71163130e-01 -1.84590399e-01 5.53251982e-01 5.06918848e-01 -7.75802791e-01 -5.79675376e-01 3.49297762e-01 4.21855658e-01 1.19099133e-01 6.17159843e-01 -9.18862700e-01 1.12510228e+00 5.73426628e+00 8.41827631e-01 -9.54011738e-01 5.89946985e-01 3.47750694e-01 -2.74747938e-01 -6.33524060e-01 9.59226862e-02 -1.10428512e+00 6.04779541e-01 1.30319941e+00 -4.61539596e-01 3.04261118e-01 1.13481390e+00 -8.85989610e-03 6.10815227e-01 -1.26856220e+00 9.74407136e-01 5.01488328e-01 -1.13292921e+00 4.22600061e-01 2.69241128e-02 6.15524650e-01 1.39139891e-01 1.15504764e-01 9.94582593e-01 2.88292468e-01 -1.09016955e+00 3.94544154e-01 2.77670801e-01 7.80047357e-01 -7.16052234e-01 6.85618758e-01 5.53361833e-01 -1.37508380e+00 -3.98691207e-01 -1.00772226e+00 1.87015776e-02 9.89775509e-02 4.92831379e-01 -8.49468827e-01 9.45910066e-02 3.08942974e-01 8.80160213e-01 -3.15780550e-01 4.15082842e-01 -4.09645140e-01 7.55194306e-01 3.18344012e-02 -5.14592886e-01 5.58127105e-01 1.59786027e-02 1.09870441e-01 1.13512278e+00 3.70445222e-01 2.22904935e-01 2.17024520e-01 6.69340432e-01 -4.39735532e-01 4.40717638e-01 -8.13189864e-01 -4.27498162e-01 9.80019569e-01 1.19943023e+00 3.52592692e-02 -4.95099425e-01 -8.87016058e-01 1.07092130e+00 1.02119529e+00 4.64617580e-01 -5.03329754e-01 -4.51186538e-01 1.15327477e+00 -2.47783676e-01 7.34866440e-01 -1.89183012e-01 2.05318898e-01 -1.40069473e+00 9.09837615e-03 -5.63030720e-01 3.85626405e-01 -5.08296192e-01 -1.70342565e+00 6.85831010e-01 -2.17655361e-01 -1.14102757e+00 -3.75521451e-01 -8.02226424e-01 -8.74112487e-01 7.88207769e-01 -1.91915917e+00 -1.09062493e+00 1.77882854e-02 1.14044286e-01 1.12221026e+00 -3.80511671e-01 1.13076305e+00 9.01577901e-03 -6.82278633e-01 7.70491242e-01 5.50294220e-01 1.77475363e-01 6.16544902e-01 -1.21293712e+00 7.24286914e-01 6.97404027e-01 6.03533149e-01 8.31302583e-01 3.82763714e-01 -3.69972974e-01 -1.64521384e+00 -1.47944975e+00 1.11723757e+00 -6.04898572e-01 9.64702547e-01 -6.74737096e-01 -1.31057179e+00 9.96177435e-01 4.61119115e-02 5.96944451e-01 1.12240207e+00 3.35069120e-01 -6.04684055e-01 -3.83674838e-02 -5.65060019e-01 4.66510713e-01 8.10437977e-01 -8.57442558e-01 -1.09358931e+00 3.53661418e-01 1.43057239e+00 8.64175186e-02 -7.52338231e-01 -2.21208438e-01 3.05150241e-01 -1.17304139e-02 9.63713109e-01 -1.19020104e+00 3.87076497e-01 2.44813599e-02 -5.38822889e-01 -1.67340350e+00 -3.42963398e-01 -1.80551037e-01 -9.39808249e-01 1.17932916e+00 4.61942375e-01 -6.58379376e-01 3.46650451e-01 3.98581266e-01 -2.00531408e-01 -1.03030622e+00 -9.77165580e-01 -8.67767096e-01 3.03696275e-01 -6.94891572e-01 7.79325724e-01 9.97044206e-01 8.59991387e-02 5.50362825e-01 -4.57327515e-01 3.85567188e-01 2.65414000e-01 5.64352423e-02 7.28244424e-01 -9.72708166e-01 -3.47220123e-01 5.56362718e-02 -5.35360336e-01 -1.47829473e+00 5.97542763e-01 -1.16689909e+00 2.54744560e-01 -1.49057400e+00 1.20954044e-01 -2.99131155e-01 -7.69190133e-01 3.67801994e-01 -7.59323061e-01 -5.88147007e-02 -6.10018186e-02 1.76893905e-01 -5.51653087e-01 1.02573931e+00 7.44337380e-01 -3.32453161e-01 5.14795519e-02 -3.55385303e-01 -8.16410959e-01 5.28700352e-01 5.54276168e-01 -6.63959861e-01 -5.79833567e-01 -8.74865294e-01 8.03951770e-02 -4.60064948e-01 -3.43180969e-02 -3.43348950e-01 1.26941010e-01 -1.99909970e-01 1.64176255e-01 -3.25921625e-01 5.98619401e-01 -8.18302214e-01 -5.60039639e-01 2.02075288e-01 -5.79101384e-01 -1.25829987e-02 -1.31705284e-01 1.02386487e+00 -2.80428261e-01 -6.64603710e-01 4.36281860e-01 2.53195558e-02 -1.11139345e+00 7.41529226e-01 -1.71872079e-01 4.47583377e-01 9.66395617e-01 1.14379331e-01 -3.93709727e-02 -2.92288989e-01 -5.67768812e-01 3.26312482e-01 2.16986030e-01 8.89427066e-01 9.57878649e-01 -1.71010840e+00 -6.71573460e-01 3.68269324e-01 1.00424314e+00 7.98579380e-02 1.89750418e-01 1.20450027e-01 5.76365963e-02 3.34934771e-01 4.82340604e-01 -2.75864333e-01 -6.67602956e-01 1.05148995e+00 1.99202579e-02 -2.27351844e-01 -7.79814959e-01 8.62372160e-01 5.06164253e-01 -6.09310508e-01 2.35814154e-01 -4.41813082e-01 -3.39826614e-01 1.84002012e-01 1.04131913e+00 -1.35050207e-01 -7.32754990e-02 -5.86980164e-01 -1.63374051e-01 3.68362129e-01 -5.89752972e-01 1.73556179e-01 1.55016208e+00 -2.43776008e-01 3.38718355e-01 1.12439048e+00 1.68325007e+00 -1.96278080e-01 -1.14744341e+00 -8.30090702e-01 2.77896244e-02 -5.85430503e-01 1.24494195e-01 -2.23441079e-01 -7.66684592e-01 1.31935596e+00 3.84108633e-01 1.03713505e-01 5.51168263e-01 2.19299629e-01 1.29743981e+00 5.91407418e-01 4.13032919e-02 -1.18292785e+00 3.84658694e-01 6.52775705e-01 5.72415292e-01 -1.24795961e+00 -3.39052290e-01 1.40936315e-01 -6.07964754e-01 1.17602885e+00 7.39706397e-01 -3.83616775e-01 8.46192479e-01 -5.62934242e-02 5.73123395e-02 -4.54936922e-02 -1.33895850e+00 -1.82615563e-01 3.68283778e-01 7.04937220e-01 3.79482329e-01 3.51840816e-02 1.77049090e-03 1.03400290e+00 4.68942486e-02 -5.06477833e-01 4.67891157e-01 8.37128997e-01 -8.84331703e-01 -1.05539715e+00 7.49288350e-02 7.63543785e-01 -5.86759709e-02 -5.34434497e-01 1.22796837e-02 4.43166375e-01 -3.81250940e-02 7.74976671e-01 3.60351026e-01 -2.62968808e-01 3.66359800e-01 4.34888601e-01 -1.79365426e-02 -1.30058944e+00 -4.69474383e-02 -3.58081818e-01 -2.77243912e-01 -4.33723450e-01 6.33559227e-02 -3.06407660e-01 -1.05814767e+00 1.69180229e-01 -5.78795195e-01 2.54220337e-01 5.80639899e-01 1.23731649e+00 3.25087786e-01 6.02172971e-01 8.87725174e-01 -7.37426579e-01 -9.36467111e-01 -1.13102221e+00 -6.20320916e-01 5.08872032e-01 6.33626640e-01 -7.72806048e-01 -5.70122361e-01 1.04876064e-01]
[10.530954360961914, 8.644791603088379]
10916bc0-f6cf-403d-a642-3502629b2f1a
deep-attention-recognition-for-attack
2303.12947
null
https://arxiv.org/abs/2303.12947v1
https://arxiv.org/pdf/2303.12947v1.pdf
Deep Attention Recognition for Attack Identification in 5G UAV scenarios: Novel Architecture and End-to-End Evaluation
Despite the robust security features inherent in the 5G framework, attackers will still discover ways to disrupt 5G unmanned aerial vehicle (UAV) operations and decrease UAV control communication performance in Air-to-Ground (A2G) links. Operating under the assumption that the 5G UAV communications infrastructure will never be entirely secure, we propose Deep Attention Recognition (DAtR) as a solution to identify attacks based on a small deep network embedded in authenticated UAVs. Our proposed solution uses two observable parameters: the Signal-to-Interference-plus-Noise Ratio (SINR) and the Reference Signal Received Power (RSSI) to recognize attacks under Line-of-Sight (LoS), Non-Line-of-Sight (NLoS), and a probabilistic combination of the two conditions. In the tested scenarios, a number of attackers are located in random positions, while their power is varied in each simulation. Moreover, terrestrial users are included in the network to impose additional complexity on attack detection. To improve the systems overall performance in the attack scenarios, we propose complementing the deep network decision with two mechanisms based on data manipulation and majority voting techniques. We compare several performance parameters in our proposed Deep Network. For example, the impact of Long Short-Term-Memory (LSTM) and Attention layers in terms of their overall accuracy, the window size effect, and test the accuracy when only partial data is available in the training process. Finally, we benchmark our deep network with six widely used classifiers regarding classification accuracy. Our algorithms accuracy exceeds 4% compared with the eXtreme Gradient Boosting (XGB) classifier in LoS condition and around 3% in the short distance NLoS condition. Considering the proposed deep network, all other classifiers present lower accuracy than XGB.
['Rui Dinis', 'Sandra Lagen', 'Biljana Bojovic', 'Katerina Koutlia', 'Luis Miguel Campos', 'Pedro Sebastiao', 'Hamed Farkhari', 'Joseanne Viana']
2023-03-03
null
null
null
null
['deep-attention', 'deep-attention']
['computer-vision', 'natural-language-processing']
[-1.53824657e-01 -1.44822858e-02 2.38592215e-02 2.81127036e-01 6.04459569e-02 -8.99619043e-01 4.54762638e-01 -1.89789250e-01 -4.72644418e-01 7.59069860e-01 -4.82480526e-01 -1.07451129e+00 -5.14253020e-01 -1.17729485e+00 -6.56834900e-01 -9.22538161e-01 -6.27064466e-01 -2.85307378e-01 5.75031191e-02 -3.69922727e-01 -9.56681594e-02 1.04015040e+00 -1.23036528e+00 -5.47671206e-02 5.00936389e-01 1.50153422e+00 -2.83273011e-01 9.40849841e-01 7.23089337e-01 5.04350603e-01 -1.45644450e+00 -2.85841495e-01 7.68366277e-01 5.19515276e-02 -3.57509166e-01 -5.74369311e-01 2.29035363e-01 -5.90958834e-01 -6.35365784e-01 1.14524448e+00 8.64013076e-01 -1.54540211e-01 3.66359890e-01 -1.58016014e+00 -5.61683066e-02 5.33565462e-01 -3.64044964e-01 2.95928061e-01 2.89235741e-01 4.00601089e-01 4.06016886e-01 -9.51025933e-02 5.48667461e-03 9.05557096e-01 8.32301140e-01 1.93561837e-01 -4.78086025e-01 -9.47894096e-01 3.19173843e-01 1.68433532e-01 -1.30583119e+00 -1.41491100e-01 4.29263890e-01 -1.58233330e-01 6.85547054e-01 4.62683678e-01 6.46645963e-01 1.40818429e+00 6.67116821e-01 -3.53661925e-02 6.24541223e-01 -3.54405284e-01 2.13378102e-01 -1.13821857e-01 -1.49265021e-01 7.04045177e-01 7.15361118e-01 8.24892879e-01 1.92161396e-01 -5.44858217e-01 3.42079222e-01 -1.20118916e-01 -6.43881023e-01 2.98730373e-01 -9.91427541e-01 5.67009091e-01 7.80317128e-01 2.43406191e-01 -6.43369555e-01 4.39701676e-01 9.83125493e-02 7.78559506e-01 -7.81690925e-02 6.86686218e-01 -5.30547142e-01 2.81630278e-01 -6.77650928e-01 -9.82245877e-02 9.95612800e-01 8.22504997e-01 1.76586732e-01 6.98555171e-01 -1.74372762e-01 -1.41781136e-01 3.65599513e-01 7.83586025e-01 3.30191016e-01 -2.78115422e-01 4.63139623e-01 7.03949407e-02 3.15018505e-01 -1.45515478e+00 -7.38491952e-01 -1.52458501e+00 -9.08958018e-01 4.77629811e-01 5.02647579e-01 -1.07993758e+00 -8.22341621e-01 1.73284280e+00 2.64457494e-01 4.36504036e-01 3.44021857e-01 9.28149819e-01 6.58361793e-01 5.29345572e-01 -1.73359841e-01 -6.72828928e-02 1.00825310e+00 -5.68464100e-01 -3.97174060e-01 -1.69158384e-01 6.87425554e-01 -5.84763825e-01 3.79846692e-01 5.03817081e-01 -4.21130359e-01 -5.08815348e-01 -1.58034301e+00 8.96584153e-01 -7.02085793e-01 4.26080257e-01 3.61020476e-01 1.29297340e+00 -8.33800137e-01 4.30521399e-01 -4.62737709e-01 -3.11200380e-01 2.12884083e-01 7.29942620e-01 -6.78148270e-02 2.32680336e-01 -1.58714592e+00 6.81729078e-01 9.95900333e-02 4.58582252e-01 -1.47277796e+00 -3.54187399e-01 -3.24990273e-01 2.05312476e-01 4.41621929e-01 -7.27488101e-01 6.68698370e-01 -7.81979859e-01 -1.34640098e+00 1.42975047e-01 6.36531651e-01 -9.52412665e-01 4.68262672e-01 -2.13709250e-01 -6.27814651e-01 7.52415732e-02 -5.11398256e-01 1.97580248e-01 8.78352463e-01 -1.00005996e+00 -6.24932826e-01 -4.89197969e-01 8.46633375e-01 -2.55143866e-02 -1.70367971e-01 -1.99062631e-01 5.61796665e-01 -6.50231481e-01 -5.58353886e-02 -1.16697824e+00 -1.68465003e-01 -3.36650014e-01 -7.10381627e-01 3.21996182e-01 1.08583868e+00 -4.40388232e-01 9.78172004e-01 -1.95525253e+00 -3.49023759e-01 3.77665877e-01 -4.40387763e-02 6.24222457e-01 -2.23533720e-01 4.32752162e-01 -1.83802113e-01 2.69276559e-01 3.86412293e-01 1.67675152e-01 -2.41113618e-01 -3.80858511e-01 -7.17442274e-01 6.33453786e-01 -4.33078498e-01 4.19098884e-01 -6.50201261e-01 4.52902049e-01 1.42194629e-01 1.83486089e-01 -2.68633932e-01 2.14182764e-01 7.68569410e-02 4.09866303e-01 -6.23678207e-01 7.52652109e-01 8.62295091e-01 4.17199820e-01 -8.20356794e-03 -4.19178247e-01 -1.51089519e-01 -7.59596378e-02 -9.15679097e-01 9.22465026e-01 -5.16741931e-01 5.43072462e-01 2.00962827e-01 -9.56019104e-01 9.38820541e-01 3.73191774e-01 1.80907086e-01 -1.23543896e-01 8.84974360e-01 2.03383025e-02 5.32984018e-01 -6.86458647e-02 2.57187318e-02 3.69987249e-01 -1.13071397e-01 7.80153051e-02 1.02283508e-01 1.58704475e-01 -3.90271455e-01 -7.43323714e-02 1.33568549e+00 -5.15248358e-01 2.92175621e-01 -1.39637873e-01 3.87340546e-01 -2.52404541e-01 2.98279583e-01 1.19134593e+00 -3.28613251e-01 3.19840275e-02 7.25045025e-01 -5.21878779e-01 -1.46930605e-01 -7.14783132e-01 2.45066211e-01 5.92909157e-01 3.93019319e-01 -1.99996352e-01 -7.14649081e-01 -9.07180309e-01 -8.41388330e-02 1.00542927e+00 -4.38840389e-01 -6.44259214e-01 1.08876184e-01 -1.22956789e+00 1.37667835e+00 -6.11782074e-02 9.76222932e-01 -3.79740834e-01 -7.84587979e-01 -6.47014007e-02 3.45930576e-01 -1.36653256e+00 2.69031048e-01 4.49779093e-01 -2.44345322e-01 -1.31565547e+00 -2.68737108e-01 -2.28106201e-01 3.34756494e-01 4.55256104e-01 4.88321304e-01 1.64178997e-01 1.45822708e-02 4.74607706e-01 -4.33667511e-01 -6.87413573e-01 -1.91467971e-01 9.87776443e-02 3.88961494e-01 4.11406644e-02 1.24803213e-02 -4.85965669e-01 -4.63500977e-01 6.21891499e-01 -3.57228905e-01 -8.19955528e-01 5.27545035e-01 6.72885239e-01 -1.86857656e-01 6.33009434e-01 4.58109140e-01 -1.43746763e-01 6.28946781e-01 -6.66155815e-01 -9.70572770e-01 3.07147294e-01 -3.38274807e-01 -6.39631212e-01 9.24255371e-01 -4.69205290e-01 -2.50701964e-01 -3.62028062e-01 -3.31844687e-02 -5.19496143e-01 -3.57338905e-01 4.84912872e-01 -5.80865026e-01 -8.65211189e-01 6.93036914e-01 -9.68969464e-02 -3.70004088e-01 1.83899090e-01 -5.83113544e-02 6.90490961e-01 8.04911777e-02 -7.95404315e-02 1.10835111e+00 2.28807718e-01 4.93665218e-01 -1.07159412e+00 -5.81709087e-01 3.97143573e-01 -1.71337485e-01 -3.75984520e-01 5.67627966e-01 -9.74271357e-01 -1.04018986e+00 7.12037265e-01 -1.30246425e+00 -1.15428180e-01 1.96252346e-01 8.85690868e-01 1.50563151e-01 2.49193594e-01 -3.45052809e-01 -9.35521901e-01 -2.77878076e-01 -1.28973973e+00 5.12949169e-01 1.34822160e-01 3.15478325e-01 -8.04841518e-01 -5.50934494e-01 -5.80848083e-02 5.55743814e-01 7.26107538e-01 6.59188807e-01 -9.04579639e-01 -5.39522111e-01 -7.87855625e-01 2.73263687e-03 3.76598209e-01 5.55293076e-02 -9.18653980e-02 -1.04067266e+00 -7.51828313e-01 2.56074220e-01 1.14479981e-01 5.27701259e-01 6.15769565e-01 1.17054653e+00 -6.00894272e-01 -3.63734990e-01 1.01779163e+00 1.12661719e+00 7.13106215e-01 5.95746160e-01 2.56636828e-01 4.03916657e-01 2.25191221e-01 5.93454719e-01 4.95914072e-01 -2.67134964e-01 4.81369823e-01 1.51498389e+00 -1.64120942e-01 2.45471478e-01 -5.52746141e-03 5.08273542e-01 -2.56696671e-01 -1.05423868e-01 -1.33309269e+00 -5.98684847e-01 -1.57380372e-01 -1.17353284e+00 -9.15927172e-01 -4.49600117e-03 2.54902554e+00 -2.37454131e-01 4.29018170e-01 -1.44201994e-01 2.64350802e-01 7.82612622e-01 3.76968771e-01 -4.28586692e-01 -3.44871461e-01 -2.97676772e-01 -1.96026415e-01 1.34167254e+00 5.55596471e-01 -1.51528823e+00 7.88338482e-01 5.25971556e+00 6.94297314e-01 -1.67414200e+00 -1.31891370e-01 5.58297515e-01 -1.65310368e-01 3.14526290e-01 -1.72400728e-01 -7.41239130e-01 5.30472755e-01 1.02633011e+00 2.83657730e-01 4.00009811e-01 7.51830518e-01 9.44852084e-02 6.93667382e-02 -5.01801074e-01 9.46402311e-01 6.78768009e-02 -9.46412444e-01 1.03942871e-01 4.69147056e-01 2.69455075e-01 1.60171062e-01 4.61963505e-01 1.13362707e-01 2.72882074e-01 -9.13792014e-01 5.78703225e-01 2.66418248e-01 8.42638373e-01 -9.84734297e-01 1.39571953e+00 4.13016349e-01 -9.05934393e-01 -7.41167665e-01 -3.80119354e-01 -2.09574297e-01 -1.73285708e-01 5.47991812e-01 -9.22145903e-01 1.06817281e+00 5.38811624e-01 9.01244581e-02 -3.90508950e-01 1.07733798e+00 -4.41817760e-01 8.02217901e-01 -5.99554181e-01 -1.03286058e-01 7.22699583e-01 2.16485634e-01 1.15113831e+00 6.65015876e-01 8.25525463e-01 1.42842591e-01 1.42898336e-01 4.17987466e-01 9.49381888e-02 -4.90632296e-01 -9.27672148e-01 8.89047757e-02 7.04410255e-01 1.08606613e+00 -6.26723111e-01 1.78907841e-01 -1.27614617e-01 6.17163658e-01 -5.80969393e-01 6.23898864e-01 -1.03043771e+00 -8.57660830e-01 6.45264328e-01 1.01346418e-01 1.81356773e-01 -3.53482664e-01 1.26520619e-01 -8.54722381e-01 -3.50318730e-01 -9.89856899e-01 3.12622875e-01 -5.65017700e-01 -8.11944842e-01 9.94761825e-01 -1.46812350e-01 -1.53608882e+00 -2.23468728e-02 -8.01926911e-01 -7.16895759e-01 5.69709182e-01 -1.17911172e+00 -1.25307965e+00 -6.50687158e-01 5.30097246e-01 -1.33310750e-01 -8.37767422e-01 8.42046678e-01 2.87671536e-01 -6.50155544e-01 8.57267082e-01 -1.71230882e-01 2.38764659e-01 4.68186796e-01 -6.37119412e-01 2.64433920e-01 1.20826781e+00 -1.40023027e-02 4.14277732e-01 7.34866381e-01 -6.85724676e-01 -1.41052926e+00 -1.17514777e+00 -2.36942340e-02 1.04293954e-02 4.44789767e-01 -2.90512472e-01 -1.63681656e-02 6.29222929e-01 1.05507284e-01 1.15184620e-01 5.20555377e-01 -2.00495273e-01 7.60320872e-02 -3.16909492e-01 -1.36294603e+00 6.42646492e-01 7.68177390e-01 -1.11347094e-01 -1.06818020e-01 2.49240413e-01 6.78797424e-01 -3.35470974e-01 -4.47184086e-01 8.70268762e-01 6.05746686e-01 -1.06372631e+00 9.35550034e-01 -7.30386674e-01 -6.59268975e-01 -4.99533445e-01 -5.77375054e-01 -1.63378036e+00 -1.94494933e-01 -9.57579553e-01 1.04530931e-01 7.13610411e-01 4.83712405e-01 -1.10357893e+00 7.74576306e-01 -1.73318878e-01 -2.08585724e-01 -5.86408079e-01 -1.14768100e+00 -9.87800658e-01 -2.47305349e-01 -4.08701330e-01 9.32926655e-01 8.18280220e-01 -4.79026347e-01 4.24654931e-02 -6.37325108e-01 1.22968280e+00 4.45978284e-01 -3.12774181e-01 9.24142897e-01 -1.16814291e+00 -1.58309445e-01 -7.41614476e-02 -8.21185172e-01 -9.48626697e-01 -4.82051931e-02 -5.74455440e-01 -4.82909560e-01 -1.02324545e+00 -1.02077162e+00 -4.69397336e-01 -4.85116661e-01 4.34236556e-01 5.04201412e-01 -5.71240373e-02 3.43693905e-02 -3.32996577e-01 -1.51376039e-01 4.71319556e-01 7.08333135e-01 -3.86675447e-01 -5.73696271e-02 7.42402673e-01 -4.74618852e-01 9.70321000e-01 8.63254130e-01 -5.17753780e-01 -5.11834025e-01 -4.19101417e-01 3.48796137e-02 3.78576159e-01 5.80610275e-01 -1.72440970e+00 3.25789243e-01 1.05817318e-01 7.25785434e-01 -4.38051552e-01 3.85692596e-01 -1.29277694e+00 4.80355546e-02 9.29802239e-01 2.24665835e-01 8.21972042e-02 4.45756465e-01 6.77504182e-01 9.75261722e-03 -4.66214493e-02 6.89057827e-01 1.66330338e-01 -2.15754241e-01 3.88771415e-01 -7.66400695e-01 -5.74995339e-01 1.28976154e+00 -2.39412129e-01 -6.57745600e-01 -7.85669327e-01 -6.18341744e-01 8.34864005e-03 1.20204903e-01 2.22480029e-01 5.97194195e-01 -8.43270600e-01 -3.77074182e-01 5.42157114e-01 -2.64890879e-01 -6.41492784e-01 1.61014199e-01 7.79844999e-01 -5.70680618e-01 5.87338209e-01 -2.72602946e-01 -1.46717191e-01 -1.21251464e+00 5.30803621e-01 1.04276943e+00 5.28497547e-02 2.30140150e-01 8.60497296e-01 5.34237884e-02 -3.06341380e-01 5.22540808e-01 -3.69717777e-01 -3.21206957e-01 -2.09716782e-02 4.30717111e-01 6.13000572e-01 2.14038372e-01 -2.01294631e-01 -5.39738655e-01 4.77109045e-01 2.96225101e-01 2.47088611e-01 6.92041576e-01 -9.29937698e-03 2.36621678e-01 -2.89202690e-01 7.25610375e-01 3.36495906e-01 -6.51745260e-01 4.98750061e-01 -3.76140594e-01 -3.00036132e-01 6.24783397e-01 -1.07878304e+00 -1.19388616e+00 7.03321695e-01 9.76942122e-01 6.73404992e-01 1.10850537e+00 -7.00211763e-01 6.18515849e-01 8.78409863e-01 6.84994936e-01 -3.74310672e-01 -2.90101945e-01 5.47719419e-01 6.49277449e-01 -8.26811433e-01 6.93448037e-02 -3.62116456e-01 -8.57423693e-02 1.30713129e+00 6.00629449e-01 -2.60134399e-01 1.05659294e+00 3.30123365e-01 5.54559119e-02 -1.56570867e-01 -6.63548887e-01 -1.86828852e-01 3.16339768e-02 8.45657110e-01 -2.92720169e-01 -7.29633346e-02 1.72799870e-01 8.16435754e-01 -4.21913385e-01 -4.30153728e-01 8.08155894e-01 5.27508736e-01 -3.54550689e-01 -5.17688632e-01 -4.56801087e-01 3.43659461e-01 -7.48394549e-01 -4.33950201e-02 -4.48122412e-01 9.05032873e-01 4.34423149e-01 1.55087507e+00 -2.21644983e-01 -1.02441216e+00 3.61723900e-01 -7.43390083e-01 1.30294442e-01 -1.71759367e-01 -5.79603672e-01 -2.74309546e-01 2.80336827e-01 -5.65065503e-01 1.22664511e-01 -4.38671075e-02 -5.70164204e-01 -2.51296550e-01 -7.24696934e-01 4.18092221e-01 7.67553866e-01 7.30776846e-01 4.29446042e-01 7.47822404e-01 9.19967890e-01 -6.54088616e-01 -6.68891013e-01 -9.11714613e-01 -7.79957235e-01 -7.97041893e-01 5.69863558e-01 -8.83735359e-01 -1.10037625e+00 -9.00375843e-01]
[5.457822322845459, 7.896629333496094]
b195ac6a-4626-4ed5-a327-af6083f85522
taming-transformers-for-high-resolution-image
2012.09841
null
https://arxiv.org/abs/2012.09841v3
https://arxiv.org/pdf/2012.09841v3.pdf
Taming Transformers for High-Resolution Image Synthesis
Designed to learn long-range interactions on sequential data, transformers continue to show state-of-the-art results on a wide variety of tasks. In contrast to CNNs, they contain no inductive bias that prioritizes local interactions. This makes them expressive, but also computationally infeasible for long sequences, such as high-resolution images. We demonstrate how combining the effectiveness of the inductive bias of CNNs with the expressivity of transformers enables them to model and thereby synthesize high-resolution images. We show how to (i) use CNNs to learn a context-rich vocabulary of image constituents, and in turn (ii) utilize transformers to efficiently model their composition within high-resolution images. Our approach is readily applied to conditional synthesis tasks, where both non-spatial information, such as object classes, and spatial information, such as segmentations, can control the generated image. In particular, we present the first results on semantically-guided synthesis of megapixel images with transformers and obtain the state of the art among autoregressive models on class-conditional ImageNet. Code and pretrained models can be found at https://github.com/CompVis/taming-transformers .
['Björn Ommer', 'Robin Rombach', 'Patrick Esser']
2020-12-17
null
http://openaccess.thecvf.com//content/CVPR2021/html/Esser_Taming_Transformers_for_High-Resolution_Image_Synthesis_CVPR_2021_paper.html
http://openaccess.thecvf.com//content/CVPR2021/papers/Esser_Taming_Transformers_for_High-Resolution_Image_Synthesis_CVPR_2021_paper.pdf
cvpr-2021-1
['image-outpainting']
['computer-vision']
[ 4.96632695e-01 2.00007841e-01 -8.12187791e-02 -3.71332705e-01 -8.74489069e-01 -5.95895827e-01 9.33546662e-01 -5.07011056e-01 -1.72078490e-01 6.27938211e-01 3.05278093e-01 -2.98445731e-01 1.32203773e-01 -1.10330927e+00 -1.22498417e+00 -8.84034812e-01 2.20923856e-01 4.40038145e-01 4.10001963e-01 -3.13004136e-01 -4.77861278e-02 5.47415316e-01 -1.70239913e+00 6.70928597e-01 6.41746283e-01 8.82696390e-01 4.95090812e-01 8.79037619e-01 3.22537273e-02 1.25052476e+00 -3.56599092e-01 -5.75070083e-02 1.06856585e-01 -2.86666691e-01 -6.86984599e-01 1.93086833e-01 6.52078331e-01 -4.68937397e-01 -4.23704594e-01 8.30655456e-01 2.88046241e-01 6.42046705e-02 5.82419097e-01 -9.30605173e-01 -8.22798550e-01 8.67659748e-01 -2.93662459e-01 -6.60836175e-02 -1.71457127e-01 4.97103363e-01 1.21125555e+00 -9.00265753e-01 7.25779474e-01 1.19400060e+00 3.39643180e-01 6.15213633e-01 -1.42042410e+00 -7.61275411e-01 4.76029575e-01 -1.12178974e-01 -1.29475999e+00 -5.21557808e-01 7.27798522e-01 -4.62439001e-01 1.07609332e+00 2.54356563e-01 7.11458981e-01 1.25193918e+00 -3.96118872e-02 7.63164461e-01 9.45872426e-01 -2.68281788e-01 -2.57038958e-02 -2.13023052e-01 -2.77998805e-01 8.03524852e-01 -3.31218392e-01 3.72397274e-01 -6.15324795e-01 2.07532793e-01 1.16525483e+00 -2.55605489e-01 -8.23522508e-02 -2.13827729e-01 -1.37812519e+00 6.80241764e-01 6.85076714e-01 3.52010995e-01 -3.02208871e-01 8.62236679e-01 3.87980603e-02 -4.28102687e-02 4.56296831e-01 5.02942264e-01 -4.37479436e-01 1.55842006e-01 -7.98740864e-01 2.77042300e-01 3.43395501e-01 1.16381025e+00 7.64336705e-01 3.61764759e-01 -1.22673161e-01 7.96545863e-01 1.10919468e-01 4.26445484e-01 1.49758890e-01 -1.27184701e+00 2.61101276e-01 2.07574680e-01 -4.37747911e-02 -7.11490452e-01 -1.29786164e-01 -4.41101015e-01 -9.71792400e-01 2.55689293e-01 3.16832930e-01 5.94867840e-02 -1.21421480e+00 1.92534816e+00 2.00113207e-02 2.73478329e-01 -8.92195776e-02 8.05011153e-01 9.32937026e-01 8.94620836e-01 9.27272886e-02 8.74002054e-02 1.27860451e+00 -1.04039419e+00 -4.90054011e-01 -2.58036375e-01 4.12907779e-01 -5.86377203e-01 1.16023076e+00 3.91012609e-01 -1.30987740e+00 -5.73554337e-01 -9.58955705e-01 -3.85410935e-01 -3.78360927e-01 1.30320877e-01 6.89335406e-01 2.74580657e-01 -1.33581746e+00 4.38225836e-01 -1.04218280e+00 -4.74258512e-03 7.17516363e-01 4.38156754e-01 -1.25837564e-01 9.34915021e-02 -1.16201949e+00 6.79102838e-01 1.64941907e-01 -2.86462400e-02 -1.46521604e+00 -9.60390687e-01 -8.34075511e-01 2.34126206e-02 3.96081656e-01 -9.13008571e-01 1.36146855e+00 -1.25511444e+00 -1.58939457e+00 7.71659374e-01 -8.33286718e-02 -4.82851386e-01 5.03053069e-01 -2.32499652e-02 -1.03060473e-02 9.85024944e-02 1.42491488e-02 1.20106065e+00 8.61284256e-01 -1.40753734e+00 -5.94402194e-01 2.65245903e-02 4.22733963e-01 1.88903496e-01 -1.55776531e-01 -5.98019026e-02 -5.45082569e-01 -6.27497613e-01 -1.36517987e-01 -9.54306483e-01 -4.89600092e-01 9.42344517e-02 -5.86124420e-01 1.61567167e-03 7.32819796e-01 -3.30016792e-01 6.62975669e-01 -2.14604115e+00 3.41192991e-01 9.24631208e-02 2.70435512e-01 1.50353378e-02 -3.53684843e-01 1.76494420e-01 -2.39037186e-01 3.49935502e-01 -4.17185813e-01 -2.90928751e-01 -1.11209668e-01 3.98789674e-01 -6.13039613e-01 2.94395924e-01 4.99212116e-01 1.19179702e+00 -8.61486137e-01 -5.30031323e-01 4.59902614e-01 7.17618406e-01 -6.93532586e-01 1.08864844e-01 -8.30050945e-01 6.85495257e-01 -5.03405213e-01 3.67219418e-01 4.00932968e-01 -5.88257194e-01 1.04709581e-01 -4.64908272e-01 -2.79928714e-01 4.55231279e-01 -7.16045976e-01 1.55621004e+00 -7.50348866e-01 7.20954001e-01 -6.61648065e-02 -9.41365182e-01 6.00345552e-01 3.04935873e-01 4.07837123e-01 -7.13768542e-01 4.12158556e-02 -1.35993548e-02 -1.56349525e-01 -6.52616695e-02 5.68412662e-01 -9.14203376e-02 -9.71926469e-03 3.86298954e-01 5.94803281e-02 -5.88658690e-01 5.02893515e-02 3.08625609e-01 8.44458997e-01 5.16211271e-01 8.71527344e-02 -3.64710510e-01 2.44485945e-01 -1.53665528e-01 3.53401542e-01 7.72398889e-01 3.16345900e-01 8.95186067e-01 4.86475259e-01 -4.20417756e-01 -1.37193716e+00 -1.12709880e+00 1.23971999e-02 1.15348196e+00 -1.24566697e-01 -3.03690553e-01 -7.49395072e-01 -3.40693265e-01 -4.22980398e-01 7.39632428e-01 -7.03735650e-01 1.02375120e-01 -7.11208344e-01 -5.75181782e-01 6.62159503e-01 7.34831393e-01 5.72798729e-01 -1.02931058e+00 -5.33175290e-01 1.13055393e-01 -3.40263993e-01 -1.43044925e+00 -3.69841009e-01 2.23072678e-01 -8.10005128e-01 -7.79025018e-01 -5.05193770e-01 -6.56951189e-01 6.64494812e-01 -2.73735244e-02 1.34474194e+00 1.40508339e-01 -2.18372330e-01 2.85335600e-01 -6.21781647e-02 -1.96281061e-01 -4.26678032e-01 1.54683173e-01 -4.33818847e-01 -3.77000198e-02 -2.47087792e-01 -7.95461953e-01 -4.48083788e-01 2.83372432e-01 -1.18969238e+00 5.44324458e-01 5.68777740e-01 8.11334431e-01 8.91225457e-01 1.48808013e-03 1.95662081e-02 -1.15572572e+00 9.50287133e-02 -1.72529921e-01 -8.47437918e-01 1.56339243e-01 -4.24573757e-02 1.00506179e-01 5.43323398e-01 -5.68351150e-01 -1.32601964e+00 3.48457873e-01 -2.08489701e-01 -3.86263460e-01 -1.09543130e-01 3.00699502e-01 -8.09772909e-02 3.12882587e-02 8.03086936e-01 3.18432510e-01 -2.84393251e-01 -8.43744725e-02 7.56484985e-01 1.69429243e-01 6.40371740e-01 -9.52626944e-01 6.42305434e-01 7.97769010e-01 1.33338660e-01 -1.02164471e+00 -6.76669121e-01 9.78068635e-03 -5.27983844e-01 -1.00293629e-01 1.14360285e+00 -1.18591833e+00 -5.68497419e-01 6.77289605e-01 -1.20679390e+00 -1.09960318e+00 -4.15195256e-01 5.08212633e-02 -9.10560191e-01 -1.74409956e-01 -8.49535704e-01 -5.09131193e-01 6.62525073e-02 -1.35194874e+00 1.20952225e+00 -8.67894739e-02 -1.81740504e-02 -9.56263185e-01 -4.05767620e-01 3.75864297e-01 3.98709238e-01 2.79002875e-01 9.89691079e-01 -1.44931125e-02 -1.41151369e+00 4.51434791e-01 -2.34787777e-01 3.84193540e-01 -7.31063168e-03 3.28199595e-01 -1.08990145e+00 1.25253141e-01 -4.39832777e-01 -4.40716475e-01 9.83865678e-01 6.54536784e-01 1.58179462e+00 -2.57683963e-01 -2.16126695e-01 8.17818642e-01 1.24188364e+00 -1.42486664e-02 1.04411817e+00 5.35817929e-02 1.03204143e+00 5.79688311e-01 4.49695438e-01 3.20753604e-01 4.02092993e-01 8.24344218e-01 6.54562175e-01 -3.45874578e-01 -4.79533672e-01 -3.67338568e-01 1.55525610e-01 6.88005030e-01 -2.06899688e-01 -3.28339130e-01 -9.32571054e-01 5.95453918e-01 -1.92612696e+00 -8.68456423e-01 -6.65463284e-02 1.71617281e+00 1.03292370e+00 -4.37084995e-02 -2.11857304e-01 -3.93526375e-01 5.50414860e-01 5.09737432e-01 -4.76608217e-01 -7.36338496e-02 -3.84052843e-01 3.83771539e-01 7.50957668e-01 7.96493113e-01 -1.06058574e+00 1.52917659e+00 6.53788948e+00 8.85418773e-01 -1.22425890e+00 1.01656511e-01 9.74804461e-01 -2.83150077e-01 -7.84048021e-01 8.85490924e-02 -6.66022658e-01 1.01104438e-01 8.37707996e-01 2.81725526e-01 5.59039652e-01 6.61876738e-01 1.86594337e-01 6.48182333e-02 -1.10515177e+00 6.43552780e-01 -1.88349590e-01 -1.76266289e+00 3.17848623e-01 7.33688772e-02 9.62873280e-01 1.69876263e-01 5.35162926e-01 -7.84813166e-02 9.61540878e-01 -1.29793966e+00 1.03812361e+00 4.63406324e-01 1.00223160e+00 -6.40949667e-01 8.89653638e-02 2.41150156e-01 -1.17653239e+00 3.23673971e-02 -2.51330525e-01 3.80353816e-02 1.13378644e-01 5.63371181e-01 -5.08029759e-01 2.74807513e-01 8.62152159e-01 8.65657449e-01 -3.22001040e-01 1.66506439e-01 -5.55243492e-01 4.75453049e-01 -3.29188466e-01 1.81700274e-01 4.48050469e-01 -8.52310434e-02 2.81576872e-01 1.14747930e+00 2.96253324e-01 3.30457747e-01 -3.16585153e-02 1.28181350e+00 -2.78895736e-01 -2.83632040e-01 -7.57615209e-01 -1.69561610e-01 2.69940227e-01 1.02705884e+00 -6.28822565e-01 -4.89137679e-01 -3.21339905e-01 6.21472895e-01 3.82157505e-01 6.15750790e-01 -1.04240012e+00 1.48096964e-01 8.39936435e-01 3.24306756e-01 6.43517852e-01 -4.57530677e-01 -4.89598751e-01 -1.13018751e+00 -3.42080861e-01 -1.13481963e+00 4.38025370e-02 -1.23744118e+00 -8.41025591e-01 5.08905351e-01 1.34609506e-01 -8.61839056e-01 -2.97047853e-01 -5.35666525e-01 -2.14418784e-01 6.45664454e-01 -1.38463414e+00 -1.60793233e+00 -2.66571254e-01 7.79778361e-01 6.35330677e-01 3.98179144e-02 6.77509427e-01 2.20956221e-01 -2.07872689e-01 2.41645157e-01 -3.56895983e-01 3.05122793e-01 4.64409292e-01 -1.13052094e+00 5.47783971e-01 9.90329564e-01 2.65909433e-01 6.22871876e-01 6.85114682e-01 -4.54886824e-01 -1.22228074e+00 -1.31499946e+00 6.19638383e-01 -4.73909646e-01 6.05729103e-01 -5.55632055e-01 -6.67927206e-01 1.17506802e+00 3.21425349e-01 2.71120012e-01 2.40658343e-01 4.25396822e-02 -7.07851887e-01 -1.75081134e-01 -6.72599375e-01 1.09525084e+00 1.28926003e+00 -7.09907770e-01 7.27855340e-02 3.19597811e-01 1.01107061e+00 -7.19174564e-01 -7.82433331e-01 5.00632405e-01 5.26936591e-01 -1.07382977e+00 1.24490345e+00 -4.31842685e-01 1.07712114e+00 -3.06665063e-01 -3.49085093e-01 -1.17046380e+00 -5.00086010e-01 -4.29071754e-01 2.09326997e-01 9.92103159e-01 6.45305336e-01 -6.24062538e-01 7.25781322e-01 5.88679373e-01 -2.74506599e-01 -5.87711036e-01 -6.47830784e-01 -4.62047011e-01 3.23208064e-01 -5.99322200e-01 7.31428683e-01 6.70966804e-01 -5.78282773e-01 2.50652105e-01 -5.19652128e-01 2.89260894e-01 5.48965275e-01 8.82825181e-02 8.10649157e-01 -6.61811531e-01 -4.77793336e-01 -4.93816137e-01 -1.15268879e-01 -1.31089020e+00 4.04874116e-01 -6.25491142e-01 3.90815794e-01 -1.32529759e+00 1.28484055e-01 -6.71634138e-01 1.35839153e-02 5.61592221e-01 1.68931291e-01 4.28502113e-01 2.40597889e-01 1.07299209e-01 -3.75891685e-01 4.50027138e-01 1.58243752e+00 -3.01453441e-01 2.54882555e-02 -3.87541622e-01 -8.04471612e-01 6.08835816e-01 7.87056565e-01 -3.12547863e-01 -7.63850629e-01 -7.94233263e-01 3.28765720e-01 1.15435570e-01 6.02502108e-01 -7.43504405e-01 1.93519965e-01 -4.28823411e-01 3.53314668e-01 -3.82825166e-01 6.97462380e-01 -6.30451024e-01 4.84927952e-01 2.15077713e-01 -7.48934567e-01 -2.39077508e-01 1.60929158e-01 2.86779493e-01 -2.83743918e-01 1.85524911e-01 9.89378631e-01 -5.43767273e-01 -6.89959407e-01 6.86569393e-01 -4.93289709e-01 -3.30726206e-02 7.85466492e-01 2.09833365e-02 -5.06236136e-01 -5.71326315e-01 -5.91421843e-01 -6.40038028e-02 6.27885401e-01 4.29278582e-01 6.20942533e-01 -1.16754484e+00 -7.62229323e-01 9.23902392e-02 7.29708523e-02 5.14599383e-01 3.67523253e-01 5.19884348e-01 -5.31090140e-01 3.04370552e-01 -2.01783597e-01 -8.86355221e-01 -1.14218307e+00 4.00803745e-01 5.24634719e-01 -2.52836168e-01 -6.08274519e-01 9.32791531e-01 9.88893688e-01 -4.02042091e-01 -1.48750022e-01 -4.89193350e-01 6.93783611e-02 -1.76868409e-01 3.65632683e-01 -2.08872929e-01 -1.79452568e-01 -6.52916253e-01 -1.60989463e-01 6.92580104e-01 2.50305440e-02 -4.78238434e-01 1.39337206e+00 9.36167240e-02 -2.50885427e-01 2.94760227e-01 1.11427760e+00 -2.23102644e-01 -1.77249968e+00 -2.64396906e-01 -4.75764751e-01 -3.99814010e-01 1.65537596e-01 -5.57193995e-01 -1.30654716e+00 1.03261185e+00 9.13423002e-02 -8.77539143e-02 1.14828479e+00 1.63689017e-01 5.96894324e-01 2.19490588e-01 2.75695920e-01 -9.54807281e-01 4.58152801e-01 6.33315206e-01 1.04753006e+00 -9.35566068e-01 -1.76955506e-01 -7.31541514e-01 -6.94072306e-01 9.17947233e-01 5.33285260e-01 -3.41138244e-01 7.07875431e-01 8.06420624e-01 1.87082570e-02 -5.86135499e-02 -1.20601261e+00 -3.27419877e-01 -2.50024647e-02 7.88435996e-01 4.87102449e-01 2.15128943e-01 2.83958554e-01 4.13745232e-02 -2.82433778e-01 2.35148873e-02 5.22186577e-01 6.09402239e-01 -1.53459355e-01 -1.08525538e+00 -1.66799985e-02 2.93821216e-01 -4.04386729e-01 -4.07636821e-01 -1.68848753e-01 6.80175304e-01 1.45711973e-01 7.32985377e-01 2.90307820e-01 -3.21090549e-01 8.70881304e-02 -3.11693400e-01 6.80399656e-01 -6.31741047e-01 -4.80677158e-01 1.00977406e-01 3.29404235e-01 -7.98523843e-01 -6.35704160e-01 -4.83687520e-01 -1.09537339e+00 -4.07024950e-01 5.94519079e-02 -3.52786124e-01 5.68825126e-01 7.78704822e-01 2.91963249e-01 8.81087065e-01 2.17345968e-01 -9.85347450e-01 -2.05642819e-01 -6.48556411e-01 -2.37302065e-01 2.77190268e-01 5.19985259e-01 -3.83046836e-01 -1.14219874e-01 4.69849527e-01]
[11.371524810791016, -0.4429171681404114]
2d1bc10f-0769-4b30-b741-6fd70dcd22e0
end-to-end-trainable-network-for-degraded
2010.14266
null
https://arxiv.org/abs/2010.14266v1
https://arxiv.org/pdf/2010.14266v1.pdf
End-to-end trainable network for degraded license plate detection via vehicle-plate relation mining
License plate detection is the first and essential step of the license plate recognition system and is still challenging in real applications, such as on-road scenarios. In particular, small-sized and oblique license plates, mainly caused by the distant and mobile camera, are difficult to detect. In this work, we propose a novel and applicable method for degraded license plate detection via vehicle-plate relation mining, which localizes the license plate in a coarse-to-fine scheme. First, we propose to estimate the local region around the license plate by using the relationships between the vehicle and the license plate, which can greatly reduce the search area and precisely detect very small-sized license plates. Second, we propose to predict the quadrilateral bounding box in the local region by regressing the four corners of the license plate to robustly detect oblique license plates. Moreover, the whole network can be trained in an end-to-end manner. Extensive experiments verify the effectiveness of our proposed method for small-sized and oblique license plates. Codes are available at https://github.com/chensonglu/LPD-end-to-end.
['Xu-Cheng Yin', 'Feng Chen', 'Chun Yang', 'Qi Liu', 'Jia-Wei Ma', 'Shu Tian', 'Song-Lu Chen']
2020-10-27
null
null
null
null
['license-plate-recognition', 'license-plate-detection']
['computer-vision', 'computer-vision']
[-2.14201197e-01 -4.68971550e-01 8.28072522e-03 -2.07688168e-01 -9.05137599e-01 -8.59973490e-01 2.20837981e-01 -6.16755545e-01 -9.65209007e-02 2.74367720e-01 -3.65724295e-01 -2.82646090e-01 2.34194532e-01 -7.33714342e-01 -8.09767485e-01 -5.95484376e-01 5.11977911e-01 3.49278420e-01 8.08397651e-01 -1.34576783e-01 6.40561700e-01 5.92296064e-01 -1.22591126e+00 1.17875941e-01 8.65745068e-01 9.66998219e-01 3.62198859e-01 2.66725570e-01 2.52478361e-01 4.34104592e-01 -9.00646225e-02 -4.48822558e-01 5.27422130e-01 5.89475557e-02 -5.21367416e-02 6.26400769e-01 4.47106034e-01 -4.33253258e-01 -6.29267812e-01 1.27604592e+00 6.03892542e-02 6.53199628e-02 6.82941556e-01 -1.00153470e+00 -2.90643871e-01 -2.05153808e-01 -8.57397914e-01 1.82761326e-01 2.87957769e-02 2.86493033e-01 5.99562407e-01 -1.44117820e+00 2.97680169e-01 8.26393723e-01 8.61496091e-01 2.30461389e-01 -4.39162701e-01 -8.77146363e-01 -7.36775994e-03 2.69271106e-01 -2.02644634e+00 -7.28815436e-01 8.98988545e-01 -4.50749099e-01 3.69868100e-01 1.38873100e-01 1.34457499e-01 1.27651021e-01 -1.10947132e-01 8.48366380e-01 8.20707500e-01 -2.37577394e-01 -2.13909209e-01 2.47982547e-01 -1.82752192e-01 8.96237969e-01 2.76898026e-01 -1.53411359e-01 1.15516141e-01 3.71941954e-01 9.74218309e-01 2.48081669e-01 -2.17296213e-01 -2.42479853e-02 -8.75341773e-01 5.15890002e-01 2.25075230e-01 7.89822415e-02 -1.17755711e-01 -2.20850810e-01 5.35355099e-02 -2.46631399e-01 4.73917872e-01 -3.06027420e-02 -1.27074555e-01 2.24082954e-02 -1.17679513e+00 7.42500275e-02 3.80154163e-01 1.14030731e+00 8.46410990e-01 -3.26230228e-02 4.19795603e-01 9.63939488e-01 5.84894121e-01 5.53455472e-01 -1.73073877e-02 -4.64821637e-01 1.27495062e+00 5.35983443e-01 2.88206428e-01 -1.21435702e+00 -2.53763974e-01 -1.29382253e-01 -7.61921346e-01 2.88973182e-01 5.02912462e-01 -2.64435112e-01 -7.11282551e-01 6.41627491e-01 2.07758918e-01 4.86480266e-01 -2.90862918e-01 1.20769322e+00 5.26562154e-01 1.06847918e+00 -3.59828085e-01 4.54029553e-02 1.40852404e+00 -1.19818664e+00 -2.29105055e-01 -8.29880834e-01 4.22157228e-01 -1.01609421e+00 8.32350969e-01 -1.56900305e-02 -9.38771307e-01 -4.37959433e-01 -9.24903512e-01 1.49483616e-02 -1.23589985e-01 9.40005064e-01 -6.70282990e-02 4.68094736e-01 -7.60161459e-01 5.32443225e-02 -6.15006268e-01 1.41580850e-01 5.12614608e-01 4.81435776e-01 -2.48487428e-01 -4.03945923e-01 -8.80647480e-01 7.88592875e-01 1.58893451e-01 5.20475328e-01 -3.84040594e-01 -3.70273322e-01 -7.77683616e-01 1.02208368e-01 6.98206663e-01 3.01504508e-02 8.93982172e-01 -7.72882700e-01 -1.26327491e+00 7.43722737e-01 -2.49654710e-01 2.02880695e-01 6.62617862e-01 1.20419152e-01 -6.32778943e-01 5.40238172e-02 2.57279158e-01 1.45709798e-01 7.44215667e-01 -1.14403665e+00 -1.14496219e+00 -5.85203230e-01 -2.89993882e-01 4.31526005e-01 1.44084170e-02 1.68953732e-01 -1.15838838e+00 -3.78782213e-01 1.95728943e-01 -1.09237957e+00 -5.49420007e-02 -9.69494805e-02 -6.81721270e-01 -3.02454233e-01 9.81914401e-01 -9.33425903e-01 1.04811001e+00 -2.64160657e+00 -7.62304783e-01 4.49407518e-01 5.94892837e-02 4.75547463e-01 1.76008850e-01 1.17180660e-01 1.11513995e-01 1.00183837e-01 -3.14854860e-01 -1.89320862e-01 -8.31093267e-02 -5.45072198e-01 -1.74607307e-01 9.46735144e-01 3.14047843e-01 5.15229344e-01 -2.10062504e-01 -4.93949443e-01 3.58394921e-01 1.98135749e-01 1.30159669e-02 1.78621057e-02 2.89985001e-01 2.91906428e-02 -7.24211395e-01 1.05264711e+00 1.36581981e+00 7.48755038e-02 -2.40595758e-01 -1.09239772e-01 -6.43391311e-01 -4.60519753e-02 -1.33053160e+00 7.24405885e-01 -4.07339096e-01 9.95504200e-01 2.06104070e-01 -5.81544757e-01 1.19299793e+00 6.84024170e-02 -4.54536006e-02 -6.96937442e-01 -2.82083303e-02 5.94657242e-01 -3.44075441e-01 -5.20296216e-01 3.38374138e-01 -1.22550577e-01 -2.43531197e-01 -1.76672801e-01 -6.82072997e-01 9.24576074e-02 3.01468819e-01 -3.82819355e-01 5.67463577e-01 -1.53963581e-01 1.19865023e-01 1.50624700e-02 1.09061468e+00 1.55381903e-01 6.18510962e-01 1.74836740e-01 -1.77605852e-01 7.86330998e-01 3.52831006e-01 -3.93152446e-01 -1.14998841e+00 -7.51847684e-01 -6.39736950e-02 6.69385433e-01 7.27025390e-01 2.08632365e-01 -7.70034790e-01 -5.08314371e-01 -1.52176589e-01 3.91949892e-01 -7.24887028e-02 1.55954733e-01 -8.80689561e-01 -3.17184389e-01 5.42741835e-01 6.31146729e-01 7.96263874e-01 -6.17491603e-01 1.56855196e-01 2.91471221e-02 -2.50398397e-01 -1.35612988e+00 -9.96338487e-01 -3.82541269e-01 -6.42435610e-01 -1.04284668e+00 -8.34995806e-01 -1.42560756e+00 8.60992253e-01 7.93534219e-01 5.53998172e-01 1.19301818e-01 2.73952633e-01 -4.16236341e-01 1.76447257e-01 -1.19209461e-01 -1.61071688e-01 -1.05047815e-01 2.14272318e-03 3.57821822e-01 3.24103594e-01 -1.69862971e-01 -6.34263396e-01 1.04152691e+00 -4.88356948e-01 -6.96268976e-02 9.21701193e-01 2.69409597e-01 5.85798800e-01 6.01368666e-01 8.69937092e-02 -4.99317557e-01 3.10010076e-01 -4.88805443e-01 -1.27610755e+00 6.75838515e-02 -2.62574315e-01 -5.16557336e-01 9.46918488e-01 -1.00045325e-02 -9.64323878e-01 6.42357767e-01 -3.87263954e-01 -5.27549267e-01 -5.00774682e-01 2.56735444e-01 -6.02040529e-01 -2.68331468e-01 -9.25243087e-03 7.61337399e-01 -8.33119377e-02 -3.69069606e-01 -1.13160476e-01 9.68059897e-01 6.25339866e-01 2.09779426e-01 1.34542048e+00 5.01247942e-01 -4.74599302e-02 -9.66784358e-01 -1.80506766e-01 -9.74805713e-01 -7.20555007e-01 -5.18063068e-01 6.71970904e-01 -1.07965648e+00 -9.23046350e-01 7.39078104e-01 -1.12128997e+00 6.51488230e-02 5.29657722e-01 4.70157892e-01 -1.19525433e-01 5.03828585e-01 -5.22720754e-01 -7.08684266e-01 -2.27661282e-02 -1.16295123e+00 1.12298274e+00 6.00158572e-01 5.18289089e-01 -7.42304444e-01 -8.88657868e-02 7.44801581e-01 -8.87895469e-04 -2.33056679e-01 5.52266181e-01 -6.36794269e-01 -9.71384883e-01 -6.29045129e-01 -6.42923176e-01 2.63296902e-01 -1.98799089e-01 1.78117752e-01 -6.04093134e-01 1.02973573e-01 -2.53971398e-01 1.12667985e-01 8.01508069e-01 4.66273934e-01 6.78220868e-01 -2.81496465e-01 -4.73163933e-01 5.81679463e-01 1.58687711e+00 3.65557641e-01 8.79077494e-01 5.51494122e-01 9.58753049e-01 4.56813097e-01 9.18301463e-01 3.70502084e-01 4.83767688e-01 7.75907993e-01 3.98125529e-01 -9.46063548e-02 1.30151972e-01 -2.59366006e-01 5.87957799e-01 7.14195371e-01 -3.08057398e-01 -3.27268600e-01 -1.10681355e+00 4.70369995e-01 -1.44903910e+00 -9.59739089e-01 -6.65240884e-01 2.22607660e+00 2.56418735e-01 2.64862090e-01 1.77120551e-01 -5.19001819e-02 1.41926527e+00 4.47347611e-02 -3.88471216e-01 1.58454239e-01 9.19462293e-02 -6.87721908e-01 8.73759151e-01 4.93983626e-01 -1.26882780e+00 1.03788722e+00 4.57364607e+00 1.25062084e+00 -1.23297858e+00 -1.97266102e-01 6.75261796e-01 3.16573650e-01 1.70858413e-01 -2.00254604e-01 -1.22440672e+00 7.94680774e-01 4.60912973e-01 6.74076378e-02 2.56042659e-01 8.62667024e-01 4.44710851e-01 -5.45790680e-02 -6.27608180e-01 9.28380728e-01 1.73155814e-01 -1.42302096e+00 -4.17264998e-01 1.86433747e-01 5.40706098e-01 1.74988091e-01 1.83199733e-01 1.96199194e-01 -4.48584944e-01 -6.55156255e-01 8.90336514e-01 3.67212236e-01 1.10529244e+00 -1.09330165e+00 8.71282756e-01 7.33279526e-01 -1.54313147e+00 -8.63776281e-02 -6.28992975e-01 2.55291075e-01 -6.59834594e-02 1.62831604e-01 -1.15978909e+00 1.07941367e-01 3.01664531e-01 7.58814335e-01 -5.48635304e-01 1.37092721e+00 -2.65918434e-01 5.99635482e-01 -2.26810947e-01 -7.72761777e-02 5.19159734e-01 -5.46929419e-01 5.58148623e-01 1.24619913e+00 6.45152867e-01 1.52830973e-01 1.08649664e-01 8.32383513e-01 -2.31155097e-01 -3.63738202e-02 -5.92310071e-01 3.04335803e-01 5.20851195e-01 1.24311650e+00 -9.80031252e-01 -1.63002126e-02 -7.15770721e-01 8.24005783e-01 -8.18749443e-02 2.52924562e-01 -1.12688625e+00 -6.59698248e-01 2.65667588e-01 6.20403409e-01 7.27559149e-01 -2.59895384e-01 -2.05650449e-01 -1.16171396e+00 4.94315118e-01 -5.11377156e-01 -6.16462678e-02 -7.22518146e-01 -9.38816845e-01 3.72268766e-01 -6.70581162e-01 -1.99625051e+00 1.12248980e-01 -7.92093217e-01 -9.03958142e-01 8.55771482e-01 -1.54192817e+00 -1.16688514e+00 -3.70581180e-01 7.77412176e-01 9.20472860e-01 -2.97438115e-01 1.52280629e-01 5.74843168e-01 -8.88579130e-01 4.68811244e-01 8.21257770e-01 7.90094912e-01 5.67488611e-01 -6.04489386e-01 4.66050833e-01 1.21227551e+00 -2.15375215e-01 2.38304958e-01 3.01219404e-01 -6.17153585e-01 -1.22882640e+00 -1.48309863e+00 6.31310701e-01 -2.00768754e-01 4.56731409e-01 -5.20188570e-01 -8.59715343e-01 4.34904426e-01 -2.49490961e-01 1.90161392e-01 3.14772844e-01 -3.06563824e-01 -2.69301478e-02 -4.56254929e-01 -9.65677083e-01 5.15058756e-01 3.76473010e-01 -3.90835881e-01 -2.77337581e-01 3.91848743e-01 7.49454722e-02 -3.84500772e-01 -4.60200429e-01 2.09672824e-02 5.26752412e-01 -7.30186641e-01 8.59371006e-01 2.12339342e-01 5.99338531e-01 -8.44708383e-01 6.16816431e-02 -1.04129338e+00 -4.91395116e-01 -8.17316473e-02 4.39790279e-01 1.43368542e+00 7.24025190e-01 -4.83883023e-01 1.07264221e+00 5.90082645e-01 -4.73678917e-01 -6.24491453e-01 -1.00563681e+00 -8.12604308e-01 -2.35304266e-01 -2.38821954e-01 3.82691503e-01 6.09336317e-01 -2.14735121e-01 1.62357166e-01 -5.16499758e-01 8.86484504e-01 5.16250253e-01 1.30970657e-01 5.93158007e-01 -9.41465974e-01 1.42461672e-01 -4.69448090e-01 -7.29632676e-01 -1.36777115e+00 1.70511290e-01 -5.22297204e-01 5.24537325e-01 -1.47985208e+00 1.35680065e-01 -4.30326402e-01 -9.69386194e-03 8.63134339e-02 -1.08820684e-01 6.72353029e-01 1.72133118e-01 8.58848393e-01 -7.60096431e-01 1.37377307e-01 1.11998653e+00 -3.76763701e-01 -1.58244327e-01 6.83780789e-01 -4.98193711e-01 1.06586170e+00 8.75672579e-01 -4.84083116e-01 -2.20596008e-02 -2.70258665e-01 7.13471621e-02 3.80211949e-01 2.52342850e-01 -9.56825733e-01 7.05196977e-01 -1.46818101e-01 5.69530487e-01 -8.61521363e-01 3.12056184e-01 -8.76577318e-01 -3.96189272e-01 1.70293868e-01 3.47139865e-01 -1.32739410e-01 3.06905299e-01 5.19632995e-01 -5.35360515e-01 -4.05897021e-01 9.23467278e-01 1.54356863e-02 -1.19045329e+00 3.78391385e-01 -6.87579274e-01 -6.90214634e-02 1.38601708e+00 -5.71437657e-01 -2.08358034e-01 -2.49659792e-01 -2.04637542e-01 2.38615498e-01 4.77763951e-01 2.37953380e-01 9.24280107e-01 -1.10472965e+00 -1.02595282e+00 2.30431348e-01 2.35930383e-01 1.68949217e-02 5.07628024e-01 8.87616277e-01 -1.03321171e+00 7.21085012e-01 -1.11607686e-01 -4.97394174e-01 -1.54686570e+00 4.22468781e-01 4.97545958e-01 1.52587682e-01 -4.01876420e-01 6.90294206e-01 3.13395470e-01 -3.72507691e-01 -2.08748758e-01 -4.71775532e-02 -2.42754936e-01 -8.01383331e-02 5.21867692e-01 6.07613504e-01 5.63574247e-02 -1.04798698e+00 -7.73757935e-01 1.02592468e+00 -1.95991352e-01 1.67479739e-01 1.19184458e+00 -3.66840184e-01 -4.90639731e-03 -1.15832001e-01 1.18408942e+00 4.58619535e-01 -1.62585080e+00 -1.37723852e-02 -2.30647475e-01 -8.07802320e-01 -2.13493966e-02 -3.58058244e-01 -1.19408655e+00 8.02976429e-01 3.51305783e-01 -1.73811883e-01 9.38546300e-01 -1.08704664e-01 8.00972641e-01 3.96811664e-01 2.10230470e-01 -1.21260381e+00 -4.67029244e-01 3.97773802e-01 6.45498514e-01 -1.24651575e+00 -2.08472144e-02 -6.69927716e-01 -8.12557638e-01 1.53339469e+00 8.11210454e-01 -2.69509852e-01 4.18852061e-01 2.75296658e-01 5.12160696e-02 5.74472100e-02 -1.37199908e-01 1.92546677e-02 4.28779989e-01 3.44492137e-01 5.53422868e-02 1.08684793e-01 2.45673284e-01 5.85353613e-01 3.13850731e-01 -1.71025708e-01 6.90802395e-01 5.84587872e-01 -7.19123006e-01 -5.74361622e-01 -9.03718591e-01 3.65750998e-01 -2.61534661e-01 -1.76826313e-01 -2.79118836e-01 9.92933929e-01 4.55869406e-01 9.58586574e-01 2.33958289e-01 -3.86098623e-01 5.87698638e-01 -2.50951558e-01 -1.48748145e-01 -3.01876456e-01 8.04636925e-02 2.65216738e-01 2.06218828e-02 -1.51969686e-01 8.63147601e-02 -8.27801228e-01 -1.26186144e+00 -4.95698571e-01 -5.15412033e-01 -8.66136998e-02 5.44260204e-01 1.01379097e+00 2.92777807e-01 2.03814432e-02 1.16472065e+00 -8.37549806e-01 -2.91271001e-01 -6.91624045e-01 -8.85711014e-01 -6.44675568e-02 3.34520131e-01 -3.62563491e-01 -3.52023125e-01 1.58071235e-01]
[9.851032257080078, -4.92311954498291]
fa8ef771-21a7-4ece-ac64-4ed1a943b509
a-hierarchical-variable-autonomy-mixed
2211.14095
null
https://arxiv.org/abs/2211.14095v1
https://arxiv.org/pdf/2211.14095v1.pdf
A Hierarchical Variable Autonomy Mixed-Initiative Framework for Human-Robot Teaming in Mobile Robotics
This paper presents a Mixed-Initiative (MI) framework for addressing the problem of control authority transfer between a remote human operator and an AI agent when cooperatively controlling a mobile robot. Our Hierarchical Expert-guided Mixed-Initiative Control Switcher (HierEMICS) leverages information on the human operator's state and intent. The control switching policies are based on a criticality hierarchy. An experimental evaluation was conducted in a high-fidelity simulated disaster response and remote inspection scenario, comparing HierEMICS with a state-of-the-art Expert-guided Mixed-Initiative Control Switcher (EMICS) in the context of mobile robot navigation. Results suggest that HierEMICS reduces conflicts for control between the human and the AI agent, which is a fundamental challenge in both the MI control paradigm and also in the related shared control paradigm. Additionally, we provide statistically significant evidence of improved, navigational safety (i.e., fewer collisions), LOA switching efficiency, and conflict for control reduction.
['Manolis Chiou', 'Rustam Stolkin', 'Grigoris Nikolaou', 'Tianshu Ruan', 'Aniketh Ramesh', 'Giannis Petousakis', 'Dimitris Panagopoulos']
2022-11-25
null
null
null
null
['robot-navigation']
['robots']
[ 2.31279328e-01 6.40357792e-01 5.94240008e-03 -1.67151704e-01 -6.04178131e-01 -5.18124104e-01 5.77919602e-01 1.28857896e-01 -7.19255149e-01 7.72834480e-01 6.96182102e-02 -5.64786971e-01 -7.39247441e-01 -5.72926998e-01 -5.81449866e-01 -6.86555624e-01 -3.31002086e-01 6.99132025e-01 4.06236202e-01 -8.94472837e-01 3.96953821e-01 4.82409120e-01 -1.50705409e+00 -4.23034132e-01 9.49300051e-01 5.53986549e-01 3.55543017e-01 7.61677563e-01 8.82137001e-01 1.31428194e+00 -2.52154052e-01 8.30488384e-01 4.53381449e-01 -8.55486095e-02 -1.11864936e+00 -2.38837361e-01 1.69752389e-02 -2.84761578e-01 -7.44484924e-03 5.55523634e-01 4.23425376e-01 5.37122786e-01 6.21388912e-01 -1.98502409e+00 5.38530290e-01 2.33752057e-01 -3.34358722e-01 5.46173863e-02 4.61170733e-01 5.42818785e-01 6.15496993e-01 -2.17828929e-01 1.09714079e+00 1.30888081e+00 5.20284534e-01 3.56659710e-01 -1.32697880e+00 -5.93604267e-01 3.36798519e-01 -5.33583388e-02 -1.52378750e+00 -4.56841856e-01 2.63154685e-01 -5.16564310e-01 1.21805739e+00 1.80124626e-01 5.39147854e-01 3.30606431e-01 9.70347464e-01 4.14754570e-01 7.30808854e-01 -3.74700695e-01 4.58253831e-01 -6.31402805e-03 2.32146494e-02 6.79574430e-01 2.80183434e-01 3.04303735e-01 -2.67680526e-01 -3.66134048e-01 5.22896171e-01 -4.56751466e-01 -2.62893945e-01 -8.77883732e-01 -1.40013921e+00 7.85306454e-01 4.01765376e-01 9.90185142e-02 -8.12856138e-01 2.00543404e-01 4.69553351e-01 6.02846503e-01 -1.32386833e-01 1.09382033e+00 -5.46811521e-01 -4.35890332e-02 -5.01288809e-02 6.17465794e-01 1.15332913e+00 9.65301692e-01 6.60524845e-01 -1.17136024e-01 -2.79012825e-02 -4.81689461e-02 4.01550889e-01 5.24720371e-01 -4.40113902e-01 -1.83487558e+00 3.72748911e-01 6.21219873e-01 7.07662940e-01 -1.38881373e+00 -7.32865274e-01 2.50220299e-01 -2.48318747e-01 9.26160753e-01 -2.46796608e-01 -7.84810662e-01 -3.31332028e-01 1.74292994e+00 5.66569507e-01 -4.89195049e-01 1.16916254e-01 8.61190319e-01 -3.05126220e-01 7.04729021e-01 2.02138603e-01 -4.02953327e-01 1.03156865e+00 -7.20530689e-01 -6.84446275e-01 -5.11141300e-01 9.81344283e-01 -1.55332834e-01 5.25928557e-01 4.31828380e-01 -9.54637051e-01 -2.71222562e-01 -8.91108394e-01 2.64225841e-01 6.28653616e-02 -5.45338929e-01 2.70372838e-01 -1.97281167e-01 -1.38038325e+00 -1.39077613e-02 -9.65602338e-01 -8.12161386e-01 -4.71938550e-01 6.33250296e-01 -6.67445362e-01 1.01464558e-02 -1.18340635e+00 1.50836658e+00 2.13760495e-01 -2.06611976e-01 -1.16585577e+00 -3.69698942e-01 -1.06823933e+00 -5.83230071e-02 1.04137862e+00 -6.93261921e-01 1.54608476e+00 -3.76112849e-01 -1.37963450e+00 1.60213947e-01 2.07517132e-01 -4.31036919e-01 3.83214593e-01 -4.85900074e-01 1.90166861e-01 2.30592668e-01 8.94545317e-01 8.96035612e-01 3.12176734e-01 -1.41909516e+00 -1.12819815e+00 -1.41381562e-01 2.69963175e-01 7.07144082e-01 2.96037018e-01 -1.78852841e-01 3.21772933e-01 3.75837572e-02 -1.60752073e-01 -1.51415265e+00 -8.77415299e-01 -8.87949485e-04 -1.43114716e-01 -1.71744153e-01 7.96488106e-01 -9.30003375e-02 1.27145565e+00 -2.08574891e+00 4.93937105e-01 3.50068122e-01 1.11292034e-01 -1.40072182e-01 -2.24633869e-02 7.67444015e-01 1.08661518e-01 -1.48173317e-01 -3.33721012e-01 2.20812052e-01 2.24572904e-02 1.61158770e-01 -1.47935525e-01 4.75388944e-01 -2.31357306e-01 7.73366615e-02 -9.36146796e-01 -5.12617290e-01 2.46309280e-01 -1.43443942e-02 -9.43895936e-01 3.73170406e-01 -1.10383853e-01 8.80216956e-01 -7.30434418e-01 3.30687791e-01 4.87659089e-02 3.03983897e-01 5.72175264e-01 1.09746873e-01 -1.04354620e+00 9.11737829e-02 -9.43424523e-01 1.48476648e+00 -4.01539385e-01 5.61692178e-01 1.03326690e+00 -5.59107661e-01 6.24764681e-01 5.45775592e-01 8.51832449e-01 -5.38436830e-01 4.77616340e-01 9.31703579e-03 7.47745261e-02 -4.44532096e-01 7.47668326e-01 1.76141247e-01 -7.44393289e-01 7.36467540e-01 -4.20066923e-01 -5.51625311e-01 7.11061954e-02 4.29325610e-01 1.50884211e+00 -1.39443412e-01 6.87977135e-01 -1.00789499e+00 2.09000796e-01 1.01703596e+00 6.97955310e-01 1.13811445e+00 -7.57133484e-01 -4.51738358e-01 2.78911084e-01 -3.05838317e-01 -4.70253080e-01 -6.86274588e-01 4.43961829e-01 1.25915527e+00 7.22383440e-01 -2.12879166e-01 -6.91419959e-01 -1.85492471e-01 4.83822078e-02 1.19573212e+00 -4.83719170e-01 -4.94803995e-01 -9.62118924e-01 9.16636607e-04 3.37510675e-01 2.24304602e-01 5.43285131e-01 -7.14311540e-01 -1.74894261e+00 4.67136532e-01 -4.67983425e-01 -8.41728389e-01 -4.88516957e-01 4.79055643e-01 -4.47976649e-01 -1.26005208e+00 3.08259368e-01 -7.38667309e-01 6.07788980e-01 5.38460910e-01 8.89108360e-01 1.21959537e-01 -3.16265136e-01 1.03297102e+00 -3.73566747e-01 -4.62834775e-01 -4.29756314e-01 1.26816481e-01 5.20075321e-01 -6.12473369e-01 -2.13216871e-01 -9.71824676e-02 -5.69742978e-01 7.29711831e-01 -4.17652577e-01 -1.47867516e-01 3.15411448e-01 5.49458027e-01 1.59283370e-01 2.60981172e-01 3.28975827e-01 -3.27121556e-01 9.31449115e-01 -6.50824726e-01 -7.47320473e-01 1.28603265e-01 -7.77402937e-01 2.43077651e-02 -1.45993475e-02 -1.30962417e-01 -1.37125731e+00 2.94023484e-01 7.28878200e-01 3.09637427e-01 -4.12948221e-01 5.57068169e-01 -4.32203524e-03 -1.63297772e-01 6.75071239e-01 -4.34044659e-01 4.17465806e-01 1.57101557e-01 3.12340289e-01 5.18321872e-01 5.25219142e-01 -6.78505719e-01 4.81269121e-01 3.79870683e-01 -6.18710704e-02 -7.17082381e-01 -8.46892316e-03 -3.93738240e-01 -4.63974237e-01 -3.65874469e-01 1.25839078e+00 -9.02445853e-01 -1.02842641e+00 2.88090855e-01 -1.15842330e+00 -9.85526264e-01 -1.00327790e-01 4.44586337e-01 -8.49723995e-01 -1.39367461e-01 -5.79725266e-01 -1.05841613e+00 -5.54797286e-03 -1.27789807e+00 8.78102899e-01 3.21361385e-02 -7.38404453e-01 -6.67110622e-01 4.45720524e-01 1.67413011e-01 6.57925129e-01 4.13455367e-01 7.59743512e-01 -5.52948892e-01 -6.75002456e-01 -1.57614723e-01 1.91855729e-01 -5.52044988e-01 4.39732568e-03 -1.48807928e-01 -1.16965622e-01 -6.73366547e-01 1.85273632e-01 -3.15534502e-01 -1.70676455e-01 3.37368190e-01 -5.35890937e-01 -4.32375789e-01 -9.02709186e-01 -3.03210795e-01 1.03026032e+00 8.38368833e-01 3.64607960e-01 7.32492387e-01 3.72603655e-01 1.22892880e+00 1.63226879e+00 6.09441698e-01 9.19262409e-01 7.77787626e-01 6.59950793e-01 4.75643761e-03 8.22661877e-01 2.00131193e-01 6.37856603e-01 5.34302592e-01 1.24249041e-01 -3.97549450e-01 -1.31050754e+00 5.97827733e-01 -2.20293808e+00 -7.36849308e-01 7.67697170e-02 1.81678247e+00 4.80869830e-01 -5.45544438e-02 -9.58667025e-02 -1.13730967e-01 7.80583024e-01 -2.85499215e-01 -5.01748681e-01 -4.87011254e-01 7.92386651e-01 -5.23284376e-01 7.76019871e-01 8.96306157e-01 -1.11263883e+00 1.10732150e+00 6.63586426e+00 1.74781054e-01 -7.79915094e-01 -4.91849743e-02 2.12112488e-03 1.12659492e-01 2.71520019e-01 2.86311150e-01 -4.12850559e-01 -2.53412813e-01 8.57734144e-01 -2.33467489e-01 3.96677613e-01 9.34216380e-01 5.66091359e-01 -8.33758235e-01 -1.06182718e+00 1.82324156e-01 -2.11317420e-01 -6.38870835e-01 -5.00156641e-01 1.31830975e-01 4.06347960e-01 -6.99048638e-02 -4.17891800e-01 4.75338012e-01 9.97381806e-01 -5.78413129e-01 1.06972110e+00 5.96922815e-01 3.56302828e-01 -1.00413561e+00 7.39168704e-01 7.78548002e-01 -1.38085330e+00 -5.71145952e-01 3.75235140e-01 -4.44535255e-01 7.00052500e-01 -2.37655222e-01 -8.09686124e-01 2.65159667e-01 7.72565901e-01 6.26454130e-02 -6.44573569e-02 5.37306964e-01 -1.13034546e-02 -1.72893450e-01 -3.68340313e-01 1.29084572e-01 4.36526716e-01 1.34054730e-02 1.01862502e+00 6.05866373e-01 -1.76555455e-01 6.92867756e-01 9.32368934e-01 5.55488825e-01 9.08131540e-01 -5.23007214e-01 -1.10822976e+00 3.37822556e-01 7.58094311e-01 7.87176073e-01 -6.65927291e-01 -1.41065121e-01 1.98307365e-01 5.40319383e-01 -1.92199141e-01 1.74880221e-01 -7.46604621e-01 -8.60700846e-01 8.51504207e-01 1.80936176e-02 -1.83856770e-01 -5.46503425e-01 -3.12724672e-02 -3.30792993e-01 -5.09064376e-01 -9.78495002e-01 1.81581274e-01 -7.89137423e-01 -6.43072963e-01 6.34469092e-01 7.56262422e-01 -1.27398276e+00 -7.90916085e-01 -1.63440779e-01 -4.24531639e-01 2.98017591e-01 -8.68094087e-01 -6.73115015e-01 -4.34978098e-01 4.66413051e-01 5.26377857e-01 -5.39919622e-02 6.55357063e-01 -1.24534257e-02 -4.46334958e-01 -1.88346922e-01 -7.48949289e-01 -4.05678779e-01 7.79972672e-01 -7.52104878e-01 7.86391944e-02 8.59858990e-01 -1.10824978e+00 8.99487555e-01 1.08745396e+00 -9.33649302e-01 -1.80787039e+00 -8.43451560e-01 4.85126197e-01 -5.81349552e-01 4.25671488e-01 1.69780001e-01 -3.92018557e-01 8.88061166e-01 3.60769510e-01 -5.88450730e-01 1.09059297e-01 -2.99005151e-01 3.96778554e-01 8.42712000e-02 -1.28680813e+00 9.74385858e-01 9.57960665e-01 -4.05533195e-01 -6.08097672e-01 -3.42128389e-02 1.05811989e+00 -1.28087699e-01 -6.14822209e-01 6.71279132e-01 3.52025062e-01 -7.69526243e-01 5.42900026e-01 -5.38761653e-02 -2.70290524e-01 -7.20490873e-01 -3.00571889e-01 -1.25714886e+00 -5.28379619e-01 -6.68520212e-01 7.69686460e-01 5.47468603e-01 3.11881304e-01 -7.65237570e-01 1.39738888e-01 1.22598910e+00 -4.99215931e-01 -1.93324924e-01 -8.98693085e-01 -6.10001624e-01 -3.35276842e-01 4.93394583e-02 -1.32842183e-01 1.07751310e+00 5.35075843e-01 4.86394763e-01 -2.12647483e-01 5.31505406e-01 3.62021565e-01 -5.15958011e-01 1.13354886e+00 -1.21404409e+00 1.86921775e-01 -7.10246786e-02 -2.10082740e-01 -5.98552406e-01 2.43765712e-01 -2.77965724e-01 1.13091838e+00 -1.52248621e+00 -3.90183538e-01 -6.11840606e-01 1.70900121e-01 7.21854806e-01 3.83489758e-01 -6.22073591e-01 4.41935956e-01 3.44684571e-01 -1.05280614e+00 4.83762622e-01 7.57059634e-01 -3.44511792e-02 -6.63528979e-01 -4.64167863e-01 -5.69041252e-01 8.37551713e-01 8.58343184e-01 -5.04024208e-01 -7.74364769e-01 -5.78706443e-01 1.29702240e-01 7.95822322e-01 2.70261347e-01 -1.24340737e+00 9.73806977e-01 -6.68762207e-01 -9.46645975e-01 -1.15411490e-01 6.82116859e-03 -1.35757232e+00 3.44849229e-01 1.06219900e+00 -5.43169081e-01 6.88667774e-01 2.40590364e-01 7.83693612e-01 -4.61158305e-02 2.32816651e-01 6.84731543e-01 -1.58244431e-01 -8.21772754e-01 -4.22526181e-01 -1.33094668e+00 -8.39732140e-02 1.48606837e+00 1.73707791e-02 -4.41230357e-01 -5.66063523e-01 -4.88248378e-01 8.30341041e-01 5.29422879e-01 1.53567106e-01 3.64103168e-01 -5.81966341e-01 -3.98301780e-01 7.97760766e-03 8.70435834e-02 2.13566422e-01 -2.35624939e-01 1.13353384e+00 -6.96538568e-01 4.61103052e-01 -5.46426952e-01 -6.39345527e-01 -1.02281892e+00 1.37896687e-01 3.53287101e-01 -2.17541307e-01 -3.17655087e-01 4.52481717e-01 3.71022642e-01 -7.76969731e-01 2.21369192e-01 -1.56118244e-01 -1.37916801e-03 -1.87165998e-02 2.30298370e-01 9.29442644e-01 -3.43271136e-01 -5.60870349e-01 -7.01293945e-01 3.69925529e-01 7.43992850e-02 -6.75499558e-01 9.11723793e-01 -5.39625704e-01 -4.34366107e-01 3.90689164e-01 3.87180895e-01 -3.06067288e-01 -1.23722064e+00 3.72257888e-01 3.71108830e-01 -7.28928344e-03 1.19323865e-01 -6.59896851e-01 -4.71289515e-01 4.65409458e-02 6.95345461e-01 2.27120500e-02 6.19935751e-01 -4.04898822e-01 3.56467605e-01 1.23618543e+00 1.27349532e+00 -1.31207800e+00 1.68429166e-01 9.21186209e-01 1.01375782e+00 -1.08856225e+00 -6.24791831e-02 -4.16127920e-01 -1.01664317e+00 4.85104918e-01 1.21787059e+00 -1.95269436e-01 5.77812731e-01 4.50438261e-01 3.27403188e-01 -4.85425174e-01 -1.18080926e+00 2.12379336e-01 -7.44827807e-01 5.76718211e-01 -1.39509201e-01 -9.03583169e-02 -4.10252623e-02 2.36520953e-02 1.53246269e-01 -1.18762940e-01 7.93178022e-01 1.95673037e+00 -9.65091467e-01 -4.53970730e-01 -4.90673453e-01 -1.24088153e-01 4.26251411e-01 4.98829424e-01 -3.53640050e-01 1.23130000e+00 9.83106121e-02 1.52887332e+00 3.11495572e-01 -5.47239006e-01 6.61697745e-01 -2.29986757e-01 -2.20154554e-01 -8.55764031e-01 -6.71007872e-01 -2.25210518e-01 3.79241705e-01 -1.05556393e+00 -5.16485214e-01 -6.05551362e-01 -1.61300516e+00 -1.31127849e-01 -3.04585904e-01 4.64604944e-01 5.23956239e-01 8.59551668e-01 6.90761626e-01 6.93469942e-01 7.30799735e-01 -1.26960063e+00 -3.67685050e-01 -6.59010172e-01 -1.82480395e-01 -2.07071349e-01 4.74878401e-01 -9.93294954e-01 -6.06926560e-01 -2.11128712e-01]
[4.827230930328369, 1.3341054916381836]
07555cc6-cb90-4f2d-80b7-9fd437755f1a
merging-diverging-hybrid-transformer-networks
2307.03427
null
https://arxiv.org/abs/2307.03427v1
https://arxiv.org/pdf/2307.03427v1.pdf
Merging-Diverging Hybrid Transformer Networks for Survival Prediction in Head and Neck Cancer
Survival prediction is crucial for cancer patients as it provides early prognostic information for treatment planning. Recently, deep survival models based on deep learning and medical images have shown promising performance for survival prediction. However, existing deep survival models are not well developed in utilizing multi-modality images (e.g., PET-CT) and in extracting region-specific information (e.g., the prognostic information in Primary Tumor (PT) and Metastatic Lymph Node (MLN) regions). In view of this, we propose a merging-diverging learning framework for survival prediction from multi-modality images. This framework has a merging encoder to fuse multi-modality information and a diverging decoder to extract region-specific information. In the merging encoder, we propose a Hybrid Parallel Cross-Attention (HPCA) block to effectively fuse multi-modality features via parallel convolutional layers and cross-attention transformers. In the diverging decoder, we propose a Region-specific Attention Gate (RAG) block to screen out the features related to lesion regions. Our framework is demonstrated on survival prediction from PET-CT images in Head and Neck (H&N) cancer, by designing an X-shape merging-diverging hybrid transformer network (named XSurv). Our XSurv combines the complementary information in PET and CT images and extracts the region-specific prognostic information in PT and MLN regions. Extensive experiments on the public dataset of HEad and neCK TumOR segmentation and outcome prediction challenge (HECKTOR 2022) demonstrate that our XSurv outperforms state-of-the-art survival prediction methods.
