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
|
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
2c25c152-acee-441f-9576-03a4f57fdb5f
|
fuzzy-expert-system-for-stock-portfolio
|
2204.13385
| null |
https://arxiv.org/abs/2204.13385v2
|
https://arxiv.org/pdf/2204.13385v2.pdf
|
Fuzzy Expert System for Stock Portfolio Selection: An Application to Bombay Stock Exchange
|
Selection of proper stocks, before allocating investment ratios, is always a crucial task for the investors. Presence of many influencing factors in stock performance have motivated researchers to adopt various Artificial Intelligence (AI) techniques to make this challenging task easier. In this paper a novel fuzzy expert system model is proposed to evaluate and rank the stocks under Bombay Stock Exchange (BSE). Dempster-Shafer (DS) evidence theory is used for the first time to automatically generate the consequents of the fuzzy rule base to reduce the effort in knowledge base development of the expert system. Later a portfolio optimization model is constructed where the objective function is considered as the ratio of the difference of fuzzy portfolio return and the risk free return to the weighted mean semi-variance of the assets that has been used. The model is solved by applying Ant Colony Optimization (ACO) algorithm by giving preference to the top ranked stocks. The performance of the model proved to be satisfactory for short-term investment period when compared with the recent performance of the stocks.
|
['Rupak Bhattacharyya', 'Seema Sarkar', 'Gour Sundar Mitra Thakur']
|
2022-04-28
| null | null | null | null |
['portfolio-optimization']
|
['time-series']
|
[-2.00231194e-01 -1.57402039e-01 2.55618393e-01 -1.10241622e-01
2.80588895e-01 -7.92433858e-01 3.90610099e-01 1.40996754e-01
-5.15086174e-01 1.03319776e+00 -1.66241363e-01 -5.11335611e-01
-9.26401913e-01 -1.11557364e+00 6.04812466e-02 -5.56578636e-01
3.17477226e-01 7.09498882e-01 2.57042795e-01 -5.77144980e-01
1.08713734e+00 7.51130164e-01 -1.46935415e+00 9.58060008e-03
8.06194782e-01 9.84569967e-01 2.30042025e-01 2.59371698e-01
-1.58061594e-01 9.21236694e-01 -6.90481007e-01 -5.99416256e-01
6.09015584e-01 -3.20511013e-01 -3.16258848e-01 1.26758066e-04
-6.08861685e-01 -9.46744457e-02 4.86371845e-01 1.20597601e+00
3.34903002e-01 2.61149764e-01 8.62363517e-01 -1.11630261e+00
-4.34213161e-01 6.29411340e-01 -4.76684988e-01 6.86380506e-01
-1.10942364e-01 -4.50136960e-01 7.77149439e-01 -7.46242642e-01
2.25563526e-01 8.32413554e-01 2.68855512e-01 -1.75368413e-01
-5.07476568e-01 -4.68080908e-01 -2.51143277e-01 4.79232341e-01
-1.08929420e+00 9.39102322e-02 6.63336515e-01 -3.87329310e-01
1.13056552e+00 2.65526175e-01 9.43728447e-01 -5.44614971e-01
5.00479043e-01 -1.22452684e-01 1.27750063e+00 -7.58332789e-01
3.61531168e-01 5.83521068e-01 7.95083269e-02 4.11319025e-02
1.07226014e+00 1.93804830e-01 5.87587133e-02 9.98575240e-02
3.44574928e-01 1.26621183e-02 4.34361249e-02 2.43580475e-01
-4.00089681e-01 9.28750753e-01 1.30044878e-01 7.91399002e-01
-8.33954275e-01 -4.34884638e-01 7.06983283e-02 3.28781784e-01
-5.08508682e-02 5.92809141e-01 -2.64381230e-01 1.18982410e-02
-9.93685424e-01 1.36317641e-01 8.55555594e-01 2.94201642e-01
2.85121858e-01 3.13453943e-01 3.52751285e-01 2.20036268e-01
5.88473201e-01 4.13528651e-01 4.61436987e-01 -6.88397467e-01
3.61699194e-01 9.86429453e-01 3.54866713e-01 -1.32425094e+00
6.08715676e-02 -5.88926375e-01 -1.43597424e-01 7.45987177e-01
1.50345534e-01 -2.42750198e-01 -4.32772160e-01 8.09203088e-01
2.98049580e-02 -1.34641930e-01 2.91194856e-01 5.18299043e-01
2.81510919e-01 8.07862222e-01 -3.05874825e-01 -5.37738979e-01
1.00666535e+00 -5.47669888e-01 -8.41228604e-01 1.81417912e-01
-2.45489031e-01 -7.19804108e-01 -1.93158500e-02 6.85764492e-01
-1.13497472e+00 -1.38695255e-01 -1.27963746e+00 9.15315509e-01
-6.79751933e-01 -1.08806975e-01 4.92753595e-01 7.51718998e-01
-8.24886501e-01 5.64768136e-01 -3.71349722e-01 1.96065187e-01
-9.74732190e-02 6.30821168e-01 9.75966305e-02 3.80291253e-01
-1.05778611e+00 1.52677262e+00 8.12916934e-01 2.57018149e-01
-7.77709782e-02 -1.63409024e-01 -2.29375675e-01 1.24311276e-01
4.70746547e-01 -1.34848535e-01 6.13463461e-01 -1.08976102e+00
-1.24845088e+00 3.51563871e-01 6.10646963e-01 -7.34782338e-01
3.62412542e-01 3.35408598e-01 -4.46251839e-01 2.86065429e-01
-1.91008337e-02 -4.78507690e-02 1.70374081e-01 -8.47668588e-01
-1.02943254e+00 -2.39055946e-01 1.89504832e-01 4.48665082e-01
-2.37908110e-01 4.10491526e-01 1.92842498e-01 -6.54852092e-01
5.41679449e-02 -6.21359766e-01 -1.17290735e-01 -8.69221210e-01
3.85703266e-01 1.60300471e-02 2.95022786e-01 -6.71504319e-01
1.57938159e+00 -1.48012757e+00 -3.01055551e-01 9.48903263e-01
-4.43875372e-01 5.76925695e-01 6.57873094e-01 3.40653598e-01
-2.44974852e-01 6.44106865e-02 -1.91625625e-01 9.45486009e-01
3.68867926e-02 8.26185271e-02 -1.08099379e-01 1.94903614e-03
7.49204680e-03 2.05625087e-01 -6.21400714e-01 -3.37206453e-01
2.73194551e-01 6.48475289e-02 -1.59635603e-01 5.09975962e-02
7.51383081e-02 -3.83694559e-01 -5.35556078e-01 7.20763445e-01
5.92245400e-01 2.46645167e-01 1.35472044e-01 -1.99073330e-01
-5.42144775e-01 -7.96844959e-02 -1.64773762e+00 2.69858032e-01
-7.20207840e-02 2.50135183e-01 -1.46205500e-01 -1.13745356e+00
1.19879520e+00 6.06890857e-01 4.35738176e-01 -5.71755171e-01
5.13070583e-01 5.51533520e-01 4.44806963e-01 -5.57514191e-01
4.48132783e-01 -5.00813842e-01 3.44272941e-01 8.26729476e-01
-2.79632896e-01 -1.27423346e-01 8.60645950e-01 -2.55431414e-01
4.38801825e-01 5.15690260e-02 7.60166526e-01 -4.47094083e-01
7.21415937e-01 1.56198129e-01 5.88292420e-01 1.35767311e-01
-5.86132193e-03 -8.93004686e-02 3.14719081e-01 -3.91645938e-01
-6.36945069e-01 -5.27712047e-01 -3.40742767e-02 3.29942852e-01
-1.18266739e-01 5.04831553e-01 -5.25522411e-01 -2.19856009e-01
4.26431093e-03 1.12407219e+00 -3.67289156e-01 1.71753258e-01
-2.94087946e-01 -1.10455894e+00 -3.90444137e-02 3.19811292e-02
5.26475370e-01 -1.32637405e+00 -1.28560412e+00 6.16745174e-01
4.73532259e-01 -3.04769695e-01 2.38448173e-01 1.39073744e-01
-8.13885152e-01 -1.19536257e+00 -6.62726343e-01 -2.70676047e-01
7.50147820e-01 4.28571692e-03 8.01357150e-01 3.36325765e-01
-6.91920221e-02 -9.94285643e-02 -5.53075016e-01 -1.06844330e+00
-2.68221945e-01 -3.67867589e-01 -9.86133143e-02 -8.24256465e-02
4.39085066e-01 -2.95142144e-01 -2.51066118e-01 1.42795652e-01
-8.85016859e-01 -4.78510678e-01 6.94592118e-01 3.69580150e-01
2.85241902e-01 1.26760590e+00 1.08556592e+00 -5.73202133e-01
9.96880889e-01 -4.73273665e-01 -1.24379003e+00 7.67919838e-01
-1.12844145e+00 3.43369283e-02 1.16843142e-01 -5.18882088e-02
-1.33241189e+00 -4.79787081e-01 4.57282096e-01 2.18271971e-01
3.41920555e-01 9.73770201e-01 -1.16763718e-01 -3.03050458e-01
1.03418283e-01 4.14728783e-02 1.14677669e-02 -2.02771693e-01
-4.26288247e-01 6.92583919e-01 2.17020407e-01 -2.98228920e-01
5.73502302e-01 5.32883033e-03 3.64682108e-01 -2.41595283e-02
-4.53736454e-01 -2.15605512e-01 -3.18642169e-01 -6.07066512e-01
5.20050466e-01 -1.92599013e-01 -6.42511606e-01 2.76757538e-01
-6.96347594e-01 3.91590029e-01 2.00132355e-01 7.20544934e-01
-2.53318340e-01 -1.34566724e-01 -8.06366727e-02 -1.48495615e+00
-6.18674874e-01 -1.10001111e+00 -4.19907540e-01 4.50194806e-01
-7.29005411e-02 -1.01061058e+00 -4.74705920e-02 4.14455026e-01
4.68895942e-01 6.57508910e-01 8.61321211e-01 -8.03157330e-01
-4.65338975e-01 -5.57386816e-01 2.05823518e-02 5.69381654e-01
3.76812547e-01 3.28366935e-01 -4.63158041e-02 2.15596855e-01
5.98369777e-01 3.25076371e-01 4.02068585e-01 4.31319773e-01
1.25786871e-01 -3.45456541e-01 8.54009166e-02 6.89589977e-02
1.92640138e+00 1.23391807e+00 6.79894686e-01 1.15368140e+00
-9.97059643e-02 7.58071661e-01 1.33954489e+00 7.50373721e-01
8.76436308e-02 3.19898516e-01 2.06352606e-01 5.99804819e-01
5.06072104e-01 6.37030065e-01 2.62676984e-01 6.00744307e-01
-4.70001996e-01 -1.18819676e-01 -8.92879903e-01 5.04422724e-01
-1.60614443e+00 -1.28978395e+00 1.38943225e-01 2.00312257e+00
6.43345475e-01 6.13948882e-01 1.85816377e-01 8.39541793e-01
8.92415524e-01 -4.93049920e-01 -7.55527988e-02 -7.43000984e-01
-1.56645641e-01 2.88220912e-01 7.42208004e-01 5.50287962e-01
-5.78741312e-01 3.16104174e-01 5.08929110e+00 6.80106044e-01
-1.21927536e+00 -5.17756641e-01 4.35675204e-01 -1.35030419e-01
-4.72871214e-01 5.71187139e-02 -7.34386265e-01 8.08380783e-01
8.04404080e-01 -8.38979244e-01 5.18818557e-01 3.92150313e-01
2.26343527e-01 -6.72769427e-01 -2.43596986e-01 2.31089056e-01
1.19094513e-01 -1.21233702e+00 -2.30541341e-02 8.80852044e-02
8.41883540e-01 -6.26548409e-01 -4.46813405e-02 -2.70714462e-01
1.86055839e-01 -7.31933653e-01 1.05378401e+00 9.86437321e-01
-5.77872731e-02 -1.34318769e+00 1.40730739e+00 3.41394246e-01
-1.10168195e+00 -4.04393971e-01 -4.45009947e-01 -1.67415783e-01
2.25051984e-01 4.02025044e-01 -9.85590398e-01 6.59455538e-01
5.97449720e-01 1.59661490e-02 -3.44548911e-01 1.22067761e+00
1.18234798e-01 1.61376968e-01 -4.79650974e-01 -4.70062912e-01
3.27006012e-01 -8.70390654e-01 3.80971342e-01 4.87435073e-01
8.24713349e-01 6.00690603e-01 -6.87067270e-01 7.09753871e-01
4.03106630e-01 2.99836934e-01 -3.93028408e-01 -2.73445666e-01
8.12407494e-01 9.50357020e-01 -1.26938450e+00 -2.99057305e-01
-1.26840562e-01 8.10843136e-04 -2.74999082e-01 -6.00857101e-02
-4.79327083e-01 -7.32322037e-01 -2.95258015e-01 1.63252234e-01
4.44728106e-01 2.59436756e-01 -6.60881817e-01 -3.42691958e-01
-8.93004984e-02 -7.91023135e-01 5.13642192e-01 -5.27977586e-01
-8.72969031e-01 6.19179547e-01 3.15193206e-01 -1.14386308e+00
-3.23083520e-01 -6.91267431e-01 -8.82606387e-01 1.11312354e+00
-1.39177203e+00 -2.95520931e-01 2.95683980e-01 2.01999590e-01
4.53428291e-02 -1.02182078e+00 4.16589558e-01 1.85957119e-01
-4.66842383e-01 1.31270643e-02 1.43311962e-01 -8.88373330e-02
6.51792288e-02 -1.21349919e+00 -4.91231024e-01 1.12092912e+00
-3.05366188e-01 5.29648125e-01 7.50454426e-01 -1.16040719e+00
-6.08763456e-01 -3.15977424e-01 1.00169861e+00 1.83560044e-01
7.49017537e-01 7.51199782e-01 -5.29509187e-01 1.39341459e-01
3.88671279e-01 -7.42053032e-01 7.60060787e-01 -8.43384147e-01
4.15620297e-01 -4.66040492e-01 -1.62450981e+00 4.16780263e-01
-2.31662840e-01 1.64238825e-01 -1.04956412e+00 -4.31068122e-01
-4.90259714e-02 1.76964298e-01 -9.17822778e-01 3.90365243e-01
5.50268471e-01 -1.13553560e+00 8.56412888e-01 -1.13795921e-01
-1.20735183e-01 -6.24171317e-01 -4.59251143e-02 -9.12292600e-01
-3.04858595e-01 -3.38636696e-01 2.64097035e-01 1.29104304e+00
6.95253730e-01 -8.21251452e-01 5.60586154e-01 8.94281089e-01
1.48195714e-01 -8.03518593e-01 -7.52079368e-01 -4.32203770e-01
-3.81502301e-01 3.02399963e-01 9.36464489e-01 7.73043990e-01
-1.83066186e-02 -1.92273900e-01 8.58987048e-02 -1.12514861e-01
8.17798436e-01 2.81137437e-01 -1.29998982e-01 -1.35035789e+00
-7.38292783e-02 -7.96729028e-01 -1.51650593e-01 5.39560854e-01
-2.44866744e-01 -5.55266082e-01 -3.13838303e-01 -1.53725207e+00
-2.15497255e-01 -3.76619875e-01 -7.78794646e-01 2.61637896e-01
-2.62298286e-01 -6.86922222e-02 4.88576174e-01 5.29398993e-02
-2.08512228e-02 -4.82492596e-02 9.74328995e-01 2.24328816e-01
-2.33075202e-01 4.65588748e-01 -9.74210083e-01 7.31271923e-01
9.89993274e-01 -5.04761994e-01 -5.67122102e-01 -2.92283203e-03
9.05816555e-01 3.44664633e-01 -3.37426782e-01 -9.12827075e-01
3.51234466e-01 -6.60320759e-01 3.22957277e-01 -9.04203415e-01
-1.51173279e-01 -1.27011037e+00 8.97799850e-01 6.92169487e-01
-1.44180387e-01 7.54074752e-01 -3.74653116e-02 3.23253423e-02
-4.03435737e-01 -1.24427104e+00 5.08683443e-01 -3.35643262e-01
-4.59467798e-01 -1.94636047e-01 -4.47808832e-01 -3.39560330e-01
1.50666976e+00 -8.05266619e-01 -4.48735803e-02 -7.79111683e-02
-3.87390167e-01 2.04979971e-01 2.35691711e-01 -1.11210652e-01
5.06757677e-01 -1.09064388e+00 -6.97809219e-01 -2.95948029e-01
-5.25952458e-01 -4.09024239e-01 1.90448891e-02 6.01132333e-01
-1.28585768e+00 6.37596250e-01 -8.86326194e-01 5.47861099e-01
-1.09994638e+00 4.13895071e-01 3.65690529e-01 -4.47486609e-01
1.36598915e-01 7.45230973e-01 -6.63402736e-01 3.96137685e-01
-3.88156697e-02 2.26187110e-01 -1.15915501e+00 6.22062922e-01
6.19990349e-01 9.46860850e-01 2.46902138e-01 -7.20819592e-01
-4.30017382e-01 7.04217374e-01 2.45123044e-01 -5.83089173e-01
1.82803535e+00 2.28196960e-02 -6.31213248e-01 1.58910424e-01
4.07162845e-01 1.14665173e-01 -8.01749706e-01 1.77501589e-01
5.97462356e-01 -5.00286579e-01 3.12427193e-01 -1.18277287e+00
-1.19363976e+00 3.89546245e-01 2.99617589e-01 5.15596092e-01
1.41137207e+00 -8.89120102e-01 4.17518198e-01 5.35245836e-01
3.72826546e-01 -1.65351105e+00 -4.25014555e-01 1.12537786e-01
9.07387435e-01 -9.04774249e-01 5.22960246e-01 -1.32617697e-01
-7.94354796e-01 1.48067105e+00 1.38128459e-01 -4.58405316e-01
1.06350005e+00 5.01674712e-01 -2.04161108e-02 -1.22400895e-01
-5.60414493e-01 -4.32825908e-02 3.44303250e-01 2.52622008e-01
2.33479157e-01 7.18075559e-02 -1.01757634e+00 9.57717001e-01
-3.46710980e-01 3.07725608e-01 7.96246469e-01 1.11662495e+00
-8.66135299e-01 -1.09080601e+00 -8.44026506e-01 6.21595085e-01
-1.06107891e+00 -1.52653411e-01 -2.93112457e-01 7.06259787e-01
4.14472282e-01 1.16151083e+00 4.44859862e-02 -3.43400016e-02
2.10846096e-01 -3.52010548e-01 2.71524847e-01 -3.64992142e-01
-8.96030903e-01 2.38128990e-01 1.51517466e-01 3.81166697e-01
-7.52504528e-01 -8.26602817e-01 -1.45238876e+00 -3.75422418e-01
-5.96613109e-01 9.05587912e-01 8.63967776e-01 1.12781882e+00
-2.96617806e-01 4.45867389e-01 7.53235459e-01 -4.35483634e-01
-6.45878136e-01 -7.35958219e-01 -8.54574621e-01 -9.15890783e-02
-3.24149758e-01 -8.80400777e-01 -4.73793328e-01 -8.24637637e-02]
|
[5.241966724395752, 3.816481351852417]
|
bced6dac-5484-42f2-85b5-a2837cd9be7a
|
how-far-is-language-model-from-100-few-shot
|
2307.00186
| null |
https://arxiv.org/abs/2307.00186v1
|
https://arxiv.org/pdf/2307.00186v1.pdf
|
How far is Language Model from 100% Few-shot Named Entity Recognition in Medical Domain
|
Recent advancements in language models (LMs) have led to the emergence of powerful models such as Small LMs (e.g., T5) and Large LMs (e.g., GPT-4). These models have demonstrated exceptional capabilities across a wide range of tasks, such as name entity recognition (NER) in the general domain. (We define SLMs as pre-trained models with fewer parameters compared to models like GPT-3/3.5/4, such as T5, BERT, and others.) Nevertheless, their efficacy in the medical section remains uncertain and the performance of medical NER always needs high accuracy because of the particularity of the field. This paper aims to provide a thorough investigation to compare the performance of LMs in medical few-shot NER and answer How far is LMs from 100\% Few-shot NER in Medical Domain, and moreover to explore an effective entity recognizer to help improve the NER performance. Based on our extensive experiments conducted on 16 NER models spanning from 2018 to 2023, our findings clearly indicate that LLMs outperform SLMs in few-shot medical NER tasks, given the presence of suitable examples and appropriate logical frameworks. Despite the overall superiority of LLMs in few-shot medical NER tasks, it is important to note that they still encounter some challenges, such as misidentification, wrong template prediction, etc. Building on previous findings, we introduce a simple and effective method called \textsc{RT} (Retrieving and Thinking), which serves as retrievers, finding relevant examples, and as thinkers, employing a step-by-step reasoning process. Experimental results show that our proposed \textsc{RT} framework significantly outperforms the strong open baselines on the two open medical benchmark datasets
|
['Rui Zhang', 'Mingchen Li']
|
2023-07-01
| null | null | null | null |
['few-shot-ner', 'cg']
|
['natural-language-processing', 'natural-language-processing']
|
[ 1.52944000e-02 2.90036023e-01 -1.43962666e-01 -5.35778403e-02
-9.91350234e-01 -1.02571681e-01 4.24158275e-01 3.62449676e-01
-8.48122656e-01 6.41190171e-01 3.83966327e-01 -4.12747771e-01
-5.09825587e-01 -7.26264358e-01 -4.30615634e-01 -4.50113058e-01
1.69582173e-01 4.36723888e-01 1.63785264e-01 -3.69257927e-01
1.42124653e-01 1.77684352e-01 -1.15208900e+00 2.75913358e-01
1.13057506e+00 5.05547881e-01 1.72321111e-01 4.32979584e-01
-4.10607666e-01 9.62514579e-01 -6.56677485e-01 -8.03842425e-01
-1.38798967e-01 -1.59451023e-01 -9.45500672e-01 -3.88709754e-01
-1.19864583e-01 -1.32314429e-01 -2.64301568e-01 9.51458335e-01
1.18452191e+00 2.11725563e-01 6.04496121e-01 -7.42919981e-01
-8.72451603e-01 1.05284905e+00 -2.36460179e-01 3.95607680e-01
3.96953017e-01 7.45838657e-02 7.32263803e-01 -9.99139786e-01
8.56646240e-01 9.31323528e-01 1.05296206e+00 8.79233360e-01
-5.72617590e-01 -7.20380962e-01 -2.14735016e-01 1.30632073e-01
-1.53499198e+00 -5.05822837e-01 8.99493396e-02 -1.29484564e-01
1.18184090e+00 4.13048953e-01 5.43281250e-02 1.41752398e+00
3.01375061e-01 6.98092401e-01 9.03389513e-01 -4.63460267e-01
2.87022352e-01 2.82578856e-01 3.20407450e-01 6.22158051e-01
3.28701138e-01 -2.53812909e-01 -4.37684208e-01 -5.15559316e-01
3.83407414e-01 1.37406722e-01 -3.54122669e-01 4.32444811e-01
-1.23866689e+00 5.01538157e-01 2.68427610e-01 8.43567610e-01
-6.50852144e-01 -4.65321481e-01 4.31224167e-01 2.25117467e-02
3.03119838e-01 7.18040645e-01 -5.93483508e-01 -5.56393191e-02
-1.12674975e+00 -1.85460597e-01 1.02104950e+00 1.00770235e+00
2.29797419e-02 -2.72343516e-01 -6.03117406e-01 9.70300913e-01
-6.59323558e-02 4.07614559e-01 8.54587018e-01 -4.19053465e-01
4.08392906e-01 5.70832074e-01 -8.82492959e-02 -7.26841033e-01
-6.92028880e-01 -7.28739560e-01 -1.14108896e+00 -6.51761234e-01
-6.92518130e-02 -2.98267156e-01 -1.04299378e+00 1.56623733e+00
1.74825534e-01 4.78520244e-01 4.12346154e-01 4.08309937e-01
1.50057232e+00 3.64463449e-01 7.26842821e-01 -1.38591588e-01
1.76957107e+00 -8.06782901e-01 -9.35880482e-01 -8.09904784e-02
9.69129860e-01 -8.48545015e-01 8.49706411e-01 8.77192765e-02
-8.94418955e-01 -3.61091256e-01 -6.71963274e-01 6.80541061e-03
-7.51731277e-01 3.06582689e-01 5.91452837e-01 5.97015560e-01
-1.09046984e+00 5.28279066e-01 -7.53615797e-01 -8.72936964e-01
2.40420967e-01 2.11156309e-01 -4.21191871e-01 -7.03893751e-02
-1.67865431e+00 1.13101590e+00 6.00618958e-01 1.30825341e-01
-6.42609596e-01 -8.22428167e-01 -7.86598623e-01 2.43641764e-01
5.72858155e-01 -1.05289054e+00 1.03846419e+00 -4.57978286e-02
-1.06938434e+00 8.50987136e-01 2.75166109e-02 -6.13224566e-01
3.42713743e-01 -1.96904600e-01 -8.62763464e-01 9.16165113e-02
3.28609049e-01 4.82724726e-01 7.88125247e-02 -8.24310362e-01
-6.81111693e-01 -6.43579960e-02 6.44388273e-02 2.56917216e-02
-6.33255601e-01 4.46999460e-01 -5.04644990e-01 -6.26407683e-01
-2.23831698e-01 -7.03241944e-01 -4.40305889e-01 -5.68659723e-01
-7.02202678e-01 -5.06328523e-01 6.07982874e-02 -8.10772181e-01
1.89638138e+00 -2.09973073e+00 -4.29333895e-01 -7.30697140e-02
3.23364347e-01 6.33802712e-01 -2.23402362e-02 5.72157204e-01
-1.98840991e-01 4.98733521e-01 -2.07847685e-01 -1.60838842e-01
-1.21634856e-01 8.95850658e-02 -1.07909344e-01 -1.74945612e-02
9.20698196e-02 1.13970470e+00 -9.78828490e-01 -8.79346609e-01
-4.53702733e-02 6.12345755e-01 -1.49582207e-01 6.00309148e-02
1.92717180e-01 1.27286971e-01 -7.47096479e-01 8.58855188e-01
3.86070758e-01 -6.95840955e-01 -3.93192470e-03 -1.04399927e-01
-1.44715980e-02 2.18494743e-01 -1.05678773e+00 1.47585392e+00
-2.64966518e-01 -6.27314076e-02 -2.76820272e-01 -5.58998466e-01
6.48131967e-01 8.05896461e-01 3.50494504e-01 -4.94756848e-01
1.55831024e-01 3.84583503e-01 -6.67961240e-02 -9.18463290e-01
5.03477633e-01 -1.33376777e-01 -1.80058569e-01 1.45378083e-01
9.27653685e-02 6.63644433e-01 2.98702568e-01 4.70037043e-01
1.53847647e+00 -2.41670787e-01 8.24688017e-01 -1.38754189e-01
6.52002454e-01 1.45052848e-02 6.85562074e-01 1.03677475e+00
-1.94157586e-01 3.59434158e-01 5.35614267e-02 -1.30995154e-01
-4.81655240e-01 -6.09688342e-01 -3.58093023e-01 9.46384132e-01
-7.19800815e-02 -6.77969754e-01 -7.95788348e-01 -8.08060229e-01
-4.10130113e-01 9.72201049e-01 -4.88528401e-01 -1.61111742e-01
-5.09480357e-01 -1.19023335e+00 1.18666887e+00 5.78676283e-01
5.61019063e-01 -1.22578049e+00 -5.63018262e-01 3.73950928e-01
-3.36054355e-01 -1.21844637e+00 -3.34370077e-01 1.76193446e-01
-6.81398749e-01 -1.09433544e+00 -9.82499182e-01 -7.52638757e-01
7.52155423e-01 7.16917291e-02 1.11773515e+00 2.18028381e-01
-4.93472159e-01 4.05214727e-01 -5.59506893e-01 -5.52127004e-01
-4.42654133e-01 3.80853534e-01 -9.54076126e-02 -4.38046962e-01
6.73096061e-01 -1.89878523e-01 -6.28196061e-01 1.70570150e-01
-1.05387163e+00 -8.62726718e-02 1.09452963e+00 8.38621616e-01
5.30936539e-01 7.54977688e-02 6.75375879e-01 -1.40401387e+00
8.61222863e-01 -7.41230667e-01 2.04577163e-01 9.82666731e-01
-9.13178921e-01 5.48180044e-02 6.58635020e-01 -4.25092429e-01
-1.17585135e+00 -4.63843018e-01 -5.35281122e-01 -6.20089322e-02
-2.43304029e-01 7.82747746e-01 1.15174681e-01 1.86167896e-01
9.02958274e-01 3.13292474e-01 -5.55243433e-01 -7.24414527e-01
9.58285779e-02 9.91275072e-01 4.69400465e-01 -3.72671783e-01
5.47648132e-01 2.04996124e-01 -3.63687336e-01 -6.94504797e-01
-1.07906485e+00 -7.28439331e-01 -1.82016224e-01 2.11361289e-01
1.02222228e+00 -9.67491388e-01 -4.12302166e-01 2.34579563e-01
-9.89961624e-01 7.06757531e-02 -6.09594621e-02 5.64020514e-01
1.18458946e-03 3.93924028e-01 -1.02255023e+00 -7.20504522e-01
-9.57129657e-01 -8.51972401e-01 1.06007397e+00 4.57795233e-01
-3.32818478e-01 -1.12914681e+00 -1.31767094e-01 4.23152447e-01
5.53180635e-01 1.22573860e-02 1.14519429e+00 -1.32575297e+00
-1.19228296e-01 -2.02328891e-01 1.18616614e-02 -8.18643123e-02
5.55386208e-03 -3.13527942e-01 -7.64931500e-01 -7.81306475e-02
1.81058198e-01 -5.47830835e-02 8.74933302e-01 1.52430594e-01
1.01018691e+00 -1.37837693e-01 -7.88585901e-01 5.55412114e-01
1.20014620e+00 2.66479701e-01 7.10546613e-01 4.38620448e-01
3.93590599e-01 4.07280564e-01 6.19784176e-01 2.55089253e-01
5.96631348e-01 1.44887552e-01 -1.49980739e-01 -2.55365193e-01
-2.39055738e-01 -2.10629880e-01 1.31625578e-01 9.80570614e-01
-2.11466581e-01 -3.18477780e-01 -1.13698912e+00 4.25367624e-01
-1.61264861e+00 -7.44085073e-01 1.16594754e-01 1.90857875e+00
9.67908442e-01 1.09304473e-01 -4.64102507e-01 -2.62098432e-01
7.82840848e-01 -1.06846206e-01 -4.47414786e-01 -7.31449202e-02
-1.78118080e-01 4.54936713e-01 3.84192377e-01 -5.86581193e-02
-9.88915086e-01 8.78635526e-01 5.96705246e+00 1.15384471e+00
-8.73105764e-01 2.87367374e-01 6.18657649e-01 1.98995113e-01
-6.41144589e-02 -2.12245211e-01 -1.26498246e+00 4.10440415e-01
1.17567670e+00 -3.18580210e-01 -6.07100911e-02 6.56902075e-01
-1.15539603e-01 1.70344561e-01 -9.61934447e-01 9.12487388e-01
3.08859915e-01 -1.21603656e+00 1.46288663e-01 -4.90984805e-02
5.33688486e-01 9.29852501e-02 -1.83082044e-01 9.16374445e-01
2.08192289e-01 -1.04719436e+00 1.34200916e-01 7.41257131e-01
7.92082727e-01 -3.30257028e-01 1.23600769e+00 6.74920499e-01
-1.06836140e+00 -3.33889723e-02 -4.76165175e-01 6.43110752e-01
2.14899346e-01 5.78602672e-01 -9.01642442e-01 1.07898915e+00
7.22510338e-01 3.68122488e-01 -5.95244348e-01 1.16859055e+00
-2.61640042e-01 5.01849830e-01 -8.13372955e-02 4.38011773e-02
1.18428431e-01 4.32979196e-01 4.33825046e-01 1.50577712e+00
6.10312879e-01 5.31383634e-01 -2.95365583e-02 6.74526513e-01
-3.10522884e-01 4.87559766e-01 -2.24768609e-01 -1.79150075e-01
6.44937813e-01 1.53709602e+00 -8.75626445e-01 -5.24000287e-01
-4.40052837e-01 7.84033358e-01 1.74222082e-01 2.43915126e-01
-8.38611901e-01 -6.01375580e-01 -2.28751381e-03 -1.11081293e-02
1.15239710e-01 3.47095847e-01 8.40733666e-03 -1.18929887e+00
-1.27578452e-01 -9.96496797e-01 9.44051504e-01 -5.27751088e-01
-1.50824690e+00 1.02899110e+00 -1.64527193e-01 -1.16237521e+00
-1.55799061e-01 -4.16942984e-01 -4.93501216e-01 5.61465919e-01
-1.55523968e+00 -1.02093852e+00 -3.17775644e-02 6.00668252e-01
4.98288751e-01 -9.18960646e-02 1.11467600e+00 6.16731167e-01
-7.77228713e-01 8.94762218e-01 -6.18801005e-02 4.71551955e-01
1.08949661e+00 -1.02393746e+00 2.68740505e-01 7.56183743e-01
7.80174416e-03 1.23143589e+00 4.92382914e-01 -8.27944696e-01
-1.12790632e+00 -9.87336636e-01 1.62064195e+00 -5.99881351e-01
4.95684296e-01 3.38798128e-02 -1.05153239e+00 4.89084810e-01
1.10261254e-01 -1.25319824e-01 1.02370310e+00 7.45902359e-02
-6.48185536e-02 1.25749797e-01 -1.27418292e+00 7.07287192e-01
1.18092883e+00 -3.58035028e-01 -9.67295706e-01 4.19183582e-01
7.13750124e-01 -4.30399269e-01 -1.20662570e+00 6.98388219e-01
1.95989475e-01 -6.19274139e-01 1.03844154e+00 -7.43359447e-01
2.57187158e-01 2.99236644e-02 1.91128552e-01 -1.01041830e+00
-3.04092020e-01 -5.22458434e-01 -9.28708315e-02 1.40180898e+00
6.53717756e-01 -6.69212580e-01 2.36160189e-01 9.59640682e-01
-1.72049955e-01 -1.00411582e+00 -7.61056542e-01 -6.79380834e-01
-9.72608626e-02 -1.16139308e-01 5.92585087e-01 1.19251025e+00
1.01411454e-01 3.12195718e-01 -2.64274806e-01 1.93805903e-01
2.68577546e-01 -2.24091813e-01 1.19422406e-01 -1.24601400e+00
-2.71743864e-01 -2.61414140e-01 -8.97955745e-02 -6.58656716e-01
5.78508265e-02 -1.03045928e+00 -4.57329303e-02 -1.89321733e+00
6.45343602e-01 -3.07238072e-01 -7.61492610e-01 7.76876926e-01
-6.77022398e-01 -7.46378452e-02 -2.94370409e-02 4.08447295e-01
-9.46123600e-01 1.80858999e-01 9.08265948e-01 8.05322230e-02
-3.06459870e-02 -7.44419813e-04 -1.10835373e+00 7.78279841e-01
4.43875223e-01 -7.73165941e-01 -2.27631014e-02 -2.99549073e-01
3.62004668e-01 1.77614778e-01 -1.78597569e-02 -8.84098172e-01
7.88876414e-01 2.10419565e-01 2.64040262e-01 -3.29732865e-01
-1.09079778e-01 -5.45323908e-01 1.75667241e-01 5.74604750e-01
-3.98022771e-01 3.37508880e-02 2.76079718e-02 4.75246161e-01
-1.96631804e-01 -5.08964419e-01 2.60338873e-01 -4.00979370e-01
-8.89339983e-01 2.91301459e-01 -1.52545199e-01 2.45049208e-01
9.21762645e-01 -1.37620391e-02 -6.22773468e-01 2.48828102e-02
-8.50262225e-01 4.45369303e-01 -7.93949813e-02 4.09321874e-01
4.14784044e-01 -9.65045691e-01 -8.18217099e-01 -1.95938289e-01
2.87780732e-01 -1.91752180e-01 6.34815991e-01 1.19387579e+00
-1.35516062e-01 7.22106993e-01 1.89553767e-01 -1.08584478e-01
-1.24290264e+00 6.34474814e-01 2.01166824e-01 -8.81026089e-01
-8.57507527e-01 8.79495859e-01 1.38245821e-01 -5.69117486e-01
3.53572637e-01 -1.22507006e-01 -5.54728329e-01 -2.92525161e-02
8.73848021e-01 4.49315012e-01 1.86563283e-01 -4.08949018e-01
-5.86774528e-01 3.37954432e-01 -2.21390590e-01 3.44470263e-01
1.43598580e+00 -6.00880012e-02 -1.77795425e-01 1.78766668e-01
5.75363874e-01 4.10774946e-02 -3.29732627e-01 -4.29916471e-01
4.38333750e-01 2.03430161e-01 -1.34765103e-01 -1.26581013e+00
-7.17839599e-01 6.06828213e-01 5.06112933e-01 -7.79361054e-02
1.15472579e+00 1.59807831e-01 1.07525933e+00 5.45321286e-01
4.71997559e-01 -1.00512469e+00 -3.48752618e-01 5.40640235e-01
3.81896794e-01 -1.27721119e+00 -1.27365738e-01 -3.66209835e-01
-7.15534747e-01 9.84609723e-01 4.95326698e-01 4.89451468e-01
5.62038660e-01 4.02738512e-01 3.60939167e-02 -4.03249234e-01
-8.53397667e-01 -3.38228524e-01 4.16220069e-01 1.37226716e-01
6.83339715e-01 -3.83250266e-02 -7.95345426e-01 1.30328810e+00
-1.07886024e-01 3.45823646e-01 1.30673885e-01 9.12183583e-01
-2.88900971e-01 -1.15921617e+00 -1.90003604e-01 6.00039840e-01
-1.06851113e+00 -6.01978660e-01 -1.04831077e-01 7.93500423e-01
3.58452141e-01 1.00584662e+00 -6.03282392e-01 -3.04954469e-01
5.48362792e-01 4.42958713e-01 -6.22622781e-02 -8.50353777e-01
-1.08636594e+00 -5.66824675e-02 1.71803564e-01 -3.03432345e-01
-3.93478751e-01 -3.10975820e-01 -1.37925780e+00 -1.00505553e-01
-4.96330470e-01 3.79458278e-01 3.19787741e-01 1.09895313e+00
5.84365129e-01 6.99123323e-01 -5.27651832e-02 1.47510141e-01
-9.37605500e-01 -1.10015595e+00 -4.36891019e-01 2.79169500e-01
-1.41340405e-01 -3.08538169e-01 -3.85990173e-01 -2.49237880e-01]
|
[8.57150936126709, 8.849961280822754]
|
16d5fdfb-0c02-4fe7-a27d-ab36b58e8735
|
face-fast-accurate-and-context-aware-audio
|
2303.03666
| null |
https://arxiv.org/abs/2303.03666v1
|
https://arxiv.org/pdf/2303.03666v1.pdf
|
Face: Fast, Accurate and Context-Aware Audio Annotation and Classification
|
This paper presents a context-aware framework for feature selection and classification procedures to realize a fast and accurate audio event annotation and classification. The context-aware design starts with exploring feature extraction techniques to find an appropriate combination to select a set resulting in remarkable classification accuracy with minimal computational effort. The exploration for feature selection also embraces an investigation of audio Tempo representation, an advantageous feature extraction method missed by previous works in the environmental audio classification research scope. The proposed annotation method considers outlier, inlier, and hard-to-predict data samples to realize context-aware Active Learning, leading to the average accuracy of 90% when only 15% of data possess initial annotation. Our proposed algorithm for sound classification obtained average prediction accuracy of 98.05% on the UrbanSound8K dataset. The notebooks containing our source codes and implementation results are available at https://github.com/gitmehrdad/FACE.
|
['Saeed Bagheri Shouraki', 'Hoda Mohammadzade', 'M. Mehrdad Morsali']
|
2023-03-07
| null | null | null | null |
['audio-classification', 'environmental-sound-classification', 'sound-classification']
|
['audio', 'audio', 'audio']
|
[ 2.02125385e-01 -2.80778669e-02 3.68188322e-02 -4.86048013e-01
-1.24412477e+00 -4.11103755e-01 3.20682488e-02 5.92219710e-01
-3.46667975e-01 5.93968391e-01 2.04425976e-01 2.25448206e-01
-6.16380453e-01 -6.37324512e-01 -1.06535472e-01 -7.15766847e-01
-3.54435354e-01 1.36950627e-01 7.29672834e-02 1.09905869e-01
5.00499189e-01 3.73525858e-01 -2.23240733e+00 4.25904810e-01
7.99202740e-01 1.15102375e+00 1.54175103e-01 9.02565420e-01
5.93279749e-02 5.59826672e-01 -5.99166691e-01 1.17980212e-01
3.02412778e-01 -2.23972261e-01 -6.87570870e-01 -2.27701619e-01
1.80263460e-01 8.13809335e-02 1.60836592e-01 6.07669055e-01
1.00411499e+00 3.88496190e-01 4.39263880e-01 -1.49403501e+00
2.68630832e-01 9.02560592e-01 -3.44747081e-02 3.78194988e-01
7.28185356e-01 6.49994165e-02 1.09355354e+00 -1.07234728e+00
2.09742710e-01 4.64980990e-01 6.50065005e-01 -1.07998841e-01
-8.37432384e-01 -9.68699813e-01 -2.13069230e-01 7.63979971e-01
-1.98557985e+00 -9.10955429e-01 1.05564940e+00 -4.77020979e-01
9.63349044e-01 7.63474584e-01 1.19268703e+00 7.96546876e-01
-1.56822413e-01 5.51544428e-01 8.32086802e-01 -7.12238133e-01
3.79221112e-01 1.89802140e-01 4.03883189e-01 2.29145005e-01
1.93069100e-01 1.52056381e-01 -1.05080140e+00 -4.57389802e-01
1.10250503e-01 -2.48273209e-01 -2.16253668e-01 2.25568414e-01
-7.73690045e-01 4.79294747e-01 -1.15172252e-01 3.73218358e-01
-5.13186038e-01 -1.86241895e-01 6.16169333e-01 4.91313398e-01
4.85286057e-01 4.48878944e-01 -5.54200172e-01 -6.21247232e-01
-1.16077161e+00 1.25435516e-01 6.49929047e-01 9.25162256e-01
7.09268749e-01 4.03427035e-01 1.18553847e-01 8.85238886e-01
3.31666768e-01 1.49851441e-01 3.89955223e-01 -8.68281066e-01
2.56475568e-01 7.45773077e-01 -9.39867944e-02 -9.60856140e-01
-5.42583883e-01 -6.36452734e-01 -2.53066659e-01 7.17446879e-02
1.61678076e-01 -1.37842894e-01 -3.27166200e-01 1.13138807e+00
5.73750913e-01 2.41910398e-01 -7.71618187e-02 5.67499816e-01
8.48042488e-01 6.12544119e-01 1.15208216e-01 -6.02954268e-01
1.00676608e+00 -3.27816993e-01 -7.51910150e-01 3.36504608e-01
5.49001157e-01 -1.09662557e+00 1.11329615e+00 8.20684373e-01
-7.36569464e-01 -6.28827453e-01 -9.69258249e-01 4.53067511e-01
-3.52288634e-01 2.29728654e-01 6.73266053e-01 8.02966297e-01
-6.92576230e-01 3.58805001e-01 -7.38331437e-01 -2.73180693e-01
2.89229363e-01 3.72500509e-01 -1.54008135e-01 5.90781450e-01
-1.02003276e+00 3.43061447e-01 6.88497126e-01 3.27165648e-02
-7.40551472e-01 -8.50862920e-01 -3.83566231e-01 2.65291706e-02
3.59967589e-01 -6.71648532e-02 1.24718511e+00 -1.03460479e+00
-1.38789129e+00 3.98201585e-01 -4.92021628e-02 -4.34922785e-01
3.57584476e-01 -5.25475025e-01 -8.47634792e-01 1.85249910e-01
-2.12479532e-01 1.78180754e-01 7.22335517e-01 -8.31725776e-01
-8.65087092e-01 -2.02212498e-01 -4.21658278e-01 2.96599507e-01
-6.48256123e-01 1.51322544e-01 3.03697467e-01 -5.99742889e-01
3.01843494e-01 -7.20787108e-01 1.76125810e-01 -3.89829606e-01
-2.46181026e-01 -2.49611791e-02 7.81774640e-01 -7.09237099e-01
1.78781998e+00 -2.25237131e+00 -3.54281723e-01 3.91646922e-01
-2.55302370e-01 -5.98886795e-03 3.01450700e-01 7.72187591e-01
-3.12216431e-01 -1.44829035e-01 2.29317266e-02 9.50196385e-02
2.10281392e-03 -1.64249122e-01 -3.93298119e-01 3.29289526e-01
-8.04423466e-02 1.39646590e-01 -7.99356759e-01 -6.47946775e-01
2.89757371e-01 4.68751818e-01 -5.76228321e-01 2.90982395e-01
1.71690002e-01 4.76733744e-01 -2.91331738e-01 1.10533285e+00
2.68568605e-01 5.07582784e-01 -1.65031880e-01 -2.17359394e-01
-5.69186449e-01 4.00425464e-01 -1.86210966e+00 1.43463683e+00
-3.69728059e-01 6.99274778e-01 -3.88393998e-01 -8.72680843e-01
1.23909235e+00 7.94407666e-01 8.88600826e-01 -3.60964417e-01
1.30112141e-01 5.05644202e-01 2.66500637e-02 -7.48475194e-01
8.17030728e-01 4.65611726e-01 -1.08476073e-01 1.76416263e-01
1.56748146e-01 -2.96405051e-02 2.44713590e-01 -6.41458109e-02
9.47927594e-01 1.57096416e-01 6.91852093e-01 -2.71808267e-01
4.45785671e-01 5.95264733e-02 7.06831932e-01 5.60512006e-01
-3.67270768e-01 3.38973343e-01 -1.07091963e-01 -3.00522506e-01
-7.16201365e-01 -5.95342040e-01 -3.02526683e-01 1.11337304e+00
-4.68124926e-01 -6.05312228e-01 -4.41543460e-01 -1.65284097e-01
-3.51020545e-01 7.39607930e-01 -1.18232049e-01 -7.56836385e-02
-3.88186514e-01 -6.14278257e-01 8.46395671e-01 5.09288013e-02
3.72977734e-01 -1.07823431e+00 -9.78323877e-01 3.13667446e-01
-3.01261127e-01 -4.63783503e-01 1.83335155e-01 6.61546171e-01
-8.33912849e-01 -1.06832576e+00 -1.84824914e-02 -4.95871007e-01
8.58899280e-02 -2.05879778e-01 9.39145923e-01 8.57927743e-03
-5.05555272e-01 4.94613767e-01 -9.83678997e-01 -8.43162000e-01
-2.30149016e-01 2.68478960e-01 8.41091126e-02 -6.45704418e-02
4.09232855e-01 -8.77537072e-01 -6.17935538e-01 1.45110175e-01
-3.93099666e-01 -3.53169471e-01 1.46304563e-01 4.62921798e-01
7.49499619e-01 3.37116450e-01 8.32267642e-01 -4.19805527e-01
3.82562548e-01 -6.19565785e-01 -3.65435123e-01 -9.66193213e-04
-6.75960839e-01 -6.50775254e-01 4.78616208e-01 -4.89917964e-01
-8.58271539e-01 5.53756952e-01 -4.38437074e-01 7.98598398e-03
-6.37741268e-01 2.75735170e-01 -1.38154373e-01 5.30393496e-02
8.56891751e-01 1.23872809e-01 -5.53592682e-01 -5.26544034e-01
-2.35357657e-01 1.06855237e+00 1.05212256e-01 -4.41564739e-01
4.81362551e-01 1.15748912e-01 -2.61989504e-01 -1.10889721e+00
-5.91847837e-01 -7.29681611e-01 -9.46270823e-01 -7.24208236e-01
3.43585312e-01 -9.21279490e-01 -5.01111865e-01 3.44878495e-01
-6.76036119e-01 -1.06531652e-02 -6.17997408e-01 8.08474243e-01
-5.95854640e-01 -3.15413214e-02 5.94622493e-02 -1.51366055e+00
-5.66414475e-01 -8.38556826e-01 8.20562363e-01 1.78453118e-01
-9.16835666e-01 -4.77214336e-01 2.57980824e-01 2.35025391e-01
2.81491667e-01 3.33623797e-01 5.42101324e-01 -1.12302077e+00
-3.27782124e-01 -3.14042121e-01 5.63282907e-01 1.36189843e-02
1.56433001e-01 3.40532422e-01 -1.58566749e+00 -8.47251043e-02
-2.44688123e-01 -1.76800061e-02 4.88057047e-01 1.66582868e-01
1.06247497e+00 -3.42851251e-01 2.73857675e-02 4.54484761e-01
1.26399279e+00 7.74780750e-01 5.33926189e-01 3.72682840e-01
2.48452783e-01 2.99041420e-01 1.18149030e+00 9.83251989e-01
5.03773168e-02 7.47440815e-01 4.30394828e-01 2.41710201e-01
-1.03370912e-01 -3.59128080e-02 3.25382501e-01 1.19616139e+00
-1.35673136e-01 -7.47767836e-02 -1.14927161e+00 5.00710964e-01
-1.51878750e+00 -1.24978375e+00 -2.34549180e-01 2.39488387e+00
7.47538626e-01 1.22327268e-01 3.98105264e-01 1.46145904e+00
5.52592218e-01 -8.42248052e-02 -1.52568996e-01 -4.58429486e-01
1.21452309e-01 3.97595882e-01 1.00420937e-01 4.25672412e-01
-1.22365725e+00 4.77210462e-01 5.46498346e+00 8.91080976e-01
-1.09015906e+00 1.48313627e-01 2.55653739e-01 -4.08410370e-01
1.47134736e-01 1.67486027e-01 -8.21037173e-01 4.23719168e-01
1.26478684e+00 -2.08285034e-01 3.57565396e-02 9.48990107e-01
6.98164701e-01 -2.06080407e-01 -6.69318438e-01 1.05628610e+00
-2.26708144e-01 -8.71463001e-01 -2.72372991e-01 -7.55954012e-02
2.21514627e-01 -2.65506148e-01 -1.74809813e-01 1.46786690e-01
-3.68358105e-01 -5.15156031e-01 1.03117752e+00 7.68602490e-01
4.93152976e-01 -1.06849170e+00 5.50607920e-01 2.58222073e-01
-1.51262391e+00 -6.31980181e-01 -8.81747082e-02 -3.65986079e-01
4.21265773e-02 7.40596235e-01 -1.24153113e+00 7.02599645e-01
9.03609991e-01 5.31645060e-01 -5.47902763e-01 1.65188932e+00
2.25530028e-01 1.17943001e+00 -5.07588565e-01 -1.16646783e-02
-3.85323912e-01 7.20802024e-02 1.01952887e+00 1.41846097e+00
8.88450682e-01 6.04023039e-02 2.51806855e-01 1.53031930e-01
5.98972440e-01 7.99870908e-01 -4.18557554e-01 8.08830783e-02
1.05926347e+00 1.19939971e+00 -9.38789785e-01 1.78418849e-02
8.26811641e-02 4.06160414e-01 -8.69786516e-02 -2.67649312e-02
-7.47718453e-01 -6.83982491e-01 3.90280902e-01 2.49360427e-01
5.11567555e-02 -2.80501336e-01 -4.90914881e-01 -6.71999335e-01
-1.16460256e-01 -7.28519678e-01 7.01449275e-01 -5.55066526e-01
-6.67386472e-01 7.32246578e-01 2.28182361e-01 -1.89668787e+00
-3.17639291e-01 7.91212842e-02 -6.62545323e-01 3.30565065e-01
-8.16114128e-01 -9.90558147e-01 -5.46750963e-01 4.69162971e-01
7.18333423e-01 -4.89703864e-01 1.22296727e+00 7.77601838e-01
-3.69676292e-01 6.55594647e-01 -1.39294297e-01 -2.17794910e-01
6.03409588e-01 -9.97916341e-01 -2.90068686e-01 8.85738015e-01
4.37504321e-01 2.97079027e-01 8.61544490e-01 -7.10101068e-01
-1.21120119e+00 -8.85678411e-01 9.91389513e-01 -1.17096551e-01
5.62883973e-01 -1.11121401e-01 -6.81960821e-01 2.59165347e-01
-2.32175998e-02 -3.00504804e-01 1.23119247e+00 1.75971866e-01
-3.28169540e-02 -4.93856996e-01 -1.18019915e+00 1.91911787e-01
1.02968919e+00 -4.59355474e-01 -2.92934060e-01 1.10488862e-01
3.37908357e-01 -3.26377958e-01 -1.12579882e+00 3.94463360e-01
7.86466718e-01 -9.50572908e-01 7.24657238e-01 -2.95411974e-05
-1.70637742e-01 -3.56177956e-01 -4.76397991e-01 -8.87220919e-01
-1.31920621e-01 -6.73045993e-01 -1.71043620e-01 1.67381084e+00
6.10650182e-01 -4.36833858e-01 4.47770953e-01 2.11209014e-01
-4.42614853e-01 -6.27143025e-01 -1.13951087e+00 -5.26316106e-01
-6.80434942e-01 -1.17502391e+00 7.29298353e-01 9.95515764e-01
9.91665274e-02 -2.10676640e-01 -3.76590312e-01 1.87243164e-01
5.36935806e-01 2.92729717e-02 6.03919566e-01 -1.48206151e+00
-2.17435285e-01 -1.27680406e-01 -8.35914075e-01 2.11974122e-02
-2.67541587e-01 -8.96247685e-01 -1.48803338e-01 -1.08510637e+00
-1.51614979e-01 -6.42124593e-01 -5.50786257e-01 8.10648918e-01
3.48291546e-01 4.85292375e-01 1.09574370e-01 9.77749825e-02
-3.94404888e-01 1.73171937e-01 5.22734761e-01 1.02936536e-01
-7.04463720e-01 4.77548420e-01 -4.94803846e-01 6.80397987e-01
1.21654487e+00 -7.71258175e-01 -6.33524954e-01 1.46308139e-01
3.53144705e-01 -1.39651876e-02 2.52726942e-01 -1.47097886e+00
1.27232790e-01 -1.82237536e-01 4.00941402e-01 -7.53641665e-01
4.37904149e-01 -1.15158987e+00 8.58925939e-01 3.09855938e-01
-4.90289867e-01 -1.01883478e-01 1.34243101e-01 1.86671183e-01
-3.53308171e-01 -3.43543977e-01 5.62059999e-01 9.64718685e-02
-9.33505774e-01 -2.43317392e-02 -6.04447305e-01 -2.89211124e-01
8.70467663e-01 -5.82475305e-01 1.07571818e-01 -2.81821489e-01
-1.17771924e+00 -3.47599298e-01 1.71937533e-02 3.62913996e-01
5.04623950e-01 -1.20947981e+00 -5.98563731e-01 1.80423662e-01
1.20660953e-01 -3.54769319e-01 3.77909184e-01 7.96018839e-01
-3.75245273e-01 1.28223717e-01 -2.72420824e-01 -7.06802368e-01
-1.79293382e+00 -1.53644174e-01 8.87040421e-02 1.69295147e-01
-4.81906056e-01 6.75735295e-01 -8.53196740e-01 -4.08046804e-02
4.18902993e-01 -1.86535388e-01 -6.60559177e-01 7.43812919e-01
4.12646145e-01 8.80395472e-01 5.90247273e-01 -7.08570719e-01
-5.19417882e-01 3.72015566e-01 4.40809935e-01 -7.10462108e-02
1.49323189e+00 -2.44237512e-01 3.48989934e-01 8.90153646e-01
1.00109899e+00 2.82062948e-01 -6.98783457e-01 4.28255126e-02
3.48548323e-01 -7.07617998e-01 1.59078106e-01 -7.25144267e-01
-7.73988008e-01 5.13196051e-01 1.17351985e+00 4.83769327e-01
1.47597086e+00 -4.12911564e-01 2.83513755e-01 3.50045711e-01
4.98955607e-01 -1.42126381e+00 -3.47239047e-01 4.74145353e-01
1.00637496e+00 -9.29399133e-01 3.12112510e-01 -2.31194332e-01
-6.45411968e-01 1.14392459e+00 4.26377326e-01 4.33998629e-02
9.54189897e-01 4.18133944e-01 1.28790304e-01 8.08895528e-02
-8.67941022e-01 -3.19046944e-01 3.55260938e-01 5.75403988e-01
7.21428990e-01 1.53915167e-01 -4.78713721e-01 7.64754713e-01
-7.25842834e-01 2.52435226e-02 3.55961800e-01 1.06137824e+00
-7.51570940e-01 -9.20134187e-01 -3.58956873e-01 4.59771365e-01
-4.14967448e-01 5.27379885e-02 -2.43321255e-01 5.61136782e-01
3.40863913e-01 1.13795578e+00 9.74729434e-02 -6.32525027e-01
4.13283706e-01 6.07989848e-01 1.78251982e-01 -5.21671653e-01
-1.01081586e+00 3.94774914e-01 4.40226167e-01 -3.21427286e-01
-4.61384416e-01 -8.30668151e-01 -1.22268963e+00 1.55086860e-01
-4.77046221e-01 5.08665383e-01 7.35281229e-01 6.15089476e-01
4.67600435e-01 4.82136875e-01 9.15117025e-01 -7.86527574e-01
-6.70618862e-02 -9.74175692e-01 -5.46729982e-01 -5.21409586e-02
1.67203262e-01 -6.38414025e-01 -4.85045344e-01 2.96509624e-01]
|
[15.624749183654785, 5.22084379196167]
|
ffefc513-01e3-4ea6-bc18-7e2a123cb995
|
neural-ordinary-differential-equations
|
1806.07366
| null |
https://arxiv.org/abs/1806.07366v5
|
https://arxiv.org/pdf/1806.07366v5.pdf
|
Neural Ordinary Differential Equations
|
We introduce a new family of deep neural network models. Instead of specifying a discrete sequence of hidden layers, we parameterize the derivative of the hidden state using a neural network. The output of the network is computed using a black-box differential equation solver. These continuous-depth models have constant memory cost, adapt their evaluation strategy to each input, and can explicitly trade numerical precision for speed. We demonstrate these properties in continuous-depth residual networks and continuous-time latent variable models. We also construct continuous normalizing flows, a generative model that can train by maximum likelihood, without partitioning or ordering the data dimensions. For training, we show how to scalably backpropagate through any ODE solver, without access to its internal operations. This allows end-to-end training of ODEs within larger models.
|
['Yulia Rubanova', 'David Duvenaud', 'Jesse Bettencourt', 'Ricky T. Q. Chen']
|
2018-06-19
|
neural-ordinary-differential-equations-1
|
http://papers.nips.cc/paper/7892-neural-ordinary-differential-equations
|
http://papers.nips.cc/paper/7892-neural-ordinary-differential-equations.pdf
|
neurips-2018-12
|
['multivariate-time-series-imputation']
|
['time-series']
|
[-2.54693359e-01 2.90161550e-01 -6.05012812e-02 -6.04509674e-02
-3.16991597e-01 -8.88094008e-01 6.15517437e-01 -5.25890946e-01
-3.87865901e-01 7.55975008e-01 -2.25900009e-01 -7.22172856e-01
-4.66610007e-02 -8.84638667e-01 -5.18694639e-01 -7.42938221e-01
-4.71350402e-01 6.73406899e-01 -1.08948641e-01 8.25925991e-02
-1.00835823e-01 8.88093948e-01 -8.33383620e-01 -3.66472341e-02
5.36461413e-01 9.11856949e-01 -4.93275762e-01 1.41256309e+00
5.49229383e-02 1.05819952e+00 -4.78976578e-01 -3.53522897e-02
7.00490534e-01 -5.77572405e-01 -8.25343966e-01 2.56251618e-02
1.40411422e-01 -7.58771062e-01 -5.81640661e-01 6.59075260e-01
3.73527378e-01 2.72255749e-01 6.81928396e-01 -1.43943846e+00
-4.80676770e-01 3.03647369e-01 1.69357538e-01 1.08585274e-02
-3.63048822e-01 7.57059872e-01 7.60573626e-01 -7.09510386e-01
8.11786890e-01 1.15191364e+00 8.55542600e-01 6.94849133e-01
-1.80390728e+00 -5.33828378e-01 1.72395140e-01 -6.18085682e-01
-1.12858915e+00 -3.98361385e-01 3.98718089e-01 -7.19294012e-01
1.36764359e+00 6.09044619e-02 7.40729749e-01 9.30653989e-01
3.58373135e-01 2.85781920e-01 6.50172412e-01 3.70818339e-02
5.33523381e-01 -1.29666582e-01 6.92400485e-02 1.00451338e+00
-1.91054925e-01 1.04650475e-01 -1.36049271e-01 -4.72034395e-01
1.50925720e+00 2.12294027e-01 -1.64930284e-01 -5.69715023e-01
-8.57001781e-01 1.08611906e+00 1.17266953e-01 -3.12282920e-01
-1.36434674e-01 9.04758990e-01 2.40712479e-01 5.59005737e-01
2.92321086e-01 5.37116885e-01 -7.69955933e-01 -2.80221909e-01
-1.08784783e+00 4.27709967e-01 1.29793203e+00 8.60307217e-01
8.56479585e-01 5.46419322e-01 2.72441730e-02 1.72171175e-01
1.92992240e-01 1.93881974e-01 5.06505966e-01 -1.81566024e+00
1.16990596e-01 2.77454525e-01 2.08558491e-03 -3.88124615e-01
-5.15975177e-01 -5.31982005e-01 -9.62343574e-01 7.94876397e-01
5.25899410e-01 -9.57068563e-01 -1.29052043e+00 1.86074448e+00
2.67545432e-01 2.24295527e-01 2.22533524e-01 5.95238566e-01
3.56278382e-02 1.00330281e+00 -2.24983096e-01 -1.78484723e-01
5.89204133e-01 -1.11038268e+00 -3.75679821e-01 4.50571161e-03
8.17623734e-01 -1.83533996e-01 6.45190358e-01 3.24376464e-01
-1.65513635e+00 -1.51975542e-01 -9.03593361e-01 -4.92343277e-01
-4.17435080e-01 6.42718151e-02 7.71116853e-01 4.65896390e-02
-1.62313318e+00 1.22337854e+00 -1.69391584e+00 2.62380809e-01
1.69039711e-01 7.11761117e-01 -7.05793053e-02 7.72538126e-01
-1.00553513e+00 7.80175149e-01 1.60563767e-01 1.77963987e-01
-1.40159869e+00 -9.70005929e-01 -7.99825013e-01 2.92352855e-01
-1.03707634e-01 -1.08476579e+00 1.62874043e+00 -8.18102777e-01
-2.19196796e+00 4.18304056e-01 -2.91267335e-01 -7.28015542e-01
6.45934403e-01 3.54770347e-02 1.67992651e-01 4.40459177e-02
-2.09067211e-01 6.08155966e-01 8.11299920e-01 -8.12221467e-01
6.23504445e-03 8.61120224e-02 2.00703979e-01 1.63349491e-02
-9.01669636e-02 -2.14474097e-01 -2.39952654e-01 -3.90208274e-01
1.31429106e-01 -1.10918677e+00 -6.92855358e-01 6.33255422e-01
-4.23839152e-01 1.93332136e-01 8.32204580e-01 -5.81660509e-01
1.13392484e+00 -1.76910782e+00 4.84187454e-01 2.48568445e-01
5.52851737e-01 7.04930946e-02 1.56306028e-01 4.37579393e-01
-2.61815697e-01 2.39008158e-01 -6.06632590e-01 -6.90648377e-01
2.45937824e-01 4.79754359e-01 -6.68195128e-01 5.21756411e-01
4.04818445e-01 1.02614665e+00 -6.19059086e-01 -2.43997782e-01
3.85737345e-02 7.35695243e-01 -1.03103197e+00 3.54087442e-01
-4.68228281e-01 5.06827176e-01 -3.18265855e-01 -5.68905808e-02
3.51466656e-01 -6.07252419e-01 1.88514233e-01 3.34406823e-01
-3.39238286e-01 6.31554961e-01 -1.38886142e+00 1.59299576e+00
-7.66341865e-01 7.14325488e-01 3.08445543e-01 -6.64027095e-01
3.94630671e-01 4.95511323e-01 3.41286123e-01 -1.09795041e-01
1.79581836e-01 2.06916332e-01 -6.28391653e-02 -1.31468013e-01
2.14330271e-01 -1.49376482e-01 -1.19979782e-02 5.95882177e-01
1.76169544e-01 -1.94146723e-01 2.39246786e-01 3.50992322e-01
1.10027754e+00 3.47705364e-01 -7.97739923e-02 -8.81636217e-02
1.39303818e-01 2.07386628e-01 4.85315621e-01 4.75762248e-01
4.19506729e-01 5.29185057e-01 1.29296517e+00 -4.26391035e-01
-1.30540538e+00 -1.15156484e+00 -7.34213218e-02 9.54680741e-01
-4.44761336e-01 -4.16671753e-01 -7.35557675e-01 -7.63930976e-02
2.37728097e-02 5.07177174e-01 -7.43720353e-01 -1.35181338e-01
-8.57906759e-01 -5.75014949e-01 8.25233936e-01 8.72518003e-01
3.55270952e-01 -7.11631060e-01 -8.15195262e-01 3.65535915e-01
5.63870370e-01 -5.92083812e-01 -3.34593773e-01 8.41410339e-01
-1.33641732e+00 -7.40208924e-01 -7.00684309e-01 -7.33157873e-01
7.52033949e-01 -6.41436279e-01 9.29023862e-01 9.05541331e-02
-1.95114732e-01 -8.12660009e-02 6.87318683e-01 3.25134099e-01
-6.54636979e-01 3.96516204e-01 -1.18863449e-01 -4.34435904e-01
-4.17026430e-01 -9.70248938e-01 -5.71246624e-01 -3.06019261e-02
-8.91772330e-01 2.74998665e-01 -4.77767698e-02 9.21907425e-01
4.82277244e-01 -2.45763972e-01 -1.22923590e-01 -6.38685882e-01
7.59422660e-01 -3.54100049e-01 -1.29656231e+00 -1.89476624e-01
-3.47691387e-01 7.18788743e-01 1.05701780e+00 -6.46808982e-01
-6.89026058e-01 2.36531690e-01 -1.35224402e-01 -8.32864463e-01
1.94216192e-01 3.33530486e-01 1.70590565e-01 1.40446369e-02
5.01157165e-01 -2.83156638e-03 2.23540649e-01 -5.16617596e-01
6.66272461e-01 8.73260647e-02 7.66711175e-01 -7.35718906e-01
8.54001284e-01 3.77813131e-01 3.55421603e-01 -2.78138816e-01
-3.95365387e-01 1.74337804e-01 -7.05293298e-01 4.02464986e-01
9.34512496e-01 -8.30384612e-01 -1.16245341e+00 6.74437702e-01
-1.32410538e+00 -1.21636355e+00 -5.65550923e-01 9.74393263e-02
-6.58009112e-01 -3.13947141e-01 -1.30338228e+00 -7.16843486e-01
-2.20960021e-01 -1.24607229e+00 7.73062706e-01 -1.28932819e-02
-3.94855797e-01 -1.36537826e+00 2.67223299e-01 -7.53658712e-01
7.39547193e-01 2.85408556e-01 8.26625049e-01 -4.45215017e-01
-8.97022009e-01 -7.79592320e-02 1.20950617e-01 3.61381352e-01
-3.56903464e-01 4.49047893e-01 -9.22926188e-01 -1.89590484e-01
7.61978850e-02 -1.03127271e-01 8.66853058e-01 2.92688876e-01
1.17533445e+00 -8.64248037e-01 -2.59280026e-01 1.45458579e+00
1.37712693e+00 4.17572670e-02 2.99708366e-01 -3.27317952e-03
6.82178438e-01 1.21019363e-01 -5.18552423e-01 4.07111079e-01
1.36734143e-01 1.43188000e-01 3.45044732e-01 -2.14436024e-01
1.35079235e-01 -1.19892240e-01 4.94156629e-01 6.23678565e-01
-7.52107576e-02 1.42177977e-02 -9.75057125e-01 2.93763161e-01
-1.77244735e+00 -7.97497213e-01 5.68364300e-02 1.92998219e+00
1.11225235e+00 1.98056638e-01 1.87522799e-01 -1.30334511e-01
1.65647805e-01 -3.52043249e-02 -9.94491756e-01 -6.94860578e-01
3.04242790e-01 5.82359493e-01 9.39905822e-01 1.07967758e+00
-8.35707843e-01 1.04478586e+00 7.74040556e+00 2.60741234e-01
-1.49111021e+00 9.28358957e-02 6.47986472e-01 -5.32442272e-01
-3.05109888e-01 2.83996761e-01 -8.72267723e-01 2.22762689e-01
1.29562628e+00 -1.95357338e-01 8.09174359e-01 9.82083738e-01
2.41991639e-01 3.01779985e-01 -1.35812867e+00 3.92824233e-01
-5.33497453e-01 -1.41871846e+00 -3.33341092e-01 2.44140953e-01
8.79300535e-01 8.37831795e-02 1.68931991e-01 3.84457707e-01
1.02774036e+00 -1.35594833e+00 5.50089121e-01 4.23666865e-01
7.23741412e-01 -6.95275843e-01 5.94439544e-03 3.48131031e-01
-9.15589154e-01 -3.05944514e-02 -3.38517547e-01 -5.74842691e-01
4.04470950e-01 1.85613602e-01 -5.41460335e-01 -2.40306601e-01
1.29417002e-01 4.59080309e-01 -1.42491892e-01 7.59157658e-01
-2.09797442e-01 6.14534736e-01 -6.47738278e-01 2.63836741e-01
4.59319592e-01 -4.15193915e-01 3.66882533e-01 1.04317379e+00
3.07093889e-01 -1.66735221e-02 -5.30629838e-03 1.55661976e+00
-8.21783766e-02 -7.03417003e-01 -3.43052983e-01 -2.73119688e-01
2.66232431e-01 1.13799047e+00 -5.27750671e-01 -6.38690591e-01
-8.30491707e-02 1.09565639e+00 3.68014783e-01 1.03591299e+00
-1.07774425e+00 -5.20372272e-01 1.15403926e+00 2.97030411e-03
4.78631943e-01 -8.48123610e-01 -5.61238706e-01 -1.39695156e+00
-1.85872182e-01 -5.41624665e-01 -4.39708531e-02 -7.72119164e-01
-8.16507399e-01 4.85417038e-01 -3.71067151e-02 -8.63127530e-01
-9.73617554e-01 -8.48470032e-01 -6.75116837e-01 1.40724289e+00
-1.18339860e+00 -4.35869753e-01 1.17440306e-01 4.23838019e-01
2.16579903e-02 1.76870748e-01 1.01363158e+00 3.97893153e-02
-8.51599395e-01 4.33475167e-01 2.89622873e-01 3.58675927e-01
1.22355424e-01 -1.34450281e+00 9.84412134e-01 8.94782722e-01
-4.78896052e-01 1.07383466e+00 7.04172194e-01 -5.67437887e-01
-1.40421355e+00 -1.10559416e+00 7.06173003e-01 -3.31589311e-01
1.03515947e+00 -4.59080935e-01 -1.02772915e+00 1.48237908e+00
2.25047663e-01 2.17686549e-01 1.28244594e-01 -3.94826531e-02
-3.04957271e-01 2.67401844e-01 -9.06347752e-01 6.96261764e-01
9.99448776e-01 -6.55390680e-01 -1.96008667e-01 1.68626636e-01
9.43112314e-01 -1.05242002e+00 -9.17821825e-01 -4.84997369e-02
4.78531212e-01 -6.69893682e-01 1.07024312e+00 -9.87843633e-01
6.35648191e-01 -1.85492128e-01 3.83228123e-01 -1.16772366e+00
-2.01968297e-01 -1.18825972e+00 -8.56941104e-01 6.64142728e-01
6.23654425e-01 -9.24235046e-01 8.28878701e-01 1.33479142e+00
-1.71804830e-01 -1.05777800e+00 -5.15335202e-01 -6.71123445e-01
7.94612348e-01 -3.06766689e-01 6.32626176e-01 7.05168009e-01
-1.04914442e-01 5.57053760e-02 -1.58766881e-01 3.69557381e-01
2.55734295e-01 7.25636929e-02 7.52715290e-01 -9.16872859e-01
-9.47245479e-01 -7.85292566e-01 -6.67022690e-02 -1.46018326e+00
9.12275836e-02 -9.22152102e-01 -1.70767978e-01 -1.48867035e+00
-2.64011800e-01 -2.24307150e-01 1.16664261e-01 7.40775645e-01
2.40851372e-01 3.83250341e-02 9.75703597e-02 1.48477510e-01
-2.27617189e-01 4.47366118e-01 1.17128181e+00 3.33049655e-01
-5.12599707e-01 -1.48670122e-01 -1.66844562e-01 9.04192507e-01
8.81144583e-01 -5.06415606e-01 -4.20126349e-01 -6.30092740e-01
2.89590269e-01 5.64116061e-01 8.52088392e-01 -9.97421265e-01
4.17919308e-01 -2.32263446e-01 5.28011382e-01 -4.03811604e-01
6.00549400e-01 -3.84167880e-01 3.26139599e-01 5.73717952e-01
-6.56912744e-01 3.88380080e-01 2.66299933e-01 -1.64146703e-02
1.45228401e-01 -1.78402320e-01 8.43961000e-01 -1.68818727e-01
8.21317360e-02 5.71214378e-01 -8.55184078e-01 3.06633294e-01
5.62855124e-01 -7.53491223e-02 -1.32319063e-01 -4.68630224e-01
-1.16028666e+00 2.61572272e-01 7.67112553e-01 -2.75675774e-01
3.94439772e-02 -1.14722633e+00 -1.19411565e-01 4.34055597e-01
-7.90806413e-01 5.80801666e-01 -1.27821416e-01 5.41429937e-01
-8.68964493e-01 2.67257363e-01 7.71755958e-03 -3.05815756e-01
-5.80085218e-01 3.45845819e-01 1.25193012e+00 -4.47816551e-01
-6.66176498e-01 7.53583848e-01 1.26205772e-01 -4.96061444e-01
3.43329400e-01 -6.04572654e-01 5.92013538e-01 -3.05453151e-01
2.90880203e-01 5.05188406e-01 -3.68037671e-01 3.63177247e-02
-3.38116921e-02 3.41641039e-01 4.21103597e-01 -6.58775210e-01
1.35249686e+00 1.73362792e-01 -2.92811066e-01 5.61828732e-01
1.66459787e+00 -3.55565637e-01 -1.91856194e+00 2.35164553e-01
-5.59801280e-01 2.76029736e-01 3.31126899e-03 -5.45473039e-01
-1.32474732e+00 1.29056764e+00 1.44232363e-01 1.24080114e-01
7.81290948e-01 -5.26981413e-01 8.88972342e-01 6.54687703e-01
-1.17333092e-01 -8.32230449e-01 -2.30446711e-01 9.93552506e-01
5.26639640e-01 -5.61871290e-01 -1.47938788e-01 2.59013083e-02
-1.04511611e-01 1.41503978e+00 4.50460464e-01 -7.10670948e-01
8.06140125e-01 1.07043588e+00 2.76811272e-02 -3.65445167e-02
-1.13641095e+00 2.13460445e-01 -2.89301462e-02 2.06612036e-01
4.47112471e-01 -4.22456920e-01 1.67176023e-01 4.41023558e-01
-3.67598444e-01 2.29608059e-01 6.24439418e-01 8.47513258e-01
-2.04061821e-01 -1.15319204e+00 1.65419970e-02 2.01380253e-01
-2.80234098e-01 -2.40700945e-01 -3.22884731e-02 7.92174757e-01
-3.05198103e-01 1.99943662e-01 3.95353526e-01 7.04254583e-02
-1.53046340e-01 5.30007362e-01 3.37859690e-01 -4.52191591e-01
-6.87390089e-01 7.88270403e-03 -1.27899140e-01 -6.37898505e-01
2.79907167e-01 -4.57022309e-01 -1.75738585e+00 -6.22390330e-01
1.67185254e-02 -1.96645856e-02 5.29140830e-01 8.24242592e-01
6.05944157e-01 7.15233982e-01 3.50012213e-01 -1.07252276e+00
-9.00322974e-01 -5.40682614e-01 -1.77381039e-01 -6.36741817e-02
7.53387690e-01 -1.43194601e-01 -6.88103557e-01 1.65724337e-01]
|
[6.591987133026123, 3.436844825744629]
|
80d28c27-5e14-4179-9869-3257b691efce
|
provable-multi-instance-deep-auc-maximization
|
2305.08040
| null |
https://arxiv.org/abs/2305.08040v4
|
https://arxiv.org/pdf/2305.08040v4.pdf
|
Provable Multi-instance Deep AUC Maximization with Stochastic Pooling
|
This paper considers a novel application of deep AUC maximization (DAM) for multi-instance learning (MIL), in which a single class label is assigned to a bag of instances (e.g., multiple 2D slices of a CT scan for a patient). We address a neglected yet non-negligible computational challenge of MIL in the context of DAM, i.e., bag size is too large to be loaded into {GPU} memory for backpropagation, which is required by the standard pooling methods of MIL. To tackle this challenge, we propose variance-reduced stochastic pooling methods in the spirit of stochastic optimization by formulating the loss function over the pooled prediction as a multi-level compositional function. By synthesizing techniques from stochastic compositional optimization and non-convex min-max optimization, we propose a unified and provable muli-instance DAM (MIDAM) algorithm with stochastic smoothed-max pooling or stochastic attention-based pooling, which only samples a few instances for each bag to compute a stochastic gradient estimator and to update the model parameter. We establish a similar convergence rate of the proposed MIDAM algorithm as the state-of-the-art DAM algorithms. Our extensive experiments on conventional MIL datasets and medical datasets demonstrate the superiority of our MIDAM algorithm.
|
['Dixian Zhu', 'Tianbao Yang', 'Xiaodong Wu', 'Milan Sonka', 'Yaxing Wang', 'Zhi Chen', 'Bokun Wang']
|
2023-05-14
| null | null | null | null |
['stochastic-optimization']
|
['methodology']
|
[ 3.89641851e-01 1.94368333e-01 -1.61770880e-02 -6.12007022e-01
-1.60523605e+00 -1.12255089e-01 3.23882326e-02 3.07843447e-01
-6.55053735e-01 9.21182513e-01 1.43832844e-02 -1.15917087e-01
-2.16857746e-01 -6.85892701e-01 -1.06645298e+00 -1.02657592e+00
5.86657934e-02 5.68573475e-01 6.85001761e-02 3.78724426e-01
1.62736867e-02 2.43295163e-01 -1.08086896e+00 5.09016275e-01
8.73182774e-01 1.29151225e+00 3.44882846e-01 5.81148148e-01
-7.34831691e-02 8.87394011e-01 -3.55579913e-01 -6.72756433e-01
7.53215104e-02 -3.01083207e-01 -7.75956333e-01 2.28745967e-01
6.73516393e-01 -2.91454464e-01 2.93434225e-02 8.45876813e-01
8.31493497e-01 3.04743767e-01 6.39948666e-01 -1.07760191e+00
-5.27804978e-02 7.12961018e-01 -7.65402555e-01 2.67539531e-01
-2.79434085e-01 9.86317359e-03 1.01682925e+00 -9.52741444e-01
3.50064188e-01 1.14008009e+00 7.37232983e-01 3.73985291e-01
-1.19635177e+00 -3.42178911e-01 4.87053692e-01 -2.43342608e-01
-1.12197459e+00 -2.56612629e-01 4.80642259e-01 -3.62556040e-01
7.48482347e-01 4.32798505e-01 3.69785398e-01 7.91164279e-01
4.37497735e-01 1.36884737e+00 8.48140419e-01 -7.85902441e-02
2.73945928e-01 2.70944774e-01 3.49430591e-01 8.53225708e-01
1.39367670e-01 -4.12430465e-01 -2.94999480e-01 -4.19143230e-01
5.11658192e-01 1.83977544e-01 -6.75810874e-03 -1.73959702e-01
-1.18848896e+00 9.16346908e-01 4.52984631e-01 -1.04525901e-01
-6.24016225e-01 4.93974090e-01 5.78005552e-01 -2.68105716e-01
9.54212904e-01 7.67304301e-02 -5.93065560e-01 1.29827574e-01
-1.25102973e+00 4.68904555e-01 8.12852740e-01 8.56881559e-01
5.40332377e-01 -2.34294772e-01 -7.26856589e-01 6.81125462e-01
2.23541215e-01 3.33287090e-01 3.10691357e-01 -5.18774271e-01
7.86825061e-01 3.09251308e-01 -1.43320272e-02 -7.98192918e-01
-6.08186424e-01 -6.62922621e-01 -8.63296986e-01 -2.45201424e-01
3.30245435e-01 -3.17776859e-01 -6.76998973e-01 1.84536231e+00
6.17660403e-01 4.91563261e-01 -2.03118533e-01 8.11115921e-01
8.01422834e-01 6.62218988e-01 1.74403727e-01 -3.50837082e-01
1.27988791e+00 -1.43354082e+00 -4.53566492e-01 -4.28075552e-01
7.24116623e-01 -4.44720864e-01 8.80881071e-01 3.39570284e-01
-1.34078181e+00 -8.03598017e-02 -9.09257770e-01 2.47016512e-02
-1.42896160e-01 1.88689083e-01 7.06547379e-01 5.94385624e-01
-6.79962277e-01 6.71033084e-01 -1.01680779e+00 2.31560081e-01
7.68357575e-01 3.02252054e-01 -4.01051566e-02 -1.53985173e-01
-8.10578525e-01 6.16068602e-01 3.54970008e-01 3.87440264e-01
-1.00296450e+00 -1.16881299e+00 -9.85501707e-01 5.34918606e-02
5.34133196e-01 -7.86576152e-01 1.17765617e+00 -7.01190710e-01
-1.39934540e+00 7.70876706e-01 -4.28751230e-01 -5.90676665e-01
8.52214754e-01 -3.88578236e-01 2.83519328e-01 -4.51979414e-02
7.45027363e-02 2.58869946e-01 8.83273602e-01 -1.07832205e+00
-5.07266343e-01 -3.86445254e-01 1.62947029e-02 1.96052387e-01
-1.39474928e-01 7.83225894e-03 -4.92648333e-01 -5.50172687e-01
1.21172145e-01 -6.84205949e-01 -7.74524927e-01 -3.04177046e-01
-7.77914762e-01 9.33845993e-03 2.36737341e-01 -7.33817816e-01
1.43556094e+00 -1.96429491e+00 2.30836302e-01 3.02696656e-02
3.26133013e-01 -9.16691422e-02 5.89177907e-02 3.17543931e-02
2.27338344e-01 9.82612185e-03 -7.06191897e-01 -1.08931661e+00
1.53393850e-01 5.41341379e-02 -2.05655962e-01 6.22307122e-01
1.70104891e-01 9.41213071e-01 -1.05580306e+00 -8.52074325e-01
-3.21887583e-02 4.80420798e-01 -7.41843402e-01 2.04982743e-01
-3.02874714e-01 3.80959511e-01 -6.77575886e-01 5.31321883e-01
1.01638484e+00 -5.21070659e-01 -1.13102555e-01 -3.31632078e-01
2.15410128e-01 2.18408883e-01 -1.23648286e+00 1.73671222e+00
-9.12931621e-01 2.81462539e-02 -8.58987123e-02 -1.04540753e+00
4.20009136e-01 1.18192025e-01 5.83897650e-01 -1.20047435e-01
1.97822854e-01 3.22349608e-01 -5.05451679e-01 -2.64111042e-01
4.24054056e-01 -4.02824819e-01 -3.03323328e-01 5.41067600e-01
5.64750805e-02 -1.29856631e-01 1.76276177e-01 -2.20466219e-02
9.05841947e-01 3.83299217e-02 1.96246132e-01 -1.91619992e-01
4.91414994e-01 -3.37898165e-01 6.57978833e-01 1.11004019e+00
-1.41561538e-01 7.93207288e-01 4.31428522e-01 -3.74535948e-01
-8.55011523e-01 -1.14067674e+00 -4.16869283e-01 1.08490777e+00
-1.08058698e-01 -1.39611363e-01 -9.79984879e-01 -8.54731917e-01
2.01956667e-02 6.30861759e-01 -4.90066111e-01 1.39904022e-02
-6.42714977e-01 -1.79820371e+00 2.50341594e-01 5.24206519e-01
3.17832202e-01 -9.79540884e-01 -7.23952889e-01 5.47634721e-01
-6.71440139e-02 -1.13448763e+00 -7.77822614e-01 3.53438020e-01
-1.03395772e+00 -5.72884440e-01 -8.58561754e-01 -7.06027985e-01
7.42648840e-01 -2.50364691e-01 1.29156041e+00 -2.89076686e-01
-3.99618685e-01 2.17716470e-01 1.69142544e-01 -4.26816881e-01
3.74342948e-02 2.97174037e-01 -2.03889161e-01 3.65179867e-01
-4.15958837e-03 -4.61050510e-01 -7.75297582e-01 6.35108911e-03
-8.66465330e-01 2.43733093e-01 5.87226570e-01 1.07993889e+00
1.13987303e+00 -1.93593621e-01 7.11962163e-01 -1.15815532e+00
6.45891309e-01 -7.48786807e-01 -5.95095634e-01 4.27448869e-01
-2.90474862e-01 1.05576724e-01 4.96242881e-01 -5.01635730e-01
-9.56460655e-01 2.18915433e-01 -1.61961779e-01 -2.81867594e-01
3.28843981e-01 6.95234478e-01 -5.83653525e-03 4.24011145e-03
1.18239507e-01 3.61237705e-01 -1.49514884e-01 -4.23015058e-01
3.86465669e-01 5.49413085e-01 2.16819093e-01 -7.61176825e-01
1.29071191e-01 5.46241641e-01 3.00730914e-02 -4.17951345e-01
-1.45041466e+00 -3.32571507e-01 -3.87484610e-01 5.19340225e-02
8.93653512e-01 -8.18888843e-01 -8.81437182e-01 6.21962428e-01
-1.17365527e+00 -4.62363750e-01 -2.94159412e-01 5.12728453e-01
-7.81836569e-01 1.77500620e-01 -8.19330513e-01 -9.70217645e-01
-7.18695879e-01 -1.46066308e+00 1.45975852e+00 1.66785195e-01
2.99419016e-01 -1.06533635e+00 3.74674574e-02 5.18295765e-01
2.40081072e-01 5.48238277e-01 7.15495706e-01 -6.14736915e-01
-6.19391203e-01 -4.04078066e-01 -2.97440410e-01 5.15364766e-01
-2.18146026e-01 -5.34462154e-01 -1.02233958e+00 -3.16581786e-01
3.27572614e-01 -4.78610754e-01 9.97065127e-01 8.49291921e-01
1.71753573e+00 -3.08096379e-01 -3.33965123e-01 8.54153037e-01
1.69739068e+00 -1.00716591e-01 2.00658321e-01 -9.58554447e-04
7.64281154e-01 1.20307662e-01 5.14608741e-01 7.59524465e-01
4.47136402e-01 4.96802002e-01 2.91335583e-01 -1.05299987e-01
3.10850710e-01 5.20554557e-03 1.22320503e-01 8.15502882e-01
2.77237773e-01 -2.52025753e-01 -6.77261651e-01 6.95355654e-01
-2.07794571e+00 -4.94887412e-01 1.29634857e-01 2.39808083e+00
9.95784700e-01 1.10198446e-01 -3.40741836e-02 -3.35084140e-01
5.72385311e-01 3.78065616e-01 -6.36494279e-01 -3.83504301e-01
1.06637307e-01 2.52310127e-01 7.59979546e-01 5.91566145e-01
-1.41427779e+00 7.38124311e-01 5.55789185e+00 1.16791558e+00
-8.73636603e-01 5.58730900e-01 1.22794878e+00 -5.40384769e-01
-8.54355097e-02 -4.98598754e-01 -1.13702846e+00 6.59682751e-01
1.05056679e+00 -3.49347405e-02 3.30737233e-01 9.61074114e-01
1.23531252e-01 -2.69999057e-01 -1.21018827e+00 1.02154207e+00
1.32759407e-01 -1.34404242e+00 -1.33156985e-01 -1.72546178e-01
8.61480117e-01 1.62343621e-01 1.99417412e-01 3.07785273e-01
1.40840024e-01 -1.01780641e+00 6.93028867e-01 5.81453681e-01
3.81871164e-01 -8.97880793e-01 8.52997780e-01 4.35996264e-01
-9.93178546e-01 -3.08253206e-02 -5.35923302e-01 3.63008738e-01
5.04485905e-01 1.11621761e+00 -5.99617720e-01 6.03413045e-01
4.86137509e-01 3.34794760e-01 -1.12069026e-01 1.12473500e+00
8.29735547e-02 5.86561263e-01 -6.71143532e-01 -1.64020211e-01
6.13934636e-01 -2.53718764e-01 5.37376106e-01 1.43862855e+00
1.58877268e-01 -9.79103073e-02 3.75779152e-01 8.96016955e-01
-2.37605184e-01 3.64429235e-01 -2.66959742e-02 1.98896661e-01
1.28824800e-01 1.29945707e+00 -6.13340497e-01 -3.81462961e-01
-3.78593683e-01 1.04945922e+00 6.01478219e-01 7.11163878e-02
-1.10186958e+00 -2.97910452e-01 4.45188344e-01 -2.09899813e-01
4.28866953e-01 2.06357747e-01 -6.12593889e-01 -1.09546626e+00
3.93595576e-01 -5.31547427e-01 6.45259380e-01 -3.67473245e-01
-1.54936981e+00 7.65922129e-01 -2.40304433e-02 -8.66028309e-01
-5.81491739e-02 -5.10168195e-01 -6.07157648e-01 1.06567621e+00
-1.59254909e+00 -1.22637951e+00 -1.38038754e-01 3.41525555e-01
5.15145659e-01 1.86176658e-01 5.64448655e-01 5.37100255e-01
-8.85976672e-01 8.68582368e-01 2.99615294e-01 -1.44684553e-01
3.76198351e-01 -1.34193945e+00 1.85502082e-01 5.97252786e-01
-1.59511417e-01 1.24186933e-01 5.16897261e-01 -4.30960953e-01
-1.26171374e+00 -1.43143415e+00 9.64564085e-01 -4.73027408e-01
6.19944096e-01 -5.99412322e-01 -6.60240531e-01 7.58448601e-01
-3.70305151e-01 4.18624282e-01 6.43149614e-01 -1.37120292e-01
2.02611640e-01 -2.83106834e-01 -1.37635648e+00 4.30910617e-01
8.03460717e-01 -2.07119167e-01 -2.15410829e-01 8.79690766e-01
1.04338777e+00 -8.11936498e-01 -9.33884323e-01 3.01883370e-01
3.04865479e-01 -5.69209218e-01 1.11893797e+00 -8.36095452e-01
6.57215178e-01 1.58458039e-01 -2.68557787e-01 -1.18045974e+00
9.60435718e-02 -5.76503158e-01 -3.72753203e-01 8.99734676e-01
6.42164886e-01 -7.10528374e-01 7.95621753e-01 6.69920087e-01
-2.20200956e-01 -1.57348156e+00 -1.13738275e+00 -5.51553011e-01
2.89335608e-01 -7.03293383e-01 6.04148984e-01 5.27420104e-01
-1.96723565e-01 -1.41362518e-01 -5.75727344e-01 3.56280982e-01
1.09191906e+00 2.58639663e-01 3.86982143e-01 -5.51910639e-01
-7.14542270e-01 -2.31956452e-01 -2.90379822e-01 -9.16042328e-01
1.74595743e-01 -1.02218747e+00 3.54590237e-01 -1.51136708e+00
6.33902550e-01 -5.81931710e-01 -6.74621701e-01 1.82801753e-01
-5.08128643e-01 8.98605436e-02 1.58908233e-01 -1.75119072e-01
-8.16483378e-01 5.29469073e-01 1.19357276e+00 -1.96978420e-01
-1.92977905e-01 3.56953442e-01 -5.93404770e-01 6.05938554e-01
5.01466990e-01 -7.67719328e-01 -3.61843050e-01 -4.19940948e-01
2.15350732e-01 2.42454499e-01 3.43151748e-01 -8.36923361e-01
2.06781834e-01 1.55639686e-02 1.16351038e-01 -7.12431371e-01
3.29223931e-01 -5.41848421e-01 -2.44078800e-01 3.66079837e-01
-4.21458602e-01 -2.33704925e-01 2.37566695e-01 5.22880256e-01
-1.73936322e-01 -4.29833978e-01 8.24456453e-01 -2.35674366e-01
-1.72698691e-01 7.71147490e-01 4.43317965e-02 4.36092913e-01
8.77769709e-01 2.12434053e-01 -1.19694769e-02 3.26094851e-02
-9.25932288e-01 4.99787390e-01 -2.49201789e-01 -1.29248545e-01
5.81638157e-01 -1.26954126e+00 -1.03725743e+00 -2.42844939e-01
-1.19273365e-01 5.27131319e-01 7.05654323e-01 1.36137390e+00
-3.61371577e-01 2.49963492e-01 3.53602886e-01 -6.61469042e-01
-8.84982347e-01 4.94071305e-01 5.45043945e-01 -1.00148189e+00
-6.30407691e-01 1.31098557e+00 5.59095025e-01 -5.68550825e-01
2.89400399e-01 -4.87682283e-01 2.43113458e-01 8.81909728e-02
6.86293364e-01 3.53002846e-01 3.18763077e-01 -2.02039167e-01
-4.31627214e-01 1.64232954e-01 -3.47155690e-01 -2.51763463e-02
1.65089452e+00 6.14538118e-02 -1.88128293e-01 6.93188787e-01
1.44835901e+00 -3.55183214e-01 -1.33617377e+00 -3.45127821e-01
-2.07495481e-01 -2.18267024e-01 3.47896427e-01 -7.09858060e-01
-1.22179949e+00 8.62986982e-01 5.55976212e-01 -3.23567361e-01
1.08250368e+00 6.69085830e-02 1.14717376e+00 3.06335688e-01
3.45581859e-01 -1.05043316e+00 -2.59350777e-01 2.66219765e-01
7.15004981e-01 -1.34641957e+00 7.68766403e-02 -2.99056351e-01
-6.46988392e-01 8.90461504e-01 3.61567020e-01 -2.14783818e-01
9.30639863e-01 4.60432976e-01 -3.85468423e-01 -1.22988373e-01
-6.80193484e-01 1.19580515e-01 4.48961198e-01 -6.00798652e-02
3.52497309e-01 2.30999529e-01 -4.11872238e-01 1.03579688e+00
1.18828639e-01 1.44772530e-01 2.21813053e-01 9.74362731e-01
-1.20696269e-01 -5.50839722e-01 -6.93483427e-02 7.42071331e-01
-8.25864732e-01 -5.02984583e-01 4.00659084e-01 3.96959335e-01
-2.54152603e-02 3.88993859e-01 9.82507989e-02 -2.75995340e-02
5.47947213e-02 8.31135511e-02 5.39229035e-01 -5.62286556e-01
-7.72878706e-01 -6.08502775e-02 -1.34583637e-01 -6.23586059e-01
-1.66012138e-01 -8.14821243e-01 -1.05087948e+00 7.98788667e-02
-3.06035131e-01 -4.97337580e-02 7.49458373e-01 1.13361537e+00
1.44395828e-01 7.13696957e-01 6.23627782e-01 -9.53946173e-01
-9.39730823e-01 -8.80210519e-01 -6.12078846e-01 2.32777372e-01
3.88244271e-01 -4.43589151e-01 -5.06460786e-01 -2.29848966e-01]
|
[14.439160346984863, -2.0214924812316895]
|
b44e1b18-cf8d-4c09-b81d-c9c3222b84c1
|
performance-aware-approximation-of-global
|
2303.11923
| null |
https://arxiv.org/abs/2303.11923v1
|
https://arxiv.org/pdf/2303.11923v1.pdf
|
Performance-aware Approximation of Global Channel Pruning for Multitask CNNs
|
Global channel pruning (GCP) aims to remove a subset of channels (filters) across different layers from a deep model without hurting the performance. Previous works focus on either single task model pruning or simply adapting it to multitask scenario, and still face the following problems when handling multitask pruning: 1) Due to the task mismatch, a well-pruned backbone for classification task focuses on preserving filters that can extract category-sensitive information, causing filters that may be useful for other tasks to be pruned during the backbone pruning stage; 2) For multitask predictions, different filters within or between layers are more closely related and interacted than that for single task prediction, making multitask pruning more difficult. Therefore, aiming at multitask model compression, we propose a Performance-Aware Global Channel Pruning (PAGCP) framework. We first theoretically present the objective for achieving superior GCP, by considering the joint saliency of filters from intra- and inter-layers. Then a sequentially greedy pruning strategy is proposed to optimize the objective, where a performance-aware oracle criterion is developed to evaluate sensitivity of filters to each task and preserve the globally most task-related filters. Experiments on several multitask datasets show that the proposed PAGCP can reduce the FLOPs and parameters by over 60% with minor performance drop, and achieves 1.2x$\sim$3.3x acceleration on both cloud and mobile platforms.
|
['Bin Wang', 'Jiayuan Fan', 'Tao Chen', 'Bo Zhang', 'Hancheng Ye']
|
2023-03-21
| null | null | null | null |
['model-compression']
|
['methodology']
|
[ 5.46477973e-01 -2.25778952e-01 -1.67944565e-01 -2.44250610e-01
-5.35514951e-01 -1.17757179e-01 -1.16866417e-02 2.76139975e-01
-5.04934072e-01 7.03587890e-01 -5.67669943e-02 -3.48508537e-01
-3.06859553e-01 -5.77885926e-01 -7.66496778e-01 -7.03368783e-01
-3.84122953e-02 -5.32550998e-02 8.34667683e-01 9.61836576e-02
3.90307456e-01 1.30969435e-01 -1.58331537e+00 7.09396660e-01
1.01785243e+00 1.41493738e+00 8.68175149e-01 3.29647124e-01
-1.24365836e-01 5.38581789e-01 -7.47810900e-01 -3.75835478e-01
5.20775795e-01 -2.46256893e-03 -6.29417658e-01 -2.08755806e-01
2.65472680e-01 -9.23119411e-02 -3.31130577e-03 1.18372095e+00
4.55759823e-01 -1.49113685e-01 3.37788105e-01 -9.79286313e-01
1.87915474e-01 6.46196663e-01 -7.04983532e-01 6.92915499e-01
-4.00171727e-01 -1.35728702e-01 9.09578562e-01 -6.60609603e-01
1.61348641e-01 1.05560946e+00 5.90112269e-01 1.66417673e-01
-1.13318646e+00 -9.62289631e-01 6.69118822e-01 2.19027072e-01
-1.47792113e+00 -4.66952890e-01 4.69296902e-01 -3.23910177e-01
8.98235500e-01 4.88073975e-01 4.29109812e-01 6.87686205e-01
6.20232403e-01 7.09396362e-01 1.07000327e+00 -2.64733583e-01
4.86194976e-02 3.12813282e-01 3.94799739e-01 5.37917793e-01
4.89858061e-01 -2.51461953e-01 -8.52364659e-01 -2.37612814e-01
5.77812493e-01 1.24365434e-01 -3.47457200e-01 -2.16174871e-01
-9.45418000e-01 4.82137442e-01 1.74833089e-01 8.40609893e-02
-5.37888467e-01 5.61802424e-02 6.20820820e-01 3.82080168e-01
3.04190129e-01 2.67091334e-01 -8.29477429e-01 -1.41870514e-01
-9.46635902e-01 2.83732206e-01 4.81712401e-01 1.06810164e+00
9.20787990e-01 -1.82460383e-01 -6.82456970e-01 9.28319156e-01
-3.29010151e-02 1.29893094e-01 4.26252872e-01 -6.23099446e-01
9.25717413e-01 6.78917468e-01 -3.23841184e-01 -1.08986890e+00
-3.54234010e-01 -9.36885118e-01 -8.31849337e-01 -1.55000046e-01
6.53612167e-02 -1.96797118e-01 -7.37079263e-01 1.51471186e+00
1.29949167e-01 1.28817171e-01 -1.94641054e-01 6.76179528e-01
4.34292227e-01 4.28347856e-01 1.70079142e-01 -5.00958920e-01
1.50908673e+00 -9.43714201e-01 -5.66173613e-01 -4.14823562e-01
5.70267200e-01 -1.07704198e+00 1.04662013e+00 5.67351758e-01
-1.01687467e+00 -7.13534474e-01 -1.14678359e+00 2.07841143e-01
-7.97892585e-02 1.47658825e-01 5.10513783e-01 7.23756790e-01
-4.87165093e-01 3.59068304e-01 -6.56885982e-01 5.18328063e-02
6.55394137e-01 4.84558374e-01 3.08810562e-01 -1.76727206e-01
-1.06413996e+00 4.46023256e-01 5.21425605e-01 -9.92293656e-02
-9.18975174e-01 -9.12445307e-01 -3.16623479e-01 3.91842425e-01
7.32110202e-01 -4.89214391e-01 1.09870934e+00 -8.71950507e-01
-9.51735795e-01 1.65140390e-01 -5.40200233e-01 -6.72700346e-01
4.24496710e-01 -3.01180959e-01 -3.87000442e-01 -1.95732452e-02
1.26852229e-01 5.26131809e-01 1.17054498e+00 -1.06155443e+00
-1.28055823e+00 -3.68646473e-01 5.12484647e-02 4.85673398e-01
-7.20269322e-01 -4.25344780e-02 -9.19992983e-01 -9.14390862e-01
2.71883726e-01 -8.08440387e-01 -1.40040621e-01 -5.17992675e-01
-6.41498446e-01 1.21783003e-01 9.67477262e-01 -6.47886217e-01
1.68264639e+00 -2.24147105e+00 -1.67707756e-01 3.55755389e-01
4.28779602e-01 4.04632270e-01 1.07512444e-01 -1.36725813e-01
1.47321776e-01 2.05907598e-01 -5.14294021e-02 -3.43084633e-01
-4.95386899e-01 9.06804502e-02 -2.33528882e-01 -1.13165611e-02
-1.97150573e-01 5.41000843e-01 -3.74238223e-01 -5.82440197e-01
-1.49729639e-01 1.89847469e-01 -5.89871287e-01 -2.51254737e-01
6.18007146e-02 2.31856257e-01 -5.27138591e-01 6.95936739e-01
9.63316381e-01 -6.06471449e-02 2.18826354e-01 -2.50881314e-01
-9.85895395e-02 5.29202759e-01 -1.40007150e+00 1.33089936e+00
-6.18335903e-01 3.13464075e-01 2.82344878e-01 -9.32301581e-01
8.38190734e-01 -3.81080657e-02 3.92953426e-01 -1.00198662e+00
-1.35234976e-02 2.67618209e-01 1.96423605e-01 -1.56177416e-01
4.40951526e-01 1.92617118e-01 -9.18804258e-02 1.92412868e-01
-3.50035131e-01 4.92175907e-01 7.96304792e-02 1.44600257e-01
1.14993620e+00 -3.12685937e-01 1.35353170e-02 -5.72143197e-01
5.95835388e-01 -1.40950382e-01 1.06354678e+00 1.00295651e+00
-1.69936255e-01 2.95807421e-01 5.45822144e-01 -4.51691747e-01
-4.74012733e-01 -6.71874940e-01 -2.74561018e-01 1.33939910e+00
3.74367416e-01 -6.45369828e-01 -5.31274080e-01 -7.75910497e-01
1.68867633e-01 3.78840864e-01 -3.10874850e-01 -2.16595024e-01
-6.51760399e-01 -1.00842583e+00 3.78031254e-01 3.69708061e-01
7.96072960e-01 -7.53888488e-01 -1.08972228e+00 3.83019179e-01
-2.61979640e-01 -1.14614546e+00 -4.22640413e-01 7.73937225e-01
-1.22320068e+00 -1.02291107e+00 -5.42471468e-01 -6.96849644e-01
5.41637659e-01 7.03740001e-01 8.42151821e-01 1.75381973e-01
-1.49572104e-01 -3.82566184e-01 -2.85364509e-01 -6.37615681e-01
1.05558842e-01 5.12152016e-01 -1.83777839e-01 2.23976597e-01
3.35921526e-01 -4.99876469e-01 -5.61491609e-01 5.64892292e-01
-5.84886670e-01 1.63545683e-01 1.02958417e+00 6.74452424e-01
1.03885984e+00 6.12586379e-01 5.12499928e-01 -1.25226712e+00
7.71259010e-01 -1.95725396e-01 -4.42534536e-01 2.27020025e-01
-9.03542399e-01 -5.61749525e-02 7.74978340e-01 -3.37849289e-01
-1.06517243e+00 1.40854955e-01 3.08817267e-01 -4.94436353e-01
1.79500848e-01 2.47444823e-01 -3.70997518e-01 -1.17920995e-01
4.31043297e-01 4.24468040e-01 -3.65817606e-01 -6.36366904e-01
-2.77470380e-01 6.17097735e-01 1.55457303e-01 -5.13973951e-01
3.81027043e-01 4.76159871e-01 3.95140424e-02 -6.48755014e-01
-5.73554277e-01 -4.94473100e-01 -4.40325290e-01 9.07606855e-02
4.80628908e-01 -1.07247174e+00 -5.56218505e-01 2.62166470e-01
-8.95359099e-01 -1.95393339e-01 -2.79064663e-02 3.17904621e-01
-1.37810186e-01 3.59304190e-01 -1.84348315e-01 -5.71239769e-01
-4.55286741e-01 -1.43420470e+00 8.74868691e-01 1.51038408e-01
-4.56567220e-02 -4.70290869e-01 -6.24752164e-01 2.08558232e-01
3.49732459e-01 -2.64466286e-01 1.27741683e+00 -6.34586692e-01
-7.77392983e-01 -2.02900600e-02 -4.04621005e-01 4.94969547e-01
-5.38880937e-02 -5.93692482e-01 -8.24963450e-01 -3.29153031e-01
1.53621078e-01 9.56564024e-02 1.26045763e+00 5.57198107e-01
1.66488087e+00 -2.32835263e-01 -7.96306670e-01 7.91646004e-01
1.28287840e+00 3.76803339e-01 4.83951360e-01 4.21891958e-01
6.90640867e-01 4.64133620e-01 9.15512085e-01 5.20595193e-01
2.20951021e-01 7.22333968e-01 5.27299762e-01 5.02452329e-02
-1.48154333e-01 -7.45968893e-02 2.76184499e-01 6.28713608e-01
1.37809411e-01 -9.62573960e-02 -6.37913108e-01 5.18905461e-01
-1.87297940e+00 -3.19254845e-01 -2.84706444e-01 2.58741450e+00
6.34942293e-01 7.45745420e-01 -5.56488447e-02 4.05578822e-01
7.82121062e-01 2.61963904e-02 -6.20146096e-01 -4.23500180e-01
-9.80813056e-02 2.31428489e-01 8.36660326e-01 1.72583893e-01
-1.11414599e+00 8.69085968e-01 5.47634888e+00 1.23270154e+00
-1.08571804e+00 4.29867208e-01 8.88104677e-01 -5.41543663e-01
4.94954251e-02 1.55150235e-01 -1.34816039e+00 7.31240630e-01
5.39468706e-01 -8.55502561e-02 3.75099897e-01 7.65957475e-01
1.27694115e-01 -3.81037265e-01 -7.54066408e-01 1.02063584e+00
-2.13613808e-01 -1.10350406e+00 1.61296740e-01 2.12742567e-01
6.34826779e-01 -1.64634958e-01 1.31420612e-01 3.57914507e-01
-2.57805288e-01 -6.23603523e-01 8.97860765e-01 2.21633688e-01
6.00182414e-01 -9.80391800e-01 6.08436048e-01 5.53886533e-01
-1.38177681e+00 -5.89540720e-01 -7.34878719e-01 1.51050091e-02
-1.23780265e-01 1.00418663e+00 -6.51041389e-01 5.58468521e-01
1.22313619e+00 3.89028132e-01 -7.02863216e-01 1.26792598e+00
1.53659344e-01 5.07364750e-01 -3.38265687e-01 1.66522786e-01
3.31467912e-02 2.26133749e-01 6.26291156e-01 1.26283765e+00
2.85091281e-01 -2.20220178e-01 2.85085768e-01 4.03990507e-01
-8.80988687e-02 3.68755370e-01 -1.06046170e-01 4.88422453e-01
7.46727169e-01 1.08192003e+00 -8.64721298e-01 -2.84798592e-01
-3.83377612e-01 7.83590496e-01 1.84831381e-01 3.33512872e-01
-8.51119936e-01 -4.73101914e-01 7.99661756e-01 3.51592153e-01
5.64287841e-01 4.03586999e-02 -7.89129972e-01 -7.26755083e-01
3.77238959e-01 -8.72735381e-01 4.31624353e-01 -1.18464984e-01
-6.01793110e-01 5.82085609e-01 -1.25103682e-01 -1.13150799e+00
4.67287034e-01 -1.24288328e-01 -4.92076606e-01 1.01219952e+00
-1.82170427e+00 -6.79853976e-01 -3.41201574e-01 7.62313068e-01
8.47817600e-01 -2.26593480e-01 2.71846235e-01 5.37770033e-01
-7.69748688e-01 9.75204349e-01 -1.36861324e-01 -5.57975471e-01
6.12375259e-01 -7.31775045e-01 9.97481681e-03 9.97687459e-01
-1.78899556e-01 6.38244629e-01 3.31540227e-01 -8.66637528e-01
-1.05078471e+00 -1.38936675e+00 9.75992680e-01 1.06390320e-01
-1.65098486e-03 -4.60203797e-01 -9.23271298e-01 4.25401479e-01
-1.68753505e-01 -1.55146345e-01 4.24258173e-01 2.45475575e-01
-8.84787813e-02 -5.38931727e-01 -9.51077878e-01 3.88154030e-01
1.25091588e+00 -1.06048398e-01 3.75872999e-02 2.80220360e-01
5.99362254e-01 -2.96115458e-01 -6.14690125e-01 6.37247324e-01
4.65807736e-01 -1.09255075e+00 8.12834501e-01 -2.05338657e-01
7.51053169e-02 -1.82128161e-01 -2.55647331e-01 -9.18193102e-01
-4.27843720e-01 -7.94035673e-01 -2.59709746e-01 1.08855402e+00
4.90837514e-01 -5.34204245e-01 8.58153343e-01 1.91253275e-01
-4.82432187e-01 -1.02741528e+00 -1.03377879e+00 -9.12643850e-01
-2.64057428e-01 -5.11075199e-01 6.91131711e-01 5.60538709e-01
-5.19953549e-01 2.75327533e-01 -5.14811456e-01 3.04630697e-01
3.70463669e-01 5.24330176e-02 6.65230095e-01 -1.35308969e+00
-3.69814098e-01 -5.73274553e-01 3.17630009e-03 -1.46668363e+00
-3.45877737e-01 -8.22312474e-01 2.72864252e-02 -1.24081039e+00
3.15203637e-01 -9.16849673e-01 -6.30529821e-01 6.28275454e-01
-3.08807284e-01 5.26969954e-02 3.79860610e-01 5.98579288e-01
-6.07258022e-01 2.15071604e-01 1.28226316e+00 3.21605913e-02
-4.89233732e-01 3.42377931e-01 -1.05789924e+00 7.09091485e-01
7.08795369e-01 -8.65161538e-01 -5.94327033e-01 -6.11370802e-01
6.02691285e-02 -1.52478293e-01 7.07806572e-02 -1.34168994e+00
3.57660592e-01 1.34938762e-01 4.53518122e-01 -7.39801764e-01
2.03502461e-01 -9.02119398e-01 -9.41287130e-02 6.99562073e-01
-1.73345387e-01 1.85952783e-02 3.53519320e-01 7.92694867e-01
-1.17680632e-01 -3.11662585e-01 7.85031080e-01 5.05325943e-02
-7.49809980e-01 1.05271280e-01 -3.34925056e-01 -2.00693771e-01
1.19160843e+00 -4.89677101e-01 -2.47284472e-01 9.87340584e-02
-6.85174346e-01 4.11287546e-01 6.28423095e-02 4.51915681e-01
4.76323307e-01 -8.98358703e-01 -4.84503567e-01 4.18402284e-01
-1.63757596e-02 5.91230802e-02 4.75314021e-01 1.00499094e+00
-2.01945245e-01 7.24106073e-01 -1.65913224e-01 -6.44007504e-01
-1.60686553e+00 3.23034495e-01 1.45916715e-01 -6.02518201e-01
-6.35443270e-01 1.14462864e+00 6.30350649e-01 1.14879861e-01
5.07293701e-01 -5.21671534e-01 -1.92825526e-01 3.93490307e-02
3.94404978e-01 6.01146042e-01 5.21299720e-01 -3.35469276e-01
-4.12270635e-01 3.22017998e-01 -4.97424692e-01 3.33094418e-01
9.10615087e-01 -1.58702299e-01 1.21272653e-01 5.41191250e-02
9.78093982e-01 5.27716614e-02 -1.31443214e+00 -4.53302234e-01
-1.20524071e-01 -6.21281624e-01 3.84513974e-01 -8.90747368e-01
-1.39485526e+00 9.25545692e-01 7.23101139e-01 1.87698364e-01
1.63604844e+00 -3.11797947e-01 1.10030353e+00 1.23316146e-01
3.45136076e-01 -1.32224202e+00 7.78405890e-02 4.85980719e-01
4.81292397e-01 -6.32623613e-01 1.55124173e-01 -9.57039833e-01
-7.47109950e-01 7.79259682e-01 7.96089232e-01 2.47667074e-01
7.27629781e-01 2.35754520e-01 -4.33542162e-01 -6.78318143e-02
-7.82725334e-01 -9.44984332e-02 9.85366032e-02 4.89986479e-01
6.15557209e-02 7.18732923e-02 -5.71248293e-01 9.85027254e-01
-2.19419808e-03 -1.52996257e-01 1.72298774e-01 1.03852320e+00
-7.38784552e-01 -1.18599677e+00 -2.57330924e-01 1.05381787e+00
-7.38656580e-01 -2.88930953e-01 -7.72916898e-02 4.85388100e-01
7.55414426e-01 8.56159210e-01 8.74900520e-02 -6.04425371e-01
5.31449556e-01 6.83292747e-02 7.62400925e-02 -7.51250386e-01
-8.40931535e-01 5.66384554e-01 4.51899059e-02 -5.68289936e-01
1.78628904e-03 -6.40081286e-01 -1.02131605e+00 -1.12744503e-01
-3.60802650e-01 1.22689363e-02 6.14600062e-01 6.82259440e-01
8.30413580e-01 9.19469357e-01 4.58353400e-01 -3.13846588e-01
-3.46222311e-01 -7.73914278e-01 -5.52321553e-01 6.06593601e-02
1.31580122e-02 -8.53868902e-01 1.93266496e-02 -2.04460919e-01]
|
[8.63611888885498, 3.0218727588653564]
|
a67ade42-5846-42d1-9742-57975680e7d4
|
using-deepfake-technologies-for-word-emphasis
|
2305.07791
| null |
https://arxiv.org/abs/2305.07791v1
|
https://arxiv.org/pdf/2305.07791v1.pdf
|
Using Deepfake Technologies for Word Emphasis Detection
|
In this work, we consider the task of automated emphasis detection for spoken language. This problem is challenging in that emphasis is affected by the particularities of speech of the subject, for example the subject accent, dialect or voice. To address this task, we propose to utilize deep fake technology to produce an emphasis devoid speech for this speaker. This requires extracting the text of the spoken voice, and then using a voice sample from the same speaker to produce emphasis devoid speech for this task. By comparing the generated speech with the spoken voice, we are able to isolate patterns of emphasis which are relatively easy to detect.
|
['Lee-Ad Gottlieb', 'Eran Kaufman']
|
2023-05-12
| null | null | null | null |
['face-swapping']
|
['computer-vision']
|
[ 3.26654285e-01 2.41452292e-01 2.58545280e-01 -1.90658972e-01
-5.04796982e-01 -7.50535488e-01 5.59264243e-01 -1.47485957e-01
-2.30560824e-01 6.04571044e-01 4.78891551e-01 -4.07220334e-01
3.66426826e-01 -3.02744627e-01 -2.01328650e-01 -6.58243060e-01
2.26867586e-01 1.48588628e-01 2.82843322e-01 -3.97438437e-01
5.01458824e-01 5.07287085e-01 -1.35763788e+00 3.10332447e-01
4.57511336e-01 3.47155958e-01 2.06602529e-01 9.26745713e-01
-3.70975703e-01 7.83790290e-01 -1.37908936e+00 -8.38728994e-03
-9.26548466e-02 -8.51539433e-01 -8.48219872e-01 4.05348808e-01
4.96424794e-01 -1.40199378e-01 1.05486855e-01 1.30733752e+00
4.14369047e-01 8.41172710e-02 5.74451208e-01 -8.49708438e-01
-7.22272918e-02 9.87277210e-01 -3.45798701e-01 5.13427496e-01
4.96677816e-01 -1.71026766e-01 7.48296857e-01 -1.03759086e+00
5.38255990e-01 1.26709878e+00 4.11432058e-01 6.82990015e-01
-1.02257574e+00 -6.34989440e-01 -2.53307298e-02 -2.95278877e-01
-1.48125517e+00 -9.97032583e-01 1.13512087e+00 -3.42999876e-01
5.81600010e-01 5.09715021e-01 2.94132560e-01 1.09868801e+00
1.13268770e-01 7.09996462e-01 9.93343472e-01 -1.08246922e+00
3.64795327e-01 5.29793620e-01 3.13631386e-01 3.30163687e-01
-3.66777897e-01 -1.54145360e-01 -3.96006167e-01 -1.29527807e-01
1.70736969e-01 -5.18607855e-01 -7.13909209e-01 3.26464355e-01
-1.19774735e+00 6.94986224e-01 -2.91311089e-02 5.96147001e-01
-4.94780391e-01 -2.21280858e-01 3.31323713e-01 3.80838066e-01
5.77592015e-01 5.30099034e-01 -2.51854241e-01 -1.83049679e-01
-1.24761140e+00 2.50638545e-01 1.38363612e+00 6.25217557e-01
3.92407328e-01 2.82443255e-01 4.99525703e-02 8.29000473e-01
2.68040359e-01 4.31121767e-01 5.84369779e-01 -6.33848727e-01
3.15365851e-01 -7.93053061e-02 1.75342724e-01 -8.10080826e-01
-7.30684549e-02 -2.11307347e-01 -4.07741755e-01 3.09420645e-01
6.77082419e-01 -5.86453259e-01 -1.04562211e+00 1.49681842e+00
5.60851932e-01 -1.24766842e-01 3.19490582e-01 8.54229033e-01
6.29098654e-01 8.23403478e-01 -2.00711191e-01 -3.80361676e-01
1.45032597e+00 -9.37075317e-01 -1.16683769e+00 -4.81247067e-01
5.96463978e-01 -1.23368514e+00 1.12111449e+00 4.58137542e-01
-8.50563645e-01 -3.41832250e-01 -1.11345601e+00 2.85616308e-01
-2.79783905e-01 -1.82047691e-02 -3.02026927e-01 6.81346357e-01
-7.44375527e-01 1.64974555e-01 -2.91159004e-01 -2.83898145e-01
-1.70080200e-01 2.70549506e-01 -1.70130461e-01 3.24304253e-01
-1.29377544e+00 9.31321144e-01 2.08420649e-01 5.60165085e-02
-3.63757253e-01 -5.37463486e-01 -8.59489322e-01 1.69378564e-01
2.13749006e-01 1.59924746e-01 1.51812696e+00 -1.59180903e+00
-1.76721537e+00 6.86767757e-01 -2.26182610e-01 -2.08555818e-01
4.47950393e-01 -2.17248306e-01 -6.60616219e-01 2.83763975e-01
-1.52773112e-01 1.27938941e-01 1.57556725e+00 -1.13166034e+00
-6.08384371e-01 -1.79283574e-01 -2.04009026e-01 -1.29657730e-01
-1.91231191e-01 5.98685563e-01 1.46257788e-01 -8.23080182e-01
-2.62035485e-02 -8.41439664e-01 2.41319939e-01 -5.14002800e-01
-7.52078056e-01 -2.10445032e-01 1.30793071e+00 -9.94030535e-01
1.20101881e+00 -2.92280030e+00 -3.37493308e-02 3.71413529e-01
1.10581562e-01 5.63897014e-01 -5.92829734e-02 4.35152829e-01
-1.21899813e-01 7.57860541e-02 -2.43271470e-01 -4.02327269e-01
-1.08714506e-01 1.57310292e-01 -6.67744994e-01 5.12753427e-01
4.46554065e-01 4.32528496e-01 -8.93728375e-01 -5.08614004e-01
-8.93654488e-03 5.41971803e-01 -3.69067878e-01 4.62016851e-01
-3.09278872e-02 2.78500438e-01 -7.95745105e-02 3.25858861e-01
5.49819112e-01 6.02129042e-01 2.46085122e-01 1.18329495e-01
-4.34588104e-01 8.57261419e-01 -1.16707051e+00 9.52467203e-01
-4.15777862e-01 9.58720803e-01 4.89399225e-01 -6.94967151e-01
1.17077041e+00 7.53696799e-01 -1.80626675e-01 -1.77586451e-01
8.09875801e-02 4.42993462e-01 5.11831462e-01 -5.22728741e-01
7.24086821e-01 -2.79049248e-01 5.05773984e-02 4.29877013e-01
-8.20532516e-02 -5.56249321e-01 -8.42631757e-02 -5.91871180e-02
8.54013741e-01 -3.88852775e-01 4.74544972e-01 -5.57633638e-01
6.12579167e-01 -1.82682246e-01 2.82935828e-01 3.01237941e-01
-4.12520766e-01 7.36615300e-01 4.49850827e-01 -1.70246705e-01
-9.40686345e-01 -7.82159567e-01 -9.64396596e-02 9.69385266e-01
-3.18906873e-01 -2.94342816e-01 -1.16015351e+00 -8.31526399e-01
-2.29991317e-01 9.72852767e-01 -2.71198064e-01 -2.02064574e-01
-9.43331182e-01 3.70330215e-02 7.08850205e-01 1.42099485e-01
6.54760823e-02 -1.26321685e+00 -5.47383964e-01 2.42738456e-01
-1.95658699e-01 -8.85761797e-01 -9.54690576e-01 4.16646063e-01
-1.88221470e-01 -5.96715212e-01 -8.94816518e-01 -1.06277204e+00
5.96463025e-01 3.47116329e-02 8.67353797e-01 1.59093633e-01
-6.20356947e-02 1.13899767e-01 -4.67155844e-01 -6.30098641e-01
-1.01142585e+00 1.60226375e-01 1.64990053e-01 4.02498364e-01
3.85177433e-01 -3.02152902e-01 -6.59710392e-02 2.05296189e-01
-9.81357038e-01 -5.02498686e-01 9.54372585e-02 7.16064692e-01
-6.57483488e-02 2.07994238e-01 6.17732108e-01 -9.20490861e-01
9.12498653e-01 -3.55753273e-01 -3.95593882e-01 -2.46684682e-02
-1.01959996e-03 -7.29239061e-02 7.48751760e-01 -8.40872705e-01
-9.44366574e-01 1.24666980e-02 -4.51474488e-01 -2.76479602e-01
-4.85614091e-01 4.66919392e-01 -2.62799859e-01 1.19653910e-01
7.04127789e-01 2.91909128e-01 -2.19993410e-03 -5.03641605e-01
8.41626301e-02 1.28457880e+00 5.43219745e-01 -6.97972551e-02
4.97088671e-01 -9.39925984e-02 -4.10646915e-01 -1.77248490e+00
-6.83696210e-01 -3.70868742e-01 -6.92975163e-01 -2.54746944e-01
5.95607996e-01 -3.83865356e-01 -5.95095679e-02 4.82682616e-01
-1.57592547e+00 -1.69851527e-01 -2.93566585e-01 5.80126047e-01
-1.19549781e-01 5.74503541e-01 -3.90890568e-01 -1.20201957e+00
-2.68727839e-01 -1.19320631e+00 9.54965234e-01 5.55894710e-02
-8.53352368e-01 -9.42287624e-01 1.48534328e-01 5.10219522e-02
4.07374889e-01 -2.92853177e-01 7.49054015e-01 -1.07272780e+00
-8.21451936e-03 -1.86435089e-01 3.67521048e-01 8.36877286e-01
5.86051941e-01 4.51298058e-01 -1.21291947e+00 -3.51786464e-02
3.99830788e-01 -3.64140011e-02 3.97339880e-01 1.55563146e-01
3.47704351e-01 -5.56817293e-01 2.61962861e-02 2.49797970e-01
6.95664942e-01 1.80231780e-01 6.80624545e-01 7.29619116e-02
4.75924522e-01 1.09906113e+00 5.60008049e-01 3.30730230e-01
-1.63640499e-01 5.63239396e-01 -1.35656282e-01 2.65912544e-02
-3.63666564e-01 -3.67544219e-02 6.56108379e-01 9.94505107e-01
4.17727739e-01 -3.74077439e-01 -8.49573791e-01 6.84313178e-01
-1.28230369e+00 -8.85877788e-01 -2.65518278e-01 2.06870461e+00
9.44343150e-01 1.94749787e-01 1.83170319e-01 6.26924276e-01
9.84455585e-01 4.41689283e-01 1.05278917e-01 -9.14684594e-01
1.14848182e-01 3.03017020e-01 1.23517923e-01 9.04360116e-01
-1.02896094e+00 1.02312839e+00 6.89442730e+00 5.58637679e-01
-1.50949383e+00 -2.42096797e-01 1.96254909e-01 2.18152493e-01
-2.81089425e-01 -1.31550491e-01 -7.94595480e-01 7.28884697e-01
1.09271407e+00 -5.71242720e-02 4.63216990e-01 5.41403890e-01
3.71960700e-01 -2.00805113e-01 -1.08996284e+00 6.34516001e-01
3.00143123e-01 -6.35823011e-01 -1.54794261e-01 -6.76767752e-02
2.15666220e-01 -4.86591429e-01 -7.35873431e-02 7.07790777e-02
-1.19776033e-01 -7.50385463e-01 9.99229908e-01 2.21354201e-01
4.01491582e-01 -7.93650627e-01 6.48595929e-01 5.15392959e-01
-7.55355954e-01 4.24652517e-01 -1.22796699e-01 -2.57764131e-01
1.26124397e-01 8.40092063e-01 -1.14264643e+00 -1.17070250e-01
3.05197597e-01 -1.11566987e-02 -3.23423743e-02 8.05740118e-01
-5.56814134e-01 1.13140690e+00 -4.56624269e-01 -1.09926604e-01
1.58119112e-01 -1.30137587e-02 1.03719020e+00 1.58532012e+00
3.99982512e-01 -8.50313231e-02 1.55864462e-01 8.79728377e-01
-4.31385785e-02 1.42722070e-01 -7.95838892e-01 -4.04613823e-01
4.64391053e-01 1.06068182e+00 -5.48190296e-01 -3.53818804e-01
-1.73803717e-01 1.26568437e+00 4.03735414e-02 4.26272988e-01
-3.89435977e-01 -9.05694962e-01 8.18398952e-01 -5.86067960e-02
3.11341316e-01 -3.32052261e-01 8.83173645e-02 -8.92714262e-01
7.64739439e-02 -1.22188723e+00 -6.49153590e-02 -5.58547974e-01
-1.00399518e+00 8.89001489e-01 -4.34466243e-01 -9.83005404e-01
-5.99635780e-01 -3.63703549e-01 -8.86996090e-01 1.43985331e+00
-1.32494414e+00 -7.02356100e-01 1.87809959e-01 2.92646617e-01
1.00535607e+00 -1.08378597e-01 7.23835230e-01 9.03082564e-02
-4.34917599e-01 5.30615211e-01 -2.59214193e-01 2.99260914e-01
8.95682216e-01 -1.34766483e+00 5.70220470e-01 9.62256789e-01
2.40219668e-01 4.90927070e-01 1.23264050e+00 -4.99824464e-01
-1.06501126e+00 -7.34986067e-01 1.69966710e+00 -3.34195405e-01
7.42027819e-01 -6.19819224e-01 -1.13906300e+00 4.22693133e-01
3.85771841e-01 -1.68103635e-01 6.68126702e-01 -1.15520723e-01
-1.40533417e-01 1.84975654e-01 -1.10309815e+00 6.66129053e-01
3.02609801e-01 -7.95080304e-01 -1.19539237e+00 1.50044978e-01
8.59436631e-01 -2.75612265e-01 -2.86277652e-01 -1.24567471e-01
4.39188898e-01 -7.38301277e-01 5.45251369e-01 -4.35534120e-01
3.23923379e-01 -4.44021046e-01 -1.70307785e-01 -1.75300682e+00
-1.54629245e-01 -7.88791597e-01 1.45337611e-01 1.73970020e+00
4.86584723e-01 -5.49045146e-01 5.22163093e-01 5.24743497e-01
-1.46714121e-01 -1.13289598e-02 -7.04061866e-01 -6.32345676e-01
-7.07582012e-02 -6.59735128e-02 4.54728752e-01 1.10801184e+00
3.71807754e-01 6.37352705e-01 -4.51249123e-01 5.15818477e-01
2.93112636e-01 -1.46453276e-01 7.17835844e-01 -9.88798499e-01
-1.25436977e-01 -6.79841638e-02 -1.65685073e-01 -9.03096437e-01
3.32620054e-01 -4.00055975e-01 6.04207158e-01 -7.65026212e-01
-6.71212912e-01 -2.92182434e-02 5.52252121e-02 1.62953258e-01
-3.34427744e-01 1.32009432e-01 2.29334489e-01 8.83074477e-02
2.07157046e-01 4.29152787e-01 9.04769242e-01 -2.43941858e-01
-4.86671388e-01 2.76414573e-01 -3.10099244e-01 8.46501052e-01
8.73712420e-01 -6.44445181e-01 -8.65633860e-02 -1.13788813e-01
-3.81477833e-01 1.38701191e-02 4.04169075e-02 -8.00900459e-01
6.70948178e-02 -1.03219852e-01 -9.36929584e-02 -4.32376474e-01
2.09195122e-01 -8.94488990e-01 -3.76540303e-01 4.59059417e-01
-4.19104129e-01 1.12429120e-01 1.93095118e-01 -7.70985708e-02
-5.80163181e-01 -7.91444898e-01 9.38343406e-01 -8.17598368e-04
-4.57698852e-01 -5.79949737e-01 -9.77295876e-01 -5.43636680e-02
6.92556322e-01 -8.63901079e-02 -1.14350602e-01 -4.59396780e-01
-4.76582259e-01 -3.55028182e-01 2.41816431e-01 2.56789476e-01
5.04417479e-01 -8.74514878e-01 -7.25477159e-01 5.15548825e-01
-1.68371066e-01 -1.61613315e-01 -3.12328786e-01 7.66471922e-01
-5.16603529e-01 9.31285545e-02 8.22419226e-02 -3.90308797e-01
-1.56991339e+00 6.09499037e-01 4.59436834e-01 4.03723001e-01
-5.30899286e-01 7.41501033e-01 1.77780449e-01 -3.79411876e-01
2.70969421e-01 -3.71982843e-01 -2.36436948e-01 1.38987049e-01
9.10109401e-01 1.41499758e-01 9.29072052e-02 -7.04213023e-01
-3.95993352e-01 5.96414730e-02 -1.76536739e-01 -3.45674753e-01
9.22275603e-01 -2.00291961e-01 -1.57480434e-01 7.39473403e-01
1.08535397e+00 9.04786587e-01 -8.57195437e-01 -1.04068540e-01
1.93328157e-01 -5.25922060e-01 2.16244683e-01 -6.27147317e-01
-6.73897564e-01 8.16001415e-01 1.68550879e-01 7.29795873e-01
8.75717640e-01 -9.85484198e-02 7.33260036e-01 2.63111651e-01
-3.71549539e-02 -1.19063497e+00 -2.09882379e-01 7.82460093e-01
9.21993136e-01 -1.06991589e+00 -5.02171814e-01 -5.25080323e-01
-5.64973772e-01 1.34754777e+00 2.93427616e-01 2.11166102e-03
6.63726747e-01 4.37205374e-01 8.12692940e-01 -1.64793611e-01
-3.45388502e-01 -1.05840623e-01 3.32670957e-01 5.08251429e-01
7.18114138e-01 -7.75461202e-04 -3.05967152e-01 1.20231323e-01
-6.12249255e-01 -3.37486386e-01 8.69930446e-01 1.03293180e+00
-7.06035674e-01 -1.18257916e+00 -9.59900498e-01 -4.07604463e-02
-6.59399509e-01 -3.81809801e-01 -1.15160382e+00 4.63343889e-01
-1.47568852e-01 1.11741745e+00 1.33366331e-01 -3.23350370e-01
3.21744382e-01 5.72386622e-01 7.30611235e-02 -8.68696332e-01
-6.73421681e-01 3.18577975e-01 3.70082140e-01 1.22884706e-01
-1.74993128e-01 -6.67315364e-01 -1.18774068e+00 -1.80880889e-01
-3.58832091e-01 4.54332769e-01 8.65262210e-01 1.21597457e+00
-9.94741991e-02 4.06051934e-01 9.04168546e-01 -8.65882814e-01
-7.26941288e-01 -1.13777363e+00 -1.04413962e+00 3.78164649e-01
9.82655287e-01 -2.05835268e-01 -9.43090141e-01 1.64343894e-01]
|
[14.660669326782227, 6.4081315994262695]
|
5e0aacbc-8a36-424d-a00b-c83ef6c39123
|
contactart-learning-3d-interaction-priors-for
|
2305.01618
| null |
https://arxiv.org/abs/2305.01618v1
|
https://arxiv.org/pdf/2305.01618v1.pdf
|
ContactArt: Learning 3D Interaction Priors for Category-level Articulated Object and Hand Poses Estimation
|
We propose a new dataset and a novel approach to learning hand-object interaction priors for hand and articulated object pose estimation. We first collect a dataset using visual teleoperation, where the human operator can directly play within a physical simulator to manipulate the articulated objects. We record the data and obtain free and accurate annotations on object poses and contact information from the simulator. Our system only requires an iPhone to record human hand motion, which can be easily scaled up and largely lower the costs of data and annotation collection. With this data, we learn 3D interaction priors including a discriminator (in a GAN) capturing the distribution of how object parts are arranged, and a diffusion model which generates the contact regions on articulated objects, guiding the hand pose estimation. Such structural and contact priors can easily transfer to real-world data with barely any domain gap. By using our data and learned priors, our method significantly improves the performance on joint hand and articulated object poses estimation over the existing state-of-the-art methods. The project is available at https://zehaozhu.github.io/ContactArt/ .