['Jinman Kim', 'Dagan Feng', 'Michael Fulham', 'Lei Bi', 'Mingyuan Meng']
2023-07-07
null
null
null
null
['tumor-segmentation']
['computer-vision']
[ 2.50934139e-02 6.24046996e-02 -6.44652665e-01 -4.72197950e-01 -1.57841027e+00 -7.19396397e-02 3.10438454e-01 1.97651163e-01 -5.14684618e-01 7.95759022e-01 7.04184234e-01 -6.19507968e-01 -8.67574066e-02 -7.10609674e-01 -3.63100618e-01 -1.19852698e+00 2.70288847e-02 6.45156324e-01 2.78765589e-01 -2.11462870e-01 -2.67769009e-01 6.51268542e-01 -9.99388039e-01 6.01466715e-01 6.39896810e-01 1.17551732e+00 4.97455746e-01 7.87889659e-01 -6.77500069e-02 9.73660767e-01 -1.68655708e-01 -6.72406405e-02 -4.74252164e-01 -3.28372121e-01 -1.15812874e+00 -4.91508752e-01 -1.34003982e-01 -4.54089999e-01 -6.23733580e-01 5.47790408e-01 7.77397037e-01 -4.84865010e-01 7.09112406e-01 -1.24212015e+00 -2.48508483e-01 5.16686916e-01 -6.02122486e-01 3.03058684e-01 -1.64942816e-01 2.26819769e-01 9.82076705e-01 -7.83110261e-01 4.53798294e-01 6.56736374e-01 9.40967500e-01 7.28443444e-01 -7.49263883e-01 -3.85329545e-01 -1.80365250e-01 2.54091799e-01 -1.20445931e+00 -1.42915681e-01 2.66564548e-01 -1.57813132e-01 9.72241521e-01 5.74086905e-01 8.22386384e-01 1.27843380e+00 9.27671611e-01 1.31323266e+00 8.04821074e-01 1.14438385e-01 -1.23548716e-01 -6.05386309e-02 1.58566847e-01 7.81653166e-01 -4.21064049e-01 4.35477525e-01 -4.96397242e-02 -8.02568421e-02 4.79801387e-01 5.23676276e-01 -4.89111215e-01 1.10238127e-01 -1.01711524e+00 9.43269074e-01 1.17635977e+00 4.94595140e-01 -2.70395875e-01 2.24888057e-01 7.21432745e-01 -4.29487601e-02 5.40185690e-01 -4.25216824e-01 -7.32418776e-01 3.96376669e-01 -1.02563536e+00 -1.84900388e-01 4.93164778e-01 6.38717532e-01 2.29292922e-02 -4.80009645e-01 -7.54609883e-01 7.13093340e-01 4.73915756e-01 2.29454830e-01 1.16315103e+00 -7.68508911e-02 8.45856741e-02 7.88033009e-01 -3.61857682e-01 -1.40532702e-02 -1.30485106e+00 -7.51345813e-01 -1.22078252e+00 -2.11557318e-02 2.76029289e-01 1.74246252e-01 -1.39442992e+00 1.63934112e+00 2.89696962e-01 2.17081681e-01 1.24152720e-01 8.87629867e-01 1.47041273e+00 4.15308446e-01 6.01185322e-01 -2.11753920e-01 1.78232086e+00 -1.18169260e+00 -3.60869646e-01 -7.83675015e-02 1.53726053e+00 -2.66068459e-01 6.70569301e-01 -3.63876015e-01 -7.74096549e-01 -1.49993628e-01 -4.12617743e-01 -4.09970671e-01 -3.76362920e-01 3.18446904e-01 6.40712321e-01 3.85922343e-01 -1.29319620e+00 4.99331474e-01 -9.56363380e-01 -4.61572617e-01 7.97281981e-01 7.43493915e-01 -6.63342357e-01 3.42227928e-02 -1.25557494e+00 1.09698868e+00 4.35383052e-01 -1.41170517e-01 -1.42585027e+00 -1.02738416e+00 -8.32486033e-01 1.98417813e-01 1.37818143e-01 -1.22389567e+00 1.38485956e+00 -9.48752642e-01 -1.36249852e+00 8.33663046e-01 -2.13659093e-01 -3.34258646e-01 3.48724812e-01 1.99829042e-01 -3.63350093e-01 2.17798054e-01 1.36294141e-01 7.48406410e-01 3.37643325e-01 -8.39885294e-01 -1.01473594e+00 -8.33465874e-01 -7.13560522e-01 3.03793967e-01 -3.68603379e-01 -5.26591092e-02 -4.34966147e-01 -6.19736493e-01 -4.60815467e-02 -7.16324389e-01 -4.86505538e-01 5.44135757e-02 -4.86678779e-01 -3.70222241e-01 9.83589947e-01 -9.81071353e-01 1.26301789e+00 -1.94730854e+00 3.23185563e-01 8.91250670e-02 4.31552470e-01 -6.88217282e-02 -1.19481139e-01 -1.47233486e-01 -2.19482377e-01 1.01568811e-01 -1.04166001e-01 -7.71702170e-01 -3.98477525e-01 2.85869867e-01 3.52966607e-01 5.66320121e-01 -2.71064434e-02 1.41883755e+00 -8.59874547e-01 -9.89182413e-01 8.69335532e-02 5.84849119e-01 -4.64005232e-01 3.85001540e-01 3.60871367e-02 8.16917300e-01 -5.58621585e-01 1.14048135e+00 2.42743745e-01 -5.70667267e-01 -9.45605338e-02 -1.92794293e-01 1.10240519e-01 5.99461198e-02 1.02048866e-01 1.65599227e+00 -5.32269239e-01 2.42114648e-01 -2.06422955e-02 -7.53295302e-01 2.97796279e-01 6.69341207e-01 9.48870897e-01 -7.99681485e-01 7.57698238e-01 3.53477359e-01 -1.13992348e-01 -4.52040374e-01 7.71974325e-02 -5.59207201e-01 -2.15685636e-01 1.42153636e-01 2.80840635e-01 -7.36094033e-03 -4.30923134e-01 1.74301311e-01 1.40432847e+00 -3.04005831e-01 5.80902874e-01 -5.54140918e-02 7.54903734e-01 -6.14903010e-02 6.89893782e-01 2.62956351e-01 -5.37299573e-01 6.38337016e-01 4.77764606e-01 -3.52964550e-01 -6.43750489e-01 -7.80406415e-01 -4.23115641e-01 1.07585859e+00 -2.27838501e-01 -9.58450437e-02 -3.09674412e-01 -1.18719828e+00 4.50548204e-03 5.25072515e-01 -1.16653967e+00 -4.00782466e-01 -4.99541342e-01 -1.25477874e+00 5.56694508e-01 1.05820966e+00 1.44406915e-01 -1.05140007e+00 -4.15214717e-01 4.15464431e-01 -2.32352510e-01 -7.33413219e-01 -4.17056382e-01 7.41680443e-01 -1.18673658e+00 -1.13893807e+00 -1.16705298e+00 -7.47695565e-01 5.33927619e-01 -1.13944590e-01 1.16666520e+00 2.36801147e-01 -2.78490990e-01 5.90343401e-03 -2.16288075e-01 -1.75096810e-01 -4.38259482e-01 3.51734549e-01 -5.59898138e-01 -1.43796578e-01 2.09055245e-01 -2.15372279e-01 -8.54507327e-01 4.57560420e-01 -5.83515823e-01 4.61014986e-01 9.27294791e-01 1.40655863e+00 9.62692618e-01 -1.77228138e-01 5.77982843e-01 -7.19093621e-01 2.33940594e-02 -9.32623208e-01 -7.41642490e-02 2.59344965e-01 -2.48111039e-01 -1.74891695e-01 8.39492261e-01 1.96899232e-02 -8.29934418e-01 2.92028338e-01 -7.08424807e-01 -8.13127384e-02 -1.16471820e-01 7.02919304e-01 -2.71762282e-01 -1.34993091e-01 2.34172806e-01 1.44794852e-01 -4.15814333e-02 -3.33669245e-01 -4.52580117e-02 6.57822549e-01 3.44411075e-01 -6.25758767e-02 1.53915241e-01 5.14009893e-01 2.68768251e-01 -1.80991367e-01 -9.66620862e-01 -5.26241541e-01 -6.12568915e-01 -7.93334842e-02 9.96230543e-01 -8.69839668e-01 -8.66883576e-01 7.05927372e-01 -6.23322248e-01 -4.25180554e-01 -1.50955006e-01 3.31893027e-01 -6.36418462e-01 -1.46569163e-01 -1.00887668e+00 -2.54850179e-01 -8.11621785e-01 -1.49413896e+00 1.57141709e+00 5.16837180e-01 2.51126122e-02 -1.12628388e+00 1.13620590e-02 4.17416722e-01 4.82299805e-01 3.18287492e-01 1.10275435e+00 -8.39533329e-01 -1.88199461e-01 -3.01181614e-01 -4.11051124e-01 -2.01503232e-01 -1.51681423e-01 -1.68107286e-01 -9.72136796e-01 -4.42805171e-01 -3.96572769e-01 -4.10936326e-01 1.29671383e+00 8.52697849e-01 1.64585781e+00 4.71275067e-04 -1.28633392e+00 1.06934440e+00 1.40073764e+00 2.44078174e-01 5.69033325e-01 2.28023678e-01 9.50908124e-01 3.51747721e-01 2.81157643e-01 2.35591561e-01 9.11093771e-01 5.93733191e-01 7.40178704e-01 -4.01822567e-01 -2.46062323e-01 -4.02177870e-02 9.70554128e-02 3.85724962e-01 1.66994825e-01 -5.54857433e-01 -1.14223719e+00 8.08783531e-01 -1.61163270e+00 -5.40745437e-01 -1.92075789e-01 1.67988420e+00 8.76167893e-01 -5.58853522e-02 -6.44532293e-02 1.22847766e-01 4.03520077e-01 -1.20162509e-01 -7.01496363e-01 -1.00391150e-01 -2.50541512e-02 2.17978075e-01 6.15811527e-01 2.26587921e-01 -1.27657998e+00 6.19889200e-01 5.26162863e+00 1.08014107e+00 -1.15139401e+00 6.02764666e-01 1.11172163e+00 -2.64566362e-01 -2.58567691e-01 -2.24932298e-01 -8.22525263e-01 4.02095258e-01 1.27360487e+00 2.63258606e-01 -5.83929300e-01 7.27216780e-01 5.28891273e-02 -9.80606675e-02 -1.05789256e+00 7.16807306e-01 -1.55985191e-01 -1.22511101e+00 -1.25235483e-01 2.32404798e-01 2.79385388e-01 3.63308191e-01 2.77625956e-02 5.66959143e-01 4.27431583e-01 -1.29461527e+00 3.91984612e-01 6.57479763e-01 1.03338528e+00 -7.74435103e-01 1.29053771e+00 3.31426054e-01 -1.15306318e+00 -4.85867769e-01 1.92414567e-01 7.69099653e-01 3.92627925e-01 3.29127282e-01 -8.91981721e-01 8.52830291e-01 7.61463702e-01 9.15909886e-01 -7.17107415e-01 9.93235052e-01 5.53379729e-02 6.81264758e-01 -2.99044847e-01 1.60679862e-01 3.13979983e-01 6.09318197e-01 4.52841632e-02 9.14411485e-01 6.15216076e-01 6.45325661e-01 -2.14677632e-01 4.27764505e-01 -1.40732080e-01 -3.02050472e-03 1.33895606e-01 1.62935123e-01 -3.03467326e-02 1.35610402e+00 -7.38673508e-01 -3.93958926e-01 -5.96102297e-01 8.67641509e-01 3.12134296e-01 -6.40343875e-02 -1.01895547e+00 1.86503440e-01 2.06686750e-01 1.96356907e-01 9.11692753e-02 5.37080348e-01 -6.68488681e-01 -9.82916176e-01 -8.61702383e-01 -4.57262665e-01 9.73769784e-01 -6.67629123e-01 -1.28768206e+00 8.86235654e-01 -3.85905147e-01 -1.12049651e+00 1.93966120e-01 -5.16489565e-01 -7.52826869e-01 9.20171440e-01 -1.72802579e+00 -1.80200505e+00 -5.60736716e-01 8.47511768e-01 4.48241889e-01 -1.57634825e-01 1.00158143e+00 3.16459209e-01 -9.58665609e-01 1.04273164e+00 -3.20978425e-02 1.39751434e-01 4.74307060e-01 -1.34041810e+00 -6.87453449e-02 8.16996247e-02 -7.16026664e-01 -1.36772439e-01 2.33134124e-02 -5.27710080e-01 -9.98145998e-01 -1.44632411e+00 8.34437251e-01 -4.64774102e-01 4.74001408e-01 1.35271832e-01 -9.28112149e-01 6.04578853e-01 -1.45747766e-01 3.89911622e-01 1.06188607e+00 -2.61185884e-01 1.97966620e-01 6.12439737e-02 -1.25864327e+00 2.94185936e-01 6.72816157e-01 -2.68371582e-01 -1.47608966e-01 4.03049946e-01 7.14062870e-01 -5.42926908e-01 -1.24887836e+00 1.02087176e+00 5.40015936e-01 -7.29365051e-01 1.12035251e+00 -4.59050536e-01 6.10632122e-01 -1.24838673e-01 -3.97036262e-02 -1.31354618e+00 -6.20366514e-01 2.51719743e-01 -1.85795739e-01 8.81720960e-01 5.90781033e-01 -5.15609384e-01 8.61675560e-01 3.27821225e-01 -7.24419594e-01 -1.51618028e+00 -1.14582908e+00 -1.94640666e-01 3.68497074e-01 -1.35872528e-01 6.73296213e-01 5.85946560e-01 -2.54451871e-01 1.67199329e-01 -8.34198073e-02 2.66101271e-01 2.93941766e-01 3.07851862e-02 4.28573117e-02 -1.14950323e+00 -2.00400501e-01 -7.65779138e-01 -3.57275724e-01 -6.66218579e-01 1.66672394e-01 -1.12938464e+00 -2.46160552e-01 -1.84347332e+00 8.42572749e-01 -5.38514733e-01 -9.70114112e-01 9.32338774e-01 -3.35544735e-01 2.01007023e-01 -2.00258166e-01 3.05035681e-01 -3.84603471e-01 8.05520535e-01 1.40479779e+00 -2.37353519e-01 4.26506326e-02 1.76430330e-01 -6.38516247e-01 6.40105069e-01 7.03983545e-01 -5.35740495e-01 1.03342779e-01 -5.10911494e-02 -2.08346993e-01 9.87532973e-01 6.96743727e-01 -8.82225573e-01 2.55113810e-01 -8.71856809e-02 8.25738490e-01 -1.09901392e+00 1.95675224e-01 -9.93872881e-01 7.64435809e-03 6.96602106e-01 -7.45465234e-02 -1.30626991e-01 1.91173568e-01 4.82271492e-01 -4.38937783e-01 1.85885176e-01 9.05321181e-01 -2.89793015e-02 -6.79469109e-01 9.71685827e-01 -2.42248625e-01 -3.85010839e-01 1.31394172e+00 -1.97849303e-01 -3.90323699e-01 4.78945561e-02 -9.64643300e-01 7.00311542e-01 1.98722199e-01 2.51124501e-01 7.19436407e-01 -1.44477201e+00 -7.51612723e-01 1.95640430e-01 2.45263800e-01 1.21736199e-01 7.32683599e-01 1.48024499e+00 -3.73003334e-01 4.75125581e-01 -1.91540435e-01 -5.36960959e-01 -1.32595432e+00 4.92338032e-01 9.02763903e-01 -9.42899406e-01 -5.83650768e-01 1.44585812e+00 6.34464204e-01 -4.02195662e-01 1.62196100e-01 -4.65509742e-01 -5.78533113e-01 4.55972515e-02 4.95137423e-01 -6.15086146e-02 2.04463273e-01 -8.37592244e-01 -5.44282317e-01 3.04967612e-01 -3.63771200e-01 3.26790780e-01 1.30059803e+00 -2.97150780e-02 -1.52020931e-01 -3.18763703e-02 1.66718864e+00 -7.68010318e-01 -1.11785543e+00 -2.28374824e-01 5.82322420e-04 5.39077222e-02 6.07004583e-01 -1.31181824e+00 -1.51180100e+00 8.55795622e-01 8.82226050e-01 -3.96347612e-01 1.74824321e+00 4.45303708e-01 1.40803456e+00 -5.64728141e-01 1.57754451e-01 -5.33607066e-01 -3.33157405e-02 4.57106531e-01 8.40015590e-01 -1.55677986e+00 -3.20474893e-01 -2.20930502e-01 -6.59706235e-01 1.08187425e+00 7.22733319e-01 3.46047670e-01 1.00772119e+00 3.92207980e-01 -2.48138700e-02 -4.68432546e-01 -9.59936917e-01 -3.24161291e-01 1.72163725e-01 1.73699021e-01 5.83730996e-01 5.71609557e-01 -1.50211111e-01 1.22593892e+00 -1.15440346e-01 1.81576863e-01 8.54337364e-02 7.61179864e-01 -4.46551293e-01 -1.02353561e+00 -1.80151567e-01 7.82060683e-01 -6.43388629e-01 -4.42421377e-01 -1.40456989e-01 7.53537476e-01 1.73084527e-01 4.96808618e-01 -1.94627389e-01 -4.99478996e-01 1.87431440e-01 -3.35817225e-02 1.76875129e-01 -4.30544823e-01 -1.05347860e+00 8.75567347e-02 -2.14631364e-01 -4.35610771e-01 -4.28485125e-02 -4.85728681e-01 -1.52520490e+00 -1.92877501e-01 -3.33981127e-01 6.30054809e-03 5.43433964e-01 9.65201139e-01 1.24962524e-01 1.18957448e+00 5.42553723e-01 -5.63416362e-01 -3.25616002e-01 -9.76978719e-01 -5.38535357e-01 -9.63247195e-02 5.47408581e-01 -6.41094863e-01 -1.28100097e-01 -4.46519703e-01]
[14.983065605163574, -2.5990729331970215]
46ba42e2-137c-4918-94c9-3a47c73cb6b2
beyond-spectral-gap-extended-the-role-of-the
2301.02151
null
https://arxiv.org/abs/2301.02151v1
https://arxiv.org/pdf/2301.02151v1.pdf
Beyond spectral gap (extended): The role of the topology in decentralized learning
In data-parallel optimization of machine learning models, workers collaborate to improve their estimates of the model: more accurate gradients allow them to use larger learning rates and optimize faster. In the decentralized setting, in which workers communicate over a sparse graph, current theory fails to capture important aspects of real-world behavior. First, the `spectral gap' of the communication graph is not predictive of its empirical performance in (deep) learning. Second, current theory does not explain that collaboration enables larger learning rates than training alone. In fact, it prescribes smaller learning rates, which further decrease as graphs become larger, failing to explain convergence dynamics in infinite graphs. This paper aims to paint an accurate picture of sparsely-connected distributed optimization. We quantify how the graph topology influences convergence in a quadratic toy problem and provide theoretical results for general smooth and (strongly) convex objectives. Our theory matches empirical observations in deep learning, and accurately describes the relative merits of different graph topologies. This paper is an extension of the conference paper by Vogels et. al. (2022). Code: https://github.com/epfml/topology-in-decentralized-learning.
['Martin Jaggi', 'Hadrien Hendrikx', 'Thijs Vogels']
2023-01-05
null
null
null
null
['distributed-optimization']
['methodology']
[-5.25739491e-01 4.04008180e-01 -1.28833190e-01 -1.59328684e-01 -5.25952578e-01 -4.88025993e-01 4.15754706e-01 5.22327423e-01 -4.99147147e-01 8.39652956e-01 1.60554379e-01 -3.59166741e-01 -4.69712436e-01 -6.34062588e-01 -9.69144821e-01 -6.77598417e-01 -6.39799297e-01 7.55289853e-01 -1.86936378e-01 -1.00022301e-01 -2.34561842e-02 3.17573607e-01 -8.10981512e-01 -2.02176884e-01 5.79829156e-01 3.99317294e-01 2.06233308e-01 1.09687376e+00 1.42036900e-01 8.80211055e-01 -6.10798120e-01 -3.01515311e-01 5.67074478e-01 -5.32296956e-01 -9.42865610e-01 8.03744718e-02 4.87194180e-01 3.47823352e-02 -6.04821384e-01 1.11995518e+00 4.94140744e-01 2.14259513e-02 5.28731346e-01 -1.31467664e+00 -5.90447247e-01 1.10508597e+00 -7.86685705e-01 4.20504898e-01 -2.47965947e-01 9.25091654e-02 1.27412307e+00 -3.64364445e-01 5.22336543e-01 1.26304829e+00 9.18195307e-01 3.78847480e-01 -1.29896820e+00 -3.83210510e-01 3.31802756e-01 -1.85563028e-01 -1.25382090e+00 -4.33499306e-01 5.57971954e-01 -4.07134116e-01 7.97056973e-01 1.02442518e-01 9.04854655e-01 6.54163122e-01 3.45094532e-01 5.35647988e-01 9.62116659e-01 -2.69280165e-01 1.67784527e-01 4.81439680e-02 -1.38092777e-02 1.07573009e+00 7.89004505e-01 7.59878159e-02 -9.32538927e-01 -2.36443236e-01 9.24105942e-01 6.75656796e-02 -4.33423042e-01 -4.61616367e-01 -1.15730870e+00 9.33212638e-01 7.90378928e-01 3.79921854e-01 -4.64829654e-01 9.66908395e-01 1.56109780e-01 9.87436056e-01 7.82130063e-01 5.00583529e-01 -4.90838796e-01 -2.23935112e-01 -6.67201996e-01 1.60343975e-01 1.20560324e+00 7.69440293e-01 1.06542826e+00 -1.00179456e-01 3.61513317e-01 4.30239350e-01 4.50871080e-01 3.51606488e-01 1.59536749e-01 -1.39251339e+00 6.57776773e-01 2.93307155e-01 -2.52895500e-03 -1.17751133e+00 -8.15414250e-01 -9.37087059e-01 -1.04461884e+00 5.88296503e-02 9.51152742e-01 -7.36567795e-01 -1.19859457e-01 1.71542048e+00 2.26393133e-01 -4.05175872e-02 -1.07112698e-01 1.15025020e+00 1.65693343e-01 4.67270911e-01 -4.55022186e-01 -2.11715460e-01 6.46029890e-01 -1.04346526e+00 -3.91774446e-01 -4.04098183e-01 1.18473828e+00 -5.03135502e-01 7.92658687e-01 3.70169580e-01 -1.10888696e+00 -3.76169272e-02 -7.39423752e-01 5.93404919e-02 -8.07103738e-02 -3.50950211e-01 9.58907425e-01 6.45310581e-01 -1.78429163e+00 1.01151907e+00 -1.03556550e+00 -3.42992157e-01 6.58676267e-01 6.39102578e-01 -1.00264624e-01 -2.12137531e-02 -7.72075534e-01 5.62664092e-01 2.30668802e-02 1.14563014e-02 -9.85915780e-01 -7.28807569e-01 -4.59984601e-01 4.81036715e-02 3.17018777e-01 -9.55729187e-01 1.26687515e+00 -1.33143640e+00 -1.24319971e+00 6.72189295e-01 9.34692994e-02 -6.64812386e-01 7.32372582e-01 -1.58439115e-01 3.98989111e-01 -3.28049734e-02 -1.62234560e-01 3.27811748e-01 6.59351408e-01 -1.13220215e+00 -3.04622829e-01 -6.26296878e-01 3.48225683e-01 5.83234251e-01 -5.51765501e-01 -4.15647835e-01 -9.28162113e-02 -2.57867903e-01 -2.82068610e-01 -1.07295156e+00 -5.66699862e-01 -7.11618084e-03 -2.04087391e-01 -1.96574062e-01 3.89624774e-01 -2.17863142e-01 9.25515890e-01 -1.87215579e+00 3.43747646e-01 4.34819102e-01 1.11836600e+00 -3.79297107e-01 -2.77521789e-01 8.26085091e-01 3.13682824e-01 2.65387207e-01 4.21356931e-02 -6.38665557e-01 1.15050785e-01 1.73034564e-01 2.46806115e-01 1.16724157e+00 -3.54247391e-01 9.96420860e-01 -1.04870903e+00 -4.25349653e-01 -1.88148275e-01 2.80581743e-01 -8.06682587e-01 -3.08296740e-01 -2.79292669e-02 3.94813061e-01 -7.43607938e-01 4.39086482e-02 2.97624975e-01 -1.01425028e+00 5.63684106e-01 3.78084838e-01 9.96493921e-02 8.64903480e-02 -1.07541573e+00 1.68151820e+00 -3.93205523e-01 1.02403688e+00 8.56463313e-01 -1.37575281e+00 5.16067207e-01 8.05666596e-02 5.94014764e-01 -3.01544517e-01 1.53346494e-01 2.07812652e-01 2.02355891e-01 -2.15597823e-01 5.44240810e-02 1.22287109e-01 2.63057262e-01 8.38944435e-01 -7.31650516e-02 -2.26775616e-01 3.63542475e-02 6.28049493e-01 1.42733872e+00 -6.59670949e-01 1.03680016e-02 -8.93359244e-01 -1.11392260e-01 -1.02080278e-01 1.50772810e-01 1.11593783e+00 -1.42654806e-01 2.81425476e-01 8.88745368e-01 -4.44123477e-01 -1.05665469e+00 -7.21791804e-01 3.27776730e-01 1.40660560e+00 2.00534955e-01 -5.26105225e-01 -9.31353569e-01 -5.99767923e-01 3.81774068e-01 -5.32983337e-03 -6.25257909e-01 1.54497147e-01 -4.37236965e-01 -8.89245331e-01 2.35519066e-01 -1.34308152e-02 1.20000765e-01 -5.65703332e-01 -2.08854765e-01 1.15085198e-02 1.36953652e-01 -7.05966175e-01 -8.46216559e-01 1.24451607e-01 -1.17371750e+00 -1.33835065e+00 -7.19593704e-01 -6.99426889e-01 8.64541113e-01 4.61511105e-01 1.32094145e+00 6.78400695e-01 -1.42148763e-01 9.68381345e-01 -2.02649478e-02 -4.27913547e-01 -3.79475862e-01 5.39577842e-01 1.74663931e-01 -9.12112370e-02 -4.66898531e-02 -6.91400468e-01 -8.37728500e-01 5.51737547e-02 -5.84801555e-01 -1.04421631e-01 6.53366864e-01 7.65887797e-01 1.87878579e-01 -7.77252838e-02 5.53142309e-01 -1.03704822e+00 1.24176633e+00 -6.80620551e-01 -6.33692920e-01 1.89223990e-01 -8.89253378e-01 2.72963792e-01 8.75177026e-01 -3.48662257e-01 -4.51075941e-01 -9.08515155e-02 6.05341077e-01 -2.90107757e-01 2.99377561e-01 6.51675045e-01 7.04030633e-01 -4.30891842e-01 9.25398648e-01 -3.89583521e-02 4.11473393e-01 -3.79296750e-01 3.83914828e-01 3.63097399e-01 -1.59685940e-01 -7.77223825e-01 7.03226328e-01 4.73675549e-01 2.58143663e-01 -9.50688422e-01 -5.80578208e-01 -1.89988703e-01 -4.40575093e-01 -3.75875264e-01 2.95326054e-01 -1.02720225e+00 -1.11427379e+00 2.80480444e-01 -1.10070038e+00 -1.04067671e+00 -4.13541615e-01 6.67650044e-01 -5.01608729e-01 2.58817881e-01 -8.91227722e-01 -8.43056440e-01 -2.32002988e-01 -7.49214828e-01 8.24987054e-01 2.15492085e-01 9.41225663e-02 -1.69630206e+00 2.83050030e-01 2.24614009e-01 7.17079937e-01 -1.63807906e-02 2.73572594e-01 -4.96995866e-01 -7.05639005e-01 1.36256978e-01 -9.41840336e-02 8.15884769e-02 -2.10054398e-01 -1.89262658e-01 -5.73707223e-01 -8.20734799e-01 -6.50057942e-02 -4.76356149e-01 8.46481502e-01 6.69909656e-01 1.06744421e+00 -7.69279182e-01 -3.61991376e-01 6.96391821e-01 1.36162055e+00 -5.95946610e-01 -1.84191197e-01 -2.04224442e-03 7.53952265e-01 6.72124147e-01 -1.79695711e-02 6.06494665e-01 6.51755929e-01 2.46575654e-01 6.15204632e-01 -8.64275619e-02 -6.01025596e-02 -1.98802084e-01 4.07810479e-01 1.09944475e+00 -2.84422278e-01 -3.45120877e-01 -1.19556224e+00 4.22460377e-01 -2.06736326e+00 -5.83007932e-01 -3.82337004e-01 2.07340884e+00 8.40933204e-01 -1.41840437e-02 3.42678547e-01 -4.68728751e-01 8.05207074e-01 2.57181108e-01 -6.57414019e-01 -1.68667465e-01 -2.74727922e-02 -2.16528103e-01 1.10203767e+00 1.08667731e+00 -5.60205579e-01 7.09926844e-01 6.42596340e+00 7.11385727e-01 -9.60313797e-01 5.30355871e-01 9.57524240e-01 -6.05678916e-01 -3.86487573e-01 -2.59782374e-02 -3.87966245e-01 3.12345296e-01 9.71683860e-01 -5.06313264e-01 1.26358366e+00 6.39253259e-01 2.90041596e-01 8.32843632e-02 -1.21263850e+00 1.04140997e+00 -1.30897254e-01 -1.56160760e+00 -4.83421892e-01 6.37464941e-01 1.15919113e+00 8.17378879e-01 7.96754807e-02 -7.83020630e-02 9.42665994e-01 -1.16865730e+00 4.61108267e-01 5.12739718e-01 3.99939597e-01 -6.37880862e-01 4.67743754e-01 7.52852380e-01 -8.15806329e-01 -2.46759489e-01 -5.36053777e-01 -4.84653085e-01 -1.61770150e-01 8.69383216e-01 -7.02818155e-01 1.06821619e-01 5.92720032e-01 8.45837176e-01 -6.00194037e-01 8.24964941e-01 1.20964684e-01 6.92257941e-01 -6.51004553e-01 -4.71807361e-01 2.97369927e-01 -4.70168591e-01 6.35103762e-01 1.05948722e+00 2.39418577e-02 -1.34170102e-03 2.26105973e-01 7.21435666e-01 -5.91879904e-01 1.46968290e-01 -7.94491708e-01 -4.34165746e-01 3.90025884e-01 1.16107559e+00 -7.35995233e-01 -1.08063065e-01 -4.40465748e-01 6.48293197e-01 9.68459487e-01 6.16404355e-01 -3.40963751e-01 1.01276310e-02 5.99838138e-01 2.33605042e-01 -1.31745571e-02 -6.46540940e-01 -2.74758130e-01 -1.24031568e+00 -2.82519441e-02 -6.96873128e-01 2.55758405e-01 -4.60350364e-02 -1.47192621e+00 9.14247334e-02 -3.21273059e-01 -4.89862770e-01 -5.75438514e-02 -3.91631901e-01 -6.02458239e-01 5.75697660e-01 -1.25542259e+00 -7.29530871e-01 -5.24244793e-02 5.95694482e-01 2.88991451e-01 -1.80259153e-01 2.88272291e-01 1.43837988e-01 -4.90455568e-01 4.44639325e-01 5.93647897e-01 4.55870703e-02 4.90265429e-01 -1.55007792e+00 2.88391054e-01 6.25268698e-01 4.96192932e-01 5.65438449e-01 8.08305740e-01 -4.54569221e-01 -1.96318209e+00 -8.48590851e-01 6.08983338e-01 -5.82372248e-01 1.02786660e+00 -4.02081430e-01 -5.21612346e-01 8.10700834e-01 3.26462299e-01 2.50877440e-01 3.45190734e-01 4.77843344e-01 -6.18390031e-02 -2.37003535e-01 -7.71438599e-01 3.21029663e-01 1.26472676e+00 -3.67268056e-01 3.46755207e-01 9.44478750e-01 6.35137975e-01 -3.80149931e-01 -9.61172223e-01 -2.81930536e-01 2.33843461e-01 -9.65059519e-01 6.61522150e-01 -5.83388686e-01 2.19687641e-01 3.25361639e-01 1.67554826e-01 -1.62254417e+00 -1.72096252e-01 -1.15922415e+00 -4.43556905e-01 5.52517176e-01 5.46879709e-01 -1.23250830e+00 1.10554087e+00 3.10872108e-01 1.23626322e-01 -9.32016909e-01 -9.01364267e-01 -7.56431520e-01 4.73250389e-01 -1.40526062e-02 1.70712546e-01 1.18525875e+00 1.27496272e-01 3.18196476e-01 -1.80675060e-01 1.09406903e-01 9.51824069e-01 -1.53312340e-01 9.09063935e-01 -1.19377279e+00 -7.01123953e-01 -8.90465796e-01 -2.22079784e-01 -1.38057506e+00 1.20725170e-01 -1.10098112e+00 -3.00620705e-01 -1.42947578e+00 2.14680746e-01 -7.20219195e-01 -6.79920539e-02 2.81537741e-01 1.08665086e-01 2.51233261e-02 3.65758419e-01 5.37567019e-01 -9.92942452e-01 3.70673209e-01 1.37346923e+00 -3.75107452e-02 -1.56573623e-01 1.57086879e-01 -8.50163460e-01 5.05429447e-01 1.11985755e+00 -6.36519670e-01 -4.46260929e-01 -8.96858215e-01 8.22231174e-01 1.00024246e-01 3.93987745e-01 -6.73824668e-01 7.70851493e-01 -1.80078834e-01 1.38525710e-01 2.86159515e-01 6.00934513e-02 -6.24906361e-01 -4.12847735e-02 8.20828497e-01 -6.75926208e-01 3.93001139e-01 -2.80166656e-01 9.47903574e-01 2.61501461e-01 -5.80687597e-02 6.44185305e-01 -2.56397069e-01 1.01690374e-01 4.97568399e-01 -9.04263034e-02 6.08287454e-01 8.73076797e-01 2.13455051e-01 -4.87040460e-01 -9.53566134e-01 -6.28879130e-01 6.30226135e-01 6.76300883e-01 -1.19701941e-02 3.12811762e-01 -9.18747783e-01 -1.01729357e+00 -1.76222607e-01 -5.87484479e-01 1.86374381e-01 6.20802231e-02 1.31744027e+00 -8.73702049e-01 1.20541312e-01 3.86158645e-01 -5.67579567e-01 -9.07349885e-01 3.01592320e-01 7.92989075e-01 -2.94199437e-01 -4.77977872e-01 1.08384752e+00 1.17385164e-01 -4.25316244e-01 3.72804582e-01 -9.85441655e-02 3.85259241e-01 -1.98157758e-01 1.03823662e-01 5.08950055e-01 -2.71184117e-01 -6.60082474e-02 -1.60187110e-01 2.94151485e-01 2.79908013e-02 -1.42307311e-01 1.50299299e+00 -3.57020885e-01 -3.50851685e-01 5.02417803e-01 1.39829004e+00 -3.75130470e-03 -1.42694068e+00 -3.89424324e-01 -1.96866900e-01 -4.19007659e-01 3.05445552e-01 -2.36691296e-01 -1.47867584e+00 7.72628546e-01 1.54367745e-01 8.64768326e-01 4.83380049e-01 3.44374448e-01 4.72804099e-01 7.26958811e-01 4.03639287e-01 -1.13952947e+00 2.66572624e-01 4.75992471e-01 8.20567071e-01 -1.24888062e+00 2.08920330e-01 -4.56196479e-02 -4.24420208e-01 1.15216565e+00 4.11594898e-01 -3.78150195e-01 9.62437212e-01 2.44403899e-01 -2.10837275e-01 -6.43710375e-01 -1.08849525e+00 1.03153318e-01 -8.88056755e-02 4.76885527e-01 4.06355023e-01 1.78413570e-01 -1.83981687e-01 -1.16949417e-01 -4.66383934e-01 -4.69567299e-01 6.01093829e-01 4.45013165e-01 -5.92820168e-01 -7.82910407e-01 -1.39444485e-01 7.78130770e-01 -5.04740238e-01 3.06655560e-03 -7.45974123e-01 6.66345239e-01 -3.80020827e-01 9.17085230e-01 7.79185593e-02 -4.36278023e-02 -2.34631702e-01 -5.46135485e-01 6.25898778e-01 -6.75759435e-01 -6.57165885e-01 -1.47505507e-01 -2.39009652e-02 -7.04914808e-01 -3.63632470e-01 -5.81717908e-01 -9.84319985e-01 -1.06571889e+00 -2.48962447e-01 4.87804413e-01 7.33891487e-01 6.98587537e-01 5.47391117e-01 2.82823682e-01 7.45405257e-01 -9.13166940e-01 -9.66144204e-01 -7.53959537e-01 -8.33540857e-01 9.34927836e-02 7.10041583e-01 -5.87068647e-02 -7.68524408e-01 -3.12473893e-01]
[6.445381164550781, 5.213738918304443]
dbf234f2-0114-4b92-856d-7d6defa4a7af
gandlf-a-generally-nuanced-deep-learning
2103.01006
null
https://arxiv.org/abs/2103.01006v4
https://arxiv.org/pdf/2103.01006v4.pdf
GaNDLF: A Generally Nuanced Deep Learning Framework for Scalable End-to-End Clinical Workflows in Medical Imaging
Deep Learning (DL) has the potential to optimize machine learning in both the scientific and clinical communities. However, greater expertise is required to develop DL algorithms, and the variability of implementations hinders their reproducibility, translation, and deployment. Here we present the community-driven Generally Nuanced Deep Learning Framework (GaNDLF), with the goal of lowering these barriers. GaNDLF makes the mechanism of DL development, training, and inference more stable, reproducible, interpretable, and scalable, without requiring an extensive technical background. GaNDLF aims to provide an end-to-end solution for all DL-related tasks in computational precision medicine. We demonstrate the ability of GaNDLF to analyze both radiology and histology images, with built-in support for k-fold cross-validation, data augmentation, multiple modalities and output classes. Our quantitative performance evaluation on numerous use cases, anatomies, and computational tasks supports GaNDLF as a robust application framework for deployment in clinical workflows.
['Peter Mattson', 'Renato Umeton', 'Alexandros Karargyris', 'Prashant Shah', 'Yong Fan', 'Joel H. Saltz', 'Sezgin Er', 'Aimilia Gastounioti', 'Tahsin M. Kurc', 'Shahira Abousamra', 'Rhea Chitalia', 'Babak Haghighi', 'Yuemeng Li', 'Chunrui Zou', 'Vinayak Ahluwalia', 'Ravi Panchumarthy', 'Deepthi Karkada', 'Junwen Wu', 'Alexander Getka', 'Camila González', 'David Lang', 'Sofia Mouchtaris', 'Orhun Güley', 'Bhakti Baheti', 'Ujjwal Baid', 'İbrahim Ethem Hamamcı', 'Sotirios A. Tsaftaris', 'Anirban Mukhopadhyay', 'Vishnu Bashyam', 'Karol Gotkowski', 'Spyridon Thermos', 'Spyridon Bakas', 'Christos Davatzikos', 'Despina Kontos', 'Bjoern Menze', 'Mark Bergman', 'Micah Sheller', 'Brandon Edwards', 'Caleb Grenko', 'Megh Bhalerao', 'Siddhesh P. Thakur', 'Sarthak Pati']
2021-02-26
null
null
null
null
['unet-segmentation']
['computer-vision']
[-4.95089814e-02 2.26106048e-01 -2.74660796e-01 -6.15744948e-01 -1.12287295e+00 -5.16923070e-01 1.71700731e-01 4.91488338e-01 -2.89888918e-01 5.66557646e-01 3.37731868e-01 -6.56806231e-01 -2.69653827e-01 -4.78139579e-01 -4.96272773e-01 -5.04690111e-01 -1.26092404e-01 6.80296004e-01 -3.97462249e-01 2.58066863e-01 -2.86969572e-01 7.80751646e-01 -7.18657076e-01 7.16971159e-01 5.06232858e-01 7.22922802e-01 2.45397091e-02 8.94635022e-01 6.35085702e-02 1.01452076e+00 -4.64468807e-01 -3.34142238e-01 2.08698198e-01 -2.04577222e-01 -8.65144789e-01 -1.22753032e-01 4.54036325e-01 -5.12955308e-01 -2.40338128e-02 2.80388385e-01 1.15366554e+00 -4.51806039e-01 4.20929521e-01 -1.06968212e+00 -5.52444100e-01 5.47514796e-01 -2.64631957e-01 2.30746225e-01 -2.25011021e-01 8.26819301e-01 5.51442087e-01 -4.84459490e-01 8.18950474e-01 8.41136336e-01 1.14695942e+00 7.97607064e-01 -1.37753642e+00 -4.57825094e-01 -3.05210441e-01 -3.23367476e-01 -1.04123068e+00 -3.50267708e-01 7.67373592e-02 -7.52043307e-01 9.92310047e-01 1.33457541e-01 7.00168073e-01 1.09372723e+00 7.92100847e-01 8.23482513e-01 9.58872378e-01 -2.26150185e-01 3.21820378e-01 -6.95622936e-02 4.74161468e-02 9.48254526e-01 3.44064325e-01 -1.46536767e-01 -3.67900491e-01 -3.90755922e-01 8.17854285e-01 1.03551954e-01 -1.99274480e-01 -1.15921117e-01 -1.22123230e+00 5.22180855e-01 2.25908130e-01 2.11068109e-01 -5.10495782e-01 3.79251242e-01 6.65110409e-01 7.17872679e-02 4.17153448e-01 6.71847284e-01 -6.01031959e-01 -3.07281673e-01 -9.62775290e-01 1.10919312e-01 7.23360896e-01 6.87558115e-01 1.22349493e-01 -1.11022547e-01 -4.68967825e-01 6.72424316e-01 2.82631606e-01 3.28221470e-01 4.06156301e-01 -1.14265823e+00 -5.42525910e-02 6.17830038e-01 -5.34043014e-01 -5.57623565e-01 -8.36473107e-01 -8.78195107e-01 -1.04717076e+00 4.23781097e-01 1.76334575e-01 -2.83036679e-01 -1.01084340e+00 1.64623702e+00 2.60279506e-01 -1.09297641e-01 -1.23993054e-01 6.45926416e-01 1.04978657e+00 2.45745201e-02 5.28960288e-01 1.26863331e-01 1.20615125e+00 -8.57248008e-01 -5.17499864e-01 -2.15418130e-01 1.10183084e+00 -6.53621078e-01 1.36389017e+00 5.06127954e-01 -1.40939510e+00 -2.64125407e-01 -7.83968031e-01 -4.37393785e-01 7.82249942e-02 3.50627929e-01 7.65508175e-01 6.25806153e-01 -1.28083372e+00 5.03873050e-01 -1.47604406e+00 -3.35190386e-01 9.28012550e-01 4.43785101e-01 -6.00469232e-01 -2.45210379e-01 -4.10995901e-01 9.27290142e-01 -9.26425457e-02 8.29631314e-02 -1.04100084e+00 -1.40927994e+00 -5.78288436e-01 -5.20051681e-02 -1.89313740e-01 -1.49883056e+00 1.44491589e+00 -4.47595835e-01 -1.05903769e+00 1.28327000e+00 1.13726199e-01 -4.60698843e-01 7.37384796e-01 -7.12434128e-02 -1.57186329e-01 -9.76002887e-02 1.35046899e-01 6.05687261e-01 2.12221637e-01 -7.26407647e-01 -2.26877600e-01 -4.53818291e-01 -3.09637278e-01 2.49161869e-02 -1.10212624e-01 -2.43431613e-01 -4.38359678e-01 -5.14595866e-01 -3.00800353e-01 -9.07156527e-01 -6.08621657e-01 5.30754089e-01 -1.67383075e-01 9.85067934e-02 4.85259324e-01 -8.11098933e-01 1.07950008e+00 -2.29886842e+00 -1.88662946e-01 1.49257988e-01 8.48735631e-01 2.76788622e-01 -1.90439913e-02 1.71536058e-01 -1.84559077e-02 4.02689517e-01 -1.28709614e-01 -5.52339077e-01 -3.71740073e-01 2.53926784e-01 2.75900662e-01 5.43929458e-01 1.48280993e-01 1.30511904e+00 -9.44259286e-01 -6.65848851e-01 3.20152968e-01 5.45152485e-01 -8.24810624e-01 2.18776166e-01 -1.16541050e-01 8.00679147e-01 -4.61532623e-01 8.09920907e-01 3.65774989e-01 -9.71003413e-01 4.94522244e-01 -3.97139043e-01 1.09328419e-01 2.46529862e-01 -6.57208800e-01 2.21916914e+00 -6.78213894e-01 5.31187952e-01 3.98961186e-01 -5.34879327e-01 5.99297762e-01 2.60875046e-01 8.60571444e-01 -4.49544340e-01 2.46385023e-01 2.36261934e-01 1.69196129e-01 -7.06494093e-01 -1.44550160e-01 -2.52757818e-01 1.59822613e-01 5.98733008e-01 2.78309047e-01 -7.64575452e-02 -1.04112104e-02 2.65886009e-01 1.60387790e+00 -1.22721367e-01 2.68196315e-01 -3.31125706e-01 2.38073729e-02 7.06942976e-02 3.83783102e-01 8.68054748e-01 -2.70539582e-01 6.91415429e-01 3.68401408e-01 -6.39752865e-01 -9.81624424e-01 -1.11248112e+00 -3.42320859e-01 8.34759831e-01 -6.08701110e-01 -3.90174687e-01 -5.97442865e-01 -8.54545772e-01 2.05199108e-01 5.08372545e-01 -7.07427084e-01 -1.29761502e-01 -4.00229305e-01 -8.60465169e-01 8.44122827e-01 7.28826582e-01 7.79243261e-02 -9.04031694e-01 -6.07295692e-01 3.12452018e-01 -9.46542099e-02 -1.09245276e+00 -2.53222495e-01 1.32753670e-01 -1.15080523e+00 -1.09265339e+00 -4.32839811e-01 -5.56440949e-01 6.50767565e-01 -2.59234697e-01 1.38169861e+00 5.26141487e-02 -9.42336857e-01 6.51130855e-01 4.38048355e-02 -7.02899933e-01 -7.73240983e-01 2.53259808e-01 -2.82873452e-01 -7.07931757e-01 8.20754841e-02 -5.25051832e-01 -9.41976011e-01 -1.65517420e-01 -8.53393018e-01 2.38918573e-01 1.01012409e+00 1.00509167e+00 8.19166481e-01 -6.21118009e-01 4.19133008e-01 -1.18798089e+00 7.29206324e-01 -5.01407385e-01 -8.93758386e-02 3.58890623e-01 -1.10196829e+00 8.94071758e-02 4.07384098e-01 6.02370054e-02 -6.83572471e-01 -1.44564778e-01 -4.72176909e-01 -1.69444680e-01 -5.45420982e-02 7.28635728e-01 2.57524878e-01 -7.98955336e-02 1.17489672e+00 -1.29349753e-01 6.00390494e-01 -4.57677424e-01 3.07197034e-01 4.38984513e-01 5.32074988e-01 -7.65490711e-01 4.48217280e-02 4.05892551e-01 1.36221558e-01 -4.19020504e-01 -6.60192430e-01 -1.14099778e-01 -3.44523042e-01 -7.39625022e-02 7.08585143e-01 -9.45944905e-01 -5.62821627e-01 2.40651175e-01 -9.00774539e-01 -5.83230197e-01 -5.61251402e-01 3.79289091e-01 -3.35846603e-01 6.86025545e-02 -9.09919202e-01 -1.46296918e-01 -1.16796327e+00 -1.46197724e+00 1.23194087e+00 -3.30414474e-02 -6.61729097e-01 -1.30914700e+00 1.46074817e-01 4.65452611e-01 7.01820731e-01 6.63198233e-01 1.26666117e+00 -6.34988666e-01 -4.72152680e-01 -3.37489486e-01 -9.84503925e-02 6.25503421e-01 2.83165962e-01 1.25553623e-01 -9.11353171e-01 -3.31071824e-01 -2.97678232e-01 -5.20636261e-01 3.28460991e-01 7.17080057e-01 1.75312138e+00 -1.40880734e-01 -3.86560887e-01 9.85025465e-01 1.19111729e+00 -1.74223498e-01 4.79028702e-01 2.09789366e-01 4.46198970e-01 4.14274028e-03 1.22479416e-01 3.63245845e-01 2.62252808e-01 1.17269315e-01 2.33797804e-01 -7.60529935e-01 -3.71505171e-01 -3.38894967e-03 -2.10497111e-01 4.78782207e-01 -1.30012445e-02 -8.98523629e-02 -1.45218599e+00 4.13006663e-01 -1.59058273e+00 -5.03690481e-01 -1.21990228e-02 1.92208481e+00 9.61780727e-01 5.16763404e-02 -1.22273989e-01 -3.94032121e-01 1.48651659e-01 -3.23683590e-01 -6.78424835e-01 -3.43533009e-01 2.33448923e-01 6.82621777e-01 3.73457551e-01 2.27996603e-01 -6.40055716e-01 5.22814214e-01 7.58354616e+00 5.93542337e-01 -1.53258121e+00 4.16473895e-01 1.11732042e+00 -3.25551182e-01 -2.93454260e-01 -4.64171767e-01 -3.93963695e-01 2.28925213e-01 1.02893162e+00 -2.47245312e-01 7.17474222e-02 1.03051090e+00 4.98408616e-01 2.39410475e-01 -1.36653364e+00 7.95908451e-01 -2.31334016e-01 -2.02582884e+00 -1.39859870e-01 3.80254276e-02 4.86760497e-01 4.85756367e-01 2.12660715e-01 2.88418949e-01 4.50208545e-01 -1.40151811e+00 1.26557261e-01 3.04758638e-01 1.28259933e+00 -2.35842049e-01 8.93596351e-01 3.47883739e-02 -4.44381446e-01 2.56204635e-01 1.98955968e-01 3.39823693e-01 2.89712697e-02 8.40878427e-01 -1.60289097e+00 4.40085292e-01 6.60533607e-01 7.40802824e-01 -5.98009348e-01 8.17084670e-01 6.32232726e-02 8.26875150e-01 -1.50819555e-01 1.93489149e-01 3.31883989e-02 3.36435080e-01 1.08776227e-01 1.57078183e+00 -1.88664813e-02 -9.30135250e-02 2.40513414e-01 7.71554649e-01 -1.61380813e-01 1.31185487e-01 -2.10014671e-01 -3.47749799e-01 2.60988951e-01 1.63231492e+00 -6.14849567e-01 -2.13121489e-01 -1.74630463e-01 5.38779259e-01 1.98174208e-01 -3.26498114e-02 -7.65734494e-01 1.82339743e-01 8.19234610e-01 5.01074493e-01 -5.08143604e-01 -1.81666896e-01 -9.62471843e-01 -8.80981266e-01 -7.12571666e-02 -1.29238224e+00 6.59662962e-01 -5.90323269e-01 -1.43669343e+00 5.46077907e-01 -3.28422368e-01 -1.04551458e+00 -3.90162498e-01 -7.75982440e-01 -3.07049036e-01 9.87020254e-01 -1.26122940e+00 -1.51115870e+00 -6.36293590e-01 4.70182955e-01 5.63013442e-02 -1.82456136e-01 1.23476624e+00 4.90704089e-01 -3.72155011e-01 8.37218046e-01 6.20390698e-02 1.89231247e-01 8.74768972e-01 -1.28428495e+00 2.76736856e-01 3.55414778e-01 -5.24309635e-01 9.99867201e-01 3.92276257e-01 -5.04754424e-01 -1.52614963e+00 -1.30186915e+00 4.47076082e-01 -7.60984242e-01 4.98636007e-01 -1.34804249e-01 -4.57152396e-01 9.28466856e-01 -1.24555580e-01 2.78794736e-01 1.47914767e+00 4.90824342e-01 -1.76735803e-01 -2.33903885e-01 -1.47369397e+00 3.84065479e-01 8.37362349e-01 -3.90249103e-01 -2.13931143e-01 6.14952803e-01 4.74645376e-01 -8.00730765e-01 -1.31616390e+00 7.07498193e-01 7.29039848e-01 -6.48738861e-01 8.45027626e-01 -8.90847445e-01 8.23360741e-01 -6.75640479e-02 2.42876798e-01 -1.06895387e+00 -4.31003839e-01 -3.53951961e-01 -1.30892515e-01 8.07578266e-01 5.90622187e-01 -4.49479073e-01 8.87272060e-01 8.12394738e-01 -6.94546342e-01 -1.23325109e+00 -6.49232864e-01 -2.24913180e-01 2.83940345e-01 -5.24934530e-01 3.00535500e-01 9.64756191e-01 -1.87798887e-01 1.48494840e-01 -3.21032368e-02 1.06804594e-02 4.74772185e-01 -8.56698006e-02 9.66824710e-01 -9.30302382e-01 -7.13561237e-01 -4.52801973e-01 -5.40050685e-01 -4.59859580e-01 -3.14482003e-01 -1.29421997e+00 -2.98183322e-01 -2.02539158e+00 3.01352948e-01 -7.90366352e-01 -3.58686209e-01 1.07421792e+00 -1.40363753e-01 1.29640892e-01 -1.10377744e-01 1.27796367e-01 -3.51770818e-01 6.34450838e-02 1.35064459e+00 -1.78116143e-01 -5.97186908e-02 -1.93191186e-01 -1.09137988e+00 4.26351637e-01 6.18691862e-01 -3.51276249e-01 -3.52200210e-01 -8.27937782e-01 4.78064902e-02 -1.92368925e-01 3.71943980e-01 -1.05074644e+00 1.57668263e-01 2.18340829e-02 7.60974944e-01 -1.34663984e-01 -2.06045918e-02 -5.02489924e-01 4.19443667e-01 8.71143937e-01 -4.46978599e-01 1.65272713e-01 5.61539888e-01 7.67357126e-02 6.17494956e-02 3.55093181e-01 9.92769599e-01 -3.75942677e-01 -2.13201389e-01 5.63215971e-01 -1.96771279e-01 2.78806746e-01 8.85579228e-01 9.78734419e-02 -4.58404839e-01 -1.61132336e-01 -9.53262866e-01 4.66529846e-01 3.69469464e-01 1.84861526e-01 3.78224015e-01 -8.85229290e-01 -1.01665866e+00 1.88845709e-01 1.82046980e-01 4.65200931e-01 6.55550241e-01 1.03311956e+00 -1.04887295e+00 1.29907027e-01 -1.82942271e-01 -9.87709939e-01 -1.12578058e+00 1.89150959e-01 7.34332442e-01 -7.11077809e-01 -7.56964862e-01 8.56308877e-01 -3.95896323e-02 -7.33554065e-01 1.99121863e-01 -3.84117723e-01 5.61312139e-01 -5.72219908e-01 5.67442834e-01 1.66395754e-01 7.22284853e-01 4.11230594e-01 -4.56692994e-01 -4.06015888e-02 -4.48771358e-01 1.66327417e-01 1.57553089e+00 3.47008437e-01 -1.40902936e-01 2.15188786e-01 1.03506804e+00 -1.75046876e-01 -7.45482087e-01 9.46048573e-02 -3.37126464e-01 -5.05601913e-02 3.51845086e-01 -1.41909611e+00 -1.11352646e+00 8.96334469e-01 7.74622500e-01 -3.81907344e-01 1.05302095e+00 -5.79406992e-02 7.94187903e-01 5.21531962e-02 3.39151084e-01 -7.49215841e-01 -7.09639564e-02 7.52079189e-02 7.10975826e-01 -1.08063900e+00 2.29952618e-01 -2.90834814e-01 -4.61722791e-01 1.15018606e+00 6.40140057e-01 2.77081668e-01 6.85078561e-01 8.54232788e-01 3.69288057e-01 -3.52644801e-01 -8.13071609e-01 3.97188425e-01 1.27875820e-01 4.95839804e-01 1.02428043e+00 9.26406085e-02 -3.15210372e-01 5.76099455e-01 -1.31068304e-01 6.83765709e-01 2.42280304e-01 1.25818598e+00 1.96594998e-01 -1.13454318e+00 5.83310463e-02 7.59777665e-01 -6.53445005e-01 -2.91910827e-01 -5.97191863e-02 7.41508961e-01 1.48537904e-01 4.55484241e-01 -2.67188191e-01 -2.39794955e-01 3.25499803e-01 -6.64381981e-02 5.83594739e-01 -7.87764072e-01 -9.25981283e-01 -3.33034899e-03 -1.57988500e-02 -7.12035060e-01 3.80114280e-02 -5.06606579e-01 -1.28164947e+00 -4.35419381e-01 2.29638353e-01 -2.05950484e-01 9.09591317e-01 8.07215035e-01 1.13008094e+00 1.00492394e+00 2.14464411e-01 -2.51358122e-01 -5.28781414e-01 -7.95968235e-01 -2.43606478e-01 2.62730002e-01 2.45902359e-01 -1.59319267e-01 -6.64104894e-02 1.82301477e-01]
[14.834442138671875, -2.428736925125122]
429288c9-717a-4be3-a4c4-ea9b01408365
ju_cse_nlp-multi-grade-classification-of
null
null
https://aclanthology.org/S12-1083
https://aclanthology.org/S12-1083.pdf
JU\_CSE\_NLP: Multi-grade Classification of Semantic Similarity between Text Pairs
null
['er', 'B', 'Alex Gelbukh', 'Snehasis Neogi', 'Partha Pakray', 'Sivaji yopadhyay']
2012-07-01
null
null
null
semeval-2012-7
['video-description']
['computer-vision']
[-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]
[-7.315940856933594, 3.831088066101074]
55ffeab7-d31f-45d0-991b-905ce76d5a66
distributed-contrastive-learning-for-medical
2208.03808
null
https://arxiv.org/abs/2208.03808v1
https://arxiv.org/pdf/2208.03808v1.pdf
Distributed Contrastive Learning for Medical Image Segmentation
Supervised deep learning needs a large amount of labeled data to achieve high performance. However, in medical imaging analysis, each site may only have a limited amount of data and labels, which makes learning ineffective. Federated learning (FL) can learn a shared model from decentralized data. But traditional FL requires fully-labeled data for training, which is very expensive to obtain. Self-supervised contrastive learning (CL) can learn from unlabeled data for pre-training, followed by fine-tuning with limited annotations. However, when adopting CL in FL, the limited data diversity on each site makes federated contrastive learning (FCL) ineffective. In this work, we propose two federated self-supervised learning frameworks for volumetric medical image segmentation with limited annotations. The first one features high accuracy and fits high-performance servers with high-speed connections. The second one features lower communication costs, suitable for mobile devices. In the first framework, features are exchanged during FCL to provide diverse contrastive data to each site for effective local CL while keeping raw data private. Global structural matching aligns local and remote features for a unified feature space among different sites. In the second framework, to reduce the communication cost for feature exchanging, we propose an optimized method FCLOpt that does not rely on negative samples. To reduce the communications of model download, we propose the predictive target network update (PTNU) that predicts the parameters of the target network. Based on PTNU, we propose the distance prediction (DP) to remove most of the uploads of the target network. Experiments on a cardiac MRI dataset show the proposed two frameworks substantially improve the segmentation and generalization performance compared with state-of-the-art techniques.