|
['Xiaolong Wang', 'Varun Jampani', 'Deqing Sun', 'Yuzhe Qin', 'Jiashun Wang', 'Zehao Zhu']
|
2023-05-02
| null | null | null | null |
['hand-pose-estimation']
|
['computer-vision']
|
[-1.93413302e-01 3.51311229e-02 -2.89109945e-01 -1.28458560e-01
-5.97384930e-01 -7.52456307e-01 3.18790257e-01 -7.06026554e-01
-1.58487797e-01 4.49334681e-01 2.87461579e-01 1.06480375e-01
7.23747164e-02 -2.17006266e-01 -7.74512887e-01 -5.23434758e-01
9.20471773e-02 1.22887647e+00 2.99504429e-01 6.08293787e-02
3.43621615e-03 7.71345377e-01 -1.43769324e+00 3.64126302e-02
3.62845004e-01 8.62795830e-01 7.57479310e-01 8.92337441e-01
2.58489341e-01 5.93811512e-01 -6.47615910e-01 -5.58464937e-02
5.14514983e-01 -1.14243969e-01 -9.83645499e-01 3.59506756e-01
2.49468789e-01 -8.71491194e-01 -7.44764924e-01 6.69719219e-01
9.44443643e-01 1.09570548e-01 6.57562256e-01 -1.35216236e+00
-3.62544566e-01 3.50285828e-01 -6.37781799e-01 -2.94191390e-01
5.95615625e-01 5.83831429e-01 7.98080921e-01 -8.44475091e-01
9.94331300e-01 1.38383996e+00 4.34184283e-01 7.92422116e-01
-9.97736752e-01 -5.59084475e-01 3.08514267e-01 1.41672432e-01
-1.32923269e+00 -3.19065541e-01 8.01447332e-01 -8.30902517e-01
5.20039022e-01 2.19890699e-01 9.46315110e-01 1.61854923e+00
-1.52701899e-01 1.29375446e+00 5.96101582e-01 -1.80751756e-01
3.54603253e-04 -6.08603917e-02 -1.69699371e-01 7.50716269e-01
-1.93974048e-01 7.99920559e-02 -4.26507622e-01 -2.21119538e-01
1.59209192e+00 3.05001944e-01 -6.02980554e-01 -7.79482484e-01
-1.45649517e+00 2.02920258e-01 4.04142678e-01 -1.29339412e-01
-4.87329781e-01 4.29228783e-01 1.76519409e-01 -2.10669354e-01
2.36350726e-02 -5.77441305e-02 -6.09342873e-01 -4.65182453e-01
-3.20010096e-01 4.83876079e-01 9.66901004e-01 1.73719537e+00
2.44597554e-01 -3.39329302e-01 -4.45026249e-01 6.26305878e-01
3.88947189e-01 6.03515387e-01 -1.52512267e-01 -1.31946969e+00
4.67472374e-01 2.05965862e-01 5.57860672e-01 -4.77922052e-01
-1.99087590e-01 -4.63697985e-02 -4.39085186e-01 4.72433239e-01
5.53242803e-01 -1.21890299e-01 -1.16664410e+00 1.41395855e+00
5.47246695e-01 -5.77949211e-02 -6.59199297e-01 1.37718260e+00
6.68788314e-01 2.97648460e-01 -1.29265204e-01 2.55446583e-01
1.43522966e+00 -1.16345346e+00 -8.15718889e-01 -2.10972816e-01
1.32655188e-01 -8.40198517e-01 1.52072322e+00 5.74731648e-01
-1.11947191e+00 -6.44337595e-01 -6.18367374e-01 -2.99132675e-01
1.87017083e-01 3.01736146e-01 8.70271921e-01 2.91478992e-01
-5.80455780e-01 5.63573956e-01 -1.37963974e+00 -1.51329413e-01
6.51922405e-01 4.50414538e-01 -3.39469552e-01 -4.33892570e-03
-4.64221150e-01 8.38166535e-01 6.53639436e-02 4.01422232e-01
-1.29503739e+00 -7.07959294e-01 -6.54755950e-01 -2.89728612e-01
7.09279180e-01 -8.07568967e-01 1.68015289e+00 -3.90407622e-01
-1.96351218e+00 7.57081389e-01 -4.67163958e-02 3.39128733e-01
1.00872362e+00 -6.18365705e-01 4.22435880e-01 3.08568198e-02
-2.22161904e-01 7.90175855e-01 9.51517940e-01 -1.77116001e+00
-3.66865546e-01 -4.13252711e-01 1.72653630e-01 3.18183988e-01
8.75469595e-02 1.01776384e-02 -1.15154278e+00 -7.34148026e-01
-2.55200993e-02 -1.29649711e+00 3.67157198e-02 5.95731914e-01
-5.20475984e-01 -3.34869027e-01 1.06338918e+00 -1.02761638e+00
7.77701557e-01 -2.00091124e+00 7.67128289e-01 1.08792633e-01
2.74339288e-01 -2.41275690e-02 -2.86408179e-02 2.77870774e-01
3.13017845e-01 -3.26750785e-01 -5.99996336e-02 -7.59665012e-01
2.57756382e-01 3.69534552e-01 -2.26142406e-01 5.14554024e-01
-3.59243959e-01 1.05940211e+00 -9.11335647e-01 -4.94089127e-01
2.45827556e-01 7.34936953e-01 -6.49530113e-01 7.72020102e-01
-5.29659092e-01 1.08724236e+00 -4.65228349e-01 8.44905674e-01
5.47906041e-01 -2.39582866e-01 1.92681894e-01 -3.27664286e-01
1.59840986e-01 2.56281644e-01 -1.32062137e+00 2.14816880e+00
-4.32821214e-01 4.67930883e-01 3.93025964e-01 -3.20352823e-01
4.99865472e-01 5.18435359e-01 5.44143617e-01 1.25507623e-01
3.54646772e-01 -6.33125380e-02 -1.58409983e-01 -6.33298039e-01
1.14204057e-01 2.50868350e-01 9.61996317e-02 5.64644635e-01
1.49064407e-01 -4.81755614e-01 -3.35355997e-01 -3.59368022e-03
8.74668837e-01 6.69123113e-01 -1.03718368e-02 -2.41647996e-02
-1.41104892e-01 -2.03951627e-01 4.07142133e-01 4.62924600e-01
-3.08679231e-02 8.12657297e-01 2.19150499e-01 -2.85642654e-01
-1.13090169e+00 -1.38429677e+00 7.59981051e-02 1.14619493e+00
3.39636177e-01 -2.33927891e-01 -6.74243569e-01 -7.73179412e-01
2.30983853e-01 2.58665264e-01 -5.31630754e-01 2.20440134e-01
-7.14824378e-01 1.20125085e-01 1.59335047e-01 1.22971082e+00
2.96084285e-01 -1.27781200e+00 -6.57809019e-01 -8.10885206e-02
-2.50730366e-01 -9.57878172e-01 -1.15449226e+00 -1.05457209e-01
-7.95566499e-01 -1.04009247e+00 -9.34406757e-01 -8.58748496e-01
8.48332942e-01 -1.28817573e-01 9.43000853e-01 -8.08955953e-02
-7.13983834e-01 7.94890165e-01 -1.45370588e-01 -6.01635039e-01
-1.72236145e-01 -3.10032554e-02 8.29844773e-02 -3.92453581e-01
-1.81692779e-01 -6.20895743e-01 -8.73150170e-01 5.79297841e-01
-3.42040658e-01 1.34722948e-01 4.73562628e-01 7.19894230e-01
5.51080763e-01 -4.65732455e-01 -1.00401953e-01 -3.99239659e-01
2.17848435e-01 -4.58752736e-02 -5.59267163e-01 1.00772105e-01
6.88789636e-02 -5.16943038e-02 1.09689301e-02 -9.45288718e-01
-1.23782563e+00 5.41102231e-01 -1.53456451e-02 -8.79324675e-01
-2.17745915e-01 -3.66788097e-02 -5.10833919e-01 5.14820255e-02
3.90189946e-01 -1.31334767e-01 8.44933093e-03 -9.48449790e-01
5.05502522e-01 9.27600563e-01 7.46133864e-01 -8.96529973e-01
7.00091898e-01 7.29068577e-01 -3.34426403e-01 -5.55385351e-01
-6.78564668e-01 -4.45782781e-01 -1.14739037e+00 -3.37379098e-01
7.97891855e-01 -7.81397283e-01 -1.32996666e+00 6.08331382e-01
-1.51205480e+00 -8.36277723e-01 -3.88063461e-01 5.87609172e-01
-9.52575266e-01 3.16766202e-01 -8.03751647e-01 -8.78588080e-01
-9.58898738e-02 -1.15439701e+00 1.47585738e+00 -7.99479261e-02
-3.99096787e-01 -7.46644258e-01 -1.83696866e-01 3.69483590e-01
-6.08689748e-02 2.38267243e-01 4.69303101e-01 2.50220783e-02
-9.16423261e-01 -1.64753914e-01 7.07367510e-02 2.80885220e-01
3.85178119e-01 -2.29579657e-01 -8.20167124e-01 -5.83522797e-01
-1.88432962e-01 -4.01622683e-01 1.79571629e-01 4.33993071e-01
1.57991564e+00 -2.75147408e-01 -6.86080754e-01 5.88038146e-01
6.01013839e-01 9.78177860e-02 4.59771961e-01 -1.52791917e-01
1.15041733e+00 5.72601438e-01 7.31832206e-01 7.16348708e-01
3.94314110e-01 1.10176528e+00 4.60957497e-01 4.21438552e-02
-4.14612740e-01 -5.28746307e-01 1.66486993e-01 6.52036428e-01
-7.60201275e-01 -2.54825085e-01 -9.36321795e-01 5.23316622e-01
-1.80744159e+00 -6.45070255e-01 1.69559866e-01 2.10414267e+00
1.02760172e+00 -1.57891363e-01 4.61765707e-01 -1.61673546e-01
6.53235793e-01 -2.46486649e-01 -8.51827323e-01 5.19735277e-01
5.14306188e-01 2.04097629e-01 3.32080930e-01 6.00225389e-01
-8.71657729e-01 9.56822932e-01 6.24491978e+00 4.49528545e-01
-8.16893756e-01 2.31407642e-01 -2.12068394e-01 -3.17967087e-01
4.58702445e-03 -1.62479833e-01 -7.28855431e-01 3.10587972e-01
1.41221523e-01 1.66451126e-01 6.88647807e-01 1.00919378e+00
1.30571485e-01 1.33855671e-01 -1.62516439e+00 1.23893046e+00
-8.31515267e-02 -8.82742107e-01 -1.22970633e-01 2.70048141e-01
4.67283338e-01 -1.06043085e-01 -4.06436138e-02 -6.55391663e-02
4.52950627e-01 -9.62716937e-01 1.14779091e+00 7.42434680e-01
9.72076416e-01 -3.38107407e-01 3.78186941e-01 5.31897068e-01
-1.26009023e+00 -1.70460977e-02 5.25626726e-02 -1.53266871e-02
5.04757881e-01 1.47484941e-02 -8.48905087e-01 2.32955173e-01
8.34254920e-01 4.98921901e-01 3.64555530e-02 9.21174049e-01
-5.07216454e-01 2.67826706e-01 -4.73845094e-01 1.28801525e-01
-3.81124258e-01 1.00516520e-01 7.91729271e-01 9.28248882e-01
1.02308899e-01 2.11273894e-01 6.25758946e-01 9.16014016e-01
-3.49116884e-02 -4.49707985e-01 -3.50251645e-01 1.31553411e-01
5.80750823e-01 9.76278067e-01 -5.44834971e-01 -3.63874942e-01
-8.13960433e-02 1.46938109e+00 2.28569925e-01 4.07594442e-01
-1.00656533e+00 -2.55067050e-01 8.37415755e-01 4.33043897e-01
3.93569469e-01 -7.73541152e-01 -1.03551075e-01 -1.04976511e+00
4.18752402e-01 -8.10747325e-01 9.15593803e-02 -9.17715013e-01
-1.33158004e+00 3.65703702e-01 3.80659342e-01 -1.23762643e+00
-2.08072439e-01 -8.37395310e-01 -1.21311940e-01 8.69890928e-01
-6.60352528e-01 -1.33020556e+00 -8.66993845e-01 6.72868490e-01
7.18916297e-01 1.70326039e-01 6.94065332e-01 9.97796953e-02
-1.50761098e-01 3.86077046e-01 -4.67101395e-01 2.75693178e-01
7.26802468e-01 -1.23834312e+00 5.31901717e-01 1.35957450e-01
1.40495673e-01 7.40214646e-01 5.65401196e-01 -8.19446266e-01
-1.68663061e+00 -6.79777980e-01 -6.96440712e-02 -1.24470246e+00
3.47639978e-01 -9.09036696e-01 -5.58805823e-01 1.26645494e+00
3.65784168e-02 1.97553426e-01 2.36025125e-01 -4.25351001e-02
-1.42733693e-01 2.79764652e-01 -1.01349843e+00 6.67642415e-01
1.92753196e+00 -5.91796279e-01 -6.43778980e-01 6.46085262e-01
4.92638439e-01 -1.15482926e+00 -8.71949017e-01 3.12267661e-01
1.03944671e+00 -5.10520458e-01 1.02657390e+00 -6.65660620e-01
8.86228979e-02 -4.52787310e-01 8.30689967e-02 -1.20756936e+00
-3.77072841e-01 -9.82303262e-01 -6.42976642e-01 8.76252413e-01
-1.61482431e-02 -3.42544436e-01 9.84934568e-01 7.28065729e-01
-1.05228029e-01 -6.87544763e-01 -6.90592527e-01 -9.58855093e-01
-3.25845718e-01 -3.39803278e-01 5.80624104e-01 4.71691787e-01
-1.04173549e-01 1.05489038e-01 -4.34604436e-01 4.07304585e-01
7.40665197e-01 3.88759263e-02 1.17955959e+00 -1.25193930e+00
-6.47681177e-01 -1.31457880e-01 -3.01782876e-01 -1.85906851e+00
3.41265589e-01 -6.37206256e-01 4.62326586e-01 -1.60524845e+00
3.86296898e-01 -5.11461377e-01 3.76996517e-01 5.85845351e-01
-4.58913371e-02 5.24684787e-02 2.71556616e-01 4.72719491e-01
-2.59254098e-01 6.50367975e-01 1.78152192e+00 2.32822467e-02
-3.65534961e-01 2.65285254e-01 -1.02792509e-01 9.57163453e-01
5.19183517e-01 -3.76645625e-01 -3.63909423e-01 -7.62925446e-01
-3.74686599e-01 2.05818728e-01 8.99160445e-01 -7.95361221e-01
2.29880482e-01 -2.43073449e-01 3.93377334e-01 -6.30900145e-01
6.23120129e-01 -1.05871630e+00 3.98851156e-01 3.88913840e-01
-1.38405994e-01 -5.71186543e-02 2.18225703e-01 6.61150992e-01
3.24865043e-01 9.39491168e-02 5.02233565e-01 -2.47459948e-01
-4.28701282e-01 7.40359843e-01 -7.01530427e-02 -8.70991200e-02
1.06162930e+00 -1.94630668e-01 -5.08782566e-02 -5.79628885e-01
-1.10969198e+00 2.91942477e-01 5.07850707e-01 7.61913478e-01
5.19086957e-01 -1.26723313e+00 -3.78339469e-01 1.82279095e-01
-2.65737269e-02 5.93173504e-01 2.19804391e-01 6.33983493e-01
-4.45746601e-01 8.32788870e-02 -2.03848481e-01 -1.01870799e+00
-1.40186977e+00 3.69396418e-01 2.74082124e-01 2.86121488e-01
-1.02367973e+00 1.09175479e+00 2.43130341e-01 -7.21677959e-01
7.59736598e-01 -5.41691124e-01 4.08688605e-01 -5.13573527e-01
2.85214901e-01 6.00787282e-01 -3.17619115e-01 -3.37073445e-01
-3.44120979e-01 7.17948079e-01 1.62186638e-01 -2.77092457e-01
1.21530354e+00 1.01169653e-01 1.01101108e-01 5.52066147e-01
9.17550325e-01 1.33120179e-01 -2.06971359e+00 4.63514263e-03
-3.79581779e-01 -8.28522801e-01 -1.83048904e-01 -8.69140506e-01
-8.82590294e-01 8.93576562e-01 4.61088717e-01 -4.45572913e-01
7.36307859e-01 6.40284657e-01 7.20875263e-01 4.56432641e-01
8.18048596e-01 -9.88356829e-01 4.58002955e-01 3.98938566e-01
1.61472261e+00 -9.87002969e-01 -1.17205456e-01 -8.22353959e-01
-7.55280316e-01 8.27199221e-01 7.44668484e-01 4.54327539e-02
7.90162921e-01 8.39114547e-01 7.89179727e-02 -2.54563749e-01
-3.82657945e-01 -2.21595466e-02 3.86050284e-01 8.57617021e-01
1.47227585e-01 2.18100473e-01 2.25722328e-01 5.98264933e-01
-2.93747902e-01 1.81911215e-01 9.47929621e-02 1.25000107e+00
7.76792085e-03 -1.26511002e+00 -3.64170402e-01 2.07731500e-01
-4.87580057e-03 3.05160254e-01 -3.72286409e-01 8.48692000e-01
-1.14039518e-02 3.33848834e-01 9.85055268e-02 -3.09618413e-01
8.09649885e-01 -1.09503597e-01 1.13489568e+00 -8.71528268e-01
-2.78590560e-01 8.95515755e-02 -2.01299295e-01 -7.77232885e-01
-1.70417517e-01 -6.04262054e-01 -1.35525560e+00 -2.14452386e-01
-3.48586589e-01 -1.57155871e-01 5.80979407e-01 6.27121031e-01
5.60880244e-01 4.99509394e-01 2.59668201e-01 -1.91065896e+00
-8.08010280e-01 -1.23709488e+00 -5.89331985e-01 3.50338668e-01
3.36465150e-01 -1.16706014e+00 -1.51168063e-01 4.58961278e-01]
|
[6.446499347686768, -0.9875685572624207]
|
c06d3ec2-d518-4ee5-9d7c-00ab4951991a
|
fisr-deep-joint-frame-interpolation-and-super
|
1912.07213
| null |
https://arxiv.org/abs/1912.07213v2
|
https://arxiv.org/pdf/1912.07213v2.pdf
|
FISR: Deep Joint Frame Interpolation and Super-Resolution with a Multi-scale Temporal Loss
|
Super-resolution (SR) has been widely used to convert low-resolution legacy videos to high-resolution (HR) ones, to suit the increasing resolution of displays (e.g. UHD TVs). However, it becomes easier for humans to notice motion artifacts (e.g. motion judder) in HR videos being rendered on larger-sized display devices. Thus, broadcasting standards support higher frame rates for UHD (Ultra High Definition) videos (4K@60 fps, 8K@120 fps), meaning that applying SR only is insufficient to produce genuine high quality videos. Hence, to up-convert legacy videos for realistic applications, not only SR but also video frame interpolation (VFI) is necessitated. In this paper, we first propose a joint VFI-SR framework for up-scaling the spatio-temporal resolution of videos from 2K 30 fps to 4K 60 fps. For this, we propose a novel training scheme with a multi-scale temporal loss that imposes temporal regularization on the input video sequence, which can be applied to any general video-related task. The proposed structure is analyzed in depth with extensive experiments.
|
['Munchurl Kim', 'Soo Ye Kim', 'Jihyong Oh']
|
2019-12-16
| null | null | null | null |
['space-time-video-super-resolution']
|
['computer-vision']
|
[ 5.26055455e-01 -3.01165432e-01 -6.36513457e-02 -1.20703794e-01
-8.20137143e-01 -1.96788609e-01 2.33210072e-01 -4.60757494e-01
-2.99584210e-01 7.72659600e-01 1.39231324e-01 -2.04233110e-01
-2.36534420e-02 -5.93488812e-01 -6.78230286e-01 -5.74712217e-01
-4.30350155e-02 -6.31458521e-01 5.71132362e-01 -1.38111800e-01
1.53893540e-02 4.86365676e-01 -1.79173434e+00 5.18856287e-01
7.33215809e-01 9.76443768e-01 6.43070161e-01 7.78563142e-01
4.26738679e-01 9.03125942e-01 -4.18934673e-01 -1.32758737e-01
4.24639612e-01 -4.52126294e-01 -4.80465800e-01 1.09086156e-01
6.08859062e-01 -8.91385198e-01 -6.41092777e-01 9.90307152e-01
3.91736686e-01 4.27145034e-01 1.39351308e-01 -7.89118946e-01
-5.55884063e-01 2.16975093e-01 -8.95363867e-01 6.50258064e-01
7.33923078e-01 4.85469513e-02 5.81617713e-01 -7.03461587e-01
7.47184217e-01 1.28836262e+00 3.70495498e-01 5.50870419e-01
-1.07966876e+00 -5.69674194e-01 -1.75270531e-02 4.13509965e-01
-1.37723219e+00 -5.03932834e-01 6.70258880e-01 -2.35756144e-01
4.81121868e-01 3.58770996e-01 5.26223063e-01 1.15653551e+00
2.42457554e-01 4.58962113e-01 8.89250636e-01 -8.35430399e-02
1.83780063e-02 -1.03786908e-01 -3.26611668e-01 8.12562555e-02
1.78251296e-01 8.15919191e-02 -4.52528805e-01 1.77764922e-01
1.53181005e+00 -6.82432670e-03 -6.85438335e-01 -9.09112208e-03
-1.16874671e+00 2.76985168e-01 2.26580814e-01 4.00647670e-01
-3.85235190e-01 -1.31633192e-01 4.29783106e-01 2.91346908e-01
4.55262631e-01 8.36529657e-02 -1.17909499e-01 -3.33584607e-01
-9.92199123e-01 1.28650472e-01 -8.16661716e-02 1.18113983e+00
2.49908701e-01 2.53682166e-01 -1.91447571e-01 9.41439509e-01
-4.39032577e-02 2.63456702e-01 3.77872050e-01 -1.40095007e+00
5.91753662e-01 -1.12462237e-01 6.20003760e-01 -1.05661702e+00
-2.08961412e-01 -1.80370465e-01 -1.22183204e+00 1.66717112e-01
3.86902064e-01 2.03332515e-03 -4.36076730e-01 1.53411782e+00
4.11738485e-01 6.11435354e-01 6.73251674e-02 1.48884869e+00
6.24279261e-01 9.90750670e-01 -7.81513527e-02 -8.82633924e-01
1.41441894e+00 -5.74107111e-01 -1.11993301e+00 1.55262098e-01
1.37358621e-01 -8.53635132e-01 1.40912712e+00 6.11223400e-01
-1.34149408e+00 -1.11477327e+00 -1.08690178e+00 -2.82463580e-01
3.15527588e-01 -3.00213099e-02 1.59711003e-01 3.08130473e-01
-8.89006615e-01 7.25182593e-01 -6.72886193e-01 -4.48558629e-02
1.79054871e-01 -2.13929992e-02 -2.90377796e-01 -2.94884264e-01
-1.49138725e+00 5.43446779e-01 3.01429063e-01 1.91460744e-01
-5.44045925e-01 -7.46282935e-01 -8.21259558e-01 -2.01374888e-01
5.03609717e-01 -4.01237875e-01 1.03933406e+00 -8.63694131e-01
-1.60359359e+00 5.01308322e-01 -1.42724216e-01 -3.46950471e-01
7.41733670e-01 -4.86814260e-01 -7.83359826e-01 3.99552673e-01
-2.69944400e-01 1.61063924e-01 1.32565880e+00 -1.24276030e+00
-9.29549396e-01 -1.82173654e-01 3.47115546e-01 3.11133236e-01
-2.86781967e-01 3.18613976e-01 -5.76293945e-01 -8.84792149e-01
5.65212369e-02 -7.07001448e-01 5.68321832e-02 -3.93445045e-02
3.72284837e-02 4.78779748e-02 1.11456954e+00 -1.05807376e+00
1.45834327e+00 -2.40719271e+00 2.22317338e-01 -3.71113807e-01
2.28797525e-01 3.85944486e-01 -4.54052031e-04 -2.71489203e-01
1.60419904e-02 -2.62485474e-01 -6.94009960e-02 -5.32067418e-02
-5.51205635e-01 1.65900718e-02 -2.52939165e-01 4.05958861e-01
-7.71796256e-02 3.89770448e-01 -9.92249072e-01 -5.10426283e-01
4.67784524e-01 1.03099453e+00 -3.57819766e-01 3.40438604e-01
3.08153003e-01 8.42183530e-01 -2.83926874e-01 4.65018630e-01
9.66865301e-01 -2.96321154e-01 3.05522997e-02 -6.44109786e-01
-4.50807810e-01 -7.25972950e-02 -1.29403234e+00 1.64001858e+00
-6.98195517e-01 7.05860376e-01 2.07281798e-01 -4.20629740e-01
8.04385722e-01 3.64710152e-01 5.87555885e-01 -9.77330625e-01
-1.06995940e-01 1.03313044e-01 -3.22125524e-01 -6.02424622e-01
8.11561584e-01 -7.23441988e-02 2.83707917e-01 -6.65877163e-02
-4.71471190e-01 2.14876026e-01 1.99245498e-01 1.32907396e-02
9.09548044e-01 3.45516473e-01 2.14119419e-01 1.01658277e-01
8.05092692e-01 -5.87220609e-01 7.11329043e-01 2.89093316e-01
-1.77616015e-01 9.03367698e-01 1.89026341e-01 -2.94577509e-01
-1.46706450e+00 -1.00790191e+00 -3.19087982e-01 8.92159581e-01
5.67066491e-01 -2.71138042e-01 -6.28648520e-01 -1.48828372e-01
-4.70237017e-01 3.92431885e-01 -1.86402142e-01 -9.47363973e-02
-9.47916865e-01 -4.65453833e-01 1.21203713e-01 4.17866290e-01
7.10270226e-01 -7.76897311e-01 -9.26740766e-01 3.12739819e-01
-4.55567032e-01 -1.64998949e+00 -6.44642115e-01 -4.61848050e-01
-9.46109951e-01 -7.78706193e-01 -1.25638056e+00 -5.54853022e-01
3.40511948e-01 6.84236169e-01 7.54066706e-01 -2.89966196e-01
-1.14450000e-01 5.32647483e-02 -6.32920206e-01 4.64965194e-01
-3.86574835e-01 -4.30113375e-01 2.59243667e-01 3.11843067e-01
-2.04221338e-01 -5.12582064e-01 -8.29090059e-01 5.21717191e-01
-1.19700289e+00 4.75067973e-01 4.74596322e-01 5.44767082e-01
8.20185125e-01 3.27293783e-01 5.41200697e-01 -5.27966559e-01
3.85187238e-01 -1.28049582e-01 -6.67451918e-01 5.17211109e-02
-1.93453684e-01 -3.93541485e-01 1.16670310e+00 -7.87342310e-01
-1.38066185e+00 -2.52796322e-01 -7.28688166e-02 -9.85901833e-01
2.16999906e-03 5.53761492e-04 -2.21299678e-01 -7.62894750e-02
4.18050170e-01 2.10006326e-01 -3.25178832e-01 -5.31429350e-01
2.04609081e-01 7.80278862e-01 8.23639691e-01 -1.41361222e-01
8.54679048e-01 4.35924590e-01 1.93694700e-02 -1.07975399e+00
-6.43589497e-01 -2.47274220e-01 -3.97604614e-01 -3.38033527e-01
1.02324915e+00 -1.12881339e+00 -6.10491753e-01 3.00788105e-01
-9.61044848e-01 -2.64692575e-01 -1.72340587e-01 7.86857188e-01
-6.68886006e-01 6.97561800e-01 -9.29038346e-01 -7.14069545e-01
-1.05282173e-01 -1.01762867e+00 9.48375523e-01 3.35458845e-01
8.82554799e-02 -4.06510711e-01 -2.71603853e-01 4.39271539e-01
4.69450027e-01 3.29241008e-01 4.77443695e-01 5.30667782e-01
-7.11250544e-01 2.76174396e-01 -6.68682754e-01 5.24250865e-01
3.58650446e-01 -1.23287797e-01 -6.96082950e-01 -5.28263569e-01
3.12710106e-01 -1.97237916e-02 4.12937433e-01 5.24766028e-01
1.33265603e+00 -4.19316560e-01 1.62744716e-01 8.72187316e-01
1.45769966e+00 3.98943990e-01 1.07705402e+00 3.63642275e-01
7.66906321e-01 3.95594120e-01 1.23556209e+00 6.76858544e-01
5.11465222e-02 1.23804998e+00 1.37775615e-01 -1.36152878e-01
-2.83978790e-01 -9.07423794e-02 4.65133876e-01 6.47729635e-01
-6.41260862e-01 -1.46209344e-01 -2.45010361e-01 1.14033118e-01
-1.68030512e+00 -1.03858197e+00 -2.40338504e-01 2.54972434e+00
7.22202420e-01 4.68732528e-02 1.14711136e-01 2.01983869e-01
1.02880371e+00 3.17393631e-01 -6.76332653e-01 -1.72916800e-01
-2.21209601e-01 -1.25343546e-01 3.43455762e-01 3.22822660e-01
-9.86059904e-01 5.31163275e-01 5.13243675e+00 1.03538978e+00
-1.18604219e+00 1.82538241e-01 7.27637827e-01 -2.57686883e-01
-3.77666093e-02 -4.06055957e-01 -5.65594494e-01 7.42881894e-01
9.63984728e-01 -2.00348869e-01 5.98619223e-01 5.56299329e-01
9.13881063e-01 -1.46624073e-02 -8.88074398e-01 1.61018085e+00
-1.96377724e-01 -1.01351595e+00 1.06061421e-01 -3.64084262e-03
6.67087734e-01 -6.50660872e-01 1.94713861e-01 4.58064191e-02
-6.21350050e-01 -7.15851724e-01 6.56270444e-01 3.50949705e-01
1.43918133e+00 -8.06040764e-01 5.14258325e-01 5.80972582e-02
-1.49111199e+00 -3.56831215e-02 -6.55123711e-01 2.00154096e-01
6.23567343e-01 5.39680660e-01 -5.64949168e-03 7.28909135e-01
1.04184628e+00 8.91128778e-01 -2.14843720e-01 7.41963029e-01
1.52320117e-01 3.61458398e-02 -1.27278298e-01 7.20923901e-01
-1.30923048e-01 -3.27733129e-01 5.81831694e-01 1.07643056e+00
7.24398315e-01 7.21629679e-01 -9.39878374e-02 4.00866032e-01
-4.73207384e-02 2.53610406e-02 -4.34087425e-01 2.37352014e-01
2.31970266e-01 1.21192539e+00 -4.40357834e-01 -2.94902265e-01
-7.09888577e-01 1.37749934e+00 -1.61663994e-01 4.77618068e-01
-1.30313563e+00 -3.47144753e-01 6.82958424e-01 5.35264373e-01
2.90059626e-01 -3.83926988e-01 2.00168490e-01 -1.40637469e+00
1.65556893e-01 -9.14193988e-01 3.72728288e-01 -9.49431062e-01
-8.92415345e-01 6.47649109e-01 -1.25856191e-01 -1.90514517e+00
-1.57580033e-01 -1.45572126e-01 -6.77310675e-02 6.52124703e-01
-1.67725778e+00 -6.66670859e-01 -6.24975920e-01 8.19467068e-01
9.61258054e-01 3.56433988e-01 7.96683207e-02 8.69924188e-01
-4.37720269e-01 5.23042142e-01 8.94014090e-02 -2.39991412e-01
8.25717866e-01 -7.20636010e-01 6.30142614e-02 1.00483191e+00
-3.42087239e-01 3.30599099e-01 9.04756725e-01 -5.47846794e-01
-1.47561753e+00 -1.13769865e+00 3.55156034e-01 -6.84386045e-02
3.12767297e-01 -9.22849402e-02 -1.31091797e+00 2.64421731e-01
-2.27197781e-01 3.63566309e-01 1.73031986e-01 -5.91235638e-01
-2.36052256e-02 -3.75179172e-01 -1.28765094e+00 6.93468750e-01
1.13612151e+00 -4.77664828e-01 -1.22056037e-01 -6.01133481e-02
9.76975083e-01 -5.99433482e-01 -1.30154026e+00 6.30272627e-01
6.70804560e-01 -1.26790106e+00 1.20759141e+00 -3.38105746e-02
5.32642186e-01 -7.51322329e-01 -2.50101686e-01 -6.84461176e-01
-4.37372237e-01 -8.20261300e-01 -6.44298732e-01 9.95222390e-01
-1.54504836e-01 -2.76000500e-01 4.84778702e-01 5.13267338e-01
3.70922945e-02 -5.32673359e-01 -1.05789232e+00 -8.72827291e-01
-6.62494361e-01 -2.86905080e-01 2.54882276e-01 8.67227077e-01
-1.58901900e-01 9.47089940e-02 -1.00185895e+00 4.93266791e-01
8.70292604e-01 -3.57028283e-03 5.99052310e-01 -7.08930314e-01
-4.23228413e-01 -4.18551918e-03 -3.94551367e-01 -1.44216120e+00
-3.92734170e-01 -6.42687008e-02 -1.85651049e-01 -1.15439701e+00
4.24725562e-02 -2.51820773e-01 -2.74422884e-01 -3.22190553e-01
-2.17992201e-01 5.47064424e-01 3.05130988e-01 3.50874811e-01
-6.03615403e-01 5.13429761e-01 1.59498131e+00 3.43594581e-01
-4.63467121e-01 -3.51081640e-02 -2.77477533e-01 5.72971523e-01
4.58129287e-01 1.58178449e-01 -5.18227816e-01 -3.85775149e-01
5.82581647e-02 7.21417964e-01 2.91802436e-01 -1.15807307e+00
-1.17083758e-01 -2.14966506e-01 5.28391123e-01 -3.09168249e-01
5.20392776e-01 -8.27092171e-01 4.99757618e-01 2.13644609e-01
-3.33492935e-01 -7.07857311e-02 2.53176037e-02 6.12697899e-01
-3.81657749e-01 2.01093316e-01 1.19555485e+00 3.76376472e-02
-8.21692526e-01 4.15817559e-01 -2.37211272e-01 -1.32013083e-01
1.03881359e+00 -4.59995985e-01 -2.12045789e-01 -5.06382883e-01
-7.11949587e-01 -4.96205948e-02 6.52712286e-01 5.93276381e-01
1.08030343e+00 -1.34852278e+00 -6.85439885e-01 1.28595158e-01
-1.67384028e-01 -1.49539588e-02 1.06953323e+00 8.99083078e-01
-4.81136262e-01 2.21130133e-01 -3.68057638e-01 -4.64059740e-01
-1.45663738e+00 8.36376786e-01 1.84548611e-04 -7.47760609e-02
-1.26082182e+00 3.41766745e-01 4.36276227e-01 5.72244167e-01
1.44403294e-01 -4.31111097e-01 -4.30856794e-01 -1.57949284e-01
1.09612858e+00 6.76563978e-01 -2.86131740e-01 -6.68177843e-01
-6.81933612e-02 7.31013715e-01 2.30597202e-02 8.65238607e-02
1.13871813e+00 -7.63490200e-01 3.77914667e-01 4.84676033e-01
1.16787410e+00 -1.03459395e-01 -1.75873220e+00 -9.25904363e-02
-2.92248160e-01 -1.08301818e+00 8.89764056e-02 -4.93867695e-02
-9.51304018e-01 6.21770322e-01 8.46703470e-01 1.75688192e-01
1.56468797e+00 -2.02406108e-01 1.21600211e+00 -1.38485640e-01
7.67175078e-01 -1.12621713e+00 -4.04154323e-03 8.01524892e-02
8.57861996e-01 -1.01697075e+00 5.22406399e-02 -5.35076737e-01
-6.35089040e-01 1.20803201e+00 5.43417454e-01 6.39979728e-03
5.99775314e-02 1.23796545e-01 -3.03595006e-01 3.99028391e-01
-5.83504975e-01 -1.29043052e-04 1.43391162e-01 4.98449564e-01
4.56553936e-01 -2.81799793e-01 -4.21251327e-01 2.48565570e-01
2.01383591e-01 3.10003400e-01 9.29500878e-01 5.80181301e-01
-3.85465205e-01 -6.23381555e-01 -5.77784896e-01 2.40811646e-01
-5.75473189e-01 2.00990736e-01 6.30230486e-01 5.10353267e-01
1.80478945e-01 9.36766744e-01 2.17069238e-02 -3.11543226e-01
4.81421113e-01 -5.09848654e-01 5.38876295e-01 -1.27632588e-01
1.11640669e-01 3.70640129e-01 -1.14313804e-01 -8.58625770e-01
-7.46705115e-01 -5.50353348e-01 -1.12581682e+00 -4.24670964e-01
6.36536255e-02 -1.49975821e-01 3.31902862e-01 3.67471457e-01
2.61269033e-01 7.06843078e-01 7.99434483e-01 -1.03611362e+00
-1.65160060e-01 -6.51041865e-01 -9.03440297e-01 5.27484715e-01
5.49997747e-01 -5.45167208e-01 -4.16595578e-01 3.41981173e-01]
|
[11.020593643188477, -1.9612853527069092]
|
53a1aa50-c7e9-4951-bb64-eeff4aed4bb2
|
detection-of-sub-cellular-changes-by-use-of-l
|
2103.13484
| null |
https://arxiv.org/abs/2103.13484v2
|
https://arxiv.org/pdf/2103.13484v2.pdf
|
Continuous monitoring of plant sub-cellular structural changes for plant and crop diseases detection by use of Intelligent Laser Speckle Classification (AI) technique
|
The continuous online monitoring of early signs of plant and crop diseases, at their early stages before a potential spread, is of high importance and necessitates multi-disciplinary techniques. Within this study a proposed technique achieves this goal by exploiting laser physics, textural image analysis, and AI for Shot hole disease. In this technique, specific laser light with a wavelength shorter than a sub-cellular component of an inspected plant, produces an interaction within the sub-cellular components and generates laser speckle patterns which can characterize those specific plant cells' features. The generated laser speckle image data then be quantized by texture analysis and classified by Bayesian networks. Such comparative methods manage to detect the differences at sub-cellular scales, such as nuclei modification, cellular shape, or size deformation, etc. for Shot hole disease with high classification accuracy between the healthy and diseased plants. The technique is capable of continuous online observation and monitoring of the plant or crop diseases via a wireless network at low instrumental cost and may replace the costly ground-truth field works
|
['Ahmet Orun']
|
2021-03-23
| null | null | null | null |
['texture-classification']
|
['computer-vision']
|
[ 5.35670578e-01 -5.57477809e-02 -1.37437701e-01 3.16531301e-01
-1.40473144e-02 -6.28315210e-01 2.04590812e-01 4.51453120e-01
1.64682195e-01 8.18016827e-01 -5.98361492e-01 -1.36619851e-01
-5.61644793e-01 -1.27637315e+00 -1.23507574e-01 -1.22726810e+00
-6.34003282e-02 4.58394736e-01 5.46019137e-01 -1.33166522e-01
3.13381404e-01 9.53516424e-01 -1.59244883e+00 4.36108820e-02
7.00844586e-01 1.19951618e+00 7.09643662e-01 7.11993515e-01
-1.85930133e-01 2.52829075e-01 -5.15486300e-01 4.53435123e-01
-4.97895032e-02 -4.08706695e-01 -4.73468870e-01 4.82280761e-01
-1.99358702e-01 -2.68937677e-01 3.52182508e-01 1.33520460e+00
2.49259204e-01 -3.76123548e-01 9.15444493e-01 -6.40189052e-01
-1.01969993e+00 2.78543890e-01 -9.00925338e-01 6.48974404e-02
3.65155339e-01 4.11434621e-01 5.19744813e-01 -5.12597561e-01
7.29503810e-01 9.80835259e-01 5.98231971e-01 1.24085240e-01
-1.42507041e+00 -1.51088517e-02 -6.74771905e-01 5.55582881e-01
-1.23909426e+00 -1.25137484e-02 6.26345217e-01 -5.95698357e-01
1.36277184e-01 2.35755399e-01 6.60025001e-01 4.03413355e-01
6.64644122e-01 3.22724432e-02 1.35871828e+00 -4.50340092e-01
4.88278210e-01 -1.90921336e-01 8.59023258e-02 8.35185111e-01
5.47169685e-01 1.91305086e-01 -1.12165123e-01 -6.61350414e-02
8.84605408e-01 3.59809212e-02 -5.12571692e-01 -3.90443718e-04
-8.27683926e-01 3.62069368e-01 2.01540828e-01 1.03182566e+00
-8.23902965e-01 -2.26822868e-01 -4.22568880e-02 4.11568098e-02
4.17572707e-01 2.92167276e-01 -6.80560529e-01 2.28854984e-01
-8.41680586e-01 -3.86369288e-01 7.53357530e-01 2.62022346e-01
7.46697843e-01 -2.60328323e-01 -2.73527354e-01 2.93609947e-01
3.34358364e-01 1.07550561e+00 1.59006774e-01 -1.08364928e+00
-7.49280274e-01 7.87953854e-01 1.35121301e-01 -1.41870987e+00
-1.23101830e-01 -4.37381975e-02 -1.09917498e+00 5.32139003e-01
5.97806513e-01 1.79383859e-01 -6.22855663e-01 1.24310637e+00
6.12582266e-01 1.48721710e-01 3.74863185e-02 5.55674970e-01
6.98166788e-01 8.50932539e-01 1.40963465e-01 -8.11136782e-01
1.85688269e+00 6.82867095e-02 -1.02088392e+00 4.55453873e-01
5.61299741e-01 -9.86337602e-01 7.13227212e-01 4.75561172e-01
-9.06562984e-01 -4.04776543e-01 -6.33093297e-01 7.41736710e-01
-5.07898867e-01 6.39960170e-01 4.75555182e-01 5.58786750e-01
-7.85129189e-01 7.97974527e-01 -6.84745848e-01 -7.37951517e-01
3.83778811e-01 7.49878772e-03 -1.97595000e-01 7.88717493e-02
-6.54542208e-01 8.56800854e-01 3.55491161e-01 6.65439665e-01
-4.99659151e-01 -5.69752336e-01 -8.60044733e-02 4.31051068e-02
1.91322863e-01 -2.04881668e-01 4.47719425e-01 -6.34844601e-01
-2.04829025e+00 1.00308537e+00 -7.35168904e-02 9.19825882e-02
-2.29851127e-01 3.18359643e-01 -2.05402717e-01 6.49958074e-01
-3.84855755e-02 -4.80040200e-02 7.49885857e-01 -1.35194504e+00
-5.17649412e-01 -8.62387776e-01 -4.26524818e-01 -3.83805513e-01
-3.04238081e-01 -1.12521902e-01 4.15216029e-01 -1.89782426e-01
5.85749686e-01 -7.98161983e-01 -5.75971901e-02 3.76340628e-01
-3.70377868e-01 6.05606753e-03 1.41458261e+00 -5.85748255e-01
4.81111974e-01 -2.11403322e+00 7.83686712e-02 2.43834287e-01
1.29635826e-01 5.45803368e-01 -4.27611582e-02 3.12037349e-01
2.41918519e-01 1.48020759e-01 -3.49962711e-01 6.17609739e-01
-5.47260642e-01 2.20664397e-01 4.69460897e-02 6.31653428e-01
1.43746242e-01 4.72191453e-01 -7.30766118e-01 -5.96831143e-01
2.40260631e-01 3.33781362e-01 2.93385118e-01 9.24944803e-02
-1.79552227e-01 4.63783145e-01 -7.70770729e-01 1.13855100e+00
1.18369293e+00 -1.34854913e-01 1.66688651e-01 -5.48187494e-01
-5.91828167e-01 -8.77452374e-01 -9.56718922e-01 9.59918737e-01
-8.21226463e-02 4.54671532e-01 7.42134094e-01 -1.32009530e+00
1.28755617e+00 4.99276966e-01 6.36242688e-01 -1.71929121e-01
1.22847781e-01 4.20329154e-01 -3.04996818e-01 -7.99162030e-01
-1.40995324e-01 1.72685966e-01 6.54819965e-01 1.15484759e-01
-1.36713967e-01 -5.22404134e-01 1.81876615e-01 -4.00822103e-01
1.13606572e+00 1.02926873e-01 2.34134078e-01 -5.66693306e-01
6.92537487e-01 1.55407295e-01 3.05982828e-01 3.84517223e-01
-4.40124035e-01 -1.13341875e-01 4.81331110e-01 -1.53940871e-01
-8.09467256e-01 -8.33156765e-01 -2.38386542e-01 4.71135378e-01
3.57156038e-01 5.86541951e-01 -6.84139013e-01 -5.76502867e-02
6.89969286e-02 2.10176602e-01 -5.91975629e-01 -9.98483673e-02
7.58744627e-02 -1.03267789e+00 2.50907511e-01 -1.14429139e-01
7.84694195e-01 -1.33707261e+00 -9.33412910e-01 3.09569508e-01
1.65225014e-01 -9.73277807e-01 5.72207451e-01 1.38310105e-01
-1.13259888e+00 -1.28592396e+00 -5.98968148e-01 -7.05626667e-01
5.69315135e-01 2.09195197e-01 8.17980289e-01 1.89284086e-01
-8.45171034e-01 1.75998032e-01 -7.22921789e-01 -3.04482579e-01
-5.51986575e-01 -1.64493635e-01 -4.14760441e-01 1.46531612e-01
3.32307637e-01 -6.53903365e-01 -4.98232007e-01 8.38030800e-02
-7.38927186e-01 -4.72792149e-01 5.60461283e-01 7.56662428e-01
7.31801331e-01 5.93097687e-01 1.34207308e-01 -7.01222420e-01
2.02489406e-01 -4.34978455e-01 -9.85186696e-01 5.27184784e-01
-4.03106600e-01 -4.21437323e-01 5.48036039e-01 -2.08676770e-01
-9.53852654e-01 -1.52376384e-01 5.24379134e-01 1.77110165e-01
-9.09397125e-01 8.52267563e-01 -2.37522498e-02 -3.10596377e-01
7.82929242e-01 3.21922749e-01 3.23689491e-01 -3.21691394e-01
-4.41443883e-02 5.80435812e-01 4.56939399e-01 -1.23762526e-01
1.02579665e+00 7.99286067e-01 6.99727237e-01 -1.44425428e+00
-6.43439174e-01 -2.99555421e-01 -8.15455556e-01 -5.89498281e-01
1.04346037e+00 -2.65802920e-01 -1.18484044e+00 8.97624612e-01
-1.32008004e+00 -1.67216137e-01 -3.68796378e-01 6.62951946e-01
-3.05298239e-01 5.46169877e-01 -8.22065473e-01 -9.91824806e-01
-3.06473315e-01 -6.45518303e-01 9.91412938e-01 6.95026457e-01
2.40161002e-01 -1.15004063e+00 8.20583552e-02 1.03729859e-01
5.16061306e-01 5.47504783e-01 1.07887709e+00 2.32456950e-04
-6.78963184e-01 -2.90410846e-01 -5.07029653e-01 4.45618927e-02
4.98348117e-01 9.71779227e-01 -6.98604643e-01 1.07613169e-01
1.98977336e-01 -2.15056434e-01 4.00839895e-01 1.05621636e+00
8.72520924e-01 2.97461689e-01 -4.35389310e-01 3.40836048e-01
1.62928319e+00 4.32678968e-01 6.54515266e-01 9.05426890e-02
1.59544289e-01 9.79700863e-01 9.37218308e-01 6.38954937e-01
-4.38515842e-01 2.55161911e-01 7.33982146e-01 -1.96352571e-01
3.76864970e-02 7.84713745e-01 1.31016433e-01 6.97722554e-01
-5.34133136e-01 -3.66601825e-01 -7.52118886e-01 4.25354332e-01
-1.39637327e+00 -1.35823631e+00 -7.21800923e-01 1.91603518e+00
7.21029520e-01 -9.71358269e-02 -4.56458658e-01 2.84167796e-01
1.04442370e+00 -1.00795172e-01 -4.11762774e-01 -2.18286023e-01
-5.48990309e-01 3.79929036e-01 6.18999243e-01 2.92693615e-01
-6.27805233e-01 7.96535313e-01 6.30452967e+00 8.84961545e-01
-1.16883671e+00 -8.40553865e-02 5.54347932e-01 8.70649576e-01
-9.37544405e-02 3.14257115e-01 -5.57871938e-01 1.82531804e-01
6.60422504e-01 3.25497389e-01 2.57206678e-01 3.72733533e-01
7.29823411e-01 -9.02815938e-01 -2.25130916e-01 4.33758914e-01
-4.50158328e-01 -1.18258214e+00 -2.55148172e-01 1.38948515e-01
5.00778556e-01 -4.82954800e-01 -1.84113637e-01 -4.91873652e-01
1.08572215e-01 -5.75563967e-01 -1.05753131e-02 1.03427613e+00
6.89218581e-01 -2.94438332e-01 8.40861320e-01 2.84806132e-01
-1.27737308e+00 -1.04644887e-01 -6.26717627e-01 1.18683368e-01
-8.12574569e-03 1.31630182e+00 -4.83408898e-01 5.65516472e-01
6.61397576e-01 6.51743948e-01 -5.07631898e-01 8.48116696e-01
-6.54509962e-02 8.29256833e-01 -4.54610109e-01 -5.18323481e-02
-2.33532814e-03 -8.06291759e-01 6.48195505e-01 5.46531141e-01
8.19542766e-01 2.80965984e-01 7.42112771e-02 1.22778511e+00
7.02398121e-01 6.89024627e-02 -6.11147344e-01 -4.11922216e-01
5.73952377e-01 1.55039084e+00 -1.40328920e+00 -2.65971422e-01
1.10455193e-01 7.64142752e-01 -5.61252475e-01 1.63028568e-01
-2.86267102e-01 -3.70676219e-01 -1.97950541e-03 3.17117921e-03
2.69470751e-01 -2.07038447e-01 -4.21689153e-01 -6.58316672e-01
-4.07831013e-01 -3.25701773e-01 -1.73162416e-01 -1.08854723e+00
-1.17854714e+00 9.55896154e-02 -3.52366835e-01 -8.79010499e-01
3.17463666e-01 -9.40357685e-01 -6.41452134e-01 1.00323355e+00
-1.14498138e+00 -1.23495722e+00 -8.38557839e-01 4.35817152e-01
2.28733160e-02 -5.43870814e-02 1.39369571e+00 -2.40198091e-01
-6.36264801e-01 -4.67489272e-01 5.69728673e-01 -3.91275555e-01
2.80396074e-01 -9.88949776e-01 -5.35040081e-01 5.84527433e-01
-4.04873431e-01 -2.68417895e-01 7.77314723e-01 -9.99401927e-01
-1.24383378e+00 -6.47875190e-01 9.06409144e-01 5.58336452e-02
7.08131075e-01 4.78577644e-01 -8.19669306e-01 -1.84298605e-01
1.46868512e-01 1.36873171e-01 5.54762244e-01 -4.06966984e-01
3.50294262e-01 -2.21672863e-01 -1.42988336e+00 3.71313125e-01
5.45427442e-01 -3.75935793e-01 -5.15902750e-02 6.78111076e-01
7.20809922e-02 1.65615648e-01 -1.18161297e+00 6.49547577e-01
5.64863563e-01 -9.39891458e-01 7.16616929e-01 -1.44064143e-01
1.98781982e-01 -5.26978433e-01 -1.26636505e-01 -9.75733459e-01
-6.15226090e-01 -4.88115638e-01 2.44186252e-01 1.37564957e+00
-2.36598644e-02 -3.81268442e-01 7.53614366e-01 -2.05486149e-01
3.23091775e-01 -2.05657616e-01 -6.64605856e-01 -5.92982292e-01
-4.42038298e-01 3.59871477e-01 1.35271564e-01 9.58568215e-01
-3.85649204e-01 -2.30199859e-01 1.48908347e-01 6.25193655e-01
9.63942409e-01 1.83778197e-01 2.40933359e-01 -1.98782980e+00
-3.65541875e-02 -3.24731112e-01 -6.10093236e-01 -1.79164529e-01
6.69208467e-02 -3.67460668e-01 6.03543036e-02 -1.33199799e+00
-4.12797220e-02 -1.93350747e-01 1.22935034e-01 1.80934981e-01
1.60677075e-01 2.73793966e-01 -2.04142958e-01 4.11099851e-01
3.01373571e-01 1.63945436e-01 1.52057171e+00 -9.33809951e-03
-2.78561205e-01 2.80941576e-01 -6.98684752e-02 7.25620329e-01
8.11301053e-01 -3.17834944e-01 -2.36126691e-01 1.46364614e-01
1.19357035e-02 3.94553661e-01 7.81239927e-01 -1.14445651e+00
-5.32719446e-03 -5.07218778e-01 1.70459956e-01 -6.03174508e-01
-4.33055172e-03 -1.19149375e+00 2.49118298e-01 9.92580414e-01
6.42557964e-02 -3.56240183e-01 -8.43401626e-02 5.92086971e-01
-2.51297802e-02 -6.31091475e-01 1.10527635e+00 -2.15170726e-01
-5.50721645e-01 -2.14134753e-02 -9.43829656e-01 -5.22440612e-01
1.41302669e+00 -5.88883221e-01 -4.73401099e-01 -6.69778064e-02
-9.46031809e-01 -3.51703733e-01 2.16754243e-01 -3.97037417e-01
1.77971229e-01 -9.64560986e-01 -5.18277347e-01 9.34718996e-02
-2.78907001e-01 -3.08105797e-01 3.06387514e-01 1.04115999e+00
-1.30630398e+00 1.07340589e-01 -8.11358750e-01 -1.00441432e+00
-1.25972378e+00 1.97457761e-01 1.71254367e-01 -1.60159871e-01
-3.41730230e-02 5.66267610e-01 -4.03222263e-01 1.32354274e-01
-2.66951293e-01 -2.55225092e-01 -8.33697736e-01 2.85100549e-01
1.51474521e-01 5.85503459e-01 -1.93849042e-01 -5.34200311e-01
-4.55973297e-02 9.97433543e-01 6.29685938e-01 2.65006036e-01
1.34624326e+00 -3.32998157e-01 -8.25028062e-01 8.05949926e-01
4.71651107e-01 -1.31198823e-01 -1.01973116e+00 2.66205780e-02
1.77449644e-01 -4.32268143e-01 3.65778297e-01 -5.94062984e-01
-1.08993542e+00 6.94878519e-01 1.08286333e+00 1.15467381e+00
1.52570713e+00 -8.57865959e-02 4.30189222e-01 4.21752542e-01
1.99889913e-01 -1.20547593e+00 -9.65559632e-02 2.83549190e-01
4.47347015e-01 -1.08765948e+00 -9.19046402e-02 -9.34164882e-01
1.82147309e-01 1.51941216e+00 3.58561873e-01 -2.89435059e-01
1.02466810e+00 4.66401339e-01 -2.18353029e-02 -4.60046202e-01
-6.35998964e-01 -4.60309505e-01 -2.12837785e-01 1.09843278e+00
3.81339431e-01 1.97905451e-01 -5.77781916e-01 -8.08644593e-02
4.98508930e-01 3.92056346e-01 5.69785833e-01 9.48857903e-01
-8.98093402e-01 -9.10605788e-01 -9.34279501e-01 3.11890781e-01
-3.24860215e-01 1.85410291e-01 -3.34458560e-01 5.91688812e-01
4.32201564e-01 7.77226746e-01 5.43982461e-02 1.29634724e-03
1.49739549e-01 -3.32525298e-02 6.58775508e-01 -5.56107938e-01
-1.07208587e-01 2.30268404e-01 -5.92805088e-01 -3.14326704e-01
-1.05788231e+00 -7.04926074e-01 -9.39326108e-01 -3.87019545e-01
-7.15897083e-01 -2.22672168e-02 7.02828705e-01 1.08297563e+00
2.79139280e-01 3.13688040e-01 8.56915295e-01 -7.59980202e-01
-3.19864184e-01 -1.00087678e+00 -1.41960156e+00 1.41941503e-01
1.82398818e-02 -7.94068575e-01 -4.45043236e-01 4.54594851e-01]
|
[9.189830780029297, -1.6011667251586914]
|
0fad5b1f-a4c6-4d3d-be90-950d167970a9
|
contrastive-energy-prediction-for-exact
|
2304.12824
| null |
https://arxiv.org/abs/2304.12824v2
|
https://arxiv.org/pdf/2304.12824v2.pdf
|
Contrastive Energy Prediction for Exact Energy-Guided Diffusion Sampling in Offline Reinforcement Learning
|
Guided sampling is a vital approach for applying diffusion models in real-world tasks that embeds human-defined guidance during the sampling procedure. This paper considers a general setting where the guidance is defined by an (unnormalized) energy function. The main challenge for this setting is that the intermediate guidance during the diffusion sampling procedure, which is jointly defined by the sampling distribution and the energy function, is unknown and is hard to estimate. To address this challenge, we propose an exact formulation of the intermediate guidance as well as a novel training objective named contrastive energy prediction (CEP) to learn the exact guidance. Our method is guaranteed to converge to the exact guidance under unlimited model capacity and data samples, while previous methods can not. We demonstrate the effectiveness of our method by applying it to offline reinforcement learning (RL). Extensive experiments on D4RL benchmarks demonstrate that our method outperforms existing state-of-the-art algorithms. We also provide some examples of applying CEP for image synthesis to demonstrate the scalability of CEP on high-dimensional data.
|
['Jun Zhu', 'Chongxuan Li', 'Hang Su', 'Jianfei Chen', 'Huayu Chen', 'Cheng Lu']
|
2023-04-25
| null | null | null | null |
['d4rl']
|
['robots']
|
[ 2.03221187e-01 2.01664209e-01 -3.62969607e-01 -7.64429569e-02
-1.03799880e+00 -4.06971246e-01 8.99341226e-01 -6.65570647e-02
-3.69745612e-01 6.29155636e-01 1.22600637e-01 -2.51777261e-01
1.94086302e-02 -6.78472698e-01 -9.47000742e-01 -8.50943744e-01
-3.00328415e-02 3.58765155e-01 9.80363488e-02 -1.78324386e-01
4.71016765e-01 2.83312976e-01 -1.30039632e+00 -6.86236843e-02
9.85688984e-01 1.06023145e+00 5.62662303e-01 7.65075028e-01
-6.03281744e-02 6.63167834e-01 -6.18228018e-01 7.03733265e-02
4.40910608e-01 -8.05188775e-01 -8.06067944e-01 2.82516271e-01
2.44582772e-01 -7.28498220e-01 -7.75678009e-02 1.16210175e+00
6.24275744e-01 3.17570806e-01 7.53254712e-01 -1.04818773e+00
-7.59268582e-01 4.87854213e-01 -5.46084881e-01 4.29575108e-02
8.90190527e-02 2.00202435e-01 9.60343778e-01 -8.89463603e-01
8.05007696e-01 1.08460009e+00 2.92200476e-01 7.52316833e-01
-1.38159907e+00 -3.55017126e-01 4.38036710e-01 1.92680746e-01
-1.08669209e+00 -3.86875570e-01 8.40892613e-01 -3.98589820e-01
7.10141480e-01 -1.72313526e-01 7.32050717e-01 1.14918590e+00
1.23318441e-01 1.06212413e+00 1.34894121e+00 -5.62629223e-01
7.32453704e-01 7.14278920e-03 -1.86874002e-01 9.18632925e-01
-2.05359504e-01 4.01151508e-01 -5.87649763e-01 -1.90106586e-01
1.10018051e+00 -3.60616505e-01 -4.68734920e-01 -6.24339879e-01
-1.18958688e+00 1.03696907e+00 2.73329288e-01 -2.11471334e-01
-1.63686156e-01 5.07028699e-01 2.32380733e-01 3.64872187e-01
7.02014983e-01 2.70343959e-01 -1.93745196e-01 -3.35617781e-01
-9.52594817e-01 4.92720246e-01 9.16380107e-01 9.61344361e-01
8.51472974e-01 3.95083055e-02 -3.49701434e-01 8.35182309e-01
3.11493516e-01 4.00451332e-01 4.77061361e-01 -1.38150811e+00
3.74802619e-01 2.35916488e-02 4.84917611e-01 -5.53827405e-01
-4.02674116e-02 -4.04660463e-01 -7.57185042e-01 4.22594815e-01
5.10179162e-01 -1.51336595e-01 -7.90299058e-01 1.98098540e+00
5.76505721e-01 4.41863000e-01 -1.60283059e-01 9.61014569e-01
1.36504427e-01 7.00231016e-01 -3.32165062e-01 -3.27776760e-01
7.67254710e-01 -1.36369419e+00 -7.13325441e-01 -4.88164723e-02
4.78876323e-01 -5.18928170e-01 1.51400948e+00 5.04649043e-01
-1.29281223e+00 -4.89134014e-01 -1.28534710e+00 -1.26534272e-02
1.52767286e-01 6.49359673e-02 4.67658460e-01 6.27857506e-01
-1.14978302e+00 9.56084430e-01 -1.05760205e+00 -2.10277095e-01
3.94360691e-01 2.79750824e-01 3.60340059e-01 1.87349897e-02
-9.00818110e-01 6.30588055e-01 -1.41572088e-01 -1.80271000e-01
-1.35665917e+00 -7.20042706e-01 -6.63580775e-01 -1.86396256e-01
5.28013587e-01 -6.11353993e-01 1.67458594e+00 -8.51055920e-01
-2.31677389e+00 5.19102275e-01 1.90345142e-02 -6.14690781e-01
7.66832829e-01 -2.31226638e-01 5.19174039e-02 2.92803645e-01
-8.33156779e-02 7.32525170e-01 1.16722059e+00 -1.16249561e+00
-5.84626675e-01 -9.66615379e-02 2.52925038e-01 3.01960915e-01
-3.10346216e-01 -5.25786638e-01 -4.82812464e-01 -8.65302205e-01
-3.07072759e-01 -9.75443661e-01 -4.65745956e-01 3.58868390e-01
-2.46970296e-01 -2.29558736e-01 6.85763419e-01 -4.21997249e-01
1.10032535e+00 -1.93956029e+00 2.93209493e-01 1.82859153e-01
9.44396630e-02 -3.32356319e-02 -2.43879870e-01 3.70477349e-01
5.26549935e-01 -2.11103205e-02 -5.39247334e-01 -4.76833642e-01
3.06254804e-01 2.57630467e-01 -4.97811973e-01 6.61904156e-01
1.50572464e-01 8.30746710e-01 -1.11199570e+00 -3.33989352e-01
1.08265020e-01 4.86748278e-01 -8.44595790e-01 3.85000467e-01
-5.14199972e-01 5.20402431e-01 -4.11134332e-01 3.23259920e-01
5.59301615e-01 -4.06993896e-01 1.14780866e-01 -1.19305581e-01
4.44339626e-02 2.30749071e-01 -1.21827185e+00 2.07456303e+00
-6.21091068e-01 4.75419223e-01 1.91464618e-01 -1.09857202e+00
8.26402605e-01 -6.55772164e-02 5.48404694e-01 -6.97130978e-01
-1.63919002e-01 2.29826570e-01 -3.22024465e-01 -2.04038769e-01
4.32722062e-01 2.94447541e-02 1.67658612e-01 8.88653040e-01
7.29020359e-03 -5.05302489e-01 7.33335242e-02 1.49430022e-01
1.14807451e+00 4.36171055e-01 2.06885666e-01 -5.72347641e-01
3.26602042e-01 -2.79840350e-01 3.58893424e-01 8.70225549e-01
-9.68533903e-02 5.07792950e-01 4.92558300e-01 -4.75050882e-02
-1.15058124e+00 -1.01800251e+00 -3.19286175e-02 7.91956186e-01
2.85059959e-01 -4.57151413e-01 -1.14004886e+00 -9.73522425e-01
-1.33807078e-01 7.17050672e-01 -7.37399459e-01 -4.66557741e-02
-4.61407959e-01 -5.36001742e-01 1.15588643e-01 3.88895512e-01
5.34859776e-01 -7.20663846e-01 -7.02234089e-01 3.86276066e-01
1.16554638e-02 -9.84976232e-01 -9.93937790e-01 4.05938402e-02
-8.19920063e-01 -9.46174681e-01 -9.50215340e-01 -3.99799168e-01
7.63020992e-01 1.53630435e-01 1.09803665e+00 5.46280481e-02
4.75603342e-02 7.47927070e-01 -2.15119243e-01 -2.14776546e-01
-5.90246916e-01 1.44740552e-01 -1.85619816e-01 -1.79025754e-02
-1.72106475e-01 -4.27397370e-01 -1.03236377e+00 4.57738549e-01
-1.05745113e+00 1.72773406e-01 3.79156768e-01 1.12822235e+00
1.07553637e+00 -1.36927024e-01 7.13183343e-01 -8.48762333e-01
8.83182526e-01 -2.68242508e-01 -9.04718399e-01 7.06057772e-02
-9.57813442e-01 5.11177599e-01 6.37717485e-01 -6.74164712e-01
-9.58306849e-01 1.35758463e-02 -1.53888121e-01 -4.10696805e-01
4.26080137e-01 3.15881819e-01 5.55813648e-02 -2.56649792e-01
5.86720169e-01 1.94185570e-01 1.57693118e-01 -3.46765131e-01
6.10725582e-01 4.46625680e-01 3.42321366e-01 -9.63255644e-01
5.76430261e-01 7.29402661e-01 8.85773972e-02 -6.62631691e-01
-9.00579453e-01 -6.58480003e-02 -3.73036832e-01 -1.51023895e-01
4.95795131e-01 -6.82447433e-01 -5.98694086e-01 2.61139154e-01
-9.66771305e-01 -1.12944889e+00 -7.62930691e-01 4.09096569e-01
-1.16884100e+00 4.60955977e-01 -5.59511602e-01 -8.48628938e-01
-3.13450903e-01 -1.51467645e+00 1.14246976e+00 -2.29887962e-01
8.63160789e-02 -1.15610516e+00 2.85632193e-01 -6.87083304e-02
4.04844940e-01 1.77937523e-01 9.17437911e-01 -1.59197107e-01
-8.34915757e-01 3.34501237e-01 6.39405698e-02 6.32568777e-01
4.86015417e-02 -1.25503257e-01 -6.90615594e-01 -3.36945057e-01
3.11231077e-01 -5.62532246e-01 1.08530092e+00 4.70846295e-01
1.42008686e+00 -3.73929083e-01 -1.14408676e-02 7.16393232e-01
1.32964051e+00 -1.39116019e-01 5.12755752e-01 1.20941252e-01
3.75576317e-01 2.91599393e-01 9.18570876e-01 6.77545786e-01
3.14612269e-01 8.57839942e-01 4.39245850e-01 1.47217274e-01
-2.79890388e-01 -4.23760444e-01 6.61881864e-01 9.65593398e-01
3.68600935e-02 -1.94142386e-01 -4.89579350e-01 4.38737035e-01
-1.96014380e+00 -8.52057397e-01 3.16707522e-01 2.25844789e+00
1.19110584e+00 3.96544114e-02 2.00790882e-01 6.21473864e-02
4.05442268e-01 2.81773984e-01 -1.09564006e+00 -3.63219619e-01
1.29793823e-01 3.76472980e-01 5.32521427e-01 8.04328859e-01
-8.84284735e-01 7.29787052e-01 7.19089890e+00 1.21700525e+00
-1.00892556e+00 2.57987559e-01 7.13981152e-01 -3.26242968e-02
-5.38082123e-01 -1.27861604e-01 -8.60730886e-01 4.57529902e-01
8.05007577e-01 -2.18788370e-01 1.02323210e+00 9.37206388e-01
3.23771089e-01 -1.31464437e-01 -1.27304101e+00 8.44167888e-01
-8.95121843e-02 -1.54477441e+00 -9.83021855e-02 2.70590901e-01
1.05516100e+00 2.17823144e-02 3.96393359e-01 1.57892570e-01
5.84251761e-01 -7.92156577e-01 8.39530528e-01 4.92518008e-01
8.93926442e-01 -7.80217648e-01 -3.63831371e-02 4.63630527e-01
-9.79027808e-01 1.46485474e-02 -5.71390927e-01 2.18175784e-01
6.63336739e-02 7.36411333e-01 -7.49685526e-01 1.92223653e-01
1.70138642e-01 8.70088816e-01 -1.61554307e-01 7.79975176e-01
-4.17938054e-01 6.69211209e-01 -3.77545327e-01 -3.48812006e-02
3.27381402e-01 -2.99059331e-01 4.45980251e-01 1.01877141e+00
6.13278210e-01 -1.52001038e-01 2.33688653e-01 1.11220706e+00
-2.50441164e-01 9.76092890e-02 -3.12562346e-01 3.37451845e-02
3.51938516e-01 9.91586685e-01 -5.73894739e-01 -1.56099930e-01
-2.85669237e-01 1.22001636e+00 5.34923792e-01 6.10126019e-01
-9.80916858e-01 -1.10856198e-01 5.49619377e-01 4.52999361e-02
4.53147918e-01 -5.10323882e-01 5.45489565e-02 -1.10091197e+00
-8.54849368e-02 -9.77478325e-01 1.21989019e-01 -5.16035140e-01
-1.28702438e+00 4.00627971e-01 -1.72417462e-02 -1.10700643e+00
-5.46123624e-01 -4.62930739e-01 -3.07597607e-01 7.05940485e-01
-1.81025422e+00 -6.28029108e-01 -2.91306853e-01 6.25399053e-01
8.82542729e-01 -5.98868094e-02 6.85388863e-01 2.31672246e-02
-3.45192075e-01 5.40822923e-01 2.99148947e-01 -4.09074843e-01
6.52731419e-01 -1.58551311e+00 4.40867782e-01 5.85960150e-01
1.92158282e-01 7.64029324e-02 6.87146366e-01 -5.28722525e-01
-1.55653632e+00 -1.13530076e+00 8.35452378e-02 -1.79370940e-01
7.82852769e-01 -3.40214193e-01 -6.48995459e-01 4.81082529e-01
3.00425977e-01 1.93452522e-01 4.05631304e-01 -4.86118376e-01
-6.73396438e-02 -8.37019235e-02 -1.10921597e+00 6.74761593e-01
1.20193815e+00 -4.52547729e-01 -5.79778031e-02 4.80787843e-01
8.17206442e-01 -7.16621816e-01 -8.75674009e-01 1.31260276e-01
3.46926391e-01 -1.02334607e+00 8.68412852e-01 -2.03794748e-01
5.07292211e-01 -1.03489101e-01 -3.11657518e-01 -1.64576101e+00
8.43102485e-02 -1.07586288e+00 -8.32857549e-01 9.03597176e-01
2.55418330e-01 -4.60615814e-01 8.09057355e-01 4.10355777e-01
-1.26388092e-02 -1.28761923e+00 -8.01339746e-01 -9.22784388e-01
1.70860887e-01 -5.82289279e-01 6.16494119e-01 5.64472377e-01
-2.38598660e-01 9.99500230e-02 -4.20037240e-01 -1.10704191e-01
1.00359404e+00 7.12003782e-02 7.75038004e-01 -7.18758225e-01
-7.35507071e-01 -4.59481061e-01 8.45910236e-02 -1.96312332e+00
2.56760865e-01 -7.26592422e-01 1.11183055e-01 -1.50967526e+00
-4.41719778e-02 -6.30694091e-01 -1.27039805e-01 -9.68245268e-02
-4.78342772e-02 -1.65133938e-01 6.01347499e-02 1.75526306e-01
-4.91142392e-01 1.09233236e+00 1.74482191e+00 -1.24710210e-01
-3.74806881e-01 1.05230808e-01 -4.46227640e-01 5.99859715e-01
7.03929722e-01 -4.37931806e-01 -1.00057912e+00 -4.80648071e-01
1.90396890e-01 4.07902570e-03 4.09931302e-01 -8.25233936e-01
4.04009074e-01 -3.80024463e-01 -8.61875117e-02 -5.86147964e-01
4.78652447e-01 -6.19146764e-01 -2.40705714e-01 4.54552501e-01
-6.86857939e-01 -1.47694767e-01 -2.35165507e-01 8.47045004e-01
8.20809305e-02 -3.53572726e-01 8.59198451e-01 3.90502401e-02
-3.65908206e-01 5.45030236e-01 4.54246160e-03 3.91632736e-01
9.27866757e-01 6.69102818e-02 -2.33243883e-01 -6.60623670e-01
-5.88761151e-01 1.88313842e-01 6.26258075e-01 -3.66971418e-02
8.17817807e-01 -1.56416464e+00 -5.56368470e-01 2.11805195e-01
-1.70690417e-01 1.50151001e-02 -1.20924436e-01 7.99132109e-01
-3.83719116e-01 4.80825938e-02 2.20962092e-01 -6.15666449e-01
-6.37734175e-01 5.99978387e-01 5.02619624e-01 -5.31139195e-01
-7.20343173e-01 6.13217056e-01 1.61635280e-01 -1.61054522e-01
4.77275074e-01 -6.13946855e-01 2.71311462e-01 -2.67401725e-01
5.09209692e-01 3.23783010e-01 -1.02813192e-01 -1.14886604e-01
9.01846290e-02 5.59404135e-01 -1.29437922e-02 -5.47423124e-01
1.37019897e+00 -3.59330386e-01 3.53595525e-01 4.40338075e-01
1.22730970e+00 -5.47856539e-02 -2.05957794e+00 -3.25353086e-01
-2.75415331e-01 -5.08612990e-01 3.34691048e-01 -5.65279901e-01
-1.18922937e+00 9.70922351e-01 6.07111573e-01 2.05538645e-01
1.09159696e+00 -1.63680151e-01 1.02239633e+00 4.28314179e-01
4.79755819e-01 -1.54408753e+00 5.13441980e-01 2.29144916e-01
9.82411802e-01 -9.88812506e-01 9.00279451e-03 -1.85726658e-01
-6.13025129e-01 9.95993435e-01 3.28363776e-01 -3.20387870e-01
8.64111662e-01 3.58166039e-01 -3.01854849e-01 9.44712609e-02
-9.23462331e-01 -1.57974437e-01 2.14614674e-01 6.81912482e-01
1.13047838e-01 -2.24281713e-01 -2.98719317e-01 3.83714527e-01
4.74814372e-03 1.40147522e-01 5.17946899e-01 9.52725768e-01
-4.06774700e-01 -1.36438715e+00 -4.90333065e-02 3.12642515e-01
-4.20110345e-01 1.11878991e-01 -2.22216528e-02 6.51280403e-01
-2.94752836e-01 7.73869216e-01 -6.06704019e-02 -1.24574646e-01
1.44411594e-01 -3.22362632e-01 8.59036028e-01 -4.36963290e-01
-2.26733014e-01 2.10887447e-01 -1.99608177e-01 -8.92237842e-01
-6.80706263e-01 -5.53826094e-01 -1.21684372e+00 -2.67552614e-01
-1.36785969e-01 -2.86224037e-02 8.22331607e-01 8.05601597e-01
3.77229780e-01 4.51050490e-01 8.43035161e-01 -1.01617229e+00
-1.20832229e+00 -5.15493333e-01 -6.63827837e-01 2.32025564e-01
5.17739594e-01 -7.64733851e-01 -4.09943283e-01 3.66960317e-02]
|
[4.144830226898193, 2.127885103225708]
|
07ca06d3-db23-4060-8b28-255a1ea6d549
|
vector-based-representation-is-the-key-a
|
2305.18063
| null |
https://arxiv.org/abs/2305.18063v1
|
https://arxiv.org/pdf/2305.18063v1.pdf
|
Vector-based Representation is the Key: A Study on Disentanglement and Compositional Generalization
|
Recognizing elementary underlying concepts from observations (disentanglement) and generating novel combinations of these concepts (compositional generalization) are fundamental abilities for humans to support rapid knowledge learning and generalize to new tasks, with which the deep learning models struggle. Towards human-like intelligence, various works on disentangled representation learning have been proposed, and recently some studies on compositional generalization have been presented. However, few works study the relationship between disentanglement and compositional generalization, and the observed results are inconsistent. In this paper, we study several typical disentangled representation learning works in terms of both disentanglement and compositional generalization abilities, and we provide an important insight: vector-based representation (using a vector instead of a scalar to represent a concept) is the key to empower both good disentanglement and strong compositional generalization. This insight also resonates the neuroscience research that the brain encodes information in neuron population activity rather than individual neurons. Motivated by this observation, we further propose a method to reform the scalar-based disentanglement works ($\beta$-TCVAE and FactorVAE) to be vector-based to increase both capabilities. We investigate the impact of the dimensions of vector-based representation and one important question: whether better disentanglement indicates higher compositional generalization. In summary, our study demonstrates that it is possible to achieve both good concept recognition and novel concept composition, contributing an important step towards human-like intelligence.
|
['Nanning Zheng', 'Yan Lu', 'Cuiling Lan', 'Yuwang Wang', 'Tao Yang']
|
2023-05-29
| null | null | null | null |
['disentanglement']
|
['methodology']
|
[ 1.98694751e-01 -1.24367699e-01 -1.96648300e-01 -1.68379620e-01
9.09113064e-02 -6.89614415e-01 9.64459777e-01 1.17241561e-01
-3.04712594e-01 7.02844381e-01 2.88370192e-01 -6.20027818e-02
-5.17130673e-01 -9.17002618e-01 -3.03129077e-01 -9.92893696e-01
-9.88449603e-02 3.25679213e-01 -3.34704995e-01 -5.45251727e-01
1.45053223e-01 6.35338187e-01 -1.92200005e+00 2.36840367e-01
1.00666523e+00 6.63325489e-01 8.20553117e-03 5.60901701e-01
-5.00864722e-02 4.53806996e-01 -8.43720615e-01 -4.44638193e-01
3.22778285e-01 -5.25248766e-01 -3.62902761e-01 -3.33294541e-01
1.57628328e-01 9.71496571e-03 -2.86804885e-01 1.03549480e+00
4.18772876e-01 2.27470636e-01 9.11882699e-01 -1.37622678e+00
-1.14865363e+00 7.76865005e-01 -4.65358794e-01 3.68708730e-01
1.88868850e-01 1.05457544e-01 1.14852965e+00 -6.25941157e-01
1.03503786e-01 1.24676251e+00 2.57081896e-01 7.23291755e-01
-1.31507015e+00 -1.22957993e+00 2.46229663e-01 3.69017512e-01
-1.21017408e+00 -1.32393464e-01 9.86228228e-01 -5.86410701e-01
8.71321201e-01 4.58163440e-01 1.05165362e+00 1.48138714e+00
7.63270780e-02 7.86281049e-01 1.35367954e+00 -2.99979419e-01
2.92467862e-01 1.52393058e-01 2.69835591e-01 5.82952380e-01
9.08936501e-01 5.08176744e-01 -6.82348669e-01 1.70331970e-02
9.57640946e-01 3.19546789e-01 -5.11586368e-01 -4.84774709e-01
-1.42299306e+00 1.10032713e+00 4.72630024e-01 4.63968784e-01
-2.79041767e-01 1.19293347e-01 4.41375941e-01 4.65462357e-01
1.18271880e-01 1.12640882e+00 -4.32353348e-01 -7.79816583e-02
-9.00531530e-01 2.16833219e-01 5.31244695e-01 4.27134514e-01
6.69240534e-01 7.36849785e-01 -3.74711826e-02 5.61192095e-01
1.19017273e-01 5.88643551e-01 1.07745409e+00 -5.96099257e-01
2.19105631e-01 8.95622969e-01 -6.36241376e-01 -1.11764443e+00
-4.11572963e-01 -9.64534163e-01 -1.26098156e+00 4.81701523e-01
6.43376410e-02 -2.78315805e-02 -6.16595984e-01 2.10760093e+00
-2.19098538e-01 3.16959709e-01 4.11522865e-01 8.09390306e-01
9.48381722e-01 5.11100888e-01 -1.41331777e-01 -1.95889562e-01
1.47292459e+00 -5.58511913e-01 -7.82315791e-01 -1.22642569e-01
4.30400431e-01 -1.24730177e-01 9.44863498e-01 6.35251284e-01
-8.55805755e-01 -6.75393283e-01 -1.72191918e+00 1.16608121e-01
-6.65672600e-01 1.66911986e-02 1.15489066e+00 8.70332420e-01
-7.37379134e-01 4.98591751e-01 -7.19490349e-01 -8.53164569e-02
4.55869496e-01 3.56021464e-01 -6.82543218e-01 5.67013584e-02
-1.37055886e+00 9.51730788e-01 7.62795150e-01 -7.99095780e-02
-6.03282869e-01 -6.37753189e-01 -9.06821489e-01 2.65018582e-01
2.71366894e-01 -1.16451406e+00 5.41715622e-01 -8.06426883e-01
-1.47487688e+00 5.32291472e-01 2.53432482e-01 -4.68058437e-01
-3.40516772e-03 -8.72305930e-02 -4.69723642e-01 -6.46166056e-02
-3.66200924e-01 4.91745412e-01 8.50815952e-01 -1.31953490e+00
-2.21790045e-01 -7.02469587e-01 1.40823990e-01 3.05579126e-01
-5.81948578e-01 -3.95657480e-01 5.46918392e-01 -9.26796198e-01
1.73703894e-01 -6.00429535e-01 6.83219433e-02 -1.54312372e-01
-5.59588671e-02 -2.30207548e-01 5.20486891e-01 -2.87445843e-01
1.05891860e+00 -2.08495760e+00 7.49169052e-01 -1.69858232e-01
8.57919931e-01 4.56795096e-01 -3.78779352e-01 3.64705622e-01
-3.41305107e-01 1.48801431e-01 -1.36214942e-01 -7.23056197e-02
3.84077057e-02 3.96089017e-01 -6.62380040e-01 4.35612917e-01
3.58831853e-01 1.18546033e+00 -9.98263240e-01 1.31427735e-01
5.49318418e-02 5.87434530e-01 -5.89281559e-01 1.92085996e-01
3.47760110e-03 3.14563602e-01 -3.40658516e-01 2.74556488e-01
4.84027088e-01 7.77896196e-02 2.23708406e-01 -4.02139574e-01
8.75766277e-02 9.83956829e-02 -1.14474678e+00 1.57985234e+00
-3.45228106e-01 8.70255232e-01 -4.74906087e-01 -1.47540581e+00
1.29324389e+00 3.31081450e-01 7.32581988e-02 -5.79568088e-01
3.34089488e-01 1.38005260e-02 4.93664324e-01 -3.50501329e-01
3.75573456e-01 -5.14854908e-01 2.67366245e-02 4.85745579e-01
3.75237793e-01 -3.15542072e-01 1.19382910e-01 1.50140107e-01
7.67504156e-01 -1.24773700e-02 8.08353841e-01 -3.63578469e-01
3.33460808e-01 -4.62422520e-01 4.80489552e-01 5.44858694e-01
1.97668727e-02 4.11164552e-01 5.20089984e-01 -3.66756499e-01
-6.58560216e-01 -1.42099023e+00 1.58110589e-01 9.45425451e-01
-1.07185151e-02 -3.01179856e-01 -2.17067972e-01 -2.73277014e-01
-8.05048645e-02 8.43084276e-01 -8.99019837e-01 -7.05267906e-01
-4.66180712e-01 -9.59525108e-01 7.97190428e-01 8.78051996e-01
4.11428005e-01 -8.76544595e-01 -7.41768956e-01 -2.26989061e-01
1.63200542e-01 -7.36427784e-01 5.83854783e-03 3.90743196e-01
-1.10478759e+00 -9.60612357e-01 -6.77769661e-01 -5.63801944e-01
4.86244410e-01 3.09185773e-01 1.05069268e+00 -1.18594408e-01
-2.32153282e-01 1.45607546e-01 -4.94672537e-01 -4.67344403e-01
-7.77363852e-02 -2.22708553e-01 4.79230136e-01 -1.37742475e-01
3.56693864e-01 -1.04922676e+00 -3.73625338e-01 -9.12113488e-02
-1.24763739e+00 1.38386786e-01 8.68834376e-01 1.05377257e+00
8.57020468e-02 -1.98428463e-02 8.39271963e-01 -5.44753015e-01
1.20667470e+00 -4.39324260e-01 -1.68222398e-01 2.28208363e-01
-7.36013114e-01 5.68551898e-01 6.54658377e-01 -8.95452082e-01
-8.36348236e-01 -4.74018663e-01 9.63809490e-02 -4.75204408e-01
-1.41536191e-01 6.12725139e-01 -2.04873592e-01 2.87808925e-01
8.65550876e-01 6.51453316e-01 2.21727192e-01 -2.30446383e-01
9.16825593e-01 3.86836290e-01 4.03163075e-01 -7.01973498e-01
7.39873946e-01 3.80095780e-01 -2.61932183e-02 -7.20987260e-01
-4.96934503e-01 -8.40752646e-02 -6.44018590e-01 2.89986283e-01
7.98200607e-01 -8.67252588e-01 -9.12213027e-01 9.27251279e-02
-1.16573501e+00 1.94137156e-01 -6.61224663e-01 7.92794526e-01
-5.54716110e-01 3.27276975e-01 -3.04480821e-01 -7.48120844e-01
-3.22345197e-01 -1.12001872e+00 6.80578232e-01 3.85685056e-01
-3.32249105e-01 -1.03417706e+00 2.96793699e-01 1.35805875e-01
4.81625021e-01 3.35566938e-01 1.28774107e+00 -1.06894767e+00
-2.35729903e-01 -4.00889404e-02 -2.21291274e-01 3.31076592e-01
4.97782081e-02 -3.67355913e-01 -9.70711172e-01 -4.38432783e-01
3.20304811e-01 -3.64227623e-01 1.17884529e+00 1.99441556e-02
8.34248602e-01 -3.95804346e-01 -1.40633270e-01 6.00770950e-01
1.18383408e+00 1.69869155e-01 5.76076746e-01 1.23016024e-02
5.10884643e-01 5.77144027e-01 -1.37043208e-01 2.97933817e-01
2.35485971e-01 4.87979412e-01 3.49018514e-01 1.28307179e-01
-3.11903298e-01 -1.13715202e-01 3.90137136e-01 1.40675116e+00
-5.27868271e-01 -1.98243752e-01 -5.89404821e-01 2.24658549e-01
-1.56622303e+00 -9.95345652e-01 2.28216037e-01 2.10099339e+00
8.02991450e-01 -5.66954687e-02 -2.62578893e-02 5.40481985e-01
3.32584262e-01 2.49222115e-01 -6.14773989e-01 -5.08162320e-01
-4.79694039e-01 4.45651799e-01 -2.07749590e-01 2.73442894e-01
-4.33606148e-01 6.11671627e-01 6.05327511e+00 8.92293036e-01
-1.27960074e+00 -3.48908603e-02 1.21202104e-01 -1.78395107e-01
-7.86872864e-01 -2.47232884e-01 -4.47443306e-01 6.15786277e-02
6.16420925e-01 -3.95353228e-01 4.42901731e-01 5.48828781e-01
-3.85971904e-01 3.29334974e-01 -1.52068925e+00 1.43982649e+00
5.37151337e-01 -1.40293801e+00 6.58023238e-01 5.08019514e-02
5.49054265e-01 -3.14447105e-01 4.43622530e-01 7.60302722e-01
2.85255194e-01 -1.41750288e+00 4.22877938e-01 5.76190889e-01
6.57350838e-01 -5.57814598e-01 6.80115044e-01 4.39053684e-01
-1.08141959e+00 -1.42552674e-01 -3.55299234e-01 -5.20555794e-01
-2.47034416e-01 5.23826897e-01 -6.44724786e-01 7.83639908e-01
1.64221495e-01 6.58782840e-01 -6.19585037e-01 6.33512735e-01
-5.76337516e-01 4.03873593e-01 1.46246821e-01 -3.56533945e-01
-2.34526172e-01 -2.10919023e-01 5.97824097e-01 9.68150318e-01
3.32231075e-01 2.54918128e-01 -1.86784923e-01 1.28957188e+00
1.92309693e-01 -1.08237982e-01 -6.91166520e-01 -4.74154592e-01
4.35383171e-01 9.67533290e-01 -6.45830214e-01 -3.59783769e-01
-2.59769619e-01 8.87018263e-01 5.58982670e-01 5.84636152e-01
-5.48127770e-01 -3.90464425e-01 9.98221695e-01 -4.32248175e-01
1.46390930e-01 -6.75185502e-01 -5.14447749e-01 -1.61074400e+00
-2.43453607e-01 -1.10784578e+00 2.19774082e-01 -7.03071833e-01
-1.44090307e+00 8.47029328e-01 1.68181181e-01 -9.61613476e-01
-1.77756950e-01 -8.60882163e-01 -5.08886039e-01 8.16609681e-01
-1.26076794e+00 -9.64418590e-01 -1.17349885e-01 4.15817082e-01
5.18359065e-01 -5.60810268e-01 1.15428722e+00 -8.31859261e-02
-4.82469231e-01 7.42234647e-01 4.56568226e-03 2.76255287e-05
1.57866120e-01 -1.10692453e+00 1.41628921e-01 6.16943538e-01
7.06839144e-01 1.08740270e+00 6.35083556e-01 -3.92771572e-01
-1.37762403e+00 -5.32736659e-01 5.31302929e-01 -6.49550557e-01
5.24850488e-01 -5.72280586e-01 -9.05570149e-01 4.37279433e-01
1.87745452e-01 -2.32141003e-01 1.12278187e+00 3.54270935e-01
-9.52075481e-01 -9.81235504e-03 -9.14558828e-01 8.84738445e-01
1.03682554e+00 -5.42424262e-01 -1.26130688e+00 -3.03577423e-01
1.11389637e+00 2.10812122e-01 -7.93544412e-01 6.10203505e-01
9.11890328e-01 -1.05202520e+00 1.22520852e+00 -8.13720047e-01
3.77713859e-01 -1.83421358e-01 -3.53507638e-01 -1.77769446e+00
-6.54540598e-01 -1.05602443e-01 -5.48270881e-01 8.65428090e-01
2.34334767e-01 -1.05643475e+00 5.67602515e-01 2.09786892e-01
1.54483438e-01 -1.09774363e+00 -7.55152106e-01 -1.11352003e+00
4.29371029e-01 -4.19649780e-01 8.87257218e-01 1.26096368e+00
3.08231771e-01 6.99944913e-01 -1.59168661e-01 -1.33635309e-02
3.26449752e-01 2.85147309e-01 4.35659707e-01 -1.57145357e+00
-4.57172334e-01 -9.88000631e-01 -9.74635303e-01 -9.90700722e-01
2.09202334e-01 -1.18711376e+00 -4.98378396e-01 -1.34022295e+00
3.96071255e-01 -1.85954601e-01 -4.70641881e-01 3.89757067e-01
-2.20141366e-01 1.05189078e-01 3.75021726e-01 1.12372048e-01
-1.68592632e-01 1.00136733e+00 1.29481781e+00 -3.34776163e-01
-1.37375429e-01 -2.76542604e-01 -1.17026722e+00 5.96601307e-01
7.41442263e-01 -2.10328475e-01 -7.63124049e-01 -4.31888252e-01
5.51159382e-01 -2.06063777e-01 2.80196846e-01 -9.76878524e-01
1.33000627e-01 -1.34436667e-01 4.29609686e-01 -1.69473469e-01
6.41762972e-01 -6.64509714e-01 -9.40440223e-02 7.29099870e-01
-4.03421283e-01 1.56159773e-01 1.97151780e-01 7.00286150e-01
-2.23724201e-01 -7.29580894e-02 4.91056889e-01 -1.07114792e-01
-6.97284281e-01 1.85074881e-01 -3.89311850e-01 7.73602948e-02
9.54715073e-01 -4.78419989e-01 -4.96769309e-01 -2.69486576e-01
-6.09534264e-01 -1.66435301e-01 -2.44645346e-02 7.30833471e-01
9.69403207e-01 -1.32619560e+00 -9.54414606e-01 6.91411853e-01
1.56894118e-01 -3.69314790e-01 7.20823780e-02 5.22823691e-01
-9.90383625e-02 5.21790624e-01 -5.73486626e-01 -3.57320905e-01
-8.81183386e-01 7.87581503e-01 1.22057378e-01 -6.85479566e-02
-3.33714366e-01 1.13284492e+00 5.88797688e-01 -5.10868013e-01
9.32941958e-02 -5.93057275e-01 -7.14886189e-01 2.99687833e-01
5.90312898e-01 4.69798058e-01 -1.06212214e-01 -4.16516095e-01
-2.22419634e-01 5.49059331e-01 3.96540097e-04 -8.18968341e-02
1.32743943e+00 2.20222637e-01 -9.65539441e-02 6.45847917e-01
1.09609652e+00 2.83881411e-04 -9.11607921e-01 -2.36508250e-01
-3.92285347e-01 -3.03302616e-01 -1.47340894e-01 -7.53802240e-01
-9.73453283e-01 1.22590172e+00 6.16635144e-01 2.44617939e-01
1.07083118e+00 -4.69718128e-02 2.86498755e-01 4.54571337e-01
4.10064280e-01 -3.84566933e-01 5.83153367e-01 4.25451607e-01
1.14569008e+00 -9.88819897e-01 2.37242654e-01 -2.45564550e-01
-6.38874531e-01 1.14817810e+00 7.14071929e-01 -2.98972130e-01
5.67831874e-01 1.29669026e-01 -2.26011589e-01 -3.60914946e-01
-7.12520957e-01 -2.38591805e-01 6.42735720e-01 9.59828019e-01
4.36687022e-01 3.97843391e-01 -3.98651570e-01 1.14791954e+00
-6.79702878e-01 -3.28579664e-01 4.54815537e-01 4.80265468e-01
-3.80590141e-01 -1.00186789e+00 -1.75479412e-01 4.31528240e-01
1.37586966e-01 -2.02616915e-01 -2.98669040e-01 6.79211140e-01
4.11273271e-01 8.88711810e-01 1.29371351e-02 -7.20677912e-01
1.39900461e-01 1.30015109e-02 8.01147461e-01 -8.57995749e-01
-3.53228420e-01 -6.99216664e-01 -3.29821795e-01 -2.89256364e-01
-3.59453887e-01 -2.54358560e-01 -1.08728337e+00 -1.12852596e-01
-5.59438407e-01 3.28621536e-01 6.43852651e-01 1.15562749e+00
2.89321214e-01 8.38843644e-01 1.56309605e-01 -6.91798389e-01
-6.57875359e-01 -8.92497838e-01 -6.59680367e-01 3.97441208e-01
3.21078986e-01 -9.85543668e-01 -4.52530056e-01 -2.21431553e-01]
|
[9.232056617736816, 4.860423564910889]
|
d1c84c53-44a0-4bc7-98a8-39c20216a786
|
a-probabilistic-end-to-end-task-oriented
|
2009.08115
| null |
https://arxiv.org/abs/2009.08115v3
|
https://arxiv.org/pdf/2009.08115v3.pdf
|
A Probabilistic End-To-End Task-Oriented Dialog Model with Latent Belief States towards Semi-Supervised Learning
|
Structured belief states are crucial for user goal tracking and database query in task-oriented dialog systems. However, training belief trackers often requires expensive turn-level annotations of every user utterance. In this paper we aim at alleviating the reliance on belief state labels in building end-to-end dialog systems, by leveraging unlabeled dialog data towards semi-supervised learning. We propose a probabilistic dialog model, called the LAtent BElief State (LABES) model, where belief states are represented as discrete latent variables and jointly modeled with system responses given user inputs. Such latent variable modeling enables us to develop semi-supervised learning under the principled variational learning framework. Furthermore, we introduce LABES-S2S, which is a copy-augmented Seq2Seq model instantiation of LABES. In supervised experiments, LABES-S2S obtains strong results on three benchmark datasets of different scales. In utilizing unlabeled dialog data, semi-supervised LABES-S2S significantly outperforms both supervised-only and semi-supervised baselines. Remarkably, we can reduce the annotation demands to 50% without performance loss on MultiWOZ.
|
['Yichi Zhang', 'Huixin Wang', 'Junlan Feng', 'Zhijian Ou']
|
2020-09-17
| null |
https://aclanthology.org/2020.emnlp-main.740
|
https://aclanthology.org/2020.emnlp-main.740.pdf
|
emnlp-2020-11
|
['end-to-end-dialogue-modelling']
|
['natural-language-processing']
|
[-2.54549861e-01 7.06485033e-01 -4.03089553e-01 -8.53078663e-01
-1.15056801e+00 -8.31564009e-01 6.77298367e-01 -1.84404388e-01
-3.68679047e-01 6.80522561e-01 5.16236782e-01 -2.51124918e-01
4.50278968e-01 -3.39927942e-01 -4.52894777e-01 -2.35964239e-01
3.79344463e-01 7.74754226e-01 3.10577065e-01 -5.41017830e-01
-2.40488201e-01 -2.72552252e-01 -7.54261136e-01 2.41814941e-01
5.52401841e-01 8.33603919e-01 3.17723900e-01 8.34232688e-01
-2.27047846e-01 1.22842288e+00 -4.32279795e-01 -5.28009474e-01
-1.20821893e-01 -3.75879318e-01 -1.06335890e+00 2.00381801e-01
1.63872570e-01 -5.86405098e-01 -3.37677747e-01 9.44284499e-01
3.07104975e-01 3.59684408e-01 4.42818910e-01 -1.23036110e+00
-5.23356915e-01 7.65944004e-01 -7.28414953e-02 -3.22749376e-01
2.15336889e-01 4.60461617e-01 1.45029819e+00 -9.24766481e-01
3.79305005e-01 1.81205451e+00 3.38145763e-01 9.62169945e-01
-1.54449141e+00 -2.52482265e-01 5.76511621e-01 -2.10258737e-01
-1.02029991e+00 -8.43836427e-01 7.66865373e-01 -5.55229783e-01
8.56783986e-01 2.01040223e-01 3.25531028e-02 1.44094539e+00
-3.35105151e-01 1.46951890e+00 1.01019251e+00 -3.16744000e-01
3.80628467e-01 5.56485713e-01 7.73068070e-01 1.10202169e+00
-7.15673923e-01 -1.57906249e-01 -9.38361228e-01 -4.26138073e-01
5.22432864e-01 4.26860079e-02 -5.37595339e-02 -3.78872395e-01
-1.11930430e+00 1.23647606e+00 7.19920248e-02 -1.37389392e-01
-2.43416816e-01 1.43364489e-01 3.26760203e-01 1.57304004e-01
6.21520400e-01 1.57967672e-01 -7.57813931e-01 -3.86751890e-01
-6.15354002e-01 1.01288855e-01 1.10085142e+00 1.28527939e+00
6.19747877e-01 -1.34559959e-01 -7.69646764e-01 1.07856953e+00
8.57901633e-01 7.28981853e-01 2.69447774e-01 -1.37608528e+00
4.33421791e-01 5.10032773e-01 3.96997660e-01 -2.68565863e-02
-3.11815977e-01 2.08539978e-01 -3.68324548e-01 -1.84191898e-01
4.83658433e-01 -5.52639425e-01 -6.75985396e-01 2.11494303e+00
3.63495827e-01 -2.40183562e-01 3.66970837e-01 7.67975032e-01
8.28728378e-01 6.74679935e-01 3.45607430e-01 -2.89138943e-01
1.36300492e+00 -1.36842227e+00 -1.10181940e+00 -2.83228040e-01
6.07617080e-01 -3.03115427e-01 1.32275522e+00 2.52061844e-01
-1.12508786e+00 -4.53083068e-01 -5.61594546e-01 -1.64377093e-01
-1.33079365e-02 1.06975392e-01 6.19809091e-01 6.08992100e-01
-1.14638102e+00 2.62655281e-02 -1.29702389e+00 -1.01414518e-02
-5.66078797e-02 2.68206567e-01 1.52714727e-02 4.50813621e-01
-1.33228528e+00 7.84032643e-01 1.64958298e-01 8.04411173e-02
-1.24188769e+00 -2.77210176e-01 -1.17758584e+00 2.80251503e-01
8.72841239e-01 -3.15450072e-01 2.15354466e+00 -1.36453331e-01
-2.31506109e+00 5.30812085e-01 -6.12146854e-01 -5.47921419e-01
4.07009959e-01 -3.63250136e-01 9.37772840e-02 -2.60871440e-01
-2.22001702e-01 9.33230519e-01 6.47598565e-01 -1.13551998e+00
-5.62636912e-01 -3.10198963e-01 2.86748171e-01 1.82343170e-01
-3.86657268e-01 -1.61163196e-01 -7.69891322e-01 -7.96561129e-03
-7.20086023e-02 -1.18055379e+00 -2.73403943e-01 -3.64450574e-01
-6.43975198e-01 -9.45168078e-01 7.90240109e-01 -6.32992625e-01
9.80609238e-01 -2.02187157e+00 3.98450315e-01 -3.06513727e-01
5.22693992e-01 9.49603170e-02 5.84693067e-02 3.39988977e-01
6.42149568e-01 -2.48138338e-01 -5.00626005e-02 -1.20816135e+00
5.86980641e-01 3.30575854e-01 -7.25510597e-01 4.78436500e-02
1.38446046e-02 1.10571277e+00 -1.00737464e+00 -7.00242639e-01
3.93686086e-01 9.44425017e-02 -7.76203275e-01 8.79548967e-01
-1.03182435e+00 5.44530392e-01 -4.27247703e-01 5.52924931e-01
1.98214844e-01 -4.46550041e-01 4.70638275e-01 3.66420858e-02
-5.80992475e-02 5.86084604e-01 -6.56146705e-01 2.08428502e+00
-5.99066079e-01 5.18913150e-01 3.69935244e-01 -6.70584619e-01
8.05673897e-01 5.85338473e-01 2.09032997e-01 -3.20853531e-01
7.72085320e-03 -3.74660313e-01 -4.64436084e-01 -3.65417302e-01
6.01807296e-01 -3.81144956e-02 -5.63863575e-01 3.42773438e-01
6.73478067e-01 -8.56544524e-02 -1.30666777e-01 6.64732039e-01
7.05915987e-01 4.09294181e-02 -8.00960436e-02 -9.26908627e-02
4.60437357e-01 -1.92471862e-01 4.77618605e-01 1.06886137e+00
-6.47661209e-01 1.74489930e-01 7.81039238e-01 1.01087391e-01
-4.06532794e-01 -1.21575189e+00 3.15030478e-02 1.98832560e+00
-3.39335785e-03 -5.16665697e-01 -7.67569304e-01 -1.06672561e+00
-6.48104819e-03 1.08562946e+00 -3.46779555e-01 -1.19879492e-01
-1.61072269e-01 -4.94033813e-01 6.07116342e-01 4.98390019e-01
4.32628274e-01 -9.17154551e-01 -2.56656915e-01 2.16207966e-01
-3.52950126e-01 -1.25592685e+00 -6.19249284e-01 3.69894236e-01
-6.03119791e-01 -6.87151730e-01 -6.59843385e-01 -4.13657397e-01
8.20419937e-02 5.60953245e-02 1.21746802e+00 -4.60515738e-01
3.38559955e-01 7.19533265e-01 -2.56708622e-01 -3.97151530e-01
-6.45510674e-01 1.01940095e-01 3.56950611e-01 -1.08267203e-01
5.58228731e-01 -1.38361856e-01 -4.41859603e-01 4.07804966e-01
-2.93183118e-01 2.67550826e-01 3.32085848e-01 1.13474238e+00
1.94113836e-01 -6.12344503e-01 6.02980614e-01 -1.10693526e+00
6.77647412e-01 -4.29010510e-01 -8.36783648e-01 4.87093359e-01
-6.32275522e-01 4.22703505e-01 2.72208065e-01 -3.22918594e-01
-1.63536382e+00 2.18762115e-01 -2.76444823e-01 -3.94543409e-01
-2.02054963e-01 4.88968700e-01 -1.61494270e-01 5.64497173e-01
4.31010485e-01 3.58642429e-01 1.74774945e-01 -6.80526316e-01
8.39206100e-01 1.11727560e+00 5.88617504e-01 -6.86832905e-01
2.92402744e-01 2.90306747e-01 -7.13514745e-01 -8.59612644e-01
-1.30123878e+00 -6.18009269e-01 -5.88779569e-01 -2.04604074e-01
1.17790401e+00 -1.15714073e+00 -1.24244356e+00 3.62465709e-01
-1.24043989e+00 -9.15502071e-01 -5.53045236e-02 3.08560610e-01
-6.09663606e-01 2.48380914e-01 -9.37097371e-01 -1.49030805e+00
-2.45302752e-01 -1.26258075e+00 1.37719762e+00 1.75066143e-01
-3.29842806e-01 -1.26579666e+00 3.61425310e-01 9.04617131e-01
3.14270765e-01 -4.50391680e-01 6.11837745e-01 -1.10185158e+00
-3.72399926e-01 -1.23500690e-01 1.65283233e-01 3.88172179e-01
7.89864212e-02 -4.62255299e-01 -1.32115853e+00 -2.74629235e-01
1.13931589e-01 -1.10470271e+00 6.69021487e-01 1.85679138e-01
8.04289460e-01 -3.31424356e-01 -1.47053689e-01 -2.44554635e-02
6.26297295e-01 4.49989736e-03 -1.79665104e-01 -4.63714033e-01
6.36342287e-01 7.34279454e-01 5.87567151e-01 4.16685581e-01
8.00060868e-01 9.41534221e-01 2.36362532e-01 1.19028158e-01
1.97808892e-01 -4.72426862e-01 7.45496094e-01 9.74532545e-01
5.56186736e-01 -3.58228356e-01 -8.74742150e-01 4.21620518e-01
-2.35161066e+00 -6.36177957e-01 -6.78543150e-02 1.85415220e+00
1.35166669e+00 2.79099286e-01 3.32853526e-01 -6.97546601e-01
4.87170339e-01 3.91607374e-01 -8.20838451e-01 1.24117486e-01
2.26283520e-01 -1.75925806e-01 2.30134472e-01 1.06534755e+00
-1.22237098e+00 1.46123934e+00 5.88114882e+00 5.85533857e-01
-7.28655517e-01 5.53831160e-01 4.40614730e-01 -1.72573730e-01
-6.19594157e-02 -3.37805711e-02 -1.20001221e+00 3.33258569e-01
1.32448304e+00 1.57177240e-01 3.30397576e-01 1.20077920e+00
2.32743606e-01 -5.20947278e-02 -1.27818394e+00 6.87185764e-01
-2.01594159e-01 -1.07221413e+00 -3.96433115e-01 -2.22110581e-02
4.84710097e-01 1.81935638e-01 1.93840668e-01 1.15897954e+00
1.27836192e+00 -6.90010428e-01 6.63162708e-01 3.75434548e-01
5.11780620e-01 -1.69010058e-01 5.56729794e-01 8.54676127e-01
-7.84851074e-01 2.09187806e-01 -2.23640412e-01 1.43022954e-01
5.44258714e-01 4.44431566e-02 -1.08640027e+00 2.52199676e-02
3.40287268e-01 5.14177322e-01 -1.52349010e-01 -2.46142671e-02
-4.51242357e-01 1.12840307e+00 -2.88128853e-01 -2.52380371e-01
4.05518323e-01 -5.60078062e-02 4.57682729e-01 1.25909019e+00
-4.54273164e-01 1.66361824e-01 6.06755972e-01 1.16218460e+00
-2.50289291e-01 -2.87990481e-01 -2.32750177e-01 -2.24507868e-01
5.87519228e-01 1.10039186e+00 -9.46320221e-02 -5.86324215e-01
-5.05443871e-01 1.22302544e+00 5.51591158e-01 3.83830667e-01
-7.79740691e-01 1.72546491e-01 6.37843609e-01 -5.60397744e-01
5.82971573e-02 -2.45774791e-01 8.85335356e-02 -1.33194649e+00
-4.03851300e-01 -6.95759177e-01 4.00951952e-01 -5.87096810e-01
-1.15418291e+00 4.60659057e-01 1.97626024e-01 -4.78545666e-01
-8.74827921e-01 -5.48397064e-01 -3.56325209e-01 1.04407907e+00
-1.21245611e+00 -1.29750097e+00 -1.31838426e-01 7.21985459e-01
1.24372470e+00 -2.28449970e-01 1.17345035e+00 -2.35100344e-01
-7.58023798e-01 7.11386800e-01 5.50721996e-02 3.19287956e-01
9.54585791e-01 -1.82074916e+00 4.58176434e-01 4.65563953e-01
2.67508149e-01 8.92314434e-01 8.21032226e-01 -6.42331123e-01
-1.45975816e+00 -9.56245899e-01 6.47634029e-01 -1.07814729e+00
7.06261396e-01 -8.91698658e-01 -9.35313284e-01 1.09348309e+00
4.55742747e-01 -2.83207119e-01 8.13337743e-01 7.37043142e-01
-3.49489450e-01 5.03211915e-01 -7.99374998e-01 5.95713496e-01
6.46161616e-01 -9.93933141e-01 -8.35478008e-01 6.84789896e-01
1.34476495e+00 -6.33410454e-01 -8.98035884e-01 1.07765505e-02
3.07830662e-01 -7.64120460e-01 8.38692546e-01 -9.50736403e-01
1.07094653e-01 1.44437611e-01 -3.13004017e-01 -1.16635752e+00
4.96649230e-03 -9.62800980e-01 -7.81776369e-01 1.32631159e+00
6.11529350e-01 -1.90353155e-01 9.89460111e-01 1.27312744e+00
-1.51483625e-01 -4.74692017e-01 -7.82147288e-01 -6.58907175e-01
1.26856580e-01 -3.69455874e-01 4.64612320e-02 8.32010329e-01
4.86563057e-01 8.46287012e-01 -8.25740457e-01 1.77988067e-01
6.72057033e-01 1.23651046e-02 1.03108764e+00 -1.02564573e+00
-6.67386234e-01 -9.09070969e-02 3.46569419e-01 -1.98472774e+00
6.36937261e-01 -5.05155563e-01 6.38034105e-01 -1.14404774e+00
2.57934332e-01 -2.52463698e-01 -1.22506410e-01 6.74237490e-01
-6.07093990e-01 -3.88576686e-01 7.17234984e-02 3.50725025e-01
-1.34475267e+00 1.09358799e+00 6.48417532e-01 -2.57057995e-01
-6.65481567e-01 3.43016982e-01 -4.89957750e-01 8.00727963e-01
5.66046834e-01 -2.84672230e-01 -6.03095829e-01 -2.05519348e-01
-2.61225879e-01 6.73771501e-01 2.06432700e-01 -3.79564106e-01
4.89799827e-01 -2.77793407e-01 -4.25453037e-01 -7.11177707e-01
9.47633862e-01 -5.50914705e-01 -4.84558046e-01 1.35768116e-01
-1.02020395e+00 -5.39438188e-01 2.89748795e-02 9.27274227e-01
-1.56793416e-01 -1.78520977e-01 5.28878570e-01 -4.35002148e-02
-7.86568344e-01 3.31434272e-02 -5.32901525e-01 1.76726654e-01
6.22145951e-01 6.63186431e-01 -3.65565747e-01 -9.38039243e-01
-1.00504625e+00 8.57206225e-01 -6.69640079e-02 4.06539917e-01
2.98929751e-01 -9.43505883e-01 -4.35073644e-01 -1.48246348e-01
3.87965053e-01 8.62131342e-02 2.69330233e-01 7.12078571e-01
2.85307348e-01 9.25511837e-01 4.26475793e-01 -8.10511947e-01
-1.23989785e+00 4.26454216e-01 2.27542609e-01 -4.13602501e-01
-1.94865882e-01 1.27892089e+00 2.96796262e-01 -9.63236988e-01
8.00382793e-01 -2.37445712e-01 -1.82611063e-01 -1.38115376e-01
2.79934198e-01 2.26533203e-03 -3.19792777e-01 -3.89356226e-01
-4.35282849e-03 -5.99613488e-01 -3.76889110e-01 -7.36236036e-01
9.95695829e-01 -4.49630648e-01 3.10057908e-01 9.82539892e-01
9.15709436e-01 -1.41711161e-01 -1.80988979e+00 -7.82484591e-01
1.97769672e-01 6.61918074e-02 2.40335628e-01 -8.98463905e-01
-5.08749187e-01 1.05037355e+00 4.70736533e-01 2.61486143e-01
3.59665662e-01 3.35411847e-01 8.47041845e-01 1.02457905e+00
5.23452759e-01 -1.04762983e+00 3.20931852e-01 8.32396448e-01
5.96101344e-01 -1.92032027e+00 -6.59354448e-01 -2.90234983e-01
-1.24493575e+00 4.92763489e-01 9.12048757e-01 4.34289187e-01
5.93607605e-01 -2.91734003e-02 2.64919221e-01 -3.79788488e-01
-1.24530709e+00 -2.66941041e-01 1.86633259e-01 2.84446508e-01
5.78080118e-01 1.18526787e-01 2.86725968e-01 1.08602130e+00
-4.52917032e-02 -2.52973706e-01 1.16223447e-01 1.10265279e+00
-5.91683447e-01 -1.17207158e+00 -1.24153100e-01 3.32134515e-02
-1.13034800e-01 -2.47891992e-01 -5.92609525e-01 3.89823586e-01
-7.35927939e-01 1.27900922e+00 -1.41591102e-01 -4.52394873e-01
2.32125521e-01 7.45639205e-01 1.90894511e-02 -9.32326853e-01
-4.20337170e-01 3.07252228e-01 1.52990997e-01 -5.68251371e-01
-8.46026242e-02 -4.64081079e-01 -1.11568177e+00 3.36471270e-03
-5.46924889e-01 5.67324638e-01 5.30306399e-01 1.06568658e+00
1.98644176e-01 5.06035268e-01 5.22464633e-01 -5.97199857e-01
-1.39160955e+00 -1.46986938e+00 -3.54359120e-01 2.68865377e-01
4.03010488e-01 -6.89912260e-01 -1.69827685e-01 3.55380446e-01]
|
[12.779973030090332, 7.853922367095947]
|
6b3b244b-e9ca-449c-9033-f48aafaad827
|
dense-fully-convolutional-network-for-skin
|
1712.10207
| null |
https://arxiv.org/abs/1712.10207v4
|
https://arxiv.org/pdf/1712.10207v4.pdf
|
Dense Pooling layers in Fully Convolutional Network for Skin Lesion Segmentation
|
One of the essential tasks in medical image analysis is segmentation and accurate detection of borders. Lesion segmentation in skin images is an essential step in the computerized detection of skin cancer. However, many of the state-of-the-art segmentation methods have deficiencies in their border detection phase. In this paper, a new class of fully convolutional network is proposed, with new dense pooling layers for segmentation of lesion regions in skin images. This network leads to highly accurate segmentation of lesions on skin lesion datasets which outperforms state-of-the-art algorithms in the skin lesion segmentation.
|
['Nader Karimi', 'Ebrahim Nasr-Esfahani', 'Mohammad H. Jafari', 'Kayvan Najarian', 'James S. Wrobel', 'Shima Rafiei', 'Shadrokh Samavi', 'S. M. Reza Soroushmehr']
|
2017-12-29
| null | null | null | null |
['skin-lesion-segmentation']
|
['medical']
|
[ 5.90450406e-01 8.44050199e-02 -2.63688803e-01 -1.44463345e-01
-4.64896262e-01 -3.20778221e-01 1.72299325e-01 2.15717584e-01
-6.86158657e-01 5.56849182e-01 -3.61446917e-01 -1.60193488e-01
1.29130349e-01 -8.33137333e-01 -5.34696169e-02 -8.03617656e-01
2.61709571e-01 1.30168684e-02 9.46066558e-01 -1.41692027e-01
2.06133306e-01 8.61891270e-01 -9.88108516e-01 3.97610009e-01
9.35236514e-01 9.81992126e-01 6.21888135e-03 8.93501341e-01
-5.54398000e-01 2.41795078e-01 -1.18623585e-01 -7.71038756e-02
1.16599165e-01 -4.91595596e-01 -1.10339665e+00 2.80559480e-01
3.16586167e-01 -2.09300473e-01 2.87807025e-02 1.30811024e+00
4.88395393e-01 -5.03389537e-01 7.58439183e-01 -6.78753376e-01
-2.22342610e-01 1.12710088e-01 -7.78773129e-01 3.36509138e-01
-7.23484680e-02 -3.48682493e-01 3.23549211e-01 -4.94488865e-01
7.57181227e-01 5.68293989e-01 8.77333522e-01 7.13079035e-01
-7.08415568e-01 -2.80327320e-01 -3.45827043e-01 1.12454653e-01
-1.36398566e+00 7.61520192e-02 4.69091058e-01 -4.22004670e-01
5.32251656e-01 5.32038033e-01 8.35168779e-01 4.74520802e-01
2.25342423e-01 8.32526088e-01 1.37459421e+00 -4.66643244e-01
2.09808320e-01 3.63514982e-02 1.66222647e-01 8.12485814e-01
3.78252506e-01 -3.29164863e-01 1.08979531e-01 4.45082299e-02
1.37554955e+00 4.68648411e-02 5.71852401e-02 3.50481756e-02
-6.33474052e-01 6.71481729e-01 5.82471311e-01 9.29516971e-01
-5.94142854e-01 4.46426030e-03 4.39992219e-01 -3.54254276e-01
4.53487158e-01 8.18731412e-02 -5.93134090e-02 1.50197014e-01
-1.26372492e+00 -1.20610982e-01 4.83369380e-01 1.78238153e-01
4.30943370e-01 -3.68249774e-01 -4.18276638e-01 6.41129792e-01
8.26665312e-02 1.44288577e-02 5.04877865e-01 -4.51045901e-01
-1.06734894e-01 1.16013789e+00 -2.74904579e-01 -3.80179822e-01
-7.16522634e-01 -2.16661558e-01 -1.09717476e+00 4.67535466e-01
7.25718081e-01 -2.32062370e-01 -1.73979390e+00 8.18784654e-01
6.65497661e-01 2.22956878e-03 -1.10825896e-01 8.78280520e-01
8.92072201e-01 1.20450146e-01 4.23783779e-01 5.52591756e-02
1.48699701e+00 -1.05644441e+00 -9.06599998e-01 -1.61383018e-01
4.34202045e-01 -9.54152107e-01 3.27105790e-01 3.12485516e-01
-1.07126343e+00 -3.77125084e-01 -8.13453615e-01 -6.77828491e-02
-5.26102245e-01 4.17636335e-01 1.07662153e+00 8.95644963e-01
-9.99879360e-01 4.74034399e-01 -9.18160200e-01 -1.06108618e+00
8.65836084e-01 4.38001364e-01 -7.20363379e-01 6.58337176e-02
-9.24973965e-01 8.97227466e-01 5.22209883e-01 2.46834710e-01
-4.78289187e-01 -4.28257376e-01 -6.15683913e-01 -3.70426476e-01
2.62365907e-01 -4.44614798e-01 1.19163096e+00 -1.13478053e+00
-1.26836252e+00 1.39478457e+00 -1.95785508e-01 -3.68565440e-01
7.35423207e-01 8.89325812e-02 -2.94978559e-01 7.83963263e-01
-1.42656803e-01 6.57270491e-01 5.55655360e-01 -1.05303180e+00
-7.83190012e-01 -3.63022923e-01 -2.50766426e-01 5.57263047e-02
1.14284353e-02 1.45582229e-01 -5.90042889e-01 -2.82538146e-01
1.99334204e-01 -6.17170811e-01 -8.54363203e-01 4.04043347e-01
-6.30891323e-01 -1.79102540e-01 9.14378405e-01 -1.11514366e+00
1.26337516e+00 -1.93644810e+00 -2.57489532e-01 4.26420540e-01
1.51552156e-01 8.85419071e-01 1.31649435e-01 3.08863044e-01
-3.02684382e-02 4.17039782e-01 -4.96541172e-01 1.94973368e-02
-4.05535221e-01 -5.28464541e-02 5.28762341e-01 5.32669008e-01
1.99275956e-01 1.01892340e+00 -4.97694731e-01 -8.77398074e-01
5.89519203e-01 6.42745316e-01 4.53186706e-02 1.02363095e-01
-1.23430602e-02 5.25048971e-01 -6.88994408e-01 9.55748558e-01
8.85529220e-01 -6.31394237e-02 -9.36855935e-03 -3.47304523e-01
-1.93360671e-01 -6.10869586e-01 -1.14437783e+00 1.74300516e+00
4.19558063e-02 3.70581448e-01 4.83507872e-01 -7.59315133e-01
4.93436784e-01 5.99919736e-01 8.93239439e-01 -2.80270100e-01
4.56023306e-01 3.04373354e-01 8.08328018e-02 -9.08198595e-01
2.59902924e-01 -2.74248898e-01 3.69898081e-01 4.13591303e-02
-7.59331360e-02 -1.01961888e-01 7.66347587e-01 -1.02534376e-01
9.26708341e-01 -1.09842122e-01 5.11453927e-01 -3.39972302e-02
7.65887082e-01 2.36287683e-01 3.32775146e-01 2.10506558e-01
-6.42797649e-01 8.03587794e-01 4.23005521e-01 -4.57678556e-01
-1.00325346e+00 -9.74628866e-01 -4.78047460e-01 3.34561616e-01
-8.58201180e-04 1.34920165e-01 -1.69615591e+00 -5.51161885e-01
-1.18941404e-01 -5.09559289e-02 -1.00780499e+00 5.13470173e-01
-4.35770452e-01 -9.47306156e-01 9.24035192e-01 5.91961265e-01
1.14430428e+00 -1.04577816e+00 -6.95381522e-01 1.93576381e-01
9.14820582e-02 -1.05822778e+00 -1.01598807e-01 -8.95658731e-02
-7.78249323e-01 -1.74220848e+00 -1.27086186e+00 -1.11449575e+00
1.27058923e+00 -4.59966734e-02 4.82692748e-01 4.78099644e-01
-1.41368079e+00 -5.03493510e-02 -2.41834804e-01 -2.46777937e-01
-4.86522615e-01 1.32400379e-01 -6.68667316e-01 2.08257977e-02
4.96704489e-01 5.44416048e-02 -7.19547331e-01 5.65784201e-02
-1.21214032e+00 -1.12429447e-01 1.00406504e+00 6.02671027e-01
8.26936185e-01 3.39379489e-01 1.62441880e-01 -1.33821201e+00
6.45977676e-01 -2.00950086e-01 -3.44902962e-01 4.56608236e-01
-1.27840787e-01 -5.01755297e-01 2.90633768e-01 -5.07165529e-02
-1.06558895e+00 4.93318111e-01 -6.50137603e-01 2.01334521e-01
-7.68778384e-01 3.42399836e-01 3.06993365e-01 -5.88408172e-01
7.84931064e-01 9.42041446e-03 1.53443560e-01 -4.88259763e-01
7.83036798e-02 7.69107640e-01 3.54558825e-01 2.23453134e-01
3.24031621e-01 6.91740990e-01 4.03901279e-01 -1.19275749e+00
-6.23773873e-01 -9.97400641e-01 -1.08854425e+00 -2.16621235e-01
1.49387443e+00 -3.23814273e-01 -2.06457123e-01 1.13486218e+00
-8.66874754e-01 -4.16812807e-01 -2.84323454e-01 1.00252815e-01
-1.86680600e-01 5.76841116e-01 -9.96967375e-01 -7.58571684e-01
-5.24024487e-01 -1.08253050e+00 8.72362196e-01 1.13303816e+00
1.45758286e-01 -1.33079994e+00 1.09440155e-01 3.84619892e-01
4.94908541e-01 6.56265259e-01 6.57253504e-01 -2.95136213e-01
-2.44887322e-01 -8.14603806e-01 -4.12110776e-01 5.09913981e-01
3.33310664e-01 5.03288686e-01 -9.59635615e-01 -5.64535074e-02
-6.11719131e-01 -2.44762879e-02 1.14151025e+00 9.64330971e-01
1.13805056e+00 3.81310076e-01 -7.58731186e-01 6.77142382e-01
1.88340771e+00 1.62188932e-01 7.87014544e-01 2.47341990e-02
4.49070483e-01 5.44355512e-01 4.77758050e-01 -4.05579247e-02
-9.97796208e-02 -1.07975319e-01 5.25084436e-01 -1.11597586e+00
-3.70898992e-01 -1.55808320e-02 -3.66522074e-01 2.42942318e-01
-2.75198430e-01 7.34050050e-02 -9.11602259e-01 9.55051541e-01
-1.52254581e+00 -6.52022719e-01 -5.77059150e-01 1.84508824e+00
7.93962181e-01 7.55356848e-02 2.45707929e-01 2.89667666e-01
1.09934890e+00 -1.57939911e-01 -4.56214696e-01 -5.10346770e-01
1.05160989e-01 7.60790884e-01 8.30121577e-01 3.75529289e-01
-1.64451432e+00 1.13194025e+00 7.28977108e+00 8.22443664e-01
-1.21003067e+00 -5.19699603e-02 6.39709473e-01 6.87563300e-01
3.38441074e-01 -2.87792772e-01 -5.98579407e-01 1.08397491e-01
2.86685377e-01 2.61232138e-01 -3.49536315e-02 5.60470104e-01
4.13372852e-02 -9.18719471e-01 -5.00867665e-01 4.34051454e-01
-4.59535532e-02 -1.30521202e+00 -2.65121967e-01 -4.44338620e-02
9.34589505e-01 -3.30322921e-01 -1.28616378e-01 -3.83513004e-01
-5.36910109e-02 -1.35264134e+00 -2.99064219e-01 6.88705325e-01
1.04479063e+00 -7.47093976e-01 1.22863293e+00 2.18899235e-01
-1.08675957e+00 1.41585782e-01 -4.01844442e-01 7.70833567e-02
2.68363774e-01 7.32898235e-01 -1.00885367e+00 4.37760741e-01
4.94877964e-01 1.81567386e-01 -8.54919195e-01 1.78281307e+00
-4.23462033e-01 6.57025278e-01 -3.25054586e-01 -1.11226015e-01
4.33246583e-01 -3.13285701e-02 2.34338902e-02 1.39696157e+00
-3.87187265e-02 4.02224660e-02 1.79891422e-01 7.22867429e-01
2.67929792e-01 3.91328871e-01 -2.27200940e-01 -2.99535215e-01
6.90601841e-02 1.84035861e+00 -1.34003925e+00 -2.03690946e-01
-5.74798658e-02 9.35553908e-01 -8.36854354e-02 2.33106270e-01
-4.06112850e-01 -6.48728192e-01 3.50353748e-01 1.67004421e-01
-5.64302094e-02 3.23792547e-02 -4.47462738e-01 -4.15554821e-01
-3.11096519e-01 -2.98807085e-01 5.39939761e-01 -2.48315722e-01
-1.22490525e+00 2.62671798e-01 -3.82126123e-01 -6.44209206e-01
1.87360287e-01 -8.33392620e-01 -1.11805868e+00 9.40363348e-01
-1.76983964e+00 -1.50258672e+00 -6.77283227e-01 5.27898729e-01
4.77501452e-01 2.20627546e-01 1.03647006e+00 1.05594069e-01
-6.70080721e-01 2.05642402e-01 -1.32027879e-01 6.26675129e-01
6.07563674e-01 -1.48687398e+00 4.15824771e-01 9.56020117e-01
-4.70769405e-01 2.78576285e-01 1.68653935e-01 -8.44919264e-01
-6.40649140e-01 -9.65441346e-01 5.11652827e-01 3.38298231e-01
4.15260583e-01 1.04849316e-01 -6.20877326e-01 2.51652211e-01
4.47666943e-01 2.53536731e-01 1.04427111e+00 -4.07279909e-01
3.40824157e-01 2.61310190e-01 -1.72725511e+00 4.65925097e-01
3.13835412e-01 -2.19553977e-01 -2.44608164e-01 5.29842317e-01
-3.20148990e-02 -6.76187217e-01 -9.24113393e-01 3.72490048e-01
5.36712527e-01 -1.00168121e+00 9.18147624e-01 -5.55647612e-01
3.32040817e-01 -2.15345085e-01 5.27899265e-01 -1.05757189e+00
-2.95732856e-01 -2.04698086e-01 4.15939301e-01 9.39566493e-01
1.36564791e-01 -2.92906135e-01 1.27840555e+00 3.68506134e-01
1.69122946e-02 -8.69252622e-01 -8.23444009e-01 -9.01099220e-02
3.29823941e-01 1.15341552e-01 2.24035718e-02 7.47603178e-01
-1.33041963e-01 -3.77846867e-01 2.74623096e-01 -1.39570028e-01
7.62033880e-01 -4.01943833e-01 2.52325088e-01 -1.06888199e+00
4.92430717e-01 -6.57717347e-01 -6.33110404e-01 -1.65176496e-01
-3.25729042e-01 -7.27396965e-01 -2.07661167e-01 -2.44369793e+00
1.43214896e-01 -1.70714721e-01 -3.08156341e-01 6.15507185e-01
-3.93075377e-01 7.29981720e-01 1.68490619e-03 -2.45402277e-01
-2.78802663e-01 -4.88338262e-01 1.74121797e+00 -5.05612828e-02
9.68028605e-03 1.74829125e-01 -2.93089896e-01 9.91811216e-01
1.19453216e+00 -5.06607816e-02 1.68024320e-02 1.45969152e-01
-2.46201023e-01 -1.86069876e-01 2.86890119e-01 -1.18631172e+00
5.19147396e-01 -2.42787465e-01 7.80733407e-01 -8.82758081e-01
2.58687101e-02 -6.75599158e-01 -1.08785681e-01 7.34332204e-01
-1.46301433e-01 -4.87083554e-01 2.59424686e-01 7.70113170e-02
-3.41752201e-01 -5.07973373e-01 1.24188137e+00 -8.52288246e-01
-9.13702667e-01 2.72914976e-01 -5.90591967e-01 -1.99292839e-01
1.68031538e+00 -5.97209811e-01 -3.07768375e-01 3.37767154e-01
-9.42768872e-01 5.73556125e-02 3.71430427e-01 -4.97313626e-02
4.73173976e-01 -9.02196050e-01 -7.62015164e-01 2.21368819e-01
-2.16504380e-01 3.22691441e-01 7.28891015e-01 1.17460954e+00
-1.40660298e+00 4.38971549e-01 -6.45547211e-01 -4.54923958e-01
-1.49977839e+00 1.18864879e-01 7.68202186e-01 -4.85041946e-01
-2.48856857e-01 1.14404452e+00 -2.64773786e-01 -9.46580321e-02
1.41008481e-01 -3.65111500e-01 -5.83158195e-01 -2.09947042e-02
5.41269004e-01 5.32143712e-01 5.78938387e-02 -5.76252043e-01
-2.81414688e-01 7.96613097e-01 -2.01500773e-01 2.80747443e-01
8.23493421e-01 1.14057980e-01 -4.98382807e-01 9.72449780e-02
7.72424519e-01 -1.06375478e-01 -1.07186079e+00 1.92730963e-01
-1.03869513e-01 -4.15641367e-01 3.17684084e-01 -1.30999720e+00
-1.16975391e+00 1.13195682e+00 1.02033460e+00 3.48298311e-01
1.18189323e+00 -4.51438218e-01 1.06233454e+00 -1.60763636e-01
2.22879931e-01 -1.60928178e+00 -2.47529536e-01 -3.06579415e-02
3.19156885e-01 -1.33162820e+00 1.49018571e-01 -1.03779352e+00
-5.24788976e-01 1.45095003e+00 6.64321899e-01 -3.43043089e-01
7.33879566e-01 5.37903368e-01 5.37216425e-01 -2.04935655e-01
2.06170976e-02 -7.48283744e-01 3.22293550e-01 7.17011094e-01
6.60953999e-01 2.55402684e-01 -9.17912602e-01 4.31060761e-01
5.57693362e-01 1.98832333e-01 3.45591992e-01 1.06232274e+00
-6.04827225e-01 -1.27382922e+00 -3.84798884e-01 5.38458049e-01
-1.07116854e+00 1.37638032e-01 -9.42482591e-01 1.10009992e+00
4.83652920e-01 7.60281265e-01 -1.61358610e-01 1.53120205e-01
8.22446570e-02 2.46698856e-02 5.66020429e-01 -5.14061570e-01
-1.13588452e+00 1.63226649e-01 4.72805416e-03 -4.52463567e-01
-4.54054385e-01 -4.19935286e-01 -1.69759047e+00 1.29590347e-01
-4.22816336e-01 -1.15128934e-01 1.14278066e+00 8.09080184e-01
-2.05125496e-01 6.29798055e-01 2.16615856e-01 -5.65064788e-01
-1.98161155e-01 -9.54767048e-01 -1.06914937e+00 4.40057635e-01
1.19660720e-01 -2.74410605e-01 1.38867781e-01 9.56839845e-02]
|
[15.582956314086914, -3.0372767448425293]
|
94f28d4b-1b82-4d82-87a4-d7a400c8c509
|
efficient-ensembles-of-graph-neural-networks
| null | null |
https://openreview.net/forum?id=lTiW8Jet8t
|
https://openreview.net/pdf?id=lTiW8Jet8t
|
Efficient Ensembles of Graph Neural Networks
|
Graph Neural Networks (GNNs) have enabled the power of deep learning to be applied to inputs beyond the Euclidean domain, with applications ranging from social networks and product recommendation engines to the life sciences. GNNs, like other classes of machine learning models, benefit from ensemble learning, wherein multiple models are combined to provide higher accuracy and robustness than single models. However, ensembles suffer from significantly higher inference processing and storage requirements, limiting their use in practical applications. In this work, we leverage the unique characteristics of GNNs to overcome these overheads, creating efficient ensemble GNNs that are faster than even single models at inference time. We observe that during message passing, nodes that are incorrectly classified (error nodes) also end up adversely affecting the representations of other nodes in their neighborhood. This error propagation also makes GNNs more difficult to approximate (e.g., through pruning) for efficient inference. We propose a technique to create ensembles of diverse models, and further propose Error Node Isolation (ENI), which prevents error nodes from sending messages to (and thereby influencing) other nodes. In addition to improving accuracy, ENI also leads to a significant reduction in the memory footprint and the number of arithmetic operations required to evaluate the computational graphs of all neighbors of error nodes. Remarkably, these savings outweigh even the overheads of using multiple models in the ensemble. A second key benefit of ENI is that it enhances the resilience of GNNs to approximations. Consequently, we propose Edge Pruning and Network Pruning techniques that target both the input graph and the neural networks used to process the graph. Our experiments on GNNs for transductive and inductive node classification demonstrate that ensembles with ENI are simultaneously more accurate (by up to 4.6% and 3.8%) and faster (by up to 2.8$\times$ and 5.7$\times$) when compared to the best-performing single models and ensembles without ENI, respectively. In addition, GNN ensembles with ENI are consistently more accurate than single models and ensembles without ENI when subject to pruning, leading to additional speedups of up to 5$\times$ with no loss in accuracy.
|
['Anand Raghunathan', 'Jacob R. Stevens', 'Amrit Nagarajan']
|
2021-09-29
| null | null | null | null |
['product-recommendation']
|
['miscellaneous']
|
[ 2.96166658e-01 1.11504301e-01 -6.06223382e-02 -1.67899951e-01
-6.17274940e-02 -4.08985734e-01 2.08656132e-01 4.80723202e-01
-2.56902426e-01 7.01243222e-01 -3.29262257e-01 -6.22615695e-01
-2.93238878e-01 -1.50425386e+00 -9.74401891e-01 -4.95090783e-01
-4.07078236e-01 1.86971247e-01 -2.07488965e-02 -2.11863011e-01
-4.01149206e-02 5.43471932e-01 -1.49994862e+00 7.40550905e-02
1.09492624e+00 1.09932566e+00 -4.13545340e-01 6.96003437e-01
-1.18977681e-01 7.70496666e-01 -6.32948518e-01 -5.20177305e-01
1.74361765e-01 -2.22301632e-01 -4.31340486e-01 -4.94938344e-01
4.23687607e-01 -3.59200895e-01 -6.09169185e-01 1.02558887e+00
3.98105621e-01 2.30603427e-01 4.82582211e-01 -1.35852385e+00
-3.82855624e-01 8.50648284e-01 -5.39770603e-01 -1.22684956e-01
-4.85563390e-02 -7.60617256e-02 1.13021266e+00 -6.35907829e-01
2.64621645e-01 1.08675539e+00 1.13164127e+00 4.39294666e-01
-1.59887469e+00 -1.09145701e+00 2.47385278e-01 2.16504764e-02
-1.59413099e+00 -4.89324927e-01 6.71402216e-01 1.51484579e-01
1.17487538e+00 4.09771711e-01 6.18362963e-01 8.94277811e-01
2.38002330e-01 5.89090526e-01 6.45121276e-01 -1.50332317e-01
3.65569830e-01 -6.79912418e-02 3.12554270e-01 8.71612787e-01
8.32791507e-01 3.05734016e-02 -3.71989906e-01 -1.72752559e-01
6.50766969e-01 2.89865971e-01 -8.57311860e-02 3.36780250e-02
-5.73249340e-01 7.34419107e-01 9.75530505e-01 2.51045525e-02
-4.96452421e-01 6.98421061e-01 3.65816623e-01 2.31275111e-01
5.04885793e-01 4.18000966e-01 -3.17443073e-01 1.82687476e-01
-8.20130944e-01 2.57263277e-02 1.08562493e+00 8.00561130e-01
9.03157473e-01 2.78781205e-01 1.08725876e-01 7.41926193e-01
1.13878168e-01 6.49245501e-01 -1.88623562e-01 -7.02092767e-01
3.51620764e-01 1.16189528e+00 -4.48057681e-01 -1.60029876e+00
-6.44620597e-01 -9.71287906e-01 -1.50063479e+00 1.68931901e-01
2.33897895e-01 -2.30534539e-01 -9.36793506e-01 1.88765204e+00
2.43264943e-01 4.28214103e-01 -1.60029065e-02 4.03023452e-01
7.47392654e-01 6.70226216e-01 1.53043523e-01 5.65866381e-02
9.79932964e-01 -7.68242717e-01 -2.26331308e-01 -2.37735927e-01
1.03755093e+00 -3.75445396e-01 5.25005996e-01 3.38553280e-01
-9.69036758e-01 -3.50840837e-01 -1.30612767e+00 1.24886543e-01
-4.66101885e-01 -2.05919355e-01 8.99397671e-01 6.33777559e-01
-1.18845248e+00 9.09022987e-01 -8.52495134e-01 -3.85273248e-02
7.41671264e-01 8.62481833e-01 -1.71065167e-01 -2.23804459e-01
-1.10992944e+00 7.39409566e-01 2.77509630e-01 2.51438618e-01
-4.46001619e-01 -1.00145578e+00 -9.32050943e-01 4.19170380e-01
3.77983570e-01 -6.96648359e-01 7.69876242e-01 -1.04996109e+00
-1.05176115e+00 -9.85843316e-02 -2.78458267e-01 -6.69295132e-01
1.35280550e-01 1.19501740e-01 -5.90318203e-01 -6.90093935e-02
-3.65816385e-01 5.22769272e-01 4.61353749e-01 -1.08885705e+00
-4.75073636e-01 -4.08756077e-01 1.71894491e-01 -2.50442810e-02
-7.34732151e-01 -5.66064119e-01 -2.33295217e-01 -5.13402045e-01
2.30455935e-01 -9.49884295e-01 -3.26956451e-01 -1.56736858e-02
-4.49575394e-01 -1.86876714e-01 9.14878428e-01 -4.36765075e-01
1.50291228e+00 -1.88183236e+00 -2.27348283e-02 9.29503262e-01
1.01000607e+00 4.91484135e-01 -1.51793212e-01 2.36646667e-01
2.09274158e-01 3.98458987e-01 -1.38114184e-01 -2.62956589e-01
-2.26350605e-01 5.24282396e-01 1.42717864e-02 2.29685277e-01
1.49266466e-01 1.08246994e+00 -8.85739267e-01 -1.81899816e-01
1.08412132e-01 6.51830554e-01 -6.18879020e-01 -3.90182436e-01
-1.81642368e-01 -1.70243420e-02 -2.09206522e-01 5.36043346e-01
7.07482517e-01 -6.28341019e-01 4.22768265e-01 -2.60882288e-01
4.37719852e-01 4.21849698e-01 -1.27263439e+00 8.86956871e-01
-7.58053124e-01 5.17376602e-01 1.25558734e-01 -1.13176954e+00
9.11898613e-01 5.78133799e-02 3.29666197e-01 -7.76555657e-01
1.36822313e-01 3.59859794e-01 2.97799408e-01 2.51052409e-01
4.75137085e-01 2.54911453e-01 5.00333868e-02 6.91874504e-01
-4.56277877e-02 4.04070675e-01 3.45156312e-01 4.14947063e-01
1.37817323e+00 -4.75568861e-01 -6.59632832e-02 -2.03499317e-01
3.14212203e-01 -3.35009813e-01 6.07100427e-01 9.62984860e-01
1.95261866e-01 -9.34575424e-02 4.42858934e-01 -5.55023909e-01
-9.12367284e-01 -9.17881727e-01 5.94464093e-02 1.03589988e+00
3.08251768e-01 -6.22283220e-01 -6.51048064e-01 -7.29263186e-01
3.72328162e-01 7.74124384e-01 -3.39014471e-01 -5.55534363e-01
-6.36012614e-01 -7.78409541e-01 9.92102444e-01 9.71762598e-01
6.71616137e-01 -5.72675049e-01 -1.28636748e-01 3.42716873e-01
6.76567405e-02 -9.88903642e-01 -2.91629024e-02 1.73987910e-01
-1.21955860e+00 -1.01436710e+00 -8.80194679e-02 -4.59260881e-01
9.82471764e-01 1.20801322e-01 1.18595123e+00 8.32400382e-01
-1.22265846e-01 4.67374958e-02 -2.10276581e-02 -4.33428556e-01
-5.36936045e-01 3.20502937e-01 1.80521443e-01 -5.96370623e-02
3.46912861e-01 -9.30325568e-01 -5.95605373e-01 2.98070908e-01
-7.59594083e-01 1.61281914e-01 6.07009709e-01 9.52772617e-01
3.55429322e-01 3.60022247e-01 6.63674235e-01 -1.00805521e+00
4.08812910e-01 -5.42047262e-01 -6.44494772e-01 2.80386835e-01
-1.07131600e+00 2.76536435e-01 9.61964130e-01 -3.55802625e-01
-5.61493278e-01 -3.01045120e-01 -6.62040338e-02 -2.34371126e-01
3.53591949e-01 5.42109907e-01 2.66692728e-01 -5.46089947e-01
7.45394409e-01 -1.21395558e-01 1.39826730e-01 -1.78911299e-01
5.89424856e-02 3.92793149e-01 1.25438094e-01 -4.09983188e-01
6.39931321e-01 2.52347946e-01 5.45179427e-01 -8.37939799e-01
-4.59459245e-01 9.46083963e-02 -7.78612196e-02 -2.71820694e-01
2.38732263e-01 -7.71395564e-01 -1.15977180e+00 4.05010194e-01
-1.02695799e+00 -1.96483657e-01 6.50248155e-02 2.84638315e-01
3.10087949e-01 1.19505472e-01 -8.29521477e-01 -8.97224784e-01
-7.41973996e-01 -9.66827691e-01 4.70571309e-01 3.36311042e-01
-4.14648622e-01 -1.23814213e+00 -6.51356578e-01 -4.41551283e-02
7.02130854e-01 3.29152554e-01 1.17860401e+00 -7.30547249e-01
-7.32934535e-01 -3.28075558e-01 -4.52444881e-01 4.85648811e-01
-9.60829854e-02 -1.40534490e-02 -8.47742260e-01 -4.38572019e-01
-5.49940646e-01 -1.14790369e-02 1.02234972e+00 2.09893093e-01
1.33156908e+00 -4.46946651e-01 -7.32416987e-01 7.01302469e-01
1.43512678e+00 5.99686652e-02 5.26993871e-01 -2.72712976e-01
8.95925045e-01 1.40205741e-01 -1.10854335e-01 1.05659537e-01
2.59414434e-01 3.30286026e-01 4.80457306e-01 -2.83140004e-01
-1.38599887e-01 -3.84984672e-01 1.97444886e-01 9.23540831e-01
-2.15845734e-01 -5.89932740e-01 -9.02101696e-01 3.42651099e-01
-1.78360486e+00 -7.19160974e-01 -2.42391318e-01 2.27416635e+00
6.45382822e-01 2.91584820e-01 -2.57049352e-01 4.31172073e-01
6.63114548e-01 -9.18067321e-02 -8.15513372e-01 -5.66514015e-01
8.19193199e-02 7.33032763e-01 7.73640454e-01 4.53432053e-01
-8.52264464e-01 5.88531852e-01 5.66850805e+00 8.18865240e-01
-1.08959472e+00 -1.92657128e-01 9.11994517e-01 -2.57128656e-01
-4.77719992e-01 -1.17596231e-01 -9.22776818e-01 3.46363753e-01
1.20231199e+00 -5.63593842e-02 7.61589825e-01 7.96459019e-01
-2.51166970e-01 2.41140984e-02 -1.24315786e+00 8.30894470e-01
-1.71396285e-01 -1.58077204e+00 1.12309381e-01 1.96639225e-01
9.09337997e-01 1.37004480e-01 -1.90651268e-01 4.95932043e-01
6.66592658e-01 -1.25021958e+00 2.43006378e-01 2.20420867e-01
7.66524374e-01 -1.15554869e+00 8.34544420e-01 2.71104693e-01
-1.28003228e+00 -1.93612084e-01 -3.11309397e-01 -3.51131618e-01
-1.11105144e-01 1.06302619e+00 -7.76277125e-01 7.02925265e-01
6.91648424e-01 5.13014615e-01 -4.27096635e-01 6.81280255e-01
-8.53309706e-02 7.51124203e-01 -8.74914885e-01 -5.39699614e-01
1.62478268e-01 -8.44438002e-02 5.25484622e-01 1.04401708e+00
3.76021236e-01 7.76162818e-02 -3.19593661e-02 1.06160629e+00
-6.69412315e-01 -1.75495744e-01 -6.29528821e-01 -4.43802997e-02
8.42342019e-01 1.20921516e+00 -6.21139705e-01 -4.11366135e-01
-2.68436968e-01 6.91193819e-01 5.90490699e-01 5.29666841e-01
-8.57779443e-01 -7.59672046e-01 8.27043116e-01 6.14517815e-02
2.15146303e-01 -9.64741707e-02 -7.15065956e-01 -7.27949679e-01
1.28409356e-01 -7.65865564e-01 2.49890462e-01 -1.75483420e-01
-1.21341860e+00 4.61427331e-01 -3.58955085e-01 -7.67659009e-01
1.75836030e-02 -6.84354782e-01 -5.23216367e-01 8.26335013e-01
-1.18915880e+00 -8.79025459e-01 -4.75132376e-01 2.33095393e-01
-2.36339927e-01 2.53065228e-01 8.64054501e-01 5.69849253e-01
-7.82914460e-01 1.16148818e+00 1.73438936e-01 3.10704291e-01
2.30020627e-01 -8.86201143e-01 5.63897789e-01 8.25896859e-01
2.03218907e-01 8.94764781e-01 2.47827172e-01 -6.72559559e-01
-1.55330408e+00 -1.36198425e+00 8.46980095e-01 -1.44869968e-01
3.25447142e-01 -4.79962796e-01 -9.34900463e-01 6.05734169e-01
-2.93428391e-01 1.60610959e-01 5.84248483e-01 5.73540926e-01
-5.70782602e-01 -5.35759628e-01 -1.18406260e+00 8.95963967e-01
1.35160613e+00 -3.06918293e-01 2.37085879e-01 3.45834903e-02
5.90591490e-01 -3.86233330e-01 -1.06155896e+00 7.91053176e-01
6.62950933e-01 -8.57613266e-01 1.03025293e+00 -5.08877099e-01
4.54989582e-01 -1.93848953e-01 -5.09351026e-03 -1.22109056e+00
-3.26672077e-01 -4.56878453e-01 -6.81599438e-01 1.05877578e+00
7.82115817e-01 -1.16404235e+00 8.17506850e-01 6.44905567e-01
2.72801220e-02 -1.07378411e+00 -6.90400243e-01 -6.66421354e-01
-1.37186721e-01 -6.64599776e-01 8.52483630e-01 7.71758497e-01
-2.21508726e-01 3.96992475e-01 -8.86941850e-02 2.36464933e-01
6.96173668e-01 -1.16695672e-01 6.56791151e-01 -1.43461883e+00
-3.72373253e-01 -6.29363954e-01 -5.93904912e-01 -1.09852791e+00
3.43962871e-02 -1.22198558e+00 -2.29439497e-01 -1.49923933e+00
-1.16725549e-01 -1.10294008e+00 -4.86824811e-01 7.04235017e-01
-2.03391820e-01 5.43685377e-01 2.52840086e-03 -1.56143591e-01
-4.54029500e-01 1.78989783e-01 9.13914144e-01 -9.90641341e-02
-2.33093441e-01 7.92533066e-03 -8.35730493e-01 7.34001040e-01
9.22148824e-01 -4.01325643e-01 -4.41244930e-01 -5.56531012e-01
5.44316888e-01 -3.07022959e-01 5.21333933e-01 -1.10161293e+00
3.72026950e-01 3.66502404e-01 7.61236906e-01 -4.26070601e-01
2.55379766e-01 -7.14085460e-01 2.66122878e-01 6.47254527e-01
-1.29346430e-01 1.06227864e-03 4.83531326e-01 7.35281587e-01
9.53015015e-02 5.86233772e-02 5.62985361e-01 2.43992850e-01
-3.91123056e-01 3.33499014e-01 -1.80501297e-01 -1.40484795e-01
9.88648713e-01 -2.32246771e-01 -4.64029133e-01 -4.76093292e-01
-3.97773832e-01 4.50623244e-01 2.55325735e-01 1.02962680e-01
5.28944075e-01 -1.22701406e+00 -4.62855369e-01 4.82982844e-01
-1.35402799e-01 2.59548485e-01 2.77468264e-01 9.54158843e-01
-5.72739780e-01 2.14135766e-01 1.59065351e-01 -5.77373862e-01
-1.23236406e+00 2.63071984e-01 3.81784409e-01 -4.33548599e-01
-4.35603023e-01 9.45601165e-01 -2.02435791e-01 -5.99762678e-01
3.67212474e-01 -3.63186508e-01 3.30030650e-01 -4.44494963e-01
4.15024877e-01 9.25419807e-01 3.28223318e-01 -9.76108611e-02
-3.09878945e-01 1.60907775e-01 -2.34996110e-01 5.43143213e-01
1.26580787e+00 2.58913815e-01 -3.99908602e-01 -1.08724218e-02
1.15783465e+00 -1.01568609e-01 -7.36618698e-01 -3.35474670e-01
-2.08498240e-01 2.61067934e-02 3.11177105e-01 -8.26850653e-01
-1.31694603e+00 7.63828695e-01 1.51056781e-01 3.79108667e-01
1.25646472e+00 -3.19705278e-01 1.06681406e+00 5.10647535e-01
4.15156722e-01 -9.95665669e-01 -3.24972183e-01 4.93267059e-01
3.30303341e-01 -9.06256199e-01 1.53295353e-01 -5.90177357e-01
6.93848636e-03 1.11707175e+00 6.92663848e-01 -1.79840356e-01
6.05600595e-01 5.31487286e-01 -4.32957917e-01 -1.38225600e-01
-7.98480153e-01 1.93909571e-01 2.35437959e-01 4.33635652e-01
2.14607760e-01 2.29199156e-01 -7.02032000e-02 5.58569133e-01
-3.03914566e-02 -9.70363468e-02 1.43582597e-01 6.99748516e-01
-2.75169879e-01 -1.05485928e+00 -7.73526123e-03 1.13174987e+00
-1.50350586e-01 -3.38345528e-01 -3.06379735e-01 7.17312694e-01
5.72833940e-02 1.12284279e+00 3.79912972e-01 -8.55525970e-01
2.10902095e-01 -3.50538380e-02 3.84755164e-01 -2.14568391e-01
-7.08025992e-01 -5.16797841e-01 3.81648481e-01 -6.58225060e-01
1.28311859e-02 -2.18090340e-02 -1.48653328e+00 -1.23364341e+00
-5.57884037e-01 -9.26230941e-03 6.08191311e-01 7.27372289e-01
8.57818663e-01 7.75802732e-01 4.28549051e-01 -6.00925446e-01
-6.54434800e-01 -7.23184049e-01 -3.43755513e-01 8.23145732e-02
1.18650965e-01 -5.48507035e-01 -4.56221312e-01 -6.40059471e-01]
|
[6.981621265411377, 6.0033135414123535]
|
5cb3819e-a4d1-44b9-9a35-a132218cbedc
|
entity-and-evidence-guided-relation
|
2008.12283
| null |
https://arxiv.org/abs/2008.12283v1
|
https://arxiv.org/pdf/2008.12283v1.pdf
|
Entity and Evidence Guided Relation Extraction for DocRED
|
Document-level relation extraction is a challenging task which requires reasoning over multiple sentences in order to predict relations in a document. In this paper, we pro-pose a joint training frameworkE2GRE(Entity and Evidence Guided Relation Extraction)for this task. First, we introduce entity-guided sequences as inputs to a pre-trained language model (e.g. BERT, RoBERTa). These entity-guided sequences help a pre-trained language model (LM) to focus on areas of the document related to the entity. Secondly, we guide the fine-tuning of the pre-trained language model by using its internal attention probabilities as additional features for evidence prediction.Our new approach encourages the pre-trained language model to focus on the entities and supporting/evidence sentences. We evaluate our E2GRE approach on DocRED, a recently released large-scale dataset for relation extraction. Our approach is able to achieve state-of-the-art results on the public leaderboard across all metrics, showing that our E2GRE is both effective and synergistic on relation extraction and evidence prediction.
|
['Guangtao Wang', 'Tengyu Ma', 'Jing Huang', 'Kevin Huang']
|
2020-08-27
| null | null | null | null |
['document-level-relation-extraction']
|
['natural-language-processing']
|
[ 1.40874296e-01 9.48631167e-01 -4.57148999e-01 -2.89935887e-01
-9.52315688e-01 -5.80689490e-01 9.22105849e-01 8.26602995e-01
-4.23011720e-01 9.69142199e-01 4.02620137e-01 -5.67898691e-01
-2.85990953e-01 -8.61662090e-01 -9.88774538e-01 1.53955771e-02
-3.31105530e-01 8.20950449e-01 5.38173854e-01 -2.18965188e-01
1.31115586e-01 5.85747182e-01 -1.08138621e+00 7.28062868e-01
6.87899947e-01 8.35472584e-01 5.90896569e-02 8.87289703e-01
-2.29599312e-01 1.47812450e+00 -3.29322755e-01 -7.29879439e-01
-1.64376855e-01 1.07506141e-02 -1.50698018e+00 -1.00503676e-01
1.09678745e-01 -2.43825559e-02 -1.86634406e-01 6.52046800e-01
9.57263485e-02 -9.76927727e-02 7.24739373e-01 -8.59075248e-01
-2.50375122e-01 1.43036580e+00 -4.48575199e-01 3.68209362e-01
6.05800569e-01 -1.88951209e-01 1.63727248e+00 -1.18986940e+00
1.13559616e+00 1.07982671e+00 4.08330679e-01 1.61532149e-01
-1.11318958e+00 -2.87987798e-01 5.13993084e-01 4.21960205e-01
-1.21169162e+00 -5.08378923e-01 5.88714659e-01 -4.71086562e-01
1.88392389e+00 3.57846916e-01 3.06517780e-01 7.65947163e-01
1.65944338e-01 9.58878517e-01 7.29924440e-01 -7.87265360e-01
-5.14296517e-02 2.79346943e-01 4.87099111e-01 8.01921070e-01
1.97804675e-01 -2.28559881e-01 -7.81633258e-01 -1.49141103e-01
5.21053262e-02 -5.42372406e-01 -1.77633166e-02 7.23765492e-02
-1.05044615e+00 4.52671528e-01 2.88010210e-01 3.46513778e-01
-8.69458854e-01 -1.63445622e-02 4.82330948e-01 5.92583045e-02
5.70177257e-01 7.37226605e-01 -1.00000429e+00 -4.70101237e-02
-8.03604066e-01 2.83160448e-01 1.17025471e+00 9.95693743e-01
4.32477266e-01 -9.65017498e-01 -6.75255239e-01 7.59769559e-01
5.69168150e-01 -1.25422046e-01 -3.90165076e-02 -4.63430226e-01
9.78802681e-01 9.23717618e-01 1.44795135e-01 -7.92029202e-01
-4.78078455e-01 -6.17496133e-01 -4.66179878e-01 -1.64604470e-01
1.40826344e-01 -2.73137391e-01 -6.17677033e-01 1.45654953e+00
4.93079722e-01 8.41340348e-02 3.93974304e-01 4.34416324e-01
1.00437510e+00 7.45981336e-01 3.23563099e-01 -2.25837395e-01
1.60621822e+00 -1.24601007e+00 -7.86492944e-01 -6.67050302e-01
1.02778220e+00 -6.81556225e-01 5.35682261e-01 3.57596904e-01
-1.17113328e+00 -2.85438240e-01 -1.11229098e+00 -1.86018124e-01
-6.31517291e-01 4.06035304e-01 5.89642823e-01 -1.20853595e-01
-5.79900861e-01 5.77480137e-01 -6.66822970e-01 -3.25458795e-01
4.95014578e-01 3.63561451e-01 -5.62589943e-01 -1.22349858e-01
-1.45121109e+00 1.30315685e+00 8.87962759e-01 1.90224946e-01
-4.95332122e-01 -4.30694163e-01 -9.38753784e-01 2.97181010e-01
9.63410258e-01 -6.44485354e-01 1.35141218e+00 -2.91909538e-02
-1.15948248e+00 1.02696669e+00 -2.72964746e-01 -7.92141974e-01
1.22437999e-01 -7.36435533e-01 -3.67001772e-01 -1.03546688e-02
1.21257797e-01 4.34784651e-01 2.83418447e-01 -1.16161335e+00
-8.24358761e-01 -2.62813624e-02 2.62975574e-01 2.69643255e-02
2.18001325e-02 5.05659699e-01 -8.11202526e-01 -5.04709542e-01
-1.58961177e-01 -6.05711520e-01 -1.43558457e-01 -7.24950433e-01
-1.07349336e+00 -7.78944612e-01 4.40956891e-01 -8.23420584e-01
1.70256186e+00 -1.56482124e+00 2.99570143e-01 3.95606637e-01
3.17430794e-01 3.32701951e-01 -1.86472505e-01 7.30774701e-01
-1.89801753e-01 1.90452918e-01 -1.18524820e-01 -4.76114839e-01
-1.89138260e-02 2.69429415e-01 -3.74710470e-01 -1.66547388e-01
9.53172326e-01 1.08601296e+00 -1.04425657e+00 -1.00036395e+00
-1.77432567e-01 1.53944254e-01 -5.20171106e-01 4.51925993e-01
-6.73578441e-01 7.14482293e-02 -5.51920116e-01 4.09569174e-01
2.57982701e-01 -5.68104029e-01 5.22493243e-01 -2.06309959e-01
2.31238385e-03 1.22031522e+00 -1.04749918e+00 1.35645366e+00
-6.07005477e-01 4.84198809e-01 -4.04938847e-01 -7.10921288e-01
7.50721097e-01 3.98666978e-01 1.85936779e-01 -9.19533297e-02
-2.77052484e-02 3.13586563e-01 2.31667429e-01 -7.42495954e-01
4.40266550e-01 2.60416180e-01 2.08383426e-02 4.29046363e-01
3.49099934e-01 1.28284395e-01 7.03943908e-01 7.49468029e-01
1.49767125e+00 3.57701957e-01 7.23470807e-01 6.45518005e-02
8.73790264e-01 1.74582284e-02 3.84860218e-01 6.11844480e-01
5.81176460e-01 7.39322156e-02 8.33567142e-01 -2.25364789e-01
-6.82634532e-01 -3.90919209e-01 -9.74012073e-03 1.06373751e+00
-2.74724305e-01 -1.15887654e+00 -4.39469099e-01 -1.36785436e+00
-1.18258476e-01 9.33688998e-01 -7.26928830e-01 4.48322929e-02
-7.53881156e-01 -4.61044461e-01 3.62173587e-01 7.26039708e-01
1.10757828e-01 -1.23762333e+00 -3.33950669e-01 6.92900658e-01
-3.56177419e-01 -1.55099845e+00 8.62516239e-02 7.54525423e-01
-4.10682172e-01 -1.08719552e+00 -4.45597386e-03 -5.68367362e-01
4.88676518e-01 -5.40043950e-01 1.63101995e+00 3.03429663e-01
1.56079888e-01 -4.69844230e-02 -4.91117567e-01 -4.85914826e-01
-4.69107807e-01 5.31329274e-01 -4.68740165e-01 -1.93931401e-01
6.36001527e-01 -2.98103362e-01 -6.99255615e-02 2.62586121e-02
-4.62384313e-01 2.70897865e-01 8.08626533e-01 7.17927814e-01
7.12063253e-01 -4.68328074e-02 3.94832194e-01 -1.36434603e+00
7.17920780e-01 -5.83013892e-01 -4.93361980e-01 7.10559070e-01
-7.17446804e-01 4.81068075e-01 3.95587087e-01 -9.89330634e-02
-8.92705321e-01 -5.19630723e-02 -2.08881617e-01 2.59223521e-01
-5.88089712e-02 1.20136118e+00 -3.00701916e-01 6.02472603e-01
4.40135360e-01 -2.57095069e-01 -8.04639220e-01 -4.36304599e-01
5.41066706e-01 6.52045667e-01 6.24960303e-01 -8.17618489e-01
7.49689043e-01 -6.51954934e-02 1.80497259e-01 -3.25731516e-01
-1.40895903e+00 -4.90747541e-01 -9.88650918e-01 6.33393005e-02
7.74533629e-01 -7.33085752e-01 -5.82494378e-01 -1.45167053e-01
-1.69057035e+00 -4.91185129e-01 -3.23516488e-01 2.15751126e-01
-1.87504917e-01 -1.98118086e-03 -7.53482997e-01 -9.06193554e-01
-5.39391875e-01 -7.69914091e-01 1.12893963e+00 -3.24958339e-02
-5.77700377e-01 -9.65932131e-01 3.08283508e-01 3.35833013e-01
-2.23820642e-01 5.95092289e-02 9.12937641e-01 -1.25328791e+00
-7.25463152e-01 -3.58777076e-01 -4.02921528e-01 -2.30607344e-03
-1.08132865e-02 1.88819781e-01 -8.90174985e-01 3.03283423e-01
-6.10630989e-01 -6.09196723e-01 8.53526592e-01 -7.50940666e-02
1.02298629e+00 -4.66932565e-01 -8.64496529e-01 9.33838338e-02
8.50070536e-01 -1.02965415e-01 4.73902345e-01 5.80814123e-01
5.33682704e-01 8.98482859e-01 1.14203215e+00 1.63828522e-01
7.77013063e-01 8.46922398e-01 2.75403708e-01 -5.84430471e-02
-1.62250876e-01 -4.86261755e-01 1.72069550e-01 5.02410531e-01
-2.27001294e-01 -5.98232448e-01 -1.08193111e+00 5.65959394e-01
-2.10539103e+00 -8.39735091e-01 -3.04229736e-01 1.57324147e+00
1.30577838e+00 6.70421124e-01 -1.42667457e-01 2.67984867e-01
2.08884895e-01 -2.41592410e-03 -1.99215546e-01 -3.85995209e-01
-1.17927097e-01 3.32533121e-01 1.48211271e-01 8.10311079e-01
-1.31630194e+00 1.35597754e+00 5.26845264e+00 9.11607504e-01
-6.99097514e-01 -6.56233057e-02 6.10483646e-01 2.28675991e-01
-2.46028855e-01 2.80071199e-01 -1.34483576e+00 3.97967407e-03
1.11080444e+00 -1.31720370e-02 2.07277164e-02 5.92197895e-01
-7.72211775e-02 -1.66487798e-01 -1.51692295e+00 2.97822326e-01
-9.34989229e-02 -1.55303478e+00 -1.08998843e-01 2.48324834e-02
3.81747216e-01 1.53292745e-01 -5.62706530e-01 4.61648226e-01
6.87521756e-01 -1.09606814e+00 6.08518839e-01 5.52117825e-01
5.56398988e-01 -4.90805954e-01 1.06202173e+00 5.21369874e-01
-1.22743344e+00 1.00032151e-01 -8.00670162e-02 9.98993963e-02
4.59226042e-01 8.33334029e-01 -1.41413724e+00 9.41694081e-01
3.45254213e-01 8.11939597e-01 -6.71504796e-01 6.26052916e-01
-1.06224406e+00 8.38522971e-01 -4.53114867e-01 -1.05576515e-01
1.66626051e-01 2.31733829e-01 5.34464478e-01 1.64715922e+00
-1.40761182e-01 3.10815513e-01 1.14372097e-01 7.01304197e-01
-4.48158264e-01 1.00634038e-01 -3.80923927e-01 -3.03227723e-01
3.97395194e-01 1.47760284e+00 -5.95780611e-01 -7.18406379e-01
-4.61878300e-01 6.40227675e-01 9.47781086e-01 5.77543341e-02
-5.70284843e-01 -4.17178065e-01 4.24807668e-02 3.92651111e-02
4.79779154e-01 6.93086311e-02 -2.21542478e-01 -1.03038287e+00
6.30664453e-02 -8.01360130e-01 7.31395781e-01 -9.32587981e-01
-9.92899060e-01 9.68989015e-01 3.78738157e-02 -7.06465542e-01
-6.83966339e-01 -6.14306509e-01 -4.28497910e-01 1.00095773e+00
-1.72701883e+00 -1.34188092e+00 2.19772160e-01 1.59126252e-01
4.61679369e-01 8.62426311e-03 8.78640831e-01 2.37845570e-01
-6.63666308e-01 4.00613070e-01 -7.33963430e-01 4.02243763e-01
6.31458461e-01 -1.56636190e+00 6.51432335e-01 1.10075510e+00
6.70583427e-01 8.10799539e-01 6.24925315e-01 -1.00070500e+00
-1.11172080e+00 -1.01795995e+00 1.94446409e+00 -8.66348684e-01
9.89713430e-01 -3.01466405e-01 -9.27854359e-01 1.01291382e+00
3.28973383e-01 9.98337939e-02 6.20769143e-01 7.05319226e-01
-3.97981465e-01 2.23460913e-01 -7.23067820e-01 4.02357906e-01
1.00819993e+00 -6.28608584e-01 -1.03312063e+00 4.43098158e-01
7.57267058e-01 -4.73106205e-01 -9.99455869e-01 5.84013164e-01
3.71541768e-01 -3.28209609e-01 9.28558230e-01 -1.06806922e+00
9.15609896e-01 -2.13900819e-01 7.28524998e-02 -1.04535282e+00
-2.95309037e-01 -6.88317120e-01 -1.06200838e+00 1.55247021e+00
1.30559659e+00 -3.33462119e-01 4.59684193e-01 5.38463652e-01
-8.11370015e-02 -1.31793237e+00 -3.67562115e-01 -4.34431881e-01
-2.34865174e-01 -6.42038524e-01 5.49335361e-01 6.20479703e-01
4.59085643e-01 9.59206164e-01 -9.94020253e-02 4.74871278e-01
1.17403395e-01 2.72663116e-01 7.82146513e-01 -1.22499835e+00
-6.51984572e-01 -1.00090921e-01 9.28178430e-02 -1.14365149e+00
3.36364895e-01 -1.21371257e+00 1.15323104e-01 -1.95429492e+00
3.31279725e-01 -4.41745400e-01 -4.02342260e-01 9.39336956e-01
-5.60220778e-01 -1.67253464e-01 -6.57825842e-02 3.12349424e-02
-8.26866984e-01 2.35051513e-01 1.01832581e+00 -1.37609303e-01
-1.37577400e-01 7.41547346e-02 -7.70168602e-01 7.35873044e-01
4.91223276e-01 -7.28273690e-01 -1.68719575e-01 -2.03987747e-01
9.08682644e-01 -5.54422615e-03 9.33841541e-02 -5.04456460e-01
4.40920889e-01 -1.70065705e-02 1.39479965e-01 -9.04112875e-01
9.87388939e-02 -6.03685439e-01 -3.51341188e-01 2.94000417e-01
-5.99159598e-01 -1.47356480e-01 3.12859386e-01 2.32235715e-01
-3.10184687e-01 -2.36185387e-01 1.05017804e-01 2.10061327e-01
-3.13815027e-01 1.90409571e-01 -1.35562018e-01 1.07751459e-01
6.08503580e-01 3.19302022e-01 -4.86708581e-01 -1.36016697e-01
-7.65420735e-01 5.33645213e-01 -2.64470875e-01 3.04583788e-01
4.58394766e-01 -9.67625856e-01 -9.32083547e-01 -1.57988697e-01
3.21852505e-01 2.87638605e-01 -2.78381795e-01 9.25267816e-01
-9.33388546e-02 7.17692196e-01 4.10924107e-01 -2.28122845e-01
-1.54853678e+00 7.17415273e-01 4.49883454e-02 -1.41921246e+00
-5.59157670e-01 1.18828487e+00 -3.18520427e-01 -2.90784985e-01
8.32969248e-02 -7.17630208e-01 -7.08799064e-01 1.40016019e-01
4.46008831e-01 -9.83309150e-02 3.19477260e-01 -4.46455151e-01
-6.81502759e-01 2.75608808e-01 -4.28682834e-01 -2.13854864e-01
1.68479919e+00 9.18713883e-02 -4.08718556e-01 2.24466741e-01
7.36700714e-01 5.61555743e-01 -9.56710279e-01 -4.68767941e-01
7.78345287e-01 1.12696193e-01 -2.61828136e-02 -1.26725602e+00
-5.00678718e-01 5.60069323e-01 -3.63516301e-01 4.07352507e-01
7.89279878e-01 6.76163733e-01 4.78912294e-01 6.91571534e-01
1.70028701e-01 -8.24589133e-01 -1.56265289e-01 8.18771720e-01
1.15189552e+00 -1.09516037e+00 2.58706361e-01 -9.58499432e-01
-5.57856917e-01 9.14566994e-01 5.62596142e-01 8.21670145e-03
5.50315380e-01 5.87994814e-01 -2.78029293e-01 -6.36121273e-01
-1.32122183e+00 -4.31023121e-01 8.73945832e-01 1.82575762e-01
8.74236107e-01 -4.29037064e-02 -4.21285897e-01 9.55395460e-01
-2.90735066e-01 5.05192466e-02 1.16108947e-01 9.57882822e-01
-1.79765046e-01 -1.56130981e+00 -1.01542063e-01 4.55587387e-01
-5.98794341e-01 -4.22003359e-01 -7.88975894e-01 5.53793490e-01
1.65180191e-01 1.01382899e+00 -3.02379251e-01 -4.25123215e-01
4.58407253e-01 3.41428071e-01 4.97916102e-01 -1.07548618e+00
-8.16120327e-01 -6.29641255e-03 1.03925395e+00 -5.41531384e-01
-3.88648510e-01 -5.84044993e-01 -1.36640704e+00 3.49302441e-01
-5.00022352e-01 3.12227070e-01 3.59539926e-01 1.43238008e+00
4.98372316e-01 9.35966492e-01 1.18964612e-01 -3.69452238e-01
-1.26117721e-01 -1.27892339e+00 9.64207668e-03 1.97196618e-01
1.35447145e-01 -4.66165155e-01 1.30160615e-01 1.52170047e-01]
|
[9.369006156921387, 8.671226501464844]
|
97bd2316-e98c-484a-8b40-8c5a83717549
|
language-free-training-for-zero-shot-video
|
2210.12977
| null |
https://arxiv.org/abs/2210.12977v1
|
https://arxiv.org/pdf/2210.12977v1.pdf
|
Language-free Training for Zero-shot Video Grounding
|
Given an untrimmed video and a language query depicting a specific temporal moment in the video, video grounding aims to localize the time interval by understanding the text and video simultaneously. One of the most challenging issues is an extremely time- and cost-consuming annotation collection, including video captions in a natural language form and their corresponding temporal regions. In this paper, we present a simple yet novel training framework for video grounding in the zero-shot setting, which learns a network with only video data without any annotation. Inspired by the recent language-free paradigm, i.e. training without language data, we train the network without compelling the generation of fake (pseudo) text queries into a natural language form. Specifically, we propose a method for learning a video grounding model by selecting a temporal interval as a hypothetical correct answer and considering the visual feature selected by our method in the interval as a language feature, with the help of the well-aligned visual-language space of CLIP. Extensive experiments demonstrate the prominence of our language-free training framework, outperforming the existing zero-shot video grounding method and even several weakly-supervised approaches with large margins on two standard datasets.
|
['Kwanghoon Sohn', 'Seongheon Park', 'Jiyoung Lee', 'Jungin Park', 'Dahye Kim']
|
2022-10-24
| null | null | null | null |
['video-grounding']
|
['computer-vision']
|
[ 3.02432120e-01 2.57229954e-02 -5.57096124e-01 -2.45055765e-01
-8.71345699e-01 -6.52979791e-01 4.85006213e-01 -1.76466450e-01
-3.54772985e-01 5.54782450e-01 4.68545407e-02 -2.94705868e-01
2.33359665e-01 -4.74955857e-01 -1.22544193e+00 -3.70794624e-01
2.87437835e-03 1.24928355e-01 3.03284079e-01 1.47724245e-03
8.31145197e-02 -8.54575112e-02 -1.36759269e+00 5.18090665e-01
4.43099141e-01 1.20922887e+00 1.51166871e-01 5.47123909e-01
1.04935737e-02 1.31656063e+00 -6.15706623e-01 -6.08124793e-01
2.43724376e-01 -7.70483196e-01 -7.54266024e-01 5.45543730e-01
7.59898543e-01 -6.64427936e-01 -8.69479299e-01 1.05545545e+00
-1.30788401e-01 2.49307901e-01 1.74336061e-01 -1.70513284e+00
-8.59600604e-01 4.60463375e-01 -3.24082315e-01 1.22711405e-01
9.42998588e-01 1.94286391e-01 1.05675876e+00 -9.00596976e-01
9.82698202e-01 1.04507685e+00 4.08650041e-01 4.90061253e-01
-1.01311457e+00 -5.96176922e-01 3.20967615e-01 4.33520079e-01
-1.79263771e+00 -2.92948574e-01 1.00142491e+00 -6.66511357e-01
5.00927985e-01 1.22956097e-01 7.93660522e-01 1.51092541e+00
-1.03015818e-01 9.48919892e-01 7.12748706e-01 -3.88192415e-01
1.82504982e-01 5.01056649e-02 -3.32784086e-01 1.11546671e+00
-1.76199481e-01 -1.17345273e-01 -9.31481361e-01 -2.23873332e-01
8.64098728e-01 2.35563383e-01 -5.76611876e-01 -5.50446212e-01
-1.72350764e+00 7.69281387e-01 2.11541206e-01 3.29435498e-01
-1.60175368e-01 2.98261136e-01 3.84396017e-01 3.34560513e-01
6.02577984e-01 3.65080953e-01 -1.90570176e-01 -3.14695239e-02
-1.26817334e+00 2.32830614e-01 7.42268384e-01 1.14438641e+00
9.31768775e-01 6.99992105e-02 -5.11834919e-01 1.53670469e-02
-6.93601295e-02 3.74609590e-01 4.58358526e-01 -9.96355176e-01
8.11882198e-01 3.82555902e-01 5.36895812e-01 -1.32024121e+00
1.35456830e-01 7.45838583e-02 -5.70689142e-01 -3.96746367e-01
6.34627104e-01 -4.42557558e-02 -6.13681614e-01 1.91470265e+00
8.55346844e-02 9.66400027e-01 2.53924653e-02 1.22941279e+00
5.88496685e-01 8.56463194e-01 -2.97916755e-02 -5.42132676e-01
1.35905755e+00 -1.00075698e+00 -8.94994795e-01 -1.81385756e-01
5.94761491e-01 -3.72541189e-01 1.31607068e+00 3.24088156e-01
-7.70368993e-01 -5.08558273e-01 -1.10437441e+00 -2.18492404e-01
-3.34913999e-01 2.20392168e-01 3.02926183e-01 6.14570640e-02
-9.31571543e-01 4.50258732e-01 -5.59217334e-01 -4.33349043e-01
1.78793162e-01 -2.33547157e-03 -5.40653706e-01 -1.15400836e-01
-1.49365199e+00 3.20943981e-01 5.65855086e-01 8.41715485e-02
-1.04639530e+00 -3.78651053e-01 -1.06695175e+00 -1.40717223e-01
9.91411567e-01 -6.14968061e-01 1.12970400e+00 -1.50604773e+00
-1.18248451e+00 1.04065967e+00 -1.92319691e-01 -7.94317186e-01
8.27876806e-01 -3.33850086e-01 -4.29166585e-01 9.25060511e-01
3.46229613e-01 8.46740723e-01 1.26987505e+00 -1.07932782e+00
-3.82061243e-01 -7.14761689e-02 3.98238540e-01 1.04037747e-01
-4.15212333e-01 -2.80679744e-02 -7.35784471e-01 -8.46339583e-01
-1.43585177e-02 -9.32932317e-01 1.41494304e-01 4.21011180e-01
-4.21351165e-01 -1.08527429e-01 9.21771944e-01 -8.84121478e-01
1.38931715e+00 -2.13185930e+00 1.74967021e-01 -1.90761536e-01
3.26676369e-01 -4.65588123e-02 -8.35684035e-03 3.49435151e-01
-3.93318720e-02 9.50794518e-02 -1.14026651e-01 -2.13734046e-01
-1.03386953e-01 1.78200915e-01 -7.74163604e-01 7.08262384e-01
3.85349959e-01 1.07163227e+00 -1.38466954e+00 -8.22761238e-01
1.61009550e-01 2.61084765e-01 -4.39779311e-01 6.48157656e-01
-6.15873277e-01 5.35626113e-01 -4.27147925e-01 6.46033585e-01
5.25695719e-02 -3.55638742e-01 -3.67188491e-02 -3.25477988e-01
-3.00264615e-03 -1.89486355e-01 -8.10389876e-01 2.11062932e+00
-1.53123245e-01 9.54745770e-01 -2.85910070e-01 -1.15201235e+00
6.94422841e-01 6.87287629e-01 6.62854016e-01 -3.83702666e-01
-1.16964743e-01 4.52772789e-02 -6.81584537e-01 -1.09545517e+00
4.63402331e-01 3.35422643e-02 -2.20465809e-01 5.06623566e-01
2.42746666e-01 1.16507210e-01 6.44730404e-02 4.28371996e-01
1.01649189e+00 4.77663785e-01 3.32167953e-01 1.99204281e-01
4.04478401e-01 1.05055198e-02 2.36946136e-01 8.38593304e-01
-3.91388774e-01 7.86220372e-01 6.39724672e-01 -6.01126134e-01
-1.00781953e+00 -6.42466068e-01 4.44840968e-01 1.19237745e+00
4.20485288e-01 -6.30861938e-01 -8.57804060e-01 -8.05272102e-01
-3.07434440e-01 4.99681264e-01 -7.27921486e-01 -1.71045855e-01
-6.62931323e-01 1.28570318e-01 6.29724562e-01 2.09119603e-01
4.60649967e-01 -9.78234649e-01 -6.83064938e-01 -4.91162986e-02
-8.09636891e-01 -1.81621635e+00 -9.36739564e-01 -2.49224097e-01
-5.08002758e-01 -1.06254125e+00 -7.71641493e-01 -9.52139378e-01
6.10736787e-01 6.23566449e-01 1.04163778e+00 2.76634336e-01
1.28031997e-02 6.34980500e-01 -5.04496753e-01 2.15355203e-01
-2.05374464e-01 -3.62644672e-01 -4.31446321e-02 5.97321689e-01
3.15951854e-01 -1.62018761e-01 -6.17431939e-01 2.74912834e-01
-1.18599367e+00 4.17898208e-01 1.56706750e-01 8.44563782e-01
5.29657662e-01 -9.15835872e-02 2.92828262e-01 -5.58733225e-01
1.68434933e-01 -8.02832007e-01 -4.20632660e-01 4.05286610e-01
-1.32096052e-01 -1.22338809e-01 7.28738606e-01 -8.25614452e-01
-5.85301816e-01 3.89205694e-01 4.38654333e-01 -1.26488698e+00
-7.92673901e-02 3.49829406e-01 -8.50957111e-02 1.11530364e-01
5.41500509e-01 4.29236114e-01 -8.93394202e-02 1.30930757e-02
5.48800826e-01 3.57616007e-01 8.43577087e-01 -5.61494827e-01
1.00823855e+00 8.02556694e-01 -2.60989368e-01 -7.40983307e-01
-1.18449664e+00 -6.73911989e-01 -7.85528719e-01 -4.36522901e-01
1.07841146e+00 -1.05926490e+00 -5.34947336e-01 4.64153178e-02
-1.50461316e+00 -1.74912617e-01 -1.67553604e-01 4.17263389e-01
-9.74576175e-01 6.66193187e-01 -3.36319625e-01 -6.50682747e-01
7.04889297e-02 -9.15292680e-01 1.36442554e+00 -2.28891954e-01
-1.88574478e-01 -8.52593362e-01 -2.83553243e-01 4.44931656e-01
-1.74183458e-01 5.27400911e-01 5.87150633e-01 -7.38724589e-01
-9.04796541e-01 -3.26282859e-01 -2.66959071e-01 1.22960784e-01
-1.02653399e-01 -9.42251012e-02 -8.25286806e-01 -1.40662596e-01
2.06339911e-01 -6.23800993e-01 6.10707939e-01 1.13896281e-01
1.37807703e+00 -8.25552344e-01 -8.61152336e-02 6.24781072e-01
1.40033627e+00 5.86533025e-02 4.10885483e-01 2.03809246e-01
8.92506540e-01 6.23216927e-01 8.24804008e-01 3.91363800e-01
2.90383905e-01 8.02654326e-01 6.01089835e-01 8.02089646e-03
2.01613262e-01 -9.44539607e-01 5.24592638e-01 3.86702359e-01
2.48091802e-01 -5.37357926e-01 -6.52988195e-01 4.69630271e-01
-2.25585794e+00 -1.40835798e+00 3.70841026e-01 2.25397635e+00
7.68684983e-01 -1.55925769e-02 1.67627707e-01 -2.58135498e-02
9.77674901e-01 4.55584526e-01 -4.76155519e-01 2.96129137e-01
-1.67375773e-01 -3.85392666e-01 2.42321014e-01 3.70986611e-01
-1.31789136e+00 1.10758579e+00 5.54242659e+00 6.33046329e-01
-1.25224340e+00 1.68891743e-01 6.84069157e-01 -4.92117368e-02
-1.36220157e-01 4.85655405e-02 -3.46399814e-01 5.72123647e-01
8.96860063e-01 -3.93494934e-01 5.71638465e-01 7.59192646e-01
5.73477268e-01 7.26463571e-02 -1.44618464e+00 1.27687991e+00
6.46783113e-01 -1.49867129e+00 2.60498434e-01 -3.07649910e-01
5.39671063e-01 -4.28620338e-01 -1.33482190e-02 2.75349319e-01
-3.97536278e-01 -8.50837648e-01 1.29805732e+00 3.72905701e-01
1.17402530e+00 -1.73636094e-01 4.29354697e-01 3.35585594e-01
-1.21279931e+00 -1.04896113e-01 -4.76894416e-02 -9.59123001e-02
2.38628358e-01 2.95884702e-02 -6.15239203e-01 3.48735839e-01
5.57950079e-01 1.03039932e+00 -4.88470167e-01 6.79141104e-01
-1.97663665e-01 5.76835215e-01 -2.77649835e-02 2.06818163e-01
3.98704469e-01 -1.17977597e-01 6.22545063e-01 1.15659869e+00
5.09802580e-01 1.43979952e-01 5.74599981e-01 8.97502303e-01
-3.32560897e-01 1.37642980e-01 -1.05595124e+00 -2.10444421e-01
3.71149778e-01 9.61185038e-01 -8.08311403e-01 -5.66336215e-01
-6.14293396e-01 1.31107497e+00 1.16749480e-01 7.49998093e-01
-1.23250413e+00 -1.54976100e-01 9.03959423e-02 4.21211421e-01
2.68399984e-01 -1.77534997e-01 4.68120605e-01 -1.66994810e+00
3.21591318e-01 -7.71655858e-01 4.09593821e-01 -1.34439421e+00
-9.51413512e-01 6.01145923e-01 1.76497102e-01 -1.76828277e+00
-5.43717802e-01 -4.86839414e-01 -3.80544990e-01 2.49645218e-01
-1.26139188e+00 -1.26223278e+00 -4.80368614e-01 1.00675666e+00
8.69483650e-01 -4.48393114e-02 4.50183570e-01 1.93257213e-01
-2.87371367e-01 3.50059062e-01 -3.78404260e-01 4.20302153e-01
8.78240943e-01 -1.03177810e+00 -3.75173241e-02 1.01645553e+00
6.89724505e-01 4.69545484e-01 8.47827256e-01 -4.85557407e-01
-1.52849126e+00 -1.24618661e+00 9.50634301e-01 -4.98314142e-01
1.22864330e+00 -5.36383808e-01 -9.02171493e-01 9.56651330e-01
1.87652141e-01 2.94379175e-01 4.11976397e-01 -6.13547683e-01
-5.88511050e-01 3.03225089e-02 -8.15948784e-01 6.91961050e-01
1.12906623e+00 -1.02819896e+00 -7.06645846e-01 7.60750115e-01
1.23206055e+00 -5.40102005e-01 -3.40788275e-01 1.50308944e-02
4.30794835e-01 -7.59466887e-01 9.10989225e-01 -8.11551332e-01
7.13610828e-01 -3.69474620e-01 -2.23470107e-01 -7.13839173e-01
1.92674682e-01 -1.15809965e+00 -4.08109754e-01 9.40633714e-01
-1.24902688e-02 -3.20857810e-03 7.93548465e-01 3.58656377e-01
8.38691443e-02 -4.79805827e-01 -1.04188764e+00 -7.39664853e-01
-5.15420139e-01 -5.34299970e-01 2.04228178e-01 1.06053197e+00
6.93851709e-02 3.05254966e-01 -1.00724983e+00 1.82009414e-01
4.02655929e-01 2.55570173e-01 6.70971692e-01 -7.48173952e-01
-2.48201266e-01 1.42978672e-02 -5.77862322e-01 -1.38297367e+00
5.96914887e-01 -5.61644375e-01 2.84465998e-01 -9.06838179e-01
1.88517615e-01 4.31304246e-01 -1.90406576e-01 3.46438169e-01
-2.74103489e-02 4.24456298e-01 2.01264396e-01 4.81193364e-01
-1.20800865e+00 3.91784489e-01 1.10145175e+00 -2.53622323e-01
-3.39651518e-02 -1.87260345e-01 -3.07501823e-01 7.59431720e-01
3.46337885e-01 -4.61782962e-01 -7.20049143e-01 -4.25889283e-01
3.08298796e-01 6.30739927e-01 8.60522032e-01 -7.98472762e-01
3.66235465e-01 -3.29657495e-01 1.24405613e-02 -3.79995197e-01
4.34368134e-01 -9.92899716e-01 -3.78493853e-02 2.19821408e-01
-6.82538509e-01 2.32468531e-01 -1.37535512e-01 1.08446729e+00
-4.91170198e-01 -1.48803234e-01 2.48942554e-01 -2.25027069e-01
-9.92174625e-01 4.77821290e-01 -2.29268312e-01 1.54571280e-01
1.25683570e+00 -3.43079090e-01 -3.27406079e-01 -8.17646146e-01
-8.01918983e-01 4.55042161e-02 5.30076087e-01 6.61353767e-01
7.54254699e-01 -1.37571931e+00 -3.60338926e-01 5.79834543e-02
4.66252059e-01 -1.95617437e-01 2.09150776e-01 7.02709973e-01
-5.98162174e-01 5.02614796e-01 4.80560884e-02 -7.94811487e-01
-9.71968770e-01 1.06853211e+00 2.34111845e-01 7.20211565e-02
-7.30559528e-01 5.54925084e-01 4.10795867e-01 2.96362728e-01
4.65681940e-01 -4.01516557e-01 2.44066399e-02 2.36707851e-02
5.11991560e-01 -1.24505207e-01 -3.58650327e-01 -9.75327253e-01
-1.30003989e-01 4.20699716e-01 3.85995030e-01 -1.58985198e-01
8.67110968e-01 -3.21074873e-01 -6.06071949e-03 8.12323213e-01
1.43648732e+00 -2.05846578e-01 -1.53086412e+00 -4.48170900e-01
1.71647161e-01 -6.09048724e-01 -2.66105533e-01 -2.76046127e-01
-9.60292459e-01 7.49056101e-01 1.26540437e-01 2.82822162e-01
1.01329350e+00 1.82609871e-01 8.53096068e-01 5.14789641e-01
4.02489543e-01 -7.88195193e-01 6.58079088e-01 1.16135113e-01
8.63298774e-01 -1.45817685e+00 -1.31374687e-01 -2.04248041e-01
-6.95694983e-01 1.17350054e+00 6.78019881e-01 2.07619462e-03
4.19469267e-01 -2.84132928e-01 1.12358555e-02 -1.40822351e-01
-8.02254438e-01 -4.21014801e-02 3.56944531e-01 3.28683913e-01
5.00590615e-02 -2.36922026e-01 4.98980209e-02 5.21536469e-01
7.68541396e-02 2.28216827e-01 5.60494184e-01 7.69061267e-01
-2.39203572e-01 -4.46244180e-01 -2.71538407e-01 4.32444662e-02
-5.13974190e-01 2.77362708e-02 -3.78510922e-01 7.52321959e-01
1.77053764e-01 8.68071258e-01 1.37587026e-01 -4.52227414e-01
1.95463356e-02 1.90375730e-01 2.03382701e-01 -6.41851366e-01
-1.60863683e-01 7.78905824e-02 -2.06973121e-01 -8.62657547e-01
-7.44313002e-01 -3.73086840e-01 -9.64773238e-01 4.03225981e-02
-1.00079514e-01 2.50986516e-01 2.06658810e-01 1.10168898e+00
2.14785337e-01 -4.33445834e-02 6.49017572e-01 -8.95595968e-01
-3.29650432e-01 -5.57107687e-01 -3.06847453e-01 8.92080665e-01
6.79965079e-01 -5.59258878e-01 -5.87791264e-01 7.49280393e-01]
|
[10.049249649047852, 0.7395883798599243]
|
8d4c14d5-2474-4a01-a151-16c689c028ee
|
hymo-vulnerability-detection-in-smart
|
2304.13103
| null |
https://arxiv.org/abs/2304.13103v1
|
https://arxiv.org/pdf/2304.13103v1.pdf
|
HyMo: Vulnerability Detection in Smart Contracts using a Novel Multi-Modal Hybrid Model
|
With blockchain technology rapidly progress, the smart contracts have become a common tool in a number of industries including finance, healthcare, insurance and gaming. The number of smart contracts has multiplied, and at the same time, the security of smart contracts has drawn considerable attention due to the monetary losses brought on by smart contract vulnerabilities. Existing analysis techniques are capable of identifying a large number of smart contract security flaws, but they rely too much on rigid criteria established by specialists, where the detection process takes much longer as the complexity of the smart contract rises. In this paper, we propose HyMo as a multi-modal hybrid deep learning model, which intelligently considers various input representations to consider multimodality and FastText word embedding technique, which represents each word as an n-gram of characters with BiGRU deep learning technique, as a sequence processing model that consists of two GRUs to achieve higher accuracy in smart contract vulnerability detection. The model gathers features using various deep learning models to identify the smart contract vulnerabilities. Through a series of studies on the currently publicly accessible dataset such as ScrawlD, we show that our hybrid HyMo model has excellent smart contract vulnerability detection performance. Therefore, HyMo performs better detection of smart contract vulnerabilities against other approaches.
|
['Jafar Tahmoresnezhad', 'Mohammad Khodadadi']
|
2023-04-25
| null | null | null | null |
['vulnerability-detection']
|
['miscellaneous']
|
[-1.34448871e-01 -2.83868730e-01 -1.03920951e-01 1.25058427e-01
-7.56668270e-01 -9.59266961e-01 8.35408270e-01 4.23478037e-02
-1.45827487e-01 3.56252313e-01 8.33871841e-01 -9.39626157e-01
2.13672131e-01 -1.02278388e+00 -2.58787900e-01 -9.05337036e-01
1.03015795e-01 6.39971852e-01 2.39792448e-02 -6.43969774e-01
5.52918673e-01 1.40520141e-01 -1.02600920e+00 6.87917888e-01
5.08328855e-01 8.30451727e-01 -4.43008304e-01 6.43046856e-01
-3.24294865e-01 1.05071557e+00 -2.63769031e-01 -9.19498622e-01
4.91269261e-01 -1.00752786e-01 -8.09650660e-01 -1.17268384e+00
-2.14708433e-01 -6.96230352e-01 -5.46689510e-01 1.47417819e+00
8.08795929e-01 -7.03581989e-01 4.93978381e-01 -1.22526240e+00
-7.28900433e-01 1.42710257e+00 -5.84910989e-01 7.64979869e-02
2.42972776e-01 3.47128808e-01 1.31400132e+00 -4.13733453e-01
4.56504077e-01 1.29503953e+00 7.34247684e-01 1.87506109e-01
-8.40404987e-01 -1.06439328e+00 -5.48945904e-01 5.06473720e-01
-6.95871770e-01 1.25908218e-02 9.85768080e-01 -4.64893907e-01
1.43156421e+00 1.61961377e-01 4.70840096e-01 1.28311014e+00
7.51737654e-01 8.68216753e-01 1.04618907e+00 -1.01038776e-01
1.81979761e-02 -1.71614543e-01 3.41092229e-01 1.33777380e-01
3.66387099e-01 7.22547770e-01 1.46698937e-01 -7.36345768e-01
2.07285658e-01 3.65853101e-01 1.80106610e-01 -7.46078789e-02
-1.04589808e+00 1.31117404e+00 -1.01638034e-01 7.89812088e-01
-2.86032319e-01 2.31469393e-01 1.22818470e+00 7.76452959e-01
6.35580495e-02 1.07374027e-01 -7.57509530e-01 -5.97613931e-01
-6.72701776e-01 4.65528756e-01 1.06561649e+00 4.95520681e-01
3.16925496e-01 -9.49663594e-02 3.35663036e-02 2.17164814e-01
5.74575484e-01 6.07901871e-01 6.05822504e-01 -5.27734160e-01
9.48675871e-01 1.01091588e+00 -1.68756738e-01 -9.87389147e-01
-3.33220929e-01 -4.13614482e-01 -9.52973366e-01 5.59549510e-01
3.33115995e-01 -1.27233103e-01 -6.80533826e-01 1.26787794e+00
2.61733592e-01 -1.41750231e-01 2.31178522e-01 4.79761124e-01
4.74658191e-01 7.65072405e-01 2.09455028e-01 3.69492397e-02
1.52833641e+00 -5.78414679e-01 -7.52569556e-01 3.49891812e-01
1.14455366e+00 -1.19272721e+00 5.80480754e-01 6.17273331e-01
-7.26114810e-01 -6.69770241e-02 -1.03322709e+00 4.03082818e-02
-6.73951328e-01 -6.50564492e-01 7.47244298e-01 1.26050627e+00
-5.54087222e-01 6.03470623e-01 -5.81048429e-01 4.97509493e-03
4.26077455e-01 3.69871765e-01 -2.29719847e-01 1.88588917e-01
-1.71242392e+00 1.00722027e+00 7.99756050e-01 -1.07981920e-01
-8.48812640e-01 -2.51120687e-01 -6.90938056e-01 3.41596752e-01
1.41718075e-01 -2.41346776e-01 1.04305804e+00 -7.91787982e-01
-1.26956332e+00 5.46410918e-01 5.27705967e-01 -5.02948761e-01
6.51492059e-01 -1.41794428e-01 -6.53226376e-01 -7.64449984e-02
-4.00115728e-01 -2.30670318e-01 6.34965360e-01 -6.56148851e-01
-5.79536617e-01 -4.30337876e-01 3.59227836e-01 -2.61844516e-01
-1.02282979e-01 6.45074606e-01 5.08269370e-01 -9.02145684e-01
-1.45715296e-01 -9.85754251e-01 -6.22519553e-02 -9.93587613e-01
-7.42630363e-02 -4.19169813e-01 1.13886321e+00 -9.50493217e-01
1.45509744e+00 -1.92478323e+00 -2.13175938e-01 3.75141710e-01
1.49369106e-01 7.62499392e-01 -1.41011894e-01 1.21225071e+00
-3.17529261e-01 3.93389583e-01 -3.69686037e-01 3.16623092e-01
4.29832131e-01 2.15955153e-02 -5.61205208e-01 3.81783605e-01
-4.20395315e-01 1.29434681e+00 -7.98278928e-01 -3.89674008e-01
9.79754403e-02 3.15954417e-01 -4.96830434e-01 2.42998153e-02
-6.53078184e-02 1.07458070e-01 -6.61764979e-01 8.63125026e-01
8.95960748e-01 1.23805486e-01 2.59191394e-01 -1.21976763e-01
-9.13182274e-03 2.81095922e-01 -9.89192784e-01 1.42365932e+00
-1.75108045e-01 2.39082247e-01 -1.12318210e-01 -1.13324344e+00
6.96059048e-01 8.35884273e-01 7.30557501e-01 -8.15409243e-01
7.01708198e-01 3.73392105e-01 3.30756515e-01 -8.86767447e-01
1.22082002e-01 -3.71246874e-01 -2.41147026e-01 1.25538087e+00
-4.87237930e-01 3.14143747e-01 -2.31166273e-01 2.96093017e-01
1.38157403e+00 7.54475817e-02 2.55688846e-01 1.40589684e-01
6.43512666e-01 -9.95245799e-02 6.20585024e-01 4.22521919e-01
-4.22997087e-01 2.36227855e-01 7.43990004e-01 -1.10240149e+00
-1.32315636e+00 -6.86348379e-01 1.90174207e-01 8.02270889e-01
-2.49346480e-01 -1.42082036e-01 -6.54924393e-01 -1.08029079e+00
2.82897949e-01 3.49814832e-01 -4.86013472e-01 -3.25713515e-01
-6.85730994e-01 -9.20471072e-01 9.45840240e-01 5.56269169e-01
6.83181226e-01 -1.58189356e+00 -5.86674035e-01 4.21132743e-01
-1.04481459e-01 -5.58619082e-01 -3.53782922e-01 -3.37577388e-02
-7.41990387e-01 -1.42737079e+00 -4.13682431e-01 -7.17133045e-01
8.34085867e-02 -2.50397861e-01 7.49231458e-01 3.33590209e-01
-2.49789149e-01 -3.40908706e-01 -6.60187960e-01 -2.10182909e-02
-8.04017186e-01 6.55406341e-02 -1.08366534e-01 -1.61897302e-01
8.44019532e-01 -5.98162055e-01 -6.01705551e-01 -6.33474439e-02
-1.09702086e+00 -2.71809548e-01 1.03283548e+00 8.48023534e-01
-2.02731147e-01 2.10962400e-01 6.99628949e-01 -1.16929615e+00
8.13246548e-01 -7.14362860e-01 -6.28164530e-01 3.16533059e-01
-9.69727933e-01 2.45312929e-01 4.14890587e-01 -2.67019957e-01
-8.61239851e-01 -2.04521358e-01 -4.39767689e-01 -1.44895479e-01
1.51135564e-01 6.96165264e-01 -3.16428930e-01 -1.06634691e-01
2.44936883e-01 2.25411922e-01 -1.58441886e-01 -5.43325961e-01
6.42850041e-01 1.09232306e+00 6.76148087e-02 -2.54956573e-01
1.13819635e+00 2.64351010e-01 -3.32887441e-01 -4.36904952e-02
1.77464053e-01 -3.14577252e-01 -3.46657872e-01 2.25171179e-01
8.10045838e-01 -3.36046696e-01 -1.44712698e+00 9.99165475e-01
-1.35795820e+00 1.88913792e-01 4.38112855e-01 6.40195847e-01
-3.69354308e-01 9.08889592e-01 -9.67548788e-01 -8.76523674e-01
-1.09528613e+00 -1.31496453e+00 5.26757598e-01 -9.16061252e-02
-2.97394514e-01 -9.12572861e-01 5.75802624e-01 5.95862150e-01
4.89910692e-01 3.75479817e-01 1.39550340e+00 -1.04392445e+00
-5.97693324e-01 -4.63707864e-01 -3.32075626e-01 3.73411447e-01
1.11504324e-01 -2.97118962e-01 -8.78492475e-01 -5.56932032e-01
3.81934971e-01 -1.98681980e-01 8.53535354e-01 5.56160808e-02
5.12526214e-01 -8.52344990e-01 -7.04455152e-02 5.92936873e-01
1.63391471e+00 7.33977616e-01 9.85857964e-01 6.74533725e-01
5.57873607e-01 2.92438567e-01 1.42126396e-01 4.23594147e-01
3.17478865e-01 1.36697814e-01 7.75000334e-01 2.35284120e-01
5.21563292e-01 -1.85292214e-01 4.63292658e-01 7.92636633e-01
-1.01647802e-01 1.18142292e-01 -1.16629994e+00 3.90047789e-01
-1.97903943e+00 -1.50379002e+00 -5.11666620e-03 1.83847952e+00
7.11491823e-01 2.77762979e-01 1.13392055e-01 4.40582246e-01
4.88948673e-01 2.61426836e-01 -2.71946043e-01 -8.82724762e-01
-7.88323060e-02 2.74531662e-01 7.87076533e-01 3.73108387e-01
-1.09251761e+00 7.43317664e-01 6.21692419e+00 7.17334270e-01
-1.04602754e+00 3.45357150e-01 3.47860634e-01 7.05643296e-02
-7.70477057e-01 3.72899264e-01 -3.03120524e-01 9.06064510e-01
6.74683452e-01 -1.30619392e-01 4.83701050e-01 6.25247478e-01
1.16535716e-01 3.73004556e-01 -7.40620255e-01 1.04851806e+00
-1.81333661e-01 -1.47192526e+00 8.54966044e-02 3.25315058e-01
6.42280400e-01 2.95259565e-01 1.99608579e-02 3.02853465e-01
9.87484038e-01 -1.03698766e+00 5.57363927e-01 2.32273322e-02
2.72111684e-01 -8.62266362e-01 1.43513465e+00 6.27049953e-02
-1.01724911e+00 -6.76187575e-01 -1.36369839e-01 -2.02263549e-01
8.44962671e-02 1.53165147e-01 -5.99402905e-01 7.95991838e-01
5.75626493e-01 3.81286591e-01 -1.18390329e-01 8.57662261e-01
-1.54791757e-01 6.96440876e-01 -6.21036217e-02 -1.43486902e-01
4.67437118e-01 -1.87206820e-01 3.34336251e-01 1.04907453e+00
3.05567861e-01 2.55596459e-01 1.71756335e-02 5.90202451e-01
1.36828601e-01 1.33283645e-01 -7.03738511e-01 -4.31211948e-01
3.67878288e-01 1.12067175e+00 -4.21019107e-01 -3.43028724e-01
-7.25650132e-01 3.75970721e-01 -1.30363524e-01 3.42463106e-01
-6.87241614e-01 -6.72290742e-01 6.39534771e-01 -2.03652099e-01
3.12474996e-01 -1.72673032e-01 -6.69762254e-01 -1.00805616e+00
-1.07949764e-01 -1.42990196e+00 8.26346934e-01 -2.49715030e-01
-1.06892693e+00 4.12213862e-01 -3.09164941e-01 -1.35064173e+00
-2.97130287e-01 -4.30517465e-01 -9.61455822e-01 7.65073597e-01
-1.51572108e+00 -1.51633132e+00 5.59999347e-01 5.39980829e-01
1.71671629e-01 -6.66474581e-01 7.04161942e-01 5.32621205e-01
-3.05909663e-01 5.97698867e-01 4.12605971e-01 7.57124364e-01
2.70833164e-01 -8.56726706e-01 8.15211952e-01 7.63775706e-01
1.37885749e-01 7.37689376e-01 6.78169966e-01 -9.62784827e-01
-1.69011867e+00 -4.34552729e-01 1.13977611e+00 -2.99440086e-01
9.81154084e-01 -2.19297722e-01 -7.78246105e-01 7.05831885e-01
6.53075218e-01 -4.90078032e-01 5.58577776e-01 -1.06453501e-01
-8.05197060e-01 -1.34367615e-01 -1.14756465e+00 2.11278558e-01
6.09620571e-01 -7.92990983e-01 -9.70581591e-01 -1.38267232e-02
6.57284856e-01 8.88919756e-02 -7.30067670e-01 2.55723089e-01
9.97840822e-01 -9.87221539e-01 7.44144499e-01 -8.74218702e-01
3.70490760e-01 -9.33165923e-02 -4.29405235e-02 -8.45160007e-01
-3.53471130e-01 -8.67624938e-01 -2.35530406e-01 1.30255866e+00
2.11022943e-01 -9.56797004e-01 8.36449146e-01 5.17304361e-01
2.84405828e-01 -6.46369338e-01 -1.22387111e+00 -6.61026895e-01
3.75791997e-01 -2.87706167e-01 1.23315656e+00 1.25089645e+00
5.31331301e-01 -1.50461569e-01 -3.34818870e-01 -9.75492895e-02
7.06129313e-01 3.18858683e-01 5.07116556e-01 -1.04064119e+00
-2.32672721e-01 -7.95879364e-01 -5.31609774e-01 -2.56801218e-01
-6.05512783e-02 -9.17382240e-01 -6.85496807e-01 -1.35700262e+00
3.39448571e-01 -3.39792579e-01 -7.83773601e-01 6.99798942e-01
9.26291943e-02 2.88822502e-02 6.38336837e-02 2.72204340e-01
9.83011872e-02 3.72367710e-01 7.76833713e-01 -8.07295144e-01
7.69219175e-02 -4.15184051e-02 -6.37154639e-01 4.11196381e-01
7.48897552e-01 -6.32026196e-01 -1.31505325e-01 -5.50660610e-01
9.02720273e-01 9.45195779e-02 -7.47142136e-02 -3.97186846e-01
1.57823086e-01 -1.97156474e-01 -8.43511745e-02 -8.45536292e-01
-4.22924340e-01 -1.00989676e+00 3.98810476e-01 1.13055086e+00
-3.33535634e-02 3.57822955e-01 -1.81794479e-01 2.40077078e-01
-2.58208364e-01 -1.96896404e-01 4.50453132e-01 -3.20660114e-01
-5.17633915e-01 2.99584627e-01 -2.25431979e-01 -2.36342177e-01
9.36395168e-01 -1.40395418e-01 -5.35774469e-01 -3.06521833e-01
-3.21365058e-01 3.86275053e-01 3.13008606e-01 8.54771912e-01
4.12064940e-01 -1.42963159e+00 -9.92261410e-01 3.28900158e-01
-1.22136153e-01 -6.30320013e-01 2.57221729e-01 8.27862144e-01
-9.35966969e-01 4.14956421e-01 -3.29473406e-01 8.71489495e-02
-1.43035173e+00 8.64848197e-01 6.13166019e-02 -8.40422988e-01
-7.72109091e-01 4.36508119e-01 -4.60380726e-02 -3.09691995e-01
1.65233925e-01 -3.09232533e-01 -4.76476938e-01 1.65008098e-01
6.69413626e-01 4.78779644e-01 -9.39295962e-02 -6.75620258e-01
-3.96969855e-01 7.27890193e-01 -3.69919240e-01 -1.00576960e-01
1.50742602e+00 2.92445898e-01 -4.81468320e-01 -3.34960461e-01
1.41912711e+00 -9.59687084e-02 -6.19735777e-01 -1.79699391e-01
5.14891267e-01 -4.44624752e-01 -1.35021582e-01 -9.81011629e-01
-1.24154603e+00 9.29305851e-01 6.98042154e-01 3.62908721e-01
7.64938891e-01 -1.54130533e-01 1.52848327e+00 2.14312479e-01
4.82833862e-01 -1.16719282e+00 -3.10213238e-01 7.06137478e-01
6.07760608e-01 -1.19213486e+00 -2.51941860e-01 3.04960102e-01
-4.98539954e-01 1.09457839e+00 -2.53313482e-01 -1.37016416e-01
7.84255385e-01 4.51564223e-01 4.52103049e-01 -1.85727760e-01
-4.42551076e-01 2.40112960e-01 -4.38043118e-01 3.94030064e-01
1.53634578e-01 2.03579381e-01 -7.53504515e-01 7.83665657e-01
-2.76513755e-01 1.40606061e-01 2.42695451e-01 8.68833065e-01
-4.54855561e-01 -1.75820863e+00 -4.95780796e-01 2.40046889e-01
-1.00015473e+00 -3.43298465e-01 -2.61520624e-01 3.29784065e-01
7.32261539e-02 9.37907040e-01 -5.04513919e-01 -8.38339508e-01
-5.04742749e-02 2.22192496e-01 9.00924671e-03 -8.40277374e-02
-1.17684066e+00 -2.80662447e-01 4.70760167e-02 -4.90943819e-01
-3.49618010e-02 -6.74393952e-01 -1.25068355e+00 -7.86858797e-01
-1.74998254e-01 1.12887166e-01 6.33156598e-01 1.05947888e+00
9.39448550e-02 -1.88307017e-01 1.00503409e+00 -5.31954393e-02
-1.01551700e+00 -1.01534390e+00 -5.18002629e-01 7.67989576e-01
4.41794932e-01 -3.20592105e-01 -2.04154730e-01 -2.51272053e-01]
|
[6.81016206741333, 7.253413677215576]
|
303588f4-02d4-4964-a219-812fb6b79214
|
airrl-a-reinforcement-learning-approach-to
|
2003.12205
| null |
https://arxiv.org/abs/2003.12205v1
|
https://arxiv.org/pdf/2003.12205v1.pdf
|
AirRL: A Reinforcement Learning Approach to Urban Air Quality Inference
|
Urban air pollution has become a major environmental problem that threatens public health. It has become increasingly important to infer fine-grained urban air quality based on existing monitoring stations. One of the challenges is how to effectively select some relevant stations for air quality inference. In this paper, we propose a novel model based on reinforcement learning for urban air quality inference. The model consists of two modules: a station selector and an air quality regressor. The station selector dynamically selects the most relevant monitoring stations when inferring air quality. The air quality regressor takes in the selected stations and makes air quality inference with deep neural network. We conduct experiments on a real-world air quality dataset and our approach achieves the highest performance compared with several popular solutions, and the experiments show significant effectiveness of proposed model in tackling problems of air quality inference.
|
['Cunxiang Yin', 'Jinchang Luo', 'JiaWei He', 'Xiaohui Wu', 'Huiqiang Zhong']
|
2020-03-27
| null | null | null | null |
['air-quality-inference']
|
['miscellaneous']
|
[ 7.65166432e-02 -4.90602255e-01 -1.77920207e-01 -2.14956343e-01
-1.01428235e+00 -3.01972210e-01 2.69950211e-01 2.74645329e-01
-4.75283116e-01 1.17187595e+00 2.42089316e-01 -5.72110176e-01
-6.09030068e-01 -1.69356596e+00 -6.93639040e-01 -7.74723470e-01
3.63012284e-01 4.71094966e-01 3.29587132e-01 9.64211151e-02
-1.38052046e-01 5.18485308e-01 -1.43066573e+00 -3.32796723e-02
1.08238375e+00 8.13049495e-01 1.75570220e-01 1.00347447e+00
-3.68880555e-02 3.40555787e-01 -6.94516838e-01 3.20823967e-01
4.52850163e-01 -1.30395159e-01 -6.89528167e-01 -3.82861078e-01
4.29456085e-01 -2.66210198e-01 -3.64059433e-02 1.04738498e+00
6.13566518e-01 6.79924667e-01 3.39969873e-01 -1.16650105e+00
-2.47718184e-03 3.11376929e-01 -1.17699794e-01 7.72109807e-01
-2.53476929e-02 4.71285164e-01 9.60005641e-01 -6.43083826e-02
-3.70470077e-01 1.32936966e+00 4.88183886e-01 8.71398598e-02
-7.66809881e-01 -1.06864333e+00 4.92856413e-01 2.30238587e-01
-1.15346563e+00 -2.61053383e-01 2.98801780e-01 -2.63514102e-01
8.54900897e-01 6.89612567e-01 9.37407970e-01 3.46499562e-01
3.52317959e-01 3.17195386e-01 1.25651991e+00 5.61295450e-02
5.10156989e-01 8.24961886e-02 2.60014623e-01 5.23840129e-01
5.36042809e-01 3.92919958e-01 3.66286546e-01 -2.30249390e-01
4.09993380e-01 5.10431409e-01 -2.35955659e-02 7.27616727e-01
-6.19285703e-01 8.14105690e-01 9.62089062e-01 1.01058610e-01
-5.23230135e-01 5.73570669e-01 1.86061189e-02 3.33442956e-01
5.27709246e-01 2.44780540e-01 -4.19221818e-01 2.25675091e-01
-8.26568067e-01 5.89647830e-01 2.07357988e-01 6.00139141e-01
9.93596017e-01 -8.97855014e-02 -3.09215814e-01 7.06740856e-01
9.53838527e-01 1.28263867e+00 -2.55161151e-02 -6.74650431e-01
6.00638688e-01 3.85862172e-01 4.46364015e-01 -1.10539794e+00
-4.03767884e-01 -6.92538023e-01 -6.80205584e-01 2.78435975e-01
-3.13702822e-02 -7.03420579e-01 -7.57304430e-01 1.38649821e+00
7.25590646e-01 7.11416781e-01 -1.97573647e-01 6.57799065e-01
9.48421776e-01 1.43834317e+00 6.02855265e-01 -5.83229624e-02
1.36267996e+00 -8.06156874e-01 -7.08266199e-01 -3.06087974e-02
-2.46703833e-01 -2.15824202e-01 7.77935386e-01 1.07139759e-01
-9.58215952e-01 -8.08428526e-01 -8.23898435e-01 5.32910705e-01
-1.01260412e+00 -2.21351400e-01 3.09191465e-01 7.67871022e-01
-5.98642290e-01 3.89921457e-01 -6.38384640e-01 -1.10732719e-01
3.05541277e-01 6.10513091e-01 1.46078929e-01 1.95247307e-01
-1.65440524e+00 5.36707878e-01 3.72807622e-01 2.69861430e-01
-1.13819790e+00 -6.92782521e-01 -6.31630063e-01 1.87829304e-02
2.26200491e-01 -9.07354474e-01 1.26504278e+00 -5.26528597e-01
-1.43279541e+00 9.21573490e-02 -1.60828292e-01 -1.43480033e-01
1.36145754e-02 -3.60189170e-01 -1.07615304e+00 -6.42472357e-02
2.63223588e-01 1.27995953e-01 5.74074924e-01 -1.16177893e+00
-1.48216450e+00 -3.74049723e-01 2.84021199e-01 6.66096061e-02
-1.03766367e-01 -2.74655106e-03 -5.24029247e-02 5.02790026e-02
-4.72298920e-01 -8.38071764e-01 -5.17712355e-01 -2.60612071e-01
-1.11963101e-01 -6.85434341e-01 3.95511776e-01 -6.02622151e-01
1.61433876e+00 -1.54192030e+00 -5.37243605e-01 5.77774763e-01
-2.80320495e-02 5.82084477e-01 1.10868566e-01 1.92590520e-01
4.05017100e-02 1.40051484e-01 -1.53809160e-01 2.85402369e-02
1.01967953e-01 3.58137876e-01 -1.58268157e-02 4.76946861e-01
1.12233326e-01 6.49020970e-01 -1.27521634e+00 -4.40412432e-01
4.49493229e-01 2.63360053e-01 -3.45494241e-01 4.00402993e-01
-6.70225978e-01 5.79250216e-01 -1.06928480e+00 2.85792381e-01
9.26259220e-01 1.92912817e-02 -4.32322294e-01 3.59211192e-02
-4.51404154e-01 5.02078831e-01 -1.58242238e+00 9.92456377e-01
-8.59735072e-01 3.37051660e-01 1.25701278e-01 -5.84991813e-01
5.34529030e-01 5.64679384e-01 1.26820430e-01 -8.92885447e-01
1.91660598e-01 -1.09855264e-01 -1.72205821e-01 -9.90055799e-01
2.36741722e-01 -4.13192391e-01 7.47372359e-02 3.49250823e-01
-5.30459404e-01 7.94000030e-02 2.22960617e-02 -4.16393220e-01
8.90274346e-01 -1.64313108e-01 -4.84175198e-02 -1.44649774e-01
1.03471327e+00 -2.84949318e-02 6.61616206e-01 7.04408586e-01
-3.77347142e-01 -1.40505657e-01 -1.76167846e-01 -4.80962813e-01
-6.60102963e-01 -1.14826095e+00 -1.45230427e-01 1.17677474e+00
-1.33088920e-02 1.57526776e-01 -6.49618864e-01 -6.68506444e-01
2.34725382e-02 8.15601528e-01 -5.47417045e-01 8.04635882e-02
-5.45227170e-01 -1.05025899e+00 4.81578141e-01 3.47138852e-01
9.95984972e-01 -8.98298383e-01 -3.35663021e-01 2.17661858e-01
-1.04954623e-01 -6.36850178e-01 -7.32313767e-02 -1.75785959e-01
-7.77902484e-01 -8.75461638e-01 1.91610698e-02 -4.06734496e-01
4.56223756e-01 4.77563888e-01 1.20424354e+00 1.26642063e-01
1.25585675e-01 -6.84762001e-02 6.90477854e-03 -9.51841116e-01
-4.14748430e-01 3.48379046e-01 -1.98257476e-01 1.37344003e-01
3.23503584e-01 -5.11989236e-01 -9.46619511e-01 2.10228145e-01
-1.01266503e+00 -4.45934236e-01 4.91948426e-01 8.81177709e-02
8.36524367e-01 1.02245224e+00 9.74363983e-01 -1.00511968e+00
5.15810192e-01 -9.04850543e-01 -9.80745018e-01 -9.65768918e-02
-4.56341714e-01 1.13028824e-01 9.69657660e-01 2.68180490e-01
-1.22160423e+00 -5.96656501e-02 -7.82674134e-01 2.51461178e-01
-6.21978462e-01 1.77186519e-01 -5.29939950e-01 4.61465061e-01
3.75397146e-01 1.09957240e-03 -7.31196225e-01 -6.70944750e-01
4.81515676e-02 9.83372271e-01 8.93362612e-02 -6.05291545e-01
1.16937542e+00 4.16153133e-01 5.27532324e-02 -8.54347706e-01
-1.28248131e+00 -7.17434287e-01 -4.09335852e-01 -2.95367599e-01
1.24248672e+00 -1.16602981e+00 -8.42547238e-01 -7.10144863e-02
-8.38847876e-01 -1.61766306e-01 -3.52963693e-02 3.99314374e-01
1.73546746e-01 -2.26115778e-01 -1.46542877e-01 -9.48407412e-01
-4.92065430e-01 -1.01997960e+00 9.02833581e-01 6.09692276e-01
2.80639917e-01 -1.43945932e+00 5.65848887e-01 5.55928230e-01
5.16065001e-01 5.54393172e-01 9.04712141e-01 -1.00526989e-01
-8.55083585e-01 -9.76067185e-02 -2.58060485e-01 1.58632398e-01
4.97703612e-01 -1.46571681e-01 -1.29209518e+00 -1.88745737e-01
-2.55493671e-01 1.30757794e-01 8.97544444e-01 2.22169235e-01
1.22081900e+00 -4.80320036e-01 -3.48369718e-01 3.96261126e-01
1.66440809e+00 4.23978001e-01 5.37462234e-01 4.05035347e-01
5.48910439e-01 1.64707720e-01 4.59875256e-01 7.13110808e-03
3.28652292e-01 1.98428020e-01 7.15847731e-01 -2.71593809e-01
5.11054695e-02 -2.80950814e-01 1.21341355e-01 7.13450015e-01
-4.83217184e-03 -3.73432934e-01 -6.84515238e-01 6.07280254e-01
-1.80130279e+00 -1.29444122e+00 -4.45312947e-01 2.11696863e+00
8.61919403e-01 2.17784271e-01 2.83165216e-01 1.70462102e-01
4.63996738e-01 3.43231052e-01 -4.76610720e-01 -4.05226499e-01
6.35469377e-01 5.93859673e-01 6.56037748e-01 9.73365009e-01
-1.38177836e+00 6.07077956e-01 6.33933973e+00 5.82025766e-01
-8.65273535e-01 3.79034042e-01 3.90165299e-01 -2.42674984e-02
-3.07779402e-01 -2.03677505e-01 -1.11023283e+00 6.25573814e-01
1.36839700e+00 3.80274117e-01 2.85004437e-01 4.61365372e-01
1.03019702e+00 -1.04534261e-01 -7.92030215e-01 4.01500016e-01
-7.64250636e-01 -1.03122687e+00 1.78694487e-01 -7.28476793e-02
8.59740138e-01 2.82656103e-01 -9.31350812e-02 2.05939204e-01
8.02128553e-01 -1.28549898e+00 4.62930799e-02 1.02852201e+00
2.83352673e-01 -1.14451766e+00 7.31843054e-01 4.31415170e-01
-1.49757111e+00 -1.16476253e-01 -5.72583437e-01 -2.10670128e-01
-2.13343069e-01 7.66155005e-01 -7.49936402e-01 6.89654768e-01
6.75469816e-01 5.24785936e-01 -3.76569182e-01 1.23366821e+00
-3.63350511e-01 1.13902235e+00 -5.06642699e-01 -3.40142161e-01
5.62876344e-01 -2.16472656e-01 3.41185242e-01 1.43516219e+00
1.05992876e-01 1.88542679e-01 3.00499201e-01 8.86330664e-01
1.22808874e-01 9.21637490e-02 -5.84939182e-01 5.21928966e-01
4.83111143e-01 1.13469768e+00 -1.85173571e-01 -5.54110765e-01
-4.06721205e-01 2.18382120e-01 -8.86876360e-02 2.97632396e-01
-1.11294270e+00 -4.42174524e-01 9.46885526e-01 1.43133000e-01
1.05582424e-01 4.15515061e-03 4.84204590e-02 -4.99468058e-01
-4.69678760e-01 -8.46874714e-01 5.42400539e-01 -5.65914631e-01
-1.06038129e+00 1.00866757e-01 1.67102084e-01 -9.08005238e-01
3.70770097e-01 -4.20382082e-01 -1.17478585e+00 1.11779034e+00
-2.12621880e+00 -6.72731340e-01 -6.49166763e-01 6.13303661e-01
5.81329644e-01 1.81032017e-01 6.56882405e-01 7.53964424e-01
-1.08625555e+00 2.75220007e-01 2.89942414e-01 1.92782313e-01
1.24006212e-01 -1.41080570e+00 3.29688847e-01 7.48337507e-01
-3.78578901e-01 4.53002036e-01 6.25239074e-01 -5.82744777e-01
-1.31085896e+00 -2.00850630e+00 7.04996347e-01 -3.68424952e-01
3.39073420e-01 4.45322581e-02 -4.34100628e-01 5.31949103e-01
3.96018386e-01 -1.32146716e-01 7.68593967e-01 -1.92896754e-01
2.47983173e-01 -8.07550013e-01 -1.34841192e+00 4.00115252e-01
5.53123832e-01 -1.68288231e-01 -4.83460724e-01 6.04467809e-01
9.57440436e-01 8.16362277e-02 -9.99117315e-01 2.73399670e-02
3.31945300e-01 -6.49596155e-01 1.00775838e+00 -6.11884832e-01
1.34383336e-01 -8.84023428e-01 -6.58236817e-02 -1.64166081e+00
-7.82437682e-01 -8.25454220e-02 2.85683393e-01 1.13005471e+00
6.45267129e-01 -6.52315736e-01 6.68326855e-01 6.42744303e-01
7.11410791e-02 -4.26919520e-01 -6.79430425e-01 -4.34791625e-01
2.41447404e-01 -3.83808285e-01 1.36610579e+00 5.15692651e-01
-9.27744031e-01 5.31073213e-01 -1.99516609e-01 9.41819429e-01
6.59835815e-01 9.23051238e-02 6.97991312e-01 -1.41090858e+00
-4.66044962e-01 -1.99322402e-01 2.09961027e-01 -9.85738337e-01
7.71133378e-02 -7.66222775e-01 3.61363471e-01 -2.14618397e+00
-4.11100127e-02 -6.97439134e-01 -8.52234066e-01 3.56460452e-01
-4.72598135e-01 -1.28857955e-01 -3.63343179e-01 -5.52961469e-01
-7.11529493e-01 5.28769791e-01 1.36527717e+00 -6.18180394e-01
-4.70721692e-01 6.54805541e-01 -5.87432325e-01 6.61146820e-01
1.24659252e+00 -7.13366091e-01 -5.85390270e-01 -7.75616169e-01
4.97771710e-01 -2.09993534e-02 2.40107074e-01 -1.36509252e+00
2.31349409e-01 -7.73353338e-01 7.57066429e-01 -7.25736201e-01
8.57716650e-02 -1.32910371e+00 4.19279963e-01 6.99616432e-01
-2.94474065e-01 2.09139317e-01 3.56063932e-01 7.31800318e-01
1.38617113e-01 -2.48037830e-01 9.00678635e-01 -2.97040075e-01
-4.52164143e-01 6.55118346e-01 -7.05249906e-01 -8.39747414e-02
8.28476787e-01 1.96814865e-01 -1.70027271e-01 -1.28882453e-02
-5.26049018e-01 6.79774523e-01 -2.59527951e-01 2.31620252e-01
3.72521609e-01 -1.46829975e+00 -6.49234414e-01 2.25790560e-01
-1.05307356e-01 2.76227713e-01 -1.99814215e-02 2.66086906e-01
-4.92919207e-01 5.05727112e-01 2.05432981e-01 -2.92662233e-01
-1.11796808e+00 3.35181564e-01 9.64167058e-01 -4.55278128e-01
-3.88723969e-01 5.04897058e-01 -4.99783568e-02 -5.94430804e-01
1.04696624e-01 -8.38391185e-01 -4.08536643e-01 -2.39527002e-01
7.81396627e-01 9.08878505e-01 1.44958943e-01 -4.54578102e-01
-3.44064951e-01 6.85709238e-01 5.20030200e-01 9.21047702e-02
1.39562142e+00 -1.69718340e-01 -1.95244282e-01 5.34505069e-01
1.16097224e+00 2.24823177e-01 -1.00262845e+00 -1.97692826e-01
-3.62172157e-01 -4.04466808e-01 5.34503937e-01 -9.00734425e-01
-1.20042098e+00 8.39061320e-01 9.97301280e-01 3.64615679e-01
1.10980570e+00 -5.92152953e-01 1.37769198e+00 6.87723637e-01
1.57317892e-02 -1.19699681e+00 -3.42387855e-01 5.10962069e-01
4.14719790e-01 -1.40518999e+00 6.78458959e-02 1.31167360e-02
4.10354525e-01 6.63293302e-01 6.79786861e-01 -4.08744901e-01
1.29787552e+00 -8.78949240e-02 7.01455474e-02 -4.60017830e-01
-6.84986293e-01 -5.64509213e-01 1.06242254e-01 4.36610937e-01
1.89907640e-01 4.80769783e-01 4.75555174e-02 2.02115893e-01
-2.22303152e-01 -1.00301616e-01 -1.53688669e-01 5.14961779e-01
-1.26249266e+00 -9.58801627e-01 -9.11848009e-01 5.78930616e-01
-5.49937487e-01 9.06348526e-02 4.86672483e-02 4.51040268e-01
8.87657225e-01 1.56996620e+00 7.05283284e-02 -1.46628141e-01
6.34752035e-01 -3.99657227e-02 -4.78933714e-02 -5.88843226e-01
-1.03562117e+00 -4.23580110e-01 1.55705631e-01 -3.55633587e-01
-5.98866463e-01 -3.13316017e-01 -1.27799332e+00 -4.26812083e-01
-2.13394940e-01 5.99948347e-01 5.75119078e-01 1.11752748e+00
-3.12277302e-02 9.92767036e-01 1.02791274e+00 -1.90800771e-01
-2.42497981e-01 -7.57058620e-01 -5.15938640e-01 3.17339927e-01
9.23140585e-01 -5.20123839e-01 -1.36196345e-01 -3.71578634e-01]
|
[6.276653289794922, 2.500866174697876]
|
54b832e9-e5ef-440d-8f3f-bd419d749969
|
explain-to-me-salience-based-explainability
|
2303.11969
| null |
https://arxiv.org/abs/2303.11969v2
|
https://arxiv.org/pdf/2303.11969v2.pdf
|
Explain To Me: Salience-Based Explainability for Synthetic Face Detection Models
|
The performance of convolutional neural networks has continued to improve over the last decade. At the same time, as model complexity grows, it becomes increasingly more difficult to explain model decisions. Such explanations may be of critical importance for reliable operation of human-machine pairing setups, or for model selection when the "best" model among many equally-accurate models must be established. Saliency maps represent one popular way of explaining model decisions by highlighting image regions models deem important when making a prediction. However, examining salience maps at scale is not practical. In this paper, we propose five novel methods of leveraging model salience to explain a model behavior at scale. These methods ask: (a) what is the average entropy for a model's salience maps, (b) how does model salience change when fed out-of-set samples, (c) how closely does model salience follow geometrical transformations, (d) what is the stability of model salience across independent training runs, and (e) how does model salience react to salience-guided image degradations. To assess the proposed measures on a concrete and topical problem, we conducted a series of experiments for the task of synthetic face detection with two types of models: those trained traditionally with cross-entropy loss, and those guided by human salience when training to increase model generalizability. These two types of models are characterized by different, interpretable properties of their salience maps, which allows for the evaluation of the correctness of the proposed measures. We offer source codes for each measure along with this paper.
|
['Adam Czajka', 'Kevin Bowyer', 'Timothy Kelley', 'Christopher Sweet', 'Jacob Piland', 'Aidan Boyd', 'Patrick Tinsley', 'Colton Crum']
|
2023-03-21
| null | null | null | null |
['face-detection']
|
['computer-vision']
|
[ 4.40605819e-01 2.75257140e-01 -1.44559711e-01 -4.10445243e-01
-2.28235483e-01 -2.71805316e-01 7.60608077e-01 4.79447305e-01
-3.13925564e-01 5.13474643e-01 3.44818868e-02 -1.27298996e-01
-2.59964794e-01 -4.60155249e-01 -7.73306370e-01 -5.95155656e-01
-3.49326879e-01 2.45057732e-01 8.51799324e-02 -2.75119632e-01
4.08142447e-01 6.57453477e-01 -1.84687686e+00 3.18653941e-01
7.00095296e-01 1.12135315e+00 3.38592112e-01 4.40126240e-01
1.45481214e-01 5.36762297e-01 -4.36467499e-01 -4.53236401e-01
3.06167245e-01 -5.19711971e-01 -7.03729331e-01 1.56129912e-01
5.50406337e-01 7.92806000e-02 5.63593730e-02 1.29871321e+00
2.76768714e-01 -5.50781898e-02 8.96481335e-01 -1.31363642e+00
-6.06895387e-01 4.25914586e-01 -6.00219190e-01 3.66079777e-01
1.64028004e-01 4.42440152e-01 9.42772329e-01 -7.98641682e-01
5.70053458e-01 1.36833751e+00 5.31750441e-01 6.09686375e-01
-1.48136747e+00 -6.55814350e-01 1.61883280e-01 3.23112130e-01
-1.31983268e+00 -4.99835759e-01 8.28215897e-01 -4.97385800e-01
4.96347964e-01 4.68397111e-01 5.51883638e-01 9.23393667e-01
1.82768077e-01 3.15622061e-01 1.33967602e+00 -5.10484993e-01
3.20565641e-01 5.91602087e-01 7.65943900e-02 5.43825328e-01
3.45346838e-01 3.37476999e-01 -6.67814016e-01 -4.22188640e-02
6.52002513e-01 -3.05223495e-01 -4.21497226e-01 -5.03670573e-01
-1.06441712e+00 7.32854962e-01 7.02448070e-01 2.77105838e-01
-2.82608956e-01 8.06290135e-02 1.33583993e-01 1.18634440e-01
4.19959068e-01 7.99620569e-01 -3.39936376e-01 2.21105456e-01
-1.00504303e+00 3.09063315e-01 4.77086127e-01 6.92855418e-01
8.05886626e-01 9.47809964e-02 -1.42700717e-01 6.94968224e-01
1.89804763e-01 2.44965136e-01 4.75686520e-01 -6.39824510e-01
1.22940943e-01 4.58651751e-01 1.41046509e-01 -1.14468372e+00
-4.40020412e-01 -4.93523359e-01 -9.27699924e-01 3.93215239e-01
6.35919392e-01 1.80603445e-01 -7.67938733e-01 2.10740948e+00
1.60457969e-01 7.17373788e-02 -1.17198855e-01 1.04524004e+00
6.40015244e-01 3.99651438e-01 3.58731598e-01 -2.65752196e-01
1.52784383e+00 -6.32079005e-01 -4.75277007e-01 -3.73545766e-01
1.57802641e-01 -9.18679059e-01 1.20150805e+00 6.20571971e-02
-1.05875313e+00 -8.28871429e-01 -1.02957892e+00 2.76060253e-01
-3.05859029e-01 1.18955344e-01 3.75886440e-01 6.19923532e-01
-1.05291772e+00 9.00725245e-01 -4.41919088e-01 -5.67758858e-01
3.17247123e-01 3.82670730e-01 -2.11580172e-01 4.51080829e-01
-1.15466666e+00 1.22072923e+00 3.68986845e-01 1.35558575e-01
-8.51202309e-01 -6.47614241e-01 -6.47168398e-01 4.97292221e-01
3.12490556e-02 -4.93480295e-01 1.02771091e+00 -1.38754106e+00
-9.57789242e-01 9.17413116e-01 -2.20296919e-01 -5.58384359e-01
5.49648821e-01 9.46771875e-02 -3.86657476e-01 1.35275245e-01
-1.50258511e-01 1.02227926e+00 1.27734458e+00 -1.58930695e+00
-3.27286273e-01 -2.47014716e-01 1.62622005e-01 3.05550303e-02
-2.67566383e-01 9.38917100e-02 1.97510794e-02 -7.12917864e-01
1.93447009e-01 -1.02528381e+00 -2.51015365e-01 3.79453808e-01
-4.66794789e-01 2.23043516e-01 4.93474483e-01 -5.34009457e-01
1.11178434e+00 -2.09410334e+00 -8.65921453e-02 2.46812716e-01
1.93223104e-01 1.69956222e-01 -1.84500337e-01 1.14905894e-01
-4.91568357e-01 4.16304320e-01 -2.69938409e-01 -8.36480409e-02
3.06093097e-02 -2.00953528e-01 -2.98560977e-01 2.89687902e-01
5.93372226e-01 5.97137570e-01 -5.60364246e-01 -5.21685660e-01
1.95557326e-01 4.67544496e-01 -5.56602836e-01 1.44485936e-01
-7.75223374e-02 4.44745123e-01 -1.48949429e-01 3.32975417e-01
6.40062749e-01 -2.90531307e-01 -2.96508204e-02 -5.07969439e-01
-8.60474724e-03 -7.78647314e-04 -1.15154457e+00 8.41473818e-01
-2.15607852e-01 9.54247057e-01 -2.16799736e-01 -9.30816174e-01
1.05263782e+00 1.00678772e-01 1.74971104e-01 -6.58055305e-01
2.16420386e-02 9.02413856e-03 2.53744036e-01 -3.15462440e-01
5.38868487e-01 -3.16865832e-01 3.08441103e-01 3.58121514e-01
-1.70611605e-01 4.38405424e-02 -1.21493833e-02 -1.14737473e-01
6.38412535e-01 -1.58651158e-01 5.66590548e-01 -6.15963757e-01
4.51916039e-01 -2.84109950e-01 4.15210217e-01 6.23748362e-01
-4.42943484e-01 8.93701255e-01 6.67020261e-01 -6.83676481e-01
-1.15674722e+00 -1.07378888e+00 -3.28927845e-01 7.24430144e-01
5.21724641e-01 -2.61836927e-02 -9.71710086e-01 -3.75270367e-01
-3.29131074e-02 8.36405218e-01 -9.12604749e-01 -4.66850221e-01
-2.86156416e-01 -6.50038362e-01 2.09896401e-01 2.23191291e-01
5.41948020e-01 -1.26078784e+00 -9.03215110e-01 -2.41481349e-01
-1.32406294e-01 -9.27348137e-01 -3.27562839e-01 1.88359290e-01
-7.91250765e-01 -1.09708953e+00 -5.94164968e-01 -6.30450785e-01
9.85294759e-01 3.43799084e-01 1.27356482e+00 4.08455610e-01
-1.19458914e-01 -1.80058610e-02 -1.64338335e-01 -5.81253469e-01
-5.38839221e-01 -1.88971665e-02 2.48991519e-01 1.43070683e-01
1.55459464e-01 -3.26955140e-01 -6.10814273e-01 5.23377955e-01
-9.28906679e-01 2.54082292e-01 5.68363607e-01 8.19609821e-01
4.71610695e-01 5.30187488e-02 4.46178645e-01 -6.72001660e-01
6.71383798e-01 -3.36530805e-01 -4.56890702e-01 4.29134518e-01
-8.06506753e-01 3.07249576e-01 5.83722830e-01 -4.25357163e-01
-9.43168461e-01 -1.69861186e-02 6.05860725e-02 -4.86940444e-01
-3.34902346e-01 2.56008923e-01 -3.83438617e-02 -5.60030639e-02
9.65540588e-01 -6.92172348e-02 -3.13795893e-03 -2.30412364e-01
2.08814576e-01 2.18979448e-01 2.74751306e-01 -3.57871503e-01
1.04498100e+00 2.48259827e-01 -1.67848095e-02 -8.97135437e-01
-7.10129917e-01 -6.55741394e-02 -5.90005100e-01 -4.35435563e-01
8.12909544e-01 -5.20174384e-01 -5.79490185e-01 2.00131431e-01
-1.15362978e+00 -1.71054378e-01 -3.28493595e-01 2.45688856e-01
-4.99695957e-01 1.55137926e-01 -1.58807904e-01 -8.29889476e-01
-1.57177627e-01 -1.40116572e+00 7.99702048e-01 4.42295015e-01
-5.12866437e-01 -1.01016235e+00 -4.21247572e-01 1.29247278e-01
4.41398740e-01 3.50351363e-01 1.16407311e+00 -5.20610750e-01
-5.22795558e-01 -7.53546879e-02 -4.83053446e-01 2.00826302e-01
7.75993913e-02 1.52538985e-01 -1.26526725e+00 -3.88424873e-01
5.85863441e-02 -1.29195958e-01 7.75899351e-01 6.30571604e-01
1.26461351e+00 -4.96820390e-01 -1.56652883e-01 4.19319957e-01
1.15235293e+00 -5.44812977e-02 5.62142551e-01 3.72668058e-01
2.94802964e-01 8.91683757e-01 6.75249815e-01 2.70460665e-01
-1.00092165e-01 9.03511107e-01 6.78574204e-01 -3.81624877e-01
-6.11164756e-02 -2.88523108e-01 2.14617059e-01 3.79714668e-01
4.93655764e-02 -4.42910101e-03 -8.65042508e-01 3.34181726e-01
-1.58625054e+00 -1.02751875e+00 1.50212526e-01 2.25039196e+00
4.88704443e-01 4.21695858e-01 8.49059820e-02 5.26098050e-02
1.00200081e+00 3.00065309e-01 -6.67628288e-01 -5.31742811e-01
-9.07157809e-02 -4.51015607e-02 2.77781248e-01 3.99143219e-01
-9.87980604e-01 7.04401433e-01 6.11686516e+00 8.60140622e-01
-1.28742468e+00 -1.40116677e-01 1.17003167e+00 4.87111844e-02
-3.14075172e-01 1.45470336e-01 -6.22089148e-01 5.13802409e-01
6.28854573e-01 -3.59973788e-01 3.13695073e-01 1.00791073e+00
3.77618432e-01 -2.76370585e-01 -1.26594198e+00 8.40532720e-01
-1.42874062e-01 -1.31308079e+00 1.48239672e-01 -2.34194100e-02
6.39022112e-01 -4.29212153e-01 4.47865009e-01 -8.28717947e-02
-2.25060046e-01 -1.29595780e+00 1.00470471e+00 4.99839664e-01
6.13812268e-01 -6.18271768e-01 5.66931248e-01 2.89211571e-01
-1.10227108e+00 -9.73143950e-02 -4.93794560e-01 -8.08389485e-03
-1.86466083e-01 3.90221924e-01 -7.86081851e-01 1.02238201e-01
6.32472396e-01 3.25552493e-01 -9.42473888e-01 1.15785527e+00
-1.96030647e-01 5.65040648e-01 -1.09281152e-01 -9.04853493e-02
1.03194453e-02 1.26921251e-01 5.92268050e-01 1.05788541e+00
2.07544714e-01 -2.81693816e-01 -2.93806940e-01 1.26302671e+00
2.78228279e-02 2.37911362e-02 -4.77577478e-01 1.83750838e-01
6.13461792e-01 1.18580937e+00 -1.03229773e+00 -3.56516838e-02
7.59125361e-03 6.00892425e-01 2.47001037e-01 3.93065482e-01
-9.73768413e-01 6.96611777e-02 7.66718268e-01 3.99933368e-01
5.22308461e-02 1.07975610e-01 -5.45711577e-01 -8.37302387e-01
4.79035340e-02 -8.83188307e-01 6.46645129e-02 -9.41387415e-01
-1.09536481e+00 8.68731678e-01 2.45078817e-01 -1.43157637e+00
-1.98197708e-01 -4.38937098e-01 -7.20248342e-01 9.28504348e-01
-1.51736152e+00 -8.06200266e-01 -4.26082879e-01 1.32004157e-01
5.23814082e-01 -1.82433620e-01 6.71001554e-01 -4.42260690e-02
-4.80584711e-01 7.05675840e-01 -1.53977871e-01 -1.77410364e-01
5.46470881e-01 -9.93392229e-01 5.64002573e-01 7.95915425e-01
3.26423913e-01 6.40675843e-01 1.26170516e+00 -2.75492787e-01
-5.97080767e-01 -9.62034762e-01 8.20055842e-01 -2.89773494e-01
3.08859915e-01 -3.06088001e-01 -1.04790533e+00 2.40017861e-01
3.61203887e-02 -1.86103880e-01 3.09489787e-01 1.44619644e-01
-3.15897912e-01 -3.11753273e-01 -1.31574810e+00 7.63436019e-01
7.36550868e-01 -4.70754147e-01 -3.57565612e-01 2.20529824e-01
5.55534661e-01 -9.62817371e-02 -5.01114309e-01 5.22657394e-01
6.14831269e-01 -1.49575496e+00 9.60014284e-01 -6.91749215e-01
6.56133771e-01 -2.58368254e-01 2.41076853e-02 -1.36170852e+00
-4.65873212e-01 -3.42346847e-01 2.15486944e-01 1.13000858e+00
6.13470972e-01 -4.58833575e-01 7.79100418e-01 8.14665139e-01
1.41568854e-01 -1.01917195e+00 -8.33263636e-01 -6.02810085e-01
-1.20123420e-02 -2.53702968e-01 6.17343068e-01 9.46039021e-01
-1.73452988e-01 7.44061098e-02 -2.78306782e-01 1.95184946e-01
5.64124465e-01 -2.15193238e-02 5.94838500e-01 -1.20455015e+00
-9.72958505e-02 -6.76023126e-01 -6.89764440e-01 -5.70010602e-01
5.41331172e-02 -6.01558387e-01 -7.82947093e-02 -1.01697397e+00
3.67478371e-01 -3.73815238e-01 -3.74713033e-01 2.34255195e-01
-3.28536838e-01 1.93881571e-01 5.25411487e-01 4.44042623e-01
-2.05352187e-01 4.68171507e-01 1.13482893e+00 -7.65713602e-02
5.01697548e-02 -6.84032589e-02 -7.00967908e-01 8.48629057e-01
8.97682369e-01 -3.81208360e-01 -4.55786556e-01 -1.29127324e-01
6.25235215e-02 -8.36202726e-02 6.70634031e-01 -1.29724264e+00
-4.14486527e-02 -3.07665914e-01 5.03249824e-01 -4.66119535e-02
3.81134301e-01 -8.93119156e-01 2.13588238e-01 5.87951481e-01
-7.46010482e-01 2.14352265e-01 3.12935919e-01 3.04003894e-01
-2.18117431e-01 -3.30370694e-01 1.37568045e+00 5.17985262e-02
-7.70156741e-01 7.58557841e-02 -1.25714675e-01 7.80921755e-03
9.13106501e-01 -4.60012555e-01 -1.56597301e-01 -5.63787758e-01
-7.51293600e-01 -2.09951207e-01 7.13120222e-01 5.52748680e-01
5.70710540e-01 -1.33569038e+00 -5.21454573e-01 2.60356218e-01
3.24484617e-01 -4.60645854e-01 2.12849841e-01 6.95972204e-01
-3.66624981e-01 2.25249484e-01 -3.78100365e-01 -6.41705751e-01
-1.46927643e+00 5.53695560e-01 5.93984842e-01 -9.39304605e-02
-8.23512673e-02 9.11023974e-01 6.14727676e-01 3.71037931e-05
2.42644697e-01 -3.65395069e-01 -2.28165269e-01 2.10671388e-02
3.17057371e-01 1.28314003e-01 -1.43152997e-01 -8.97865295e-01
-1.81384563e-01 5.50991893e-01 -6.83248788e-02 1.60962656e-01
1.01069498e+00 -3.81330587e-02 1.15466550e-01 2.91739613e-01
8.46141040e-01 -4.37938064e-01 -1.44868863e+00 4.39088345e-02
9.49305817e-02 -4.68796223e-01 -1.06548078e-01 -7.38501370e-01
-1.01748133e+00 1.08081567e+00 9.49413896e-01 4.49247986e-01
1.21155274e+00 -1.17179818e-01 1.57880068e-01 6.13355115e-02
2.25724548e-01 -9.77387190e-01 2.04886407e-01 1.64871499e-01
1.23849487e+00 -1.39674544e+00 2.57589351e-02 -3.93604904e-01
-7.80750573e-01 1.06150842e+00 6.95722103e-01 6.70526773e-02
7.10060894e-01 -1.62386283e-01 2.96965446e-02 -2.26410553e-01
-7.35511005e-01 -1.83677852e-01 5.79038382e-01 5.08531988e-01
4.03787225e-01 7.54033327e-02 -8.28002319e-02 3.90560269e-01
-4.47866350e-01 -2.89057761e-01 3.43770891e-01 3.86920571e-01
-5.98762572e-01 -5.80073237e-01 -4.63548392e-01 5.28716266e-01
-3.32723707e-02 -1.06295809e-01 -5.40557981e-01 8.97921562e-01
2.73526251e-01 8.87392104e-01 -2.05370039e-02 -4.79612172e-01
3.80509287e-01 1.74884277e-03 2.64731228e-01 -3.66342902e-01
-5.08675456e-01 -3.59978676e-01 -1.29705980e-01 -3.79755467e-01
-2.48862356e-01 -7.82829702e-01 -9.57356751e-01 -3.21442306e-01
-5.21375239e-01 1.36559874e-01 6.41548574e-01 9.46413279e-01
2.01499030e-01 2.87217826e-01 7.27123201e-01 -9.56890225e-01
-6.70223713e-01 -9.63832796e-01 -4.22719419e-01 7.63942122e-01
2.80543804e-01 -9.17314529e-01 -4.85927850e-01 1.40514702e-01]
|
[9.939043045043945, 2.1740570068359375]
|
1eb63d36-db9e-4703-8ce3-8a77c9230d2f
|
date-dual-attentive-tree-aware-embedding-for
| null | null |
https://dl.acm.org/doi/10.1145/3394486.3403339
|
https://dl.acm.org/doi/pdf/10.1145/3394486.3403339
|
DATE: Dual Attentive Tree-aware Embedding for Customs Fraud Detection
|
Intentional manipulation of invoices that lead to undervaluation of trade goods is the most common type of customs fraud to avoid ad valorem duties and taxes. To secure government revenue without interrupting legitimate trade flows, customs administrations around the world strive to develop ways to detect illicit trades. This paper proposes DATE, a model of Dual-task Attentive Tree-aware Embedding, to classify and rank illegal trade flows that contribute the most to the overall customs revenue when caught. The strength of DATE comes from combining a tree-based model for interpretability and transaction-level embeddings with dual attention mechanisms. To accurately identify illicit transactions and predict tax revenue, DATE learns simultaneously from illicitness and surtax of each transaction. With a five-year amount of customs import data with a test illicit ratio of 2.24%, DATE shows a remarkable precision of 92.7% on illegal cases and a recall of 49.3% on revenue after inspecting only 1% of all trade flows. We also discuss issues on deploying DATE in Nigeria Customs Service, in collaboration with the World Customs Organization.