['Jingtong Hu', 'Yiyu Shi', 'Zhepeng Wang', 'Dewen Zeng', 'Yawen Wu']
2022-08-07
null
null
null
null
['volumetric-medical-image-segmentation']
['medical']
[-1.68522701e-01 6.27183774e-03 -4.31558907e-01 -7.16288090e-01 -7.87120759e-01 -4.12991196e-01 -2.03913331e-01 1.39604464e-01 -4.66086358e-01 7.82274008e-01 -1.87235922e-01 -3.32456231e-01 -2.55828857e-01 -8.73528242e-01 -6.46718383e-01 -1.02049375e+00 -3.56387272e-02 5.76188684e-01 4.92244363e-01 4.41251665e-01 -3.18040490e-01 5.60018003e-01 -1.04109609e+00 3.57934117e-01 8.91283453e-01 1.35010779e+00 6.24308586e-01 2.10626960e-01 -3.58904064e-01 9.17056799e-01 -3.52568537e-01 1.55268395e-02 3.94153684e-01 -1.85931489e-01 -8.80049765e-01 -1.67740285e-01 8.99770558e-02 -6.76801980e-01 -3.16225678e-01 1.02172554e+00 8.30797255e-01 -2.30181187e-01 1.31213456e-01 -1.28248203e+00 2.82429326e-02 7.20877767e-01 -3.52416009e-01 2.99835593e-01 -3.52548629e-01 1.44232750e-01 5.15377223e-01 -3.89320910e-01 8.25845838e-01 5.67984104e-01 8.06624293e-01 5.60479105e-01 -9.12594199e-01 -8.68999422e-01 -2.94920020e-02 2.22520113e-01 -1.27465725e+00 -2.67239958e-01 7.07261741e-01 -1.28548115e-01 4.13395286e-01 4.27671999e-01 7.01507151e-01 4.67733026e-01 1.79504558e-01 7.33048379e-01 1.07740140e+00 -1.09851204e-01 4.62265939e-01 1.61211684e-01 2.08592474e-01 7.83217847e-01 2.04925984e-01 -3.08229655e-01 -5.71117043e-01 -5.96692741e-01 8.50443661e-01 5.41442692e-01 -3.84424239e-01 -6.23408675e-01 -1.14326560e+00 5.96055686e-01 6.05465829e-01 3.18026334e-01 -4.72958624e-01 -5.84728234e-02 7.99421430e-01 3.44998240e-01 4.78428692e-01 -3.00963432e-01 -8.34593117e-01 7.75109082e-02 -9.70690787e-01 -4.51759249e-01 7.36837924e-01 8.65995288e-01 1.12408197e+00 -4.87187147e-01 -1.07665703e-01 7.79014051e-01 1.52403027e-01 5.62144101e-01 9.05948043e-01 -8.29657435e-01 4.63482738e-01 7.08047807e-01 -2.45669603e-01 -8.71994019e-01 -6.91676974e-01 -5.27011991e-01 -1.22351551e+00 -2.30873153e-01 1.40923634e-01 -4.65756267e-01 -5.11705995e-01 1.61296618e+00 9.02401924e-01 2.52716988e-01 -1.34847328e-01 1.01260412e+00 7.25851893e-01 4.18588340e-01 -1.28733903e-01 -5.47744334e-01 1.03746426e+00 -1.19900322e+00 -5.11342704e-01 1.85110390e-01 1.16237700e+00 -4.90273446e-01 8.58619869e-01 2.80818164e-01 -7.55826712e-01 -2.16943473e-01 -8.37374151e-01 1.78682491e-01 1.98864955e-02 1.07954212e-01 8.57377172e-01 6.90355361e-01 -1.10040760e+00 7.83737242e-01 -1.30293655e+00 -7.05199540e-02 7.86745310e-01 7.08949447e-01 -4.94858623e-01 -2.63719738e-01 -8.73565137e-01 2.52284020e-01 1.40199631e-01 4.28695232e-02 -8.30853641e-01 -8.87131631e-01 -4.21860009e-01 7.08302297e-03 2.60372400e-01 -7.19471335e-01 1.10536873e+00 -1.10332668e+00 -1.33987796e+00 5.06980956e-01 -2.72704344e-02 -3.57049465e-01 7.87906408e-01 3.45048636e-01 -2.57404894e-01 5.73867559e-01 2.62301296e-01 4.63478088e-01 6.17263019e-01 -8.85542035e-01 -6.55957758e-01 -6.41563892e-01 -3.03383857e-01 1.76238388e-01 -8.42166126e-01 -2.70377189e-01 -5.43199122e-01 -3.91563147e-01 3.63101423e-01 -8.86167943e-01 -3.78667742e-01 4.59852099e-01 -2.77606666e-01 -8.06619376e-02 1.08840609e+00 -5.63766837e-01 1.00156724e+00 -2.28948116e+00 -4.77434635e-01 6.00288033e-01 5.81638098e-01 1.13881312e-01 5.46443649e-03 -1.07086733e-01 1.07489020e-01 -8.96132290e-02 -7.29146972e-02 -2.47769713e-01 -4.63608682e-01 3.75063151e-01 2.97460586e-01 6.19763076e-01 -4.98264879e-01 7.40633547e-01 -6.83458090e-01 -1.08731771e+00 -1.33162037e-01 2.64445603e-01 -6.44371450e-01 1.21210054e-01 5.73246628e-02 7.92981625e-01 -9.38634574e-01 5.47236323e-01 9.50616539e-01 -8.25515568e-01 6.26832366e-01 -3.89402360e-01 7.79830590e-02 -1.60862505e-02 -1.26012564e+00 1.97033942e+00 -4.97038513e-01 -1.72059044e-01 4.51121300e-01 -1.15938377e+00 8.22009146e-01 4.45425212e-01 1.13809669e+00 -6.21690154e-01 1.17473505e-01 6.20945752e-01 -3.49180698e-01 -4.72111911e-01 -1.84709936e-01 2.33206153e-01 2.84239113e-01 8.00482094e-01 5.38182817e-02 5.22522688e-01 -3.64700079e-01 3.74898911e-01 1.40145516e+00 -2.20770776e-01 -1.31563693e-01 -3.35460186e-01 5.74102819e-01 2.86247805e-02 1.04927599e+00 7.02294111e-01 -4.01892722e-01 2.04613999e-01 2.55812138e-01 -7.32397914e-01 -7.67125845e-01 -7.28169620e-01 -1.62961498e-01 8.66565287e-01 3.62833411e-01 -3.13636333e-01 -1.00014341e+00 -1.10529339e+00 -1.01537794e-01 -1.65264815e-01 -1.76363260e-01 1.46531574e-02 -6.00557208e-01 -6.14572227e-01 3.13444018e-01 3.38194966e-01 6.96925521e-01 -9.51521575e-01 -6.55725420e-01 3.12144071e-01 -3.72964054e-01 -7.44626582e-01 -6.25065565e-01 1.24463968e-01 -1.34949505e+00 -1.16713250e+00 -6.83707416e-01 -8.48901629e-01 9.37086523e-01 4.01004970e-01 6.72805905e-01 3.80406797e-01 -4.05471355e-01 2.22281560e-01 -2.45288432e-01 -7.83303753e-02 -1.40054762e-01 4.18732792e-01 2.98174452e-02 2.16191024e-01 6.86609447e-02 -6.40145600e-01 -1.07181799e+00 5.04950762e-01 -8.42060030e-01 9.88218188e-02 4.48297620e-01 1.01351666e+00 1.08249187e+00 -9.07453746e-02 7.72166252e-01 -1.30388653e+00 3.10875066e-02 -6.09651089e-01 -4.64768887e-01 3.63873690e-01 -8.85135233e-01 -1.77449822e-01 1.02853715e+00 -3.45048696e-01 -9.65648592e-01 4.04219121e-01 2.06556365e-01 -5.95910609e-01 9.02751908e-02 4.40273732e-01 -1.13853000e-01 -3.97788376e-01 5.29110193e-01 1.98935568e-01 3.82836372e-01 -6.72421277e-01 6.09242879e-02 9.91199911e-01 1.67338237e-01 -4.37070966e-01 1.42137900e-01 6.94124758e-01 -1.57992169e-02 -3.02876055e-01 -6.36236966e-01 -4.61171359e-01 -4.46079612e-01 -2.06498370e-01 4.19214457e-01 -9.64287877e-01 -8.64438415e-01 4.37339962e-01 -7.94639409e-01 -4.13861305e-01 -4.79856968e-01 7.71114528e-01 -5.27348340e-01 4.17521894e-01 -8.64467859e-01 -2.00324476e-01 -1.01918745e+00 -1.30584323e+00 6.23446345e-01 1.89660847e-01 3.22634399e-01 -7.72427380e-01 -1.69606641e-01 5.52778304e-01 6.79832816e-01 7.34513327e-02 7.63255179e-01 -7.52529085e-01 -8.01454246e-01 -6.38120025e-02 -2.71232814e-01 3.52721602e-01 2.90203869e-01 -5.85547984e-01 -6.83820844e-01 -6.37622774e-01 3.05233210e-01 -3.17463487e-01 3.29086781e-01 4.52774256e-01 1.67608535e+00 -4.30331439e-01 -7.34795392e-01 7.92414546e-01 1.43235517e+00 -1.06500447e-01 1.96192026e-01 2.11978495e-01 6.95879161e-01 2.22138345e-01 7.87487924e-01 7.43434727e-01 3.95787537e-01 2.99715340e-01 3.08510333e-01 -3.33121151e-01 -4.88686711e-02 1.10439546e-02 -9.18837339e-02 1.31239820e+00 2.30687723e-01 3.55407059e-01 -8.26233804e-01 3.82375538e-01 -2.04607201e+00 -4.29183155e-01 1.78928208e-02 2.22154260e+00 1.01966250e+00 -2.47850820e-01 -5.24558127e-02 -2.52222300e-01 8.34687889e-01 -2.60062873e-01 -1.13853264e+00 2.43253663e-01 1.31108448e-01 1.76025197e-01 8.05851042e-01 -8.64899904e-02 -8.19120109e-01 5.16093731e-01 4.81567240e+00 1.14556527e+00 -1.52967536e+00 8.88505995e-01 9.88639474e-01 -3.82064939e-01 -1.09592803e-01 -9.00960192e-02 -5.06573439e-01 6.20231330e-01 8.22220623e-01 -1.07959993e-01 3.85721177e-01 1.15728652e+00 2.59969741e-01 6.74817385e-03 -8.64677966e-01 1.29118919e+00 -2.22804755e-01 -1.69347394e+00 -2.11011112e-01 7.92454258e-02 7.11655796e-01 5.90316176e-01 -4.42671835e-01 -1.49070606e-01 6.86630653e-03 -3.23859245e-01 3.23812932e-01 6.15177512e-01 1.04969823e+00 -7.74982512e-01 8.19882035e-01 6.61623061e-01 -1.08220375e+00 -1.56103224e-01 -5.73482573e-01 5.34765542e-01 -1.35694653e-01 9.10372078e-01 -4.79597062e-01 7.51600325e-01 1.05878973e+00 6.34236217e-01 -3.61288279e-01 9.97185290e-01 3.77122432e-01 5.40875316e-01 -6.14536464e-01 9.45103243e-02 -2.55207904e-02 -8.93876851e-02 1.46846205e-01 7.57438242e-01 1.28735647e-01 1.98028967e-01 6.77496791e-01 3.93853396e-01 -2.64922112e-01 6.36240959e-01 -1.73036858e-01 4.85738099e-01 7.69417167e-01 1.64735627e+00 -7.81185329e-01 -4.25542533e-01 -3.52380633e-01 9.36039031e-01 4.98541713e-01 -4.45306376e-02 -6.98486328e-01 -1.67275801e-01 1.33834079e-01 9.73837301e-02 9.19332206e-02 1.05335213e-01 -2.12454110e-01 -1.18818319e+00 2.93868721e-01 -7.03398764e-01 6.95443451e-01 -2.86839783e-01 -1.56724334e+00 6.81092083e-01 -4.07727540e-01 -1.34619319e+00 1.51354417e-01 1.58705357e-02 -2.11597666e-01 7.17599571e-01 -1.49438405e+00 -1.20530987e+00 -5.98839343e-01 1.31603277e+00 8.03612620e-02 -9.34973434e-02 7.86453068e-01 7.10820317e-01 -5.04484653e-01 7.75965333e-01 3.90482157e-01 1.30966365e-01 9.28814113e-01 -5.74242353e-01 -3.13134789e-01 3.67346197e-01 -2.65798002e-01 5.10423481e-01 -2.11312205e-01 -6.80815816e-01 -1.48332930e+00 -1.35199499e+00 4.80510801e-01 3.38632286e-01 3.18307728e-01 -1.09865457e-01 -7.71160185e-01 5.03676951e-01 -4.02563483e-01 9.10047472e-01 1.00085747e+00 -2.31298685e-01 -1.13018965e-02 -8.07858109e-01 -1.66277850e+00 3.92262787e-02 8.29240382e-01 -2.69284755e-01 2.64612973e-01 9.46592808e-01 7.45667636e-01 -3.92204821e-01 -1.21181023e+00 2.50222981e-01 1.98673710e-01 -8.59839082e-01 4.73883331e-01 -2.89480954e-01 -1.32413968e-01 -3.02116364e-01 4.80775014e-02 -8.63636613e-01 -2.51215875e-01 -6.19574904e-01 -1.57878287e-02 1.09577954e+00 2.92037159e-01 -1.18472624e+00 1.25348771e+00 9.20544267e-01 -9.47574377e-02 -9.14732814e-01 -1.31116378e+00 -6.97210193e-01 -2.14334115e-01 -1.48922339e-01 8.14184129e-01 1.27383101e+00 -1.62266996e-02 -2.27320865e-02 -1.26352251e-01 1.29416272e-01 8.09802413e-01 2.70399243e-01 6.55089200e-01 -1.24927092e+00 -3.99863452e-01 3.10145497e-01 -3.20471793e-01 -9.00015712e-01 3.03200204e-02 -1.09608459e+00 -3.34299892e-01 -1.37677085e+00 5.11228144e-01 -1.53192091e+00 -4.37598646e-01 1.09116936e+00 2.10413471e-01 1.63235307e-01 1.19386382e-01 7.53123283e-01 -8.29930127e-01 2.68069983e-01 1.36162663e+00 -3.74585018e-02 -1.49202153e-01 7.49907568e-02 -3.31354350e-01 4.93520468e-01 9.01670158e-01 -7.80070662e-01 -5.23100019e-01 -5.23379326e-01 -1.63376197e-01 3.62499952e-01 9.66216251e-02 -1.04722476e+00 7.29728222e-01 2.04717200e-02 4.19564843e-01 -4.16067243e-01 -4.30219285e-02 -1.19848096e+00 3.07170123e-01 6.78062975e-01 -1.26316231e-02 -2.73697108e-01 -2.34099612e-01 5.63756645e-01 -7.78464675e-02 -1.02037363e-01 7.39444673e-01 -2.21977770e-01 -1.89093441e-01 9.08171713e-01 9.28519666e-02 -2.10256353e-01 1.13310158e+00 -3.09917089e-02 -2.60554671e-01 -9.04030278e-02 -9.47810352e-01 5.54116130e-01 3.75092834e-01 -7.26132467e-03 4.01568383e-01 -1.08889759e+00 -3.65357280e-01 2.58821130e-01 -3.00824821e-01 4.03589875e-01 9.18700993e-01 1.22289252e+00 -5.79798460e-01 1.33656412e-01 -1.29874527e-01 -9.03107405e-01 -1.20002484e+00 5.74379027e-01 4.50843424e-01 -3.71758431e-01 -8.71736109e-01 4.62920129e-01 4.22230409e-03 -8.50753248e-01 1.03268825e-01 4.10728604e-02 1.47378698e-01 -2.26688206e-01 5.00457942e-01 4.09356594e-01 4.85717237e-01 -2.32434779e-01 -4.17179018e-01 1.22945383e-01 -2.13518947e-01 2.95189202e-01 1.52171612e+00 -3.46016318e-01 -5.45442700e-01 1.54604912e-01 1.40980268e+00 -2.34920651e-01 -1.22291064e+00 -4.51802105e-01 -4.59092259e-01 -3.47137064e-01 3.73015106e-01 -7.21019983e-01 -1.79634738e+00 4.88046348e-01 1.00408399e+00 -2.32779652e-01 1.32531571e+00 -1.22253641e-01 1.19400954e+00 3.09210390e-01 9.61293399e-01 -1.14104581e+00 -2.81477630e-01 -3.08570024e-02 1.35409489e-01 -1.11398375e+00 -2.85773836e-02 -5.82212448e-01 -5.35403967e-01 1.16518486e+00 6.35459900e-01 2.31427819e-01 1.04846847e+00 3.08381855e-01 1.83650047e-01 -1.78131089e-01 -5.56873024e-01 4.50873673e-01 -3.27708662e-01 3.32717210e-01 -2.96905395e-02 1.89104125e-01 -3.97616982e-01 8.22168648e-01 1.22842558e-01 4.05793756e-01 1.63213730e-01 1.16304481e+00 -2.38479719e-01 -1.24507856e+00 -1.75222456e-01 8.76461923e-01 -5.25696695e-01 1.15161976e-02 2.67623693e-01 2.47379616e-01 2.26333976e-01 8.16175222e-01 4.37559113e-02 -7.07372725e-02 1.06166750e-02 -8.46293494e-02 1.44842207e-01 -4.38357234e-01 -7.82987297e-01 2.86981881e-01 -3.59760255e-01 -9.00226951e-01 -4.19574082e-01 -4.74794954e-01 -1.62029338e+00 -2.39932701e-01 -5.60736775e-01 2.49387592e-01 8.92528117e-01 7.88758576e-01 8.06442976e-01 2.12308824e-01 1.10338199e+00 -4.79503632e-01 -7.10413277e-01 -6.75336301e-01 -9.27369773e-01 2.84094870e-01 -1.95309171e-04 -2.06536531e-01 -8.50438178e-02 -8.41518641e-02]
[6.074367046356201, 6.476650714874268]
84801c94-ce28-4e29-950a-fbce58fb061c
ada-tta-towards-adaptive-high-quality-text-to
2306.03504
null
https://arxiv.org/abs/2306.03504v1
https://arxiv.org/pdf/2306.03504v1.pdf
Ada-TTA: Towards Adaptive High-Quality Text-to-Talking Avatar Synthesis
We are interested in a novel task, namely low-resource text-to-talking avatar. Given only a few-minute-long talking person video with the audio track as the training data and arbitrary texts as the driving input, we aim to synthesize high-quality talking portrait videos corresponding to the input text. This task has broad application prospects in the digital human industry but has not been technically achieved yet due to two challenges: (1) It is challenging to mimic the timbre from out-of-domain audio for a traditional multi-speaker Text-to-Speech system. (2) It is hard to render high-fidelity and lip-synchronized talking avatars with limited training data. In this paper, we introduce Adaptive Text-to-Talking Avatar (Ada-TTA), which (1) designs a generic zero-shot multi-speaker TTS model that well disentangles the text content, timbre, and prosody; and (2) embraces recent advances in neural rendering to achieve realistic audio-driven talking face video generation. With these designs, our method overcomes the aforementioned two challenges and achieves to generate identity-preserving speech and realistic talking person video. Experiments demonstrate that our method could synthesize realistic, identity-preserving, and audio-visual synchronized talking avatar videos.
['Zhou Zhao', 'Zejun Ma', 'Xiang Yin', 'Chen Zhang', 'Jinglin Liu', 'Yi Ren', 'Ziyue Jiang', 'Zhenhui Ye']
2023-06-06
null
null
null
null
['zero-shot-multi-speaker-tts', 'video-generation', 'neural-rendering']
['audio', 'computer-vision', 'computer-vision']
[ 5.54737806e-01 -8.78122598e-02 2.20950544e-01 -1.37681276e-01 -1.20601106e+00 -3.27551931e-01 6.30955815e-01 -8.85474980e-01 2.60466129e-01 4.78976101e-01 6.14515007e-01 1.54196441e-01 2.32292667e-01 -2.70778507e-01 -6.00425184e-01 -7.92374313e-01 5.02748072e-01 3.11820477e-01 -9.78521928e-02 -4.12531734e-01 -3.48503709e-01 3.11630696e-01 -2.08985543e+00 6.38255835e-01 6.97342515e-01 1.17157590e+00 2.37658665e-01 9.47315216e-01 -4.76479530e-03 7.61815250e-01 -8.53632629e-01 -5.99794090e-01 1.39731407e-01 -8.66412938e-01 -3.77753407e-01 2.65998513e-01 7.27030337e-01 -4.95086670e-01 -5.16647041e-01 8.01656365e-01 8.68761420e-01 1.02847770e-01 5.98567247e-01 -1.57528245e+00 -5.83789945e-01 4.69142109e-01 -4.34214741e-01 -3.01080912e-01 8.06792319e-01 3.63693923e-01 7.27824926e-01 -7.12327838e-01 6.12113774e-01 1.57868052e+00 5.99269271e-01 1.11469722e+00 -1.18228745e+00 -8.99350822e-01 -2.47991711e-01 1.31304339e-01 -1.66429496e+00 -1.19618714e+00 9.11418796e-01 -3.41708094e-01 3.01494688e-01 6.22530818e-01 8.47023726e-01 1.94406986e+00 -7.14041144e-02 5.81706643e-01 8.82274330e-01 -2.43682951e-01 -1.02362402e-01 5.94761260e-02 -6.91961050e-01 1.97027281e-01 -7.76320040e-01 9.30488482e-02 -1.00374162e+00 -1.20658956e-01 6.78232491e-01 -3.70343387e-01 -6.54415786e-01 4.72206250e-02 -1.30442417e+00 5.79793870e-01 -3.09374332e-01 2.94613123e-01 -3.12643707e-01 3.11147749e-01 6.65496349e-01 3.03707808e-01 4.04356390e-01 -2.86539160e-02 3.80714864e-01 -5.30851126e-01 -1.14919627e+00 3.45336020e-01 6.63888931e-01 9.62467551e-01 3.80006768e-02 6.24499977e-01 -2.47082412e-01 1.17147374e+00 4.75308150e-02 8.76307607e-01 4.85800564e-01 -1.25767803e+00 1.83928445e-01 -2.34586060e-01 4.52297144e-02 -7.68451154e-01 8.01226869e-02 1.27571508e-01 -9.41182375e-01 2.14532822e-01 3.02610964e-01 -1.35952190e-01 -4.37980205e-01 1.90315008e+00 3.20995271e-01 3.71213168e-01 4.49966602e-02 1.08534062e+00 1.12983549e+00 9.86838102e-01 -3.18787694e-01 -6.77996635e-01 1.48578644e+00 -7.98580527e-01 -1.39823878e+00 3.49983573e-02 -2.40517616e-01 -8.67638171e-01 1.46217036e+00 4.41389769e-01 -1.31976569e+00 -8.80258262e-01 -8.85943949e-01 -2.16006100e-01 3.19715083e-01 -6.55004159e-02 1.48411587e-01 9.55203354e-01 -1.21621919e+00 2.17203587e-01 -3.07673782e-01 -1.77375287e-01 1.30432814e-01 8.62925351e-02 -3.50727260e-01 1.25542611e-01 -1.18390572e+00 4.90208089e-01 -1.41195431e-01 -3.55539061e-02 -1.13166952e+00 -7.96095014e-01 -9.73872900e-01 -1.07060345e-02 4.75894153e-01 -8.73496294e-01 1.40398753e+00 -1.45964313e+00 -2.31091452e+00 8.69824111e-01 -5.09642541e-01 -4.99723218e-02 7.78236687e-01 -1.75200060e-01 -7.88443804e-01 3.04800808e-01 -1.45693496e-01 6.53677821e-01 1.38716865e+00 -1.47702169e+00 -3.22032928e-01 -6.79452792e-02 -3.37825686e-01 4.52108979e-01 -4.96344566e-01 3.65960270e-01 -3.89915347e-01 -8.66833270e-01 -3.90751153e-01 -7.39753783e-01 5.00908911e-01 3.26090217e-01 -4.29699838e-01 1.61105424e-01 1.27006543e+00 -6.26739681e-01 9.42702770e-01 -2.37195873e+00 2.26618558e-01 -3.85560155e-01 2.63542533e-01 2.81303942e-01 -2.05397278e-01 4.40279126e-01 -3.20628062e-02 -1.69356063e-01 3.84444632e-02 -7.69276857e-01 -3.88975739e-02 -8.29101820e-03 -7.43800938e-01 4.02733117e-01 -1.02980413e-01 6.90761805e-01 -7.01007843e-01 -6.80433571e-01 3.30011517e-01 1.10001540e+00 -5.10500729e-01 3.55412036e-01 -1.12155944e-01 7.83220291e-01 -4.32149693e-02 5.51727951e-01 5.17352998e-01 3.44188958e-01 -8.19872413e-03 -3.65933359e-01 -2.00863518e-02 -1.98389776e-02 -1.10096252e+00 1.74938536e+00 -5.55411220e-01 9.68904734e-01 4.52844083e-01 -5.50575316e-01 1.02408397e+00 8.82634580e-01 5.86456418e-01 -5.99352896e-01 2.59454340e-01 1.31328642e-01 -3.44377816e-01 -8.25207770e-01 5.84004939e-01 -5.67978561e-01 -1.35582685e-01 2.05030963e-01 1.70353919e-01 -5.29377103e-01 -2.16472223e-01 -9.22798589e-02 6.36877239e-01 1.47895694e-01 -1.44015148e-01 9.22409594e-02 5.17857790e-01 -5.32777905e-01 4.99009877e-01 2.33558267e-01 -1.79855138e-01 1.11095822e+00 3.38038146e-01 -6.53393194e-03 -1.18112791e+00 -1.30854607e+00 1.90895528e-01 1.09611559e+00 3.87566872e-02 -5.08482099e-01 -9.98965561e-01 1.99612767e-01 -4.38396722e-01 7.68526375e-01 -3.27489555e-01 -9.07974914e-02 -5.55849254e-01 -4.76715714e-02 1.08066523e+00 8.31517354e-02 3.27929974e-01 -9.03126240e-01 -2.65838772e-01 2.98848897e-02 -8.47906590e-01 -1.44230688e+00 -1.19240808e+00 -7.15390623e-01 -9.80558395e-02 -4.71552402e-01 -1.19164681e+00 -7.45320499e-01 4.56460752e-02 2.21619070e-01 6.71570241e-01 -5.63951790e-01 -1.92338228e-01 4.61103231e-01 -9.90001261e-02 -3.40409040e-01 -9.51794863e-01 -5.96139669e-01 4.76710975e-01 7.29399621e-01 -2.19511434e-01 -8.09365928e-01 -4.07049268e-01 5.63683510e-01 -7.75855958e-01 4.33973402e-01 -6.12506308e-02 9.28516090e-01 3.19810182e-01 -2.20116019e-01 8.33838105e-01 -1.97914869e-01 8.66905928e-01 -3.36431675e-02 -1.43512636e-01 1.41851291e-01 1.93978652e-01 -6.34775102e-01 9.44693565e-01 -1.13467073e+00 -1.34200692e+00 -1.10629886e-01 -5.29530764e-01 -1.03655756e+00 -6.80098543e-03 -1.65079892e-01 -6.35239124e-01 4.75715995e-02 5.40982604e-01 3.83228183e-01 4.03067112e-01 -1.40242875e-01 4.69532281e-01 1.30724239e+00 1.28668976e+00 -6.62921011e-01 6.25154078e-01 6.62204683e-01 -2.36436546e-01 -1.38346350e+00 -2.35471725e-01 -1.10466100e-01 -1.61378279e-01 -8.31256807e-01 9.54627156e-01 -1.09671509e+00 -1.47855818e+00 8.31384301e-01 -1.23428023e+00 -4.62197781e-01 -4.53740269e-01 3.47453296e-01 -1.18758166e+00 4.60544318e-01 -7.15293944e-01 -1.06987846e+00 -4.15558219e-01 -1.33519173e+00 1.33760428e+00 8.51249024e-02 -3.56586605e-01 -5.33936024e-01 -5.01193777e-02 7.67915130e-01 4.82869536e-01 3.81415576e-01 3.78899604e-01 6.23205826e-02 -2.74542987e-01 1.98982641e-01 1.96610659e-01 2.73961455e-01 3.01006764e-01 2.42920265e-01 -1.60114384e+00 -2.21076235e-01 2.83445388e-01 -6.27236784e-01 3.35459322e-01 4.83572543e-01 1.04359424e+00 -5.12535691e-01 1.51022226e-02 7.40136266e-01 6.12493992e-01 2.35423341e-01 7.15242982e-01 -5.92196584e-01 7.05131471e-01 8.63785446e-01 4.96868551e-01 4.55254167e-01 2.78868377e-01 1.17267191e+00 1.66212499e-01 -6.60285279e-02 -6.07857585e-01 -6.13732934e-01 6.11006021e-01 1.03210211e+00 -1.02463834e-01 -5.07904589e-01 -4.84647840e-01 4.32895690e-01 -1.51682103e+00 -1.37327838e+00 1.61887392e-01 2.04081821e+00 9.54487205e-01 -3.21963847e-01 3.81203651e-01 2.24535197e-01 1.04544818e+00 2.56448716e-01 -6.61331415e-01 -3.79813403e-01 -4.24808979e-01 -8.77893195e-02 -2.62049526e-01 3.81165504e-01 -6.85371518e-01 6.97476208e-01 5.67975187e+00 1.28519917e+00 -1.43317401e+00 2.40520760e-01 6.33748829e-01 -5.62783480e-01 -4.23956990e-01 -5.18294036e-01 -5.00395119e-01 5.05138814e-01 1.25059426e+00 -5.52879691e-01 6.92678511e-01 5.06708682e-01 5.97278953e-01 3.64702791e-01 -1.20515299e+00 1.60063839e+00 5.46854258e-01 -1.31299210e+00 3.27432096e-01 -6.52351379e-02 4.93934363e-01 -7.07176685e-01 4.05524313e-01 6.22691959e-02 -1.85090482e-01 -1.25113344e+00 1.25272310e+00 4.12704110e-01 1.74685848e+00 -7.27266788e-01 2.61995066e-02 2.92164415e-01 -1.36998129e+00 3.17588523e-02 1.01438381e-01 4.07614648e-01 7.73082018e-01 8.93353019e-03 -5.36131322e-01 4.43906605e-01 6.73478723e-01 4.00601149e-01 3.40948910e-01 6.54522598e-01 2.57946521e-01 3.43446195e-01 -2.16183841e-01 2.28494704e-01 -2.85839289e-01 -9.17331427e-02 8.69062483e-01 1.00379527e+00 5.32366753e-01 1.45472035e-01 -7.49718249e-02 9.89429176e-01 -2.29284525e-01 4.57758084e-02 -7.45401442e-01 -1.30733222e-01 5.79942882e-01 9.81319726e-01 -2.17671201e-01 -1.66236266e-01 -2.23153666e-01 1.20549560e+00 -4.17252719e-01 2.88851529e-01 -1.00381601e+00 -3.23586464e-01 1.10176945e+00 3.60560596e-01 -1.22765593e-01 6.04155213e-02 1.27672508e-01 -9.90945458e-01 -4.69865091e-02 -1.31255519e+00 -5.67690283e-02 -1.27108836e+00 -1.03659713e+00 8.79264414e-01 -1.72013953e-01 -1.38695574e+00 -5.68693936e-01 -1.63617119e-01 -5.50524533e-01 9.58065569e-01 -9.49022889e-01 -1.52793598e+00 -3.93445164e-01 9.46850955e-01 9.86729026e-01 -1.64979026e-01 9.63739216e-01 5.50428033e-01 -2.75832385e-01 8.92798722e-01 7.86908716e-02 -1.16708770e-01 9.95095730e-01 -5.45704722e-01 4.05521184e-01 4.73596305e-01 -1.47552803e-01 2.69650757e-01 1.03663945e+00 -2.38887787e-01 -1.57762909e+00 -7.79498994e-01 5.39795160e-01 -1.77041322e-01 4.48891819e-01 -7.93779671e-01 -8.13973784e-01 4.18633789e-01 3.21596205e-01 -1.61447749e-01 7.20513880e-01 -5.00673711e-01 -2.69904524e-01 -4.14003402e-01 -1.26641119e+00 8.53812456e-01 1.04715610e+00 -9.48487461e-01 -2.93369412e-01 7.71728083e-02 8.92742455e-01 -5.43855727e-01 -8.85672569e-01 1.75829291e-01 9.16203141e-01 -1.05141056e+00 1.05431652e+00 -6.20303489e-02 3.69999200e-01 -1.83599427e-01 -1.95136771e-01 -1.30292296e+00 3.29171538e-01 -1.58317685e+00 1.38366073e-01 1.70672250e+00 -1.52250737e-01 -4.93501663e-01 3.77839208e-01 3.29303026e-01 -3.18603396e-01 -3.38120699e-01 -1.18708253e+00 -7.31394112e-01 -1.09957688e-01 -4.53984588e-01 7.27875113e-01 8.77624869e-01 1.16771355e-01 3.92275959e-01 -1.20304906e+00 -8.11317712e-02 5.83658934e-01 -5.73970191e-02 9.94291842e-01 -1.07331038e+00 -2.98098981e-01 -3.33928734e-01 -1.50203347e-01 -1.29362321e+00 2.88749009e-01 -4.26698297e-01 2.49720231e-01 -1.04605806e+00 2.81050038e-02 -5.53298816e-02 5.18037915e-01 -7.50347823e-02 1.53361365e-01 3.29419345e-01 4.27202404e-01 2.03156501e-01 -1.59817532e-01 1.10020065e+00 1.48854387e+00 -6.71318397e-02 -7.26970658e-02 5.24552055e-02 -5.22823751e-01 5.73590815e-01 2.24570438e-01 -7.89922476e-02 -6.72571898e-01 -1.76299632e-01 -2.91973591e-01 1.01963782e+00 4.42402750e-01 -9.14782643e-01 2.18637317e-01 -9.63680148e-02 -4.35145088e-02 -3.63088429e-01 1.31511641e+00 -6.55018747e-01 6.97702348e-01 9.65934321e-02 -4.39516068e-01 -1.84942842e-01 1.16270453e-01 3.83654207e-01 -2.43151635e-01 3.35729361e-01 9.84214246e-01 7.35240802e-02 -1.91392377e-01 3.42672437e-01 -4.82008994e-01 1.95705235e-01 9.14646149e-01 -4.91541535e-01 -4.08317298e-01 -1.16941786e+00 -6.09940350e-01 -9.42556038e-02 5.11170089e-01 7.08287120e-01 7.64947534e-01 -1.52641082e+00 -9.55141425e-01 2.56241173e-01 -9.95308626e-03 -2.55272746e-01 7.66861498e-01 4.42686796e-01 -2.17423856e-01 1.59428760e-01 -4.53865379e-01 -7.83961892e-01 -1.72248828e+00 5.34951091e-01 4.55025345e-01 5.76471329e-01 -8.38152230e-01 6.79907322e-01 4.99638021e-01 -5.55541217e-02 3.77572566e-01 2.84758240e-01 6.07955493e-02 4.12139893e-02 7.89056718e-01 2.92572260e-01 -2.53445715e-01 -1.11920345e+00 4.20014858e-02 7.22608447e-01 3.76234591e-01 -5.00431061e-01 1.06347704e+00 -4.34988469e-01 3.64357203e-01 8.55617344e-01 1.18660617e+00 1.75484627e-01 -1.37555087e+00 1.63447417e-04 -8.99868190e-01 -7.22104788e-01 -1.70349389e-01 -2.85913885e-01 -1.03111184e+00 1.10995281e+00 3.75247180e-01 6.47293404e-02 1.17605996e+00 -1.21371955e-01 1.20160341e+00 -6.06995337e-02 2.86193520e-01 -8.87293160e-01 5.03038883e-01 3.36478427e-02 1.15705359e+00 -6.75919116e-01 -6.17110074e-01 -5.94442368e-01 -9.71484959e-01 1.08357203e+00 4.05882418e-01 5.80388248e-01 2.93682098e-01 5.95571160e-01 3.87197495e-01 1.67739987e-01 -7.96404004e-01 9.66749787e-02 1.33643910e-01 9.48658824e-01 1.54081002e-01 -1.53832123e-01 2.29646608e-01 4.80788857e-01 -7.46569932e-01 1.17946155e-01 5.57890832e-01 2.23784775e-01 -1.25275388e-01 -6.42569423e-01 -7.14878261e-01 -1.14712812e-01 -5.34114540e-01 9.72565487e-02 -2.43255481e-01 4.13649708e-01 -6.87073171e-02 1.35710561e+00 1.11282170e-01 -5.40998042e-01 3.86563331e-01 1.62422433e-01 5.15753210e-01 5.25639579e-03 -5.01636624e-01 4.19242233e-01 1.98664635e-01 -4.79416072e-01 -2.55247295e-01 -6.08741820e-01 -8.03768337e-01 -7.40462601e-01 -7.66199306e-02 4.56114719e-03 7.30976164e-01 6.05624139e-01 3.61211091e-01 6.99240327e-01 7.36667871e-01 -1.32082868e+00 -3.36977363e-01 -8.98981869e-01 -5.94893396e-01 7.06725419e-01 6.23996139e-01 -5.21299779e-01 -3.98673952e-01 3.86953592e-01]
[13.230497360229492, -0.4383385181427002]
bb638376-4ae8-4808-b59f-ab6ae8ddc131
development-of-a-hand-gesture-based-control
null
null
https://link.springer.com/chapter/10.1007%2F978-3-030-46140-9_14
https://dennishnf.bitbucket.io/research/2019%20-%20simbig%202019%20-%20development%20of%20a%20hand%20gesture%20interface.pdf
Development of a hand gesture based control interface using Deep Learning
This paper describes the implementation of a control system based on ten different hand gestures, providing a useful approach for the implementation of better user-friendly human-machine interfaces. Hand detection is achieved using fast detection and tracking algorithms, and classification by a light convolutional neural network. The experimental results show a real-time response with an accuracy of 95.09%, and making use of low power consumption. These results demonstrate that the proposed system could be applied in a large range of applications such as virtual reality, robotics, autonomous driving systems, human-machine interfaces, augmented reality among others.
['Dennis Núñez-Fernández']
2020-04-23
null
null
null
international-conference-on-information
['hand-detection']
['computer-vision']
[-2.92234063e-01 -5.91383338e-01 -4.60353851e-01 -2.68373907e-01 4.87231106e-01 -3.35487902e-01 6.28115773e-01 -6.85419619e-01 -8.77326548e-01 5.16479433e-01 -4.28571999e-01 -5.35508752e-01 -7.49811381e-02 -6.43254936e-01 -5.62035851e-02 -5.62155604e-01 -6.88285977e-02 3.70322078e-01 5.62755883e-01 -3.44165742e-01 5.23393214e-01 1.26500607e+00 -1.95592797e+00 -2.88809955e-01 7.58968666e-02 9.65701222e-01 4.48231339e-01 8.03650022e-01 1.06106766e-01 6.20487988e-01 -5.23538768e-01 -3.81738357e-02 2.20348299e-01 3.33900720e-01 -3.13436776e-01 -5.10445237e-01 1.87219426e-01 -9.65964496e-01 -5.88346303e-01 9.23720121e-01 1.04439628e+00 2.91718304e-01 4.17295754e-01 -1.32792342e+00 -3.70089650e-01 3.91096950e-01 -6.17876112e-01 4.63444516e-02 6.40658081e-01 3.96599203e-01 4.33632106e-01 -9.48477924e-01 7.26445615e-01 1.29848266e+00 4.09871906e-01 7.60864675e-01 -4.70517308e-01 -1.15802133e+00 -1.49145678e-01 4.76896048e-01 -1.60049117e+00 -3.16680633e-02 5.12202382e-01 -5.35290658e-01 1.28663814e+00 3.60940248e-01 5.10280192e-01 9.96933877e-01 4.50008005e-01 7.31217206e-01 9.30245221e-01 -5.38264453e-01 1.90362078e-03 2.20808327e-01 7.73428798e-01 9.23934519e-01 3.99487376e-01 4.08618778e-01 -3.31897855e-01 -6.69300184e-02 1.14658082e+00 3.87379855e-01 -2.81483322e-01 -1.70216024e-01 -1.25750327e+00 5.03744245e-01 7.38267362e-01 6.69300377e-01 -5.12161672e-01 2.14919090e-01 4.61452633e-01 1.26374096e-01 -3.39270055e-01 -2.25978391e-03 -3.38139862e-01 -2.08262131e-01 -3.70168924e-01 2.77442604e-01 7.65299380e-01 1.27702224e+00 -1.12734607e-03 1.25460953e-01 -1.66237622e-01 4.71425533e-01 6.63921535e-01 8.04866493e-01 5.07227182e-01 -5.32016516e-01 3.24528515e-01 3.07922333e-01 4.15253073e-01 -6.11297369e-01 -1.03677583e+00 2.04324797e-01 -7.85530865e-01 1.11001194e+00 3.17725152e-01 -2.05277026e-01 -1.10131884e+00 1.00823808e+00 1.63357854e-01 -2.81662047e-01 -4.36072201e-01 1.31204641e+00 1.08257151e+00 2.76463360e-01 1.95404217e-01 1.82735220e-01 1.67673039e+00 -6.72305107e-01 -1.08068168e+00 3.01376015e-01 -5.47830611e-02 -8.72126102e-01 1.17351198e+00 7.65221179e-01 -5.36523640e-01 -9.81208980e-01 -1.15414310e+00 2.13964842e-02 -5.70667326e-01 7.38654554e-01 9.84603107e-01 1.07790923e+00 -7.33814001e-01 4.53343123e-01 -9.83040094e-01 -7.76267171e-01 2.53245413e-01 7.74447024e-01 -1.49870366e-01 4.60956305e-01 -7.93098688e-01 1.11669338e+00 4.05284673e-01 3.50361615e-01 -1.98581278e-01 1.21819340e-01 -3.78520310e-01 -1.85549464e-02 -6.28820360e-02 -4.10362393e-01 1.31474042e+00 -2.46000335e-01 -1.89667809e+00 6.88693941e-01 -1.30172983e-01 -3.23122293e-02 8.07745159e-01 -2.48433158e-01 -5.45494795e-01 -2.45374620e-01 -3.86586398e-01 4.34812218e-01 7.26822495e-01 -5.25637627e-01 -1.07648075e+00 -7.02727616e-01 -9.59214866e-02 -5.88593744e-02 -2.63671845e-01 5.08844674e-01 -3.82688940e-01 -1.91218913e-01 4.77037840e-02 -9.84863162e-01 9.37514529e-02 3.42321724e-01 -2.41202176e-01 -7.76313126e-01 1.24686921e+00 -5.99169970e-01 1.03308821e+00 -1.71597338e+00 -5.08648083e-02 4.17861402e-01 1.78613603e-01 8.23731661e-01 2.09547639e-01 1.77905723e-01 3.45293909e-01 -3.20235223e-01 5.93376815e-01 1.39337271e-01 3.79480459e-02 -1.44161090e-01 -1.75029203e-01 3.86242121e-01 -3.28362703e-01 8.43693256e-01 -6.51480615e-01 -1.02240309e-01 1.08408403e+00 8.57612610e-01 2.17822522e-01 2.84523308e-01 2.68236697e-01 4.23316598e-01 -4.42936212e-01 7.46791124e-01 5.17705739e-01 3.16059947e-01 -1.54712126e-01 -8.27547014e-02 -5.51541984e-01 -1.12106781e-02 -1.53505182e+00 1.17219985e+00 -4.23530519e-01 1.33870852e+00 1.14820711e-01 -2.30795488e-01 1.02917147e+00 3.55466098e-01 7.59372711e-02 -6.24442160e-01 8.46219063e-01 4.00474042e-01 8.70597586e-02 -9.56965744e-01 5.77253878e-01 3.68921041e-01 5.94420016e-01 4.71824914e-01 -1.38553768e-01 1.72863752e-01 -2.01075599e-02 -3.50744992e-01 6.42337561e-01 2.07107410e-01 3.44745010e-01 8.22914764e-02 5.50098956e-01 -2.66237259e-01 -1.78169578e-01 7.22295582e-01 -4.74810272e-01 1.90174952e-01 -4.54513639e-01 -7.35383868e-01 -8.46171439e-01 -5.52211046e-01 -2.03593224e-01 1.18789160e+00 2.31963888e-01 2.78152764e-01 -4.83247459e-01 -2.35241577e-01 1.51875570e-01 2.49053374e-01 -2.94904858e-01 3.03347528e-01 -9.11764503e-01 -2.80344516e-01 6.85231984e-01 1.11473882e+00 6.64519787e-01 -1.61782920e+00 -1.35249698e+00 1.58509985e-01 5.11797190e-01 -1.05333447e+00 -6.79674512e-03 7.30632022e-02 -6.46436036e-01 -1.08561265e+00 -1.08902586e+00 -1.22686851e+00 1.57582536e-01 4.07868028e-01 4.92690772e-01 3.28874290e-01 -7.06703782e-01 2.84504771e-01 -2.82344878e-01 -6.53539836e-01 1.84900641e-01 2.44704753e-01 2.99975246e-01 -4.06953216e-01 9.31474447e-01 -2.21159205e-01 -5.05448699e-01 4.56561923e-01 -1.25539660e-01 -3.98181707e-01 5.17148614e-01 9.52118814e-01 7.02657849e-02 -3.90708894e-01 2.87457287e-01 -3.63709986e-01 9.10925925e-01 2.18186334e-01 -1.01543808e+00 2.15401724e-01 -7.57688940e-01 -1.61214367e-01 3.43636036e-01 -7.80656695e-01 -1.15409279e+00 4.86978143e-01 -5.46537757e-01 -1.39230505e-01 -4.96093363e-01 -1.07756920e-01 9.55989119e-03 -6.74699247e-01 7.89057434e-01 1.12942345e-01 -1.28740042e-01 -6.49784386e-01 4.59654123e-01 1.78180194e+00 6.41568184e-01 -2.25957856e-02 3.22934747e-01 3.64950716e-01 -2.00064052e-02 -1.16104984e+00 5.13194025e-01 -7.94946551e-01 -9.93638873e-01 -5.07466972e-01 7.92656660e-01 -5.99813581e-01 -1.37654245e+00 8.05767059e-01 -1.53725564e+00 -2.33166501e-01 3.74425024e-01 9.66123402e-01 -3.58025432e-01 1.40896231e-01 -6.64748132e-01 -1.12435997e+00 -8.60798776e-01 -1.19119620e+00 9.75603342e-01 5.19333124e-01 -3.98958534e-01 -4.30084944e-01 -1.56754658e-01 -4.67046201e-02 3.98691505e-01 -2.32724562e-01 3.52981240e-01 -3.38978291e-01 -5.54577827e-01 -7.52557456e-01 -5.12439549e-01 -2.04425558e-01 2.28372112e-01 3.55403006e-01 -1.24394464e+00 -4.33179945e-01 -6.22200787e-01 -1.28619835e-01 4.98625606e-01 2.60447174e-01 1.02762878e+00 1.17328435e-01 -8.47788453e-01 4.47052896e-01 1.30850768e+00 8.86379063e-01 6.50311470e-01 3.38522762e-01 8.06900740e-01 2.52724409e-01 6.05232239e-01 5.33794403e-01 -7.90445730e-02 9.12335396e-01 2.79715717e-01 -4.55497392e-02 -3.27887505e-01 2.78842241e-01 -1.21618174e-01 2.19885319e-01 -9.45852637e-01 -7.56745487e-02 -9.90720332e-01 8.82721096e-02 -1.65766561e+00 -9.76495922e-01 -5.41889608e-01 2.19838428e+00 2.83814400e-01 -9.00460333e-02 3.68726522e-01 4.74962533e-01 7.75510848e-01 -2.40004659e-01 -5.46458900e-01 -4.97502506e-01 5.69533646e-01 5.86856842e-01 7.35736787e-01 4.94758278e-01 -1.24919379e+00 1.01845098e+00 6.95455837e+00 2.77024865e-01 -1.77493334e+00 6.90432116e-02 -3.98429036e-01 2.04846352e-01 8.56788397e-01 -1.07368791e+00 -1.05351019e+00 3.79220724e-01 3.89327288e-01 1.69682473e-01 7.01596320e-01 1.20972228e+00 2.00365230e-01 4.18371260e-02 -9.97846305e-01 1.27331519e+00 -1.03274442e-01 -8.60042274e-01 -2.22599745e-01 -1.23960726e-01 2.08660021e-01 1.21272728e-01 -6.49688169e-02 2.17607945e-01 -1.38202548e-01 -8.98503602e-01 7.52940714e-01 2.16936275e-01 1.09424078e+00 -6.71501160e-01 1.05503511e+00 3.78628999e-01 -1.51190770e+00 -2.64062494e-01 -3.28950852e-01 -5.71982801e-01 -7.49442726e-02 -2.49333382e-01 -1.07632530e+00 1.73628852e-02 9.06362236e-01 1.53494880e-01 -7.52347559e-02 1.20517671e+00 -2.77087510e-01 -5.42245433e-02 -2.55994767e-01 -1.12617958e+00 -2.66264677e-01 1.90092921e-01 3.35686058e-01 1.65143347e+00 6.36599660e-02 3.53329360e-01 -1.45422239e-02 5.76339185e-01 2.64384419e-01 1.57542944e-01 -7.58779585e-01 1.54120713e-01 5.78527033e-01 1.14764643e+00 -9.03492451e-01 -4.62852955e-01 -3.09591293e-01 9.74116683e-01 -2.31942236e-01 1.95246398e-01 -6.66219592e-01 -1.42950344e+00 5.32365978e-01 -1.74567819e-01 1.69902310e-01 -5.63238978e-01 -7.93720841e-01 -7.88053334e-01 2.52503544e-01 -5.15936196e-01 -1.05149612e-01 -8.56799603e-01 -6.23429239e-01 7.50589907e-01 -3.69822204e-01 -1.35127759e+00 -6.15831137e-01 -1.50176799e+00 -4.17755038e-01 1.13549221e+00 -1.22830963e+00 -1.04085720e+00 -9.60845232e-01 6.32716537e-01 5.41795611e-01 -6.33347332e-01 1.02747345e+00 4.33365464e-01 -5.20211935e-01 8.33974600e-01 5.06563531e-03 3.40470254e-01 5.12049854e-01 -9.75241423e-01 5.05192697e-01 5.55728853e-01 -1.25490725e-01 8.52121115e-01 3.60340953e-01 -3.66591781e-01 -1.68042588e+00 -5.24410367e-01 5.54098487e-01 -2.76561081e-01 2.74503678e-01 -2.79688626e-03 -3.81675005e-01 6.57772124e-01 2.91588485e-01 -1.04085624e-01 1.31941721e-01 7.06836022e-03 -1.84924468e-01 7.46558281e-03 -1.47234440e+00 5.99124849e-01 1.12217450e+00 -4.39465582e-01 -5.28510332e-01 1.55239895e-01 7.24661872e-02 -7.36679852e-01 -4.48336750e-01 1.50539428e-01 1.59330130e+00 -7.16442049e-01 9.21478033e-01 -5.09142458e-01 -4.33999479e-01 -3.43736261e-01 6.36022761e-02 -8.86734545e-01 -6.65924609e-01 -3.82746130e-01 -2.20875919e-01 7.30724394e-01 -1.84611101e-02 -7.53912508e-01 5.85754633e-01 6.10578716e-01 2.86318094e-01 -2.83897072e-01 -8.56943667e-01 -8.35832655e-01 -6.65254533e-01 -4.27702039e-01 6.93507135e-01 5.41265190e-01 3.23542833e-01 2.69381195e-01 -2.24009350e-01 1.73899874e-01 3.26497942e-01 -2.33832374e-01 8.60873222e-01 -1.58827710e+00 1.43016875e-01 -8.09047759e-01 -9.17906165e-01 -1.02646136e+00 -1.35994598e-01 -2.23095745e-01 2.49063179e-01 -1.40038633e+00 -1.36968046e-01 -4.52504575e-01 -1.74476296e-01 4.58080828e-01 9.33402926e-02 5.41895807e-01 2.20091954e-01 1.96158633e-01 -5.13798408e-02 -1.73170879e-01 9.41418469e-01 -2.02132732e-01 -4.73808348e-01 4.55114454e-01 2.40013421e-01 7.50203788e-01 9.13490593e-01 9.38155428e-02 9.07925591e-02 -2.25761771e-01 -4.14662182e-01 -2.72449970e-01 3.04940552e-01 -1.18820930e+00 4.08433229e-01 -2.21609265e-01 5.80057383e-01 -6.43504977e-01 2.83847153e-01 -1.15861654e+00 -5.25622107e-02 1.09991276e+00 3.56212184e-02 -1.13780253e-01 1.91763222e-01 -3.42513733e-02 2.25167125e-01 -2.34066814e-01 6.84510827e-01 8.54240358e-02 -1.37379575e+00 -2.45880987e-02 -6.07990324e-01 -1.14230633e+00 1.17991245e+00 -3.76307547e-01 -2.66769886e-01 -2.58393437e-02 -5.22772431e-01 -2.06713945e-01 1.60021912e-02 7.54594982e-01 6.71010256e-01 -1.24373686e+00 -6.01618811e-02 6.33033156e-01 5.01439162e-02 -6.05400503e-01 -2.79261261e-01 3.61115932e-01 -1.18135488e+00 1.06392300e+00 -1.02115595e+00 -6.59201920e-01 -1.78212774e+00 4.43107992e-01 4.22280192e-01 5.17587841e-01 -8.65867138e-01 7.53787458e-01 -7.18953907e-01 -4.91578579e-01 1.31629336e+00 -6.05407238e-01 -6.93009079e-01 -2.19947278e-01 1.13078523e+00 9.35359657e-01 2.67913818e-01 -6.11992955e-01 -5.09403765e-01 1.02682328e+00 4.00771946e-01 -2.65019864e-01 9.13567424e-01 2.98829496e-01 2.25909844e-01 2.82299012e-01 6.50142193e-01 -2.51079291e-01 -8.45210969e-01 8.22690204e-02 -5.71961738e-02 -6.68598235e-01 1.55698702e-01 -1.08799267e+00 -9.41684961e-01 1.31338477e+00 1.63324571e+00 -2.47713756e-02 8.79776955e-01 -4.46122795e-01 5.24256408e-01 1.15268588e+00 1.10761070e+00 -1.05039108e+00 -4.68233764e-01 5.81202984e-01 1.05259490e+00 -1.28791904e+00 -1.29806697e-01 -2.32599676e-01 -2.42083579e-01 1.59053123e+00 8.30629408e-01 -1.10069308e-02 6.40155733e-01 5.11148930e-01 4.40287411e-01 -1.22595737e-02 7.64733460e-03 -5.50191998e-01 2.81186610e-01 9.60203052e-01 5.66029668e-01 5.73878646e-01 -8.63224030e-01 4.30257350e-01 1.28450431e-02 6.97245896e-01 2.67988682e-01 1.26564598e+00 -6.55498981e-01 -8.87037814e-01 -6.81621432e-01 3.42834353e-01 -2.94636548e-01 2.70945996e-01 -3.26283395e-01 9.26753402e-01 2.07153901e-01 8.24139595e-01 5.50268032e-02 -1.01327872e+00 6.23412907e-01 1.86011493e-01 6.94996119e-01 -2.96828777e-01 -4.91618782e-01 -1.84203267e-01 -3.72470409e-01 -5.82091510e-01 -1.08012713e-01 -1.11248024e-01 -1.48430264e+00 -5.00348806e-01 -5.93079925e-01 -4.96503025e-01 1.17171884e+00 7.26310074e-01 2.16538280e-01 5.34475029e-01 4.27582562e-01 -1.52009058e+00 -6.52046084e-01 -1.39366555e+00 -6.55542433e-01 -2.00491980e-01 4.11214471e-01 -1.09287357e+00 1.61820143e-01 -1.95777476e-01]
[6.447032928466797, -0.31830930709838867]
5b858c82-da64-496a-977e-b9ac4bfddebd
a-human-centered-safe-robot-reinforcement
2302.13137
null
https://arxiv.org/abs/2302.13137v2
https://arxiv.org/pdf/2302.13137v2.pdf
A Human-Centered Safe Robot Reinforcement Learning Framework with Interactive Behaviors
Deployment of reinforcement learning algorithms for robotics applications in the real world requires ensuring the safety of the robot and its environment. Safe robot reinforcement learning (SRRL) is a crucial step towards achieving human-robot coexistence. In this paper, we envision a human-centered SRRL framework consisting of three stages: safe exploration, safety value alignment, and safe collaboration. We examine the research gaps in these areas and propose to leverage interactive behaviors for SRRL. Interactive behaviors enable bi-directional information transfer between humans and robots, such as conversational robot ChatGPT. We argue that interactive behaviors need further attention from the SRRL community. We discuss four open challenges related to the robustness, efficiency, transparency, and adaptability of SRRL with interactive behaviors.
['Alois Knoll', 'Jan Peters', 'Guang Chen', 'Yali Du', 'Alap Kshirsagar', 'Shangding Gu']
2023-02-25
null
null
null
null
['safe-exploration']
['robots']
[-3.15033734e-01 9.44182336e-01 -2.37936556e-01 -3.21383737e-02 -3.09150636e-01 -4.35398668e-01 5.62607646e-01 -2.84056336e-01 -3.59494686e-01 1.10330236e+00 1.51965931e-01 -2.40346909e-01 -2.03264207e-01 -4.57270712e-01 -4.69558716e-01 -5.62545538e-01 -6.83802664e-01 2.07493737e-01 2.49836788e-01 -6.96563840e-01 1.11981682e-01 3.79988194e-01 -1.31011641e+00 -3.60377252e-01 8.19131017e-01 5.24236023e-01 1.43266812e-01 7.52721369e-01 5.18516004e-01 1.52538633e+00 -3.15072626e-01 -5.43305986e-02 3.33147347e-01 -2.14842841e-01 -1.39503241e+00 -5.95547855e-02 -6.43711269e-01 -9.01485980e-01 -3.41433525e-01 6.79411292e-01 1.76269591e-01 2.12653890e-01 5.25462270e-01 -2.45934844e+00 -3.35120529e-01 9.06759024e-01 -3.51343781e-01 -7.35843182e-01 4.55502212e-01 9.49839294e-01 6.99861050e-01 1.14176599e-02 8.50763023e-01 1.47795415e+00 4.68952954e-01 9.71173704e-01 -7.34594464e-01 -6.46651566e-01 8.90417099e-02 2.60338783e-01 -9.91979361e-01 -2.31204778e-01 2.46704757e-01 -4.43933100e-01 1.10751021e+00 -1.43558279e-01 7.61760712e-01 1.55369830e+00 7.04594135e-01 6.46091938e-01 8.03782880e-01 -2.65326351e-01 3.60413939e-01 1.69496052e-02 -4.57168557e-02 4.45059091e-01 3.39486182e-01 6.45224810e-01 -5.02343535e-01 -2.04208136e-01 8.93038452e-01 -4.84721541e-01 3.15322310e-01 -8.96318376e-01 -1.29573441e+00 6.43419683e-01 4.32776034e-01 4.52872477e-02 -5.61072588e-01 7.22787976e-01 6.08637094e-01 5.47153294e-01 -4.26721603e-01 7.25076377e-01 -1.61997512e-01 -4.89314586e-01 3.79365325e-01 4.24642205e-01 1.01291001e+00 1.45221472e+00 5.72085798e-01 1.54566258e-01 1.98567227e-01 3.91969442e-01 6.29313052e-01 3.92517924e-01 -2.46409759e-01 -1.85856855e+00 5.86514361e-02 2.75554597e-01 6.28440022e-01 -7.50018120e-01 -6.92713141e-01 3.36580843e-01 -6.72536567e-02 7.74281502e-01 -1.93130337e-02 -6.54569387e-01 5.68940975e-02 1.74092484e+00 6.82845473e-01 -3.70202899e-01 9.42974985e-01 6.96308613e-01 4.08712178e-01 5.81384838e-01 2.73892581e-01 -1.20971352e-01 9.17527139e-01 -1.04241776e+00 -6.91525877e-01 -1.82225645e-01 8.99379909e-01 -6.95502525e-03 1.03425968e+00 3.88746679e-01 -8.30096304e-01 4.29201387e-02 -1.15405250e+00 7.74822384e-02 1.64449766e-01 -4.90001410e-01 4.51814979e-01 1.50367334e-01 -1.03127933e+00 2.59331554e-01 -1.03412819e+00 -9.27087009e-01 -4.32024747e-02 4.00965363e-01 -4.96570855e-01 5.19561172e-01 -1.16744161e+00 1.35730195e+00 5.14461696e-01 -2.50950843e-01 -1.49859536e+00 1.00269035e-01 -8.72883201e-01 -2.34185129e-01 7.24467278e-01 -2.73916513e-01 1.86630940e+00 -5.78931391e-01 -1.99447632e+00 4.50212717e-01 7.37583399e-01 -8.18321288e-01 6.40313923e-01 -5.79571724e-01 -7.35225454e-02 2.48140723e-01 2.97579855e-01 8.80979419e-01 3.66574645e-01 -1.66640031e+00 -9.26197231e-01 3.61319520e-02 3.57874513e-01 4.32078153e-01 -9.04449597e-02 1.33194849e-02 2.16947228e-01 1.08573496e-01 -5.34635782e-01 -1.31187034e+00 -5.32646060e-01 -1.42611325e-01 -1.22274868e-01 -5.03556252e-01 1.02959228e+00 -2.57881552e-01 5.80946684e-01 -2.28505611e+00 9.16992053e-02 -1.99763221e-03 1.70064092e-01 -2.97010034e-01 -2.71242470e-01 9.04472470e-01 5.13925135e-01 -2.31342241e-01 1.42125010e-01 2.69952297e-01 4.10594732e-01 3.98477584e-01 -3.98663372e-01 4.97020990e-01 7.68962875e-03 7.44403720e-01 -1.28913069e+00 -7.12070644e-01 1.80439308e-01 5.89156486e-02 -5.79153776e-01 8.54283154e-01 -2.49413908e-01 8.34677041e-01 -5.99583447e-01 4.62253511e-01 1.53659463e-01 2.72078604e-01 6.15195751e-01 5.07452011e-01 -5.19783556e-01 2.36682594e-01 -6.98985875e-01 1.22763300e+00 -2.87812293e-01 5.25657654e-01 6.70529366e-01 -3.28547925e-01 9.37033415e-01 4.69801694e-01 7.62863874e-01 -6.39274418e-01 2.59738028e-01 1.80147350e-01 2.19982654e-01 -7.19652891e-01 5.14752626e-01 1.48212850e-01 -6.08331621e-01 8.44933093e-01 1.01463623e-01 -4.78905320e-01 -1.77079856e-01 2.84948260e-01 1.11965489e+00 5.99730790e-01 6.21275425e-01 -2.46293381e-01 6.04515150e-02 4.32667434e-01 5.99629939e-01 8.85305583e-01 -1.04971242e+00 -4.79467571e-01 7.72067249e-01 -2.05228254e-01 -8.35496426e-01 -9.61480916e-01 4.75940287e-01 1.24175382e+00 7.33904779e-01 -2.52142638e-01 -9.50553358e-01 -9.69955862e-01 -7.97783509e-02 1.26262319e+00 -4.28484499e-01 -3.82526904e-01 -7.01340258e-01 2.43486181e-01 8.17455053e-01 2.78917551e-01 5.80821931e-01 -1.53554082e+00 -1.78137696e+00 1.94460481e-01 -3.15978140e-01 -1.17636502e+00 -9.14767161e-02 4.05134797e-01 -3.36645901e-01 -1.15768147e+00 2.06238553e-01 -7.80657470e-01 2.09207848e-01 4.47391301e-01 4.04007047e-01 6.87894821e-02 2.53021549e-02 9.18434560e-01 -6.74843907e-01 -4.43571270e-01 -8.21956158e-01 -1.74980670e-01 4.67620850e-01 -6.97482884e-01 -9.93342251e-02 -6.27180099e-01 -4.73014891e-01 5.84011376e-01 -3.45040977e-01 1.48219436e-01 4.85033125e-01 6.30254030e-01 -1.60544574e-01 -9.30450559e-02 7.28340387e-01 -1.38886869e-01 8.46504927e-01 -7.38358200e-01 -5.72860360e-01 2.70890266e-01 -4.03137147e-01 -6.38611168e-02 4.31528777e-01 -2.90448636e-01 -1.12596178e+00 1.19505413e-01 1.40507206e-01 1.91024095e-01 -2.93281585e-01 -1.18800737e-01 -1.13207221e-01 -1.81759357e-01 8.47756743e-01 -7.54049122e-02 5.76859176e-01 2.63387769e-01 6.69997633e-01 9.98076379e-01 4.32621002e-01 -8.88607562e-01 4.16085929e-01 1.95519939e-01 -2.84000456e-01 -7.32727706e-01 1.28069028e-01 6.12208992e-03 -3.08668643e-01 -7.46613801e-01 7.67093301e-01 -8.17554951e-01 -1.55817914e+00 3.76586437e-01 -1.11391449e+00 -1.14671338e+00 -3.15553516e-01 4.00292218e-01 -1.39205205e+00 4.11621362e-01 -6.95295393e-01 -1.29336309e+00 -3.21150929e-01 -1.32403386e+00 6.09482348e-01 2.94661701e-01 -7.26482391e-01 -3.35158557e-01 1.24776267e-01 1.75345510e-01 4.19956297e-01 2.47534335e-01 7.28056788e-01 -5.67384720e-01 -5.02807081e-01 3.69507700e-01 4.49840687e-02 -1.39598921e-01 1.31115928e-01 1.10900495e-02 -4.32050437e-01 -2.92271793e-01 5.33493422e-02 -1.16838241e+00 -4.06586468e-01 -5.00376076e-02 3.50954831e-01 -8.29521418e-01 -3.48811299e-01 -3.96810174e-02 7.89435983e-01 7.91448236e-01 5.48998475e-01 7.22359598e-01 4.02891845e-01 1.28132200e+00 1.55275941e+00 6.69771850e-01 9.42877829e-01 4.34940845e-01 8.83031011e-01 4.41630751e-01 3.87601942e-01 -4.30825591e-01 8.18466604e-01 3.12597156e-01 9.54036266e-02 8.96491408e-02 -1.07845926e+00 3.70191962e-01 -2.55253625e+00 -7.45665669e-01 1.61186248e-01 1.73772812e+00 6.73520625e-01 -6.31672740e-02 4.46667641e-01 -3.18309426e-01 4.66308028e-01 -4.34395552e-01 -8.08998287e-01 -7.53060162e-01 4.99415934e-01 -8.69027734e-01 4.71061409e-01 5.46654642e-01 -9.21885192e-01 1.38663316e+00 7.01050138e+00 2.23811775e-01 -8.58460665e-01 1.42384037e-01 1.64825827e-01 1.97741210e-01 1.22003175e-01 2.33551174e-01 -3.36480767e-01 -6.14209622e-02 8.61554146e-01 -2.55774289e-01 6.66817605e-01 1.40840721e+00 4.67222303e-01 -4.12385821e-01 -1.24175119e+00 6.11719489e-01 -4.50119823e-01 -6.98502779e-01 -4.48242784e-01 -9.77171585e-03 2.69137204e-01 4.80876081e-02 -2.85551071e-01 5.70979238e-01 1.36795497e+00 -1.03687990e+00 1.19271982e+00 1.35875300e-01 3.61255616e-01 -1.06302512e+00 3.36330414e-01 5.87478042e-01 -9.09941018e-01 -6.68850720e-01 1.07790142e-01 -2.74670631e-01 2.19062239e-01 -5.42999864e-01 -1.10547626e+00 3.64832699e-01 1.04045367e+00 5.29431581e-01 6.81768432e-02 4.96913344e-01 -3.98048490e-01 9.69628394e-02 4.19588834e-02 -4.74019170e-01 4.06673461e-01 -3.64916444e-01 6.20419443e-01 7.37800479e-01 -9.68628842e-03 1.58278629e-01 3.19103718e-01 8.06232691e-01 5.90509057e-01 -2.64174074e-01 -1.13155401e+00 -1.46899700e-01 7.87565529e-01 1.06507909e+00 -5.61992168e-01 2.05169439e-01 -1.37391120e-01 7.07296073e-01 5.15870750e-01 2.32676774e-01 -1.01626694e+00 -4.41296279e-01 9.01267707e-01 -5.10894656e-01 -2.27157697e-01 -5.78825235e-01 -3.49579096e-01 -4.40222979e-01 -3.77325416e-01 -1.27222729e+00 1.86305955e-01 -8.30306411e-01 -9.59614158e-01 5.26140630e-01 7.47095570e-02 -1.38066792e+00 -6.18871331e-01 -3.23569804e-01 -3.31543565e-01 -1.70492634e-01 -9.51408803e-01 -1.22188926e+00 -2.87289321e-01 4.25197691e-01 4.32873696e-01 -2.04177722e-01 7.75734961e-01 -5.65562725e-01 -4.29775178e-01 3.24866772e-01 -2.36398369e-01 -2.44418591e-01 6.86635494e-01 -8.41968596e-01 3.08283448e-01 4.32060689e-01 -8.08240831e-01 6.20009840e-01 9.74416912e-01 -8.69827449e-01 -1.78417969e+00 -9.68317270e-01 1.71669066e-01 -3.26141477e-01 7.00461805e-01 -3.24845076e-01 -3.05620015e-01 1.00442052e+00 6.76413596e-01 -6.59161091e-01 3.78675997e-01 -1.02028124e-01 -1.22531816e-01 -2.19053738e-02 -1.31578588e+00 1.42889750e+00 1.12131941e+00 -2.93192863e-01 -5.65944552e-01 -3.65200490e-02 1.22369516e+00 -8.78537744e-02 -6.05716527e-01 3.57093185e-01 7.03966141e-01 -9.64835227e-01 5.88529885e-01 -4.64064419e-01 3.55313659e-01 -3.14658701e-01 -1.38007313e-01 -1.27484000e+00 -1.89383239e-01 -1.15777528e+00 -4.84445430e-02 8.40845287e-01 -1.27645321e-02 -7.81160831e-01 3.86140436e-01 9.10001218e-01 -2.55859077e-01 -2.42351398e-01 -7.54258811e-01 -1.12809837e+00 1.83350235e-01 -2.97834367e-01 4.60462838e-01 7.22299218e-01 1.30007744e+00 2.63190120e-01 -7.13042855e-01 1.54103443e-01 5.68417251e-01 -4.29914236e-01 1.26974165e+00 -7.32743800e-01 -5.23194596e-02 -2.19394892e-01 1.36987880e-01 -5.75534642e-01 3.88229430e-01 -2.34825850e-01 9.47027385e-01 -1.56277299e+00 1.71876475e-01 -6.15570962e-01 2.62422770e-01 7.19911575e-01 4.44576919e-01 -5.09738743e-01 4.18844074e-01 3.46600413e-01 -1.19063735e+00 9.25818980e-01 1.12506270e+00 2.54318446e-01 -5.32468557e-01 -4.35249895e-01 -7.61234641e-01 7.55293250e-01 1.21328616e+00 -2.24110216e-01 -6.67999387e-01 -1.45650655e-01 -1.06271744e-01 5.54522276e-01 3.55595648e-01 -1.02765059e+00 4.42207605e-01 -1.10729003e+00 -5.73440373e-01 -2.36896388e-02 2.00331777e-01 -1.03799939e+00 2.38081768e-01 1.13626683e+00 -5.58140278e-01 -1.77942347e-02 -5.68084903e-02 5.33067703e-01 2.73210138e-01 6.97440356e-02 7.50496089e-01 -6.26848489e-02 -8.36440146e-01 -3.90417129e-01 -1.35866439e+00 -1.94031850e-01 1.93111241e+00 -1.10055514e-01 -4.61323738e-01 -7.55285442e-01 -3.37778270e-01 1.04526484e+00 6.13636494e-01 6.00866079e-01 6.20673418e-01 -9.33706045e-01 -2.60042965e-01 -2.13360474e-01 2.99153090e-01 -1.93798885e-01 -3.10044326e-02 6.82318866e-01 -6.61099732e-01 2.44146273e-01 -9.06653821e-01 -2.76889026e-01 -1.14546287e+00 4.62085009e-01 4.48439986e-01 9.34711024e-02 -7.38168418e-01 3.39409322e-01 2.61669248e-01 -9.70130503e-01 8.04784775e-01 1.45729616e-01 -2.30582014e-01 -6.06153011e-01 2.37989992e-01 7.01608837e-01 -7.29456246e-01 -5.58399498e-01 -4.90260214e-01 -2.77534015e-02 8.62802714e-02 -6.27159119e-01 1.10668004e+00 -6.11242950e-01 -1.71239093e-01 3.80240917e-01 4.22753185e-01 -5.21689057e-01 -1.71778119e+00 2.85535634e-01 4.09401864e-01 6.71940390e-03 -5.36863923e-01 -6.92182422e-01 -3.38515967e-01 3.04962009e-01 3.51361930e-01 1.43884733e-01 5.58906317e-01 -2.25793235e-02 7.11275816e-01 8.25939953e-01 1.34854293e+00 -1.60679889e+00 7.07368791e-01 8.04466069e-01 1.19685996e+00 -1.14176071e+00 -3.76277208e-01 -4.16867077e-01 -1.37586057e+00 9.02546883e-01 1.44568825e+00 -1.02828793e-01 2.63845116e-01 4.08565223e-01 2.43589684e-01 -2.95076538e-02 -1.02431750e+00 -5.72858043e-02 -8.47899079e-01 1.15605617e+00 -1.42422646e-01 1.46464705e-01 -2.09384695e-01 4.66180056e-01 -2.00886443e-01 -4.99309860e-02 8.94962370e-01 1.43792677e+00 -7.88633525e-01 -9.41797495e-01 -3.30414116e-01 -1.73827231e-01 1.53068393e-01 7.19816744e-01 -6.87458813e-01 9.27251995e-01 -2.46032685e-01 1.63180649e+00 -3.65852952e-01 -6.60824001e-01 2.62111008e-01 -2.53262222e-01 2.40551203e-01 -3.58254939e-01 -5.83070993e-01 -2.77228355e-01 4.48181093e-01 -1.12056887e+00 -1.69153214e-01 -6.26547456e-01 -1.94013488e+00 -3.08341324e-01 1.54698923e-01 2.68143862e-01 3.86977196e-01 8.40933979e-01 3.94775301e-01 1.47634327e-01 8.73059690e-01 -6.49907947e-01 -8.21424127e-01 -5.47229290e-01 -3.20405185e-01 6.52321195e-03 5.54168284e-01 -7.30744958e-01 -2.03655109e-01 -2.10579962e-01]
[4.681305885314941, 1.210062026977539]
b475d5d3-8d8f-478a-93e1-3476c3969b11
modeling-the-mistakes-of-boundedly-rational
2106.13249
null
https://arxiv.org/abs/2106.13249v1
https://arxiv.org/pdf/2106.13249v1.pdf
Modeling the Mistakes of Boundedly Rational Agents Within a Bayesian Theory of Mind
When inferring the goals that others are trying to achieve, people intuitively understand that others might make mistakes along the way. This is crucial for activities such as teaching, offering assistance, and deciding between blame or forgiveness. However, Bayesian models of theory of mind have generally not accounted for these mistakes, instead modeling agents as mostly optimal in achieving their goals. As a result, they are unable to explain phenomena like locking oneself out of one's house, or losing a game of chess. Here, we extend the Bayesian Theory of Mind framework to model boundedly rational agents who may have mistaken goals, plans, and actions. We formalize this by modeling agents as probabilistic programs, where goals may be confused with semantically similar states, plans may be misguided due to resource-bounded planning, and actions may be unintended due to execution errors. We present experiments eliciting human goal inferences in two domains: (i) a gridworld puzzle with gems locked behind doors, and (ii) a block-stacking domain. Our model better explains human inferences than alternatives, while generalizing across domains. These findings indicate the importance of modeling others as bounded agents, in order to account for the full richness of human intuitive psychology.
['Joshua B. Tenenbaum', 'Vikash K. Mansinghka', 'Tan Zhi-Xuan', 'Joie Le', 'Gloria Z. Lin', 'Arwa Alanqary']
2021-06-24
null
null
null
null
['game-of-chess']
['playing-games']
[-8.99610817e-02 5.34022987e-01 2.45382801e-01 -3.31574768e-01 -1.49785176e-01 -3.71584713e-01 5.78369021e-01 2.73321986e-01 -3.89016062e-01 7.50113368e-01 3.16783905e-01 -5.15935242e-01 -2.93274224e-01 -8.32675815e-01 -3.06815922e-01 -3.47402602e-01 2.66808599e-01 8.06826890e-01 5.45129366e-02 -2.32105702e-01 6.73570037e-01 3.31013650e-01 -1.21448767e+00 1.82648320e-02 6.20092630e-01 1.22653447e-01 2.50373453e-01 4.94798660e-01 2.44926035e-01 1.65185654e+00 -4.71340835e-01 -5.91084301e-01 -1.85145229e-01 -5.09618580e-01 -1.16478109e+00 4.33666050e-01 -2.21834004e-01 -9.58068848e-01 -2.53823102e-01 1.27112293e+00 -2.10508630e-01 5.17291665e-01 6.10070050e-01 -1.41659570e+00 -7.75067747e-01 9.04136181e-01 -2.72170484e-01 -1.24132685e-01 8.20512891e-01 5.34604013e-01 9.20334339e-01 -3.89695279e-02 1.64536700e-01 1.65533257e+00 3.08420867e-01 6.54274821e-01 -1.43649709e+00 -3.06351900e-01 3.88287872e-01 3.91460389e-01 -1.20958948e+00 -5.40878534e-01 6.63640857e-01 -5.43803334e-01 1.11274874e+00 2.59480059e-01 5.32381356e-01 1.15418792e+00 3.69829237e-01 4.30615693e-01 1.26824081e+00 -3.12668502e-01 5.69477022e-01 -7.07409605e-02 2.77431905e-01 5.33454180e-01 6.59155309e-01 4.30279672e-01 -5.25363207e-01 -5.96417725e-01 7.99352407e-01 -8.30702409e-02 -8.99369642e-02 -2.15175420e-01 -1.19798791e+00 7.65962958e-01 8.67731869e-02 1.85462907e-01 -8.32568467e-01 3.24156553e-01 -1.17579371e-01 5.25828116e-02 -1.77517101e-01 8.08062971e-01 -1.93943386e-03 -1.70352384e-01 -3.75002831e-01 7.11776733e-01 9.06467021e-01 7.29202509e-01 4.58327591e-01 1.05026580e-01 2.65374124e-01 2.34296381e-01 6.17553711e-01 2.22725794e-01 5.37917435e-01 -1.54122663e+00 6.57073259e-02 4.91908789e-01 9.08234358e-01 -1.13096249e+00 -7.39465535e-01 -7.08145872e-02 -3.06697786e-01 7.20329344e-01 7.80187190e-01 9.24978554e-02 -3.71222079e-01 2.02847958e+00 2.73542684e-02 -9.48681682e-02 1.37713894e-01 9.94189441e-01 2.33793661e-01 1.80831060e-01 3.71343255e-01 -2.45738685e-01 1.46923602e+00 -4.74494815e-01 -5.12157500e-01 -7.86792696e-01 7.33001292e-01 -5.00882864e-01 1.20176744e+00 5.08915603e-01 -1.41935885e+00 -8.15797374e-02 -8.07619631e-01 -1.30046606e-01 2.02191114e-01 -4.92224663e-01 8.77178490e-01 6.79975688e-01 -8.55771482e-01 7.37153828e-01 -1.13574052e+00 -3.41159761e-01 1.62177444e-01 -1.10841066e-01 -1.65308431e-01 -1.35896862e-01 -7.71842241e-01 1.29850662e+00 5.48516989e-01 5.29667586e-02 -1.08684945e+00 1.21156033e-02 -9.28499103e-01 4.06852901e-01 6.69301271e-01 -8.78085792e-01 1.45971584e+00 -8.67561877e-01 -1.24187791e+00 9.72170889e-01 -4.44221385e-02 -4.76417959e-01 4.47691530e-01 1.24120466e-01 -1.35001406e-01 -2.53406733e-01 3.31893116e-01 3.51543307e-01 5.42939365e-01 -9.61216092e-01 -2.36404568e-01 -8.85746539e-01 7.09441543e-01 5.88468909e-01 5.75821638e-01 1.47643492e-01 7.23906159e-01 -2.68391311e-01 6.92551136e-01 -8.69028866e-01 -3.34088862e-01 -3.76428634e-01 -5.61707616e-01 -2.78183997e-01 -1.13684274e-01 -4.36874390e-01 9.57718670e-01 -2.01958346e+00 -1.05836093e-01 -2.29791738e-02 5.27950346e-01 -4.50229555e-01 1.61952198e-01 4.27019507e-01 -5.91631085e-02 2.85674393e-01 1.59214810e-01 1.60537995e-02 6.12819970e-01 2.20683813e-01 -2.58847684e-01 6.54830337e-01 -2.34732166e-01 6.62156761e-01 -8.21758211e-01 -2.04404995e-01 2.19751552e-01 -1.03712842e-01 -5.09379268e-01 1.99251279e-01 -3.36855143e-01 2.60863483e-01 -5.27720749e-01 3.40881228e-01 4.66166735e-01 -3.42195541e-01 5.27511358e-01 5.54481983e-01 1.84384510e-02 9.47334766e-01 -1.16014898e+00 1.04012597e+00 1.96700394e-02 3.00885499e-01 -1.42690614e-01 -5.62817991e-01 2.86038488e-01 3.45764071e-01 -4.46770728e-01 -3.18635285e-01 2.39349410e-01 -1.17393598e-01 5.96741080e-01 -4.27874982e-01 4.48426187e-01 -7.66477823e-01 -3.43159661e-02 1.01130033e+00 -5.28287828e-01 -5.29371083e-01 -6.82559162e-02 4.41965789e-01 1.08872986e+00 1.04890898e-01 7.21731007e-01 -2.28270128e-01 1.30766451e-01 2.73764908e-01 7.33697772e-01 1.12446332e+00 -5.06818116e-01 1.62582353e-01 8.31659555e-01 -6.93276644e-01 -9.46897984e-01 -1.12327409e+00 4.10975933e-01 9.94544923e-01 3.00693542e-01 -1.10847250e-01 -7.43256807e-01 -5.87958060e-02 -3.23878407e-01 1.91660178e+00 -3.37105185e-01 -2.80062467e-01 -1.69614717e-01 -4.58101481e-01 5.58236122e-01 4.75927591e-01 4.41977650e-01 -9.42923069e-01 -1.44218695e+00 2.28608161e-01 -5.79904318e-01 -7.97914803e-01 -1.29799664e-01 -1.60604239e-01 -7.95378447e-01 -1.29122901e+00 1.81263298e-01 -5.00600189e-02 5.56178093e-01 4.49516892e-01 1.24649155e+00 7.30484366e-01 1.70988336e-01 5.83858907e-01 -1.26809478e-01 -4.64901745e-01 -5.71540356e-01 -9.04725134e-01 3.36648732e-01 -7.38480389e-01 7.12717533e-01 -6.05657518e-01 -2.20324785e-01 3.95529002e-01 -4.86265332e-01 3.90550703e-01 -3.97918373e-02 6.67883933e-01 -1.02409713e-01 4.82895404e-01 -7.93819353e-02 -6.82381034e-01 9.50246334e-01 -5.45165062e-01 -6.25364661e-01 1.87148258e-01 -4.08430278e-01 -1.87775791e-02 3.85221392e-01 -3.35620940e-01 -1.35009694e+00 -6.57086372e-01 2.60183096e-01 1.67792603e-01 -6.13555729e-01 3.90086144e-01 -1.71509579e-01 3.63900006e-01 9.16487217e-01 1.29646868e-01 7.27887526e-02 -8.59898608e-03 5.47694229e-03 1.70952961e-01 2.05295742e-01 -1.11715233e+00 4.34469968e-01 3.58643472e-01 -1.71153113e-01 -5.84859729e-01 -8.03237319e-01 1.38607159e-01 -3.12716663e-02 -2.10537329e-01 6.21623755e-01 -8.60686302e-01 -1.51096809e+00 4.94861245e-01 -9.75506246e-01 -6.37113810e-01 -1.28831221e-02 7.30127215e-01 -1.16614473e+00 4.85020280e-01 -7.53062963e-01 -1.27532196e+00 5.79140782e-01 -1.23594952e+00 2.70037085e-01 2.33910143e-01 -9.33210969e-01 -7.28172302e-01 -3.55498374e-01 6.42608404e-01 2.34662682e-01 -4.74039055e-02 1.21213877e+00 -9.49716210e-01 -6.37771726e-01 1.95002690e-01 1.16116054e-01 -1.98200494e-01 -5.76123595e-02 -3.71099114e-01 -5.60250282e-01 6.62166923e-02 7.41003990e-01 -6.96384072e-01 1.91427574e-01 2.86798239e-01 7.58970678e-01 -4.89224821e-01 -9.53716561e-02 4.26159427e-02 1.01397002e+00 4.68316287e-01 8.85873199e-01 4.17210847e-01 4.46617045e-02 7.80303717e-01 5.87896705e-01 6.45402431e-01 9.09648776e-01 5.29407740e-01 6.78407490e-01 7.26684511e-01 4.45207566e-01 -4.32654053e-01 3.09328765e-01 -2.85230786e-01 -2.63227016e-01 -2.12508336e-01 -9.64956105e-01 4.91090953e-01 -1.90471530e+00 -1.42436862e+00 -2.59895593e-01 2.29059339e+00 5.12583137e-01 3.57945293e-01 7.11648259e-04 -1.99197590e-01 7.45114088e-01 1.78094395e-02 -6.61988616e-01 -4.82779115e-01 3.13017726e-01 -4.15868789e-01 1.54618248e-01 9.16875958e-01 -4.69795763e-01 1.08955550e+00 6.65543318e+00 2.09495679e-01 -2.14707151e-01 3.06950361e-02 7.51294076e-01 5.01853134e-03 -4.90729511e-01 4.15050834e-01 -4.07816380e-01 1.99930891e-01 7.25858927e-01 -3.23693782e-01 9.70281065e-01 8.31150413e-01 3.15735072e-01 -6.64999366e-01 -1.46697855e+00 6.36592925e-01 -1.51950464e-01 -6.66496277e-01 -3.17647576e-01 3.09474729e-02 -4.69097681e-02 -5.51004708e-01 -2.07115293e-01 3.19224745e-01 1.34444714e+00 -1.29943228e+00 1.13720906e+00 2.49135792e-01 -1.20717913e-01 -5.13836324e-01 5.50517619e-01 1.18408465e+00 -2.17919275e-01 -9.47652981e-02 -6.13198400e-01 -1.26243854e+00 1.76994875e-01 2.10260861e-02 -7.72100687e-01 -1.69156089e-01 3.98434401e-01 -8.35063085e-02 2.62692198e-03 4.61906165e-01 -7.25672901e-01 2.86897629e-01 -4.18191224e-01 -1.06570207e-01 1.76619485e-01 -4.29062009e-01 6.24375939e-01 1.91970184e-01 2.32014671e-01 7.87634790e-01 4.17127088e-02 1.55373085e+00 5.35957932e-01 -4.26846445e-01 -4.69770223e-01 -1.37924001e-01 6.39403939e-01 6.38415575e-01 -8.11615169e-01 -3.16410303e-01 -2.53688805e-02 7.02263713e-01 3.22072268e-01 4.03685480e-01 -9.40999448e-01 2.92540371e-01 1.01681507e+00 7.32321367e-02 -4.04773980e-01 -2.88366109e-01 -5.34211278e-01 -1.31902635e+00 -3.52891058e-01 -1.00346649e+00 4.74059992e-02 -1.46130359e+00 -1.07104361e+00 1.21139750e-01 1.29195660e-01 -4.66840744e-01 -4.50846910e-01 -4.88464981e-01 -5.47964394e-01 9.78145480e-01 -6.96009994e-01 -8.81004095e-01 -2.15927884e-02 3.70866358e-01 4.94475335e-01 3.20966333e-01 9.08632636e-01 -7.47650206e-01 -2.81859010e-01 -9.22658741e-02 -6.81467116e-01 -1.52050585e-01 4.34883624e-01 -1.13501525e+00 4.38634932e-01 8.01289439e-01 -1.61440596e-01 1.05618036e+00 1.14395225e+00 -7.48278260e-01 -9.47346866e-01 -1.63234144e-01 9.27519739e-01 -4.88626093e-01 6.80336297e-01 1.89930543e-01 -1.04505026e+00 1.32209504e+00 5.12366779e-02 -7.23379731e-01 7.24296451e-01 4.02420104e-01 -5.83460033e-01 7.35224068e-01 -1.58808947e+00 1.13439608e+00 1.09294617e+00 -3.62931371e-01 -1.42179358e+00 4.89633948e-01 3.20016950e-01 -5.21184802e-01 -2.14471802e-01 -4.04397905e-01 4.76579219e-01 -1.78158748e+00 9.63099837e-01 -1.17787635e+00 5.31638086e-01 -3.64842832e-01 -1.89206883e-01 -1.45561635e+00 -7.95803607e-01 -2.39173278e-01 4.60867256e-01 6.39270008e-01 -2.94593144e-02 -8.08187664e-01 6.32346034e-01 1.65533674e+00 5.31766983e-03 -2.03887224e-02 -6.93751633e-01 -5.20198882e-01 7.87884146e-02 -8.21037292e-01 8.30691874e-01 9.61976647e-01 6.18333519e-01 1.11970574e-01 -2.89353967e-01 4.35319334e-01 1.02423084e+00 -9.06615034e-02 5.69722474e-01 -1.29952157e+00 -5.13404250e-01 -5.78058720e-01 -2.59710342e-01 -1.06279993e+00 2.07181886e-01 -3.00369650e-01 2.19819635e-01 -1.35419345e+00 4.77317512e-01 -4.05986309e-01 3.32083553e-01 6.00731492e-01 -1.04685001e-01 -3.65552336e-01 4.18182254e-01 2.10406840e-01 -4.57442731e-01 7.06539899e-02 9.26517069e-01 1.35214612e-01 1.50746688e-01 6.09523728e-02 -1.40157938e+00 1.60275686e+00 9.09967065e-01 -5.42013049e-01 -5.53298116e-01 -3.64713997e-01 7.35681236e-01 5.93531191e-01 8.18928897e-01 -7.09529459e-01 3.59095842e-01 -1.03058147e+00 2.74923354e-01 -7.04456717e-02 4.47900116e-01 -8.77900720e-01 4.66757715e-01 5.89350462e-01 -4.23950016e-01 9.91028454e-03 5.58541343e-02 5.07472396e-01 3.31463277e-01 -6.80338502e-01 7.59614468e-01 -7.02227890e-01 -5.77701032e-01 -5.61627090e-01 -1.09553254e+00 -1.61914676e-02 1.09294784e+00 -3.64169598e-01 -5.83187938e-01 -9.35345769e-01 -1.08770299e+00 2.80275166e-01 9.11232710e-01 -1.07124344e-01 6.87750459e-01 -8.54374468e-01 -4.29846823e-01 2.08620839e-02 -1.21787898e-01 -2.84256399e-01 5.18009484e-01 7.21605420e-01 -7.62993574e-01 3.04821223e-01 -2.62326002e-01 1.03693837e-02 -1.03034961e+00 7.11621463e-01 4.63574111e-01 -5.29543944e-02 -4.18088198e-01 9.12936151e-01 6.83889747e-01 -2.95316190e-01 -1.18628323e-01 -2.24106014e-01 -2.51539759e-02 -5.48635960e-01 8.32121670e-01 5.04210353e-01 -3.97613704e-01 -4.32751566e-01 -4.16381508e-01 -1.11277848e-01 -1.38101056e-01 -1.32346138e-01 9.59770381e-01 -5.97199023e-01 -4.28846210e-01 4.74134862e-01 -1.97855487e-01 -6.90316260e-02 -1.09187651e+00 -5.76707609e-02 -1.23289146e-01 -8.07017446e-01 -2.32420206e-01 -8.88872504e-01 -2.83738319e-02 5.88311911e-01 -2.81343549e-01 4.52794522e-01 7.34949648e-01 -1.28626439e-03 1.50736317e-01 6.58777297e-01 1.25908387e+00 -9.01092649e-01 6.48248792e-02 4.77135807e-01 6.72636569e-01 -1.08275640e+00 1.25276238e-01 -2.83410728e-01 -9.17227864e-01 9.84259784e-01 7.00146377e-01 -2.66699065e-02 4.74719480e-02 4.48480323e-02 -2.82473624e-01 -2.83353657e-01 -1.15849674e+00 -2.14515533e-02 -5.42182446e-01 7.00053036e-01 3.19949985e-01 6.11891925e-01 -3.20812315e-01 8.63175452e-01 -4.49997813e-01 -4.53827716e-02 1.41190672e+00 8.33708167e-01 -8.83822501e-01 -6.68434858e-01 -9.64564621e-01 2.81759411e-01 -4.96633142e-01 -5.84192872e-02 -4.17627066e-01 7.12344468e-01 -2.79038131e-01 1.37693667e+00 1.64345860e-01 3.91560122e-02 6.03684410e-02 2.50069112e-01 7.73004889e-01 -8.09645653e-01 -2.16932207e-01 -1.55557841e-01 1.32915005e-01 -7.79260874e-01 -1.55083269e-01 -9.02459621e-01 -1.24121892e+00 -1.00311053e+00 -1.65710747e-01 2.81250924e-02 1.14517219e-01 1.20980823e+00 -7.29154870e-02 2.69331634e-01 -1.83488622e-01 -4.20575082e-01 -1.03187370e+00 -8.66059661e-01 -8.44648898e-01 4.07955229e-01 7.79609242e-03 -8.42854083e-01 -5.56758702e-01 -2.66711771e-01]
[9.520100593566895, 7.234376430511475]
39e0a94b-1eb3-44b5-8b8e-965ffaeca8d3
visual-story-post-editing
1906.01764
null
https://arxiv.org/abs/1906.01764v1
https://arxiv.org/pdf/1906.01764v1.pdf
Visual Story Post-Editing
We introduce the first dataset for human edits of machine-generated visual stories and explore how these collected edits may be used for the visual story post-editing task. The dataset, VIST-Edit, includes 14,905 human edited versions of 2,981 machine-generated visual stories. The stories were generated by two state-of-the-art visual storytelling models, each aligned to 5 human-edited versions. We establish baselines for the task, showing how a relatively small set of human edits can be leveraged to boost the performance of large visual storytelling models. We also discuss the weak correlation between automatic evaluation scores and human ratings, motivating the need for new automatic metrics.