|
['Cheng-Te Li', 'Yu-Che Tsai', 'Karandeep Singh', 'Etim Ibok', 'Yeonsoo Choi', 'Sundong Kim', 'Meeyoung Cha']
|
2020-08-23
| null | null | null |
kdd-2020-8
|
['value-prediction', 'multi-target-regression']
|
['computer-code', 'miscellaneous']
|
[-4.18016404e-01 -1.41008884e-01 -7.47260213e-01 -3.12133431e-01
-6.12652779e-01 -1.11993515e+00 5.87069035e-01 3.72948676e-01
-4.22447324e-01 4.36341584e-01 4.35133129e-01 -1.17821968e+00
-2.71955192e-01 -1.10202134e+00 -5.17749310e-01 -3.94336939e-01
-1.93427548e-01 7.69650161e-01 -5.08282125e-01 -3.62895876e-01
6.96545303e-01 6.05684519e-01 -6.78977311e-01 5.13980865e-01
7.12502956e-01 1.00743902e+00 -4.61847126e-01 5.22882164e-01
-1.20647758e-01 7.76992738e-01 -6.14692867e-01 -1.12402189e+00
1.13542485e+00 3.24949682e-01 -8.30214202e-01 -5.23722135e-02
5.16314685e-01 -8.43628705e-01 -3.30354661e-01 1.07138944e+00
-1.04472019e-01 -4.61476177e-01 1.05938780e+00 -1.20612097e+00
-1.59324312e+00 6.91833615e-01 -9.98765409e-01 7.30734527e-01
5.14421649e-02 3.77948791e-01 1.55866039e+00 -8.23972225e-01
5.03559530e-01 1.05116212e+00 7.99615145e-01 -7.58282244e-02
-1.40132296e+00 -9.73745584e-01 -1.31276220e-01 -2.14982163e-02
-7.54255950e-01 -2.80067056e-01 4.92243916e-01 -5.22420704e-01
1.31660211e+00 1.24947309e-01 5.18591285e-01 8.00854206e-01
6.26827121e-01 4.94662613e-01 1.31174600e+00 -6.54780716e-02
-3.34682316e-01 3.65584999e-01 2.81464458e-01 7.35874295e-01
1.10860324e+00 1.58082455e-01 -1.80928126e-01 -4.03415650e-01
7.25374222e-01 2.80724406e-01 3.43153715e-01 9.65168327e-02
-7.85064518e-01 1.34442115e+00 6.30568147e-01 3.06604773e-01
-5.98385513e-01 1.55083716e-01 6.16995513e-01 7.54029989e-01
4.83581603e-01 5.76505065e-01 -8.78944397e-01 3.43358936e-03
-7.93356478e-01 3.13310027e-01 8.29061329e-01 5.89962304e-01
4.49480444e-01 3.60533416e-01 4.17341769e-01 5.11415601e-01
5.90369582e-01 7.85073161e-01 2.89098769e-01 -3.99819851e-01
1.08088720e+00 9.52148855e-01 3.12184304e-01 -1.20788574e+00
1.62188366e-01 -3.01007599e-01 -5.39107859e-01 2.72295922e-01
8.27920914e-01 -1.07052187e-02 -1.16101015e+00 9.73284602e-01
-1.49986133e-01 -6.67758524e-01 3.72174531e-02 7.33359814e-01
-2.91310996e-01 5.29295921e-01 2.11567506e-01 2.55001098e-01
1.65109909e+00 -3.90881360e-01 -6.05126023e-01 -1.86844207e-02
5.48775434e-01 -7.48193502e-01 9.47252989e-01 3.60194266e-01
-8.44992340e-01 1.36815384e-01 -8.51079583e-01 -1.10769324e-01
-1.06539309e+00 -1.96874797e-01 1.21681690e+00 7.48081863e-01
-3.07309717e-01 5.62078893e-01 -7.72729874e-01 1.84227172e-02
9.81438220e-01 4.57491189e-01 -3.42145205e-01 1.89359978e-01
-1.12390423e+00 1.25968385e+00 8.13399479e-02 1.11997440e-01
-6.39952958e-01 -7.89928377e-01 -6.89922631e-01 1.93569407e-01
1.70423478e-01 1.04212709e-01 9.93564665e-01 -1.01793563e+00
-3.03372771e-01 1.14312553e+00 3.58563572e-01 -9.40502703e-01
3.46127957e-01 -1.46712348e-01 -8.45533431e-01 -1.79438174e-01
5.69696784e-01 9.74868387e-02 5.28387606e-01 -8.20759296e-01
-9.67354774e-01 -8.39012802e-01 1.52851462e-01 -8.54202136e-02
-1.56617001e-01 2.03020781e-01 5.86692810e-01 -8.56534600e-01
-1.90024711e-02 -6.96851611e-01 6.57047108e-02 -2.76959598e-01
-9.45131034e-02 -1.90344572e-01 9.68775690e-01 -1.10655534e+00
1.32702625e+00 -1.60405612e+00 -6.02414429e-01 2.68530846e-01
5.43931574e-02 4.23509508e-01 2.58213460e-01 5.89563489e-01
-1.70206875e-01 6.86017394e-01 -1.58674002e-01 4.66603011e-01
2.97956288e-01 3.31055403e-01 -7.89833546e-01 6.44013345e-01
4.75055963e-01 1.00636816e+00 -8.67561162e-01 -6.74547479e-02
2.25257024e-01 -4.91809957e-02 -4.12680328e-01 -2.60623276e-01
9.68419760e-02 -3.52694839e-01 -4.05405760e-01 1.62508380e+00
8.19477975e-01 2.29587834e-02 3.91673356e-01 -2.52511561e-01
-2.67196298e-01 7.72478223e-01 -7.99810827e-01 8.46884191e-01
-3.18316132e-01 5.43403268e-01 2.28827193e-01 -9.02549028e-01
4.35987890e-01 1.21003710e-01 3.24665427e-01 -9.29922104e-01
1.60042107e-01 4.55305636e-01 1.62894353e-01 -4.26295787e-01
8.03292155e-01 -3.96715462e-01 -5.82721531e-01 8.59677315e-01
2.85558011e-02 4.49980855e-01 1.28402248e-01 1.97479025e-01
8.74868631e-01 7.45206093e-03 5.41288793e-01 -4.35485810e-01
3.67500968e-02 5.52948594e-01 6.60204291e-01 3.49628866e-01
-4.22223449e-01 -3.67721945e-01 7.43957520e-01 -1.15693247e+00
-1.31995225e+00 -1.32400978e+00 -2.27916852e-01 1.09103322e+00
-3.65375549e-01 2.80942112e-01 -1.75596073e-01 -1.37627494e+00
8.77515852e-01 6.76646709e-01 -6.58325613e-01 2.81250060e-01
-5.65152466e-01 -9.90523756e-01 7.19560742e-01 6.35803461e-01
6.38625681e-01 -7.68151939e-01 -6.10676050e-01 3.59163135e-01
7.20458776e-02 -4.72050428e-01 -5.66038966e-01 5.73477805e-01
-9.18106973e-01 -1.42658436e+00 -2.45861843e-01 -5.18473089e-01
6.73433483e-01 7.65190125e-02 8.77006233e-01 1.43403858e-02
-3.24282944e-01 -3.39248002e-01 1.52530633e-02 -5.25111735e-01
-2.77334720e-01 8.94788951e-02 2.60672301e-01 -1.63852721e-01
1.07065141e+00 -4.25063550e-01 -3.78144681e-01 -1.64936721e-01
-9.54853296e-01 -3.41919571e-01 6.39363170e-01 8.41578841e-01
-4.49995697e-02 1.74982086e-01 5.82335353e-01 -1.19489121e+00
7.22388506e-01 -8.80189538e-01 -9.05186594e-01 1.98570386e-01
-1.23573029e+00 5.13964659e-03 6.40238047e-01 -2.69586146e-01
-9.20445800e-01 -4.11024064e-01 2.08091035e-01 -1.98044300e-01
-6.57536002e-05 3.88551742e-01 3.60537946e-01 7.13414252e-02
3.43848467e-01 -1.23794796e-02 -1.75964192e-01 -5.83410740e-01
6.26561418e-02 1.05369687e+00 3.57063711e-01 -3.10790002e-01
1.05579245e+00 3.71355951e-01 -2.83700317e-01 -4.06457275e-01
-5.54366946e-01 -5.48917532e-01 -6.21413529e-01 2.67839074e-01
4.99547809e-01 -6.23734415e-01 -1.02036273e+00 4.29033905e-01
-9.90668118e-01 3.74280810e-02 5.52350879e-02 5.38924277e-01
-8.63265023e-02 1.17647119e-01 -1.01348293e+00 -1.28968716e+00
-4.32353973e-01 -7.04661727e-01 5.83184063e-01 -1.91883475e-01
-2.73809344e-01 -1.22321057e+00 -1.23334691e-01 7.44237602e-01
5.18323958e-01 4.11694854e-01 1.24862778e+00 -1.07017684e+00
-6.18683457e-01 -8.86509418e-02 -5.14827311e-01 4.86648500e-01
5.93979001e-01 -2.30545461e-01 -6.01870954e-01 -2.97417670e-01
7.43941441e-02 -2.30492532e-01 7.60428429e-01 -4.48045768e-02
3.44672203e-01 -1.05956304e+00 -1.57666598e-02 4.65309143e-01
2.05166388e+00 6.16923034e-01 4.06852335e-01 6.49539471e-01
5.63166678e-01 5.94985306e-01 6.61122441e-01 2.99080163e-01
3.34348410e-01 1.49849832e-01 3.93676549e-01 1.40830606e-01
3.65541250e-01 -5.37219048e-01 5.85704923e-01 4.70715374e-01
-1.26195148e-01 -1.45463347e-01 -1.02127171e+00 1.01707411e+00
-1.24336123e+00 -1.21899295e+00 -1.96943730e-01 2.00234365e+00
5.70839763e-01 2.37091452e-01 1.24835208e-01 2.39890590e-01
5.27760446e-01 7.50165433e-02 -4.47275788e-01 -1.17194247e+00
2.24801660e-01 1.96136653e-01 1.76157558e+00 5.68192005e-01
-1.02664435e+00 1.15523553e+00 6.23384476e+00 4.48700339e-01
-8.77735078e-01 3.55368890e-02 6.49481058e-01 1.51879443e-02
-5.78375101e-01 5.51188625e-02 -7.78820693e-01 8.20388258e-01
9.80591357e-01 -1.98689565e-01 6.67401016e-01 5.74328184e-01
1.65095046e-01 6.63557416e-03 -7.34677017e-01 3.47561628e-01
-1.81908324e-01 -1.38359869e+00 4.46196467e-01 7.75630057e-01
5.99596858e-01 -6.29994646e-03 4.00037706e-01 3.64450306e-01
8.69526565e-01 -9.87222791e-01 7.48866916e-01 9.51497406e-02
7.63417721e-01 -8.57722938e-01 9.06168222e-01 -8.14593676e-03
-1.05386055e+00 -5.47541559e-01 -6.99394196e-03 -2.89798975e-01
1.58753186e-01 7.28509128e-02 -1.05238938e+00 3.22883815e-01
7.01608062e-01 4.98440355e-01 -1.80267602e-01 1.65714681e-01
-7.45017603e-02 6.63085341e-01 -3.34261566e-01 2.45984361e-01
1.08722126e+00 -5.14196455e-01 2.39971578e-01 1.21196234e+00
-7.78236054e-03 2.68180519e-02 -2.71443993e-01 9.65218663e-01
-4.58034545e-01 -2.24742219e-01 -1.05658269e+00 -5.59188426e-01
4.43926930e-01 8.41262698e-01 -5.71046054e-01 -2.76927322e-01
-6.44670606e-01 8.12938213e-01 1.55679986e-01 1.14316076e-01
-8.81895602e-01 -6.04068637e-01 1.10500133e+00 3.69426817e-01
3.99465442e-01 -2.94699252e-01 -3.99807334e-01 -1.12256324e+00
9.84422863e-02 -8.54694724e-01 5.53140879e-01 -2.54168361e-01
-1.33987474e+00 -1.33472458e-01 -8.90072435e-02 -8.73572052e-01
-2.52418518e-02 -8.70344222e-01 -4.05247033e-01 1.08520329e+00
-1.73318875e+00 -1.41360521e+00 6.90349281e-01 3.96219164e-01
7.35438049e-01 -3.33727866e-01 6.30395412e-01 3.16081762e-01
-1.16984926e-01 4.99509841e-01 8.44345242e-02 6.93191111e-01
3.37736696e-01 -1.52946150e+00 8.00769210e-01 6.53482676e-01
4.36332375e-02 1.11048830e+00 2.66420811e-01 -1.07300544e+00
-1.58040988e+00 -1.10346282e+00 1.37977695e+00 -7.16998160e-01
1.22583556e+00 -3.75832558e-01 -7.86126077e-01 1.38295460e+00
1.60960495e-01 -4.15434748e-01 6.80397153e-01 1.45845041e-01
-7.31571555e-01 -2.46776938e-01 -1.80360079e+00 2.49037549e-01
6.97628617e-01 -6.99090421e-01 -1.27693629e+00 2.92070359e-01
3.43829364e-01 1.63627923e-01 -1.06203222e+00 -2.17225119e-01
1.00171459e+00 -7.72749066e-01 7.99712360e-01 -1.01843667e+00
2.07768992e-01 4.48649526e-02 -2.78759122e-01 -9.90140617e-01
-4.06619281e-01 -3.90984952e-01 6.41410500e-02 1.22856593e+00
5.24044394e-01 -1.03967321e+00 8.74940813e-01 6.62179291e-01
2.55950630e-01 -4.96950209e-01 -8.95906389e-01 -8.06601107e-01
2.02427268e-01 -2.60642797e-01 6.11539423e-01 1.47183609e+00
-8.72110501e-02 1.65425800e-02 -3.56019199e-01 2.74580628e-01
7.72189677e-01 1.74762338e-01 3.75299633e-01 -9.55536962e-01
1.93873852e-01 -2.22808257e-01 -2.85610050e-01 -4.69834566e-01
-1.75466076e-01 -1.07363796e+00 -7.58070469e-01 -1.20775092e+00
1.92353815e-01 -4.94362354e-01 -4.25293803e-01 7.56922185e-01
6.27157688e-01 1.53279915e-01 2.68090636e-01 4.19208050e-01
5.18989004e-02 -2.33945832e-01 8.58003139e-01 -3.48248005e-01
4.96766977e-02 -2.32775018e-01 -1.33093619e+00 7.31335640e-01
9.22104597e-01 -7.83500612e-01 6.01926446e-02 -8.20830762e-01
2.62187868e-01 -2.56737154e-02 2.48662725e-01 -1.07384518e-01
1.12845182e-01 -5.46222448e-01 5.05781293e-01 -8.56969893e-01
3.09960432e-02 -1.29233563e+00 9.83611643e-02 8.78671110e-01
-2.57225484e-01 5.94601631e-01 -3.32193822e-02 6.35291874e-01
-1.52202602e-02 -1.44302398e-01 6.26660466e-01 -5.14733613e-01
-3.72339576e-01 1.97508141e-01 -6.88312709e-01 -1.66027784e-01
1.01951611e+00 -3.89763087e-01 -3.30467850e-01 1.78395644e-01
-2.91322917e-01 3.92596632e-01 4.67212439e-01 5.30067682e-01
2.90783614e-01 -1.28655434e+00 -7.31968939e-01 4.68693137e-01
-1.02911681e-01 -6.75596774e-01 -4.14024144e-01 4.94029760e-01
-8.04474711e-01 8.43412936e-01 -5.68264902e-01 1.86682165e-01
-9.78079557e-01 4.47591692e-01 7.74417222e-02 -6.35546029e-01
-2.22273633e-01 3.50026309e-01 -3.96851152e-01 -2.42744863e-01
-1.82636261e-01 -6.62953258e-01 2.84771472e-02 4.45072889e-01
3.45155180e-01 6.45685673e-01 -1.82235371e-02 -7.22556114e-01
-4.55703616e-01 2.92054862e-01 -5.44139862e-01 -1.95094064e-01
1.57094705e+00 1.30100012e-01 -8.07718560e-02 1.87866151e-01
1.32076943e+00 2.11057708e-01 -8.40189099e-01 1.15344763e-01
3.95052671e-01 -9.93291795e-01 1.19057886e-01 -1.38287795e+00
-1.18920875e+00 6.27372980e-01 2.27570385e-01 4.40266788e-01
7.19484508e-01 -5.01235425e-01 1.15911531e+00 4.24472123e-01
5.66249371e-01 -1.23802209e+00 -5.80220997e-01 2.93901771e-01
6.23614430e-01 -1.29674828e+00 8.02538469e-02 3.41088742e-01
-6.35232270e-01 9.28273022e-01 2.47757465e-01 -5.77714980e-01
3.38359833e-01 2.56701529e-01 -3.48647423e-02 -4.75252867e-01
-7.29334652e-01 7.32125714e-02 -3.88539076e-01 6.05558991e-01
3.52110386e-01 3.50716263e-01 -5.39255559e-01 5.94656646e-01
8.17344487e-02 -1.21543027e-01 5.61130762e-01 8.35469246e-01
-4.22085285e-01 -9.90513325e-01 -4.77835149e-01 1.05985081e+00
-1.43201697e+00 -4.10472095e-01 -6.02997363e-01 1.39901912e+00
2.94014990e-01 7.03976035e-01 3.91807199e-01 -2.13558316e-01
3.13467890e-01 1.46817893e-01 2.14872152e-01 -4.12094325e-01
-1.48443186e+00 -2.42147908e-01 3.51269871e-01 -2.55923539e-01
-7.18312338e-02 -8.12526584e-01 -1.03163731e+00 -1.14604402e+00
-5.03900886e-01 2.56687611e-01 8.63966763e-01 3.51581901e-01
1.29006043e-01 -2.54169255e-01 8.51744354e-01 -2.61885554e-01
-9.91133094e-01 -9.48137760e-01 -9.76363242e-01 6.05793595e-01
4.32693541e-01 -6.27577901e-01 -6.81844234e-01 -7.57198855e-02]
|
[7.336564064025879, 5.875798225402832]
|
5c62b8b3-6c90-4531-b316-3631390f4a13
|
noise-robust-morphological-disambiguation-for
| null | null |
https://aclanthology.org/N18-1087
|
https://aclanthology.org/N18-1087.pdf
|
Noise-Robust Morphological Disambiguation for Dialectal Arabic
|
User-generated text tends to be noisy with many lexical and orthographic inconsistencies, making natural language processing (NLP) tasks more challenging. The challenging nature of noisy text processing is exacerbated for dialectal content, where in addition to spelling and lexical differences, dialectal text is characterized with morpho-syntactic and phonetic variations. These issues increase sparsity in NLP models and reduce accuracy. We present a neural morphological tagging and disambiguation model for Egyptian Arabic, with various extensions to handle noisy and inconsistent content. Our models achieve about 5{\%} relative error reduction (1.1{\%} absolute improvement) for full morphological analysis, and around 22{\%} relative error reduction (1.8{\%} absolute improvement) for part-of-speech tagging, over a state-of-the-art baseline.
|
['er', 'Alex Erdmann', 'Nizar Habash', 'Nasser Zalmout']
|
2018-06-01
| null | null | null |
naacl-2018-6
|
['lexical-normalization', 'morphological-disambiguation', 'morphological-tagging']
|
['natural-language-processing', 'natural-language-processing', 'natural-language-processing']
|
[-1.59002244e-02 -2.85439901e-02 2.16284543e-01 -4.57800299e-01
-1.22898614e+00 -9.03099477e-01 1.35047957e-01 6.73784494e-01
-8.43741059e-01 8.85409594e-01 5.36454439e-01 -2.49759525e-01
2.07059562e-01 -7.91083395e-01 -4.45470273e-01 -6.16801918e-01
5.99640831e-02 6.09280229e-01 -1.13715805e-01 -5.77373207e-01
6.80660307e-02 2.66344160e-01 -1.09722757e+00 4.30048794e-01
1.10610497e+00 6.03947282e-01 3.13143551e-01 3.72723848e-01
-6.44856453e-01 3.54680955e-01 -1.01121438e+00 -7.90416896e-01
-8.08875263e-02 -3.00456852e-01 -9.70059991e-01 6.26663715e-02
6.47892416e-01 2.06196755e-01 4.95740287e-02 1.50260627e+00
7.68191814e-01 3.56351614e-01 6.04309082e-01 -3.66334200e-01
-1.03281939e+00 9.66610670e-01 -5.09770989e-01 2.10488275e-01
2.40526795e-01 -1.75779089e-01 1.16326094e+00 -1.13349903e+00
6.01987898e-01 1.25307810e+00 8.13569248e-01 5.05984485e-01
-1.08374715e+00 -5.67817628e-01 2.28076950e-01 -2.46850416e-01
-1.46205747e+00 -4.52566445e-01 3.04472387e-01 -3.83507721e-02
1.23322868e+00 3.59395593e-02 2.46972203e-01 7.42238641e-01
2.49006730e-02 7.15039849e-01 1.15004444e+00 -7.35927761e-01
1.73093542e-01 -1.18649445e-01 1.62346646e-01 6.55513823e-01
3.51250380e-01 -5.61025560e-01 -2.94152200e-01 8.63484573e-03
4.84574527e-01 -5.49469054e-01 -2.80478206e-02 6.58739567e-01
-9.66813922e-01 8.27036202e-01 -4.52091135e-02 4.61755663e-01
-3.11540335e-01 -3.37930530e-01 3.02365184e-01 2.04089105e-01
4.62045699e-01 8.04201126e-01 -7.66499639e-01 -4.62181270e-01
-1.08615959e+00 1.01515427e-01 9.03273940e-01 8.98695588e-01
5.77191114e-01 5.56177318e-01 1.36816353e-01 1.61336434e+00
1.19421579e-01 9.60715413e-01 7.30985463e-01 -8.14048648e-01
7.11737275e-01 3.35481346e-01 9.12713110e-02 -9.23615515e-01
-5.99264205e-01 -3.60256553e-01 -8.03123355e-01 -2.39609450e-01
7.39790380e-01 -5.37756026e-01 -1.16017973e+00 1.77633238e+00
5.44165634e-02 -7.78496265e-01 2.24211052e-01 4.31843519e-01
7.58444071e-01 8.98191214e-01 3.19860607e-01 -3.77433747e-01
1.67203712e+00 -6.44800544e-01 -9.01969433e-01 -7.65105009e-01
5.72281063e-01 -1.53900504e+00 1.44022036e+00 3.97023946e-01
-1.36837471e+00 -2.55076438e-01 -7.50762582e-01 -1.50927007e-01
-5.06235600e-01 2.65974194e-01 7.58032426e-02 8.76928985e-01
-9.71139014e-01 4.17807102e-01 -7.56196618e-01 -2.79085636e-01
1.07803471e-01 3.65759999e-01 -3.93205404e-01 -1.56362787e-01
-1.22044396e+00 8.93878400e-01 4.16468263e-01 -1.77221522e-02
-1.23325452e-01 -6.34566784e-01 -1.17224038e+00 -1.00618966e-01
3.99283677e-01 1.85168296e-01 1.28835404e+00 -7.33640969e-01
-1.29894984e+00 9.41224456e-01 -3.86368185e-01 -1.71540648e-01
-2.06468314e-01 -3.22082341e-01 -8.92416239e-01 -2.67059743e-01
2.55184203e-01 5.50678074e-01 2.61047274e-01 -1.16447496e+00
-7.39582360e-01 -3.78993392e-01 -4.33975428e-01 3.48385662e-01
-2.25393668e-01 6.78473592e-01 -3.61696571e-01 -1.30988503e+00
3.67864221e-01 -8.96443784e-01 -2.12302133e-01 -6.28993332e-01
-8.73543229e-03 -2.10385725e-01 4.08511400e-01 -1.41114581e+00
1.43328464e+00 -1.97535765e+00 -2.16429278e-01 1.52486891e-01
-2.80742973e-01 5.67839086e-01 -4.16478187e-01 3.69430810e-01
3.62170547e-01 4.49521929e-01 -4.38635975e-01 -3.15296590e-01
-6.57729432e-02 5.36407709e-01 2.44515762e-03 1.39320076e-01
5.95057487e-01 6.00568891e-01 -1.02954423e+00 -2.34635904e-01
-1.81350797e-01 4.55405295e-01 -3.94952714e-01 -2.95065463e-01
-1.02206960e-01 1.40396222e-01 1.08289257e-01 9.81224477e-01
8.51064205e-01 4.37745601e-01 4.58713949e-01 1.53654307e-01
-2.52245963e-01 9.28351045e-01 -1.51687622e+00 1.43520796e+00
-6.33943796e-01 3.54206681e-01 4.28789824e-01 -6.59651637e-01
1.07502198e+00 2.63714671e-01 6.19827732e-02 -7.37470269e-01
1.37457922e-01 6.02626622e-01 2.48719424e-01 -2.57279485e-01
1.17796564e+00 -2.29502961e-01 -4.73324865e-01 3.14731717e-01
2.92364657e-01 -2.98049062e-01 3.19403350e-01 -9.92430598e-02
9.16600108e-01 -1.66345581e-01 4.33656514e-01 -6.22317910e-01
1.93171546e-01 1.38445809e-01 9.47123230e-01 5.98298192e-01
-1.48966715e-01 7.82494307e-01 3.00657034e-01 -1.43553689e-01
-9.65792298e-01 -8.59513044e-01 -2.54053861e-01 1.33463013e+00
-2.32060686e-01 -4.43775594e-01 -9.16298926e-01 -7.04570293e-01
-2.45109081e-01 7.95918286e-01 -2.55217433e-01 2.40752727e-01
-1.24644601e+00 -1.27343202e+00 1.08157969e+00 6.68648243e-01
4.21091616e-01 -1.39993095e+00 1.14861354e-01 6.71230018e-01
-6.29154384e-01 -1.24199867e+00 -4.94729608e-01 4.98596370e-01
-6.72483444e-01 -6.01163208e-01 -4.20324624e-01 -1.30341637e+00
4.91111159e-01 7.54557624e-02 1.40442085e+00 5.22209629e-02
1.16123497e-01 -1.81638509e-01 -7.29738832e-01 -6.75071776e-01
-7.56820738e-01 1.68527395e-01 1.83387235e-01 -4.77128804e-01
7.49261200e-01 -7.81313330e-02 -1.97951496e-01 1.64892405e-01
-1.02790403e+00 -7.28581429e-01 5.56099653e-01 9.71450388e-01
8.66669118e-01 1.67667568e-01 5.22127032e-01 -1.08662307e+00
5.30623198e-01 -2.81409413e-01 -4.11911011e-01 1.06912173e-01
-4.34440047e-01 6.98082000e-02 8.16115558e-01 -2.05880940e-01
-1.45342195e+00 -1.93147380e-02 -7.22216129e-01 5.44922471e-01
-4.50936556e-01 6.81115508e-01 -5.41979790e-01 3.29288453e-01
8.24627101e-01 5.50360046e-02 -2.59811789e-01 -7.92116165e-01
3.43647778e-01 9.06196535e-01 7.56322682e-01 -6.52349412e-01
4.82343256e-01 1.61955208e-01 -4.49439883e-01 -1.08665156e+00
-9.08557296e-01 -5.26509285e-01 -7.42478728e-01 4.35274750e-01
7.87134886e-01 -9.07793880e-01 -5.32011967e-03 6.42271876e-01
-1.06766510e+00 -3.94953251e-01 -1.63210824e-01 3.29765618e-01
6.45203963e-02 6.10347748e-01 -1.02868795e+00 -6.34513199e-01
-4.09207880e-01 -1.03962374e+00 1.00024140e+00 1.42575562e-01
-5.23095131e-01 -9.56884205e-01 -1.23340070e-01 3.88282955e-01
2.44959339e-01 -1.21504389e-01 1.16373527e+00 -8.92550826e-01
1.82814926e-01 5.61712980e-02 -9.86087173e-02 6.63406134e-01
3.28821570e-01 -7.14546442e-02 -7.02235878e-01 -1.84085086e-01
-7.62973949e-02 -2.26227194e-01 5.24429262e-01 2.53046423e-01
5.91034472e-01 -4.91387844e-01 1.68131784e-01 5.96587360e-02
1.31462109e+00 3.40605021e-01 5.19816875e-01 2.93226898e-01
5.23096502e-01 7.04792440e-01 8.06473732e-01 3.14127982e-01
4.02544349e-01 3.92151088e-01 1.02146603e-01 1.64269999e-01
-3.66697103e-01 -1.25799812e-02 6.06198430e-01 1.18786585e+00
2.21383005e-01 -3.66676033e-01 -1.29609227e+00 9.41571951e-01
-1.31835210e+00 -6.85751081e-01 -3.16128433e-01 2.06880498e+00
1.48579729e+00 -1.34761810e-01 -8.35951790e-02 1.56615108e-01
8.44948292e-01 1.17829435e-01 1.80239096e-01 -6.90543234e-01
-6.76895142e-01 7.40990222e-01 4.07983541e-01 6.73960328e-01
-1.16047370e+00 1.51813567e+00 5.92578077e+00 1.09507310e+00
-1.02768397e+00 -6.35584723e-03 4.15520787e-01 1.65625528e-01
-2.12936148e-01 -2.77566612e-01 -1.16340995e+00 4.33761328e-01
1.09455168e+00 3.85542274e-01 2.58470178e-01 5.84304810e-01
1.97248727e-01 -3.71563673e-01 -5.34545481e-01 7.05068171e-01
1.94130719e-01 -9.76051629e-01 -1.11673241e-02 -6.15343228e-02
1.02122569e+00 2.68466562e-01 -9.72136930e-02 2.00782284e-01
8.01696956e-01 -9.95948255e-01 8.24605644e-01 -2.65386552e-01
7.04534650e-01 -1.07460666e+00 1.12750471e+00 8.73533487e-02
-1.02649581e+00 2.69204050e-01 -5.84308147e-01 -1.17584141e-02
3.67808849e-01 7.05894232e-01 -6.60377622e-01 2.36409158e-01
9.55126464e-01 8.66418704e-02 -5.78797519e-01 8.02708149e-01
-6.05249286e-01 1.01323390e+00 -2.99375772e-01 -2.31628165e-01
4.58244890e-01 -4.02444005e-01 5.86265504e-01 1.83629894e+00
3.73274624e-01 2.05437273e-01 1.82137877e-01 1.65445209e-01
-4.15511310e-01 5.56118965e-01 -6.84383810e-02 -1.74766839e-01
8.64520252e-01 1.14516163e+00 -1.07895780e+00 -3.19384724e-01
-3.87758404e-01 1.11229610e+00 4.03264463e-01 1.29045770e-01
-4.07108724e-01 -8.23994637e-01 8.85336041e-01 -1.51137754e-01
3.90726507e-01 -4.96207684e-01 -6.90248907e-01 -9.32411134e-01
2.91387707e-01 -1.16764200e+00 5.08585751e-01 -2.43035272e-01
-1.57968187e+00 8.55263233e-01 -5.08573949e-01 -7.12318361e-01
-7.92501867e-02 -8.62058461e-01 -1.23056918e-01 9.98845220e-01
-1.32714152e+00 -9.53251719e-01 1.33654326e-01 8.57013687e-02
8.46957266e-01 -2.34587550e-01 1.15650010e+00 4.76119041e-01
-3.63740683e-01 8.61669481e-01 4.40930605e-01 6.93280160e-01
1.08469868e+00 -1.62897491e+00 5.70159376e-01 1.15114117e+00
4.48612541e-01 6.86256111e-01 5.41546583e-01 -7.82209337e-01
-9.11094189e-01 -1.39643502e+00 1.71524096e+00 -5.71139872e-01
8.66863430e-01 -3.67167920e-01 -1.16331387e+00 5.78194916e-01
3.21799070e-01 -3.96897137e-01 1.00181842e+00 4.17052329e-01
-5.01209676e-01 9.08525735e-02 -1.24984932e+00 9.26628828e-01
9.56563771e-01 -4.90460634e-01 -7.61845827e-01 1.73118263e-01
6.13180399e-01 -5.17893791e-01 -8.08765292e-01 8.72187093e-02
3.35871518e-01 -3.62539470e-01 4.31487769e-01 -4.73866910e-01
7.08342269e-02 -2.78782517e-01 -3.81856799e-01 -1.65148854e+00
-4.51497883e-01 -7.23509133e-01 5.70809007e-01 1.67721367e+00
7.86037982e-01 -3.08445722e-01 5.39476395e-01 4.86627191e-01
-5.42693675e-01 -2.46047676e-01 -8.21066737e-01 -8.56788933e-01
6.09179735e-01 -6.18167877e-01 4.32652354e-01 1.13159394e+00
1.92505017e-01 4.25017416e-01 -7.75750875e-02 2.17387095e-01
6.62697479e-02 -3.74502569e-01 -1.30312994e-01 -1.00103652e+00
-1.30102217e-01 -5.36013305e-01 -1.90742970e-01 -7.52334356e-01
1.33559525e-01 -7.73069382e-01 4.89152938e-01 -1.44715464e+00
-3.68431658e-01 -5.65662980e-01 -6.30740076e-02 7.88564920e-01
-2.75727361e-01 7.01486349e-01 2.32920691e-01 -1.69659227e-01
-3.56380075e-01 2.46209309e-01 6.00199282e-01 -1.39108077e-02
-4.55174595e-01 -2.57127911e-01 -7.75911033e-01 8.28239918e-01
1.01793122e+00 -6.56663477e-01 2.18810827e-01 -1.02930927e+00
5.04047751e-01 -3.47340912e-01 -4.07934338e-01 -7.66152740e-01
-9.57769975e-02 -2.34772310e-01 4.00940984e-01 -3.62802267e-01
1.26273468e-01 -3.10546845e-01 -4.75056022e-01 1.95970222e-01
-2.43762434e-02 6.80789232e-01 6.07236803e-01 8.63863304e-02
-3.60533535e-01 -6.22971117e-01 9.36127126e-01 -3.53864968e-01
-5.97765982e-01 -4.29010950e-03 -9.47238147e-01 6.22151554e-01
2.46763974e-01 5.03633134e-02 -2.77623504e-01 -1.85223967e-01
-7.66070604e-01 -1.87994093e-02 3.29685241e-01 4.20533687e-01
1.66452378e-01 -1.02303731e+00 -8.80654812e-01 9.02002007e-02
-4.25492525e-02 1.13690913e-01 6.39581680e-03 5.38994849e-01
-7.51112401e-01 2.21419290e-01 -8.67462605e-02 1.55509040e-01
-1.30395043e+00 8.62571504e-03 1.94930495e-03 -3.26260239e-01
-1.51666760e-01 1.10105813e+00 -4.19101268e-01 -8.22097600e-01
2.36954182e-01 -2.84834176e-01 -1.48653343e-01 3.09885561e-01
4.69034135e-01 5.09925663e-01 7.05462337e-01 -1.03246772e+00
-3.49173844e-01 3.63072813e-01 -1.91063553e-01 -4.65673357e-01
1.19372141e+00 -2.36455977e-01 -4.38003957e-01 1.78567380e-01
7.62681663e-01 8.70853484e-01 -6.83565199e-01 -3.59115392e-01
4.30925578e-01 -6.63691461e-02 -1.71512648e-01 -1.32434392e+00
-5.63979268e-01 8.82085264e-01 2.33665913e-01 1.11857481e-01
8.27308416e-01 -5.25965728e-02 1.04637265e+00 6.13829792e-01
1.86321825e-01 -1.80663586e+00 -2.85225153e-01 1.38348246e+00
5.83564997e-01 -1.20883656e+00 -3.88120830e-01 -4.92000252e-01
-8.21720302e-01 9.56313848e-01 4.59784478e-01 4.71148640e-02
5.86181819e-01 5.86568952e-01 6.27743244e-01 -3.38930190e-02
-7.14865103e-02 -3.85734260e-01 1.51501626e-01 8.14647973e-01
9.47738588e-01 2.34570965e-01 -8.20332348e-01 9.51236963e-01
-8.34230900e-01 -8.03026795e-01 5.37278354e-01 8.31544280e-01
-6.40083432e-01 -1.60857463e+00 -3.99620652e-01 2.22402453e-01
-1.07825816e+00 -7.09264457e-01 -4.23744202e-01 7.97301292e-01
3.13824415e-01 1.28872061e+00 3.27483565e-01 5.42590022e-02
3.64594698e-01 4.88976181e-01 3.23658913e-01 -9.78235424e-01
-8.75977159e-01 5.42734861e-01 5.45338690e-01 -2.41930380e-01
-1.01250418e-01 -8.40765655e-01 -1.82319069e+00 -1.88177288e-01
-2.77299315e-01 2.82801241e-01 6.77546024e-01 8.98935437e-01
2.30067015e-01 1.97191939e-01 1.41121790e-01 -3.78163993e-01
-6.45354986e-01 -1.20257640e+00 -7.30111480e-01 3.93663734e-01
-1.44483894e-01 -2.20811129e-01 -1.14240482e-01 2.14679867e-01]
|
[10.433847427368164, 10.161334991455078]
|
f41d3e79-7bc4-426b-a498-d2681f4ffe2c
|
robust-recovery-for-stochastic-block-models
|
2111.08568
| null |
https://arxiv.org/abs/2111.08568v1
|
https://arxiv.org/pdf/2111.08568v1.pdf
|
Robust recovery for stochastic block models
|
We develop an efficient algorithm for weak recovery in a robust version of the stochastic block model. The algorithm matches the statistical guarantees of the best known algorithms for the vanilla version of the stochastic block model. In this sense, our results show that there is no price of robustness in the stochastic block model. Our work is heavily inspired by recent work of Banks, Mohanty, and Raghavendra (SODA 2021) that provided an efficient algorithm for the corresponding distinguishing problem. Our algorithm and its analysis significantly depart from previous ones for robust recovery. A key challenge is the peculiar optimization landscape underlying our algorithm: The planted partition may be far from optimal in the sense that completely unrelated solutions could achieve the same objective value. This phenomenon is related to the push-out effect at the BBP phase transition for PCA. To the best of our knowledge, our algorithm is the first to achieve robust recovery in the presence of such a push-out effect in a non-asymptotic setting. Our algorithm is an instantiation of a framework based on convex optimization (related to but distinct from sum-of-squares), which may be useful for other robust matrix estimation problems. A by-product of our analysis is a general technique that boosts the probability of success (over the randomness of the input) of an arbitrary robust weak-recovery algorithm from constant (or slowly vanishing) probability to exponentially high probability.
|
['David Steurer', 'Rajai Nasser', "Tommaso d'Orsi", 'Jingqiu Ding']
|
2021-11-16
| null | null | null | null |
['stochastic-block-model']
|
['graphs']
|
[ 5.00885248e-01 8.41350257e-02 -1.78417832e-01 2.15778574e-01
-1.24732625e+00 -7.58709192e-01 5.50441444e-01 7.91398659e-02
-2.48293146e-01 6.79014742e-01 2.31415227e-01 -4.38785851e-01
-6.21263206e-01 -5.39478064e-01 -9.90546703e-01 -1.43438900e+00
-2.98647672e-01 3.79947990e-01 1.87805757e-01 -4.67455596e-01
2.08551437e-01 4.96214896e-01 -9.97940838e-01 -5.68194762e-02
4.70555514e-01 6.48950338e-01 1.43763393e-01 8.58031690e-01
3.26890439e-01 5.69497466e-01 -1.29736200e-01 -3.56517553e-01
7.41796613e-01 -5.01426101e-01 -6.80497289e-01 1.17701419e-01
1.24193490e-01 5.44935353e-02 -4.54545259e-01 1.37874091e+00
5.23989439e-01 -1.00645736e-01 4.57957476e-01 -1.00646174e+00
-1.56740561e-01 9.15734231e-01 -8.79528165e-01 2.14109346e-01
1.76816404e-01 -1.23284414e-01 1.10529542e+00 -7.14071691e-01
5.63368738e-01 8.37222695e-01 6.93868756e-01 1.53448015e-01
-1.49032998e+00 -4.74761426e-01 -6.35369048e-02 7.91517794e-02
-1.57283843e+00 -6.96214139e-01 5.96646249e-01 -3.76485616e-01
2.32023090e-01 5.28879881e-01 2.23529622e-01 1.01305413e+00
-1.92719042e-01 9.40399051e-01 1.46790624e+00 -7.81764388e-01
1.91565037e-01 -1.04306052e-02 2.98408985e-01 4.75846320e-01
6.57032371e-01 3.35854977e-01 -3.52551669e-01 -3.96328717e-01
4.75447208e-01 -1.79276839e-01 -6.09771848e-01 -9.42331970e-01
-1.32929981e+00 9.09059346e-01 1.58759579e-01 5.45933008e-01
-2.03914136e-01 2.81290412e-01 1.46957800e-01 4.21919852e-01
2.09543049e-01 2.00099528e-01 -3.73884797e-01 -1.21747307e-01
-1.21666539e+00 4.50849742e-01 1.06361938e+00 7.96051443e-01
5.41424990e-01 -3.97514366e-02 -1.56201094e-01 3.18410367e-01
1.80569947e-01 7.27024674e-01 5.51235080e-02 -8.70678067e-01
6.26325488e-01 -4.22249317e-01 2.80259013e-01 -1.06902575e+00
-2.79180765e-01 -8.67801845e-01 -1.13240516e+00 1.97298035e-01
9.79024351e-01 2.20514694e-03 -3.75146568e-01 1.94643521e+00
1.30989909e-01 3.80593181e-01 8.58473778e-02 7.32790768e-01
-7.57549927e-02 4.63541120e-01 -7.06974924e-01 -6.81610286e-01
9.68527734e-01 -5.57358682e-01 -6.21834159e-01 -1.49371549e-01
4.84557033e-01 -7.27305293e-01 5.51143408e-01 6.67484045e-01
-1.09370744e+00 -3.65937129e-02 -1.20423722e+00 3.81330758e-01
2.23652914e-01 -2.30251253e-01 5.03764153e-01 1.04247224e+00
-9.75434840e-01 8.85571957e-01 -6.56672239e-01 -3.45197678e-01
1.73346847e-01 3.84869665e-01 -5.31238854e-01 -3.32200289e-01
-6.43097579e-01 6.11571431e-01 1.04289457e-01 3.95893782e-01
-5.34574986e-01 -4.17960048e-01 -5.30766129e-01 -1.56706106e-03
8.17622304e-01 -4.31789994e-01 1.02794302e+00 -8.61732304e-01
-1.30681515e+00 7.76149869e-01 -3.78766239e-01 -7.23876715e-01
7.36469090e-01 4.78076003e-02 1.06273770e-01 1.32474340e-02
7.14883395e-03 -3.50221246e-01 1.10032487e+00 -1.18420589e+00
8.90467875e-03 -5.81674933e-01 -7.53986686e-02 -2.67523378e-01
2.19737187e-01 5.24463039e-03 -3.21639806e-01 -8.45229626e-01
3.75266641e-01 -1.26817918e+00 -6.74953818e-01 -4.69211668e-01
-3.95129472e-01 4.00524378e-01 8.73415172e-02 -4.13556755e-01
1.01514256e+00 -2.15327311e+00 6.77518964e-01 6.19994879e-01
2.58308887e-01 3.18526067e-02 -1.86018065e-01 8.39797437e-01
-4.85337615e-01 1.09009117e-01 -5.85878611e-01 -3.56620282e-01
1.83757648e-01 1.86598331e-01 -7.50281274e-01 1.18334031e+00
-1.52461067e-01 6.34929240e-01 -9.15224552e-01 -7.63264522e-02
-1.22573068e-02 5.10000139e-02 -5.76072991e-01 -2.41406977e-01
3.90447855e-01 3.17616135e-01 -1.59719840e-01 3.77205402e-01
9.85038042e-01 -1.73373640e-01 3.74998689e-01 2.29777507e-02
-5.16263507e-02 -5.66627979e-02 -1.75053632e+00 1.46924746e+00
-1.08152509e-01 5.17872810e-01 6.46935642e-01 -1.48567283e+00
4.08713013e-01 3.12890410e-01 5.49975753e-01 -2.06285596e-01
1.32972717e-01 4.63849425e-01 1.45897567e-01 -1.55371234e-01
5.29541314e-01 -3.69199276e-01 -1.88033238e-01 5.81901550e-01
-7.48846158e-02 7.25252135e-03 1.83276758e-01 3.82275075e-01
1.32173777e+00 -1.72118992e-01 5.74259579e-01 -6.24746680e-01
3.55771989e-01 -3.53241891e-01 3.68247271e-01 1.44695401e+00
-1.15365967e-01 8.31925333e-01 7.59896338e-01 1.20177820e-01
-1.03662777e+00 -8.86178732e-01 -2.72155285e-01 7.09881723e-01
-1.38127401e-01 -6.15547776e-01 -4.47944999e-01 -4.15916145e-01
-8.87903292e-03 1.49372801e-01 -6.63733840e-01 1.78414229e-02
-4.75145817e-01 -1.14912128e+00 3.33412111e-01 7.55858645e-02
4.04006317e-02 -5.00191569e-01 -3.55075710e-02 1.69897929e-01
-8.06622878e-02 -1.30090415e+00 -2.30843991e-01 5.52764118e-01
-9.69841897e-01 -1.09169471e+00 -8.08580577e-01 -2.70583957e-01
4.71424609e-01 5.31249344e-01 8.40203762e-01 -8.04616958e-02
1.44066885e-01 4.01248872e-01 -3.35165381e-01 -2.36826122e-01
-5.65254927e-01 -9.69343856e-02 1.68528631e-01 2.83953786e-01
-1.19461045e-01 -6.23083711e-01 -2.48060271e-01 1.31955251e-01
-1.05535483e+00 -2.82409608e-01 5.48622191e-01 9.41314936e-01
4.38690722e-01 2.71542400e-01 2.25091390e-02 -9.27374005e-01
3.96721870e-01 -6.89258158e-01 -8.18516135e-01 -3.22897509e-02
-5.77697933e-01 4.42836285e-01 5.15696645e-01 -2.19641343e-01
-3.36229593e-01 3.45040679e-01 5.77368401e-03 -2.07181454e-01
3.50400567e-01 4.58572596e-01 -1.91751879e-03 -3.03145111e-01
6.44789159e-01 3.49941969e-01 -9.04399008e-02 -5.40780723e-01
5.73035717e-01 4.92994040e-01 6.45711839e-01 -7.29347050e-01
1.50877607e+00 8.29753995e-01 6.21575952e-01 -7.21105814e-01
-7.04181552e-01 -6.11271024e-01 -5.51474392e-01 8.07714537e-02
1.74568594e-01 -8.68319571e-01 -6.14449978e-01 4.52209651e-01
-8.10925007e-01 -2.81674474e-01 -4.17720109e-01 3.62873822e-01
-6.92928433e-01 8.50986660e-01 -3.89076084e-01 -1.23702991e+00
1.46874517e-01 -1.08475375e+00 9.40128207e-01 -2.51686543e-01
1.98258191e-01 -6.03282273e-01 2.84406573e-01 1.90627545e-01
2.94212818e-01 1.96009859e-01 4.57974315e-01 -4.54666197e-01
-6.74561679e-01 -4.19456899e-01 -1.49863020e-01 3.21610540e-01
-4.38365519e-01 -1.86373144e-01 -8.67494702e-01 -5.07908583e-01
4.63170141e-01 7.08401278e-02 1.16147602e+00 4.82145488e-01
8.40521693e-01 -4.97601837e-01 -1.77766144e-01 6.18871272e-01
1.60314941e+00 -3.99476230e-01 7.16975093e-01 3.27233464e-01
5.24651468e-01 4.98070776e-01 2.17916831e-01 4.33286786e-01
-2.79746223e-02 1.07269919e+00 4.16296810e-01 1.65685579e-01
3.02922606e-01 4.38926630e-02 5.66439033e-01 9.18096781e-01
-1.35421515e-01 -5.88863194e-02 -5.93298495e-01 6.18836880e-01
-2.07007122e+00 -1.27360010e+00 -5.42181671e-01 2.75183773e+00
8.61033797e-01 3.63914706e-02 3.89880151e-01 5.22041738e-01
6.43761694e-01 3.48648340e-01 4.24274839e-02 -3.00004065e-01
-4.35997009e-01 5.42509437e-01 1.12044096e+00 7.19302893e-01
-1.03934562e+00 4.42725003e-01 6.35495615e+00 1.09829581e+00
-5.68769336e-01 3.64585310e-01 1.79064244e-01 -2.13109836e-01
-2.29274482e-01 3.82521719e-01 -5.04262209e-01 4.60238636e-01
9.13224220e-01 -1.26029298e-01 7.69003868e-01 5.64462185e-01
5.52423149e-02 -1.04367144e-01 -9.13505197e-01 9.70199883e-01
1.65226743e-01 -1.12838268e+00 -7.62157559e-01 6.31857276e-01
8.68661404e-01 9.59545299e-02 4.82068658e-02 5.83779998e-03
5.27058423e-01 -8.59182775e-01 6.30473554e-01 3.81514966e-01
3.54348898e-01 -6.20026529e-01 6.73606634e-01 3.83060098e-01
-8.65937591e-01 -1.89098775e-01 -2.45280653e-01 -1.31750062e-01
1.67750180e-01 1.02710235e+00 -3.33441973e-01 7.38316953e-01
2.41321340e-01 4.36732113e-01 -2.59935677e-01 1.15182436e+00
-1.45784453e-01 8.26489747e-01 -7.63787389e-01 1.95224032e-01
1.11208960e-01 -5.06285191e-01 1.06973445e+00 1.18164217e+00
4.01497096e-01 -7.51196593e-02 -9.92662385e-02 3.17277730e-01
1.60405766e-02 1.68089852e-01 -7.62090743e-01 8.74294862e-02
1.85336992e-01 9.67114151e-01 -7.64340699e-01 5.58963455e-02
-3.14901203e-01 8.75262916e-01 1.41866431e-01 3.14573884e-01
-6.02056921e-01 -2.88963486e-02 2.88085580e-01 2.02208757e-02
8.71241450e-01 -4.07713860e-01 -3.47725064e-01 -1.33642936e+00
4.46875334e-01 -1.12970233e+00 2.95541108e-01 -2.42945790e-01
-1.22246194e+00 2.60783937e-02 -1.40878141e-01 -1.11326659e+00
-2.69980401e-01 -4.58420724e-01 -2.57497221e-01 6.48747206e-01
-1.32910800e+00 -7.35345304e-01 2.73378313e-01 6.08593225e-01
-9.80183780e-02 9.33704227e-02 7.55024850e-01 1.66979179e-01
-5.14997900e-01 4.53544766e-01 9.04851913e-01 -1.65474772e-01
5.84611475e-01 -1.38802385e+00 7.11068064e-02 1.60991418e+00
6.34026289e-01 5.88755608e-01 1.11556375e+00 -2.56477773e-01
-1.90367770e+00 -5.38328648e-01 7.52772450e-01 -6.09142542e-01
1.16511476e+00 -5.88324070e-01 -5.76496124e-01 5.84374726e-01
-7.48842657e-02 -4.71299589e-02 4.73722190e-01 3.32372010e-01
-5.34280539e-01 -1.57605901e-01 -7.89809346e-01 5.33378482e-01
9.20837402e-01 -5.06253481e-01 -3.65336090e-01 6.48817122e-01
2.66627133e-01 -2.47531787e-01 -5.01943469e-01 2.85706490e-01
5.24604857e-01 -9.90818560e-01 1.09813786e+00 -3.91838789e-01
3.29003423e-01 -3.68095517e-01 -6.40133619e-01 -8.40457797e-01
-2.66319543e-01 -1.37705553e+00 -4.38291341e-01 9.16087508e-01
5.00786342e-02 -4.70929086e-01 6.33379579e-01 1.05261728e-01
2.48402312e-01 -4.51403499e-01 -1.25017190e+00 -1.04483330e+00
2.46077240e-01 -8.15464199e-01 1.60473332e-01 7.18587279e-01
-8.17731246e-02 1.28370672e-01 -7.90660441e-01 3.88578713e-01
9.93946612e-01 2.38119736e-01 9.77265060e-01 -1.09206140e+00
-1.10081220e+00 -6.54381096e-01 -4.65276450e-01 -1.17462564e+00
-6.94416165e-02 -1.01351166e+00 8.81013423e-02 -8.94241035e-01
3.63236129e-01 -5.12346208e-01 -3.66171539e-01 -1.96164474e-02
-1.51246578e-01 4.06018585e-01 5.85325241e-01 3.41621965e-01
-4.29282844e-01 1.59960642e-01 6.89523935e-01 -3.80257070e-02
-8.21212009e-02 4.73281264e-01 -1.10116041e+00 3.86467367e-01
5.24249196e-01 -8.22934210e-01 -9.88478214e-03 -7.33549371e-02
6.30833924e-01 1.77836567e-01 4.16296065e-01 -8.73595893e-01
3.13258678e-01 1.68991581e-01 -3.62303287e-01 -3.12733322e-01
2.51984954e-01 -8.99515092e-01 2.40202263e-01 5.60397089e-01
-8.99733081e-02 -1.67442933e-01 -1.51590094e-01 1.00776505e+00
1.53572321e-01 -8.03987026e-01 6.19800329e-01 1.16475366e-01
-4.12659273e-02 7.69326538e-02 -4.35278863e-01 5.94019517e-02
7.06943870e-01 -1.72158461e-02 -1.80915028e-01 -7.42202222e-01
-6.53195024e-01 -2.10075751e-01 4.78352427e-01 -1.33616865e-01
3.27236205e-02 -1.04828703e+00 -1.05913055e+00 1.05483606e-01
-8.37585479e-02 -3.10341150e-01 2.20498621e-01 1.44971764e+00
-1.60724774e-01 3.03373456e-01 3.85802209e-01 -4.97439563e-01
-1.16883206e+00 8.49094808e-01 4.55915034e-02 -6.96666002e-01
-4.59122509e-01 6.58846676e-01 5.06665111e-02 -4.31822762e-02
-3.93669531e-02 -9.07338485e-02 4.35471922e-01 -3.41857858e-02
6.37362957e-01 4.13660884e-01 2.22263992e-01 -7.07537651e-01
-2.72460163e-01 5.59450984e-01 1.25034600e-01 -5.16652644e-01
1.37365317e+00 -1.56952381e-01 -2.37717703e-01 4.04275119e-01
1.18682885e+00 1.04570127e+00 -8.43719363e-01 -4.25859928e-01
-4.04475927e-02 -4.84262347e-01 3.53578962e-02 -2.63219446e-01
-9.91291225e-01 6.63977027e-01 4.26679671e-01 5.16126096e-01
1.02926147e+00 5.65641709e-02 4.17162925e-01 4.49153572e-01
5.52630365e-01 -6.23546004e-01 -4.41247642e-01 3.61917853e-01
7.66655803e-01 -1.08021867e+00 3.73232305e-01 -3.28189045e-01
-2.36620069e-01 9.66439664e-01 -4.59839940e-01 -3.73522609e-01
6.14739954e-01 4.92614895e-01 -5.04277885e-01 -4.80758660e-02
-3.99838448e-01 -4.66643542e-01 1.01944737e-01 6.21317208e-01
-1.77766681e-01 1.48297310e-01 -4.97449726e-01 5.39953530e-01
-1.78610742e-01 -2.68213987e-01 8.37037802e-01 7.55536556e-01
-1.76289737e-01 -1.40037429e+00 -7.17945457e-01 1.58564523e-01
-7.67874837e-01 -3.44048142e-01 -2.09099889e-01 8.03312838e-01
-3.13843697e-01 1.03506708e+00 -4.66177434e-01 -1.48155347e-01
7.49759283e-03 -4.94424924e-02 7.64867246e-01 -3.71569037e-01
-4.21813011e-01 3.59717578e-01 -1.31371513e-01 -5.18669844e-01
-6.52019441e-01 -1.03047645e+00 -4.51824158e-01 -6.01694167e-01
-4.57458228e-01 1.57867506e-01 5.68806887e-01 1.04790211e+00
2.69734915e-02 1.13152368e-02 8.71716738e-01 -7.41874933e-01
-9.06175494e-01 -6.80338025e-01 -9.47376311e-01 1.05343603e-01
5.30638695e-01 -2.90499806e-01 -7.12140918e-01 -3.10083747e-01]
|
[6.912405014038086, 4.694620609283447]
|
9ca024c8-3607-4ae4-8045-fbe8e92c0837
|
efficient-deep-learning-for-stereo-matching
| null | null |
http://openaccess.thecvf.com/content_cvpr_2016/html/Luo_Efficient_Deep_Learning_CVPR_2016_paper.html
|
http://openaccess.thecvf.com/content_cvpr_2016/papers/Luo_Efficient_Deep_Learning_CVPR_2016_paper.pdf
|
Efficient Deep Learning for Stereo Matching
|
In the past year, convolutional neural networks have been shown to perform extremely well for stereo estimation. However, current architectures rely on siamese networks which exploit concatenation followed by further processing layers, requiring a minute of GPU computation per image pair. In contrast, in this paper we propose a matching network which is able to produce very accurate results in less than a second of GPU computation. Towards this goal, we exploit a product layer which simply computes the inner product between the two representations of a siamese architecture. We train our network by treating the problem as multi-class classification, where the classes are all possible disparities. This allows us to get calibrated scores, which result in much better matching performance when compared to existing approaches.
|
['Raquel Urtasun', 'Wenjie Luo', 'Alexander G. Schwing']
|
2016-06-01
| null | null | null |
cvpr-2016-6
|
['stereo-matching']
|
['computer-vision']
|
[ 1.28031611e-01 -1.58014223e-01 2.04477042e-01 -3.67932647e-01
-6.12249672e-01 -4.52298403e-01 8.85284960e-01 6.68308511e-03
-9.38941300e-01 4.86620873e-01 4.70551774e-02 -1.35303557e-01
2.65974611e-01 -9.83831584e-01 -7.69989252e-01 -3.24827969e-01
3.56919356e-02 3.74332964e-01 4.28253770e-01 -4.04324025e-01
4.33973908e-01 6.72540069e-01 -1.78045821e+00 3.69914979e-01
7.54587233e-01 1.19494569e+00 -1.89605668e-01 1.03731000e+00
5.20103201e-02 5.82573473e-01 -4.09475386e-01 -6.99860215e-01
7.07956791e-01 -2.21757665e-01 -7.28314340e-01 -2.84456939e-01
1.40159023e+00 -5.22612393e-01 -4.80862945e-01 1.03558660e+00
4.99063939e-01 7.93540701e-02 3.22027326e-01 -1.08456910e+00
1.46148680e-02 3.76869291e-02 -4.52401936e-01 5.81008643e-02
3.38756621e-01 1.41477585e-02 9.40452337e-01 -8.83221030e-01
6.88351572e-01 1.10282123e+00 1.02251172e+00 2.70629942e-01
-1.32751417e+00 -5.63350916e-01 -4.39378947e-01 2.68892854e-01
-1.34444368e+00 -4.85818177e-01 5.63757718e-01 -2.17737630e-01
1.25245273e+00 5.86216301e-02 9.64842677e-01 6.09068274e-01
2.72065431e-01 5.79397917e-01 9.16745126e-01 -2.26739034e-01
1.00687526e-01 -3.38643283e-01 -1.54800089e-02 7.51628101e-01
2.26224601e-01 3.90148401e-01 -6.48048997e-01 1.40917689e-01
1.00892150e+00 -1.03130816e-02 -1.50082663e-01 -5.85582674e-01
-1.41161728e+00 8.74802709e-01 9.51536417e-01 1.32772103e-01
-2.76731580e-01 5.03635585e-01 6.29471362e-01 2.96414554e-01
2.74319381e-01 4.45935786e-01 -8.60474706e-02 -1.09315850e-01
-1.41261256e+00 5.11552453e-01 8.46764088e-01 7.30990112e-01
9.38290119e-01 -1.78819120e-01 1.16018571e-01 6.19584918e-01
-7.50744864e-02 1.83159083e-01 3.55249524e-01 -1.29421735e+00
6.22828841e-01 3.95053953e-01 -6.35303138e-03 -1.16777146e+00
-3.99958760e-01 -4.59422737e-01 -1.03869510e+00 7.78553903e-01
6.75625741e-01 5.69086932e-02 -7.54743099e-01 1.55713129e+00
2.35198498e-01 4.13908333e-01 8.53127018e-02 9.33873475e-01
4.23202008e-01 3.74072790e-01 -3.40885580e-01 5.22940516e-01
1.16074955e+00 -1.29151154e+00 -2.02447087e-01 -1.40247956e-01
7.29473829e-01 -1.02839029e+00 8.26884747e-01 5.16894341e-01
-1.37507784e+00 -6.71060920e-01 -1.62563169e+00 -4.81822371e-01
-4.62845057e-01 -3.22773643e-02 9.07262146e-01 6.62035465e-01
-1.38981009e+00 1.11518943e+00 -8.81418467e-01 -2.23620445e-01
5.49133837e-01 5.84510028e-01 -4.48389202e-01 -2.15823233e-01
-8.65761757e-01 8.42449188e-01 3.59821320e-01 2.52679348e-01
-2.96337843e-01 -7.11392164e-01 -8.69298339e-01 1.05992824e-01
-9.79283527e-02 -9.47254777e-01 1.24899507e+00 -1.07126772e+00
-1.51389325e+00 9.56950963e-01 -9.15050060e-02 -7.71587849e-01
8.30303729e-01 -1.70733958e-01 3.41599225e-03 1.41281232e-01
-8.18005130e-02 1.12862492e+00 5.00746012e-01 -6.70424759e-01
-8.32127988e-01 -2.13427454e-01 3.66037130e-01 8.94580558e-02
-2.24249244e-01 -1.59738153e-01 -6.33166909e-01 -5.54466307e-01
3.52536649e-01 -1.07929754e+00 -4.19436932e-01 5.03101349e-01
-4.64536846e-02 4.09706607e-02 5.64912498e-01 -4.22761768e-01
7.95841217e-01 -2.09831500e+00 -2.49930341e-02 2.78900504e-01
3.44265103e-01 3.63008022e-01 -1.98377833e-01 1.32259101e-01
5.58247678e-02 -4.19553876e-01 -2.53400028e-01 -6.86177194e-01
-1.20347645e-02 2.54086584e-01 -2.55792230e-01 5.58789194e-01
1.98307633e-01 7.42909491e-01 -7.40536392e-01 -4.48219866e-01
3.15365493e-01 6.71944857e-01 -1.03793645e+00 1.27406254e-01
1.56041101e-01 1.22127190e-01 1.69238672e-01 2.16416359e-01
1.10851252e+00 -1.37787789e-01 1.49530366e-01 -3.78110856e-01
-2.15599567e-01 3.57470870e-01 -1.26584733e+00 2.12307835e+00
-7.49431431e-01 1.02752030e+00 -3.88633609e-02 -7.81022251e-01
7.07154512e-01 -1.05930693e-01 3.65414619e-01 -7.93970227e-01
1.81928203e-01 6.35695875e-01 9.46080964e-03 1.34568185e-01
7.92790949e-01 1.30993649e-01 1.90796316e-01 2.20783055e-01
-5.70190959e-02 -3.84830803e-01 4.08126682e-01 1.11629143e-01
9.15719688e-01 2.18538031e-01 6.58579320e-02 -3.31573427e-01
5.87916493e-01 1.30920574e-01 2.90086329e-01 5.95326841e-01
-4.64128517e-02 1.03398156e+00 4.06951010e-01 -8.07912290e-01
-1.37693942e+00 -1.13242316e+00 -1.35096043e-01 6.01795256e-01
2.27455243e-01 -3.52516413e-01 -8.57546389e-01 -2.15252921e-01
-9.40437615e-02 2.87713800e-02 -4.87181783e-01 9.12355632e-02
-1.07771993e+00 -5.02422333e-01 7.15506911e-01 9.03639615e-01
1.05366993e+00 -4.61913228e-01 -9.10233378e-01 2.09556207e-01
8.69159400e-02 -1.13447058e+00 -4.77201521e-01 2.26602852e-01
-1.05782020e+00 -1.09117329e+00 -9.14685369e-01 -8.20738614e-01
4.95927840e-01 2.81097263e-01 1.37803471e+00 1.63024321e-01
-2.53383547e-01 -2.34191313e-01 1.87289372e-01 4.82987501e-02
-8.48046765e-02 1.72199950e-01 -2.38718182e-01 -2.02004239e-01
2.37908140e-01 -7.34767616e-01 -9.47222888e-01 7.18535781e-02
-1.00538540e+00 2.84270346e-01 5.94941616e-01 9.28435445e-01
4.89107132e-01 -4.79068637e-01 -2.69322664e-01 -7.99054086e-01
3.23063076e-01 -1.21144764e-02 -9.32313800e-01 -2.38923412e-02
-4.86224353e-01 2.50095755e-01 9.10916030e-01 -1.89881057e-01
-6.26503646e-01 3.11551601e-01 -4.22995836e-01 -3.56344402e-01
5.26222214e-02 2.18464002e-01 3.76010507e-01 -5.80869615e-01
6.82632804e-01 -6.51722997e-02 1.72033146e-01 -2.01961726e-01
2.30944827e-01 3.20605248e-01 8.35713327e-01 -3.69371176e-01
6.31990075e-01 7.23339200e-01 5.02297103e-01 -5.70382595e-01
-7.70801783e-01 -4.53857899e-01 -7.42302001e-01 -7.79906884e-02
8.54154944e-01 -8.62956524e-01 -8.79026055e-01 5.55157065e-01
-1.30303872e+00 -2.94673353e-01 -1.05833143e-01 6.29425645e-01
-7.31695056e-01 3.84957194e-01 -6.67531908e-01 -1.79292724e-01
-2.16411486e-01 -1.12744117e+00 1.15302014e+00 1.10379301e-01
-1.38918430e-01 -1.00987303e+00 1.76232487e-01 1.47136852e-01
6.93768978e-01 2.62636602e-01 4.16048467e-01 -1.78999037e-01
-8.49572003e-01 -3.09389830e-01 -6.69681191e-01 3.04826319e-01
-4.85438377e-01 -7.66387731e-02 -9.85830188e-01 -4.10499483e-01
-2.01472759e-01 -3.05562258e-01 1.06043565e+00 2.32917026e-01
1.03686392e+00 1.28674284e-01 -5.76606439e-03 1.22649515e+00
1.74418068e+00 -2.38878608e-01 9.18815136e-01 5.54806054e-01
6.04170084e-01 4.35981125e-01 4.61972505e-01 2.26409018e-01
3.76000226e-01 9.72926915e-01 5.28845191e-01 -4.08007950e-01
-5.51771522e-01 -1.45747453e-01 -6.82396367e-02 6.48892403e-01
-2.75834620e-01 6.95412830e-02 -1.11316860e+00 2.57967770e-01
-1.77474487e+00 -8.00116539e-01 -3.59952986e-01 2.31066084e+00
3.62863094e-01 2.24511877e-01 -1.37369586e-02 1.50740683e-01
3.21874440e-01 3.84849876e-01 -3.00554067e-01 -5.68096161e-01
-2.36363277e-01 7.80990601e-01 8.61120284e-01 6.32929206e-01
-1.15742004e+00 7.74115026e-01 7.12273359e+00 6.52767062e-01
-1.21244001e+00 -9.00607258e-02 5.23381472e-01 -2.00915128e-01
-4.75340821e-02 1.58030223e-02 -5.84582090e-01 2.58885831e-01
7.54360378e-01 5.50519451e-02 3.97786260e-01 6.87005103e-01
-1.79961532e-01 -4.05342251e-01 -1.23569167e+00 1.32254136e+00
2.16986656e-01 -1.63334882e+00 1.13546677e-01 1.21453121e-01
8.95532548e-01 3.20365041e-01 1.42179564e-01 -9.77493357e-03
2.52495825e-01 -1.05751801e+00 6.96562648e-01 4.14016396e-01
7.09946871e-01 -8.21128190e-01 8.89583409e-01 1.69980958e-01
-1.29667592e+00 1.38435677e-01 -5.34339070e-01 -3.09582621e-01
-2.23773997e-02 5.22377491e-01 -3.84125710e-01 5.20863056e-01
7.21115947e-01 7.04215884e-01 -5.73714972e-01 1.33413637e+00
1.26206338e-01 -1.35641977e-01 -5.14275551e-01 1.52159631e-01
6.46703243e-01 -4.60818112e-01 -7.52504449e-03 1.16471386e+00
5.70160568e-01 -2.76512265e-01 -7.46345371e-02 4.84422415e-01
-1.43421263e-01 -4.15729312e-03 -6.14051521e-01 5.10600150e-01
-2.67286934e-02 1.04629445e+00 -6.52392268e-01 -6.05335653e-01
-4.90287900e-01 1.37777305e+00 6.18992865e-01 7.94646963e-02
-7.59580970e-01 -7.97462702e-01 9.08796668e-01 -1.59794241e-01
3.24530542e-01 -6.04274929e-01 -5.07564127e-01 -1.29745722e+00
1.71421826e-01 -5.65584481e-01 1.54776528e-01 -6.60798788e-01
-8.38891625e-01 6.08384669e-01 -4.23050344e-01 -1.36869061e+00
-5.30875564e-01 -1.04193723e+00 -6.32194996e-01 9.00025249e-01
-1.74449253e+00 -9.99839842e-01 -5.30446351e-01 4.74064589e-01
2.99532622e-01 3.38976784e-03 6.51474237e-01 6.74204886e-01
-2.35086009e-01 6.08468533e-01 1.47898808e-01 1.66216195e-01
5.90915442e-01 -1.13139093e+00 8.42859805e-01 8.78354013e-01
2.65317917e-01 5.61331630e-01 6.22286022e-01 -1.03479445e-01
-1.16164362e+00 -8.22814584e-01 1.15265286e+00 -2.16409471e-02
5.85890710e-01 -2.65799463e-01 -5.66277266e-01 3.47853899e-01
3.33901137e-01 3.38982731e-01 3.35673511e-01 -4.17227969e-02
-7.16090322e-01 -2.30473042e-01 -8.90413284e-01 6.54474318e-01
1.19229496e+00 -6.79203629e-01 -1.95601389e-01 1.84413865e-01
2.16344014e-01 -7.28372753e-01 -7.44000375e-01 2.17381924e-01
8.21408331e-01 -1.73539329e+00 1.16352344e+00 -3.34732562e-01
7.26849020e-01 -3.19810539e-01 -2.19185144e-01 -1.05812991e+00
-7.42710531e-02 -2.95910269e-01 1.16364367e-01 5.12636304e-01
2.49613643e-01 -6.72766089e-01 1.21607649e+00 4.33413148e-01
-1.15704410e-01 -7.60082245e-01 -1.07569540e+00 -9.03596044e-01
1.62208900e-01 -2.94841200e-01 6.16973996e-01 7.62699187e-01
-3.84663641e-01 -1.00097395e-01 -1.73835889e-01 -6.17964659e-03
7.28770137e-01 3.41682076e-01 8.57359886e-01 -1.13905466e+00
-3.53010833e-01 -7.81118810e-01 -8.84924293e-01 -1.55497968e+00
1.98924065e-01 -8.48015189e-01 -8.39915648e-02 -1.26119113e+00
1.07003234e-01 -5.19417524e-01 -8.17942023e-02 2.64061838e-02
4.96447533e-02 1.04844534e+00 3.15777332e-01 1.59243718e-01
-3.54816198e-01 1.77046701e-01 1.05770850e+00 -1.02955446e-01
1.24935322e-01 -1.27344206e-01 -2.28506133e-01 7.61346340e-01
7.75570452e-01 -1.13414302e-01 -5.93515448e-02 -9.91009712e-01
3.69965136e-01 -1.21644199e-01 5.05741000e-01 -1.63646638e+00
5.49491227e-01 2.10780159e-01 4.78619069e-01 -4.92617905e-01
7.33465075e-01 -8.07609320e-01 2.00981930e-01 7.91525900e-01
-2.03184143e-01 2.94241637e-01 1.16300978e-01 1.12707570e-01
-6.78882658e-01 -1.73129037e-01 9.24782813e-01 1.31254233e-02
-6.19579375e-01 3.64468694e-01 2.73428001e-02 -2.09470987e-01
7.00674415e-01 -5.92554629e-01 -1.06532902e-01 -2.95245796e-01
-4.04868990e-01 -7.30914176e-02 8.72108638e-01 5.04265986e-02
5.24814606e-01 -1.51370692e+00 -5.90438068e-01 3.60801548e-01
-1.52193725e-01 2.81700399e-02 1.58662066e-01 7.53883541e-01
-1.31121933e+00 4.77702081e-01 -5.58369398e-01 -7.54578173e-01
-1.23802114e+00 3.07712793e-01 5.58367252e-01 -2.88935393e-01
-4.34265256e-01 8.47256541e-01 -5.07027619e-02 -3.65928501e-01
7.94638097e-02 -3.12324166e-01 6.36136606e-02 -3.27231735e-02
3.50801915e-01 4.72757906e-01 3.20517957e-01 -6.35859072e-01
-2.71221787e-01 9.34485614e-01 1.17801130e-01 -1.96525916e-01
1.18067896e+00 2.64099002e-01 -1.14938289e-01 1.26102287e-02
1.71276891e+00 -1.59172893e-01 -1.41883504e+00 3.06084007e-02
-1.48469016e-01 -7.68414319e-01 2.13770077e-01 -2.23075867e-01
-1.22528446e+00 1.24988234e+00 7.62934446e-01 -7.66393319e-02
1.21913445e+00 -4.07602966e-01 1.08928907e+00 5.02781510e-01
3.82593721e-01 -1.10195839e+00 1.36082422e-03 7.70933270e-01
5.80909967e-01 -1.39029944e+00 -3.67223695e-02 -4.91075963e-01
-3.52855213e-02 1.46260476e+00 4.51989889e-01 -6.82619810e-01
4.27517563e-01 4.09462303e-01 7.84262866e-02 1.35467574e-01
-4.36663151e-01 -3.40842932e-01 2.37974808e-01 3.68621647e-01
5.36844254e-01 -1.94577947e-01 -2.98387140e-01 -4.44045275e-01
-5.15587747e-01 -3.20855454e-02 3.97265851e-01 8.84602726e-01
-2.25068957e-01 -1.27955651e+00 -2.64516324e-01 4.06144351e-01
-2.27139890e-01 -3.13667655e-01 -9.31259319e-02 6.37061715e-01
7.39372149e-02 5.61468780e-01 5.84851205e-01 -3.16238731e-01
4.61417198e-01 -2.77666390e-01 6.77446425e-01 -2.25156218e-01
-8.19563329e-01 -2.40671203e-01 6.75624534e-02 -1.14314747e+00
-4.55932558e-01 -3.78760308e-01 -1.02050638e+00 -7.36209512e-01
9.92142558e-02 -2.12008968e-01 8.08241367e-01 5.02004325e-01
2.49429956e-01 2.50571311e-01 5.23879528e-01 -1.30811918e+00
-4.97005969e-01 -4.69803900e-01 -3.53592455e-01 4.92556512e-01
3.65281582e-01 -5.32758117e-01 -2.77110755e-01 -2.41962194e-01]
|
[8.847197532653809, -2.1265339851379395]
|
1d2e7e32-cbe4-4b21-ba8b-2d46336e19ed
|
detecting-and-simulating-artifacts-in-gan
|
1907.06515
| null |
https://arxiv.org/abs/1907.06515v2
|
https://arxiv.org/pdf/1907.06515v2.pdf
|
Detecting and Simulating Artifacts in GAN Fake Images
|
To detect GAN generated images, conventional supervised machine learning algorithms require collection of a number of real and fake images from the targeted GAN model. However, the specific model used by the attacker is often unavailable. To address this, we propose a GAN simulator, AutoGAN, which can simulate the artifacts produced by the common pipeline shared by several popular GAN models. Additionally, we identify a unique artifact caused by the up-sampling component included in the common GAN pipeline. We show theoretically such artifacts are manifested as replications of spectra in the frequency domain and thus propose a classifier model based on the spectrum input, rather than the pixel input. By using the simulated images to train a spectrum based classifier, even without seeing the fake images produced by the targeted GAN model during training, our approach achieves state-of-the-art performances on detecting fake images generated by popular GAN models such as CycleGAN.
|
['Shih-Fu Chang', 'Xu Zhang', 'Svebor Karaman']
|
2019-07-15
| null | null | null | null |
['gan-image-forensics']
|
['computer-vision']
|
[ 6.67744160e-01 3.27569455e-01 2.98467219e-01 3.33431393e-01
-9.25810099e-01 -8.72506559e-01 6.90575123e-01 -4.98359770e-01
2.24705949e-01 4.66541409e-01 -3.60542089e-01 -1.98247224e-01
7.49767244e-01 -8.90499830e-01 -1.19305992e+00 -8.06309760e-01
2.57773250e-01 1.44474149e-01 1.02504350e-01 8.42805654e-02
1.30617410e-01 2.62082934e-01 -1.21257854e+00 5.25619626e-01
5.71486235e-01 8.51181865e-01 -2.93178558e-01 9.66524839e-01
2.98151940e-01 1.05969203e+00 -1.57615805e+00 -5.09184182e-01
5.65261602e-01 -1.35228765e+00 -2.55832732e-01 9.44843814e-02
3.69514346e-01 -7.57643223e-01 -2.38150418e-01 1.37529171e+00
2.11607650e-01 -7.49463975e-01 5.88358223e-01 -1.60327435e+00
-3.88151377e-01 6.99006557e-01 -5.95613778e-01 -1.92289978e-01
4.00706768e-01 7.50818551e-01 4.24753249e-01 -2.99916744e-01
6.62110150e-01 8.47500145e-01 6.55905545e-01 6.08729303e-01
-1.39419460e+00 -9.59628105e-01 -7.60257423e-01 5.81048988e-02
-1.12901032e+00 -1.67870551e-01 1.24168873e+00 -2.90401012e-01
4.21981841e-01 3.44806314e-01 6.46979451e-01 1.85791624e+00
2.85138786e-01 3.89422476e-01 1.55109060e+00 -5.86884499e-01
3.02902997e-01 4.71930057e-01 -3.59602779e-01 8.00330877e-01
2.79109180e-01 4.98591810e-01 -5.77951431e-01 -4.62391227e-01
7.72130966e-01 -4.53223288e-01 -5.96546531e-01 -1.09530073e-02
-9.61008132e-01 8.02699387e-01 1.35991201e-01 1.19435750e-01
-2.63066173e-01 7.81069398e-01 2.85926372e-01 4.35668766e-01
1.34282112e-01 6.15959466e-01 2.81380061e-02 -1.04835898e-01
-1.01910388e+00 -2.72793591e-01 6.83564305e-01 9.23326552e-01
8.75538886e-01 4.17000145e-01 4.47996259e-02 -2.84925960e-02
5.46215549e-02 6.64465547e-01 4.40098077e-01 -5.69048405e-01
1.52936384e-01 1.71081841e-01 6.48446903e-02 -7.83126593e-01
1.29721746e-01 -4.95739430e-01 -6.58873260e-01 4.36047703e-01
6.79918051e-01 -1.79863438e-01 -7.96966076e-01 1.42647636e+00
1.71943903e-01 6.11023307e-01 1.01478390e-01 8.74191880e-01
2.60740757e-01 4.19252604e-01 -3.11973840e-01 -1.34445697e-01
1.28990161e+00 -1.00852668e+00 -6.65706694e-01 -7.72684664e-02
7.43643582e-01 -7.91030467e-01 1.22559035e+00 6.02693737e-01
-8.23182821e-01 -4.81504649e-01 -1.57452178e+00 3.73599380e-01
-2.68709421e-01 2.00918332e-01 1.01700060e-01 1.47575486e+00
-7.88700998e-01 4.36113894e-01 -5.05945146e-01 -6.15366781e-03
4.29923892e-01 -8.74349102e-02 -9.24995616e-02 1.35650933e-01
-1.06174386e+00 6.96862936e-01 -4.71048523e-03 -1.86245471e-01
-1.33524978e+00 -7.83896625e-01 -4.61239308e-01 -7.14228898e-02
1.62158147e-01 -4.13997382e-01 1.00842047e+00 -1.62719893e+00
-1.77727127e+00 1.00151956e+00 3.82506639e-01 -6.66783750e-01
1.34979177e+00 -1.26234442e-02 -6.08356237e-01 5.39613247e-01
-2.16194242e-01 -9.52498335e-03 1.69674838e+00 -1.35048330e+00
-4.17607464e-02 7.31645897e-02 -2.41204441e-01 -6.24853849e-01
-8.13055113e-02 -3.52896415e-02 1.87703297e-01 -6.69733465e-01
-4.28772062e-01 -8.98602724e-01 3.47125262e-01 -2.53200501e-01
-8.63406956e-01 7.02121615e-01 1.40487862e+00 -7.17500746e-01
5.50162554e-01 -2.35811591e+00 -6.56966507e-01 3.75102729e-01
6.80417717e-02 3.20279986e-01 -1.23400174e-01 6.45174086e-01
-2.91553050e-01 2.65503049e-01 -1.57997072e-01 -3.04273546e-01
-1.51066750e-01 -3.25799793e-01 -8.39940190e-01 9.75034595e-01
2.77940750e-01 8.87773037e-01 -9.05192494e-01 7.65562057e-02
1.39448538e-01 3.69469315e-01 8.46444145e-02 4.63800579e-01
-3.04276526e-01 7.81452835e-01 -1.74672902e-01 4.66368616e-01
1.04363692e+00 -2.91040212e-01 3.07092965e-01 -3.12762827e-01
4.06943947e-01 3.52371365e-01 -8.30769897e-01 1.29869235e+00
-3.51867139e-01 9.65637207e-01 -1.43368885e-01 -7.82040238e-01
7.57345796e-01 3.99249852e-01 -1.14018749e-02 -5.87611139e-01
2.56891966e-01 3.63256305e-01 -1.06545419e-01 -3.55501473e-01
1.05182223e-01 3.24515291e-02 6.03440043e-04 9.00960088e-01
7.01357871e-02 -3.29535872e-01 -5.23826659e-01 6.51766211e-02
1.55848479e+00 2.29301453e-01 2.44243473e-01 -9.97701939e-03
2.96035349e-01 2.17931420e-01 -3.44188847e-02 1.05113709e+00
-1.11819655e-02 7.16932952e-01 8.28852117e-01 -2.81799078e-01
-1.24785197e+00 -9.35868502e-01 2.61892259e-01 2.44908482e-01
2.10443571e-01 -2.86863834e-01 -1.24837554e+00 -1.01219881e+00
-2.00015426e-01 6.96936488e-01 -6.41082644e-01 -4.18532491e-01
-5.68455458e-01 -5.60588837e-01 1.36587656e+00 6.49435222e-02
8.77637565e-01 -9.28977191e-01 -9.92440939e-01 3.43187004e-02
-7.40784034e-02 -1.43664753e+00 -3.71245831e-01 4.65672538e-02
-3.81255388e-01 -1.36391079e+00 -2.74456948e-01 -2.91024506e-01
7.28700221e-01 9.76544470e-02 1.02603364e+00 4.45344329e-01
-4.53872174e-01 3.17141116e-01 -4.85861510e-01 -4.37936515e-01
-1.39122534e+00 -2.94680864e-01 -4.61203933e-01 3.19003433e-01
-5.33645339e-02 -4.55385238e-01 -4.04401422e-01 2.37843722e-01
-1.05589926e+00 2.00124919e-01 4.99093622e-01 7.97937214e-01
1.91294298e-01 2.94976830e-01 3.36323202e-01 -1.12519121e+00
2.56853074e-01 -2.10286245e-01 -1.00085092e+00 8.22315440e-02
-4.34660614e-01 -1.08353525e-01 1.09098506e+00 -7.96101928e-01
-7.72063613e-01 -4.20552604e-02 1.15698390e-01 -6.76166415e-01
-3.88254225e-01 -3.35015021e-02 -3.54975075e-01 -4.27866548e-01
9.22623634e-01 4.59127784e-01 -3.93652320e-02 -1.14659272e-01
2.43208990e-01 5.57869315e-01 6.91367745e-01 -5.97616993e-02
1.42257702e+00 6.72005475e-01 2.33791918e-01 -8.76334429e-01
-2.44702548e-01 1.68463558e-01 2.85510011e-02 -4.30988997e-01
6.03541613e-01 -8.72374117e-01 -4.72976565e-01 1.21534550e+00
-1.43300223e+00 -5.42069316e-01 -3.00502241e-01 1.75414369e-01
-4.80248392e-01 3.83873880e-01 -6.52899802e-01 -9.34567809e-01
-3.02025318e-01 -1.15308750e+00 1.16896117e+00 -8.26462209e-02
9.35033057e-03 -7.62248337e-01 -1.96662709e-01 3.70495826e-01
4.21177298e-01 7.31518507e-01 8.89569163e-01 -5.74669838e-01
-8.92507970e-01 -4.13601458e-01 7.21634999e-02 4.75960016e-01
2.82489836e-01 6.87462837e-02 -1.58957160e+00 -2.78244704e-01
6.82164252e-01 -3.11615527e-01 6.33735299e-01 -8.97000581e-02
1.08691502e+00 -5.06735682e-01 -6.99755549e-02 7.55970478e-01
1.75114918e+00 -1.67010985e-02 1.03287566e+00 6.12418652e-02
6.82211101e-01 3.47399205e-01 9.78606343e-02 8.17307979e-02
-4.64532912e-01 5.04626513e-01 7.69919813e-01 -2.59690076e-01
-3.05366218e-01 -5.63034296e-01 7.66183436e-01 1.81971014e-01
4.18750525e-01 -4.90907788e-01 -4.40109015e-01 9.80816185e-02
-1.25240064e+00 -1.03389060e+00 -5.54365277e-01 2.26082659e+00
4.86287624e-01 1.36218742e-01 8.83180872e-02 2.67540604e-01
8.43342423e-01 1.11242823e-01 -4.98100460e-01 -2.69140512e-01
-2.80095398e-01 5.31831801e-01 9.15744185e-01 1.66754141e-01
-9.16131139e-01 8.81248355e-01 6.90493965e+00 9.10342991e-01
-1.50220430e+00 3.77128631e-01 6.34240627e-01 2.51762003e-01
-2.62974262e-01 7.64817894e-02 -2.22090073e-02 7.84176469e-01
1.00742853e+00 2.02442870e-01 5.53861439e-01 8.50066662e-01
1.64636582e-01 -2.52528965e-01 -1.05706346e+00 9.94490504e-01
3.20435762e-01 -1.33804166e+00 -1.40181825e-01 3.25612783e-01
7.08499730e-01 -4.50646937e-01 1.87355965e-01 -2.85306871e-01
2.71019727e-01 -1.15215313e+00 8.06711793e-01 2.01439202e-01
1.05185318e+00 -4.90574986e-01 7.53882706e-01 4.38691199e-01
-7.08271623e-01 3.91105324e-01 -2.37635784e-02 1.38719589e-01
-3.06411237e-02 8.48983526e-01 -1.24559450e+00 5.28826356e-01
1.97320506e-01 1.21238433e-01 -6.59483612e-01 5.76176345e-01
-8.08720529e-01 1.32898390e+00 -2.03948259e-01 2.76130676e-01
-9.61167365e-02 -1.42236188e-01 5.63259661e-01 9.07717705e-01
5.53228498e-01 -6.94117308e-01 -3.82184029e-01 1.35417652e+00
-3.19651723e-01 -5.09808719e-01 -7.63115168e-01 -2.45860025e-01
3.03343475e-01 1.21961296e+00 -8.18129063e-01 -2.52070129e-01
-1.95113763e-01 1.59063244e+00 -1.77106380e-01 2.84705669e-01
-1.33869720e+00 -6.50116205e-02 3.12270701e-01 8.99044275e-02
2.06723735e-01 1.88056543e-01 -2.35560089e-01 -1.04374623e+00
3.39032203e-01 -1.24187505e+00 -2.42976192e-02 -9.65679884e-01
-1.23085165e+00 3.79683733e-01 -4.11342144e-01 -1.37139666e+00
-3.48452330e-01 -4.48912382e-01 -9.02658880e-01 9.00161803e-01
-1.28028035e+00 -1.39748955e+00 -6.75131440e-01 4.64719832e-01
1.42064020e-01 -1.13583095e-01 9.11782980e-01 -8.07211772e-02
-1.68188274e-01 6.26547992e-01 -7.38294721e-02 3.75086963e-01
7.95686841e-01 -1.11762702e+00 7.66252220e-01 1.16408491e+00
3.24835926e-01 -3.66146385e-04 9.57201898e-01 -6.87948227e-01
-1.55970490e+00 -1.10674512e+00 8.14369991e-02 -3.39077383e-01
6.55657589e-01 -6.21178806e-01 -7.08037913e-01 7.28676975e-01
4.13984865e-01 6.00960664e-02 4.78161693e-01 -9.60755587e-01
-7.43772805e-01 1.55714840e-01 -1.54895663e+00 1.81984916e-01
7.20353544e-01 -9.10340786e-01 -9.33437981e-03 3.74594003e-01
4.82012659e-01 -4.00987834e-01 -2.01615289e-01 -2.61007130e-01
5.48433363e-01 -1.32196808e+00 5.99982798e-01 -3.01140875e-01
4.56307083e-01 -4.41328257e-01 2.27015495e-01 -1.49059176e+00
3.04917067e-01 -9.86813188e-01 -2.91980982e-01 1.12549639e+00
1.71496034e-01 -9.20256078e-01 8.31993461e-01 -1.69406950e-01
2.88721740e-01 5.34087885e-04 -8.28491867e-01 -1.06130040e+00
-1.45631939e-01 -3.17629308e-01 8.24058294e-01 1.07038462e+00
-3.49541813e-01 -6.66733384e-02 -8.68377686e-01 5.67018151e-01
9.14138377e-01 -3.72134186e-02 1.23395276e+00 -6.58662379e-01
-7.34784007e-01 -7.61575773e-02 -7.00614750e-01 -7.67667651e-01
4.83001173e-02 -4.24874574e-01 -2.28821067e-03 -5.56486368e-01
-1.52225792e-01 -1.38483331e-01 2.18077123e-01 1.98017120e-01
4.42067347e-02 7.74930656e-01 1.48738965e-01 3.45287651e-01
2.22425774e-01 3.09486501e-02 8.80093038e-01 -2.29309902e-01
1.84432894e-01 -2.48240799e-01 -3.41048956e-01 6.30415618e-01
7.60751903e-01 -9.31980371e-01 -2.70927310e-01 -2.03554884e-01
2.71752864e-01 5.69400303e-02 1.02198648e+00 -1.40368116e+00
-2.89803483e-02 1.30517274e-01 4.64200735e-01 -4.31512669e-02
2.39772126e-01 -9.37292457e-01 7.44235694e-01 7.47407973e-01
2.26497254e-03 -2.54047245e-01 -9.69784558e-02 4.91303831e-01
-1.21260835e-02 -2.50299245e-01 9.65034366e-01 -1.72379956e-01
9.75807849e-03 -2.97810346e-01 -5.20200372e-01 -1.89837292e-01
1.12045586e+00 -1.10378437e-01 -9.60358262e-01 -5.34818470e-01
-1.04972757e-01 -6.49763346e-01 9.71989751e-01 -7.11987093e-02
2.81043410e-01 -1.00123811e+00 -4.85023260e-01 5.03023326e-01
1.33140478e-02 -4.72365916e-01 8.68263319e-02 5.52114904e-01
-8.87495697e-01 -2.09987402e-01 -1.59821674e-01 -4.70333338e-01
-1.14310670e+00 6.78807080e-01 5.88443339e-01 -2.89495140e-01
-6.88048184e-01 3.55769783e-01 4.72872294e-02 1.12633638e-01
-3.50818276e-01 -1.10414229e-01 6.19442105e-01 -3.54438812e-01
5.03306448e-01 2.39492804e-01 3.03451508e-01 -5.09559751e-01
-1.97068945e-01 1.96882531e-01 5.37609220e-01 -2.41861969e-01
7.46486962e-01 4.02592689e-01 -5.79781551e-03 2.38455206e-01
1.24629748e+00 5.52971959e-01 -1.16661119e+00 1.93072751e-01
-3.45205009e-01 -5.82572699e-01 -1.86115891e-01 -9.39905047e-01
-1.11723399e+00 7.17324197e-01 7.47296810e-01 6.05859876e-01
1.24173105e+00 -1.72983363e-01 7.63788879e-01 -1.65897593e-01
6.94470942e-01 -8.16433966e-01 2.99539208e-01 -2.68732578e-01
5.60012579e-01 -8.55935454e-01 -2.28921294e-01 -7.75435805e-01
-2.13079885e-01 1.21305048e+00 3.25332403e-01 -4.21131611e-01
2.09366724e-01 4.14318204e-01 4.50468332e-01 -3.47691625e-02
-2.86195099e-01 3.28727722e-01 -2.87091136e-01 8.41894507e-01
-2.32463643e-01 1.24604687e-01 1.05295993e-01 2.30042383e-01
-4.71766591e-01 -1.98447123e-01 1.19116688e+00 8.20279181e-01
8.53504241e-02 -1.10065067e+00 -7.83045650e-01 2.19052017e-01
-5.80501556e-01 -2.43359879e-02 -8.41008723e-01 7.32666254e-01
2.20394477e-01 1.06440115e+00 -6.07981384e-02 -5.11331737e-01
-1.41514137e-01 2.25231588e-01 5.35614014e-01 -2.13888034e-01
-9.40563202e-01 3.15950401e-02 -1.05779208e-02 -7.45652676e-01
-3.10264766e-01 -3.09672624e-01 -8.04808795e-01 -2.27414608e-01
-5.76471090e-01 -1.33864239e-01 8.03380728e-01 9.81802106e-01
2.83129930e-01 4.95782673e-01 9.99603748e-01 -5.90973377e-01
-6.61129713e-01 -1.02046347e+00 -7.79564261e-01 5.78342319e-01
6.43823385e-01 -1.89584047e-01 -1.04804862e+00 3.90061349e-01]
|
[12.385367393493652, 1.0355428457260132]
|
736b4286-bf87-4e0e-b0f4-15d4b8233450
|
conflict-based-search-for-connected-multi
|
2006.03280
| null |
https://arxiv.org/abs/2006.03280v1
|
https://arxiv.org/pdf/2006.03280v1.pdf
|
Conflict-Based Search for Connected Multi-Agent Path Finding
|
We study a variant of the multi-agent path finding problem (MAPF) in which agents are required to remain connected to each other and to a designated base. This problem has applications in search and rescue missions where the entire execution must be monitored by a human operator. We re-visit the conflict-based search algorithm known for MAPF, and define a variant where conflicts arise from disconnections rather than collisions. We study optimizations, and give experimental results in which we compare our algorithms with the literature.
|
['François Schwarzentruber', 'Ocan Sankur', 'Arthur Queffelec']
|
2020-06-05
| null | null | null | null |
['multi-agent-path-finding']
|
['playing-games']
|
[ 7.95360878e-02 1.64578080e-01 -2.35573709e-01 -5.12544066e-02
-1.93766758e-01 -9.88363147e-01 5.04608750e-01 6.96490288e-01
-9.17623937e-01 1.45508313e+00 -9.90945473e-02 -1.64751142e-01
-1.09226048e+00 -1.13473010e+00 -2.82779038e-01 -6.86216056e-01
-8.36110234e-01 1.40974045e+00 8.05402517e-01 -8.96609306e-01
3.53171706e-01 8.34974706e-01 -8.91042054e-01 -3.46611381e-01
5.05294561e-01 5.12004316e-01 6.91443160e-02 8.14192235e-01
2.24010393e-01 4.01408374e-01 -8.52372825e-01 1.29814863e-01
4.62381899e-01 -2.39458218e-01 -1.41585350e+00 1.26695961e-01
-6.10180914e-01 -1.66669533e-01 -1.68495730e-01 6.64003015e-01
3.70854616e-01 5.85444450e-01 5.68042397e-01 -2.17350864e+00
1.93762869e-01 4.17365283e-01 -6.75055861e-01 3.62818718e-01
8.49090576e-01 -2.30687112e-01 6.54731512e-01 -1.12351306e-01
9.12745297e-01 9.93033409e-01 5.69846690e-01 5.16900659e-01
-1.01313710e+00 1.14966901e-02 4.42028493e-01 5.68757236e-01
-1.22409189e+00 -3.51388961e-01 1.43817335e-01 3.73820141e-02
1.45270789e+00 5.25991380e-01 5.30857861e-01 4.50842917e-01
4.44030315e-01 1.46646425e-01 6.94052100e-01 -4.78317410e-01
3.51744711e-01 -4.36390340e-01 1.63884968e-01 6.15693748e-01
6.05740309e-01 2.44175777e-01 -2.54899710e-01 -7.11796284e-01
4.63977218e-01 -2.53866702e-01 -6.43298566e-01 -6.06440425e-01
-1.26488161e+00 1.07280171e+00 3.68495911e-01 1.91815749e-01
-7.87011743e-01 2.77392536e-01 1.15596585e-01 5.32577395e-01
-1.96293846e-01 4.62936580e-01 -4.05240536e-01 1.08067854e-03
-3.16064268e-01 5.08785903e-01 1.25670350e+00 8.85049462e-01
7.58896589e-01 -4.59238738e-01 4.71478105e-01 3.03379029e-01
1.20447919e-01 2.16826171e-01 -2.48908401e-01 -1.29408550e+00
3.48115861e-01 4.88821447e-01 6.31976664e-01 -1.02024829e+00
-1.05880487e+00 9.61444080e-02 -3.86803716e-01 7.63656020e-01
3.19691859e-02 -2.41993919e-01 -3.18969011e-01 1.51137841e+00
6.08142436e-01 5.18005490e-02 4.67402548e-01 7.97713399e-01
2.00103939e-01 6.94987476e-01 -4.36026871e-01 -8.07053924e-01
9.28631783e-01 -1.27833068e+00 -5.55438042e-01 -2.71320283e-01
6.30627692e-01 -3.89796048e-01 1.90526500e-01 3.44407797e-01
-1.21928751e+00 4.85927701e-01 -1.06983495e+00 5.31181097e-01
-4.02595073e-01 -1.03885520e+00 4.16897327e-01 3.07083249e-01
-1.42540693e+00 4.89734948e-01 -8.37256134e-01 -6.83965683e-01
-2.17387110e-01 8.12303782e-01 -6.25835240e-01 -5.96956573e-02
-8.82785976e-01 1.24669755e+00 5.95750332e-01 2.65617762e-02
-8.84534299e-01 1.26642957e-01 -6.81796789e-01 -2.87503123e-01
9.64421332e-01 -9.12228942e-01 1.25209785e+00 -4.02040422e-01
-9.70330775e-01 3.98007512e-01 -8.82050768e-02 -4.27053273e-01
3.85402828e-01 2.68922597e-01 -2.85826623e-01 2.23560110e-01
4.84315544e-01 1.33807391e-01 -1.70657143e-01 -1.57087970e+00
-1.07464218e+00 -1.88024282e-01 6.76594317e-01 3.99174958e-01
1.67964920e-01 1.73213586e-01 -2.67938673e-01 -5.58360256e-02
-8.13196227e-02 -1.23255050e+00 -7.86133528e-01 -2.82903552e-01
-3.66934866e-01 -2.74251252e-01 8.25803280e-01 3.10692713e-02
1.19697213e+00 -1.34346247e+00 6.16072655e-01 7.24529147e-01
8.63035843e-02 -3.46022606e-01 -3.09747726e-01 1.29436338e+00
3.66709858e-01 1.34964615e-01 -2.98781097e-01 -1.87362522e-01
-9.83544216e-02 9.24867213e-01 9.36742797e-02 6.68811858e-01
-4.58852798e-01 4.01931316e-01 -1.12404108e+00 -5.56141973e-01
-2.65137196e-01 -3.49679232e-01 -4.56258595e-01 -1.19755752e-01
-2.01469198e-01 6.81813359e-02 -6.37239456e-01 6.67425632e-01
4.62876469e-01 -1.09474525e-01 3.19594055e-01 5.74664533e-01
-5.25666118e-01 -9.09849778e-02 -1.25402570e+00 1.69529641e+00
-1.10559069e-01 2.76229650e-01 8.07935059e-01 -9.21750546e-01
4.25874501e-01 2.75277823e-01 9.11181986e-01 -4.10114557e-01
1.09367892e-01 1.50266752e-01 -1.29795521e-01 -3.73560578e-01
6.74935818e-01 -1.08400255e-01 -4.56696928e-01 9.85490739e-01
-6.38071895e-01 -1.75057948e-02 6.61412954e-01 3.99250299e-01
1.80597794e+00 -5.29689670e-01 5.19798219e-01 -3.49866182e-01
3.58911693e-01 8.92714739e-01 7.20003307e-01 9.85842407e-01
-3.18661094e-01 -2.74163038e-01 1.87017500e-01 -6.34549558e-01
-3.74864608e-01 -1.11276627e+00 4.62709069e-01 9.54846799e-01
1.00347984e+00 -5.67011356e-01 -3.01094979e-01 -6.48739100e-01
-6.98034791e-03 5.62783599e-01 -6.56915903e-01 1.13033332e-01
-1.13583100e+00 -6.18603826e-01 2.01189682e-01 1.72712073e-01
2.21797913e-01 -8.91894758e-01 -1.17290020e+00 7.10300386e-01
-4.51298326e-01 -8.42901111e-01 -4.84427840e-01 1.59623772e-01
-3.61874729e-01 -1.72143757e+00 -1.91087410e-01 -7.72491157e-01
6.21154249e-01 6.96538925e-01 9.75199997e-01 6.18522704e-01
-1.14522390e-01 8.17991972e-01 -6.84679985e-01 -2.69167215e-01
-3.16340357e-01 1.79711238e-01 4.08889264e-01 -5.95817566e-01
-2.59881467e-01 -5.39849937e-01 -3.90116066e-01 8.76040280e-01
-6.56966746e-01 -4.24176216e-01 6.96185157e-02 5.25252402e-01
6.37080312e-01 6.42321110e-01 6.35531127e-01 -3.94556612e-01
9.66630220e-01 -7.08978474e-01 -6.74068987e-01 5.55277288e-01
-6.84325755e-01 -3.16116273e-01 3.01562786e-01 -5.07472381e-02
-3.54530275e-01 1.05317403e-02 3.65059704e-01 1.71050839e-02
9.96578336e-02 6.14016473e-01 -2.39035673e-02 -6.08754933e-01
3.07430238e-01 -1.64833501e-01 -1.00745186e-02 -5.55624999e-02
-2.92270556e-02 1.44294500e-01 6.03308320e-01 -5.85402489e-01
6.72304451e-01 5.82992792e-01 4.55477387e-01 -4.22178924e-01
-2.04019900e-02 -3.80521148e-01 -3.47140849e-01 -3.93430680e-01
5.36500335e-01 -1.81086093e-01 -1.33088934e+00 -1.09446079e-01
-1.52596569e+00 -3.88715804e-01 -4.24249396e-02 1.94422379e-01
-8.55748713e-01 3.14653188e-01 -3.67940873e-01 -9.07745063e-01
1.38044273e-02 -8.89502466e-01 5.83246708e-01 2.34171122e-01
-1.84133321e-01 -1.09426594e+00 7.08238184e-01 1.48305576e-02
4.06335026e-01 7.81713665e-01 7.65120864e-01 -7.85646617e-01
-5.25811136e-01 3.65175977e-02 2.17570096e-01 -9.25591171e-01
3.63352209e-01 -3.76363248e-01 2.24286333e-01 -8.00230443e-01
-2.24401265e-01 8.00093338e-02 2.86511481e-01 1.95669085e-01
2.04920903e-01 -7.30148613e-01 -1.20981801e+00 6.77712336e-02
1.61556351e+00 8.09724510e-01 3.80790830e-01 1.19104731e+00
-1.87686250e-01 8.50816011e-01 9.63062763e-01 6.19047880e-01
8.63622069e-01 8.52926731e-01 8.82421553e-01 2.21856460e-01
5.80634952e-01 4.40125555e-01 -3.44756357e-02 1.73897639e-01
-3.81453425e-01 -1.04763973e+00 -1.21076119e+00 7.49595404e-01
-2.51995826e+00 -1.00236821e+00 -1.62126511e-01 1.84795105e+00
2.99698740e-01 1.61142033e-02 5.09974301e-01 2.34566331e-01
7.95275390e-01 -1.35923028e-01 -3.93389374e-01 -4.95806396e-01
-1.35428101e-01 -1.47773668e-01 8.45677078e-01 1.10448694e+00
-8.94058883e-01 7.04846203e-01 7.55060577e+00 1.40145883e-01
-4.86282140e-01 2.17971370e-01 -2.12349683e-01 -3.55270386e-01
-6.57578558e-02 1.63609684e-01 -3.50798517e-01 -4.41663787e-02
7.96947122e-01 -5.58074892e-01 7.82646477e-01 3.10412109e-01
4.36566740e-01 -5.22295117e-01 -9.11388218e-01 6.15978003e-01
1.60721259e-03 -1.19641268e+00 -3.40680271e-01 3.33544672e-01
4.88950193e-01 -5.86868264e-03 -6.94933057e-01 -3.35415900e-01
7.18441546e-01 -9.18119788e-01 5.91634274e-01 1.59901455e-01
1.85325712e-01 -1.10164559e+00 5.14111221e-01 5.96013129e-01
-1.44985986e+00 -4.54835773e-01 8.43232498e-02 -2.60637522e-01
1.15708792e+00 -1.80910796e-01 -9.49202001e-01 1.14439690e+00
7.00872600e-01 4.61077355e-02 3.90968621e-01 1.46145713e+00
-1.06131351e-02 -4.02169019e-01 -6.00671947e-01 -1.78230539e-01
5.48660219e-01 -3.29531699e-01 1.07382774e+00 7.17497826e-01
2.20856726e-01 6.34299099e-01 7.23994136e-01 1.10771269e-01
4.76647228e-01 -2.22010270e-01 -6.24004602e-01 3.80192667e-01
7.43246078e-01 9.05994296e-01 -1.16956043e+00 7.17160553e-02
-2.65364815e-02 1.06208014e+00 1.13714680e-01 1.82178825e-01
-8.02486300e-01 -7.22741902e-01 9.51322973e-01 5.98073155e-02
-2.08061486e-02 -6.41818047e-01 1.05444707e-01 -3.26889098e-01
-6.34517148e-02 -5.41181386e-01 8.42547596e-01 -5.04412293e-01
-8.30990493e-01 1.16304278e+00 4.95360821e-01 -1.06821036e+00
-4.69313025e-01 -1.47673324e-01 -8.40424180e-01 4.31720376e-01
-1.51128650e+00 -7.17441738e-01 -1.75252661e-01 8.55969429e-01
3.71035129e-01 -1.18609548e-01 1.06694233e+00 1.47773311e-01
-4.62502778e-01 -1.71592180e-02 -2.61395156e-01 -4.91551936e-01
5.18220477e-02 -8.81963432e-01 9.85089242e-02 8.90132844e-01
-4.22053277e-01 3.17580521e-01 9.19220984e-01 -9.96111572e-01
-1.60271955e+00 -7.35521019e-01 8.14206600e-01 -1.08212464e-01
6.20906532e-01 2.11464256e-01 -2.28268817e-01 1.00891364e+00
4.57763880e-01 -2.68871099e-01 5.12813926e-01 -1.59982592e-01
3.75051618e-01 1.55112088e-01 -1.39708364e+00 5.12540877e-01
1.39407265e+00 2.85341084e-01 -4.91155535e-01 3.05200100e-01
6.14153683e-01 -2.66867787e-01 -3.63499403e-01 4.70391423e-01
2.30258659e-01 -8.16034973e-01 9.78966415e-01 -7.70539224e-01
-5.45409918e-01 -7.70338178e-01 -1.74121782e-01 -1.80179369e+00
-4.52015430e-01 -9.85272586e-01 3.47674996e-01 6.72996759e-01
5.90111017e-01 -1.12591088e+00 7.94803381e-01 5.49336255e-01
-3.77480030e-01 -6.44281030e-01 -1.53646374e+00 -1.13106728e+00
-3.23639631e-01 1.60578936e-01 7.80116260e-01 8.53315830e-01
4.73140419e-01 2.52433151e-01 -3.40174377e-01 7.15052187e-01
6.32501185e-01 3.59463066e-01 5.70103526e-01 -1.25542414e+00
-2.66299158e-01 -4.34008479e-01 -4.20675054e-02 -6.98704541e-01
2.05900803e-01 -5.34703493e-01 3.11773866e-01 -2.20536900e+00
-2.54919678e-01 -9.88684297e-01 -3.52043808e-02 7.61280715e-01
5.86708665e-01 -1.68905661e-01 3.46958041e-01 2.75870055e-01
-9.53681409e-01 3.18409443e-01 7.82854199e-01 -2.13479996e-01
-3.91090035e-01 2.03649893e-01 -4.74812269e-01 5.17253160e-01
9.98524964e-01 -8.45607996e-01 -4.53202188e-01 -5.84442556e-01
4.35621172e-01 8.21542561e-01 3.99095900e-02 -7.77672112e-01
8.93051803e-01 -1.03499854e+00 -5.73042214e-01 -5.21918654e-01
6.78126752e-01 -1.24109340e+00 7.08384991e-01 1.06716144e+00
1.57930121e-01 1.06728935e+00 3.79530750e-02 6.12548769e-01
-5.43877445e-02 -5.44740081e-01 3.75926793e-01 -2.06397772e-01
-7.29737818e-01 2.87783414e-01 -7.95298159e-01 -3.37292641e-01
1.79916501e+00 -3.35862309e-01 -9.65189934e-01 -6.21889412e-01
-7.50857294e-01 9.79516923e-01 6.54239416e-01 -1.24040665e-02
9.08262253e-01 -1.08142078e+00 -6.43656313e-01 -5.22790194e-01
-4.23052609e-02 -1.85005516e-01 -1.46433860e-01 9.29687917e-01
-8.05845916e-01 3.30406070e-01 -3.05107534e-01 7.66690299e-02
-1.44321597e+00 8.88154924e-01 3.73094589e-01 -2.28613481e-01
-4.82095778e-01 4.72280979e-01 -3.65145206e-01 -1.12679325e-01
2.05341667e-01 1.49772391e-01 -2.03816056e-01 -8.24468136e-02
7.70978749e-01 9.38647866e-01 -2.85257131e-01 -5.35360038e-01
-1.12180102e+00 6.14095569e-01 2.22335845e-01 -5.43739617e-01
1.61038697e+00 -5.32284141e-01 -5.41439235e-01 -8.14987272e-02
5.12979686e-01 1.13699265e-01 -4.56241131e-01 8.40213746e-02
2.90105432e-01 -5.49765170e-01 -3.07570875e-01 -7.95826018e-01
-8.53498280e-01 -2.45545328e-01 -9.22491029e-03 7.75105476e-01
1.27655900e+00 2.91866124e-01 7.18700886e-01 7.84832120e-01
1.23903680e+00 -9.50734198e-01 -1.03773728e-01 7.17129886e-01
1.02123487e+00 -6.42108381e-01 9.37378183e-02 -5.82162261e-01
-4.48724508e-01 9.80873227e-01 7.06488013e-01 -1.74686506e-01
4.69768733e-01 5.56759596e-01 -1.57859102e-01 -3.79293352e-01
-9.83464956e-01 -2.75735825e-01 -7.49955714e-01 8.77606153e-01
-7.57274210e-01 -2.22180746e-02 -5.93407214e-01 2.82371640e-01
-1.29188448e-01 -2.59619683e-01 1.05585575e+00 1.84592414e+00
-8.59560490e-01 -1.56173086e+00 -5.94386995e-01 5.22395559e-02
1.52114585e-01 2.97961205e-01 -7.33916223e-01 8.85841787e-01
9.45562869e-02 1.61857295e+00 3.96028571e-02 -3.82721007e-01
6.45128846e-01 -4.62685436e-01 5.14192224e-01 -2.77083278e-01
-5.53844690e-01 -1.39560029e-01 8.19703043e-01 -6.36999309e-01
-8.12149942e-01 -7.34874964e-01 -1.79737306e+00 -6.28564060e-01
-1.63498163e-01 9.02477205e-01 4.56212908e-01 8.33555400e-01
6.96639270e-02 3.79856706e-01 7.96706557e-01 -8.22460711e-01
1.01240743e-02 -3.52063090e-01 -4.89299327e-01 -5.61847329e-01
6.10727310e-01 -8.99873972e-01 -2.57343948e-01 -6.29902244e-01]
|
[4.937862873077393, 1.7267073392868042]
|
eb89581c-5022-48f8-9f06-3340402b902f
|
depth-adaptive-computational-policies-for
|
1801.00508
| null |
http://arxiv.org/abs/1801.00508v1
|
http://arxiv.org/pdf/1801.00508v1.pdf
|
Depth-Adaptive Computational Policies for Efficient Visual Tracking
|
Current convolutional neural networks algorithms for video object tracking
spend the same amount of computation for each object and video frame. However,
it is harder to track an object in some frames than others, due to the varying
amount of clutter, scene complexity, amount of motion, and object's
distinctiveness against its background. We propose a depth-adaptive
convolutional Siamese network that performs video tracking adaptively at
multiple neural network depths. Parametric gating functions are trained to
control the depth of the convolutional feature extractor by minimizing a joint
loss of computational cost and tracking error. Our network achieves accuracy
comparable to the state-of-the-art on the VOT2016 benchmark. Furthermore, our
adaptive depth computation achieves higher accuracy for a given computational
cost than traditional fixed-structure neural networks. The presented framework
extends to other tasks that use convolutional neural networks and enables
trading speed for accuracy at runtime.
|
['Katerina Fragkiadaki', 'Chris Ying']
|
2018-01-01
| null | null | null | null |
['video-object-tracking']
|
['computer-vision']
|
[-5.29385172e-02 -3.80833179e-01 -4.18627828e-01 -1.80170491e-01
-6.02599919e-01 -6.82406008e-01 2.98577040e-01 -1.24749809e-03
-9.32064235e-01 2.33744055e-01 -1.20816521e-01 2.90148985e-02
1.77315295e-01 -5.79747558e-01 -1.04502249e+00 -4.94926363e-01
-3.21960121e-01 2.86316723e-01 8.65778029e-01 2.75873542e-01
-1.01955710e-02 6.71718657e-01 -1.38938916e+00 1.63435847e-01
3.07584792e-01 1.46464241e+00 2.95307904e-01 9.00057733e-01
4.51858900e-02 1.00901484e+00 -4.76212531e-01 -3.41578096e-01
4.58535552e-01 -1.06496245e-01 -6.14317417e-01 2.04212189e-01
1.17052352e+00 -6.07768595e-01 -8.32349598e-01 1.36476934e+00
2.22913176e-01 2.25354433e-01 3.43669444e-01 -1.25064838e+00
-5.76487184e-01 3.87581646e-01 -4.46565807e-01 7.34312057e-01
-1.58773139e-01 3.56693715e-01 8.91467094e-01 -7.25245893e-01
7.17794955e-01 1.09576142e+00 8.06347609e-01 7.86723912e-01
-1.04991698e+00 -7.56746233e-01 6.62604511e-01 2.67580003e-01
-1.15290451e+00 -3.37755293e-01 4.38424677e-01 -6.34435058e-01
9.87231195e-01 -2.25559130e-01 9.64422464e-01 9.49566424e-01
1.36210576e-01 1.12552607e+00 2.56638259e-01 1.07437633e-01
2.24399462e-01 -3.89774024e-01 3.91792804e-02 9.59266305e-01
4.15738791e-01 2.28950679e-01 -4.33450878e-01 1.68393478e-01
9.31680143e-01 2.85673141e-01 -2.92401582e-01 -6.90087199e-01
-1.17073309e+00 7.18420208e-01 7.18715727e-01 9.93474722e-02
-2.14064375e-01 1.08260119e+00 6.36724055e-01 1.92428693e-01
1.14090420e-01 2.64519602e-01 -8.16705763e-01 -2.16342881e-01
-1.16731739e+00 3.89366060e-01 5.35107613e-01 1.12315392e+00
4.07054126e-01 3.44193131e-01 -4.29985017e-01 1.45753592e-01
3.46522450e-01 3.40557665e-01 2.53338337e-01 -1.16986120e+00
4.25944924e-01 4.95008826e-01 1.81479678e-01 -6.92821801e-01
-3.36535841e-01 -6.15825713e-01 -4.81519729e-01 5.28172433e-01
7.89175391e-01 -1.54914051e-01 -1.08166730e+00 1.85619235e+00
3.45478684e-01 4.85339612e-01 -1.55561537e-01 1.11269414e+00
8.35979819e-01 4.07163650e-01 2.62706012e-01 7.14697242e-02
1.41792309e+00 -1.29264152e+00 -5.40827215e-01 -3.40484589e-01
5.98110616e-01 -5.15111864e-01 7.58293867e-01 2.09088758e-01
-1.33411968e+00 -6.10057831e-01 -1.05083179e+00 -1.64030477e-01
-1.66389406e-01 2.40594238e-01 7.37006783e-01 5.86199105e-01
-9.53153968e-01 6.62117481e-01 -1.23881865e+00 -5.92640601e-02
1.00791883e+00 6.47838414e-01 -1.25014767e-01 1.45911977e-01
-8.07704151e-01 5.10775983e-01 4.31413293e-01 1.18594363e-01
-1.23289287e+00 -9.54515219e-01 -8.60362113e-01 2.67088652e-01
3.83617222e-01 -7.02878952e-01 1.33356380e+00 -1.37562454e+00
-1.25496459e+00 7.05424488e-01 1.57081917e-01 -8.56568575e-01
7.98471987e-01 -5.42529762e-01 -9.70823988e-02 1.92756653e-01
-1.23773724e-01 9.60004091e-01 9.98440802e-01 -7.26080239e-01
-8.52159679e-01 -3.60942572e-01 3.13458830e-01 -1.31807411e-02
-4.02604610e-01 1.52250707e-01 -1.09178948e+00 -7.98665285e-01
-1.25551224e-01 -9.53363776e-01 -3.44857275e-01 7.67761409e-01
1.10706091e-01 -2.10052475e-01 1.11384583e+00 -2.77972162e-01
1.09812558e+00 -2.21847415e+00 1.52516901e-01 -2.51988620e-01
4.63357627e-01 5.07401288e-01 -2.73349673e-01 -3.54986519e-01
3.19790959e-01 -2.67031848e-01 4.85434681e-02 -2.54553527e-01
-1.33543804e-01 -2.09588185e-02 7.73319453e-02 6.84476852e-01
2.97371179e-01 1.23395836e+00 -9.06663001e-01 -5.22573471e-01
1.68766484e-01 6.08908951e-01 -9.21942770e-01 1.88383441e-02
-4.82437909e-01 1.86878979e-01 -4.46110547e-01 6.73384666e-01
6.20007694e-01 -6.33581579e-01 1.09298728e-01 -2.57934481e-01
-1.28597513e-01 -5.19475304e-02 -1.10280561e+00 1.87540388e+00
-1.10319234e-01 1.09955621e+00 1.05686620e-01 -6.49352670e-01
4.51090515e-01 1.12749487e-01 4.91815120e-01 -5.16431570e-01
4.17113900e-01 5.99394776e-02 1.58290580e-01 -3.77140552e-01
5.51456690e-01 3.57899398e-01 1.69236496e-01 -2.74774488e-02
2.58633494e-01 3.34030420e-01 3.13141257e-01 6.87481742e-03
1.18214941e+00 3.69306743e-01 -3.02553773e-02 -2.79455304e-01
3.59771609e-01 -4.45428342e-02 6.31092966e-01 7.01903880e-01
-5.91321528e-01 3.49795341e-01 3.72812271e-01 -9.30216253e-01
-1.04554367e+00 -7.19557643e-01 1.92050561e-02 1.25098073e+00
3.63340586e-01 -1.53510749e-01 -6.96680605e-01 -7.37716317e-01
2.20425993e-01 9.65336785e-02 -8.41094255e-01 -4.81534377e-02
-9.29112494e-01 -2.40341142e-01 6.59717321e-01 1.11451447e+00
4.59331781e-01 -1.03359294e+00 -1.24359632e+00 4.32776093e-01
2.66068190e-01 -1.58540690e+00 -8.97528887e-01 2.41780668e-01
-1.02141559e+00 -1.23158920e+00 -6.03578091e-01 -9.39271986e-01
4.84176666e-01 2.44349048e-01 1.26229596e+00 2.75593489e-01
-3.08304697e-01 2.22752094e-01 -3.74472104e-02 -2.75752902e-01
3.88046652e-02 2.55622774e-01 -2.16023907e-01 -1.30722836e-01
4.48084235e-01 -5.54288477e-02 -9.36709940e-01 2.36787185e-01
-8.65740716e-01 -3.71485025e-01 2.52820492e-01 7.11019158e-01
5.34275115e-01 -7.70163536e-02 4.92417477e-02 -5.49539506e-01
-3.52216959e-02 -1.53664008e-01 -1.32020748e+00 2.82422360e-02
-1.63404003e-01 7.35279620e-02 3.88556033e-01 -8.45837057e-01
-4.79102314e-01 4.47367072e-01 9.91267934e-02 -1.08210087e+00
1.48694664e-01 -9.46333334e-02 1.02764837e-01 -4.02874857e-01
3.22585762e-01 -2.22767610e-03 -7.38072693e-02 -1.84262410e-01
2.17552140e-01 -1.67359889e-01 5.78186452e-01 -2.26169020e-01
7.44565189e-01 5.78176439e-01 1.15672462e-01 -4.15528774e-01
-9.98468161e-01 -2.56720841e-01 -6.22551978e-01 -3.14918071e-01
1.18202770e+00 -1.17167878e+00 -1.23572218e+00 4.63960141e-01
-1.18115342e+00 -5.24156570e-01 -2.24487782e-01 6.14485681e-01
-4.24485207e-01 2.04497986e-02 -6.62439764e-01 -4.95089352e-01
-3.59302640e-01 -1.33193135e+00 1.09131169e+00 2.89690465e-01
1.48634880e-03 -1.05303013e+00 -2.19993532e-01 -4.02385481e-02
4.53718483e-01 3.34013909e-01 3.84771913e-01 -3.27296436e-01
-1.23963547e+00 -2.99093634e-01 -4.85289484e-01 1.51832357e-01
-2.45429650e-01 1.76795736e-01 -8.30285132e-01 -5.79209268e-01
-3.38741273e-01 -3.39815170e-01 1.30547559e+00 8.38170052e-01
1.38828576e+00 -3.98400575e-01 -3.93735200e-01 1.14378679e+00
1.57666636e+00 1.55863181e-01 3.26369196e-01 5.83561540e-01
8.87198806e-01 8.48990530e-02 4.45382386e-01 2.71046817e-01
9.79671553e-02 7.12581396e-01 9.00168002e-01 -5.79706393e-02
-2.83533543e-01 7.43852258e-02 3.15497369e-01 1.22257300e-01
-6.23812899e-02 -2.30945811e-01 -7.24731028e-01 6.59048319e-01
-1.93716502e+00 -1.10728061e+00 6.44276366e-02 2.11940098e+00
3.99160028e-01 4.94947076e-01 3.83703828e-01 -2.06009433e-01
6.01684213e-01 3.24714839e-01 -7.40132511e-01 3.71453026e-03
-5.49701340e-02 5.63668273e-02 1.00185156e+00 3.06934357e-01
-1.53087342e+00 1.13555992e+00 6.36668682e+00 5.37162185e-01
-1.29053342e+00 2.18547404e-01 3.87729853e-01 -6.96599722e-01
2.28376344e-01 -2.86286950e-01 -1.15662038e+00 4.25010115e-01
7.22405374e-01 1.34919509e-01 2.56054997e-01 1.12640047e+00
-1.19003899e-01 2.63413697e-01 -1.35941660e+00 9.76964533e-01
-3.64121236e-02 -1.78633773e+00 -9.86563712e-02 -8.58639255e-02
8.27009618e-01 4.67961490e-01 2.01428115e-01 1.97054625e-01
2.59128094e-01 -8.65595102e-01 1.18803704e+00 2.08200738e-01
6.06228888e-01 -6.93883240e-01 5.31331956e-01 -8.76561403e-02
-1.65319681e+00 -3.96111757e-01 -3.31193119e-01 1.25767495e-02
3.92712094e-02 2.45775040e-02 -3.02324951e-01 -1.64768994e-01
1.08468246e+00 6.83952928e-01 -5.06245673e-01 1.32505393e+00
1.28191486e-01 2.57989615e-01 -2.56796747e-01 -1.34260476e-01
7.88794756e-01 1.81246534e-01 4.40588981e-01 1.37135923e+00
1.52469292e-01 -1.97424471e-01 4.78496909e-01 7.16945589e-01
-5.03973961e-01 -3.07544678e-01 -3.62457663e-01 -3.36183347e-02
4.67161685e-01 1.19552493e+00 -8.44442368e-01 -5.03198028e-01
-5.80098629e-01 9.06085789e-01 4.54148799e-01 2.82092333e-01
-1.06893432e+00 -1.09071471e-01 1.02747726e+00 1.91278636e-01
9.94419098e-01 -3.46910030e-01 -8.97482485e-02 -1.11594224e+00
1.38251325e-02 -6.36063516e-01 4.58448112e-01 -3.79172772e-01
-9.34195161e-01 7.27462173e-01 -4.15795147e-01 -1.35273826e+00
1.32388715e-02 -9.29804027e-01 -4.00924295e-01 2.95665801e-01
-1.60304141e+00 -1.02460301e+00 -4.19745415e-01 5.94675303e-01
7.72459030e-01 -1.30256429e-01 3.23472142e-01 5.46417773e-01
-4.91435498e-01 7.35036135e-01 -5.61974496e-02 6.40666008e-01
3.08182627e-01 -1.11648631e+00 6.86480582e-01 8.69342804e-01
1.29451081e-01 3.92702132e-01 5.00019968e-01 -4.18207914e-01
-1.60325038e+00 -1.46067321e+00 4.92983341e-01 -6.74512506e-01
8.92273962e-01 -3.98331076e-01 -8.64460349e-01 9.09218788e-01
1.11112364e-01 8.29016924e-01 2.09506467e-01 -8.60939249e-02
-5.89933336e-01 -6.51945844e-02 -8.75823200e-01 5.93763590e-01
1.26543403e+00 -3.80181283e-01 -1.44255891e-01 2.56200403e-01
9.76403177e-01 -8.55474293e-01 -7.93148816e-01 3.15187097e-01
7.88608551e-01 -6.79798365e-01 1.06819403e+00 -9.24492896e-01
2.24023968e-01 -5.49215734e-01 -1.33711591e-01 -5.58869779e-01
-5.57301283e-01 -5.20199180e-01 -6.37603879e-01 6.11264765e-01
2.96769738e-01 -2.41907407e-02 1.27968895e+00 5.05082846e-01
4.17119265e-02 -8.72848153e-01 -8.53438199e-01 -9.88799095e-01
2.76648290e-02 -4.84447658e-01 3.41424912e-01 6.86752260e-01
-7.87866414e-01 1.45222917e-01 -2.40329519e-01 3.95683765e-01
9.13406014e-01 1.07448332e-01 7.16432035e-01 -1.19397151e+00
-3.35989445e-01 -7.51243591e-01 -7.99194694e-01 -1.38368642e+00
1.52895048e-01 -7.00193167e-01 -8.37860107e-02 -1.11384702e+00
1.26592249e-01 -2.57843554e-01 -4.84711856e-01 3.85511905e-01
-2.45437160e-01 4.95269954e-01 5.39903462e-01 6.42788932e-02
-1.07639015e+00 4.44530100e-01 1.26432872e+00 -3.23829353e-01
-1.31332666e-01 -3.64329815e-02 -1.85859591e-01 9.13162827e-01
3.83499831e-01 -6.62428617e-01 -1.54573828e-01 -9.49221194e-01
3.51759116e-03 5.00410050e-02 6.35816693e-01 -1.20385504e+00
5.11371791e-01 -5.08943480e-03 7.37709820e-01 -7.40897238e-01
3.82142931e-01 -9.88873899e-01 -1.24452591e-01 8.84990275e-01
-4.07181740e-01 3.11700016e-01 3.74234378e-01 6.78392947e-01
-1.53823093e-01 8.53539852e-04 1.15725136e+00 -1.49737060e-01
-9.55818534e-01 9.15532351e-01 -2.84707367e-01 3.20506215e-01
1.04758406e+00 -4.72243220e-01 -1.14176966e-01 7.40124285e-02
-7.20465422e-01 2.84020543e-01 4.97130662e-01 6.56676233e-01
5.34633815e-01 -1.42099392e+00 -4.50802982e-01 -2.00734157e-02
-5.00751995e-02 -3.34927738e-02 1.35624751e-01 7.61547685e-01
-7.55040586e-01 3.25290203e-01 -3.18431646e-01 -9.21454966e-01
-1.34393036e+00 7.82698870e-01 6.87547922e-01 -2.52153605e-01
-8.42680514e-01 1.23552716e+00 3.41364652e-01 1.33545414e-01
6.75342739e-01 -5.01952231e-01 3.37812044e-02 -6.58727586e-02
6.94807649e-01 3.03710461e-01 -1.47978827e-01 -4.72531229e-01
-5.34472048e-01 7.70542562e-01 -2.61704445e-01 2.85771728e-01
1.19343197e+00 2.07766756e-01 3.43930304e-01 1.13628149e-01
1.26962042e+00 -4.29448128e-01 -2.08038735e+00 -3.49970311e-01
1.39293700e-01 -6.46960795e-01 3.72667044e-01 -2.40242004e-01
-1.71029460e+00 7.29758322e-01 8.59599650e-01 -1.64388523e-01
9.48875248e-01 -1.43238083e-01 7.36846149e-01 5.41777670e-01
1.96014583e-01 -1.05021143e+00 3.38830441e-01 5.89913547e-01
4.01832134e-01 -1.39531624e+00 -1.07433461e-01 -1.17775030e-01
-4.38158423e-01 1.11411333e+00 9.00952399e-01 -5.31987250e-01
6.17805302e-01 6.43385172e-01 7.67006539e-03 -2.07988366e-01
-8.79886091e-01 -2.10654154e-01 3.82522106e-01 4.23997432e-01
4.32846576e-01 -3.16772580e-01 2.07297578e-01 1.17993094e-01
2.21197277e-01 1.81179985e-01 1.04197070e-01 9.35115218e-01
-5.07442296e-01 -5.73306918e-01 -2.11357139e-02 2.93032974e-01
-8.60125601e-01 -1.06960848e-01 4.08840179e-03 9.55115080e-01
2.41376728e-01 5.09433687e-01 4.68397796e-01 -3.30682024e-02
2.44400159e-01 -3.11917543e-01 7.58686960e-01 -3.72495472e-01
-8.69656980e-01 1.54439822e-01 -1.80504858e-01 -9.53076303e-01
-6.79148257e-01 -6.79846525e-01 -1.34952462e+00 -2.73425698e-01
-3.66878688e-01 -2.44368136e-01 5.13183653e-01 8.48972321e-01
2.13522524e-01 7.76685894e-01 1.46159902e-01 -9.23880517e-01
-3.81322831e-01 -5.95577598e-01 -3.01826984e-01 3.47589284e-01
6.83112085e-01 -7.86476672e-01 4.20916453e-02 1.11256711e-01]
|
[8.954293251037598, -0.19921939074993134]
|
af6b2960-f49a-4c51-9746-da3ec44be886
|
yolo-and-mask-r-cnn-for-vehicle-number-plate
|
2207.13165
| null |
https://arxiv.org/abs/2207.13165v2
|
https://arxiv.org/pdf/2207.13165v2.pdf
|
YOLO and Mask R-CNN for Vehicle Number Plate Identification
|
License plate scanners have grown in popularity in parking lots during the past few years. In order to quickly identify license plates, traditional plate recognition devices used in parking lots employ a fixed source of light and shooting angles. For skewed angles, such as license plate images taken with ultra-wide angle or fisheye lenses, deformation of the license plate recognition plate can also be quite severe, impairing the ability of standard license plate recognition systems to identify the plate. Mask RCNN gadget that may be utilised for oblique pictures and various shooting angles. The results of the experiments show that the suggested design will be capable of classifying license plates with bevel angles larger than 0/60. Character recognition using the suggested Mask R-CNN approach has advanced significantly as well. The proposed Mask R-CNN method has also achieved significant progress in character recognition, which is tilted more than 45 degrees as compared to the strategy of employing the YOLOv2 model. Experiment results also suggest that the methodology presented in the open data plate collecting is better than other techniques (known as the AOLP dataset).
|
['Siddharth Ganjoo']
|
2022-07-26
| null | null | null | null |
['license-plate-recognition']
|
['computer-vision']
|
[-1.21058479e-01 -3.88271034e-01 2.01024771e-01 -2.55337000e-01
-1.66802660e-01 -6.56529248e-01 3.87402564e-01 -8.40622723e-01
-5.41000068e-01 5.04746675e-01 -5.07853627e-01 -4.63842034e-01
3.39824766e-01 -6.31129742e-01 -4.17179316e-01 -6.72854245e-01
4.23709452e-01 5.44956267e-01 5.19510865e-01 -4.50461864e-01
8.98292601e-01 1.03596461e+00 -1.53867531e+00 -9.89725292e-02
2.26630062e-01 7.49589443e-01 2.40266144e-01 5.48599362e-01
-2.13801131e-01 2.43472904e-01 -5.15965939e-01 -4.17320728e-01
7.85027683e-01 2.58332640e-01 -2.02381328e-01 5.24520993e-01
4.11094308e-01 -2.74137735e-01 -3.23945165e-01 9.89113450e-01
2.12128758e-01 6.70181811e-02 7.21681058e-01 -7.50322461e-01
-6.02420390e-01 -6.33751750e-02 -6.72020853e-01 4.70607311e-01
4.23822924e-02 8.70533660e-02 3.30474705e-01 -1.10323775e+00
3.25373977e-01 6.75422132e-01 8.77341807e-01 4.27774578e-01
-6.31877184e-01 -6.44763172e-01 -5.16747475e-01 3.01936388e-01
-1.83339334e+00 -4.18574512e-01 7.81041205e-01 -3.89852494e-01
1.19180357e+00 2.03581110e-01 4.02141601e-01 3.86467695e-01
2.58169562e-01 4.64643449e-01 1.14825451e+00 -7.20889270e-01
-7.45772421e-02 4.77621168e-01 9.70728025e-02 5.77470124e-01
3.82402122e-01 -1.54037371e-01 1.57341808e-01 6.90287292e-01
1.13899183e+00 1.95753545e-01 8.44322443e-02 1.80071950e-01
-8.09095681e-01 7.09264219e-01 9.71949473e-02 4.83283937e-01
6.78237751e-02 -2.16311783e-01 9.66788754e-02 -3.45917307e-02
1.91491663e-01 5.15310168e-01 -1.61723644e-01 -4.30499353e-02
-1.22519755e+00 1.15985490e-01 5.30691385e-01 1.19800210e+00
7.52496362e-01 6.00911617e-01 6.40719473e-01 1.03273773e+00
3.86150062e-01 5.36173105e-01 4.74568129e-01 -3.27553511e-01
7.53332794e-01 6.00548267e-01 1.22282967e-01 -1.13470268e+00
-3.47091526e-01 -4.52362746e-02 -6.33363903e-01 6.37214959e-01
5.21406949e-01 3.73931527e-02 -1.09071827e+00 4.69427347e-01
-3.25428694e-01 -1.40424028e-01 -1.26305923e-01 8.26143861e-01
5.70513904e-01 8.63239646e-01 -4.17521626e-01 2.18650401e-01
1.31295204e+00 -8.85193229e-01 -6.14263773e-01 -3.81256104e-01
4.41355765e-01 -1.13689840e+00 8.15470755e-01 3.69978428e-01
-9.06884968e-01 -6.99365914e-01 -1.29323459e+00 1.79159299e-01
-7.35228121e-01 5.53418219e-01 -5.67342453e-02 1.15566492e+00
-1.03053284e+00 2.21787170e-01 -4.26340133e-01 -3.90599757e-01
2.49277607e-01 7.54988670e-01 -4.93505388e-01 -1.61481962e-01
-7.22353697e-01 1.39240003e+00 1.74594745e-01 5.18726945e-01
4.82876934e-02 -4.15474065e-02 -4.99584258e-01 -4.72940356e-02
5.04176579e-02 3.52222174e-01 7.05933154e-01 -8.19178343e-01
-1.71875644e+00 9.85078037e-01 1.93178102e-01 -1.38060346e-01
5.52823544e-01 -5.16965128e-02 -9.08872545e-01 7.26416558e-02
-1.91986158e-01 5.66258311e-01 8.06081891e-01 -8.56042087e-01
-5.62925935e-01 -5.72277367e-01 -2.56516397e-01 4.24816906e-01
-9.69143305e-03 3.89780283e-01 -7.39270389e-01 -1.51203334e-01
6.78032041e-02 -1.24382663e+00 -1.59767956e-01 -3.00123245e-01
-6.32266223e-01 -1.09967375e-02 1.26123905e+00 -6.43155992e-01
7.32076883e-01 -2.04814267e+00 -9.86349642e-01 5.41996717e-01
-2.38883272e-01 9.17438567e-01 4.44426298e-01 3.46812308e-01
-2.07064718e-01 1.81411952e-01 -2.31349505e-02 -1.53799564e-01
-2.07765058e-01 8.49109665e-02 5.96515685e-02 6.95264161e-01
9.03016850e-02 5.78947544e-01 2.93945611e-01 -3.45046639e-01
5.72568834e-01 3.43324572e-01 2.86276229e-02 -1.86609447e-01
4.84597355e-01 -1.72601461e-01 -1.25331655e-01 8.21639717e-01
1.23309016e+00 1.60516977e-01 -2.28763297e-01 1.37526423e-01
-8.00128758e-01 -3.50469977e-01 -1.16968882e+00 7.37237215e-01
-1.96183637e-01 1.36140275e+00 -1.91738456e-01 -7.61745274e-01
1.55811810e+00 2.58243710e-01 -5.70692942e-02 -8.09246957e-01
1.16838835e-01 2.81740099e-01 -9.12529826e-02 -5.99726319e-01
8.85050535e-01 -1.24886885e-01 2.42839426e-01 -1.23209283e-01
-4.95286524e-01 -2.60545444e-02 9.09406170e-02 -5.39270461e-01
3.38972539e-01 -4.86565419e-02 -1.20153800e-02 -2.82188058e-01
8.63553405e-01 2.10060582e-01 1.43119290e-01 4.96253371e-01
-1.95508301e-01 8.13503861e-01 9.91722867e-02 -7.68405139e-01
-1.64164722e+00 -5.52370727e-01 -2.42158368e-01 4.43679392e-01
2.70852089e-01 5.44112861e-01 -7.50151753e-01 -4.09147777e-02
-2.65787154e-01 3.36459816e-01 -2.76507944e-01 5.20970523e-01
-9.70863044e-01 -6.66456759e-01 1.01970303e+00 6.78445697e-01
1.08130276e+00 -8.91292036e-01 -4.56790149e-01 2.83375401e-02
3.11741471e-01 -1.32894158e+00 -3.23883623e-01 4.57161553e-02
-1.00817978e+00 -8.82966220e-01 -9.10731912e-01 -1.13946760e+00
8.23438168e-01 5.98449647e-01 6.76986992e-01 -5.14391325e-02
-1.03069521e-01 -8.18990692e-02 -2.06760719e-01 -3.09184670e-01
-2.37823665e-01 -8.94685015e-02 -4.50815521e-02 6.69127563e-03
8.49270284e-01 -3.58179063e-02 -5.58827043e-01 8.53095710e-01
-8.15071106e-01 -2.29670167e-01 7.28350282e-01 2.62901723e-01
5.29591218e-02 1.14762895e-01 2.63399482e-01 -6.01509392e-01
5.49721062e-01 -1.82168290e-01 -9.24944937e-01 -4.66106534e-02
-8.63126040e-01 -4.49645162e-01 7.84528971e-01 -2.12375317e-02
-1.00558794e+00 2.21355520e-02 -4.11910325e-01 -4.11339164e-01
-6.53883934e-01 1.89359998e-03 -1.34220764e-01 -5.39126337e-01
3.68574530e-01 4.29863811e-01 1.85705841e-01 -3.10317814e-01
-3.46582115e-01 9.45010006e-01 5.71438372e-01 2.58277923e-01
1.05621171e+00 4.42652881e-01 8.45249519e-02 -1.46317267e+00
3.42770875e-01 -8.99569273e-01 -8.03632557e-01 -6.85298026e-01
9.97864187e-01 -6.38782859e-01 -7.70670950e-01 1.00649559e+00
-8.19483817e-01 2.43926361e-01 3.09646368e-01 4.57744807e-01
-8.84848535e-02 5.15661776e-01 -5.02630413e-01 -8.39630842e-01
1.15513481e-01 -1.27510214e+00 6.78659081e-01 6.36597693e-01
2.08936125e-01 -1.11785626e+00 -4.40253094e-02 6.56802177e-01
4.58105803e-01 -2.14293361e-01 5.68972945e-01 -9.45691049e-01
-6.12513900e-01 -8.11782122e-01 -3.10733736e-01 5.13416708e-01
-1.55501410e-01 4.17770356e-01 -1.08337760e+00 2.92532235e-01
-7.67992288e-02 1.90816708e-02 5.61799705e-01 4.00583833e-01
6.25935435e-01 3.31628397e-02 -6.40341491e-02 5.36657035e-01
1.83524990e+00 7.59338260e-01 1.42594135e+00 9.57148790e-01
6.90818906e-01 3.64629388e-01 4.07674372e-01 1.26924500e-01
-2.67286040e-02 7.19008684e-01 3.25515926e-01 -2.17509121e-01
3.12356710e-01 4.50080186e-02 2.50782639e-01 7.35365033e-01
-5.96343219e-01 -2.81405717e-01 -1.08804452e+00 2.15919852e-01
-9.65776622e-01 -1.08422160e+00 -6.22133315e-01 1.97657669e+00
9.97124687e-02 2.24099651e-01 1.36939377e-01 4.33212429e-01
1.22947848e+00 4.32357453e-02 -1.47743151e-01 -1.04292822e+00
-2.25148007e-01 -5.31428456e-02 1.22135735e+00 3.92361462e-01
-1.01782072e+00 9.42451954e-01 6.58528185e+00 6.42903566e-01
-1.59618866e+00 -4.76358980e-01 4.68367964e-01 4.72411454e-01
1.42032087e-01 -2.63907313e-01 -1.11348319e+00 6.64580941e-01
8.77611756e-01 5.09254813e-01 1.58992246e-01 8.34640801e-01
2.33652875e-01 -2.91145593e-01 -5.63881159e-01 1.18652081e+00
3.66113245e-01 -1.40371394e+00 -2.99831539e-01 5.34726083e-01
4.97772008e-01 1.47968262e-01 3.84487063e-01 1.36098847e-01
-3.13175201e-01 -1.06677639e+00 4.06820267e-01 3.43433619e-01
9.73472297e-01 -8.81010711e-01 1.11474049e+00 4.28570360e-01
-8.66491675e-01 7.69694597e-02 -7.53907204e-01 -7.11918697e-02
-1.99844807e-01 -9.68585983e-02 -1.69607651e+00 1.15623996e-01
3.33582103e-01 3.66168618e-01 -5.97795844e-01 1.18610728e+00
3.18276972e-01 6.13784075e-01 -6.07780159e-01 -3.89218301e-01
7.11170673e-01 -6.40689909e-01 3.82743061e-01 1.47984624e+00
6.37980402e-01 -1.40189394e-01 -5.56368351e-01 6.46741688e-01
2.01784149e-01 -3.54527985e-03 -8.98027599e-01 2.28044391e-01
2.88299769e-01 1.03532422e+00 -1.06455815e+00 5.65640293e-02
-7.52635121e-01 7.02834785e-01 -5.51793039e-01 1.92028120e-01
-6.47012472e-01 -3.59488457e-01 2.12570325e-01 3.77848119e-01
7.76812136e-01 -4.19465095e-01 -3.91782850e-01 -8.76339972e-01
-4.82796095e-02 -3.51093680e-01 -2.10696444e-01 -9.69079494e-01
-6.77443027e-01 5.22055566e-01 -1.46059141e-01 -1.75431228e+00
-1.74994424e-01 -1.50993276e+00 -8.15440357e-01 9.72859740e-01
-1.31722295e+00 -1.03446078e+00 -8.12910870e-03 7.15007424e-01
9.77617919e-01 -7.18397200e-01 5.96052349e-01 2.79033005e-01
-7.29080498e-01 3.95046502e-01 5.69101036e-01 5.74051678e-01
4.12101895e-01 -1.03243113e+00 3.13554823e-01 9.20814216e-01
-1.65006533e-01 3.78894627e-01 7.06181049e-01 -4.71360028e-01
-1.22718561e+00 -9.17675912e-01 7.67773092e-01 -1.65106267e-01
2.59127229e-01 -2.60073125e-01 -6.82866812e-01 6.31105959e-01
2.61841416e-01 -4.64572087e-02 7.73285270e-01 -4.51070845e-01
-1.65234968e-01 -2.74385154e-01 -1.20455992e+00 4.14845228e-01
7.34074861e-02 -3.18958282e-01 -5.48298120e-01 -6.76647108e-03
-4.19948012e-01 -1.48334712e-01 -3.20637643e-01 -1.35366559e-01
8.39932084e-01 -1.12602556e+00 8.16992819e-01 -1.57297868e-02
3.08404207e-01 -3.71684521e-01 -1.08100340e-01 -9.05457735e-01
-2.27081820e-01 -1.26490787e-01 8.87274981e-01 9.14418280e-01
6.27032042e-01 -7.27427125e-01 1.18622160e+00 7.60509312e-01
-3.16272289e-01 -3.23550195e-01 -1.04987216e+00 -8.48152459e-01
2.57052463e-02 -3.03149521e-01 1.50740057e-01 7.20493138e-01
-1.96394101e-01 -1.05118155e-01 -4.71530318e-01 2.91342229e-01
4.13884819e-01 -4.28564936e-01 6.64007485e-01 -1.13597381e+00
2.06266761e-01 -5.21388292e-01 -1.04837537e+00 -9.23287392e-01
-1.12261914e-01 -4.59366202e-01 1.28382549e-01 -1.18037868e+00
-3.52434874e-01 -3.52981508e-01 4.14868481e-02 -5.65572083e-02
3.52225214e-01 8.43951404e-01 4.97173399e-01 7.13292956e-01
-1.08363301e-01 -2.19355837e-01 1.13825476e+00 -1.56406268e-01
3.23717482e-02 4.03127283e-01 -2.36889601e-01 9.29310739e-01
9.11455452e-01 -1.78814560e-01 -1.52625158e-01 -2.49409050e-01
3.34338456e-01 2.08190978e-02 2.28653271e-02 -1.16121328e+00
5.30129433e-01 1.92342505e-01 8.31120968e-01 -9.73367751e-01
5.23491561e-01 -1.02895772e+00 -4.81015351e-03 2.06667811e-01
1.07355706e-01 4.80066746e-01 1.96148008e-01 2.26093218e-01
-3.94035906e-01 -6.84882224e-01 8.95384789e-01 -4.34105605e-01
-1.07084751e+00 -5.94974346e-02 -9.92032409e-01 -4.41976905e-01
1.11476755e+00 -1.54063773e+00 -2.09569529e-01 -2.56504238e-01
-5.00171721e-01 -3.46662611e-01 5.75058579e-01 2.03846857e-01
8.78686726e-01 -1.02471852e+00 -4.88659322e-01 4.86116379e-01
-7.64540732e-02 -2.09317610e-01 2.62496591e-01 6.88423336e-01
-1.55137789e+00 9.19197619e-01 -7.89649665e-01 -5.74962139e-01
-1.32557440e+00 1.16881892e-01 3.27764869e-01 -2.60149650e-02
-6.08073652e-01 7.13699877e-01 -2.39493698e-01 -2.38685742e-01
-2.28455499e-01 -8.61589760e-02 -7.67365754e-01 -1.73240334e-01
3.64646733e-01 6.85637474e-01 5.29000342e-01 -8.72988760e-01
-4.32546854e-01 9.71697390e-01 -2.77468950e-01 -2.25298658e-01
1.46043313e+00 -4.40856665e-02 2.35049397e-01 2.86368698e-01
9.56199110e-01 3.94080132e-01 -1.21402085e+00 2.71936715e-01
-1.20373353e-01 -6.99381590e-01 -2.29989871e-01 -5.28504968e-01
-9.33850586e-01 9.00991976e-01 7.13003516e-01 3.69804591e-01
8.09991658e-01 -6.54683709e-01 4.85463470e-01 6.99238658e-01
2.49115705e-01 -1.56758702e+00 -4.75413650e-01 6.18345201e-01
6.19869351e-01 -1.08113444e+00 -1.66376978e-01 -2.25785151e-01
-8.54790032e-01 1.92060423e+00 6.61397874e-01 -4.99351859e-01
5.07339001e-01 5.82170665e-01 2.36274406e-01 -4.45240103e-02
-1.24297798e-01 9.34942514e-02 -2.17816811e-02 7.46344388e-01
2.76396245e-01 1.68140635e-01 4.13837470e-02 -1.16678812e-01
-2.85557695e-02 -1.93307355e-01 1.13414180e+00 8.70491207e-01
-7.67218173e-01 -7.61543691e-01 -1.06277192e+00 3.34955633e-01
-5.60896873e-01 -6.66426867e-02 -2.80942589e-01 1.50645280e+00
2.07274765e-01 7.37209678e-01 6.53946638e-01 -3.74218196e-01
3.13506931e-01 2.24831253e-01 1.27227843e-01 -2.50305295e-01
-3.26199174e-01 2.27679715e-01 6.58346564e-02 -3.23581249e-02
-4.48986143e-01 -6.54471219e-01 -9.23367441e-01 -5.07528365e-01
-5.60886085e-01 -5.96200638e-02 1.17783475e+00 1.11310470e+00
-2.40475640e-01 -1.02108538e-01 7.04158604e-01 -9.68639970e-01
-3.53760362e-01 -1.02142918e+00 -9.64966238e-01 -1.25273630e-01
6.13286607e-02 -3.50279689e-01 -2.99341291e-01 2.32079998e-01]
|
[9.819615364074707, -4.968945503234863]
|
e78a0002-bb72-4cb5-a633-2d7729511b52
|
evaluating-counterfactual-explanations-using
|
2301.02499
| null |
https://arxiv.org/abs/2301.02499v1
|
https://arxiv.org/pdf/2301.02499v1.pdf
|
Evaluating counterfactual explanations using Pearl's counterfactual method
|
Counterfactual explanations (CEs) are methods for generating an alternative scenario that produces a different desirable outcome. For example, if a student is predicted to fail a course, then counterfactual explanations can provide the student with alternate ways so that they would be predicted to pass. The applications are many. However, CEs are currently generated from machine learning models that do not necessarily take into account the true causal structure in the data. By doing this, bias can be introduced into the CE quantities. I propose in this study to test the CEs using Judea Pearl's method of computing counterfactuals which has thus far, surprisingly, not been seen in the counterfactual explanation (CE) literature. I furthermore evaluate these CEs on three different causal structures to show how the true underlying causal structure affects the CEs that are generated. This study presented a method of evaluating CEs using Pearl's method and it showed, (although using a limited sample size), that thirty percent of the CEs conflicted with those computed by Pearl's method. This shows that we cannot simply trust CEs and it is vital for us to know the true causal structure before we blindly compute counterfactuals using the original machine learning model.
|
['Bevan I. Smith']
|
2023-01-06
| null | null | null | null |
['counterfactual-explanation']
|
['miscellaneous']
|
[ 1.62570924e-01 6.54399216e-01 -6.06645286e-01 -2.98243254e-01
-3.40147883e-01 -5.99505067e-01 1.02837110e+00 1.79561540e-01
-4.58535463e-01 1.44514489e+00 5.39653420e-01 -1.19873071e+00
-4.56149697e-01 -8.85164082e-01 -8.75922978e-01 -4.44192529e-01
4.12910990e-02 3.36130023e-01 -1.50802091e-01 2.97820661e-02
8.08181047e-01 3.68473470e-01 -1.68784189e+00 2.92517513e-01
1.07944858e+00 6.22484460e-02 -2.62801617e-01 7.35100150e-01
2.44698208e-02 8.93017471e-01 -8.09843957e-01 -6.60311222e-01
1.39028028e-01 -9.47915912e-01 -9.59243894e-01 -3.10644448e-01
4.34030533e-01 -3.69111329e-01 6.24606982e-02 9.18506145e-01
4.97686416e-02 6.09612390e-02 1.06251514e+00 -1.64954674e+00
-4.95135754e-01 9.46945965e-01 -4.82074648e-01 9.97437388e-02
5.30733705e-01 1.33708835e-01 8.75185788e-01 -9.25150514e-02
4.68334466e-01 1.56140065e+00 3.37488979e-01 6.30038500e-01
-1.56350625e+00 -1.16968489e+00 2.19890907e-01 9.69560072e-02
-6.07375979e-01 -5.10408320e-02 6.82161570e-01 -3.96676838e-01
2.89164335e-01 5.96099913e-01 8.98297012e-01 1.10619318e+00
4.74890381e-01 3.44725102e-01 1.80300510e+00 -7.28061914e-01
4.60429162e-01 4.00643647e-01 -1.08903237e-01 3.64813775e-01
1.01297319e+00 9.66368616e-01 -4.30471569e-01 -3.60273033e-01
7.80636609e-01 -2.65117466e-01 -4.47422683e-01 -5.91168344e-01
-1.28976142e+00 1.25296295e+00 3.93952489e-01 2.35323071e-01
-3.76341373e-01 3.31094295e-01 8.96538422e-02 3.57529312e-01
1.75735757e-01 9.06305134e-01 -5.78428328e-01 -3.57957594e-02
-8.93320620e-01 7.45074511e-01 8.05444896e-01 2.98557639e-01
1.71854421e-01 4.68795858e-02 1.22762308e-03 -1.58456843e-02
4.45385903e-01 2.41287142e-01 7.46631384e-01 -1.23474121e+00
3.59341532e-01 4.83156145e-01 5.93290031e-01 -7.87872493e-01
-9.54313204e-02 -1.89129144e-01 -2.94962347e-01 7.74193168e-01
9.63028669e-01 -4.10403639e-01 -5.54262102e-01 1.90964782e+00
1.87882885e-01 3.00796866e-01 3.16974223e-01 8.04024875e-01
3.00145417e-01 3.38461757e-01 5.22584617e-01 -4.77578104e-01
1.00948846e+00 -1.33291930e-01 -6.91328943e-01 -2.21395288e-02
1.06096518e+00 -7.64482260e-01 1.14152431e+00 4.98128265e-01
-9.20638800e-01 -2.05276638e-01 -1.13968956e+00 5.65318286e-01
-1.56280577e-01 -5.16094863e-01 1.26825380e+00 1.16926098e+00
-5.37688613e-01 9.77236569e-01 -4.37745631e-01 -5.43239191e-02
1.84910670e-01 1.03575602e-01 -1.91223800e-01 1.95696279e-01
-1.36290812e+00 9.74807620e-01 5.35210788e-01 -3.79669756e-01
-7.71251678e-01 -9.64791715e-01 -6.39905870e-01 3.70496631e-01
3.58679324e-01 -7.80479550e-01 1.39138579e+00 -1.23572862e+00
-1.06677175e+00 5.15065908e-01 5.47715370e-03 -5.11456549e-01
8.61720979e-01 8.35166201e-02 -2.86782086e-01 -2.00926006e-01
3.56780827e-01 5.59916973e-01 3.06064367e-01 -1.51382613e+00
-7.19735444e-01 -4.14080650e-01 2.17093140e-01 1.73367664e-01
3.32486808e-01 -1.26166731e-01 9.58225250e-01 -4.54875231e-01
9.32280421e-02 -9.29222703e-01 -2.61197686e-01 -4.14884448e-01
-3.97218466e-01 -2.74653852e-01 3.55212063e-01 -3.86954360e-02
9.11922395e-01 -1.64225388e+00 -6.88124001e-01 2.55457848e-01
2.22754129e-03 -1.50040880e-01 2.52471685e-01 1.66786209e-01
-9.88519609e-01 6.01320088e-01 -1.42605931e-01 4.39737290e-01
1.94058448e-01 1.01293093e-02 -7.10366607e-01 5.44340551e-01
-1.48976207e-01 4.76567864e-01 -9.46348011e-01 -4.32551354e-01
3.31258893e-01 -8.94581303e-02 -6.69183373e-01 1.30272537e-01
2.11662903e-01 4.59251925e-02 -9.38603580e-02 -1.15081050e-01
4.52822983e-01 2.08963677e-01 4.52433825e-01 4.24967915e-01
-3.04093421e-01 7.48016596e-01 -1.28645837e+00 8.52478147e-01
-2.93714702e-01 6.97515905e-01 -6.03880584e-01 -1.14362288e+00
5.61324596e-01 7.42354989e-01 -3.21247540e-02 -3.46750766e-01
-1.41807701e-02 4.44162071e-01 5.72416306e-01 -2.28199854e-01
3.24161559e-01 -1.10877919e+00 -6.78157732e-02 7.89337158e-01
-3.86530459e-01 -5.50219238e-01 3.50289010e-02 2.09051341e-01
5.81461728e-01 6.92515597e-02 6.19865656e-01 -4.12304550e-01
1.10614426e-01 3.76716346e-01 5.77505410e-01 9.95268345e-01
-1.06873676e-01 3.31473321e-01 9.49680328e-01 -4.28300172e-01
-8.60169590e-01 -1.28272867e+00 -2.25904837e-01 3.46109331e-01
-3.34479868e-01 2.72957534e-01 -6.50796533e-01 -1.06329370e+00
1.81342036e-01 1.91724873e+00 -7.66844511e-01 -3.43956321e-01
-8.07691291e-02 -8.67842317e-01 2.96402872e-01 1.96154281e-01
1.96792617e-01 -8.06623876e-01 -9.63510990e-01 -1.06163330e-01
-2.07562864e-01 -9.05580223e-02 -4.59225811e-02 -1.26447886e-01
-1.23230278e+00 -1.49326885e+00 -4.13583994e-01 6.57577291e-02
6.05275095e-01 3.06669950e-01 1.11392558e+00 1.34230554e-01
2.46486455e-01 2.31943607e-01 -1.58293489e-02 -1.10357273e+00
-7.40009189e-01 -6.92051768e-01 1.40385509e-01 -6.71843767e-01
5.38654804e-01 -6.04844868e-01 -4.73974913e-01 -1.11946516e-01
-7.81322598e-01 5.23404311e-03 5.31838715e-01 9.46819305e-01
-1.05074003e-01 1.79451376e-01 8.56861949e-01 -1.33170462e+00
7.43238986e-01 -4.73820508e-01 -7.11059928e-01 1.29677996e-01
-1.25048506e+00 2.64511555e-01 4.23919648e-01 -5.49152017e-01
-1.58899474e+00 -4.94051486e-01 3.37196440e-01 1.48959279e-01
-6.67723119e-01 3.03544372e-01 -4.26447131e-02 4.34083700e-01
8.46681893e-01 -3.18258643e-01 -1.87717035e-01 -1.38492391e-01
3.03323925e-01 3.85693997e-01 2.14112133e-01 -6.57771468e-01
7.20617712e-01 1.84703454e-01 2.28776857e-01 -1.32393807e-01
-9.20695841e-01 2.44236469e-01 -2.34731302e-01 -1.88824445e-01
6.14321113e-01 -7.29486525e-01 -1.07254636e+00 -3.82096738e-01
-9.72675443e-01 -2.71959215e-01 -5.33381879e-01 1.29906225e+00
-8.68062556e-01 -7.31314346e-02 -1.50091901e-01 -1.08272290e+00
4.20190871e-01 -8.60130906e-01 2.23502338e-01 2.50739336e-01
-8.25183570e-01 -1.11038625e+00 1.61286443e-01 3.72281343e-01
-1.21469880e-02 3.13405722e-01 1.20169044e+00 -8.02822113e-01
-3.51089925e-01 -1.24992944e-01 -2.16722097e-02 -3.57709110e-01
1.45729616e-01 1.02594480e-01 -9.55392182e-01 -7.04989731e-02
2.52809286e-01 -5.18192202e-02 5.02705693e-01 6.50408387e-01
9.77142513e-01 -8.68675351e-01 -3.69849056e-01 -7.52954781e-02
1.49690270e+00 5.03974795e-01 6.02171957e-01 3.96665275e-01
2.29306087e-01 1.11272407e+00 6.96342826e-01 6.67607263e-02
2.65884161e-01 2.97849834e-01 3.87753099e-01 -3.31191272e-02
2.57651389e-01 -7.45319128e-01 2.91648507e-01 -1.18002512e-01
-1.43970504e-01 7.31111616e-02 -6.41946375e-01 7.10862458e-01
-1.47043610e+00 -1.57522023e+00 -6.99094534e-01 2.55803943e+00
9.14912879e-01 4.46804315e-01 1.56804603e-02 3.86217892e-01
4.77518380e-01 -1.49715379e-01 -1.43550277e-01 -1.02825689e+00
3.94418478e-01 4.38701138e-02 7.46942043e-01 6.73909962e-01
-3.79628450e-01 2.62758583e-01 7.04175615e+00 3.04701477e-01
-7.43496537e-01 -1.61024377e-01 9.30271029e-01 -1.18138686e-01
-1.05844760e+00 6.94764674e-01 -2.96845645e-01 5.17791033e-01
1.27351284e+00 -8.05886686e-01 -1.67105734e-01 6.91045702e-01
6.84455514e-01 -6.27381086e-01 -1.30106556e+00 1.48151770e-01
-4.52624977e-01 -1.32892096e+00 1.97583973e-01 2.81100571e-01
1.03927636e+00 -9.37282681e-01 -1.32276788e-01 3.44379932e-01
1.20713854e+00 -1.26354909e+00 8.13940644e-01 3.74853551e-01
4.21311915e-01 -1.04963696e+00 9.14257050e-01 5.69536209e-01
-2.46484160e-01 -1.58212334e-01 -1.99545398e-01 -8.25038433e-01
-2.83534318e-01 6.89505994e-01 -1.19880927e+00 3.62440705e-01
2.07384750e-01 -4.69247662e-02 -2.48839483e-01 8.52331221e-01
-7.44947076e-01 1.11535323e+00 2.02117532e-01 -2.67524719e-02
1.79530323e-01 -1.53358668e-01 4.21107471e-01 6.96774006e-01
4.49021220e-01 3.24606806e-01 -5.31078637e-01 1.18495142e+00
1.96974665e-01 -1.93460301e-01 -1.09284925e+00 6.61554784e-02
5.85290611e-01 5.80726564e-01 -6.89799547e-01 -4.23742175e-01
-3.14269692e-01 4.16630805e-01 -2.22367540e-01 3.43703479e-01
-6.73718333e-01 -1.27333134e-01 7.46900022e-01 2.52754956e-01
-3.87427509e-01 3.67959678e-01 -7.18620062e-01 -9.28498328e-01
-5.42986810e-01 -8.06767285e-01 5.22858441e-01 -9.54993606e-01
-1.09835649e+00 -4.18188632e-01 5.39414048e-01 -8.80747676e-01
-5.45312881e-01 -3.74766797e-01 -1.06046391e+00 1.20534694e+00
-1.18254626e+00 -3.86407942e-01 4.65910941e-01 1.01275817e-01
2.61138856e-01 2.85031229e-01 6.96058452e-01 -4.23058629e-01
-1.90295249e-01 2.83073336e-01 -1.15779877e-01 -1.32231593e-01
9.03779566e-01 -1.76346433e+00 -4.04373296e-02 8.41722786e-01
8.30742344e-02 8.59018147e-01 1.33610809e+00 -8.23125780e-01
-4.72512007e-01 -4.73264635e-01 1.33978593e+00 -3.99462551e-01
4.56357956e-01 4.07559276e-01 -7.31605828e-01 9.68157113e-01
3.04992318e-01 -5.88594973e-01 6.69427156e-01 4.20203716e-01
-1.64324358e-01 1.99440479e-01 -1.33204162e+00 1.03660357e+00
6.57468021e-01 -9.94696021e-02 -1.35221517e+00 4.58869934e-02
6.72328711e-01 -3.52667384e-02 -4.51770484e-01 4.09464799e-02
6.17360294e-01 -1.43601084e+00 8.46397400e-01 -1.16155136e+00
8.79494488e-01 -1.42100826e-01 2.10047826e-01 -1.80747426e+00
-7.94488378e-03 -1.80367246e-01 6.12566531e-01 1.22083068e+00
5.57195663e-01 -8.78976047e-01 9.39848602e-01 1.27591980e+00
1.81802958e-01 -3.53735089e-01 -8.90555620e-01 -6.74185097e-01
5.44344783e-01 -5.24048448e-01 9.98737872e-01 1.56624603e+00
4.03922260e-01 1.11686982e-01 -3.71572264e-02 3.26414272e-04
9.83842373e-01 3.87007117e-01 7.30218232e-01 -1.41415942e+00
-5.62785789e-02 -4.95465040e-01 -2.58055106e-02 -2.14889705e-01
1.87783495e-01 -5.97323000e-01 -2.54108399e-01 -1.31487322e+00
4.65926528e-01 -4.32976693e-01 -3.16493982e-03 3.32695186e-01
-6.14458978e-01 -4.80739892e-01 3.62715185e-01 -9.72379744e-02
3.89785081e-01 3.96567315e-01 1.25514841e+00 3.51204485e-01
-1.71771184e-01 1.97636425e-01 -1.23146379e+00 1.12580335e+00
9.41340864e-01 -8.56429338e-01 -5.61299205e-01 3.03341776e-01
2.15519562e-01 5.04926264e-01 8.63572657e-01 -5.94812393e-01
-1.58257470e-01 -9.15949702e-01 6.99373126e-01 -1.55626118e-01
-3.08647722e-01 -8.67222190e-01 4.75345105e-01 1.02032709e+00
-8.61269772e-01 1.41689181e-01 2.29165748e-01 5.05292714e-01
2.62584947e-02 -6.77733243e-01 4.87161905e-01 -4.64873642e-01
-2.51671463e-01 -5.47320724e-01 -7.37524569e-01 -1.48015916e-01
1.04553390e+00 -9.45682526e-02 -3.30190957e-01 -6.76415086e-01
-3.57882768e-01 -2.52897162e-02 5.05523026e-01 8.59891251e-02
3.86660367e-01 -1.47124767e+00 -6.74990475e-01 -3.26347053e-01
-2.01170295e-01 -5.70037901e-01 1.24236263e-01 5.74159324e-01
-1.61389843e-01 6.82238817e-01 -2.26706445e-01 1.34116203e-01
-1.11338687e+00 6.88913047e-01 3.32573235e-01 -3.29920232e-01
-2.00597942e-01 4.27642643e-01 4.31058735e-01 -4.00883079e-01
-3.61143231e-01 -1.20855443e-01 -2.68247187e-01 6.52425811e-02
5.16432285e-01 4.90769476e-01 -4.77682859e-01 -9.67457816e-02
2.32272074e-02 -1.89020187e-01 2.04062417e-01 -5.43939531e-01
1.21740508e+00 -7.82985985e-02 1.42725885e-01 6.94973290e-01
5.32739222e-01 3.03447843e-01 -1.05997884e+00 4.27148074e-01
1.13939866e-01 -1.12223494e+00 -8.83729756e-02 -9.74673331e-01
-4.56836432e-01 7.38295496e-01 5.01734138e-01 4.35456306e-01
8.12658608e-01 -3.99123311e-01 -1.28961399e-01 -6.78885207e-02
3.57930094e-01 -8.22128773e-01 -3.51632684e-01 -3.22785974e-01
7.92533994e-01 -1.36400843e+00 3.59826267e-01 -2.30968475e-01
-6.00151777e-01 8.68505716e-01 5.19222915e-01 -4.46491456e-03
2.68337756e-01 -1.79332376e-01 -6.72858506e-02 -1.50342479e-01
-1.08723891e+00 1.37731835e-01 6.33420423e-02 5.24360955e-01
9.63554204e-01 6.78202748e-01 -1.11839390e+00 4.38384831e-01
-7.72369683e-01 1.93077087e-01 1.40726542e+00 5.58789730e-01
-2.54804522e-01 -1.08468997e+00 -1.00178039e+00 6.98976934e-01
-6.24246418e-01 8.41168091e-02 -4.48058516e-01 1.35590434e+00
5.84409535e-02 1.20393836e+00 1.32191166e-01 9.21226889e-02
3.00107956e-01 3.27525675e-01 3.59069407e-01 -5.96242011e-01
-2.94663519e-01 -3.48469645e-01 3.09038907e-01 -5.77719808e-01
-6.57157063e-01 -8.57694805e-01 -1.09194160e+00 -8.04398477e-01
-2.62869507e-01 6.95212126e-01 4.77837533e-01 1.14733493e+00
-2.43351564e-01 4.32107180e-01 5.00571609e-01 -3.77941400e-01
-8.50947976e-01 -8.14887583e-01 -6.09151483e-01 3.98283452e-01
2.54565656e-01 -8.51289451e-01 -7.92237937e-01 -2.16052577e-01]
|
[8.563305854797363, 5.629896640777588]
|
52440856-4cc7-4ced-a167-6178b9705601
|
interventional-and-counterfactual-inference
|
2302.00860
| null |
https://arxiv.org/abs/2302.00860v2
|
https://arxiv.org/pdf/2302.00860v2.pdf
|
Interventional and Counterfactual Inference with Diffusion Models
|
We consider the problem of answering observational, interventional, and counterfactual queries in a causally sufficient setting where only observational data and the causal graph are available. Utilizing the recent developments in diffusion models, we introduce diffusion-based causal models (DCM) to learn causal mechanisms, that generate unique latent encodings. These encodings enable us to directly sample under interventions and perform abduction for counterfactuals. Diffusion models are a natural fit here, since they can encode each node to a latent representation that acts as a proxy for exogenous noise. Our empirical evaluations demonstrate significant improvements over existing state-of-the-art methods for answering causal queries. Furthermore, we provide theoretical results that offer a methodology for analyzing counterfactual estimation in general encoder-decoder models, which could be useful in settings beyond our proposed approach.
|
['Shiva Prasad Kasiviswanathan', 'Patrick Blöbaum', 'Patrick Chao']
|
2023-02-02
| null | null | null | null |
['counterfactual-inference']
|
['miscellaneous']
|
[ 6.10231400e-01 6.24248922e-01 -1.20084918e+00 -2.74099320e-01
-8.71277153e-01 -4.87407953e-01 1.03068912e+00 2.33841240e-01
2.69869845e-02 1.17978656e+00 1.30837691e+00 -7.95579135e-01
-4.29388672e-01 -1.08118927e+00 -1.14323902e+00 -3.59036356e-01
-5.65280199e-01 3.57469976e-01 -4.69033957e-01 1.71762988e-01
3.76455747e-02 -3.77475061e-02 -1.03553975e+00 3.08006644e-01
9.23261344e-01 2.61796087e-01 1.52939111e-02 7.04121768e-01
2.76359290e-01 1.32000077e+00 -3.04258108e-01 -7.82852054e-01
-1.54309034e-01 -5.19615471e-01 -6.90587580e-01 -4.75149781e-01
2.44307548e-01 -7.06077754e-01 -9.13046241e-01 9.45837021e-01
5.43581843e-01 -2.73129195e-01 1.01051164e+00 -1.16209376e+00
-1.18860364e+00 1.40313184e+00 -5.18537641e-01 3.69218946e-01
6.90791011e-01 4.04166013e-01 1.25246227e+00 -2.89891958e-01
8.18890333e-01 1.84248114e+00 3.65857363e-01 6.15020633e-01
-1.55773318e+00 -7.24394441e-01 2.70594299e-01 7.10644126e-02
-6.73438549e-01 -4.48369861e-01 6.49211884e-01 -5.63059926e-01
7.21178174e-01 2.16477826e-01 5.13575375e-01 1.92146099e+00
3.95638704e-01 9.89216626e-01 1.10910475e+00 -1.72838360e-01
3.21763605e-01 -5.66644192e-01 -1.69237301e-01 6.55706644e-01
5.98328650e-01 8.72379065e-01 -7.71823585e-01 -6.03129447e-01
1.01812649e+00 3.10935508e-02 -4.46474105e-01 -2.42656454e-01
-1.31572700e+00 1.32516313e+00 4.82966512e-01 -4.10045207e-01
-8.34841609e-01 8.70205045e-01 2.97033161e-01 2.83364177e-01
5.82644343e-01 2.87374079e-01 -3.14456493e-01 -1.61437064e-01
-5.74185371e-01 5.27818620e-01 8.50559711e-01 1.01000237e+00
9.86635387e-02 -1.68526739e-01 -7.75482297e-01 3.68056208e-01
4.60951477e-01 7.17271805e-01 -3.49358772e-03 -1.06937468e+00
6.74871266e-01 1.85168996e-01 3.17432851e-01 -6.69146895e-01
-1.44753486e-01 -1.98387012e-01 -9.50950921e-01 -5.73552251e-01
1.20876759e-01 -4.96049225e-01 -7.14717567e-01 2.18720102e+00
3.43455940e-01 6.17758036e-01 -8.60527810e-03 6.99252129e-01
5.01643896e-01 4.63568628e-01 2.91846037e-01 -6.66136801e-01
1.24254906e+00 -4.11314070e-01 -1.07438791e+00 -1.58260107e-01
6.35395050e-01 -4.83310550e-01 1.05641842e+00 9.87216309e-02
-1.12438059e+00 6.52367249e-02 -7.37856090e-01 7.18307197e-02
6.10808283e-02 -2.76048183e-01 1.31000400e+00 7.70938396e-01
-8.56482267e-01 4.63199884e-01 -7.65041828e-01 5.25378529e-03
7.23280013e-01 6.93017691e-02 1.71242401e-01 -3.00337374e-01
-1.69276047e+00 6.08121991e-01 1.42543793e-01 -1.69350892e-01
-1.84544837e+00 -1.18647289e+00 -9.29442048e-01 1.62281215e-01
8.59944880e-01 -1.40312338e+00 1.55772674e+00 -2.70948172e-01
-1.19233561e+00 3.94678473e-01 -2.32433915e-01 -8.97916317e-01
6.32719755e-01 -2.88625564e-02 -3.75548899e-01 -1.86868869e-02
4.96197015e-01 4.47181106e-01 5.93113780e-01 -9.41904366e-01
-4.76042330e-01 -3.34757894e-01 4.58027840e-01 -6.80199564e-02
-7.04674562e-03 -1.18424430e-01 -1.51358113e-01 -9.57154453e-01
-5.79150438e-01 -5.55669606e-01 -4.91899967e-01 -3.40216249e-01
-9.42515552e-01 -2.27286965e-01 2.64049143e-01 -4.08294410e-01
1.66065264e+00 -1.58512735e+00 2.89262384e-01 -5.81905991e-02
3.84909630e-01 -4.24563438e-01 -2.33269870e-01 6.36072814e-01
-2.37223059e-01 5.17832220e-01 -4.37556982e-01 -5.84923215e-02
9.12052989e-02 1.10932685e-01 -7.15483129e-01 4.24754053e-01
1.85955614e-01 1.31547773e+00 -1.31920063e+00 -3.70441198e-01
-8.37889612e-02 8.61670449e-02 -9.11082506e-01 2.64861494e-01
-7.25963473e-01 2.85109997e-01 -5.30408680e-01 3.48992974e-01
3.02773923e-01 -3.47073048e-01 6.08859956e-01 2.22400576e-01
1.23870738e-01 8.86486769e-01 -1.02380502e+00 1.68422616e+00
-5.56487620e-01 3.70642573e-01 -2.98080444e-01 -9.34553206e-01
1.40729710e-01 5.63930929e-01 3.93771887e-01 -5.86628377e-01
-1.12882905e-01 -1.75452903e-01 -1.47365049e-01 -7.03077734e-01
1.69740424e-01 -3.54322433e-01 -2.91169971e-01 8.48103344e-01
-4.34331931e-02 1.91701040e-01 1.45666093e-01 5.79433024e-01
1.28049684e+00 -8.42952132e-02 5.56101441e-01 -1.16382368e-01
-2.28522494e-01 -1.34131536e-01 3.65281910e-01 1.34946144e+00
2.77885944e-01 1.88370898e-01 1.10654867e+00 -1.31448373e-01
-9.48283911e-01 -1.30476928e+00 3.94221433e-02 6.03343129e-01
-3.75984102e-01 -4.40302104e-01 -5.50993800e-01 -7.29363978e-01
2.24805579e-01 1.06956112e+00 -9.99242008e-01 -4.33957666e-01
-2.74085253e-01 -1.02959871e+00 6.60749018e-01 6.17387354e-01
4.54597138e-02 -6.67199075e-01 -3.97018522e-01 2.67648667e-01
-3.65139931e-01 -4.79957998e-01 -5.08832157e-01 -1.17546357e-01
-9.89762366e-01 -1.31685388e+00 -5.74832618e-01 1.68886438e-01
1.80170015e-01 4.50460203e-02 1.10701203e+00 -3.31194758e-01
4.12830450e-02 4.68950659e-01 2.43151307e-01 -5.53799629e-01
-7.10317433e-01 -1.37495160e-01 3.08403615e-02 -2.04346478e-01
1.20682634e-01 -5.21742940e-01 -8.82745743e-01 -1.80762470e-01
-1.04154050e+00 1.34631962e-01 4.15975630e-01 9.59855855e-01
3.17384541e-01 -2.66435891e-01 8.95402670e-01 -1.46764314e+00
1.21094537e+00 -1.15964293e+00 -7.15632141e-01 2.33840063e-01
-8.68308365e-01 3.69184196e-01 5.02804995e-01 -4.22554851e-01
-1.42584527e+00 -5.45745552e-01 1.79743007e-01 -1.31887004e-01
6.55971691e-02 1.04458404e+00 -2.12355003e-01 8.69070053e-01
7.47417927e-01 -4.12194967e-01 -1.45748481e-01 -4.17708606e-01
1.13676250e+00 5.40371239e-01 4.18237984e-01 -5.92078507e-01
2.71948278e-01 7.33563542e-01 -8.83387681e-03 -1.39418274e-01
-1.03602016e+00 -1.57401830e-01 -5.69792539e-02 4.91450205e-02
7.32990086e-01 -1.10549414e+00 -1.18136632e+00 -2.26537585e-01
-1.29077661e+00 -4.71051216e-01 -4.27554488e-01 6.34355962e-01
-8.26368570e-01 2.61318162e-02 -5.54169178e-01 -1.13351345e+00
4.11603749e-02 -9.22228038e-01 1.13915515e+00 -2.04089150e-01
-3.63499671e-01 -1.39976835e+00 1.91892922e-01 5.96224032e-02
4.11268920e-02 2.81415582e-01 1.30158937e+00 -4.16642521e-03
-8.05761039e-01 5.88764511e-02 -1.69201612e-01 -4.28658456e-01
1.93734452e-01 -1.46335751e-01 -7.04958379e-01 -2.06926093e-02
-2.03311786e-01 -2.19367087e-01 1.21039128e+00 9.61739242e-01
1.15361905e+00 -9.83277023e-01 -7.83360839e-01 2.94300228e-01
1.22547710e+00 -1.64152812e-02 5.60765922e-01 -1.79086506e-01
4.79259729e-01 4.55506027e-01 3.86591434e-01 7.46609867e-01
8.02691519e-01 4.74263698e-01 5.64970195e-01 -3.28557901e-02
-1.78346351e-01 -1.12464130e+00 3.45590293e-01 2.55622298e-01
5.74476048e-02 -5.43137491e-01 -7.05802441e-01 8.48795056e-01
-2.00250196e+00 -1.33421886e+00 -4.82901782e-01 2.23856926e+00
1.21726263e+00 -5.61718419e-02 3.17079782e-01 -2.76223391e-01
5.78533351e-01 2.29750454e-01 -6.25904143e-01 -1.93887278e-01
-1.52712598e-01 1.69116393e-01 8.06534946e-01 6.76972628e-01
-9.50514078e-01 6.77063465e-01 7.97951412e+00 6.69616759e-01
-6.93106294e-01 4.86809939e-01 6.87739611e-01 -3.59356582e-01
-1.31734514e+00 2.34468162e-01 -3.83724272e-01 5.79637229e-01
1.31224358e+00 -4.08269227e-01 5.71625710e-01 2.79024273e-01
6.68909311e-01 -4.61539663e-02 -1.39008415e+00 4.95593965e-01
-5.21019995e-01 -1.94551289e+00 3.46491218e-01 1.94314107e-01
1.17651165e+00 -3.96116897e-02 2.07122028e-01 6.35748729e-02
1.31337559e+00 -1.06412625e+00 6.41334832e-01 7.72271574e-01
1.11624730e+00 -4.28892374e-01 3.81047964e-01 3.75687301e-01
-4.94409859e-01 -4.13616121e-01 -4.28523242e-01 -2.92636544e-01
5.10110378e-01 8.72160017e-01 -1.00712252e+00 7.70869792e-01
2.28292346e-01 8.29269409e-01 -3.73307094e-02 7.29471028e-01
-6.72125638e-01 1.27789414e+00 -9.03511345e-02 -5.52159399e-02
-5.75522222e-02 2.00533390e-01 4.87935156e-01 9.78370786e-01
2.14070380e-01 1.31403863e-01 -3.41603667e-01 1.22823322e+00
-4.37980473e-01 -2.76139319e-01 -1.18028462e+00 -2.55756706e-01
6.31085694e-01 2.76600987e-01 -1.08646229e-01 -5.25339603e-01
-4.53393877e-01 6.56018019e-01 3.56341571e-01 7.36272812e-01
-9.73502100e-01 5.04795671e-01 6.43823564e-01 -2.69409511e-02
-7.99683109e-02 1.00948311e-01 -3.59245390e-01 -1.35490835e+00
-4.26267862e-01 -8.91196191e-01 9.43911433e-01 -6.92100644e-01
-1.40643489e+00 -4.65673983e-01 5.59867144e-01 -5.91800928e-01
-6.62854850e-01 -7.39187375e-02 -4.49330419e-01 8.39233577e-01
-1.46361673e+00 -9.01640534e-01 4.30719584e-01 3.11252713e-01
4.09239054e-01 3.07982236e-01 6.87887192e-01 1.43276364e-01
-5.54672956e-01 2.31153190e-01 -5.60839251e-02 -2.68369704e-01
4.75905031e-01 -1.33352745e+00 5.49069881e-01 9.31645751e-01
2.19738886e-01 1.03037977e+00 7.11145639e-01 -1.15066385e+00
-1.58345294e+00 -1.10155988e+00 9.54287767e-01 -5.19548833e-01
8.50313783e-01 -4.56871450e-01 -3.56230944e-01 9.00828004e-01
3.62799436e-01 -6.60544574e-01 7.54410684e-01 6.23011827e-01
-3.64988178e-01 1.89248189e-01 -7.02667654e-01 1.02817917e+00
1.57387316e+00 -7.23973095e-01 -6.33527815e-01 4.48333293e-01
1.20871007e+00 -2.29432672e-01 -6.10392809e-01 1.45683110e-01
4.15566504e-01 -5.37909329e-01 1.06336331e+00 -1.33201110e+00
1.16444135e+00 1.18790440e-01 4.50962149e-02 -1.52174354e+00
-3.16574544e-01 -1.04437363e+00 -4.94957626e-01 9.59094405e-01
6.19100690e-01 -6.31245434e-01 3.54298472e-01 4.51738715e-01
3.03757995e-01 -5.47176838e-01 -8.54571819e-01 -4.26820874e-01
1.85555890e-02 -8.04779232e-01 1.02516842e+00 1.14111054e+00
-4.51878179e-03 5.01742780e-01 -7.12030053e-01 2.80554563e-01
7.42968619e-01 1.94772601e-01 4.59090352e-01 -1.00972080e+00
-5.38868070e-01 -3.79832804e-01 2.64311045e-01 -1.24663734e+00
3.02452415e-01 -9.67451155e-01 -4.05178010e-01 -1.65913892e+00
5.18311262e-01 -2.07641184e-01 -9.13986266e-02 2.67443419e-01
-5.78188717e-01 -2.69084871e-01 -2.00652450e-01 -1.02523461e-01
-2.04115853e-01 7.56575406e-01 1.27536559e+00 -3.37921888e-01
6.75431341e-02 5.81805967e-02 -1.08252716e+00 3.79586577e-01
4.22203481e-01 -7.74813175e-01 -9.57028747e-01 -5.19993305e-01
7.50405133e-01 8.52863252e-01 7.70663023e-01 1.85228854e-01
1.69996127e-01 -5.74310660e-01 -7.92280883e-02 -3.03169042e-01
-2.28684619e-01 -3.50726753e-01 2.70781785e-01 6.27227783e-01
-9.66425717e-01 3.48286144e-02 -5.51349856e-02 1.27658355e+00
2.82685012e-02 6.46002069e-02 -1.36115223e-01 -7.77518153e-02
-1.86159566e-01 5.07115841e-01 -4.10521626e-01 2.85746217e-01
5.80544293e-01 4.40123409e-01 -6.21175766e-01 -9.75335717e-01
-5.22285283e-01 4.48402911e-01 4.15188707e-02 4.97208476e-01
5.54897189e-01 -1.42510486e+00 -9.04040217e-01 -2.57213682e-01
7.49568567e-02 -3.34627718e-01 3.17330509e-01 9.32622135e-01
2.43837878e-01 6.78117394e-01 3.18905622e-01 -1.42505258e-01
-4.25509632e-01 1.12018776e+00 7.19975308e-02 -3.90455037e-01
-3.23358506e-01 5.75622857e-01 6.58907115e-01 -1.73038095e-01
6.66358471e-02 -4.54880118e-01 -8.69429410e-02 -2.68525854e-02
4.02398169e-01 4.18988317e-01 -4.80081409e-01 2.43210390e-01
-9.41451117e-02 -3.68667692e-01 3.19493383e-01 -5.27856827e-01
1.28397048e+00 -2.15018079e-01 -7.57904053e-02 4.95128393e-01
9.50720072e-01 7.39821717e-02 -1.23633492e+00 -1.56589270e-01
1.75065532e-01 -4.66867983e-01 9.21332687e-02 -1.04299545e+00
-4.40280050e-01 8.42362463e-01 1.70371950e-01 4.66772437e-01
9.77961004e-01 2.43330732e-01 1.66563138e-01 -2.02360630e-01
2.04121470e-01 -6.00470245e-01 -2.30148479e-01 -6.33925125e-02
9.84699726e-01 -1.01372266e+00 -1.41973540e-01 -5.49344361e-01
-3.31657469e-01 5.15905976e-01 -5.69529608e-02 1.65481463e-01
5.02580464e-01 3.68630081e-01 -5.09711087e-01 -5.19041240e-01
-1.52073801e+00 -1.89326271e-01 2.05382705e-01 7.31809139e-01
7.89734781e-01 6.63880050e-01 -7.04242229e-01 6.72312379e-01
-2.01248720e-01 2.25656688e-01 7.32493997e-01 2.78334588e-01
3.25305015e-01 -1.05072045e+00 -1.77592278e-01 7.76944458e-01
-5.02409697e-01 -4.72542077e-01 -3.84766072e-01 6.53149068e-01
-3.21140945e-01 1.11655402e+00 5.97094942e-04 6.65602610e-02
3.48606020e-01 -1.17514417e-01 5.33556879e-01 -6.65946662e-01
-1.00775525e-01 5.68464883e-02 4.35805261e-01 -7.10615456e-01
-5.34534752e-01 -8.08720112e-01 -8.25619996e-01 -5.14447153e-01
-3.55342850e-02 9.81813413e-04 2.32731700e-01 7.25945234e-01
4.71271574e-01 7.69743562e-01 5.43175101e-01 3.80105190e-02
-9.40393448e-01 -1.03826320e+00 -2.07040817e-01 2.77820349e-01
5.85061431e-01 -8.03096831e-01 3.68882902e-02 1.19491644e-01]
|
[8.108266830444336, 5.42254638671875]
|
06ea026e-3db4-4c38-a15e-bdea4528c6c6
|
survey-on-software-isp-methods-based-on-deep
|
2305.11994
| null |
https://arxiv.org/abs/2305.11994v2
|
https://arxiv.org/pdf/2305.11994v2.pdf
|
ISP meets Deep Learning: A Survey on Deep Learning Methods for Image Signal Processing
|
The entire Image Signal Processor (ISP) of a camera relies on several processes to transform the data from the Color Filter Array (CFA) sensor, such as demosaicing, denoising, and enhancement. These processes can be executed either by some hardware or via software. In recent years, Deep Learning has emerged as one solution for some of them or even to replace the entire ISP using a single neural network for the task. In this work, we investigated several recent pieces of research in this area and provide deeper analysis and comparison among them, including results and possible points of improvement for future researchers.
|
['Claudio Filipi Gonçalves dos Santos', 'Rodolfo Coelho Dalapicola', 'Mayara Costa Regazio', 'Bruno Melo de Souza', 'Lucas Borges Rondon', 'Iago Oliveira Lima', 'Guilherme Augusto Bileki', 'Leonardo Tadeu Lopes', 'Wladimir Barroso Guedes de Araújo Neto', 'Rodrigo Reis Arrais', 'Jhessica Victoria Santos da Silva', 'Matheus Henrique Marques da Silva']
|
2023-05-19
| null | null | null | null |
['demosaicking']
|
['computer-vision']
|
[ 2.44056195e-01 -3.59969705e-01 5.63508987e-01 -4.58023012e-01
-3.73420209e-01 -4.28320497e-01 3.32056373e-01 -2.41496846e-01
-6.14428341e-01 2.69789189e-01 -1.55688897e-01 -2.10093081e-01
1.37065828e-01 -7.77703881e-01 -6.50225580e-01 -9.47762430e-01
6.51889294e-02 -6.83613941e-02 4.16048318e-01 -2.85978224e-02
2.20058367e-01 7.64506638e-01 -1.53314006e+00 2.39447832e-01
6.29083633e-01 1.50085318e+00 1.45098880e-01 8.11272025e-01
-7.52912238e-02 7.05734789e-01 -5.80606341e-01 -4.90947843e-01
5.85157037e-01 -1.84002459e-01 -7.45478570e-02 1.57830134e-01
5.22427917e-01 -7.45780766e-01 -3.60587269e-01 1.41380727e+00
4.10985112e-01 -2.10395992e-01 2.51675069e-01 -9.68274653e-01
-3.05900514e-01 3.40232849e-01 -6.20649457e-01 1.61247123e-02
-1.84222627e-02 9.98007134e-02 3.45503479e-01 -9.66251254e-01
1.29732415e-01 9.58901107e-01 7.44036376e-01 1.18486300e-01
-7.07075715e-01 -7.51516879e-01 -4.01382655e-01 4.46053714e-01
-1.14962029e+00 -4.15752083e-01 1.02306533e+00 -3.68484050e-01
7.82468915e-01 -2.86877770e-02 8.54033172e-01 7.51686990e-01
5.13071001e-01 4.64917362e-01 1.30435205e+00 -4.60751414e-01
3.94335687e-01 -1.42087862e-01 2.28322580e-01 3.91653270e-01
4.39069867e-01 1.70036778e-01 -3.61449361e-01 6.17812537e-02
8.47374678e-01 -1.10829428e-01 -4.16290939e-01 -1.35777414e-01
-6.57591820e-01 6.35418475e-01 2.30054170e-01 3.40974480e-01
-6.25142992e-01 2.68943399e-01 2.51723588e-01 4.95195508e-01
4.78266686e-01 3.19984525e-01 -4.27521974e-01 -3.92933339e-02
-1.17817664e+00 -9.96684358e-02 7.90761411e-01 4.55911666e-01
9.79163885e-01 4.07765716e-01 5.06259263e-01 3.28234106e-01
3.98947030e-01 2.36147657e-01 3.24676067e-01 -9.71820056e-01
7.36798495e-02 2.87069261e-01 3.74520235e-02 -1.14435852e+00
-5.03765106e-01 -4.23224568e-01 -9.57141757e-01 6.40822530e-01
3.38639587e-01 -6.28840029e-01 -8.98741424e-01 9.88469422e-01
9.97428596e-02 3.86140645e-01 -2.08089665e-01 1.04946530e+00
6.38688505e-01 9.62485373e-01 -2.54548281e-01 1.22362234e-01
1.17404580e+00 -8.14711869e-01 -8.40616167e-01 -5.23283184e-01
-6.56192228e-02 -9.88291323e-01 5.30461013e-01 1.17497456e+00
-1.10712755e+00 -8.98924112e-01 -1.43712366e+00 -1.68117300e-01
-2.78450876e-01 4.62543100e-01 6.42148077e-01 7.15040743e-01
-1.34921575e+00 7.41457403e-01 -1.13108218e+00 -2.54556566e-01
9.42725688e-02 3.89190048e-01 -2.91838069e-02 -7.84722269e-02
-1.05256236e+00 8.43031466e-01 2.88792431e-01 6.58513427e-01
-7.28780329e-01 -2.46053740e-01 -5.69939375e-01 1.95668787e-01
2.23177910e-01 -5.36278009e-01 1.02987206e+00 -1.32984376e+00
-1.68103552e+00 6.28366470e-01 2.86719114e-01 -6.21175647e-01
3.23758632e-01 -6.62626266e-01 -5.00200748e-01 4.08234030e-01
-5.16474664e-01 2.36286193e-01 1.44821632e+00 -1.02954733e+00
-7.95262039e-01 -5.37997425e-01 -7.84071721e-03 -1.99239925e-01
-2.83247560e-01 2.59876251e-01 -5.22177815e-01 -5.81551731e-01
4.50229257e-01 -7.21078336e-01 -1.74165949e-01 1.83863699e-01
-4.80512157e-02 1.85586989e-01 9.20778155e-01 -1.10240698e+00
9.09223020e-01 -2.40000224e+00 1.06256969e-01 1.10701784e-01
1.70261145e-01 6.41361892e-01 5.97012304e-02 3.26209813e-01
-1.44710064e-01 -2.77955502e-01 -1.72811419e-01 -4.10621822e-01
-4.02074665e-01 -1.78069957e-02 -2.42531747e-01 8.86716068e-01
1.73977137e-01 4.53924090e-01 -2.20405236e-01 -9.66966450e-02
4.46308970e-01 6.02717996e-01 -1.48863733e-01 1.12021349e-01
3.68565805e-02 2.85741806e-01 -1.93977118e-01 5.55974901e-01
1.12533391e+00 2.36609086e-01 1.91348001e-01 -6.69372916e-01
-4.14897025e-01 2.19483059e-02 -1.60783744e+00 1.47768533e+00
-3.17764312e-01 1.04809117e+00 8.73714745e-01 -1.31640100e+00
9.14301157e-01 2.03762725e-01 4.64337856e-01 -5.30002594e-01
5.37302554e-01 3.49885494e-01 -1.12391569e-01 -6.75697267e-01
5.63924253e-01 -1.03050597e-01 3.57798129e-01 2.50561059e-01
9.78005752e-02 1.68122962e-01 -4.52071764e-02 -2.40857691e-01
1.07416582e+00 1.35001212e-01 2.40382731e-01 3.01079405e-03
6.37966514e-01 6.75104111e-02 5.36375344e-01 6.06338382e-01
-2.49804035e-01 5.34051120e-01 1.23768918e-01 -6.63065970e-01
-1.03004396e+00 -8.63049805e-01 1.63023412e-01 5.39302588e-01
1.79224744e-01 -5.39524369e-02 -7.94125855e-01 -9.17317420e-02
-1.80803433e-01 2.72718132e-01 -1.92455232e-01 -7.94929937e-02
-7.96246171e-01 -7.86128283e-01 4.22197849e-01 5.72184384e-01
1.01019597e+00 -7.55906701e-01 -1.04312837e+00 3.45568240e-01
1.99903071e-01 -1.25010192e+00 1.39000982e-01 4.65456128e-01
-1.10812354e+00 -1.02487862e+00 -5.31687617e-01 -7.88422287e-01
3.84355098e-01 5.39245725e-01 9.46328700e-01 7.04742745e-02
-1.31072877e-02 2.32650027e-01 -6.46820068e-01 -4.26848978e-01
-1.83845013e-01 -4.11596119e-01 -3.75156924e-02 3.43449473e-01
3.42381120e-01 -6.79798484e-01 -4.93199080e-01 -8.64469111e-02
-9.97759402e-01 -1.17993422e-01 9.12617862e-01 5.30372798e-01
3.81143898e-01 5.48119783e-01 -3.79739441e-02 -3.44710678e-01
4.97413814e-01 -1.13876723e-01 -1.13686454e+00 -6.03793710e-02
-3.41183722e-01 -1.30604491e-01 7.06183016e-01 -1.10621773e-01
-1.17992163e+00 3.65688443e-01 -3.57983589e-01 -2.64164984e-01
-2.40924701e-01 6.31244242e-01 -3.16938072e-01 -4.28053111e-01
2.41270676e-01 3.16179663e-01 1.85959741e-01 -7.29299307e-01
-2.86077373e-02 8.47025752e-01 8.02491069e-01 8.07421189e-03
9.58483994e-01 7.90060282e-01 -3.82284150e-02 -1.12470412e+00
-3.75379741e-01 -2.52464175e-01 -5.33325315e-01 -4.15889412e-01
1.08999062e+00 -1.08177328e+00 -5.46292007e-01 1.03480196e+00
-1.33812761e+00 -1.66290671e-01 2.21086532e-01 7.32427597e-01
4.94755171e-02 5.47839463e-01 -8.99626791e-01 -5.59264541e-01
-3.18588883e-01 -1.19061232e+00 9.63607252e-01 6.44842803e-01
3.41046035e-01 -7.36885309e-01 -6.77764714e-02 1.25965327e-01
5.76856613e-01 2.28918195e-02 4.55797017e-01 -1.86649516e-01
-7.82083869e-01 -3.99064451e-01 -2.18483388e-01 1.02204287e+00
-9.88746583e-02 3.57299238e-01 -9.70045924e-01 -4.38549876e-01
7.37557650e-01 2.58405525e-02 1.06982589e+00 7.46012509e-01
1.03811812e+00 1.47533134e-01 -6.59860298e-02 1.07017648e+00
1.83232236e+00 5.57397425e-01 9.96204913e-01 5.67809045e-01
2.83015311e-01 3.93407553e-01 4.52742189e-01 4.03303474e-01
-7.34039536e-03 4.56909001e-01 7.42124140e-01 -3.28953475e-01
-1.81977689e-01 3.52708995e-01 6.37077093e-01 6.21362031e-01
-1.78379789e-01 -1.88980654e-01 -8.16234589e-01 9.50273052e-02
-1.45611513e+00 -6.82502866e-01 -4.04832006e-01 1.77254987e+00
1.72534287e-01 5.64498305e-02 -3.05293381e-01 3.79891664e-01
7.98400760e-01 2.03856602e-01 -3.04510385e-01 -3.73715430e-01
-1.88007846e-01 3.92064035e-01 6.43478274e-01 2.43842065e-01
-1.20352316e+00 6.02218270e-01 6.71288157e+00 4.37316954e-01
-1.62280679e+00 -2.01444253e-02 3.52341563e-01 2.22686321e-01
3.35057437e-01 3.81045118e-02 -4.92811143e-01 3.91314507e-01
7.26700187e-01 3.39681953e-01 6.62747681e-01 9.26080942e-01
3.10561568e-01 -4.96039450e-01 -7.80347168e-01 1.15896201e+00
1.31268471e-01 -1.13953590e+00 -2.16687143e-01 -3.65443677e-02
3.98069978e-01 1.74454048e-01 -6.56802952e-02 -2.33536083e-02
-9.92602333e-02 -5.45715868e-01 7.50171125e-01 6.40004396e-01
2.66841620e-01 -7.09597528e-01 1.17397523e+00 2.04845011e-01
-1.04497802e+00 -2.71908432e-01 -4.45650846e-01 -5.53444147e-01
2.18552724e-01 9.79575574e-01 -1.57631904e-01 6.86985672e-01
1.15332353e+00 4.91108567e-01 -7.44673550e-01 1.21172178e+00
-3.57657164e-01 7.79502869e-01 -3.27437967e-01 2.98498958e-01
2.71589190e-01 -5.22038162e-01 4.83955264e-01 9.64498758e-01
6.84069812e-01 5.57252392e-02 -1.08814187e-01 5.84645927e-01
2.32249200e-01 -4.66774434e-01 -2.78812289e-01 1.35360852e-01
9.13393199e-02 1.60834455e+00 -7.37971008e-01 -4.22142863e-01
-7.87163734e-01 9.50936437e-01 -2.03509301e-01 3.86432916e-01
-1.00314629e+00 -6.57627285e-01 7.78552234e-01 -7.05355257e-02
4.80583072e-01 -7.09626377e-01 -4.05669153e-01 -1.07459867e+00
8.51762593e-02 -9.01272893e-01 6.15017302e-02 -1.11358774e+00
-9.84345078e-01 5.51976204e-01 -3.92862439e-01 -1.39766002e+00
1.16524264e-01 -1.06131208e+00 -5.55044949e-01 7.00314641e-01
-1.58579195e+00 -7.07326293e-01 -6.73236489e-01 5.43903887e-01
2.83718586e-01 -1.92582533e-01 2.73775488e-01 7.43736625e-01
-5.99949300e-01 -1.98958933e-01 1.24765225e-01 2.87369996e-01
7.65872955e-01 -9.54436600e-01 3.38146150e-01 1.43501103e+00
-4.71025193e-03 2.39006221e-01 9.17146146e-01 -4.44741249e-01
-1.81903231e+00 -8.56423497e-01 3.78829837e-01 3.35937113e-01
4.72070009e-01 -1.82665676e-01 -8.81312549e-01 7.11218655e-01
7.30739892e-01 -1.04419902e-01 3.02990079e-01 -4.60434556e-01
1.08864151e-01 -5.93953967e-01 -9.14833069e-01 3.83274496e-01
4.09359992e-01 -2.36041740e-01 -5.37302673e-01 -3.25524747e-01
2.69572645e-01 -4.63606149e-01 -5.72399676e-01 3.23527366e-01
4.61324364e-01 -1.48966169e+00 8.17574561e-01 2.43547861e-03
2.57483184e-01 -6.42492831e-01 -1.95758566e-01 -1.28210461e+00
-3.32255244e-01 -4.96653289e-01 5.20675853e-02 1.04454005e+00
-1.18495308e-01 -5.66985726e-01 8.28101575e-01 3.50688457e-01
-6.13255203e-01 -1.62753195e-01 -8.02904248e-01 -6.39376044e-01
-2.74745762e-01 -7.81520426e-01 4.49892730e-01 4.46009189e-01
-6.82025552e-01 -1.43627925e-02 -4.61272448e-01 6.99160516e-01
7.27443039e-01 1.94434583e-01 7.00551212e-01 -1.23584938e+00
-2.64601231e-01 -1.84183940e-01 -4.90500182e-01 -1.03044271e+00
-2.21633881e-01 -3.67698938e-01 2.41208598e-02 -1.56905317e+00
-3.93226594e-01 2.48392612e-01 -5.03033660e-02 9.43202972e-02
9.15178806e-02 3.12323183e-01 2.37219483e-01 1.25404829e-02
-1.27254725e-01 2.29380950e-01 9.60014701e-01 -1.38777539e-01
-7.07588568e-02 -1.52939692e-01 -4.95850295e-01 1.24906790e+00
8.64131808e-01 -4.56125557e-01 1.28046721e-01 -9.68991578e-01
3.03258836e-01 7.59998113e-02 5.36992013e-01 -1.62202644e+00
6.43711865e-01 2.71952629e-01 7.18612611e-01 -7.12197661e-01
4.11882997e-01 -1.29744232e+00 3.64666402e-01 7.05146790e-01
3.14311147e-01 1.16573833e-01 1.46990404e-01 2.62571841e-01
-7.06150413e-01 -2.77623177e-01 9.06732917e-01 -1.67467907e-01
-1.23663354e+00 -2.02509120e-01 -7.35494554e-01 -6.84628248e-01
9.72462595e-01 -2.81398267e-01 -2.15764105e-01 -4.24213409e-01
-7.24450827e-01 -2.50129014e-01 2.74806023e-01 -6.06948957e-02
7.35431731e-01 -9.08001244e-01 -4.80932504e-01 4.32759345e-01
-5.91527879e-01 -1.68830484e-01 2.11102396e-01 8.78835022e-01
-1.05347061e+00 1.90101653e-01 -5.59830606e-01 -4.85974014e-01
-1.16459417e+00 4.35812086e-01 5.64270616e-01 -7.73724839e-02
-4.74911958e-01 7.17401385e-01 -3.62047374e-01 2.10493892e-01
2.72190124e-01 -4.44576323e-01 -1.50901094e-01 1.96791843e-01
6.99451327e-01 5.67463100e-01 2.90899456e-01 -2.53940731e-01
-4.17922765e-01 9.02425110e-01 3.48211914e-01 -6.20875470e-02
1.65818667e+00 -2.63299435e-01 -5.04359305e-01 1.13961259e-02
1.00103867e+00 -9.78200883e-02 -1.48136282e+00 3.52850445e-02
-1.97761878e-01 -2.91872472e-01 5.24314821e-01 -5.86217999e-01
-1.68684816e+00 1.04729915e+00 8.28615308e-01 2.63900042e-01
1.91097760e+00 -4.02290046e-01 8.29851925e-01 4.34537947e-01
4.37802762e-01 -1.35075676e+00 -7.55878910e-02 3.73887122e-01
6.98560834e-01 -1.06011641e+00 1.41040191e-01 -2.57929534e-01
-2.07803339e-01 1.69502103e+00 3.78226429e-01 -5.44063330e-01
1.01544547e+00 7.41449118e-01 1.59004778e-01 -6.96496889e-02
-3.09559762e-01 -1.27738476e-01 -1.93489417e-02 5.32519579e-01
2.90408224e-01 -2.96122551e-01 -3.84147823e-01 4.91694272e-01
5.75993471e-02 3.73679668e-01 7.00181842e-01 9.70387101e-01
-6.14551783e-01 -8.67444336e-01 -7.46699393e-01 3.75187606e-01
-4.66925353e-01 -1.86048949e-03 -2.95731872e-01 7.43968189e-01
4.09921110e-01 1.20854568e+00 1.03175148e-01 -7.89948881e-01
2.37243563e-01 -1.57578632e-01 2.95942634e-01 -7.28946179e-02
-7.00216949e-01 2.38458261e-01 -8.14574491e-03 -7.03051150e-01
-5.44564128e-01 -5.39183795e-01 -1.07392192e+00 -2.19017223e-01
-7.31879622e-02 -1.96476400e-01 1.00756502e+00 7.99142957e-01
1.45261273e-01 5.99994540e-01 5.28426886e-01 -1.11641455e+00
-3.09800714e-01 -8.24472964e-01 -9.38179672e-01 -1.40501365e-01
4.42738086e-01 -3.79736215e-01 -5.72677076e-01 2.01200828e-01]
|
[11.374403953552246, -2.3213202953338623]
|
7f7bce46-ed8d-4261-9afc-e64110340640
|
aec-in-a-netshell-on-target-and-topology
|
2103.09007
| null |
https://arxiv.org/abs/2103.09007v1
|
https://arxiv.org/pdf/2103.09007v1.pdf
|
AEC in a NetShell: On Target and Topology Choices for FCRN Acoustic Echo Cancellation
|
Acoustic echo cancellation (AEC) algorithms have a long-term steady role in signal processing, with approaches improving the performance of applications such as automotive hands-free systems, smart home and loudspeaker devices, or web conference systems. Just recently, very first deep neural network (DNN)-based approaches were proposed with a DNN for joint AEC and residual echo suppression (RES)/noise reduction, showing significant improvements in terms of echo suppression performance. Noise reduction algorithms, on the other hand, have enjoyed already a lot of attention with regard to DNN approaches, with the fully convolutional recurrent network (FCRN) architecture being among state of the art topologies. The recently published impressive echo cancellation performance of joint AEC/RES DNNs, however, so far came along with an undeniable impairment of speech quality. In this work we will heal this issue and significantly improve the near-end speech component quality over existing approaches. Also, we propose for the first time-to the best of our knowledge-a pure DNN AEC in the form of an echo estimator, that is based on a competitive FCRN structure and delivers a quality useful for practical applications.
|
['Tim Fingscheidt', 'Ernst Seidel', 'Jan Franzen']
|
2021-03-16
| null | null | null | null |
['acoustic-echo-cancellation', 'acoustic-echo-cancellation']
|
['medical', 'speech']
|
[ 2.28354558e-01 -1.16607778e-01 6.97270155e-01 -9.55925435e-02
-8.16010177e-01 -1.79604560e-01 5.80587149e-01 -1.60476983e-01
-6.00199640e-01 4.14609283e-01 4.93507832e-01 -3.03564459e-01
-4.19037253e-01 -3.43167752e-01 -3.29042315e-01 -7.79872656e-01
-2.13495776e-01 -4.40201722e-02 2.35679194e-01 -5.92866898e-01
-3.66103090e-02 5.49819231e-01 -1.90683568e+00 1.41715840e-01
5.47536552e-01 1.18963075e+00 1.73024356e-01 8.35723519e-01
3.47074419e-02 8.46210241e-01 -6.91099524e-01 -5.38619757e-01
1.38172686e-01 -3.63940626e-01 -1.13120303e-01 -6.33846402e-01
3.06013435e-01 -3.03370297e-01 -5.52379072e-01 9.05235648e-01
1.10715806e+00 1.58799961e-01 3.97759944e-01 -5.62730968e-01
-2.48755008e-01 7.32705712e-01 -1.12716407e-01 1.63945973e-01
1.57605901e-01 -1.25351707e-02 8.78556311e-01 -1.09158325e+00
2.20020145e-01 9.88477349e-01 9.67113495e-01 6.78762436e-01
-8.70997131e-01 -6.87334836e-01 -1.31172210e-01 4.04915810e-01
-1.12799680e+00 -9.31411088e-01 9.32566345e-01 1.52717531e-01
1.18168819e+00 3.84870231e-01 5.89865983e-01 1.30888200e+00
-2.78731473e-02 8.84822905e-01 9.72202957e-01 -5.90683043e-01
3.36025178e-01 -8.91059861e-02 -1.43523797e-01 1.62586153e-01
-1.05782069e-01 2.73737311e-01 -5.92155993e-01 2.37289414e-01
4.36844945e-01 -2.43247449e-01 -5.95929086e-01 2.99377684e-02
-7.46856391e-01 4.79790360e-01 2.91958958e-01 9.27950442e-01
-5.42646170e-01 3.90167236e-01 5.26283979e-01 5.44449091e-01
5.07119358e-01 2.67313629e-01 -2.93030620e-01 -4.17142034e-01
-1.14151394e+00 2.93447942e-01 1.08774006e+00 4.30949479e-01
-1.81663223e-02 8.07142198e-01 5.32808788e-02 1.17912483e+00
8.05429965e-02 7.08254755e-01 7.05707848e-01 -7.91092694e-01
3.22002292e-01 -2.39420712e-01 -9.07140747e-02 -1.06113315e+00
-5.49642682e-01 -1.05259264e+00 -1.29855096e+00 4.36714381e-01
2.44659871e-01 -2.31038764e-01 -6.55914664e-01 1.79334271e+00
-4.09568138e-02 1.86653361e-01 2.62695819e-01 1.02081215e+00
6.15443051e-01 6.40969455e-01 -1.83077082e-01 -2.13945895e-01
1.01799548e+00 -7.37234771e-01 -1.21963823e+00 -1.01360291e-01
1.72595948e-01 -9.72116232e-01 4.04260963e-01 8.40470850e-01
-1.23837864e+00 -6.07172728e-01 -1.06848824e+00 2.65272468e-01
-2.92463899e-01 -1.01115875e-01 2.19720080e-01 1.02366972e+00
-1.46230757e+00 8.43757868e-01 -6.72751129e-01 2.39408808e-03
-1.52136892e-01 4.37059969e-01 -1.83508098e-01 7.46337250e-02
-1.44047546e+00 8.91224444e-01 -1.60623163e-01 6.09820366e-01
-6.80983722e-01 -7.77707279e-01 -5.17980814e-01 2.73161829e-01
3.44077706e-01 -2.50891954e-01 1.48032653e+00 -1.06654227e+00
-2.06976175e+00 3.25515151e-01 7.42803961e-02 -9.77222621e-01
4.30579454e-01 -6.26161933e-01 -1.02780962e+00 -1.21769467e-02
-5.60988843e-01 3.85133952e-01 1.13127697e+00 -1.04107845e+00
-5.55389643e-01 -1.93392798e-01 -2.68797964e-01 1.25079915e-01
-4.17129695e-01 1.96851697e-03 -1.15735844e-01 -8.26794207e-01
-9.50678345e-03 -6.33894205e-01 -2.66538173e-01 -2.75057614e-01
-1.50809348e-01 -1.64911434e-01 6.04216039e-01 -9.43886817e-01
1.48313928e+00 -2.27189326e+00 1.37711048e-01 6.49026036e-02
1.25586957e-01 1.08504629e+00 -3.31489623e-01 5.62049091e-01
-3.07020992e-01 -2.67959535e-01 -2.58598298e-01 -6.04330659e-01
7.54233673e-02 -2.02465609e-01 -3.23630959e-01 4.53575104e-01
1.32821873e-01 6.09558105e-01 -5.40272295e-01 1.82721078e-01
5.05017936e-01 1.02832818e+00 -4.26880568e-01 1.08687848e-01
2.42546663e-01 4.24646540e-03 1.52041212e-01 1.42340451e-01
7.85645723e-01 5.73027790e-01 4.04807739e-02 -5.37853576e-02
-4.71820772e-01 2.45855838e-01 -1.31937158e+00 1.54947150e+00
-6.89237475e-01 1.04458570e+00 8.62122953e-01 -1.07873309e+00
1.11617637e+00 7.23088980e-01 4.24356073e-01 -1.29581070e+00
8.87529030e-02 9.68261123e-01 4.18848366e-01 -2.23819926e-01
5.83428681e-01 -3.40287268e-01 3.85518134e-01 -1.85929164e-02
1.89628392e-01 3.36890072e-02 -1.72317475e-01 -8.15731063e-02
1.21257937e+00 -2.16860190e-01 2.04660352e-02 -3.12279910e-01
8.42929184e-01 -7.28048980e-01 1.98547944e-01 8.48048627e-01
-2.38349691e-01 8.64076972e-01 4.34153192e-02 -1.55170739e-01
-9.83029366e-01 -7.51700759e-01 4.63867821e-02 8.47795546e-01
-4.16168809e-01 -2.18888029e-01 -9.98957574e-01 7.66004696e-02
-2.88923383e-01 7.05543756e-01 -1.98655594e-02 -1.02956943e-01
-1.04038799e+00 -2.12492403e-02 8.55044186e-01 2.58523494e-01
4.04307336e-01 -1.37405777e+00 -4.62721467e-01 8.18826139e-01
4.29368429e-02 -1.09249246e+00 -1.44687504e-01 5.53524137e-01
-6.50313735e-01 -5.09154499e-01 -1.39119184e+00 -8.06225479e-01
-2.02653408e-01 1.28231451e-01 9.76853251e-01 -1.92044511e-01
-8.64094719e-02 3.74396980e-01 -3.13308895e-01 -4.70681310e-01
-5.57480872e-01 5.55323362e-02 3.67675841e-01 1.83653444e-01
6.75772429e-02 -9.79147971e-01 -9.03456032e-01 1.27278045e-01
-9.91404235e-01 -3.76547813e-01 6.92635953e-01 5.61244607e-01
-1.99911892e-02 -4.10455279e-03 9.21907842e-01 -4.66133386e-01
9.31313574e-01 -9.70097259e-02 -5.32131732e-01 -1.18767120e-01
-4.78757530e-01 -9.84464586e-02 1.06770146e+00 -2.34303728e-01
-1.25937116e+00 -2.50217229e-01 -1.07484448e+00 -6.22838318e-01
-2.94809729e-01 4.57762033e-01 -2.26083212e-02 3.38973179e-02
5.05264223e-01 3.09519440e-01 -6.70541450e-02 -8.27464521e-01
1.66625246e-01 9.02269900e-01 7.74960220e-01 6.01898059e-02
6.01046860e-01 2.13201836e-01 4.68760133e-02 -1.15866041e+00
-4.07396197e-01 -6.55424178e-01 -6.63407370e-02 -2.99874693e-01
4.49709326e-01 -8.22165489e-01 -7.52892733e-01 9.62341070e-01
-1.28368568e+00 -1.59759045e-01 -2.84616947e-01 7.63626993e-01
-2.71023005e-01 3.30887616e-01 -6.94043756e-01 -1.31509459e+00
-5.81685901e-01 -9.95500684e-01 6.16483867e-01 1.39857516e-01
4.37125713e-02 -7.94234097e-01 1.99343950e-01 -1.60314098e-01
1.26723373e+00 -2.00250462e-01 3.52933824e-01 -5.67394257e-01
-2.90515214e-01 -1.80042997e-01 7.30037391e-02 7.96722174e-01
-1.24485120e-01 -5.30878782e-01 -1.35575557e+00 -2.69421190e-01
4.25505906e-01 6.93773776e-02 1.15657294e+00 5.89276016e-01
7.44616926e-01 -1.68268401e-02 1.40545085e-01 4.32069391e-01
1.30162060e+00 3.43694627e-01 8.88241172e-01 -1.17816992e-01
2.17198625e-01 4.88729924e-01 -8.10011697e-04 4.28748429e-01
-3.98707896e-01 7.49225497e-01 3.11207145e-01 -2.12912783e-01
-7.60095239e-01 5.95828220e-02 3.48103791e-01 1.51972115e+00
-1.21188685e-01 -7.15845525e-01 -5.86275816e-01 4.28062946e-01
-1.44781482e+00 -9.93044853e-01 -2.03762069e-01 2.19375968e+00
4.30580765e-01 2.22596645e-01 -3.15806061e-01 5.86308599e-01
4.98942494e-01 4.11328912e-01 -4.69789118e-01 -6.81666076e-01
-2.40510598e-01 7.73609936e-01 3.72651130e-01 5.71082950e-01
-8.38013649e-01 4.10663486e-01 5.51095915e+00 1.12997866e+00
-1.18782914e+00 1.27728701e-01 3.00866693e-01 -1.06937841e-01
-1.88883379e-01 -5.62782407e-01 -6.06448114e-01 2.46907935e-01
1.42933142e+00 4.40795779e-01 6.17892444e-01 6.11593604e-01
2.35899791e-01 1.94784075e-01 -7.45307684e-01 1.28792620e+00
1.91182792e-01 -8.98289084e-01 -3.34525168e-01 -4.69395556e-02
3.43659937e-01 1.12308949e-01 2.52405137e-01 5.48196554e-01
-2.89943486e-01 -9.45866048e-01 8.54137301e-01 6.53056979e-01
7.85609841e-01 -1.01021647e+00 8.10159326e-01 3.02642196e-01
-1.21555054e+00 -2.09725112e-01 -2.81040341e-01 -9.93684009e-02
4.32711065e-01 1.00391388e+00 -2.72291183e-01 5.64904511e-01
6.36506855e-01 2.86875188e-01 -2.28696573e-03 1.25805914e+00
-3.66811045e-02 9.36567426e-01 -4.64777023e-01 -2.26644233e-01
3.57904375e-01 1.24131385e-02 1.01802588e+00 1.47947681e+00
4.92669374e-01 1.31318912e-01 -6.27746701e-01 4.92210597e-01
-2.40818813e-01 1.05881151e-02 -4.30148125e-01 1.14138126e-01
1.72509417e-01 1.17964506e+00 -4.38854337e-01 1.09361876e-02
-3.13733160e-01 9.18190539e-01 6.72130361e-02 4.34939712e-01
-5.47111750e-01 -8.42825174e-01 8.16503108e-01 -2.18963586e-02
4.83166069e-01 -2.70736545e-01 1.60194725e-01 -7.96710432e-01
2.53672808e-01 -9.70387697e-01 -2.84010321e-01 -4.56591606e-01
-1.02928102e+00 8.36873233e-01 -6.04575872e-01 -1.24695039e+00
-4.22445953e-01 -8.10376406e-01 -4.75027204e-01 9.44899857e-01
-1.67677534e+00 -4.10426468e-01 1.74492195e-01 3.24609458e-01
6.64966881e-01 -2.78648108e-01 9.38233793e-01 8.34060311e-01
-2.42861569e-01 6.35198474e-01 6.39746487e-01 -4.71584089e-02
6.58888519e-01 -1.13891983e+00 5.50872445e-01 8.62123847e-01
8.19666833e-02 5.26311100e-01 9.07997072e-01 -2.50659883e-01
-1.38004231e+00 -7.06779480e-01 9.85765457e-01 3.05073142e-01
4.68831062e-01 -6.44214332e-01 -9.08059418e-01 -2.34332353e-01
7.12520301e-01 -1.97356910e-01 2.61645764e-01 6.78523853e-02
-1.07718177e-01 -5.33083797e-01 -7.27952838e-01 6.35125339e-01
9.96227980e-01 -6.69423163e-01 -3.01542372e-01 -1.56194374e-01
6.57168448e-01 -3.50679666e-01 -3.86736512e-01 4.78563756e-01
7.79568255e-01 -1.58864987e+00 9.58282173e-01 2.25185640e-02
2.80332267e-01 -1.13431655e-01 -1.41339108e-01 -1.44937384e+00
-2.02948034e-01 -1.02708125e+00 -1.69808194e-01 1.35028660e+00
3.46665561e-01 -5.43783903e-01 6.70022190e-01 2.33096674e-01
-6.31254792e-01 -5.82758725e-01 -1.23107135e+00 -8.00377131e-01
-3.80003266e-02 -8.37994635e-01 2.84295768e-01 2.17356265e-01
-2.86033630e-01 3.12686771e-01 -7.67187655e-01 9.44512803e-03
4.80958819e-01 -4.13894147e-01 8.88684243e-02 -1.13495433e+00
-4.12661433e-01 -7.01191604e-01 -2.77375281e-01 -1.15705180e+00
-6.44635111e-02 -5.80744326e-01 3.57845664e-01 -1.47050250e+00
-5.02891600e-01 -6.77050129e-02 -5.09582222e-01 -2.24274933e-01
1.74081191e-01 2.57508397e-01 4.16320115e-01 -1.80232391e-01
-4.07491684e-01 6.07426643e-01 7.63937771e-01 -5.05909249e-02
-2.61030465e-01 3.09869885e-01 -3.22504371e-01 5.83060145e-01
7.94216037e-01 -3.70512009e-01 -2.45378509e-01 -4.58041966e-01
2.29556456e-01 3.05423856e-01 2.70174116e-01 -1.60567307e+00
6.61659956e-01 8.06514084e-01 1.50707483e-01 -5.29366612e-01
6.16415918e-01 -1.03997910e+00 3.18935484e-01 5.88916063e-01
-4.88786280e-01 -2.18433097e-01 2.01344356e-01 5.49088120e-01
-5.73341012e-01 -3.40367764e-01 8.46956968e-01 -4.81702015e-02
-5.04824221e-01 -1.47658825e-01 -9.50578332e-01 -2.32817590e-01
2.12686285e-01 -6.67345524e-02 1.72222391e-01 -8.11661124e-01
-5.61167419e-01 -3.40458930e-01 -4.75523323e-01 4.15270597e-01
7.33073652e-01 -1.09179676e+00 -8.80607605e-01 2.25224361e-01
-4.70164657e-01 -3.68968457e-01 6.26278460e-01 7.73877800e-01
-1.52788103e-01 7.89509952e-01 1.76110640e-02 -2.93099970e-01
-1.04833972e+00 2.73863614e-01 6.26850307e-01 -3.47051054e-01
-6.85078025e-01 7.61886179e-01 5.66258989e-02 -3.72292310e-01
6.56180143e-01 -2.01851845e-01 -3.49229842e-01 6.47684336e-02
7.46526837e-01 6.33325517e-01 6.97544992e-01 -4.31151658e-01
-3.16910177e-01 5.08305430e-01 7.82555267e-02 -5.29175162e-01
1.54080725e+00 -8.31255987e-02 1.60984740e-01 2.90571481e-01
1.37586427e+00 2.59973675e-01 -9.29361105e-01 -1.93901002e-01
-3.07357237e-02 3.30805779e-02 3.85478914e-01 -1.04171765e+00
-1.27033877e+00 1.05419266e+00 1.09180570e+00 3.38849932e-01
1.42803526e+00 -3.84250909e-01 1.02918899e+00 4.50957507e-01
2.05060095e-01 -1.11327136e+00 9.33820475e-03 6.98591232e-01
9.48033333e-01 -7.23531127e-01 -4.72974777e-01 1.47446901e-01
-1.98156670e-01 1.40540290e+00 1.11694038e-02 -2.54497021e-01
8.47706258e-01 7.65782416e-01 2.07219511e-01 5.53712919e-02
-5.21073818e-01 -3.58286589e-01 2.02334702e-01 5.53224564e-01
6.76536918e-01 -7.28155077e-02 -3.54466558e-01 6.14378870e-01
-5.06034680e-03 -5.68041094e-02 7.78758824e-02 5.78989029e-01
-5.48830390e-01 -1.12112474e+00 -3.26329857e-01 2.61511743e-01
-7.17237353e-01 -4.15209174e-01 -1.68246612e-01 5.97410321e-01
-1.43181548e-01 1.16875529e+00 -7.01080933e-02 -3.42206955e-01
7.52621293e-01 6.23224005e-02 4.20868322e-02 2.23086074e-01
-1.09732354e+00 4.33107257e-01 1.60266325e-01 -5.81181824e-01
-3.61634701e-01 -5.03690958e-01 -9.56176996e-01 -2.34067231e-01
-4.95528221e-01 1.57599583e-01 1.01700175e+00 6.78585708e-01
2.56965458e-01 9.08607841e-01 6.58226848e-01 -9.94725645e-01
-6.08403623e-01 -1.14666069e+00 -8.60662997e-01 1.58159286e-01
7.06289113e-01 -1.04154609e-01 -4.76591259e-01 -4.58294034e-01]
|
[15.050735473632812, 5.930543422698975]
|
054d3260-562f-4c2a-8e64-3ced295fec4e
|
scene-coordinate-regression-with-angle-based
|
1808.04999
| null |
http://arxiv.org/abs/1808.04999v2
|
http://arxiv.org/pdf/1808.04999v2.pdf
|
Scene Coordinate Regression with Angle-Based Reprojection Loss for Camera Relocalization
|
Image-based camera relocalization is an important problem in computer vision
and robotics. Recent works utilize convolutional neural networks (CNNs) to
regress for pixels in a query image their corresponding 3D world coordinates in
the scene. The final pose is then solved via a RANSAC-based optimization scheme
using the predicted coordinates. Usually, the CNN is trained with ground truth
scene coordinates, but it has also been shown that the network can discover 3D
scene geometry automatically by minimizing single-view reprojection loss.
However, due to the deficiencies of the reprojection loss, the network needs to
be carefully initialized. In this paper, we present a new angle-based
reprojection loss, which resolves the issues of the original reprojection loss.
With this new loss function, the network can be trained without careful
initialization, and the system achieves more accurate results. The new loss
also enables us to utilize available multi-view constraints, which further
improve performance.
|
['Juho Kannala', 'Xiaotian Li', 'Juha Ylioinas', 'Jakob Verbeek']
|
2018-08-15
| null | null | null | null |
['camera-relocalization']
|
['computer-vision']
|
[-7.40239546e-02 -1.94846153e-01 -2.03956962e-01 -4.22682434e-01
-3.84613186e-01 -4.74923283e-01 1.99369252e-01 -4.01763678e-01
-6.70868576e-01 4.89465624e-01 -8.06817710e-02 5.13552837e-02
1.83516100e-01 -7.95985520e-01 -1.00635016e+00 -5.35851419e-01
7.61330605e-01 2.48413011e-01 8.87376666e-02 -8.07750002e-02
3.03455949e-01 8.05168390e-01 -1.18003035e+00 -3.56846660e-01
6.37497485e-01 9.69839096e-01 3.53004694e-01 4.23468441e-01
2.07174942e-01 4.36002821e-01 -6.08408563e-02 -3.32686543e-01
4.99035686e-01 -7.76470527e-02 -5.15628874e-01 3.11542332e-01
8.21272135e-01 -6.57077551e-01 -5.50265431e-01 1.25633240e+00
2.91533321e-01 4.15020697e-02 2.78746575e-01 -9.82037306e-01
-4.31198508e-01 -1.01936363e-01 -7.25832224e-01 -2.50532120e-01
3.46605450e-01 -1.94422558e-01 1.04864991e+00 -1.10290849e+00
5.90265274e-01 1.07686281e+00 7.95749187e-01 3.07455808e-01
-1.22375906e+00 -5.63748837e-01 1.16425551e-01 -5.57408249e-03
-1.58765709e+00 -8.47690776e-02 1.11995256e+00 -3.54958922e-01
7.72569597e-01 -6.53740391e-02 7.91317523e-01 8.18170488e-01
2.80572027e-01 5.92433512e-01 7.40410149e-01 -2.72454262e-01
-3.45302597e-02 6.91134408e-02 -3.01316857e-01 9.23439920e-01
1.99840650e-01 1.23567820e-01 -2.19200492e-01 2.51527876e-01
1.32431853e+00 4.71505433e-01 -4.81679767e-01 -1.21169806e+00
-1.31397700e+00 9.36486542e-01 1.02788401e+00 4.45012003e-03
-2.62779713e-01 2.37026393e-01 7.83301890e-02 -1.91471115e-01
4.42662984e-01 7.93013513e-01 -4.27470833e-01 1.55311942e-01
-7.36161411e-01 -5.33154197e-02 5.38101673e-01 9.04095650e-01
1.15671921e+00 1.31290674e-01 5.47199309e-01 9.00153935e-01
4.11436588e-01 6.07519567e-01 1.16013862e-01 -1.26061213e+00
5.16385376e-01 6.55325711e-01 1.10144831e-01 -1.39769971e+00
-5.07885933e-01 -5.38375616e-01 -9.28059280e-01 2.60520756e-01
3.03525597e-01 -8.30967426e-02 -8.26242447e-01 1.48840237e+00
2.97700197e-01 7.83702508e-02 -3.73188257e-02 1.36383617e+00
4.56227630e-01 5.37121177e-01 -4.40312713e-01 1.68184236e-01
9.21212852e-01 -1.10283244e+00 -4.76361752e-01 -3.70054603e-01
3.65168124e-01 -8.47112715e-01 8.32440376e-01 3.58601421e-01
-8.38790894e-01 -5.87185383e-01 -1.25042629e+00 -3.67976636e-01
-7.83036053e-02 2.61693716e-01 6.27343416e-01 2.38087058e-01
-1.02266502e+00 3.49195987e-01 -7.82649994e-01 -4.38759208e-01
2.07471535e-01 3.29050511e-01 -5.74952602e-01 -2.13206008e-01
-9.69745398e-01 1.01615071e+00 4.04587120e-01 2.38328710e-01
-6.59014761e-01 -5.01121581e-01 -1.19698346e+00 3.12529542e-02
4.21455890e-01 -8.69509995e-01 9.81905699e-01 -9.19717133e-01
-1.66680038e+00 8.20988357e-01 -1.28287211e-01 -1.19853772e-01
5.42064250e-01 -4.80087429e-01 1.72201153e-02 8.00936967e-02
2.47686431e-01 6.96160793e-01 7.79190540e-01 -1.50689566e+00
-4.83528137e-01 -5.42947114e-01 3.90168935e-01 5.54755151e-01
-1.70313284e-01 -4.95560080e-01 -9.42095757e-01 -5.03812969e-01
7.93743670e-01 -1.15328252e+00 -5.07965147e-01 3.09553504e-01
-4.54678446e-01 1.80111006e-01 6.21964514e-01 -4.15226728e-01
5.87442398e-01 -2.20496774e+00 2.78095424e-01 1.15593493e-01
1.96104929e-01 -5.35660945e-02 -1.26507461e-01 1.05414666e-01
-1.60052970e-01 -2.03412786e-01 -1.04034223e-01 -4.67087597e-01
-5.25236845e-01 3.04563701e-01 -3.34000736e-01 6.75825775e-01
9.55766961e-02 7.61529744e-01 -8.68244469e-01 -2.17486367e-01
6.21581078e-01 6.46859825e-01 -7.15008914e-01 2.43919656e-01
-3.54223587e-02 6.49079144e-01 -4.25920755e-01 2.92127043e-01
8.44273269e-01 -3.05427581e-01 -1.04347177e-01 -6.15153491e-01
-1.90589711e-01 -1.18570805e-01 -1.04407632e+00 2.19290519e+00
-7.98342228e-01 6.79664671e-01 8.05239975e-02 -9.59174633e-01
1.07191741e+00 1.82956997e-02 6.64240181e-01 -4.30548280e-01
2.39418671e-01 5.79908825e-02 -4.98744518e-01 -2.78552324e-01
7.01461494e-01 -1.64533162e-03 1.66452572e-01 1.04290582e-01
-6.98693469e-02 -4.59807426e-01 -2.96076804e-01 -9.25221369e-02
4.17141706e-01 4.36076641e-01 3.29898506e-01 1.34289697e-01
6.20313704e-01 1.67629600e-01 7.43365943e-01 2.61012167e-01
2.18417868e-01 1.06305218e+00 1.52479857e-01 -8.04459572e-01
-1.30716169e+00 -1.02552152e+00 -1.23610847e-01 5.26536644e-01
7.16006100e-01 -5.30272312e-02 -4.83968496e-01 -6.74572647e-01
-1.09301187e-01 3.44476074e-01 -2.00690702e-01 -8.58646333e-02
-7.60244668e-01 -3.55337709e-01 3.73678841e-02 6.16566956e-01
7.76523769e-01 -4.90334183e-01 -5.67775071e-01 9.91300941e-02
-3.60277742e-01 -1.26970243e+00 -5.99764109e-01 -2.37778458e-03
-1.05745995e+00 -1.13193822e+00 -7.89799631e-01 -7.20757008e-01
1.09454155e+00 7.27704346e-01 7.84134328e-01 -4.26642597e-02
-9.04406309e-02 3.18319052e-01 -1.92171618e-01 -1.49403708e-02
2.42929067e-02 1.00342602e-01 1.13239981e-01 6.29644021e-02
1.91060528e-01 -5.69439232e-01 -5.18207371e-01 4.74494696e-01
-6.23814166e-01 2.59574860e-01 5.00689507e-01 1.00484884e+00
9.03397262e-01 -2.77556658e-01 -1.91407904e-04 -7.04938710e-01
6.58734962e-02 -5.86333452e-03 -1.14459479e+00 1.14115678e-01
-6.45081699e-01 1.31720349e-01 6.64610088e-01 -2.20104039e-01
-7.98028171e-01 7.37758398e-01 -1.78218439e-01 -9.98173058e-01
-9.38774943e-02 4.04487729e-01 -1.81003407e-01 -4.36015874e-01
4.20913696e-01 2.06166223e-01 -4.14215639e-04 -4.70402837e-01
4.54591483e-01 3.39189440e-01 6.04160190e-01 -1.65894344e-01
8.17767680e-01 7.25430191e-01 1.05798006e-01 -5.54114640e-01
-1.27675688e+00 -6.06678069e-01 -9.72949088e-01 -1.49940789e-01
1.06283414e+00 -1.30125225e+00 -7.89054930e-01 3.08250993e-01
-1.46722758e+00 8.65100026e-02 -4.89190221e-02 9.68214095e-01
-6.76131248e-01 5.67983091e-01 -3.23324710e-01 -3.45880866e-01
-1.22919194e-01 -1.38807571e+00 1.05508888e+00 2.73497939e-01
1.57374486e-01 -8.73233616e-01 -4.22845893e-02 3.04357558e-01
6.80169687e-02 1.92631960e-01 5.34874022e-01 -6.84242770e-02
-8.36575985e-01 -3.04391474e-01 -4.80094820e-01 4.68861371e-01
6.11978211e-02 -4.15450335e-02 -8.82697403e-01 -4.50005621e-01
2.22839668e-01 -1.53493449e-01 7.92135894e-01 4.81417567e-01
1.23338819e+00 9.91283432e-02 -1.85880706e-01 1.29358196e+00
1.65373051e+00 9.07547697e-02 3.45353693e-01 5.57384729e-01
1.05205321e+00 2.87870497e-01 5.09313166e-01 3.10963660e-01
6.31599724e-01 7.15768397e-01 9.07440484e-01 -3.52444321e-01
1.77567288e-01 -5.07578850e-01 6.40383065e-02 7.87533939e-01
5.45714349e-02 -1.09001666e-01 -8.04415166e-01 3.43321562e-01
-1.86382616e+00 -5.72299719e-01 -1.17058782e-02 2.19846487e+00
3.65056306e-01 -1.38983592e-01 -4.98967558e-01 -1.80339456e-01
7.54625797e-01 2.75918573e-01 -9.85526741e-01 -1.05817758e-01
2.36365441e-02 -3.17742050e-01 9.13862944e-01 7.14782655e-01
-1.27352035e+00 1.06348062e+00 6.12779093e+00 3.26162219e-01
-1.51874626e+00 -1.72085151e-01 4.24762636e-01 1.42665640e-01
-1.57176435e-01 7.36627653e-02 -8.00554752e-01 5.19078374e-02
4.49462682e-02 2.31738567e-01 5.68454504e-01 1.11169612e+00
7.13954493e-02 -2.50414103e-01 -1.16431510e+00 1.37735951e+00
4.91955996e-01 -1.27693939e+00 -3.87438945e-03 1.26515865e-01
9.12404597e-01 3.08199018e-01 5.15393019e-02 -1.26161411e-01
1.91015005e-01 -7.01543033e-01 6.45931423e-01 4.64769691e-01
8.24118316e-01 -7.81823397e-01 7.07968235e-01 4.05063242e-01
-9.93401587e-01 -6.37453794e-03 -6.84930801e-01 -4.28169109e-02
2.91963637e-01 6.76781833e-01 -8.38998735e-01 6.72919452e-01
6.88791811e-01 1.09433424e+00 -3.96894485e-01 1.09558880e+00
-4.13532704e-01 -8.82846415e-02 -3.55951339e-01 1.70053050e-01
1.94951653e-01 -6.34887636e-01 4.81262445e-01 4.55354840e-01
4.76391882e-01 -2.02769622e-01 3.88301969e-01 9.85026896e-01
-1.15345277e-01 -1.60866395e-01 -8.33025157e-01 4.16352838e-01
2.86217868e-01 1.33678257e+00 -5.88269889e-01 2.42760126e-02
-5.05076766e-01 1.09293091e+00 5.38924158e-01 4.85033363e-01
-6.15926325e-01 -1.81138203e-01 5.98170161e-01 -5.34760468e-02
1.09472401e-01 -5.88223815e-01 -3.67259890e-01 -1.56571150e+00
8.32813233e-03 -3.56394529e-01 -1.96888089e-01 -1.01749039e+00
-1.16940606e+00 5.39146602e-01 -2.75361300e-01 -1.48405230e+00
-1.28880486e-01 -6.03489697e-01 -2.70463407e-01 8.20063889e-01
-1.67304575e+00 -9.88597393e-01 -4.90919322e-01 5.26149988e-01
4.73806292e-01 -2.24126969e-02 5.37004292e-01 2.39267945e-01
-5.06374955e-01 3.34964901e-01 1.39528856e-01 3.18014473e-01
8.46712589e-01 -1.22664797e+00 2.71965593e-01 7.31113851e-01
1.14280403e-01 6.23955429e-01 4.45053220e-01 -3.04985046e-01
-1.41379988e+00 -1.29047394e+00 7.08227694e-01 -3.76257777e-01
3.40770572e-01 -3.36644083e-01 -6.53215468e-01 8.05894077e-01
-2.59794798e-02 2.42705300e-01 2.42194474e-01 -3.49190049e-02
-3.36740375e-01 -3.28927428e-01 -8.97699594e-01 5.50085068e-01
8.50533664e-01 -5.89754879e-01 -2.77908146e-01 2.93662846e-01
8.08333874e-01 -8.10179412e-01 -7.88288593e-01 4.22843575e-01
5.83442450e-01 -9.90160763e-01 1.14805973e+00 -1.72052160e-01
3.67809147e-01 -5.18528342e-01 -1.24263406e-01 -1.33688653e+00
-1.40927181e-01 -7.34221488e-02 4.11904335e-01 7.91541576e-01
3.46728176e-01 -7.19980061e-01 9.73565400e-01 4.45691496e-01
-1.47776708e-01 -7.61221588e-01 -8.98526669e-01 -5.90495944e-01
-1.71303451e-01 -2.01545238e-01 4.36672360e-01 9.25355852e-01
-5.07927179e-01 4.56138968e-01 -5.82111061e-01 5.15415668e-01
5.04868150e-01 3.00911814e-01 9.50472653e-01 -1.21726131e+00
7.88867027e-02 -5.03017008e-02 -4.29628372e-01 -1.58387268e+00
1.98762104e-01 -7.71354496e-01 3.24122667e-01 -1.49460983e+00
1.51046887e-01 -4.18855518e-01 -5.75876189e-03 2.26778939e-01
8.40345249e-02 3.60441536e-01 2.31808439e-01 2.68292040e-01
-4.02289629e-01 7.57039845e-01 1.61934626e+00 -2.94475332e-02
-2.44802475e-01 8.39565769e-02 -3.35897207e-01 1.03713429e+00
8.24036598e-01 -4.00661319e-01 -3.45302910e-01 -9.61590290e-01
3.69123906e-01 1.03423022e-01 5.58514059e-01 -9.58072662e-01
4.50094610e-01 -1.96835995e-01 5.82118809e-01 -8.59165192e-01
7.46919096e-01 -1.18513918e+00 4.94332872e-02 4.01662916e-01
-9.70979333e-02 2.42633939e-01 -2.12836921e-01 6.08066618e-01
-4.13163573e-01 -1.43655777e-01 9.41404045e-01 -1.81695417e-01
-7.05940723e-01 6.17637932e-01 5.37627190e-02 -3.26869696e-01
1.00507355e+00 -3.38957280e-01 1.59241796e-01 -4.86282617e-01
-5.46297491e-01 2.21343592e-01 9.17572677e-01 4.75664228e-01
8.63775551e-01 -1.43469572e+00 -2.28464767e-01 4.50008601e-01
1.42852768e-01 5.93658507e-01 9.86776426e-02 7.67109811e-01
-9.66120899e-01 3.94874901e-01 -1.52633756e-01 -9.18866456e-01
-9.35554206e-01 5.80680609e-01 4.70126837e-01 1.79561600e-03
-5.98798096e-01 7.03647196e-01 4.25735712e-01 -8.71660709e-01
2.00799659e-01 -2.50472248e-01 -1.98140457e-01 -3.01377803e-01
6.81221262e-02 1.12645693e-01 -1.71598196e-01 -8.79686058e-01
-1.61678597e-01 1.25998342e+00 1.04570622e-02 -9.15761758e-03
1.33514404e+00 -3.16172153e-01 -1.92075089e-01 1.64899364e-01
1.48793280e+00 -1.41134024e-01 -1.59275377e+00 -4.14796025e-01
-2.74769813e-01 -7.09219217e-01 2.70099491e-01 -1.93637446e-01
-1.36127567e+00 1.08047938e+00 5.44988573e-01 -3.11601490e-01
9.02394474e-01 -3.21612805e-01 8.19375336e-01 6.62489116e-01
4.73116547e-01 -9.52578723e-01 2.39182293e-01 8.79353344e-01
9.00199413e-01 -1.36440086e+00 1.24841563e-01 -3.89224648e-01
-5.07838786e-01 1.34992647e+00 9.13599312e-01 -4.11394596e-01
6.64636374e-01 -8.53790045e-02 2.49525592e-01 -7.87179619e-02
2.43730713e-02 -1.74056347e-02 1.71471313e-01 3.52256060e-01
1.06124453e-01 -1.92318380e-01 3.74925248e-02 6.02227927e-04
-2.07527772e-01 -2.02953100e-01 5.31353414e-01 4.65439796e-01
-3.23529005e-01 -9.19967830e-01 -3.59364837e-01 8.77808630e-02
-1.00263603e-01 5.71258180e-02 -2.15102062e-01 7.26773560e-01
5.72001822e-02 7.06386387e-01 2.04833388e-01 -4.23063427e-01
4.66125101e-01 -4.56426293e-01 3.09625238e-01 -6.57124937e-01
-4.76657152e-02 2.21484303e-01 -4.33253497e-01 -8.62732887e-01
-5.73632717e-01 -2.87008435e-01 -1.13539541e+00 -6.57225102e-02
-7.03991711e-01 -7.13252202e-02 9.02539551e-01 7.11306274e-01
3.22832316e-01 2.39988983e-01 1.02722681e+00 -1.05056381e+00
-3.59325290e-01 -5.63897431e-01 -4.75689381e-01 1.74092963e-01
5.69566667e-01 -6.61681414e-01 -1.98195338e-01 -3.23041081e-02]
|
[7.952687740325928, -2.2922425270080566]
|
edd97948-735e-47a6-aee3-308a8d66a5e3
|
mul-gad-a-semi-supervised-graph-anomaly
|
2212.05478
| null |
https://arxiv.org/abs/2212.05478v1
|
https://arxiv.org/pdf/2212.05478v1.pdf
|
Mul-GAD: a semi-supervised graph anomaly detection framework via aggregating multi-view information
|
Anomaly detection is defined as discovering patterns that do not conform to the expected behavior. Previously, anomaly detection was mostly conducted using traditional shallow learning techniques, but with little improvement. As the emergence of graph neural networks (GNN), graph anomaly detection has been greatly developed. However, recent studies have shown that GNN-based methods encounter challenge, in that no graph anomaly detection algorithm can perform generalization on most datasets. To bridge the tap, we propose a multi-view fusion approach for graph anomaly detection (Mul-GAD). The view-level fusion captures the extent of significance between different views, while the feature-level fusion makes full use of complementary information. We theoretically and experimentally elaborate the effectiveness of the fusion strategies. For a more comprehensive conclusion, we further investigate the effect of the objective function and the number of fused views on detection performance. Exploiting these findings, our Mul-GAD is proposed equipped with fusion strategies and the well-performed objective function. Compared with other state-of-the-art detection methods, we achieve a better detection performance and generalization in most scenarios via a series of experiments conducted on Pubmed, Amazon Computer, Amazon Photo, Weibo and Books. Our code is available at https://github.com/liuyishoua/Mul-Graph-Fusion.
|
['Jingzhang Sun', 'Chunjie Cao', 'Zhiyuan Liu']
|
2022-12-11
| null | null | null | null |
['graph-anomaly-detection']
|
['graphs']
|
[-1.65324524e-01 -1.51498273e-01 -1.28336370e-01 -2.58731134e-02
-1.46019623e-01 -4.45739925e-01 7.52324998e-01 7.23992467e-01
-1.04832217e-01 2.84246504e-01 -8.21027905e-02 -2.03498632e-01
-2.80174792e-01 -1.11886859e+00 -4.07451779e-01 -5.86918175e-01
-3.26277316e-01 1.72185391e-01 4.08427328e-01 -2.57176369e-01
1.77823365e-01 6.36251986e-01 -1.53620124e+00 -5.06578274e-02
9.60257947e-01 1.06323147e+00 -4.47216153e-01 4.50238496e-01
-2.64378160e-01 5.38808703e-01 -4.99481767e-01 -6.64388716e-01
3.86234969e-01 -3.21128577e-01 -3.66017848e-01 4.72158007e-02
3.87162864e-01 -1.92145303e-01 -6.53978825e-01 1.39450848e+00
4.71820116e-01 -4.78114281e-03 5.47761321e-01 -1.64660203e+00
-6.44536674e-01 4.75052893e-01 -8.46801460e-01 7.36758828e-01
5.60414732e-01 -1.14415511e-02 1.06623816e+00 -8.57467711e-01
4.69833940e-01 1.01222181e+00 5.45053482e-01 2.28851467e-01
-8.97530317e-01 -5.49229383e-01 4.31789190e-01 5.29670119e-01
-1.35592496e+00 4.09405269e-02 9.19471860e-01 -2.26155177e-01
7.73382902e-01 2.43149787e-01 7.50304282e-01 1.19068921e+00
3.44107181e-01 7.90756643e-01 8.51695955e-01 -3.99293154e-01
1.10184930e-01 -9.62866172e-02 2.54711986e-01 9.96574104e-01
8.91474545e-01 -9.54062566e-02 -4.69856173e-01 -3.74168247e-01
4.07787919e-01 4.71446663e-01 -2.10620642e-01 -4.53469217e-01
-7.71108866e-01 8.17598164e-01 4.61829394e-01 5.91811955e-01
-5.60910821e-01 -1.86250642e-01 6.90587044e-01 4.13169891e-01
6.84943676e-01 2.22254574e-01 4.80796024e-02 1.82432875e-01
-7.92418718e-01 -1.90191567e-02 8.81102145e-01 6.17605627e-01
6.39065981e-01 2.30128706e-01 -1.72388554e-01 5.59756875e-01
2.76028544e-01 1.51294142e-01 5.04465938e-01 -3.06095600e-01
2.40321904e-01 1.32298005e+00 -4.72942770e-01 -1.47573543e+00
-4.70340073e-01 -6.90689921e-01 -1.07138884e+00 4.37217876e-02
4.50808614e-01 -4.46708687e-02 -8.93335938e-01 1.44790041e+00
4.61489588e-01 3.80996078e-01 -1.07164174e-01 6.07372046e-01
7.17485309e-01 1.68427199e-01 -1.59806132e-01 -1.62360128e-02
1.36080229e+00 -8.46209347e-01 -8.23490918e-01 -1.07376210e-01
8.86480272e-01 -5.07142246e-01 7.84866035e-01 5.03130138e-01
-6.51020348e-01 -1.53216466e-01 -1.28548717e+00 4.62658405e-01
-8.51258934e-01 -1.05866656e-01 8.04700971e-01 8.40038300e-01
-9.76786137e-01 5.67697525e-01 -8.05945158e-01 -6.97770894e-01
5.19326508e-01 9.10918340e-02 -5.93173265e-01 -2.93398529e-01
-1.10521150e+00 6.67813838e-01 6.34570599e-01 -3.10667902e-02
-6.17181420e-01 -3.26088101e-01 -7.74663806e-01 8.60851854e-02
8.70283425e-01 -6.26004696e-01 7.50995934e-01 -8.27118337e-01
-8.32521915e-01 6.67848110e-01 1.42687798e-01 -7.39646375e-01
4.17300999e-01 -1.42543584e-01 -7.96622336e-01 1.99288160e-01
3.79724288e-03 -3.02640703e-02 7.39181817e-01 -1.07988036e+00
-5.69076240e-01 -8.26640666e-01 1.54585332e-01 4.16471362e-02
-5.92009783e-01 -2.52061665e-01 -4.55055565e-01 -7.02848554e-01
3.55182707e-01 -6.82424545e-01 -1.07979886e-01 -2.62738168e-01
-5.54626584e-01 -4.85064685e-01 1.11986268e+00 -5.11519074e-01
1.63683856e+00 -2.23814678e+00 -2.49926522e-02 6.43985271e-01
7.71008849e-01 4.15751606e-01 1.01019837e-01 7.35499799e-01
-4.08048518e-02 1.33506611e-01 -2.40925804e-01 -3.48517120e-01
-1.80784106e-01 3.07074100e-01 9.45856422e-02 6.42120659e-01
-1.22487366e-01 7.59740412e-01 -8.92837107e-01 -3.35384727e-01
1.56960621e-01 1.40073225e-01 -3.81434917e-01 1.54695466e-01
4.85853069e-02 1.45357385e-01 -4.18030769e-01 1.20913196e+00
5.78849375e-01 -4.14310485e-01 1.11156851e-01 -8.23298469e-02
1.96712464e-01 -3.91815007e-01 -1.25636363e+00 1.52666926e+00
5.06235100e-02 2.65947342e-01 -2.81286202e-02 -1.21704078e+00
9.78305340e-01 9.98825878e-02 4.92245197e-01 -7.18050241e-01
1.72997162e-01 4.34341192e-01 2.84706533e-01 -3.27105373e-01
3.61419439e-01 3.88655633e-01 3.13446447e-02 3.80819261e-01
2.79312849e-01 4.47846264e-01 4.60895866e-01 5.98577201e-01
1.41822624e+00 -3.49835306e-01 7.35031605e-01 -1.03105016e-01
8.13618004e-01 -2.59967506e-01 2.38678098e-01 1.01930511e+00
-3.39287132e-01 4.04880136e-01 7.94954717e-01 -5.36744297e-01
-6.26631141e-01 -9.55798566e-01 1.80569351e-01 1.02133501e+00
2.04131216e-01 -8.56909394e-01 -8.15022945e-01 -1.18279672e+00
2.03909382e-01 6.09686732e-01 -7.22626328e-01 -4.14742708e-01
-2.88818717e-01 -8.39336455e-01 7.23983526e-01 3.10673594e-01
6.69248343e-01 -9.60177004e-01 -2.76105821e-01 -6.03529066e-02
1.80712923e-01 -1.18316054e+00 -2.18961284e-01 -8.55995864e-02
-8.21685493e-01 -1.30270493e+00 -2.53033847e-01 -2.04182312e-01
6.15166426e-01 3.62508267e-01 1.06668866e+00 6.24702215e-01
-3.89048040e-01 8.64726424e-01 -6.85741246e-01 -5.47151029e-01
-2.69871861e-01 1.77052841e-01 2.08297491e-01 3.42799425e-01
5.98524570e-01 -9.32617426e-01 -6.19432509e-01 1.46622568e-01
-1.11331737e+00 -5.15928686e-01 7.34120846e-01 5.88469028e-01
4.14602280e-01 1.00354619e-01 4.87146109e-01 -1.02346671e+00
9.44966912e-01 -7.90170550e-01 -5.13137043e-01 2.28327543e-01
-1.04174483e+00 -7.82571360e-02 4.86822516e-01 -1.58229411e-01
-5.29118359e-01 -3.32071275e-01 -5.65302037e-02 -6.35145307e-01
-3.16674352e-01 7.49354899e-01 -5.49486764e-02 -2.47407585e-01
7.02449262e-01 4.08276618e-01 5.64835742e-02 -3.30649167e-01
1.87599868e-01 3.28080714e-01 3.61207634e-01 -3.06036115e-01
7.30350733e-01 4.65644658e-01 1.22695088e-01 -8.91912043e-01
-6.14797413e-01 -6.06101096e-01 -4.97850060e-01 -3.36967379e-01
6.16942108e-01 -5.68705022e-01 -5.40431440e-01 6.23531103e-01
-7.43225455e-01 2.91289330e-01 -1.24363020e-01 2.55977005e-01
2.07598470e-02 8.82851064e-01 -2.69831657e-01 -7.86682963e-01
-4.67614561e-01 -7.86518455e-01 6.57591164e-01 1.70359969e-01
9.32527855e-02 -1.18679857e+00 1.40854031e-01 4.13777269e-02
5.50437510e-01 4.96617973e-01 8.34812224e-01 -1.52229440e+00
-3.99244964e-01 -6.86692059e-01 -3.36450428e-01 2.42521286e-01
2.18850166e-01 -4.87427339e-02 -7.35282481e-01 -4.78646547e-01
-6.67049512e-02 1.65176049e-01 9.22915697e-01 1.45874575e-01
1.32804620e+00 -2.46559694e-01 -3.94125789e-01 4.89660025e-01
1.46553814e+00 1.21226355e-01 5.15239894e-01 4.47281182e-01
1.04649997e+00 2.88112640e-01 2.86838204e-01 5.39264858e-01
4.00917530e-01 4.64596659e-01 8.91168535e-01 5.25534153e-02
1.22125618e-01 -1.70827717e-01 2.72669286e-01 6.41147733e-01
-2.89504588e-01 -6.52933240e-01 -1.06520259e+00 3.36188197e-01
-1.98453438e+00 -8.35285783e-01 -7.52738863e-02 2.21347022e+00
-1.08874179e-01 4.49794769e-01 4.50794131e-01 3.07195812e-01
7.77646899e-01 3.04454535e-01 -5.44351339e-01 -2.99062520e-01
-1.51748508e-01 -1.49781629e-01 4.17821854e-01 1.38296112e-01
-1.10876417e+00 6.26638770e-01 5.52408028e+00 9.89204824e-01
-1.00847352e+00 1.10844716e-01 3.98500115e-01 1.69883505e-01
-2.40536764e-01 -1.17990427e-01 -4.78218824e-01 4.47518080e-01
7.81111896e-01 -2.72140920e-01 3.00360918e-01 8.63906801e-01
-2.51104414e-01 -2.15247851e-02 -9.63241398e-01 9.55019772e-01
3.94741029e-01 -1.07204270e+00 3.22611928e-01 3.22805971e-01
4.05933082e-01 2.28330433e-01 -1.25612140e-01 3.80477041e-01
-4.82675880e-02 -8.77367198e-01 1.73228398e-01 4.52661425e-01
9.01848152e-02 -7.11523294e-01 1.04840016e+00 3.32377017e-01
-1.33531058e+00 -3.38903964e-01 -5.92492297e-02 7.47323558e-02
-1.47671580e-01 8.55929732e-01 -5.77132881e-01 1.28096282e+00
6.55267656e-01 7.91305423e-01 -1.04164445e+00 1.05611110e+00
-1.69321924e-01 5.68217635e-01 -3.76007378e-01 6.08291477e-02
2.78568268e-01 -3.60539436e-01 9.15035367e-01 1.10543764e+00
5.26091516e-01 -9.86016691e-02 4.14387375e-01 5.73885024e-01
-8.47366899e-02 3.63647223e-01 -1.04221451e+00 -2.14144379e-01
4.05700445e-01 1.44979978e+00 -1.00628042e+00 -2.80613929e-01
-7.80337334e-01 8.34166646e-01 3.89756262e-01 2.54942030e-01
-6.57220125e-01 -1.78986639e-01 2.81832308e-01 2.37947926e-01
1.97775155e-01 -1.15911447e-01 -2.18173005e-02 -1.26686263e+00
2.44914100e-01 -9.70185578e-01 9.80700195e-01 -8.04951116e-02
-1.60142183e+00 6.50746465e-01 9.73312557e-02 -1.23102307e+00
-7.05104470e-02 -6.96200132e-01 -8.68769169e-01 3.68941993e-01
-1.13795996e+00 -1.07332206e+00 -4.96042609e-01 6.91366076e-01
2.91203529e-01 -5.79405844e-01 7.22895086e-01 4.00401562e-01
-9.29458201e-01 7.99236715e-01 -1.49085552e-01 1.98320493e-01
4.62735802e-01 -1.34909534e+00 3.37894291e-01 1.26821589e+00
4.07473773e-01 3.72853070e-01 6.11835897e-01 -7.39651918e-01
-1.31485558e+00 -7.43918598e-01 2.23920718e-01 -3.24986666e-01
7.71283090e-01 -3.24806303e-01 -1.29463816e+00 6.29773974e-01
1.60039529e-01 4.22540635e-01 6.51799560e-01 3.03785503e-01
-4.89536226e-01 -8.14954564e-02 -1.16916072e+00 4.66466278e-01
1.19261539e+00 -4.13077533e-01 -2.59781510e-01 1.45593747e-01
4.06421185e-01 -3.00726056e-01 -8.01048279e-01 6.39842212e-01
3.29066902e-01 -1.39384747e+00 8.04916799e-01 -4.62797672e-01
6.28775656e-02 -1.45765722e-01 -2.83917606e-01 -1.24741685e+00
-2.64863104e-01 -2.43140832e-01 -7.43240416e-01 1.10201514e+00
2.58431882e-01 -1.11325121e+00 7.99453139e-01 1.35035008e-01
-1.06603540e-01 -1.08607018e+00 -1.00618708e+00 -8.34528863e-01
-4.30897564e-01 -2.96838462e-01 5.64382672e-01 1.04452014e+00
-1.32227689e-01 6.65328726e-02 -2.91957170e-01 3.59719187e-01
5.56577504e-01 -8.58828425e-02 6.75069213e-01 -1.60945404e+00
-1.21862128e-01 -6.47771418e-01 -9.98432636e-01 -3.04782331e-01
-4.36087474e-02 -9.80947614e-01 -6.75829589e-01 -1.39576089e+00
1.31453663e-01 1.07437707e-01 -6.23847961e-01 4.57981557e-01
-3.08560461e-01 1.03829652e-01 -6.15295116e-03 2.13215962e-01
-8.64648283e-01 4.55933273e-01 8.71695876e-01 8.27055722e-02
-1.17683470e-01 6.20698333e-02 -6.87539160e-01 9.30763245e-01
8.87434125e-01 -3.56182575e-01 -3.89732957e-01 2.85137612e-02
2.80474335e-01 -2.26204470e-01 3.07913601e-01 -1.20395696e+00
5.15343726e-01 3.13110203e-01 2.75313526e-01 -5.56063533e-01
2.39440929e-02 -7.88795173e-01 -1.62874773e-01 6.63847625e-01
1.70612469e-01 6.87827289e-01 1.88489154e-01 1.05670989e+00
-3.27697456e-01 -8.76719058e-02 3.61784607e-01 -1.57324687e-01
-9.21479225e-01 5.69702864e-01 -2.88168162e-01 -5.03143817e-02
1.15761590e+00 -3.08261663e-01 -3.35662961e-01 -4.59813923e-01
-6.94576859e-01 2.45027736e-01 3.14809501e-01 5.55279493e-01
5.93706071e-01 -1.35241532e+00 -6.87387466e-01 3.09206694e-01
4.63188589e-01 -3.29713136e-01 3.52821261e-01 1.03159142e+00
-4.13545847e-01 9.00941789e-02 -2.52090424e-01 -6.64460242e-01
-1.17578375e+00 7.58940458e-01 3.88593525e-01 -5.50303221e-01
-6.24939203e-01 4.62284535e-01 1.66191239e-04 -2.83959299e-01
1.96680143e-01 7.30682835e-02 -3.96948010e-01 2.61602044e-01
3.39156449e-01 6.09126866e-01 2.92491585e-01 -4.25251633e-01
-4.66094941e-01 2.12663144e-01 -3.51570308e-01 3.64643574e-01
1.10468435e+00 -5.93611933e-02 -1.98793054e-01 3.37032914e-01
7.03479528e-01 1.15284115e-01 -5.26879072e-01 -3.18248630e-01
1.87220037e-01 -4.57146049e-01 -1.30013242e-01 -4.85675514e-01
-1.23056901e+00 5.46553373e-01 8.03379595e-01 8.66709232e-01
1.21737504e+00 -2.92073172e-02 4.29236233e-01 3.31776619e-01
1.85851619e-01 -7.10479736e-01 3.28901261e-01 3.49121511e-01
6.53053641e-01 -1.47021663e+00 1.42150164e-01 -5.79245269e-01
-3.71949047e-01 1.00990951e+00 9.39800262e-01 -3.00608516e-01
7.38451421e-01 7.63185602e-03 -2.01245591e-01 -7.82000363e-01
-4.46080834e-01 -2.93255568e-01 5.02589405e-01 3.97035807e-01
2.73098469e-01 -1.36879422e-02 -3.20947260e-01 4.87090498e-01
1.27074271e-01 -4.28130329e-01 4.08833385e-01 7.71691024e-01
-3.21224511e-01 -1.10159636e+00 -1.88317150e-01 8.35654557e-01
-6.71900809e-01 1.35227796e-02 -6.08109474e-01 1.03245687e+00
3.43048084e-03 7.07808197e-01 -1.08712740e-01 -5.52983940e-01
5.35987258e-01 2.30296209e-01 2.17143878e-01 -4.82609868e-01
-4.52560008e-01 -2.25689471e-01 -5.05795665e-02 -7.65674889e-01
-2.31098980e-01 -4.31459188e-01 -8.90855968e-01 -4.99216229e-01
-4.29649502e-01 1.56576887e-01 3.02271962e-01 1.02480972e+00
5.08205056e-01 6.16354823e-01 4.06306386e-01 -5.17328143e-01
-4.36325401e-01 -9.27182674e-01 -7.88451374e-01 4.09868062e-01
1.81115314e-01 -8.06086838e-01 -6.75842524e-01 -8.89851749e-01]
|
[6.690584659576416, 5.826527118682861]
|
50d53c77-b637-4f09-8e32-5cbdf267d03a
|
why-so-pessimistic-estimating-uncertainties-1
|
2205.13703
| null |
https://arxiv.org/abs/2205.13703v1
|
https://arxiv.org/pdf/2205.13703v1.pdf
|
Why So Pessimistic? Estimating Uncertainties for Offline RL through Ensembles, and Why Their Independence Matters
|
Motivated by the success of ensembles for uncertainty estimation in supervised learning, we take a renewed look at how ensembles of $Q$-functions can be leveraged as the primary source of pessimism for offline reinforcement learning (RL). We begin by identifying a critical flaw in a popular algorithmic choice used by many ensemble-based RL algorithms, namely the use of shared pessimistic target values when computing each ensemble member's Bellman error. Through theoretical analyses and construction of examples in toy MDPs, we demonstrate that shared pessimistic targets can paradoxically lead to value estimates that are effectively optimistic. Given this result, we propose MSG, a practical offline RL algorithm that trains an ensemble of $Q$-functions with independently computed targets based on completely separate networks, and optimizes a policy with respect to the lower confidence bound of predicted action values. Our experiments on the popular D4RL and RL Unplugged offline RL benchmarks demonstrate that on challenging domains such as antmazes, MSG with deep ensembles surpasses highly well-tuned state-of-the-art methods by a wide margin. Additionally, through ablations on benchmarks domains, we verify the critical significance of using independently trained $Q$-functions, and study the role of ensemble size. Finally, as using separate networks per ensemble member can become computationally costly with larger neural network architectures, we investigate whether efficient ensemble approximations developed for supervised learning can be similarly effective, and demonstrate that they do not match the performance and robustness of MSG with separate networks, highlighting the need for new efforts into efficient uncertainty estimation directed at RL.