["Ting-Hao 'Kenneth' Huang", 'Yen-Chia Hsu', 'Chieh-Yang Huang', 'Ting-Yao Hsu']
2019-06-05
visual-story-post-editing-1
https://aclanthology.org/P19-1658
https://aclanthology.org/P19-1658.pdf
acl-2019-7
['visual-storytelling']
['natural-language-processing']
[ 2.20914200e-01 5.70679665e-01 -1.80159643e-01 -3.09054703e-01 -6.27802730e-01 -8.82436931e-01 1.18818176e+00 3.44297707e-01 -2.21795246e-01 5.87144673e-01 9.87135887e-01 1.31670618e-02 4.19256896e-01 -5.06925344e-01 -8.39795053e-01 1.98375627e-01 5.18925153e-02 4.79142904e-01 3.62767816e-01 -1.94634423e-01 5.47078192e-01 -1.52440555e-02 -1.48765302e+00 8.63915145e-01 6.70138836e-01 3.35386485e-01 2.63737231e-01 1.19554818e+00 -2.53965229e-01 1.18540215e+00 -1.16090870e+00 -8.95955324e-01 -5.19116707e-02 -5.21315992e-01 -7.37998009e-01 -6.22010557e-04 7.81776190e-01 -2.51903147e-01 -4.66590375e-01 4.88454700e-01 5.83758950e-01 3.23982805e-01 1.07797635e+00 -1.40964639e+00 -1.40444410e+00 1.00739193e+00 -6.17759705e-01 1.22533694e-01 8.02876055e-01 4.55979705e-01 1.25805211e+00 -1.28029954e+00 1.53646874e+00 1.34074092e+00 7.42069006e-01 4.18649882e-01 -1.25885153e+00 -3.61761242e-01 -1.44586980e-01 3.31534743e-01 -8.54572177e-01 -5.17065942e-01 5.42728961e-01 -1.10528111e+00 1.18042386e+00 3.02973449e-01 8.94932151e-01 1.48435068e+00 -2.22207963e-01 7.79275954e-01 5.87219179e-01 -3.84920806e-01 -1.65870458e-01 1.48150817e-01 -1.23394325e-01 4.92636025e-01 2.45723408e-02 6.47560135e-02 -9.19426143e-01 8.49955063e-03 7.22993314e-01 -6.71501756e-01 -2.09677413e-01 -5.54615021e-01 -1.54985666e+00 7.90260494e-01 4.33384031e-01 8.54928195e-02 -1.21659100e-01 2.32407033e-01 9.44644868e-01 2.56222516e-01 4.82205927e-01 1.17411923e+00 1.28811300e-01 -5.43657362e-01 -1.09434462e+00 5.45203567e-01 2.56356597e-01 1.33528078e+00 3.06103557e-01 5.19590490e-02 -1.06439817e+00 1.03780401e+00 -1.60784245e-01 2.32548043e-01 7.89002050e-03 -9.97713387e-01 6.53738320e-01 4.44367290e-01 1.72980413e-01 -9.60735381e-01 -3.01917922e-02 3.57115678e-02 -4.32468444e-01 4.65876460e-01 7.98111483e-02 1.07066013e-01 -9.38643694e-01 1.38066399e+00 4.80886735e-03 -3.87072444e-01 -5.09732902e-01 4.09772903e-01 1.27897358e+00 6.73883438e-01 6.97583929e-02 -7.54401982e-02 9.44421649e-01 -1.43600297e+00 -7.34799743e-01 -1.69935986e-01 4.19178128e-01 -1.03752053e+00 1.69711161e+00 2.38369018e-01 -1.43155968e+00 -7.35353649e-01 -1.20000660e+00 -4.21379328e-01 -2.67810345e-01 2.02510059e-01 2.92959571e-01 1.54598057e-01 -1.20390165e+00 5.48774421e-01 -2.62638897e-01 -6.40095949e-01 6.60582125e-01 -6.00973129e-01 -4.89625067e-01 9.66744050e-02 -7.24938631e-01 1.28274262e+00 3.42117012e-01 -4.70963240e-01 -1.20095253e+00 -9.67138171e-01 -8.70816827e-01 -2.82348931e-01 1.26525640e-01 -6.42698407e-01 1.65612459e+00 -7.63040602e-01 -9.12657440e-01 1.16124296e+00 -1.80080548e-01 -1.91981763e-01 1.12212968e+00 -4.27629441e-01 -3.90027612e-01 1.33009657e-01 4.78803366e-01 9.46226597e-01 9.85439360e-01 -1.61446559e+00 -5.00229418e-01 4.67125207e-01 9.37173963e-02 1.61415413e-01 -2.92112291e-01 2.60711133e-01 -5.80551744e-01 -1.27555239e+00 -8.10658455e-01 -5.51315844e-01 1.86332520e-02 2.01064646e-01 -5.51900268e-01 3.31580155e-02 7.73930013e-01 -1.03166628e+00 1.42395699e+00 -2.10838747e+00 4.18323338e-01 -8.06693062e-02 4.97153938e-01 -1.47729263e-01 -5.39627373e-01 7.45421886e-01 -1.09159105e-01 4.66305792e-01 -8.51851255e-02 -8.46808612e-01 -7.18829706e-02 -3.48751307e-01 -3.98498237e-01 -8.45503137e-02 2.36920416e-02 1.22948718e+00 -9.98039007e-01 -6.52600765e-01 2.92553753e-01 3.10639709e-01 -5.17130613e-01 5.05079389e-01 -3.78463179e-01 1.81167558e-01 3.28542203e-01 2.94907629e-01 2.72261858e-01 -7.62472600e-02 -1.56604499e-01 -9.30345133e-02 -1.90312684e-01 2.30708659e-01 -4.86337990e-01 1.68582070e+00 -1.22854918e-01 1.59087884e+00 -5.29370129e-01 2.60387140e-04 8.28642130e-01 3.18096668e-01 -1.29774317e-01 -6.77961469e-01 -6.36548176e-02 -3.91571969e-01 -5.76871812e-01 -6.94362164e-01 1.20313561e+00 2.85974741e-02 -2.32262388e-01 7.35800385e-01 1.47161648e-01 -5.46905339e-01 6.32161796e-01 8.41873288e-01 1.08964837e+00 2.55536765e-01 4.97713774e-01 1.78978845e-01 -3.18132192e-01 3.85213107e-01 -3.56770097e-03 1.18090916e+00 7.12665692e-02 1.07799256e+00 6.18973374e-01 -4.05460596e-01 -1.72699630e+00 -1.33797121e+00 2.39551157e-01 1.34163988e+00 -1.02563508e-01 -9.28906918e-01 -6.03362322e-01 -6.21361375e-01 5.87891489e-02 1.23169088e+00 -1.17073023e+00 -1.66106328e-01 -3.85346919e-01 -4.80014272e-02 7.28269160e-01 7.72798717e-01 -1.10692419e-01 -1.50113618e+00 -7.64447451e-01 -5.26042134e-02 -2.59464175e-01 -8.55253696e-01 -6.46690607e-01 -3.41889054e-01 -3.82312626e-01 -1.05675101e+00 -8.91562939e-01 -6.36979997e-01 4.31555599e-01 3.61843109e-01 1.78250968e+00 3.04567456e-01 -1.75825447e-01 5.45733392e-01 -8.48232031e-01 -4.68137681e-01 -9.74495530e-01 8.29264224e-02 -1.38626710e-01 -5.65571606e-01 -1.65785640e-01 -5.23490071e-01 -1.53636754e-01 1.50148153e-01 -6.95830226e-01 8.13143611e-01 2.16234148e-01 6.59417987e-01 1.61466062e-01 -5.73429048e-01 1.58094272e-01 -9.84025180e-01 9.31381226e-01 -2.28320733e-01 1.77574772e-02 4.31534231e-01 -2.73628324e-01 -1.37876377e-01 2.87182361e-01 -5.35304666e-01 -1.22286057e+00 -3.47182304e-01 4.27875429e-01 -4.21623290e-01 2.68887848e-01 4.68495995e-01 2.72316456e-01 3.11527282e-01 1.06503999e+00 -3.92851979e-01 -5.70550025e-01 -4.47269559e-01 1.06721604e+00 3.43520641e-01 1.13005865e+00 -3.92813295e-01 1.14198983e+00 1.10129908e-01 -5.56646228e-01 -7.08703399e-01 -5.99288404e-01 -7.37284124e-02 -6.54060185e-01 -8.01337481e-01 9.12729263e-01 -8.61662149e-01 -1.76724881e-01 3.61786753e-01 -1.59377408e+00 -7.37613022e-01 -6.51374519e-01 -1.48282066e-01 -7.45455861e-01 3.27896416e-01 -6.12943590e-01 -5.15233576e-01 -2.08413392e-01 -6.84391141e-01 8.65011334e-01 -4.61293245e-03 -1.21133947e+00 -5.95399916e-01 5.23315728e-01 3.10912251e-01 2.41594478e-01 7.41522193e-01 1.08969474e+00 -2.11942926e-01 -2.75066853e-01 -1.34244204e-01 -5.33254385e-01 -1.43761143e-01 -8.20332915e-02 7.26718009e-01 -8.65915954e-01 -8.91290531e-02 -8.76693189e-01 -7.13718176e-01 9.16695714e-01 2.87451029e-01 8.53771091e-01 -1.99229136e-01 -2.77583420e-01 3.84800494e-01 9.31786776e-01 3.49143110e-02 9.50927615e-01 4.37676251e-01 1.18924308e+00 3.53840053e-01 7.86577404e-01 7.87664831e-01 4.02654380e-01 7.39174902e-01 3.21059734e-01 -1.11529782e-01 -5.37694395e-01 -1.02938676e+00 2.53785878e-01 8.18528235e-01 -4.28042650e-01 -5.45622766e-01 -9.92666006e-01 9.28528666e-01 -1.87738574e+00 -1.27020431e+00 -3.73679221e-01 1.84265196e+00 8.59890282e-01 3.44085187e-01 2.68389195e-01 5.78829553e-03 9.05779362e-01 6.84007168e-01 -3.63883168e-01 -6.90727949e-01 -4.33180958e-01 6.73434362e-02 1.38541490e-01 4.58609134e-01 -8.59308422e-01 1.02841914e+00 8.34043407e+00 7.43974745e-01 -4.42138404e-01 1.73474014e-01 3.03143978e-01 -5.74565172e-01 -7.20874608e-01 2.15393543e-01 -1.17833845e-01 3.25744003e-01 3.86807680e-01 -4.36066598e-01 4.28089529e-01 9.07250226e-01 4.05669868e-01 -1.32678971e-01 -1.35883296e+00 9.46575224e-01 3.55729401e-01 -1.82330096e+00 5.23832142e-01 -3.17731917e-01 1.23300314e+00 -2.29832947e-01 -5.39832413e-02 4.57553476e-01 6.81322455e-01 -1.15127587e+00 1.55136251e+00 7.61437595e-01 1.47915256e+00 -6.16894662e-01 1.58028573e-01 -2.29183376e-01 -1.16135764e+00 2.90088981e-01 -3.95929329e-02 -2.70100325e-01 7.33398438e-01 4.64423835e-01 -9.72394645e-01 2.86362451e-02 8.03029716e-01 1.20635569e+00 -1.19587111e+00 1.01021039e+00 -5.46432674e-01 3.31076533e-01 4.28056985e-01 -8.53082761e-02 -1.93678588e-01 4.07270461e-01 8.84862185e-01 1.64017773e+00 2.92879045e-01 -3.40806961e-01 -3.93349111e-01 9.15515900e-01 -3.88239086e-01 3.88702042e-02 -7.88154662e-01 -5.12814879e-01 6.75712526e-01 1.08390677e+00 -5.35264432e-01 -7.69315720e-01 3.49126547e-03 1.27695322e+00 7.59211957e-01 4.32548225e-01 -1.27594900e+00 -4.94060814e-01 3.80266696e-01 4.27527219e-01 2.65644073e-01 -1.31267712e-01 -7.14913905e-01 -9.95021045e-01 -6.92804530e-02 -1.01255870e+00 2.37835318e-01 -1.76757920e+00 -1.22308552e+00 5.58333397e-01 8.90094042e-02 -1.31913102e+00 -1.26335204e-01 -1.46747991e-01 -1.10071242e+00 3.29525143e-01 -6.00403845e-01 -1.18889737e+00 -7.17781901e-01 1.33951440e-01 8.84442389e-01 -1.40649736e-01 4.86234039e-01 -9.24097523e-02 -2.13534459e-01 7.98425376e-01 -1.35864988e-01 4.41650450e-02 1.29168224e+00 -1.33660471e+00 1.29514599e+00 9.95684147e-01 1.92755997e-01 1.60582602e-01 1.18502021e+00 -1.01312959e+00 -4.28603739e-01 -1.04525650e+00 9.91482079e-01 -1.01189578e+00 5.36524773e-01 -5.27996063e-01 -7.81908512e-01 9.03990865e-01 8.87517512e-01 -7.32387662e-01 5.25940120e-01 1.28303811e-01 -7.16826975e-01 6.95769966e-01 -8.57209444e-01 1.07789540e+00 1.59999287e+00 -7.63783753e-01 -8.53557229e-01 1.34835139e-01 1.05444860e+00 -5.09067774e-01 -7.00054228e-01 1.28711700e-01 7.10562050e-01 -9.51947689e-01 8.97747040e-01 -8.97469044e-01 1.47786546e+00 -2.39866897e-01 3.08649153e-01 -1.72857201e+00 -6.37340963e-01 -7.66028404e-01 -1.96074679e-01 1.43924761e+00 7.03344285e-01 1.10679157e-01 2.45226413e-01 5.38577378e-01 -3.25805575e-01 -1.63649559e-01 -5.42803943e-01 -5.69349349e-01 -2.50192940e-01 -4.13652539e-01 5.26423931e-01 1.31071579e+00 3.86119097e-01 4.38575596e-01 -8.06956232e-01 -7.90183306e-01 2.80465305e-01 -3.12241197e-01 1.27837694e+00 -1.01595712e+00 -4.05412704e-01 -6.22409403e-01 -2.78665453e-01 -6.50306523e-01 -2.24980906e-01 -8.35807562e-01 1.26373038e-01 -1.99328649e+00 7.18088210e-01 9.24456716e-02 1.50694743e-01 4.95022833e-01 -3.97098422e-01 3.26138467e-01 5.50193369e-01 1.73428386e-01 -8.11021507e-01 7.32342005e-01 1.33925259e+00 -3.54503363e-01 -2.72944123e-01 -7.11077511e-01 -5.94078124e-01 5.26045740e-01 5.15257120e-01 -5.09311557e-01 -3.46437722e-01 -7.53114700e-01 6.08375788e-01 -2.46092245e-01 3.05288672e-01 -1.02905393e+00 -3.67441736e-02 -3.35551709e-01 4.84519124e-01 -7.19891310e-01 3.49561512e-01 -6.25075623e-02 4.56874102e-01 -1.16036236e-01 -8.26348782e-01 4.79737669e-01 1.66375160e-01 3.93133551e-01 -9.04733315e-03 -6.10902235e-02 6.22534931e-01 -1.04002230e-01 -7.34357119e-01 -1.22758076e-01 -5.69065630e-01 4.78350699e-01 1.11684692e+00 -5.86936235e-01 -8.48457873e-01 -8.14416945e-01 -6.10082746e-01 1.39944717e-01 1.00142586e+00 9.61974263e-01 6.81030571e-01 -1.85078561e+00 -9.53969717e-01 -5.07934093e-01 7.63630867e-01 -3.66171688e-01 1.96532995e-01 1.08540684e-01 -7.76344657e-01 -2.00232849e-01 -5.99464655e-01 -1.97504178e-01 -1.47917783e+00 7.11551368e-01 -3.19641411e-01 -1.10092968e-01 -6.54643655e-01 9.62788284e-01 1.47171870e-01 1.81029178e-02 1.14401370e-01 -1.17757462e-01 -1.47152826e-01 3.08339685e-01 7.79129624e-01 6.61900938e-01 -3.71435523e-01 -6.68674231e-01 -3.29428166e-02 3.35075915e-01 -7.46262893e-02 -5.87965906e-01 1.33806002e+00 -1.40441228e-02 1.87799677e-01 7.60385334e-01 8.88797760e-01 1.37789622e-01 -1.30211866e+00 1.68411493e-01 -1.09280862e-01 -6.53570354e-01 -4.81540352e-01 -8.23135734e-01 -7.32084513e-01 8.14298689e-01 3.66384089e-02 7.07976818e-02 6.55464649e-01 2.86058277e-01 5.66427112e-01 2.25526616e-01 1.43712685e-01 -1.39769244e+00 5.01447201e-01 5.78456879e-01 1.70285475e+00 -1.13916600e+00 2.78775275e-01 -2.88947493e-01 -1.24151552e+00 8.68507206e-01 8.07893693e-01 5.67513220e-02 -5.36864810e-02 1.71476960e-01 1.46722309e-02 -2.88979828e-01 -9.04648542e-01 1.91801116e-01 4.95856136e-01 8.55604053e-01 5.27631879e-01 2.67203361e-01 -7.93597177e-02 4.93306369e-01 -7.13414311e-01 -8.46835971e-02 8.91017139e-01 7.48150587e-01 -2.96680599e-01 -5.61627626e-01 -1.30093560e-01 5.53984404e-01 2.09091902e-01 -1.76068082e-01 -1.11572230e+00 7.34754801e-01 -1.64439037e-01 8.71616781e-01 1.27910480e-01 -5.98625898e-01 5.16464114e-01 -1.69046804e-01 4.91536766e-01 -7.21826792e-01 -7.78176427e-01 -5.05227625e-01 6.75740361e-01 -4.80557650e-01 -9.61637050e-02 -6.65517151e-01 -8.46109927e-01 -5.32174051e-01 4.59091999e-02 -3.95365804e-01 1.30468041e-01 4.77209806e-01 3.33130002e-01 5.69948375e-01 6.49598092e-02 -1.17930293e+00 4.07697201e-01 -1.19077098e+00 -1.25391126e-01 1.00503910e+00 1.12547502e-01 -6.88094199e-01 -2.88500339e-01 6.24606490e-01]
[11.199010848999023, 0.865492045879364]
6546f864-c575-4b66-b1ba-e71d920556a9
end-to-end-segmentation-with-recurrent
1812.02068
null
https://arxiv.org/abs/1812.02068v2
https://arxiv.org/pdf/1812.02068v2.pdf
Brain Segmentation from k-space with End-to-end Recurrent Attention Network
The task of medical image segmentation commonly involves an image reconstruction step to convert acquired raw data to images before any analysis. However, noises, artifacts and loss of information due to the reconstruction process are almost inevitable, which compromises the final performance of segmentation. We present a novel learning framework that performs magnetic resonance brain image segmentation directly from k-space data. The end-to-end framework consists of a unique task-driven attention module that recurrently utilizes intermediate segmentation estimation to facilitate image-domain feature extraction from the raw data, thus closely bridging the reconstruction and the segmentation tasks. In addition, to address the challenge of manual labeling, we introduce a novel workflow to generate labeled training data for segmentation by exploiting imaging modality simulators and digital phantoms. Extensive experimental results show that the proposed method outperforms several state-of-the-art methods.
['Mariappan S. Nadar', 'Xiao Chen', 'Qiaoying Huang', 'Dimitris Metaxas']
2018-12-05
null
null
null
null
['brain-image-segmentation']
['medical']
[ 6.20540679e-01 2.16897860e-01 2.49745294e-01 -7.74036467e-01 -1.16649365e+00 -2.23307863e-01 8.39877278e-02 -6.00009002e-02 -7.80789375e-01 5.37991345e-01 -4.05427031e-02 -2.71792024e-01 1.19595401e-01 -4.23306972e-01 -6.45494282e-01 -7.66808331e-01 2.46954784e-01 4.05887097e-01 9.13949460e-02 2.84464240e-01 1.79532707e-01 4.03433532e-01 -8.96137655e-01 3.49088281e-01 1.02327025e+00 1.00464046e+00 6.21390522e-01 2.47960880e-01 -2.76305169e-01 7.71306157e-01 -4.21893418e-01 -8.29414930e-03 1.50511652e-01 -4.78261858e-01 -1.08023381e+00 5.04585564e-01 -7.82774314e-02 -4.94680822e-01 -1.61554754e-01 1.28956807e+00 7.97444701e-01 1.06188385e-02 4.16805744e-01 -7.72609651e-01 -4.14377868e-01 6.55279279e-01 -5.91124654e-01 3.05149496e-01 2.90826205e-02 2.43680045e-01 2.82308847e-01 -7.58766472e-01 6.07544780e-01 6.76015496e-01 6.29869282e-01 4.83452797e-01 -1.21171212e+00 -6.68025136e-01 -1.29097076e-02 -5.16418256e-02 -1.26181412e+00 -2.64880747e-01 9.77380335e-01 -7.16253698e-01 4.23976153e-01 -3.98901403e-02 6.82481945e-01 7.55904198e-01 2.39695400e-01 1.04439938e+00 1.33564305e+00 -2.18520403e-01 2.81562865e-01 -1.43321678e-01 4.31151658e-01 6.06972456e-01 -7.55188987e-02 -2.07129389e-01 1.52485505e-01 2.49467358e-01 9.81907248e-01 4.74038385e-02 -5.34720778e-01 -3.38362098e-01 -1.34295201e+00 4.69469965e-01 6.69300199e-01 4.40812320e-01 -7.87830889e-01 -3.20461541e-02 4.29403096e-01 -1.35316029e-01 3.42041790e-01 1.80988505e-01 -2.51237929e-01 3.10337573e-01 -1.30700147e+00 -9.09012258e-02 3.71442437e-01 7.15043366e-01 6.78378105e-01 -4.95216772e-02 -3.33623409e-01 6.68424368e-01 2.37061501e-01 2.03619421e-01 8.34236801e-01 -6.95340037e-01 2.72587806e-01 4.80116844e-01 -6.84594736e-02 -4.99929488e-01 -8.02644074e-01 -6.11276686e-01 -8.74610484e-01 7.38751749e-03 4.44142371e-01 -2.28690654e-01 -1.37204969e+00 1.45029938e+00 5.09785891e-01 4.08584923e-01 -9.85923856e-02 1.32711756e+00 1.10583735e+00 2.17095688e-01 3.61353010e-01 -3.15494925e-01 1.19971418e+00 -1.05436099e+00 -8.87458861e-01 -2.56019413e-01 5.58317661e-01 -6.06084585e-01 1.10847533e+00 1.44577637e-01 -1.07532990e+00 -6.67677820e-01 -1.02125311e+00 -1.48052290e-01 2.81726345e-02 2.45708764e-01 5.30388594e-01 4.81954187e-01 -8.25171113e-01 5.25378823e-01 -1.05969703e+00 1.80124745e-01 8.35895777e-01 3.97469252e-01 -3.36090356e-01 -8.16635862e-02 -9.58857894e-01 6.71967447e-01 6.69183969e-01 4.33063269e-01 -9.20667648e-01 -1.03464258e+00 -7.73916125e-01 -2.75008649e-01 3.82711232e-01 -5.93689084e-01 1.34473681e+00 -8.71623218e-01 -1.55623651e+00 1.17659426e+00 9.94392335e-02 -3.71347427e-01 8.05685163e-01 -2.43941769e-01 -2.17482954e-01 3.87658030e-01 3.78254086e-01 6.36797905e-01 8.82007301e-01 -1.28729379e+00 -1.47857964e-01 -4.52668488e-01 -2.90263712e-01 1.67048872e-01 3.64070684e-01 -1.27892122e-01 -6.70824468e-01 -6.68213010e-01 4.43556666e-01 -7.16298640e-01 -6.87860906e-01 -1.78784505e-01 -5.85178256e-01 2.42185250e-01 4.81123745e-01 -1.02879798e+00 7.63806224e-01 -2.25169516e+00 -4.25894819e-02 2.01112717e-01 4.22135293e-01 1.45910606e-01 -5.35306856e-02 -3.22003961e-01 -3.70445013e-01 -1.83108747e-01 -7.29738116e-01 -3.36522132e-01 -3.65121871e-01 -1.00686945e-01 -1.85941868e-02 6.92606449e-01 1.09320894e-01 1.08197141e+00 -8.56546521e-01 -6.19143367e-01 4.94513124e-01 4.31525409e-01 -4.47791159e-01 4.48892891e-01 3.42312977e-02 1.45740199e+00 -5.85649729e-01 3.48616451e-01 8.10702801e-01 -4.37668204e-01 5.45062609e-02 -6.61600053e-01 8.12247396e-02 1.31577961e-02 -9.30925250e-01 2.33169293e+00 -5.28477073e-01 1.08610094e-01 3.02820474e-01 -1.48596358e+00 4.76038367e-01 4.33754593e-01 9.72477555e-01 -8.26612234e-01 5.90248704e-01 2.47813269e-01 -1.18692983e-02 -8.37674737e-01 4.14956175e-02 -2.50788927e-01 -6.94634616e-02 5.41754901e-01 1.55468196e-01 -3.32031608e-01 -1.16539046e-01 3.69881280e-02 8.11724424e-01 2.65570253e-01 1.38848409e-01 -2.85923988e-01 6.42887294e-01 5.99579327e-02 6.52617931e-01 6.33624673e-01 -4.40615863e-01 8.97395670e-01 2.14768201e-01 -4.23890710e-01 -9.70303595e-01 -1.04221523e+00 -3.77113193e-01 5.21028876e-01 2.42251962e-01 2.31156886e-01 -1.38135469e+00 -7.89223552e-01 -4.82633948e-01 5.96067965e-01 -6.66481197e-01 -4.50819060e-02 -6.89270735e-01 -8.77266347e-01 2.73272306e-01 4.83553857e-01 7.18138814e-01 -1.31612802e+00 -8.12942863e-01 6.02318645e-01 -4.97312635e-01 -1.51593673e+00 -5.87409973e-01 7.43117258e-02 -9.73052323e-01 -1.10651934e+00 -9.76440787e-01 -8.99078071e-01 1.06146789e+00 1.18866697e-01 9.37806427e-01 3.74034047e-02 -6.53590798e-01 2.15621322e-01 -1.22415282e-01 -1.18591227e-01 -4.48421061e-01 -2.03830898e-02 -3.82541448e-01 1.12993695e-01 1.27413375e-02 -4.74796891e-01 -8.58690083e-01 5.84036410e-02 -1.03667438e+00 4.58876759e-01 8.55055571e-01 8.85440648e-01 9.53546107e-01 -1.11748174e-01 6.26482427e-01 -1.20503187e+00 4.73683685e-01 -3.67790818e-01 -5.57752490e-01 2.67110288e-01 -2.12014437e-01 6.42633140e-02 6.37659013e-01 -2.88561434e-01 -1.07591295e+00 5.79458117e-01 -4.37750190e-01 -3.88095826e-01 -4.23979163e-01 5.42590201e-01 -1.82946920e-01 -1.43141314e-01 3.96087646e-01 4.89250779e-01 2.70107538e-01 -4.56069410e-01 3.41497362e-01 6.42314374e-01 9.63218927e-01 -2.93139905e-01 3.78887206e-01 6.80885613e-01 -3.22083116e-01 -7.20822096e-01 -9.67427969e-01 -4.01953340e-01 -9.38964546e-01 -3.73067647e-01 1.20194376e+00 -7.92351007e-01 -4.27473336e-01 6.24735713e-01 -1.17542899e+00 -4.81970787e-01 -2.19772130e-01 6.67789698e-01 -7.27163851e-01 5.63969016e-01 -7.08810031e-01 -3.26014578e-01 -6.99134886e-01 -1.91384721e+00 1.12590528e+00 2.19899938e-01 1.31742761e-01 -7.55843639e-01 -2.59371281e-01 8.27359974e-01 3.12052637e-01 3.10574740e-01 8.29479694e-01 -6.00535095e-01 -6.45616472e-01 -1.49149656e-01 -3.86643589e-01 4.68333364e-01 2.22439393e-01 -7.97092438e-01 -9.86375868e-01 -2.78373033e-01 4.56472397e-01 -3.61580223e-01 6.24376476e-01 8.53736460e-01 1.54333985e+00 2.19919488e-01 -2.45365530e-01 8.39512646e-01 1.32484078e+00 1.70368850e-01 6.40609622e-01 6.12169057e-02 8.78796458e-01 5.09675443e-01 4.38863009e-01 8.54542851e-02 3.30625445e-01 3.71514380e-01 1.52619436e-01 -7.08983004e-01 -3.40254724e-01 1.36862472e-01 -1.29112497e-01 8.96474361e-01 3.47640514e-01 3.19957554e-01 -9.94730055e-01 5.22816658e-01 -1.61664033e+00 -5.22339165e-01 -8.11642781e-02 1.91939187e+00 9.34144855e-01 -1.72123071e-02 -1.77440256e-01 9.43988562e-02 8.09687436e-01 -1.22058056e-01 -7.72026122e-01 3.32651347e-01 4.79497880e-01 3.60050440e-01 4.97175485e-01 4.69797969e-01 -1.26633930e+00 1.00057769e+00 6.37399530e+00 7.69038379e-01 -1.47563529e+00 4.49443340e-01 8.33437026e-01 2.67667472e-01 -3.82715203e-02 -3.42885494e-01 -1.86785027e-01 5.93477845e-01 5.54965496e-01 1.90462649e-01 3.67471129e-01 7.37196982e-01 3.85154366e-01 -1.51624471e-01 -1.00024307e+00 1.11822331e+00 -1.26225978e-01 -1.23003793e+00 -2.40997478e-01 -3.01197678e-01 6.16394162e-01 1.37290135e-01 -7.87545964e-02 3.90460230e-02 7.70396218e-02 -8.93983901e-01 7.56387770e-01 5.01633883e-01 8.17613780e-01 -5.29227257e-01 6.96081579e-01 3.61172140e-01 -8.62636924e-01 2.05758333e-01 -1.02242984e-01 4.76284772e-01 4.97076899e-01 9.33898211e-01 -8.89729679e-01 6.48086250e-01 4.63058710e-01 5.01261175e-01 -3.87069613e-01 1.17352057e+00 -1.92167655e-01 5.12572825e-01 -3.96896675e-02 6.32769406e-01 3.54557008e-01 -2.55367458e-01 2.27014482e-01 1.15641284e+00 2.25097197e-03 3.08705807e-01 4.51257646e-01 1.09691644e+00 -7.62573928e-02 1.89367041e-01 -1.73855528e-01 2.53200948e-01 1.04364000e-01 1.43812025e+00 -1.23346031e+00 -5.02612293e-01 -4.13597018e-01 1.17947984e+00 2.07196653e-01 4.25263852e-01 -9.48984563e-01 -2.66671121e-01 6.97793588e-02 1.19800568e-01 5.52972071e-02 -1.43587813e-01 -5.28077483e-01 -1.11918581e+00 2.97707263e-02 -6.81841671e-01 8.70851800e-02 -6.80406392e-01 -1.11121213e+00 7.72085309e-01 -2.03197643e-01 -1.07675469e+00 -4.54813167e-02 -1.72318697e-01 -4.93391305e-01 8.61977458e-01 -1.61928058e+00 -9.96917367e-01 -5.15852273e-01 6.08318508e-01 5.45206487e-01 1.58512294e-01 5.33394098e-01 6.07227027e-01 -5.83322227e-01 3.62661600e-01 -1.09520676e-02 5.45159698e-01 3.87385160e-01 -9.40820098e-01 3.87551516e-01 8.84429216e-01 -2.11565360e-01 3.75063419e-01 3.15096378e-01 -6.56852961e-01 -9.53410387e-01 -1.19822192e+00 8.52355585e-02 7.36416578e-02 3.78534257e-01 -2.01333702e-01 -9.57375944e-01 5.33640265e-01 2.49106847e-02 3.88704062e-01 7.22241104e-01 -6.76511943e-01 1.95117474e-01 8.16377476e-02 -1.40005493e+00 4.21533078e-01 7.60030985e-01 -5.22299290e-01 -6.19746804e-01 3.57251853e-01 6.69193447e-01 -7.12849677e-01 -9.05337811e-01 4.61132705e-01 3.07371318e-01 -7.40630984e-01 1.06980097e+00 -2.62793601e-01 3.16114396e-01 -3.22645605e-01 2.71214604e-01 -1.21352684e+00 -3.36246014e-01 -3.50554109e-01 4.47737217e-01 8.62395525e-01 2.14320779e-01 -3.55435938e-01 7.60518968e-01 8.18520427e-01 -4.66220111e-01 -5.72579622e-01 -9.49946821e-01 -3.16220015e-01 1.23947553e-01 -5.72292864e-01 4.71338928e-01 1.00604856e+00 -3.60382318e-01 2.03024402e-01 -9.07570124e-02 2.38877252e-01 1.05380452e+00 3.31566125e-01 2.15698168e-01 -9.10331786e-01 -1.87783360e-01 -3.39767814e-01 -2.85099655e-01 -1.07042503e+00 2.87395447e-01 -1.17296827e+00 4.33504134e-01 -1.61953104e+00 1.87482670e-01 -5.29163361e-01 -4.06625688e-01 2.29910716e-01 -3.92630845e-01 5.02675891e-01 -8.03846575e-04 1.75307974e-01 -5.45423687e-01 3.85784000e-01 1.81915426e+00 -1.66349977e-01 -2.14731425e-01 -9.32711288e-02 -5.41576564e-01 7.03209460e-01 7.54802823e-01 -5.32242537e-01 -5.10249019e-01 -6.63159966e-01 -6.30942166e-01 4.25338328e-01 3.92102748e-01 -1.14795208e+00 2.35153154e-01 1.79151714e-01 4.56204206e-01 -6.00419879e-01 -4.50796075e-02 -9.11228716e-01 2.00396199e-02 5.46434760e-01 -4.92997855e-01 -3.72198790e-01 1.65261030e-02 3.47834647e-01 -1.30301803e-01 -3.02798331e-01 1.27660680e+00 -3.99994522e-01 -5.53972542e-01 5.01905441e-01 -4.28061396e-01 2.08688855e-01 1.00478995e+00 -1.39326587e-01 4.01715845e-01 -2.30458658e-02 -1.23166978e+00 1.77408516e-01 8.21904317e-02 7.51737580e-02 5.71036696e-01 -1.03335965e+00 -6.82777822e-01 3.76737982e-01 -2.29119167e-01 3.54680032e-01 7.18719661e-01 1.21745205e+00 -6.29432976e-01 1.47214711e-01 -3.27043444e-01 -7.63247490e-01 -7.59283364e-01 5.58814764e-01 5.68406880e-01 -3.28785002e-01 -1.11653674e+00 7.20154047e-01 3.13540816e-01 -5.16543329e-01 1.42218590e-01 -5.59367418e-01 -1.26394048e-01 -1.83895886e-01 5.50012290e-01 -5.99055886e-02 3.64959121e-01 -6.66384816e-01 -3.59569967e-01 5.36116302e-01 -3.21475863e-01 -1.19769782e-01 1.32715821e+00 -2.00568289e-01 -7.93391541e-02 8.25264081e-02 1.30168259e+00 -6.00891590e-01 -1.35105622e+00 -3.81163746e-01 -2.72151250e-02 -1.83395550e-01 4.40302730e-01 -9.34892356e-01 -1.52157068e+00 1.03395665e+00 9.04270530e-01 -2.16233656e-01 1.18834269e+00 -1.38888866e-01 1.14788842e+00 2.50608418e-02 4.84807879e-01 -1.12411809e+00 -1.63904354e-01 1.11167550e-01 6.86929286e-01 -1.34204900e+00 -1.96899533e-01 -5.30420303e-01 -7.08920479e-01 8.58010709e-01 4.24230933e-01 -1.14485033e-01 7.51626194e-01 4.64879453e-01 4.32171881e-01 -2.10820287e-01 1.28511086e-01 -1.45853102e-01 2.00714722e-01 5.32788098e-01 4.36865181e-01 -5.43906577e-02 -4.06426102e-01 9.60559368e-01 2.98074465e-02 4.88049150e-01 3.04732651e-01 8.89787138e-01 -1.62724555e-01 -8.96814227e-01 -3.39173675e-01 4.56423879e-01 -7.04353988e-01 -1.16399676e-01 2.10875750e-01 3.80517632e-01 8.52765664e-02 7.33107090e-01 -7.95997083e-02 -1.81538593e-02 2.17752293e-01 3.47331278e-02 4.95814025e-01 -5.28539121e-01 -6.65065706e-01 3.26640785e-01 -3.98604274e-01 -7.31478989e-01 -3.76013279e-01 -4.48432863e-01 -1.80503047e+00 2.31508270e-01 -1.66491881e-01 1.19438604e-01 7.13825285e-01 1.14073026e+00 2.45927572e-01 1.00088692e+00 5.56947052e-01 -1.07189894e+00 -3.42332006e-01 -9.02522326e-01 -4.49148864e-01 6.86676681e-01 2.47124687e-01 -5.30972242e-01 1.78851739e-01 3.60783219e-01]
[14.544795989990234, -2.2090208530426025]
dbffc506-8709-4631-8b1e-01823992f3ec
the-power-of-typed-affine-decision-structures
2304.14888
null
https://arxiv.org/abs/2304.14888v1
https://arxiv.org/pdf/2304.14888v1.pdf
The Power of Typed Affine Decision Structures: A Case Study
TADS are a novel, concise white-box representation of neural networks. In this paper, we apply TADS to the problem of neural network verification, using them to generate either proofs or concise error characterizations for desirable neural network properties. In a case study, we consider the robustness of neural networks to adversarial attacks, i.e., small changes to an input that drastically change a neural networks perception, and show that TADS can be used to provide precise diagnostics on how and where robustness errors a occur. We achieve these results by introducing Precondition Projection, a technique that yields a TADS describing network behavior precisely on a given subset of its input space, and combining it with PCA, a traditional, well-understood dimensionality reduction technique. We show that PCA is easily compatible with TADS. All analyses can be implemented in a straightforward fashion using the rich algebraic properties of TADS, demonstrating the utility of the TADS framework for neural network explainability and verification. While TADS do not yet scale as efficiently as state-of-the-art neural network verifiers, we show that, using PCA-based simplifications, they can still scale to mediumsized problems and yield concise explanations for potential errors that can be used for other purposes such as debugging a network or generating new training samples.
['Bernhard Steffen', 'Alnis Murtovi', 'Maximilian Schlüter', 'Gerrit Nolte']
2023-04-28
null
null
null
null
['dimensionality-reduction']
['methodology']
[ 4.94121939e-01 3.20276678e-01 -8.27094316e-02 -1.24748953e-01 -1.63668260e-01 -9.49497759e-01 5.03739774e-01 -8.80618915e-02 -1.28116772e-01 6.22002780e-01 -4.91398335e-01 -9.90631461e-01 -2.48479083e-01 -6.61227643e-01 -1.28324234e+00 -7.50874937e-01 -2.83147603e-01 8.12501535e-02 8.49400535e-02 -1.15482032e-01 6.67765364e-02 9.45544302e-01 -1.66892254e+00 -7.43607655e-02 3.19688171e-01 7.63585865e-01 -6.22745156e-01 1.12537158e+00 3.54214430e-01 5.58234930e-01 -7.24874854e-01 -6.14850402e-01 4.86375540e-01 -1.44745424e-01 -6.68472588e-01 -3.48317415e-01 7.40425348e-01 -3.57180208e-01 -3.75780940e-01 1.31268883e+00 1.32589608e-01 -2.45369792e-01 2.43401363e-01 -1.96858454e+00 -5.08661330e-01 8.75114381e-01 1.32076768e-02 -9.96324122e-02 9.04280543e-02 1.04894519e-01 9.45471883e-01 -5.25704026e-01 5.17854035e-01 1.22740400e+00 1.06171656e+00 9.71216917e-01 -1.68140209e+00 -8.39959443e-01 -1.37156865e-04 -6.50644749e-02 -1.09682429e+00 -6.72996700e-01 7.60270178e-01 -4.53316450e-01 8.35573137e-01 7.07036316e-01 3.29558045e-01 1.08008695e+00 1.91065535e-01 6.68622851e-01 9.16477978e-01 -4.60485548e-01 2.80077368e-01 2.36506194e-01 6.47496760e-01 9.73772407e-01 7.56428540e-01 3.63863975e-01 -1.64858073e-01 -2.08301768e-01 6.26665056e-01 -1.57461822e-01 -4.66491818e-01 -6.03964686e-01 -1.04680824e+00 7.57780612e-01 1.64133400e-01 1.36108279e-01 2.29462996e-01 5.68281651e-01 6.53422296e-01 6.30456448e-01 -1.12794951e-01 6.52707577e-01 -7.04656303e-01 -1.61680073e-01 -5.47788918e-01 5.84363759e-01 1.02075839e+00 9.75864410e-01 5.65232038e-01 4.42451149e-01 2.68975079e-01 3.46597619e-02 -4.26012278e-03 5.60966134e-01 2.94387549e-01 -1.21436334e+00 2.47296289e-01 2.90782273e-01 -1.89345613e-01 -1.06039667e+00 -3.67463380e-01 -4.48643953e-01 -8.47933829e-01 6.82733834e-01 6.26327336e-01 -2.47725859e-01 -7.42817283e-01 2.31188631e+00 2.28931084e-02 2.01348484e-01 4.32309151e-01 3.87097925e-01 2.77294606e-01 4.06903684e-01 -5.14514506e-01 -4.31193076e-02 1.07272553e+00 -4.97221261e-01 -3.83357018e-01 -7.83721283e-02 9.22806680e-01 6.15734570e-02 1.04152524e+00 7.35896587e-01 -1.16981030e+00 -3.38880479e-01 -1.41238725e+00 5.02133667e-02 -6.13363445e-01 3.85044180e-02 8.26411843e-01 1.23694098e+00 -1.08588886e+00 8.87062073e-01 -9.59235311e-01 -1.00290634e-01 4.19679433e-01 7.29926586e-01 -6.12880588e-01 2.59801805e-01 -1.05277574e+00 8.25724483e-01 3.89793843e-01 1.73978046e-01 -1.11358285e+00 -9.29249346e-01 -1.07657099e+00 1.27636865e-01 3.22140753e-01 -5.30983806e-01 1.35718489e+00 -7.49010384e-01 -1.32568574e+00 4.91626173e-01 -1.33152246e-01 -1.02677941e+00 3.35391313e-01 1.64058343e-01 -4.66463447e-01 -6.72223978e-03 -4.31734383e-01 4.30853307e-01 8.61317575e-01 -1.12538052e+00 -2.68281460e-01 -3.38717669e-01 2.99765021e-01 -6.96309328e-01 -3.39900345e-01 -1.06591331e-02 -2.34447958e-04 -3.55221182e-01 -7.37662837e-02 -1.15818906e+00 -2.16157079e-01 3.39186430e-01 -7.91998684e-01 1.91335648e-01 9.44223166e-01 -3.24029267e-01 8.18456352e-01 -2.04600883e+00 1.10906094e-01 6.83771849e-01 5.58599114e-01 6.78318322e-01 -6.44153878e-02 8.74256790e-02 -4.64620233e-01 6.04209304e-01 -2.42089048e-01 -3.55458021e-01 4.40765947e-01 4.08460200e-01 -8.36957753e-01 5.95674455e-01 5.13366461e-01 8.60585392e-01 -5.39747238e-01 6.06715046e-02 1.30116686e-01 3.19376677e-01 -5.91872752e-01 -6.11080863e-02 -1.86568782e-01 -9.41028595e-02 -9.89410356e-02 3.34433258e-01 6.48483098e-01 1.24746487e-01 -4.52472270e-02 -1.61195341e-02 2.42696628e-01 3.07036877e-01 -1.23264527e+00 1.16023493e+00 -3.79214138e-01 1.19602501e+00 1.77408069e-01 -9.72970963e-01 7.49773741e-01 1.77123040e-01 -2.49803841e-01 5.24790809e-02 4.84981388e-01 9.00297426e-03 1.50593400e-01 -2.04097122e-01 1.98761567e-01 -1.32117823e-01 -1.00289784e-01 6.39248431e-01 2.61258613e-02 -1.01505942e-03 2.39407733e-01 1.13376386e-01 1.21959758e+00 -2.96675384e-01 1.11046918e-01 -4.90582660e-02 5.26302457e-01 -6.87814504e-02 5.40737689e-01 1.04137254e+00 -7.26358518e-02 3.32338452e-01 1.05268538e+00 -2.02112436e-01 -1.20339835e+00 -1.02607715e+00 -1.09524906e-01 5.74366093e-01 -2.56142467e-01 -5.15835285e-01 -1.00973105e+00 -7.22032964e-01 1.98704109e-01 7.29595721e-01 -9.19154823e-01 -5.37549019e-01 -1.97842598e-01 -2.85722911e-01 1.42359364e+00 7.54572213e-01 3.07967305e-01 -5.95302463e-01 -4.93176818e-01 -3.51598740e-01 6.64266586e-01 -1.07776713e+00 -8.90226476e-03 5.65853059e-01 -9.57040966e-01 -1.12506473e+00 -2.43107349e-01 -6.35445356e-01 8.96625280e-01 3.76549214e-02 6.43834174e-01 2.69100010e-01 -1.33942738e-01 2.74427801e-01 3.26539725e-01 -4.16066855e-01 -9.91788566e-01 -1.24148697e-01 5.57032824e-01 -3.69023472e-01 7.35413879e-02 -7.76306391e-01 3.04634601e-01 2.33596638e-01 -1.00604832e+00 -4.75360826e-02 2.83887416e-01 8.79691541e-01 1.96658432e-01 4.17979658e-01 1.51742041e-01 -1.19952905e+00 8.49883318e-01 -1.09041199e-01 -1.16657043e+00 2.95142382e-01 -9.49284554e-01 5.04931450e-01 1.06603503e+00 -7.24930823e-01 -2.11700752e-01 6.42673448e-02 -9.59362835e-02 -7.21353114e-01 -3.03818852e-01 5.00048399e-01 -3.29598963e-01 -5.60703576e-01 8.04509819e-01 1.55906022e-01 4.29864079e-01 -2.14754224e-01 4.64949965e-01 2.65388340e-01 8.78648162e-01 -5.83051085e-01 1.49589097e+00 1.55860901e-01 5.98409534e-01 -3.85566592e-01 -4.16410118e-01 2.08329663e-01 -2.91393220e-01 2.36803532e-01 2.60569602e-01 -5.37147701e-01 -1.37016833e+00 4.84824121e-01 -1.10589969e+00 -3.30499828e-01 -3.72801274e-01 2.08611414e-01 -5.06102800e-01 3.15161288e-01 -6.73807919e-01 -8.30724657e-01 -1.35109395e-01 -1.27521336e+00 5.53243339e-01 -2.92073749e-02 -2.17264876e-01 -1.01695037e+00 -9.67943445e-02 -2.22039461e-01 2.63265610e-01 2.00529397e-01 1.07742727e+00 -1.15337384e+00 -3.85971606e-01 -5.76806366e-01 -9.54965800e-02 7.83164978e-01 -2.87891626e-01 4.95488912e-01 -1.24611092e+00 -2.22449854e-01 7.07343817e-02 -1.45768374e-01 6.80308938e-01 4.32355516e-03 1.29020524e+00 -6.80119157e-01 -2.11260602e-01 7.26110876e-01 1.01452053e+00 -1.15340516e-01 5.40278316e-01 1.99490443e-01 6.87303782e-01 3.63331318e-01 5.69644533e-02 -2.73491768e-03 -1.41667813e-01 4.12410766e-01 6.24156117e-01 1.60717204e-01 2.90321946e-01 -3.63024980e-01 5.75722992e-01 3.22937638e-01 1.97414011e-01 3.19982246e-02 -7.23006487e-01 2.80847132e-01 -1.65441036e+00 -1.02143192e+00 -2.27311477e-01 2.24358964e+00 8.14411819e-01 5.25394678e-01 9.66919884e-02 8.34484339e-01 5.62498093e-01 -2.46264085e-01 -5.70100546e-01 -8.02569926e-01 -1.75414458e-01 2.99319774e-01 8.62633467e-01 6.13056540e-01 -8.56034517e-01 5.59308231e-01 7.23755550e+00 5.68021238e-01 -1.05252314e+00 -1.25895634e-01 2.21816391e-01 2.84321368e-01 -4.92817700e-01 4.23777811e-02 -6.53399825e-01 1.97232943e-02 1.15624070e+00 -2.89716929e-01 7.08864987e-01 1.19766176e+00 -1.63941219e-01 5.25399506e-01 -1.66469073e+00 7.90359914e-01 -4.30925563e-02 -1.46232009e+00 -1.61809534e-01 1.07053228e-01 4.27697003e-01 -4.05304253e-01 3.60711753e-01 2.01200545e-01 4.34649915e-01 -1.18169284e+00 7.71041453e-01 1.34430304e-01 6.12265766e-01 -8.12452555e-01 6.74860001e-01 2.72943288e-01 -6.61874354e-01 -2.37248287e-01 -4.66702670e-01 -2.05014110e-01 -4.35521066e-01 4.10279304e-01 -1.12132919e+00 2.16212630e-01 1.25870153e-01 5.24084032e-01 -6.72525406e-01 8.93350899e-01 -5.86074889e-01 7.40147114e-01 -4.51697111e-01 -1.39574870e-01 -8.53787959e-02 2.05881208e-01 6.52391315e-01 1.11857653e+00 2.79644549e-01 -3.87722701e-01 -5.76735675e-01 1.34066594e+00 -1.58954009e-01 -6.45638824e-01 -9.60000038e-01 -2.07156450e-01 5.20778656e-01 1.12614739e+00 -3.40833396e-01 -2.86726445e-01 1.28477141e-01 6.33144140e-01 1.99986309e-01 3.56343329e-01 -9.29174960e-01 -7.35110044e-01 9.99378502e-01 -4.52006578e-01 4.08418089e-01 -3.19190383e-01 -4.46351111e-01 -1.17775762e+00 9.39773023e-02 -1.17061460e+00 -6.16586506e-02 -8.76139402e-01 -9.04055536e-01 4.16359246e-01 -2.59956755e-02 -1.10522676e+00 -5.05476117e-01 -1.16407454e+00 -7.07336783e-01 7.37026274e-01 -1.10386026e+00 -8.67940903e-01 4.72890511e-02 7.05171824e-01 -1.70831501e-01 -1.91924289e-01 1.03014302e+00 -1.73091199e-02 -9.94939208e-01 1.16337216e+00 -1.46382704e-01 2.81998456e-01 1.68434784e-01 -1.35934293e+00 7.72363186e-01 1.44060910e+00 3.50855559e-01 1.19909763e+00 1.12889874e+00 -2.83618450e-01 -2.02807283e+00 -1.06155801e+00 6.39605522e-01 -3.81121963e-01 1.01995552e+00 -7.26664722e-01 -7.64141619e-01 1.00490069e+00 -4.06715602e-01 1.76260963e-01 3.27098191e-01 2.94544250e-01 -9.10696149e-01 -2.67735183e-01 -1.21714151e+00 9.86667871e-01 9.58150208e-01 -9.16171253e-01 -5.65319121e-01 1.53671488e-01 9.65970278e-01 -4.66188222e-01 -7.17927277e-01 1.29004449e-01 6.60093129e-01 -7.58797407e-01 9.41015899e-01 -1.08275521e+00 3.01925331e-01 -5.28411269e-01 -1.63865417e-01 -1.11103308e+00 1.24616414e-01 -1.14731193e+00 -3.90828639e-01 1.32130790e+00 7.43049085e-01 -9.63391602e-01 7.94681013e-01 1.00850356e+00 -1.19296707e-01 -3.76337290e-01 -9.09391105e-01 -1.06905746e+00 1.64462611e-01 -1.05817115e+00 7.98682511e-01 8.49049330e-01 2.88761020e-01 -1.01217613e-01 -3.28578055e-01 6.08188629e-01 4.31523591e-01 -2.43491575e-01 1.06067896e+00 -1.21653831e+00 -4.90259320e-01 -7.03927100e-01 -9.57315683e-01 -7.68846154e-01 4.83627379e-01 -8.51183534e-01 1.07643925e-01 -6.02373898e-01 -7.49321133e-02 -4.36705440e-01 -1.32872313e-01 8.81679237e-01 2.73870438e-01 1.34425551e-01 1.82133168e-01 -1.63827166e-01 -4.86344881e-02 9.30075720e-03 5.94431102e-01 -1.95547268e-01 6.37742877e-02 3.08305383e-01 -8.57785046e-01 7.38917410e-01 8.77314568e-01 -5.91679931e-01 -5.59453666e-01 -1.05500080e-01 4.27451193e-01 -1.17588721e-01 8.45411956e-01 -1.21564794e+00 3.40117574e-01 7.42921755e-02 1.67840421e-02 1.64765911e-03 2.64902681e-01 -1.07682395e+00 3.26142788e-01 7.38896608e-01 -7.65633821e-01 1.42283455e-01 5.90994596e-01 4.37885344e-01 1.33871943e-01 -5.45990348e-01 2.78450161e-01 4.80368555e-01 -2.48628572e-01 1.79201633e-01 -4.34446841e-01 -2.14171082e-01 8.80906343e-01 -1.51812524e-01 -7.80819297e-01 -4.41136628e-01 -6.64246023e-01 -3.75509262e-02 7.07733214e-01 4.77874279e-02 6.48456037e-01 -1.12414432e+00 -1.73884109e-01 6.32745802e-01 -5.83644956e-02 -2.22037256e-01 -1.62009805e-01 8.54932308e-01 -3.80029589e-01 3.94566566e-01 -3.08584988e-01 -4.55444276e-01 -1.66575909e+00 8.95145893e-01 5.23283184e-01 7.69729018e-02 -4.17454183e-01 7.14099765e-01 -4.04148772e-02 -5.03371537e-01 5.58072627e-01 -1.01114035e+00 3.33697528e-01 -5.56560457e-01 6.79110885e-01 8.69507045e-02 1.39702946e-01 -3.79348807e-02 -3.09082478e-01 4.07001287e-01 7.37751275e-02 -1.47872984e-01 1.16641760e+00 2.68227577e-01 -7.02969059e-02 4.42266971e-01 1.23079789e+00 4.66671437e-02 -1.06133246e+00 9.64135900e-02 -2.11116374e-01 -1.52928114e-01 -1.03869058e-01 -5.62653303e-01 -1.18851113e+00 1.25540960e+00 3.94548893e-01 6.30512178e-01 1.18514359e+00 -2.18741208e-01 4.86425877e-01 1.11628985e+00 1.22812569e-01 -3.41800272e-01 -6.32088244e-01 4.31708783e-01 5.88267803e-01 -6.34350538e-01 -1.70521036e-01 -3.92429322e-01 -2.96429127e-01 1.44516718e+00 2.16633394e-01 -1.63610950e-01 4.06439185e-01 6.91154540e-01 -2.92227536e-01 1.28574848e-01 -1.04055583e+00 2.28756309e-01 1.22245148e-01 8.53631437e-01 -2.20904619e-01 -1.27014995e-01 4.05950457e-01 7.09922373e-01 -5.59529364e-01 2.74227411e-02 1.04075623e+00 7.31411397e-01 -2.01455176e-01 -1.07607591e+00 -3.15264553e-01 3.16236436e-01 -4.11996663e-01 -3.46639566e-02 -4.51905340e-01 1.13348055e+00 -7.41938725e-02 6.87951744e-01 -2.42574707e-01 -9.78261411e-01 1.45774424e-01 1.24848343e-01 4.68237698e-01 -2.34971479e-01 -4.28020626e-01 -6.96123898e-01 4.20148432e-01 -5.43277085e-01 -1.45084545e-01 -6.78265154e-01 -1.22275925e+00 -8.55218947e-01 -5.15209973e-01 4.04392108e-02 9.19020057e-01 9.87834513e-01 2.99038380e-01 3.17172378e-01 4.92266685e-01 -5.63809335e-01 -8.97233844e-01 -4.86232102e-01 -4.95631576e-01 2.30971072e-02 8.28395963e-01 -3.56135845e-01 -7.37625301e-01 -6.68374673e-02]
[6.1553730964660645, 7.565533638000488]
f820ef54-d11a-4366-aa68-0cea1d037c5f
localizing-objects-with-self-supervised
2109.14279
null
https://arxiv.org/abs/2109.14279v1
https://arxiv.org/pdf/2109.14279v1.pdf
Localizing Objects with Self-Supervised Transformers and no Labels
Localizing objects in image collections without supervision can help to avoid expensive annotation campaigns. We propose a simple approach to this problem, that leverages the activation features of a vision transformer pre-trained in a self-supervised manner. Our method, LOST, does not require any external object proposal nor any exploration of the image collection; it operates on a single image. Yet, we outperform state-of-the-art object discovery methods by up to 8 CorLoc points on PASCAL VOC 2012. We also show that training a class-agnostic detector on the discovered objects boosts results by another 7 points. Moreover, we show promising results on the unsupervised object discovery task. The code to reproduce our results can be found at https://github.com/valeoai/LOST.
['Jean Ponce', 'Renaud Marlet', 'Patrick Pérez', 'Andrei Bursuc', 'Spyros Gidaris', 'Simon Roburin', 'Huy V. Vo', 'Gilles Puy', 'Oriane Siméoni']
2021-09-29
null
null
null
null
['single-object-discovery']
['computer-vision']
[-9.82093662e-02 2.88551241e-01 -1.86540201e-01 -4.77307200e-01 -8.85345638e-01 -7.89121211e-01 8.68014336e-01 1.17928624e-01 -6.32057071e-01 4.23882365e-01 -1.58600882e-01 1.74400121e-01 6.40926836e-03 -4.21242923e-01 -1.15043116e+00 -5.46464145e-01 1.12519510e-01 7.51996756e-01 7.85970807e-01 8.89319479e-02 2.92514294e-01 4.93182212e-01 -1.68981242e+00 3.13382506e-01 3.07804227e-01 1.04763818e+00 4.54830736e-01 5.53179860e-01 2.33159482e-01 5.28500557e-01 -4.55001086e-01 -3.28033060e-01 4.46104884e-01 1.86215565e-02 -9.89984572e-01 2.90778786e-01 8.25656056e-01 -2.68636346e-01 -7.09864572e-02 1.19227791e+00 2.47937530e-01 -6.91365600e-02 4.03021604e-01 -1.15603173e+00 -4.43086296e-01 4.24721330e-01 -4.01577175e-01 3.35573941e-01 -1.09592546e-02 4.45268691e-01 1.06418145e+00 -1.34758067e+00 6.99545264e-01 8.54642928e-01 5.23720741e-01 4.40387160e-01 -1.39561784e+00 -5.06711662e-01 1.43826842e-01 1.83905512e-01 -1.50999725e+00 -7.22288668e-01 6.69831276e-01 -4.15461004e-01 7.82339513e-01 6.13403730e-02 7.06744194e-01 9.26392794e-01 -5.01636624e-01 1.03981972e+00 1.02964449e+00 -4.74686593e-01 2.54356891e-01 4.76316392e-01 1.06617510e-01 1.02231228e+00 3.68083119e-01 1.09250396e-01 -6.10009670e-01 -2.49283612e-02 5.49108505e-01 -4.78299707e-02 -9.14427862e-02 -8.51599157e-01 -1.18936849e+00 5.99741161e-01 6.85464442e-01 2.77644008e-01 -3.61256629e-01 3.75086427e-01 3.14375848e-01 -1.63251579e-01 3.17032546e-01 7.39028454e-01 -5.11981964e-01 -1.57828012e-03 -7.68608451e-01 5.12980185e-02 5.23361087e-01 1.12575269e+00 1.07929265e+00 -3.50682855e-01 -1.66628398e-02 4.96948004e-01 2.38167971e-01 3.01044136e-01 1.45282164e-01 -1.14833307e+00 -5.94031513e-02 5.84770679e-01 2.59783387e-01 -4.08446670e-01 -2.09569216e-01 -5.21745145e-01 3.15737445e-03 3.27797413e-01 4.40327257e-01 1.49722770e-01 -9.57313061e-01 1.33175123e+00 5.35487056e-01 -1.80578269e-02 -3.06676596e-01 9.74438012e-01 6.67335808e-01 1.09785303e-01 -1.17538407e-01 3.69810522e-01 1.42469132e+00 -1.33565760e+00 -2.49554187e-01 -2.69202441e-01 4.71824557e-01 -6.58833206e-01 1.03862739e+00 4.49441493e-01 -1.01980615e+00 -5.03549576e-01 -9.04127538e-01 2.68935715e-03 -3.92653823e-01 6.07678413e-01 8.28845322e-01 3.33857596e-01 -1.05103624e+00 2.83305109e-01 -9.36502039e-01 -7.41346002e-01 1.03778601e+00 3.82244945e-01 -4.49188232e-01 4.97679077e-02 -2.90745944e-01 7.04609096e-01 5.42765677e-01 -8.04347843e-02 -1.57388830e+00 -6.66100085e-01 -4.68674511e-01 -1.13432534e-01 8.32229793e-01 -4.85883027e-01 1.45627725e+00 -1.03694856e+00 -1.23922324e+00 1.29829085e+00 -1.49270281e-01 -5.75796545e-01 4.40797210e-01 -6.23584569e-01 1.66229054e-01 3.18594813e-01 3.67893159e-01 1.16571236e+00 8.04600775e-01 -1.44584537e+00 -8.63006890e-01 -2.79711604e-01 2.14010686e-01 -4.72747386e-02 -2.57036507e-01 1.53701916e-01 -8.31019700e-01 -2.75821328e-01 -1.45520672e-01 -9.73119795e-01 -2.06010401e-01 3.32392961e-01 -4.56741840e-01 -4.66568828e-01 7.40681112e-01 -9.83929783e-02 2.92183906e-01 -2.18597841e+00 -2.67180383e-01 -5.21712899e-02 1.30774796e-01 3.86685848e-01 -1.81231409e-01 9.06560123e-02 9.85138118e-02 -1.32769430e-02 -1.17532834e-01 -5.57196975e-01 -1.28456116e-01 7.87701085e-02 -3.66971046e-01 7.24959016e-01 4.56476867e-01 9.97105479e-01 -1.06260741e+00 -4.37830478e-01 3.27155054e-01 2.09029481e-01 -4.83680636e-01 1.61421150e-01 -5.32914996e-01 4.81707305e-01 -3.67903173e-01 9.03865159e-01 4.65777129e-01 -5.47130704e-01 3.98104824e-02 -2.81030178e-01 -2.11339861e-01 4.78968889e-01 -8.64593029e-01 1.95632637e+00 -8.62043910e-03 6.91953599e-01 -4.20577675e-02 -1.13530874e+00 7.57292330e-01 3.74926254e-02 4.32743341e-01 -3.81000191e-01 9.51786861e-02 1.84474766e-01 -1.56586900e-01 -3.70374024e-01 2.03007445e-01 3.79068226e-01 2.15907544e-01 3.58108014e-01 6.26196027e-01 -1.20194644e-01 2.96116054e-01 2.89840728e-01 1.34883583e+00 3.16824228e-01 1.82412744e-01 -3.35768968e-01 2.24262327e-01 3.31947476e-01 4.70799804e-01 1.00683463e+00 -2.79277146e-01 6.02615237e-01 1.56918705e-01 -5.38859785e-01 -1.15326691e+00 -1.11096227e+00 -3.32775265e-01 1.05929804e+00 2.23756716e-01 -5.86209297e-01 -5.86829782e-01 -1.08434999e+00 -1.79766007e-02 4.35957283e-01 -7.02843845e-01 1.32457331e-01 -3.05108845e-01 -3.04561615e-01 6.59135580e-01 6.36247516e-01 4.52717096e-01 -1.06709659e+00 -7.38782644e-01 -1.71246871e-01 4.47701523e-03 -1.34235013e+00 -1.60840914e-01 4.56632465e-01 -7.09850907e-01 -1.23739386e+00 -3.97294670e-01 -7.50993013e-01 1.07840872e+00 4.31601733e-01 1.14147449e+00 1.58572435e-01 -6.04523897e-01 7.97682226e-01 -2.38473609e-01 -7.36565113e-01 -2.58745216e-02 2.22678721e-01 6.01426437e-02 -6.00575358e-02 7.42935479e-01 -3.17910790e-01 -5.48082888e-01 3.92215073e-01 -5.15424788e-01 -1.03974484e-01 6.68780804e-01 6.39171124e-01 8.05202782e-01 -2.64061242e-01 2.63523012e-01 -6.57747388e-01 -1.33959860e-01 -2.68629849e-01 -1.06297958e+00 7.62702897e-02 -5.49598873e-01 8.30803514e-02 2.23492220e-01 -2.60945112e-01 -7.52503395e-01 7.64276028e-01 2.61878312e-01 -6.78421199e-01 -5.33792257e-01 -1.11939721e-01 1.13594487e-01 -3.13550621e-01 8.19164634e-01 1.70119986e-01 -1.52283460e-01 -6.04411900e-01 4.89212960e-01 3.50814432e-01 7.50915647e-01 -5.56886971e-01 1.05880141e+00 9.72087562e-01 -2.63523608e-01 -5.47409892e-01 -1.11006534e+00 -9.75728393e-01 -8.63725901e-01 -2.09279597e-01 6.94856584e-01 -1.10997248e+00 -7.91140676e-01 1.10662997e-01 -1.21578062e+00 -4.93084639e-01 -6.33878469e-01 4.43005174e-01 -6.22663915e-01 4.92272712e-02 -1.71099797e-01 -7.56580472e-01 -2.87108183e-01 -7.74702489e-01 1.19983566e+00 2.25214794e-01 -1.49248332e-01 -5.14417350e-01 -9.19276942e-03 5.18788099e-01 1.97569475e-01 -2.08020836e-01 1.22801676e-01 -8.63653719e-01 -1.23200631e+00 -2.86066264e-01 -2.96842664e-01 4.03376877e-01 1.78441107e-02 -1.45118758e-01 -1.28943670e+00 -2.37806872e-01 -1.39762655e-01 -7.35446393e-01 1.10958469e+00 7.47926086e-02 1.32494009e+00 -2.64663935e-01 -5.97151101e-01 4.88573670e-01 1.30461943e+00 -2.89838940e-01 3.84813905e-01 5.68899274e-01 6.17620111e-01 6.22602284e-01 8.46423090e-01 3.97730857e-01 2.95316786e-01 7.08418608e-01 8.60090017e-01 -2.45240375e-01 -2.91265905e-01 -1.91702247e-01 1.61434218e-01 -4.73458022e-02 -3.41819286e-01 1.76282644e-01 -9.29964602e-01 1.09658587e+00 -1.91169906e+00 -8.58400881e-01 -4.50039096e-02 2.20852280e+00 8.41061294e-01 1.53878093e-01 2.05383629e-01 -2.44623259e-01 5.60347557e-01 -1.07666992e-01 -4.58555788e-01 2.40802422e-01 -2.06011683e-02 2.99861431e-01 7.71411896e-01 3.32699716e-01 -1.25258160e+00 1.35150862e+00 6.29819393e+00 4.45209444e-01 -9.51813698e-01 4.25934017e-01 3.76571715e-01 -3.63973349e-01 8.07801560e-02 4.51348364e-01 -1.09879112e+00 2.07165018e-01 4.47397828e-01 1.19544357e-01 3.84815514e-01 1.18518519e+00 -8.15129131e-02 -4.00734127e-01 -1.36551344e+00 8.42620313e-01 2.36908987e-01 -1.55052149e+00 -2.43867740e-01 1.36706889e-01 6.25861883e-01 6.73214257e-01 -6.37006611e-02 5.54324761e-02 4.78212535e-01 -7.51721084e-01 8.98983359e-01 3.99703830e-01 3.77670527e-01 -1.88946977e-01 5.25031924e-01 3.08148652e-01 -9.09513533e-01 -3.34134549e-02 -4.50234115e-01 7.79733136e-02 -1.43428221e-01 5.92244267e-01 -1.35435951e+00 2.09469289e-01 1.19955218e+00 6.46902800e-01 -1.13260102e+00 1.46530616e+00 -6.57569289e-01 7.84817278e-01 -5.66429973e-01 1.45870492e-01 1.89709708e-01 2.54926324e-01 7.54091442e-01 1.14882112e+00 2.80571543e-02 -2.00402081e-01 2.96846300e-01 1.14137435e+00 -1.91200227e-01 -2.08942249e-01 -6.35485947e-01 9.25761908e-02 4.77126151e-01 1.51737750e+00 -7.18943954e-01 -3.52467269e-01 -1.68370113e-01 9.28079367e-01 6.20849073e-01 2.50173241e-01 -8.26980233e-01 -2.20018998e-01 2.28546232e-01 2.64210105e-01 9.62182939e-01 -1.38512626e-01 -9.55142602e-02 -1.07976425e+00 3.06952119e-01 -6.01945639e-01 1.99517384e-01 -8.13316524e-01 -1.03412139e+00 3.00155491e-01 -8.27561691e-02 -1.11488366e+00 1.69664063e-02 -6.62711561e-01 -6.16850436e-01 3.39555144e-01 -1.41741312e+00 -1.32689142e+00 -5.42673886e-01 5.57577014e-01 4.50642228e-01 -1.94549903e-01 7.67697334e-01 1.04406103e-01 -3.76939029e-01 4.09146100e-01 -2.13814065e-01 3.60857427e-01 8.03462625e-01 -1.24923325e+00 2.42939502e-01 9.50515747e-01 6.71262741e-01 5.66082001e-01 7.55463183e-01 -3.60058278e-01 -1.35701513e+00 -1.12454855e+00 5.83917379e-01 -9.36051548e-01 6.59588993e-01 -5.91216087e-01 -7.69272804e-01 1.02459717e+00 2.57032126e-01 6.99611068e-01 3.52962971e-01 3.75938058e-01 -5.68023682e-01 -1.84697270e-01 -9.06170547e-01 2.49596938e-01 1.17105591e+00 -3.87955099e-01 -4.73355889e-01 7.48134613e-01 6.10335529e-01 -2.34535843e-01 -6.41473174e-01 4.49455082e-01 2.69923359e-01 -8.13112497e-01 9.11275566e-01 -5.31746864e-01 1.45530298e-01 -6.98922813e-01 -1.38215885e-01 -8.47224832e-01 -3.17661554e-01 -4.72064883e-01 -1.63573682e-01 1.16665375e+00 4.56174344e-01 -5.68872929e-01 8.89995575e-01 1.87328413e-01 -2.40231961e-01 -6.29758596e-01 -8.60430479e-01 -1.06610394e+00 -5.24067998e-01 -3.86967093e-01 2.68122703e-01 7.14303732e-01 -2.27837607e-01 3.97276431e-01 1.24647720e-02 4.04302835e-01 9.36356604e-01 7.05825537e-02 1.23341346e+00 -1.09778249e+00 -4.97594506e-01 -2.74401397e-01 -5.94547570e-01 -9.02895510e-01 5.78729808e-02 -1.01369667e+00 2.89345056e-01 -1.35599625e+00 4.16175902e-01 -7.62639225e-01 -3.62193048e-01 1.07238042e+00 9.15246010e-02 5.40046215e-01 1.78704023e-01 6.12500966e-01 -1.30680847e+00 5.43894172e-01 6.23973191e-01 -2.49013111e-01 9.61987823e-02 -9.94823650e-02 -6.68862104e-01 9.49643672e-01 8.59357715e-01 -7.96247244e-01 -8.82100165e-02 -4.29741472e-01 2.72022653e-02 -7.99310803e-01 9.67946887e-01 -1.24484921e+00 3.95117372e-01 9.29923058e-02 3.98554593e-01 -6.09408975e-01 4.71557200e-01 -7.65277445e-01 -1.49910137e-01 4.16470379e-01 -2.56795645e-01 -4.52883661e-01 2.99887776e-01 5.75195014e-01 -2.09645927e-02 -4.79912639e-01 7.52229810e-01 -3.24381292e-01 -8.92175794e-01 2.27311477e-01 -8.31277370e-02 -1.80079609e-01 1.19415557e+00 3.77888512e-03 -6.47081792e-01 3.90857123e-02 -5.23671031e-01 2.34778062e-01 7.65920162e-01 3.57003480e-01 2.81049043e-01 -1.15238750e+00 -5.34825206e-01 -6.80875108e-02 5.88678479e-01 2.99331069e-01 -2.78052062e-01 9.57833052e-01 -1.93617880e-01 4.68712687e-01 -7.22575337e-02 -1.04572332e+00 -1.26651371e+00 7.02535212e-01 3.58815074e-01 6.96139708e-02 -5.63188791e-01 1.06732678e+00 3.49074095e-01 -4.55133379e-01 3.46607447e-01 9.71646607e-03 1.66553497e-01 -2.38222674e-01 4.92860943e-01 9.67622474e-02 1.30158216e-01 -2.60778934e-01 -6.33643806e-01 3.51399779e-01 -3.80643845e-01 -1.78307533e-01 1.41686499e+00 1.03991047e-01 -1.30148068e-01 3.44276428e-01 9.35235918e-01 2.41130814e-02 -1.53678322e+00 -5.12803972e-01 2.19357610e-01 -6.03278160e-01 1.31102160e-01 -7.90291488e-01 -9.07102823e-01 4.61569995e-01 6.82684481e-01 -1.40658673e-03 9.50849950e-01 6.99193835e-01 2.12367460e-01 8.51245999e-01 5.33830523e-01 -9.90829468e-01 5.15177131e-01 3.22097749e-01 9.14289355e-01 -1.63946831e+00 1.27532735e-01 -3.77674431e-01 -4.02951121e-01 7.80712306e-01 8.57625544e-01 -2.76394248e-01 4.85102028e-01 1.67366311e-01 -1.74242526e-01 -6.31077111e-01 -7.96221673e-01 -5.73314965e-01 1.54655337e-01 6.48230791e-01 1.59337208e-01 -1.43440202e-01 1.60580292e-01 2.27438852e-01 2.96396334e-02 1.63608000e-01 3.43280613e-01 1.01267374e+00 -5.77941895e-01 -9.35323596e-01 -3.70974690e-01 2.51626611e-01 -3.26703191e-01 -1.15433648e-01 -5.67984521e-01 6.20087206e-01 2.36676931e-01 5.91037869e-01 1.70976773e-01 -1.40330166e-01 1.34991661e-01 -6.27608374e-02 5.86384356e-01 -9.49794888e-01 -3.82886350e-01 2.07338296e-02 9.08450633e-02 -7.96086550e-01 -7.42058814e-01 -7.81220496e-01 -1.16159749e+00 2.08184898e-01 -6.23335540e-01 -2.18633115e-02 8.16620052e-01 7.64533937e-01 5.84753692e-01 2.74901778e-01 4.91816312e-01 -1.03415680e+00 -3.90591145e-01 -8.78021240e-01 -1.47906154e-01 2.65835583e-01 2.18577340e-01 -6.73283577e-01 -2.98595369e-01 2.57973254e-01]
[9.365816116333008, 1.1099289655685425]
a97f8585-20d4-4ecc-a0e9-95231d2e534d
ranking-sentences-for-extractive
1802.08636
null
http://arxiv.org/abs/1802.08636v2
http://arxiv.org/pdf/1802.08636v2.pdf
Ranking Sentences for Extractive Summarization with Reinforcement Learning
Single document summarization is the task of producing a shorter version of a document while preserving its principal information content. In this paper we conceptualize extractive summarization as a sentence ranking task and propose a novel training algorithm which globally optimizes the ROUGE evaluation metric through a reinforcement learning objective. We use our algorithm to train a neural summarization model on the CNN and DailyMail datasets and demonstrate experimentally that it outperforms state-of-the-art extractive and abstractive systems when evaluated automatically and by humans.