|
['Ofir Nachum', 'Shixiang Shane Gu', 'Seyed Kamyar Seyed Ghasemipour']
|
2022-05-27
| null | null | null | null |
['d4rl']
|
['robots']
|
[-1.30740389e-01 3.27629566e-01 -1.79104745e-01 -4.04128313e-01
-1.16158426e+00 -8.45654428e-01 5.29508233e-01 2.09284890e-02
-6.58562124e-01 1.19519353e+00 3.35900113e-02 -6.07539177e-01
-2.69235730e-01 -4.80270535e-01 -8.80209923e-01 -6.71502590e-01
-4.44528311e-01 4.58762795e-01 -2.43779257e-01 -2.57702738e-01
2.10420623e-01 2.58167028e-01 -1.44136202e+00 -1.31977245e-01
8.40505958e-01 1.16556835e+00 -3.55475873e-01 4.93701577e-01
4.29605633e-01 9.11417484e-01 -8.52364838e-01 -2.55833626e-01
3.05710465e-01 -3.85314643e-01 -5.61313212e-01 -3.60101432e-01
1.89703092e-01 -6.17858589e-01 -1.56459451e-01 9.46748435e-01
5.17952681e-01 4.65807438e-01 8.05207074e-01 -1.26359749e+00
-1.72788069e-01 1.05132174e+00 -2.81967074e-01 1.04235642e-01
-1.26819819e-01 6.44389212e-01 1.21700299e+00 -3.66693199e-01
1.32701054e-01 1.29062808e+00 7.41430581e-01 4.24865067e-01
-1.48034251e+00 -8.34758818e-01 3.67021233e-01 -1.52629316e-01
-1.15974057e+00 -5.35176814e-01 3.98714602e-01 -1.91337615e-01
1.12817180e+00 -1.15056053e-01 3.66370767e-01 1.23496830e+00
3.55873048e-01 7.08175540e-01 1.39245450e+00 -1.46448150e-01
5.98994136e-01 -9.08741280e-02 -2.47211158e-01 6.50048137e-01
2.47399226e-01 9.18872833e-01 -3.11865777e-01 -2.39674643e-01
5.03035188e-01 -4.81236458e-01 2.26234362e-04 -3.59551609e-01
-8.28498721e-01 1.06335056e+00 2.04738453e-01 -3.15682366e-02
-1.23234272e-01 7.72647023e-01 5.59561908e-01 4.57669318e-01
4.93426830e-01 9.32072103e-01 -6.21726930e-01 -4.94503647e-01
-7.62082100e-01 5.50158799e-01 8.76734734e-01 5.10555685e-01
4.48533684e-01 4.21541542e-01 -1.80722058e-01 4.12415057e-01
2.45543212e-01 5.22356510e-01 2.16828570e-01 -1.45012522e+00
3.65284860e-01 2.43625388e-01 6.02375329e-01 -6.76069498e-01
-4.57996309e-01 -7.30370760e-01 -3.33629876e-01 7.84122467e-01
7.59096205e-01 -7.08755136e-01 -6.61008358e-01 2.10428667e+00
1.15586661e-01 7.20410645e-02 2.86198199e-01 6.24143243e-01
-1.85196355e-01 3.31922859e-01 6.70529380e-02 -2.97126979e-01
7.92520702e-01 -5.63620985e-01 -2.36843884e-01 -3.67010981e-01
7.76565373e-01 -1.90012500e-01 9.85812306e-01 6.79882050e-01
-1.03204668e+00 -2.67324239e-01 -1.29137969e+00 5.33087432e-01
-6.58495501e-02 -8.06825086e-02 6.76228464e-01 6.77321672e-01
-7.89704144e-01 1.27871346e+00 -1.02215219e+00 2.78860360e-01
4.92656261e-01 4.69319642e-01 8.99513066e-02 3.29322398e-01
-1.22569394e+00 1.30941415e+00 5.17727554e-01 2.46085912e-01
-1.51594245e+00 -4.64342684e-01 -6.78324640e-01 2.02805459e-01
7.90366709e-01 -2.66685903e-01 1.69585633e+00 -1.11691284e+00
-1.76842380e+00 1.21203221e-01 3.57749999e-01 -9.87761497e-01
6.54461205e-01 -2.88897455e-01 -2.13766932e-01 -1.91790596e-01
-2.62768924e-01 5.05845368e-01 1.01028645e+00 -1.12374556e+00
-5.26404560e-01 -3.54314029e-01 1.51075989e-01 2.57173449e-01
1.53741568e-01 -3.69993329e-01 5.11260629e-01 -4.41667706e-01
-3.21503252e-01 -1.05224144e+00 -4.81631696e-01 -5.68654656e-01
-1.39254138e-01 -3.54580969e-01 3.24273080e-01 -3.21355492e-01
1.29924154e+00 -1.79537630e+00 8.26146454e-02 5.36006689e-01
2.86687668e-02 1.82785541e-02 -1.63509950e-01 3.42254877e-01
8.41677003e-03 1.68380171e-01 -2.38131195e-01 -1.96497023e-01
5.19301653e-01 2.54761249e-01 -5.71476221e-01 4.06774879e-01
9.15095583e-02 8.79298568e-01 -1.02443135e+00 -1.21136323e-01
2.47927886e-02 2.60583404e-02 -5.84222138e-01 3.34308371e-02
-7.33897746e-01 3.75107616e-01 -4.02122825e-01 2.70576179e-01
1.16684183e-01 -2.29090691e-01 6.19665444e-01 -3.18107428e-03
-1.14058992e-02 3.85178268e-01 -1.28238595e+00 1.36714602e+00
-5.28700650e-01 3.23129207e-01 -1.60698399e-01 -1.08765483e+00
7.24189162e-01 8.89855176e-02 3.21967632e-01 -4.51621085e-01
3.58022243e-01 3.97987455e-01 2.47848272e-01 4.91923802e-02
4.45700943e-01 -2.90089786e-01 -3.80728006e-01 8.81469727e-01
2.11142272e-01 -4.05661076e-01 -2.41825078e-03 -8.69242027e-02
1.12105703e+00 5.03699183e-01 1.32620156e-01 -3.96599233e-01
1.50324628e-01 -7.02001527e-02 5.22698820e-01 1.06384718e+00
-3.25671136e-01 6.06557634e-03 8.23310733e-01 -3.21065128e-01
-9.70684230e-01 -1.16624439e+00 2.85984203e-02 1.20701122e+00
-1.63624182e-01 -3.00375104e-01 -6.20933533e-01 -1.05745077e+00
3.91213328e-01 1.17518973e+00 -6.69626296e-01 -4.58832383e-01
-3.51380527e-01 -7.86987841e-01 7.69748926e-01 6.92806304e-01
2.43877739e-01 -8.61737370e-01 -9.59675252e-01 3.63714039e-01
1.43158153e-01 -7.38085508e-01 -2.25534201e-01 5.16494215e-01
-8.10608685e-01 -1.01390946e+00 -4.00494814e-01 6.54134825e-02
3.50740820e-01 -4.70052987e-01 1.36732805e+00 -2.54670680e-01
1.84819594e-01 6.23053312e-01 -2.25372314e-02 -5.95064878e-01
-6.29188597e-01 1.76229160e-02 4.90994126e-01 -5.63007712e-01
9.71383154e-02 -8.39196682e-01 -7.15308607e-01 2.98620820e-01
-5.22108734e-01 -4.26089436e-01 4.16106671e-01 9.91602421e-01
2.38790900e-01 8.33249912e-02 9.69508708e-01 -5.76660991e-01
9.96909320e-01 -3.88876110e-01 -1.15201390e+00 2.22886071e-01
-9.77982759e-01 8.30370784e-01 6.80735648e-01 -4.21905249e-01
-8.85534763e-01 -3.33808959e-01 6.63991794e-02 -4.03658301e-01
2.92690217e-01 3.99242848e-01 3.72339904e-01 -2.49634255e-02
9.79891539e-01 2.13358086e-03 3.83212090e-01 -8.97547379e-02
4.62375194e-01 3.82772326e-01 2.51215279e-01 -1.15390635e+00
3.78374398e-01 1.19996801e-01 5.24954200e-02 -8.01140741e-02
-1.08661532e+00 3.30317497e-01 1.48477420e-01 -2.15948001e-01
4.18257356e-01 -9.49829400e-01 -1.41606200e+00 1.04494363e-01
-6.85671151e-01 -7.71923721e-01 -4.12269741e-01 4.81206357e-01
-1.08462226e+00 1.13527499e-01 -3.73079330e-01 -1.26233268e+00
-2.61005938e-01 -1.29339087e+00 7.16672301e-01 1.62420988e-01
-2.74059415e-01 -8.35668445e-01 2.07995877e-01 4.04947251e-02
3.83101404e-01 2.90904850e-01 8.72930050e-01 -8.59766364e-01
-3.04894090e-01 1.58300266e-01 1.55315965e-01 8.38953674e-01
-2.58388013e-01 4.03712364e-03 -9.99715805e-01 -4.28580910e-01
-1.73750669e-01 -9.72676635e-01 8.33246887e-01 4.09846634e-01
1.17215455e+00 -3.26136172e-01 7.83188418e-02 2.37723812e-01
1.34968328e+00 3.06627274e-01 2.96144366e-01 3.20246756e-01
8.61484334e-02 3.64881456e-01 6.28311932e-01 7.73021996e-01
2.37837717e-01 3.04139227e-01 6.19420409e-01 6.93650603e-01
6.36703312e-01 -3.69828731e-01 7.21327126e-01 2.06927270e-01
-1.18685715e-01 -1.02580324e-01 -7.68046260e-01 2.71662086e-01
-1.94891524e+00 -9.94416595e-01 7.60735810e-01 2.57674170e+00
1.03635108e+00 5.39722264e-01 4.27991718e-01 -1.84263945e-01
2.38119632e-01 2.46811405e-01 -1.06154919e+00 -7.12170303e-01
2.44084090e-01 5.40490270e-01 7.16155887e-01 5.41921914e-01
-9.20297265e-01 7.46407986e-01 6.92483711e+00 9.94449615e-01
-9.23413396e-01 -1.03736944e-01 9.83575642e-01 -4.33053315e-01
-2.89268553e-01 -8.95721242e-02 -7.67274857e-01 5.44598520e-01
1.28931069e+00 -1.85906410e-01 8.99046063e-01 9.67920959e-01
1.23665906e-01 -3.20006609e-01 -1.41130960e+00 6.90097153e-01
-4.46782023e-01 -1.25984859e+00 -6.90671086e-01 -1.49169192e-02
9.75850046e-01 3.05925280e-01 2.91871250e-01 9.71968651e-01
1.18297267e+00 -1.48665881e+00 6.68310165e-01 3.92885536e-01
5.61775982e-01 -1.10586119e+00 6.11015737e-01 5.11461020e-01
-5.76059043e-01 -4.21029866e-01 -3.13143969e-01 -8.91979337e-02
-2.00549588e-01 3.32067579e-01 -8.68914187e-01 2.51602888e-01
3.76384735e-01 2.00673953e-01 -1.13952264e-01 4.19677079e-01
-2.84798145e-01 6.98043525e-01 -6.92382514e-01 -3.40604246e-01
6.43961370e-01 -1.40943170e-01 3.20619643e-01 8.12878191e-01
2.21504956e-01 -3.43208909e-02 -7.79294670e-02 1.06817567e+00
6.60826713e-02 -3.55072111e-01 -5.83504558e-01 -4.00276512e-01
4.60460722e-01 9.71406460e-01 -2.37485006e-01 -1.84894398e-01
1.01168834e-01 3.08909923e-01 6.46355450e-01 3.97358418e-01
-1.08337498e+00 2.47873424e-04 7.50109076e-01 -5.05460918e-01
4.12746906e-01 -4.04863179e-01 -1.95039734e-01 -1.00289774e+00
-2.00452089e-01 -1.36769783e+00 4.70837444e-01 -4.61676747e-01
-1.19760239e+00 3.28803599e-01 7.92962313e-02 -1.10480034e+00
-9.67096269e-01 -6.17971778e-01 -4.99310136e-01 7.78808475e-01
-1.13667715e+00 -4.59279716e-01 3.65897000e-01 2.72798538e-01
2.32599363e-01 -5.10592647e-02 8.07275712e-01 -4.88984704e-01
-5.76625586e-01 7.22235680e-01 4.98631001e-01 -2.35964999e-01
5.73334217e-01 -1.34367526e+00 2.23303571e-01 4.88479942e-01
2.08257154e-01 4.07785237e-01 1.02374554e+00 -4.27600324e-01
-1.29916441e+00 -7.22171903e-01 -1.23157971e-01 -7.29517043e-01
8.33510518e-01 -9.43620950e-02 -4.50162739e-01 8.07051122e-01
-2.69148257e-02 -8.24813023e-02 2.77096659e-01 4.41330522e-01
-3.43215883e-01 -1.94055945e-01 -1.24724007e+00 7.92711139e-01
8.87659907e-01 -2.71014452e-01 -5.76149106e-01 1.09943904e-01
6.48884833e-01 -5.23785770e-01 -9.13145304e-01 5.94711423e-01
8.19261789e-01 -1.13618803e+00 7.53909051e-01 -9.49446797e-01
5.09634614e-01 1.26788408e-01 -2.99235344e-01 -1.77420235e+00
3.01734865e-01 -8.23312759e-01 -4.16741967e-01 7.21953034e-01
4.25118387e-01 -8.44723284e-01 7.72827506e-01 7.94466555e-01
4.76607792e-02 -1.15698934e+00 -1.06542385e+00 -9.47835982e-01
5.14541268e-01 -7.51891315e-01 6.10918999e-01 5.01226842e-01
1.45503268e-01 8.57284889e-02 -2.50242352e-01 -1.41126243e-02
7.69415021e-01 2.77625509e-02 6.34294868e-01 -8.33832979e-01
-8.02868426e-01 -6.65453970e-01 7.18412772e-02 -8.00754488e-01
4.87190187e-01 -4.78288144e-01 2.34027982e-01 -7.49964774e-01
-2.67900020e-01 -7.13548720e-01 -5.36354542e-01 4.32807505e-01
4.26643007e-02 -2.23504499e-01 3.44174981e-01 -2.38018245e-01
-7.91795611e-01 7.42769122e-01 9.82754946e-01 2.48093605e-02
-2.77285606e-01 1.27473339e-01 -8.42433989e-01 8.67482543e-01
8.95766437e-01 -3.69956046e-01 -7.33203351e-01 -5.97054958e-02
6.57707095e-01 2.82041430e-01 2.28363931e-01 -9.61307526e-01
-1.40976876e-01 -4.32880789e-01 2.95307159e-01 -2.04584360e-01
2.96259075e-01 -6.53399646e-01 -1.29631504e-01 4.17952091e-01
-5.07120609e-01 8.22322369e-02 5.09038150e-01 6.79673016e-01
2.03462675e-01 -2.47950494e-01 6.78932130e-01 -2.49823585e-01
-4.74662513e-01 5.87473921e-02 -2.68763572e-01 4.22150493e-01
9.39087093e-01 1.82461709e-01 -3.08681756e-01 -6.31943047e-01
-6.58041835e-01 6.39399171e-01 1.66891649e-01 1.91271212e-02
3.78400475e-01 -1.01957715e+00 -5.65347850e-01 1.75491429e-03
-1.97766036e-01 -9.13193151e-02 6.24395013e-02 7.35071123e-01
-1.88751053e-02 3.54266971e-01 6.90751225e-02 -4.48902518e-01
-6.58397019e-01 3.09468806e-01 8.34357917e-01 -6.22540832e-01
-1.47560760e-01 7.59674132e-01 -6.54222593e-02 -4.96124029e-01
3.10059398e-01 -4.75975603e-01 3.14786345e-01 -1.50181115e-01
2.98637897e-01 4.97850746e-01 -1.01778954e-01 9.66096669e-02
-2.29112431e-01 -1.14496216e-01 4.50597517e-02 -4.95554924e-01
1.28198433e+00 1.21829458e-01 4.16148216e-01 5.00082314e-01
7.90067494e-01 -2.86718905e-01 -1.86487114e+00 8.21925774e-02
5.31594194e-02 -8.17639530e-02 6.32110760e-02 -1.28745520e+00
-6.69519126e-01 6.50007129e-01 5.08898616e-01 -7.52431899e-03
9.09179509e-01 -3.35529298e-01 3.00526142e-01 7.95333028e-01
6.44719183e-01 -1.43927884e+00 8.25220719e-02 6.29295647e-01
7.96931922e-01 -1.41775739e+00 1.79325402e-01 5.77433944e-01
-8.78958464e-01 1.06289303e+00 7.51315951e-01 -4.58884954e-01
3.74110371e-01 2.87717104e-01 -2.10695818e-01 7.08114505e-02
-1.22527504e+00 -4.24563214e-02 5.99122941e-02 2.21623614e-01
1.52650729e-01 2.55843729e-01 -1.56654194e-01 6.66664779e-01
-2.13952169e-01 4.34641838e-02 4.46515918e-01 8.24493408e-01
-5.07060349e-01 -9.34244156e-01 -2.76045918e-01 5.45518160e-01
-5.31969666e-01 -1.64259132e-02 1.81166545e-01 1.01288891e+00
-2.16514170e-01 8.22389185e-01 1.54865593e-01 -4.44950879e-01
-4.47209589e-02 2.21735626e-01 8.89554799e-01 -3.56684685e-01
-6.26835108e-01 -2.37170562e-01 3.99193048e-01 -8.34400654e-01
-8.96969810e-02 -5.20746231e-01 -1.23016953e+00 -3.49416673e-01
-1.79362491e-01 3.41842353e-01 5.79863071e-01 1.23730671e+00
3.85324687e-01 2.73321152e-01 7.14193761e-01 -8.39515448e-01
-1.90373409e+00 -9.56332445e-01 -5.94798565e-01 -5.38989455e-02
3.47771019e-01 -8.58357072e-01 -7.62854338e-01 -7.67788649e-01]
|
[4.171448707580566, 2.429891586303711]
|
43d9f557-3651-43e6-a99c-e3d08eaf227e
|
sparsity-based-morphological-identification
|
2301.06538
| null |
https://arxiv.org/abs/2301.06538v1
|
https://arxiv.org/pdf/2301.06538v1.pdf
|
Sparsity based morphological identification of heartbeats
|
The electrocardiogram (ECG) is one of the most common primary tests to evaluate the health of the heart. Reliable automatic interpretation of ECG records is crucial to the goal of improving public health. It can enable a safe inexpensive monitoring. This work presents a new methodology for morphological identification of heartbeats, which is placed outside the usual machine learning framework. The proposal considers the sparsity of the representation of a heartbeat as a parameter for morphological identification. The approach involves greedy algorithms for selecting elements from redundant dictionaries, which should be previously learnt from examples of the classes to be identified. Using different metrics of sparsity, the dictionary rendering the smallest sparsity value, for the equivalent approximation quality of a new heartbeat, classifies the morphology of that beat. This study focuses on a procedure of learning the dictionaries for representing heartbeats and compares several metrics of sparsity for morphological identification on the basis of those metrics. The suitability of the method is illustrated by binary differentiation of Normal and Ventricular heartbeats in the MIT-BIH Arrhythmia data set. In general classification 99.7% of the Normal beats and 97.6% of the Ventricular beats in the testing sets are correctly identified. In interpatient assessment 91.8% of the Normal beats and 91.0% of Ventricular beats are correctly identified. Even more important than these scores is the fact that they are produced on the bases of a single parameter. The numerical tests, designed to emphasise the interpretability and reliability of the approach, demonstrate the potential of the method to contribute towards the development of a well grounded expert system for classification of heartbeats in ECG records.
|
['Amadou Sidi Watt', 'Khalil Battikh', 'Laura Rebollo-Neira']
|
2023-01-16
| null | null | null | null |
['classification']
|
['methodology']
|
[ 4.42961991e-01 1.40257284e-01 1.23937331e-01 -3.32420677e-01
-3.20866525e-01 -5.38055241e-01 1.25832513e-01 5.54986596e-01
-2.46078789e-01 6.59635007e-01 -3.28051411e-02 -3.18578482e-01
-4.95832950e-01 -4.93389785e-01 3.83361951e-02 -8.19467247e-01
-3.07511896e-01 6.39182031e-01 -2.37257063e-01 2.13055476e-03
2.63835251e-01 7.52333879e-01 -1.40782225e+00 2.88525194e-01
6.67001009e-01 1.07874000e+00 -4.96699028e-02 8.05423796e-01
2.29237735e-01 2.00966805e-01 -8.91064703e-01 -2.93655577e-03
3.42377305e-01 -9.04697657e-01 -6.25925124e-01 2.19031692e-01
8.21457710e-03 9.04735550e-02 3.00246149e-01 7.49835193e-01
8.06367695e-01 -3.18635046e-01 7.53263891e-01 -7.26082563e-01
2.95668632e-01 4.80510503e-01 3.72470990e-02 5.34503937e-01
3.78035218e-01 -1.25247285e-01 8.97824466e-01 -7.41032362e-01
4.30180669e-01 4.52549428e-01 8.24927688e-01 1.74296230e-01
-1.20832407e+00 -2.76523113e-01 -6.28767908e-01 -4.99955304e-02
-1.65863478e+00 -3.79619420e-01 8.36361527e-01 -6.43488646e-01
4.32961017e-01 8.21266472e-01 1.00099421e+00 2.32390091e-01
2.89796174e-01 2.39914730e-01 1.06738079e+00 -6.76387966e-01
3.78466427e-01 2.74710685e-01 2.24467263e-01 5.85589111e-01
6.15173042e-01 9.69116539e-02 -2.58673102e-01 -3.58103096e-01
8.26146841e-01 -9.82287526e-02 -2.62118876e-01 -3.07080954e-01
-1.23951638e+00 6.72700584e-01 -1.13782004e-01 7.82473505e-01
-4.99301046e-01 -2.28428647e-01 5.19323826e-01 2.90271461e-01
1.06204674e-01 6.67259514e-01 -3.35614473e-01 -1.19747460e-01
-1.12384224e+00 6.40564337e-02 9.35726345e-01 2.75349468e-01
4.05567020e-01 2.89945483e-01 -3.34118046e-02 6.43855214e-01
1.09752171e-01 4.58398908e-01 6.56230092e-01 -8.37128162e-01
-8.09502080e-02 8.73807847e-01 -1.07365027e-01 -1.11726296e+00
-5.54212391e-01 -8.14107239e-01 -9.36484218e-01 1.55570030e-01
4.88177478e-01 -7.19990507e-02 -5.16502619e-01 1.26021993e+00
2.14473486e-01 9.56541300e-03 1.75129309e-01 6.77380800e-01
7.26532519e-01 3.50019038e-01 -2.75298327e-01 -6.89874291e-01
1.29305279e+00 7.29836449e-02 -6.33777499e-01 3.27603519e-01
5.62701762e-01 -7.61116922e-01 6.37729228e-01 7.01604903e-01
-8.95955026e-01 -7.37974346e-01 -1.03847170e+00 8.09789479e-01
8.30155835e-02 4.92382526e-01 3.57057601e-01 7.65175402e-01
-6.89825416e-01 6.03500664e-01 -5.92450678e-01 -2.33219311e-01
6.28179237e-02 4.75194931e-01 -1.57758474e-01 5.04547656e-01
-1.01301467e+00 9.13308918e-01 6.66135967e-01 1.64738610e-01
-6.10659897e-01 -4.23093945e-01 -6.09620273e-01 1.52955865e-02
-1.26116216e-01 -4.16515559e-01 7.83892334e-01 -1.16258764e+00
-9.72861707e-01 1.14005053e+00 3.09095345e-02 -6.67783916e-01
4.70780909e-01 2.53004044e-01 -5.45108438e-01 3.74732465e-01
5.17843999e-02 3.73213924e-02 1.02537417e+00 -1.03665686e+00
-5.01783133e-01 -1.86948642e-01 -5.03839552e-01 2.90085245e-02
-1.62272155e-01 -1.16085745e-01 -5.54587841e-02 -7.68129289e-01
4.28445429e-01 -8.19322050e-01 -1.59794465e-01 -4.14645851e-01
-1.83142409e-01 3.15818116e-02 5.36650240e-01 -7.77974367e-01
1.65222168e+00 -2.20531559e+00 1.45413920e-01 8.31976116e-01
2.70501792e-01 3.67504805e-01 4.12881285e-01 3.51869375e-01
-3.11635256e-01 -5.71439229e-02 -3.90260547e-01 8.22739229e-02
-3.33184391e-01 3.27565223e-01 -2.34759241e-01 5.87698877e-01
3.79270921e-03 3.63467723e-01 -6.58059955e-01 -5.08879483e-01
3.34566742e-01 4.05592471e-01 -1.27682239e-01 1.61775693e-01
3.86208773e-01 5.88014901e-01 -2.90260226e-01 4.42015111e-01
3.19517893e-03 -1.17399074e-01 5.15955031e-01 -4.24212456e-01
-2.48813741e-02 3.83470468e-02 -1.56767833e+00 1.10032547e+00
1.89252645e-02 3.50987971e-01 -3.47117305e-01 -1.17186403e+00
1.34069729e+00 8.27142596e-01 7.80226469e-01 -3.05569887e-01
2.99985081e-01 6.41250312e-01 4.35348511e-01 -4.65107322e-01
-5.83100505e-02 -2.78242916e-01 -1.40886540e-02 4.55494404e-01
2.83114687e-02 -1.31429285e-01 4.00090247e-01 -8.53236020e-02
7.22863436e-01 -3.56211066e-01 9.53076184e-01 -4.92888242e-01
7.76396751e-01 -2.62352377e-01 4.95859146e-01 5.14959991e-01
4.09028307e-03 7.14290857e-01 4.44883853e-01 -9.90928054e-01
-9.55057144e-01 -6.99079096e-01 -4.16907758e-01 2.75358886e-01
-2.78470725e-01 -3.61140013e-01 -7.03226089e-01 -4.58060026e-01
-1.42496839e-01 2.79360890e-01 -4.26842421e-01 -2.48694122e-02
-5.92682064e-01 -9.88287568e-01 5.84939480e-01 1.49258673e-01
7.75536243e-03 -1.17248940e+00 -1.41683662e+00 3.53437930e-01
-2.49179587e-01 -7.09012687e-01 1.21770874e-01 2.95894176e-01
-1.31493807e+00 -1.22246814e+00 -5.70671678e-01 -6.15207314e-01
7.51713276e-01 -3.78614157e-01 1.07830358e+00 4.96637434e-01
-6.87312126e-01 2.42928058e-01 -3.80043477e-01 -3.93148631e-01
-8.46289277e-01 -1.25586972e-01 1.70995146e-01 3.78627121e-01
5.38259037e-02 -4.18960005e-01 -4.21895623e-01 3.41456383e-01
-5.95678389e-01 -2.46780857e-01 4.86619323e-01 7.55651712e-01
9.57094848e-01 2.24991083e-01 6.65453434e-01 -8.71730983e-01
6.52370095e-01 -2.79259145e-01 -3.78506899e-01 1.01873986e-01
-8.47174525e-01 -1.25059187e-01 6.68052554e-01 -2.63056904e-01
-2.13938564e-01 3.43206137e-01 -9.25373882e-02 -1.27340093e-01
-3.15698117e-01 5.48509538e-01 1.07002512e-01 -1.96426675e-01
8.70305657e-01 2.98544437e-01 6.27424419e-02 -4.44275826e-01
-1.76451281e-01 4.98034447e-01 6.05817020e-01 -4.56902206e-01
4.30875748e-01 2.57691711e-01 2.92882323e-01 -1.00658190e+00
-3.46864611e-01 -8.13195527e-01 -6.76217437e-01 -4.35991645e-01
6.08340383e-01 -3.72172832e-01 -5.75237453e-01 1.55678332e-01
-7.98421204e-01 1.94127515e-01 -8.50129902e-01 5.66107810e-01
-5.85675299e-01 6.27317846e-01 -1.19499244e-01 -1.06983340e+00
-6.10668123e-01 -8.45059216e-01 7.44537234e-01 -1.41625628e-01
-8.76921117e-01 -9.18676972e-01 1.32628858e-01 8.16355050e-02
7.89472535e-02 6.44160688e-01 9.71666157e-01 -9.87269163e-01
1.38836831e-01 -6.75486088e-01 5.26263177e-01 7.15896964e-01
3.40562820e-01 -9.34262946e-03 -8.59189510e-01 -3.99655163e-01
5.36468804e-01 1.65709138e-01 4.64840889e-01 3.96925449e-01
7.12167621e-01 -3.12883228e-01 -5.89603819e-02 4.60013181e-01
1.41921270e+00 6.99046195e-01 5.88371038e-01 8.81525874e-02
3.49822730e-01 3.40845704e-01 5.37361860e-01 6.59245253e-01
-2.10196152e-01 5.38824260e-01 2.69830137e-01 -4.07002956e-01
5.98962158e-02 2.85313696e-01 1.16446845e-01 9.99442637e-01
-4.28866446e-01 1.96011558e-01 -1.14482474e+00 5.76936185e-01
-1.42908108e+00 -9.75501776e-01 -3.19295466e-01 2.51521921e+00
6.45902038e-01 3.01116198e-01 3.15233022e-01 1.21579862e+00
6.71590686e-01 -2.19137758e-01 -2.09547698e-01 -5.29075325e-01
-1.00585602e-01 6.53280795e-01 1.86392576e-01 4.93125409e-01
-1.05056834e+00 -5.46821253e-03 6.34620047e+00 2.91974574e-01
-1.24866188e+00 -2.43962556e-01 7.52159953e-01 2.68580914e-01
1.16961136e-01 -1.11512482e-01 -7.16210365e-01 4.43298906e-01
1.00228417e+00 -1.23239318e-02 1.34566933e-01 5.25275052e-01
4.60002244e-01 -1.49661794e-01 -1.07902050e+00 1.16580820e+00
2.06222147e-01 -1.25184619e+00 4.68033925e-02 -1.27393201e-01
5.16887903e-01 -4.81347144e-01 -1.31988063e-01 -3.31490159e-01
-7.02605903e-01 -1.09222722e+00 7.04812407e-01 6.88687265e-01
8.49120915e-01 -6.91887856e-01 1.08296680e+00 3.34143132e-01
-9.91631985e-01 -1.69489637e-01 -6.73711747e-02 -1.64535806e-01
1.77518707e-02 7.62281179e-01 -1.11381602e+00 5.23876250e-01
2.77737767e-01 6.02531374e-01 -5.44322908e-01 1.19690931e+00
-4.36568223e-02 9.77892339e-01 -4.74722475e-01 2.42793441e-01
-1.67024672e-01 -1.52923569e-01 7.84175992e-01 1.25169861e+00
3.44572961e-01 9.51513499e-02 2.54252404e-01 5.47224581e-01
4.91666943e-01 5.51498234e-01 -5.28651595e-01 1.17824927e-01
3.73566121e-01 1.12364995e+00 -1.01555884e+00 -4.25910026e-01
6.79421946e-02 4.77624327e-01 -3.52147788e-01 -9.78737883e-03
-4.39961195e-01 -4.26363111e-01 6.23190776e-02 3.52573335e-01
1.47751018e-01 2.65334100e-01 -7.09670067e-01 -7.51123726e-01
9.20311809e-02 -1.26435483e+00 6.41551733e-01 -1.99683592e-01
-9.05642450e-01 9.03397739e-01 -5.25816204e-03 -1.58986223e+00
-5.57287097e-01 -2.67653763e-01 -5.78187704e-01 1.05548894e+00
-8.14017177e-01 -5.69707870e-01 -1.84558690e-01 4.54276204e-01
1.63740039e-01 -3.74303460e-01 1.29765701e+00 2.93590814e-01
-1.73494309e-01 2.99220681e-01 -1.80871442e-01 1.23987034e-01
3.01546991e-01 -1.19153476e+00 -1.06820099e-01 7.84460664e-01
4.03148592e-01 5.44670224e-01 7.96450913e-01 -4.89699662e-01
-6.71701133e-01 -6.83511913e-01 1.38077295e+00 -2.24467322e-01
9.94101539e-02 2.36892939e-01 -8.46110225e-01 2.73928996e-02
-2.01600298e-01 -1.01512738e-01 9.85238731e-01 -1.73812971e-01
2.66039759e-01 -4.09666896e-01 -1.05351377e+00 6.15672469e-02
1.36843771e-01 -3.43576014e-01 -8.55735242e-01 1.52764827e-01
-3.81428689e-01 -2.46752590e-01 -1.13734531e+00 5.22614539e-01
7.22647488e-01 -1.11836636e+00 9.61618483e-01 -3.91560823e-01
-7.72538260e-02 -4.99027908e-01 1.05919719e-01 -8.41857553e-01
-1.42832294e-01 -7.93336332e-01 -1.74087033e-01 8.23944569e-01
2.11726516e-01 -5.56168556e-01 6.23441219e-01 2.70537231e-02
1.03701137e-01 -9.14210677e-01 -1.03185856e+00 -6.00218773e-01
-5.46244383e-01 -1.80179313e-01 2.40581214e-01 9.05967295e-01
-3.66908908e-02 3.34205568e-01 -3.48002911e-01 -4.72083986e-02
6.88777506e-01 1.78294703e-01 3.89264762e-01 -1.69853902e+00
-5.83478332e-01 -3.59766603e-01 -9.41217244e-01 -8.10854957e-02
-5.08064151e-01 -9.06939149e-01 -2.29540363e-01 -1.20179760e+00
-8.84190798e-02 -6.71756446e-01 -6.47624314e-01 3.65204096e-01
7.13694002e-03 4.92898196e-01 1.24951169e-01 4.30425733e-01
7.19529390e-02 -4.24723357e-01 7.99590588e-01 3.59557830e-02
-4.83545840e-01 4.96963978e-01 -4.33812648e-01 9.16990161e-01
1.05781043e+00 -6.68748021e-01 -4.79393989e-01 2.37985715e-01
2.61497766e-01 3.18663627e-01 2.77668953e-01 -1.25771546e+00
-2.01022968e-01 2.94392735e-01 4.48828876e-01 -4.45240915e-01
4.83610108e-02 -9.51072633e-01 7.97874689e-01 1.06400466e+00
-3.53218257e-01 2.90091515e-01 -1.26610547e-01 2.26866588e-01
-3.43484312e-01 -6.09387696e-01 9.14913177e-01 -1.47214696e-01
-3.61798942e-01 -1.28132045e-01 -4.66277987e-01 1.43501416e-01
9.33555841e-01 -6.34613514e-01 6.26355648e-01 -2.51458913e-01
-1.17980969e+00 -4.43011582e-01 9.62335989e-02 -1.37459099e-01
6.52760446e-01 -9.88165677e-01 -9.06861722e-01 4.97682363e-01
-5.43321148e-02 -3.22687745e-01 -1.48715347e-01 1.11259520e+00
-8.82317364e-01 2.75202096e-01 -4.40581352e-01 -7.12841034e-01
-1.70568180e+00 2.13193253e-01 5.30057192e-01 -3.11301857e-01
-6.31158471e-01 4.58087265e-01 -2.24607706e-01 3.55452150e-01
3.58397871e-01 -5.36927104e-01 -6.19368970e-01 2.60413110e-01
4.06929702e-01 5.10808170e-01 3.32050949e-01 -7.40234196e-01
-3.78019303e-01 6.60965204e-01 4.49698657e-01 1.00037985e-01
1.13298273e+00 2.30193704e-01 -3.21007460e-01 7.35170662e-01
7.33695447e-01 2.16422945e-01 -5.75976491e-01 2.24084884e-01
1.46081060e-01 -2.08878666e-01 -2.42204070e-01 -9.18893158e-01
-8.74319375e-01 8.05097282e-01 8.88697863e-01 5.68986237e-01
1.35745370e+00 -5.00927627e-01 3.03233981e-01 2.82133609e-01
3.79405797e-01 -8.68144572e-01 -2.56766081e-01 2.30943155e-03
7.34170735e-01 -7.27636337e-01 2.43563518e-01 -3.33127797e-01
-5.56951940e-01 1.43810391e+00 -3.31876963e-01 -1.52891561e-01
5.71145177e-01 1.11625291e-01 2.64977217e-01 -2.52298355e-01
-2.05586061e-01 4.87383120e-02 4.51280266e-01 4.07167435e-01
4.70027775e-01 2.71756500e-01 -1.04088545e+00 5.16744852e-01
-6.30055964e-02 4.81657721e-02 3.66303474e-01 8.62549663e-01
-4.80788708e-01 -1.11375368e+00 -4.71347243e-01 5.51271737e-01
-6.95661902e-01 -8.86027515e-02 -4.31337297e-01 5.44018447e-01
5.20990610e-01 6.80314600e-01 -3.02438647e-01 -1.61795616e-01
3.02935660e-01 3.42159629e-01 4.82063979e-01 -6.54560924e-01
-1.02591538e+00 3.09389412e-01 1.28135413e-01 -6.07593060e-02
-4.10963625e-01 -7.25553453e-01 -1.21079659e+00 2.30514690e-01
-1.73152521e-01 6.20846212e-01 5.77933908e-01 6.87778354e-01
3.75609361e-02 4.35164660e-01 5.79195559e-01 -4.25941557e-01
-7.36187637e-01 -6.81055427e-01 -8.50487828e-01 5.78749299e-01
2.62669504e-01 -3.31706375e-01 -4.07331944e-01 6.12148523e-01]
|
[14.206507682800293, 3.2180705070495605]
|
4dc202b6-a584-4c56-8928-72e5e1d12df9
|
towards-ontology-reshaping-for-kg-generation
|
2209.11067
| null |
https://arxiv.org/abs/2209.11067v1
|
https://arxiv.org/pdf/2209.11067v1.pdf
|
Towards Ontology Reshaping for KG Generation with User-in-the-Loop: Applied to Bosch Welding
|
Knowledge graphs (KG) are used in a wide range of applications. The automation of KG generation is very desired due to the data volume and variety in industries. One important approach of KG generation is to map the raw data to a given KG schema, namely a domain ontology, and construct the entities and properties according to the ontology. However, the automatic generation of such ontology is demanding and existing solutions are often not satisfactory. An important challenge is a trade-off between two principles of ontology engineering: knowledge-orientation and data-orientation. The former one prescribes that an ontology should model the general knowledge of a domain, while the latter one emphasises on reflecting the data specificities to ensure good usability. We address this challenge by our method of ontology reshaping, which automates the process of converting a given domain ontology to a smaller ontology that serves as the KG schema. The domain ontology can be designed to be knowledge-oriented and the KG schema covers the data specificities. In addition, our approach allows the option of including user preferences in the loop. We demonstrate our on-going research on ontology reshaping and present an evaluation using real industrial data, with promising results.
|
['Evgeny Kharlamov', 'Egor V. Kostylev', 'Gong Cheng', 'Jieying Chen', 'Baifan Zhou', 'Dongzhuoran Zhou']
|
2022-09-22
| null | null | null | null |
['general-knowledge']
|
['miscellaneous']
|
[ 2.46809214e-01 5.09029686e-01 -1.86778456e-01 -3.73363137e-01
6.76283613e-02 -4.84393388e-01 3.92714590e-01 3.88839781e-01
-3.91439348e-02 5.84925175e-01 -5.94585063e-03 -2.06942484e-01
-7.49920189e-01 -1.51487410e+00 -3.84771645e-01 -3.20391476e-01
1.60723343e-01 7.86946118e-01 5.91737688e-01 -4.94811922e-01
2.15565816e-01 5.06164014e-01 -2.05215669e+00 1.82211742e-01
1.13202238e+00 1.05787754e+00 3.48286718e-01 1.35686830e-01
-6.17201030e-01 5.38884938e-01 -4.60392565e-01 -3.92251104e-01
1.90886751e-01 -3.05978298e-01 -1.14714360e+00 3.10550660e-01
-1.10419713e-01 4.48987871e-01 4.35957640e-01 1.30550325e+00
7.02738687e-02 -4.04258072e-02 3.08024883e-01 -1.53219247e+00
-3.99168372e-01 5.60610831e-01 2.69409478e-01 -4.69704807e-01
4.74221170e-01 -4.26507711e-01 1.03768110e+00 -3.00711304e-01
9.08580422e-01 1.04060757e+00 1.52437136e-01 4.02802914e-01
-1.02841628e+00 -2.30829820e-01 1.69656556e-02 3.21383238e-01
-1.39005804e+00 -1.14174083e-01 6.88067377e-01 -7.43881345e-01
6.81464553e-01 3.34691793e-01 8.23486328e-01 3.63530636e-01
1.08796895e-01 3.31979357e-02 8.94874036e-01 -8.65367353e-01
5.63434184e-01 7.43783534e-01 2.59611636e-01 2.67820835e-01
7.74654925e-01 -2.27763280e-01 -3.55976105e-01 -9.04598311e-02
6.26068652e-01 -2.41895676e-01 -2.73891091e-01 -1.11314011e+00
-7.53172457e-01 5.86608768e-01 -9.48794745e-03 6.74355626e-01
-4.12339687e-01 -3.70475054e-01 4.25467581e-01 4.75433171e-01
-3.26284729e-02 6.84805453e-01 -5.50649822e-01 7.12785348e-02
-3.34344506e-01 4.46439713e-01 1.31292880e+00 1.29234278e+00
1.04125547e+00 -2.32519999e-01 5.13724148e-01 7.41046309e-01
4.87053841e-01 1.61134005e-01 4.40105319e-01 -5.82147956e-01
2.28828207e-01 1.35294628e+00 2.37762198e-01 -9.18179274e-01
-1.77082434e-01 -3.64213705e-01 -2.74466693e-01 3.74056160e-01
3.03830445e-01 2.81760246e-01 -6.29416823e-01 1.60041404e+00
4.95453566e-01 -4.89824086e-01 5.68124354e-01 5.43831050e-01
7.07703531e-01 2.96031147e-01 4.05817181e-01 -2.17587397e-01
1.59463215e+00 -3.50751609e-01 -9.55717325e-01 -4.53985520e-02
7.69423306e-01 -4.79292661e-01 8.86930704e-01 4.51328814e-01
-8.25256467e-01 -5.00945091e-01 -1.15436900e+00 1.08173363e-01
-1.06282091e+00 -1.95134833e-01 4.34112221e-01 7.19323277e-01
-7.41182625e-01 5.38351834e-01 -1.60104483e-01 -9.64887083e-01
-1.51052549e-01 5.32353163e-01 -4.85271752e-01 9.43867266e-02
-1.55539012e+00 1.00204062e+00 1.42062092e+00 -2.62007952e-01
-1.09448871e-02 -4.29751933e-01 -7.68267334e-01 2.52391607e-01
8.71260643e-01 -7.21370816e-01 1.02159238e+00 -1.05579460e+00
-1.42775869e+00 6.85190797e-01 4.61975485e-01 -3.31526607e-01
3.39320540e-01 1.99582979e-01 -1.13791871e+00 -2.46463776e-01
4.18356434e-02 8.20942670e-02 4.60574865e-01 -1.35890245e+00
-1.24906147e+00 -3.81633669e-01 4.64849383e-01 1.14048958e-01
-5.08231163e-01 -1.91792011e-01 -4.35855508e-01 -2.28185162e-01
7.29313567e-02 -6.18365765e-01 -1.69870518e-02 -5.01522660e-01
-1.30105972e-01 -1.84637398e-01 7.93927491e-01 -3.66688818e-01
1.57350636e+00 -1.92511964e+00 2.12452322e-01 6.59927368e-01
-5.29422313e-02 2.18351811e-01 1.94331199e-01 8.19586635e-01
-3.05169374e-01 1.21165983e-01 -7.20905364e-02 5.11285186e-01
4.15104270e-01 4.67536360e-01 -2.59609699e-01 -2.58449435e-01
-4.27238196e-02 4.11362469e-01 -8.87984574e-01 -4.30700809e-01
2.61269748e-01 9.93021950e-02 -5.92622757e-01 1.02125235e-01
-6.43602490e-01 1.98396847e-01 -6.46462560e-01 4.26641434e-01
5.18355668e-01 -1.07438147e-01 9.02650297e-01 -5.15642762e-01
-2.14300960e-01 1.51542187e-01 -1.67542481e+00 1.49730492e+00
-5.56707919e-01 -2.59025723e-01 -1.26536712e-01 -9.32586014e-01
1.26587296e+00 5.71139216e-01 5.47212899e-01 -5.56618214e-01
1.33198440e-01 6.39365315e-01 -3.41899693e-02 -6.74461842e-01
6.17993355e-01 -2.56960154e-01 -1.73090279e-01 1.72767997e-01
3.92633257e-03 -2.49000981e-01 7.90366352e-01 -1.73453733e-01
6.35843873e-01 5.29476047e-01 9.34190750e-01 -5.83762288e-01
8.17812443e-01 3.58499199e-01 5.97705364e-01 9.57650021e-02
4.77085561e-01 -3.44992802e-02 6.95488214e-01 -6.96006119e-01
-8.28467965e-01 -7.40253985e-01 -2.02499647e-02 7.48794258e-01
1.13965809e-01 -7.47993231e-01 -8.95165443e-01 -7.58453369e-01
-5.17792515e-02 9.05942976e-01 -3.35071027e-01 -1.63470522e-01
-1.83434948e-01 -5.16301572e-01 -7.07469732e-02 1.44026205e-01
3.11775535e-01 -1.14895785e+00 -9.69314933e-01 4.76453125e-01
-6.23809993e-02 -1.23684359e+00 3.12401086e-01 -3.53497751e-02
-7.41856933e-01 -1.33367741e+00 7.36192912e-02 -6.81433678e-01
7.29491115e-01 -2.46528566e-01 1.29061604e+00 1.15403116e-01
6.11440763e-02 2.74604023e-01 -7.57656038e-01 -7.88684130e-01
-9.49383855e-01 3.68111432e-01 -8.42555538e-02 1.76480308e-01
6.09914660e-01 -6.73074424e-01 -1.24120638e-01 4.70278531e-01
-1.56509531e+00 1.18934296e-01 5.56769848e-01 1.29209191e-01
5.90452433e-01 7.66087651e-01 7.99055457e-01 -1.19969141e+00
7.37533450e-01 -4.00839984e-01 -8.47579598e-01 6.79224730e-01
-8.72290075e-01 2.50510365e-01 6.22463763e-01 -9.77632627e-02
-1.02954829e+00 2.74550080e-01 -6.60204664e-02 2.03783199e-01
-2.77409077e-01 8.85968387e-01 -9.25270557e-01 9.83577743e-02
5.01391470e-01 -3.77353057e-02 -1.68260157e-01 -5.42843640e-01
2.97533244e-01 7.41638839e-01 3.38130057e-01 -9.38159406e-01
7.10289538e-01 2.03879282e-01 1.25940740e-01 -6.46927655e-01
-4.01641428e-01 -2.68354446e-01 -8.77813458e-01 -1.77658364e-01
8.28786492e-01 -2.46566653e-01 -6.25193894e-01 -4.21583243e-02
-9.82492507e-01 1.25160515e-02 -8.34414840e-01 2.83452094e-01
-7.89322734e-01 2.97444433e-01 3.65339398e-01 -6.11959755e-01
-1.30972400e-01 -1.03707349e+00 6.03953183e-01 9.90992188e-02
-3.34043711e-01 -9.72677410e-01 -6.45800680e-03 4.91150767e-02
2.96712071e-01 2.89065570e-01 1.49899483e+00 -9.22954619e-01
-5.79212368e-01 -3.58681053e-01 -2.23960113e-02 2.49845430e-01
6.06956422e-01 -1.12069443e-01 -5.66970468e-01 -3.13159004e-02
-2.61814684e-01 2.19638020e-01 -2.37601653e-01 -3.13095331e-01
7.99449384e-01 -9.39518884e-02 -2.66537428e-01 1.08368015e-02
1.80487907e+00 5.86556375e-01 9.03588533e-01 6.59571409e-01
3.92457336e-01 1.09374034e+00 9.25862432e-01 3.03556681e-01
4.71732259e-01 1.24533665e+00 4.90175128e-01 1.96080342e-01
1.61443353e-01 -1.96099415e-01 -1.31617129e-01 6.62981212e-01
-3.22639406e-01 -1.59696713e-01 -1.08608723e+00 4.64235008e-01
-2.18914890e+00 -7.38845944e-01 -8.48236755e-02 2.58532476e+00
6.52468026e-01 1.81894988e-01 1.10209361e-01 3.99941444e-01
5.48477530e-01 -4.60707486e-01 -1.10171258e-01 -5.04162312e-01
2.69162983e-01 7.68070593e-02 2.84490138e-01 5.02331614e-01
-7.43612647e-01 8.49573970e-01 5.27927542e+00 5.36889434e-01
-9.35388386e-01 -1.06064208e-01 -3.56760353e-01 5.22657931e-01
-3.14289033e-01 3.03484201e-01 -8.83633554e-01 3.18938613e-01
8.54991972e-01 -7.94328272e-01 3.52975935e-01 9.54194486e-01
6.93370625e-02 5.73300896e-03 -1.08770800e+00 5.60960531e-01
-2.24887192e-01 -1.23561811e+00 4.69968647e-01 4.07258630e-01
2.76315480e-01 -7.85935521e-01 -6.36267483e-01 1.86674759e-01
3.26800436e-01 -5.43565750e-01 8.57367277e-01 7.29992628e-01
4.79182631e-01 -9.33657646e-01 8.18416774e-01 3.28522772e-01
-1.37584603e+00 -1.53563172e-01 -2.35942036e-01 1.36797875e-01
1.64234966e-01 5.10422051e-01 -8.98187041e-01 1.37191474e+00
7.23165393e-01 2.84807473e-01 -1.92958042e-01 9.04236376e-01
-1.22617789e-01 -3.28804284e-01 -9.21810865e-02 1.43719733e-01
-3.04458499e-01 -5.44643760e-01 3.82852912e-01 8.74978185e-01
5.36578774e-01 -1.00596517e-01 2.11573437e-01 7.93146431e-01
2.24864811e-01 7.27803767e-01 -7.05119371e-01 -2.41467401e-01
4.28456992e-01 1.15212691e+00 -5.32073379e-01 -4.81671602e-01
-4.53926414e-01 4.63152647e-01 8.80941525e-02 1.29411012e-01
-4.26654488e-01 -5.95514119e-01 6.84205770e-01 5.25746226e-01
2.47133777e-01 -1.96533389e-02 1.24809310e-01 -8.81005168e-01
7.24388808e-02 -1.03845370e+00 5.16261816e-01 -6.84279025e-01
-1.08348024e+00 7.30446458e-01 3.98013383e-01 -1.40587234e+00
-3.82407755e-01 -7.58204222e-01 -1.10478122e-02 8.72342050e-01
-1.35635114e+00 -1.24824417e+00 -5.04055083e-01 3.55539918e-01
1.90460935e-01 8.33746642e-02 1.11132336e+00 5.46397448e-01
-1.98162079e-01 -1.31855994e-01 -3.27833921e-01 -2.68768817e-01
4.79702055e-01 -1.29754424e+00 6.51495457e-02 5.99555910e-01
-2.26304159e-01 7.07471728e-01 8.37267339e-01 -8.74448240e-01
-1.18434346e+00 -1.00618792e+00 1.14746964e+00 -1.76731020e-01
7.33446896e-01 -1.80341989e-01 -1.02420771e+00 7.10570931e-01
-5.37791289e-02 -1.76363349e-01 8.07248771e-01 8.00475199e-03
-1.75679594e-01 -3.68923515e-01 -1.16281927e+00 5.15129447e-01
9.45906878e-01 -1.80150211e-01 -7.34121740e-01 1.62938610e-01
4.29816097e-01 -2.34872416e-01 -1.56574774e+00 5.18105626e-01
6.21816456e-01 -9.06650484e-01 6.76501095e-01 -6.90935671e-01
-6.30829930e-02 -8.02201867e-01 -3.44537914e-01 -1.28604448e+00
-2.53180802e-01 -2.69389153e-01 1.05556473e-01 1.44055235e+00
3.53323162e-01 -8.22034836e-01 5.43315291e-01 7.58585274e-01
-1.36631913e-02 -3.75899613e-01 -5.79290569e-01 -1.02926660e+00
-3.40391010e-01 -3.75821501e-01 1.43220150e+00 9.88655806e-01
5.19808352e-01 2.29905367e-01 -1.19032480e-01 3.85398597e-01
3.35555345e-01 2.78816760e-01 9.59838748e-01 -1.99214327e+00
-8.59042555e-02 -2.43779466e-01 -8.30458581e-01 -2.92356610e-01
-1.30516335e-01 -7.19948649e-01 -2.81504720e-01 -2.01038885e+00
-3.46817970e-01 -4.88528669e-01 -2.80213058e-01 4.37662452e-01
4.74811196e-01 -3.43628347e-01 2.25273535e-01 1.27672538e-01
-2.89249390e-01 1.41189381e-01 1.10303855e+00 1.43020108e-01
-5.35096109e-01 -6.71061128e-02 -7.76766598e-01 6.97301269e-01
8.95814419e-01 -4.49121028e-01 -9.23476636e-01 -2.79187895e-02
6.24077082e-01 -3.15010399e-01 -6.05459465e-03 -1.09320819e+00
1.83226839e-01 -5.29768705e-01 -4.08872217e-01 -1.42879203e-01
6.88289776e-02 -1.60160148e+00 1.03157759e+00 2.68919349e-01
7.08891600e-02 -1.89807385e-01 -5.39937848e-03 2.31442556e-01
-3.23506325e-01 -5.81071794e-01 5.80909133e-01 -2.39976868e-01
-1.30057895e+00 1.08847082e-01 -1.36132678e-02 -2.82768250e-01
1.26440227e+00 -5.86815417e-01 -8.81990418e-02 6.00892380e-02
-9.54022765e-01 -1.80901159e-02 7.58312583e-01 6.95704997e-01
7.64868408e-02 -1.35977554e+00 -1.66761190e-01 4.04487044e-01
6.95723295e-01 -3.45597602e-02 -1.32741943e-01 4.18758690e-01
-5.42537391e-01 4.82048541e-01 -5.29658258e-01 -1.74850821e-01
-1.03257203e+00 6.84801877e-01 5.07875323e-01 -4.29478288e-01
-5.00161648e-01 -2.91996181e-01 1.01664662e-01 -4.95199531e-01
-6.71989471e-02 -3.42959434e-01 -7.83413410e-01 1.21335261e-01
4.47345555e-01 2.86319017e-01 3.90278220e-01 -6.30836189e-01
-2.73246378e-01 7.20483959e-01 2.17491522e-01 9.18844491e-02
1.27525961e+00 -1.33010030e-01 -4.32219893e-01 2.23793298e-01
5.02148271e-01 2.02379242e-01 -5.17250538e-01 -1.68503925e-01
6.54354990e-01 -3.82685244e-01 -2.40377396e-01 -7.48535037e-01
-8.71925652e-01 3.16985667e-01 3.76203448e-01 1.02180588e+00
1.13940120e+00 -7.03593493e-02 2.66277581e-01 3.79336357e-01
8.67073834e-01 -1.40223789e+00 -4.74018991e-01 2.70751178e-01
9.18111026e-01 -5.42411208e-01 3.30799855e-02 -1.02348268e+00
-5.13531744e-01 1.37544775e+00 4.54197437e-01 3.43590736e-01
6.94662154e-01 1.45395577e-01 5.18018492e-02 -6.43439412e-01
-2.57601678e-01 -5.09049118e-01 3.78465146e-01 7.87242234e-01
2.25685030e-01 2.18019187e-02 -7.12684333e-01 7.25418866e-01
-2.89294004e-01 6.72188997e-01 4.38619852e-01 1.10407102e+00
-6.20668769e-01 -1.86022079e+00 -3.10119361e-01 1.53010741e-01
-1.84463665e-01 4.00104284e-01 -3.79794031e-01 1.14620733e+00
5.76982975e-01 8.89197230e-01 -1.76906884e-01 -2.74603993e-01
1.10254526e+00 3.61903638e-01 3.01859885e-01 -9.21228588e-01
-2.25916579e-01 -2.53505707e-01 5.08823812e-01 -2.74516821e-01
-5.33413827e-01 -3.27703536e-01 -1.31888688e+00 -6.26730099e-02
-3.48792434e-01 5.86858928e-01 9.25387681e-01 1.00885320e+00
4.09086049e-01 6.60451055e-01 2.12310538e-01 -2.42840812e-01
-3.19221944e-01 -4.29653674e-01 -8.50422978e-01 6.30074501e-01
-2.97659665e-01 -8.49720776e-01 9.96682569e-02 4.05756980e-01]
|
[9.105035781860352, 7.814087390899658]
|
cb56c813-b90a-48fc-a25e-5a03a13caf9c
|
em-decipherment-for-large-vocabularies
| null | null |
https://aclanthology.org/P14-2123
|
https://aclanthology.org/P14-2123.pdf
|
EM Decipherment for Large Vocabularies
| null |
['Hermann Ney', 'Malte Nuhn']
|
2014-06-01
| null | null | null |
acl-2014-6
|
['decipherment']
|
['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]
|
[-7.410221099853516, 3.768235921859741]
|
0ebe6476-6120-4a8b-ad05-30f0b4609e29
|
automatic-parameter-tying-in-neural-networks
| null | null |
https://openreview.net/forum?id=HkinqfbAb
|
https://openreview.net/pdf?id=HkinqfbAb
|
Automatic Parameter Tying in Neural Networks
|
Recently, there has been growing interest in methods that perform neural network compression, namely techniques that attempt to substantially reduce the size of a neural network without significant reduction in performance. However, most existing methods are post-processing approaches in that they take a learned neural network as input and output a compressed network by either forcing several parameters to take the same value (parameter tying via quantization) or pruning irrelevant edges (pruning) or both. In this paper, we propose a novel algorithm that jointly learns and compresses a neural network. The key idea in our approach is to change the optimization criteria by adding $k$ independent Gaussian priors over the parameters and a sparsity penalty. We show that our approach is easy to implement using existing neural network libraries, generalizes L1 and L2 regularization and elegantly enforces parameter tying as well as pruning constraints. Experimentally, we demonstrate that our new algorithm yields state-of-the-art compression on several standard benchmarks with minimal loss in accuracy while requiring little to no hyperparameter tuning as compared with related, competing approaches.
|
['Vibhav Gogate', 'Nicholas Ruozzi', 'Yibo Yang']
|
2018-01-01
| null | null | null |
iclr-2018-1
|
['l2-regularization']
|
['methodology']
|
[ 5.60902476e-01 1.85930163e-01 -3.11451674e-01 -5.96653223e-01
-4.59565341e-01 -3.90040874e-01 3.71648580e-01 2.40414903e-01
-8.55426013e-01 7.30767787e-01 -3.08952779e-02 -3.00154567e-01
-3.86451066e-01 -8.89891505e-01 -1.11240053e+00 -6.29613280e-01
7.74702281e-02 4.80777949e-01 2.58626699e-01 1.20463230e-01
3.78321856e-01 5.90335369e-01 -1.59790194e+00 1.50402695e-01
6.69823468e-01 1.18438053e+00 1.09880986e-02 5.14939129e-01
1.23860955e-01 7.46628046e-01 -4.36890781e-01 -5.01520991e-01
5.85418224e-01 -1.08396158e-01 -9.36557531e-01 -7.20254518e-03
6.96051359e-01 -1.25789002e-01 -5.12299955e-01 1.19588792e+00
3.14702243e-01 4.19272095e-01 4.76692051e-01 -7.49810636e-01
-4.46788400e-01 7.76380241e-01 -3.75700384e-01 1.66727081e-01
-3.85190308e-01 -5.49684241e-02 9.29260135e-01 -6.63553119e-01
5.31155348e-01 1.02235103e+00 8.94514978e-01 4.03110474e-01
-1.70019877e+00 -7.39818394e-01 3.22780684e-02 1.14827203e-02
-1.61109912e+00 -7.21849203e-01 8.70148301e-01 -1.15802482e-01
1.25047767e+00 2.39113733e-01 5.90449631e-01 5.51308870e-01
-1.10783495e-01 5.87538004e-01 6.68612897e-01 -6.34589195e-01
3.95642459e-01 1.24075152e-01 1.24670751e-01 1.06449330e+00
3.05841386e-01 -4.46121097e-02 -4.50919241e-01 -6.49175122e-02
8.12624991e-01 -5.65376319e-02 -1.63178056e-01 -4.77495849e-01
-8.29189241e-01 8.44523907e-01 4.12826389e-01 1.47724539e-01
-1.47236243e-01 5.77674270e-01 4.85423803e-01 4.01176840e-01
5.15779018e-01 6.02031827e-01 -4.78861481e-01 -3.75654623e-02
-1.27182341e+00 3.69369090e-01 8.23116004e-01 8.11828673e-01
7.50348806e-01 1.56798080e-01 -2.43547503e-02 1.22774255e+00
-1.13137238e-01 -1.52754754e-01 5.02988577e-01 -1.39926898e+00
5.69659770e-01 5.49231529e-01 -2.60354370e-01 -1.11933100e+00
-1.20357074e-01 -4.33729649e-01 -9.43795621e-01 1.86117157e-01
1.91403449e-01 -1.81548595e-01 -1.07592618e+00 2.00873899e+00
-5.33405729e-02 2.32511237e-01 -1.79456342e-02 4.30903226e-01
4.17616099e-01 6.17179871e-01 -1.84905067e-01 -2.00155795e-01
8.25116873e-01 -1.05083191e+00 -4.14412528e-01 -2.50175655e-01
5.28091669e-01 -5.49513340e-01 1.12383378e+00 6.07265294e-01
-1.71742642e+00 -3.83523166e-01 -1.22019231e+00 -1.32633671e-01
-3.50411922e-01 1.75141677e-01 6.72625363e-01 5.81286132e-01
-1.11149800e+00 1.18545151e+00 -9.78770554e-01 -5.56717329e-02
5.67501128e-01 8.81363511e-01 -2.68888652e-01 -2.18619183e-01
-7.99899817e-01 7.77663827e-01 9.37771142e-01 -1.47872820e-01
-4.82678622e-01 -6.63372695e-01 -8.17705393e-01 4.60208118e-01
5.56609571e-01 -5.90898931e-01 1.16497564e+00 -9.13677454e-01
-1.49562562e+00 5.58130205e-01 7.64319720e-03 -8.81305635e-01
1.95032269e-01 -2.57404178e-01 -5.46886660e-02 2.11859196e-01
-5.87825418e-01 9.96453166e-01 9.59865808e-01 -1.01380968e+00
-4.24571127e-01 -1.38696775e-01 -7.88947791e-02 1.58981666e-01
-9.05358553e-01 -8.76322761e-02 -9.09587383e-01 -9.98865306e-01
9.52090174e-02 -7.58908153e-01 -4.10324663e-01 2.23428726e-01
-3.30862522e-01 -1.97426319e-01 7.91448772e-01 -2.55065411e-01
1.49231637e+00 -2.13023424e+00 3.95717144e-01 5.10601938e-01
2.67527908e-01 5.12977540e-01 -2.53554195e-01 7.02397749e-02
-9.31880623e-02 3.36532295e-01 -5.99224925e-01 -6.44936025e-01
-1.15694173e-01 6.92670107e-01 -1.24094799e-01 3.95018637e-01
1.65291980e-01 6.17885411e-01 -4.93792027e-01 -5.00564337e-01
4.00699209e-03 5.87197721e-01 -1.09904981e+00 6.08545169e-03
-2.23118469e-01 -2.43990377e-01 -3.00244223e-02 3.84763569e-01
3.58874381e-01 -3.35705668e-01 1.44967332e-01 -2.69214511e-01
1.47021795e-02 4.27212477e-01 -1.37875104e+00 1.64454627e+00
-1.01951584e-01 6.08680606e-01 2.54032731e-01 -1.45191860e+00
8.82451415e-01 6.24063723e-02 3.85742605e-01 -2.40487710e-01
3.42620701e-01 1.34296343e-01 -1.81183785e-01 -1.18371278e-01
4.78161544e-01 7.23700151e-02 1.71421126e-01 4.03474152e-01
3.49597186e-01 -2.72490859e-01 5.28076887e-01 -9.54507291e-02
1.23387349e+00 -2.08402693e-01 1.54678702e-01 -1.95731819e-01
2.82380253e-01 -1.34503424e-01 7.09944963e-01 7.72409022e-01
2.56438076e-01 5.35759091e-01 5.01419187e-01 -4.49911416e-01
-1.26263261e+00 -6.87910676e-01 -1.36336520e-01 1.16664684e+00
-3.27566922e-01 -5.70555747e-01 -1.06755078e+00 -5.07556975e-01
9.74624828e-02 5.92298150e-01 -5.86247444e-01 -3.83687884e-01
-8.65420163e-01 -8.01564097e-01 8.37998152e-01 6.55839920e-01
5.71532190e-01 -9.82613385e-01 -5.49710214e-01 1.91198453e-01
2.54458547e-01 -8.41039240e-01 -4.72037137e-01 7.89754272e-01
-1.36619115e+00 -6.84230745e-01 -5.27411222e-01 -9.66716290e-01
9.66093898e-01 -9.93760824e-02 1.08210552e+00 2.27144614e-01
-3.50928128e-01 -1.60051763e-01 1.15266174e-01 -7.40445629e-02
-1.25336528e-01 3.88399154e-01 2.22441535e-02 -3.63491982e-01
2.46800900e-01 -7.77424753e-01 -3.40995550e-01 2.26804409e-02
-1.23130572e+00 -2.92263255e-02 6.84684098e-01 8.59879136e-01
8.83321702e-01 3.62002730e-01 2.20717415e-01 -1.17878652e+00
8.94861877e-01 -2.28459015e-01 -7.57509291e-01 5.12087680e-02
-1.13452411e+00 5.66688597e-01 8.53488684e-01 -4.77197349e-01
-6.24356866e-01 3.00110936e-01 -1.08858226e-02 -8.54829013e-01
-1.58027917e-01 6.33029461e-01 1.05937444e-01 -4.08899933e-01
6.57890260e-01 7.10996017e-02 1.16708130e-01 -6.29087508e-01
3.91058385e-01 2.32073560e-01 7.96339095e-01 -7.08680332e-01
7.37879217e-01 2.43344426e-01 5.14894053e-02 -6.00632668e-01
-7.54500151e-01 -2.91614205e-01 -6.22231245e-01 2.71157801e-01
4.07072216e-01 -4.98492032e-01 -4.46155757e-01 2.16279656e-01
-9.73492861e-01 -4.08137023e-01 -5.91446936e-01 3.38471800e-01
-6.32111013e-01 3.60796005e-01 -8.79969060e-01 -2.89016485e-01
-4.67069566e-01 -1.11729491e+00 4.27556008e-01 -8.55503231e-03
-1.45289794e-01 -7.78503895e-01 -6.92929104e-02 2.49418542e-02
5.95741272e-01 -2.33603586e-02 1.15620303e+00 -6.05897725e-01
-4.81135428e-01 -2.55555123e-01 -3.45445782e-01 8.42151403e-01
-8.20810348e-02 -1.59429833e-02 -5.41971087e-01 -4.44477946e-01
-2.41144951e-02 -5.23142040e-01 1.24732983e+00 3.98458064e-01
1.83139431e+00 -7.61817336e-01 -1.80902824e-01 1.15648150e+00
1.57120395e+00 5.07138036e-02 5.78846991e-01 2.58690298e-01
4.29936200e-01 3.59555453e-01 4.33313586e-02 4.25934196e-01
-2.29317158e-01 4.44268435e-01 4.28263664e-01 -6.75110593e-02
1.89232957e-02 -8.01917464e-02 7.42443651e-02 8.56474519e-01
-2.26709306e-01 -2.82489121e-01 -6.92113996e-01 5.33443332e-01
-1.78886652e+00 -8.87694716e-01 5.63274086e-01 2.28801870e+00
1.24768519e+00 4.77345765e-01 -1.75423041e-01 5.29990613e-01
4.81633067e-01 1.67828277e-01 -5.97039580e-01 -5.11336446e-01
7.75016025e-02 6.10871673e-01 9.70732927e-01 6.11525118e-01
-1.10727322e+00 1.02149618e+00 7.28082037e+00 8.12397301e-01
-1.10518229e+00 -1.69826657e-01 7.30801582e-01 -5.53261817e-01
-4.79297601e-02 -1.96776077e-01 -9.28454220e-01 7.96840340e-02
9.66907799e-01 -4.62381281e-02 7.99077451e-01 9.83164191e-01
-1.48710191e-01 1.63990542e-01 -1.06435549e+00 8.33330512e-01
1.51577532e-01 -1.71998608e+00 1.67802796e-01 1.83813702e-02
6.76959872e-01 9.30210724e-02 1.76685408e-01 1.06118925e-01
4.59379882e-01 -1.21320009e+00 4.57979411e-01 4.33087230e-01
8.27356398e-01 -1.14991057e+00 5.19433379e-01 3.42256933e-01
-9.81026173e-01 -1.38325319e-01 -5.39460957e-01 -9.71940532e-02
-1.81121215e-01 4.14706081e-01 -6.60529852e-01 -1.01470441e-01
6.83382571e-01 5.62779486e-01 -5.58118939e-01 1.13580406e+00
-2.06558723e-02 7.96940029e-01 -6.68374658e-01 5.60142845e-02
3.78219098e-01 -1.41464114e-01 2.91018277e-01 1.37398410e+00
2.02020019e-01 1.42195784e-02 1.70080811e-01 7.81339467e-01
-5.32311201e-01 -2.70851539e-03 -4.34101999e-01 -1.27538264e-01
5.82706511e-01 8.21655333e-01 -7.55534351e-01 -4.37285513e-01
-2.16542751e-01 8.20236325e-01 6.61193192e-01 3.40921998e-01
-6.04128540e-01 -7.86036909e-01 3.75754446e-01 1.66384950e-01
6.83569789e-01 -2.53317147e-01 -4.28164274e-01 -8.64572346e-01
5.11950962e-02 -7.91080236e-01 3.06646734e-01 -2.43549779e-01
-9.06382501e-01 5.74281335e-01 8.79938677e-02 -7.15971231e-01
-3.12573940e-01 -7.30510116e-01 -3.34919482e-01 6.58798933e-01
-1.44800007e+00 -5.99021673e-01 -1.32025585e-01 5.60704827e-01
5.40384829e-01 -2.68220663e-01 7.25511193e-01 4.82641339e-01
-7.78492153e-01 7.87530363e-01 2.25517169e-01 5.18976301e-02
5.93019307e-01 -1.10176086e+00 1.32384643e-01 7.80327082e-01
1.45257562e-01 8.93958807e-01 5.93511403e-01 -3.53518486e-01
-1.11575830e+00 -1.12862062e+00 8.41042697e-01 2.69030422e-01
4.25258666e-01 -2.73014933e-01 -1.10787630e+00 7.39754498e-01
1.03195161e-01 2.74587959e-01 6.50002956e-01 4.71445434e-02
-3.64905179e-01 -2.43503466e-01 -1.22501910e+00 4.06038672e-01
9.07508850e-01 -2.13652030e-01 -4.94941473e-01 3.74357998e-01
8.49226952e-01 -3.62569869e-01 -9.75694478e-01 5.62893689e-01
3.76988918e-01 -7.95375764e-01 1.08521152e+00 -7.94410348e-01
5.06992042e-01 -4.42669205e-02 -3.08507860e-01 -1.07938290e+00
-4.86203134e-01 -5.60390294e-01 -5.41033030e-01 9.22156096e-01
5.02196968e-01 -3.34853411e-01 1.20250869e+00 6.85910940e-01
-2.66182929e-01 -1.14106393e+00 -8.43350112e-01 -7.47321844e-01
4.67038564e-02 -4.54107970e-01 4.45040703e-01 8.59335363e-01
-1.85609281e-01 7.06537813e-02 -3.72801423e-01 -1.44875363e-01
5.35989583e-01 -2.11618185e-01 4.87881035e-01 -1.29534137e+00
-5.26194394e-01 -7.72094071e-01 -1.08000048e-01 -1.26033580e+00
1.48728088e-01 -8.83480608e-01 5.81620680e-03 -1.07959127e+00
-8.02837685e-02 -6.60598516e-01 -4.90047932e-01 9.01278317e-01
2.95232922e-01 3.62083197e-01 1.55202942e-02 2.53344893e-01
-4.03416425e-01 4.41030413e-01 7.66533971e-01 -5.89760765e-02
-3.62993389e-01 -1.29858747e-01 -6.16645873e-01 9.78224397e-01
8.45978618e-01 -5.50884068e-01 -4.60331023e-01 -7.80987442e-01
2.55717963e-01 -2.44266614e-01 1.32477105e-01 -1.28406692e+00
5.32505333e-01 -8.54094848e-02 3.55076015e-01 -3.92418116e-01
4.57168400e-01 -7.61449814e-01 -1.46822125e-01 4.56058502e-01
-7.24747121e-01 1.71954319e-01 3.39049220e-01 4.83761728e-01
-2.18473911e-01 -6.92050278e-01 1.09735537e+00 -1.52582735e-01
-3.40174317e-01 3.28203291e-01 -1.55466408e-01 2.35580076e-02
6.73425198e-01 -1.14162415e-01 -1.12684071e-01 -1.90167874e-02
-6.45718277e-01 -1.75489821e-02 5.06127656e-01 7.45834187e-02
6.72752619e-01 -1.06803358e+00 -3.40229243e-01 3.70950371e-01
-3.66087943e-01 3.07576716e-01 -3.28986049e-01 3.71168911e-01
-6.91749930e-01 4.87237364e-01 -8.24657604e-02 -3.27188551e-01
-1.47420239e+00 5.24285316e-01 3.04097205e-01 -3.68209720e-01
-6.17866635e-01 1.07435799e+00 -5.11169374e-01 -3.69270861e-01
8.63885880e-01 -4.64742929e-01 3.26140374e-02 -1.78636938e-01
3.83165210e-01 3.34393233e-01 1.70753762e-01 -2.33474433e-01
1.47836268e-01 4.61500108e-01 -3.37400556e-01 -4.90877964e-02
1.54910481e+00 3.99152517e-01 -1.45921126e-01 1.43019661e-01
1.34835613e+00 -2.22038522e-01 -1.33246708e+00 -4.61267054e-01
-1.44266009e-01 -3.20833057e-01 5.28084755e-01 -5.14506161e-01
-1.44597590e+00 7.09749758e-01 3.38171512e-01 1.34239525e-01
1.38841546e+00 -1.85970142e-01 8.52686524e-01 1.00653720e+00
2.33509298e-02 -1.37714958e+00 -3.01347896e-02 6.21703863e-01
5.44936895e-01 -7.75569022e-01 2.51659900e-01 -3.92318368e-01
-5.01500852e-02 1.14284086e+00 5.32321274e-01 -5.10230005e-01
6.65276051e-01 5.77455699e-01 -5.01565635e-01 -6.27543628e-02
-9.04005289e-01 1.49222657e-01 2.91152179e-01 2.32422620e-01
4.03094649e-01 -3.27025771e-01 -3.57642621e-01 3.06088477e-01
-3.07381988e-01 2.62421072e-01 1.50451124e-01 1.07093036e+00
-6.87507272e-01 -1.23497880e+00 -6.62167743e-02 7.97359586e-01
-6.41268730e-01 -3.00404847e-01 -1.70695156e-01 7.13074386e-01
2.00076416e-01 3.80026847e-01 2.00068966e-01 -4.35662240e-01
1.92530215e-01 2.49034941e-01 4.20364618e-01 -6.25535369e-01
-6.63102567e-01 1.02579556e-01 -2.40042452e-02 -5.32917261e-01
-2.56022543e-01 -4.62842345e-01 -1.17598450e+00 -3.47700745e-01
-3.99572641e-01 1.13329671e-01 6.82777822e-01 8.17007244e-01
4.85005856e-01 5.50850451e-01 2.07077831e-01 -8.49479258e-01
-7.63819039e-01 -7.19868600e-01 -3.61804783e-01 1.79068983e-01
7.56886527e-02 -4.89867866e-01 -4.11377639e-01 2.36114934e-01]
|
[8.517987251281738, 3.2845938205718994]
|
d02f8cd4-5fd5-4874-963d-3929569a4583
|
cryo-electron-microscopy-image-analysis-using
|
1904.07772
| null |
http://arxiv.org/abs/1904.07772v1
|
http://arxiv.org/pdf/1904.07772v1.pdf
|
Cryo-Electron Microscopy Image Analysis Using Multi-Frequency Vector Diffusion Maps
|
Cryo-electron microscopy (EM) single particle reconstruction is an entirely
general technique for 3D structure determination of macromolecular complexes.
However, because the images are taken at low electron dose, it is extremely
hard to visualize the individual particle with low contrast and high noise
level. In this paper, we propose a novel approach called multi-frequency vector
diffusion maps (MFVDM) to improve the efficiency and accuracy of cryo-EM 2D
image classification and denoising. This framework incorporates different
irreducible representations of the estimated alignment between similar images.
In addition, we propose a graph filtering scheme to denoise the images using
the eigenvalues and eigenvectors of the MFVDM matrices. Through both simulated
and publicly available real data, we demonstrate that our proposed method is
efficient and robust to noise compared with the state-of-the-art cryo-EM 2D
class averaging and image restoration algorithms.
|
['Zhizhen Zhao', 'Yifeng Fan']
|
2019-04-16
| null | null | null | null |
['cryogenic-electron-microscopy-cryo-em']
|
['computer-vision']
|
[ 3.73937309e-01 -5.38809121e-01 7.57321715e-01 -1.08986825e-01
-5.19223988e-01 -4.43716496e-01 5.61265767e-01 2.39426434e-01
-4.86073524e-01 7.16674030e-01 -1.54805392e-01 -4.16551232e-02
-1.85712248e-01 -4.79909867e-01 -5.33848464e-01 -1.14176929e+00
1.53761134e-01 5.95879138e-01 1.23561904e-01 2.69461237e-02
4.70869690e-01 8.08296144e-01 -1.20306039e+00 7.71365166e-02
7.57692516e-01 4.94254321e-01 7.74320781e-01 6.58946872e-01
-7.35787675e-02 5.13220668e-01 -4.64817196e-01 2.05727406e-02
-2.31621843e-02 -5.43154001e-01 -7.24256277e-01 3.27252924e-01
3.46308142e-01 1.29455631e-03 -2.24992454e-01 1.34300530e+00
6.61449015e-01 3.50223899e-01 7.82236576e-01 -5.31664550e-01
-6.37879848e-01 -2.79731065e-01 -7.51731038e-01 4.70013976e-01
4.47387993e-01 1.08907595e-02 3.20861340e-01 -1.01829946e+00
1.26963294e+00 1.26471305e+00 5.07007957e-01 3.78166378e-01
-1.74326754e+00 -2.69929320e-01 -1.24028720e-01 4.41423535e-01
-1.13887751e+00 -2.96567887e-01 9.23567533e-01 -5.05596638e-01
1.00870490e+00 1.75719142e-01 5.44368982e-01 6.19441390e-01
8.21655452e-01 1.76035687e-01 1.53448522e+00 -5.36042392e-01
2.98259318e-01 -3.82140726e-01 3.39973003e-01 6.33067012e-01
2.59793907e-01 -1.84476063e-01 -1.94058910e-01 -5.55005968e-01
6.68649018e-01 2.81938314e-01 -4.41356868e-01 -6.22769475e-01
-1.20448530e+00 4.41212714e-01 -2.96966713e-02 4.06813711e-01
-5.69845796e-01 -2.56908089e-01 1.84124187e-01 1.04262404e-01
4.10642803e-01 2.78910846e-01 2.00063840e-01 5.73028140e-02
-7.97625959e-01 2.78044134e-01 2.71598607e-01 4.08201069e-01
7.92259276e-01 -1.89218730e-01 4.85073328e-01 7.04653561e-01
2.86421269e-01 3.96711349e-01 3.55256885e-01 -1.10189533e+00
5.45481406e-02 4.26764578e-01 2.29026452e-01 -1.16761482e+00
-5.58994114e-01 2.46098991e-02 -1.04370081e+00 5.91231704e-01
2.72932708e-01 3.40404391e-01 -8.39549005e-01 1.43883073e+00
5.56956291e-01 1.45479187e-01 -6.51103910e-03 9.74456191e-01
6.94800377e-01 7.60450482e-01 -1.45212814e-01 -7.68172026e-01
1.16731548e+00 -5.77111661e-01 -1.06873846e+00 1.50181532e-01
2.28021488e-01 -9.56027627e-01 5.69711208e-01 3.01042199e-01
-1.11980820e+00 -5.12046278e-01 -1.21515965e+00 -3.31037268e-02
-1.14165634e-01 5.17854206e-02 3.33703369e-01 1.52322441e-01
-7.36636341e-01 1.08207786e+00 -1.30535352e+00 -3.83017927e-01
7.61372447e-02 2.83562899e-01 -9.67004061e-01 -1.18145645e-01
-5.54700613e-01 1.10249913e+00 1.50168404e-01 4.66649532e-02
-7.70992875e-01 -6.07212722e-01 -4.38949108e-01 -2.36030713e-01
-2.64110774e-01 -5.00816643e-01 4.72411305e-01 -2.11672574e-01
-1.46273422e+00 9.99855578e-01 -5.42476177e-01 -1.59199014e-01
2.14139730e-01 2.84781754e-01 -2.36672074e-01 4.85243708e-01
-6.84284568e-02 3.49144042e-02 8.55944812e-01 -1.39265430e+00
4.71573137e-02 -8.70040119e-01 -5.50090969e-01 1.02030113e-01
1.44630834e-01 9.65747088e-02 1.75502136e-01 -2.99246162e-01
7.67989576e-01 -5.83627462e-01 -3.18791509e-01 -1.44061282e-01
-1.58955932e-01 2.04644740e-01 1.19414985e+00 -8.69975269e-01
6.64109588e-01 -2.14977884e+00 7.00949967e-01 1.51911199e-01
6.05952263e-01 3.26154530e-01 8.65756944e-02 5.66210568e-01
-2.42514685e-01 -3.78092825e-01 -2.83749133e-01 -4.22068179e-01
-3.18875045e-01 1.89184129e-01 1.18223988e-01 1.09061027e+00
-8.29901695e-02 3.33913803e-01 -7.21688271e-01 -4.07590061e-01
5.57563961e-01 9.37353253e-01 -1.85336396e-01 1.59194455e-01
3.89707059e-01 1.00502467e+00 -3.20419967e-01 3.11195940e-01
1.34322202e+00 -3.16440344e-01 7.21477330e-01 -6.10621989e-01
-2.88093477e-01 2.30914019e-02 -1.33105850e+00 1.66509509e+00
6.80397451e-02 3.34173530e-01 5.96098483e-01 -1.21644616e+00
8.89345348e-01 2.07311705e-01 6.06264710e-01 -5.01680970e-01
2.32399553e-02 -3.42531800e-02 -4.51805964e-02 -2.47165397e-01
3.84855807e-01 -4.49600220e-01 5.81019759e-01 5.57610989e-01
1.91095859e-01 -7.56363645e-02 2.08917782e-01 2.93105006e-01
1.10906708e+00 6.05787076e-02 2.16661006e-01 -6.23660743e-01
7.79927909e-01 -1.28294945e-01 5.13009608e-01 4.31329459e-01
-3.53703350e-01 7.64380217e-01 1.79527983e-01 -5.88725686e-01
-1.51099360e+00 -1.15832758e+00 -4.78788555e-01 2.46653214e-01
3.26823086e-01 -4.76690263e-01 -1.06170976e+00 -1.36682138e-01
-3.31219546e-02 5.73707297e-02 -3.75913203e-01 5.65107204e-02
-5.68782210e-01 -1.29758167e+00 2.39470582e-02 -2.59009749e-02
2.78667420e-01 -6.88214421e-01 -3.82744819e-01 4.71892118e-01
-1.69865534e-01 -1.06896663e+00 -2.52650082e-01 1.05290949e-01
-1.05452883e+00 -1.18900514e+00 -6.26787245e-01 -7.05711305e-01
1.05589855e+00 6.89630389e-01 8.56603086e-01 7.12786242e-02
-5.65221608e-01 2.18022496e-01 -1.24066286e-01 3.87939334e-01
-6.68524444e-01 -6.04675353e-01 4.73768502e-01 7.87045732e-02
2.49053136e-01 -9.17225122e-01 -6.07362270e-01 3.70647788e-01
-8.80177200e-01 7.56442472e-02 2.50882775e-01 1.07987356e+00
1.20075214e+00 4.28764462e-01 8.71495679e-02 -6.94527626e-01
8.07779789e-01 4.80706319e-02 -6.24909937e-01 2.83552647e-01
-5.41252315e-01 1.38800189e-01 7.22378910e-01 -2.97405988e-01
-1.15123761e+00 1.73625857e-01 -1.28119320e-01 -3.24465662e-01
-2.61362880e-01 9.41452757e-02 -1.15025871e-01 -6.80881381e-01
4.47085708e-01 5.62634289e-01 5.12384295e-01 -8.18076670e-01
-2.73758806e-02 5.49886703e-01 6.67264163e-01 -5.38690269e-01
5.21300435e-01 9.27548528e-01 4.40758049e-01 -1.18547046e+00
-2.29214206e-01 -5.37873268e-01 -8.30400527e-01 -1.85080767e-01
8.96583796e-01 -6.17482007e-01 -9.19911683e-01 6.07132673e-01
-1.08532071e+00 2.25109905e-01 3.20432842e-01 6.26031995e-01
-6.35448456e-01 1.27530432e+00 -9.95451570e-01 -7.28433728e-01
-4.37544495e-01 -1.54478967e+00 9.02147472e-01 -1.10716633e-02
-4.11833543e-03 -9.29025412e-01 3.33131164e-01 3.54469329e-01
2.71184146e-01 4.17727113e-01 1.07207322e+00 1.22243837e-01
-4.05230343e-01 4.07696189e-03 7.02800974e-03 2.59494841e-01
2.27739498e-01 1.59477875e-01 -5.77453375e-01 -7.52491832e-01
6.06049299e-01 9.10103172e-02 7.31232643e-01 5.94604671e-01
7.30759799e-01 2.14921504e-01 -4.17102873e-01 5.83179712e-01
1.60930979e+00 2.22298548e-01 9.14204061e-01 3.88341963e-01
6.73496604e-01 5.56372225e-01 5.08866608e-01 3.86673927e-01
-1.57307550e-01 7.46965826e-01 2.24531487e-01 7.75259659e-02
-1.64508633e-02 1.57144800e-01 1.32425174e-01 1.28736937e+00
-6.18592560e-01 9.05646384e-02 -5.29267669e-01 2.07543150e-01
-1.88497603e+00 -1.23349738e+00 -4.36557651e-01 2.11160040e+00
4.67199951e-01 -1.91530496e-01 -1.76827684e-01 2.83558547e-01
1.01278043e+00 7.20603764e-02 -3.46484751e-01 -8.56336504e-02
-5.09976864e-01 2.32471377e-01 2.87593812e-01 6.26299083e-01
-9.61189747e-01 4.31859493e-01 6.99662066e+00 6.82317495e-01
-7.79494107e-01 2.47534171e-01 5.78012951e-02 2.41288066e-01
-1.55033991e-01 9.55609679e-02 -6.56894743e-01 6.48731351e-01
7.18275249e-01 -3.83105949e-02 4.15928811e-01 4.23192382e-01
3.60129237e-01 -2.81765670e-01 -6.31793797e-01 1.17723227e+00
-7.83421472e-02 -1.60871875e+00 5.72065078e-02 2.89483875e-01
4.95253265e-01 -1.47418752e-01 -2.48359233e-01 -6.69168770e-01
-7.01540634e-02 -6.99061155e-01 2.87443548e-01 7.04863310e-01
3.72134805e-01 -8.60608995e-01 6.58347428e-01 3.11607420e-01
-1.07342708e+00 4.03273076e-01 -9.76780355e-01 3.63329016e-02
5.22335112e-01 8.82435918e-01 -3.44909877e-01 5.49019217e-01
6.48986578e-01 7.20916390e-01 -2.12886706e-01 6.69665933e-01
2.59867579e-01 3.66228931e-02 -1.50109097e-01 2.49335274e-01
-3.84818506e-03 -1.02821553e+00 7.10848331e-01 9.46599066e-01
2.32189119e-01 2.94969380e-01 -4.15584864e-03 8.13426197e-01
1.34725064e-01 -7.58190453e-02 -4.79659855e-01 -8.66293162e-02
1.95780560e-01 1.43007386e+00 -9.70653296e-01 -1.81295618e-01
-3.73537660e-01 1.12772346e+00 4.35868293e-01 2.56414562e-01
-3.53301734e-01 -2.65539408e-01 8.85516942e-01 2.17179984e-01
3.38221997e-01 -6.14666343e-01 1.29884750e-01 -1.35588086e+00
-7.06086904e-02 -7.57292867e-01 4.44609225e-02 -8.55066955e-01
-1.33741093e+00 4.51087534e-01 -3.06534469e-01 -9.88994122e-01
1.49324819e-01 -8.51018727e-01 -4.55532998e-01 8.42792749e-01
-1.19730997e+00 -7.05361128e-01 -1.76601171e-01 4.89724725e-01
1.33054540e-03 1.08837405e-04 1.05188656e+00 2.97336549e-01
-4.35532779e-01 -7.21046627e-02 1.00572908e+00 -4.02916312e-01
5.65582216e-01 -1.17737329e+00 3.65007132e-01 8.59916210e-01
-2.45588049e-01 8.32671702e-01 1.05549562e+00 -7.45743573e-01
-1.83709514e+00 -5.80805123e-01 5.10803759e-01 -2.79365242e-01
4.03675824e-01 -2.95867473e-01 -1.24044752e+00 4.17655915e-01
2.88447529e-01 7.07980245e-03 6.37400329e-01 -5.18070996e-01
9.38204303e-02 3.04667056e-01 -1.65891826e+00 2.70605356e-01
7.85246432e-01 -5.65507293e-01 -6.67612910e-01 5.32183468e-01
9.27627981e-02 -2.48728901e-01 -1.43193650e+00 1.95834726e-01
4.07211572e-01 -1.07750595e+00 1.20477235e+00 -4.65972483e-01
-3.96368988e-02 -8.50397468e-01 -2.86903352e-01 -1.32331526e+00
-7.20726669e-01 -5.51308215e-01 -1.04376825e-03 7.98796892e-01
-1.28355369e-01 -5.46069562e-01 7.72315562e-01 2.20407650e-01
-1.19588226e-01 -3.88153076e-01 -1.08806443e+00 -6.12827301e-01
-1.51738450e-01 2.99995273e-01 2.74298072e-01 9.28081274e-01
6.68657497e-02 3.38302016e-01 -3.72500777e-01 2.60459930e-01
1.38572431e+00 3.61359328e-01 5.15985847e-01 -1.38942826e+00
-2.42642909e-01 1.35197699e-01 -9.38226581e-01 -7.19391048e-01
2.79816866e-01 -7.08169043e-01 -2.43578449e-01 -1.48453641e+00
6.81566775e-01 1.59989253e-01 -6.91898987e-02 -1.85325339e-01
-7.50816194e-03 2.34063491e-01 8.50166157e-02 6.08692944e-01
-5.42087257e-01 5.68574905e-01 1.48662567e+00 -1.09854594e-01
8.86586979e-02 -4.72941309e-01 -9.31101516e-02 5.12851536e-01
4.57005143e-01 -4.48451370e-01 -1.32501870e-02 -1.52165741e-01
-2.14919403e-01 -5.92130348e-02 2.43208498e-01 -1.00053382e+00
2.37299085e-01 1.22300824e-02 5.29725254e-01 -9.78884757e-01
6.09032929e-01 -6.48567617e-01 8.48656595e-01 4.87118751e-01
2.10825384e-01 2.94355243e-01 -1.10695370e-01 7.52449870e-01
-2.92027295e-01 -2.62426376e-01 1.26571333e+00 -5.82664669e-01
-3.89088959e-01 5.99766225e-02 -7.05665171e-01 -3.74245733e-01
8.44233096e-01 -1.74825042e-01 -5.90918124e-01 -1.60613611e-01
-9.56744671e-01 -1.99687198e-01 1.15034473e+00 -3.80762726e-01
1.06897783e+00 -1.29032254e+00 -3.58309478e-01 3.40355188e-01
-2.41511151e-01 -1.97443232e-01 6.82466328e-01 1.07649314e+00
-9.00120437e-01 2.09004790e-01 -7.32531548e-01 -8.10724616e-01
-1.55002189e+00 6.33414567e-01 4.00364757e-01 -2.98233926e-01
-8.84528220e-01 3.20541203e-01 5.34056649e-02 -5.83292902e-01
-4.00561392e-01 1.42162368e-01 -1.01499155e-01 -4.05107915e-01
9.07595873e-01 5.14529645e-01 3.72956783e-01 -1.05529082e+00
-3.74055862e-01 1.03176916e+00 -2.64232755e-01 2.06456140e-01
1.63046229e+00 -4.34477568e-01 -6.77757978e-01 2.06838213e-05
1.27916658e+00 -2.71720618e-01 -1.31030238e+00 9.81012881e-02
-2.85327703e-01 -5.74490666e-01 1.54479548e-01 -1.45488186e-02
-7.97601640e-01 8.88394475e-01 8.70761514e-01 1.09025687e-01
9.64565635e-01 -2.22268924e-01 7.17520773e-01 4.48115498e-01
4.68453616e-01 -1.13742864e+00 -2.96046019e-01 2.36372650e-01
4.95767981e-01 -9.24852967e-01 4.54854578e-01 -4.68522489e-01
-1.92139119e-01 1.17304528e+00 6.00749291e-02 -3.64554971e-01
5.53009391e-01 3.01218450e-01 9.02575441e-03 -4.80012894e-01
-6.35006964e-01 1.62367269e-01 -2.60928750e-01 8.47967446e-01
3.75807673e-01 -1.03884108e-01 -5.76306462e-01 4.11576517e-02
3.19859475e-01 -2.28114352e-01 7.59679735e-01 1.22434449e+00
-5.77545404e-01 -1.30148613e+00 -7.08338261e-01 2.04882458e-01
-6.32065833e-01 2.80239999e-01 -3.05265576e-01 3.64054203e-01
-3.35925013e-01 7.24066257e-01 -3.38931940e-02 -3.41502696e-01
2.54692793e-01 1.37908727e-01 9.65632737e-01 -1.70011163e-01
-7.73009211e-02 1.85454845e-01 -2.72623092e-01 -4.84657586e-01
-1.10526752e+00 -6.62136257e-01 -1.40665007e+00 -6.94598377e-01
-4.47190464e-01 4.57931250e-01 8.30531895e-01 7.76624441e-01
6.96393669e-01 2.67498702e-01 6.14747584e-01 -1.26179707e+00
-3.92066479e-01 -8.60007584e-01 -1.18714726e+00 7.97550261e-01
1.84164450e-01 -9.72531259e-01 -4.47616458e-01 1.21145315e-01]
|
[13.251298904418945, -3.0353035926818848]
|
d8dfb577-635d-404a-bac7-726b02ff2f8c
|
multi-stage-framework-with-refinement-based
| null | null |
https://openreview.net/forum?id=z7KsNClgofB
|
https://openreview.net/pdf?id=z7KsNClgofB
|
Multi-Stage Framework with Refinement based Point Set Registration for Unsupervised Bi-Lingual Word Alignment
|
Cross-lingual alignment of word embeddings play an important role in knowledge transfer across languages, for improving machine translation and other multi-lingual applications. Current unsupervised approaches rely on learning structure preserving linear transformations using adversarial networks and refinement strategies. However, such techniques, tend to suffer from instability and convergence issues, requiring tedious fine-tuning of parameter setting. This paper proposes BioSpere, a novel multi-stage framework for unsupervised mapping of bi-lingual word embeddings onto a shared vector space, by combining adversarial initialization, refinement procedure and point set registration algorithm. We show that our framework alleviates the above shortcomings, and is robust against variable adversarial learning performance and parameter choices. Experiments for parallel dictionary induction, sentence translation and word similarity demonstrate state-of-the-art results for BioSpere on diverse language pairs.
|
['Anonymous']
|
2021-11-16
| null | null | null |
acl-arr-november-2021-11
|
['word-similarity']
|
['natural-language-processing']
|
[ 2.65851747e-02 -3.99447262e-01 -4.22549963e-01 -2.98222899e-01
-1.20121908e+00 -1.03314114e+00 7.13579714e-01 6.91958666e-02
-5.71244657e-01 5.90564907e-01 3.02823544e-01 -5.33489406e-01
1.22528449e-01 -4.89308745e-01 -7.37358391e-01 -7.33682096e-01
2.54779488e-01 7.74275362e-01 -1.83859304e-01 -4.76825953e-01
-1.47968512e-02 3.76843750e-01 -6.76184952e-01 -1.57198131e-01
9.73935366e-01 2.44932696e-01 -1.18818663e-01 4.77267265e-01
-3.08389813e-01 2.89667305e-02 -2.30712458e-01 -9.65110302e-01
4.43627477e-01 -3.66313249e-01 -6.90219939e-01 -2.60853678e-01
6.61040783e-01 1.02951497e-01 -1.26830682e-01 1.38350546e+00
8.60952556e-01 3.65917832e-02 7.02434421e-01 -1.11774063e+00
-1.18684709e+00 5.58390796e-01 -4.33251947e-01 1.26109734e-01
2.19562352e-01 -3.14035127e-03 1.18328452e+00 -1.25475013e+00
5.73560596e-01 1.15477824e+00 1.09931529e+00 5.31314969e-01
-1.52032113e+00 -7.73912132e-01 -1.58455253e-01 1.74063846e-01
-1.62515581e+00 -6.28619075e-01 7.68240273e-01 -5.64838409e-01
1.00864255e+00 1.17213890e-01 2.98335582e-01 1.31577098e+00
4.38908219e-01 3.03872585e-01 1.07519495e+00 -6.90943301e-01
-7.33276457e-02 4.10689861e-01 -1.50568441e-01 7.48082817e-01
2.11374581e-01 -2.01760046e-03 -4.83877331e-01 -4.00875539e-01
8.48497093e-01 -3.79786134e-01 -1.08868472e-01 -6.46251976e-01
-1.29605091e+00 1.06243181e+00 4.03574444e-02 4.84993786e-01
-1.74091160e-01 -6.21444769e-02 4.86824930e-01 6.27518415e-01
6.48510575e-01 6.26573563e-01 -5.18226504e-01 -9.98372883e-02
-6.73108280e-01 -5.96435368e-02 4.77466226e-01 1.00329947e+00
7.39126801e-01 1.37015060e-01 1.27959371e-01 1.12667704e+00
2.52003759e-01 8.59720767e-01 8.64086449e-01 -3.84313732e-01
5.25462270e-01 1.91865966e-01 -2.14519903e-01 -1.07304275e+00
-9.01577249e-02 -2.21976474e-01 -8.52711201e-01 -1.55360430e-01
2.26283580e-01 -4.47812490e-02 -6.49031818e-01 1.89432585e+00
3.21169853e-01 1.75035566e-01 2.09937885e-01 7.58622289e-01
5.13438046e-01 4.54738766e-01 2.15734802e-02 -9.42635685e-02
1.25111449e+00 -9.60918546e-01 -7.90420890e-01 -2.76563406e-01
7.60053575e-01 -1.32631612e+00 1.32475984e+00 -4.17342260e-02
-9.48899031e-01 -5.53956866e-01 -1.00607932e+00 -1.92965895e-01
-3.27169716e-01 -2.43829727e-01 5.80004632e-01 7.01602280e-01
-9.50242877e-01 3.85667145e-01 -8.39683235e-01 -4.48544800e-01
1.41629562e-01 5.73581815e-01 -7.42918015e-01 -8.82654563e-02
-1.36012828e+00 1.29088569e+00 1.36123583e-01 -1.56010821e-01
-3.73346955e-01 -9.11239803e-01 -1.10236275e+00 -2.91579962e-01
-1.62779689e-01 -5.46575963e-01 8.55405211e-01 -9.82013822e-01
-1.79228342e+00 1.05176699e+00 -2.36082390e-01 -4.19818252e-01
2.89073706e-01 -3.61385375e-01 -5.47949016e-01 -4.36408281e-01
7.84119740e-02 3.63216162e-01 8.85879338e-01 -9.10664260e-01
-4.99175899e-02 -2.45201379e-01 -2.93862253e-01 3.91819865e-01
-7.31451690e-01 3.01451653e-01 -5.73690176e-01 -1.03516328e+00
-5.51283769e-02 -1.05497253e+00 -3.91829014e-01 -1.91197872e-01
-5.69529459e-02 -1.96052864e-01 4.90065932e-01 -7.26364195e-01
9.93215859e-01 -1.94327748e+00 5.53250909e-01 2.39351660e-01
-2.82907158e-01 4.81728852e-01 -4.81960803e-01 4.23246086e-01
-1.55007884e-01 2.75357105e-02 -3.10107201e-01 -4.77509350e-01
8.51915106e-02 3.69006693e-01 -3.96393090e-01 7.54414260e-01
8.71702060e-02 1.01576817e+00 -9.73462522e-01 -6.02444828e-01
3.22157860e-01 6.31672800e-01 -7.22758055e-01 2.92803317e-01
1.10474519e-01 5.47360778e-01 -2.46299431e-01 4.40325081e-01
5.45517147e-01 2.42687255e-01 3.32323939e-01 -3.23701411e-01
1.88441887e-01 4.64016050e-01 -1.08633482e+00 2.25979710e+00
-7.40856469e-01 5.24557531e-01 -6.21678643e-02 -1.05315435e+00
9.86075163e-01 4.30258304e-01 4.27456349e-01 -5.79933763e-01
2.08733901e-01 4.38595682e-01 5.39929187e-03 -2.69455731e-01
5.59583843e-01 -3.65096331e-01 -3.17126691e-01 6.67714357e-01
4.17133749e-01 -3.88417095e-01 -3.21499288e-01 -7.96990376e-03
6.09593928e-01 2.56152123e-01 4.30671066e-01 -5.20261049e-01
7.12700188e-01 -7.60177299e-02 5.83174765e-01 1.77331954e-01
-1.14970721e-01 5.36130488e-01 -5.64428717e-02 -3.85357052e-01
-1.42271650e+00 -1.09587371e+00 -2.51557797e-01 1.19753599e+00
1.57134794e-02 -3.33804280e-01 -7.61598289e-01 -7.27387488e-01
8.32365230e-02 5.75878441e-01 -3.83088410e-01 -2.52271712e-01
-1.00707793e+00 -7.62462020e-01 8.99870098e-01 3.81989568e-01
-7.91452900e-02 -8.70813727e-01 4.49185103e-01 3.62878740e-01
-2.55663782e-01 -1.41608095e+00 -1.15702176e+00 -1.11205736e-02
-7.33130217e-01 -8.74982834e-01 -6.47218764e-01 -1.29847884e+00
8.35242569e-01 1.84653893e-01 1.27337301e+00 -1.96330264e-01
-1.29096821e-01 3.37320983e-01 -1.06080040e-01 -1.41116351e-01
-7.78355479e-01 3.43964607e-01 8.59834909e-01 5.14801266e-03
6.07162774e-01 -7.65848696e-01 -2.39599273e-01 4.41516638e-01
-6.87832892e-01 -2.55542815e-01 3.63404840e-01 1.06829929e+00
8.94069910e-01 -4.67097282e-01 5.07902622e-01 -9.08165276e-01
8.90045047e-01 -2.95161545e-01 -6.85377836e-01 2.61451930e-01
-7.10330486e-01 3.12465489e-01 8.25578690e-01 -6.73187852e-01
-6.62563026e-01 -1.09257936e-01 -2.42721528e-01 -5.10462761e-01
1.68627083e-01 2.62969226e-01 -2.80041367e-01 -4.34018672e-01
7.97748506e-01 4.09233898e-01 8.79820287e-02 -4.92593169e-01
8.91551673e-01 8.06535959e-01 7.71600187e-01 -8.27480376e-01
1.36926401e+00 6.55564740e-02 -3.94561499e-01 -6.16537094e-01
-4.79697227e-01 -4.78470445e-01 -1.08004558e+00 3.73228490e-01
8.21084261e-01 -1.08647096e+00 -6.24748394e-02 3.90570819e-01
-1.19281399e+00 -1.50938109e-01 -9.81851444e-02 6.87260270e-01
-5.45361757e-01 4.41637248e-01 -4.31777149e-01 -9.52036679e-02
-7.01950252e-01 -1.16310251e+00 7.92648196e-01 -2.48344153e-01
-4.61426854e-01 -1.56701899e+00 6.52047396e-01 3.70188177e-01
5.02953351e-01 -3.10450047e-02 9.05701399e-01 -8.11313391e-01
-2.55065203e-01 -2.00980827e-01 2.67890036e-01 6.40720904e-01
4.84023154e-01 -2.79555500e-01 -5.76418936e-01 -6.34197533e-01
-1.21363796e-01 -2.83956051e-01 1.63150996e-01 -8.64517316e-02
6.82872653e-01 -3.16776812e-01 -9.55027863e-02 1.07811463e+00
1.37993455e+00 -2.57093757e-01 4.33315635e-01 4.57525641e-01
1.08924496e+00 1.92770630e-01 4.43410367e-01 -1.00618452e-01
3.33570629e-01 9.54410851e-01 -4.67723049e-02 -2.99895078e-01
-2.10597768e-01 -2.03105390e-01 6.82591021e-01 1.73854375e+00
4.83261747e-03 1.60478413e-01 -9.58281398e-01 9.11702216e-01
-1.54110575e+00 -7.58873284e-01 2.37372383e-01 2.22741032e+00
1.37757933e+00 -1.53408527e-01 -1.10246807e-01 -1.40896514e-01
7.44009376e-01 1.89515129e-01 -2.89293736e-01 -8.36824119e-01
-2.24842519e-01 7.17720389e-01 7.69158423e-01 8.43550980e-01
-9.94262218e-01 1.66235566e+00 6.45604181e+00 8.69727969e-01
-1.07034469e+00 5.38703024e-01 -2.07642000e-02 3.29026222e-01
-5.92557549e-01 -2.43619621e-01 -7.39660263e-01 1.66734472e-01
7.52443433e-01 -3.84607106e-01 6.50307834e-01 5.54159999e-01
-1.33868009e-01 7.13527381e-01 -1.18917990e+00 8.97178173e-01
3.94193769e-01 -1.27155364e+00 4.27031927e-02 -1.57756403e-01
1.09178579e+00 2.68664449e-01 8.80272985e-02 1.69265717e-01
6.50155723e-01 -1.09763479e+00 2.58375436e-01 -1.94326621e-02
1.21152532e+00 -8.48404944e-01 6.86739802e-01 2.46461201e-02
-1.14791012e+00 6.01301193e-01 -3.44324678e-01 2.41896510e-01
1.10260077e-01 1.80365220e-01 -7.27096677e-01 4.47327882e-01
3.26382309e-01 6.79933965e-01 -2.80470997e-01 4.17841434e-01
-3.78876537e-01 5.14490187e-01 -2.50477880e-01 1.55251071e-01
1.87731773e-01 -4.93364304e-01 6.35476708e-01 1.40629923e+00
2.29298785e-01 -2.70120382e-01 8.84668753e-02 5.36763012e-01
-2.38784537e-01 5.34567952e-01 -8.03650558e-01 -2.85410602e-02
6.46274090e-01 1.05320215e+00 -1.79567859e-01 -1.25902936e-01
-5.46079040e-01 1.27087247e+00 5.69713831e-01 2.86659718e-01
-9.08152699e-01 -3.48614365e-01 1.38762403e+00 -1.06042661e-01
2.56967485e-01 -6.72091067e-01 -3.45055908e-01 -1.45039999e+00
9.23497677e-02 -1.38756263e+00 8.67130384e-02 -1.21985242e-01
-1.49142110e+00 6.35917783e-01 -2.90453672e-01 -1.29294920e+00
-2.90217578e-01 -6.38735235e-01 -6.07509494e-01 1.08310735e+00
-1.44421506e+00 -1.29182792e+00 2.88358510e-01 9.37894762e-01
5.72716296e-01 -6.74344599e-01 1.45217478e+00 5.59855163e-01
-4.37648267e-01 1.44111311e+00 4.64678675e-01 2.74942666e-01
1.26177788e+00 -1.08080316e+00 8.90029550e-01 9.55677450e-01
6.10432327e-01 8.22611213e-01 6.37361526e-01 -3.18901688e-01
-1.70618415e+00 -1.31664872e+00 1.28759634e+00 -7.04849184e-01
1.13445938e+00 -5.39662719e-01 -1.01439977e+00 8.74653637e-01
3.57558846e-01 1.63434297e-01 1.02233577e+00 3.69716585e-01
-6.96502328e-01 -3.16235982e-02 -9.89892960e-01 8.82437050e-01
9.56252694e-01 -1.02612877e+00 -8.29860449e-01 5.93058169e-01
7.16480136e-01 -4.97696638e-01 -1.36573827e+00 1.81044370e-01
5.02996624e-01 -2.52025187e-01 1.27788532e+00 -8.41089070e-01
1.94594443e-01 -2.07141429e-01 -2.09370285e-01 -1.65807331e+00
-4.17468399e-01 -1.05011547e+00 8.58002976e-02 1.37076771e+00
4.66059744e-01 -8.17861378e-01 4.03255612e-01 1.58232719e-01
-1.53204128e-01 -5.58294296e-01 -1.04564261e+00 -8.65838468e-01
6.77252293e-01 -3.28444093e-01 6.58011913e-01 1.59899867e+00
-1.09331861e-01 5.50678790e-01 -6.21200681e-01 3.15739006e-01
6.92856491e-01 -1.38748169e-01 1.09772801e+00 -7.41325438e-01
-2.61572391e-01 -4.28524613e-01 -4.79087591e-01 -8.22996974e-01
5.89950144e-01 -1.32214177e+00 3.22730914e-02 -1.13186848e+00
-2.93041207e-03 -6.78741097e-01 -4.89894807e-01 4.89218265e-01
-4.22111958e-01 6.82490289e-01 -8.13013241e-02 4.04708594e-01
-2.47194082e-01 5.53933918e-01 1.07026911e+00 -1.96727321e-01
-1.34507775e-01 -3.70798290e-01 -5.08626163e-01 7.08765864e-01
8.68533850e-01 -6.82192624e-01 -2.85584092e-01 -9.17935729e-01
1.05705541e-02 -5.04387736e-01 -4.89003733e-02 -5.20992160e-01
1.54833004e-01 -2.25021064e-01 5.65996161e-03 -5.97287454e-02
1.23095073e-01 -7.57991493e-01 9.14502982e-03 3.00333887e-01
-2.28612766e-01 5.11476755e-01 3.04099768e-01 3.32996309e-01
-2.23464400e-01 -7.97354057e-02 7.77403712e-01 2.64131546e-01
-4.63206470e-01 4.40998852e-01 -1.13779739e-01 3.66926491e-01
7.08578825e-01 -7.74710849e-02 2.90651601e-02 5.96090257e-02
-4.81310487e-01 6.68476969e-02 5.19657075e-01 9.92734671e-01
1.93583637e-01 -1.92879999e+00 -1.01625836e+00 5.61839402e-01
2.14482367e-01 -2.59055436e-01 -3.12305212e-01 7.11687088e-01
-4.81281340e-01 2.30884597e-01 -1.50440186e-01 -4.90651041e-01
-1.38468075e+00 4.07177866e-01 2.49409884e-01 -4.36378568e-01
-5.16267300e-01 9.66324747e-01 -7.51407519e-02 -9.99129951e-01
1.00154787e-01 1.07767187e-01 7.90799782e-02 -1.23327166e-01
1.35025188e-01 2.94635864e-03 1.45541310e-01 -1.08843434e+00
-5.25525928e-01 1.20969069e+00 -1.95169315e-01 -1.48953006e-01
1.17816651e+00 -1.63488910e-01 -1.86832592e-01 2.72047758e-01
1.44171417e+00 4.37229931e-01 -5.70185006e-01 -8.85977149e-01
-1.56151891e-01 -3.47915322e-01 -5.96269732e-03 -2.75455296e-01
-8.83559525e-01 7.60283053e-01 6.18947268e-01 -4.08156186e-01
7.68190563e-01 -2.51454026e-01 1.13596070e+00 3.98684233e-01
4.37631458e-01 -9.78060484e-01 -1.75333425e-01 7.48336792e-01
8.42523217e-01 -1.43582332e+00 1.01906836e-01 -2.64739782e-01
-4.84517097e-01 9.17646706e-01 3.79975617e-01 -3.81144851e-01
5.23370028e-01 2.26484999e-01 6.04404688e-01 2.03892857e-01
-2.62069017e-01 1.43319249e-01 6.73862934e-01 7.34344363e-01
6.55432582e-01 2.38244042e-01 -4.95197237e-01 4.11353201e-01
-5.11878610e-01 -5.16089737e-01 -8.38297009e-02 6.08488798e-01
2.70309877e-02 -1.85862839e+00 -4.63281691e-01 -2.68718004e-01
-5.99837065e-01 -6.78026736e-01 -2.26115391e-01 7.78367817e-01
7.62678608e-02 4.75929588e-01 1.05968654e-01 -4.60655570e-01
2.57454067e-01 2.63540030e-01 7.03006148e-01 -5.01502514e-01
-7.20512211e-01 6.19627163e-03 -1.91694461e-02 -3.07606339e-01
-3.03692520e-01 -5.88624060e-01 -9.50915456e-01 -3.63700122e-01
-3.86563212e-01 2.60557145e-01 7.62492836e-01 9.08651531e-01
5.01625717e-01 2.50355959e-01 7.75305331e-01 -6.87125802e-01
-7.86804676e-01 -1.01179743e+00 -8.48520175e-02 7.19487965e-01
1.69569597e-01 -4.29335415e-01 -1.75975010e-01 3.10696900e-01]
|
[11.112120628356934, 10.105268478393555]
|
d3777594-b1b2-4ad1-abda-cbc1172238ab
|
assessing-dialogue-systems-with-distribution
|
2105.02573
| null |
https://arxiv.org/abs/2105.02573v3
|
https://arxiv.org/pdf/2105.02573v3.pdf
|
Assessing Dialogue Systems with Distribution Distances
|
An important aspect of developing dialogue systems is how to evaluate and compare the performance of different systems. Existing automatic evaluation metrics are based on turn-level quality evaluation and use average scores for system-level comparison. In this paper, we propose to measure the performance of a dialogue system by computing the distribution-wise distance between its generated conversations and real-world conversations. Specifically, two distribution-wise metrics, FBD and PRD, are developed and evaluated. Experiments on several dialogue corpora show that our proposed metrics correlate better with human judgments than existing metrics.
|
['Lemao Liu', 'Defu Lian', 'Huayang Li', 'Deng Cai', 'Yahui Liu', 'Jiannan Xiang']
|
2021-05-06
| null |
https://aclanthology.org/2021.findings-acl.193
|
https://aclanthology.org/2021.findings-acl.193.pdf
|
findings-acl-2021-8
|
['dialogue-evaluation']
|
['natural-language-processing']
|
[-3.25559109e-01 1.50894597e-01 8.52463916e-02 -6.06260896e-01
-8.69603038e-01 -9.23719704e-01 1.14890265e+00 4.60711449e-01
-3.51440430e-01 9.95876431e-01 7.33494878e-01 -1.82718590e-01
-6.96602911e-02 -5.90713799e-01 4.55136836e-01 -1.65217608e-01
2.14223817e-01 6.21482849e-01 3.51693362e-01 -7.75307238e-01
6.55046940e-01 1.62113294e-01 -1.14585292e+00 3.35726649e-01
1.12788129e+00 7.10224807e-01 -8.37927125e-03 1.16857040e+00
-4.62239653e-01 9.41234112e-01 -1.36158586e+00 -5.82421601e-01
-1.45474985e-01 -9.45010781e-01 -1.26356208e+00 -9.83682945e-02
4.04879451e-03 -2.48182878e-01 -1.16275422e-01 8.93181264e-01
4.19363022e-01 3.12150210e-01 8.69019628e-01 -1.31610119e+00
-2.54760712e-01 7.92003095e-01 2.21612290e-01 2.50685960e-01
1.18849432e+00 1.44892424e-01 1.09640002e+00 -5.37918448e-01
5.20212531e-01 1.55034065e+00 3.47848415e-01 5.58365464e-01
-9.23873961e-01 -1.72658667e-01 -3.29358578e-01 3.22568305e-02
-7.95024097e-01 -5.57301581e-01 4.27337259e-01 -5.22803485e-01
9.82681274e-01 4.57942516e-01 2.81855017e-01 8.16232026e-01
-1.53984707e-02 7.33960152e-01 1.15761173e+00 -5.22769570e-01
1.00579150e-01 5.61731100e-01 5.25319457e-01 5.51304519e-01
-3.12687188e-01 -4.15222019e-01 -4.02722061e-01 -3.22528154e-01
2.27379873e-01 -6.71770573e-01 -4.55351681e-01 8.48283619e-02
-1.11223972e+00 8.52378547e-01 -1.67567760e-01 7.95294464e-01
-2.19482258e-01 -5.99733651e-01 7.11957097e-01 4.90688026e-01
5.02542913e-01 8.22586358e-01 -3.07483017e-01 -1.11012495e+00
-5.83580315e-01 4.49260801e-01 1.65489197e+00 7.71763802e-01
3.48598272e-01 -4.07492161e-01 -6.91074550e-01 1.23959410e+00
1.46829531e-01 1.97169438e-01 7.00257719e-01 -1.12343979e+00
5.45825541e-01 7.89152980e-01 4.25970078e-01 -1.05294716e+00
-4.10737008e-01 4.58303988e-01 -4.69340801e-01 -1.12840459e-01
7.26502538e-01 -4.58963364e-01 -7.21095083e-03 1.47956526e+00
1.89310223e-01 -7.03135610e-01 4.47437555e-01 6.57243192e-01
1.13443685e+00 8.54209960e-01 -2.17274979e-01 -5.25453687e-01
8.80458057e-01 -1.25453293e+00 -1.14906144e+00 3.73569310e-01
7.03704834e-01 -1.23556495e+00 1.24454939e+00 2.62965411e-01
-1.27355766e+00 -5.80297768e-01 -8.67034018e-01 2.60575980e-01
-2.27998972e-01 -5.12422808e-02 6.76881224e-02 9.05646861e-01
-1.14618242e+00 4.64602768e-01 -8.53506550e-02 -5.67833006e-01
-6.82510912e-01 -1.51705578e-01 -1.86037019e-01 5.09411454e-01
-1.38199925e+00 1.42117035e+00 1.55601829e-01 -2.28249714e-01
-4.71724898e-01 -1.04391634e-01 -5.66779196e-01 -1.17561653e-01
5.54064438e-02 -1.75686449e-01 1.93647742e+00 -5.42155206e-01
-2.25354362e+00 6.68830037e-01 2.30048131e-02 -2.56687760e-01
6.69271767e-01 -8.40053931e-02 -5.15755773e-01 -4.25776746e-03
-6.90558553e-02 2.08820581e-01 1.21758297e-01 -1.18298423e+00
-9.08925653e-01 1.64452791e-01 5.90690494e-01 5.50743759e-01
-3.86202514e-01 9.14315209e-02 -2.55165994e-01 -1.22730508e-01
-3.45524818e-01 -7.59728730e-01 6.31987303e-02 -5.45612872e-01
-3.88916880e-01 -8.65760505e-01 6.26769841e-01 -6.94180608e-01
1.48381436e+00 -1.83839393e+00 6.25973791e-02 2.97153462e-02
1.65881962e-01 6.22494221e-01 -1.21433817e-01 8.54448438e-01
5.51027119e-01 6.53919429e-02 7.70338774e-02 -7.32909814e-02
3.20675433e-01 -2.99824905e-02 1.12882458e-01 -2.52336320e-02
-1.13896504e-01 3.56237561e-01 -1.12941015e+00 -5.60396731e-01
3.15263838e-01 5.72269075e-02 -3.15780729e-01 7.87881315e-01
-2.49362126e-01 3.22562218e-01 -2.05983669e-01 1.85155664e-02
2.21210510e-01 1.35479540e-01 4.64163631e-01 -7.41614774e-02
-3.22419494e-01 8.89517784e-01 -7.22580075e-01 1.44542658e+00
-8.01903844e-01 9.02614892e-01 -1.26478866e-01 -4.75677550e-01
1.23670793e+00 5.12381494e-01 2.94600315e-02 -6.69254422e-01
3.27291429e-01 6.95614070e-02 4.18966293e-01 -5.44551969e-01
1.16900337e+00 1.55068710e-01 -3.73234063e-01 9.37170029e-01
1.85524479e-01 -3.83616209e-01 7.26398468e-01 4.85673010e-01
1.03829992e+00 -4.76716518e-01 5.26922107e-01 -3.13014448e-01
9.40003276e-01 -1.27939776e-01 1.60004541e-01 6.35424197e-01
-5.98861039e-01 2.72891015e-01 1.08330500e+00 1.80653334e-02
-8.76700819e-01 -7.87527680e-01 -2.41914927e-03 1.15581024e+00
3.02744322e-02 -6.85750127e-01 -1.22668839e+00 -9.67168868e-01
-2.84489363e-01 9.54491735e-01 -4.10202444e-01 5.31763360e-02
-3.89943749e-01 -3.46609592e-01 8.05655360e-01 8.34421962e-02
5.92470825e-01 -9.12054360e-01 -2.97294855e-01 4.33676720e-01
-8.16975534e-01 -1.06166661e+00 -5.26053488e-01 -4.73397046e-01
-2.25283489e-01 -1.10467947e+00 -7.14369655e-01 -4.92859781e-01
1.62499268e-02 3.29910845e-01 1.37424219e+00 1.82983384e-01
1.67430669e-01 3.71173859e-01 -8.32851946e-01 -1.15538634e-01
-1.17676675e+00 2.93701798e-01 5.30587956e-02 -4.35362577e-01
2.59256512e-01 -1.45825654e-01 -3.10325325e-01 7.61038363e-01
-4.56937045e-01 -1.15533583e-01 2.13392153e-01 7.76419103e-01
-4.28855836e-01 -2.14229822e-01 7.97096133e-01 -8.52647305e-01
1.90721381e+00 -2.22001478e-01 -7.99086764e-02 6.40288234e-01
-5.52114606e-01 1.40771151e-01 4.49557096e-01 -4.40842845e-02
-1.20920813e+00 -7.41824806e-01 -2.62992620e-01 3.59410971e-01
-1.91378430e-01 3.84176314e-01 -1.78301528e-01 2.75113761e-01
7.72789657e-01 -6.08889572e-02 1.08243585e-01 -1.50967479e-01
4.66164649e-01 1.23601830e+00 1.50738150e-01 -6.66277885e-01
3.91600654e-02 -4.02645260e-01 -7.91325867e-01 -8.98888469e-01
-7.33270764e-01 -6.21255457e-01 -6.01040661e-01 -8.10180247e-01
6.04131401e-01 -3.92749965e-01 -9.09584880e-01 3.48530978e-01
-1.40876210e+00 -4.02251661e-01 7.87324905e-02 5.11771262e-01
-6.32995129e-01 4.29750681e-01 -6.57730222e-01 -1.14198232e+00
-4.49386269e-01 -1.14132726e+00 7.22775161e-01 3.58582795e-01
-8.59573364e-01 -1.20541704e+00 6.98538899e-01 3.81605566e-01
4.81834620e-01 -5.85183688e-02 7.56741047e-01 -8.67944777e-01
3.08087260e-01 -2.53295749e-01 -2.36726999e-01 5.45140564e-01
4.13518399e-01 4.63243783e-01 -8.05691421e-01 1.36114255e-01
-5.61964214e-02 -4.60311264e-01 1.81250960e-01 -1.17117591e-01
5.90953827e-01 -5.36885977e-01 3.09694000e-02 -3.95741254e-01
7.58151591e-01 4.74470109e-01 5.25792480e-01 7.86481351e-02
1.98613480e-01 9.59022641e-01 1.03163469e+00 5.68955719e-01
5.68871796e-01 7.69370615e-01 -1.74358606e-01 3.90349805e-01
-6.59655780e-03 -9.05053988e-02 4.07139629e-01 1.60069716e+00
-4.87129875e-02 -7.02972531e-01 -1.02832305e+00 3.57657760e-01
-1.92871964e+00 -9.26328778e-01 -2.04962820e-01 1.84811461e+00
1.03594708e+00 3.28907490e-01 7.56112218e-01 3.04469883e-01
8.02555501e-01 2.28166342e-01 8.96534324e-02 -9.28047597e-01
9.26400274e-02 -1.10050432e-01 -3.29482019e-01 9.28831220e-01
-6.84471726e-01 7.53651202e-01 7.38553524e+00 5.71871877e-01
-6.70118332e-01 5.05306460e-02 4.10934836e-01 8.52367356e-02
-1.72372237e-01 -2.89104491e-01 -5.08844018e-01 4.75976259e-01
1.27314079e+00 -9.30752754e-01 9.16336775e-02 6.97752953e-01
1.79992840e-01 -2.16197148e-01 -1.15923572e+00 7.22481191e-01
1.53237388e-01 -9.81118619e-01 -1.35071859e-01 -1.32743508e-01
5.16351044e-01 -5.26571870e-01 -2.47560099e-01 5.37737429e-01
7.16809094e-01 -6.00067675e-01 4.59096491e-01 6.92451477e-01
2.85859585e-01 -8.63624990e-01 1.02367032e+00 3.61351222e-01
-8.41070294e-01 4.95304376e-01 -2.13633507e-01 -1.31160319e-01
2.65306532e-01 3.93206149e-01 -1.43579459e+00 5.05003095e-01
2.08005100e-01 2.60677099e-01 -2.57910937e-01 8.56505811e-01
-2.77881533e-01 3.72164726e-01 8.26532990e-02 -9.50199544e-01
3.33041936e-01 -2.24397495e-01 4.49782521e-01 1.68837678e+00
8.57861191e-02 1.50731936e-01 1.22594513e-01 4.37292725e-01
-5.53313419e-02 5.16336381e-01 -6.51772320e-01 -1.28927529e-01
7.10344017e-01 1.44362175e+00 -2.95743197e-01 -4.04571831e-01
-1.15484938e-01 8.29510808e-01 3.49553883e-01 -2.24888831e-01
-7.18961418e-01 -6.95940852e-01 7.83246934e-01 -3.84021282e-01
-4.27154958e-01 -3.41028750e-01 -9.19729844e-02 -7.30708718e-01
-1.03579007e-01 -1.13131690e+00 1.46472797e-01 -3.09390754e-01
-1.37617683e+00 1.09853482e+00 -2.53683627e-02 -1.26657081e+00
-6.76341832e-01 -3.98026675e-01 -8.20980906e-01 7.89331019e-01
-7.98282325e-01 -3.33448321e-01 -4.47065771e-01 3.42577189e-01
7.51846135e-01 -4.80874389e-01 1.13467574e+00 1.30831987e-01
-5.77222407e-01 7.93702662e-01 8.76386464e-03 2.58926183e-01
8.96462083e-01 -1.43987870e+00 2.22931176e-01 2.48690441e-01
-1.22641668e-01 4.12031591e-01 1.02385306e+00 -2.01079935e-01
-6.92532539e-01 -6.26972616e-01 1.04767823e+00 -5.60072780e-01
7.79573262e-01 -6.23399206e-02 -8.05505216e-01 -6.99337048e-04
7.41114914e-01 -1.01112115e+00 1.05027211e+00 5.03203213e-01
-2.80231416e-01 1.46277264e-01 -1.27627742e+00 4.05424386e-01
7.06118405e-01 -6.47064865e-01 -8.90779078e-01 4.74444866e-01
5.24639606e-01 -4.18349534e-01 -1.13887596e+00 1.53178528e-01
7.14969635e-01 -1.33284771e+00 3.29359204e-01 -5.12531638e-01
4.70157415e-01 5.57774715e-02 -1.89727575e-01 -1.87436664e+00
8.80273953e-02 -7.31614053e-01 1.28296569e-01 1.52952802e+00
7.24643826e-01 -4.77325380e-01 1.82882100e-01 6.98706508e-01
9.14584771e-02 -4.42524642e-01 -5.17615855e-01 -7.21363246e-01
1.62193090e-01 4.15729173e-02 6.39128089e-01 9.72616196e-01
9.18765187e-01 9.35993731e-01 -2.36177459e-01 -3.74895245e-01
1.54545277e-01 -7.40237385e-02 1.21371460e+00 -1.00702691e+00
-5.92080690e-02 -8.75816464e-01 -4.56215769e-01 -9.86094773e-01
2.59616286e-01 -3.41277450e-01 4.53920275e-01 -1.62270927e+00
-1.86413806e-02 -2.52391040e-01 7.18586966e-02 -1.81346074e-01
-3.43521506e-01 -2.26225644e-01 3.66939753e-01 1.54640719e-01
-8.88681531e-01 7.39746749e-01 1.10658824e+00 -6.18947521e-02
-2.88592100e-01 4.17564809e-02 -3.97708178e-01 6.70884490e-01
1.08501720e+00 -1.29939094e-01 -3.02260935e-01 -2.01587096e-01
-1.74235463e-01 3.85726243e-01 -4.25082684e-01 -1.12465632e+00
1.62064880e-01 -1.71615019e-01 -5.07721603e-01 -5.22442937e-01
2.16813326e-01 -1.71320230e-01 -4.77893561e-01 2.52070278e-01
-7.94116557e-01 2.72983640e-01 -2.44932011e-01 1.95377335e-01
-6.90305173e-01 -5.17658293e-01 7.72047460e-01 -2.01164670e-02
-4.09140140e-01 -3.76455873e-01 -7.04513669e-01 2.65195340e-01
9.65835571e-01 2.17482254e-01 -6.22628212e-01 -1.02474749e+00
-4.41819012e-01 2.76667088e-01 2.19575077e-01 6.52256846e-01
3.21851075e-01 -1.16398704e+00 -8.16981733e-01 -5.16190529e-01
2.61655331e-01 -7.32393265e-01 -1.72836989e-01 5.11955678e-01
-7.03401685e-01 6.82534218e-01 -3.15321386e-01 -4.90757495e-01
-1.58965600e+00 -1.94604278e-01 4.12871450e-01 -5.77547848e-01
1.22996464e-01 7.47269094e-01 -1.84515581e-01 -9.65869069e-01
3.71322960e-01 -1.63411349e-01 -5.39112747e-01 2.45810747e-01
7.33095050e-01 6.37948036e-01 1.38406038e-01 -6.25945270e-01
-1.50741652e-01 -9.98879820e-02 -1.16979264e-01 -6.13020241e-01
7.88659871e-01 -1.80085331e-01 -1.34384245e-01 8.99871945e-01
1.10348046e+00 -9.34440345e-02 -5.65045357e-01 -3.10802966e-01
2.48731777e-01 -5.21254182e-01 -3.50608885e-01 -9.73754704e-01
-4.24951047e-01 7.94917762e-01 4.08341944e-01 1.18002141e+00
6.25364244e-01 -2.17941418e-01 6.38890803e-01 7.29947627e-01
3.73238027e-01 -1.47852731e+00 4.16668564e-01 1.13718116e+00
1.07179368e+00 -1.15934217e+00 -2.70050853e-01 -3.70360613e-01
-1.01396406e+00 1.12798142e+00 8.71576965e-01 2.53350019e-01
5.21045327e-01 2.60690488e-02 4.75314945e-01 7.96032995e-02
-1.07103312e+00 -3.27183515e-01 2.37292796e-01 4.21453148e-01
1.14294314e+00 2.92226404e-01 -9.04920876e-01 3.11667413e-01
-7.55354345e-01 -3.21227401e-01 6.64962769e-01 8.46697927e-01
-7.69576848e-01 -1.38506579e+00 -4.09779474e-02 3.25870067e-01
-1.83481112e-01 3.20863247e-01 -1.11471534e+00 6.69989884e-01
-6.27309918e-01 1.67635846e+00 -2.96107996e-02 -7.27929652e-01
7.27642894e-01 2.67692655e-01 5.15459657e-01 -6.80555046e-01
-1.01449561e+00 -4.70718443e-01 9.81332779e-01 -1.54927865e-01
-7.07002640e-01 -5.57226002e-01 -8.41826499e-01 -7.27386475e-01
-6.39432490e-01 8.55752468e-01 6.75160766e-01 1.00845563e+00
2.64128875e-02 4.97132927e-01 1.12885177e+00 -4.41716731e-01
-6.38522267e-01 -1.63124824e+00 -3.45646054e-01 5.03933609e-01
-2.90878508e-02 -5.46052575e-01 -3.57951581e-01 -3.66016477e-01]
|
[12.856679916381836, 8.161293983459473]
|
b80779e9-59f9-4b58-b080-9ab67fde9f45
|
distinguishability-calibration-to-in-context
|
2302.06198
| null |
https://arxiv.org/abs/2302.06198v3
|
https://arxiv.org/pdf/2302.06198v3.pdf
|
Distinguishability Calibration to In-Context Learning
|
Recent years have witnessed increasing interests in prompt-based learning in which models can be trained on only a few annotated instances, making them suitable in low-resource settings. When using prompt-based learning for text classification, the goal is to use a pre-trained language model (PLM) to predict a missing token in a pre-defined template given an input text, which can be mapped to a class label. However, PLMs built on the transformer architecture tend to generate similar output embeddings, making it difficult to discriminate between different class labels. The problem is further exacerbated when dealing with classification tasks involving many fine-grained class labels. In this work, we alleviate this information diffusion issue, i.e., different tokens share a large proportion of similar information after going through stacked multiple self-attention layers in a transformer, by proposing a calibration method built on feature transformations through rotation and scaling to map a PLM-encoded embedding into a new metric space to guarantee the distinguishability of the resulting embeddings. Furthermore, we take the advantage of hyperbolic embeddings to capture the hierarchical relations among fine-grained class-associated token embedding by a coarse-to-fine metric learning strategy to enhance the distinguishability of the learned output embeddings. Extensive experiments on the three datasets under various settings demonstrate the effectiveness of our approach. Our code can be found at https://github.com/donttal/TARA.
|
['Lin Gui', 'Yulan He', 'Li Qian', 'Yanran Li', 'Hanqi Yan', 'Hongjing Li']
|
2023-02-13
| null | null | null | null |
['metric-learning', 'metric-learning']
|
['computer-vision', 'methodology']
|
[ 1.23551689e-01 -2.47101132e-02 -2.20767468e-01 -5.24672925e-01
-7.97835112e-01 -6.99424863e-01 7.69062757e-01 5.54636538e-01
-5.21413267e-01 3.01668286e-01 2.91257918e-01 -1.95762828e-01
-1.75838873e-01 -8.12782705e-01 -4.12576735e-01 -6.80044055e-01
3.94284278e-01 4.54467773e-01 -2.40912903e-02 1.38425663e-01
2.94830114e-01 3.22466910e-01 -1.35897446e+00 4.14707780e-01
8.89189720e-01 1.04761887e+00 2.59582639e-01 3.91696244e-01
-5.14557719e-01 5.24183214e-01 -5.29086590e-01 -4.92751747e-01
2.12444782e-01 -3.30134064e-01 -1.04618311e+00 -2.39343550e-02
3.32310557e-01 -1.37504518e-01 -2.96356082e-01 1.06061399e+00
3.48618418e-01 3.08735460e-01 8.60533714e-01 -1.41362977e+00
-7.94492841e-01 6.63555562e-01 -3.00736487e-01 1.57920092e-01
2.36497566e-01 -7.75789171e-02 1.48484659e+00 -1.31579912e+00
2.27899000e-01 1.25072730e+00 4.93944526e-01 4.08144921e-01
-1.29805803e+00 -7.40090132e-01 1.85648978e-01 3.26101571e-01
-1.51721811e+00 -2.80595452e-01 6.86228335e-01 -3.86211962e-01
8.80928755e-01 1.88898951e-01 3.16804171e-01 1.05959344e+00
5.45877703e-02 7.14757383e-01 8.98723602e-01 -4.17779565e-01
9.71948430e-02 3.52478027e-01 3.19587111e-01 6.45973086e-01
1.64942041e-01 -3.80157351e-01 -5.32778323e-01 -1.61681786e-01
4.09795970e-01 4.25306916e-01 -2.04588175e-01 -2.45408878e-01
-1.24419618e+00 1.01946259e+00 6.45436585e-01 5.57270646e-01
-1.62164658e-01 -2.23159283e-01 5.18389523e-01 4.19247746e-01
5.77731669e-01 5.67912877e-01 -4.64842588e-01 -1.65707171e-01
-8.07521820e-01 -1.20352544e-01 5.34561515e-01 8.67745399e-01
9.49840665e-01 -4.59018916e-01 -3.67713302e-01 9.69983339e-01
3.14312100e-01 8.76970589e-02 1.00129282e+00 -5.10015309e-01
7.72508860e-01 9.06229854e-01 -2.67065335e-02 -1.02532494e+00
-2.77486950e-01 -1.98175520e-01 -9.05614138e-01 -5.97536750e-02
5.89597166e-01 1.44042701e-01 -6.52592301e-01 1.62163031e+00
2.85925061e-01 1.07827857e-01 2.33402792e-02 7.58386493e-01
4.78366852e-01 7.35452354e-01 6.66942894e-02 2.48927861e-01
1.36339808e+00 -1.03482389e+00 -3.55552942e-01 -2.61514187e-01
1.19329345e+00 -7.48341441e-01 1.43226504e+00 -4.47346754e-02
-5.59809923e-01 -4.31328416e-01 -8.88658404e-01 -3.18103671e-01
-7.09751368e-01 8.57019126e-02 2.01851457e-01 4.03818846e-01
-6.91005409e-01 6.80945396e-01 -6.86408222e-01 -3.77037704e-01
4.52851832e-01 2.43467867e-01 -4.18216348e-01 -3.11592162e-01
-1.22426975e+00 7.44378090e-01 6.14004850e-01 -1.43225426e-02
-4.89080638e-01 -6.81850493e-01 -8.35323989e-01 3.21058601e-01
7.29525536e-02 -1.83505476e-01 1.06822944e+00 -7.95777857e-01
-1.22657084e+00 8.14225078e-01 -1.70656443e-01 5.61243109e-03
4.52585071e-01 7.28934351e-03 -2.20277563e-01 -1.04636244e-01
3.46493840e-01 5.38225234e-01 7.81270266e-01 -7.73921192e-01
-7.33096004e-01 -2.20632523e-01 1.49286434e-01 2.31917754e-01
-1.01327753e+00 -6.04637451e-02 -3.70215438e-02 -6.65788710e-01
1.35596842e-01 -7.78544843e-01 9.13267210e-02 8.27501416e-02
-4.21140701e-01 -8.09951782e-01 8.21866155e-01 -3.40705335e-01
1.16710997e+00 -2.36688471e+00 5.85053898e-02 6.84109554e-02
3.00285995e-01 2.49032870e-01 -4.98747617e-01 4.86558378e-01
-9.97210294e-03 3.20577264e-01 -7.73236081e-02 -4.23508227e-01
2.00936615e-01 1.66157454e-01 -4.74115491e-01 4.33822930e-01
4.41653371e-01 7.64054120e-01 -1.10612345e+00 -4.44889277e-01
5.09009697e-02 2.76507378e-01 -2.59188712e-01 4.47824001e-01
4.88118343e-02 1.75179407e-01 -4.83734250e-01 3.95590276e-01
5.46800077e-01 -4.72585320e-01 5.16455509e-02 -1.38847694e-01
1.10793792e-01 6.72526598e-01 -1.10814011e+00 1.57884860e+00
-9.46812928e-01 6.22851133e-01 -4.06819701e-01 -1.20520461e+00
1.02736580e+00 2.83727109e-01 2.31113806e-01 -5.80444098e-01
1.91216469e-01 3.28277230e-01 9.16108415e-02 -2.43953437e-01
5.19315600e-01 -1.90365985e-02 -2.37637311e-01 9.36494708e-01
2.00400010e-01 3.02156180e-01 9.19866264e-02 2.23005936e-01
1.11389387e+00 -3.85665625e-01 1.02242023e-01 -1.83408290e-01
4.36673760e-01 -2.94361740e-01 5.00545204e-01 6.20877445e-01
-1.33917734e-01 5.90994716e-01 3.55473191e-01 -3.52487743e-01
-1.01542366e+00 -1.00278234e+00 -3.73823285e-01 1.34945905e+00
1.41084492e-01 -5.37062228e-01 -4.04222131e-01 -1.06480265e+00
1.35800421e-01 6.07020676e-01 -7.18714952e-01 -4.55006599e-01
-3.47109944e-01 -6.49402380e-01 4.97284472e-01 5.74818015e-01
2.76761383e-01 -1.01240921e+00 -3.68459612e-01 2.92963177e-01
-3.90572697e-01 -9.38840210e-01 -7.42465198e-01 5.16796112e-01
-7.32633829e-01 -9.88239288e-01 -6.17772400e-01 -8.60446513e-01
8.95247579e-01 4.61512029e-01 9.06374812e-01 1.32297695e-01
-1.27665177e-01 1.56264409e-01 -5.69120109e-01 -4.99108844e-02
-3.75129938e-01 4.54941005e-01 1.49848476e-01 3.71347427e-01
7.40074456e-01 -2.69561946e-01 -4.33726668e-01 4.21391934e-01
-9.83703613e-01 -1.68222934e-03 4.44687009e-01 1.11456358e+00
3.05610150e-01 1.74569905e-01 5.41399598e-01 -7.41913080e-01
6.99468911e-01 -6.88098490e-01 -2.57079273e-01 1.72680929e-01
-6.56561971e-01 2.95434266e-01 1.06786263e+00 -7.48525083e-01
-7.13793099e-01 -2.31034592e-01 1.21292144e-01 -2.70013601e-01
-3.26017477e-02 4.55358624e-01 -2.02900320e-01 1.82725444e-01
5.44966757e-01 2.52352506e-01 -2.46830836e-01 -4.89646018e-01
4.70602095e-01 1.06612074e+00 7.17827603e-02 -5.60202956e-01
8.69162619e-01 1.52570158e-01 -4.79776263e-01 -4.85493034e-01
-1.17250097e+00 -6.05097413e-01 -8.38008881e-01 1.52290657e-01
5.49230158e-01 -7.17554986e-01 -3.39103550e-01 4.25067514e-01
-1.06104112e+00 -3.71435672e-01 -3.40237141e-01 4.50002044e-01
-1.58939868e-01 1.87299445e-01 -6.21118784e-01 -3.92556548e-01
-2.26818874e-01 -1.03281355e+00 8.59211683e-01 2.61413932e-01
-4.48596537e-01 -1.29488993e+00 -1.06888423e-02 1.31265923e-01
3.94135475e-01 -3.44382465e-01 1.12726915e+00 -1.20701838e+00
-3.26454788e-01 -3.34294885e-01 -4.84336734e-01 3.44710201e-01
4.97447163e-01 -1.89165741e-01 -9.43289042e-01 -4.76314485e-01
-1.77499533e-01 -5.66407084e-01 7.91469991e-01 -1.42067134e-01
1.24820733e+00 -4.25210446e-01 -4.09129977e-01 4.09871548e-01
1.17336273e+00 -2.60426253e-01 1.88227817e-01 4.89946485e-01
9.45324898e-01 4.89469677e-01 6.78685367e-01 4.87488419e-01
3.87784421e-01 6.75693929e-01 1.36772811e-01 1.21097535e-01
-1.37247965e-01 -5.16081393e-01 3.08407187e-01 1.05634856e+00
5.78076363e-01 -3.77894461e-01 -8.68699372e-01 6.76479101e-01
-1.71823740e+00 -6.96553349e-01 1.96604282e-01 2.21852255e+00
1.05118227e+00 9.07766074e-02 -1.58175468e-01 4.01779294e-01
8.76664519e-01 1.33143589e-01 -5.76967180e-01 -3.99192303e-01
1.63995847e-01 1.16965562e-01 2.38154680e-01 3.52054000e-01
-9.97440338e-01 8.81505907e-01 4.76084280e+00 1.05199909e+00
-1.32891226e+00 2.10683182e-01 5.57230830e-01 1.13064814e-02
-3.80413353e-01 -8.29020217e-02 -9.93855715e-01 7.27363884e-01
9.95500565e-01 -4.80041713e-01 1.79124773e-01 6.51874602e-01
4.22327332e-02 1.60420388e-01 -1.34278059e+00 9.02356446e-01
3.03779878e-02 -1.04293358e+00 1.07191831e-01 -1.55587299e-02
4.90557462e-01 1.31084397e-02 1.39642537e-01 4.34343010e-01
2.85530299e-01 -9.24421549e-01 5.59556186e-01 2.09873632e-01
8.94982100e-01 -6.26982927e-01 6.78205132e-01 4.10001069e-01
-1.31774116e+00 -1.95970550e-01 -7.28840470e-01 -7.78948143e-02
-3.10971200e-01 6.21658623e-01 -1.13882065e+00 3.21859002e-01
5.68613291e-01 8.74176085e-01 -6.00290537e-01 9.27047729e-01
-3.48182857e-01 5.18440962e-01 -1.90625876e-01 -2.13439763e-01
2.85389274e-01 -1.05903625e-01 7.40298927e-02 1.30203128e+00
4.63260710e-01 -2.66030401e-01 1.31433174e-01 7.86633253e-01
-4.82628942e-01 3.17196637e-01 -5.32423913e-01 -7.29344562e-02
7.89963603e-01 1.50179458e+00 -6.73038185e-01 -4.01418447e-01
-5.23110569e-01 1.16302407e+00 8.10742259e-01 1.82751521e-01
-7.30609596e-01 -7.06767499e-01 7.83996463e-01 -6.29437268e-02
2.59300828e-01 -4.98937257e-02 -2.62489438e-01 -1.31512594e+00
2.07989350e-01 -6.14599526e-01 4.72401977e-01 -3.16350877e-01
-1.64451408e+00 5.62879980e-01 -3.81778032e-01 -1.35691774e+00
-1.10676937e-01 -5.17596006e-01 -7.09426343e-01 9.53745127e-01
-1.50474250e+00 -9.50712621e-01 -2.94380188e-01 5.06982267e-01
5.78209221e-01 -5.13219200e-02 9.75933194e-01 3.97715896e-01
-6.75084591e-01 1.10097396e+00 4.36757267e-01 4.57975924e-01
9.96510804e-01 -1.32675183e+00 3.15874457e-01 7.05582857e-01
4.52441633e-01 4.52766687e-01 2.89244980e-01 -2.57746637e-01
-1.09664369e+00 -1.53979242e+00 1.34518039e+00 -5.31595945e-01
8.58381391e-01 -6.69830918e-01 -1.35381949e+00 6.69622719e-01
-8.66497904e-02 3.56407195e-01 9.41583633e-01 2.33724654e-01
-8.53735447e-01 -1.29076585e-01 -9.77300227e-01 3.77877116e-01
8.28209758e-01 -8.65697980e-01 -6.34501696e-01 4.07922924e-01
6.14837885e-01 -9.12048593e-02 -9.40414131e-01 -1.24644227e-01
4.16144729e-01 -5.53032160e-01 5.73295474e-01 -5.91325998e-01
4.56583828e-01 -2.40949765e-01 -6.75218999e-02 -1.58150625e+00
-5.26094258e-01 -1.19313873e-01 1.17895499e-01 1.48513198e+00
4.27004337e-01 -8.84679556e-01 5.36542356e-01 7.17814505e-01
3.73615883e-02 -7.19790101e-01 -1.02402961e+00 -9.69030380e-01
3.22804570e-01 -2.06775725e-01 7.01058209e-01 1.30564702e+00
2.10994855e-01 4.54498023e-01 -7.61394426e-02 1.14885963e-01
3.22269320e-01 2.45030612e-01 5.55575132e-01 -1.32691920e+00
5.84549308e-02 -4.90619987e-01 -5.69447100e-01 -1.17688751e+00
5.39560318e-01 -1.47655344e+00 5.76165244e-02 -1.28072953e+00
3.66294593e-01 -8.77847433e-01 -6.76639020e-01 8.71004403e-01
-3.12995553e-01 3.27184677e-01 2.36990988e-01 3.73475373e-01
-5.97210169e-01 7.28625000e-01 9.11668360e-01 -3.84070039e-01
-3.34661007e-02 -9.19413045e-02 -7.52619088e-01 4.69018310e-01
9.28554416e-01 -9.40319300e-01 -2.66775399e-01 -5.73875904e-01
1.37198165e-01 -4.14618790e-01 2.11083278e-01 -7.43499398e-01
3.14977318e-01 -6.06918335e-02 3.74864250e-01 -1.40798196e-01
1.25640020e-01 -8.58794510e-01 -4.21019167e-01 2.93587506e-01
-7.81769872e-01 -3.98220355e-03 -2.41487101e-02 3.82119328e-01
-2.91481376e-01 -4.96246606e-01 7.63796806e-01 2.36591741e-01
-5.22619188e-01 5.18105745e-01 -2.53659785e-01 2.90790051e-01
7.83403933e-01 -1.37991309e-02 -2.65729249e-01 -1.56995151e-02
-3.92550707e-01 1.60902843e-01 5.04684210e-01 7.82832205e-01
5.04021227e-01 -1.68945503e+00 -6.95812583e-01 2.59839386e-01
4.73715156e-01 4.41342965e-02 -3.42304558e-02 6.19005203e-01
3.28365713e-02 3.52561593e-01 -6.61105290e-02 -5.39560914e-01
-1.05630493e+00 3.85395527e-01 1.86524168e-01 -3.90655518e-01
-5.81771016e-01 7.78088808e-01 2.45684683e-01 -7.95802414e-01
9.45093036e-02 -3.80452991e-01 -3.55856679e-02 3.24054241e-01
5.74563146e-01 2.09633082e-01 1.89776659e-01 -5.95673144e-01
-4.16496068e-01 4.97173935e-01 -3.67460042e-01 1.43109024e-01
1.12542212e+00 -2.11802512e-01 1.73383225e-02 6.59768999e-01
1.72574961e+00 -1.13717794e-01 -1.14680910e+00 -7.25241244e-01
1.63755357e-01 -6.03258371e-01 1.12428423e-02 -4.36340660e-01
-9.16742563e-01 1.05363977e+00 4.82774585e-01 3.94875437e-01
8.87135804e-01 9.04161409e-02 8.56041908e-01 4.49840695e-01
-7.03796521e-02 -1.06665564e+00 3.14227164e-01 7.30167389e-01
5.92511475e-01 -1.42216790e+00 -4.45269912e-01 -3.58466282e-02
-3.25769931e-01 1.32581484e+00 5.27183533e-01 5.43570779e-02
5.59654891e-01 -5.67172281e-03 1.74557745e-01 1.38688028e-01
-7.76438296e-01 5.50593659e-02 1.56007975e-01 4.02356327e-01
6.14229918e-01 1.61579922e-01 -1.90340672e-02 5.35688698e-01
-2.13600799e-01 -3.70401204e-01 5.32557786e-01 8.25970531e-01
-4.30452913e-01 -1.32546866e+00 -1.51867926e-01 6.56643391e-01
-2.27826431e-01 -2.74297923e-01 -1.96141198e-01 4.09919381e-01
-1.00006409e-01 9.74589229e-01 3.03780496e-01 -3.83724242e-01
2.22733766e-01 3.09474200e-01 2.40991294e-01 -8.87098968e-01
-4.40199703e-01 -4.82698113e-01 -3.46879363e-01 -2.99812108e-01
2.15440840e-02 -5.93638957e-01 -1.26360488e+00 -2.20247969e-01
-3.86759698e-01 2.68611640e-01 3.87444854e-01 1.02224922e+00
4.49403763e-01 3.47885251e-01 1.08358526e+00 -6.28381431e-01
-9.01929200e-01 -1.02433884e+00 -6.59655154e-01 7.31256068e-01
3.32165033e-01 -6.75313771e-01 -6.41923308e-01 -1.61297783e-01]
|
[10.295755386352539, 6.960758686065674]
|
35d4daea-487c-4b57-b342-e6658b291f00
|
polka-lines-learning-structured-illumination
|
2011.13117
| null |
https://arxiv.org/abs/2011.13117v2
|
https://arxiv.org/pdf/2011.13117v2.pdf
|
Polka Lines: Learning Structured Illumination and Reconstruction for Active Stereo
|
Active stereo cameras that recover depth from structured light captures have become a cornerstone sensor modality for 3D scene reconstruction and understanding tasks across application domains. Existing active stereo cameras project a pseudo-random dot pattern on object surfaces to extract disparity independently of object texture. Such hand-crafted patterns are designed in isolation from the scene statistics, ambient illumination conditions, and the reconstruction method. In this work, we propose the first method to jointly learn structured illumination and reconstruction, parameterized by a diffractive optical element and a neural network, in an end-to-end fashion. To this end, we introduce a novel differentiable image formation model for active stereo, relying on both wave and geometric optics, and a novel trinocular reconstruction network. The jointly optimized pattern, which we dub "Polka Lines," together with the reconstruction network, achieve state-of-the-art active-stereo depth estimates across imaging conditions. We validate the proposed method in simulation and on a hardware prototype, and show that our method outperforms existing active stereo systems.
|
['Felix Heide', 'Seung-Hwan Baek']
|
2020-11-26
| null |
http://openaccess.thecvf.com//content/CVPR2021/html/Baek_Polka_Lines_Learning_Structured_Illumination_and_Reconstruction_for_Active_Stereo_CVPR_2021_paper.html
|
http://openaccess.thecvf.com//content/CVPR2021/papers/Baek_Polka_Lines_Learning_Structured_Illumination_and_Reconstruction_for_Active_Stereo_CVPR_2021_paper.pdf
|
cvpr-2021-1
|
['3d-scene-reconstruction']
|
['computer-vision']
|
[ 6.56992972e-01 1.24043236e-02 2.82489240e-01 -4.89140034e-01
-6.79068923e-01 -5.37634254e-01 4.12972391e-01 -6.28136933e-01
-4.09713775e-01 4.99238729e-01 3.47822756e-01 2.50790529e-02
-1.12524390e-01 -4.81140256e-01 -9.88098145e-01 -9.67903674e-01
6.47633851e-01 3.25812072e-01 2.26518214e-01 1.60989597e-01
6.79430306e-01 3.48144591e-01 -1.75414443e+00 2.85855651e-01
8.25324416e-01 1.10511923e+00 6.14497721e-01 6.46860301e-01
1.11020654e-01 1.06717777e+00 7.01152952e-03 -1.33722126e-01
5.67685544e-01 -9.64953564e-03 -1.70019493e-01 6.89646244e-01
7.44516134e-01 -9.25254762e-01 -4.70082134e-01 1.03674603e+00
4.95483458e-01 -2.70291809e-02 4.42176759e-01 -6.53991401e-01
-3.92132252e-01 -1.98573768e-01 -5.10221004e-01 -1.91545084e-01
4.71073598e-01 5.01355350e-01 6.51491880e-01 -1.21549833e+00
5.32348037e-01 9.78961408e-01 3.96369427e-01 6.53775990e-01
-1.28890479e+00 -5.64456046e-01 -3.33715305e-02 -7.57177919e-02
-1.18838322e+00 -9.96189952e-01 1.20795512e+00 -5.52455783e-01
9.06262934e-01 -1.14647672e-01 8.34816933e-01 9.31841433e-01
2.14389294e-01 8.15677464e-01 1.24129176e+00 -5.39303362e-01
3.30788761e-01 3.76437791e-02 -6.91063628e-02 7.55259931e-01
2.59403408e-01 4.75517631e-01 -1.03228295e+00 -5.66704609e-02
1.10723233e+00 2.53458142e-01 -7.74531364e-01 -9.50112462e-01
-9.14425969e-01 4.33171868e-01 9.26878229e-02 -4.20402318e-01
-4.48617548e-01 1.04244769e-01 -3.73210013e-01 -1.60389096e-01
5.93813717e-01 2.93480158e-01 -6.94788992e-02 2.86172461e-02
-5.25870502e-01 -4.96238805e-02 6.66112542e-01 9.26276922e-01
1.11520374e+00 -7.23255426e-02 1.83997288e-01 8.65511239e-01
6.93644345e-01 7.77722180e-01 1.23217106e-02 -1.44929886e+00
5.19183159e-01 6.21878564e-01 4.74751562e-01 -6.82694376e-01
-1.68700516e-01 -1.54236004e-01 -5.35991907e-01 6.09017015e-01
3.58683676e-01 -1.21391997e-01 -8.00729096e-01 1.49569905e+00
3.80524486e-01 3.30170333e-01 -1.42796174e-01 1.17247355e+00
6.39878631e-01 4.52071428e-01 -8.55685472e-01 -3.40314239e-01
5.80754697e-01 -8.01475942e-01 -5.18764079e-01 -3.74866009e-01
1.25669166e-01 -7.94134617e-01 8.41601789e-01 8.19317937e-01
-1.54129410e+00 -2.69476205e-01 -1.05105793e+00 -4.43945795e-01
4.58879828e-01 4.22110781e-02 4.64414001e-01 3.75439852e-01
-1.21774948e+00 3.50454241e-01 -8.72932315e-01 -1.45679280e-01
3.89036715e-01 4.12575573e-01 -9.08796787e-02 -3.52087051e-01
-2.85083264e-01 3.11321646e-01 2.70710094e-03 1.42245889e-01
-1.01519239e+00 -7.21842527e-01 -8.75167906e-01 -2.18720481e-01
4.48257506e-01 -1.12442803e+00 1.06931818e+00 -8.76602590e-01
-2.13173103e+00 1.21015048e+00 -4.98294711e-01 -2.61360765e-01
2.25099221e-01 -5.06096840e-01 1.31587818e-01 3.97174865e-01
-2.23178029e-01 4.98178899e-01 9.53783393e-01 -1.57819510e+00
-6.46782458e-01 -7.30063021e-01 2.04390541e-01 5.87699056e-01
-6.94793314e-02 -3.63430172e-01 -8.52250695e-01 -5.33753298e-02
3.62159222e-01 -8.10606480e-01 -1.93504006e-01 5.65342844e-01
-4.26813364e-01 4.63772535e-01 5.79683602e-01 -1.58476546e-01
5.90385556e-01 -2.17790389e+00 3.34162831e-01 -6.92945868e-02
4.89512086e-01 -1.54262614e-02 -5.25566265e-02 5.85891455e-02
2.90538400e-01 -7.78788090e-01 -2.14736119e-01 -8.49026263e-01
-3.64129752e-01 1.39461622e-01 -3.66138995e-01 7.52236187e-01
-1.82897481e-03 5.37149429e-01 -8.67379963e-01 -1.78887486e-01
6.73131824e-01 5.92598259e-01 -8.18780601e-01 5.28529942e-01
-2.24697664e-01 7.52387583e-01 -3.18354219e-01 6.40009284e-01
8.44640970e-01 -4.01801854e-01 3.96131799e-02 -1.25518471e-01
-6.51633620e-01 3.61955076e-01 -1.20398772e+00 2.17791653e+00
-5.85039020e-01 6.76794410e-01 3.79074931e-01 -5.92194915e-01
9.74598765e-01 1.37932047e-01 5.81254423e-01 -9.98235822e-01
2.04566289e-02 3.82135153e-01 -4.82148200e-01 -4.29088235e-01
5.43510377e-01 1.63985565e-01 5.08735895e-01 4.27994788e-01
-1.20429553e-01 -5.10501921e-01 -9.52561572e-02 -3.39126252e-02
1.10835719e+00 3.45185935e-01 1.35727912e-01 -2.08868921e-01
4.93946284e-01 -3.14902872e-01 5.91032147e-01 6.68246448e-01
2.19469562e-01 9.74389076e-01 6.56323805e-02 -3.46017450e-01
-1.20144939e+00 -1.24888885e+00 -1.48987174e-01 3.44029844e-01
6.01783931e-01 1.17907733e-01 -5.23180008e-01 -1.66144028e-01
-1.79180816e-01 4.41453695e-01 -1.67242602e-01 1.86456516e-01
-5.98192394e-01 -2.76280046e-01 -1.79614991e-01 1.24781489e-01
7.93124676e-01 -5.63914955e-01 -7.67172515e-01 1.60385147e-01
-2.95576192e-02 -1.27511990e+00 -4.35452700e-01 7.10586086e-02
-9.61286247e-01 -1.25095046e+00 -6.95040941e-01 -5.48465610e-01
7.27083266e-01 8.90990257e-01 1.17388666e+00 -3.20872933e-01
-2.01219872e-01 7.86800444e-01 -3.09004681e-03 -2.41197199e-01
1.94811106e-01 -4.38582271e-01 -7.28245229e-02 5.46369553e-01
1.35230705e-01 -9.89488721e-01 -1.12405121e+00 4.18210506e-01
-8.13139617e-01 7.28102446e-01 4.71560627e-01 7.43760884e-01
8.23645771e-01 -4.12371039e-01 -5.13353825e-01 -9.55267847e-01
-7.12495521e-02 -1.00069739e-01 -1.22826016e+00 -7.53660128e-02
-2.87502319e-01 -2.86995573e-03 3.58517021e-01 -2.09562361e-01
-1.38116503e+00 5.90018630e-01 6.32384717e-02 -7.49567211e-01
-9.97140631e-02 -1.53259173e-01 -4.76755112e-01 -3.98768544e-01
7.08627343e-01 3.22305948e-01 -6.17242642e-02 -4.43940520e-01
-1.15697481e-01 6.59137964e-01 6.38429463e-01 -4.45720285e-01
7.42173493e-01 1.39636469e+00 1.37496099e-01 -9.96959627e-01
-1.18454731e+00 -8.33663523e-01 -5.88971078e-01 -3.79119575e-01
6.98690653e-01 -1.22274935e+00 -8.79728436e-01 9.61482048e-01
-1.39122725e+00 -6.22117817e-01 -3.26686174e-01 7.86772668e-01
-8.58228028e-01 3.04569066e-01 -4.05062586e-01 -9.48142111e-01
-1.23093814e-01 -1.16721070e+00 1.66390586e+00 2.19406679e-01
3.61750364e-01 -9.40916777e-01 2.09679008e-01 6.75657392e-01
2.43528381e-01 -7.47335181e-02 4.73231763e-01 5.49012840e-01
-1.55718648e+00 3.54624063e-01 -3.71994466e-01 4.15161818e-01
6.92223758e-02 -2.16342762e-01 -1.43510795e+00 -1.79988131e-01
5.35494506e-01 -2.30955884e-01 8.20408404e-01 7.30432272e-01
1.17936063e+00 -5.27485879e-03 -1.20069683e-01 1.38729060e+00
1.85331488e+00 3.22463185e-01 8.54547799e-01 7.59417564e-02
8.98408175e-01 5.89798212e-01 3.15854818e-01 7.18337297e-01
2.58444965e-01 8.99415672e-01 6.48623526e-01 -7.07477406e-02
-1.87418118e-01 -1.51149198e-01 3.72615486e-01 8.88446569e-01
-1.07685380e-01 -3.78677756e-01 -7.96194732e-01 1.94317997e-01
-1.51382363e+00 -8.40426564e-01 -2.21146271e-01 2.53267145e+00
6.84564471e-01 8.87393057e-02 -4.31145877e-01 1.51612684e-01
4.91980463e-01 1.53678462e-01 -9.15560722e-01 1.96005374e-01
-3.81191134e-01 2.66258061e-01 6.36053205e-01 8.92845452e-01
-6.92173302e-01 5.77251494e-01 5.27199316e+00 4.01646137e-01
-1.40802789e+00 -7.41898045e-02 1.80737108e-01 -3.43942672e-01
-6.31484568e-01 1.47031799e-01 -9.72201169e-01 3.92855912e-01
2.66233861e-01 2.24488154e-01 6.82705522e-01 4.52383280e-01
3.87672603e-01 -3.61503780e-01 -1.27043402e+00 1.54075146e+00
3.03101212e-01 -1.32076299e+00 -1.20404981e-01 3.92140388e-01
9.67757225e-01 1.64361790e-01 1.45738393e-01 -5.28884590e-01
2.51384765e-01 -5.73827028e-01 7.89259553e-01 8.04251254e-01
7.50629663e-01 -2.15218246e-01 1.30685300e-01 4.90250528e-01
-7.35957563e-01 -4.43172380e-02 -2.42808640e-01 -1.56627178e-01
4.00458694e-01 9.27208543e-01 -3.68531942e-01 1.71195149e-01
5.03904879e-01 1.16894686e+00 -1.01654410e-01 1.23005867e+00
-2.81300366e-01 3.20948869e-01 -3.40308785e-01 2.28495300e-01
-1.95424974e-01 -3.92915696e-01 8.45627546e-01 5.01274347e-01
3.64850640e-01 3.55656296e-01 -8.52520019e-02 1.07470500e+00
6.42674267e-02 -3.39793622e-01 -8.29249084e-01 2.88952529e-01
2.94765621e-01 9.49573934e-01 -2.69004375e-01 9.68884677e-02
-6.44071698e-01 1.06881034e+00 2.75905430e-01 5.93620002e-01
-3.46357733e-01 -7.79357925e-02 6.74308777e-01 4.25384969e-01
8.69521797e-02 -4.66038585e-01 -4.57686067e-01 -1.61383545e+00
2.71058261e-01 -4.74701881e-01 -2.37618476e-01 -1.30169213e+00
-1.15984333e+00 -1.35661922e-02 -1.84959233e-01 -1.49227118e+00
-1.20168179e-01 -9.74870682e-01 -2.75305718e-01 8.55691373e-01
-1.83073175e+00 -7.30078578e-01 -6.24739349e-01 9.82611179e-01
6.62763774e-01 5.77462912e-02 3.91517699e-01 2.99164057e-01
-2.51158863e-01 -4.07012086e-03 3.98203999e-01 -1.46472290e-01
5.21145284e-01 -8.92508805e-01 1.23522878e-01 8.79671812e-01
2.93166749e-03 5.49024761e-01 4.89899158e-01 -3.91498029e-01
-2.03141832e+00 -7.64370143e-01 3.07505608e-01 -2.35991538e-01
1.62645981e-01 -5.86331308e-01 -5.14995277e-01 4.59428847e-01
6.91486374e-02 7.82820284e-02 3.15204233e-01 -1.80975124e-01
-1.92613408e-01 -4.19048965e-01 -9.65803862e-01 5.73107600e-01
1.52864993e+00 -7.44645894e-01 -3.19421023e-01 2.10104510e-01
5.12099922e-01 -7.85192728e-01 -9.43083391e-02 4.93854582e-01
6.72886014e-01 -1.67076445e+00 1.01252425e+00 2.14853808e-01
7.03097045e-01 -2.28684574e-01 -3.85519713e-01 -1.04867136e+00
-8.80432948e-02 -7.67808616e-01 -1.76053837e-01 8.24134290e-01
1.30980462e-01 -7.76769280e-01 1.19080937e+00 5.93842983e-01
-6.28451526e-01 -5.92577934e-01 -7.57059932e-01 -4.58041877e-01
-6.17957294e-01 -4.91562694e-01 1.54743597e-01 5.36292732e-01
-5.80619693e-01 4.70680118e-01 -3.46185893e-01 3.71972620e-01
1.40853500e+00 1.82969302e-01 1.09110487e+00 -1.27761567e+00
-7.72048354e-01 -3.40584926e-02 -3.74044508e-01 -2.04007554e+00
5.72720580e-02 -2.84543365e-01 3.44618797e-01 -1.18475270e+00
3.14630747e-01 -3.69221240e-01 2.08032846e-01 -3.26530963e-01
9.55498293e-02 2.40981147e-01 -9.78146270e-02 4.01819378e-01
-4.70829159e-01 6.36047661e-01 1.41851604e+00 5.41324690e-02
-4.37203497e-01 -4.94199879e-02 -3.54664832e-01 9.73067522e-01
3.03024977e-01 -1.60638064e-01 -7.65611053e-01 -1.15608859e+00
5.08883893e-01 4.12555248e-01 5.34388185e-01 -1.01517260e+00
6.07881665e-01 -2.84944564e-01 2.05535740e-01 -6.04178369e-01
1.00980639e+00 -8.52867365e-01 2.20159456e-01 -6.81172078e-03
-1.86958715e-01 -5.82472742e-01 -2.02962309e-01 6.61919892e-01
-1.62366375e-01 -6.85636327e-02 9.70711529e-01 -1.10969961e-01
-6.13993764e-01 5.25295436e-01 1.18222192e-01 2.16479868e-01
6.56683505e-01 -7.49507487e-01 -4.67168242e-01 -3.28330725e-01
-1.99102014e-01 2.33551022e-02 9.75933492e-01 -4.49770167e-02
1.06424034e+00 -1.04759657e+00 -4.44940925e-01 6.36735559e-01
1.79798156e-01 5.27681887e-01 3.17524344e-01 7.86940217e-01
-8.16533208e-01 4.80391651e-01 -1.18899807e-01 -1.08385623e+00
-1.14826751e+00 2.18168274e-01 5.41501045e-01 1.47055537e-01
-8.12694728e-01 8.58439445e-01 9.05194521e-01 -2.92864829e-01
2.79467762e-01 -2.73591846e-01 2.81175464e-01 -5.44566333e-01
4.55596119e-01 2.14802012e-01 5.39623126e-02 -2.73793310e-01
2.10407585e-01 1.07075357e+00 1.61182150e-01 -4.09185052e-01
1.50248051e+00 -6.01294279e-01 8.42345953e-02 7.04093874e-01
1.12169468e+00 2.01340124e-01 -1.93543231e+00 -5.77630460e-01
-6.54333115e-01 -1.05498910e+00 3.79464656e-01 -2.42467314e-01
-1.11119819e+00 1.07584274e+00 4.62247789e-01 -3.43027234e-01
1.24478531e+00 -1.96276218e-01 6.90994263e-01 4.59904343e-01
8.70111108e-01 -8.96628141e-01 3.06266993e-01 5.23823857e-01
5.67994356e-01 -1.33906150e+00 -1.25694096e-01 -6.56201303e-01
-2.45242774e-01 9.71015573e-01 5.35343409e-01 -1.88722700e-01
6.99331224e-01 3.39048177e-01 -2.24495213e-02 -3.04115385e-01
-8.83574247e-01 -1.10957369e-01 9.92227793e-02 5.78788698e-01
7.06239641e-02 -3.61089349e-01 2.94042468e-01 -8.60660225e-02
1.81953833e-01 2.56090373e-01 4.77656424e-01 7.95226753e-01
-5.71195185e-01 -9.10270691e-01 -3.92020464e-01 1.55173868e-01
-1.80973887e-01 -4.79343422e-02 -2.44298965e-01 3.97391468e-01
9.74534228e-02 7.85123348e-01 3.41400445e-01 -1.31878972e-01
4.30917442e-01 -4.16835457e-01 9.64648306e-01 -8.70667219e-01
2.79467436e-04 7.15163648e-02 -5.52369654e-02 -8.31129909e-01
-6.49874032e-01 -7.29566455e-01 -7.63292015e-01 -1.27890646e-01
-2.10521087e-01 -4.21063453e-01 6.46427572e-01 7.69078434e-01
4.15099651e-01 -2.08629310e-01 1.03302336e+00 -1.12846267e+00
-1.49636626e-01 -5.38276196e-01 -8.64137173e-01 3.01541060e-01
6.15843058e-01 -6.08581841e-01 -7.40498185e-01 1.34400964e-01]
|
[9.375216484069824, -2.730773448944092]
|
ecd0f20d-50a3-45c7-b166-9f1080ea0ffa
|
test-time-fast-adaptation-for-dynamic-scene
| null | null |
http://openaccess.thecvf.com//content/CVPR2021/html/Chi_Test-Time_Fast_Adaptation_for_Dynamic_Scene_Deblurring_via_Meta-Auxiliary_Learning_CVPR_2021_paper.html
|
http://openaccess.thecvf.com//content/CVPR2021/papers/Chi_Test-Time_Fast_Adaptation_for_Dynamic_Scene_Deblurring_via_Meta-Auxiliary_Learning_CVPR_2021_paper.pdf
|
Test-Time Fast Adaptation for Dynamic Scene Deblurring via Meta-Auxiliary Learning
|
In this paper, we tackle the problem of dynamic scene deblurring. Most existing deep end-to-end learning approaches adopt the same generic model for all unseen test images. These solutions are sub-optimal, as they fail to utilize the internal information within a specific image. On the other hand, a self-supervised approach, SelfDeblur, enables internal-training within a test image from scratch, but it does not fully take advantages of large external dataset. In this work, we propose a novel self-supervised meta-auxiliary learning to improve the performance of deblurring by integrating both external and internal learning. Concretely, we build a self-supervised auxiliary reconstruction task which shares a portion of the network with the primary deblurring task. The two tasks are jointly trained on an external dataset. Furthermore, we propose a meta-auxiliary training scheme to further optimize the pre-trained model as a base learner which is applicable for fast adaptation at test time. During training, the performance of both tasks is coupled. Therefore, we are able to exploit the internal information at test time via the auxiliary task to enhance the performance of deblurring. Extensive experimental results across evaluation datasets demonstrate the effectiveness of test-time adaptation of the proposed method.
|
['Jin Tang', 'Yuanhao Yu', 'Yang Wang', 'Zhixiang Chi']
|
2021-06-19
| null | null | null |
cvpr-2021-1
|
['auxiliary-learning']
|
['methodology']
|
[ 2.18034536e-03 -3.50150645e-01 -8.92826542e-02 -2.15552256e-01
-7.88145423e-01 -3.89366657e-01 4.30465609e-01 -4.67834294e-01
-4.92260963e-01 4.86317515e-01 1.83747455e-01 2.44511967e-03
2.58214712e-01 -4.87178445e-01 -7.54414856e-01 -9.47269022e-01
4.57872480e-01 6.11320324e-02 2.73154765e-01 2.41180230e-02
6.95321187e-02 1.07052706e-01 -1.22592103e+00 1.54514119e-01
1.24657047e+00 8.98435414e-01 6.85741782e-01 5.29106319e-01
3.63474727e-01 8.27140331e-01 -4.36174810e-01 1.36608593e-02
4.99081314e-02 -5.97610295e-01 -7.91418016e-01 2.43603677e-01
5.36897540e-01 -7.36785650e-01 -5.98221183e-01 1.06511176e+00
7.07504094e-01 2.56259710e-01 2.88090467e-01 -7.94393599e-01
-6.87063396e-01 3.33517313e-01 -6.03505194e-01 4.52908516e-01
-1.33740306e-01 2.93216884e-01 5.10477304e-01 -1.00575435e+00
4.56933796e-01 8.83333564e-01 6.06907725e-01 6.46158576e-01
-8.93355548e-01 -5.92770934e-01 2.60846585e-01 5.70321202e-01
-1.35572314e+00 -6.25673473e-01 1.04239464e+00 -1.91718742e-01
4.89318401e-01 -4.13143784e-02 4.09975499e-01 1.12583399e+00
-1.18515901e-01 1.10908914e+00 1.28756952e+00 -2.44390592e-01
1.47063151e-01 9.33659673e-02 1.79505497e-01 5.11784017e-01
-7.05953687e-02 2.09623352e-01 -3.24555278e-01 1.48645326e-01
7.82749295e-01 4.23134165e-03 -8.30288589e-01 -4.16172355e-01
-1.25017643e+00 2.53729254e-01 6.19683862e-01 3.85437429e-01
-3.11886907e-01 3.88728306e-02 3.38156968e-01 3.68318319e-01
7.62113214e-01 1.22152112e-01 -5.32286167e-01 1.93906371e-02
-1.14696527e+00 -2.28395298e-01 4.46309507e-01 6.73413217e-01
8.97675514e-01 2.19820011e-02 -1.88122898e-01 1.06648672e+00
2.72877753e-01 3.55144650e-01 8.36699128e-01 -4.88686293e-01
6.36713982e-01 2.35548154e-01 1.22325867e-01 -8.19347620e-01
-1.47256956e-01 -1.05105197e+00 -1.06248236e+00 -1.16174646e-01
1.52009204e-01 -1.61183402e-01 -1.02935624e+00 1.86204755e+00
4.80596721e-01 8.70132089e-01 2.45688081e-01 1.10469270e+00
6.07933700e-01 7.38120258e-01 -1.50681511e-01 -2.35411301e-01
8.80604625e-01 -1.65774024e+00 -6.30981386e-01 -4.36939210e-01
6.28989279e-01 -6.47337079e-01 9.86618340e-01 3.24773669e-01
-1.00275016e+00 -9.68994141e-01 -1.13558114e+00 -4.07889709e-02
2.55009253e-02 3.95328134e-01 1.35382921e-01 3.21007460e-01
-1.02168941e+00 5.30109167e-01 -9.73828316e-01 -4.01091576e-02
3.80071580e-01 2.03163788e-01 -2.27527902e-01 -5.11270344e-01
-1.15346825e+00 8.43050778e-01 5.60591638e-01 4.41253126e-01
-1.31489480e+00 -4.51187611e-01 -7.77575493e-01 -1.97569709e-02
3.08034420e-01 -7.48096228e-01 1.03524578e+00 -1.21401000e+00
-1.63161922e+00 6.30359054e-01 -8.54744464e-02 -1.03355289e-01
6.97061419e-01 -7.39313304e-01 -4.01848018e-01 1.43297613e-01
7.67003372e-02 4.47466522e-01 1.31801033e+00 -1.57874727e+00
-3.78981441e-01 -3.55224073e-01 -1.39721751e-01 3.11352193e-01
-6.80568337e-01 -4.04019445e-01 -9.90088284e-01 -1.04043996e+00
1.35878161e-01 -8.71147633e-01 -1.52415216e-01 -3.97601724e-01
-3.07209700e-01 2.36420825e-01 1.18760002e+00 -9.19464469e-01
1.33226573e+00 -2.27238941e+00 5.27350605e-01 -1.24004833e-01
1.05358794e-01 6.72516584e-01 -5.88988423e-01 3.45021524e-02
-2.39351541e-01 -2.37927616e-01 -3.78856540e-01 -5.83767831e-01
-4.42216516e-01 6.07104115e-02 -2.46287227e-01 7.56750286e-01
1.03368662e-01 1.05771685e+00 -1.00312722e+00 -2.95575112e-01
3.57002288e-01 5.38745999e-01 -5.72596252e-01 6.64967060e-01
3.62191140e-03 8.79184902e-01 -4.57599163e-01 4.19718653e-01
9.62083578e-01 -4.95293468e-01 -1.47018299e-01 -3.58123094e-01
1.34371407e-02 -5.81431091e-02 -1.05609322e+00 2.02131510e+00
-8.22902620e-01 4.38391656e-01 2.93018937e-01 -1.16461837e+00
6.27924263e-01 2.70108968e-01 2.82682270e-01 -5.80768943e-01
3.30107324e-02 3.31695676e-01 -1.43887118e-01 -7.58883536e-01
4.08478260e-01 -1.41619574e-02 3.55028719e-01 5.95970154e-01
9.19460580e-02 1.41210854e-01 -1.55021548e-01 -3.12855877e-02
8.98149669e-01 3.91311407e-01 1.12609275e-01 -7.59662548e-03
9.98059273e-01 -2.55893558e-01 4.87818778e-01 6.54357553e-01
-8.59631076e-02 9.12296474e-01 -2.36644343e-01 -3.51785511e-01
-9.35294449e-01 -6.20757639e-01 -1.46326900e-01 8.65462899e-01
6.61941230e-01 -2.62786865e-01 -9.06444669e-01 -9.81047332e-01
-5.36815226e-01 5.48990369e-01 -5.42660773e-01 -3.29258442e-01
-8.09889972e-01 -1.09376550e+00 1.97266385e-01 3.63481253e-01
1.01501286e+00 -7.29566336e-01 -3.21851641e-01 3.26805085e-01
-4.60903168e-01 -1.23906910e+00 -7.83312023e-01 -2.44808160e-02
-1.10388458e+00 -8.25107515e-01 -1.28681064e+00 -9.06960726e-01
8.72089863e-01 8.47777545e-01 8.79144490e-01 3.34645927e-01
2.12442011e-01 3.59465599e-01 -4.81622577e-01 4.70696926e-01
-5.95166504e-01 2.11624682e-01 -2.04423621e-01 4.25277561e-01
-6.65815026e-02 -7.29405522e-01 -8.86507690e-01 6.77357674e-01
-1.09999347e+00 2.98381984e-01 7.34864473e-01 1.15818822e+00
4.30650800e-01 7.44959563e-02 6.03946447e-01 -5.21450460e-01
2.99991280e-01 -5.59856296e-01 -3.00213724e-01 4.24077898e-01
-4.44737792e-01 1.30258530e-01 6.88506722e-01 -7.00609028e-01
-1.22272539e+00 -6.95689321e-02 -8.39480907e-02 -7.45748401e-01
-1.49574950e-01 6.05853319e-01 -3.92022282e-01 -1.63164705e-01
4.39438820e-01 7.27072954e-01 -1.41281456e-01 -7.29458690e-01
2.03699857e-01 7.65706539e-01 7.00072467e-01 -4.31051552e-01
1.03264058e+00 3.65909249e-01 -3.67748946e-01 -5.57730317e-01
-1.08136690e+00 -5.37389934e-01 -6.84099376e-01 -2.72240490e-01
6.41271055e-01 -1.33180976e+00 -1.39702976e-01 1.13720393e+00
-9.83052909e-01 -7.36290991e-01 -4.50970866e-02 5.85711420e-01
-4.36702669e-01 9.27873909e-01 -4.31074113e-01 -3.33822131e-01
-3.59678686e-01 -1.27665520e+00 1.07202542e+00 1.56960502e-01
3.68382841e-01 -1.06347275e+00 1.94568738e-01 5.81738591e-01
5.95922291e-01 -1.58543438e-01 5.07999718e-01 -4.53029692e-01
-6.32728279e-01 -2.37286776e-01 -3.54138821e-01 7.62117922e-01
3.26069713e-01 -6.76461220e-01 -1.04568493e+00 -8.18998337e-01
3.30401182e-01 -5.60441554e-01 1.24886906e+00 2.05620840e-01
1.31622386e+00 -1.50953904e-01 -2.55341768e-01 8.41200411e-01
1.38589537e+00 -3.39499623e-01 7.76195526e-01 6.36829436e-01
1.02957261e+00 3.66995871e-01 5.94733417e-01 1.90450639e-01
4.96051162e-01 9.14220870e-01 4.50676650e-01 -2.22697660e-01
-5.25012970e-01 -9.49081108e-02 7.55971193e-01 1.04192150e+00
-4.17432822e-02 -3.32915962e-01 -8.25508654e-01 5.74522495e-01
-1.96253514e+00 -8.22182894e-01 -1.64239351e-02 2.35330343e+00
9.83267307e-01 -1.42819121e-01 -3.60237181e-01 -1.55833796e-01
8.73128772e-01 4.21095997e-01 -7.75374472e-01 4.87128258e-01
-1.67107657e-01 -1.59433216e-01 2.48094395e-01 4.83620852e-01
-1.31381476e+00 1.02793515e+00 5.24476910e+00 9.06168878e-01
-1.33319783e+00 4.81612384e-01 6.99691415e-01 1.28289297e-01
-1.76529020e-01 1.07472707e-02 -6.17813587e-01 6.56131566e-01
7.02171326e-01 2.53028125e-01 4.75154281e-01 7.02251077e-01
2.54477888e-01 -3.71830314e-02 -1.04725897e+00 8.69202673e-01
3.08821052e-01 -9.56113577e-01 -8.56619403e-02 -1.19909883e-01
1.11981976e+00 2.19696701e-01 1.36956766e-01 2.68466353e-01
-4.34101522e-02 -5.98699927e-01 6.43352747e-01 4.10943180e-01
7.19088733e-01 -4.85997081e-01 7.95920849e-01 6.02307081e-01
-9.79434311e-01 -1.03377558e-01 -1.88775375e-01 3.14309657e-01
2.31219366e-01 6.38188303e-01 -2.04585776e-01 8.64184260e-01
5.46798885e-01 1.11485004e+00 -7.24149048e-01 1.15288484e+00
-4.85452235e-01 6.89000547e-01 -1.79310292e-02 7.83993661e-01
2.44671643e-01 -1.05847463e-01 6.65214539e-01 1.08121920e+00
4.27594304e-01 4.44136448e-02 1.70554549e-01 6.28890455e-01
-1.14116728e-01 -5.80602884e-02 -1.49129003e-01 3.19944263e-01
1.62736192e-01 1.11796498e+00 -3.13351512e-01 -5.19515097e-01
-5.26911318e-01 1.57378292e+00 3.75930727e-01 5.85840285e-01
-9.23915684e-01 -6.90361038e-02 3.13064009e-01 -2.05901638e-01
5.20246446e-01 -2.98376679e-01 8.28184858e-02 -1.87760496e+00
1.09224811e-01 -1.00610089e+00 2.16794521e-01 -7.56813526e-01
-1.26089799e+00 6.38444424e-01 -2.29747608e-01 -1.62758303e+00
-1.88842565e-01 -2.13484839e-01 -7.79269278e-01 1.00059354e+00
-1.98175120e+00 -1.32830906e+00 -7.23861039e-01 7.18817294e-01
7.81030595e-01 -1.07198916e-01 4.20643955e-01 4.38021749e-01
-1.04246235e+00 6.69686437e-01 4.30917114e-01 7.90077224e-02
1.06855595e+00 -9.55776989e-01 3.76098365e-01 1.31998801e+00
-7.86989927e-02 5.25807500e-01 4.30408448e-01 -7.15520740e-01
-1.32667649e+00 -1.37339449e+00 2.97326684e-01 -6.69913888e-02
3.87623250e-01 1.27739698e-01 -1.28212428e+00 4.21428949e-01
2.79185146e-01 4.26637501e-01 5.05126178e-01 -1.01285368e-01
-3.76667857e-01 -1.23794027e-01 -7.33982623e-01 3.55071664e-01
8.35825026e-01 -5.25569618e-01 -4.88264531e-01 2.20011920e-01
6.03961349e-01 -4.63873357e-01 -6.85253501e-01 5.95548153e-01
2.50914186e-01 -9.00027931e-01 1.08052087e+00 -3.72421205e-01
6.43074393e-01 -4.92732584e-01 1.08337052e-01 -1.51653373e+00
-3.22741121e-01 -5.12554228e-01 -5.46232700e-01 1.22777414e+00
-4.52059060e-02 -5.94824970e-01 6.29353344e-01 2.16167316e-01
-4.02137488e-01 -5.88262260e-01 -6.94723547e-01 -8.80804300e-01
5.65192327e-02 -2.23883376e-01 2.91836679e-01 1.12834930e+00
-3.86221886e-01 3.23931217e-01 -7.89303184e-01 5.01827538e-01
8.54662359e-01 3.06998581e-01 9.19281244e-01 -7.18164086e-01
-6.26171172e-01 -1.29832417e-01 -7.01340809e-02 -1.67342937e+00
2.49277070e-01 -7.93353319e-01 1.86773553e-01 -1.32920861e+00
4.96400326e-01 -5.48341393e-01 -5.88883221e-01 3.79257321e-01
-7.73514569e-01 3.36217135e-01 6.02528565e-02 6.97877407e-01
-5.84703386e-01 8.96738172e-01 1.61035430e+00 -2.67388731e-01
-3.04501623e-01 -6.61131814e-02 -5.38827479e-01 5.45572698e-01
6.94686472e-01 -4.88091975e-01 -4.38370168e-01 -9.26243365e-01
-2.56103992e-01 9.38026085e-02 6.59198225e-01 -1.04080951e+00
4.15274292e-01 5.05359396e-02 3.77552778e-01 -5.53767622e-01
1.00816399e-01 -7.59114027e-01 1.02149071e-02 2.80351907e-01
-1.91545054e-01 -3.70669216e-01 1.98760107e-01 8.07897866e-01
-4.40232337e-01 -3.12620133e-01 9.99570191e-01 2.99648251e-02
-7.71914780e-01 3.88292342e-01 1.05595432e-01 -5.42387590e-02
7.71186113e-01 -1.12094820e-01 -3.08917373e-01 -3.16821247e-01
-6.80088818e-01 2.41331592e-01 6.51279271e-01 4.28637952e-01
5.81352115e-01 -1.25764489e+00 -7.52553821e-01 1.16395555e-01
-6.58439286e-03 1.78563491e-01 7.61998594e-01 1.15327346e+00
-1.82629153e-01 -4.11907807e-02 -2.82234907e-01 -6.71505868e-01
-1.01411712e+00 7.66546488e-01 4.25264746e-01 -3.98794234e-01
-6.17440104e-01 7.46729732e-01 5.90745807e-01 -1.86712757e-01
1.20043620e-01 8.38945135e-02 -1.34030849e-01 -1.12658009e-01
7.78796434e-01 2.50522524e-01 -5.20008057e-03 -7.70237684e-01
-7.36985505e-02 7.87324429e-01 -4.09842402e-01 -2.88400650e-02
1.47183239e+00 -6.22631133e-01 -2.02161312e-01 5.39956577e-02
1.48918200e+00 -3.20154913e-02 -1.66170967e+00 -7.42446005e-01
-3.57459515e-01 -5.85457683e-01 4.87159431e-01 -6.66930139e-01
-1.44703829e+00 9.32496071e-01 7.05190659e-01 -3.02010685e-01
1.58666122e+00 -1.47243261e-01 9.31043804e-01 3.57092142e-01
1.19629659e-01 -9.09486294e-01 4.90292341e-01 3.08858782e-01
8.98503006e-01 -1.37333345e+00 5.18663563e-02 -2.50987798e-01
-3.63107473e-01 1.15931392e+00 8.73733103e-01 -2.44295932e-02
4.70651150e-01 -4.64616790e-02 3.47204916e-02 8.91760960e-02
-4.88356203e-01 -1.37585342e-01 3.30127895e-01 1.68438762e-01
7.59061351e-02 -4.25059825e-01 -5.06179854e-02 4.27312970e-01
4.20074761e-01 1.84957877e-01 3.84874523e-01 8.47517908e-01
-3.34298044e-01 -1.13750684e+00 -4.60949421e-01 -7.67642707e-02
-3.00073832e-01 -2.53746033e-01 -1.68355610e-02 4.49472576e-01
7.53101632e-02 7.98366845e-01 -2.82034129e-01 -4.98811930e-01
7.68229738e-02 -2.74241805e-01 3.77932906e-01 -4.75912392e-01
-4.15473193e-01 2.17332318e-01 -2.39141032e-01 -4.61825103e-01
-6.68557882e-01 -5.45284390e-01 -6.15059197e-01 2.23206878e-02
-5.84476471e-01 1.07756555e-01 3.62374246e-01 9.54075754e-01
4.93045688e-01 4.92374420e-01 1.05466819e+00 -1.09526432e+00
-6.65047526e-01 -1.05734527e+00 -2.05898657e-01 2.26074502e-01
7.25483835e-01 -3.93768132e-01 -3.95353884e-01 2.98750371e-01]
|
[11.495757102966309, -2.5158350467681885]
|
d1141a2e-281f-4a72-b7b5-86ff2340468c
|
self-supervised-representations-improve-end
|
2006.12124
| null |
https://arxiv.org/abs/2006.12124v2
|
https://arxiv.org/pdf/2006.12124v2.pdf
|
Self-Supervised Representations Improve End-to-End Speech Translation
|
End-to-end speech-to-text translation can provide a simpler and smaller system but is facing the challenge of data scarcity. Pre-training methods can leverage unlabeled data and have been shown to be effective on data-scarce settings. In this work, we explore whether self-supervised pre-trained speech representations can benefit the speech translation task in both high- and low-resource settings, whether they can transfer well to other languages, and whether they can be effectively combined with other common methods that help improve low-resource end-to-end speech translation such as using a pre-trained high-resource speech recognition system. We demonstrate that self-supervised pre-trained features can consistently improve the translation performance, and cross-lingual transfer allows to extend to a variety of languages without or with little tuning.
|
['Juan Pino', 'Changhan Wang', 'Jiatao Gu', 'Anne Wu']
|
2020-06-22
| null | null | null | null |
['speech-to-text-translation']
|
['natural-language-processing']
|
[ 1.95861697e-01 1.91898614e-01 -5.49438357e-01 -5.18722534e-01
-1.52528858e+00 -6.43944800e-01 7.71066189e-01 -4.28931206e-01
-4.57965881e-01 8.37125123e-01 7.51136899e-01 -7.94927061e-01
5.18046737e-01 -2.91635752e-01 -7.25630760e-01 -2.10268974e-01
4.17589903e-01 8.06583643e-01 -2.24635080e-01 -4.40989792e-01
-5.35606325e-01 1.57698780e-01 -9.41551149e-01 3.49512398e-01
8.27804029e-01 4.16355491e-01 3.30919981e-01 4.91975874e-01
-1.38548061e-01 4.69896436e-01 -4.98387992e-01 -3.82137686e-01
3.43231440e-01 -6.42184258e-01 -8.36454928e-01 -7.82255083e-02
3.67902845e-01 -5.34603417e-01 -4.81975108e-01 5.53366363e-01
9.09356475e-01 1.43029690e-01 5.10214329e-01 -6.81015193e-01
-1.05323613e+00 7.94747651e-01 -9.71763879e-02 2.38273188e-01
2.77903110e-01 2.24624544e-01 8.72706294e-01 -1.27383769e+00
7.33965158e-01 1.22872961e+00 5.18488348e-01 8.34640324e-01
-1.10318816e+00 -6.50782287e-01 -3.26311082e-01 -1.02174386e-01
-9.92493868e-01 -1.35645497e+00 3.32461208e-01 -1.15424350e-01
1.47013414e+00 -8.87563452e-03 -2.86837686e-02 1.48575675e+00
-3.72542530e-01 8.33416581e-01 1.11290193e+00 -6.77703738e-01
-1.68341875e-01 1.09708205e-01 -4.52526659e-01 5.16511559e-01
-2.19517976e-01 2.80784935e-01 -8.17499816e-01 1.26116395e-01
3.63881081e-01 -1.87952235e-01 -2.05522016e-01 2.50320077e-01
-1.58016670e+00 8.52291286e-01 2.77395129e-01 4.82202321e-01
-1.02267697e-01 -3.92491464e-03 6.00861132e-01 6.57500088e-01
8.73658419e-01 6.63612306e-01 -8.73806298e-01 -5.97865820e-01
-1.09449875e+00 -3.69309217e-01 8.27168763e-01 1.12120903e+00
8.28211367e-01 5.41844726e-01 -2.57789135e-01 1.12804592e+00
-1.40173197e-01 9.62566972e-01 8.88200819e-01 -5.98562360e-01
1.03321695e+00 9.80848595e-02 -1.38460621e-01 -1.02065153e-01
-1.58153698e-01 -3.25771958e-01 -6.36342466e-01 -2.70122230e-01
3.37130338e-01 -5.71653128e-01 -1.02765036e+00 1.83806133e+00
2.14356761e-02 -6.25378489e-02 5.31991601e-01 9.75359619e-01
7.62537181e-01 9.28081930e-01 -1.31048426e-01 -3.00816894e-01
1.01967275e+00 -1.48043633e+00 -6.66083276e-01 -6.90451801e-01
8.67226064e-01 -1.07236111e+00 1.53402603e+00 -2.93856949e-01
-1.18325949e+00 -4.72800434e-01 -7.55554914e-01 -2.51456469e-01
-2.41325870e-01 3.71053219e-01 4.94426161e-01 6.80023849e-01
-1.33493280e+00 3.93392652e-01 -1.00883305e+00 -6.42411292e-01
3.62765372e-01 2.63053328e-01 -5.28259456e-01 -4.24612790e-01
-1.21021307e+00 1.23831666e+00 -2.02570166e-02 -1.75056189e-01
-8.71935189e-01 -6.25357747e-01 -7.83767521e-01 1.52996495e-01
4.58004475e-01 -7.76696026e-01 1.41878974e+00 -1.36256111e+00
-2.09164000e+00 7.77859390e-01 -3.83786112e-01 -4.03248727e-01
3.66995096e-01 -2.66272664e-01 -3.48482341e-01 -2.71982085e-02
5.60149625e-02 6.00601733e-01 7.50677586e-01 -6.22002065e-01
-2.68787116e-01 -2.53205419e-01 -2.54586607e-01 5.69320858e-01
-6.48533762e-01 3.59518439e-01 -4.68212783e-01 -6.49517477e-01
-1.38829067e-01 -1.12326348e+00 -9.33440700e-02 -2.65974760e-01
-2.42945492e-01 -7.42950067e-02 6.79444909e-01 -1.05608690e+00
6.19147360e-01 -2.05034399e+00 1.55735686e-01 -4.28726017e-01
-4.57833409e-01 6.78647041e-01 -4.81280386e-01 7.62189627e-01
2.91705757e-01 3.11750919e-01 -6.60744831e-02 -5.55771649e-01
-1.69507757e-01 1.96109861e-01 -3.62118244e-01 1.94589332e-01
6.26656890e-01 1.13167524e+00 -8.22267771e-01 -1.35284036e-01
3.32709984e-03 5.24148464e-01 -4.11857247e-01 4.35989648e-01
-1.68382525e-01 7.14455247e-01 -9.70102474e-02 5.13925195e-01
1.51656643e-01 -1.45561635e-01 1.25580162e-01 3.59700292e-01
1.14245698e-01 1.02454007e+00 -2.78605223e-01 2.00091434e+00
-1.01677823e+00 8.65007222e-01 7.08399490e-02 -8.39569807e-01
9.09762323e-01 6.22695506e-01 1.01061784e-01 -8.47097516e-01
-1.06700361e-02 5.65334141e-01 1.33294821e-01 -3.31408471e-01
5.12906790e-01 -4.01419252e-01 6.20208606e-02 7.57120788e-01
4.23928648e-01 -2.40679055e-01 -2.13214621e-01 1.56898387e-02
1.09891987e+00 1.39571935e-01 1.25573367e-01 -1.00960329e-01
1.38784215e-01 1.24614835e-01 3.69705468e-01 6.06117964e-01
-1.79440610e-03 6.36658669e-01 -2.55457431e-01 -1.24809310e-01
-1.29516113e+00 -8.48124087e-01 8.65932554e-02 1.63516748e+00
-3.95519972e-01 -1.92362890e-01 -6.49366617e-01 -7.71479189e-01
-8.19289908e-02 5.87775767e-01 1.40358144e-02 -2.30617136e-01
-6.46319032e-01 -4.16071385e-01 8.78456414e-01 6.18284166e-01
2.71156549e-01 -9.54217613e-01 1.18964188e-01 3.13782513e-01
-3.14095527e-01 -1.53637493e+00 -8.07150722e-01 3.42785597e-01
-7.81096101e-01 -1.25025496e-01 -1.13876832e+00 -9.87743318e-01
5.62581658e-01 5.09374380e-01 1.18302727e+00 -4.23616134e-02
3.34302604e-01 7.40865096e-02 -5.62993586e-01 -1.99581981e-01
-9.51233625e-01 7.12149143e-01 5.04319847e-01 -2.67300636e-01
2.21518874e-01 -4.95680064e-01 -2.36576363e-01 5.87043345e-01
-3.95772845e-01 9.24537778e-02 7.32424378e-01 1.14809668e+00
2.27876261e-01 -7.03596711e-01 9.81226683e-01 -7.65183628e-01
5.52079797e-01 -3.81947756e-01 -2.69460469e-01 2.93543309e-01
-7.20395148e-01 2.30481997e-01 9.92373765e-01 -5.78642845e-01
-9.74576473e-01 1.92258619e-02 -2.12917194e-01 -4.67726231e-01
2.27891337e-02 7.18919337e-01 -1.86470598e-01 1.65986925e-01
8.71355951e-01 2.23157704e-01 1.11739486e-01 -5.77060640e-01
8.68316352e-01 1.29209590e+00 3.52314353e-01 -6.08373940e-01
8.75040948e-01 4.05583754e-02 -4.53060001e-01 -6.87078238e-01
-7.03761518e-01 -3.31700981e-01 -6.05195820e-01 4.25980181e-01
6.32119656e-01 -1.46884584e+00 -6.36969646e-03 4.86034378e-02
-9.73285854e-01 -9.07524705e-01 -2.03412190e-01 6.31913722e-01
-6.48259878e-01 1.23882562e-01 -6.86348677e-01 -5.60971737e-01
-5.95665932e-01 -1.11315513e+00 1.15571880e+00 -8.11714977e-02
-1.20332256e-01 -8.90525877e-01 2.02366058e-02 7.26909935e-01
8.24486732e-01 -6.92339718e-01 6.83157086e-01 -1.02105975e+00
-4.48695600e-01 -2.20521484e-02 -1.61986709e-01 6.28399193e-01
4.17007446e-01 -2.60231525e-01 -1.08114171e+00 -6.23037636e-01
-3.79836529e-01 -7.95240343e-01 7.29671121e-01 -1.89318489e-02
4.30235118e-01 -5.55354059e-01 -5.56586459e-02 6.26590192e-01
8.02900851e-01 -1.81957126e-01 3.79209578e-01 -4.32268605e-02
7.44909763e-01 4.87400770e-01 4.15789008e-01 -1.62067175e-01
4.71439898e-01 9.79786932e-01 -3.22960079e-01 -3.07654977e-01
-6.62363589e-01 -4.50767756e-01 8.00417244e-01 1.54357624e+00
2.12209210e-01 -3.53635281e-01 -1.11651158e+00 6.49453163e-01
-1.66832256e+00 -7.77881980e-01 3.74161959e-01 2.11982775e+00
1.22288537e+00 -1.71563715e-01 2.48329490e-01 -4.70125407e-01
6.69778705e-01 7.71274790e-02 -5.52334547e-01 -4.43587661e-01
-2.35608742e-01 3.42038840e-01 4.90017980e-01 4.81077105e-01
-7.28873074e-01 1.64509130e+00 7.01120520e+00 8.11441660e-01
-1.51916039e+00 7.71064043e-01 5.69881380e-01 -7.80856609e-02
-4.00844425e-01 1.54882818e-01 -7.27503359e-01 2.94999182e-01
1.57265556e+00 -2.57165462e-01 8.45907509e-01 8.31415892e-01
2.13327676e-01 4.70034510e-01 -1.25202167e+00 9.25291002e-01
2.19117135e-01 -1.30812633e+00 -1.26454026e-01 4.97595482e-02
8.46119761e-01 7.42697358e-01 7.85082206e-02 7.04708517e-01
5.53830445e-01 -1.13872278e+00 5.25573671e-01 -2.01713979e-01
1.47273576e+00 -3.90993148e-01 7.18213379e-01 4.88536566e-01
-7.77347803e-01 1.77227646e-01 -4.53060001e-01 -1.71048909e-01
2.82999128e-01 2.21302748e-01 -1.52876234e+00 5.19318044e-01
1.98854670e-01 5.85890889e-01 -2.75614083e-01 4.36099976e-01
-3.61675739e-01 1.06512928e+00 -3.93814087e-01 -2.76514977e-01
3.33427489e-01 2.38062944e-02 2.41046622e-01 1.32465541e+00
6.36742949e-01 -3.27119082e-01 2.55249798e-01 5.02855122e-01
-5.76654613e-01 4.56182480e-01 -8.78154576e-01 -5.66871941e-01
6.12308443e-01 8.34538102e-01 -2.06108794e-01 -4.68052715e-01
-6.04049385e-01 1.24688172e+00 6.56203866e-01 5.84157526e-01
-3.39672267e-01 1.85550493e-03 8.38446021e-01 8.81390050e-02
2.46179089e-01 -4.40773606e-01 -3.16824347e-01 -1.51568377e+00
-6.23980956e-03 -1.27999210e+00 9.62691475e-03 -7.40404069e-01
-1.43123853e+00 8.92930627e-01 -5.81955075e-01 -1.12196565e+00
-7.31875062e-01 -4.73999590e-01 -4.32355672e-01 1.23163521e+00
-1.65261340e+00 -1.53967464e+00 2.69349009e-01 6.57320857e-01
9.71800208e-01 -7.68893480e-01 1.07980454e+00 3.06612730e-01
-4.43975270e-01 1.02841735e+00 5.53698897e-01 3.67726475e-01
1.10231161e+00 -8.00072670e-01 8.62300038e-01 9.58024740e-01
6.48849964e-01 4.90282923e-01 3.13203216e-01 -5.24135172e-01
-1.66652703e+00 -1.08744657e+00 1.10098243e+00 -5.47689199e-01
5.88557661e-01 -6.50265038e-01 -7.37894773e-01 7.89157033e-01
3.51486295e-01 2.60288179e-01 6.85378373e-01 5.90298355e-01
-6.06219947e-01 9.50126629e-03 -9.52194929e-01 8.13372791e-01
1.24925828e+00 -1.14644003e+00 -4.13496345e-01 5.97070694e-01
1.10348988e+00 -4.54840451e-01 -6.61115587e-01 1.26775846e-01
3.07732522e-01 -1.82327867e-01 8.50698650e-01 -1.05822790e+00
4.55963910e-01 2.37527609e-01 -4.54133421e-01 -1.81420386e+00
-1.77459687e-01 -1.19842970e+00 2.35402539e-01 1.40389144e+00
1.03193605e+00 -7.29602039e-01 4.96971816e-01 3.98191690e-01
-4.82560396e-01 -5.02826691e-01 -1.29959977e+00 -1.24512196e+00
5.55664539e-01 -1.51549950e-01 6.28856659e-01 1.05497742e+00
3.11595291e-01 8.33977878e-01 -6.60735786e-01 -4.46118042e-02
-5.82287759e-02 -5.85901476e-02 1.00881422e+00 -6.91000879e-01
-4.90069211e-01 -3.04451823e-01 -2.61202335e-01 -1.09525692e+00
6.61227107e-01 -1.40318143e+00 3.28806698e-01 -1.45731366e+00
2.84800418e-02 -6.07199907e-01 -2.49634199e-02 7.41192043e-01
-3.95980835e-01 1.49620220e-01 3.37258667e-01 4.34356213e-01
-1.84134513e-01 7.40653694e-01 1.14560902e+00 -1.24074645e-01
-2.34360084e-01 -8.67912844e-02 -7.55564094e-01 4.38937917e-02
7.64417708e-01 -6.60716772e-01 -3.31900567e-01 -9.52347100e-01
-1.39757678e-01 2.84171999e-01 -4.61983085e-01 -6.52431607e-01
5.19705536e-05 -5.97343780e-02 -9.67243984e-02 9.90418047e-02
4.44768727e-01 -4.85183537e-01 -3.39449227e-01 -6.19589631e-03
-4.35237348e-01 8.35038871e-02 2.39738703e-01 2.87904024e-01
-2.90110111e-01 -5.83099239e-02 5.39991438e-01 -2.02689454e-01
-1.11331120e-01 2.04196081e-01 -3.75321895e-01 4.35504258e-01
4.14000481e-01 2.87206680e-01 -5.68879902e-01 -8.72916639e-01
-4.14693058e-01 -7.41091073e-02 5.90672255e-01 7.12077916e-01
2.35795155e-01 -1.49976063e+00 -1.11434090e+00 2.17242941e-01
1.56317160e-01 -4.38540339e-01 -2.33452216e-01 7.29489803e-01
-1.30548537e-01 4.76974130e-01 2.85188239e-02 -6.58066988e-01
-9.76153672e-01 4.53029007e-01 1.58464357e-01 -1.62136197e-01
-5.32746971e-01 7.83763170e-01 -1.15003191e-01 -9.49806333e-01
8.30250084e-02 -1.56660900e-01 5.13402045e-01 -3.00680399e-01
3.50795776e-01 -7.03412741e-02 2.91064143e-01 -9.30759490e-01
-4.25564289e-01 1.87086374e-01 -2.41595320e-02 -6.10287845e-01
1.43449593e+00 -2.56121039e-01 4.17031497e-01 4.26003069e-01
1.31957042e+00 7.84125552e-02 -1.20775962e+00 -7.11342871e-01
-1.87661231e-01 -3.14324498e-01 1.89494327e-01 -1.09854639e+00
-9.96819317e-01 1.18454123e+00 4.02988970e-01 -2.60150105e-01
8.29964757e-01 9.60105211e-02 1.12565899e+00 7.59720087e-01
5.18914223e-01 -1.06430244e+00 1.35363743e-01 7.98950911e-01
7.85709918e-01 -1.66236591e+00 -3.85583967e-01 -2.90865958e-01
-9.70139384e-01 9.42370415e-01 2.53587753e-01 2.51310259e-01
3.15398008e-01 3.84264410e-01 6.00067794e-01 4.43097979e-01
-8.27336729e-01 -5.27181625e-01 2.66190171e-01 7.66364813e-01
7.18461812e-01 2.08002999e-01 -1.38304764e-02 5.10655046e-01
-4.41809833e-01 6.77837012e-03 3.71826410e-01 6.44076526e-01
-1.30151421e-01 -1.27174425e+00 -2.08067056e-02 2.17280209e-01
-5.22847116e-01 -5.97600043e-01 -4.97014552e-01 4.32257086e-01
-4.38772291e-01 1.32508457e+00 -3.84330675e-02 -4.08020377e-01
2.28122011e-01 5.41681647e-01 3.50584120e-01 -1.01105702e+00
-7.32625067e-01 2.52893239e-01 6.74168348e-01 -4.10414129e-01
-5.69347851e-02 -5.09777546e-01 -9.76350904e-01 -2.80488312e-01
-4.84997660e-01 -1.63762830e-02 9.68267798e-01 1.05372071e+00
8.23783159e-01 1.67113930e-01 8.53511930e-01 -8.70223463e-01
-9.80261624e-01 -1.35627437e+00 8.26302320e-02 3.72446835e-01
4.36258852e-01 -2.25719109e-01 -3.86327147e-01 1.11506410e-01]
|
[14.479103088378906, 7.139556407928467]
|
02f2be1a-a74e-4e26-a4ff-1940e7c525d6
|
multi-task-audio-source-separation
|
2107.06467
| null |
https://arxiv.org/abs/2107.06467v1
|
https://arxiv.org/pdf/2107.06467v1.pdf
|
Multi-Task Audio Source Separation
|
The audio source separation tasks, such as speech enhancement, speech separation, and music source separation, have achieved impressive performance in recent studies. The powerful modeling capabilities of deep neural networks give us hope for more challenging tasks. This paper launches a new multi-task audio source separation (MTASS) challenge to separate the speech, music, and noise signals from the monaural mixture. First, we introduce the details of this task and generate a dataset of mixtures containing speech, music, and background noises. Then, we propose an MTASS model in the complex domain to fully utilize the differences in spectral characteristics of the three audio signals. In detail, the proposed model follows a two-stage pipeline, which separates the three types of audio signals and then performs signal compensation separately. After comparing different training targets, the complex ratio mask is selected as a more suitable target for the MTASS. The experimental results also indicate that the residual signal compensation module helps to recover the signals further. The proposed model shows significant advantages in separation performance over several well-known separation models.
|
['Xiaorui Wang', 'Feng Deng', 'Chenxing Li', 'Lu Zhang']
|
2021-07-14
| null | null | null | null |
['audio-source-separation', 'music-source-separation']
|
['audio', 'music']
|
[ 2.45199472e-01 -7.53341556e-01 2.06555873e-01 9.41184722e-03
-1.29518473e+00 -3.94592285e-01 3.67813855e-01 -2.69807160e-01
-5.48337847e-02 4.57123339e-01 4.91215289e-01 6.62977174e-02
-2.27583125e-01 -2.62280833e-02 -2.62831450e-01 -1.02133965e+00
5.66171780e-02 -1.94132522e-01 9.34692845e-02 -1.12194844e-01
3.17052044e-02 3.57343763e-01 -1.80976117e+00 3.15540820e-01
1.19780254e+00 1.14158404e+00 3.61096889e-01 8.34179401e-01
-1.04087859e-01 6.83342934e-01 -9.07774568e-01 -6.44229958e-03
2.41023302e-01 -6.88141286e-01 -1.46684736e-01 -1.10470086e-01
3.65687042e-01 -1.99655145e-01 -3.84191960e-01 1.26259911e+00
1.14408863e+00 4.93566900e-01 5.40594935e-01 -1.21402264e+00
-1.51689604e-01 1.00524843e+00 -5.14310956e-01 3.86291534e-01
5.63648716e-02 -1.44391119e-01 7.98509657e-01 -1.10523021e+00
-3.06499869e-01 1.20261335e+00 6.70902848e-01 2.43737519e-01
-9.86167133e-01 -1.16028273e+00 -6.03397600e-02 3.29361349e-01
-1.43442702e+00 -9.99908328e-01 1.13011873e+00 -4.26759452e-01
6.42742574e-01 4.40383404e-01 3.71972561e-01 8.96075249e-01
-3.58943731e-01 1.06486392e+00 9.46274221e-01 -3.96106243e-01
9.40586701e-02 4.92668524e-02 2.79886518e-02 -1.29987761e-01
-1.41274720e-01 -1.89755727e-02 -7.34014988e-01 -1.99520394e-01
5.61150372e-01 -2.30586082e-01 -7.42301464e-01 1.35012373e-01
-1.05648315e+00 2.71250963e-01 2.13096872e-01 5.94445467e-01
-4.14842486e-01 -8.67381766e-02 2.10062563e-01 -5.28933387e-03
4.20097023e-01 2.85427570e-01 -1.95716754e-01 -1.93040799e-02
-1.38676798e+00 1.81625217e-01 4.78069156e-01 6.96739972e-01
1.47124827e-01 8.60210657e-01 -3.26729536e-01 1.53678513e+00
3.26777726e-01 6.00914776e-01 7.97175467e-01 -7.60609686e-01
5.41190803e-01 -1.32537827e-01 9.93983224e-02 -8.01706195e-01
-1.49979055e-01 -1.10658503e+00 -1.03210294e+00 8.53969380e-02
1.76372439e-01 -2.57717818e-01 -8.19759965e-01 1.67883015e+00
3.44265044e-01 7.73434818e-01 2.70350754e-01 1.04654872e+00
1.00038970e+00 9.42001820e-01 -2.69701183e-01 -3.79083872e-01
1.06528306e+00 -1.19990730e+00 -1.01840472e+00 -3.35561901e-01
-3.08925152e-01 -1.18142796e+00 6.53685212e-01 6.56115830e-01
-1.29792523e+00 -1.02361250e+00 -1.21284771e+00 1.52778059e-01
1.12514935e-01 5.70540726e-01 1.08638398e-01 6.16801739e-01
-8.43411684e-01 4.77149725e-01 -6.63518965e-01 3.42532575e-01
1.53512225e-01 1.13190971e-01 1.54312253e-01 1.08865000e-01
-1.19395912e+00 4.50778902e-01 8.13872293e-02 2.83100426e-01
-1.05267489e+00 -6.54658675e-01 -8.35995674e-01 4.72724378e-01
1.80580288e-01 -3.79543185e-01 1.42183089e+00 -1.13181114e+00
-1.60524380e+00 2.33746409e-01 -3.69626403e-01 -3.04251045e-01
1.81339905e-01 -4.96413469e-01 -9.20723200e-01 4.61891554e-02
2.37578750e-02 1.59532055e-01 1.43929970e+00 -1.28781211e+00
-7.51133561e-01 -1.31254464e-01 -5.97950101e-01 3.71535331e-01
-2.63288379e-01 4.29919660e-01 -4.50260222e-01 -1.27850831e+00
3.60692114e-01 -5.70916951e-01 2.24291570e-02 -6.36329353e-01
-4.72296536e-01 2.07025513e-01 3.75796676e-01 -9.87498939e-01
1.38764751e+00 -2.80357456e+00 2.30186790e-01 7.39309937e-02
8.99109021e-02 4.54109401e-01 -3.01462591e-01 1.43837541e-01
-4.69680667e-01 -3.58196735e-01 -2.58683473e-01 -7.19669759e-01
5.80376238e-02 -4.67635959e-01 -5.73878050e-01 1.96115226e-01
1.56650081e-01 2.91567326e-01 -6.06441617e-01 -1.41478390e-01
-4.67211381e-02 7.04525530e-01 -2.88356483e-01 3.85038227e-01
3.32843214e-01 5.58104455e-01 8.92045721e-02 4.72961277e-01
9.78751123e-01 2.09013045e-01 -1.40141949e-01 -3.23237896e-01
-1.30362839e-01 6.06667161e-01 -1.63049197e+00 1.42466342e+00
-2.18113512e-01 7.56301880e-01 8.94159079e-01 -7.22471297e-01
9.27005112e-01 6.80816650e-01 4.21310067e-01 -3.58854502e-01
2.49084368e-01 6.17317557e-01 4.49586362e-01 -3.34851563e-01
4.02141720e-01 -4.02813643e-01 2.37526432e-01 1.34752288e-01
2.43021101e-01 -2.50212938e-01 4.70873751e-02 -1.02018774e-01
6.21731758e-01 -1.62069708e-01 5.81932105e-02 9.21758860e-02
7.57331908e-01 -5.83854735e-01 8.45554113e-01 4.16206419e-01
-3.35548282e-01 8.38202953e-01 -9.51547772e-02 4.66942251e-01
-5.09516776e-01 -1.22183669e+00 9.42073539e-02 1.20975256e+00
3.93048227e-02 -3.39994997e-01 -8.31340909e-01 3.13875861e-02
-1.34090751e-01 6.78980768e-01 7.67625794e-02 -2.94099927e-01
-6.68196201e-01 -7.67101526e-01 8.13882291e-01 4.77913767e-01
5.09016573e-01 -7.98596978e-01 6.39454126e-02 1.80197239e-01
-5.63440263e-01 -9.86277044e-01 -7.31584191e-01 1.60704896e-01
-5.78747988e-01 -6.43672407e-01 -1.11207116e+00 -9.14408863e-01
9.14935693e-02 7.61129379e-01 5.07399440e-01 -3.77790689e-01
-3.60030048e-02 6.53079450e-02 -1.96497306e-01 -6.93639040e-01
-4.00068402e-01 -2.37214193e-01 4.54841137e-01 6.15319073e-01
-2.13853642e-02 -9.79267001e-01 -5.30800104e-01 2.44628549e-01
-8.42765570e-01 -8.79061744e-02 6.28353059e-01 4.97768581e-01
4.46047574e-01 6.90306067e-01 9.52914298e-01 -6.30439594e-02
8.94196749e-01 -4.52967465e-01 -3.55614007e-01 -1.64630964e-01
-1.67613238e-01 -3.56619209e-01 7.33127654e-01 -8.17066371e-01
-1.39329588e+00 -5.54414168e-02 -3.78621966e-01 -6.78163052e-01
-2.45599657e-01 4.76353288e-01 -6.28056705e-01 2.52136230e-01
4.60291564e-01 4.65777308e-01 -1.91743761e-01 -1.00648522e+00
1.07484888e-02 1.14067662e+00 8.01631093e-01 -2.18729138e-01
7.98835278e-01 2.02355072e-01 -4.49123621e-01 -9.48525846e-01
-6.90621078e-01 -7.39828706e-01 -2.87155509e-01 -6.19809888e-02
5.41219234e-01 -1.16814506e+00 -2.90084153e-01 1.01440501e+00
-1.11638486e+00 -2.13890225e-01 -1.81902096e-01 1.00940657e+00
-2.06268132e-01 3.67773980e-01 -6.73683465e-01 -1.33237505e+00
-4.10060674e-01 -1.28670013e+00 9.32097197e-01 4.37289238e-01
-2.75527164e-02 -4.73080248e-01 4.83518802e-02 4.26701188e-01
6.38064682e-01 -3.83967996e-01 5.63475132e-01 -8.13861072e-01
-4.09939200e-01 -9.92775038e-02 1.10645644e-01 7.78358102e-01
4.44632262e-01 -1.60829172e-01 -1.51669967e+00 -2.57190168e-01
5.04864156e-01 2.02978432e-01 9.23257113e-01 6.89898074e-01
8.72247159e-01 -6.19217828e-02 -7.46126920e-02 8.40839982e-01
7.72586405e-01 6.19715452e-01 4.93666112e-01 4.52374928e-02
5.66263258e-01 4.92441267e-01 4.32043552e-01 2.43940338e-01
6.06491752e-02 6.28025711e-01 1.77728087e-01 -3.31566423e-01
-5.20367861e-01 -6.57755183e-03 7.44196653e-01 1.45684791e+00
2.04884782e-01 -2.69896183e-02 -5.87521374e-01 6.12886667e-01
-1.44735909e+00 -9.77568269e-01 -1.10202298e-01 2.20305729e+00
9.86797273e-01 -3.70345563e-02 2.34624535e-01 7.61894882e-01
9.85649407e-01 2.66695231e-01 -5.72170019e-01 1.09371722e-01
-3.31616223e-01 3.90988797e-01 2.99258456e-02 4.27326262e-01
-1.22377360e+00 4.73673314e-01 6.41361523e+00 1.23625827e+00
-1.34771585e+00 1.06662586e-01 1.43624872e-01 -5.04962385e-01
1.30968653e-02 -3.66895288e-01 -5.81701279e-01 5.28746903e-01
8.70734215e-01 -1.38213709e-01 7.00447738e-01 4.55975533e-01
3.38129163e-01 1.08545750e-01 -8.16023111e-01 1.36943185e+00
2.31503099e-01 -6.57064557e-01 -1.08660616e-01 -4.44768041e-01
4.71923113e-01 -4.17767875e-02 3.27376932e-01 3.86205077e-01
-1.07259072e-01 -9.24155056e-01 1.06460750e+00 3.63658518e-01
5.78870118e-01 -7.59751320e-01 4.70235199e-01 5.91475248e-01
-1.34207797e+00 -4.05838609e-01 -1.39857978e-01 9.80108231e-02
1.94917068e-01 8.20274830e-01 -6.81843817e-01 8.03848863e-01
7.12045968e-01 5.50938129e-01 -1.54598266e-01 1.55912626e+00
-2.32404321e-01 9.92287338e-01 -1.72524214e-01 5.68712056e-01
-3.02797824e-01 -2.54830718e-01 1.07579172e+00 1.42870605e+00
5.97857773e-01 -4.68997732e-02 -7.93493763e-02 7.98087478e-01
8.06638226e-02 1.13534734e-01 -4.17477936e-02 -1.15881063e-01
5.45190334e-01 1.29009438e+00 -3.37061167e-01 -3.50191087e-01
-1.09798387e-01 7.27287054e-01 -2.66757011e-01 6.83444202e-01
-9.65755165e-01 -7.89601326e-01 9.00288999e-01 -1.90433994e-01
3.22711915e-01 -8.65914375e-02 -1.59873515e-01 -1.10966694e+00
1.93755403e-02 -1.25171602e+00 6.99498951e-02 -9.88256454e-01
-1.02291882e+00 7.24337339e-01 -1.98231220e-01 -1.49078095e+00
-1.40036106e-01 -3.83948714e-01 -8.46534669e-01 1.31400764e+00
-1.57987702e+00 -7.13519454e-01 -1.90768763e-01 6.08924806e-01
8.28889310e-01 -3.76108915e-01 5.24616241e-01 8.71332049e-01
-8.16240370e-01 5.45447707e-01 3.72730017e-01 1.51443809e-01
9.52378333e-01 -1.00166965e+00 1.50768533e-01 1.11557615e+00
1.15032613e-01 6.11826241e-01 6.47102416e-01 -4.40597832e-01
-9.37156022e-01 -1.02081573e+00 5.06093323e-01 1.56682417e-01
4.48335141e-01 -3.11428994e-01 -1.07649255e+00 2.95402825e-01
1.73777878e-01 -3.65720153e-01 9.15622890e-01 -2.11891741e-01
-3.25768024e-01 -5.27267396e-01 -6.82410121e-01 4.21517998e-01
6.73612356e-01 -5.98273635e-01 -7.76896536e-01 -1.99633494e-01
7.01711357e-01 -4.41362679e-01 -3.53801191e-01 3.89446497e-01
3.88851404e-01 -8.99567306e-01 1.20766616e+00 -3.08447927e-01
2.15356618e-01 -6.93062782e-01 -2.82927632e-01 -1.78032982e+00
-4.17090654e-01 -8.55879009e-01 -1.67859152e-01 1.78505898e+00
2.18863189e-01 -5.07077336e-01 9.40943584e-02 2.67441701e-02
-4.33415771e-01 -1.42718107e-01 -7.36024380e-01 -8.10785711e-01
-2.53630638e-01 -5.81665277e-01 7.48272181e-01 8.88425946e-01
-7.53612891e-02 5.09165764e-01 -4.85964984e-01 4.38949913e-01
5.53313076e-01 1.96135685e-01 6.84641540e-01 -1.21865058e+00
-6.10496521e-01 -7.37978876e-01 4.20790911e-02 -1.20677054e+00
2.97789872e-02 -8.62137139e-01 3.99055541e-01 -1.49194193e+00
-1.77825987e-01 -1.10103436e-01 -6.23493671e-01 -1.36887338e-02
-6.58099592e-01 4.28441204e-02 3.77594560e-01 3.08797330e-01
-2.59823769e-01 7.98180699e-01 9.00708854e-01 -2.86116958e-01
-4.95951653e-01 5.45050204e-01 -9.41288590e-01 9.30551052e-01
6.84705496e-01 -3.06737274e-01 -4.64313030e-01 -4.52111721e-01
-2.92851329e-01 2.38143131e-01 1.65467948e-01 -1.26008296e+00
3.39053214e-01 9.33635980e-02 3.54162365e-01 -5.78599036e-01
8.07302713e-01 -7.31096506e-01 2.58186579e-01 4.16897573e-02
-3.35360169e-01 -6.23838484e-01 5.11060297e-01 3.71181726e-01
-7.43705511e-01 -1.13681719e-01 9.79245901e-01 2.88758934e-01
-1.28062233e-01 -9.15170461e-02 -4.74444926e-01 -1.45052925e-01
4.74597335e-01 -3.13349403e-02 -4.87250574e-02 -6.87549651e-01
-9.23079729e-01 -4.20594290e-02 -3.41767281e-01 3.50933045e-01
6.30610526e-01 -1.39716291e+00 -9.61197495e-01 3.86698574e-01
-4.87057775e-01 4.56223972e-02 5.92946470e-01 1.10205197e+00
1.87882960e-01 1.86463252e-01 -5.59932925e-02 -4.06671554e-01
-1.44690478e+00 4.82218057e-01 6.54624999e-01 1.06436864e-01
-2.09766954e-01 1.04354918e+00 5.52600682e-01 -9.13267210e-02
5.64983845e-01 -3.00647259e-01 -5.18985391e-01 1.53726295e-01
8.72389197e-01 8.82358849e-01 9.89618674e-02 -8.79494190e-01
-2.50768393e-01 4.00429338e-01 2.40140811e-01 -3.61020923e-01
1.33118784e+00 -2.69778430e-01 -2.65942097e-01 6.18816316e-01
9.69605327e-01 7.58818448e-01 -1.03316891e+00 -3.48267585e-01
-1.55249521e-01 -4.60720658e-01 1.79231301e-01 -7.76783466e-01
-1.02758265e+00 1.21952081e+00 5.50412953e-01 4.26457167e-01
1.72351432e+00 -2.63089955e-01 8.70909572e-01 -2.75657386e-01
-1.41969487e-01 -9.26404297e-01 5.59517033e-02 4.99386132e-01
1.01227951e+00 -7.09058821e-01 -4.02107179e-01 -4.63621080e-01
-4.48287398e-01 7.78640449e-01 4.65256095e-01 6.11142516e-02
4.54605430e-01 4.83460605e-01 3.45954180e-01 3.25848460e-01
-2.90082276e-01 -3.59855890e-01 7.26033568e-01 5.64226389e-01
4.64062423e-01 -1.57303393e-01 1.98478028e-01 1.40064561e+00
-4.32391763e-01 -3.28108400e-01 2.08482161e-01 4.14946526e-01
-6.32320344e-01 -8.90136719e-01 -1.15650213e+00 1.48395255e-01
-7.09817052e-01 -3.46725792e-01 -4.48560685e-01 -6.45885840e-02
5.97261898e-02 1.37015486e+00 -1.78681150e-01 -4.17960703e-01
5.52460730e-01 3.83196741e-01 8.77065137e-02 -4.43126261e-01
-5.82761824e-01 1.03723049e+00 -2.09288284e-01 -9.47733670e-02
-3.64049673e-01 -4.94839251e-01 -1.19441462e+00 2.47929513e-01
-5.59503734e-01 4.05247629e-01 6.64836705e-01 6.67214215e-01
4.14132506e-01 1.06912470e+00 7.86011875e-01 -1.24671793e+00
-6.00869596e-01 -1.25482082e+00 -8.38714719e-01 2.48003498e-01
9.09585834e-01 -6.01833105e-01 -6.96523666e-01 6.93300441e-02]
|
[14.995284080505371, 5.818770408630371]
|
06812e5e-831b-4bc8-bca3-298c9ce235b1
|
non-linearities-improve-originet-based-on
|
2005.07991
| null |
https://arxiv.org/abs/2005.07991v1
|
https://arxiv.org/pdf/2005.07991v1.pdf
|
Non-Linearities Improve OrigiNet based on Active Imaging for Micro Expression Recognition
|
Micro expression recognition (MER)is a very challenging task as the expression lives very short in nature and demands feature modeling with the involvement of both spatial and temporal dynamics. Existing MER systems exploit CNN networks to spot the significant features of minor muscle movements and subtle changes. However, existing networks fail to establish a relationship between spatial features of facial appearance and temporal variations of facial dynamics. Thus, these networks were not able to effectively capture minute variations and subtle changes in expressive regions. To address these issues, we introduce an active imaging concept to segregate active changes in expressive regions of a video into a single frame while preserving facial appearance information. Moreover, we propose a shallow CNN network: hybrid local receptive field based augmented learning network (OrigiNet) that efficiently learns significant features of the micro-expressions in a video. In this paper, we propose a new refined rectified linear unit (RReLU), which overcome the problem of vanishing gradient and dying ReLU. RReLU extends the range of derivatives as compared to existing activation functions. The RReLU not only injects a nonlinearity but also captures the true edges by imposing additive and multiplicative property. Furthermore, we present an augmented feature learning block to improve the learning capabilities of the network by embedding two parallel fully connected layers. The performance of proposed OrigiNet is evaluated by conducting leave one subject out experiments on four comprehensive ME datasets. The experimental results demonstrate that OrigiNet outperformed state-of-the-art techniques with less computational complexity.
|
['Monu Verma', 'Santosh Kumar Vipparthi', 'Girdhari Singh']
|
2020-05-16
| null | null | null | null |
['micro-expression-recognition']
|
['computer-vision']
|
[ 1.88113943e-01 -2.26669878e-01 -2.25183547e-01 -5.13616264e-01
-1.05487816e-01 -1.92032710e-01 3.31530511e-01 -5.48657835e-01
-5.46540856e-01 5.96910059e-01 1.01570040e-01 4.69646335e-01
-2.16302320e-01 -4.93332267e-01 -7.30407894e-01 -9.43653941e-01
-3.27246815e-01 -4.90689754e-01 9.96944457e-02 -6.63418472e-01
-7.05148801e-02 7.99941480e-01 -1.30916965e+00 3.49884182e-01
5.79156399e-01 1.26855946e+00 2.97396183e-02 3.32093149e-01
-2.55947784e-02 1.21038008e+00 -2.39051685e-01 -9.74035710e-02
1.62640467e-01 -3.79206806e-01 -3.52725595e-01 1.32054165e-01
4.44504231e-01 -5.75895786e-01 -6.49151623e-01 9.73155558e-01
5.74634790e-01 2.52211541e-01 3.67971718e-01 -1.30267298e+00
-8.22785735e-01 1.17430694e-01 -9.81268942e-01 4.84163105e-01
1.68678239e-01 -3.36037129e-02 6.73135281e-01 -1.09591544e+00
8.24336410e-01 1.09475255e+00 8.89192224e-01 6.13154292e-01
-1.08871007e+00 -8.27714920e-01 2.93751419e-01 2.95057476e-01
-1.47027218e+00 -6.06612802e-01 1.26612186e+00 -2.08193794e-01
9.75969911e-01 1.37462094e-01 7.14740157e-01 1.21422482e+00
3.33017826e-01 7.89128661e-01 1.07932925e+00 -1.50849074e-01
-1.42499477e-01 -2.46872976e-02 6.49216473e-02 9.95607734e-01
-3.05195570e-01 -1.57279428e-02 -6.64956152e-01 4.05446514e-02
1.13617957e+00 2.19716251e-01 -5.91183126e-01 -2.59995311e-01
-7.74497867e-01 5.99806249e-01 6.52399361e-01 5.75155079e-01
-5.32915890e-01 2.81073540e-01 4.03411537e-01 3.78665209e-01
6.49189591e-01 -4.49932292e-02 -4.08333063e-01 -1.17304899e-01
-8.14376771e-01 -8.56329054e-02 2.91453391e-01 5.00869334e-01
8.58284235e-01 4.54519510e-01 -3.01560597e-03 9.00550842e-01
7.88687989e-02 1.67145893e-01 5.29902101e-01 -9.67381716e-01
1.27997711e-01 7.97480524e-01 -2.42302582e-01 -1.39915442e+00
-5.18312216e-01 -2.66870946e-01 -1.13753176e+00 4.87159401e-01
9.00998339e-02 -2.14361444e-01 -7.92738616e-01 2.04804826e+00
1.93214640e-01 3.49078625e-01 -6.19144887e-02 1.04401636e+00
9.60938096e-01 8.17696691e-01 -2.95333471e-02 -5.74839711e-01
9.06980395e-01 -9.75438774e-01 -1.06688571e+00 1.17995478e-01
4.32303488e-01 -4.21166569e-01 7.63860643e-01 2.21593186e-01
-1.06699860e+00 -7.32673287e-01 -8.83865178e-01 -6.82824552e-02
-3.01845551e-01 8.38487446e-02 8.48989785e-01 1.39992610e-01
-1.17980385e+00 6.59111381e-01 -8.46372545e-01 -7.09615275e-02
4.98183131e-01 8.24847281e-01 -9.50968146e-01 3.50885093e-01
-1.19231796e+00 5.91064453e-01 4.26494814e-02 6.67546213e-01
-6.14996672e-01 -6.80076718e-01 -1.04245031e+00 1.97152868e-02
2.47235984e-01 -1.91681668e-01 8.39723587e-01 -1.91080093e+00
-1.86158741e+00 7.00088322e-01 -1.88163444e-01 -8.78279284e-02
3.97954375e-01 -1.55073211e-01 -5.43444991e-01 3.93471926e-01
-4.68795508e-01 7.08393395e-01 9.49544549e-01 -8.91077578e-01
-2.01674193e-01 -5.54756880e-01 -1.82073843e-02 1.58633247e-01
-6.57763362e-01 2.24428132e-01 -2.64577806e-01 -8.10940564e-01
2.57786885e-02 -7.31970668e-01 -8.25851113e-02 5.07561326e-01
2.28712529e-01 -5.77291921e-02 1.29802787e+00 -6.34455264e-01
1.14742255e+00 -2.41178060e+00 3.43235046e-01 1.72428992e-02
3.91151696e-01 3.76639873e-01 -3.11316907e-01 5.13386615e-02
-4.07512188e-01 -1.07324712e-01 -6.95705712e-02 -2.35199332e-01
-3.17575812e-01 2.80395448e-01 -2.39986047e-01 6.39222324e-01
3.82084370e-01 1.14508855e+00 -6.77557826e-01 -3.73399705e-01
1.37344189e-02 9.89567876e-01 -4.81270820e-01 2.28143960e-01
1.27370223e-01 5.53220391e-01 -3.73190939e-01 7.45916843e-01
7.49144375e-01 9.21053961e-02 -1.22283474e-01 -3.90473396e-01
-2.01579556e-01 -4.40767586e-01 -1.00899386e+00 1.77932072e+00
-3.50231707e-01 7.05444038e-01 3.96242917e-01 -1.20483792e+00
1.01850462e+00 4.85995710e-01 8.46004426e-01 -9.36161697e-01
2.94914156e-01 1.47180617e-01 -8.46319720e-02 -7.94802666e-01
2.17248201e-01 -1.48552462e-01 3.30536753e-01 1.42104000e-01
4.29407775e-01 5.81758559e-01 -1.57311708e-01 -2.22477585e-01
9.08192098e-01 3.93422455e-01 2.50218600e-01 -6.50839880e-02
6.06372714e-01 -5.10360956e-01 9.51472998e-01 1.62800491e-01
-3.54000449e-01 4.16034967e-01 5.74420750e-01 -6.91954970e-01
-7.16116071e-01 -7.64522076e-01 -1.77982189e-02 1.19416153e+00
1.00450359e-01 -1.38278231e-01 -5.04004538e-01 -6.67967975e-01
-2.18788102e-01 -4.30381298e-02 -1.01319039e+00 -2.72064269e-01
-6.84022605e-01 -8.53476107e-01 5.80216527e-01 7.31220901e-01
9.46405649e-01 -1.25922692e+00 -5.33009291e-01 1.46762088e-01
1.23696951e-02 -1.08142483e+00 -6.52702391e-01 1.18781216e-01
-8.08955371e-01 -7.59083390e-01 -8.17697883e-01 -8.41710567e-01
8.12385678e-01 1.78074464e-01 5.34741163e-01 -1.31235987e-01
-3.80221099e-01 1.68008432e-01 -4.04328138e-01 -6.41585216e-02
7.72038102e-02 -3.63602042e-01 1.25413388e-01 6.00152493e-01
2.85190940e-01 -9.43985641e-01 -5.85615695e-01 1.98299706e-01
-9.18088555e-01 4.03658599e-02 6.72572851e-01 1.00398421e+00
5.60009420e-01 1.67874387e-03 5.41158915e-01 -5.72387695e-01
3.84058326e-01 -3.64551365e-01 -3.19978863e-01 1.04871340e-01
-2.91640431e-01 -1.55557588e-01 6.26390994e-01 -8.52730870e-01
-1.24890113e+00 2.59345472e-01 -2.06769422e-01 -7.68389881e-01
3.12721692e-02 3.95627350e-01 -1.27431944e-01 -6.40491307e-01
3.47159654e-01 3.52023244e-01 3.00290763e-01 -2.85525173e-01
2.71248836e-02 4.88626450e-01 5.85475028e-01 -2.44256601e-01
4.50257778e-01 6.51715338e-01 5.80287073e-03 -9.78158593e-01
-6.22940421e-01 -2.86066175e-01 -8.43275547e-01 -3.64100456e-01
6.70309484e-01 -9.34675872e-01 -7.44712889e-01 7.64254868e-01
-1.07165241e+00 -3.77810448e-01 -2.94747561e-01 3.85091513e-01
-6.15372539e-01 3.96774888e-01 -8.48779500e-01 -7.44101286e-01
-3.68676662e-01 -9.56232667e-01 9.26963568e-01 3.21258277e-01
-9.76554081e-02 -1.03889573e+00 1.54186472e-01 2.78851185e-02
5.00251710e-01 8.36743176e-01 5.89613676e-01 -5.34454435e-02
-2.43260205e-01 -1.63258567e-01 -1.45586863e-01 5.29313087e-01
2.11660951e-01 2.26035073e-01 -1.02480614e+00 -2.75182724e-01
1.36387706e-01 -4.23457742e-01 8.82236958e-01 4.48691994e-01
1.19915223e+00 -5.31336844e-01 4.35518771e-02 1.12340629e+00
1.39118469e+00 2.93827415e-01 8.18262696e-01 1.50588647e-01
6.94410563e-01 6.47459447e-01 2.17639938e-01 4.84934211e-01
-6.95660934e-02 7.91768789e-01 3.52502674e-01 -6.18220448e-01
-5.83635084e-02 -4.11962569e-02 7.24596918e-01 8.07000995e-01
-7.04773366e-01 1.94416746e-01 -3.42101574e-01 3.29988390e-01
-1.93739259e+00 -1.14318216e+00 1.38992891e-01 1.78896046e+00
7.57448614e-01 -6.88141957e-02 -9.61963311e-02 1.43590331e-01
4.22898054e-01 3.34288627e-01 -5.76247334e-01 -5.24875581e-01
-5.41908860e-01 3.93010080e-01 1.39558256e-01 3.15076441e-01
-9.78974104e-01 8.28917921e-01 5.84767485e+00 7.27972209e-01
-1.68330181e+00 2.83243746e-01 6.34693384e-01 -7.24629462e-02
8.07349011e-02 -3.93030435e-01 -4.92046148e-01 2.86819637e-01
5.69881499e-01 1.31169006e-01 3.00257832e-01 8.01266134e-01
4.75475907e-01 2.77340293e-01 -7.21904457e-01 1.19413626e+00
1.41053110e-01 -1.13351929e+00 -1.00454371e-02 -9.12331939e-02
6.96085632e-01 -2.57646799e-01 2.38233000e-01 1.67329669e-01
-3.30348134e-01 -1.20242143e+00 4.92290586e-01 7.61864066e-01
9.55146432e-01 -8.25793326e-01 7.62281537e-01 1.25100002e-01
-1.23671877e+00 -2.94615269e-01 -2.95256138e-01 -3.37657303e-01
-7.80300349e-02 6.61628991e-02 -1.64282605e-01 2.79308498e-01
7.58926034e-01 1.05152214e+00 -3.27211410e-01 5.31835616e-01
3.17628384e-02 3.82308155e-01 -2.16251686e-01 2.04898104e-01
3.74197572e-01 -2.53365397e-01 5.01391470e-01 1.24064386e+00
2.56415695e-01 3.33161116e-01 -1.86193734e-01 8.91955137e-01
-8.10021311e-02 2.94791520e-01 -7.32162535e-01 -1.12873100e-01
-3.16311121e-01 1.58253312e+00 -3.98780644e-01 1.54287353e-01
-5.11616349e-01 1.39520979e+00 4.91870791e-01 6.94235981e-01
-7.89042592e-01 -3.56969535e-01 5.90726435e-01 1.53723583e-01
2.14122653e-01 -2.40345165e-01 3.07836235e-01 -1.25336993e+00
1.37613997e-01 -5.89052916e-01 1.28933027e-01 -7.65618443e-01
-9.29731071e-01 8.31945300e-01 -1.33639559e-01 -1.05436623e+00
-1.86002746e-01 -7.56145060e-01 -7.44971335e-01 5.82561672e-01
-1.61290610e+00 -1.34252214e+00 -7.51127183e-01 1.00313890e+00
4.92747545e-01 -2.07833305e-01 8.68327141e-01 4.39808011e-01
-7.66789556e-01 7.36016810e-01 -3.86925116e-02 3.36309642e-01
5.70693076e-01 -7.62139797e-01 -3.06944251e-01 5.77763081e-01
3.88764217e-03 5.72150528e-01 3.98204237e-01 -2.73742527e-01
-1.43977034e+00 -9.44489241e-01 6.22587740e-01 1.88911870e-01
6.15405500e-01 -3.85617882e-01 -9.93693352e-01 7.61628807e-01
1.72403008e-01 6.82345271e-01 5.77423930e-01 -2.47512862e-01
-3.23643446e-01 -4.66605604e-01 -1.14715171e+00 3.96397471e-01
1.06419241e+00 -4.57375735e-01 -3.21389735e-01 -6.60105869e-02
4.08679008e-01 -2.78343379e-01 -7.98310757e-01 8.82453918e-01
7.47712493e-01 -1.10616910e+00 7.17322230e-01 -5.28398275e-01
4.61268306e-01 -8.64441246e-02 -1.17431052e-01 -1.07551861e+00
-5.05999923e-01 -8.09213817e-01 -3.95811886e-01 1.15675032e+00
4.43160310e-02 -5.60592830e-01 7.32457161e-01 2.75470912e-01
-1.73047636e-04 -1.27206206e+00 -1.01023018e+00 -4.02848810e-01
-2.77333945e-01 -3.76784839e-02 1.52269617e-01 1.12795401e+00
3.04480549e-02 1.68006346e-01 -5.81566632e-01 -1.73678368e-01
2.98304677e-01 -2.01973096e-01 4.22247976e-01 -9.08981860e-01
-1.21384382e-01 -4.17924255e-01 -8.98260951e-01 -8.35122883e-01
4.80571032e-01 -6.07416391e-01 -2.84737319e-01 -9.59163070e-01
3.57443601e-01 -1.09461375e-01 -5.85853517e-01 7.21834362e-01
1.06541537e-01 6.69781744e-01 -3.44194658e-02 1.04046524e-01
-3.64489138e-01 9.47031081e-01 1.31080556e+00 -1.54875860e-01
-3.07878613e-01 -2.81153291e-01 -3.69571030e-01 8.61456752e-01
6.44038796e-01 -8.96179006e-02 -5.42613745e-01 -3.47951204e-01
6.61617890e-02 2.85951234e-02 4.90889341e-01 -8.01903486e-01
2.06128180e-01 -7.58883506e-02 7.10153341e-01 -2.04806566e-01
5.51763237e-01 -9.00666833e-01 1.33303910e-01 1.09230757e-01
-2.51675546e-01 7.75811598e-02 3.30161721e-01 3.48901719e-01
-5.56399345e-01 1.66340813e-01 9.01037335e-01 -1.00120865e-01
-1.18680918e+00 6.64056838e-01 -2.06145376e-01 -3.98164660e-01
1.17734420e+00 -5.91782928e-01 1.11922391e-01 -3.56831938e-01
-8.48818958e-01 -4.99970354e-02 1.65106967e-01 5.39149880e-01
9.38680410e-01 -1.49067545e+00 -5.02962530e-01 4.36978638e-01
-8.11664760e-02 -2.61962354e-01 6.31170630e-01 1.14065087e+00
-5.59597671e-01 1.01846583e-01 -7.50003755e-01 -4.20174092e-01
-1.44471478e+00 4.44885194e-01 7.96922684e-01 -3.27646621e-02
-7.17169285e-01 8.53598475e-01 4.93376762e-01 -2.69181281e-01
2.64024615e-01 -7.30208308e-02 -6.01695597e-01 1.60268337e-01
6.79769278e-01 6.58138469e-02 -2.37582505e-01 -1.16588950e+00
-1.62992269e-01 9.60035324e-01 -1.11262955e-01 2.20159322e-01
1.65898263e+00 -1.48381561e-01 -3.43892545e-01 4.52722460e-01
1.61570358e+00 -1.67902023e-01 -1.58530200e+00 -3.02020162e-01
-3.68876487e-01 -3.91909659e-01 3.47450167e-01 -3.64143610e-01
-1.57739997e+00 8.44143569e-01 9.06191826e-01 -4.52895761e-01
1.64857018e+00 -4.82554764e-01 7.04497457e-01 1.33814141e-01
7.11348578e-02 -1.01221967e+00 4.16704267e-01 3.40913236e-01
1.06283081e+00 -1.20603120e+00 -1.82958573e-01 -3.27657580e-01
-6.11628592e-01 1.50956511e+00 8.77961457e-01 -2.38219216e-01
7.69794405e-01 5.00167549e-01 8.20306242e-02 -3.66148829e-01
-6.68716550e-01 4.04088572e-02 3.36099684e-01 2.80605108e-01
5.21222353e-01 -4.40444171e-01 -2.43837982e-01 5.52758455e-01
2.24316016e-01 4.17378932e-01 2.06248462e-01 9.36404228e-01
-1.72584355e-01 -5.84840298e-01 5.22508519e-03 -6.42578816e-03
-7.03397751e-01 1.73515528e-01 -3.16411555e-01 9.57511246e-01
4.40675825e-01 4.21326220e-01 2.97415525e-01 -5.12318134e-01
3.09143871e-01 -5.23849875e-02 5.76318324e-01 -1.03050731e-01
-3.38643372e-01 2.84764051e-01 -3.40260684e-01 -7.67358780e-01
-7.24700034e-01 -3.30573976e-01 -1.32438636e+00 -2.10008934e-01
-1.71503961e-01 -1.42085269e-01 2.89918214e-01 7.14979351e-01
2.44033664e-01 5.55536389e-01 8.80240440e-01 -9.61601138e-01
-3.26282233e-01 -9.76297140e-01 -8.38124990e-01 4.86568809e-01
6.60702229e-01 -7.76188672e-01 -3.00659984e-01 1.91042751e-01]
|
[13.635526657104492, 1.7515418529510498]
|
ce434b1b-feac-417f-934f-184c84d98449
|
groupvit-semantic-segmentation-emerges-from
|
2202.11094
| null |
https://arxiv.org/abs/2202.11094v5
|
https://arxiv.org/pdf/2202.11094v5.pdf
|
GroupViT: Semantic Segmentation Emerges from Text Supervision
|
Grouping and recognition are important components of visual scene understanding, e.g., for object detection and semantic segmentation. With end-to-end deep learning systems, grouping of image regions usually happens implicitly via top-down supervision from pixel-level recognition labels. Instead, in this paper, we propose to bring back the grouping mechanism into deep networks, which allows semantic segments to emerge automatically with only text supervision. We propose a hierarchical Grouping Vision Transformer (GroupViT), which goes beyond the regular grid structure representation and learns to group image regions into progressively larger arbitrary-shaped segments. We train GroupViT jointly with a text encoder on a large-scale image-text dataset via contrastive losses. With only text supervision and without any pixel-level annotations, GroupViT learns to group together semantic regions and successfully transfers to the task of semantic segmentation in a zero-shot manner, i.e., without any further fine-tuning. It achieves a zero-shot accuracy of 52.3% mIoU on the PASCAL VOC 2012 and 22.4% mIoU on PASCAL Context datasets, and performs competitively to state-of-the-art transfer-learning methods requiring greater levels of supervision. We open-source our code at https://github.com/NVlabs/GroupViT .
|
['Xiaolong Wang', 'Jan Kautz', 'Thomas Breuel', 'Wonmin Byeon', 'Sifei Liu', 'Shalini De Mello', 'Jiarui Xu']
|
2022-02-22
| null |
http://openaccess.thecvf.com//content/CVPR2022/html/Xu_GroupViT_Semantic_Segmentation_Emerges_From_Text_Supervision_CVPR_2022_paper.html
|
http://openaccess.thecvf.com//content/CVPR2022/papers/Xu_GroupViT_Semantic_Segmentation_Emerges_From_Text_Supervision_CVPR_2022_paper.pdf
|
cvpr-2022-1
|
['unsupervised-semantic-segmentation-with']
|
['computer-vision']
|
[ 4.60078508e-01 3.02424014e-01 -2.11424440e-01 -6.57957137e-01
-8.19537401e-01 -5.74003220e-01 4.53570753e-01 3.44910920e-02
-6.15896106e-01 2.51556635e-01 -2.26755023e-01 -4.27780151e-01
4.62215930e-01 -7.42175043e-01 -1.25099504e+00 -4.26123202e-01
3.42760146e-01 5.57258189e-01 6.64180696e-01 1.86831966e-01
3.30660604e-02 1.06435291e-01 -1.47871888e+00 5.04729986e-01
8.96054268e-01 1.21396601e+00 3.87614012e-01 7.06602514e-01
-1.92042276e-01 1.01211655e+00 -2.94713646e-01 -3.25204492e-01
3.20524305e-01 -4.51644897e-01 -1.05330586e+00 5.88895082e-01
9.42795813e-01 -3.66611898e-01 -1.67695135e-01 1.03051603e+00
1.93145573e-01 1.14118919e-01 5.09684324e-01 -1.10917759e+00
-6.94317818e-01 4.01791334e-01 -7.25678384e-01 -1.01394981e-01
-2.30003521e-01 2.64687955e-01 1.19271028e+00 -9.23744440e-01
5.31188369e-01 1.08623111e+00 5.62874496e-01 5.31077385e-01
-1.49888623e+00 -4.89835143e-01 3.69419426e-01 9.91893187e-02
-1.16345477e+00 -3.49824339e-01 6.69453621e-01 -7.15945959e-01
8.84487927e-01 -1.99922025e-02 4.98185515e-01 7.20177889e-01
-2.01811060e-01 1.21404827e+00 9.72990453e-01 -2.24645048e-01
2.24601880e-01 -1.85476705e-01 2.71797508e-01 9.49881911e-01
-1.34010706e-02 -3.08945775e-01 -3.61420065e-01 3.91002268e-01
8.10340106e-01 1.96774513e-01 -6.06226921e-03 -5.33186316e-01
-1.09598124e+00 7.87674427e-01 9.10728097e-01 6.67450577e-02
-1.33437842e-01 4.73044246e-01 4.51970190e-01 -1.87002551e-02
6.05189681e-01 1.29310980e-01 -6.63157165e-01 1.83711469e-01
-9.70506251e-01 -7.77642727e-02 4.95747477e-01 9.57587838e-01
1.10154021e+00 -7.81757385e-02 -1.97300076e-01 9.91267204e-01
1.86604366e-01 2.88855255e-01 2.24550307e-01 -1.08329475e+00
3.88463348e-01 7.50618756e-01 -8.10322687e-02 -3.89289230e-01
-2.55974203e-01 -4.21642900e-01 -7.02991903e-01 2.75876909e-01
5.25850594e-01 -1.22474797e-01 -1.64071953e+00 1.59369063e+00
2.39808843e-01 2.13957712e-01 -1.64946377e-01 9.48025763e-01
8.62347305e-01 7.28179514e-01 2.34641448e-01 2.78719306e-01
1.45964801e+00 -1.53410316e+00 -2.95826882e-01 -6.18876815e-01
6.90514624e-01 -4.01347816e-01 1.47609997e+00 3.62970978e-01
-1.02578235e+00 -7.25653231e-01 -8.76697361e-01 -5.62489390e-01
-3.60264212e-01 1.74005345e-01 4.81077850e-01 1.97028071e-01
-1.14884055e+00 3.64400983e-01 -1.16537762e+00 -3.52253467e-01
9.58078504e-01 2.99398005e-01 -2.03172192e-01 -1.15895972e-01
-5.96793890e-01 3.65392774e-01 5.07976770e-01 -4.35120724e-02
-1.09343576e+00 -6.72537208e-01 -9.97972608e-01 4.86592054e-02
7.31821299e-01 -6.96537435e-01 1.50303698e+00 -1.47074521e+00
-1.33776760e+00 1.19483900e+00 -3.40551078e-01 -6.31480455e-01
5.40225863e-01 -3.29125434e-01 3.02144110e-01 3.41685504e-01
4.28565472e-01 1.33242631e+00 7.46906281e-01 -1.38448143e+00
-7.12979555e-01 -3.33780855e-01 -4.63385172e-02 1.05941214e-01
-6.42176792e-02 -7.33115524e-02 -8.83848786e-01 -5.74496806e-01
4.64259610e-02 -8.92543077e-01 -2.43125916e-01 3.17556918e-01
-4.96388435e-01 -5.14642835e-01 1.01747334e+00 -7.14872599e-01
6.44055724e-01 -2.12030911e+00 1.13489307e-01 -2.55727530e-01
3.04605067e-01 3.10445279e-01 -1.66984662e-01 1.30380206e-02
9.83100981e-02 7.70817176e-02 -7.23945141e-01 -7.33455122e-01
2.08951961e-02 1.66212514e-01 -2.84226805e-01 3.12186331e-01
3.13572884e-01 1.35739398e+00 -7.46745229e-01 -4.34428722e-01
4.12737548e-01 2.45355889e-01 -7.32769310e-01 2.07358405e-01
-7.31831074e-01 4.67998266e-01 -3.21397871e-01 7.11672485e-01
4.26747918e-01 -6.35972917e-01 -1.39422402e-01 -2.83119887e-01
-5.25019057e-02 1.63798004e-01 -4.72483277e-01 1.90294147e+00
-3.51783454e-01 8.16708565e-01 1.90388292e-01 -1.38716030e+00
6.64578557e-01 -1.28088146e-01 3.47144574e-01 -7.67211676e-01
1.85181379e-01 1.64528742e-01 -1.74200088e-01 -1.94816753e-01
2.89005458e-01 1.18266076e-01 -2.01482087e-01 2.65822828e-01
4.01916236e-01 -1.89177513e-01 1.17036209e-01 1.87972859e-01
8.73792291e-01 2.29481176e-01 5.86528666e-02 -1.31259531e-01
1.60601139e-01 1.59982458e-01 5.95286071e-01 7.26692200e-01
-1.69499412e-01 8.99515927e-01 5.00452578e-01 -2.68906295e-01
-1.09962547e+00 -1.18074763e+00 -5.91345429e-02 1.47421658e+00
3.24905515e-01 -2.01776192e-01 -9.31600630e-01 -9.08894062e-01
8.06806460e-02 5.56827605e-01 -6.83184862e-01 1.18027858e-01
-5.13159096e-01 -5.14008820e-01 4.43148285e-01 7.72057950e-01
9.06703293e-01 -1.25871432e+00 -4.76867318e-01 1.36340097e-01
-1.12752117e-01 -1.30863273e+00 -6.63363457e-01 4.92907614e-01
-6.84148312e-01 -9.38888669e-01 -6.93191707e-01 -1.13605690e+00
8.41749847e-01 2.66148567e-01 1.10965598e+00 7.32053397e-03
-4.46637124e-01 2.20102102e-01 -2.85137266e-01 -3.11127037e-01
-2.95730419e-02 1.76940098e-01 -6.98805749e-01 8.84294361e-02
2.15470821e-01 -3.34565401e-01 -7.97083378e-01 4.44516480e-01
-8.65951896e-01 6.27627611e-01 4.94539171e-01 8.81661832e-01
9.93583322e-01 -3.14268351e-01 4.33887780e-01 -9.41490352e-01
-4.77951467e-02 -2.38855168e-01 -8.00583482e-01 1.16459407e-01
-2.52871066e-01 -1.54125154e-01 7.17412472e-01 -1.01879522e-01
-7.70662785e-01 3.73406023e-01 -1.58979103e-01 -6.58286452e-01
-4.27361339e-01 9.64490622e-02 -3.13064039e-01 2.32617501e-02
4.21516597e-01 3.85726690e-01 -2.62370855e-01 -5.12948334e-01
6.49531484e-01 7.24466324e-01 7.79364049e-01 -5.19812226e-01
4.83113766e-01 5.98089635e-01 -4.75195915e-01 -6.94038033e-01
-1.33712065e+00 -6.94209814e-01 -7.67505348e-01 -2.32240604e-03
1.45864701e+00 -1.10891223e+00 -5.87176621e-01 8.31016243e-01
-9.84226942e-01 -1.22779870e+00 -3.47398043e-01 -7.25312307e-02
-7.93465316e-01 2.09555060e-01 -7.46643662e-01 -2.98105747e-01
-3.87772143e-01 -1.13027823e+00 1.44523799e+00 1.85854435e-01
1.75782382e-01 -7.67822683e-01 -5.03264308e-01 8.54767203e-01
3.33533660e-02 9.83872190e-02 5.69660723e-01 -5.07062912e-01
-8.81139517e-01 1.61287442e-01 -7.27787077e-01 6.64346695e-01
3.58108394e-02 -2.42805198e-01 -1.10495377e+00 -1.76680610e-01
-2.63580561e-01 -6.28596842e-01 1.38908792e+00 4.29680198e-01
1.71002245e+00 -2.41286606e-01 -2.88463145e-01 1.00079417e+00
1.38249862e+00 2.19616249e-01 4.35416877e-01 2.09449694e-01
1.27739370e+00 5.95812082e-01 5.38110912e-01 3.92835028e-02
6.44400179e-01 5.91039658e-01 4.67091829e-01 -5.04098833e-01
-3.79503846e-01 -4.11745161e-01 7.75568783e-02 4.57253456e-01
4.50616986e-01 -3.08711231e-01 -1.08655262e+00 7.26039827e-01
-2.11135244e+00 -5.77529430e-01 1.53021459e-02 1.91995013e+00
9.66217041e-01 2.66107708e-01 1.29561946e-01 -2.95400679e-01
7.91907132e-01 2.33474597e-01 -1.01874745e+00 -2.45125607e-01
4.43109758e-02 4.81976978e-02 8.18737984e-01 4.91584718e-01
-1.49850965e+00 1.61337435e+00 5.01893473e+00 9.48946238e-01
-1.21375787e+00 2.77538896e-01 1.11208332e+00 1.03738174e-01
7.39589287e-03 4.26365383e-04 -7.54763365e-01 4.58309293e-01
5.06961346e-01 1.80884570e-01 4.08814609e-01 7.71680832e-01
1.28370523e-01 -1.83939978e-01 -1.07647288e+00 8.18594754e-01
2.60222051e-02 -1.24835896e+00 5.80810802e-03 -2.00336915e-03
1.03083420e+00 5.55496991e-01 -5.43681681e-02 2.41480529e-01
6.32613897e-01 -1.08377898e+00 8.70386243e-01 9.73470360e-02
7.88646579e-01 -3.71281803e-01 4.30866510e-01 3.69760871e-01
-1.25521851e+00 7.50027150e-02 -2.99431533e-01 2.29564756e-02
1.68690100e-01 5.08608043e-01 -6.54118776e-01 2.24958584e-01
7.36072779e-01 1.10895324e+00 -5.63663542e-01 9.81280625e-01
-4.83242095e-01 9.48958397e-01 -3.73505771e-01 2.92276919e-01
5.38239300e-01 -1.87229887e-01 6.52630627e-02 1.13852406e+00
-5.11602908e-02 -3.95964086e-02 5.66213548e-01 1.08174896e+00
-4.72132683e-01 -1.39962435e-01 -2.32671589e-01 4.52272706e-02
2.94956774e-01 1.16313648e+00 -1.04206920e+00 -7.12514162e-01
-4.43920285e-01 1.41434979e+00 4.91536260e-01 5.06609261e-01
-8.20262372e-01 -4.32566971e-01 4.94053125e-01 6.55343905e-02
8.19960058e-01 -1.95178449e-01 -6.47816002e-01 -1.11333334e+00
6.99575804e-03 -5.10310531e-01 3.02104771e-01 -8.98446083e-01
-1.25416136e+00 3.71280223e-01 -2.51097471e-01 -9.11280274e-01
7.21650124e-02 -6.32706940e-01 -7.29983211e-01 6.48845136e-01
-1.51842690e+00 -1.44430923e+00 -4.09497887e-01 6.30231917e-01
9.08423781e-01 2.55090743e-01 3.69506687e-01 1.26230836e-01
-6.45084381e-01 5.55410147e-01 -3.14005725e-02 5.36064804e-01
6.53694212e-01 -1.54387677e+00 7.60255396e-01 9.16320026e-01
1.61032885e-01 1.69677570e-01 2.73849249e-01 -4.59031075e-01
-9.20622289e-01 -1.76420200e+00 5.69909036e-01 -2.66100854e-01
6.72823310e-01 -8.67507160e-01 -9.72667038e-01 8.48215580e-01
2.99317688e-01 4.38114673e-01 3.83363128e-01 -1.81909446e-02
-5.33888519e-01 -1.31785944e-01 -9.57691133e-01 5.99753439e-01
1.19624484e+00 -6.32504284e-01 -5.57177186e-01 4.22414809e-01
1.08905983e+00 -3.39620978e-01 -6.63847387e-01 3.95302325e-01
1.48459509e-01 -7.50267148e-01 8.89294386e-01 -2.64274478e-01
4.56073344e-01 -5.75737894e-01 -2.90997643e-02 -1.00868797e+00
-7.34422654e-02 -3.09234381e-01 2.71626770e-01 1.09366477e+00
4.46684450e-01 -6.81211531e-01 9.81891453e-01 3.65915626e-01
-5.27629733e-01 -7.26182044e-01 -7.57487237e-01 -6.39268577e-01
2.01951325e-01 -5.05619109e-01 7.21493214e-02 8.18833053e-01
-4.47478175e-01 4.95830715e-01 -1.17143907e-01 9.07413736e-02
7.60685682e-01 3.12467694e-01 7.81777084e-01 -9.60582316e-01
-3.73792231e-01 -5.92173457e-01 -3.00572574e-01 -1.46191204e+00
2.63696700e-01 -1.03689325e+00 4.36890036e-01 -1.87224543e+00
2.62431234e-01 -3.66501182e-01 -3.17791075e-01 1.05378044e+00
-2.42284209e-01 7.19097555e-01 3.38043928e-01 2.05846593e-01
-1.14178848e+00 6.61174595e-01 1.26397634e+00 -3.63765150e-01
-1.13730557e-01 -1.69993386e-01 -6.90066993e-01 8.08353305e-01
9.45394814e-01 -3.53992701e-01 -3.95614624e-01 -6.24074280e-01
-2.10713252e-01 -1.66729465e-01 6.86735511e-01 -9.07843888e-01
2.12756827e-01 2.28735041e-02 2.75570840e-01 -6.70740128e-01
2.07394898e-01 -5.71965516e-01 -4.07604843e-01 2.78249532e-01
-4.03265595e-01 -5.46703279e-01 3.20086509e-01 4.33796406e-01
-3.76864433e-01 3.16552795e-03 9.12867248e-01 -3.49548310e-02
-9.95705903e-01 4.28155988e-01 -2.59654045e-01 3.27422857e-01
1.07871985e+00 -2.68871605e-01 -4.74272102e-01 -2.61160266e-02
-8.17040741e-01 6.26168549e-01 6.85175955e-01 4.21335876e-01
3.80169600e-01 -8.93014789e-01 -4.87832308e-01 1.50390416e-01
2.22056970e-01 7.41354406e-01 2.55686104e-01 6.87439859e-01
-5.54799855e-01 3.44958246e-01 4.07922901e-02 -1.03108346e+00
-1.15713525e+00 3.43046188e-01 5.43857157e-01 -1.56897865e-02
-7.72025406e-01 1.09250915e+00 1.02367759e+00 -5.86877644e-01
2.98213840e-01 -5.61418533e-01 1.71091631e-01 -9.63184983e-02
2.73818284e-01 -1.54966325e-01 -6.78303391e-02 -4.66382742e-01
-2.23891273e-01 7.31572270e-01 -1.86058670e-01 3.54892835e-02
1.10804868e+00 -6.93715736e-02 -1.36401311e-01 4.51359481e-01
1.37396526e+00 -4.84776735e-01 -1.93172395e+00 -3.69887382e-01
4.55415063e-02 -2.07519755e-01 1.18116878e-01 -9.83681619e-01
-1.33137143e+00 1.18919230e+00 4.14490670e-01 -7.82222226e-02
1.20482731e+00 4.23908412e-01 8.95011783e-01 3.69731218e-01
1.95282757e-01 -9.26553667e-01 3.00811738e-01 5.92139423e-01
6.00560904e-01 -1.57616138e+00 -4.81139123e-01 -5.32777131e-01
-6.92690015e-01 6.79743230e-01 7.87233055e-01 -3.64820570e-01
4.66969401e-01 2.44091034e-01 2.39766482e-02 -5.49557060e-02
-7.56767511e-01 -6.18831217e-01 3.93227875e-01 4.07127142e-01
2.94838220e-01 1.92906797e-01 3.23481560e-02 5.54462552e-01
8.62321034e-02 6.46959171e-02 2.54420668e-01 7.27120697e-01
-7.25763023e-01 -8.25715959e-01 -1.23893082e-01 5.85047960e-01
-4.51267004e-01 -2.94124663e-01 -4.21410501e-01 5.63455284e-01
2.85398602e-01 8.14396322e-01 4.53462124e-01 -1.74879670e-01
2.08160415e-01 -1.01196229e-01 2.73637384e-01 -8.77002060e-01
-4.53599781e-01 2.52570599e-01 -8.72785822e-02 -6.14055872e-01
-2.73538679e-01 -6.58702791e-01 -1.58129656e+00 1.05059184e-01
8.16175342e-03 -1.11980096e-01 5.79878390e-01 8.81277859e-01
3.21044356e-01 7.41337419e-01 3.66910696e-01 -9.40195382e-01
6.19482435e-02 -8.47718894e-01 -2.63962328e-01 3.41714829e-01
3.82392317e-01 -2.59372026e-01 -1.38190404e-01 4.52942908e-01]
|
[9.694416999816895, 0.7219445109367371]
|
fd41703c-d339-40b9-b717-3a9cfe20eb7a
|
joint-convolutional-neural-pyramid-for-depth
|
1801.00968
| null |
http://arxiv.org/abs/1801.00968v1
|
http://arxiv.org/pdf/1801.00968v1.pdf
|
Joint convolutional neural pyramid for depth map super-resolution
|
High-resolution depth map can be inferred from a low-resolution one with the
guidance of an additional high-resolution texture map of the same scene.