['Mirella Lapata', 'Shay B. Cohen', 'Shashi Narayan']
2018-02-23
ranking-sentences-for-extractive-1
https://aclanthology.org/N18-1158
https://aclanthology.org/N18-1158.pdf
naacl-2018-6
['extractive-document-summarization']
['natural-language-processing']
[ 3.51704657e-01 5.18900454e-01 -1.95456430e-01 -3.75956267e-01 -9.87829804e-01 -5.59708178e-01 6.17804766e-01 6.00552440e-01 -6.83010638e-01 8.67719531e-01 1.07513952e+00 -6.72860965e-02 2.33211983e-02 -5.34454703e-01 -8.22991073e-01 -1.75759763e-01 6.14913628e-02 5.59269905e-01 -2.89204001e-01 -3.37941021e-01 7.93140054e-01 2.56791353e-01 -1.18904030e+00 5.53157568e-01 1.21343207e+00 4.95167881e-01 7.92234112e-03 1.28143489e+00 8.73349458e-02 9.27397788e-01 -1.40943992e+00 -6.31871283e-01 3.25272745e-03 -7.78852582e-01 -1.16209197e+00 -7.50954226e-02 9.78099942e-01 -8.11424971e-01 -4.72682327e-01 9.81125593e-01 7.83382714e-01 4.10178632e-01 7.22327113e-01 -6.30270123e-01 -1.16441786e+00 1.03975844e+00 -3.31424057e-01 4.36963499e-01 4.42194253e-01 -1.38751969e-01 1.56479728e+00 -6.18120849e-01 8.53612006e-01 1.08333945e+00 4.39056873e-01 7.20797598e-01 -1.13635957e+00 -4.62840162e-02 -1.14037991e-01 5.02849631e-02 -6.54060423e-01 -4.72334921e-01 7.57035673e-01 1.64444774e-01 1.67593372e+00 4.89584088e-01 6.42244577e-01 9.32279050e-01 8.45642984e-01 1.45882058e+00 9.43145826e-02 -3.01613003e-01 2.44760349e-01 -5.56286871e-01 5.32121301e-01 7.66703904e-01 4.86981541e-01 -4.88865048e-01 -7.49224961e-01 -7.24421674e-03 -9.15152021e-03 -1.61390528e-01 -2.68664837e-01 1.96984798e-01 -1.19208848e+00 9.47954059e-01 4.35360074e-01 1.88795492e-01 -7.36250639e-01 4.92433041e-01 8.33561540e-01 4.32815582e-01 9.25969720e-01 1.17619944e+00 -2.33287483e-01 -1.51276186e-01 -1.44638872e+00 6.31322563e-01 9.87521470e-01 7.64629483e-01 4.74671900e-01 3.94658744e-01 -7.71919310e-01 9.21313941e-01 -2.15522692e-01 4.34452415e-01 8.72803628e-01 -1.12170684e+00 6.62948668e-01 4.89035904e-01 1.12230390e-01 -7.86876798e-01 -3.88823718e-01 -7.02352226e-01 -1.05804276e+00 -2.25111499e-01 -5.27050018e-01 -4.77658927e-01 -7.55410194e-01 1.32292318e+00 -3.60069811e-01 -1.95801482e-01 4.91461217e-01 5.24762750e-01 1.35049307e+00 9.43936884e-01 -2.64332235e-01 -4.46344733e-01 7.56152153e-01 -1.42406166e+00 -9.20376182e-01 -1.21734172e-01 5.44777870e-01 -4.59056348e-01 7.77507544e-01 4.07471806e-01 -1.69716966e+00 -4.79927599e-01 -1.50113344e+00 -3.80728722e-01 -1.31247446e-01 4.55039144e-01 2.87289113e-01 2.18921632e-01 -1.50327218e+00 1.17366195e+00 -7.41612017e-01 -3.97609383e-01 4.77220476e-01 3.38199139e-01 -3.60095233e-01 3.75682414e-01 -8.49565089e-01 9.73501861e-01 8.47762227e-01 -5.82983494e-02 -7.05188096e-01 -5.59534490e-01 -8.11448157e-01 6.03494108e-01 6.19556978e-02 -1.29236424e+00 1.95710635e+00 -8.37943316e-01 -1.79420865e+00 6.06831133e-01 -1.66461796e-01 -1.15812159e+00 3.20078403e-01 -8.04878950e-01 -1.90363780e-01 6.00751460e-01 1.93777084e-01 7.83090591e-01 7.93945611e-01 -9.04611468e-01 -6.93475008e-01 -1.92574248e-01 -5.91127910e-02 4.17500079e-01 -4.12079304e-01 -1.23231849e-02 7.01611117e-02 -8.03095162e-01 -4.62700486e-01 -4.82391715e-01 -1.49522036e-01 -9.51067686e-01 -9.86888945e-01 -4.46413517e-01 5.90826809e-01 -9.52819407e-01 1.34794497e+00 -1.50117922e+00 5.49029827e-01 -3.27024996e-01 4.91656959e-01 5.54224610e-01 -4.32592422e-01 8.31060350e-01 1.69155493e-01 2.49678552e-01 -3.37478846e-01 -7.65701234e-01 6.38587177e-02 -1.74618378e-01 -5.64903259e-01 1.04828611e-01 2.70883083e-01 1.35832202e+00 -1.15357447e+00 -4.61666316e-01 -2.33983353e-01 -5.50586358e-02 -6.84292138e-01 4.01033401e-01 -3.46074104e-01 -1.58149570e-01 -3.28788579e-01 1.80477247e-01 3.56862515e-01 3.96556221e-02 -8.40545632e-03 -9.13124532e-03 -1.14870500e-02 3.37976754e-01 -4.14326370e-01 1.95556855e+00 -1.73778668e-01 1.00168049e+00 -3.59846324e-01 -1.04120278e+00 8.48003387e-01 9.30116102e-02 1.35602713e-01 -6.91780388e-01 1.10071309e-01 1.68675527e-01 -4.11545783e-01 -3.71147960e-01 1.78540277e+00 1.99370414e-01 -2.37098664e-01 8.84903252e-01 4.71109837e-01 -4.40880775e-01 6.60590470e-01 7.91261733e-01 1.38909829e+00 -2.63155669e-01 4.65725541e-01 -3.87106687e-01 2.45148629e-01 -1.31594211e-01 2.28737488e-01 1.34932733e+00 1.38646662e-01 7.50245035e-01 6.39577031e-01 -3.83604795e-01 -1.19939971e+00 -1.10816109e+00 5.77384472e-01 1.25056636e+00 -1.85342476e-01 -7.98130155e-01 -1.11326361e+00 -1.05433571e+00 -6.43884838e-02 1.34928334e+00 -6.13449812e-01 -4.36797351e-01 -7.39565909e-01 -3.31868649e-01 8.21541250e-01 5.90108991e-01 6.35085762e-01 -1.49347365e+00 -9.46596265e-01 3.18817198e-01 -2.42966712e-01 -5.17766237e-01 -7.97103763e-01 8.89556706e-02 -9.85892653e-01 -6.37846291e-01 -7.17745423e-01 -8.16704392e-01 4.86919254e-01 2.12120309e-01 1.41293013e+00 2.78935224e-01 1.71387598e-01 3.66051823e-01 -4.83177483e-01 -6.97409391e-01 -8.28735232e-01 8.36679757e-01 -1.46992669e-01 -3.71217728e-01 3.70341949e-02 -4.37966347e-01 -6.60139740e-01 -7.32874274e-01 -1.25511289e+00 1.84587941e-01 6.80995345e-01 8.14223528e-01 3.52334350e-01 -4.06376481e-01 1.09373593e+00 -1.08138084e+00 1.80242562e+00 -3.01379785e-02 6.29350618e-02 4.81939167e-01 -5.46776652e-01 4.73268867e-01 8.85443568e-01 7.46117830e-02 -1.11580253e+00 -2.83400476e-01 -1.54937088e-01 4.82067838e-02 1.50050849e-01 8.23320270e-01 2.95420378e-01 6.61618829e-01 7.62887716e-01 6.10804856e-01 -2.06291854e-01 -4.04523075e-01 7.20625937e-01 7.55930781e-01 1.04636085e+00 -1.64979458e-01 4.44849610e-01 3.45359147e-01 -2.85456836e-01 -9.02576745e-01 -1.16150343e+00 -3.37166250e-01 -6.71733141e-01 -2.57694662e-01 7.47710586e-01 -4.82102513e-01 -5.17074645e-01 3.09543103e-01 -1.70705938e+00 -7.22249374e-02 -6.50472939e-01 1.08626127e-01 -8.62853169e-01 6.45706415e-01 -6.86576486e-01 -1.98463261e-01 -1.67398214e+00 -6.13437176e-01 1.35434699e+00 4.69711363e-01 -6.36650622e-01 -7.57347941e-01 5.58529258e-01 6.71485364e-02 3.77265930e-01 2.14962006e-01 7.11987734e-01 -1.01709366e+00 -7.69035816e-02 -4.17577147e-01 -7.99094960e-02 7.65001535e-01 -4.26262170e-02 1.50464550e-01 -5.36470950e-01 -3.96528184e-01 -1.90571576e-01 -4.25206929e-01 1.68766224e+00 6.15725458e-01 9.91668999e-01 -8.81014407e-01 8.77703130e-02 3.67652595e-01 1.11742091e+00 -1.30760655e-01 6.39906824e-01 2.11636543e-01 4.70904201e-01 4.27902967e-01 3.35259438e-01 5.06636381e-01 1.15993977e-01 -1.21339805e-01 3.28279644e-01 1.58451632e-01 -1.23563990e-01 -3.29758525e-01 3.67353588e-01 1.13318050e+00 1.29968613e-01 -7.83265352e-01 -4.91437107e-01 6.22502387e-01 -2.29653525e+00 -1.39301705e+00 1.85212433e-01 1.62211049e+00 8.74387383e-01 2.03528970e-01 -1.02343597e-01 -2.82773942e-01 6.70810997e-01 7.28658974e-01 -6.14161730e-01 -1.16054320e+00 -2.28626430e-01 2.38936722e-01 2.58538961e-01 4.62334424e-01 -1.00439334e+00 1.13130736e+00 7.16861963e+00 5.27188838e-01 -9.28714633e-01 -3.40319693e-01 5.21176934e-01 -3.76510918e-01 -3.70051384e-01 -4.81936306e-01 -4.13445264e-01 5.86004257e-02 1.06496930e+00 -7.26934433e-01 2.51520067e-01 6.95108533e-01 1.92436948e-01 -9.65922996e-02 -1.23702800e+00 6.33597374e-01 7.11840391e-01 -1.88572240e+00 6.27909005e-01 -3.58106077e-01 1.12527740e+00 3.12258065e-01 -6.96892664e-02 3.30674469e-01 5.69944263e-01 -9.47386682e-01 8.22649300e-01 8.47337008e-01 4.17256057e-01 -1.00156474e+00 7.67276525e-01 3.59489501e-01 -3.65963489e-01 6.88151866e-02 -4.87411529e-01 4.71591577e-02 1.33215249e-01 3.20390731e-01 -1.00143695e+00 8.31261635e-01 1.86034277e-01 1.06789482e+00 -8.23328972e-01 1.13746059e+00 -4.30742562e-01 5.51452041e-01 2.76621759e-01 -5.44099331e-01 4.23714042e-01 -1.24740321e-02 9.73719954e-01 1.75609446e+00 2.08521113e-01 -1.30091771e-01 -9.18446034e-02 7.15539098e-01 -8.57686639e-01 1.14652857e-01 -5.43841004e-01 -2.75027961e-01 7.72645995e-02 1.20318961e+00 -5.93480825e-01 -7.44692802e-01 1.30326107e-01 1.34830391e+00 5.66807568e-01 4.06036377e-01 -4.16817009e-01 -1.08103204e+00 1.71209708e-01 -5.30073702e-01 5.63350797e-01 -9.00088102e-02 -2.86961198e-01 -1.33352959e+00 2.01647412e-02 -8.27779174e-01 2.56677806e-01 -9.73228872e-01 -9.21801805e-01 8.33713830e-01 -1.24741830e-01 -8.78828049e-01 -7.43203521e-01 -9.72074941e-02 -1.10561204e+00 3.22799146e-01 -1.29328537e+00 -7.92832315e-01 -9.08285826e-02 -9.48079899e-02 1.04838324e+00 -4.85369802e-01 8.79924774e-01 -4.41004097e-01 -6.22778535e-01 4.42324042e-01 7.30624259e-01 1.07918277e-01 5.98083019e-01 -1.64200521e+00 8.52649987e-01 9.65528071e-01 1.91597328e-01 5.49111009e-01 1.19010592e+00 -6.65948868e-01 -1.09113634e+00 -1.26437497e+00 1.11584640e+00 -1.02082834e-01 3.32524896e-01 -5.52526638e-02 -5.98971903e-01 7.97842681e-01 1.25860834e+00 -8.56341600e-01 5.59732378e-01 -8.39926768e-03 -1.20184131e-01 -1.34144556e-02 -9.02251899e-01 7.26332605e-01 7.85308540e-01 -3.32553327e-01 -1.30718362e+00 4.86966282e-01 1.19777608e+00 -3.59871328e-01 -4.63666141e-01 9.03053656e-02 4.83178645e-01 -7.87134409e-01 5.89121699e-01 -1.11446297e+00 1.15163720e+00 2.42679536e-01 1.41653359e-01 -2.09214139e+00 -2.78727025e-01 -8.70323777e-01 -5.61606705e-01 1.03779161e+00 3.13599914e-01 -3.87394696e-01 7.17696786e-01 -2.39953324e-02 -6.29396141e-01 -6.36617661e-01 -7.63386667e-01 -6.54441595e-01 2.04777136e-01 1.84149057e-01 5.57929814e-01 2.99334317e-01 3.77908558e-01 9.12571251e-01 -2.48855263e-01 -5.06524503e-01 4.84422863e-01 2.80351520e-01 8.37346017e-01 -1.09154212e+00 4.41602198e-03 -8.79684210e-01 -9.45085809e-02 -1.10429740e+00 7.51838088e-01 -1.06304812e+00 3.35878313e-01 -2.23530793e+00 5.07147968e-01 1.06004214e+00 -2.21596062e-01 1.76012188e-01 -2.05574572e-01 -4.08592187e-02 2.34464809e-01 1.19980931e-01 -1.45586896e+00 1.03975236e+00 1.04519677e+00 -7.34979749e-01 -2.36846551e-01 1.14246327e-02 -1.15350795e+00 3.36827070e-01 9.74244773e-01 -5.27031600e-01 -2.41459459e-01 -8.32804441e-01 4.23466027e-01 1.39679715e-01 2.77000889e-02 -9.26254690e-01 4.45397735e-01 2.72423476e-01 2.98246354e-01 -1.17020309e+00 -6.88347891e-02 1.64723903e-01 -7.48592675e-01 5.85472643e-01 -1.16350257e+00 4.87889618e-01 1.33844689e-01 6.26449585e-01 -2.95526147e-01 -6.47714317e-01 3.79682243e-01 -1.07455455e-01 -2.81293243e-01 -8.82443041e-02 -5.77052057e-01 2.51999408e-01 4.03534085e-01 1.19146772e-01 -7.83671439e-01 -7.96207964e-01 -2.69033432e-01 3.59591216e-01 1.85878247e-01 4.27463740e-01 1.02448726e+00 -9.65133786e-01 -1.54895008e+00 -2.08994061e-01 -7.52238706e-02 -1.64218023e-01 8.25053006e-02 1.53097898e-01 -9.76476669e-01 8.27074051e-01 -3.24715793e-01 -8.11120793e-02 -1.35538280e+00 2.45160654e-01 2.19915792e-01 -6.50286615e-01 -7.24340558e-01 5.23289740e-01 -2.63427734e-01 -3.04743260e-01 1.53359085e-01 -4.96394455e-01 -4.05623376e-01 1.43053293e-01 7.24998474e-01 6.36497498e-01 3.08867127e-01 -3.43411773e-01 1.16837792e-01 -1.42922461e-01 -6.89982355e-01 -2.13688985e-01 1.65271974e+00 1.36060968e-01 -2.55378604e-01 4.96870093e-02 1.46278811e+00 -5.66383041e-02 -7.22027540e-01 -6.00567795e-02 1.17829040e-01 1.20221630e-01 1.20531935e-02 -8.29616845e-01 -6.16397142e-01 5.46129763e-01 -1.98082611e-01 5.04707277e-01 9.57763672e-01 -7.92471841e-02 1.15422451e+00 1.27979720e+00 -5.05829096e-01 -1.60142052e+00 4.99327600e-01 8.32027197e-01 1.47328019e+00 -8.11373949e-01 3.20632130e-01 3.75196248e-01 -7.98553586e-01 1.36706626e+00 4.12606120e-01 -8.04194152e-01 4.61981865e-03 -3.57189298e-01 -3.29144925e-01 -3.24721932e-01 -9.56496179e-01 1.06036954e-01 4.27715659e-01 2.94004023e-01 4.85952079e-01 1.03157043e-01 -6.91618383e-01 5.36856294e-01 -7.85056472e-01 -1.62042961e-01 1.03881025e+00 8.80507529e-01 -8.08604598e-01 -6.19994164e-01 2.25411117e-01 9.50030923e-01 -7.44193196e-01 -2.60770589e-01 -9.45860147e-01 2.91158736e-01 -7.91927993e-01 1.04198015e+00 2.14305162e-01 -3.43318254e-01 5.30635595e-01 -3.91979609e-03 4.18056697e-01 -8.71582150e-01 -1.13866878e+00 -3.47394824e-01 3.99484664e-01 -2.44832367e-01 -2.43879706e-01 -6.19850874e-01 -1.27886879e+00 -3.43118966e-01 -2.13568285e-01 4.53577936e-01 7.14537561e-01 9.40242767e-01 6.05926335e-01 9.06746149e-01 5.64003825e-01 -9.85488772e-01 -1.17003095e+00 -1.29072452e+00 -5.24514377e-01 3.66412014e-01 6.16851568e-01 3.71075571e-01 -6.96468353e-03 -4.98808697e-02]
[12.559900283813477, 9.525269508361816]
bc3d0382-8b81-4413-a38a-1debbbb55d47
mmface4d-a-large-scale-multi-modal-4d-face
2303.09797
null
https://arxiv.org/abs/2303.09797v1
https://arxiv.org/pdf/2303.09797v1.pdf
MMFace4D: A Large-Scale Multi-Modal 4D Face Dataset for Audio-Driven 3D Face Animation
Audio-Driven Face Animation is an eagerly anticipated technique for applications such as VR/AR, games, and movie making. With the rapid development of 3D engines, there is an increasing demand for driving 3D faces with audio. However, currently available 3D face animation datasets are either scale-limited or quality-unsatisfied, which hampers further developments of audio-driven 3D face animation. To address this challenge, we propose MMFace4D, a large-scale multi-modal 4D (3D sequence) face dataset consisting of 431 identities, 35,904 sequences, and 3.9 million frames. MMFace4D has three appealing characteristics: 1) highly diversified subjects and corpus, 2) synchronized audio and 3D mesh sequence with high-resolution face details, and 3) low storage cost with a new efficient compression algorithm on 3D mesh sequences. These characteristics enable the training of high-fidelity, expressive, and generalizable face animation models. Upon MMFace4D, we construct a challenging benchmark of audio-driven 3D face animation with a strong baseline, which enables non-autoregressive generation with fast inference speed and outperforms the state-of-the-art autoregressive method. The whole benchmark will be released.
['Jelo Wang', 'Xiangyuan Wang', 'Hongwei Xu', 'Junliang Xing', 'Jia Jia', 'Haozhe Wu']
2023-03-17
null
null
null
null
['3d-face-animation']
['computer-vision']
[-7.72983208e-02 -2.94422776e-01 1.34255186e-01 -2.85732776e-01 -7.50391781e-01 -2.57656366e-01 6.15546346e-01 -8.31844032e-01 2.41255328e-01 4.01483983e-01 4.26823288e-01 6.80363178e-02 2.88709164e-01 -5.42395651e-01 -6.50864959e-01 -6.90583169e-01 -4.57980424e-01 5.62035561e-01 -1.55429453e-01 -3.42354685e-01 -3.14245731e-01 7.91031718e-01 -2.11353016e+00 7.37043396e-02 3.85011792e-01 1.07976019e+00 -1.24836914e-01 8.28463316e-01 -1.15730651e-01 4.59042519e-01 -5.71329534e-01 -5.97319603e-01 1.92647949e-01 -5.88959336e-01 -3.29195231e-01 -9.00167692e-03 6.41661286e-01 -6.41915023e-01 -4.96639103e-01 6.32204294e-01 9.72668767e-01 1.49476722e-01 5.71336269e-01 -1.56806338e+00 -7.57966042e-01 1.78825259e-01 -7.65524805e-01 6.30534813e-02 8.14288259e-01 2.90724814e-01 5.02815783e-01 -1.11412227e+00 8.39179695e-01 1.92332947e+00 5.12948573e-01 1.02628839e+00 -1.09233904e+00 -1.19960439e+00 -1.77093044e-01 1.33784741e-01 -1.59697843e+00 -1.06669271e+00 9.75046694e-01 -3.98689985e-01 7.40891039e-01 2.42034703e-01 1.01476693e+00 1.64024854e+00 -1.67523026e-01 4.54231709e-01 8.76594722e-01 -7.46584460e-02 3.57268639e-02 -2.84923673e-01 -7.78888941e-01 7.19428599e-01 -4.00977135e-01 2.99940944e-01 -9.02657986e-01 -4.40488189e-01 1.06986165e+00 -4.95880187e-01 -1.59740478e-01 1.17565855e-01 -8.40174496e-01 8.07148516e-01 -1.74650669e-01 -6.01168126e-02 -5.29261708e-01 2.16220528e-01 6.95938408e-01 3.36403936e-01 8.25161576e-01 -3.23282510e-01 -2.32671723e-02 -6.41902924e-01 -9.37734067e-01 7.33759105e-01 5.23213565e-01 1.14149308e+00 2.80502468e-01 8.65008593e-01 1.60792857e-01 9.56937790e-01 4.29284573e-01 9.70220625e-01 1.84925884e-01 -1.34046280e+00 -8.71607736e-02 3.58293881e-03 -3.93917859e-01 -1.13840365e+00 -2.41770558e-02 2.90882677e-01 -1.18426108e+00 2.33153075e-01 5.10872267e-02 4.93896114e-05 -6.53519988e-01 1.94505107e+00 8.01893353e-01 8.38306487e-01 -2.54110366e-01 9.43779528e-01 1.30965364e+00 8.16811025e-01 3.68418507e-02 -5.08668065e-01 1.29072750e+00 -4.21344131e-01 -8.54826212e-01 4.32489336e-01 3.76954116e-02 -9.07944083e-01 1.11553216e+00 3.18229049e-01 -1.45366359e+00 -7.63549805e-01 -6.96125567e-01 -1.33136362e-01 2.91033715e-01 -3.63879055e-01 6.47879541e-01 8.50505352e-01 -1.30235684e+00 2.03312948e-01 -5.86444497e-01 4.67991270e-02 6.06214941e-01 3.40061247e-01 -6.98270559e-01 -5.94651736e-02 -1.15109122e+00 4.10200685e-01 -3.19605976e-01 -7.49046430e-02 -1.12846220e+00 -1.07334542e+00 -1.06576777e+00 -1.34914249e-01 4.95184101e-02 -6.17621183e-01 1.09375191e+00 -7.97779262e-01 -1.98089314e+00 1.13030064e+00 -3.52898121e-01 -7.24579319e-02 4.36968505e-01 -1.37167767e-01 -7.84776628e-01 3.61372054e-01 -1.06053717e-01 8.19337249e-01 1.43841636e+00 -1.15816271e+00 -1.11664146e-01 -4.19174343e-01 -2.96807557e-01 -5.12073785e-02 -2.95846194e-01 4.28058982e-01 -7.05785811e-01 -7.41451383e-01 -4.59746271e-01 -8.86374950e-01 2.29129955e-01 2.88772345e-01 -1.69532895e-01 -3.48702729e-01 1.19605088e+00 -5.50016642e-01 8.92406285e-01 -2.46769166e+00 2.88894475e-01 -1.35141119e-01 2.06811741e-01 2.81385660e-01 -4.09507990e-01 2.02642128e-01 -2.16298774e-01 1.56256318e-01 1.00091815e-01 -7.48657107e-01 -1.14096165e-01 1.14851154e-01 -4.12984341e-01 4.53868896e-01 3.63086313e-01 5.43529332e-01 -7.02907264e-01 -8.48684251e-01 2.96421915e-01 1.24723899e+00 -9.76467490e-01 3.40977073e-01 -9.21921954e-02 8.57887447e-01 -3.57085168e-01 1.02276933e+00 8.89693916e-01 8.06604847e-02 -3.59980851e-01 -8.31182748e-02 1.38605744e-01 -1.11500748e-01 -1.12279534e+00 1.87506759e+00 -4.97547358e-01 6.93530500e-01 3.79734546e-01 -4.15435404e-01 1.18556380e+00 6.12657309e-01 7.71921217e-01 -6.13833308e-01 2.10212424e-01 6.84149712e-02 -3.95314425e-01 -5.27276695e-01 3.62716734e-01 -1.75964653e-01 5.29619157e-02 2.48753041e-01 2.69434899e-01 -4.79384243e-01 -1.00143775e-01 1.90609649e-01 7.80014217e-01 2.94718534e-01 -1.64926097e-01 -6.93703517e-02 5.94690681e-01 -4.75999475e-01 6.72628403e-01 -8.14442560e-02 -1.48039237e-01 7.07154274e-01 3.85062933e-01 -4.35143828e-01 -1.05394590e+00 -9.16793346e-01 -1.41883567e-01 9.81206059e-01 -2.40790203e-01 -6.51830256e-01 -8.52226794e-01 -7.47470977e-03 1.91440880e-01 3.54074329e-01 -6.06869519e-01 -7.19214007e-02 -8.55462551e-01 -8.92205760e-02 8.65801811e-01 2.08007410e-01 3.73454362e-01 -1.18328857e+00 -1.04454137e-01 -2.67874673e-02 -1.35442629e-01 -1.24648952e+00 -8.15595150e-01 -8.93478632e-01 -6.56126320e-01 -8.57423306e-01 -7.75834620e-01 -7.90995777e-01 2.62150019e-01 3.21802169e-01 1.36907864e+00 1.47944927e-01 -2.41768464e-01 5.24595737e-01 -1.44017562e-01 -3.30208182e-01 -7.26383209e-01 -3.75215679e-01 6.90121174e-01 7.54871070e-02 7.72937238e-02 -9.81816590e-01 -4.63602066e-01 2.04737201e-01 -5.26674688e-01 -8.94145966e-02 4.71499041e-02 7.59578705e-01 8.13312650e-01 -3.28898817e-01 8.46928477e-01 -5.93235672e-01 4.59069133e-01 -4.90884870e-01 -5.53958654e-01 -2.30139673e-01 7.25650862e-02 -3.84239495e-01 7.70969927e-01 -8.07793021e-01 -1.14915657e+00 1.01405039e-01 -7.73653388e-01 -1.28634429e+00 -7.39617348e-02 6.63971808e-03 -3.52568448e-01 -7.83872753e-02 5.16684532e-01 2.27285326e-01 5.32391012e-01 -4.38522935e-01 4.61604178e-01 6.82011604e-01 9.83438313e-01 -6.66117966e-01 9.84851837e-01 3.48646343e-01 2.08366558e-01 -1.30476892e+00 -3.72212715e-02 1.84958577e-01 -2.19080433e-01 -6.14461720e-01 6.19101942e-01 -1.22492623e+00 -1.36505246e+00 6.78135991e-01 -1.22369087e+00 -2.62910843e-01 -2.08104432e-01 2.86243618e-01 -8.96486223e-01 2.26436242e-01 -7.34893262e-01 -9.07863140e-01 -6.02847099e-01 -1.28789568e+00 1.42259228e+00 8.73478875e-02 -2.53049314e-01 -4.85624313e-01 5.38866483e-02 2.31632635e-01 4.00166482e-01 6.06046915e-01 6.74630046e-01 -1.95394345e-02 -3.12219888e-01 -1.79931801e-02 6.35901242e-02 1.16210409e-01 1.36665538e-01 6.15880907e-01 -1.12637258e+00 -3.04284155e-01 -1.07506804e-01 -6.71979904e-01 1.66897431e-01 4.97912258e-01 1.34759796e+00 -2.59357423e-01 1.12104863e-01 1.00250328e+00 6.98352635e-01 3.10221091e-02 5.27093887e-01 -4.23374563e-01 6.60970271e-01 5.94991863e-01 6.25133455e-01 7.47255325e-01 3.91926855e-01 8.76800656e-01 5.72067678e-01 1.55816868e-01 -3.30861509e-01 -5.23120880e-01 4.29279178e-01 1.40815687e+00 -4.60817218e-01 6.00982793e-02 -4.28761929e-01 3.47650826e-01 -1.05397177e+00 -1.23233700e+00 6.65013939e-02 2.00535274e+00 9.44433331e-01 -2.56823897e-01 4.34759825e-01 6.46674260e-02 7.38580108e-01 3.08817804e-01 -6.70252740e-01 -4.28836554e-01 -1.25644535e-01 4.44456279e-01 -3.95444721e-01 3.58371139e-01 -8.16260993e-01 9.13455844e-01 6.23225546e+00 1.11064684e+00 -1.15355635e+00 6.57385588e-02 6.76739454e-01 -3.89673859e-01 -3.30895185e-01 -5.64490139e-01 -9.27661717e-01 7.25375772e-01 1.23694789e+00 -4.84692067e-01 6.04754686e-01 1.00664544e+00 4.06560272e-01 5.92881858e-01 -9.24474597e-01 1.52246189e+00 1.52674481e-01 -1.59652185e+00 3.83147120e-01 2.71177143e-01 4.79295373e-01 -3.03807795e-01 2.93722183e-01 2.97399670e-01 1.22793645e-01 -1.29961073e+00 6.49307132e-01 3.09276849e-01 1.76751804e+00 -1.13028407e+00 1.37168914e-01 7.39810541e-02 -1.57523012e+00 3.25120687e-01 -2.92563379e-01 3.61055642e-01 4.68286037e-01 1.22275785e-01 -2.08148032e-01 2.56002814e-01 8.94633472e-01 8.32877100e-01 8.25185999e-02 5.62211990e-01 1.67340487e-01 5.86828053e-01 -3.96817714e-01 2.69591421e-01 -1.65886730e-01 -1.67429999e-01 6.77527606e-01 9.97937083e-01 6.13963068e-01 5.11944950e-01 -1.61043555e-01 5.46848953e-01 -4.02640074e-01 7.96352923e-02 -9.95008111e-01 -4.51884232e-02 9.34413314e-01 1.12605166e+00 -1.79333776e-01 -1.49520099e-01 -4.13429618e-01 7.66367972e-01 -1.14120856e-01 -2.62754727e-02 -9.48655128e-01 1.08752616e-01 1.37090695e+00 5.65040857e-02 7.78118074e-02 -3.92827243e-01 3.27058941e-01 -8.66060555e-01 -2.73462981e-01 -1.33317435e+00 1.51774824e-01 -8.12813342e-01 -1.23226643e+00 1.07178295e+00 -1.51501700e-01 -1.35775983e+00 -7.27151155e-01 -1.48025319e-01 -3.89655292e-01 7.80005276e-01 -1.34062195e+00 -1.21433961e+00 -5.43477356e-01 1.10485065e+00 8.61329854e-01 -4.54170763e-01 1.18911064e+00 6.79407775e-01 -3.68562847e-01 8.96479070e-01 -3.57360452e-01 -1.49044394e-01 8.62470984e-01 -5.42759418e-01 8.56435478e-01 3.20477992e-01 7.59488121e-02 2.18287334e-01 6.09858811e-01 -4.36686605e-01 -2.19049573e+00 -1.01070297e+00 5.07343590e-01 -2.93906659e-01 3.90955806e-01 -6.60767436e-01 -1.12266982e+00 3.92757893e-01 -9.16664898e-02 2.69600838e-01 7.24243343e-01 -2.43563592e-01 -3.35359603e-01 -1.42863557e-01 -1.38122427e+00 6.08327985e-01 1.48065162e+00 -6.23969376e-01 1.45676201e-02 2.96734005e-01 9.66359317e-01 -7.13562489e-01 -1.29841101e+00 1.99121848e-01 7.69085944e-01 -9.34981883e-01 1.23697400e+00 -3.81721079e-01 5.53337276e-01 1.34584770e-01 -1.02172598e-01 -1.08637166e+00 -2.03439906e-01 -1.30230749e+00 -5.27906954e-01 1.53329396e+00 -2.98029870e-01 -2.27266774e-01 9.14973199e-01 3.79591584e-01 -1.41608819e-01 -6.20473325e-01 -1.29211402e+00 -6.15770817e-01 -6.97235838e-02 -5.32444775e-01 1.09243679e+00 8.57424736e-01 -4.67922837e-01 1.03763491e-01 -9.10090387e-01 4.52917302e-03 6.13465369e-01 -1.79650430e-02 1.20960546e+00 -1.31006229e+00 -1.07777014e-01 -3.64240587e-01 -5.79120815e-01 -1.19806755e+00 6.01599276e-01 -5.65824151e-01 -2.64723182e-01 -6.12627149e-01 -1.42003223e-01 -5.25361061e-01 5.38525641e-01 2.02487111e-01 9.24891829e-02 4.24383074e-01 1.56353965e-01 1.04136325e-01 -3.18127163e-02 1.11468482e+00 1.54607773e+00 1.23383857e-01 -1.73007756e-01 -1.11396112e-01 -2.89349407e-01 6.79660082e-01 3.30447674e-01 -1.91800237e-01 -5.00117302e-01 -2.63340354e-01 -2.29030415e-01 6.61737561e-01 2.29940593e-01 -6.91740632e-01 5.90215698e-02 -3.46392952e-02 3.65430087e-01 -4.79202837e-01 1.13755476e+00 -7.37662137e-01 5.51504254e-01 1.63930058e-01 7.06438627e-03 2.62158662e-01 3.78219783e-01 3.55719686e-01 -2.24884987e-01 4.46659774e-01 8.33464622e-01 1.51599616e-01 -5.77963412e-01 9.53590930e-01 -1.00148708e-01 2.74766684e-01 7.47617245e-01 -2.38132030e-01 -5.65432534e-02 -7.35464096e-01 -4.65644002e-01 -3.14501114e-02 5.00310361e-01 7.52759695e-01 1.03719008e+00 -1.79104078e+00 -1.27359247e+00 7.08154500e-01 -2.70271301e-01 1.68580666e-01 6.75650537e-01 4.08686996e-01 -3.78110111e-01 -1.13887787e-02 -4.90543634e-01 -7.84968436e-01 -1.75079417e+00 3.31366122e-01 1.62307604e-03 3.54367256e-01 -9.00197923e-01 8.62836361e-01 3.61161716e-02 -2.61666834e-01 1.88697159e-01 4.39272732e-01 -1.23237237e-01 1.04328617e-01 8.92770886e-01 3.25274318e-01 -2.05411419e-01 -1.34827399e+00 -2.87389517e-01 8.28377843e-01 3.30147028e-01 1.30185746e-02 1.51381421e+00 1.47005543e-01 6.60481974e-02 2.68560171e-01 1.23412693e+00 1.68150086e-02 -1.49963772e+00 -5.67376129e-02 -8.29842985e-01 -9.02821004e-01 -1.71591826e-02 4.97447997e-02 -1.49736321e+00 8.84723067e-01 3.88132244e-01 -2.69632693e-02 1.27661312e+00 -1.19346613e-03 1.14964271e+00 -2.74848819e-01 5.24206161e-01 -4.46407497e-01 2.19489455e-01 2.88397491e-01 1.22684538e+00 -9.83690798e-01 -1.28642559e-01 -6.36797130e-01 -6.89149082e-01 9.25482273e-01 5.57809532e-01 -1.90586522e-02 6.66746616e-01 5.21642208e-01 -1.19253732e-01 -8.82102251e-02 -1.00416648e+00 2.78885484e-01 8.95563811e-02 1.03329587e+00 3.27172160e-01 -2.84223229e-01 2.90536702e-01 4.75371659e-01 -5.78655601e-01 1.22764833e-01 1.85400262e-01 4.42942739e-01 1.18356809e-01 -8.55555356e-01 -3.98761362e-01 1.69829786e-01 -5.00255823e-01 1.40416026e-01 -9.25268307e-02 9.03460085e-01 -3.13743830e-01 8.47156525e-01 2.64956951e-01 -4.82996613e-01 3.21099639e-01 -6.10345751e-02 5.71782112e-01 -6.55986592e-02 -3.51799458e-01 2.48056188e-01 3.09957489e-02 -6.91478789e-01 -4.34340894e-01 -6.97062969e-01 -1.05919898e+00 -1.26829135e+00 -1.00925609e-01 9.54451133e-03 7.24145591e-01 3.64346325e-01 9.00774896e-01 1.67225450e-01 1.09260166e+00 -1.45006561e+00 -2.42591649e-01 -7.71912158e-01 -5.68770289e-01 4.33265656e-01 3.59247416e-01 -9.09880400e-01 -3.06059986e-01 2.76578575e-01]
[13.132710456848145, -0.3872746527194977]
22a88137-8a6c-4bdc-8626-1805ad794f2f
a-universal-parent-model-for-low-resource
1909.06516
null
https://arxiv.org/abs/1909.06516v2
https://arxiv.org/pdf/1909.06516v2.pdf
A Universal Parent Model for Low-Resource Neural Machine Translation Transfer
Transfer learning from a high-resource language pair `parent' has been proven to be an effective way to improve neural machine translation quality for low-resource language pairs `children.' However, previous approaches build a custom parent model or at least update an existing parent model's vocabulary for each child language pair they wish to train, in an effort to align parent and child vocabularies. This is not a practical solution. It is wasteful to devote the majority of training time for new language pairs to optimizing parameters on an unrelated data set. Further, this overhead reduces the utility of neural machine translation for deployment in humanitarian assistance scenarios, where extra time to deploy a new language pair can mean the difference between life and death. In this work, we present a `universal' pre-trained neural parent model with constant vocabulary that can be used as a starting point for training practically any new low-resource language to a fixed target language. We demonstrate that our approach, which leverages orthography unification and a broad-coverage approach to subword identification, generalizes well to several languages from a variety of families, and that translation systems built with our approach can be built more quickly than competing methods and with better quality as well.
['Mozhdeh Gheini', 'Jonathan May']
2019-09-14
null
null
null
null
['low-resource-neural-machine-translation']
['natural-language-processing']
[ 8.05506185e-02 -1.00646345e-02 -4.32144821e-01 -4.92502242e-01 -1.18110693e+00 -8.02496195e-01 3.94254357e-01 1.54537395e-01 -7.20384061e-01 1.04764152e+00 2.81834155e-01 -6.73134625e-01 3.87280285e-01 -8.72197986e-01 -8.05613279e-01 -3.59737426e-01 4.72983360e-01 1.11080849e+00 -7.09149763e-02 -7.76707947e-01 -2.94083059e-01 3.27947080e-01 -1.01607907e+00 2.04136968e-01 8.53972316e-01 1.74903944e-01 2.06235513e-01 4.34557915e-01 -2.35120252e-01 5.23350611e-02 -4.02788311e-01 -9.31875467e-01 3.97337377e-01 -4.58592504e-01 -8.91949654e-01 -4.72241729e-01 6.83362722e-01 -4.65821266e-01 -1.53536210e-02 6.49304867e-01 9.35824633e-01 -2.54991144e-01 5.66590846e-01 -7.46538520e-01 -8.51726592e-01 9.17756975e-01 -2.80084550e-01 2.75122285e-01 2.85272717e-01 7.56254513e-03 8.96541715e-01 -7.04067171e-01 7.10066855e-01 1.08195102e+00 1.06487644e+00 9.82305884e-01 -1.33591640e+00 -9.07142043e-01 -1.00336030e-01 -2.42538899e-01 -1.27110779e+00 -6.79144263e-01 2.61296421e-01 -2.82057434e-01 1.35840571e+00 2.20003538e-03 6.45185709e-01 1.21403587e+00 6.12553693e-02 5.93064606e-01 7.67035842e-01 -6.77837312e-01 -1.16486639e-01 2.10950151e-01 -6.70777857e-02 6.97332621e-01 4.11221057e-01 7.60455877e-02 -5.40301442e-01 -1.07928351e-01 6.10547900e-01 -3.64159077e-01 -1.40608824e-03 -1.89869642e-01 -1.22947419e+00 9.16558623e-01 1.29892454e-01 4.43253279e-01 -6.40552118e-02 2.10455596e-01 5.05347788e-01 6.89372420e-01 5.74753046e-01 7.09563255e-01 -6.57662213e-01 -2.93142438e-01 -9.75994825e-01 1.02670491e-02 8.34106922e-01 1.09682739e+00 7.09414124e-01 1.16539493e-01 1.39887705e-01 1.08175445e+00 2.03163829e-03 7.74390638e-01 6.74676478e-01 -5.30443192e-01 7.39782512e-01 3.24820995e-01 -2.75015056e-01 -4.74891126e-01 -4.34952497e-01 -4.43812579e-01 -6.67265654e-01 -2.66395986e-01 4.41623211e-01 -3.78401309e-01 -6.69337034e-01 2.14577174e+00 3.58792037e-01 -1.64775833e-01 3.47164810e-01 3.79242122e-01 8.04125428e-01 5.51488519e-01 -4.27359752e-02 -5.00997193e-02 1.16843379e+00 -9.13493097e-01 -1.91209555e-01 -5.96304953e-01 1.03300273e+00 -9.61501062e-01 1.01314533e+00 1.79774165e-01 -1.34437180e+00 -3.18794221e-01 -1.01858795e+00 -2.90876478e-01 -3.73057157e-01 1.50205418e-01 8.62856805e-01 9.70090151e-01 -1.32353389e+00 6.74791455e-01 -8.21109295e-01 -9.23212469e-01 1.24421552e-01 8.97214711e-01 -7.86614478e-01 -1.62638947e-01 -1.21911359e+00 1.38851190e+00 4.11743581e-01 -2.88672388e-01 -3.94021183e-01 -6.12993002e-01 -7.98875153e-01 -2.71320939e-01 -1.60420448e-01 -8.58631372e-01 1.19502997e+00 -1.09984756e+00 -1.50957859e+00 1.18155313e+00 2.90284008e-02 -2.94790447e-01 1.89939111e-01 -9.26181152e-02 -1.96599931e-01 -2.52136171e-01 2.29117095e-01 9.37268257e-01 5.03283441e-01 -5.48620343e-01 -4.59261417e-01 -3.44545156e-01 4.29338329e-02 3.28675479e-01 -6.87514484e-01 5.07211387e-01 -6.20855391e-01 -7.43150711e-01 -1.73403248e-01 -1.07733977e+00 2.14324519e-02 -2.67484963e-01 2.05845699e-01 -1.08159386e-01 -1.06051482e-01 -9.11957145e-01 1.16249263e+00 -1.84026885e+00 4.32769835e-01 4.57239337e-02 -2.13691130e-01 4.75633055e-01 -5.61813414e-01 6.45598173e-01 4.47183736e-02 4.13752645e-02 -2.53889441e-01 -4.31674004e-01 -2.36109778e-01 4.35726345e-01 -2.63482690e-01 4.56621140e-01 2.86033332e-01 1.10945320e+00 -9.90612626e-01 -4.67461526e-01 -2.90140599e-01 4.59547549e-01 -8.59388471e-01 5.86632565e-02 9.86663401e-02 4.29836333e-01 -2.67331377e-02 3.56559783e-01 3.15323681e-01 1.24295302e-01 3.07291061e-01 3.23573709e-01 1.09962739e-01 7.47564256e-01 -7.01716065e-01 1.94660389e+00 -8.37176681e-01 4.44237292e-01 -9.86347720e-02 -1.02935171e+00 9.15696800e-01 4.17983264e-01 3.84008616e-01 -5.96253812e-01 2.64467746e-01 8.30738366e-01 2.36196488e-01 -2.22576365e-01 3.55438977e-01 -4.44312453e-01 -2.35622957e-01 7.73144841e-01 3.79398733e-01 -4.10212278e-01 1.60424680e-01 -2.26454303e-01 1.07846999e+00 2.76259512e-01 3.13892066e-01 -2.75281608e-01 3.92780751e-01 9.16876569e-02 5.78955829e-01 4.29790169e-01 1.62736267e-01 6.64593279e-01 -9.95975509e-02 -4.37003940e-01 -1.42835343e+00 -1.07597029e+00 8.04190990e-03 1.46794510e+00 -4.39485878e-01 -2.25270346e-01 -8.06331933e-01 -5.05447030e-01 -2.39695549e-01 6.12627387e-01 -3.93146813e-01 -3.33460718e-01 -1.22673512e+00 -7.72282541e-01 1.13429093e+00 6.60634041e-01 1.45426109e-01 -8.48156214e-01 -2.95215547e-01 4.56051707e-01 -2.74392933e-01 -9.65849102e-01 -7.55958080e-01 3.07801783e-01 -8.97657156e-01 -4.85233247e-01 -9.92597103e-01 -1.27242243e+00 5.80532253e-01 1.43504500e-01 1.27845621e+00 2.49698505e-01 1.10778414e-01 2.56505758e-01 -2.87587404e-01 -3.72233242e-01 -8.74561131e-01 7.48345077e-01 5.26031256e-01 -4.63110328e-01 6.20805979e-01 -6.46859646e-01 -1.79771021e-01 7.42213652e-02 -6.48948371e-01 1.53848946e-01 5.37262440e-01 1.14451718e+00 2.25108311e-01 -7.33225405e-01 6.08680367e-01 -7.49462605e-01 6.34804130e-01 -5.58287680e-01 -5.44974804e-01 5.49665153e-01 -5.32009065e-01 2.90175587e-01 8.29047084e-01 -6.86967909e-01 -6.01402342e-01 8.43261182e-03 -3.86017352e-01 2.92463191e-02 2.11173996e-01 3.73513848e-01 -7.67602324e-02 5.90212122e-02 8.24656844e-01 7.73005933e-02 -1.91967729e-02 -5.88578403e-01 5.18173635e-01 9.39580262e-01 6.77432537e-01 -8.26414704e-01 8.58159900e-01 5.36993258e-02 -2.29523271e-01 -5.39153099e-01 -3.18151236e-01 -4.06721443e-01 -9.53372180e-01 4.27082211e-01 7.94962108e-01 -1.05390990e+00 -1.75859183e-01 2.40187734e-01 -1.43078268e+00 -4.97383356e-01 -1.66629225e-01 6.56466961e-01 -5.21668971e-01 1.51184380e-01 -4.54054028e-01 -1.27893388e-01 -8.28182459e-01 -1.19293499e+00 1.19153392e+00 -1.04077026e-01 -5.18535733e-01 -1.17336154e+00 5.41542172e-01 2.78918803e-01 5.49664617e-01 -3.60052139e-01 1.11341488e+00 -9.26056445e-01 -3.54218334e-02 -1.28977135e-01 -6.26319647e-02 4.63056028e-01 3.22091699e-01 -3.70838195e-01 -5.17216861e-01 -6.88810766e-01 -3.32515955e-01 -5.67165792e-01 3.50367725e-01 -1.59441531e-01 1.04168169e-01 -2.65425414e-01 -2.66591698e-01 8.01531076e-01 1.17165148e+00 -6.80382252e-02 3.47001284e-01 3.35408032e-01 6.20473802e-01 5.32640219e-01 2.13554263e-01 -1.96367621e-01 5.69255829e-01 1.03281140e+00 -3.57325882e-01 -1.93474531e-01 -3.14974099e-01 -3.61265272e-01 8.11389387e-01 1.42978311e+00 -2.38136239e-02 -3.15708891e-02 -1.39334404e+00 8.26807141e-01 -1.43109918e+00 -7.31839120e-01 3.42058957e-01 2.39839363e+00 1.40166390e+00 -1.83758795e-01 1.70351937e-01 -3.87261033e-01 5.42912006e-01 -5.14187038e-01 -2.59857118e-01 -8.88454914e-01 -1.16576798e-01 7.47643113e-01 6.64492309e-01 5.24167240e-01 -7.45043397e-01 1.50480294e+00 6.48011732e+00 7.40436614e-01 -1.42739606e+00 4.75464553e-01 2.37927780e-01 -2.53309160e-01 -4.89863873e-01 3.92230637e-02 -1.09374237e+00 2.99116224e-01 1.31686747e+00 -1.39425263e-01 7.58384168e-01 5.85823476e-01 -1.54912785e-01 3.40756953e-01 -1.48221397e+00 1.04063356e+00 3.97204459e-01 -1.12585640e+00 7.27798492e-02 -3.03927492e-02 6.09110475e-01 5.37927091e-01 -1.17959805e-01 4.87838149e-01 5.91270864e-01 -9.87704933e-01 4.98649269e-01 -1.34356514e-01 1.35177696e+00 -5.55718660e-01 5.45123279e-01 4.40315217e-01 -9.08781946e-01 1.76007152e-01 -6.01476312e-01 -2.70319339e-02 7.66297653e-02 -1.60353616e-01 -1.14042187e+00 2.91846842e-01 5.11814415e-01 2.77649403e-01 -5.85081398e-01 7.40481198e-01 -1.88021645e-01 6.12749338e-01 -6.72023535e-01 -9.81453955e-02 1.78350374e-01 -9.79463682e-02 2.07154661e-01 1.23566115e+00 4.56642061e-01 -5.38414977e-02 3.92669961e-02 1.94448695e-01 -2.72522539e-01 7.97387958e-01 -9.57645833e-01 -6.69109449e-02 4.14214730e-01 8.68818223e-01 -4.88445759e-01 -2.55079120e-01 -5.71169674e-01 1.13684642e+00 7.96466649e-01 -9.37023535e-02 -6.86177492e-01 -1.16644800e-01 5.66749990e-01 2.51675900e-02 1.51972294e-01 -3.28117043e-01 -2.13161126e-01 -1.44815397e+00 1.55751621e-02 -1.33476055e+00 2.13229194e-01 -5.12243629e-01 -1.10282111e+00 8.53511214e-01 5.50706461e-02 -1.03976941e+00 -6.95262432e-01 -6.48740709e-01 -4.07751977e-01 1.12629318e+00 -1.14129031e+00 -1.58535564e+00 3.37797344e-01 4.20106828e-01 5.41612029e-01 -6.07262731e-01 1.18907702e+00 7.48950601e-01 -3.08845758e-01 1.20730186e+00 2.45098650e-01 2.15785667e-01 9.92587745e-01 -9.87322330e-01 9.92653072e-01 8.64402533e-01 3.96262854e-01 8.80095959e-01 5.06715178e-01 -7.06103325e-01 -1.40506864e+00 -8.40566456e-01 1.36365449e+00 -7.02089667e-01 6.25537515e-01 -6.18976295e-01 -7.02832937e-01 8.95526409e-01 3.35414261e-01 -4.14956570e-01 9.14108396e-01 3.12473059e-01 -4.21573907e-01 -7.38703087e-02 -8.47752452e-01 8.27626348e-01 1.18589222e+00 -6.56052351e-01 -7.25564957e-01 4.86508906e-01 8.10158253e-01 -3.72226804e-01 -9.66745913e-01 4.23218012e-01 7.71223128e-01 -2.45203793e-01 8.86715293e-01 -9.94749248e-01 3.67356390e-01 1.11135446e-01 -2.25920662e-01 -1.43762708e+00 -1.56200677e-01 -7.29641736e-01 4.68479037e-01 1.42816496e+00 8.42520773e-01 -7.02801764e-01 5.59784353e-01 5.34586787e-01 -6.85226778e-03 -7.88133502e-01 -1.18011475e+00 -8.85499120e-01 8.99622619e-01 -5.02633333e-01 9.57114995e-01 9.15700436e-01 6.82839453e-02 8.85329068e-01 -4.35917705e-01 -5.73519655e-02 -4.29853834e-02 -1.34224847e-01 1.12158179e+00 -9.48502064e-01 -6.86360121e-01 -5.04229665e-01 -2.28790954e-01 -8.96520734e-01 2.79445887e-01 -1.51796281e+00 -1.35102570e-01 -1.39157641e+00 7.88210407e-02 -8.39912534e-01 -1.03830002e-01 7.70633399e-01 -1.51833102e-01 5.69188535e-01 1.06983751e-01 2.15010628e-01 8.20754692e-02 2.56102979e-01 7.14601040e-01 -2.28060603e-01 -3.63842726e-01 -2.54076421e-02 -7.76622653e-01 4.76909012e-01 8.37015927e-01 -5.89521289e-01 -3.88871819e-01 -1.21885407e+00 6.40969753e-01 -1.82301462e-01 -2.77281076e-01 -7.37090170e-01 8.43130350e-02 1.17534623e-02 5.04605696e-02 -1.11373052e-01 1.99927226e-01 -6.13392174e-01 1.49484336e-01 3.30620527e-01 -1.46423161e-01 5.53686917e-01 4.68168199e-01 -2.63994694e-01 1.57072395e-01 -4.86069918e-01 7.75620461e-01 -9.89792496e-02 -1.64389953e-01 2.46387810e-01 -1.59342006e-01 2.22064435e-01 6.00456357e-01 -1.72241062e-01 -3.01089231e-02 -2.30556518e-01 -2.50469327e-01 5.61014377e-02 8.24231505e-01 7.29316950e-01 1.80122048e-01 -1.23889041e+00 -1.02487850e+00 3.14220220e-01 2.13311777e-01 -3.51355553e-01 -4.62991565e-01 7.21934795e-01 -7.84873307e-01 3.92991751e-01 -3.75802696e-01 -3.11228842e-01 -1.23953784e+00 5.71525395e-01 2.39444599e-01 -3.45018238e-01 -4.65037555e-01 9.62919354e-01 -8.47580880e-02 -9.84951258e-01 -1.14070371e-01 -1.64432108e-01 -7.52111301e-02 -4.86750975e-02 3.60857189e-01 1.43255375e-03 3.01124156e-01 -1.07721925e+00 -2.88670152e-01 7.81900406e-01 -1.99551225e-01 -3.80965024e-01 1.40831268e+00 -1.48460548e-02 -2.79907733e-01 2.67899454e-01 1.19822049e+00 2.88837045e-01 -4.15072948e-01 -3.28658491e-01 -1.00897558e-01 -1.05367154e-02 -3.06595683e-01 -6.26071632e-01 -8.62395763e-01 1.03978419e+00 5.08191228e-01 -5.64806700e-01 1.02863979e+00 3.37583050e-02 1.18265045e+00 7.20541179e-01 7.15500295e-01 -1.04073799e+00 -4.11080003e-01 6.43536806e-01 6.31060600e-01 -1.19757068e+00 -1.88612062e-02 -4.71214093e-02 -3.91347736e-01 9.86430883e-01 4.17896777e-01 1.40475988e-01 3.05389613e-01 2.51045763e-01 2.63868511e-01 2.04502180e-01 -6.25082076e-01 -1.26953408e-01 2.82140106e-01 6.47983134e-01 6.56030834e-01 5.79106808e-02 -5.93950629e-01 4.12026852e-01 -8.42926025e-01 -1.79781601e-01 2.97234207e-01 7.17245817e-01 -1.16127700e-01 -1.91866732e+00 -2.59830624e-01 3.03991616e-01 -6.41612887e-01 -5.72105587e-01 -3.92585844e-01 6.42231345e-01 3.89483839e-01 6.76691115e-01 3.65188457e-02 -2.02835783e-01 1.83916196e-01 2.59733498e-01 7.97380328e-01 -1.14831269e+00 -8.68271172e-01 -1.76415354e-01 2.38270655e-01 -1.28494620e-01 -1.25339851e-01 -6.46838665e-01 -1.06480849e+00 -3.86719257e-01 -3.88404727e-01 1.14182821e-02 7.51365542e-01 1.01192451e+00 2.48060152e-01 -1.55110076e-01 2.77082175e-01 -6.10167205e-01 -5.61615288e-01 -8.52149606e-01 1.10423744e-01 2.00045034e-01 -1.89397410e-02 -4.35975075e-01 2.83024877e-01 -2.32429896e-03]
[11.557246208190918, 10.32866096496582]
9b68fc89-d294-43c0-bffa-a18726d4692c
ssd-monodtr-supervised-scale-constrained
2305.07270
null
https://arxiv.org/abs/2305.07270v3
https://arxiv.org/pdf/2305.07270v3.pdf
SSD-MonoDETR: Supervised Scale-aware Deformable Transformer for Monocular 3D Object Detection
Transformer-based methods have demonstrated superior performance for monocular 3D object detection recently, which aims at predicting 3D attributes from a single 2D image. Most existing transformer-based methods leverage both visual and depth representations to explore valuable query points on objects, and the quality of the learned query points has a great impact on detection accuracy. Unfortunately, existing unsupervised attention mechanisms in transformers are prone to generate low-quality query features due to inaccurate receptive fields, especially on hard objects. To tackle this problem, this paper proposes a novel Supervised Scale-aware Deformable Attention (SSDA) for monocular 3D object detection. Specifically, SSDA presets several masks with different scales and utilizes depth and visual features to adaptively learn a scale-aware filter for object query augmentation. Imposing the scale awareness, SSDA could well predict the accurate receptive field of an object query to support robust query feature generation. Aside from this, SSDA is assigned with a Weighted Scale Matching (WSM) loss to supervise scale prediction, which presents more confident results as compared to the unsupervised attention mechanisms. Extensive experiments on the KITTI benchmark demonstrate that SSDA significantly improves the detection accuracy, especially on moderate and hard objects, yielding state-of-the-art performance as compared to the existing approaches. Our code will be made publicly available at https://github.com/mikasa3lili/SSD-MonoDETR.
['Meng Wang', 'Haolong Fu', 'Jiacheng Lin', 'Zhiyong Li', 'Kailun Yang', 'Jin Yuan', 'Fan Yang', 'Xuan He']
2023-05-12
null
null
null
null
['monocular-3d-object-detection']
['computer-vision']
[-5.16674109e-02 -3.00421119e-01 -1.98523223e-01 -3.34668696e-01 -7.42736280e-01 -4.19607282e-01 4.29610312e-01 -3.26860137e-02 -2.60536700e-01 1.12155266e-01 5.80404792e-03 2.47721355e-02 -2.32149921e-02 -5.39122462e-01 -7.86710441e-01 -7.22387016e-01 1.42698035e-01 3.66193384e-01 7.67002761e-01 2.63294950e-02 3.63911271e-01 7.82492936e-01 -1.70421481e+00 2.42837518e-01 7.01230943e-01 1.51929688e+00 7.41042912e-01 5.25919557e-01 -7.36315176e-02 4.33723956e-01 -4.12941337e-01 -2.74184465e-01 5.80340028e-01 1.75447226e-01 -3.42390418e-01 -4.93885987e-02 6.99636102e-01 -6.91887915e-01 -6.56092286e-01 1.08415318e+00 7.15432227e-01 -1.64467674e-02 6.57248259e-01 -1.32559967e+00 -8.74202311e-01 2.06444785e-01 -6.90618455e-01 5.91828644e-01 2.38174751e-01 3.86044025e-01 9.84676659e-01 -1.46163499e+00 3.19630235e-01 1.37333739e+00 2.92554796e-01 2.76053190e-01 -9.05579865e-01 -8.06410849e-01 2.93150783e-01 3.59694332e-01 -1.59389150e+00 -4.17378128e-01 8.54432344e-01 -2.93634444e-01 1.09242129e+00 2.45825201e-01 7.10256994e-01 7.85768807e-01 1.02345034e-01 1.13041735e+00 9.44632411e-01 -1.66789100e-01 -4.03944962e-02 1.30205691e-01 -2.55086988e-01 6.65445387e-01 3.12548816e-01 3.20520550e-01 -6.07615173e-01 7.01014176e-02 1.05200708e+00 2.23819315e-01 -2.37214088e-01 -7.22582817e-01 -1.21893752e+00 6.05581164e-01 1.06043446e+00 9.81835797e-02 -3.63156348e-01 6.13658242e-02 5.88140264e-02 5.38322106e-02 4.73523289e-01 3.10129732e-01 -3.60353261e-01 2.02735230e-01 -5.19264698e-01 2.35468552e-01 2.69467440e-02 1.09998739e+00 8.08670759e-01 -1.81520224e-01 -5.80199480e-01 7.51413047e-01 3.38739485e-01 7.24106014e-01 2.83874929e-01 -7.17800975e-01 5.27750552e-01 1.09121740e+00 2.22774386e-01 -1.09278965e+00 -2.42115676e-01 -5.98807633e-01 -5.06304383e-01 3.17550659e-01 2.22160250e-01 5.07736564e-01 -1.16164362e+00 1.43976021e+00 6.17410660e-01 -1.84201047e-01 -3.01065594e-01 1.46698928e+00 1.01247644e+00 4.34529245e-01 -1.41926482e-01 1.19702198e-01 1.21403015e+00 -8.16900253e-01 -3.42359394e-01 -3.91187757e-01 2.45273784e-01 -9.72250700e-01 1.26506746e+00 1.77460060e-01 -1.13719702e+00 -7.02839971e-01 -9.31064308e-01 -3.66324216e-01 -2.99680531e-01 4.67681885e-01 5.83723307e-01 4.16517168e-01 -8.29447687e-01 1.21470708e-02 -8.31315815e-01 -2.65600979e-01 7.18801200e-01 4.16273743e-01 -1.06207661e-01 -8.44736770e-02 -9.06917751e-01 6.83756411e-01 2.85049289e-01 1.88329350e-02 -1.02754569e+00 -5.27978778e-01 -7.82835305e-01 1.36547253e-01 5.55420637e-01 -7.06666648e-01 1.21566463e+00 -5.14677525e-01 -1.19909155e+00 1.10625529e+00 -1.93745017e-01 -3.61332983e-01 5.61453283e-01 -2.94247448e-01 1.29892200e-01 3.57619494e-01 3.91956985e-01 1.02918160e+00 1.12931788e+00 -1.08895159e+00 -7.59798288e-01 -6.70167267e-01 2.98431009e-01 6.31120980e-01 -4.15973604e-01 5.79914916e-03 -8.06486249e-01 -7.17326224e-01 5.91833532e-01 -6.63948536e-01 1.30287424e-01 5.30474424e-01 -3.77018303e-01 -5.90911925e-01 7.79413998e-01 -2.27244869e-01 9.60796475e-01 -2.12388873e+00 1.16743669e-01 -1.77517727e-01 2.29870513e-01 2.22513810e-01 -4.75832634e-02 4.86133387e-03 3.43465775e-01 -1.79185256e-01 6.48649186e-02 -3.17703456e-01 -3.74878868e-02 -1.36560693e-01 -3.24356705e-01 5.04862845e-01 5.08815467e-01 1.16653299e+00 -7.01339662e-01 -5.61351180e-01 4.08127248e-01 4.46211874e-01 -5.60286045e-01 3.80078673e-01 -2.92855024e-01 2.17118278e-01 -7.21190810e-01 1.28746426e+00 8.40186238e-01 -4.45387125e-01 -5.04620194e-01 -4.20375496e-01 -4.78107482e-02 8.49828348e-02 -1.05315638e+00 1.70610654e+00 -1.41668364e-01 3.65912616e-01 -9.01993290e-02 -7.32206523e-01 1.17358208e+00 -1.43271387e-01 3.00028712e-01 -8.42579126e-01 1.69364721e-01 2.72844374e-01 -2.61198908e-01 -3.81074280e-01 3.57530028e-01 2.74053484e-01 1.03217311e-01 1.40744671e-01 -2.30173513e-01 -3.02181125e-01 -2.28680819e-01 1.94686562e-01 7.76848853e-01 9.46890116e-02 1.66129559e-01 1.55476676e-02 4.94054168e-01 -1.35985330e-01 4.71989274e-01 8.07809353e-01 -4.56273854e-01 7.47199595e-01 1.78921837e-02 -3.53903532e-01 -9.02761877e-01 -1.26889288e+00 -2.65933901e-01 1.00152349e+00 8.27032506e-01 -6.08249940e-02 -2.67987072e-01 -7.88065851e-01 3.64985973e-01 2.14991793e-01 -5.91169953e-01 -2.32240468e-01 -3.57077926e-01 -3.75537783e-01 1.90294430e-01 7.28682220e-01 5.35265863e-01 -1.12189472e+00 -9.46285009e-01 3.01094744e-02 1.77802201e-02 -1.16719890e+00 -7.08782077e-01 5.84512241e-02 -7.10182726e-01 -1.05453813e+00 -9.14180338e-01 -8.88524055e-01 6.89631999e-01 9.61049557e-01 8.26297879e-01 2.39698254e-02 -6.10144436e-01 4.28855002e-01 -3.13712716e-01 -6.10193789e-01 1.56288818e-01 1.12599373e-01 1.50130212e-01 1.13444023e-01 5.55317819e-01 -2.89657295e-01 -1.11065817e+00 6.14529490e-01 -7.72656679e-01 1.97298825e-02 9.54338670e-01 7.87822008e-01 9.17774677e-01 -3.10842842e-01 4.08951908e-01 -1.85111821e-01 5.99027909e-02 -1.94639817e-01 -8.57174754e-01 1.71627507e-01 -4.87528741e-01 -1.03248499e-01 1.88624486e-01 -6.16959095e-01 -8.60907078e-01 1.51972145e-01 1.27822965e-01 -9.93594587e-01 -9.92057472e-02 3.33578251e-02 -2.16127753e-01 -4.30103868e-01 5.96612036e-01 4.77550328e-01 -1.73630491e-01 -4.94573653e-01 8.25398788e-02 5.88963330e-01 2.59492308e-01 -3.07173312e-01 9.97009039e-01 7.18011081e-01 -2.09247410e-01 -6.06772900e-01 -9.51035321e-01 -7.74370670e-01 -4.89085138e-01 -1.69073477e-01 7.77714550e-01 -1.08913207e+00 -6.44615412e-01 6.14458263e-01 -9.38196182e-01 -1.55173540e-01 -2.24148050e-01 5.77402651e-01 -5.62773645e-01 2.39965931e-01 -4.50644672e-01 -9.33277488e-01 -4.19445395e-01 -1.16971612e+00 1.56073558e+00 4.09956783e-01 3.94977987e-01 -3.71275306e-01 -3.49004447e-01 4.37890202e-01 3.97008270e-01 -8.08277801e-02 6.32890105e-01 -3.64618063e-01 -1.28169644e+00 -2.15806141e-01 -7.44732499e-01 1.30380183e-01 1.06657045e-02 -3.89215350e-01 -1.12752354e+00 -5.60684383e-01 -7.77554512e-02 -5.25009394e-01 9.58207190e-01 4.51338381e-01 1.28976095e+00 -4.91259396e-02 -4.16347504e-01 8.21387708e-01 1.23992562e+00 1.54672310e-01 2.65606135e-01 3.55507106e-01 7.52189100e-01 2.97758579e-01 1.07880068e+00 4.95435506e-01 3.58166575e-01 9.06701088e-01 8.12287867e-01 -8.01623389e-02 -2.05201775e-01 -2.75022537e-01 1.50338218e-01 2.64583409e-01 1.75152346e-01 -2.28808280e-02 -7.40043640e-01 5.56474566e-01 -1.52142751e+00 -7.09872365e-01 2.67519474e-01 2.21695995e+00 6.29763424e-01 3.77575487e-01 1.12080537e-01 1.10891044e-01 6.12061262e-01 1.35208324e-01 -1.06840324e+00 2.88559645e-01 -3.56294155e-01 -1.52290508e-01 4.62808579e-01 1.12703875e-01 -1.19672251e+00 9.55066204e-01 4.99699926e+00 8.59702826e-01 -1.22579575e+00 8.99748132e-02 4.20768380e-01 -3.38287264e-01 -1.98649317e-01 -1.73999473e-01 -1.17740166e+00 3.05038333e-01 2.82829665e-02 -3.49087492e-02 2.12789878e-01 9.50549185e-01 6.84661865e-02 -9.77387130e-02 -8.56531024e-01 1.20366693e+00 1.35791495e-01 -1.06083643e+00 1.77507803e-01 6.14382364e-02 6.21394873e-01 2.67262369e-01 3.78061980e-01 1.87145352e-01 -2.95019448e-01 -7.24690974e-01 8.80254865e-01 3.96132141e-01 7.89199114e-01 -6.08860254e-01 5.39161086e-01 3.58639330e-01 -1.45451260e+00 -3.60861629e-01 -7.70782292e-01 1.64952487e-01 -1.34789631e-01 3.88315529e-01 -9.10463750e-01 3.02662015e-01 1.13318574e+00 7.41473675e-01 -8.70169759e-01 1.48496854e+00 -8.57558399e-02 1.58668697e-01 -5.16071916e-01 -1.50349557e-01 9.69464704e-02 2.86048084e-01 8.79132032e-01 8.28725517e-01 2.91435152e-01 1.54973686e-01 3.17260385e-01 1.00655138e+00 -9.35965851e-02 2.23762065e-01 -3.69387150e-01 8.33289400e-02 6.95020378e-01 1.13196456e+00 -7.09913731e-01 -1.96174577e-01 -4.76410240e-01 9.11338866e-01 4.12547886e-01 2.62194097e-01 -7.24563777e-01 -3.18703324e-01 5.70103645e-01 4.77810204e-01 7.66424954e-01 -1.39583781e-01 -2.01347440e-01 -1.20545566e+00 3.48371387e-01 -6.12930238e-01 4.06367272e-01 -9.86674011e-01 -1.29206979e+00 5.05257308e-01 -9.54063237e-02 -1.39786208e+00 2.99623817e-01 -6.61982298e-01 -2.78148383e-01 7.91615367e-01 -1.76052845e+00 -1.28776205e+00 -7.69467235e-01 5.71220517e-01 7.01920033e-01 -1.09614477e-01 2.33974338e-01 3.31751555e-01 -2.76711136e-01 8.42964649e-01 -3.36081862e-01 -1.35246441e-01 8.62611234e-01 -1.14400244e+00 4.93978262e-01 6.59386039e-01 2.22814947e-01 3.35509241e-01 3.04235160e-01 -6.29946709e-01 -1.57599533e+00 -1.06434536e+00 5.58235765e-01 -6.13542914e-01 3.59039456e-01 -3.26215893e-01 -1.04415357e+00 3.15073758e-01 -3.44532192e-01 3.11713964e-01 7.04217032e-02 -3.58441383e-01 -4.47743207e-01 -2.69041836e-01 -9.10970211e-01 3.31851870e-01 1.33758163e+00 -6.13741815e-01 -4.81153965e-01 3.44915897e-01 8.45653951e-01 -6.82261527e-01 -6.99676156e-01 7.13790298e-01 5.21012068e-01 -1.25403070e+00 1.27491784e+00 -1.59321502e-01 -4.11956385e-02 -5.61086297e-01 -2.10266754e-01 -6.44385993e-01 -3.55980515e-01 -1.40956566e-01 -3.33717287e-01 8.46360505e-01 4.99737561e-02 -5.26893795e-01 9.13824201e-01 1.82264939e-01 -2.02758521e-01 -1.01951754e+00 -1.05266285e+00 -6.60327852e-01 -2.99978018e-01 -2.84669727e-01 6.16286695e-01 5.35478771e-01 -6.41893983e-01 1.66913375e-01 -1.57486349e-01 4.78442848e-01 8.18023205e-01 5.41601002e-01 8.75022352e-01 -1.16111696e+00 -2.62613207e-01 -5.97013175e-01 -7.75800467e-01 -1.59213066e+00 -2.44322538e-01 -8.52709055e-01 -4.44613285e-02 -1.40716577e+00 3.40376079e-01 -6.20281696e-01 -3.96002084e-01 3.80953044e-01 -3.70880604e-01 5.98247468e-01 2.35118166e-01 5.42479753e-01 -7.37154305e-01 8.45402420e-01 1.53069282e+00 -2.13120341e-01 -3.14013243e-01 2.40345255e-01 -5.03412545e-01 4.56195831e-01 7.25064456e-01 -2.97569662e-01 -4.10632253e-01 -5.15436471e-01 7.19402507e-02 -9.12899598e-02 7.97609270e-01 -9.89983439e-01 3.80053133e-01 -8.09205845e-02 7.26249695e-01 -1.21637189e+00 5.66476882e-01 -6.32320940e-01 -4.04475510e-01 4.67807680e-01 -2.06235886e-01 -1.65713832e-01 1.39276698e-01 7.63504028e-01 -1.12372980e-01 1.76764265e-01 8.51777911e-01 -4.60095443e-02 -9.22418356e-01 7.64325976e-01 1.71056181e-01 5.34906797e-02 1.02971792e+00 -4.25913185e-01 -1.92097738e-01 -1.32085353e-01 -4.51244086e-01 3.54962498e-01 6.48012221e-01 7.31488883e-01 1.09191716e+00 -1.37671447e+00 -6.49245441e-01 5.53764462e-01 5.35083413e-01 3.44334543e-01 2.86578774e-01 7.45070994e-01 -2.92572737e-01 5.81731558e-01 -1.35265470e-01 -1.02740061e+00 -1.12658608e+00 7.22888827e-01 4.20316011e-01 2.28170410e-01 -5.39655566e-01 1.16891634e+00 7.00124443e-01 -8.24274048e-02 5.43215990e-01 -4.99059170e-01 -7.49343410e-02 -1.19068302e-01 4.30010885e-01 1.09008491e-01 -1.61689725e-02 -5.52828312e-01 -5.09084582e-01 9.33296025e-01 -2.92219192e-01 2.81308860e-01 1.02113032e+00 -3.70759934e-01 3.89728576e-01 3.06933939e-01 1.04759383e+00 -2.30994493e-01 -1.66254568e+00 -6.81453168e-01 -3.43369752e-01 -8.83266211e-01 1.99655622e-01 -6.74706578e-01 -9.07893062e-01 9.81644392e-01 9.20138240e-01 -1.45259947e-02 1.17201006e+00 4.14165646e-01 6.24498963e-01 3.43479276e-01 4.75968212e-01 -6.76656067e-01 5.92364073e-01 2.70368040e-01 1.15421176e+00 -1.69467616e+00 -3.58478650e-02 -5.15317798e-01 -4.99198794e-01 8.10061157e-01 1.04133010e+00 -5.99674881e-03 4.23660904e-01 -8.09768811e-02 2.79039284e-03 -2.09234729e-01 -5.41657209e-01 -4.38383996e-01 7.53597379e-01 4.71106768e-01 -2.13442240e-02 -1.43443108e-01 1.60395145e-01 4.46528673e-01 1.19292758e-01 -3.00307870e-01 -3.66523638e-02 8.44054520e-01 -7.02842772e-01 -6.10320091e-01 -4.20478106e-01 4.39278215e-01 -2.00573370e-01 -8.60657319e-02 -4.02045280e-01 5.75172544e-01 2.56202240e-02 6.61109626e-01 9.95610431e-02 -2.76318163e-01 4.49853987e-01 -4.02248174e-01 5.95424116e-01 -6.15111172e-01 -2.29189247e-01 2.82560706e-01 -5.69734156e-01 -6.56677365e-01 -3.35754752e-01 -5.09190202e-01 -1.22639573e+00 1.22482195e-01 -6.27064049e-01 -2.12334588e-01 4.62017834e-01 5.00913322e-01 5.25753200e-01 1.63711846e-01 7.41789043e-01 -1.12335181e+00 -8.54449511e-01 -7.82772243e-01 -3.47243905e-01 3.20200473e-01 3.19853723e-01 -1.19945157e+00 -4.21584964e-01 -4.35483873e-01]
[7.944461345672607, -2.5206286907196045]
6d882678-471a-4aa7-9bbf-aa9fd8686330
learning-multilingual-word-embeddings-using
1905.12260
null
https://arxiv.org/abs/1905.12260v1
https://arxiv.org/pdf/1905.12260v1.pdf
Learning Multilingual Word Embeddings Using Image-Text Data
There has been significant interest recently in learning multilingual word embeddings -- in which semantically similar words across languages have similar embeddings. State-of-the-art approaches have relied on expensive labeled data, which is unavailable for low-resource languages, or have involved post-hoc unification of monolingual embeddings. In the present paper, we investigate the efficacy of multilingual embeddings learned from weakly-supervised image-text data. In particular, we propose methods for learning multilingual embeddings using image-text data, by enforcing similarity between the representations of the image and that of the text. Our experiments reveal that even without using any expensive labeled data, a bag-of-words-based embedding model trained on image-text data achieves performance comparable to the state-of-the-art on crosslingual semantic similarity tasks.
['Balder ten Cate', 'Karthik Raman', 'Karan Singhal']
2019-05-29
learning-multilingual-word-embeddings-using-1
https://aclanthology.org/W19-1807
https://aclanthology.org/W19-1807.pdf
ws-2019-6
['multilingual-word-embeddings']
['methodology']
[-3.26834857e-01 -1.21090733e-01 -3.38517845e-01 -4.24870551e-01 -1.03431427e+00 -5.67172885e-01 8.98768127e-01 5.21335244e-01 -8.56865823e-01 3.06785136e-01 6.54205561e-01 -2.78489918e-01 2.75928706e-01 -4.42618966e-01 -6.83216095e-01 -2.43872851e-01 1.96513772e-01 4.51000750e-01 2.11470691e-03 -2.03526810e-01 3.85365412e-02 7.63229057e-02 -1.32738018e+00 2.13923812e-01 6.75783277e-01 5.67500234e-01 4.02570099e-01 2.71330684e-01 -4.14085537e-01 3.54525656e-01 2.60832962e-02 -6.08777285e-01 1.83994725e-01 -3.95012856e-01 -6.72389328e-01 -5.03599197e-02 9.70586419e-01 -1.13470688e-01 -6.24251962e-01 1.10864353e+00 5.33283353e-01 -1.62317529e-01 6.25874341e-01 -1.11881614e+00 -1.46683061e+00 2.35394388e-01 -3.83195013e-01 1.70340180e-01 4.32268858e-01 -1.05301179e-01 1.38646936e+00 -1.14909530e+00 9.09821928e-01 1.35337150e+00 7.92223454e-01 4.02106553e-01 -1.42825663e+00 -4.38090056e-01 9.83445048e-02 3.15535814e-01 -1.57654822e+00 -4.19023424e-01 4.77604091e-01 -5.49646676e-01 1.13488531e+00 -4.30871338e-01 4.73279268e-01 1.04736423e+00 2.45488942e-01 6.10830009e-01 1.18109703e+00 -9.59394932e-01 -3.16244870e-01 5.00591934e-01 -2.50294745e-01 9.45330024e-01 2.71066993e-01 6.19194750e-03 -4.44973707e-01 -8.45896900e-02 5.29864192e-01 -3.49755250e-02 -2.02904582e-01 -7.00337172e-01 -1.30430663e+00 1.13454986e+00 3.89475644e-01 7.35529602e-01 -9.79771614e-02 1.30694211e-01 5.86590886e-01 2.99962282e-01 1.00520003e+00 4.35091645e-01 -4.84200239e-01 -4.05419953e-02 -7.57594228e-01 3.05883326e-02 4.99224931e-01 9.68004584e-01 1.01507080e+00 -1.79649755e-01 3.61939371e-01 1.06380296e+00 3.36089224e-01 4.94708598e-01 9.14693713e-01 -4.78748053e-01 3.65785927e-01 4.95445073e-01 -2.16961414e-01 -9.10891831e-01 -1.96528826e-02 1.47453919e-01 -1.45685360e-01 -1.13823548e-01 2.78212965e-01 1.85197577e-01 -7.21773028e-01 1.90523732e+00 1.20549716e-01 4.15303893e-02 4.45476532e-01 8.19613338e-01 4.53225464e-01 4.70444262e-01 1.21193774e-01 2.96631664e-01 1.49044478e+00 -1.06833541e+00 -6.03570700e-01 -3.48697037e-01 8.58706892e-01 -1.19133210e+00 1.45823252e+00 -1.29042029e-01 -7.59580672e-01 -6.52615070e-01 -9.85106051e-01 -4.96534765e-01 -8.93366516e-01 4.14155517e-03 4.66342747e-01 4.53429371e-01 -1.18726599e+00 1.87019274e-01 -5.80763936e-01 -1.07532668e+00 8.28633755e-02 -1.18443094e-01 -9.24710929e-01 -7.31676042e-01 -1.14261758e+00 1.38294053e+00 2.38038406e-01 -4.50092316e-01 -9.05305386e-01 -8.31293583e-01 -1.41361165e+00 -2.32554942e-01 3.85566847e-03 -4.47022676e-01 9.55582201e-01 -1.20468545e+00 -8.50309849e-01 1.38485968e+00 -1.79272443e-01 -1.68313161e-01 1.68482304e-01 -3.86739820e-01 -5.87715685e-01 2.76916027e-01 5.33969998e-01 9.82501805e-01 7.39098489e-01 -1.32029653e+00 -4.68776166e-01 -5.14975965e-01 1.62561059e-01 2.26421103e-01 -8.42150033e-01 1.96048379e-01 -6.16006672e-01 -7.92862058e-01 -2.43940175e-01 -1.07353401e+00 -5.37359528e-02 3.32470179e-01 2.29240984e-01 -3.50147009e-01 8.30893934e-01 -8.65090132e-01 9.94939268e-01 -2.39131999e+00 1.00953057e-02 -3.23973030e-01 -2.28575230e-01 1.34970739e-01 -5.75868964e-01 7.39912093e-01 -2.51532733e-01 1.46306202e-01 8.75898916e-03 -4.53627437e-01 2.18977064e-01 5.66902995e-01 -2.69151062e-01 7.18459189e-01 3.40154111e-01 8.54080558e-01 -1.14282608e+00 -8.40423405e-01 1.38639972e-01 6.28697097e-01 -5.96972167e-01 3.00855368e-01 2.12002695e-02 1.42008886e-01 -2.38570571e-01 5.03719032e-01 3.47178429e-01 -8.16393569e-02 5.21788716e-01 -4.78634983e-01 -2.37194791e-01 3.10158908e-01 -5.90260744e-01 2.45467663e+00 -1.06990635e+00 8.43368888e-01 -2.93866932e-01 -9.72840548e-01 5.42120159e-01 4.93549258e-01 3.27463210e-01 -8.70970249e-01 -1.06493041e-01 4.80398417e-01 -9.74277183e-02 -6.27520502e-01 5.76035380e-01 -4.61306661e-01 -2.55066723e-01 6.90871954e-01 6.06001973e-01 -2.45480701e-01 2.26715773e-01 2.33760431e-01 6.78619385e-01 2.30979696e-01 2.57692248e-01 -4.66950744e-01 4.52776492e-01 7.42110163e-02 1.86384931e-01 2.82111079e-01 -1.54683933e-01 5.04991531e-01 5.61996512e-02 -3.59544128e-01 -1.47989941e+00 -1.37311661e+00 -4.35030878e-01 1.19722009e+00 -2.13490631e-02 -6.06893659e-01 -5.54149449e-01 -7.46395111e-01 1.81976438e-01 5.70165157e-01 -6.14237726e-01 3.82407792e-02 -3.79309893e-01 -4.88255054e-01 5.39503217e-01 5.96165597e-01 4.09343876e-02 -7.46395469e-01 -3.77897806e-02 1.76573500e-01 -5.27154021e-02 -1.62564564e+00 -7.98802137e-01 4.94412845e-03 -6.73721433e-01 -9.97722924e-01 -6.92800224e-01 -1.31577623e+00 8.42762351e-01 3.09471488e-01 1.20166898e+00 -1.29984230e-01 -6.27983093e-01 9.28800046e-01 -5.69721818e-01 -2.54419267e-01 -2.41078466e-01 -1.18817845e-02 3.10437053e-01 -4.88807075e-02 6.40854716e-01 -3.57058167e-01 -2.20783502e-01 7.63693452e-03 -1.09146166e+00 -3.49049091e-01 4.18088645e-01 8.85292232e-01 5.79887688e-01 -3.75524849e-01 4.99054283e-01 -6.12946630e-01 6.13836169e-01 -4.09867138e-01 -3.69171411e-01 4.58226264e-01 -7.05479085e-01 4.03998554e-01 4.86849099e-01 -5.68565428e-01 -7.99352288e-01 -2.94526517e-01 5.62243275e-02 -5.80371916e-01 -5.84869385e-02 5.43857455e-01 2.57429853e-02 -1.51482284e-01 4.15085524e-01 7.54167140e-02 -6.63830414e-02 -4.99267817e-01 8.58358204e-01 7.95522392e-01 4.25248295e-01 -8.46133471e-01 6.38977408e-01 6.17081881e-01 -4.55861866e-01 -9.56682086e-01 -1.02960491e+00 -6.88785672e-01 -8.79823029e-01 7.12695867e-02 1.38090885e+00 -1.38867962e+00 3.22745055e-01 1.21367067e-01 -1.17625248e+00 -4.88497987e-02 -2.89332300e-01 8.94261837e-01 -4.94138777e-01 4.76382583e-01 -6.77955449e-01 -1.99758783e-01 6.19871058e-02 -1.19141841e+00 1.12644327e+00 -2.26554230e-01 -2.44418144e-01 -1.59019160e+00 4.82735574e-01 3.30469519e-01 2.53321141e-01 -1.11825533e-01 1.05862939e+00 -7.49288917e-01 -2.80541748e-01 -2.65065879e-01 -3.97402257e-01 5.70923567e-01 3.89791399e-01 -3.05076867e-01 -1.00788748e+00 -4.86655653e-01 -3.38010848e-01 -8.01339984e-01 6.61642909e-01 5.09898327e-02 7.75210261e-01 4.88503203e-02 -1.85027525e-01 4.95666951e-01 1.82945180e+00 -3.86374742e-01 2.91993290e-01 4.83611524e-01 7.76875079e-01 6.73986554e-01 4.70478863e-01 1.11838885e-01 8.41475427e-01 6.86292529e-01 3.77113223e-02 -2.97481418e-01 -2.51053035e-01 -5.13321221e-01 4.26303327e-01 1.32856548e+00 3.51320684e-01 -2.73339953e-02 -8.90092731e-01 1.28278601e+00 -1.72626495e+00 -6.84861660e-01 1.19193278e-01 2.06936193e+00 9.74984825e-01 -2.08449200e-01 -3.15802217e-01 -3.57316345e-01 5.79961658e-01 4.22482848e-01 -1.54627413e-01 -6.49107099e-01 -9.84968916e-02 5.06785691e-01 5.03266454e-01 6.31861150e-01 -1.07919693e+00 1.25855112e+00 6.36334372e+00 6.27993345e-01 -1.04300940e+00 7.30338514e-01 8.88749212e-03 1.66023225e-01 -4.77433622e-01 1.78824499e-01 -5.53012550e-01 1.51738241e-01 9.23790514e-01 -1.85089037e-01 1.97711602e-01 7.62489140e-01 -7.98619986e-02 -5.15386276e-02 -1.33303487e+00 9.65813041e-01 8.58779192e-01 -1.04005587e+00 2.27831066e-01 1.17067337e-01 9.71897006e-01 3.61441076e-01 1.70086861e-01 2.35448629e-01 4.04255182e-01 -9.07230735e-01 6.51394248e-01 1.94220752e-01 8.85367990e-01 -5.41235626e-01 6.08389199e-01 -1.21939003e-01 -1.30394840e+00 2.65101492e-01 -5.34119189e-01 2.35820994e-01 1.90109879e-01 3.69093955e-01 -5.71872056e-01 6.72352195e-01 7.48289645e-01 9.36565697e-01 -6.23848200e-01 4.35477436e-01 -3.21550608e-01 1.44857600e-01 1.00637458e-01 3.30755502e-01 6.75130725e-01 -2.13235676e-01 1.15586771e-02 1.41961813e+00 3.44899148e-01 -4.69325513e-01 5.50783813e-01 6.20989084e-01 -2.17244491e-01 4.66911823e-01 -1.08233678e+00 -4.79220629e-01 2.38554701e-01 1.11492860e+00 -2.03092784e-01 -3.18890005e-01 -1.17514968e+00 1.17922115e+00 6.37990832e-01 1.94046766e-01 -5.29601216e-01 -2.25049675e-01 9.52829838e-01 1.07572943e-01 3.49185228e-01 -5.20057559e-01 1.48769349e-01 -1.31787741e+00 1.33801419e-02 -6.92256391e-01 3.47881943e-01 -9.60121036e-01 -1.84983039e+00 5.67856491e-01 -6.09209538e-02 -1.02377045e+00 -1.93052679e-01 -8.60511422e-01 -3.85859847e-01 9.66541588e-01 -1.88859963e+00 -1.68866479e+00 6.97479174e-02 7.86941886e-01 5.05524695e-01 -3.41952324e-01 1.19977844e+00 5.73179603e-01 -1.20524250e-01 6.22417331e-01 2.62567222e-01 2.66888559e-01 1.28764486e+00 -1.15498745e+00 2.23478869e-01 6.97592080e-01 6.08334363e-01 6.92424059e-01 5.47513425e-01 -3.92610490e-01 -1.38140595e+00 -1.03286147e+00 1.53286469e+00 -4.94509637e-01 1.34011459e+00 -4.24902737e-01 -8.40145469e-01 9.19321060e-01 7.54566014e-01 3.97663742e-01 1.11198604e+00 2.78706878e-01 -9.26306486e-01 -3.30254808e-02 -8.57276380e-01 5.74304223e-01 8.31846893e-01 -1.48385489e+00 -9.54487920e-01 3.93938154e-01 6.96335316e-01 1.32911757e-01 -1.03924251e+00 7.20305368e-02 6.02641404e-01 -5.18847466e-01 1.04230666e+00 -8.05343926e-01 3.76558721e-01 -1.82145640e-01 -7.28036165e-01 -1.56370580e+00 -9.77590531e-02 1.01839736e-01 6.15246773e-01 1.18441594e+00 3.54375422e-01 -5.17045736e-01 1.98846996e-01 -1.76910479e-02 -2.58875579e-01 -3.57863456e-01 -8.78462374e-01 -9.78436053e-01 4.67967212e-01 -3.13696861e-01 2.24817872e-01 1.49504685e+00 1.77700520e-01 3.86254638e-01 -9.08724442e-02 2.26919010e-01 4.35807496e-01 7.22543374e-02 6.53933227e-01 -9.48820591e-01 -5.98842138e-03 -1.80171356e-01 -6.31263733e-01 -5.70946157e-01 9.38859284e-01 -1.31100571e+00 -9.47107840e-03 -1.55676377e+00 3.24494928e-01 -3.41729611e-01 -3.44106048e-01 5.18706560e-01 -1.58803791e-01 4.59827572e-01 1.94851846e-01 1.20520338e-01 -6.53139830e-01 6.24745905e-01 1.01625741e+00 -2.80572712e-01 4.59144145e-01 -1.24863029e+00 -5.50044119e-01 7.52475619e-01 5.93227804e-01 -6.91658199e-01 -3.90268862e-01 -9.61815059e-01 3.29609811e-02 -4.52794760e-01 2.42307112e-01 -7.18696713e-01 -3.54346484e-02 8.59546289e-02 8.79672170e-02 -8.20320621e-02 4.03950781e-01 -8.66489768e-01 -2.89890647e-01 2.95439303e-01 -2.71794856e-01 6.58822238e-01 1.88895926e-01 5.64436674e-01 -6.48984075e-01 -3.83297682e-01 6.52775109e-01 -2.49152318e-01 -9.02659833e-01 2.80466318e-01 -2.63911694e-01 5.05159914e-01 9.21780646e-01 1.38994157e-01 -1.96774527e-01 -8.44063014e-02 -4.96390671e-01 3.74481529e-02 9.60756004e-01 8.91643226e-01 3.79206896e-01 -1.79727447e+00 -6.98074400e-01 1.27970517e-01 9.31802630e-01 -6.34403050e-01 -6.87490106e-02 5.57115376e-01 -4.29173827e-01 5.73904574e-01 -3.29577088e-01 -4.37341154e-01 -1.19751108e+00 7.13456750e-01 1.82029486e-01 -3.18031371e-01 -3.44462335e-01 4.40301359e-01 4.04464900e-01 -8.59976828e-01 -1.63615748e-01 -1.04698829e-01 2.11700395e-01 2.83119768e-01 2.06548601e-01 -1.74923539e-01 -1.69313531e-02 -1.04000950e+00 -4.38823372e-01 8.51073146e-01 -9.30327028e-02 -5.40225208e-01 1.29818487e+00 -8.13268274e-02 -1.65019885e-01 7.11618364e-01 1.66082323e+00 -2.44766138e-02 -7.22784460e-01 -5.15830994e-01 1.57537907e-01 -7.33531833e-01 1.54480472e-01 -3.43545675e-01 -7.84914374e-01 1.13717020e+00 7.82705486e-01 -2.61475235e-01 6.72762036e-01 3.30948383e-01 9.07333553e-01 2.34008178e-01 5.04274666e-01 -1.31667471e+00 3.54458302e-01 5.11667073e-01 7.24696934e-01 -1.58113205e+00 -1.19996919e-04 3.20171639e-02 -5.76072693e-01 9.82015431e-01 4.48727012e-01 -2.80686557e-01 7.99263179e-01 -5.11355102e-02 3.11067402e-01 -1.84349388e-01 -6.20882392e-01 -5.17502606e-01 3.32709044e-01 7.14187086e-01 8.60666394e-01 2.49461532e-02 -4.57981437e-01 1.99553415e-01 1.10988073e-01 -2.28730336e-01 2.18220681e-01 1.15064335e+00 -1.90150574e-01 -1.56302571e+00 -2.57510483e-01 8.17649364e-02 -6.14076912e-01 -5.74604928e-01 -1.79778710e-01 1.00421798e+00 3.55978370e-01 7.91627169e-01 4.23155010e-01 4.01471443e-02 1.98598310e-01 2.92103440e-01 9.21931565e-01 -7.79188454e-01 -3.60616803e-01 -8.19398686e-02 2.04285532e-01 -4.35419083e-01 -7.55309045e-01 -8.26429486e-01 -9.29847538e-01 -1.02585360e-01 -1.55295774e-01 1.52460620e-01 9.15473282e-01 9.48432207e-01 2.61333197e-01 5.93694448e-02 3.78716052e-01 -8.11794996e-01 -3.02391589e-01 -9.10449147e-01 -5.04789293e-01 7.40246296e-01 1.61615372e-01 -6.24510467e-01 -2.53731191e-01 3.19967330e-01]
[11.19913387298584, 1.746645450592041]
801ad881-a319-447a-8409-ef27c2eac77d
embedding-structured-dictionary-entries
null
null
https://aclanthology.org/2020.insights-1.18
https://aclanthology.org/2020.insights-1.18.pdf
Embedding Structured Dictionary Entries
Previous work has shown how to effectively use external resources such as dictionaries to improve English-language word embeddings, either by manipulating the training process or by applying post-hoc adjustments to the embedding space. We experiment with a multi-task learning approach for explicitly incorporating the structured elements of dictionary entries, such as user-assigned tags and usage examples, when learning embeddings for dictionary headwords. Our work generalizes several existing models for learning word embeddings from dictionaries. However, we find that the most effective representations overall are learned by simply training with a skip-gram objective over the concatenated text of all entries in the dictionary, giving no particular focus to the structure of the entries.
['Gareth Tyson', 'Barbara McGillivray', 'Walid Magdy', 'Steven Wilson']
null
null
null
null
emnlp-insights-2020-11
['learning-word-embeddings']
['methodology']
[-9.82778799e-03 -7.62202293e-02 -4.59259242e-01 -3.96114558e-01 -5.56289017e-01 -7.59470403e-01 5.66378653e-01 6.20312572e-01 -1.16936815e+00 4.03635263e-01 8.70300174e-01 -6.64144754e-01 9.11798999e-02 -6.04294598e-01 -3.24908406e-01 -3.68112713e-01 -5.00060171e-02 4.96156782e-01 -2.41241366e-01 -4.86347497e-01 2.40302935e-01 2.08183482e-01 -1.06375372e+00 -7.82947242e-03 6.08598709e-01 4.75853026e-01 2.08517417e-01 4.47368532e-01 -3.32327873e-01 2.75229782e-01 -4.97604698e-01 -5.20657122e-01 1.65702254e-01 -1.97030619e-01 -4.59453851e-01 -1.49209902e-01 2.08636075e-01 -3.80539507e-01 -5.32546997e-01 7.20259368e-01 4.93064106e-01 3.14404964e-01 7.30940878e-01 -6.42620087e-01 -1.21398127e+00 8.76593351e-01 -1.70862570e-01 5.48373818e-01 2.02732220e-01 2.86265723e-02 1.62468112e+00 -1.20920575e+00 4.84801680e-01 1.11145437e+00 7.51605153e-01 3.79915267e-01 -1.39928770e+00 -5.66569090e-01 3.76451880e-01 9.44870338e-02 -1.51002014e+00 -4.84486669e-01 6.86693311e-01 -5.64503670e-01 1.60823190e+00 -1.38871118e-01 3.57231647e-01 1.16574144e+00 -2.27059387e-02 4.48508382e-01 6.27381027e-01 -8.08703125e-01 -8.96229446e-02 3.79890800e-01 4.42829341e-01 6.13187790e-01 5.10854483e-01 8.71128589e-02 -1.56624392e-01 -4.47370738e-01 5.71020782e-01 -6.26050979e-02 1.29896896e-02 -4.93649334e-01 -1.09294748e+00 1.14377749e+00 1.93229035e-01 4.84941810e-01 -3.50739092e-01 2.04810560e-01 7.15670407e-01 3.40770513e-01 5.21516800e-01 8.06167603e-01 -9.27509487e-01 -1.48953229e-01 -4.84092206e-01 1.40425310e-01 6.57840014e-01 7.53631413e-01 9.68921304e-01 1.56894103e-01 1.00352339e-01 1.05708551e+00 3.33572894e-01 2.90731281e-01 1.00174105e+00 -2.89721251e-01 2.80512035e-01 2.81768143e-01 3.15702222e-02 -7.98028767e-01 -1.98076263e-01 -2.37870216e-01 3.27148736e-02 -3.65277767e-01 1.14913076e-01 -5.29587269e-01 -9.39032555e-01 1.85462773e+00 1.27866581e-01 -7.26039782e-02 2.92411763e-02 4.88467246e-01 6.18818283e-01 5.51472068e-01 5.80660999e-01 1.19489364e-01 1.35710442e+00 -8.40345919e-01 -7.20526814e-01 -6.63795173e-01 1.14134932e+00 -7.52692997e-01 1.44606650e+00 1.29474759e-01 -5.87096870e-01 -4.06421244e-01 -1.09043372e+00 -3.66074175e-01 -7.67967939e-01 1.37371704e-01 6.43508554e-01 6.58394039e-01 -6.30609870e-01 2.29491711e-01 -6.73240244e-01 -3.15208167e-01 1.03869755e-02 2.61592925e-01 -4.08854574e-01 -1.30970776e-01 -1.25686920e+00 1.35358179e+00 5.76255620e-01 -4.46181178e-01 -7.51184165e-01 -7.21647024e-01 -1.34321046e+00 4.15533818e-02 1.84731975e-01 -3.19401205e-01 1.08513868e+00 -6.94062948e-01 -9.36487019e-01 8.48602712e-01 -1.36272207e-01 -1.07305944e-01 -4.26683128e-01 -2.85992563e-01 -4.42624837e-01 -4.62257326e-01 -2.27351710e-02 4.10799444e-01 5.70378304e-01 -1.12499619e+00 -4.61290330e-01 -1.56271830e-01 3.65798235e-01 2.31832340e-01 -1.10068989e+00 2.78304428e-01 -3.72438431e-01 -1.20075655e+00 -4.51382399e-01 -7.88923025e-01 -5.35307884e-01 -4.96865101e-02 1.68661505e-01 -4.78502095e-01 4.35644507e-01 -8.01296234e-01 1.72385001e+00 -2.47373080e+00 2.65728146e-01 2.25702748e-01 -8.28636345e-03 2.46142447e-01 -6.03556812e-01 6.54394746e-01 -5.17088501e-03 1.76981270e-01 1.36646777e-01 -6.86575472e-01 3.12272906e-01 8.04632246e-01 -4.44694728e-01 2.98955500e-01 1.45789072e-01 7.71410763e-01 -1.01543260e+00 -3.06782395e-01 2.13435039e-01 5.48748672e-01 -8.95474613e-01 1.12104245e-01 -1.65288776e-01 -1.58175468e-01 -1.63128719e-01 2.21819788e-01 6.11183271e-02 6.06129281e-02 5.38238585e-01 -1.30094260e-01 1.15035094e-01 1.09684932e+00 -1.18286002e+00 1.63162351e+00 -1.03125751e+00 4.88655597e-01 -1.16732329e-01 -9.33510661e-01 6.91744149e-01 3.29300106e-01 3.54211003e-01 -5.38563848e-01 8.56958479e-02 1.82654083e-01 2.04434946e-01 -3.72183591e-01 7.43774772e-01 -5.52303255e-01 -2.91708529e-01 6.44246697e-01 5.34640431e-01 -9.16700736e-02 1.80358782e-01 6.92638904e-02 9.03204381e-01 -8.28556567e-02 6.15155339e-01 -2.94719905e-01 1.42103881e-01 -4.30680849e-02 3.38575363e-01 5.32184362e-01 2.05225617e-01 1.80189133e-01 7.39816874e-02 -3.43182296e-01 -1.28425205e+00 -8.52704704e-01 -4.41393614e-01 1.76742077e+00 -3.38143110e-01 -1.15123069e+00 -2.21425712e-01 -8.42622519e-01 3.16459447e-01 9.45819020e-01 -5.97702146e-01 -3.48754704e-01 -7.34079182e-01 -7.43904054e-01 5.19728243e-01 8.16229105e-01 -5.64108491e-01 -9.40227330e-01 -1.21332198e-01 5.36008894e-01 2.40393311e-01 -1.09162903e+00 -8.08661819e-01 7.76717186e-01 -7.30378687e-01 -6.65005445e-01 -2.50252903e-01 -8.93312693e-01 6.31897807e-01 9.52097028e-02 1.23049903e+00 3.47295761e-01 -1.62364990e-01 4.02218789e-01 -5.55386662e-01 -4.02669638e-01 -1.32318333e-01 3.37811649e-01 2.23938018e-01 -4.50857729e-01 8.85666549e-01 -5.12227952e-01 -1.06396474e-01 3.40148360e-02 -9.52130616e-01 -6.10730112e-01 4.69603390e-01 9.63925004e-01 3.98085475e-01 -1.95737571e-01 4.19078618e-01 -1.02969098e+00 9.54367518e-01 -6.56185091e-01 -1.82909697e-01 -9.56230890e-03 -7.51403034e-01 3.54572713e-01 6.81535125e-01 -8.32773566e-01 -4.41812754e-01 -2.68043011e-01 -4.88695025e-01 -2.68511206e-01 5.34212366e-02 9.47863638e-01 -1.58126336e-02 7.37458393e-02 7.91939497e-01 5.71978651e-03 -2.57306665e-01 -7.32102215e-01 7.76994824e-01 6.85365975e-01 2.00718656e-01 -8.97488654e-01 7.26496279e-01 -8.71794075e-02 -6.86947227e-01 -8.37463319e-01 -8.43777001e-01 -6.64316595e-01 -7.48164296e-01 5.03377378e-01 7.20941305e-01 -1.09086788e+00 1.52351484e-01 -1.84922770e-01 -8.91485631e-01 -4.03677851e-01 -5.64987898e-01 7.07445681e-01 -3.83073837e-02 4.37711865e-01 -5.16016901e-01 -3.51102203e-01 -1.92244947e-02 -1.00237656e+00 8.88854742e-01 -3.72077674e-01 -6.54365361e-01 -1.65167046e+00 3.67742330e-01 -1.15125723e-01 6.62353456e-01 -3.18120509e-01 1.60561228e+00 -1.14722955e+00 5.95379584e-02 -2.64061570e-01 1.67791754e-01 6.44056082e-01 4.84959066e-01 -2.47705057e-01 -7.89391518e-01 -4.48782772e-01 -1.81036070e-01 -3.31649780e-01 7.52002180e-01 -1.79757755e-02 1.07800376e+00 -4.98611659e-01 -4.02955353e-01 7.04044044e-01 1.60090899e+00 5.67251910e-03 2.32000872e-01 4.53626990e-01 8.02890062e-01 3.10256690e-01 2.87456423e-01 5.12581050e-01 5.98127842e-01 7.24259555e-01 -8.00020993e-02 -1.15065098e-01 -6.36764616e-02 -5.73169112e-01 4.41057175e-01 1.11251080e+00 4.60904837e-01 -2.97484279e-01 -9.42417800e-01 9.57147300e-01 -1.43354368e+00 -4.25028265e-01 3.10142517e-01 2.03322601e+00 1.20559835e+00 1.92323029e-01 -5.01075871e-02 -6.80877129e-04 2.02338725e-01 4.78290856e-01 -2.75020570e-01 -5.78471005e-01 8.01217258e-02 9.19766843e-01 7.19575763e-01 8.72621834e-01 -1.09040582e+00 1.08030701e+00 7.85103703e+00 6.18777633e-01 -1.10482097e+00 3.45542699e-01 -2.46072024e-01 -1.64963901e-01 -8.52601171e-01 2.33523071e-01 -9.33688998e-01 2.94989288e-01 9.27207887e-01 -2.01628283e-01 4.93736655e-01 8.19503367e-01 -9.65775251e-02 3.71034384e-01 -1.22232199e+00 7.61739254e-01 4.24862981e-01 -9.44796622e-01 4.76927042e-01 1.77104756e-01 5.91839969e-01 1.81396484e-01 -5.03755286e-02 6.43203616e-01 7.26717234e-01 -1.02458048e+00 5.83567321e-01 -7.58686811e-02 8.01798403e-01 -5.09397089e-01 5.87524176e-01 1.93940297e-01 -1.00163090e+00 -2.09984049e-01 -5.30583501e-01 -4.46241111e-01 1.88982025e-01 3.53067547e-01 -7.53262222e-01 7.23330230e-02 1.24024384e-01 8.26398849e-01 -6.25852346e-01 6.49335325e-01 -3.97566795e-01 7.25358725e-01 -3.03499192e-01 -1.08182266e-01 4.39031154e-01 -1.44728974e-01 3.43764246e-01 1.67995524e+00 7.85504803e-02 -6.45820275e-02 5.11363804e-01 3.13015282e-01 -7.70145580e-02 3.92923862e-01 -8.29534113e-01 -4.17857915e-01 5.73656797e-01 1.06019866e+00 -1.65280849e-01 -4.02644604e-01 -8.73007596e-01 5.94104469e-01 5.51120579e-01 2.81176239e-01 -5.01671135e-01 -5.63537478e-01 1.23474884e+00 1.38457745e-01 7.47715771e-01 -8.96056831e-01 -3.84147584e-01 -1.24225366e+00 -1.75959229e-01 -9.94362056e-01 4.63528812e-01 -5.43312371e-01 -1.46149218e+00 3.82486254e-01 3.84615138e-02 -5.69510818e-01 -2.47076049e-01 -9.64587569e-01 -3.84272426e-01 1.04980898e+00 -1.60619414e+00 -1.03694499e+00 3.77414674e-01 4.38748538e-01 4.68551069e-01 -2.38335237e-01 1.29627919e+00 6.94449544e-01 -3.90968770e-01 8.36507082e-01 1.90599486e-01 4.50253755e-01 8.56342614e-01 -1.42504323e+00 4.07198489e-01 7.60128379e-01 6.90494895e-01 1.09132373e+00 8.06760013e-01 -3.30735326e-01 -1.29239666e+00 -7.23393321e-01 1.29465055e+00 -8.29568088e-01 1.08056259e+00 -6.96488082e-01 -1.01908243e+00 1.16555214e+00 2.93965101e-01 -1.48812244e-02 1.26410830e+00 7.93224335e-01 -7.16148436e-01 1.56373963e-01 -6.25895917e-01 4.97326314e-01 9.24147308e-01 -1.05491495e+00 -1.20671117e+00 2.47497648e-01 8.49928677e-01 -1.87452689e-01 -8.37987542e-01 1.85310811e-01 4.01359886e-01 2.13796332e-01 9.97731328e-01 -1.23842907e+00 1.91890836e-01 -1.50838032e-01 -4.96616274e-01 -1.66130638e+00 -5.11500597e-01 -1.39642313e-01 -1.54515713e-01 1.02158666e+00 6.58829153e-01 -4.06707376e-01 3.63295734e-01 4.29679871e-01 -1.50963575e-01 -6.85074210e-01 -6.59542978e-01 -6.00327551e-01 4.80618328e-01 -3.52999091e-01 4.17875290e-01 1.12646699e+00 1.91308349e-01 6.30352318e-01 -2.28742868e-01 1.48156108e-02 6.15331121e-02 -4.45145488e-01 6.96281016e-01 -1.10413194e+00 -4.00711834e-01 -3.32965165e-01 -2.44677350e-01 -1.30525863e+00 4.56054240e-01 -1.22269034e+00 4.63812463e-02 -1.42145872e+00 -4.64909598e-02 -8.02367747e-01 -6.18386686e-01 7.66039670e-01 -4.34072345e-01 5.64239398e-02 4.20277044e-02 2.86340583e-02 -2.99327105e-01 4.35446590e-01 8.28912675e-01 -3.00284363e-02 -6.67078560e-03 -5.75574636e-01 -1.12916851e+00 5.12338877e-01 5.35282135e-01 -7.40966916e-01 -4.87528622e-01 -1.05449688e+00 4.10477817e-01 -6.13281250e-01 -6.18419386e-02 -3.51408005e-01 1.06098786e-01 -1.37156904e-01 7.28798434e-02 7.56698027e-02 2.90916830e-01 -7.48677075e-01 -2.63128459e-01 1.32743776e-01 -5.78327715e-01 5.50329864e-01 4.05706167e-01 4.03868735e-01 -2.40687296e-01 -6.72277451e-01 6.68445230e-01 -2.20960274e-01 -8.09665322e-01 1.37422577e-01 -5.91129065e-01 3.04152042e-01 6.36674285e-01 -6.73895106e-02 8.74772593e-02 -8.27616360e-03 -7.59856939e-01 9.00496766e-02 5.35511434e-01 7.03101575e-01 3.36762577e-01 -1.40096664e+00 -4.14272517e-01 4.84280556e-01 4.38872218e-01 -4.95131105e-01 -4.93625611e-01 1.51279271e-01 2.50324216e-02 3.94702524e-01 -6.81630000e-02 -2.53582355e-02 -9.38057363e-01 7.43000209e-01 2.12454543e-01 -3.09363037e-01 -5.15883863e-01 8.39180946e-01 -1.23411175e-02 -7.02483892e-01 5.10975182e-01 -5.60449839e-01 -2.48181194e-01 3.88507217e-01 2.59058863e-01 -2.88044214e-01 2.23513767e-01 -5.31000972e-01 -3.26608837e-01 4.46806371e-01 -3.56974632e-01 -1.75500244e-01 1.52154565e+00 -1.15433715e-01 8.60313922e-02 5.86190999e-01 1.30322063e+00 3.62511784e-01 -6.61000311e-01 -7.34720469e-01 1.59199357e-01 -4.89328772e-01 2.81112611e-01 -6.94888830e-01 -6.18146122e-01 8.07297826e-01 1.68834358e-01 -1.83594108e-01 6.38161957e-01 2.15453655e-02 8.64731014e-01 5.77605069e-01 3.83715034e-01 -1.00379169e+00 1.44106433e-01 8.66330147e-01 4.31926340e-01 -9.76241708e-01 -3.71573009e-02 1.41545683e-01 -5.50083458e-01 9.81747687e-01 4.67986852e-01 -3.35914552e-01 1.09823394e+00 3.83441269e-01 1.51133120e-01 -2.06586450e-01 -8.25055778e-01 -2.42639720e-01 2.35489026e-01 5.76507986e-01 8.17549109e-01 -1.34208398e-02 -5.41943789e-01 6.12916291e-01 -1.81699157e-01 -4.16282743e-01 3.06319475e-01 1.02265763e+00 -4.25609857e-01 -1.67914915e+00 5.93748316e-02 7.07158983e-01 -5.62467754e-01 -6.79065108e-01 -2.39195049e-01 6.39344513e-01 2.30152652e-01 5.39537489e-01 1.75787702e-01 -2.74943113e-01 4.61463749e-01 4.41516161e-01 3.93499970e-01 -1.38022971e+00 -6.94169104e-01 -1.70110807e-01 2.79295534e-01 -5.35857305e-02 -1.28393903e-01 -7.02262461e-01 -8.86659443e-01 -5.47976941e-02 -5.08151293e-01 3.23284835e-01 6.32628858e-01 1.12329304e+00 1.43606529e-01 4.01004910e-01 4.16454107e-01 -6.17201984e-01 -8.13132524e-01 -1.07008648e+00 -3.84770721e-01 5.20522058e-01 4.47114080e-01 -8.03104401e-01 -4.36178386e-01 -1.38001651e-01]
[10.523515701293945, 8.818354606628418]
5770f2e0-5e42-4fc6-bb4b-1baed0f9d496
towards-high-performance-exploratory-data
2306.04425
null
https://arxiv.org/abs/2306.04425v1
https://arxiv.org/pdf/2306.04425v1.pdf
Towards High-Performance Exploratory Data Analysis (EDA) Via Stable Equilibrium Point
Exploratory data analysis (EDA) is a vital procedure for data science projects. In this work, we introduce a stable equilibrium point (SEP) - based framework for improving the efficiency and solution quality of EDA. By exploiting the SEPs to be the representative points, our approach aims to generate high-quality clustering and data visualization for large-scale data sets. A very unique property of the proposed method is that the SEPs will directly encode the clustering properties of data sets. Compared with prior state-of-the-art clustering and data visualization methods, the proposed methods allow substantially improving computing efficiency and solution quality for large-scale data analysis tasks.
['Yongyu Wang', 'Yuxuan Song']
2023-06-07
null
null
null
null
['data-visualization', 'data-visualization']
['methodology', 'miscellaneous']
[-4.89482403e-01 -3.19532365e-01 2.73748636e-01 -1.44521952e-01 -2.96699136e-01 -5.12883604e-01 4.58373874e-01 4.24150556e-01 3.30661535e-02 2.70645201e-01 1.62795648e-01 -2.58514643e-01 -7.05275714e-01 -9.34985340e-01 -1.57247066e-01 -9.42956328e-01 -3.18154007e-01 3.80170614e-01 1.37256026e-01 1.39888942e-01 6.35240614e-01 7.18372703e-01 -2.03894591e+00 4.79245126e-01 1.31783640e+00 7.89815485e-01 2.90898323e-01 2.83331573e-01 -4.90048856e-01 1.52486324e-01 -6.08247995e-01 2.90630728e-01 1.78462029e-01 -3.26610416e-01 -7.78078377e-01 -5.84423821e-03 -2.79185742e-01 5.34732461e-01 2.38816634e-01 1.06603932e+00 5.48296332e-01 1.80368558e-01 6.77163243e-01 -1.65222359e+00 -5.47999680e-01 1.33320615e-01 -9.65367377e-01 -1.25968337e-01 2.88193315e-01 7.04682544e-02 7.27661073e-01 -1.09202898e+00 7.79745996e-01 1.29468226e+00 5.14033675e-01 1.13878371e-02 -1.41103089e+00 -5.61421335e-01 -3.22933029e-03 1.35622948e-01 -1.63131762e+00 -2.16412604e-01 1.08038354e+00 -7.12382376e-01 2.91968942e-01 6.09444678e-01 9.32119489e-01 1.82043254e-01 -9.12136808e-02 6.72404587e-01 1.12651992e+00 -2.91095704e-01 7.32860684e-01 1.03470488e-02 5.22987783e-01 2.91745067e-01 4.83691156e-01 -2.99717516e-01 -3.22036654e-01 -2.80032158e-01 5.79930902e-01 2.63996542e-01 -8.19153786e-02 -9.31643367e-01 -1.28214025e+00 6.78043842e-01 6.10624611e-01 4.13981974e-01 -5.69863498e-01 -3.28664154e-01 2.78285563e-01 -2.06540655e-02 2.74561018e-01 6.21879041e-01 2.55121231e-01 -2.80774623e-01 -1.01798534e+00 2.67363757e-01 2.64270157e-01 8.11221361e-01 8.23528409e-01 -1.90894037e-01 -3.18718702e-01 6.52190685e-01 3.85720730e-01 1.72897398e-01 3.76298755e-01 -9.19814527e-01 1.33570895e-01 1.51710701e+00 1.52615085e-01 -1.43647957e+00 -7.81824946e-01 -2.86561966e-01 -1.06770241e+00 5.60772479e-01 3.22232783e-01 7.15698674e-02 -6.30468726e-01 1.34476674e+00 7.97311664e-01 -2.12035850e-01 1.19817771e-01 6.54475570e-01 7.14854777e-01 6.89090550e-01 -6.76036477e-02 -5.79615831e-01 1.07612669e+00 -4.80965018e-01 -1.04712737e+00 4.60566103e-01 7.30359674e-01 -4.95806068e-01 1.42938089e+00 5.06988704e-01 -9.71049607e-01 -6.40842974e-01 -8.08823526e-01 3.14499512e-02 -4.52672571e-01 1.29433200e-01 4.80300367e-01 4.80275571e-01 -9.98816609e-01 3.86013985e-01 -9.29847836e-01 -4.05862391e-01 6.28740609e-01 1.09825619e-01 -3.60633880e-02 1.50997266e-01 -5.05647361e-01 3.54533732e-01 5.42170823e-01 1.91436857e-01 -3.23328793e-01 -9.85289514e-01 -4.18285459e-01 1.62440047e-01 1.40665606e-01 -4.78243947e-01 4.99691963e-01 -3.45106244e-01 -8.78280342e-01 6.11360550e-01 -4.34803247e-01 5.00443168e-02 4.37898487e-01 -5.09987101e-02 -2.59512663e-01 1.00056320e-01 -7.65673742e-02 4.09743041e-01 2.93387860e-01 -1.89160919e+00 -4.98401791e-01 -6.08251095e-01 -7.10615575e-01 1.61273867e-01 -7.78725863e-01 4.93128672e-02 -5.05588412e-01 -4.48277444e-01 3.74441236e-01 -4.91514981e-01 -4.90739971e-01 1.76305473e-01 -4.29855049e-01 -4.28773701e-01 1.45744777e+00 -4.42139596e-01 1.90665066e+00 -2.43878222e+00 2.51191676e-01 7.31712282e-01 5.81403255e-01 2.21052587e-01 2.85321385e-01 7.89861917e-01 -8.12704191e-02 3.00651699e-01 -3.31947774e-01 -2.28204697e-01 1.04003921e-01 -4.25331928e-02 -8.36768448e-02 3.16430271e-01 -1.27951307e-02 7.14816630e-01 -8.44061196e-01 -7.63366282e-01 3.87862921e-01 4.78200883e-01 -4.68663067e-01 3.20612431e-01 2.03962941e-02 5.39759934e-01 -2.72770882e-01 4.41913992e-01 9.55229700e-01 -3.72570425e-01 3.41012269e-01 -1.73788682e-01 -6.98231995e-01 -2.61513442e-01 -1.74280286e+00 1.71597040e+00 4.45609018e-02 4.99447286e-01 2.43445456e-01 -8.55390549e-01 1.41500676e+00 -3.16987820e-02 9.44712579e-01 -5.64206362e-01 -8.65465477e-02 -1.75546277e-02 2.82743536e-02 -5.02080560e-01 5.33291578e-01 3.33323538e-01 1.47763371e-01 7.31356621e-01 -5.84100962e-01 1.78687304e-01 6.13246500e-01 4.41487640e-01 8.13388884e-01 -2.12403953e-01 3.98947746e-01 -9.33355987e-01 3.54503602e-01 1.24936640e-01 6.05682194e-01 3.34401220e-01 -4.13547223e-03 5.48284173e-01 7.90620863e-01 -3.05786252e-01 -1.04325247e+00 -9.74533081e-01 -2.05805346e-01 6.32834077e-01 2.43240178e-01 -8.55331361e-01 -8.49322200e-01 -2.43359610e-01 4.45215497e-03 7.26273060e-01 -8.32750857e-01 -6.01719581e-02 -3.33917141e-01 -6.97348833e-01 -2.95835286e-02 4.68496829e-01 3.78879339e-01 -1.04046512e+00 -7.46748149e-01 -8.46193060e-02 -8.12176540e-02 -2.71094024e-01 -8.91657397e-02 -2.58669168e-01 -1.05346501e+00 -1.31435454e+00 -3.83982211e-01 -7.14089394e-01 1.00307059e+00 5.32226384e-01 9.30371523e-01 1.93646699e-01 -5.40912926e-01 -1.16601944e-01 -4.05786097e-01 -3.27542037e-01 -2.14051366e-01 -2.55889863e-01 1.27207339e-02 1.32820368e-01 4.22192574e-01 -6.42532766e-01 -7.33083665e-01 5.73034406e-01 -1.04048562e+00 4.20298070e-01 1.97867975e-01 4.69063073e-01 1.05623400e+00 7.15311944e-01 5.06232679e-01 -8.16932142e-01 1.19207692e+00 -5.93755305e-01 -6.55626118e-01 4.72889662e-01 -9.20308650e-01 6.41964003e-02 8.76067996e-01 -2.52280325e-01 -9.51350868e-01 1.50620073e-01 3.32517177e-01 -4.77786005e-01 -3.20397019e-01 5.65919220e-01 -3.47236484e-01 3.76839042e-01 5.95966876e-01 8.45953971e-02 1.52637914e-01 -8.41899991e-01 4.97880697e-01 6.99685693e-01 5.22773504e-01 -6.08643770e-01 5.72263360e-01 5.66679299e-01 1.97079048e-01 -6.83849037e-01 -9.47379097e-02 -6.60005629e-01 -1.10395455e+00 -3.46963018e-01 6.23975098e-01 -4.79071319e-01 -1.13751721e+00 3.82234901e-01 -6.90971553e-01 1.15862615e-01 -3.59078944e-01 6.66810526e-03 -2.52279252e-01 3.87264103e-01 1.12113766e-01 -1.06508505e+00 -2.30805486e-01 -1.08046806e+00 7.52360165e-01 1.99149773e-01 -2.51521349e-01 -1.02553511e+00 2.73062497e-01 -7.89696202e-02 3.02493393e-01 8.69706333e-01 9.14263070e-01 -2.67022103e-01 -2.53185809e-01 1.54672116e-02 -1.66613936e-01 -2.57303268e-01 3.27172726e-01 7.59330571e-01 -6.31780803e-01 -4.12969261e-01 -9.59568918e-02 1.42787620e-01 3.94031465e-01 3.86623591e-01 1.64958584e+00 -2.48273432e-01 -6.08843148e-01 6.02240264e-01 1.24185753e+00 4.81019378e-01 5.59240818e-01 1.88811332e-01 6.93722665e-01 9.39643383e-01 8.80788863e-01 8.62523675e-01 3.13258350e-01 6.18456602e-01 3.23762685e-01 -4.99860138e-01 -1.42530706e-02 -2.87810802e-01 -3.05042237e-01 7.13924110e-01 3.23044248e-02 7.75972754e-02 -1.51590788e+00 6.90046847e-01 -2.26435280e+00 -8.70125890e-01 -7.88180828e-01 2.14714479e+00 4.22849149e-01 -2.13073418e-01 5.41468084e-01 6.57210052e-01 7.76169300e-01 -2.33016625e-01 -5.69684148e-01 -2.96696965e-02 -1.22453766e-02 -2.15672612e-01 -3.55049640e-01 8.99422616e-02 -8.52631748e-01 3.19196850e-01 6.70398045e+00 7.37541020e-01 -7.52557158e-01 -3.51813138e-01 4.08793390e-01 -1.57102630e-01 -4.09341097e-01 -1.20542772e-01 -2.40189999e-01 7.22081006e-01 6.89469159e-01 -4.87184465e-01 2.07089156e-01 9.45976198e-01 7.61853874e-01 -3.82948071e-02 -1.14200974e+00 1.08780885e+00 -5.98062217e-01 -1.54189134e+00 2.29585785e-02 3.89959544e-01 7.12575018e-01 -6.49905324e-01 2.25661367e-01 -4.27643448e-01 5.56515694e-01 -9.81851220e-01 4.78658915e-01 6.88271523e-01 7.04077780e-01 -1.17985916e+00 4.36347216e-01 3.48930717e-01 -1.31139719e+00 -2.90879905e-01 -2.28871390e-01 -3.99140827e-02 7.17858523e-02 9.33982790e-01 -5.37437439e-01 9.05425310e-01 1.18217456e+00 9.77141857e-01 -8.21121335e-01 1.18591762e+00 1.67369649e-01 4.98835385e-01 -3.18493783e-01 -9.36582908e-02 -1.00535490e-01 -6.91733241e-01 2.94745386e-01 1.13199782e+00 4.26329494e-01 1.15239337e-01 -1.59975253e-02 1.24255955e+00 1.08611375e-01 2.61630535e-01 -7.13830113e-01 -1.89518943e-01 9.38439965e-01 1.35338712e+00 -9.85506415e-01 -1.47003725e-01 5.32538909e-03 3.57137680e-01 3.24929267e-01 3.19865912e-01 -2.70985991e-01 -6.18732512e-01 6.83474541e-01 2.62099922e-01 1.07479356e-01 -5.25044739e-01 -1.07985091e+00 -7.45650530e-01 4.76786084e-02 -7.14772463e-01 7.27276027e-01 -7.56557584e-01 -1.22161651e+00 3.25940192e-01 -8.77170917e-03 -1.65579236e+00 3.74554284e-02 -3.23740631e-01 -9.91071641e-01 6.98416233e-01 -7.99610853e-01 -9.54986334e-01 -7.35048890e-01 8.14792275e-01 2.50417124e-02 -1.09562734e-02 6.86008632e-01 5.33508100e-02 -8.16889405e-01 1.32917970e-01 8.44853818e-01 -1.62160769e-01 3.67528349e-01 -1.46783507e+00 2.48563349e-01 1.12470126e+00 -5.72900809e-02 1.17035580e+00 7.32717872e-01 -5.96595228e-01 -1.39866328e+00 -7.80866325e-01 3.07783246e-01 -3.53463948e-01 4.85368758e-01 -6.85776532e-01 -1.12688768e+00 1.70420296e-02 1.21298239e-01 -1.45673826e-01 9.13429797e-01 3.88955355e-01 1.04042301e-02 -2.39018738e-01 -1.27297521e+00 5.26327491e-01 9.15888488e-01 -3.58326547e-02 -4.41293985e-01 7.61412084e-02 4.04232174e-01 1.08402595e-01 -1.15475225e+00 1.65787444e-01 2.69052446e-01 -1.16342318e+00 1.00086749e+00 -4.63422954e-01 3.94472897e-01 -7.98502326e-01 2.38880619e-01 -1.33291507e+00 -5.35865128e-01 -6.49865985e-01 -2.59538800e-01 1.53086042e+00 -2.92820600e-03 -4.35915321e-01 6.14374578e-01 7.88535058e-01 -1.63091063e-01 -5.43508172e-01 -7.45545983e-01 -6.45409346e-01 1.19578587e-02 -1.73458889e-01 1.00289261e+00 1.06550562e+00 3.94097298e-01 -1.85749792e-02 -7.19272345e-02 1.25393257e-01 8.41201425e-01 6.52261376e-01 9.71659362e-01 -1.77992952e+00 2.88677186e-01 -6.91677451e-01 -1.88808382e-01 -5.93159616e-01 -5.09660900e-01 -6.94918275e-01 -3.03784013e-01 -1.98305428e+00 1.60984218e-01 -7.21476972e-01 -3.81917417e-01 4.10031140e-01 -2.64467269e-01 -8.58540535e-02 2.35615581e-01 7.08069324e-01 -7.26119459e-01 5.81304073e-01 1.16213894e+00 4.45393711e-01 -6.58310175e-01 -9.64799002e-02 -8.73973072e-01 3.38986516e-01 6.00846708e-01 -3.75444829e-01 -5.82094371e-01 -1.28681129e-02 5.48752882e-02 -3.01884532e-01 1.55924305e-01 -8.99675071e-01 4.25060928e-01 -3.96798670e-01 3.90454799e-01 -1.14169407e+00 -7.77781755e-02 -1.01230073e+00 7.50728488e-01 5.35012424e-01 -2.52188683e-01 1.67958558e-01 1.23205200e-01 4.72723871e-01 -3.46834451e-01 2.33524010e-01 7.51489639e-01 8.81452337e-02 -4.95606095e-01 -5.80544919e-02 -1.10460781e-01 -1.69968620e-01 1.50190616e+00 -4.42391396e-01 -4.50537264e-01 -4.41857846e-03 -6.32099152e-01 6.63596272e-01 8.13631535e-01 2.42490038e-01 6.28858209e-01 -1.51139712e+00 -4.74623352e-01 4.06354696e-01 1.51494697e-01 4.96381134e-01 3.02194804e-01 7.46673703e-01 -5.33405125e-01 1.49740979e-01 -5.41252673e-01 -8.37724566e-01 -1.40326464e+00 9.36677277e-01 -2.25301489e-01 -4.45647202e-02 -7.80188978e-01 3.36647421e-01 1.48423940e-01 -4.10374343e-01 2.19198883e-01 -5.18330485e-02 -2.89555192e-01 2.42516771e-01 7.16871560e-01 9.81386483e-01 -7.57495910e-02 -3.18328053e-01 -5.47777057e-01 4.53330517e-01 2.86044925e-01 9.49706957e-02 1.64420402e+00 -2.72135228e-01 -3.73748660e-01 6.66924655e-01 8.21288943e-01 -6.07657917e-02 -1.26855206e+00 1.09395467e-01 1.93315402e-01 -8.70856583e-01 -1.49911553e-01 -7.60501146e-01 -9.98718739e-01 1.08722663e+00 5.85332692e-01 7.00394273e-01 1.41796696e+00 5.26455194e-02 4.98887338e-02 9.88560691e-02 -3.03999744e-02 -1.31364751e+00 1.45059600e-01 -2.90128708e-01 9.82442498e-01 -1.01829112e+00 6.41568154e-02 -4.60900038e-01 -8.30455720e-01 1.20341480e+00 6.17110014e-01 2.49663904e-01 6.45131052e-01 3.67064059e-01 4.19989415e-02 -6.14716411e-01 -5.33683121e-01 6.11363538e-02 3.69210392e-01 8.29405367e-01 3.83306354e-01 1.88173819e-02 -5.30279160e-01 5.50863981e-01 -2.25394368e-01 -1.06753118e-01 2.80863017e-01 1.06229401e+00 -5.05833566e-01 -8.75910401e-01 -7.10212290e-01 3.56942713e-01 2.33230412e-01 3.85169238e-01 -5.67111850e-01 7.70184875e-01 -1.34692997e-01 1.09786141e+00 2.36254230e-01 -3.79856169e-01 3.63692075e-01 2.06063371e-02 -3.55052054e-01 -1.89499199e-01 -5.11156440e-01 1.61026582e-01 -5.17560780e-01 -7.41516173e-01 -5.57271719e-01 -7.79669285e-01 -1.62298870e+00 -4.94619727e-01 -9.16367993e-02 5.26486516e-01 8.24939907e-01 1.97127238e-01 1.04396784e+00 5.07530153e-01 8.60817611e-01 -4.20536131e-01 1.09493926e-01 -7.61779428e-01 -6.30269706e-01 8.69107783e-01 -7.33713061e-02 -7.85877466e-01 -9.07394737e-02 1.56765923e-01]
[7.888458728790283, 4.569932460784912]
e77b1418-9659-4c32-a862-620cf46b251b
vision-transformer-with-attentive-pooling-for
2212.05463
null
https://arxiv.org/abs/2212.05463v1
https://arxiv.org/pdf/2212.05463v1.pdf
Vision Transformer with Attentive Pooling for Robust Facial Expression Recognition
Facial Expression Recognition (FER) in the wild is an extremely challenging task. Recently, some Vision Transformers (ViT) have been explored for FER, but most of them perform inferiorly compared to Convolutional Neural Networks (CNN). This is mainly because the new proposed modules are difficult to converge well from scratch due to lacking inductive bias and easy to focus on the occlusion and noisy areas. TransFER, a representative transformer-based method for FER, alleviates this with multi-branch attention dropping but brings excessive computations. On the contrary, we present two attentive pooling (AP) modules to pool noisy features directly. The AP modules include Attentive Patch Pooling (APP) and Attentive Token Pooling (ATP). They aim to guide the model to emphasize the most discriminative features while reducing the impacts of less relevant features. The proposed APP is employed to select the most informative patches on CNN features, and ATP discards unimportant tokens in ViT. Being simple to implement and without learnable parameters, the APP and ATP intuitively reduce the computational cost while boosting the performance by ONLY pursuing the most discriminative features. Qualitative results demonstrate the motivations and effectiveness of our attentive poolings. Besides, quantitative results on six in-the-wild datasets outperform other state-of-the-art methods.
['Guodong Guo', 'Zhongsong Ma', 'Zichang Tan', 'Qiangchang Wang', 'Fanglei Xue']
2022-12-11
null
null
null
null
['facial-expression-recognition']
['computer-vision']
[ 1.01347715e-01 1.71445478e-02 5.51915020e-02 -3.93055350e-01 -6.15110278e-01 -1.09955361e-02 2.86971748e-01 -3.43532562e-01 -5.24600029e-01 7.14131892e-01 2.40954444e-01 3.56215894e-01 2.69696526e-02 -7.45200217e-01 -6.53901815e-01 -9.70930636e-01 -5.22801019e-02 -2.96650261e-01 2.41735905e-01 -3.24621499e-01 -1.73121374e-02 5.07204652e-01 -1.73049986e+00 5.37082314e-01 8.29390943e-01 1.55934012e+00 2.63416618e-01 -4.97975163e-02 -1.29557446e-01 1.01550043e+00 -4.82487917e-01 -7.22816825e-01 3.96848191e-03 -2.17881173e-01 -5.87091029e-01 -1.17339335e-01 5.34107804e-01 -3.51167977e-01 -3.29216212e-01 1.05808628e+00 7.57711172e-01 -1.85748376e-02 3.20183486e-01 -1.43231058e+00 -6.02287233e-01 2.22552568e-01 -7.55828559e-01 7.78382719e-02 9.44792479e-02 2.72678226e-01 9.53956366e-01 -1.42211270e+00 4.10295665e-01 1.27409446e+00 7.72459507e-01 5.69711149e-01 -9.02853489e-01 -7.75445580e-01 3.40888947e-01 2.99179196e-01 -1.38190246e+00 -6.22871280e-01 9.04869199e-01 -7.25602806e-02 7.98522055e-01 2.76560158e-01 7.13312387e-01 1.16802680e+00 7.07641989e-02 1.03937578e+00 1.05040884e+00 -1.65708020e-01 1.33582085e-01 1.42707527e-01 4.73883264e-02 9.06894565e-01 -1.64194718e-01 -2.67497122e-01 -8.25264990e-01 -1.31609797e-01 5.20905435e-01 3.02152634e-01 -4.43672568e-01 -1.12563789e-01 -6.27625287e-01 6.82690740e-01 7.57618308e-01 2.79755116e-01 -7.31182754e-01 2.66088516e-01 3.91703337e-01 6.62923884e-03 5.74274421e-01 2.40178958e-01 -5.37727118e-01 -1.05854705e-01 -9.13079262e-01 1.99040666e-01 3.09471697e-01 6.62315845e-01 1.21993279e+00 1.48384452e-01 -6.89222991e-01 1.06730175e+00 8.55607837e-02 2.07631424e-01 2.46878251e-01 -8.01753521e-01 3.24380010e-01 7.79772282e-01 -1.04412511e-01 -1.07966530e+00 -2.13256359e-01 -4.50928122e-01 -9.23039079e-01 1.44580022e-01 1.93148538e-01 -3.19802910e-01 -1.03935730e+00 1.79512846e+00 -8.07889923e-02 1.00633852e-01 -2.99994886e-01 9.83314574e-01 1.15447402e+00 6.24289274e-01 5.08288205e-01 -1.16227604e-01 1.45107496e+00 -1.07770014e+00 -6.35059059e-01 -2.90896058e-01 3.91100377e-01 -7.83408523e-01 1.08930254e+00 2.92051405e-01 -1.04893172e+00 -5.95834255e-01 -7.11493611e-01 -1.15872510e-01 -3.75100553e-01 4.79578942e-01 7.94102788e-01 4.61572379e-01 -1.15203536e+00 5.20832598e-01 -5.74096560e-01 -1.65266037e-01 1.07652521e+00 4.97277379e-01 -4.48718280e-01 -1.41830087e-01 -1.18640018e+00 8.21833014e-01 -3.21642049e-02 6.18485093e-01 -8.57243061e-01 -8.39484453e-01 -9.45459127e-01 3.28921974e-01 2.41196126e-01 -4.59894866e-01 9.27874088e-01 -1.57236278e+00 -1.73624873e+00 6.24001741e-01 -4.22864735e-01 -1.79159701e-01 3.58099788e-01 -4.80969429e-01 5.21915369e-02 3.55797350e-01 -1.34976819e-01 1.15753746e+00 1.17216945e+00 -1.04925394e+00 -3.61028463e-01 -2.21654490e-01 1.54844910e-01 5.54226562e-02 -7.40139902e-01 2.73717552e-01 -4.46509272e-01 -5.55418193e-01 -9.70702246e-02 -5.22212148e-01 -1.33983389e-01 4.20016497e-01 -2.07695872e-01 -3.31156403e-01 1.16122520e+00 -5.85148335e-01 1.03697193e+00 -2.28201842e+00 -1.02909967e-01 -7.97444433e-02 2.04644695e-01 5.26660323e-01 -3.06979656e-01 1.00005977e-01 -3.05538322e-03 -2.13576946e-03 -1.25161111e-01 -4.45173383e-01 -1.26951680e-01 2.08238184e-01 -2.70563394e-01 2.92398185e-01 9.73278105e-01 1.04188740e+00 -8.05645049e-01 -6.25914693e-01 1.94347605e-01 7.51765490e-01 -7.46236622e-01 2.07127646e-01 -1.03951350e-01 2.38336846e-01 -5.99925160e-01 9.01507139e-01 9.95191157e-01 8.46406519e-02 -2.58408040e-01 -5.26616335e-01 -8.22861344e-02 1.03316098e-01 -8.27394485e-01 1.37476194e+00 -4.47205007e-01 6.02057576e-01 1.58404872e-01 -1.10431194e+00 1.08047652e+00 2.41102099e-01 3.97825718e-01 -8.50923896e-01 3.50490868e-01 7.01893270e-02 -1.69218734e-01 -8.19153190e-01 5.11699259e-01 -1.33433565e-01 1.75851285e-01 -1.22932702e-01 4.58409697e-01 2.08804935e-01 -5.63068986e-02 -1.22362584e-01 9.28937197e-01 3.29070985e-01 8.03709030e-02 -2.32983276e-01 5.43139160e-01 -4.98028815e-01 9.72326875e-01 3.95077169e-01 -4.59447473e-01 7.01152503e-01 6.90740108e-01 -4.69363987e-01 -5.23156166e-01 -9.02503371e-01 -1.10704638e-02 1.20847738e+00 7.91890323e-02 -3.62980157e-01 -7.73647785e-01 -7.80284286e-01 -2.25442022e-01 1.00584552e-01 -8.99059832e-01 -1.95550993e-01 -4.55985934e-01 -8.31683815e-01 5.94396532e-01 5.97722650e-01 1.09323704e+00 -1.31555295e+00 -7.84925997e-01 -1.06251100e-03 -3.90997589e-01 -9.24777806e-01 -1.93445802e-01 2.21185446e-01 -5.98774195e-01 -7.87232161e-01 -1.06702828e+00 -8.09666991e-01 8.39743078e-01 2.09785491e-01 9.35749829e-01 9.44927782e-02 -2.99556494e-01 -2.18962990e-02 -4.22023565e-01 -5.54915905e-01 4.26493794e-01 -2.02878285e-02 -3.54580820e-01 4.66875792e-01 5.44064283e-01 -5.08081615e-01 -7.36738145e-01 4.04823750e-01 -7.21645296e-01 -1.59756795e-01 7.53923893e-01 1.20288253e+00 5.52357495e-01 -2.40176097e-01 5.08952558e-01 -5.86428702e-01 3.76283884e-01 -3.36387843e-01 -1.78574473e-01 1.44715428e-01 2.77204551e-02 -2.09913567e-01 7.56839395e-01 -4.98033047e-01 -1.20922983e+00 -2.59629600e-02 -2.29816020e-01 -6.62677348e-01 -6.32057264e-02 1.36934578e-01 -4.31793451e-01 -3.15573901e-01 3.53763103e-01 1.26056269e-01 -7.17667788e-02 -3.48305196e-01 4.12002243e-02 4.02165532e-01 6.97828829e-02 -6.11457050e-01 3.37903410e-01 4.68195707e-01 -7.92955607e-02 -8.98386955e-01 -8.85434449e-01 -1.59708396e-01 -4.40134943e-01 -3.17542166e-01 6.15242183e-01 -9.56042707e-01 -7.80975759e-01 5.88433146e-01 -1.14568341e+00 -1.44374862e-01 -4.03739512e-01 1.40898228e-01 -2.93893188e-01 1.30046681e-01 -5.68131983e-01 -9.76422966e-01 -5.05925417e-01 -1.22389889e+00 1.27543545e+00 6.03723228e-01 -1.32770747e-01 -4.50856715e-01 -4.68889415e-01 8.36767182e-02 7.41276681e-01 2.53885269e-01 6.90746427e-01 -9.64391753e-02 -4.51477945e-01 1.63980052e-02 -6.60010934e-01 6.72715545e-01 -3.59615535e-02 1.69740468e-01 -1.64308012e+00 -1.01730615e-01 -1.83776602e-01 -6.06622219e-01 1.22794950e+00 4.67598408e-01 1.51016533e+00 -2.64138788e-01 -1.85450986e-01 6.42056704e-01 1.33761597e+00 1.14835009e-01 1.04592645e+00 2.54556566e-01 4.97765273e-01 6.90543056e-01 6.24787927e-01 4.70907003e-01 -1.77441966e-02 5.48220038e-01 5.64404905e-01 -6.14509761e-01 -1.19752176e-01 -1.68552920e-01 6.15375102e-01 3.27053845e-01 -3.08831364e-01 -1.24091585e-03 -4.04777974e-01 6.24946475e-01 -1.79739845e+00 -9.10890937e-01 1.32475421e-01 1.79151249e+00 7.23864257e-01 -6.44430891e-02 -1.86376981e-02 8.70553330e-02 5.03435135e-01 3.29565793e-01 -2.20373735e-01 -4.58292663e-01 -4.20512348e-01 8.62478316e-01 1.53103068e-01 3.12700272e-02 -1.10334229e+00 1.10804462e+00 5.47629499e+00 1.05653989e+00 -1.36216831e+00 1.02415271e-01 1.02089155e+00 -2.04514086e-01 -1.71431398e-03 -9.25833732e-03 -6.82372093e-01 4.42437857e-01 3.20554137e-01 3.48917156e-01 -4.15431373e-02 1.01471519e+00 2.36199990e-01 -3.57248157e-01 -8.34683537e-01 1.04275024e+00 6.25022128e-03 -1.11765969e+00 2.21402019e-01 -2.33702540e-01 5.45293808e-01 -2.04948515e-01 2.80354619e-01 4.67742682e-01 -2.59157538e-01 -1.07017052e+00 7.69269407e-01 8.24408054e-01 5.62172353e-01 -8.73207986e-01 1.14022958e+00 -2.35165879e-01 -1.00956190e+00 -1.55096263e-01 -7.30041981e-01 -2.31887743e-01 -8.09079111e-02 7.68496811e-01 -2.19354793e-01 4.44311947e-01 1.14842141e+00 6.50985956e-01 -6.63080812e-01 1.06375813e+00 -3.48840773e-01 6.39212430e-01 -3.20533186e-01 -1.36535347e-01 5.38757682e-01 -5.13991453e-02 3.30388427e-01 1.31373429e+00 2.88461059e-01 -2.44918745e-02 -4.39217873e-02 9.40848887e-01 -2.76526064e-01 1.65955961e-01 -4.45211858e-01 2.82167077e-01 6.57305680e-03 1.56098425e+00 -3.56391728e-01 -1.57231122e-01 -3.72633040e-01 9.66985106e-01 4.62558925e-01 4.63705212e-01 -7.87980855e-01 -5.62522352e-01 8.68774056e-01 1.53572559e-01 5.60022473e-01 2.06244916e-01 -1.86851472e-01 -8.52345765e-01 2.91375548e-01 -6.62360728e-01 2.71642834e-01 -8.09102476e-01 -1.24320471e+00 8.34756434e-01 -2.32172355e-01 -1.18775642e+00 2.81943142e-01 -6.00296378e-01 -7.64002085e-01 9.13461685e-01 -2.03147745e+00 -1.29085243e+00 -6.05402589e-01 7.57672846e-01 4.50065702e-01 5.95140643e-02 7.69720316e-01 5.50781012e-01 -8.65339398e-01 9.05668139e-01 -3.47956449e-01 2.93458581e-01 7.80564010e-01 -8.03908348e-01 -3.13634604e-01 8.19598794e-01 -2.98666328e-01 6.27409339e-01 3.80495131e-01 -1.43472135e-01 -1.12289512e+00 -1.22365248e+00 8.13023686e-01 8.67399946e-02 3.54458749e-01 -4.33520526e-01 -9.09774780e-01 2.38789931e-01 2.26360261e-01 3.99747759e-01 4.14284497e-01 -4.91473712e-02 -3.74632001e-01 -5.02498865e-01 -1.21023834e+00 5.90350032e-01 9.80958462e-01 -3.13117057e-01 -2.95879036e-01 -7.77696371e-02 3.40788066e-01 2.97600441e-02 -4.71959293e-01 4.38021600e-01 5.81512690e-01 -1.25050700e+00 7.42945075e-01 -4.05419379e-01 7.58222163e-01 -1.86473802e-01 -2.25414447e-02 -1.15415657e+00 -3.21835339e-01 -8.00777078e-01 1.82101861e-01 1.51516044e+00 2.32655317e-01 -5.97934186e-01 8.65244091e-01 4.00213391e-01 -1.03684634e-01 -1.05407095e+00 -9.58387554e-01 -5.29900193e-01 -2.26770341e-01 -1.94348261e-01 5.50478697e-01 7.92275310e-01 -1.41973764e-01 1.69931784e-01 -4.73620951e-01 -1.42186537e-01 3.20253879e-01 -4.38628495e-02 5.97131848e-01 -8.94690275e-01 3.97591107e-03 -4.08999026e-01 -6.01045072e-01 -8.86214018e-01 6.70281053e-02 -5.27865827e-01 1.21352211e-01 -1.19595063e+00 2.45505288e-01 -4.27459478e-01 -4.53628242e-01 1.02346528e+00 -3.94379199e-01 6.56516254e-01 3.35197479e-01 -3.28010768e-02 -5.44438839e-01 1.01871431e+00 1.30442071e+00 -1.97574273e-01 -5.69830909e-02 -1.80119529e-01 -7.44328618e-01 8.55731010e-01 7.19703019e-01 -3.86653721e-01 -2.22063139e-01 -4.39926058e-01 -4.45583388e-02 -4.80011642e-01 6.41798675e-01 -8.44695210e-01 6.93981275e-02 7.71363080e-02 7.02187002e-01 -3.66638899e-01 6.12681925e-01 -5.68228424e-01 -2.72635251e-01 2.94854283e-01 -6.01411946e-02 -1.90702438e-01 4.88826752e-01 8.82494822e-02 -6.14746094e-01 -5.46514243e-02 1.02249348e+00 -1.05774775e-01 -8.70852113e-01 5.24881542e-01 -3.46899718e-01 -8.96005183e-02 8.32248449e-01 -4.46486443e-01 -2.08224475e-01 -3.21383476e-01 -6.16249144e-01 1.75538808e-01 -4.18592803e-02 2.56949246e-01 7.22558737e-01 -1.16381359e+00 -6.03555322e-01 2.43444398e-01 1.02447182e-01 2.84620896e-02 4.48667049e-01 1.00720990e+00 -2.17378348e-01 2.06999466e-01 -5.41221440e-01 -5.10514379e-01 -1.20553601e+00 2.70315468e-01 5.05107820e-01 -2.65443802e-01 -4.35858011e-01 1.22047389e+00 5.95995069e-01 -6.01470210e-02 4.08069879e-01 -2.54967302e-01 -3.29992205e-01 2.46530086e-01 6.38185203e-01 3.33094329e-01 1.42706245e-01 -5.01403630e-01 -4.36229199e-01 6.53486192e-01 -2.59762466e-01 3.52530152e-01 1.59839213e+00 1.93141416e-01 -1.57665625e-01 -3.97936143e-02 1.27955508e+00 -2.41035342e-01 -1.55716240e+00 -1.98646203e-01 -3.39133114e-01 -5.74530184e-01 2.36390620e-01 -6.41795814e-01 -1.42145169e+00 1.22239065e+00 6.08690977e-01 -1.86575308e-01 1.65546215e+00 -2.52449095e-01 7.65741587e-01 1.05685875e-01 2.50235617e-01 -1.03566599e+00 2.36733347e-01 3.79049212e-01 9.31754649e-01 -9.25018549e-01 -1.37843594e-01 -7.19893098e-01 -6.40470743e-01 1.17032158e+00 9.35117781e-01 -2.20429674e-01 7.04341471e-01 2.79835880e-01 -8.56536701e-02 -3.43812644e-01 -5.60314238e-01 -4.43275660e-01 9.81657133e-02 5.85863829e-01 4.02103305e-01 -3.65114510e-01 -2.96412855e-01 7.16190398e-01 5.37037440e-02 3.25305015e-01 6.48091212e-02 8.68354678e-01 -2.39390075e-01 -7.68880188e-01 -2.47556776e-01 4.21028018e-01 -6.68152750e-01 -1.55487508e-01 -3.64105314e-01 7.59293437e-01 6.12590611e-01 7.42547154e-01 1.13516189e-01 -2.70232558e-01 4.88854170e-01 2.73400769e-02 4.26550835e-01 -2.13475600e-01 -8.45731437e-01 -1.06546199e-02 -2.03882810e-02 -7.55001843e-01 -6.31403327e-01 -4.53832835e-01 -8.18193018e-01 -1.27143621e-01 -3.99902284e-01 -1.03039771e-01 2.23526105e-01 8.43973219e-01 5.20708621e-01 5.88424265e-01 7.24927902e-01 -9.98954773e-01 -2.71350712e-01 -1.08425236e+00 -4.32802379e-01 3.24879974e-01 1.74772680e-01 -9.63171959e-01 -1.21822432e-01 -1.04302593e-01]
[13.557586669921875, 1.5777231454849243]
e74217cf-f1cc-47e8-8eb8-7f4996a84263
max-margin-contrastive-learning
2112.11450
null
https://arxiv.org/abs/2112.11450v1
https://arxiv.org/pdf/2112.11450v1.pdf
Max-Margin Contrastive Learning
Standard contrastive learning approaches usually require a large number of negatives for effective unsupervised learning and often exhibit slow convergence. We suspect this behavior is due to the suboptimal selection of negatives used for offering contrast to the positives. We counter this difficulty by taking inspiration from support vector machines (SVMs) to present max-margin contrastive learning (MMCL). Our approach selects negatives as the sparse support vectors obtained via a quadratic optimization problem, and contrastiveness is enforced by maximizing the decision margin. As SVM optimization can be computationally demanding, especially in an end-to-end setting, we present simplifications that alleviate the computational burden. We validate our approach on standard vision benchmark datasets, demonstrating better performance in unsupervised representation learning over state-of-the-art, while having better empirical convergence properties.
['Anoop Cherian', 'Rama Chellappa', 'Suvrit Sra', 'Anshul Shah']
2021-12-21
null
null
null
null
['self-supervised-image-classification']
['computer-vision']
[ 4.42263901e-01 -6.29691407e-02 -4.80811596e-01 -3.29789788e-01 -7.46015728e-01 -6.24592960e-01 8.68618011e-01 1.92633107e-01 -6.70928061e-01 6.25286877e-01 -3.57412882e-02 -2.19872430e-01 1.20064449e-02 -2.63823837e-01 -5.46027958e-01 -6.43530488e-01 -1.14448711e-01 3.68582308e-01 1.23716310e-01 2.14492977e-02 3.70624781e-01 3.82806093e-01 -1.51979983e+00 1.76792338e-01 8.31626117e-01 9.95288849e-01 6.01011515e-03 4.49884504e-01 1.70749113e-01 9.21180248e-01 -6.12620175e-01 -3.97160739e-01 4.86823708e-01 -4.42014992e-01 -6.14283383e-01 2.90097415e-01 1.01045191e+00 -1.44552380e-01 -2.02708885e-01 1.11020207e+00 3.17409098e-01 3.01373720e-01 9.53268230e-01 -1.55574298e+00 -3.60154808e-01 2.26521924e-01 -7.39496291e-01 4.47922051e-01 -1.90097876e-02 2.21260116e-01 1.24962318e+00 -1.16194665e+00 5.71267486e-01 7.92022943e-01 6.64270699e-01 5.14963269e-01 -1.65258408e+00 -5.55753887e-01 1.85788214e-01 1.50900289e-01 -1.19111228e+00 -6.07877851e-01 8.72271359e-01 -5.15391588e-01 1.12392104e+00 4.00400817e-01 6.80634022e-01 1.08434737e+00 -2.56405979e-01 9.87420261e-01 1.30836582e+00 -5.50200820e-01 5.03199697e-01 5.15595853e-01 2.68026441e-01 8.74587238e-01 2.85798758e-01 8.66910070e-02 -6.90972984e-01 -3.58490795e-01 7.72981763e-01 -8.28503817e-02 -1.89945340e-01 -7.69590497e-01 -9.28135037e-01 1.21103692e+00 5.62981308e-01 -5.05049005e-02 -3.94713104e-01 -4.95193116e-02 4.51285660e-01 4.80637848e-01 5.38455665e-01 6.27816617e-01 -4.04168904e-01 -1.05923735e-01 -1.27263725e+00 2.79404342e-01 7.43481457e-01 5.27291179e-01 6.43524230e-01 1.76010236e-01 2.21489333e-02 1.02432418e+00 4.11382355e-02 2.67313302e-01 4.82944787e-01 -8.65272343e-01 2.63093650e-01 3.25087488e-01 -3.23695391e-02 -8.95460308e-01 -2.79241830e-01 -9.06612754e-01 -6.97088599e-01 6.30426705e-01 5.14313757e-01 1.67791527e-02 -8.32306027e-01 1.54206657e+00 6.65077567e-02 1.22755334e-01 1.77185591e-02 1.03523660e+00 5.93835950e-01 4.21078622e-01 1.59480304e-01 -5.06853461e-01 8.78180146e-01 -1.33526039e+00 -3.75255764e-01 -5.37453055e-01 7.71813273e-01 -5.18589199e-01 1.39984751e+00 6.54564500e-01 -9.58412170e-01 -1.86535478e-01 -1.28414178e+00 1.99058458e-01 -8.10842738e-02 2.31470585e-01 9.07768369e-01 5.54989576e-01 -7.10811973e-01 6.74617410e-01 -9.08062935e-01 8.66576564e-03 7.82765388e-01 3.69722307e-01 -3.97874236e-01 2.57504165e-01 -5.72464049e-01 8.51676762e-01 2.16881245e-01 -1.56838492e-01 -5.10621488e-01 -8.08214784e-01 -8.30827236e-01 -6.78607598e-02 3.95237595e-01 -4.88145649e-01 1.07240963e+00 -1.57310843e+00 -1.30463910e+00 1.04970956e+00 -5.64138368e-02 -7.58816242e-01 7.14471221e-01 -3.64195734e-01 1.38253374e-02 3.27498525e-01 4.31065774e-03 6.90847039e-01 1.19219851e+00 -1.26360607e+00 -3.84402871e-01 -1.52135462e-01 -3.49581763e-02 3.59313071e-01 -5.82471669e-01 -4.01074849e-02 -2.61034638e-01 -7.17899561e-01 1.72249600e-01 -1.05715013e+00 -4.58920479e-01 2.64525741e-01 -3.14008385e-01 -7.06391111e-02 8.17422271e-01 -3.48297000e-01 1.08998370e+00 -2.16267538e+00 1.11682519e-01 3.41545016e-01 4.17481333e-01 3.44995022e-01 -8.55027512e-02 3.59885320e-02 -2.99992830e-01 -3.65337998e-01 -5.08814394e-01 -5.12893796e-01 -2.07311541e-01 1.11918308e-01 -4.25195307e-01 7.41415501e-01 4.67710316e-01 7.54083335e-01 -9.84427452e-01 -6.08614624e-01 3.85502636e-01 2.53208727e-01 -7.49121487e-01 -6.48261001e-03 -1.61668181e-01 8.73554796e-02 -3.18542495e-02 6.01837277e-01 4.53187525e-01 -7.14596152e-01 2.47609615e-01 -2.31918097e-01 1.65448278e-01 2.67912209e-01 -1.39548039e+00 1.26789153e+00 -2.95382679e-01 8.16851020e-01 -2.25662678e-01 -1.40446591e+00 5.36754251e-01 -1.31506026e-01 3.89375269e-01 -5.68892658e-01 -7.18806311e-02 3.24421197e-01 5.77147603e-02 -2.58003682e-01 4.36341763e-01 -3.48653048e-01 4.62607563e-01 2.22031996e-01 1.33062169e-01 -2.31182665e-01 1.31481811e-01 3.28895003e-01 8.27729106e-01 -8.94440264e-02 5.54088950e-01 -3.72141212e-01 2.45264322e-01 1.46993130e-01 3.17003429e-01 7.57196248e-01 -2.29357988e-01 5.83012938e-01 6.91819787e-01 -1.82421923e-01 -9.24625218e-01 -1.09901309e+00 -2.42342383e-01 1.14315104e+00 -5.09521887e-02 -5.10443509e-01 -4.41762626e-01 -1.08663821e+00 5.00224298e-03 5.81941009e-01 -6.28315985e-01 -5.61691299e-02 -5.43497384e-01 -8.72995734e-01 2.73327291e-01 6.99579716e-01 8.78325999e-02 -7.82692075e-01 -7.38169312e-01 -8.15433040e-02 2.72027433e-01 -1.05585313e+00 -2.81427771e-01 5.68484902e-01 -9.76741970e-01 -9.89548564e-01 -6.79848313e-01 -8.93271327e-01 8.78200173e-01 3.39162886e-01 1.25588417e+00 1.23116508e-01 -3.62206459e-01 2.78088570e-01 -2.25455575e-02 -3.52921188e-01 -1.59592196e-01 3.66325416e-02 1.41061932e-01 -2.54042834e-01 1.29446819e-01 -4.99474615e-01 -5.75182974e-01 3.63364825e-06 -8.34699452e-01 3.09829023e-02 4.89751637e-01 1.36186957e+00 6.76885724e-01 -1.98846489e-01 4.98806089e-01 -1.10000479e+00 4.65543449e-01 -3.87058496e-01 -8.34577203e-01 -1.02000877e-01 -8.80523205e-01 1.73898935e-01 8.19867492e-01 -6.92467690e-01 -5.59086680e-01 1.87772870e-01 1.80811644e-01 -6.74052775e-01 1.49074748e-01 4.65605348e-01 3.82088631e-01 -4.24601316e-01 1.22514021e+00 2.59694397e-01 2.40073875e-01 -1.14964262e-01 3.16661149e-01 4.84906763e-01 5.05100608e-01 -3.61481249e-01 9.20549214e-01 5.08994937e-01 2.58141905e-01 -1.05120635e+00 -1.07335150e+00 -4.56030458e-01 -3.26454252e-01 -4.91510108e-02 7.67410919e-02 -9.23047245e-01 -3.56642932e-01 1.90581828e-01 -5.43661952e-01 -4.40770060e-01 -4.78160620e-01 5.83943665e-01 -5.58620691e-01 5.00636876e-01 -5.18791616e-01 -8.68234813e-01 -3.73631477e-01 -9.57653224e-01 7.97552407e-01 1.89492106e-01 -3.69518876e-01 -1.05068624e+00 3.46840546e-02 2.25680113e-01 4.05154288e-01 7.67426342e-02 7.68095374e-01 -7.31328130e-01 -2.45656893e-01 -1.72520965e-01 -2.36244485e-01 6.37508571e-01 -9.20222923e-02 1.42645866e-01 -1.10011292e+00 -5.18821836e-01 -2.16881603e-01 -7.57446885e-01 1.18158460e+00 3.91532958e-01 1.12968159e+00 -3.60127449e-01 -2.20945314e-01 6.21909738e-01 1.60812151e+00 -4.36788142e-01 4.09814596e-01 5.40822089e-01 7.09475815e-01 6.03967011e-01 7.58578122e-01 5.13158977e-01 -3.23546790e-02 6.06769443e-01 3.87423158e-01 -3.59944344e-01 7.30661154e-02 -1.07275032e-01 2.05828115e-01 3.53076518e-01 5.71280941e-02 1.24081947e-01 -7.04999566e-01 4.48443800e-01 -1.80711937e+00 -1.00802386e+00 -1.04850888e-01 2.43300843e+00 1.03780961e+00 4.27730888e-01 6.43755317e-01 3.94123167e-01 2.17254400e-01 2.97708571e-01 -5.51732421e-01 -6.78156540e-02 -3.61282289e-01 4.85366732e-01 5.00660658e-01 4.82946396e-01 -1.34129560e+00 1.02277064e+00 6.86833429e+00 8.06176007e-01 -1.65362167e+00 -7.52116144e-02 7.20303893e-01 -5.70570588e-01 -1.25326291e-01 -1.90555789e-02 -4.34077352e-01 3.60882819e-01 5.48106790e-01 1.02796294e-01 3.41752678e-01 1.12549686e+00 -1.49021238e-01 -5.80386892e-02 -1.18707883e+00 1.38743210e+00 1.01732694e-01 -1.49277079e+00 -2.15439469e-01 -1.91827327e-01 8.97774696e-01 3.19615096e-01 4.96927500e-01 2.42697239e-01 -2.04662755e-02 -1.11351609e+00 6.75436676e-01 -2.53502578e-02 6.45586729e-01 -7.08708346e-01 3.76066029e-01 2.42102504e-01 -7.05736697e-01 -5.92338927e-02 -3.54748398e-01 -2.95396715e-01 -2.72350460e-01 7.66879737e-01 -7.73912847e-01 -9.59431902e-02 3.74060571e-01 7.54613519e-01 -6.98291659e-01 1.01907682e+00 -2.00393319e-01 8.25536609e-01 -4.83467877e-01 -1.68214113e-01 3.74534696e-01 -2.61124223e-01 6.31371081e-01 1.49398732e+00 -3.81287336e-01 -3.00616622e-01 2.82372057e-01 6.93980813e-01 -5.38347550e-02 3.07689548e-01 -5.62338948e-01 -4.19378243e-02 4.12560880e-01 1.31355095e+00 -8.97546768e-01 -4.71969515e-01 -5.16492546e-01 1.13063085e+00 5.87435603e-01 3.15392971e-01 -7.16163516e-01 -1.95374608e-01 5.46687484e-01 5.82514182e-02 4.52100188e-01 -2.70588160e-01 -5.44120669e-01 -1.30746114e+00 3.07690859e-01 -1.13722551e+00 4.40200239e-01 -2.82608092e-01 -1.43670392e+00 4.13531780e-01 -7.45213479e-02 -1.46583104e+00 -2.78465211e-01 -6.18134677e-01 -3.81106615e-01 5.36469936e-01 -1.50238776e+00 -8.70351195e-01 -1.63433269e-01 3.11332345e-01 5.39158940e-01 -1.46995232e-01 6.72316968e-01 1.03075698e-01 -3.76338005e-01 9.87401009e-01 -3.60649973e-02 -1.69121057e-01 7.17024684e-01 -1.36614549e+00 1.37858480e-01 7.80213058e-01 4.84404743e-01 6.77919686e-01 9.33319688e-01 -2.64163315e-01 -1.25891984e+00 -7.73416221e-01 6.84528351e-01 -3.49836856e-01 7.13210106e-01 -1.88819498e-01 -8.94167125e-01 5.36879420e-01 -1.17619514e-01 3.97082031e-01 8.54920387e-01 1.89979106e-01 -7.29047239e-01 2.39774942e-01 -1.23351598e+00 7.67986655e-01 9.36154366e-01 -5.51857352e-01 -4.00870323e-01 5.52596152e-01 2.44817644e-01 -3.21788728e-01 -5.02478123e-01 2.81731784e-01 5.74647069e-01 -9.70053673e-01 1.00779998e+00 -6.66364312e-01 7.24698842e-01 -1.53122902e-01 -2.17807531e-01 -1.21970117e+00 -2.27981612e-01 -3.85127366e-01 -4.01140451e-01 6.76601231e-01 5.14038742e-01 -6.54072344e-01 1.22209167e+00 3.77644271e-01 2.41909072e-01 -1.11315489e+00 -6.94511533e-01 -1.00362563e+00 -9.16233361e-02 -3.35706323e-01 -2.90540218e-01 1.43652189e+00 4.41500470e-02 5.92342675e-01 -4.36471075e-01 -2.47908141e-02 9.00896847e-01 6.11902662e-02 7.07078815e-01 -1.24937594e+00 -8.30193877e-01 -7.28327096e-01 -5.84156811e-01 -1.09432471e+00 2.78705180e-01 -1.00321710e+00 -6.80416971e-02 -8.32126975e-01 4.30703163e-01 -4.73172337e-01 -2.93656111e-01 5.89103520e-01 -4.06233311e-01 6.43030226e-01 3.07038695e-01 4.12568271e-01 -5.63848674e-01 5.10005653e-01 8.21795642e-01 -2.05463618e-01 -4.49915409e-01 -1.14121437e-01 -7.73687840e-01 8.95040154e-01 6.23581409e-01 -5.24389088e-01 -5.38836181e-01 -1.44453779e-01 2.27896839e-01 -2.54224330e-01 4.34163958e-01 -8.34561646e-01 2.36851927e-02 -1.55233309e-01 5.05310714e-01 -2.60718882e-01 5.05461693e-01 -6.03706479e-01 -4.81979609e-01 5.56832671e-01 -5.69141984e-01 -1.55180499e-01 1.58330172e-01 4.57589924e-01 -1.91048935e-01 -3.80597174e-01 1.09011686e+00 1.51562065e-01 -4.24242914e-01 1.71708152e-01 -2.05997631e-01 2.27528006e-01 7.90685833e-01 -3.31370443e-01 -2.00590208e-01 -2.45814636e-01 -5.49354255e-01 -6.04805462e-02 6.54884458e-01 -9.67037678e-03 6.56384766e-01 -1.10360146e+00 -6.85841084e-01 2.61537552e-01 2.83380747e-01 -1.57828361e-01 -1.35788828e-01 9.97643650e-01 -4.74341780e-01 3.28316763e-02 -1.37524605e-01 -7.17037201e-01 -1.61922169e+00 4.05049711e-01 2.82321215e-01 -2.38334417e-01 -8.14293683e-01 1.12013316e+00 1.40340328e-01 -1.08135179e-01 4.76890564e-01 6.79368898e-02 -8.49734768e-02 1.16829574e-01 4.59725201e-01 4.08157349e-01 -3.88076901e-02 -3.61272812e-01 -5.21159470e-01 2.70493567e-01 -3.92975807e-01 -3.03856373e-01 1.41324997e+00 4.01179612e-01 1.88962936e-01 3.53599906e-01 1.34939802e+00 1.90372109e-01 -1.42654765e+00 -3.63382667e-01 -3.91825661e-02 -5.86131930e-01 2.56134659e-01 -5.49398065e-01 -9.21733081e-01 5.56001008e-01 6.86622739e-01 7.86997080e-02 1.00284040e+00 -5.39441220e-02 4.27150071e-01 6.84657991e-01 1.31659254e-01 -1.15105677e+00 2.94026494e-01 2.28280574e-01 6.43073738e-01 -1.67093813e+00 3.62318277e-01 -5.81897140e-01 -8.18282723e-01 9.77193058e-01 5.34152985e-01 -6.70557261e-01 5.53580105e-01 2.86324114e-01 3.77052799e-02 -2.04250976e-01 -8.10136437e-01 -5.27459756e-02 3.46827030e-01 5.63257873e-01 4.53401744e-01 9.60390344e-02 -3.64039272e-01 2.23075151e-01 -2.07073167e-01 -2.32706159e-01 3.84258837e-01 1.09843481e+00 -2.72298634e-01 -8.96701455e-01 -4.49424088e-02 7.75397003e-01 -5.15139878e-01 -3.76085132e-01 -4.18733597e-01 6.73634410e-01 -3.31452996e-01 7.97357917e-01 1.83757827e-01 -1.22885205e-01 1.29988402e-01 -1.10965177e-01 9.28606033e-01 -6.14634335e-01 -6.41656399e-01 4.33658063e-02 1.65448695e-01 -4.65607256e-01 -3.99701297e-01 -5.93480110e-01 -1.25714421e+00 3.98188196e-02 -5.25362968e-01 -6.32734448e-02 5.78497648e-01 1.00957346e+00 1.00174867e-01 2.07263559e-01 7.64763594e-01 -7.16143966e-01 -1.12017119e+00 -8.09014976e-01 -5.38269520e-01 6.69308960e-01 4.94087696e-01 -6.63720012e-01 -6.71674013e-01 -4.13966998e-02]
[9.315892219543457, 2.951799154281616]
4d0fffbf-cb14-4f70-b11c-e6291545c2c7
ddipnet-and-ddipnet-discriminant-deep-image
2212.10411
null
https://arxiv.org/abs/2212.10411v1
https://arxiv.org/pdf/2212.10411v1.pdf
DDIPNet and DDIPNet+: Discriminant Deep Image Prior Networks for Remote Sensing Image Classification
Research on remote sensing image classification significantly impacts essential human routine tasks such as urban planning and agriculture. Nowadays, the rapid advance in technology and the availability of many high-quality remote sensing images create a demand for reliable automation methods. The current paper proposes two novel deep learning-based architectures for image classification purposes, i.e., the Discriminant Deep Image Prior Network and the Discriminant Deep Image Prior Network+, which combine Deep Image Prior and Triplet Networks learning strategies. Experiments conducted over three well-known public remote sensing image datasets achieved state-of-the-art results, evidencing the effectiveness of using deep image priors for remote sensing image classification.
['João P. Papa', 'Leandro A. Passos', 'Rafael G. Pires', 'Daniel F. S. Santos']
2022-12-20
null
null
null
null
['remote-sensing-image-classification']
['miscellaneous']
[ 6.33747280e-01 -3.55153561e-01 -1.70068577e-01 -6.10266387e-01 -5.37290275e-01 -2.46950805e-01 7.69611835e-01 -1.29090294e-01 -4.94927794e-01 5.51578879e-01 -4.20057684e-01 -5.58696806e-01 -4.17560726e-01 -1.01500642e+00 -3.34896237e-01 -9.97434199e-01 -1.12347670e-01 4.14719373e-01 -4.15355027e-01 -1.94452614e-01 -7.69588957e-03 9.08303738e-01 -1.57985532e+00 4.01235074e-02 8.62867296e-01 1.19240260e+00 4.25498635e-01 5.41407466e-01 3.29035878e-01 4.21964943e-01 -1.85637757e-01 4.96989600e-02 6.44346058e-01 -2.83073653e-02 -5.30562460e-01 2.49430880e-01 3.55882198e-01 -6.48842275e-01 -2.12585077e-01 1.15895653e+00 6.28671050e-01 2.05597505e-01 7.38009989e-01 -7.79093385e-01 -1.00809908e+00 4.00006354e-01 -8.05641592e-01 2.15343386e-01 -6.00445509e-01 2.48932764e-01 9.29185152e-01 -6.98018909e-01 9.18430761e-02 1.05513406e+00 7.14554429e-01 -1.05014026e-01 -1.27334344e+00 -5.23119271e-01 2.25295067e-01 2.18382776e-01 -1.52143085e+00 -1.56746000e-01 6.13389969e-01 -6.84239864e-01 8.22361350e-01 1.52882293e-01 5.85509539e-01 6.80769384e-01 2.85387933e-01 4.84430164e-01 1.44570804e+00 -2.54377842e-01 3.95866074e-02 -9.20531899e-02 1.65713683e-01 3.10647458e-01 2.53691763e-01 5.08437753e-01 1.94015324e-01 4.36598621e-02 9.98831570e-01 5.34392715e-01 -1.34552374e-01 9.57495123e-02 -8.84486794e-01 9.66833889e-01 7.59744883e-01 4.54673946e-01 -8.99335206e-01 1.14167258e-01 -9.97120589e-02 2.78805494e-01 9.67295289e-01 2.75430530e-01 -8.26087534e-01 6.00169182e-01 -8.98889840e-01 -4.25143354e-02 2.28633761e-01 4.36546832e-01 1.29837835e+00 9.31086987e-02 2.62189507e-01 1.02491653e+00 5.30135691e-01 9.54315603e-01 1.62162170e-01 -6.49366260e-01 -5.89854792e-02 4.99256849e-01 1.18765369e-01 -1.27313757e+00 -5.35492361e-01 -6.75631702e-01 -1.37723589e+00 2.66648769e-01 -1.02736309e-01 -6.05671480e-02 -1.13416481e+00 1.24983394e+00 5.36776893e-02 4.89620268e-02 1.82887852e-01 9.77803230e-01 7.00376153e-01 9.64587629e-01 3.56248498e-01 2.94664294e-01 1.18332434e+00 -7.10535645e-01 -4.34638381e-01 -5.40587127e-01 5.81607223e-02 -4.75647509e-01 9.56083059e-01 4.04480547e-01 -3.06283623e-01 -1.03325045e+00 -1.02864075e+00 6.79045394e-02 -7.74336398e-01 8.98448646e-01 1.07974899e+00 4.39055711e-01 -9.00891840e-01 6.74234986e-01 -9.51659799e-01 -4.45506781e-01 5.27422607e-01 3.15228373e-01 -4.80366647e-01 -2.50310581e-02 -9.37175930e-01 9.29513454e-01 6.76260412e-01 9.95054066e-01 -1.05008733e+00 -6.53452158e-01 -7.71780491e-01 2.01687723e-01 -1.26154751e-01 -2.33632550e-01 9.28862870e-01 -1.20142448e+00 -1.54757154e+00 1.26409519e+00 4.06146616e-01 -3.62793118e-01 2.76241601e-01 -5.02457917e-01 -3.04256469e-01 5.49434274e-02 -5.82733005e-02 8.75716507e-01 8.08686733e-01 -1.18022799e+00 -7.15139329e-01 -6.00872219e-01 -1.83157027e-01 9.53519065e-03 -1.00982547e-01 -2.21854955e-01 3.54634523e-01 -4.26345378e-01 4.33125973e-01 -7.34432340e-01 -5.50017416e-01 7.21545890e-02 -6.72849417e-02 -6.35215417e-02 9.66705084e-01 -9.69084978e-01 3.67951393e-01 -2.24686193e+00 -2.75890499e-01 2.90678591e-01 7.19496906e-02 5.23221314e-01 -6.67249262e-01 2.26171255e-01 -3.91827047e-01 2.09955528e-01 -3.64693433e-01 2.63044000e-01 -3.48235965e-02 4.33543622e-01 -4.15171713e-01 8.40763032e-01 5.20829856e-01 8.41286480e-01 -8.16602826e-01 2.31436253e-01 7.39460409e-01 7.64187038e-01 -2.29939222e-02 2.76557207e-01 -2.13009432e-01 5.79029620e-01 -3.67493898e-01 5.94490111e-01 9.95731473e-01 -4.31123495e-01 1.87177155e-02 -2.35377997e-01 -3.28422308e-01 -7.00127184e-02 -7.90911496e-01 1.34830761e+00 -3.96549612e-01 6.16164804e-01 2.00651467e-01 -1.31181812e+00 1.32000673e+00 1.18714221e-01 5.00837147e-01 -6.36973917e-01 2.29432985e-01 1.29756004e-01 -2.06382036e-01 -4.69266713e-01 4.56074208e-01 -1.78448305e-01 2.75416553e-01 2.99313515e-02 -8.60093012e-02 -2.29007080e-01 -2.45730102e-01 -5.28313279e-01 2.94340342e-01 4.53267694e-01 3.83133292e-01 -5.17507315e-01 4.66601700e-01 8.90860930e-02 3.11456025e-01 7.91276395e-01 -2.19524205e-01 4.60327029e-01 -1.06543750e-01 -9.93197858e-01 -8.10282528e-01 -7.18752265e-01 -4.40503418e-01 1.34753454e+00 -2.24374235e-02 5.18302083e-01 -2.90591598e-01 -2.84947932e-01 4.99100052e-02 1.68408394e-01 -6.87723279e-01 2.76812255e-01 -6.93411380e-02 -1.34706712e+00 4.60496008e-01 5.23075104e-01 1.17201185e+00 -9.93520498e-01 -6.58191621e-01 2.93299347e-01 -1.86056152e-01 -1.02923071e+00 4.85389084e-01 3.26767862e-01 -8.34461212e-01 -1.01526070e+00 -7.82182574e-01 -5.77996433e-01 4.09359574e-01 5.28589547e-01 1.00103283e+00 -3.52801010e-02 -4.10322577e-01 4.79915068e-02 -2.59941727e-01 -3.91239822e-01 -2.03753009e-01 2.71098584e-01 -4.84840095e-01 1.12285241e-01 5.18571675e-01 -7.17982352e-01 -6.60178006e-01 1.03544235e-01 -1.05416286e+00 -3.90551388e-02 1.15784574e+00 9.68905270e-01 6.76024020e-01 4.14009243e-01 3.61192375e-01 -6.49075329e-01 -9.70890671e-02 -4.87121880e-01 -1.09832239e+00 2.51449585e-01 -5.54597020e-01 -2.82381207e-01 1.16818376e-01 -8.20474140e-03 -1.14369655e+00 3.05736750e-01 -3.94132823e-01 1.99479148e-01 -8.74045968e-01 1.12610221e+00 -1.88640326e-01 -7.78558701e-02 6.76971793e-01 2.13854209e-01 -9.50974971e-02 -4.82652336e-01 3.40775818e-01 9.62552428e-01 5.57090461e-01 -5.78706376e-02 5.93716204e-01 5.55688381e-01 1.21351629e-02 -1.40811253e+00 -1.03514862e+00 -6.73363864e-01 -7.94842064e-01 -2.89698523e-02 1.21263611e+00 -1.33443356e+00 -3.14660639e-01 9.48598087e-01 -9.42650020e-01 -6.28322542e-01 1.83789074e-01 6.86416268e-01 -2.03843355e-01 1.87884971e-01 -4.10387397e-01 -7.02079535e-01 -5.56003928e-01 -9.02498245e-01 1.19844472e+00 4.30231020e-02 4.39673424e-01 -1.00965667e+00 3.26806726e-03 1.56599388e-01 6.73442245e-01 4.08010632e-01 6.88395858e-01 -3.18660080e-01 -5.66221178e-01 -1.38772920e-01 -8.37230682e-01 7.79489040e-01 3.37769181e-01 1.60410181e-01 -1.29200411e+00 -2.35489577e-01 -2.34798148e-01 -3.28925103e-01 1.17515814e+00 6.88866556e-01 1.08134854e+00 -1.78304338e-03 -1.69727355e-01 8.53022575e-01 1.81973982e+00 5.83433583e-02 8.12524676e-01 3.18003535e-01 6.87987626e-01 6.62470996e-01 4.65196311e-01 3.00113201e-01 1.20053940e-01 2.46261016e-01 6.77778006e-01 -6.18912756e-01 3.10283482e-01 4.39964309e-02 -1.67876124e-01 2.22415328e-01 -2.13426948e-01 -1.41203567e-01 -1.21114266e+00 5.23256600e-01 -1.90231383e+00 -9.87446785e-01 -7.81180859e-02 1.65307987e+00 3.94835740e-01 -5.34745872e-01 -6.23024642e-01 1.56590208e-01 5.70045471e-01 3.69764417e-01 -4.65588301e-01 2.76635975e-01 -2.18084261e-01 3.81696701e-01 8.52981746e-01 4.49620366e-01 -1.80180991e+00 1.15762258e+00 6.55390549e+00 5.51516831e-01 -1.67652905e+00 -3.59829925e-02 1.02740324e+00 8.14970732e-01 -2.12832447e-02 -1.65437475e-01 -6.86277211e-01 -1.63246498e-01 7.14750350e-01 4.23540115e-01 1.89994723e-01 8.42458963e-01 4.36668873e-01 -3.64587784e-01 -5.65391898e-01 9.55938339e-01 -2.75274962e-01 -1.18854535e+00 2.21283197e-01 8.68102685e-02 9.02992070e-01 6.93700433e-01 2.89903373e-01 2.03971207e-01 4.32175428e-01 -1.04785991e+00 4.28217798e-01 3.85385066e-01 7.39696026e-01 -4.61239427e-01 1.07786393e+00 3.23698223e-01 -1.05829620e+00 -4.84139919e-02 -6.97488308e-01 -4.51399475e-01 -3.05487663e-01 9.28499520e-01 -4.86385047e-01 6.78309441e-01 9.05568600e-01 1.16118610e+00 -4.70146298e-01 6.29270852e-01 -6.03724480e-01 8.20796251e-01 -4.19840455e-01 4.29369748e-01 5.43249488e-01 -6.16062164e-01 6.19575419e-02 1.11055422e+00 2.67781466e-01 3.57181609e-01 4.59363520e-01 7.95036972e-01 2.12873623e-01 -1.50415421e-01 -6.60098374e-01 -3.05693239e-01 1.01063929e-01 1.44626915e+00 -8.39164734e-01 -1.39612243e-01 -5.37675396e-02 6.75224841e-01 -1.22921847e-01 5.53764999e-01 -5.75499892e-01 -1.70133248e-01 6.92502677e-01 -2.85042465e-01 3.58608216e-01 -5.82257748e-01 -3.30346823e-01 -1.05237663e+00 -3.34215015e-01 -7.54141569e-01 2.14014068e-01 -8.85504007e-01 -1.35647559e+00 5.04176736e-01 1.18031345e-01 -8.31050634e-01 1.55846909e-01 -1.02308178e+00 -3.63013119e-01 1.15044868e+00 -2.43609905e+00 -1.53477550e+00 -7.72741914e-01 2.81604767e-01 9.13058296e-02 -7.21036196e-02 1.15873146e+00 2.28077471e-01 -4.38171893e-01 -3.35646458e-02 3.87528330e-01 2.88219303e-01 2.95967579e-01 -8.72516811e-01 2.52676696e-01 1.05214393e+00 1.09620117e-01 1.44733861e-01 1.79448247e-01 -1.30196929e-01 -1.18580019e+00 -1.46248925e+00 6.57514811e-01 2.57689476e-01 4.59106266e-01 7.67639652e-02 -7.32922912e-01 7.43788838e-01 7.18524307e-02 1.01103093e-02 7.56397963e-01 -1.41941905e-01 -2.16082335e-01 -6.85357988e-01 -1.15361786e+00 8.01102519e-02 5.21404028e-01 -5.40445566e-01 -3.23222101e-01 3.37145627e-01 1.15747396e-02 9.76140425e-03 -6.85248673e-01 7.32819676e-01 5.57221591e-01 -8.25325727e-01 8.82110000e-01 -2.67647535e-01 5.36035240e-01 -2.74231523e-01 -4.72197860e-01 -1.20083809e+00 -7.11171329e-01 -2.13199005e-01 6.71219885e-01 7.51267374e-01 2.34559223e-01 -8.57828081e-01 5.09391844e-01 1.10081464e-01 -8.27364177e-02 -1.61878750e-01 -4.24496531e-01 -6.40076041e-01 7.66687021e-02 -1.57241106e-01 6.17231190e-01 1.05533206e+00 -9.60739255e-01 3.75433862e-01 -2.55122930e-01 7.85788059e-01 3.84832382e-01 3.01941514e-01 6.69220686e-01 -1.63755858e+00 -1.55162871e-01 -5.91985106e-01 -5.26862800e-01 -1.24079382e+00 3.00560296e-01 -6.94384038e-01 3.30657780e-01 -1.74006736e+00 1.43786684e-01 -5.39958119e-01 -2.80364633e-01 1.03481758e+00 -3.10767353e-01 6.78111792e-01 -2.46408686e-01 2.07339495e-01 2.12242566e-02 6.36388242e-01 8.17566335e-01 -5.09456277e-01 -1.05202831e-01 2.57700793e-02 -5.51532745e-01 7.38386810e-01 1.22975957e+00 -2.89366215e-01 -1.86825082e-01 -9.22934473e-01 2.58641183e-01 -4.06009227e-01 7.95111656e-01 -9.61012244e-01 -3.92582268e-01 -3.11078548e-01 5.38327038e-01 -7.99292445e-01 1.90034863e-02 -1.09700203e+00 2.23501280e-01 5.75056136e-01 -1.10477179e-01 -2.04609588e-01 4.45642471e-01 3.56174737e-01 -4.40690726e-01 2.54857838e-01 1.01303291e+00 -1.77617982e-01 -1.23661423e+00 5.19084573e-01 -5.68360865e-01 -7.84881234e-01 8.31577897e-01 -4.24562544e-02 -1.85389206e-01 -1.16794713e-01 -7.05960572e-01 6.98125288e-02 -1.52026623e-01 2.73197949e-01 4.48829472e-01 -9.57618117e-01 -1.26515841e+00 6.11405134e-01 9.56253931e-02 -2.87907897e-03 1.40501976e-01 6.29079282e-01 -9.03079033e-01 2.62340158e-01 -5.80149174e-01 -1.00343204e+00 -9.13049281e-01 1.29388710e-02 5.36005497e-01 -3.22549731e-01 -5.07912159e-01 8.22619557e-01 1.29487857e-01 -7.75365591e-01 -1.60374358e-01 -4.71429169e-01 -2.50346392e-01 4.82769720e-02 4.96124387e-01 4.00682420e-01 1.18086562e-01 -5.30809104e-01 -2.57247865e-01 5.14688134e-01 2.31334090e-01 -3.59088220e-02 1.84107506e+00 -9.34495926e-02 -5.07641971e-01 2.93263346e-01 9.43317771e-01 -7.11693883e-01 -1.46428275e+00 -1.89465776e-01 6.26199767e-02 -4.64691609e-01 7.87961662e-01 -9.42972839e-01 -1.17949867e+00 1.29801834e+00 1.09330153e+00 1.54088840e-01 1.36932921e+00 -4.60598141e-01 1.28559753e-01 9.69805360e-01 3.32756519e-01 -6.58805311e-01 -4.12786484e-01 8.36862445e-01 8.28425050e-01 -1.70362890e+00 1.58351660e-01 -2.29951024e-01 -1.47021621e-01 1.05966437e+00 2.13159695e-01 -1.79151759e-01 1.11302865e+00 -8.46259594e-02 6.72367990e-01 -1.57671869e-01 -5.49926311e-02 -6.23058617e-01 3.40926349e-01 7.66567647e-01 6.40705109e-01 3.13359857e-01 -1.11387104e-01 1.70347303e-01 1.48109317e-01 8.91943201e-02 -6.83255643e-02 8.71394157e-01 -5.39280891e-01 -8.09401155e-01 -2.33112216e-01 2.53250718e-01 -4.61590081e-01 -2.06277192e-01 -1.75126135e-01 6.94532514e-01 2.46047333e-01 8.88779879e-01 9.50119346e-02 -1.38139531e-01 -1.14051281e-02 -2.37221584e-01 3.29790145e-01 -5.53027391e-01 -5.20671368e-01 6.61988035e-02 -3.89356613e-01 -2.38556728e-01 -1.02727318e+00 -2.22315580e-01 -6.89311326e-01 -2.81666666e-01 -5.38378656e-01 4.80804294e-02 9.60495174e-01 9.63648021e-01 3.08055431e-01 1.73202172e-01 6.90982521e-01 -1.33824539e+00 -8.06862175e-01 -1.31861460e+00 -8.01595986e-01 5.87936938e-02 3.97566110e-01 -2.88491040e-01 -8.28301087e-02 3.28431785e-01]
[9.625642776489258, -1.4939337968826294]
aca393ec-dfc3-4386-8287-c0a58dd26f86
trade-the-event-corporate-events-detection
2105.12825
null
https://arxiv.org/abs/2105.12825v2
https://arxiv.org/pdf/2105.12825v2.pdf
Trade the Event: Corporate Events Detection for News-Based Event-Driven Trading
In this paper, we introduce an event-driven trading strategy that predicts stock movements by detecting corporate events from news articles. Unlike existing models that utilize textual features (e.g., bag-of-words) and sentiments to directly make stock predictions, we consider corporate events as the driving force behind stock movements and aim to profit from the temporary stock mispricing that may occur when corporate events take place. The core of the proposed strategy is a bi-level event detection model. The low-level event detector identifies events' existences from each token, while the high-level event detector incorporates the entire article's representation and the low-level detected results to discover events at the article-level. We also develop an elaborately-annotated dataset EDT for corporate event detection and news-based stock prediction benchmark. EDT includes 9721 news articles with token-level event labels as well as 303893 news articles with minute-level timestamps and comprehensive stock price labels. Experiments on EDT indicate that the proposed strategy outperforms all the baselines in winning rate, excess returns over the market, and the average return on each transaction.
['Han Liu', 'Liqian Ma', 'Zhihan Zhou']
2021-05-26
null
https://aclanthology.org/2021.findings-acl.186
https://aclanthology.org/2021.findings-acl.186.pdf
findings-acl-2021-8
['text-based-stock-prediction', 'event-driven-trading', 'stock-trend-prediction', 'stock-market-prediction', 'stock-price-prediction', 'stock-prediction']
['natural-language-processing', 'natural-language-processing', 'time-series', 'time-series', 'time-series', 'time-series']
[-6.77056909e-01 -2.77917027e-01 -6.14372671e-01 -2.42738962e-01 -9.03615057e-01 -8.25350285e-01 1.04434109e+00 6.29193068e-01 -3.72412890e-01 7.54627407e-01 6.83411419e-01 -2.07266226e-01 3.89594793e-01 -1.38254535e+00 -6.30670905e-01 -1.66421488e-01 -3.27555746e-01 1.79929763e-01 5.89050770e-01 -1.31161869e-01 8.57030630e-01 2.97087252e-01 -1.29346240e+00 4.68110651e-01 2.59829909e-01 1.81977928e+00 -1.64377615e-01 2.02338293e-01 -6.73111022e-01 1.93736851e+00 -9.08829570e-01 -6.62215233e-01 4.29786772e-01 -2.93580621e-01 -3.12740117e-01 -9.83197838e-02 -1.11617908e-01 -5.74223638e-01 -2.01039866e-01 9.27697718e-01 4.53706421e-02 -2.50289500e-01 4.91200566e-01 -1.08525026e+00 -4.09739226e-01 1.32126892e+00 -6.66434467e-01 1.00913966e+00 7.38677159e-02 -1.15455247e-01 1.80954099e+00 -1.13743210e+00 9.13021863e-01 7.56698191e-01 6.22708380e-01 -4.18409377e-01 -4.78994399e-01 -7.58386731e-01 4.59440082e-01 -2.47300401e-01 -5.64758599e-01 -1.49801910e-01 6.61102474e-01 -5.09305716e-01 1.14549828e+00 6.39328286e-02 1.02195513e+00 8.86209190e-01 7.54295409e-01 1.11483443e+00 9.64519441e-01 1.83702689e-02 3.61122072e-01 7.71792680e-02 2.40429237e-01 1.62813626e-02 5.98161340e-01 7.04400390e-02 -9.12247837e-01 -5.24152100e-01 6.71621859e-01 6.69102728e-01 3.40509385e-01 5.73478341e-01 -1.61985683e+00 1.30860627e+00 1.55792564e-01 2.89673716e-01 -1.05821609e+00 5.11086248e-02 9.06581879e-01 6.03019238e-01 1.11719668e+00 4.91160154e-01 -8.16612840e-01 -2.92737335e-01 -1.20318723e+00 7.41420686e-01 1.16324878e+00 9.05311525e-01 1.40555575e-01 4.12984103e-01 -4.44830954e-01 2.77941704e-01 2.38432333e-01 2.74072140e-01 9.95517194e-01 -1.94092557e-01 6.57268882e-01 7.00715840e-01 3.25348586e-01 -1.10498226e+00 -2.78618217e-01 -9.55498815e-01 -1.81451693e-01 -1.62442580e-01 7.49915093e-02 -2.95773685e-01 -4.73445266e-01 9.55838382e-01 2.79459767e-02 4.09845084e-01 3.42238188e-01 4.95162308e-01 9.05785680e-01 1.10467029e+00 -8.87771845e-02 -6.17661774e-01 1.54943466e+00 -9.17212546e-01 -8.28182638e-01 -1.81180775e-01 3.85980159e-01 -9.05287325e-01 1.56211376e-01 1.48011789e-01 -1.18373549e+00 -8.04907903e-02 -8.58040750e-01 6.33572578e-01 -7.33523548e-01 -2.39451066e-01 5.83377481e-01 -7.05911145e-02 -1.93678245e-01 5.01732051e-01 -5.67080855e-01 3.12385112e-01 6.71932623e-02 -3.33286166e-01 4.14691716e-01 9.43473458e-01 -1.66106856e+00 7.40931928e-01 6.64806843e-01 -2.32136577e-01 -6.09459996e-01 -8.98921847e-01 -6.09162271e-01 7.83856809e-02 4.78034586e-01 8.84859040e-02 1.81514955e+00 -5.28755546e-01 -1.09276962e+00 7.00144529e-01 2.79153466e-01 -1.37226629e+00 6.97385848e-01 -2.30914265e-01 -7.84800112e-01 -6.79706186e-02 5.33380508e-01 -1.07718043e-01 6.13305688e-01 -4.50583309e-01 -1.40449870e+00 1.06662765e-01 -2.81081676e-01 -1.65845856e-01 -6.71928599e-02 4.06450719e-01 1.14347048e-01 -1.51113045e+00 6.38333708e-02 -3.69250983e-01 1.52102597e-02 -1.01540697e+00 -5.41436911e-01 -4.06637162e-01 6.11599505e-01 -5.60696006e-01 1.41324914e+00 -2.00247288e+00 -9.86982346e-01 2.47515365e-01 1.47561595e-01 -5.62066317e-01 4.59735453e-01 9.65194404e-01 -1.72256976e-01 5.42191528e-02 6.20203376e-01 -7.90289938e-02 4.28137392e-01 -2.75726855e-01 -1.57903445e+00 2.40868434e-01 2.12396935e-01 1.11246824e+00 -6.71964467e-01 -2.91979343e-01 -1.14838287e-01 -3.47222507e-01 -1.66433364e-01 4.15267562e-03 -4.73875552e-01 -2.74995804e-01 -6.63028836e-01 8.73493612e-01 1.83523968e-01 -5.25527060e-01 -3.65544796e-01 5.25841936e-02 -8.73247862e-01 1.06352246e+00 -1.21295428e+00 5.47466993e-01 8.73780549e-02 4.41985101e-01 -7.49502420e-01 -4.18514669e-01 1.08844554e+00 3.89379084e-01 6.28414929e-01 -7.46774971e-01 1.97443068e-01 5.38772106e-01 -4.88548845e-01 2.53504485e-01 9.10574019e-01 -4.70553160e-01 -7.68364131e-01 8.10430527e-01 -4.08944786e-01 3.54539633e-01 3.58908653e-01 2.77464360e-01 1.14089453e+00 -4.94740903e-01 7.22723842e-01 -7.24112317e-02 -4.33430448e-03 4.67554897e-01 8.47457588e-01 7.65946209e-01 6.77583646e-03 2.01258257e-01 8.87747467e-01 -8.16035569e-01 -8.09788167e-01 -8.34477782e-01 -1.59138978e-01 1.32045543e+00 -1.16438195e-01 -3.42529744e-01 -8.62454157e-03 -6.01966321e-01 6.13626599e-01 1.05366337e+00 -6.68427289e-01 3.24841857e-01 -2.99386084e-01 -1.02135324e+00 3.09922278e-01 5.74890375e-01 5.49039960e-01 -1.36781073e+00 -8.33899319e-01 9.15165722e-01 9.54410434e-02 -1.05684066e+00 -6.88254356e-01 6.51741505e-01 -6.05366349e-01 -7.57750332e-01 -5.45366466e-01 -7.09754527e-01 3.19768786e-02 -1.85845792e-01 1.44669068e+00 -7.58445382e-01 2.06696182e-01 -1.53242081e-01 -5.03652990e-01 -1.13193583e+00 -1.86064288e-01 -7.07856566e-02 -1.87118605e-01 3.23941737e-01 4.97853339e-01 -4.43356000e-02 -6.38032317e-01 8.47719908e-02 -7.59809196e-01 -3.23126823e-01 5.61778724e-01 5.40103137e-01 6.08860731e-01 6.45575941e-01 1.19313860e+00 -1.01411724e+00 6.60142899e-01 -7.27327704e-01 -9.84086275e-01 -6.69112802e-02 -8.10262203e-01 -2.06994653e-01 2.55993843e-01 -4.32429433e-01 -1.17600310e+00 -4.39016253e-01 2.84899890e-01 -1.18563265e-01 4.14648116e-01 1.24540877e+00 6.59027338e-01 8.33284557e-01 8.76961648e-02 7.36302674e-01 -4.48860556e-01 -3.61643285e-01 -1.56820148e-01 3.55601817e-01 5.17203689e-01 -6.91451551e-03 6.89014614e-01 3.81241500e-01 -7.70332277e-01 -2.54946351e-01 -1.42551494e+00 -5.16436160e-01 -8.96954909e-02 -2.45097309e-01 5.16678691e-01 -1.50430608e+00 -4.88208205e-01 8.76635730e-01 -8.91071916e-01 2.68098027e-01 -6.91924274e-01 7.43639052e-01 -1.82677507e-01 -4.82605517e-01 -1.36867917e+00 -1.00472140e+00 -5.58476865e-01 -4.18623805e-01 8.38260531e-01 8.40168521e-02 -4.60233182e-01 -9.78323698e-01 4.15495843e-01 -1.20486692e-01 4.06647861e-01 5.98632216e-01 4.84705985e-01 -1.84647489e+00 -6.36686802e-01 -8.11985970e-01 -2.13137120e-02 1.28595576e-01 2.11877018e-01 -4.71261814e-02 -5.11245131e-01 5.69458399e-03 3.44589293e-01 -2.15059504e-01 1.14190757e+00 5.99616766e-01 3.21150154e-01 -6.32441044e-01 1.06101539e-02 5.36611862e-02 1.31328642e+00 6.30754948e-01 3.34270179e-01 1.05578721e+00 8.04386009e-03 3.83118123e-01 8.54546368e-01 1.18647921e+00 4.32727635e-01 3.22284363e-02 2.17636660e-01 2.70823330e-01 6.99945450e-01 -5.05930305e-01 7.85388708e-01 1.04795420e+00 2.14096904e-01 -1.43752158e-01 -5.96252739e-01 6.25971079e-01 -1.69227529e+00 -1.68327653e+00 -9.29242373e-02 1.62543678e+00 1.02473211e+00 1.15678895e+00 4.14741307e-01 6.50928617e-02 7.18426228e-01 8.98999631e-01 -5.25706291e-01 7.76212439e-02 -4.55916017e-01 -3.59604418e-01 9.00801241e-01 -1.42601609e-01 -1.31383443e+00 7.38468349e-01 6.44475698e+00 6.37199700e-01 -1.09597242e+00 -3.09355974e-01 7.19069719e-01 -4.19264644e-01 -3.82227957e-01 -2.34184042e-01 -1.51903772e+00 1.05130637e+00 9.47562933e-01 -6.43845797e-01 -5.53508997e-01 1.15665817e+00 3.33897799e-01 8.91822297e-03 -8.07132363e-01 6.70087278e-01 -8.22140127e-02 -2.09571004e+00 1.19395800e-01 3.36325377e-01 7.69665420e-01 2.30824471e-01 8.68474171e-02 4.18333381e-01 3.66363525e-01 -1.53414786e-01 1.59224582e+00 6.13265455e-01 2.76807565e-02 -9.70194161e-01 1.07726085e+00 2.53434986e-01 -1.55796528e+00 -8.38105306e-02 -6.11912319e-03 -2.60331422e-01 3.74720424e-01 9.93993282e-01 -7.15468287e-01 3.78413916e-01 5.76112032e-01 1.36630225e+00 -2.63841033e-01 7.11313426e-01 1.99332148e-01 8.43945026e-01 -1.72891051e-01 -3.86668444e-01 7.22387791e-01 -1.58301711e-01 3.37136656e-01 1.18367946e+00 5.46502769e-01 -3.36174271e-03 1.59586072e-01 6.73754513e-01 -3.36090475e-01 2.43191242e-01 -4.47689146e-01 -4.03189301e-01 6.25897467e-01 8.27522218e-01 -1.12975359e+00 -6.98551476e-01 -9.21284318e-01 4.69764322e-01 -2.29005039e-01 -1.08332559e-01 -9.03305233e-01 -5.59728622e-01 5.25148749e-01 1.76297098e-01 9.57386732e-01 2.15807348e-01 -3.22139114e-01 -1.41338921e+00 7.30468035e-02 -5.30427396e-01 5.71478665e-01 -4.28033918e-01 -1.74500537e+00 5.73723137e-01 -2.79187024e-01 -1.56948268e+00 -5.04006147e-01 -4.47773546e-01 -1.23325360e+00 4.84495997e-01 -1.70590901e+00 -4.29944932e-01 4.41371650e-01 3.34702849e-01 7.17575252e-01 -5.84083080e-01 1.67013049e-01 -9.41685662e-02 -5.35140216e-01 -4.02524509e-02 3.84470344e-01 7.81358421e-01 6.82629645e-01 -1.53013790e+00 9.06224191e-01 7.02368081e-01 2.23781630e-01 3.47603977e-01 5.74260533e-01 -1.33188212e+00 -8.97060871e-01 -1.37581027e+00 1.42444801e+00 -4.74438459e-01 1.72125661e+00 -9.59291402e-03 -7.35962808e-01 1.02686036e+00 1.13971435e-01 -3.20193022e-01 6.83532178e-01 -2.06970438e-01 -4.86902058e-01 -2.46790498e-01 -8.22541177e-01 2.01155171e-01 2.98721701e-01 -5.06717563e-01 -1.27107656e+00 2.62633413e-01 8.13450634e-01 -3.07319492e-01 -1.05339611e+00 -4.55915853e-02 4.78951067e-01 -7.84389257e-01 7.80638218e-01 -4.60927337e-01 5.98416328e-01 1.21976230e-02 3.35871428e-02 -1.04310453e+00 -2.70038962e-01 -6.08765423e-01 -2.20793217e-01 1.48857701e+00 7.00280130e-01 -9.95627820e-01 5.72779715e-01 2.20220059e-01 1.13039128e-02 -4.84497994e-01 -7.31033027e-01 -9.30901706e-01 -1.47084713e-01 -5.66415966e-01 9.23993409e-01 1.00191832e+00 1.21934630e-01 1.96512476e-01 -4.91363734e-01 -5.63825257e-02 5.06089568e-01 9.75622535e-01 3.12156469e-01 -1.38254607e+00 -7.28007108e-02 -7.14689374e-01 -4.97175679e-02 -7.69784570e-01 -9.17108580e-02 -6.05139256e-01 -9.10710171e-02 -1.19698298e+00 1.50347829e-01 6.07701838e-02 -7.04233050e-01 1.75064713e-01 1.30147979e-01 -7.74967894e-02 -1.03273280e-02 7.88362741e-01 -7.86927879e-01 4.07901078e-01 8.53301764e-01 -3.85199450e-02 -2.38401175e-01 1.94667444e-01 -6.06712937e-01 9.97678757e-01 6.83105230e-01 -6.76393688e-01 2.68018097e-01 3.19441855e-01 8.51773441e-01 2.72259265e-01 2.79967815e-01 -3.39959592e-01 4.73878354e-01 -2.95019627e-01 5.67218006e-01 -1.61484456e+00 -1.88605577e-01 -3.64837289e-01 1.76933452e-01 7.09314167e-01 -6.31194413e-01 5.22903860e-01 -1.93670884e-01 8.84180188e-01 -9.50971246e-01 -4.74144071e-02 4.23279405e-01 -3.29015046e-01 -6.89979494e-01 2.61161596e-01 -7.76628077e-01 3.37997437e-01 1.27906418e+00 3.44820619e-02 -4.31392759e-01 -4.04314429e-01 -5.82209229e-01 3.02161604e-01 -7.90145770e-02 4.63333070e-01 4.17384267e-01 -1.51712644e+00 -1.06377888e+00 8.74191406e-04 3.10623318e-01 -4.70104009e-01 -3.20396900e-01 6.78561926e-01 -4.85938281e-01 7.30445683e-01 6.80722445e-02 1.52037606e-01 -5.57603955e-01 3.98831099e-01 2.51868870e-02 -6.47062242e-01 -7.83696711e-01 7.28221953e-01 3.09211947e-02 4.58613276e-01 3.04435313e-01 -1.00910270e+00 -3.75948817e-01 1.22073889e+00 8.72007132e-01 2.83062577e-01 -2.11734325e-02 -4.21433747e-01 -3.69167387e-01 -1.04151607e-01 -5.42119622e-01 -1.72998652e-01 1.79020894e+00 4.48344313e-02 -2.06411004e-01 1.11638176e+00 8.56387556e-01 2.70928890e-01 -1.10075307e+00 -6.20028555e-01 9.15898740e-01 -8.02395716e-02 2.00336754e-01 -6.26744092e-01 -1.12266397e+00 -5.42753227e-02 -3.49316180e-01 1.06946445e+00 5.22695839e-01 2.69849300e-01 1.35383165e+00 1.23382986e-01 2.81515270e-01 -1.25293112e+00 7.62546882e-02 7.57815778e-01 8.86702895e-01 -1.21961129e+00 -5.16289249e-02 -2.45872466e-03 -8.93016756e-01 9.52571869e-01 7.63429031e-02 -5.24416983e-01 1.07471669e+00 4.60029691e-01 5.27147087e-04 -5.09302914e-01 -1.29226661e+00 -3.46874036e-02 2.52106905e-01 -6.34164274e-01 2.04089969e-01 -4.59059775e-02 -1.18413948e-01 1.15553880e+00 -6.28228188e-01 -2.63616800e-01 6.49707079e-01 1.32221091e+00 -7.99105644e-01 -4.95623678e-01 -3.88737619e-01 1.21392667e+00 -1.39523661e+00 -2.43981630e-01 -2.81520784e-01 8.00177097e-01 -3.15140218e-01 6.23324573e-01 6.41533971e-01 -2.05242142e-01 5.38968027e-01 2.55738437e-01 -6.72577739e-01 -7.37357020e-01 -1.10049248e+00 5.63499510e-01 2.97169119e-01 -4.00644362e-01 -3.04810315e-01 -1.25853848e+00 -1.31343222e+00 -4.71541166e-01 -1.17825910e-01 3.76230329e-01 1.34100050e-01 8.88503611e-01 1.68338895e-01 4.89839584e-01 1.11260271e+00 -3.10659677e-01 -8.41812193e-01 -9.36724901e-01 -1.33531475e+00 1.34369805e-01 6.35358036e-01 -4.62000698e-01 -7.26416469e-01 1.92477807e-01]
[4.41933012008667, 4.283189296722412]
a53468cb-bbec-4ff4-949e-4aa52f5a8d8e
single-mr-image-super-resolution-using
2207.08036
null
https://arxiv.org/abs/2207.08036v1
https://arxiv.org/pdf/2207.08036v1.pdf
Single MR Image Super-Resolution using Generative Adversarial Network
Spatial resolution of medical images can be improved using super-resolution methods. Real Enhanced Super Resolution Generative Adversarial Network (Real-ESRGAN) is one of the recent effective approaches utilized to produce higher resolution images, given input images of lower resolution. In this paper, we apply this method to enhance the spatial resolution of 2D MR images. In our proposed approach, we slightly modify the structure of the Real-ESRGAN to train 2D Magnetic Resonance images (MRI) taken from the Brain Tumor Segmentation Challenge (BraTS) 2018 dataset. The obtained results are validated qualitatively and quantitatively by computing SSIM (Structural Similarity Index Measure), NRMSE (Normalized Root Mean Square Error), MAE (Mean Absolute Error), and VIF (Visual Information Fidelity) values.
['Mehran Ebrahimi', 'Elham Shakibapour', 'Shawkh Ibne Rashid']
2022-07-16
null
null
null
null
['brain-tumor-segmentation']
['medical']
[ 8.15010369e-01 5.95107615e-01 1.18851528e-01 3.53047401e-02 -1.09389758e+00 -2.63812512e-01 5.43618679e-01 -4.60809737e-01 -5.93787074e-01 1.14440262e+00 2.59240836e-01 -4.61268611e-02 -6.29572123e-02 -7.94202209e-01 -5.40567815e-01 -8.96477520e-01 -1.45042464e-01 1.95398882e-01 3.98280233e-01 -4.03755635e-01 2.27430657e-01 5.87924600e-01 -8.45978916e-01 2.79906511e-01 1.02461839e+00 7.10904658e-01 2.58563131e-01 6.52507246e-01 2.43598327e-01 9.22956645e-01 -5.67949593e-01 -2.11405411e-01 2.53144592e-01 -6.44802332e-01 -1.16802120e+00 -2.31868237e-01 6.04154356e-02 -4.59167302e-01 -4.71161276e-01 1.28729808e+00 6.71617448e-01 1.46502212e-01 7.41285980e-01 -6.64298892e-01 -8.76184106e-01 8.93142521e-01 -9.14381087e-01 1.12790942e+00 1.20448224e-01 3.03272288e-02 -6.48056567e-02 -7.33923137e-01 1.08731508e+00 1.06439388e+00 4.55103040e-01 6.68968618e-01 -1.33817303e+00 -6.42164111e-01 -5.34235895e-01 -7.52098951e-03 -1.37113893e+00 -2.27340441e-02 5.43888450e-01 -2.38980234e-01 4.37397540e-01 2.55649865e-01 1.92472935e-01 1.09983385e+00 7.61890233e-01 1.82840824e-01 1.79339325e+00 -1.93797007e-01 8.78971219e-02 -6.12830892e-02 -3.58195186e-01 6.30713522e-01 1.94680125e-01 4.90870416e-01 6.50526509e-02 9.34871137e-02 1.59981799e+00 -1.94201156e-01 -4.91498291e-01 3.48634347e-02 -1.22059047e+00 9.33659434e-01 1.00551188e+00 7.78057694e-01 -4.61805463e-01 -1.48249209e-01 9.86183733e-02 2.57180035e-01 4.39275324e-01 5.36576033e-01 4.14274037e-01 2.00451329e-01 -7.63244390e-01 -1.21811047e-01 -3.39436978e-02 5.09082437e-01 4.61395010e-02 3.92363936e-01 -3.53926301e-01 6.42949760e-01 -1.85796842e-01 3.60357791e-01 9.28496838e-01 -1.06129384e+00 5.09610660e-02 1.29011974e-01 -1.61783099e-01 -9.14773285e-01 -4.30279613e-01 -7.52747059e-01 -1.32757318e+00 4.11757469e-01 5.01950271e-02 -8.79506618e-02 -1.26128089e+00 1.65440238e+00 3.46603394e-01 1.68879956e-01 2.95622408e-01 1.15614045e+00 1.09853625e+00 4.21659023e-01 1.69675693e-01 -4.02787358e-01 1.08418298e+00 -8.30964684e-01 -8.87799442e-01 8.87158141e-03 2.05927059e-01 -5.21166265e-01 6.48366034e-01 1.04090557e-01 -1.60564184e+00 -5.09277999e-01 -1.30411148e+00 1.92654207e-01 -1.83644205e-01 -4.07729328e-01 2.78184503e-01 6.15461051e-01 -1.11276579e+00 8.04198980e-01 -1.01785493e+00 1.34877101e-01 7.64669478e-01 4.08221871e-01 -5.28513193e-01 -2.11734802e-01 -1.44440460e+00 1.08043396e+00 3.43544692e-01 -3.05485606e-01 -1.03190041e+00 -1.05441546e+00 -7.16112375e-01 -3.65508169e-01 9.08921063e-02 -6.06475711e-01 6.80664182e-01 -7.06893742e-01 -1.36621273e+00 1.00851297e+00 5.04745185e-01 -6.96563363e-01 6.26393318e-01 5.21846354e-01 -5.67880929e-01 7.12982774e-01 -1.65014431e-01 8.72926235e-01 8.91168892e-01 -1.31356394e+00 -2.90930063e-01 -5.91198504e-01 -6.76369742e-02 2.08728552e-01 4.73211259e-01 1.33735001e-01 3.66583824e-01 -9.76442754e-01 1.96345374e-01 -7.59577990e-01 -6.10103488e-01 -3.39386821e-01 -2.92124838e-01 6.16046011e-01 6.78238392e-01 -1.19481933e+00 6.12306356e-01 -1.95498383e+00 4.62914288e-01 9.68485996e-02 6.14711225e-01 2.76097029e-01 -1.88521385e-01 -5.22161067e-01 -3.07778567e-01 3.84349406e-01 -7.61860311e-01 2.68706322e-01 -6.60281658e-01 -3.64364497e-02 1.14524579e-02 5.64708054e-01 4.18560132e-02 1.23346698e+00 -8.56886089e-01 -6.04664445e-01 3.43746960e-01 1.12594187e+00 -1.07760571e-01 -2.55955812e-02 5.47102451e-01 1.28636956e+00 -4.61272329e-01 2.61287630e-01 9.21365201e-01 -1.62123188e-01 -8.19558576e-02 -2.95854241e-01 8.88334960e-02 -2.23917633e-01 -8.09054852e-01 1.58577645e+00 -3.92743856e-01 3.56601208e-01 -9.46731791e-02 -7.96764791e-01 6.53528631e-01 2.70102561e-01 7.69889772e-01 -1.23662937e+00 2.72439033e-01 -2.62989011e-02 -1.06610532e-03 -1.69966727e-01 2.43193328e-01 -4.50443268e-01 6.47470057e-02 3.92883480e-01 -6.97793514e-02 -1.77186236e-01 -1.18212305e-01 1.35115460e-01 1.03115165e+00 -2.98058569e-01 3.16824019e-01 -9.52687487e-02 5.80490112e-01 -1.85253814e-01 7.11601600e-02 5.68705797e-01 -3.06752801e-01 1.03701723e+00 3.37372184e-01 -1.27725393e-01 -1.64558387e+00 -1.37671030e+00 -3.54256779e-01 2.11169332e-01 1.81249022e-01 3.86050791e-01 -1.09211385e+00 -6.72984004e-01 -5.69690943e-01 6.19915247e-01 -8.90462339e-01 -2.90944308e-01 -8.38363051e-01 -8.85616183e-01 6.68183148e-01 6.05839908e-01 6.70771658e-01 -1.11545444e+00 -8.62570047e-01 1.32216588e-01 -3.76940072e-01 -1.44116962e+00 -4.49373573e-01 -2.45411620e-01 -1.00174272e+00 -7.14077115e-01 -1.37226224e+00 -7.17314303e-01 7.88605034e-01 1.37491561e-02 1.00061750e+00 1.02706209e-01 -7.27818370e-01 2.54470180e-03 -3.49237502e-01 7.69328922e-02 -8.50585759e-01 -8.13993290e-02 -1.53320014e-01 -4.84788001e-01 -5.40854275e-01 -6.75307333e-01 -8.76956224e-01 3.81365210e-01 -1.38044620e+00 3.48199666e-01 8.82992446e-01 9.88272369e-01 1.24021077e+00 1.68833077e-01 5.54173231e-01 -9.96018052e-01 5.64714432e-01 -2.54263163e-01 -3.83490950e-01 9.10077393e-02 -7.91498125e-01 8.60155076e-02 4.05717880e-01 -5.30439854e-01 -8.88000846e-01 -3.06716084e-01 -2.03230292e-01 -5.82179606e-01 6.45685792e-02 2.69781947e-01 3.18673223e-01 -5.61807096e-01 7.33930111e-01 5.27695954e-01 1.23442784e-01 2.38305368e-02 2.71815181e-01 2.47084290e-01 9.62084591e-01 8.42995495e-02 8.18523347e-01 7.09800899e-01 2.43942738e-01 -2.55966872e-01 -5.74828923e-01 3.81696671e-01 -6.48184299e-01 -2.88893580e-02 1.17767406e+00 -5.83839536e-01 -2.84927189e-01 2.34068573e-01 -5.67197025e-01 -4.11614805e-01 -2.90086955e-01 4.95728225e-01 -6.53384387e-01 1.58924714e-01 -8.12253654e-01 -1.96628883e-01 -8.50161195e-01 -1.44502592e+00 6.70480967e-01 5.01435995e-01 1.67618021e-01 -9.42548752e-01 -5.40904291e-02 5.16928732e-01 9.61652339e-01 1.08403087e+00 9.48205411e-01 -2.02663064e-01 -3.93701762e-01 1.53285190e-01 -4.55935210e-01 3.57400954e-01 1.77265123e-01 -6.25713289e-01 -3.87810946e-01 -5.04914999e-01 2.21157700e-01 -4.43713255e-02 6.60668671e-01 8.76960874e-01 1.19378889e+00 -3.10875922e-01 -1.30968630e-01 8.98724735e-01 1.49248636e+00 2.03042090e-01 1.26773787e+00 2.85298944e-01 5.76512992e-01 6.54520318e-02 4.83179897e-01 1.31093666e-01 9.26234573e-03 6.90794230e-01 4.13576066e-01 -5.07066548e-01 -7.52992272e-01 -2.66959369e-02 1.02821685e-01 4.77532446e-01 -4.72310513e-01 1.51999429e-01 -4.97605056e-01 3.90876442e-01 -1.12772906e+00 -1.18367696e+00 2.07042351e-01 1.86272478e+00 6.38639331e-01 3.86093631e-02 -9.10666659e-02 6.50547594e-02 1.00499153e+00 6.42251000e-02 -7.34065652e-01 -2.14134499e-01 -2.48481318e-01 6.61663949e-01 7.57257819e-01 6.43237054e-01 -7.86444068e-01 6.73565626e-01 6.42482805e+00 8.79833341e-01 -1.32085621e+00 5.61289728e-01 9.82026577e-01 1.29750252e-01 -3.52745801e-01 -5.04430652e-01 -2.03147456e-02 4.11100358e-01 1.03414357e+00 -2.28907645e-01 7.46539712e-01 2.65057862e-01 -1.28853649e-01 -9.38524902e-02 -4.99157786e-01 9.45586085e-01 2.16383152e-02 -1.52442586e+00 3.38784605e-02 1.94064736e-01 9.65866446e-01 -1.21239580e-01 7.62791336e-01 3.20402868e-02 1.22759685e-01 -1.72734821e+00 1.69416547e-01 6.27431750e-01 1.67657816e+00 -1.26883924e+00 8.39105368e-01 -9.40380320e-02 -7.37012327e-01 3.25769871e-01 -3.60498518e-01 9.82263505e-01 1.11280359e-01 3.14594120e-01 -8.01434278e-01 6.14209712e-01 7.06634879e-01 1.02400795e-01 -5.14387667e-01 5.26170671e-01 -1.61478937e-01 1.32605895e-01 1.89421520e-01 6.16281509e-01 9.58005935e-02 -1.28608778e-01 7.75850296e-01 8.16447020e-01 3.27332407e-01 5.44492483e-01 -4.27001387e-01 1.15563071e+00 -4.51845899e-02 -7.03251362e-02 -4.22092110e-01 1.49754182e-01 2.13952661e-01 1.38511884e+00 -1.04295135e+00 -2.74052680e-01 6.18670844e-02 1.10088027e+00 -8.97154734e-02 1.84702083e-01 -1.08006573e+00 -2.30983004e-01 1.37351826e-01 3.35854441e-01 2.41436049e-01 1.41842946e-01 -2.81766176e-01 -9.06183004e-01 -4.95736033e-01 -1.02396715e+00 4.31505710e-01 -9.87229526e-01 -9.47151303e-01 1.18251419e+00 1.03076212e-01 -9.97012556e-01 -3.19900125e-01 -1.82683721e-01 -1.91754580e-01 9.81083930e-01 -1.45628250e+00 -9.59419727e-01 -4.27992225e-01 6.10213876e-01 3.92377496e-01 -1.73523188e-01 6.32113755e-01 2.71787971e-01 -1.12754993e-01 8.18685234e-01 -3.75532731e-02 1.51099026e-01 3.96425068e-01 -1.10854840e+00 3.43828797e-01 8.39943409e-01 -3.49908710e-01 2.10377946e-01 9.50367093e-01 -7.49192476e-01 -9.20319557e-01 -1.13703060e+00 3.56127098e-02 -2.08642840e-01 3.49573791e-01 2.86426723e-01 -9.52870786e-01 4.59687203e-01 2.31432036e-01 2.69643039e-01 3.62119675e-01 -1.01800692e+00 1.40747488e-01 3.87012362e-01 -2.30046940e+00 5.41895151e-01 7.83111572e-01 -2.52743155e-01 -3.89434189e-01 8.57787505e-02 1.03952098e+00 -9.58388031e-01 -1.63069713e+00 7.89494872e-01 2.89424986e-01 -8.65434885e-01 1.33147252e+00 -4.03775722e-01 8.37694705e-01 -8.29009563e-02 -7.56550431e-02 -1.57479918e+00 -3.19589585e-01 -2.10680053e-01 1.28563985e-01 4.95084077e-01 2.61534095e-01 -5.70565522e-01 4.64998573e-01 3.40235919e-01 -1.57624111e-01 -7.36422598e-01 -1.15994430e+00 -4.66687739e-01 4.07778114e-01 2.34490916e-01 7.87694037e-01 1.09434998e+00 -3.22985590e-01 -5.78574054e-02 -2.03739822e-01 1.97477117e-01 9.83403862e-01 -3.45148981e-01 -4.39377725e-02 -7.28898466e-01 -9.18173194e-02 -4.38972175e-01 -5.29089451e-01 -1.98633950e-02 -1.03344567e-01 -1.02625883e+00 -5.02022684e-01 -1.39955986e+00 5.69254458e-01 -9.30367634e-02 -6.45100057e-01 5.24734259e-02 -4.42676172e-02 7.85108328e-01 7.95160383e-02 1.86850354e-01 -2.22617760e-01 2.20993042e-01 2.07875514e+00 -1.58044696e-01 -3.92916612e-02 -1.47883698e-01 -7.66047001e-01 4.04254854e-01 7.87049413e-01 -5.48014939e-01 -3.80034536e-01 -1.49247155e-01 -1.67013288e-01 7.16477394e-01 3.99210155e-01 -1.05707395e+00 -1.85179546e-01 1.45587428e-02 8.43040347e-01 -3.37690413e-01 1.32071674e-01 -4.99226838e-01 4.49349791e-01 8.56366813e-01 -6.37035728e-01 1.84436843e-01 1.29073918e-01 8.38593319e-02 4.69612367e-02 -1.04260266e-01 1.48416889e+00 -2.50332534e-01 -3.36445034e-01 4.03727680e-01 -2.19771877e-01 -1.34100348e-01 1.03555679e+00 -1.19147241e-01 -5.59830248e-01 -2.84242421e-01 -1.02757418e+00 -3.76998514e-01 4.77645338e-01 2.33784303e-01 1.11748123e+00 -1.36168206e+00 -9.71936584e-01 -2.59396993e-02 -1.72754735e-01 -7.54777566e-02 7.58103371e-01 1.21640921e+00 -7.42044151e-01 3.46393198e-01 -8.67671311e-01 -2.32123479e-01 -1.01418674e+00 8.22196782e-01 7.10055292e-01 -8.48939538e-01 -1.13284683e+00 5.80418527e-01 3.04323316e-01 -9.25238431e-02 -2.82236993e-01 2.17323788e-02 -5.77864230e-01 -6.51889086e-01 9.22131479e-01 3.33546102e-01 -1.50413904e-02 -8.76764297e-01 -2.30815724e-01 4.62084889e-01 -3.44317406e-01 -4.44665074e-01 1.40702438e+00 -2.74917185e-01 1.58739284e-01 -2.97413379e-01 9.59474027e-01 -2.43732288e-01 -1.18928087e+00 -3.08546245e-01 -4.31476682e-01 -2.88280100e-01 6.53865159e-01 -1.06401801e+00 -1.58523488e+00 5.66916227e-01 1.50748837e+00 6.19162731e-02 1.22879565e+00 1.41445771e-01 1.07005465e+00 -5.40962815e-01 3.64336848e-01 -6.73589349e-01 8.41249600e-02 -3.83796059e-02 9.17620897e-01 -1.15996647e+00 -3.35717052e-02 -4.02602017e-01 -8.93325210e-01 6.30593002e-01 4.17756766e-01 -3.16148281e-01 2.42136315e-01 7.27968633e-01 -3.24887712e-03 -1.96540758e-01 -2.83406287e-01 3.33397947e-02 2.99247563e-01 8.95805836e-01 3.09927523e-01 4.00143564e-02 -5.87008357e-01 5.33816636e-01 -2.44158685e-01 -1.27701476e-01 8.30863833e-01 5.45778692e-01 -2.28740215e-01 -5.69120586e-01 -4.38430965e-01 2.80103266e-01 -9.73202109e-01 -4.84343320e-02 -5.87295592e-02 7.33644843e-01 -2.53444672e-01 6.33442163e-01 -1.28074782e-02 -3.75925273e-01 1.25541657e-01 -5.69527268e-01 1.03297353e+00 -6.33750781e-02 -3.05625856e-01 -1.26960054e-01 -5.27361989e-01 -5.45136631e-01 -5.52268982e-01 -2.48213366e-01 -1.67104638e+00 -3.12623978e-01 1.62954554e-01 -1.51093841e-01 9.35232639e-01 8.02345872e-01 1.15836032e-01 1.23296905e+00 8.03434610e-01 -6.69034600e-01 -2.16055542e-01 -8.77596617e-01 -5.16020358e-01 4.28128511e-01 3.07561308e-01 -6.06945574e-01 -1.28671885e-01 -1.66254282e-01]
[13.739777565002441, -2.2646074295043945]
897ba227-343b-4ae4-9e79-6766ce538195
bert-hlstms-bert-and-hierarchical-lstms-for
2012.02128
null
https://arxiv.org/abs/2012.02128v1
https://arxiv.org/pdf/2012.02128v1.pdf
BERT-hLSTMs: BERT and Hierarchical LSTMs for Visual Storytelling
Visual storytelling is a creative and challenging task, aiming to automatically generate a story-like description for a sequence of images. The descriptions generated by previous visual storytelling approaches lack coherence because they use word-level sequence generation methods and do not adequately consider sentence-level dependencies. To tackle this problem, we propose a novel hierarchical visual storytelling framework which separately models sentence-level and word-level semantics. We use the transformer-based BERT to obtain embeddings for sentences and words. We then employ a hierarchical LSTM network: the bottom LSTM receives as input the sentence vector representation from BERT, to learn the dependencies between the sentences corresponding to images, and the top LSTM is responsible for generating the corresponding word vector representations, taking input from the bottom LSTM. Experimental results demonstrate that our model outperforms most closely related baselines under automatic evaluation metrics BLEU and CIDEr, and also show the effectiveness of our method with human evaluation.
['Mian Zhou', 'Frank Guerin', 'Qingyun Dai', 'Jing Su']
2020-12-03
null
null
null
null
['visual-storytelling']
['natural-language-processing']
[ 1.67084590e-01 7.30734095e-02 -4.60610203e-02 -1.60714999e-01 -6.87340915e-01 -4.06738281e-01 1.07934165e+00 2.02560257e-02 -8.08810219e-02 6.92904055e-01 7.06514359e-01 -1.14679851e-01 5.48588216e-01 -8.73723209e-01 -8.13738823e-01 -6.17455542e-01 1.78989246e-01 3.46724600e-01 7.16589019e-02 -7.43399039e-02 3.75282198e-01 5.16580828e-02 -1.43750155e+00 8.59898567e-01 5.66937506e-01 7.38078356e-01 6.43938959e-01 7.36361802e-01 -6.56771600e-01 1.53556824e+00 -8.90319645e-01 -3.60437006e-01 -3.60715657e-01 -9.19043064e-01 -7.36870944e-01 2.78219074e-01 2.76558608e-01 -4.49770600e-01 -3.89549226e-01 8.72338295e-01 3.99281293e-01 1.36496559e-01 6.65044367e-01 -1.26528096e+00 -1.09983778e+00 1.01944411e+00 -3.34946275e-01 -2.53282726e-01 5.86432397e-01 3.43090862e-01 1.22052336e+00 -1.12737691e+00 1.09697998e+00 1.36002493e+00 1.85589314e-01 7.63039887e-01 -1.15264630e+00 -4.27311987e-01 2.60231525e-01 5.37944794e-01 -1.33991659e+00 -1.65336236e-01 8.87706757e-01 -5.74130356e-01 1.04134548e+00 -7.41103590e-02 9.35202360e-01 1.65364063e+00 1.30238533e-01 1.07415688e+00 8.13794374e-01 -3.79792303e-01 1.90380692e-01 1.69806913e-01 -2.81370282e-01 7.71201551e-01 -3.18466395e-01 -1.00696892e-01 -7.80087769e-01 1.35315493e-01 9.83069599e-01 -1.59199581e-01 -9.48655978e-02 -5.06747007e-01 -1.55818689e+00 1.04908633e+00 5.23926973e-01 5.81974745e-01 -5.65467656e-01 5.40291131e-01 6.95213497e-01 3.12950350e-02 4.01786238e-01 3.79450053e-01 3.15572292e-01 2.55369395e-02 -1.08493197e+00 4.72061396e-01 4.37441021e-01 1.24339187e+00 3.98392677e-01 4.05377477e-01 -9.46281433e-01 7.38280296e-01 2.32858285e-01 2.42754504e-01 6.67622805e-01 -8.09911251e-01 5.82344890e-01 4.39138174e-01 2.09529519e-01 -8.91044080e-01 6.72489852e-02 -1.87879190e-01 -8.23735416e-01 9.71776247e-02 -1.69726625e-01 8.89060497e-02 -1.03051937e+00 1.64088666e+00 -1.57989070e-01 6.08240552e-02 1.62715912e-01 7.97148407e-01 1.35990059e+00 1.14971673e+00 2.78077066e-01 -2.01406017e-01 1.31016588e+00 -1.29143906e+00 -9.59015846e-01 -3.23784381e-01 5.11566579e-01 -4.91145045e-01 1.48154414e+00 2.06118338e-02 -1.43376100e+00 -6.21250331e-01 -1.22120440e+00 -3.30732018e-01 -2.78219432e-01 3.05246204e-01 -1.27198845e-02 -1.49101093e-01 -1.36865807e+00 4.03403610e-01 -3.59959096e-01 -2.50187248e-01 4.40337032e-01 -3.66139650e-01 -2.17929706e-01 1.02243602e-01 -1.27183008e+00 8.97796631e-01 5.54921389e-01 -1.04946852e-01 -1.42588270e+00 -3.81046623e-01 -1.36034358e+00 2.87500232e-01 4.34531458e-02 -1.12177622e+00 1.37077761e+00 -9.54379261e-01 -1.47182834e+00 8.18157375e-01 -3.53869110e-01 -6.36125267e-01 5.49865901e-01 5.18171787e-02 6.92642555e-02 2.79158860e-01 2.98907250e-01 1.23978198e+00 8.85235012e-01 -1.60030270e+00 -3.25136423e-01 2.25442305e-01 2.67658178e-02 2.74168879e-01 -3.42020124e-01 -1.52893960e-01 -3.34093720e-01 -8.90800893e-01 -4.10388798e-01 -3.89402241e-01 -3.24538261e-01 -2.12861717e-01 -7.02559233e-01 -3.10131103e-01 6.56343758e-01 -6.90855682e-01 1.18438995e+00 -2.12077188e+00 3.78843755e-01 -2.85521716e-01 1.85203150e-01 1.05984069e-01 -4.84438449e-01 8.17536652e-01 -3.10307350e-02 1.80884957e-01 -2.13197827e-01 -7.50678778e-01 3.35518532e-02 -1.05079135e-03 -7.07468987e-01 -1.42144918e-01 2.99677163e-01 1.25823867e+00 -1.09515488e+00 -7.76193619e-01 3.58050972e-01 6.90914929e-01 -2.40670085e-01 5.41584432e-01 -6.87587321e-01 1.69394433e-01 -2.23157406e-01 9.52066407e-02 1.30890116e-01 -4.13383573e-01 1.28980726e-01 -7.24162385e-02 -1.84489623e-01 5.06258667e-01 -5.08069813e-01 2.03402019e+00 -7.02265084e-01 9.93282974e-01 -6.04034007e-01 -7.10134506e-01 1.03058326e+00 6.54331267e-01 -1.13419920e-01 -7.68111289e-01 8.28174055e-02 -1.39388189e-01 -5.33382416e-01 -7.55009174e-01 6.81286037e-01 -4.66694921e-01 -2.81602621e-01 4.91957784e-01 1.78591728e-01 -3.22178990e-01 3.18172216e-01 6.04989231e-01 8.57301176e-01 4.92078036e-01 3.13550621e-01 3.00483495e-01 5.01269817e-01 1.03796877e-01 4.88304347e-02 6.55317247e-01 1.72357306e-01 8.55205059e-01 7.46149957e-01 -4.66505289e-01 -1.31293213e+00 -1.20919240e+00 6.39007270e-01 9.43515956e-01 4.09851857e-02 -6.15360618e-01 -8.84965718e-01 -6.74986839e-01 -4.34365422e-01 1.38191128e+00 -7.80781448e-01 -1.25398055e-01 -5.54313779e-01 3.59383784e-02 3.17749828e-01 5.81677437e-01 4.45061535e-01 -1.70989847e+00 -7.56148040e-01 3.82669359e-01 -5.22015989e-01 -1.24009979e+00 -4.64192718e-01 -4.45075244e-01 -5.98251820e-01 -4.86100882e-01 -1.20651567e+00 -1.04784405e+00 6.65485144e-01 1.28500953e-01 1.22660136e+00 -1.36234492e-01 -2.50281572e-01 1.11284204e-01 -4.85657066e-01 -1.88702062e-01 -7.23725498e-01 -6.84021711e-02 -5.00574648e-01 5.26854359e-02 -7.16520175e-02 -5.25049269e-01 -3.68973732e-01 -2.82800823e-01 -1.01871324e+00 8.07705879e-01 5.37669122e-01 8.37175190e-01 3.13493282e-01 -5.15171885e-01 5.29314458e-01 -5.93092799e-01 9.09680843e-01 -2.51202017e-01 -1.81444809e-01 2.97665477e-01 1.38153753e-03 2.64911503e-01 8.08744729e-01 -3.85235459e-01 -1.03808784e+00 -6.16781041e-03 -6.76053688e-02 -6.49615943e-01 -1.41553819e-01 3.96067113e-01 -1.25247583e-01 6.02144480e-01 3.99641722e-01 6.94209874e-01 -3.33323777e-01 -1.23708509e-01 8.05988550e-01 4.66572285e-01 6.07625186e-01 -1.05673261e-01 6.74449444e-01 3.40412170e-01 -4.05372351e-01 -7.52890289e-01 -7.62942553e-01 1.38673976e-01 -4.01694924e-01 -5.88551700e-01 1.17127132e+00 -9.13848698e-01 -5.11423409e-01 9.56528783e-02 -1.88559306e+00 -3.61623138e-01 -5.40392876e-01 7.56947175e-02 -9.43264961e-01 1.71685681e-01 -6.38347685e-01 -8.24816465e-01 -4.55975860e-01 -1.13099396e+00 1.22619069e+00 4.35992889e-02 -5.17942727e-01 -1.00004590e+00 3.88440415e-02 1.33946568e-01 2.37516686e-01 4.98839974e-01 1.23013365e+00 -2.66221136e-01 -5.85523188e-01 -1.85939705e-03 -3.60837311e-01 4.90119569e-02 -1.70820430e-01 -8.07926059e-02 -8.07457626e-01 3.68048027e-02 -1.44084975e-01 -5.54435611e-01 9.81870949e-01 3.72792512e-01 1.04964721e+00 -5.86408079e-01 -2.25879133e-01 2.19136715e-01 1.35239840e+00 2.63160378e-01 7.61558533e-01 2.40292981e-01 9.08529580e-01 8.75642478e-01 4.10362333e-01 5.60909986e-01 5.81650496e-01 6.31397784e-01 4.06765074e-01 -1.75996438e-01 -4.25440550e-01 -9.04437721e-01 6.86303437e-01 8.64833295e-01 1.65868551e-01 -5.22282302e-01 -6.88460648e-01 8.83901298e-01 -1.98221397e+00 -1.22686303e+00 -8.39856360e-03 1.82982004e+00 7.03145683e-01 5.71419857e-02 1.00371599e-01 -4.55745347e-02 7.35277236e-01 6.95185363e-01 -2.83143550e-01 -6.38281763e-01 -1.43272653e-01 -6.26701191e-02 -4.93019521e-02 5.74763238e-01 -6.59100235e-01 1.39229453e+00 6.38185883e+00 8.04530144e-01 -1.06780767e+00 2.34230742e-01 4.98560995e-01 -3.66976500e-01 -7.17113912e-01 -1.67306408e-01 -2.83940792e-01 2.89001971e-01 7.33238399e-01 -4.53398198e-01 1.55278608e-01 6.98866189e-01 5.77160299e-01 1.89503446e-01 -1.18726742e+00 1.07482862e+00 5.06568253e-01 -1.74658966e+00 8.28614295e-01 -2.59851903e-01 7.16162503e-01 -4.30988163e-01 7.06979446e-03 1.73658445e-01 5.24507463e-01 -1.16394520e+00 1.30759442e+00 5.61316669e-01 8.73093605e-01 -7.34906614e-01 4.25870419e-01 2.89077431e-01 -1.28547931e+00 2.59919196e-01 -2.17187703e-01 -9.66484845e-02 7.59393513e-01 2.28402928e-01 -1.09939420e+00 3.27356249e-01 2.60863960e-01 9.34052229e-01 -3.19100738e-01 7.53420055e-01 -8.54671955e-01 3.03797245e-01 6.91560268e-01 -4.82906073e-01 5.16866982e-01 1.26762763e-01 5.86302638e-01 1.43378484e+00 5.31405985e-01 -2.43172288e-01 1.40934633e-02 1.32122648e+00 -1.85521722e-01 2.09757164e-01 -1.05669272e+00 -3.95487159e-01 2.62629807e-01 1.16689074e+00 -6.13839209e-01 -8.24802160e-01 -6.00873418e-02 1.41043031e+00 3.86946917e-01 6.02046847e-01 -9.40376580e-01 -5.14333844e-01 3.18040997e-01 1.23486735e-01 3.55865479e-01 -2.77653188e-01 -2.72283345e-01 -1.01167905e+00 -1.43127292e-02 -5.38913488e-01 7.86041692e-02 -1.55470884e+00 -9.84414279e-01 9.49214518e-01 5.92630543e-02 -1.29619443e+00 -7.59022892e-01 -2.74613529e-01 -8.19017887e-01 6.13837659e-01 -1.15202487e+00 -1.31499541e+00 -2.08576247e-01 2.87903577e-01 1.22657263e+00 -1.16517507e-01 7.18208611e-01 -1.45591035e-01 -2.36297056e-01 1.97288334e-01 -3.99634719e-01 1.77993968e-01 3.82208437e-01 -1.13953578e+00 8.91966343e-01 6.77423418e-01 4.41022307e-01 3.14951509e-01 9.40818250e-01 -6.37225211e-01 -9.54469144e-01 -1.18182743e+00 1.44008255e+00 -9.17802453e-02 6.05759561e-01 -7.93998301e-01 -6.30916655e-01 7.05274999e-01 8.61416996e-01 -5.28074205e-01 4.96517092e-01 -5.82736075e-01 -5.40042877e-01 3.49262714e-01 -7.54917800e-01 9.27201569e-01 1.09895158e+00 -6.31554842e-01 -8.00315678e-01 3.71024251e-01 1.11916637e+00 -1.13734357e-01 -2.93617308e-01 -1.38286844e-01 4.52161938e-01 -9.03948128e-01 8.73116016e-01 -5.34660101e-01 1.33660209e+00 -1.69385314e-01 1.80999085e-01 -1.49229908e+00 -3.18286180e-01 -4.41570878e-01 1.05333984e-01 1.28927362e+00 5.23873568e-01 1.13635091e-02 6.22088194e-01 -1.33182174e-02 -2.31306866e-01 -5.30885398e-01 -6.66617393e-01 -7.40773439e-01 -1.09546088e-01 -3.66564900e-01 5.17360687e-01 7.64470875e-01 2.50350863e-01 8.72036040e-01 -7.58019865e-01 -3.93901974e-01 4.78590637e-01 1.72717497e-01 6.76808834e-01 -6.36768222e-01 -1.54550716e-01 -6.24538481e-01 -2.19228819e-01 -1.02155793e+00 3.84781688e-01 -9.36828434e-01 5.27455866e-01 -2.26539040e+00 4.41583365e-01 4.75014299e-01 1.45471171e-01 4.49606419e-01 -9.82230678e-02 2.49509409e-01 6.23008549e-01 2.93817669e-02 -6.94669724e-01 1.04667437e+00 1.44031608e+00 -4.65543509e-01 -1.13597356e-01 -6.42915070e-01 -4.96526808e-01 6.21594191e-01 4.80823159e-01 -3.41575861e-01 -5.25512695e-01 -7.35428393e-01 1.93915918e-01 3.36889863e-01 5.57988226e-01 -9.60098922e-01 6.61912039e-02 -2.04984307e-01 3.61277819e-01 -6.74275100e-01 4.02643532e-01 -2.25110784e-01 4.80062962e-02 5.01967192e-01 -9.80158865e-01 2.06662714e-01 -9.32945404e-03 2.61638373e-01 -3.94827724e-01 -1.31570026e-01 5.27627885e-01 -4.07361031e-01 -5.13638258e-01 1.43651620e-01 -7.36176789e-01 -1.13018036e-01 1.05856240e+00 -6.49110302e-02 -2.36168429e-01 -8.55248630e-01 -7.28426874e-01 2.54121423e-01 3.56659502e-01 7.33576298e-01 1.15473032e+00 -1.88377142e+00 -9.79264736e-01 -5.91171794e-02 3.55843633e-01 -1.14040576e-01 2.80315012e-01 1.54433951e-01 -4.79914039e-01 4.20349687e-01 -2.68083036e-01 -4.27443773e-01 -1.24997485e+00 7.10803688e-01 3.22197974e-02 -3.97671849e-01 -7.11803377e-01 9.60234821e-01 6.15025759e-01 9.61700603e-02 2.06530929e-01 -2.65066236e-01 -4.99833196e-01 1.85960069e-01 8.13336551e-01 3.85468714e-02 -4.30287123e-01 -7.41081238e-01 -4.71490026e-02 3.84560138e-01 -6.10699914e-02 -8.13536584e-01 1.23292422e+00 -5.23860641e-02 6.64357767e-02 8.59030783e-01 1.12675869e+00 -5.73035479e-01 -1.17227173e+00 -2.97500938e-02 1.23130782e-02 -2.28657603e-01 -2.82450706e-01 -5.38733840e-01 -8.22886288e-01 1.41663933e+00 6.54456839e-02 2.31688306e-01 8.23896587e-01 1.82337582e-01 1.04333997e+00 3.25388797e-02 1.25814170e-01 -7.94835389e-01 6.54749453e-01 4.86420840e-01 1.41990864e+00 -6.33881330e-01 -5.06246328e-01 -7.74281844e-02 -1.05775297e+00 9.03108835e-01 5.69783688e-01 -1.50225326e-01 -2.51080096e-01 1.05575681e-01 2.94958055e-02 -1.83412492e-01 -1.14308524e+00 -2.04742447e-01 1.26475796e-01 4.69403118e-01 5.42691112e-01 -2.29839925e-02 -3.93625289e-01 4.44984078e-01 -2.93980926e-01 1.89311460e-01 6.17899835e-01 7.10194111e-01 -5.18773317e-01 -1.02107382e+00 -2.11135387e-01 -5.39392084e-02 -4.53423895e-02 -3.40434581e-01 -7.92215407e-01 5.09788215e-01 -7.72635564e-02 6.70692146e-01 3.13624471e-01 -4.55785990e-01 2.94384629e-01 2.43730888e-01 5.06835699e-01 -8.45835924e-01 -5.88792801e-01 -1.77944358e-02 1.04317904e-01 -4.80278105e-01 -2.68817574e-01 -4.44181353e-01 -1.35892773e+00 -1.31513579e-02 3.15294951e-01 2.30473951e-01 6.56767964e-01 1.00147533e+00 2.88264990e-01 7.95659542e-01 6.17012680e-01 -1.11299133e+00 -9.95692611e-02 -1.03358948e+00 -2.44491652e-01 5.71928680e-01 3.34010154e-01 -2.72324532e-01 -6.26785830e-02 4.18767452e-01]
[11.180764198303223, 0.7461451292037964]
0c3c7568-3f77-4bc1-8f18-74f8c79c2c46
a-step-towards-digital-operations-a-novel
2306.07772
null
https://arxiv.org/abs/2306.07772v1
https://arxiv.org/pdf/2306.07772v1.pdf
A step towards digital operations -- A novel grey-box approach for modelling the heat dynamics of Ultra-low temperature freezing chambers
Ultra-low temperature (ULT) freezers store perishable bio-contents and have high energy consumption, which highlight a demand for reliable methods for intelligent surveillance and smart energy management. This study introduces a novel grey-box modelling approach based on stochastic differential equations to describe the heat dynamics of the ULT freezing chambers. The proposed modelling approach only requires temperature data measured by the embedded sensors and uses data from the regular operation periods for model identification. The model encompasses three states: chamber temperature, envelope temperature, and local evaporator temperature. Special attention is given to the local evaporator temperature state, which is modelled as a time-variant system, to characterize the time delay and dynamic variations in cooling intensity. Two ULT freezers with different operational patterns are modelled. The unknown model parameters are estimated using the maximum likelihood method. The results demonstrate that the models can accurately predict the chamber temperature measured by the control probe (RMSE < 0.19 {\deg}C) and are promising to be applied for forecasting future states. In addition, the model for local evaporator temperature can effectively adapt to different operational patterns and provide insight into the local cooling supply status. The proposed approach greatly promotes the practical feasibility of grey-box modelling of the heat dynamics for ULT freezers and can serve several potential digital applications. A major limitation of the modelling approach is the low identifiability, which can potentially be addressed by inferring model parameters based on relative parameter changes.
['Wiebke Brix Markussen', "Francesco D'Ettorre", 'Jan Kloppenborg Møller', 'Peder Bacher', 'Tao Huang']
2023-06-13
null
null
null
null
['energy-management']
['time-series']
[ 1.34078190e-01 -4.61513638e-01 -5.64046204e-02 -4.83794101e-02 -5.34700379e-02 -7.50763893e-01 4.36861515e-01 2.74249882e-01 4.14589979e-02 6.41099215e-01 -5.11252940e-01 -3.40556115e-01 -1.19053684e-01 -7.60010481e-01 -3.91255975e-01 -1.11089969e+00 -1.45197436e-01 7.98371881e-02 3.00915297e-02 -1.16654215e-02 -4.00368795e-02 5.51311314e-01 -1.45344591e+00 -1.98682576e-01 9.27403033e-01 1.04799283e+00 4.11479443e-01 5.70482075e-01 4.29090589e-01 3.01644087e-01 -7.15618849e-01 5.44795036e-01 1.64049685e-01 -3.69022250e-01 1.55959437e-02 -1.93876430e-01 -6.49368763e-01 -4.05319273e-01 3.29313576e-01 6.09160185e-01 1.69867724e-01 1.18098304e-01 6.67130947e-01 -8.64932120e-01 1.15765974e-01 -1.44952774e-01 -6.60784310e-03 -7.57694766e-02 2.43975803e-01 3.89728576e-01 1.55658171e-01 -1.49288569e-02 1.85791478e-01 8.48695636e-01 6.82066441e-01 2.51908749e-01 -1.33658242e+00 -4.35696363e-01 7.66273290e-02 1.86211854e-01 -1.43224204e+00 -1.80663079e-01 6.30092025e-01 -4.58457649e-01 1.09053457e+00 6.76949799e-01 1.18236005e+00 5.98775983e-01 8.77618670e-01 1.58116415e-01 1.63521945e+00 -1.93588495e-01 6.17384017e-01 2.78027266e-01 -2.65825659e-01 1.94296598e-01 2.48271674e-01 5.15854299e-01 1.27806097e-01 -2.36306846e-01 4.91128564e-01 6.81290254e-02 -3.90075207e-01 3.35557833e-02 -4.77817774e-01 4.43798184e-01 2.03394115e-01 3.63298059e-01 -4.88103211e-01 -2.76741125e-02 2.62713552e-01 2.05490038e-01 6.78725839e-01 2.89175153e-01 -6.58837020e-01 -7.49048069e-02 -1.25378704e+00 -9.81551558e-02 9.10341442e-01 7.73691416e-01 4.21976596e-01 2.84831583e-01 1.07606232e-01 2.53635645e-01 6.67625427e-01 1.33012390e+00 5.39747998e-02 -8.15300822e-01 -1.53413147e-01 4.54768270e-01 6.34960890e-01 -8.83952558e-01 -5.65356374e-01 3.67463157e-02 -7.31657684e-01 -3.73189598e-02 2.51007918e-02 -5.59505999e-01 -8.00420761e-01 1.18360901e+00 5.55440724e-01 -1.36527151e-01 3.35126482e-02 7.92521596e-01 3.08237761e-01 1.28143954e+00 1.50634527e-01 -9.97621179e-01 1.24835014e+00 -1.07910270e-02 -1.23659194e+00 -6.77938610e-02 3.19395363e-01 -4.59977388e-01 2.47724935e-01 6.83531538e-02 -8.05244029e-01 -2.38960713e-01 -8.81745934e-01 6.03053927e-01 -5.14366865e-01 2.11934686e-01 4.15072851e-02 7.79795885e-01 -8.62493932e-01 6.56328440e-01 -1.38970411e+00 -5.10636091e-01 -4.48197454e-01 1.77600384e-01 4.03655589e-01 3.26284498e-01 -1.52614057e+00 1.20419049e+00 2.34795079e-01 9.37813103e-01 -9.64009643e-01 -6.89676404e-01 -8.24264288e-01 -5.37665710e-02 2.77245224e-01 -1.79655075e-01 9.63065207e-01 -2.30211645e-01 -1.97753704e+00 2.39034027e-01 -4.42609668e-01 -2.08431199e-01 4.33775127e-01 1.80757884e-02 -6.77054584e-01 1.18475787e-01 -4.75892782e-01 -4.86932129e-01 6.27309263e-01 -1.09920824e+00 -2.57610261e-01 -3.73139888e-01 -5.02100945e-01 5.88640124e-02 4.33487371e-02 2.39633229e-02 1.47929072e-01 -5.55184148e-02 1.46682924e-02 -9.77718234e-01 -2.37902075e-01 -2.79046416e-01 3.31030823e-02 -1.51657651e-03 1.00264597e+00 -9.45286334e-01 1.37859917e+00 -1.62733221e+00 -5.91487996e-03 5.20527005e-01 -6.07677042e-01 3.53563607e-01 6.15006506e-01 8.52485001e-01 1.93976209e-01 1.90001085e-01 -5.20134926e-01 2.11083531e-01 -1.08126126e-01 2.59793937e-01 1.00903235e-01 8.21772456e-01 1.56962082e-01 4.03970540e-01 -9.83662963e-01 4.27732691e-02 7.80617356e-01 6.73622847e-01 2.41445392e-01 2.68787473e-01 -1.88549235e-01 5.24090946e-01 -5.61933994e-01 6.37749434e-01 7.32907593e-01 3.90752912e-01 6.05253875e-01 -2.32497573e-01 -6.32588446e-01 -1.36824965e-01 -1.24881768e+00 9.25266385e-01 -6.94732785e-01 4.06707793e-01 7.44408786e-01 -8.03969800e-01 1.40066350e+00 5.35254598e-01 5.73235393e-01 -8.19916487e-01 2.57517427e-01 1.64492190e-01 -3.48081887e-01 -7.95995235e-01 4.71256554e-01 -5.55064976e-01 7.17944950e-02 1.28499970e-01 -3.60133886e-01 -7.15665758e-01 -1.53114036e-01 -4.65492696e-01 5.44771969e-01 3.99509847e-01 1.39804929e-01 -9.21713769e-01 5.87512016e-01 -8.90042633e-02 9.03116584e-01 -3.50777246e-02 -2.08216488e-01 -2.81252086e-01 2.91622460e-01 -3.32641095e-01 -9.82047319e-01 -3.57049644e-01 -5.61140180e-01 2.30078474e-01 5.71432829e-01 -1.79947868e-01 -6.95405006e-01 2.14046225e-01 -1.13806073e-02 9.93794203e-01 -4.92608398e-01 -4.28089350e-01 -6.37719572e-01 -1.18434763e+00 -1.35135219e-01 3.28149229e-01 3.69806081e-01 -5.11747062e-01 -1.25089455e+00 4.77310032e-01 -2.96218723e-01 -8.47371697e-01 9.34565738e-02 3.02866042e-01 -1.12135172e+00 -1.16308010e+00 -2.22388938e-01 1.14346281e-01 6.62838876e-01 -1.71505660e-01 9.98887837e-01 5.30437790e-02 -3.89955372e-01 4.62570250e-01 -3.45486492e-01 -5.64766467e-01 -6.81952775e-01 -4.04148340e-01 2.54078060e-01 -8.71547386e-02 3.51189971e-01 9.90064070e-02 -8.90024006e-01 8.82412970e-01 -9.43439662e-01 -2.27497041e-01 -3.95021327e-02 4.06562179e-01 7.78383076e-01 4.72188115e-01 3.41255456e-01 -6.47378325e-01 5.57607293e-01 -5.11968017e-01 -1.00227034e+00 6.41728401e-01 -8.32957864e-01 -8.48927163e-03 6.44364774e-01 -2.48129472e-01 -1.31843436e+00 5.45245185e-02 4.67650324e-01 -2.38255739e-01 -6.88824952e-02 3.47320080e-01 -2.19902799e-01 1.23514548e-01 9.83325467e-02 2.53228635e-01 1.07398964e-01 -4.02379483e-01 -2.51990974e-01 6.06800079e-01 9.44664702e-02 -6.42176211e-01 8.23374212e-01 3.88391256e-01 4.59307581e-01 -1.16452575e+00 1.45551875e-01 -2.33708143e-01 -3.94575924e-01 -7.23324835e-01 7.40620494e-01 -8.85264754e-01 -1.01104796e+00 8.90977859e-01 -5.51338911e-01 -7.17728019e-01 -1.53444335e-01 4.04229999e-01 -2.46887803e-01 2.66433507e-01 -3.91436428e-01 -1.81945050e+00 -6.70717537e-01 -9.59928870e-01 8.49367380e-01 5.08498371e-01 -7.87797868e-02 -1.42390370e+00 1.23191729e-01 -2.04937588e-02 7.03005195e-01 9.23911452e-01 5.43958545e-01 3.71417791e-01 -3.91912609e-01 -3.85110646e-01 6.76575541e-01 3.05836856e-01 3.29448014e-01 5.26735961e-01 -8.34829271e-01 -9.32815492e-01 6.72092676e-01 2.22438559e-01 5.30598998e-01 6.28465950e-01 6.64719284e-01 -1.69889495e-01 -4.89808649e-01 6.33567154e-01 1.89052153e+00 7.47647345e-01 3.26902658e-01 2.85147667e-01 -1.30088544e-02 7.14132965e-01 9.64446604e-01 8.06416750e-01 7.79608637e-02 3.80426198e-01 6.69120789e-01 -8.92085284e-02 8.47212315e-01 -9.16080773e-02 5.84269285e-01 4.32421803e-01 -1.99632958e-01 -2.13500068e-01 -9.15474117e-01 4.65556473e-01 -1.34734797e+00 -8.93976927e-01 -4.98553485e-01 2.66912889e+00 4.85004932e-01 -4.19069529e-01 -2.05320626e-01 5.66804968e-02 7.47366011e-01 6.31962046e-02 -7.20700026e-01 -9.81880069e-01 1.16127238e-01 3.36262435e-02 9.42040563e-01 3.55634540e-01 -8.03630173e-01 2.86785930e-01 6.50058222e+00 3.22686613e-01 -1.11280465e+00 -2.87772477e-01 6.37388885e-01 -1.56484336e-01 -4.14531752e-02 2.59263724e-01 -7.06268609e-01 8.51216078e-01 1.72931933e+00 -2.87499845e-01 4.03693974e-01 2.94865310e-01 1.03675163e+00 -7.51618803e-01 -7.73603857e-01 1.94076911e-01 -3.52140158e-01 -6.31558716e-01 -6.20037496e-01 3.13806050e-02 5.53338051e-01 -2.72647172e-01 -1.54638067e-01 -6.41114488e-02 -3.84275503e-02 -4.24566448e-01 4.17822033e-01 1.02472818e+00 9.49482679e-01 -7.47803867e-01 8.89720738e-01 6.75632060e-01 -1.46939254e+00 -3.00943196e-01 -3.84188175e-01 -5.18272892e-02 4.24737930e-01 7.05024004e-01 -6.44872308e-01 8.55228722e-01 7.91255593e-01 2.25675926e-01 -1.54189289e-01 7.56432712e-01 -1.59683958e-01 7.74034321e-01 -1.11459780e+00 -3.77718568e-01 -2.27579027e-01 -8.27204764e-01 3.96882862e-01 8.24913025e-01 4.61754978e-01 3.58026743e-01 -1.40210718e-01 9.19527352e-01 7.94727504e-01 -1.84055701e-01 -3.10997337e-01 1.77536264e-01 6.11512780e-01 1.14526427e+00 -6.63791478e-01 -1.25683650e-01 1.29746730e-02 5.64763248e-01 -8.03936899e-01 5.64849317e-01 -9.38458562e-01 -2.46468604e-01 5.34119785e-01 4.04723912e-01 2.73242500e-02 -2.48501301e-01 -1.14222057e-02 -9.63570356e-01 1.03389949e-01 -2.52756834e-01 1.39342006e-02 -5.54210186e-01 -7.20036983e-01 2.32480615e-01 4.72415447e-01 -1.12906682e+00 -5.60812771e-01 -4.54902738e-01 -8.98360848e-01 1.11186290e+00 -1.64112437e+00 -7.66900897e-01 -2.43250072e-01 3.48780036e-01 9.01640803e-02 5.70406675e-01 8.79052162e-01 -1.62396207e-01 -1.09685111e+00 -4.66312421e-03 1.18055546e+00 -4.78907615e-01 2.87449718e-01 -1.01811361e+00 1.66079383e-02 8.47209930e-01 -1.11274731e+00 5.04483819e-01 1.16127825e+00 -8.91942978e-01 -1.87071669e+00 -1.04102838e+00 5.33464372e-01 -1.89360350e-01 4.74748433e-01 -5.45378506e-01 -1.02975392e+00 6.23065885e-03 5.80737740e-02 -1.90023035e-01 4.23508286e-01 -4.72926736e-01 6.18684828e-01 -3.57326090e-01 -1.76711214e+00 -1.47524655e-01 -6.64137080e-02 -4.67984676e-01 -1.09658718e-01 3.01124185e-01 1.12025447e-01 -3.94281954e-01 -1.33681142e+00 5.46576679e-01 5.84111333e-01 -5.42329669e-01 4.30193216e-01 3.17416973e-02 -3.22558939e-01 -4.03934628e-01 2.14431703e-01 -1.37443674e+00 -5.87539673e-02 -7.12872684e-01 -3.46826106e-01 1.36571348e+00 1.53500102e-02 -7.51976252e-01 4.03706104e-01 1.43146062e+00 2.54179358e-01 -8.01688850e-01 -1.34311593e+00 -9.51368511e-01 8.27776641e-03 -8.70428607e-02 6.04591787e-01 5.94645619e-01 1.08817622e-01 -3.87161553e-01 -3.46409678e-01 5.37120581e-01 7.27763712e-01 8.22245628e-02 1.68716073e-01 -1.20031297e+00 7.37312213e-02 2.46499076e-01 2.33718306e-02 -3.36988151e-01 -8.91384035e-02 -1.36102185e-01 3.60590369e-01 -1.51468277e+00 -7.65598491e-02 -2.70060718e-01 -2.14298502e-01 1.19902901e-02 -2.27871031e-01 -3.69903058e-01 2.30782881e-01 1.09869584e-01 2.18363523e-01 8.62549365e-01 7.92923987e-01 -2.20752154e-02 -4.79279280e-01 2.62314886e-01 3.75240266e-01 3.42108965e-01 1.04418325e+00 -3.81764114e-01 -1.73528150e-01 -1.47700787e-01 1.19181141e-01 4.01718557e-01 3.53177756e-01 -1.14530194e+00 -3.72100957e-02 -4.61935729e-01 3.57533753e-01 -5.54863751e-01 1.61528841e-01 -1.58760071e+00 1.02993178e+00 8.49604964e-01 3.06816965e-01 -2.70518865e-02 5.21536827e-01 6.79055274e-01 1.53504014e-01 -1.35966301e-01 8.25088620e-01 -8.16688165e-02 -2.34853074e-01 -1.06421188e-01 -9.88257706e-01 -6.39710009e-01 1.45678556e+00 -4.63030308e-01 -2.03350946e-01 -9.52384993e-02 -6.63393795e-01 8.50344479e-01 6.99027598e-01 -1.36364596e-02 1.84418872e-01 -8.95687401e-01 -4.72490638e-01 3.50090772e-01 -4.10318583e-01 -2.18342822e-02 6.24323726e-01 7.34656513e-01 -6.48737669e-01 5.15918076e-01 8.77176225e-02 -6.54455066e-01 -1.04423964e+00 6.19380832e-01 9.28599715e-01 -2.52517015e-01 -1.97554693e-01 2.04329416e-01 -3.20504516e-01 -1.52116477e-01 -5.09146094e-01 -7.16536701e-01 9.81779844e-02 2.89149016e-01 3.65097344e-01 6.73768520e-01 1.45974278e-01 -6.30613029e-01 -4.34055954e-01 7.51973212e-01 5.90023994e-01 3.09556037e-01 1.39622426e+00 -4.09556478e-01 -1.92448348e-01 8.73960018e-01 8.90782595e-01 -7.93244481e-01 -1.56797361e+00 1.49154484e-01 -2.23111480e-01 -1.96604103e-01 3.98721904e-01 -9.66616690e-01 -9.80943203e-01 6.59410059e-01 9.61614490e-01 4.82273668e-01 1.61740446e+00 -4.72102284e-01 3.55251849e-01 2.06383929e-01 5.19904017e-01 -1.58004546e+00 -9.55495179e-01 2.57688135e-01 7.12095737e-01 -7.08198547e-01 3.56086761e-01 -1.68046609e-01 -1.75309196e-01 1.20096135e+00 3.34085554e-01 1.47244871e-01 7.24122047e-01 5.34102738e-01 2.71873511e-02 1.93535373e-01 -7.36330807e-01 3.31063688e-01 -2.66242504e-01 4.31574941e-01 3.16726893e-01 4.38583642e-01 -4.44429725e-01 1.13266066e-01 4.39377666e-01 6.20362163e-02 4.72155273e-01 1.10634255e+00 -3.15126717e-01 -8.54489148e-01 -9.56963420e-01 3.48785549e-01 -2.01459467e-01 3.59011710e-01 -6.12750910e-02 7.38934755e-01 2.49149669e-02 1.41990268e+00 1.21210337e-01 -1.75322190e-01 6.24734521e-01 3.04760695e-01 -7.12465495e-02 -1.58172883e-02 -7.96122849e-01 2.15807304e-01 -2.28312805e-01 -5.78292847e-01 -5.92199504e-01 -9.70779061e-01 -9.99161780e-01 -4.16288882e-01 -7.60731995e-01 4.53402013e-01 1.17102802e+00 6.39668405e-01 1.72472849e-01 4.90720093e-01 1.21124041e+00 -8.73054445e-01 -4.93384033e-01 -9.83489037e-01 -1.01471233e+00 -1.08302034e-01 4.16800380e-01 -6.83868527e-01 -8.04384112e-01 -1.89257070e-01]
[5.738037586212158, 2.48437762260437]
271d87fe-fd8e-496a-be67-32c061a44cdf
supervised-video-summarization-via-multiple
2104.11530
null
https://arxiv.org/abs/2104.11530v2
https://arxiv.org/pdf/2104.11530v2.pdf
Supervised Video Summarization via Multiple Feature Sets with Parallel Attention
The assignment of importance scores to particular frames or (short) segments in a video is crucial for summarization, but also a difficult task. Previous work utilizes only one source of visual features. In this paper, we suggest a novel model architecture that combines three feature sets for visual content and motion to predict importance scores. The proposed architecture utilizes an attention mechanism before fusing motion features and features representing the (static) visual content, i.e., derived from an image classification model. Comprehensive experimental evaluations are reported for two well-known datasets, SumMe and TVSum. In this context, we identify methodological issues on how previous work used these benchmark datasets, and present a fair evaluation scheme with appropriate data splits that can be used in future work. When using static and motion features with parallel attention mechanism, we improve state-of-the-art results for SumMe, while being on par with the state of the art for the other dataset.
['Ralph Ewerth', 'Sherzod Hakimov', 'Junaid Ahmed Ghauri']
2021-04-23
null
null
null
null
['supervised-video-summarization', 'automated-feature-engineering']
['computer-vision', 'methodology']
[ 3.02078545e-01 -2.09971860e-01 -3.69904995e-01 -2.52904147e-01 -6.87713861e-01 -3.62690836e-01 8.18572819e-01 4.06375527e-01 -6.82253420e-01 7.33725548e-01 8.30771029e-01 1.65518224e-01 -1.79284647e-01 -4.25934643e-01 -4.69600767e-01 -6.48531735e-01 -2.62595892e-01 -1.37660978e-02 5.43599069e-01 -5.28120762e-03 8.23353946e-01 4.10361916e-01 -2.01819158e+00 6.95960045e-01 4.18072701e-01 9.92131770e-01 2.47402310e-01 1.09434855e+00 -1.00424491e-01 1.36227906e+00 -8.94778311e-01 -1.56547487e-01 -8.91231596e-02 -6.69928551e-01 -1.14983606e+00 1.30642712e-01 7.99260139e-01 -3.35660696e-01 -2.59305596e-01 6.04158461e-01 5.21518648e-01 2.91185021e-01 9.05649364e-01 -1.14322567e+00 -6.38497531e-01 3.81543487e-01 -5.84704697e-01 8.49371433e-01 5.39718151e-01 1.79201320e-01 1.23037910e+00 -7.20506608e-01 8.67435932e-01 9.19465780e-01 4.14466888e-01 3.75210851e-01 -6.97508991e-01 -2.42101904e-02 2.06365004e-01 9.89149749e-01 -1.02420104e+00 -5.47233045e-01 7.79828310e-01 -6.38475001e-01 1.00083077e+00 7.86068738e-01 8.56951892e-01 9.97556031e-01 2.54441470e-01 1.01605475e+00 8.98895800e-01 -3.17580074e-01 1.87098935e-01 -9.06097591e-02 3.70483875e-01 4.77028847e-01 3.42711769e-02 -2.40551829e-01 -6.48090184e-01 1.07047059e-01 3.37346643e-01 -3.03627938e-01 -5.25739729e-01 -3.55599076e-01 -1.34967089e+00 7.60952532e-01 3.34721446e-01 6.60025418e-01 -6.30370259e-01 2.67651349e-01 6.48089290e-01 2.56030768e-01 4.70890909e-01 2.03904271e-01 -1.53030798e-01 -4.11802620e-01 -1.40513563e+00 2.79941887e-01 5.67675292e-01 5.53365469e-01 4.16379809e-01 -3.78785469e-02 -8.01712990e-01 6.29029214e-01 2.08247140e-01 -7.16694519e-02 6.39818490e-01 -1.13418376e+00 3.83063287e-01 4.19324011e-01 1.28765702e-01 -1.05974889e+00 -3.37804675e-01 -2.36253187e-01 -6.24390602e-01 2.23581195e-01 3.33881706e-01 1.05438828e-01 -8.73705387e-01 1.50212085e+00 1.42430648e-01 2.41723835e-01 -3.32804322e-02 1.16139019e+00 1.18222201e+00 8.38689148e-01 3.88590503e-03 -4.38167065e-01 1.24550891e+00 -1.40263367e+00 -8.63759756e-01 2.53304634e-02 2.58928746e-01 -8.12196255e-01 8.64395678e-01 4.74957496e-01 -1.39207876e+00 -7.55922496e-01 -1.17222679e+00 -2.87834793e-01 -1.66867018e-01 2.73388356e-01 4.18801427e-01 2.39108846e-01 -1.32132494e+00 7.63925076e-01 -6.05135024e-01 -6.61598206e-01 3.04101586e-01 2.81553835e-01 -3.34588617e-01 1.47333011e-01 -8.76095295e-01 9.36136246e-01 4.19131428e-01 -1.75404117e-01 -7.84454823e-01 -2.98274010e-01 -6.28209233e-01 1.93840697e-01 2.12343931e-01 -8.65195096e-01 1.14075124e+00 -1.27447498e+00 -1.31962752e+00 8.76088917e-01 -1.97519735e-01 -6.35647833e-01 6.02701187e-01 -4.17357534e-01 -1.53561845e-01 6.15623474e-01 -1.25284865e-02 7.72709370e-01 9.05220032e-01 -1.17753935e+00 -8.39732170e-01 -9.90264211e-03 3.07018399e-01 3.45186621e-01 -5.10387897e-01 3.20315927e-01 -6.93613350e-01 -8.00096691e-01 -4.53582317e-01 -5.97285926e-01 2.48415321e-02 -9.89471301e-02 -5.49714379e-02 -2.15985522e-01 7.29156315e-01 -8.36148202e-01 1.79749644e+00 -1.82521451e+00 6.27622664e-01 -2.51004905e-01 7.14976192e-02 4.19812918e-01 -1.31459579e-01 4.25781876e-01 -8.35365281e-02 1.07631333e-01 -1.54932424e-01 -3.30888540e-01 -2.22556397e-01 -1.02237336e-01 1.77780628e-01 4.55234617e-01 3.46603483e-01 6.78323448e-01 -7.52933443e-01 -7.17602015e-01 3.69136006e-01 4.07212347e-01 -5.37200451e-01 2.27665916e-01 8.60187188e-02 2.83410877e-01 -3.32231492e-01 3.88474792e-01 3.76307398e-01 -1.80442166e-02 -8.14205408e-02 -3.57999891e-01 -3.36187333e-01 8.03321507e-03 -1.01404250e+00 1.70455277e+00 -2.26548180e-01 1.06960380e+00 -2.73071647e-01 -1.01920187e+00 6.86646044e-01 3.84318113e-01 6.74418747e-01 -3.73669982e-01 2.99008280e-01 -1.66668773e-01 8.78182054e-02 -9.02353942e-01 9.37728345e-01 3.21616262e-01 6.78496733e-02 1.77968159e-01 3.31438333e-01 2.46098980e-01 6.80376232e-01 1.19403578e-01 1.05064583e+00 2.62168616e-01 7.11905241e-01 -2.92868733e-01 9.62025464e-01 -1.43906847e-02 3.00498635e-01 6.05710328e-01 -5.46676099e-01 1.01453507e+00 6.86632156e-01 -5.10152042e-01 -1.24653530e+00 -6.67415798e-01 9.32896063e-02 1.10069239e+00 1.08433522e-01 -7.04821885e-01 -8.03781211e-01 -7.43395030e-01 -3.54643911e-01 5.76126516e-01 -9.17440355e-01 1.23696469e-01 -6.55614078e-01 -7.59414375e-01 1.47873297e-01 5.94385326e-01 2.15516344e-01 -1.33640039e+00 -1.05664766e+00 2.67148823e-01 -3.55642945e-01 -8.61662447e-01 -4.41484302e-01 -1.75344184e-01 -7.11405277e-01 -1.09018147e+00 -1.08914936e+00 -5.22216380e-01 1.11275114e-01 5.18450439e-01 1.34763861e+00 3.35400105e-01 -2.11773202e-01 4.09378439e-01 -7.59956658e-01 -2.61945575e-01 -2.33389020e-01 9.49327722e-02 -4.95942235e-01 1.91381648e-01 1.53816799e-02 -3.36976409e-01 -9.21128273e-01 1.44033924e-01 -1.00047100e+00 2.22253978e-01 5.73846996e-01 6.90500796e-01 3.85574669e-01 -3.95093352e-01 5.02334893e-01 -5.32682121e-01 5.38516700e-01 -6.41654611e-01 4.16845083e-02 1.11802109e-01 -1.73093081e-01 8.26435257e-03 4.91540045e-01 -1.45470485e-01 -9.06984627e-01 -1.80653736e-01 -2.51831293e-01 -4.20828730e-01 -1.83108732e-01 4.17712063e-01 8.28362480e-02 1.65495679e-01 3.80528450e-01 1.69370875e-01 -2.80867130e-01 -4.38241452e-01 2.42139086e-01 5.91477752e-01 5.57465315e-01 -1.37081698e-01 1.34816363e-01 4.64493841e-01 -8.41635391e-02 -9.73070323e-01 -5.80899119e-01 -7.50177324e-01 -7.16836751e-01 -5.31466901e-01 1.06682479e+00 -6.84758127e-01 -3.86890680e-01 3.76135707e-01 -1.22023988e+00 -7.64842033e-02 -2.90137142e-01 4.87682462e-01 -7.43192911e-01 6.44011319e-01 -5.10765374e-01 -8.40362370e-01 -4.39866334e-01 -1.09160006e+00 1.05127752e+00 1.87055811e-01 -4.71294612e-01 -9.36584175e-01 2.32695416e-01 5.81030846e-01 3.94024223e-01 4.67043400e-01 4.99478698e-01 -5.40108144e-01 -4.96765703e-01 -6.37884215e-02 -1.70593306e-01 2.86158890e-01 -1.94777310e-01 4.49492007e-01 -1.03550255e+00 -2.70204067e-01 -1.54234588e-01 -2.80003756e-01 1.49483263e+00 6.16984248e-01 1.25118232e+00 -2.22163752e-01 5.31752175e-03 2.55157202e-01 1.43467700e+00 1.18416429e-01 8.18941593e-01 6.65250599e-01 5.48755705e-01 6.02264583e-01 9.23507512e-01 6.75393522e-01 4.70536798e-01 8.96220505e-01 5.27766168e-01 -4.95179892e-02 -4.47193503e-01 3.59593421e-01 4.59392101e-01 9.58704412e-01 -7.74109364e-01 -7.06134796e-01 -6.44215226e-01 8.91529799e-01 -2.15073395e+00 -1.44517744e+00 -3.65307331e-01 2.08591533e+00 3.76159102e-01 6.73048720e-02 4.31403637e-01 4.34149981e-01 6.77331924e-01 6.87821448e-01 -1.11685686e-01 -6.10743165e-01 -9.63200703e-02 -3.67495455e-02 2.42043734e-01 3.49303335e-01 -1.36072385e+00 4.82852608e-01 6.39764071e+00 8.36254537e-01 -1.14433551e+00 2.29965642e-01 6.53545082e-01 -3.86777788e-01 -6.22796416e-02 -1.99042499e-01 -3.13347846e-01 6.43963039e-01 1.04168212e+00 -1.03814863e-01 -3.35579365e-02 4.65322405e-01 3.64483565e-01 -3.18422794e-01 -9.33008134e-01 8.40271711e-01 6.08802855e-01 -1.40094984e+00 -1.52541352e-02 -1.28801644e-01 6.83829308e-01 -2.54526865e-02 -1.35311499e-01 1.50436265e-02 -3.30336541e-01 -8.53879213e-01 9.63224173e-01 8.61675680e-01 4.56712574e-01 -7.46918440e-01 1.04423153e+00 1.19591855e-01 -1.22240126e+00 -1.10344641e-01 -1.84520528e-01 -1.09298348e-01 2.61645883e-01 1.04582518e-01 -3.55462551e-01 9.11274195e-01 7.66235828e-01 9.91019487e-01 -8.61858547e-01 1.41354501e+00 -1.16732657e-01 5.93049467e-01 1.36071339e-01 -4.32699732e-02 4.64898348e-01 2.05046758e-01 6.81805432e-01 1.60176241e+00 5.40605962e-01 -2.69533724e-01 -1.99497547e-02 2.28988782e-01 1.72419593e-01 4.09432262e-01 -4.20282036e-01 2.03763232e-01 5.96863218e-02 1.40420628e+00 -8.34214687e-01 -7.08954513e-01 -5.97754776e-01 1.12318790e+00 3.67056161e-01 2.42583323e-02 -9.95021045e-01 -2.39702597e-01 3.49071115e-01 -1.32943718e-02 5.87738752e-01 2.50946898e-02 -1.00177459e-01 -1.27621078e+00 -9.36768055e-02 -6.34171963e-01 5.73735714e-01 -8.52814019e-01 -8.91851366e-01 6.89967632e-01 3.02509934e-01 -1.42418480e+00 -4.58050877e-01 -3.71962786e-01 -8.87758970e-01 4.86654043e-01 -1.51275635e+00 -1.01583314e+00 -4.53441948e-01 3.17636847e-01 1.01221669e+00 -1.18669428e-01 5.30517638e-01 1.87856153e-01 -5.12874722e-01 3.65587711e-01 1.05250724e-01 -1.85185060e-01 7.23879993e-01 -1.25078332e+00 2.99952656e-01 8.28825057e-01 2.63590336e-01 1.13879003e-01 7.92344093e-01 -2.13148311e-01 -1.04471040e+00 -8.01791847e-01 9.71753418e-01 -2.94177771e-01 4.31991577e-01 1.38172641e-01 -9.49550152e-01 3.79614145e-01 9.78435040e-01 -1.10867918e-01 6.85518980e-01 -2.15516746e-01 1.85132742e-01 1.47842243e-01 -9.95831966e-01 4.27790880e-01 1.01870561e+00 -8.79266113e-03 -7.93304920e-01 -1.30763546e-01 4.50973064e-01 -9.22430679e-02 -9.15046215e-01 2.94763327e-01 7.64397621e-01 -1.48467600e+00 9.77434635e-01 -4.51374382e-01 8.48523438e-01 -3.32013637e-01 -1.46913305e-01 -1.28747213e+00 -5.80716372e-01 -3.40477854e-01 -4.44087744e-01 1.32496357e+00 7.40405917e-02 4.05612774e-02 6.66388452e-01 4.38790061e-02 -3.29725295e-01 -8.00734460e-01 -8.52863729e-01 -3.82274300e-01 -1.14684671e-01 -1.93499267e-01 1.81882545e-01 6.93044364e-01 -1.21452719e-01 5.66402972e-01 -8.27237308e-01 -2.48018935e-01 2.88650334e-01 1.06906831e-01 7.20350087e-01 -1.02240431e+00 -1.62587926e-01 -7.23171234e-01 -8.56939733e-01 -6.03903234e-01 6.60863742e-02 -6.39501750e-01 -2.42460132e-01 -1.91747200e+00 6.19925499e-01 3.99485797e-01 -4.64854836e-01 2.34461829e-01 -4.55856681e-01 6.08848155e-01 7.77193785e-01 2.60415435e-01 -1.04090786e+00 5.30274570e-01 1.19064820e+00 -2.02966526e-01 -3.24713551e-02 -2.21927285e-01 -5.19124568e-01 8.46630573e-01 9.56035674e-01 -2.20874429e-01 -4.02984709e-01 -4.28891987e-01 -1.56500295e-01 1.31715879e-01 2.81447411e-01 -1.40635073e+00 5.38962148e-02 -1.17643498e-01 5.33095300e-01 -8.78628433e-01 3.60940903e-01 -5.70101500e-01 4.71001677e-02 4.68212694e-01 -4.58740264e-01 1.42173722e-01 1.25336662e-01 4.80966419e-01 -4.92120236e-01 -5.73152959e-01 7.13850498e-01 -1.50969908e-01 -1.20905030e+00 1.29426410e-02 -5.88592350e-01 -6.93160072e-02 1.13812923e+00 -3.82005900e-01 -2.54106969e-01 -5.54270744e-01 -7.36675739e-01 7.94987082e-02 5.31718791e-01 5.50368488e-01 6.87588692e-01 -1.41283226e+00 -1.08229530e+00 -1.95969060e-01 2.00596333e-01 -5.57785928e-01 5.54470122e-01 9.10611808e-01 -7.01935768e-01 4.56139892e-01 -6.84917748e-01 -6.09641314e-01 -1.71808052e+00 6.35227740e-01 -2.46674851e-01 -4.50326443e-01 -3.47662807e-01 4.27334875e-01 3.43211927e-02 3.61732811e-01 1.44141734e-01 -2.86038548e-01 -9.68474686e-01 6.16645873e-01 7.77384281e-01 6.42753065e-01 6.78002909e-02 -1.04251420e+00 -4.12264109e-01 7.04380810e-01 5.80000400e-04 -9.51576009e-02 1.39704514e+00 -4.20518875e-01 2.02131450e-01 4.89767045e-01 1.26457441e+00 -3.83104421e-02 -1.31506228e+00 1.50923833e-01 -2.19099992e-03 -6.56051755e-01 1.17656231e-01 -5.88616610e-01 -1.13040483e+00 9.99123037e-01 7.42532015e-01 4.28369403e-01 1.43590105e+00 -5.89458384e-02 4.49465185e-01 -1.46514043e-01 -5.87495454e-02 -1.01761460e+00 2.76048541e-01 3.48309278e-01 1.22203124e+00 -1.24973094e+00 2.87250280e-01 1.02420198e-02 -9.35667038e-01 1.15847456e+00 4.55516100e-01 -2.01685384e-01 7.81183913e-02 -7.54738674e-02 -1.40151322e-01 -9.21030641e-02 -1.06185758e+00 -4.81005192e-01 7.02126324e-01 4.66757447e-01 9.65354443e-01 -2.70335674e-01 -9.99384701e-01 4.40224886e-01 1.48224846e-01 2.14356020e-01 7.23541915e-01 9.83944654e-01 -6.68046772e-01 -9.12303209e-01 -4.83335882e-01 4.27681178e-01 -9.46840346e-01 -9.65271890e-03 -5.03346741e-01 8.37489128e-01 2.15735361e-01 9.42357719e-01 1.92329809e-01 -2.63950229e-01 3.07100803e-01 -1.67529270e-01 5.71508586e-01 -2.94515163e-01 -8.95089924e-01 8.62062499e-02 3.71051699e-01 -6.60564780e-01 -1.07671964e+00 -8.12949181e-01 -8.25660884e-01 -3.06239992e-01 1.10519469e-01 -8.72268677e-02 3.82047176e-01 6.14649236e-01 2.68596381e-01 9.26578760e-01 5.13901770e-01 -1.31322134e+00 -8.61007571e-02 -1.02652168e+00 -2.56991059e-01 7.66740024e-01 3.81720692e-01 -6.57666624e-01 -1.30269095e-01 3.12673002e-01]
[10.456260681152344, 0.45033958554267883]
0f8e5610-4001-4b67-8f95-adeb45f8f425
text-segmentation-by-cross-segment-attention
2004.14535
null
https://arxiv.org/abs/2004.14535v2
https://arxiv.org/pdf/2004.14535v2.pdf
Text Segmentation by Cross Segment Attention
Document and discourse segmentation are two fundamental NLP tasks pertaining to breaking up text into constituents, which are commonly used to help downstream tasks such as information retrieval or text summarization. In this work, we propose three transformer-based architectures and provide comprehensive comparisons with previously proposed approaches on three standard datasets. We establish a new state-of-the-art, reducing in particular the error rates by a large margin in all cases. We further analyze model sizes and find that we can build models with many fewer parameters while keeping good performance, thus facilitating real-world applications.
['Gonçalo Simões', 'Kishore Papineni', 'Boris Dadachev', 'Michal Lukasik']
2020-04-30
null
https://aclanthology.org/2020.emnlp-main.380
https://aclanthology.org/2020.emnlp-main.380.pdf
emnlp-2020-11
['discourse-segmentation']
['natural-language-processing']
[ 5.25713265e-01 4.11774814e-01 -5.63858032e-01 -2.55021483e-01 -1.30890954e+00 -8.91292274e-01 7.90683270e-01 4.76745635e-01 -4.07602966e-01 8.05890799e-01 6.46768630e-01 -7.16401875e-01 1.08082600e-01 -4.17888373e-01 -4.40495402e-01 -4.22623217e-01 1.23969004e-01 7.93549538e-01 5.67212820e-01 -2.74556309e-01 6.08949363e-01 7.71080256e-02 -1.19243443e+00 7.91873574e-01 1.15409732e+00 6.98572636e-01 -4.71616127e-02 8.90773237e-01 -5.11822045e-01 1.01083720e+00 -9.94912386e-01 -4.42533225e-01 -2.03214705e-01 -4.09371942e-01 -1.54327476e+00 5.58677549e-03 3.86136651e-01 -2.25777566e-01 -2.36251593e-01 7.37336159e-01 4.74084944e-01 3.09461325e-01 6.83579743e-01 -6.53597474e-01 -4.26228195e-01 1.18965852e+00 -5.69212019e-01 4.95807588e-01 5.16442657e-01 -3.01756620e-01 1.48947942e+00 -5.27754486e-01 6.95283294e-01 1.33979142e+00 4.61217552e-01 6.61715746e-01 -9.87684965e-01 -1.03005022e-01 5.94619751e-01 3.03986102e-01 -8.49876583e-01 -8.14648688e-01 5.27316689e-01 -1.92554623e-01 1.49445677e+00 8.43859911e-01 3.45505595e-01 8.54349792e-01 5.16504869e-02 1.39228368e+00 5.57699442e-01 -5.12571812e-01 -1.60441501e-03 -3.67516965e-01 5.47577381e-01 4.42301303e-01 1.63997281e-02 -8.64712417e-01 -5.58209240e-01 -2.88682524e-02 9.81952175e-02 -6.57554626e-01 -4.23344225e-01 3.87155652e-01 -1.31576526e+00 8.89913917e-01 1.50642797e-01 6.03524804e-01 -7.13208020e-02 -6.67213947e-02 7.10463703e-01 1.70853406e-01 6.90812528e-01 5.65723240e-01 -4.42373395e-01 -3.28809172e-01 -1.11834788e+00 3.13661665e-01 1.05205965e+00 1.18571568e+00 2.44101778e-01 -1.21936314e-01 -6.69617116e-01 8.48219454e-01 6.33757859e-02 7.36982673e-02 6.51368797e-01 -9.96043921e-01 1.08701050e+00 6.99783683e-01 -9.53785330e-02 -5.98988295e-01 -5.06533682e-01 -1.74864411e-01 -9.35986698e-01 -8.61045003e-01 3.51297736e-01 -1.99011311e-01 -8.49684715e-01 1.43906450e+00 2.40727946e-01 -1.66106090e-01 1.83291778e-01 4.11562443e-01 1.13032985e+00 1.11309600e+00 -1.40967695e-02 -6.34008467e-01 1.55739748e+00 -1.36581075e+00 -1.00249422e+00 -6.52854443e-01 9.67553854e-01 -9.33143497e-01 9.69328463e-01 4.61110651e-01 -1.41333270e+00 -1.58274576e-01 -8.29526126e-01 -7.37994492e-01 -1.54201567e-01 2.12818474e-01 6.35398865e-01 4.85405505e-01 -1.08506680e+00 6.39232993e-01 -8.43875945e-01 -3.05312634e-01 4.49739397e-01 2.89647043e-01 -7.84534886e-02 3.33883137e-01 -1.06922996e+00 9.28989172e-01 5.44458032e-01 1.49535403e-01 -4.43241030e-01 -5.35872579e-01 -7.85603642e-01 2.35320315e-01 4.11693633e-01 -7.87271142e-01 1.87815762e+00 -2.88034946e-01 -1.68741775e+00 9.72990274e-01 -6.80179477e-01 -7.88863063e-01 4.20289457e-01 -5.35626173e-01 -1.42453998e-01 1.05883062e-01 1.42010584e-01 5.36620855e-01 4.12966311e-01 -7.25560904e-01 -8.26909721e-01 -2.32057959e-01 2.12070823e-01 5.58484197e-01 -4.15265381e-01 3.39726716e-01 -7.73906410e-01 -6.64029062e-01 5.27526438e-02 -6.50530457e-01 -2.34095037e-01 -6.85686111e-01 -1.05650043e+00 -7.18407154e-01 5.98211825e-01 -8.51814985e-01 1.70442283e+00 -1.53555489e+00 5.92842042e-01 -4.49377179e-01 3.62930715e-01 3.90741825e-01 -6.51167855e-02 5.16064465e-01 3.37493777e-01 4.60554361e-01 -3.49708080e-01 -5.52620173e-01 9.84462574e-02 1.24146277e-02 -4.93959248e-01 2.83776551e-01 1.26869634e-01 1.09693360e+00 -6.65715337e-01 -8.14466774e-01 -1.19953342e-02 -6.00696802e-02 -4.24137414e-01 1.71957731e-01 -6.36341751e-01 2.60044008e-01 -6.56892836e-01 3.34700376e-01 2.10434079e-01 -2.49478221e-01 3.54993731e-01 -3.19384970e-02 -7.51546249e-02 1.13243377e+00 -6.84885442e-01 1.81674123e+00 -2.99300343e-01 1.13495147e+00 1.77298322e-01 -1.25552976e+00 5.78656793e-01 3.26804399e-01 2.53504634e-01 -6.10511601e-01 4.09397632e-01 1.34904906e-01 4.88820933e-02 -4.31128860e-01 1.05641246e+00 2.00615391e-01 -3.12595248e-01 6.33212864e-01 -8.51489082e-02 -1.87428713e-01 8.80645275e-01 4.34517622e-01 1.01906729e+00 -3.23522538e-01 3.79286170e-01 -4.74151552e-01 7.38406658e-01 2.02900797e-01 3.93631041e-01 6.88327014e-01 -6.85080886e-02 4.26589727e-01 8.03875089e-01 -1.29469693e-01 -8.48065317e-01 -5.58601439e-01 -1.16918728e-01 1.48875654e+00 1.46698742e-03 -7.11364210e-01 -1.15152836e+00 -7.25152731e-01 -3.87438804e-01 9.69486177e-01 -3.85088354e-01 1.99197665e-01 -1.23543119e+00 -9.49372113e-01 8.56197238e-01 5.89661598e-01 3.82667959e-01 -9.38042045e-01 -3.79049033e-01 2.84766674e-01 -7.58458257e-01 -1.32703173e+00 -5.08633018e-01 6.82343319e-02 -1.14158356e+00 -1.07279909e+00 -6.79518878e-01 -1.03993917e+00 2.94304073e-01 3.16420943e-01 1.30451882e+00 1.34371951e-01 3.75760794e-02 -1.35310143e-01 -4.31354493e-01 -3.02935213e-01 -5.94474971e-01 9.02470529e-01 -4.58705485e-01 -5.11026919e-01 2.03562260e-01 -1.26581743e-01 -4.02814478e-01 -6.46281568e-03 -8.87391448e-01 2.82987982e-01 2.72670120e-01 6.03195965e-01 4.31979090e-01 -1.11831896e-01 6.87348187e-01 -1.27122426e+00 1.08395135e+00 -2.25472167e-01 -3.35820168e-01 4.39130753e-01 -3.65373582e-01 1.82868034e-01 6.58067942e-01 -1.50823757e-01 -1.29600382e+00 -5.86170375e-01 -4.59571779e-01 5.29237747e-01 2.14619841e-02 6.56678021e-01 -1.16095789e-01 4.72463250e-01 5.11269450e-01 1.75650954e-01 -4.96734351e-01 -6.99086189e-01 6.25879824e-01 8.88039231e-01 5.11721611e-01 -6.16045177e-01 3.52931887e-01 3.33796352e-01 -3.04733276e-01 -9.50697541e-01 -1.23518789e+00 -6.59010887e-01 -5.68420172e-01 1.48976862e-01 7.36796618e-01 -6.66086972e-01 -5.97238004e-01 4.26621348e-01 -1.51967180e+00 -3.16599518e-01 -9.98542011e-02 6.72134198e-03 -3.89823556e-01 6.54593289e-01 -1.12443686e+00 -4.82935816e-01 -9.48678315e-01 -1.03383589e+00 1.14999306e+00 4.01305497e-01 -4.41601217e-01 -1.17084777e+00 1.01650171e-01 6.08936071e-01 1.87356308e-01 -6.80381507e-02 1.13306725e+00 -1.19395804e+00 -2.35502526e-01 1.28292367e-01 -1.45439968e-01 1.49519682e-01 -3.70604265e-03 9.17255059e-02 -8.05380106e-01 -2.72824705e-01 3.08230855e-02 -1.50061220e-01 1.08395863e+00 4.21510160e-01 1.20013475e+00 -5.14439940e-01 -5.52654028e-01 3.60121220e-01 9.40068841e-01 1.04516901e-01 6.83480680e-01 4.49086666e-01 6.63065791e-01 8.32899809e-01 6.44025207e-01 9.42051783e-02 4.91205245e-01 4.64798361e-01 -2.31942832e-02 4.02595438e-02 -1.39476046e-01 -1.29282817e-01 1.88245595e-01 1.39699173e+00 1.96434647e-01 -8.88314366e-01 -9.69090044e-01 7.73781300e-01 -2.14688849e+00 -8.32868636e-01 -5.76287687e-01 1.57775164e+00 1.18444777e+00 3.37132603e-01 -1.48389861e-01 1.94529429e-01 7.38252640e-01 6.47298634e-01 -4.06305283e-01 -6.92729175e-01 -1.45325422e-01 1.36315033e-01 2.74145901e-01 6.59241140e-01 -1.34026170e+00 1.34113061e+00 7.65547085e+00 1.05903363e+00 -7.97177076e-01 -7.10978433e-02 7.19768763e-01 -2.38227576e-01 -3.90347064e-01 1.56911090e-02 -1.03210330e+00 2.20771655e-01 1.02306843e+00 -5.86447418e-01 9.90838930e-02 4.76365596e-01 1.71292394e-01 -1.86181337e-01 -1.26549792e+00 5.35965443e-01 2.03508899e-01 -1.74319065e+00 3.79690230e-01 -2.94682443e-01 7.63562977e-01 -2.72883475e-02 -3.99280876e-01 3.08934838e-01 2.12614417e-01 -9.83033001e-01 5.63543797e-01 -1.49761103e-02 5.44097781e-01 -5.40213406e-01 5.96817553e-01 6.40638292e-01 -1.09391689e+00 1.35450572e-01 -2.06461564e-01 -3.55504565e-02 5.42110324e-01 5.49238145e-01 -7.97098696e-01 7.83067763e-01 4.57268685e-01 6.64516509e-01 -3.97742301e-01 8.22247684e-01 -5.17236829e-01 7.92954266e-01 -3.46451759e-01 -2.47543767e-01 5.14640033e-01 -2.77724508e-02 6.77006125e-01 1.57924438e+00 1.94048462e-03 9.96638760e-02 5.71996607e-02 3.20621133e-01 -5.40300667e-01 2.68030673e-01 -1.44201458e-01 -2.78179776e-02 5.65656722e-01 1.16418779e+00 -8.01118910e-01 -6.74599171e-01 -1.69212446e-01 8.16674113e-01 3.75069320e-01 1.90325081e-01 -7.80172527e-01 -5.72913110e-01 2.73480922e-01 -1.89502999e-01 3.17888528e-01 -1.35275781e-01 -2.94196606e-01 -1.31291544e+00 2.03472629e-01 -1.06520307e+00 4.85769004e-01 -2.90478736e-01 -8.91468108e-01 7.03841507e-01 6.90422505e-02 -8.43315780e-01 -6.04614019e-01 -4.50514078e-01 -7.50925720e-01 5.37828982e-01 -1.72594392e+00 -1.05854762e+00 3.61052565e-02 2.75239140e-01 1.05414510e+00 9.59432870e-02 7.87292302e-01 1.88861758e-01 -9.79371905e-01 4.93638635e-01 4.72793192e-01 3.49640369e-01 6.76094294e-01 -1.36470520e+00 8.93877923e-01 1.05728197e+00 3.38586926e-01 4.25867319e-01 7.20037580e-01 -3.96128654e-01 -1.25973594e+00 -8.61035883e-01 1.32726312e+00 -2.90605634e-01 6.89698517e-01 -3.58677804e-01 -9.85331774e-01 9.39656198e-01 6.00369155e-01 -6.87913179e-01 7.19461322e-01 3.54660630e-01 -1.87560067e-01 1.81456193e-01 -7.83368468e-01 6.50479257e-01 1.02709222e+00 -3.40498686e-01 -9.94018972e-01 5.46569109e-01 9.97327447e-01 -8.06475699e-01 -7.98358202e-01 2.02818483e-01 2.61711329e-01 -8.46429288e-01 7.20764935e-01 -9.35461938e-01 6.70716763e-01 3.42430398e-02 1.93530858e-01 -1.14410055e+00 -2.28213266e-01 -9.55914855e-01 -4.65741605e-01 1.58925951e+00 6.30736530e-01 -4.71796989e-01 8.17080140e-01 4.60448146e-01 -4.87761796e-01 -8.27377439e-01 -1.10729992e+00 -4.57035422e-01 5.99117219e-01 -4.01840866e-01 4.64289844e-01 6.58834040e-01 4.15516943e-01 9.86911297e-01 -1.85548849e-02 -3.13219815e-01 2.21003056e-01 4.60304588e-01 5.27922928e-01 -1.08773482e+00 1.16149127e-01 -9.34896290e-01 2.27557838e-01 -1.86320400e+00 3.42246741e-01 -8.58487844e-01 2.05884367e-01 -2.07474995e+00 3.01783830e-01 -1.49074584e-01 1.33697376e-01 3.26499283e-01 -3.42204958e-01 -6.83858842e-02 7.07032084e-02 2.95627952e-01 -1.03377342e+00 6.50163949e-01 1.14410114e+00 -4.53599423e-01 -2.10602418e-01 5.95858805e-02 -1.12907314e+00 8.92222822e-01 8.62528265e-01 -3.77017200e-01 -3.79682779e-01 -1.11844659e+00 2.04358146e-01 3.19537744e-02 -3.62305492e-01 -5.14835894e-01 4.41233963e-01 -1.95778087e-01 -9.94200781e-02 -1.00247741e+00 9.65059176e-02 -4.25305814e-02 -6.26457751e-01 1.22471452e-01 -6.26259804e-01 4.92473773e-04 3.65576446e-01 2.55115628e-01 -4.82875466e-01 -5.63102901e-01 5.45742810e-01 -5.33629134e-02 -4.38575953e-01 -4.27259542e-02 -5.36978483e-01 5.84800661e-01 9.38173831e-01 2.08140835e-01 -8.49297345e-01 -3.61886024e-01 -2.94312328e-01 6.20663583e-01 1.20544486e-01 4.45129395e-01 3.09670478e-01 -7.78117418e-01 -9.67233717e-01 -2.60542661e-01 -2.15520173e-01 5.10923982e-01 1.10458098e-01 8.54499042e-01 -7.15525329e-01 1.11142135e+00 3.66955429e-01 -5.10083556e-01 -1.39674342e+00 2.51317322e-01 1.08598679e-01 -1.00099385e+00 -6.18994951e-01 9.50055063e-01 -1.41484535e-03 -2.56864429e-01 3.42437983e-01 -4.52428401e-01 -5.89888453e-01 4.15511042e-01 6.82488859e-01 5.54241002e-01 4.10085648e-01 -5.71020484e-01 -3.45844448e-01 3.57998461e-01 -5.16863942e-01 -2.27766670e-03 1.15494978e+00 -2.83994138e-01 -4.10811514e-01 2.68415868e-01 1.03783607e+00 -7.84402788e-02 -7.02020586e-01 -3.72474730e-01 4.41000164e-01 -4.38253246e-02 8.31549093e-02 -4.94041324e-01 -7.28581846e-01 1.01533079e+00 -2.65575260e-01 7.07159877e-01 1.14625585e+00 1.68887258e-01 1.17538333e+00 7.51572847e-01 -5.90384938e-02 -1.44012320e+00 -1.86538368e-01 8.48060966e-01 7.61009395e-01 -9.23874378e-01 3.14161420e-01 -5.46822548e-01 -5.45978129e-01 1.00687897e+00 3.40799481e-01 6.49492815e-02 8.81373361e-02 2.36482516e-01 -7.77638629e-02 3.76324244e-02 -1.12841177e+00 -1.61322773e-01 4.17101264e-01 2.94563890e-01 8.64060402e-01 -1.24769002e-01 -6.30159140e-01 6.90565825e-01 -4.87660825e-01 -5.66756248e-01 5.26590168e-01 9.94296551e-01 -6.63175702e-01 -1.23428965e+00 -1.85054928e-01 6.93442047e-01 -8.25268626e-01 -3.00871581e-01 -6.22831643e-01 6.04113996e-01 -6.65913165e-01 1.20764935e+00 2.17989802e-01 6.98819384e-02 3.49700332e-01 8.20166022e-02 4.77820963e-01 -6.90216064e-01 -8.73859286e-01 1.38544470e-01 8.00803542e-01 -2.75113314e-01 -5.08513510e-01 -6.62093580e-01 -1.41354871e+00 -6.08328521e-01 -4.89030421e-01 2.55500138e-01 2.89791614e-01 1.15939915e+00 4.23688501e-01 8.36153388e-01 3.08341384e-01 -6.32099330e-01 -5.96427202e-01 -1.34175813e+00 -5.81795275e-02 1.64152682e-01 2.46030465e-01 -7.09175244e-02 -1.67083070e-01 3.21709394e-01]
[12.240033149719238, 9.405409812927246]