Recently, deep neural networks with large receptive fields are shown to benefit
applications such as image completion. Our insight is that super resolution is
similar to image completion, where only parts of the depth values are precisely
known. In this paper, we present a joint convolutional neural pyramid model
with large receptive fields for joint depth map super-resolution. Our model
consists of three sub-networks, two convolutional neural pyramids concatenated
by a normal convolutional neural network. The convolutional neural pyramids
extract information from large receptive fields of the depth map and guidance
map, while the convolutional neural network effectively transfers useful
structures of the guidance image to the depth image. Experimental results show
that our model outperforms existing state-of-the-art algorithms not only on
data pairs of RGB/depth images, but also on other data pairs like
color/saliency and color-scribbles/colorized images.
|
['Yan Zheng', 'Xianyi Zhu', 'Xiang Cao', 'Renzhi Yang', 'Yi Xiao']
|
2018-01-03
| null | null | null | null |
['depth-map-super-resolution']
|
['computer-vision']
|
[ 6.13427520e-01 3.49907637e-01 1.28637061e-01 -3.91574144e-01
-7.69419014e-01 -1.76496476e-01 2.71521896e-01 -3.73168945e-01
-9.32148620e-02 7.54310191e-01 4.23585951e-01 4.04976010e-01
1.09068938e-01 -1.23667800e+00 -1.10822356e+00 -7.12074518e-01
1.58939332e-01 1.48406014e-01 8.14536929e-01 -5.31708479e-01
4.94904608e-01 5.11776388e-01 -1.97586703e+00 9.76753294e-01
7.74331450e-01 1.12566161e+00 6.34927869e-01 5.86728454e-01
-2.00717762e-01 1.02174330e+00 -9.68534872e-02 1.05445005e-01
4.64452416e-01 6.37190640e-02 -8.29251409e-01 8.01534206e-02
9.26209986e-01 -1.02354527e+00 -5.53849638e-01 1.36909115e+00
1.62954047e-01 -3.90131871e-04 1.26744211e-01 -5.55049002e-01
-1.13884664e+00 3.51793915e-01 -9.09838259e-01 1.08477123e-01
4.96873498e-01 -2.49786407e-01 6.39490783e-01 -1.08691096e+00
7.69865155e-01 1.75191915e+00 3.55072826e-01 5.49351275e-01
-1.19988847e+00 -5.26848376e-01 2.32695118e-01 -3.89893452e-04
-1.21033180e+00 -1.27140790e-01 7.64745951e-01 -1.42520145e-01
8.35702360e-01 -1.45663694e-01 4.77263451e-01 7.13962615e-01
8.00772235e-02 5.91049254e-01 1.44888949e+00 -3.34726244e-01
2.19703004e-01 -6.78527206e-02 -1.44904613e-01 8.05074692e-01
2.38059223e-01 3.76853615e-01 -8.84323955e-01 2.38333777e-01
1.89514041e+00 4.07677263e-01 -3.61760288e-01 -2.44007573e-01
-1.17386246e+00 4.24440861e-01 9.70876992e-01 1.97920874e-01
-5.46330154e-01 2.59175479e-01 -3.42679232e-01 -2.91807912e-02
6.85259044e-01 1.43701151e-01 -3.72724086e-01 3.39221835e-01
-6.52911246e-01 1.04590669e-01 2.33374208e-01 9.92874742e-01
1.56581318e+00 2.99408380e-02 -4.00746539e-02 7.33609140e-01
-3.13463844e-02 3.13154668e-01 2.06756920e-01 -1.46970093e+00
4.29520011e-01 6.74023211e-01 4.48484063e-01 -9.41163540e-01
-2.36681089e-01 1.90840915e-01 -1.22937119e+00 6.82429850e-01
2.74634987e-01 1.21871226e-01 -1.16226125e+00 1.44919229e+00
3.05264294e-01 8.81616324e-02 1.23897672e-01 1.11750650e+00
9.78358150e-01 7.67817974e-01 -4.25087631e-01 2.78962702e-01
1.36055493e+00 -1.04630303e+00 -5.84349871e-01 -4.80693609e-01
-3.31203252e-01 -5.35234034e-01 9.90437150e-01 3.98956805e-01
-1.34872484e+00 -1.03432846e+00 -1.09311497e+00 -8.47341120e-01
-5.14856339e-01 -5.14841042e-02 8.60485196e-01 8.42860993e-03
-1.63170207e+00 7.79851615e-01 -6.48754716e-01 -1.50556073e-01
5.84700346e-01 2.89711893e-01 -5.72292030e-01 -5.20291150e-01
-1.05636251e+00 5.03241837e-01 5.74425161e-01 1.78646997e-01
-1.09257460e+00 -7.87731290e-01 -9.79991913e-01 -2.78192777e-02
-9.49143916e-02 -5.90906858e-01 9.42359447e-01 -9.44651961e-01
-1.42193282e+00 8.80226672e-01 -2.92417347e-01 -9.88490805e-02
8.44744220e-02 -4.06582206e-01 -4.89413626e-02 4.97888118e-01
3.08407485e-01 1.18983471e+00 1.09219313e+00 -1.55093706e+00
-1.23648953e+00 -6.03460371e-01 4.76802111e-01 3.18307161e-01
-2.58528031e-02 -3.22001874e-01 -6.03755951e-01 -2.68894076e-01
5.92977762e-01 -3.14078331e-01 -4.93666112e-01 1.57327786e-01
-2.68636495e-01 2.15402782e-01 5.90290546e-01 -7.14593887e-01
6.38260126e-01 -2.00004053e+00 2.47792289e-01 -1.72083154e-01
4.33193743e-01 -3.37012500e-01 -1.13093063e-01 -9.67527255e-02
-1.55141860e-01 -8.43812302e-02 -9.82850716e-02 -1.65528998e-01
-4.91520286e-01 2.87395269e-01 -5.43966472e-01 2.50831008e-01
3.56618106e-01 8.69597852e-01 -9.76279974e-01 -2.73807883e-01
4.05746341e-01 8.54783237e-01 -6.04411900e-01 3.70245218e-01
-1.68431222e-01 6.33029938e-01 -4.99027371e-01 8.17552209e-01
1.21493721e+00 -3.61693382e-01 -2.05850586e-01 -5.07526398e-01
-3.90159875e-01 1.45729169e-01 -9.93119419e-01 2.21247578e+00
-2.42958426e-01 5.48627257e-01 1.95087045e-01 -4.55525607e-01
1.08607984e+00 -1.73908845e-01 3.01807940e-01 -1.15319085e+00
-2.12156162e-01 -5.21019707e-03 -5.91074884e-01 -1.47093102e-01
9.51544464e-01 5.22281490e-02 8.88401866e-02 2.77113497e-01
-1.91910192e-02 -3.99256498e-01 -8.35203286e-03 1.67051703e-01
5.99234700e-01 4.59534734e-01 -1.44663692e-01 -1.29205197e-01
3.91499221e-01 -1.55937165e-01 5.07443547e-01 6.81248367e-01
1.95826948e-01 1.05300653e+00 3.46507519e-01 -8.22032213e-01
-1.44206619e+00 -1.23229969e+00 -1.96291953e-01 1.08553433e+00
6.53572738e-01 -3.18518393e-02 -8.56611252e-01 -3.63124572e-02
-1.74341857e-01 -5.10006538e-03 -1.17687440e+00 2.05627456e-01
-5.69966793e-01 -5.78089058e-01 1.23794183e-01 7.69926548e-01
1.11633372e+00 -1.10913801e+00 -6.38532698e-01 -2.05509458e-02
-3.41306210e-01 -1.44266653e+00 -6.76957667e-02 2.01937154e-01
-1.15508306e+00 -9.82208848e-01 -8.83132994e-01 -9.44074750e-01
9.84460771e-01 6.63633347e-01 1.29472423e+00 -1.91514477e-01
-3.51862788e-01 1.51407138e-01 -2.56906271e-01 -4.94540520e-02
5.29207364e-02 -1.89722925e-01 -2.37637177e-01 1.47352561e-01
2.38779753e-01 -8.52574587e-01 -9.38279092e-01 2.68813074e-01
-1.13549101e+00 7.20087409e-01 8.72497857e-01 6.23348117e-01
1.19920599e+00 -1.75282229e-02 9.72236991e-02 -8.46247554e-01
7.70553425e-02 -1.32112846e-01 -8.57495487e-01 -2.39316132e-02
-1.85926855e-01 3.84312332e-01 2.97507435e-01 -7.91551173e-02
-1.57313144e+00 2.35333741e-01 -1.19076163e-01 -4.06295180e-01
-2.94853330e-01 -1.69220492e-01 -7.34899491e-02 -2.62353390e-01
7.04026818e-01 3.02855074e-01 -2.62990475e-01 -7.49650061e-01
5.67408621e-01 3.23185533e-01 9.75808501e-01 -4.94253665e-01
7.22607732e-01 1.16514301e+00 1.02899209e-01 -5.76781213e-01
-1.09723449e+00 -3.02744269e-01 -1.11553228e+00 3.27558741e-02
1.03810608e+00 -1.31789994e+00 -5.00889897e-01 5.24661362e-01
-1.18475175e+00 -4.65841264e-01 -3.30985904e-01 1.43661723e-01
-6.81732774e-01 1.75316975e-01 -9.70470309e-01 -3.40082705e-01
-2.21631899e-01 -9.51371551e-01 1.61132157e+00 6.49700165e-01
6.17361426e-01 -7.00126767e-01 -7.27976561e-02 1.50955230e-01
3.95895600e-01 4.26165968e-01 6.81988060e-01 4.62317050e-01
-1.41324317e+00 1.94804102e-01 -9.85931516e-01 4.11069632e-01
2.66098589e-01 -3.54153454e-01 -1.47528923e+00 -3.94035168e-02
-1.53261542e-01 -4.45512623e-01 1.25165570e+00 6.35943770e-01
1.40839362e+00 -1.41453654e-01 -1.64201185e-01 1.14042222e+00
1.85513079e+00 -2.19280109e-01 1.15250695e+00 2.83875912e-01
1.08043730e+00 6.18076444e-01 6.00917339e-01 2.77362943e-01
4.55633879e-01 2.81567395e-01 9.08188522e-01 -5.51590204e-01
-4.27692741e-01 -4.09914404e-01 1.41899258e-01 2.98290282e-01
-5.36790490e-01 5.41330159e-01 -4.91405070e-01 5.54820538e-01
-1.69504154e+00 -7.64570117e-01 -3.74772027e-03 2.05485511e+00
9.29244339e-01 -4.43441756e-02 -3.66017550e-01 -2.06490010e-01
5.94916880e-01 3.27886939e-01 -6.83990180e-01 -2.62323409e-01
-4.77637619e-01 3.81275743e-01 6.23671472e-01 8.09421301e-01
-1.04049695e+00 1.22546947e+00 6.53178596e+00 7.14058280e-01
-8.46060693e-01 -4.83223423e-02 8.82144392e-01 6.08786754e-02
-4.86847043e-01 -1.51378021e-01 -9.94403720e-01 -1.10501155e-01
4.29188937e-01 2.68379062e-01 6.67278290e-01 8.40856433e-01
-1.18286029e-01 -5.69526613e-01 -1.15070879e+00 1.15973938e+00
3.97763327e-02 -1.64324749e+00 3.81514400e-01 5.54699935e-02
1.30441213e+00 2.40281031e-01 3.04793656e-01 9.85859428e-03
5.20384133e-01 -1.18334317e+00 5.15831530e-01 7.10684061e-01
1.31043804e+00 -6.74351990e-01 4.41833228e-01 4.11118343e-02
-1.39442527e+00 -1.16725199e-01 -1.02201784e+00 -9.78967398e-02
-2.13425070e-01 5.67557156e-01 -3.51842850e-01 5.13338745e-01
1.27770734e+00 1.14992988e+00 -7.10777521e-01 5.97553730e-01
-3.13645005e-01 -2.90286303e-01 -9.45485905e-02 6.06215596e-01
1.82647988e-01 -1.78401366e-01 -8.48784670e-02 9.26319122e-01
1.04852200e-01 3.08377624e-01 -1.40745848e-01 1.36612248e+00
2.12447904e-02 -4.31806237e-01 -6.94551945e-01 3.77273977e-01
2.18396857e-01 1.50725007e+00 -5.21023929e-01 -4.13983673e-01
-4.29649234e-01 1.26439238e+00 6.39907122e-01 7.75790691e-01
-4.10669714e-01 -3.52416933e-01 7.79657006e-01 8.29715654e-02
4.32747364e-01 -7.80212805e-02 -5.92410743e-01 -1.16887844e+00
-1.88014463e-01 -2.83324331e-01 4.40874174e-02 -1.54854095e+00
-9.98897672e-01 9.17053461e-01 -7.68233687e-02 -1.21627116e+00
-3.36070150e-01 -8.06665957e-01 -1.47787809e-01 1.34271860e+00
-2.17015481e+00 -1.20522690e+00 -9.85658824e-01 1.02684784e+00
5.57529330e-01 3.34959418e-01 8.24704945e-01 -1.99225191e-02
-4.23962623e-02 -3.78623642e-02 1.53455377e-01 2.68757790e-01
4.68875200e-01 -1.30738497e+00 4.72366065e-01 8.34114790e-01
-3.53709966e-01 4.32872683e-01 3.03786755e-01 -5.23883104e-01
-1.25751579e+00 -1.17685437e+00 1.87163398e-01 -5.42145729e-01
1.33840799e-01 -4.77401108e-01 -1.05598319e+00 6.88135982e-01
7.31695816e-02 2.85255462e-01 -2.62650512e-02 -1.76507041e-01
-5.35480917e-01 -3.55434924e-01 -1.13242102e+00 5.06430447e-01
1.02291107e+00 -7.56944597e-01 -5.15757442e-01 6.38885498e-02
1.11265469e+00 -7.40854383e-01 -9.72130060e-01 3.46646637e-01
6.74306214e-01 -1.52474666e+00 1.50565529e+00 -1.02150932e-01
9.51238871e-01 -4.33973819e-01 -4.91090089e-01 -9.89841819e-01
-6.46960974e-01 -1.89398155e-01 -4.51437617e-03 7.75003254e-01
7.03971609e-02 -2.59664327e-01 9.13244188e-01 5.66987097e-01
-5.97919635e-02 -3.90694082e-01 -7.10957050e-01 -7.48982206e-02
-4.17789705e-02 -1.06563166e-01 5.85642457e-01 5.47688186e-01
-3.21578085e-01 1.29674256e-01 -2.60004163e-01 4.69509006e-01
1.07040513e+00 3.32837135e-01 5.88324964e-01 -1.23040760e+00
3.84813137e-02 -2.43418813e-01 -1.86165407e-01 -1.28817225e+00
-3.13856333e-01 -3.00927460e-01 2.38613024e-01 -1.84721982e+00
3.23296905e-01 -1.99099019e-01 -3.17798287e-01 5.80417216e-01
-6.91368952e-02 6.44341588e-01 -2.25406855e-01 1.20108947e-01
-5.56280375e-01 4.12213236e-01 1.76276231e+00 -5.95092736e-02
-3.07265311e-01 -5.10058939e-01 -8.81871581e-01 9.11190271e-01
6.47726953e-01 4.20563994e-03 -3.84539187e-01 -6.71923518e-01
3.43446165e-01 1.32231861e-01 4.39642698e-01 -9.80685949e-01
2.01838091e-01 -1.58638254e-01 1.09914088e+00 -8.10281754e-01
6.02586627e-01 -6.39162242e-01 -2.67361104e-01 -6.04610033e-02
-2.87398785e-01 -2.99715072e-01 3.69335353e-01 6.17808402e-01
-3.30323637e-01 2.76364386e-01 8.07706654e-01 -6.44139826e-01
-1.29341388e+00 6.60048544e-01 -3.78752835e-02 -2.07316548e-01
5.71897507e-01 -4.75498408e-01 -5.83905935e-01 -2.86608785e-01
-5.49405217e-01 -4.82853577e-02 6.34067118e-01 4.94685262e-01
1.20649350e+00 -1.31216979e+00 -6.87918603e-01 6.62393689e-01
1.48839608e-01 9.23601091e-01 5.50763965e-01 3.83801311e-01
-6.78356767e-01 4.13195163e-01 -9.44352627e-01 -7.69174516e-01
-7.06395566e-01 7.03498065e-01 3.92636359e-01 8.48578475e-03
-8.89589489e-01 9.66639519e-01 1.11494482e+00 -1.94360912e-01
2.45186582e-01 -8.66511524e-01 -2.33809158e-01 -4.04274464e-01
1.23330033e+00 1.01874188e-01 -3.13852817e-01 -5.10935962e-01
-1.15831286e-01 1.09358895e+00 -3.30134541e-01 -1.71890929e-01
1.54149973e+00 -6.10894263e-01 -5.93053162e-01 2.48303965e-01
9.54159260e-01 -2.91235059e-01 -2.00104380e+00 -5.75705826e-01
-5.24556696e-01 -7.99741268e-01 3.26035291e-01 -6.05409741e-01
-1.17939019e+00 1.13130617e+00 5.98173022e-01 -8.50274563e-02
1.65429688e+00 9.03872028e-02 4.94654775e-01 2.74631917e-01
6.74566150e-01 -1.06900430e+00 3.65936875e-01 5.94956875e-01
1.04477596e+00 -1.42447436e+00 -4.86141443e-02 -4.74762678e-01
-4.55970258e-01 1.27622247e+00 1.01377273e+00 -4.99585718e-01
5.52948594e-01 3.14681262e-01 -1.73792318e-01 -1.63223788e-01
-6.75475478e-01 -5.18277645e-01 3.00883353e-01 1.05195093e+00
1.50644511e-01 -1.69714153e-01 5.87228596e-01 4.51608688e-01
-6.83668302e-03 1.91915452e-01 6.75848484e-01 6.26201212e-01
-8.82157922e-01 -6.44822598e-01 -5.84545970e-01 3.47908065e-02
-1.94665223e-01 -5.24937570e-01 -1.68252110e-01 5.19710362e-01
2.38739908e-01 5.93579948e-01 5.23511052e-01 -4.45730925e-01
1.15038395e-01 -4.66018826e-01 6.76034331e-01 -6.50598288e-01
-2.26868279e-02 2.21837267e-01 -3.00807655e-01 -1.20234489e+00
-6.07715905e-01 -8.96047056e-03 -1.33216906e+00 -3.03755790e-01
3.19142699e-01 -2.59672284e-01 6.40232146e-01 5.35641134e-01
2.62492865e-01 7.30753005e-01 5.37940443e-01 -1.65082586e+00
3.63997221e-01 -9.68535185e-01 -9.94252563e-01 2.59124160e-01
8.06868851e-01 -4.33282763e-01 -1.14574820e-01 2.58522213e-01]
|
[9.858637809753418, -2.392747640609741]
|
d626123e-eb92-4608-a143-71d9c764b3a2
|
gio-gradient-information-optimization-for
|
2306.11670
| null |
https://arxiv.org/abs/2306.11670v1
|
https://arxiv.org/pdf/2306.11670v1.pdf
|
GIO: Gradient Information Optimization for Training Dataset Selection
|
It is often advantageous to train models on a subset of the available train examples, because the examples are of variable quality or because one would like to train with fewer examples, without sacrificing performance. We present Gradient Information Optimization (GIO), a scalable, task-agnostic approach to this data selection problem that requires only a small set of (unlabeled) examples representing a target distribution. GIO begins from a natural, information-theoretic objective that is intractable in practice. Our contribution is in showing that it can be made highly scalable through a simple relaxation of the objective and a highly efficient implementation. In experiments with machine translation, spelling correction, and image recognition, we show that GIO delivers outstanding results with very small train sets. These findings are robust to different representation models and hyperparameters for GIO itself. GIO is task- and domain-agnostic and can be applied out-of-the-box to new datasets and domains.
|
['Christopher Potts', 'Dante Everaert']
|
2023-06-20
| null | null | null | null |
['machine-translation', 'spelling-correction']
|
['natural-language-processing', 'natural-language-processing']
|
[ 3.93536091e-01 -1.50218561e-01 -5.02798200e-01 -6.47840023e-01
-1.37375879e+00 -7.87088752e-01 7.52646089e-01 4.01125140e-02
-7.03554034e-01 9.07491088e-01 -1.00217871e-01 -5.32215059e-01
-2.22642794e-01 -3.28618288e-01 -7.20865309e-01 -5.26976168e-01
1.66383639e-01 1.11774135e+00 1.58323213e-01 -1.00660235e-01
3.54209304e-01 4.22369719e-01 -1.35857546e+00 2.48566225e-01
8.80677998e-01 9.40854907e-01 2.96496153e-01 7.90423512e-01
-1.94425777e-01 5.27459443e-01 -5.96080482e-01 -4.18621421e-01
3.86765003e-01 -4.96129394e-01 -8.50757718e-01 4.12756979e-01
5.99069476e-01 3.05623952e-02 -6.24886621e-03 9.23540831e-01
5.16091049e-01 7.98415691e-02 9.50197577e-01 -1.14735985e+00
-6.99055612e-01 3.30242127e-01 -3.43501270e-01 3.09040189e-01
1.03100941e-01 8.69862884e-02 1.02776182e+00 -1.09367585e+00
7.07227886e-01 9.81538534e-01 5.53326905e-01 7.20697284e-01
-1.45030904e+00 -4.33061182e-01 2.10450873e-01 5.14875688e-02
-1.19806421e+00 -5.76253474e-01 4.46737885e-01 -4.63308185e-01
8.78524244e-01 4.01067138e-01 8.83002952e-03 9.63828504e-01
-1.17596604e-01 1.07283497e+00 1.13340569e+00 -6.13760591e-01
3.43189448e-01 5.23259103e-01 1.42278045e-01 4.26073670e-01
2.62742221e-01 9.73206386e-02 -3.41714919e-01 -2.55125463e-01
4.96491671e-01 -1.13790445e-01 -1.92948937e-01 -6.14712298e-01
-1.19665170e+00 9.11968410e-01 1.95051447e-01 1.95052922e-01
-1.24674983e-01 -2.15742905e-02 3.80986333e-01 7.96411633e-01
5.68415642e-01 8.07936966e-01 -8.38801444e-01 -1.57204911e-01
-9.63887811e-01 3.37801039e-01 9.84244406e-01 1.25356448e+00
9.32511330e-01 1.19122036e-03 -1.38346866e-01 1.03911817e+00
-4.67904955e-02 6.33699656e-01 7.56897628e-01 -8.49648416e-01
6.30465686e-01 2.90549129e-01 1.88785329e-01 -3.97922784e-01
-2.39059687e-01 -4.21126992e-01 -4.38313514e-01 1.55413434e-01
6.35513365e-01 -3.29494745e-01 -1.25069451e+00 1.83658028e+00
1.20701663e-01 -4.59688991e-01 -2.77316198e-02 7.96860933e-01
4.05016959e-01 4.25248504e-01 -6.36705384e-02 -1.90426469e-01
1.01366627e+00 -9.70291913e-01 -3.39112222e-01 -7.53897607e-01
7.66235650e-01 -8.06837559e-01 1.43984747e+00 4.26110744e-01
-1.02608752e+00 -1.80213690e-01 -8.19369316e-01 -2.28851382e-02
-3.74248773e-01 1.42851755e-01 6.48372710e-01 7.55473673e-01
-9.51595366e-01 7.00884402e-01 -4.78609562e-01 -2.95483023e-01
4.92174745e-01 5.53913176e-01 -4.40334827e-01 -2.99998552e-01
-9.16230500e-01 1.14035261e+00 4.02399451e-01 -2.28374869e-01
-8.54009807e-01 -6.59222662e-01 -7.40193605e-01 -2.09815912e-02
4.90507901e-01 -5.66815078e-01 1.53622746e+00 -1.47122383e+00
-1.40088642e+00 8.10223460e-01 -2.55056471e-01 -3.04804653e-01
6.68334723e-01 -2.28130564e-01 -1.63622081e-01 -5.90929426e-02
1.78321600e-01 4.55029905e-01 9.76322591e-01 -1.03527725e+00
-6.49176538e-01 -2.90239841e-01 -3.99007350e-01 3.25182706e-01
-4.86110061e-01 1.45308971e-01 -4.53058064e-01 -6.03318036e-01
-3.18262689e-02 -1.09125042e+00 -4.28615212e-01 -1.13584861e-01
-2.90211916e-01 -1.53885573e-01 5.39034128e-01 -3.31899971e-01
1.10110688e+00 -2.10104585e+00 1.13208190e-01 2.11508244e-01
-6.81363046e-02 2.42161825e-01 -3.63726258e-01 4.11579788e-01
-7.30998442e-03 2.98370123e-01 -5.03832579e-01 -2.29752839e-01
3.95311639e-02 2.19631016e-01 -2.63856739e-01 3.42103779e-01
4.05647993e-01 6.33520484e-01 -1.04273856e+00 -4.53909963e-01
-1.25749260e-01 -8.34852364e-03 -5.04195452e-01 1.56807363e-01
-4.26598251e-01 2.87023485e-01 -5.30041099e-01 4.84528482e-01
4.52553272e-01 -5.04348814e-01 1.84660763e-01 3.03273320e-01
1.63202450e-01 4.42459017e-01 -1.23895228e+00 1.51441061e+00
-5.43322742e-01 8.78809988e-01 -1.80497468e-01 -1.08914554e+00
7.03104377e-01 1.66455999e-01 3.19380581e-01 -4.80583698e-01
4.44406681e-02 5.74760854e-01 1.78318433e-02 -4.06212389e-01
2.68922031e-01 -1.57802820e-01 -1.07200287e-01 6.57612085e-01
3.09247136e-01 -3.23872745e-01 4.70255315e-01 2.13318001e-02
1.17656207e+00 -3.32743004e-02 3.86168808e-01 -3.57929707e-01
2.51822501e-01 2.40663305e-01 5.31098425e-01 1.02762103e+00
-1.71837900e-02 7.98326015e-01 2.65354812e-01 -3.51180315e-01
-1.33585346e+00 -9.62596595e-01 -3.42315823e-01 1.37243021e+00
-2.40065783e-01 -1.56949729e-01 -5.74883997e-01 -9.82546151e-01
2.04598214e-02 7.03682363e-01 -4.68810201e-01 4.09747884e-02
-5.46850085e-01 -9.21307802e-01 2.56846309e-01 4.87873942e-01
3.67016047e-01 -8.66787493e-01 -1.02844737e-01 2.48073563e-01
7.33359829e-02 -8.70700479e-01 -6.06284857e-01 4.71826583e-01
-1.00887311e+00 -8.83876622e-01 -8.97023380e-01 -7.89222777e-01
7.79362500e-01 2.64719486e-01 1.47405791e+00 -1.22417815e-01
-7.27593601e-02 2.78553218e-01 -2.56984115e-01 -6.24401927e-01
-4.99518365e-01 4.10244733e-01 -3.42730545e-02 -1.09194607e-01
5.96078038e-01 -4.04767692e-01 -2.67336041e-01 4.81086105e-01
-8.53862405e-01 -3.48337710e-01 8.39673221e-01 1.12790525e+00
5.92663586e-01 -2.12357730e-01 5.85049570e-01 -1.49764824e+00
7.18185902e-01 -5.02924919e-01 -5.81992090e-01 3.78600419e-01
-9.08998728e-01 4.20714945e-01 7.83031821e-01 -6.09730959e-01
-8.92721415e-01 2.08084524e-01 5.46204485e-02 -4.46269512e-01
-9.27852243e-02 3.91368270e-01 3.40426490e-02 -2.19146892e-01
1.18486917e+00 2.85497189e-01 2.68389340e-02 -5.87985098e-01
4.37194616e-01 8.03243756e-01 1.40420988e-01 -6.19961441e-01
7.89088130e-01 1.27771422e-01 -2.24057853e-01 -8.07582796e-01
-9.96350527e-01 -5.18646657e-01 -5.68024874e-01 2.54201353e-01
2.43335187e-01 -8.49979103e-01 -1.83706820e-01 9.98337343e-02
-7.97144532e-01 -6.40968621e-01 -4.47149605e-01 5.54424942e-01
-6.29752338e-01 1.77568048e-01 -3.16885591e-01 -6.55325651e-01
-2.12799937e-01 -1.12478852e+00 9.05254185e-01 2.59743128e-02
-3.14276576e-01 -1.00789022e+00 4.02908474e-02 1.66457385e-01
3.97412419e-01 -1.56259611e-01 8.77199650e-01 -1.26926899e+00
-5.77089787e-01 -4.61577356e-01 -1.56283766e-01 5.61318099e-01
1.61058053e-01 -2.71674376e-02 -9.63674068e-01 -4.43289578e-01
3.64965759e-03 -6.64672911e-01 8.87674928e-01 3.00378114e-01
1.18650818e+00 -5.30587494e-01 -1.87473744e-01 5.32178879e-01
1.39436579e+00 -1.60846510e-04 2.80507565e-01 4.05531257e-01
6.10234201e-01 5.99541187e-01 6.83409095e-01 2.26992279e-01
1.53985530e-01 6.38565004e-01 -1.19990259e-01 -1.49992080e-02
-3.21992412e-02 -3.60664576e-02 2.82992810e-01 6.07533634e-01
1.92645833e-01 -2.76098520e-01 -1.02800858e+00 8.12647581e-01
-1.79466772e+00 -8.64822805e-01 4.77190018e-02 2.51405644e+00
1.17170703e+00 1.22353777e-01 3.28613430e-01 -6.57993183e-02
5.63397229e-01 -1.64815322e-01 -6.40610278e-01 -4.69597578e-01
-8.51837695e-02 2.26918116e-01 8.16789508e-01 5.50216675e-01
-1.08806050e+00 8.93831968e-01 7.78624344e+00 1.10948670e+00
-1.05756366e+00 2.36738734e-02 7.27535367e-01 -2.35372171e-01
-4.19577241e-01 -9.03996825e-02 -9.19891775e-01 5.59962928e-01
1.00287831e+00 -3.91769022e-01 5.29578090e-01 1.14861703e+00
-1.41562521e-01 -4.67995703e-02 -1.46732271e+00 8.95865440e-01
3.06783207e-02 -1.23329949e+00 -1.08530752e-01 -2.08367892e-02
8.72274399e-01 3.85391593e-01 2.89553493e-01 3.31589758e-01
6.51551545e-01 -9.65651512e-01 4.90871072e-01 -7.75579363e-02
8.51806402e-01 -6.57424867e-01 5.92988014e-01 4.71127033e-01
-5.84640801e-01 -3.09037775e-01 -5.59821844e-01 1.40815437e-01
-2.13454440e-01 6.32973373e-01 -1.05083919e+00 2.59305298e-01
4.88763094e-01 3.92741829e-01 -6.80595577e-01 1.13888240e+00
-1.41426623e-01 7.14321196e-01 -5.04331887e-01 -2.33556062e-01
3.55723888e-01 4.13033506e-03 4.99399006e-01 1.48465025e+00
2.31304213e-01 -2.12571993e-01 3.84354651e-01 4.73434985e-01
-2.32329965e-01 2.44480997e-01 -8.60800982e-01 -4.80591580e-02
5.54942429e-01 1.06311774e+00 -4.31974113e-01 -4.50474501e-01
-5.44214249e-01 9.18203473e-01 5.82970321e-01 5.02671540e-01
-3.23607743e-01 -5.08929670e-01 4.63881820e-01 2.04886068e-02
3.61248761e-01 -4.81292084e-02 -4.90965277e-01 -1.31738281e+00
1.63562059e-01 -1.25405896e+00 5.05354285e-01 -2.87447572e-01
-1.57921612e+00 6.61224902e-01 6.79413527e-02 -1.32922304e+00
-7.57778049e-01 -7.70493209e-01 -2.19999433e-01 9.01810586e-01
-1.55773330e+00 -6.27337039e-01 1.26023740e-01 4.87713128e-01
6.08467162e-01 -2.81774610e-01 7.64619946e-01 4.24260527e-01
-3.24423105e-01 8.06963921e-01 6.14239573e-01 7.43512586e-02
8.45889091e-01 -1.58017564e+00 4.35325831e-01 7.83978164e-01
3.57985824e-01 5.57645679e-01 7.83303201e-01 -2.85586685e-01
-1.20822978e+00 -9.83896375e-01 1.24783528e+00 -7.18910933e-01
6.91865563e-01 -3.92302036e-01 -9.13691282e-01 7.58540988e-01
-7.34000653e-02 -4.21575457e-02 8.45646441e-01 3.41670096e-01
-4.50940400e-01 -1.14033788e-01 -1.15420043e+00 5.60699522e-01
9.28241432e-01 -5.38886070e-01 -5.55106938e-01 8.89553249e-01
4.56436872e-01 -3.70242327e-01 -6.05646491e-01 1.81090876e-01
3.38717818e-01 -5.33357739e-01 7.81859398e-01 -1.07972133e+00
2.60361344e-01 -2.15602778e-02 -2.49407470e-01 -1.49213529e+00
-4.20782983e-01 -7.88921952e-01 1.75340608e-01 1.04778326e+00
1.00676715e+00 -5.39974868e-01 8.75032723e-01 9.49177861e-01
4.28680740e-02 -7.09296525e-01 -7.72516787e-01 -1.13753009e+00
3.36794734e-01 -3.63891363e-01 3.65040809e-01 8.42660666e-01
-2.27140993e-01 5.67257047e-01 -4.39800590e-01 -1.64188847e-01
5.91163218e-01 1.59110725e-01 9.58486974e-01 -1.12468481e+00
-5.21028578e-01 -4.10219878e-01 -1.66071311e-01 -1.21624172e+00
-9.52876806e-02 -8.70881855e-01 3.04260701e-01 -1.11442614e+00
1.37169302e-01 -6.69525325e-01 -2.91997313e-01 5.11290848e-01
-3.64741921e-01 1.80928215e-01 2.20804870e-01 4.38981593e-01
-5.27923882e-01 2.56664425e-01 1.08859336e+00 -1.46317095e-01
-3.05226982e-01 3.93832952e-01 -7.90696383e-01 5.71790218e-01
7.47172892e-01 -6.79946840e-01 -6.51531637e-01 -6.52901709e-01
1.99867681e-01 -1.74679294e-01 -3.62822674e-02 -6.93847239e-01
7.27354884e-02 -2.94145614e-01 3.16020578e-01 -1.72672600e-01
1.50348499e-01 -7.09968805e-01 -1.60324454e-01 1.05847448e-01
-7.35563934e-01 3.78413387e-02 4.56576310e-02 5.32584667e-01
-1.09237611e-01 -6.77863836e-01 9.40125763e-01 -3.00791800e-01
-6.38644755e-01 3.87200654e-01 -2.27612838e-01 6.10537887e-01
8.07952404e-01 -7.03710392e-02 -2.28805304e-01 -3.53082061e-01
-5.84526360e-01 1.57843992e-01 5.12792826e-01 1.81970030e-01
3.10639322e-01 -1.23915362e+00 -9.04508591e-01 2.06809282e-01
3.29774469e-01 -7.76870698e-02 -2.51317829e-01 5.60702145e-01
-3.47132117e-01 4.37618136e-01 8.97160918e-02 -6.28656983e-01
-1.17222381e+00 6.31531537e-01 1.49173692e-01 -3.73662621e-01
-2.82266587e-01 9.49473560e-01 1.38945222e-01 -6.71112537e-01
1.83034465e-01 3.12073175e-02 5.72985709e-02 -2.52202488e-02
5.79546750e-01 7.92207792e-02 1.76000774e-01 -2.18372732e-01
-3.72290134e-01 3.16902995e-01 -4.58150953e-01 -4.09270376e-01
1.24303675e+00 -1.85690112e-02 2.90013611e-01 5.04892170e-01
1.30402267e+00 -1.24711163e-01 -1.34034538e+00 -5.57810605e-01
4.06266838e-01 -7.95023799e-01 -1.10051017e-02 -8.40548158e-01
-7.84564912e-01 8.93823862e-01 3.13817143e-01 1.95042595e-01
1.10297918e+00 -7.94030726e-02 4.47526455e-01 8.23609948e-01
3.54459226e-01 -1.30380905e+00 7.26777166e-02 4.32918340e-01
6.81522429e-01 -1.48349631e+00 9.43536833e-02 -1.72257617e-01
-8.83989632e-01 1.04655170e+00 3.74570400e-01 -7.33028129e-02
4.39260691e-01 8.00117403e-02 1.24548234e-01 6.58792704e-02
-9.66694951e-01 -3.38264734e-01 3.98352265e-01 7.05964446e-01
4.67284083e-01 -9.53134298e-02 -3.55662704e-01 9.28207412e-02
-7.63659701e-02 1.43121211e-02 3.59284639e-01 9.86083567e-01
-5.65622091e-01 -1.41604674e+00 -1.77414075e-01 6.70073032e-01
-6.13460898e-01 -3.63259792e-01 -5.62205493e-01 8.49027514e-01
-3.93184155e-01 7.79318810e-01 -1.10101663e-01 -1.32823914e-01
1.86594337e-01 2.72091120e-01 4.96267051e-01 -8.66335511e-01
-4.62961584e-01 7.29109049e-02 3.09953094e-01 -3.56378585e-01
-1.79939210e-01 -7.18434930e-01 -7.63098359e-01 -2.62458593e-01
-4.31171447e-01 2.72045910e-01 6.77950740e-01 1.01626766e+00
5.27666926e-01 5.06724901e-02 9.48525488e-01 -6.26373112e-01
-1.00389314e+00 -8.33982468e-01 -4.12553072e-01 5.31543791e-01
3.97885680e-01 -4.44563627e-01 -4.33113724e-01 4.13477831e-02]
|
[9.513284683227539, 3.4852070808410645]
|
afbdf0db-f854-46ff-9301-de2bb3c0d2b3
|
a-review-of-sentiment-analysis-research-in
|
2005.12240
| null |
https://arxiv.org/abs/2005.12240v1
|
https://arxiv.org/pdf/2005.12240v1.pdf
|
A review of sentiment analysis research in Arabic language
|
Sentiment analysis is a task of natural language processing which has recently attracted increasing attention. However, sentiment analysis research has mainly been carried out for the English language. Although Arabic is ramping up as one of the most used languages on the Internet, only a few studies have focused on Arabic sentiment analysis so far. In this paper, we carry out an in-depth qualitative study of the most important research works in this context by presenting limits and strengths of existing approaches. In particular, we survey both approaches that leverage machine translation or transfer learning to adapt English resources to Arabic and approaches that stem directly from the Arabic language.
|
['Habib OUNELLI', 'Erik Cambria', 'Oumaima Oueslati', 'Moez Ben HajHmida']
|
2020-05-25
| null | null | null | null |
['arabic-sentiment-analysis']
|
['natural-language-processing']
|
[-5.36336452e-02 -1.52552038e-01 -4.26634014e-01 -5.07769227e-01
-4.36011910e-01 -7.24670887e-01 6.50573611e-01 5.60603499e-01
-7.30829179e-01 5.95684409e-01 2.06927717e-01 -5.18487751e-01
2.80357659e-01 -6.92834198e-01 -3.78659777e-02 -3.78268272e-01
2.92190433e-01 3.02688777e-01 -1.03290588e-01 -9.86312509e-01
7.70025373e-01 3.20676237e-01 -1.26367795e+00 4.29582536e-01
8.79002988e-01 6.85612321e-01 -2.50135452e-01 2.44747967e-01
-6.29093051e-01 8.86052847e-01 -6.60794795e-01 -9.19654608e-01
-2.36837510e-02 -6.44881725e-01 -1.00489962e+00 1.66101441e-01
-2.17445865e-01 3.78583111e-02 3.35612625e-01 1.18940413e+00
4.21804935e-01 -2.41502240e-01 4.90718603e-01 -1.13776577e+00
-8.99116158e-01 7.33650386e-01 -6.25037134e-01 2.79308837e-02
8.30596983e-01 -5.46188354e-01 8.49646986e-01 -1.06924617e+00
5.83372235e-01 1.09231031e+00 4.41306829e-01 3.36373955e-01
-4.33814228e-01 -5.30677915e-01 1.21829122e-01 3.39551628e-01
-9.77722168e-01 -1.54182777e-01 1.12672710e+00 -2.63623208e-01
9.04524744e-01 -1.74693525e-01 6.05435789e-01 6.06323957e-01
5.43501139e-01 7.20913351e-01 1.59004331e+00 -1.14682007e+00
1.03208438e-01 6.69127762e-01 8.29566866e-02 2.75966465e-01
1.88171536e-01 -7.87537217e-01 -5.65694451e-01 2.45043263e-01
8.27221125e-02 -6.50911108e-02 3.29394400e-01 -4.15118374e-02
-1.03922939e+00 1.19283390e+00 1.29581824e-01 7.74227977e-01
-3.66524488e-01 -6.80883944e-01 8.88286352e-01 6.75930738e-01
6.61299944e-01 3.21038783e-01 -7.07151890e-01 -5.53120673e-01
-6.70725524e-01 -8.30867048e-03 1.05143642e+00 9.32982087e-01
5.72416008e-01 -1.98855907e-01 6.29508197e-01 7.03458607e-01
6.46004856e-01 7.39787400e-01 7.28597224e-01 -2.95258135e-01
4.88675326e-01 1.08599448e+00 4.69346717e-02 -1.16434085e+00
-3.18489760e-01 1.58962265e-01 -2.85204351e-01 -5.34862801e-02
3.26125473e-01 -5.30379713e-01 -5.18929005e-01 1.02807462e+00
2.69671589e-01 -9.26646709e-01 4.70877796e-01 7.53844619e-01
8.72880042e-01 7.16486156e-01 1.59326807e-01 -2.14212880e-01
1.67485785e+00 -1.22748494e+00 -8.63682508e-01 -4.17824894e-01
7.28096247e-01 -1.37988758e+00 1.21030486e+00 5.67131162e-01
-1.02645397e+00 -1.49813905e-01 -9.55765307e-01 -1.99813381e-01
-1.17363513e+00 7.93399960e-02 6.99334025e-01 1.32315016e+00
-9.82172132e-01 -6.24546222e-02 -7.93917358e-01 -1.03405511e+00
5.46507388e-02 2.51223534e-01 -4.03043717e-01 -1.93192050e-01
-1.26068664e+00 1.32222009e+00 2.83057749e-01 1.90869302e-01
2.25930735e-01 1.00965023e-01 -7.79937387e-01 -5.47360063e-01
3.93971115e-01 -1.86511583e-03 1.37119520e+00 -1.62870538e+00
-1.87214482e+00 1.00247240e+00 -3.87116015e-01 -1.53553307e-01
-2.40677953e-01 -1.79883108e-01 -8.71806622e-01 5.25837578e-02
2.46457085e-01 1.60732150e-01 4.49459076e-01 -9.70866919e-01
-9.13753986e-01 -6.32893860e-01 3.23608577e-01 3.18259031e-01
-8.31398726e-01 1.11300611e+00 -1.66500077e-01 -7.49554396e-01
9.89747047e-02 -9.31209326e-01 -1.48420453e-01 -5.59745491e-01
2.65191376e-01 -3.07956696e-01 5.65398932e-01 -7.11178541e-01
1.56394815e+00 -1.92018962e+00 -6.57661781e-02 1.63557217e-01
-4.26416904e-01 1.75131947e-01 5.48576675e-02 1.22194672e+00
1.11573569e-01 2.39956215e-01 -2.74180800e-01 -1.73335224e-01
1.04860745e-01 1.61198184e-01 -4.05984074e-01 3.48357528e-01
2.81581670e-01 6.63390398e-01 -1.07425952e+00 -5.64028621e-01
-2.76503880e-02 3.51346731e-01 -2.78901845e-01 -2.66238093e-01
2.11252004e-01 2.93434143e-01 -6.25755131e-01 1.16468561e+00
4.60444540e-01 1.68006644e-01 4.53301102e-01 2.28898209e-02
-6.10406399e-01 4.22179461e-01 -7.78217256e-01 1.43789673e+00
-5.85075319e-01 5.17088473e-01 2.41510803e-03 -1.03871477e+00
9.27654266e-01 4.08920169e-01 4.40548986e-01 -6.56288445e-01
4.79644537e-01 7.16380775e-01 6.83649480e-02 -5.18765986e-01
8.46398711e-01 -5.14978766e-01 -5.05226195e-01 6.37883306e-01
-9.99119803e-02 -4.53542203e-01 6.62157476e-01 -9.80775133e-02
2.77006894e-01 3.72421861e-01 8.15937281e-01 -1.38741776e-01
1.01801181e+00 5.66688061e-01 7.57653564e-02 5.61887547e-02
-5.07610619e-01 2.63027072e-01 4.51812088e-01 -3.86489511e-01
-7.18062341e-01 -3.42924118e-01 -1.00583769e-01 1.30722034e+00
-1.55082405e-01 -8.15109968e-01 -9.51072752e-01 -8.48676920e-01
-4.82668668e-01 4.47363943e-01 -6.23570323e-01 2.76725799e-01
-3.83888900e-01 -8.73224318e-01 3.80187660e-01 2.91098386e-01
4.12233293e-01 -1.49045587e+00 -6.42580628e-01 1.90627515e-01
-3.44810903e-01 -1.10019100e+00 -1.29348377e-03 5.06216697e-02
-8.35944414e-01 -9.59107339e-01 -6.39667332e-01 -1.17129195e+00
6.61981583e-01 2.06680998e-01 1.07051766e+00 -4.92332913e-02
3.68551552e-01 3.68889004e-01 -1.33834100e+00 -1.20105803e+00
-4.87002760e-01 4.71081138e-01 3.97473983e-02 -1.15648195e-01
1.33174896e+00 3.82258259e-02 -5.59189096e-02 1.58423752e-01
-1.24047565e+00 -3.37360203e-01 6.17945790e-01 3.82077605e-01
9.41263959e-02 9.32528526e-02 9.62479115e-01 -1.11770403e+00
9.41254377e-01 -6.79091692e-01 -6.84248805e-02 2.55603999e-01
-7.11538136e-01 -4.49022472e-01 7.25586057e-01 1.48971781e-01
-1.02179182e+00 -1.50016427e-01 -4.22312468e-01 7.61424780e-01
-2.52959639e-01 1.39003193e+00 1.04248352e-01 -8.90221447e-02
4.53501910e-01 -1.75967291e-01 1.44375503e-01 -5.99199571e-02
1.48830384e-01 1.20983589e+00 -3.40848297e-01 -3.93731892e-01
5.06852269e-01 5.46960115e-01 -4.26107347e-01 -1.10783505e+00
-8.53607416e-01 -5.51470876e-01 -9.43117678e-01 -4.54743356e-01
8.41824353e-01 -6.93078339e-01 -1.80306226e-01 5.89837134e-01
-7.12158978e-01 1.45078516e-02 -3.12666632e-02 4.54592615e-01
-1.68882579e-01 4.78119552e-01 -5.74292600e-01 -7.93737829e-01
-5.07397413e-01 -1.12149894e+00 7.21906245e-01 2.61819154e-01
-4.59159911e-01 -1.29129612e+00 2.19796091e-01 4.38668221e-01
5.50772607e-01 7.55510777e-02 6.64737046e-01 -7.34195411e-01
5.25664449e-01 -5.75297296e-01 9.94011462e-02 3.13699365e-01
6.73872054e-01 2.88714439e-01 -6.78102911e-01 -2.87087232e-01
2.88522840e-01 -5.75755596e-01 2.31065497e-01 3.92895192e-02
2.24080175e-01 -1.28598630e-01 1.05494797e-01 -2.80699223e-01
1.46068251e+00 4.11209285e-01 6.52106583e-01 1.26013350e+00
1.96725488e-01 1.07344198e+00 1.25754809e+00 4.16513681e-01
7.80194461e-01 1.32204860e-01 1.16240397e-01 1.53675407e-01
5.02763808e-01 3.33160907e-02 7.85408616e-01 1.50837982e+00
-1.72499225e-01 3.29382494e-02 -1.32610297e+00 5.30411005e-01
-1.60002565e+00 -5.14075935e-01 -1.13565415e-01 1.81138551e+00
8.59105051e-01 1.61398143e-01 2.01736271e-01 4.12216425e-01
3.77166629e-01 6.45566136e-02 1.43196076e-01 -1.03464413e+00
-2.82646835e-01 5.41968942e-01 1.38494119e-01 1.75974011e-01
-1.11323273e+00 1.30676520e+00 5.85032988e+00 4.33425784e-01
-1.25945044e+00 -1.06499800e-02 4.70303148e-01 4.68251646e-01
-1.81430995e-01 2.26851851e-01 -4.74338949e-01 2.32357636e-01
1.04135466e+00 -1.46233346e-02 1.51209518e-01 7.63121963e-01
3.12581033e-01 -3.00479531e-01 -5.08143008e-01 5.88694215e-01
5.93602300e-01 -5.94139576e-01 2.59752691e-01 -3.13749462e-01
8.93734634e-01 -5.34026995e-02 1.61689013e-01 4.04442340e-01
-3.38966489e-01 -9.44020689e-01 7.17449963e-01 9.82160419e-02
3.89702260e-01 -1.10273945e+00 1.48774374e+00 1.55019268e-01
-8.86105835e-01 5.26445992e-02 -1.97820142e-01 -5.68322241e-01
2.90893614e-01 3.48051012e-01 -5.09722948e-01 6.89170539e-01
1.10711896e+00 8.95491421e-01 -6.83013260e-01 4.16782737e-01
-4.87658352e-01 7.82355130e-01 -5.62116317e-03 -6.72079146e-01
4.96035755e-01 -9.23129916e-01 1.87038332e-01 1.52767241e+00
1.73093125e-01 3.17039900e-02 1.16757162e-01 -1.94752082e-01
2.33067736e-01 1.06396997e+00 -6.88087702e-01 -5.53772807e-01
6.56296462e-02 1.38580072e+00 -1.34737647e+00 -1.78314954e-01
-1.03007329e+00 8.23967099e-01 -9.71761569e-02 8.36102143e-02
-4.89459902e-01 -8.78692627e-01 1.26099721e-01 -7.95932487e-02
1.46127239e-01 -3.12935680e-01 -5.02194941e-01 -1.12235355e+00
-3.55361514e-02 -1.34362006e+00 3.82554233e-01 -6.05055451e-01
-1.48575222e+00 8.38335752e-01 -2.08595589e-01 -1.36271310e+00
-3.44484627e-01 -9.78330910e-01 -3.31931025e-01 8.20678830e-01
-1.67119992e+00 -1.35822439e+00 1.44656345e-01 4.28337157e-01
5.60569525e-01 -3.50321859e-01 1.15106428e+00 3.62144232e-01
-4.07396197e-01 3.04420978e-01 1.27634734e-01 1.01866968e-01
1.20968807e+00 -1.09744632e+00 8.66370276e-02 8.30296040e-01
-2.10628420e-01 1.03679502e+00 6.33133411e-01 -6.22245908e-01
-1.46940589e+00 -3.33581656e-01 1.65414631e+00 -4.65264469e-01
1.12549686e+00 -1.09481387e-01 -6.01512671e-01 7.82593250e-01
9.59880829e-01 -6.20911121e-01 1.17787290e+00 2.81437207e-02
8.76038298e-02 -3.30211408e-02 -1.19191206e+00 9.08250630e-01
-5.32547534e-02 -5.57280719e-01 -7.94125080e-01 5.11651263e-02
1.15831330e-01 -2.43799776e-01 -9.27781582e-01 1.53676867e-01
7.24322021e-01 -6.19851887e-01 3.76943648e-01 -6.15378916e-01
7.50118732e-01 -4.03489769e-01 -1.20628141e-01 -1.20571530e+00
1.61481902e-01 -3.70631188e-01 3.89195204e-01 1.39099109e+00
7.05085039e-01 -7.57125735e-01 6.97571099e-01 3.65575105e-01
3.65339704e-02 -6.96932256e-01 -4.75342482e-01 -5.79813644e-02
4.98874247e-01 -4.37086463e-01 4.29438502e-01 1.18288732e+00
7.14814246e-01 5.05963981e-01 8.51478130e-02 -3.51697028e-01
-1.01828389e-01 2.05013245e-01 6.39797926e-01 -9.62548912e-01
5.51216245e-01 -5.87441623e-01 -1.84493408e-01 -5.62750041e-01
1.38635755e-01 -6.49322748e-01 1.36659704e-02 -1.78275216e+00
-9.05360747e-03 -1.00544907e-01 -1.47076517e-01 4.88665879e-01
-1.09231889e-01 6.57909930e-01 1.28507122e-01 4.88670776e-03
-4.99081880e-01 2.31242329e-01 9.17225003e-01 1.60550848e-01
-3.36578399e-01 -2.72094328e-02 -1.23230314e+00 1.05427909e+00
1.49785340e+00 -4.38752532e-01 -2.17475861e-01 -3.30271006e-01
1.09207070e+00 -5.08686960e-01 -6.71770632e-01 -5.06968856e-01
2.03852534e-01 -4.59556222e-01 -3.55393328e-02 -4.82331038e-01
-2.53586680e-01 -1.04782796e+00 -5.40416002e-01 1.74718693e-01
8.96104723e-02 8.33240747e-01 4.03576612e-01 5.19696325e-02
-8.17603886e-01 -7.73309529e-01 4.26115662e-01 -1.94366112e-01
-7.60801494e-01 -1.16133712e-01 -1.13799608e+00 -3.21359150e-02
1.05342400e+00 -3.80960673e-01 7.69135207e-02 -4.75879937e-01
-1.76597506e-01 5.81034236e-02 6.89099729e-01 6.69084609e-01
5.40390491e-01 -9.43404317e-01 -6.39822900e-01 -1.02831997e-01
3.42937142e-01 -4.30525810e-01 -1.34809777e-01 9.12141979e-01
-9.22757328e-01 6.61917150e-01 -5.21167159e-01 1.84316382e-01
-9.77357149e-01 6.29942715e-01 -2.16417000e-01 -3.22430074e-01
2.29340047e-02 2.34440714e-01 -6.54851139e-01 -7.37514496e-01
-2.50400096e-01 -4.66078185e-02 -1.16890645e+00 5.61269224e-01
4.17558312e-01 3.30685556e-01 3.31237555e-01 -1.25993657e+00
-5.66448808e-01 6.50769413e-01 -5.70226349e-02 -4.93314624e-01
1.24767220e+00 -4.37085003e-01 -9.75040913e-01 9.26564515e-01
6.94108725e-01 6.63901627e-01 -1.15235224e-01 8.05011839e-02
4.12002236e-01 -1.64026320e-01 -2.37640232e-01 -9.12548423e-01
-6.85894847e-01 7.02473462e-01 2.86580056e-01 6.11571252e-01
1.36330736e+00 -3.51440310e-01 6.48778498e-01 4.86027360e-01
5.09863257e-01 -1.55664730e+00 -2.05068905e-02 1.08515930e+00
7.30280697e-01 -1.40240085e+00 1.26750499e-01 -4.29564595e-01
-9.31351006e-01 1.43485641e+00 3.92135561e-01 -2.58031804e-02
1.09274018e+00 9.46915224e-02 7.48215020e-01 -7.85048530e-02
8.46720338e-02 -7.66518191e-02 4.79775369e-02 5.32722950e-01
1.36168778e+00 -1.48006722e-01 -1.14410448e+00 8.28541875e-01
-5.35269976e-01 1.43077761e-01 8.58502507e-01 1.64573741e+00
-3.04056048e-01 -1.69389689e+00 -5.44148982e-01 3.37952942e-01
-1.13232052e+00 -3.00075263e-01 -7.38715291e-01 9.78163302e-01
-7.50645101e-02 1.43539310e+00 -1.81878671e-01 -1.55907720e-01
4.18431997e-01 1.74386382e-01 3.17420840e-01 -8.09713900e-01
-9.84996438e-01 -1.64316550e-01 2.36359481e-02 4.69457954e-02
-1.42975831e+00 -8.59250844e-01 -1.36548233e+00 -3.43198597e-01
-2.87053645e-01 5.93024075e-01 9.89845395e-01 1.09020722e+00
-1.52684720e-02 6.81025758e-02 4.80073005e-01 -6.01728499e-01
-7.06974119e-02 -1.24362528e+00 -6.74007893e-01 6.52437508e-02
4.49004918e-02 -3.01828384e-01 -1.48991317e-01 1.84257716e-01]
|
[11.021263122558594, 6.877188205718994]
|
82d90016-b886-410a-b927-ad0ed548efd2
|
fdti-fine-grained-deep-traffic-inference-with
|
2306.10945
| null |
https://arxiv.org/abs/2306.10945v1
|
https://arxiv.org/pdf/2306.10945v1.pdf
|
FDTI: Fine-grained Deep Traffic Inference with Roadnet-enriched Graph
|
This paper proposes the fine-grained traffic prediction task (e.g. interval between data points is 1 minute), which is essential to traffic-related downstream applications. Under this setting, traffic flow is highly influenced by traffic signals and the correlation between traffic nodes is dynamic. As a result, the traffic data is non-smooth between nodes, and hard to utilize previous methods which focus on smooth traffic data. To address this problem, we propose Fine-grained Deep Traffic Inference, termed as FDTI. Specifically, we construct a fine-grained traffic graph based on traffic signals to model the inter-road relations. Then, a physically-interpretable dynamic mobility convolution module is proposed to capture vehicle moving dynamics controlled by the traffic signals. Furthermore, traffic flow conservation is introduced to accurately infer future volume. Extensive experiments demonstrate that our method achieves state-of-the-art performance and learned traffic dynamics with good properties. To the best of our knowledge, we are the first to conduct the city-level fine-grained traffic prediction.
|
['Hua Wei', 'Guanjie Zheng', 'Chumeng Liang', 'Zhanyu Liu']
|
2023-06-19
| null | null | null | null |
['traffic-prediction']
|
['time-series']
|
[-3.50158840e-01 -3.31562102e-01 -4.30227071e-01 -3.85639191e-01
-8.14425871e-02 -1.18336082e-01 5.23416817e-01 -4.04751688e-01
1.50269225e-01 8.38856339e-01 -6.88703433e-02 -8.17319751e-01
-4.62284833e-01 -1.33341825e+00 -8.13543081e-01 -5.27283132e-01
-2.36640483e-01 6.56292617e-01 6.35940909e-01 -4.21680748e-01
-3.68813681e-03 7.29990959e-01 -1.36782837e+00 -1.24717459e-01
1.12146246e+00 1.11633432e+00 9.27118436e-02 4.97770488e-01
-4.93145466e-01 6.62563443e-01 -1.28086939e-01 -2.44215876e-01
1.38966843e-01 1.75686628e-01 -4.54758674e-01 5.94608411e-02
3.57268453e-01 -3.26103240e-01 -1.25737166e+00 6.57184541e-01
-1.95723519e-01 3.09086710e-01 4.58308876e-01 -1.47319376e+00
-6.34754539e-01 4.33136344e-01 -6.10839009e-01 6.80968225e-01
-4.06027079e-01 6.81767762e-01 8.24556768e-01 -4.47597831e-01
1.11173704e-01 1.52769005e+00 4.78357971e-01 2.54904002e-01
-1.09202862e+00 -9.90660965e-01 6.89746797e-01 3.16296637e-01
-1.46877444e+00 -3.81837666e-01 9.04827535e-01 -5.96885026e-01
5.48347712e-01 2.22297981e-01 5.39930999e-01 9.70873415e-01
1.27181396e-01 5.10801733e-01 6.54252052e-01 3.32221180e-01
-1.52306706e-01 -2.43646413e-01 1.91237539e-01 6.68799102e-01
2.38875017e-01 2.35998780e-01 -1.53513312e-01 2.59676158e-01
8.65161657e-01 2.97086835e-01 1.43062159e-01 8.76516700e-02
-9.55235958e-01 4.49061841e-01 6.26746655e-01 6.38343990e-02
-3.09737235e-01 6.47505641e-01 3.93661559e-01 5.09917550e-02
6.61843359e-01 -4.05824035e-01 -3.74818683e-01 -3.77617747e-01
-8.92374933e-01 2.21622452e-01 5.00244915e-01 1.09120131e+00
1.11349809e+00 1.95807800e-01 -3.02918553e-01 6.15370095e-01
3.58896285e-01 8.06122839e-01 -1.54553682e-01 -1.04687309e+00
9.24308419e-01 4.55714166e-01 -3.15062888e-02 -1.44589174e+00
-4.49016780e-01 -4.70901012e-01 -1.18134749e+00 -1.09578080e-01
6.10802829e-01 -1.30107641e-01 -7.97258914e-01 1.62578928e+00
2.70057291e-01 1.15303004e+00 -6.25277698e-01 8.68460953e-01
3.84904832e-01 8.50286663e-01 2.84477621e-01 -8.44453555e-03
1.07363629e+00 -9.34367537e-01 -5.75122893e-01 -5.06365448e-02
5.10162473e-01 -2.84086913e-01 1.04431272e+00 -2.78708249e-01
-8.98920894e-01 -7.13977039e-01 -4.63320524e-01 4.09175754e-02
-4.69065040e-01 -2.90064573e-01 7.51678526e-01 4.71527576e-01
-6.55183613e-01 3.95503521e-01 -1.05665946e+00 -3.61971580e-03
8.16982329e-01 2.32023403e-01 1.53115690e-01 -1.51559159e-01
-1.35648048e+00 4.80322570e-01 3.45193632e-02 4.39409584e-01
-6.99972212e-01 -1.12625825e+00 -6.79908454e-01 2.20173135e-01
4.61241394e-01 -7.90119231e-01 1.10654461e+00 -3.15030754e-01
-1.33950257e+00 3.15739930e-01 -4.59003925e-01 -4.65253353e-01
6.72336936e-01 1.73954695e-01 -1.03183961e+00 -1.75473675e-01
2.70148426e-01 8.09501410e-02 5.40786445e-01 -1.16442013e+00
-9.71031785e-01 -1.90642625e-01 2.77802587e-01 -4.14808154e-01
-1.30276933e-01 -3.93415064e-01 -9.49571729e-01 -6.33388638e-01
-3.73732001e-01 -9.70170081e-01 -3.44182223e-01 1.86544266e-02
-5.08041859e-01 -3.06930602e-01 1.14861238e+00 -2.57022440e-01
1.81740415e+00 -2.02422214e+00 -6.19364619e-01 4.99913782e-01
6.08331382e-01 2.36119539e-01 -8.05748925e-02 1.34992763e-01
3.44932348e-01 2.37764567e-01 -4.53483015e-02 -1.93122953e-01
2.65959144e-01 4.33190405e-01 -4.35134917e-01 3.87347311e-01
7.06423372e-02 1.32260144e+00 -9.29921985e-01 -5.42984068e-01
6.10380471e-01 4.79789019e-01 -3.20004374e-01 -1.37667313e-01
-1.80922449e-01 7.26499379e-01 -1.12741482e+00 4.61708963e-01
1.07365882e+00 -4.18688655e-01 -3.28679413e-01 -2.39624560e-01
-3.35148484e-01 2.74630308e-01 -7.57694483e-01 1.19505394e+00
-1.00117528e+00 8.73421431e-01 3.57062183e-02 -1.15400875e+00
8.10955346e-01 7.57484837e-03 6.17637753e-01 -1.11751890e+00
-1.88392047e-02 9.45745483e-02 -7.36010894e-02 -4.02723491e-01
4.18029457e-01 2.15029567e-02 -1.72660649e-01 1.51640728e-01
-6.52646840e-01 4.54141229e-01 3.11964273e-01 1.17227107e-01
1.13032639e+00 -2.66547173e-01 -3.50051373e-01 -1.97278902e-01
6.00418210e-01 -7.06959069e-02 7.64137626e-01 5.19803107e-01
-3.23072970e-01 8.07146654e-02 4.28762108e-01 -5.26672363e-01
-8.96615684e-01 -1.34610128e+00 -1.96546376e-01 9.69633162e-01
5.32436430e-01 -1.42996192e-01 -5.08244395e-01 -3.86636615e-01
4.32700068e-01 4.58301067e-01 -6.16107166e-01 -9.53371674e-02
-1.10068822e+00 -5.45659781e-01 4.90682632e-01 5.10016441e-01
5.77892244e-01 -5.11856079e-01 2.06065699e-01 5.87185323e-01
-8.61264542e-02 -1.43293142e+00 -8.01613867e-01 -6.18180990e-01
-6.59114480e-01 -1.01278734e+00 -3.13163102e-01 -4.01846319e-01
5.08555651e-01 6.68787479e-01 1.16758335e+00 2.48614162e-01
-7.44845793e-02 -1.79180250e-01 1.76794194e-02 7.29812905e-02
-2.99658142e-02 3.76098692e-01 -1.55447543e-01 3.75540614e-01
5.42512596e-01 -9.64581311e-01 -8.54896188e-01 6.92480564e-01
-6.35667503e-01 8.25247839e-02 4.44214791e-01 5.55530429e-01
7.17110097e-01 5.84010303e-01 6.16294444e-01 -7.99540520e-01
4.83185530e-01 -9.80618775e-01 -7.62924492e-01 -6.20428585e-02
-3.29989761e-01 -7.90732913e-03 1.00049853e+00 -3.53667200e-01
-1.23503864e+00 -3.64716589e-01 -1.27681360e-01 -6.54455423e-01
-4.17129695e-01 2.44743451e-01 -3.99720669e-01 1.21089473e-01
9.60163251e-02 5.60385883e-02 -4.25207973e-01 -4.16480005e-01
4.68355417e-01 6.67873859e-01 5.19358933e-01 -8.15909922e-01
1.16608143e+00 8.28435600e-01 4.55384284e-01 -7.69470155e-01
-7.82365739e-01 -3.04867089e-01 -5.13521254e-01 -3.45100284e-01
6.20011806e-01 -7.92926013e-01 -1.29792786e+00 4.15874690e-01
-9.74110603e-01 -6.46330059e-01 -1.30482227e-01 3.67654502e-01
-4.53916937e-01 1.84478998e-01 -6.27106011e-01 -7.03494191e-01
1.72453597e-01 -1.15100551e+00 1.09506595e+00 1.46567866e-01
3.48644108e-01 -1.30057716e+00 -1.63654089e-01 3.61827910e-01
6.04187310e-01 2.55327255e-01 7.62533963e-01 1.56234384e-01
-1.31891727e+00 8.90898630e-02 -8.30762208e-01 -2.04260945e-01
3.38682026e-01 2.36596510e-01 -5.97450137e-01 1.74371526e-01
-6.89788222e-01 5.33907175e-01 1.03868985e+00 6.87909544e-01
1.82962704e+00 -2.65915126e-01 -7.92911291e-01 8.11999321e-01
9.07798052e-01 7.94498101e-02 6.60497844e-01 1.74315706e-01
1.20351779e+00 5.91265202e-01 6.57984614e-01 4.04549450e-01
1.00367594e+00 8.71095479e-01 3.30204934e-01 -1.15645669e-01
-3.50009710e-01 -5.18194020e-01 -1.14036597e-01 7.30667412e-01
2.72230338e-02 -4.31975901e-01 -9.88312185e-01 5.90797901e-01
-2.10316229e+00 -1.16593146e+00 -7.49761105e-01 1.80513692e+00
1.32506683e-01 5.43505609e-01 3.06067556e-01 -1.89482942e-01
9.17763591e-01 3.51814896e-01 -8.47715139e-01 -1.54971749e-01
1.00407243e-01 3.08855381e-02 9.85756040e-01 6.00600183e-01
-8.65470886e-01 1.23328900e+00 5.72395992e+00 1.26376402e+00
-1.04551828e+00 1.82289556e-01 7.02373743e-01 -2.58816872e-02
-5.06183743e-01 -8.89556035e-02 -8.26246858e-01 1.25545752e+00
1.21517968e+00 -3.61422002e-01 5.29724956e-01 4.81144100e-01
1.02847517e+00 3.81270826e-01 -6.51284635e-01 8.42598557e-01
-8.07855308e-01 -1.67770612e+00 1.07496440e-01 3.91232878e-01
6.19172394e-01 2.44193420e-01 2.05418661e-01 4.83246714e-01
5.11384547e-01 -1.12298489e+00 2.62438834e-01 9.92787361e-01
9.29092526e-01 -9.11001682e-01 1.88876256e-01 2.29491308e-01
-2.04483390e+00 -7.75942355e-02 -1.08745664e-01 -3.53862345e-02
7.98058689e-01 1.06419349e+00 -2.27358043e-01 3.88682306e-01
4.35387850e-01 1.06791806e+00 -2.36369297e-01 9.95855212e-01
1.87762812e-01 9.23671663e-01 -3.70614648e-01 2.67244428e-01
4.23057556e-01 -2.41242349e-01 3.63679081e-01 1.26170981e+00
4.47953790e-01 3.38664465e-02 3.64421785e-01 1.00564718e+00
-3.64889324e-01 -3.80504936e-01 -4.93926942e-01 2.70539582e-01
7.42583871e-01 1.06942499e+00 -3.55573118e-01 -3.14785212e-01
-7.41269946e-01 4.97566193e-01 2.49109462e-01 7.18465388e-01
-1.11871672e+00 -4.89266157e-01 1.47254777e+00 6.53681576e-01
3.73293549e-01 -5.57631493e-01 -3.72312427e-01 -1.18886697e+00
8.54014009e-02 -8.47992077e-02 9.27313417e-02 -4.29430395e-01
-1.53271544e+00 3.07602793e-01 -1.98913693e-01 -1.28735006e+00
1.11626245e-01 -5.85096359e-01 -1.00738013e+00 8.41224372e-01
-1.96596372e+00 -1.25695920e+00 -5.45299053e-01 7.66411126e-01
4.54912663e-01 3.50171700e-02 -7.06479475e-02 9.40390944e-01
-9.14148450e-01 5.99696875e-01 1.96459711e-01 1.84589103e-01
1.54001594e-01 -8.52464616e-01 9.60634470e-01 6.23930573e-01
-3.38136524e-01 2.94313699e-01 4.29365873e-01 -6.68356240e-01
-1.34086490e+00 -1.71107423e+00 7.19910800e-01 -4.64096069e-01
1.29602039e+00 -1.70531631e-01 -1.01993477e+00 5.19592345e-01
-5.74914694e-01 6.26882315e-01 2.86865115e-01 5.18948361e-02
-2.10687518e-01 -6.74842238e-01 -8.68859470e-01 5.20583868e-01
1.59368122e+00 -5.64910054e-01 -2.34737806e-02 2.42883578e-01
1.03989816e+00 -2.81136751e-01 -8.96067977e-01 2.49279872e-01
4.66494948e-01 -4.24935460e-01 1.10428023e+00 -5.81358910e-01
3.32180341e-03 -4.44526613e-01 -9.85092111e-03 -1.26776218e+00
-5.22112727e-01 -5.93762875e-01 -6.54089749e-01 1.17092776e+00
1.79947048e-01 -8.13401103e-01 9.29089069e-01 5.57700098e-01
-3.54720652e-01 -6.05030179e-01 -9.95936871e-01 -1.03868699e+00
1.97210670e-01 -9.26859856e-01 1.29574168e+00 6.37248218e-01
-4.56196189e-01 1.85964152e-01 -5.40740848e-01 4.39482719e-01
7.93255925e-01 3.24970663e-01 1.00074363e+00 -1.40262127e+00
3.71584110e-02 -9.44956779e-01 -5.07205665e-01 -1.73479891e+00
6.78408921e-01 -6.66615307e-01 -1.43769026e-01 -1.32743263e+00
-2.21024111e-01 -9.89915907e-01 -3.30539852e-01 -5.62068038e-02
-8.75743032e-02 6.40565231e-02 -1.46398604e-01 1.67736933e-01
-6.45428538e-01 7.40060508e-01 1.68393290e+00 -1.50854975e-01
-1.11164138e-01 3.92170280e-01 -6.83438659e-01 4.16350812e-01
1.13297355e+00 -1.82575673e-01 -4.86175120e-01 -4.67349112e-01
-1.00474842e-01 1.35726109e-01 5.49146771e-01 -8.58316243e-01
3.64280909e-01 -8.01398098e-01 -1.03153355e-01 -9.47028697e-01
3.13770115e-01 -9.39154267e-01 1.50372446e-01 9.38502178e-02
1.97717384e-03 -1.20962717e-01 2.29565457e-01 9.65520024e-01
-2.29991719e-01 8.74844730e-01 5.11640787e-01 2.71184981e-01
-8.14660907e-01 1.40767300e+00 -5.05301654e-01 2.19986215e-01
1.07652724e+00 -2.15046480e-01 -5.70302844e-01 -1.94763511e-01
-5.45915484e-01 8.55987787e-01 2.07132414e-01 5.26099205e-01
2.81909794e-01 -1.68476295e+00 -5.33362508e-01 -2.23400854e-02
1.47829816e-01 9.41014737e-02 7.73079276e-01 1.02661681e+00
-2.54097044e-01 7.62806892e-01 8.33974034e-02 -5.28374195e-01
-6.12104595e-01 6.42771602e-01 3.82761329e-01 -2.09967092e-01
-7.28271246e-01 1.82402864e-01 4.63183969e-01 -3.64523619e-01
-1.14207707e-01 -5.72221696e-01 1.44861907e-01 -3.72190177e-01
4.72598821e-01 8.81611824e-01 -1.35697111e-01 -1.06323433e+00
-3.33307654e-01 6.96886182e-01 1.59685418e-01 3.80948991e-01
1.06805193e+00 -7.96392560e-01 2.20377326e-01 3.36749882e-01
1.23121500e+00 -1.18534192e-01 -1.58504570e+00 -4.90403056e-01
-2.72822767e-01 -9.06605005e-01 2.55225778e-01 -2.97010034e-01
-1.65730345e+00 1.21904540e+00 1.95266202e-01 4.32139486e-01
1.00115359e+00 -1.87782142e-02 1.65452695e+00 3.73006701e-01
3.35027397e-01 -1.00497520e+00 -3.41681629e-01 7.76345313e-01
1.73495382e-01 -1.20165873e+00 -6.46541536e-01 -6.88867033e-01
-2.02390462e-01 9.10382986e-01 9.17960286e-01 -8.69011953e-02
1.17192423e+00 1.84136808e-01 -3.63440305e-01 -2.88375944e-01
-8.12303960e-01 -6.04542732e-01 2.49214798e-01 4.85801399e-01
-4.04589996e-03 5.18333018e-01 1.18790403e-01 3.73612672e-01
-7.00780898e-02 2.35599771e-01 1.37261420e-01 2.15736821e-01
-5.96007168e-01 -8.21889937e-01 3.94975096e-02 7.10563540e-01
-1.66154541e-02 4.67839018e-02 3.32501739e-01 7.54406452e-01
1.75220206e-01 1.08028841e+00 5.53590536e-01 -4.07488257e-01
4.95820373e-01 -6.87895060e-01 -5.54607324e-02 -2.19447151e-01
4.87049408e-02 -4.89162445e-01 -2.09675692e-02 -8.58659923e-01
-5.32215945e-02 -3.95729870e-01 -1.39840353e+00 -1.37149763e+00
1.04878627e-01 7.04046264e-02 3.47857744e-01 1.07081068e+00
6.38966203e-01 8.23122263e-01 9.79174078e-01 -8.21988106e-01
2.53365457e-01 -4.80535179e-01 -6.89932942e-01 4.33586955e-01
7.01434612e-01 -1.04500949e+00 -2.02020004e-01 -3.31940502e-01]
|
[6.466664791107178, 2.0357916355133057]
|
6c0fca20-a1eb-48fb-bf4e-4582eaf3f67d
|
cascaded-zoom-in-detector-for-high-resolution
|
2303.08747
| null |
https://arxiv.org/abs/2303.08747v1
|
https://arxiv.org/pdf/2303.08747v1.pdf
|
Cascaded Zoom-in Detector for High Resolution Aerial Images
|
Detecting objects in aerial images is challenging because they are typically composed of crowded small objects distributed non-uniformly over high-resolution images. Density cropping is a widely used method to improve this small object detection where the crowded small object regions are extracted and processed in high resolution. However, this is typically accomplished by adding other learnable components, thus complicating the training and inference over a standard detection process. In this paper, we propose an efficient Cascaded Zoom-in (CZ) detector that re-purposes the detector itself for density-guided training and inference. During training, density crops are located, labeled as a new class, and employed to augment the training dataset. During inference, the density crops are first detected along with the base class objects, and then input for a second stage of inference. This approach is easily integrated into any detector, and creates no significant change in the standard detection process, like the uniform cropping approach popular in aerial image detection. Experimental results on the aerial images of the challenging VisDrone and DOTA datasets verify the benefits of the proposed approach. The proposed CZ detector also provides state-of-the-art results over uniform cropping and other density cropping methods on the VisDrone dataset, increasing the detection mAP of small objects by more than 3 points.
|
['Marco Pedersoli', 'Eric Granger', 'Akhil Meethal']
|
2023-03-15
| null | null | null | null |
['small-object-detection']
|
['computer-vision']
|
[ 2.13500470e-01 -2.22010270e-01 2.82841265e-01 7.23731518e-02
-2.17523873e-01 -5.17140925e-01 5.21914482e-01 2.15430006e-01
-7.67754018e-01 6.18074834e-01 -4.44783330e-01 1.60304278e-01
2.04275101e-01 -1.00528383e+00 -5.70645690e-01 -1.07419300e+00
-1.35836661e-01 6.17846787e-01 1.05078030e+00 6.95501193e-02
-6.59912676e-02 5.51581800e-01 -1.88347912e+00 3.68081741e-02
1.03750789e+00 9.29812312e-01 7.23339915e-01 7.72897243e-01
-3.87819894e-02 7.86306381e-01 -8.90628457e-01 -1.61606893e-01
2.64944613e-01 -2.11729482e-01 -2.04305872e-01 4.35745984e-01
7.47892022e-01 -8.32341850e-01 2.38898277e-01 1.26809001e+00
5.07211268e-01 1.83410555e-01 8.05515528e-01 -1.01040053e+00
-2.56730974e-01 4.54853475e-01 -1.29633069e+00 3.93686324e-01
4.55074683e-02 1.46012619e-01 5.01886368e-01 -1.37033701e+00
3.10323715e-01 1.38972735e+00 5.00208139e-01 2.88645476e-01
-9.53823030e-01 -8.26208115e-01 4.49522316e-01 -2.31932670e-01
-1.87006688e+00 -5.72452471e-02 4.44851309e-01 -7.63012350e-01
5.34049451e-01 1.22738101e-01 8.47770572e-01 4.47128773e-01
-3.43465269e-01 8.87168407e-01 7.59462237e-01 -2.57907480e-01
3.92843097e-01 4.03979480e-01 -1.09653383e-01 8.66035283e-01
7.69312978e-01 2.25527156e-02 -1.15217946e-01 -1.56491041e-01
8.16011012e-01 1.56095520e-01 -2.10922435e-01 -5.09923279e-01
-9.57704365e-01 7.46713519e-01 8.44221830e-01 2.93007698e-02
-5.80923200e-01 -1.53133571e-01 6.86355531e-02 -4.17339385e-01
8.01229537e-01 2.11782351e-01 -6.42519910e-03 3.77283841e-01
-1.40316677e+00 5.24276376e-01 3.79444540e-01 8.93549621e-01
9.75406706e-01 -3.22552025e-02 -3.32638949e-01 6.58003926e-01
-1.12510361e-02 9.80745673e-01 1.00121617e-01 -4.47046340e-01
4.33536232e-01 1.11741447e+00 5.47747433e-01 -1.22946751e+00
-3.39396089e-01 -3.72071981e-01 -1.00504792e+00 4.85652447e-01
4.96011674e-01 -5.21506965e-01 -1.09077168e+00 1.02959192e+00
9.32712555e-01 2.62026638e-01 -5.80031462e-02 1.11104798e+00
6.30961955e-01 8.64866316e-01 1.38184205e-01 -8.12026039e-02
1.40520000e+00 -8.18859339e-01 -4.60983425e-01 -4.83765692e-01
5.78071713e-01 -4.91666228e-01 8.59076381e-01 3.50202769e-01
-7.60404825e-01 -6.15720153e-01 -1.08975399e+00 2.27582172e-01
-3.98525029e-01 6.47448182e-01 4.94428486e-01 4.80549544e-01
-7.18489885e-01 2.93743372e-01 -6.56490266e-01 -2.70628750e-01
8.28080177e-01 1.91713139e-01 -1.44971833e-01 -3.70676607e-01
-6.85067952e-01 6.47513509e-01 7.76548386e-01 2.01730728e-01
-1.12052035e+00 -6.74918950e-01 -8.05492043e-01 -6.96606338e-02
7.82573700e-01 -5.35713196e-01 7.74408460e-01 -6.52231991e-01
-1.00059175e+00 6.51602924e-01 -7.88732897e-03 -5.23789346e-01
6.37791395e-01 -3.57493013e-01 2.03111693e-02 2.32009545e-01
2.04622030e-01 9.36721683e-01 1.30695546e+00 -1.21613181e+00
-1.38697588e+00 -4.26332861e-01 2.30398163e-01 4.13675874e-01
-3.25290293e-01 3.39363217e-02 -4.47455972e-01 -3.53289962e-01
-7.05385506e-02 -8.89218688e-01 -1.78523675e-01 2.86283821e-01
-1.96836203e-01 -8.79560187e-02 1.25423622e+00 -4.19391721e-01
1.24675751e+00 -2.10351586e+00 1.21257722e-01 -3.33621167e-02
4.37952548e-01 5.75756669e-01 1.05017684e-01 -1.38240814e-01
3.71439308e-01 -1.40461892e-01 -5.30398965e-01 -3.09826612e-01
-4.26471829e-01 -1.30740806e-01 -1.16628319e-01 5.36969543e-01
7.33888030e-01 5.39551854e-01 -1.13289821e+00 -8.38183522e-01
6.49920821e-01 4.60696518e-01 -4.46695775e-01 3.14389825e-01
-2.78776407e-01 2.78504103e-01 -2.99831092e-01 9.92359161e-01
1.20095825e+00 -2.15674080e-02 -2.05805868e-01 -1.27760954e-02
-3.43493402e-01 -5.03298819e-01 -1.43603492e+00 9.89408791e-01
-2.36775219e-01 4.36164588e-01 1.87112600e-01 -5.91276944e-01
1.08578610e+00 -2.14422643e-01 -3.32321860e-02 7.39223137e-02
1.65504888e-01 8.90493989e-02 -4.64991443e-02 -4.14323777e-01
8.33754003e-01 1.84949469e-02 -1.71684362e-02 -7.96254948e-02
-1.25935557e-03 -3.78559142e-01 5.95561981e-01 1.83387980e-01
6.57926977e-01 7.81475753e-02 6.04311049e-01 -2.27893829e-01
4.53839749e-01 3.03039491e-01 4.60727036e-01 6.81763113e-01
-1.39487937e-01 6.37817502e-01 2.55533874e-01 -3.57101262e-01
-8.85312378e-01 -7.85942852e-01 -1.96858793e-01 1.04416633e+00
3.85056674e-01 -1.38745472e-01 -8.67577076e-01 -8.38680804e-01
7.25427046e-02 3.17552924e-01 -7.60443211e-01 7.09624067e-02
-3.98495048e-01 -1.06530166e+00 4.08216000e-01 6.58356905e-01
8.61391127e-01 -9.19382811e-01 -1.11396730e+00 6.86426610e-02
-2.40099743e-01 -1.34534657e+00 -2.30161235e-01 9.86333191e-02
-7.33314514e-01 -1.05107856e+00 -1.28730905e+00 -6.12178266e-01
9.16838408e-01 8.40813100e-01 8.48744571e-01 2.57896304e-01
-6.08694315e-01 -1.95078596e-01 -3.81979555e-01 -7.69175410e-01
-1.37727931e-01 1.42608965e-02 7.76145235e-02 2.13261604e-01
3.77425849e-01 -1.69957265e-01 -6.92440748e-01 3.33727896e-01
-7.94803619e-01 -1.81096792e-01 7.98442960e-01 7.75188148e-01
5.20822465e-01 3.88956845e-01 2.32512087e-01 -7.42878139e-01
1.16946861e-01 -4.05589163e-01 -1.26713073e+00 -4.97219898e-02
-9.42067951e-02 -3.39345187e-01 4.59443271e-01 -7.29726493e-01
-9.34377730e-01 5.51646531e-01 3.71106327e-01 -6.53034925e-01
-3.69061142e-01 9.86213982e-02 -6.43702224e-02 -1.60830393e-01
8.39221239e-01 8.21998119e-02 -3.12691092e-01 -2.38610774e-01
2.09926963e-01 7.15706229e-01 4.41142321e-01 7.81804137e-03
1.21812594e+00 6.83907688e-01 -1.64332569e-01 -1.25283420e+00
-7.93017983e-01 -8.66590917e-01 -9.71914887e-01 -3.62152219e-01
9.57216918e-01 -1.33906424e+00 -2.05208316e-01 6.28695905e-01
-9.71236467e-01 -2.87131250e-01 -4.01662827e-01 4.02649909e-01
1.34396955e-01 2.85835683e-01 -1.44647807e-01 -1.31373453e+00
-2.83356994e-01 -8.39856446e-01 1.45979059e+00 6.26994193e-01
1.43865004e-01 -7.03544259e-01 -3.32484722e-01 2.17033718e-02
1.47088990e-01 2.97498494e-01 4.53736484e-01 -1.79987147e-01
-7.12990582e-01 -4.70916569e-01 -5.48248112e-01 3.48576546e-01
4.65650186e-02 2.10159719e-01 -1.07498765e+00 -3.06261510e-01
-3.81009102e-01 -2.37397164e-01 1.11478293e+00 5.53957701e-01
8.31914485e-01 1.08432569e-01 -7.08679795e-01 5.17304957e-01
1.46445990e+00 -4.92116399e-02 3.35082620e-01 6.09266460e-02
7.96015799e-01 5.01717985e-01 1.31079519e+00 6.27621174e-01
5.63985333e-02 3.94535184e-01 7.35895276e-01 -5.77744544e-01
-1.59516916e-01 -2.18646854e-01 3.45652103e-01 7.47701600e-02
-3.28888983e-01 -1.58093989e-01 -8.27407002e-01 7.65768945e-01
-1.75509667e+00 -9.55020666e-01 -6.68907389e-02 2.10253406e+00
5.67136347e-01 -3.60780512e-03 4.61953580e-01 1.69241890e-01
9.69588220e-01 2.01577231e-01 -6.01694345e-01 4.71041173e-01
-1.71010867e-01 -1.10320896e-01 7.86865473e-01 3.59285355e-01
-1.61689079e+00 1.21574390e+00 5.29536915e+00 8.69375110e-01
-9.78206098e-01 -1.17357805e-01 3.68450552e-01 -1.33318424e-01
6.39583349e-01 -2.57404149e-01 -1.38291299e+00 3.33693415e-01
1.91621885e-01 -6.38085380e-02 1.68852732e-01 1.23981452e+00
1.24008320e-01 -8.87941539e-01 -7.26029813e-01 9.97208118e-01
1.11645453e-01 -9.19390082e-01 -2.27292231e-03 -5.21080829e-02
8.07170212e-01 -2.28583425e-01 -5.05457371e-02 3.43169123e-01
4.20494735e-01 -7.04057395e-01 6.40039027e-01 7.65419602e-02
8.19172263e-01 -8.61502171e-01 1.06764758e+00 7.08668411e-01
-1.44751644e+00 -4.35683131e-01 -1.10606706e+00 -2.35296547e-01
3.38831320e-02 8.81690502e-01 -1.14910853e+00 2.72206545e-01
9.66526091e-01 4.68300104e-01 -8.76638532e-01 1.27401233e+00
-1.62529886e-01 4.05885011e-01 -4.52935785e-01 -2.79327095e-01
3.29065591e-01 -1.33375064e-01 6.59803212e-01 1.29915011e+00
5.01816511e-01 1.43601656e-01 4.63319927e-01 9.49137092e-01
-5.63746914e-02 -6.01733662e-02 -7.49854147e-01 2.48301491e-01
5.03292799e-01 1.62581205e+00 -1.13244653e+00 -6.85931087e-01
-7.26942644e-02 9.15859520e-01 2.84341067e-01 9.06629581e-03
-1.02029145e+00 -3.64385694e-01 3.00778180e-01 4.71848220e-01
7.54271567e-01 -2.57284865e-02 2.80885696e-01 -9.56539929e-01
4.30755876e-02 -6.79625571e-01 1.57918543e-01 -5.84662080e-01
-1.03274381e+00 5.97997189e-01 4.19516623e-01 -1.35064554e+00
-8.36311132e-02 -5.72762012e-01 -4.64273155e-01 7.57109821e-01
-1.43616176e+00 -1.15148604e+00 -1.04488313e+00 4.43474680e-01
6.80781960e-01 4.86694771e-04 5.92620373e-01 2.39372566e-01
-6.26920998e-01 9.28083435e-02 -2.05030084e-01 2.02251822e-01
5.25424361e-01 -1.35905397e+00 2.51074284e-01 1.33128583e+00
1.65337499e-03 1.14477873e-01 4.05113161e-01 -8.76975596e-01
-7.27164805e-01 -1.70670605e+00 1.97366327e-01 -4.59032059e-01
3.18460494e-01 -6.74405754e-01 -8.75949562e-01 1.80189177e-01
-1.81126207e-01 3.83431971e-01 6.27947226e-02 -3.59220594e-01
1.47661313e-01 2.15450376e-02 -1.19403934e+00 3.69156867e-01
8.05615485e-01 4.92742658e-02 -3.97786766e-01 3.96765649e-01
7.07546413e-01 -3.72493088e-01 -3.14541817e-01 4.51634198e-01
2.47121766e-01 -9.85102832e-01 9.69119191e-01 5.35357408e-02
2.88317472e-01 -8.11709881e-01 9.81917977e-03 -1.36967087e+00
-3.35960329e-01 -1.26476392e-01 -2.37548277e-01 1.20367801e+00
1.92317382e-01 -3.20621818e-01 7.35676885e-01 5.81792369e-02
2.16749743e-01 -3.48310471e-01 -5.87266386e-01 -6.95397019e-01
-3.00951600e-01 2.59641379e-01 5.46637297e-01 7.26280510e-01
-4.20042098e-01 3.27297360e-01 -2.93876767e-01 6.44212306e-01
8.51070523e-01 7.91188236e-03 1.08452582e+00 -1.41655862e+00
1.19981996e-03 3.17896763e-03 -4.11946803e-01 -1.10684228e+00
-4.64421362e-01 -4.67652112e-01 4.16684806e-01 -1.42004001e+00
2.86703765e-01 -4.36942190e-01 2.77421266e-01 2.06133261e-01
-7.94576824e-01 4.66972828e-01 3.67895901e-01 1.45273566e-01
-5.81994772e-01 5.04317760e-01 1.08296883e+00 -2.73897171e-01
-3.71265888e-01 1.60929173e-01 -3.27410519e-01 1.10554421e+00
6.55973256e-01 -6.23185337e-01 -3.20971787e-01 -2.15559825e-01
-2.53477633e-01 -3.72137100e-01 6.47179425e-01 -1.52187860e+00
1.70248747e-01 3.07085775e-02 8.00956547e-01 -1.21466017e+00
4.06107664e-01 -7.93994367e-01 -5.33559084e-01 5.56626260e-01
3.29069257e-01 -1.20884717e-01 3.70086163e-01 6.51932240e-01
-5.94249256e-02 -3.10829669e-01 1.17770112e+00 -2.48094887e-01
-8.55430126e-01 2.20564023e-01 -3.55109721e-01 -5.79315051e-02
1.39865959e+00 -2.33554050e-01 -1.88238874e-01 4.09939624e-02
-5.07851839e-01 2.39700064e-01 4.43759799e-01 1.11143857e-01
6.44982338e-01 -8.46619487e-01 -8.00372601e-01 2.67594188e-01
1.01611100e-01 8.61529231e-01 1.80330992e-01 9.30459023e-01
-6.31035864e-01 4.01923656e-02 5.68879060e-02 -1.07354045e+00
-1.46969771e+00 6.79594815e-01 2.77716726e-01 -1.91226527e-01
-6.70278013e-01 9.79420125e-01 6.07561409e-01 -9.29466486e-02
2.65286654e-01 -6.75879180e-01 -3.93433988e-01 3.74412835e-01
9.88765419e-01 4.45789307e-01 -1.83197573e-01 -4.50438082e-01
-2.98242837e-01 6.65774047e-01 -5.95144741e-02 2.11031124e-01
1.07501996e+00 -1.04898192e-01 3.40236872e-02 4.13861275e-01
4.08045709e-01 -7.84470960e-02 -1.52292311e+00 -2.35396951e-01
-2.27063224e-01 -7.48408854e-01 1.66095212e-01 -4.91330147e-01
-9.75779712e-01 9.94784951e-01 8.56042504e-01 1.26874089e-01
1.12651074e+00 -6.47002906e-02 2.94323355e-01 4.27485138e-01
4.91003096e-01 -9.30412591e-01 1.47516757e-01 2.41791636e-01
7.26522505e-01 -1.43566072e+00 2.77209401e-01 -6.32937968e-01
-7.15444684e-01 7.94397354e-01 9.63671923e-01 -4.30764198e-01
3.86399627e-01 6.72472239e-01 -1.62171170e-01 -1.07253820e-01
-3.25177372e-01 -6.66492760e-01 -3.56816649e-02 9.16998029e-01
-6.18617833e-02 1.42998800e-01 2.26227179e-01 3.31985027e-01
1.55623287e-01 7.53294304e-02 6.00941479e-01 8.37747335e-01
-9.80774164e-01 -2.11358622e-01 -8.11479390e-01 4.91773605e-01
-5.46741858e-02 -2.33588621e-01 -5.20725489e-01 1.10118210e+00
4.22926992e-01 8.56500626e-01 3.43382120e-01 -2.51471221e-01
3.56449515e-01 -3.48426700e-01 3.43549907e-01 -7.59167314e-01
-3.28643501e-01 1.98513582e-01 -1.91830829e-01 -2.42918372e-01
-4.53605473e-01 -5.22118092e-01 -1.05548990e+00 -1.13949269e-01
-9.98451531e-01 -1.11404963e-01 3.79940629e-01 5.35794973e-01
9.15634856e-02 6.15256608e-01 4.65139240e-01 -1.41436100e+00
-4.65740025e-01 -1.21636891e+00 -6.35393858e-01 8.60307086e-03
3.30355197e-01 -1.03341532e+00 -4.85875756e-01 9.89268050e-02]
|
[8.685348510742188, -0.7300403118133545]
|
b1f0c602-d024-4644-8975-86f20b599b5a
|
the-role-of-user-profile-for-fake-news
|
1904.13355
| null |
http://arxiv.org/abs/1904.13355v1
|
http://arxiv.org/pdf/1904.13355v1.pdf
|
The Role of User Profile for Fake News Detection
|
Consuming news from social media is becoming increasingly popular. Social
media appeals to users due to its fast dissemination of information, low cost,
and easy access. However, social media also enables the widespread of fake
news. Because of the detrimental societal effects of fake news, detecting fake
news has attracted increasing attention. However, the detection performance
only using news contents is generally not satisfactory as fake news is written
to mimic true news. Thus, there is a need for an in-depth understanding on the
relationship between user profiles on social media and fake news. In this
paper, we study the challenging problem of understanding and exploiting user
profiles on social media for fake news detection. In an attempt to understand
connections between user profiles and fake news, first, we measure users'
sharing behaviors on social media and group representative users who are more
likely to share fake and real news; then, we perform a comparative analysis of
explicit and implicit profile features between these user groups, which reveals
their potential to help differentiate fake news from real news. To exploit user
profile features, we demonstrate the usefulness of these user profile features
in a fake news classification task. We further validate the effectiveness of
these features through feature importance analysis. The findings of this work
lay the foundation for deeper exploration of user profile features of social
media and enhance the capabilities for fake news detection.
|
['Huan Liu', 'Reza Zafarani', 'Kai Shu', 'Suhang Wang', 'Xinyi Zhou']
|
2019-04-30
| null | null | null | null |
['news-classification']
|
['natural-language-processing']
|
[-2.80128896e-01 9.69915763e-02 -6.72786713e-01 -1.58197403e-01
-2.55473167e-01 -6.30002260e-01 8.96925092e-01 6.40838444e-01
-1.29016832e-01 6.03952944e-01 4.29401875e-01 -1.49249166e-01
2.28722081e-01 -9.25589442e-01 -3.38117421e-01 -1.69001937e-01
3.73927169e-02 -1.14480898e-01 2.64364868e-01 -5.80034077e-01
6.24154210e-01 2.98975527e-01 -1.40575731e+00 6.97517574e-01
8.91682565e-01 8.35743070e-01 -1.95149690e-01 1.21162012e-01
-5.49834594e-02 8.48953009e-01 -9.03324425e-01 -6.82661593e-01
4.73546013e-02 -7.12368309e-01 -5.26335418e-01 1.31918550e-01
2.48365596e-01 -4.64396119e-01 -5.30600131e-01 1.26140344e+00
1.61475480e-01 -2.00048342e-01 4.06132132e-01 -1.47704864e+00
-8.35546076e-01 6.61793411e-01 -3.33758920e-01 4.36367333e-01
6.11879230e-01 -2.41276413e-01 1.04053593e+00 -5.27499735e-01
7.48235285e-01 1.08433664e+00 8.14886332e-01 2.65058875e-01
-7.33915269e-01 -6.89857662e-01 -2.06918895e-01 -2.77697388e-02
-1.01203907e+00 -4.42742288e-01 7.95332789e-01 -6.66553140e-01
1.96407944e-01 5.32584369e-01 9.89367127e-01 1.36168444e+00
2.48228267e-01 1.00624788e+00 1.42818892e+00 -2.25333944e-01
5.36803007e-02 8.87799442e-01 3.11660141e-01 5.44414997e-01
6.18013084e-01 9.73896533e-02 -6.45749986e-01 -8.28780830e-01
4.37176466e-01 2.64399648e-01 -4.36386377e-01 7.08921403e-02
-1.02370059e+00 1.31221902e+00 5.32966912e-01 5.66061318e-01
-2.84930408e-01 -2.01260969e-01 4.82996911e-01 7.91118026e-01
1.22965169e+00 1.03618300e+00 -1.52602777e-01 -2.28230223e-01
-7.13361144e-01 2.79291004e-01 1.05580533e+00 4.91357476e-01
5.76726258e-01 -1.81742445e-01 -3.23044397e-02 9.06560242e-01
1.36179626e-01 3.95843327e-01 8.07137311e-01 -4.02359366e-01
8.90677720e-02 6.94307148e-01 5.13113499e-01 -2.17355919e+00
-1.19926780e-01 -4.89112824e-01 -3.37306976e-01 -5.02709210e-01
4.26026523e-01 2.62106229e-02 -4.72934805e-02 1.31882381e+00
1.84449479e-01 1.37453571e-01 -2.86764652e-01 9.18340027e-01
6.73489869e-01 4.52905595e-01 -3.21957558e-01 -3.22394192e-01
1.38530684e+00 -7.27777660e-01 -9.01314497e-01 -9.43501666e-02
1.04235041e+00 -9.16600645e-01 9.47920024e-01 2.26722434e-02
-5.94547093e-01 -2.08286736e-02 -7.52377927e-01 2.47422010e-01
-6.78613663e-01 -1.63492728e-02 1.00240266e+00 9.51262534e-01
-3.91480476e-01 8.60980451e-01 -4.52246219e-01 -3.65399569e-01
6.80861771e-01 -1.19637378e-01 -2.53040254e-01 9.24966708e-02
-1.52643478e+00 8.27527165e-01 1.43078044e-01 -2.98573643e-01
-1.72015160e-01 -3.13910395e-01 -5.72597265e-01 -1.49863690e-01
4.03167754e-01 -1.66821003e-01 1.01039886e+00 -1.51355493e+00
-1.22475719e+00 7.91425467e-01 1.74475953e-01 -4.23336715e-01
8.53626370e-01 -3.31476256e-02 -7.56299973e-01 2.43009329e-01
3.51689816e-01 -1.51337579e-01 1.12728691e+00 -1.09771955e+00
-4.19370949e-01 -2.43234575e-01 -1.58313345e-02 -2.08652750e-01
-6.40615106e-01 1.66217715e-01 1.77191556e-01 -9.59551096e-01
3.60926747e-01 -8.96371245e-01 3.50350171e-01 -2.42410377e-01
-5.82364619e-01 8.90751779e-02 1.12438810e+00 -5.99380195e-01
1.19960105e+00 -2.21689963e+00 -5.40322661e-01 2.91002661e-01
6.36431873e-01 2.73257405e-01 3.15781862e-01 6.84832275e-01
3.69025707e-01 4.38303322e-01 3.00616175e-01 -9.42988496e-04
-2.74963766e-01 -1.76492304e-01 -6.02467775e-01 8.34164798e-01
-1.68038055e-01 1.02978384e+00 -1.34529114e+00 -5.71918162e-03
-2.49459356e-01 1.48317546e-01 -5.38532138e-01 -1.78426579e-01
1.04602829e-01 3.58173728e-01 -8.86366427e-01 5.55846572e-01
4.69118804e-01 -6.29414141e-01 1.41096041e-01 -3.97045538e-02
-1.57914564e-01 7.82348096e-01 -4.15775359e-01 3.92271221e-01
-1.21998303e-01 1.08645236e+00 -2.64688164e-01 -8.36808383e-01
7.92431235e-01 1.73547044e-01 2.10601583e-01 -4.63016719e-01
6.41962111e-01 4.13624436e-01 -1.05617017e-01 -5.33951640e-01
8.25090289e-01 -1.89733982e-01 -2.17814758e-01 8.88918579e-01
-4.93113369e-01 -8.96203890e-02 -2.55983680e-01 4.77690339e-01
7.76825964e-01 -5.73216617e-01 5.57985961e-01 -1.60269544e-01
1.12295516e-01 1.46254063e-01 -7.64533356e-02 9.20014739e-01
-4.44149315e-01 2.02289462e-01 6.98826253e-01 -3.25229347e-01
-6.61869526e-01 -3.52050245e-01 -2.65289992e-02 8.57048333e-01
4.83271420e-01 -5.20138443e-01 -5.16328752e-01 -8.90816271e-01
5.11448801e-01 5.35925210e-01 -7.65295208e-01 -4.09361392e-01
-8.99043828e-02 -9.14725721e-01 6.27118170e-01 -2.83740431e-01
6.82499170e-01 -7.36970663e-01 -1.64696038e-01 1.11863732e-01
-5.03387988e-01 -1.13760436e+00 -4.44539100e-01 -7.99218476e-01
-6.65964484e-01 -1.15992153e+00 -3.81522894e-01 -3.56012493e-01
5.81006110e-01 1.09265828e+00 6.29512906e-01 5.45674920e-01
1.35948449e-01 2.25819632e-01 -7.58016109e-01 -3.15596372e-01
-9.18907344e-01 -1.68243460e-02 2.82521546e-01 3.84113342e-01
3.30840200e-01 -2.62156546e-01 -2.62768626e-01 8.16244125e-01
-9.85132515e-01 5.26242182e-02 2.03272387e-01 7.22625315e-01
-3.26241821e-01 1.31611735e-01 8.33324969e-01 -1.36635900e+00
1.04485309e+00 -1.05959058e+00 -3.29920143e-01 -2.19911277e-01
-5.32918870e-01 -5.48915386e-01 4.91698325e-01 -7.40404725e-01
-6.13478124e-01 -6.83350682e-01 2.56951004e-01 2.06763998e-01
3.34737718e-01 7.75351405e-01 4.57790375e-01 -3.17581415e-01
8.45267713e-01 1.06667623e-01 2.67442286e-01 -3.09467673e-01
5.73488921e-02 1.14978027e+00 -2.21827313e-01 1.15932031e-02
7.71289468e-01 8.19441676e-01 -6.93623483e-01 -1.21430922e+00
-1.30365062e+00 -6.39815927e-01 1.79103404e-01 -2.76340753e-01
2.62996823e-01 -7.35986292e-01 -6.36406541e-01 7.21091509e-01
-1.07012165e+00 1.63436502e-01 2.07258105e-01 5.67038774e-01
-1.16449460e-01 7.69402444e-01 -9.50822949e-01 -7.46895194e-01
2.39392310e-01 -8.05164993e-01 4.19747084e-01 -3.35972756e-01
-5.17003953e-01 -1.07139015e+00 -1.71221122e-01 8.66261065e-01
6.27661228e-01 4.00803596e-01 3.36420238e-01 -1.06019664e+00
-4.20079052e-01 -8.44500721e-01 -4.90524888e-01 2.13819593e-01
4.61324960e-01 -1.80880055e-01 -8.24643075e-01 -3.32374364e-01
2.47472435e-01 -2.85452336e-01 5.89969993e-01 2.34397687e-02
6.55536711e-01 -8.65235031e-01 -4.54140067e-01 -4.43768725e-02
9.18901324e-01 -3.77975911e-01 3.84568751e-01 4.60232645e-01
5.03989339e-01 8.52983892e-01 5.73148489e-01 6.82892740e-01
2.12359324e-01 6.68843269e-01 3.43336016e-01 3.22905421e-01
3.72782767e-01 -5.11255503e-01 4.23353136e-01 7.83249736e-01
1.32468760e-01 -3.92492563e-01 -5.89477003e-01 3.10147643e-01
-1.66497886e+00 -1.25856435e+00 -5.14198303e-01 2.12546229e+00
4.18307066e-01 1.99747190e-01 3.95008087e-01 2.65950948e-01
1.04828906e+00 4.24907118e-01 6.32080734e-02 -1.02404073e-01
-1.90693542e-01 -5.69219053e-01 6.67754352e-01 3.49727869e-01
-9.64880884e-01 1.00934315e+00 5.78946400e+00 7.39633143e-01
-1.39108264e+00 4.06737477e-01 5.66460967e-01 1.87487692e-01
-3.48990113e-01 -1.15988277e-01 -5.17183185e-01 8.59351218e-01
6.83166981e-01 -1.49212703e-01 3.88855010e-01 8.84957790e-01
6.25588655e-01 -1.59734160e-01 -5.66749871e-01 7.55507648e-01
1.99387729e-01 -1.64495707e+00 -4.12302129e-02 3.79479706e-01
9.01239634e-01 3.43052484e-02 9.18520018e-02 1.54159233e-01
-9.36118811e-02 -5.64425290e-01 6.64418876e-01 -3.68413068e-02
2.62446880e-01 -4.84839588e-01 9.29888010e-01 6.01575017e-01
-2.16633081e-01 -7.55589642e-03 -2.65579730e-01 -7.34605908e-01
2.65097171e-01 9.64377284e-01 -1.03806674e+00 5.25231957e-02
2.62271106e-01 9.45701838e-01 -5.02933025e-01 7.33439386e-01
-6.41598627e-02 7.16354072e-01 -1.51136220e-01 -4.83822703e-01
2.73287386e-01 -9.94150564e-02 6.86113596e-01 1.03199327e+00
1.47042006e-01 8.45915638e-03 -3.22737396e-02 7.46023595e-01
-2.71202683e-01 3.08397800e-01 -8.93036962e-01 -9.45553362e-01
4.96239513e-01 8.86368394e-01 -9.60864127e-01 -4.47235316e-01
-4.19361591e-01 1.12445366e+00 2.78579831e-01 -4.07184064e-02
-7.24562347e-01 -1.23883285e-01 5.19174278e-01 7.35153139e-01
-1.92309991e-01 -6.37412965e-02 -5.72414026e-02 -1.61641181e+00
-2.12264374e-01 -9.75293517e-01 3.06665413e-02 -4.80437636e-01
-1.69569683e+00 3.77605826e-01 -2.43016541e-01 -1.32281792e+00
1.37709141e-01 -2.45956153e-01 -4.14818317e-01 3.24073702e-01
-1.33460271e+00 -7.70728469e-01 -3.37283820e-01 2.64801353e-01
8.22128430e-02 -4.81767282e-02 3.91373962e-01 2.05540761e-01
-1.65714711e-01 5.42730331e-01 2.63632327e-01 1.94661751e-01
6.06852710e-01 -5.42930722e-01 4.42199826e-01 2.58294284e-01
-8.42876174e-03 7.13748336e-01 9.55028653e-01 -1.03385365e+00
-1.01997089e+00 -7.95553327e-01 1.21800387e+00 -7.06568539e-01
1.18192375e+00 -2.02141881e-01 -8.46503615e-01 6.00625396e-01
-4.33083475e-01 -7.32512251e-02 9.41729963e-01 -1.02717951e-01
-6.58046007e-01 7.17811823e-01 -1.43913388e+00 6.37745023e-01
8.33717346e-01 -7.35721290e-01 -5.55647850e-01 6.73794806e-01
4.31855291e-01 -4.59932871e-02 -5.72811186e-01 -2.34518230e-01
7.58610487e-01 -1.28867471e+00 6.93431735e-01 -6.82967007e-01
6.23990357e-01 8.58628601e-02 3.18550691e-02 -1.39324594e+00
-1.44676760e-01 -6.81880951e-01 6.11298271e-02 7.98933506e-01
3.25454086e-01 -1.23863757e+00 9.20371354e-01 3.64681095e-01
1.70238197e-01 -4.66468900e-01 -4.17562068e-01 -7.82051027e-01
-3.29761982e-01 -2.64464438e-01 2.52168357e-01 1.98256338e+00
4.23819631e-01 1.34157836e-01 -7.44583189e-01 9.57909450e-02
1.88775018e-01 1.37130961e-01 8.52780640e-01 -1.34837937e+00
-4.44045991e-01 -4.40902233e-01 -5.92428505e-01 -1.04228568e+00
6.45481190e-03 -6.88595355e-01 -6.05153382e-01 -8.53545904e-01
1.09439455e-01 -4.63255852e-01 2.68882364e-01 -1.83245316e-02
-7.70946294e-02 7.67450869e-01 1.41382784e-01 7.64027894e-01
-3.87308508e-01 4.23119366e-01 1.53011835e+00 2.14244798e-02
-3.32881898e-01 3.73453796e-01 -8.58699381e-01 7.57749498e-01
8.93175244e-01 -7.23515570e-01 -1.56077340e-01 2.16196790e-01
7.43683934e-01 4.09028158e-02 4.42051381e-01 -3.98946077e-01
-2.91664511e-01 -2.30918288e-01 -1.61802053e-01 -8.17902908e-02
3.61855030e-01 -4.59239513e-01 -1.66236058e-01 5.97070336e-01
-2.12109566e-01 -1.14266679e-01 -2.49039114e-01 9.91814435e-01
-3.90335590e-01 -1.52932942e-01 7.62370169e-01 -2.68400878e-01
-1.44624561e-01 -2.73334365e-02 -8.75620842e-01 6.97109774e-02
9.03492749e-01 -4.18936491e-01 -5.95664144e-01 -1.06821346e+00
-5.90649247e-01 -3.85847926e-01 8.28032911e-01 5.20220041e-01
4.13956493e-01 -1.09991288e+00 -6.00590646e-01 1.53837144e-01
3.68265152e-01 -1.05385387e+00 -1.58195179e-02 1.09187353e+00
-4.68505532e-01 6.04818612e-02 -1.04148023e-01 -1.01751886e-01
-1.10344446e+00 4.41215903e-01 -1.46698365e-02 2.01023772e-01
-3.12299758e-01 6.08956993e-01 -2.61586338e-01 -2.62900889e-01
-4.60376680e-01 -1.55080532e-04 -1.94943607e-01 3.88531208e-01
6.56991184e-01 6.37889564e-01 -7.54826292e-02 -1.05364048e+00
-2.80362889e-02 -3.05817187e-01 -2.44567841e-01 1.35187149e-01
1.07003820e+00 -2.23795399e-01 -4.39316571e-01 3.52393925e-01
1.35057938e+00 7.41017938e-01 -4.78776723e-01 -1.26477450e-01
1.61144540e-01 -1.20248103e+00 1.58906102e-01 -4.60024685e-01
-7.72997200e-01 7.54386261e-02 -2.35370845e-01 1.30462718e+00
2.50425249e-01 8.85479525e-02 1.02957630e+00 2.84916401e-01
5.15489817e-01 -8.34463179e-01 4.02461946e-01 3.86821091e-01
6.28468692e-01 -1.60711908e+00 1.96480989e-01 -1.00209439e+00
-6.37547791e-01 8.84849608e-01 9.47755352e-02 -2.68210113e-01
7.70953476e-01 -4.96517420e-01 1.38048217e-01 -5.41990399e-01
-1.71693653e-01 2.58572191e-01 2.84937561e-01 7.21688569e-02
4.77223963e-01 2.77644277e-01 -7.79765308e-01 4.00828272e-01
-5.39727986e-01 -2.14744568e-01 1.11521077e+00 8.81912589e-01
-6.04521513e-01 -7.68876374e-01 -3.78960907e-01 9.32122409e-01
-8.14175069e-01 2.08688434e-02 -9.00659800e-01 7.93141305e-01
-2.44939983e-01 1.14936209e+00 -3.03695172e-01 -5.12623727e-01
-2.29741856e-01 -2.34929994e-01 8.32576305e-02 -7.32603490e-01
-6.85509980e-01 -4.20876533e-01 5.45390487e-01 -5.00263095e-01
-3.89153004e-01 -5.37996650e-01 -4.78990674e-01 -7.74962127e-01
-1.03473186e+00 4.79047865e-01 8.49694967e-01 1.13020623e+00
4.79177028e-01 -2.76803344e-01 8.59444976e-01 -6.00963235e-01
-6.54762566e-01 -8.87022138e-01 -5.67563772e-01 7.79285669e-01
3.65992963e-01 -7.76999414e-01 -8.95080447e-01 -4.48891014e-01]
|
[8.133410453796387, 10.265463829040527]
|
82821d53-f9b2-4124-b83a-195caadd8cf5
|
deep-active-inference-for-pixel-based
|
2109.04155
| null |
https://arxiv.org/abs/2109.04155v1
|
https://arxiv.org/pdf/2109.04155v1.pdf
|
Deep Active Inference for Pixel-Based Discrete Control: Evaluation on the Car Racing Problem
|
Despite the potential of active inference for visual-based control, learning the model and the preferences (priors) while interacting with the environment is challenging. Here, we study the performance of a deep active inference (dAIF) agent on OpenAI's car racing benchmark, where there is no access to the car's state. The agent learns to encode the world's state from high-dimensional input through unsupervised representation learning. State inference and control are learned end-to-end by optimizing the expected free energy. Results show that our model achieves comparable performance to deep Q-learning. However, vanilla dAIF does not reach state-of-the-art performance compared to other world model approaches. Hence, we discuss the current model implementation's limitations and potential architectures to overcome them.
|
['Pablo Lanillos', 'Niels van Hoeffelen']
|
2021-09-09
| null | null | null | null |
['carracing-v0']
|
['playing-games']
|
[-9.61500108e-02 6.20849133e-01 -6.13483250e-01 -3.31742704e-01
-7.95705140e-01 -4.32096809e-01 9.51532066e-01 6.50648624e-02
-7.03811646e-01 8.15779805e-01 3.97353381e-01 -2.62043923e-01
-8.96880217e-03 -6.03579938e-01 -1.03674817e+00 -6.44353509e-01
-1.80483952e-01 9.23233986e-01 2.46028289e-01 -1.50176629e-01
3.40868801e-01 3.33029002e-01 -1.48512161e+00 -9.37216356e-02
8.07081938e-01 7.19103277e-01 2.84508020e-01 9.57027853e-01
1.77858576e-01 1.48547685e+00 -2.36795202e-01 5.04207425e-02
1.67198896e-01 -1.06128059e-01 -8.76792490e-01 -2.02286065e-01
3.81159693e-01 -5.77356339e-01 -5.25961101e-01 8.39277029e-01
2.88519412e-01 6.80125833e-01 6.18247092e-01 -1.46493495e+00
-5.80875695e-01 1.56847790e-01 5.58362491e-02 5.97953796e-02
1.31676748e-01 1.03708768e+00 9.35285330e-01 -4.52383697e-01
1.09402442e+00 1.56988764e+00 -9.04226303e-02 7.89694130e-01
-1.59774458e+00 -2.84028143e-01 7.20393062e-01 4.24255103e-01
-8.48300278e-01 -6.84276283e-01 6.58983409e-01 -4.04258668e-01
1.43413675e+00 -1.40450209e-01 9.92471516e-01 1.48671806e+00
2.90122747e-01 1.26444018e+00 1.04432869e+00 -2.23495841e-01
8.74939740e-01 -2.79804707e-01 -6.50797635e-02 9.68076646e-01
-1.75957635e-01 8.11503887e-01 -6.41440153e-01 -2.97122031e-01
5.82858980e-01 -3.30485374e-01 2.20465168e-01 -1.33233011e+00
-1.20934474e+00 1.04838467e+00 5.98614156e-01 -5.79493821e-01
-4.70276266e-01 8.81898284e-01 1.70434758e-01 -4.90557477e-02
3.55210632e-01 7.85302103e-01 -3.63356888e-01 -5.23278534e-01
-3.97892684e-01 6.51916981e-01 8.24584782e-01 8.07573497e-01
7.42087901e-01 1.51368007e-01 -3.28909516e-01 3.91738236e-01
9.10741031e-01 4.42969024e-01 -2.76361495e-01 -1.73125994e+00
3.03966910e-01 3.66803169e-01 6.29924715e-01 -5.03659964e-01
-1.78086251e-01 -3.71754393e-02 -8.31186175e-02 1.13340652e+00
3.47433180e-01 -5.26232719e-01 -1.29979432e+00 1.76207948e+00
3.16634357e-01 3.80919844e-01 2.20452502e-01 1.02035630e+00
3.23592871e-01 8.93281341e-01 4.68664289e-01 6.29447848e-02
5.78844666e-01 -1.28461683e+00 -7.99160838e-01 -1.00605726e+00
5.26803657e-02 -1.21196151e-01 8.10772359e-01 2.75024652e-01
-1.32504761e+00 -4.58326936e-01 -1.10440719e+00 -1.45693615e-01
-3.19853932e-01 -2.40696207e-01 9.08842862e-01 1.24491507e-03
-9.75507021e-01 3.83139074e-01 -1.65170813e+00 -2.42169857e-01
7.38119960e-01 3.79193395e-01 -2.22363576e-01 1.53379247e-01
-1.09656310e+00 1.66539311e+00 4.40138310e-01 -1.33634433e-02
-1.89857268e+00 -6.24659717e-01 -1.06598938e+00 -2.49175988e-02
8.06360364e-01 -8.26007485e-01 1.61307120e+00 -7.50611484e-01
-2.19433641e+00 3.28436047e-01 -2.19902815e-03 -8.51562858e-01
5.94632030e-01 -3.79979253e-01 6.75813574e-03 -1.44880451e-02
-3.02092463e-01 1.12926590e+00 6.08854175e-01 -1.38994825e+00
-5.31630635e-01 -1.37132943e-01 5.97838521e-01 4.90931511e-01
4.31265622e-01 -3.56095612e-01 -5.18310785e-01 -5.09486627e-03
-4.40329999e-01 -1.22796214e+00 -7.90354908e-01 5.48223913e-01
-1.02901332e-01 -3.37385535e-01 9.39651787e-01 -2.22033143e-01
7.24516273e-01 -2.04038167e+00 4.66507345e-01 4.63691466e-02
8.45004693e-02 2.12896571e-01 -3.38328391e-01 5.66277266e-01
3.71048778e-01 -1.88780516e-01 -2.74809390e-01 -6.47122979e-01
5.26896894e-01 6.90302432e-01 -3.18348497e-01 5.34777761e-01
5.48479617e-01 1.21064901e+00 -1.18273735e+00 -4.86484975e-01
6.24571621e-01 4.79182184e-01 -8.33633304e-01 6.32914186e-01
-9.68206108e-01 6.48870111e-01 -5.47990978e-01 1.80927083e-01
1.45981282e-01 -5.03711589e-02 4.43886042e-01 2.33031154e-01
-2.36163899e-01 4.71108884e-01 -1.01292610e+00 2.12578869e+00
-4.84087437e-01 6.31999493e-01 1.59543231e-01 -8.19518268e-01
6.49790585e-01 1.05483830e-01 1.74424320e-01 -1.09457028e+00
-1.70742765e-01 -3.90511304e-01 1.43055618e-01 -6.49963737e-01
2.64195442e-01 3.67889881e-01 -1.13587990e-01 2.29800165e-01
1.10342696e-01 -4.37731653e-01 1.87720343e-01 3.62159252e-01
1.03075719e+00 8.33697081e-01 7.43196607e-02 -1.99349359e-01
-1.38952523e-01 3.38740230e-01 6.38143837e-01 9.44598734e-01
-5.07795572e-01 1.02327861e-01 7.36594677e-01 -3.46273869e-01
-9.33625400e-01 -1.53230202e+00 3.16822916e-01 1.30043280e+00
1.90948784e-01 -3.91453981e-01 -4.95593250e-01 -6.73613131e-01
1.34571046e-01 1.28716803e+00 -8.01426053e-01 -4.02620375e-01
-5.86934030e-01 -4.21978354e-01 3.78470160e-02 5.97700894e-01
3.44492167e-01 -1.17750537e+00 -1.19360077e+00 2.31821731e-01
8.33125785e-02 -6.98276281e-01 -1.24534652e-01 3.30005407e-01
-6.14515901e-01 -9.15199995e-01 -1.24879688e-01 -2.67744839e-01
5.01710176e-01 -5.30193865e-01 1.43604016e+00 -3.34656358e-01
-1.30370244e-01 6.50103092e-01 7.02854618e-02 -6.07491076e-01
-4.61200565e-01 -7.86523074e-02 -2.45225608e-01 -4.45815951e-01
-9.14048404e-02 -1.53707579e-01 -9.99212563e-01 -1.08368494e-01
-3.78018439e-01 1.74509093e-01 4.43467766e-01 7.77908504e-01
5.40964365e-01 -3.78378183e-01 2.64973283e-01 -8.02405596e-01
4.76700574e-01 -4.55335498e-01 -1.04923034e+00 4.57481220e-02
-6.30712748e-01 4.74907637e-01 1.46177977e-01 -4.16460484e-01
-1.56604195e+00 5.06319463e-01 1.69073656e-01 -3.54449362e-01
-2.13190466e-01 3.30062836e-01 -1.22439899e-02 3.53048414e-01
5.43248057e-01 -2.85609394e-01 3.32105279e-01 -2.26388991e-01
6.93842709e-01 6.33698106e-02 5.61886907e-01 -8.43043029e-01
4.62679535e-01 4.49837953e-01 -1.66951016e-01 -1.32975802e-01
-8.35979998e-01 -1.31988600e-01 -4.29019332e-01 -1.71721727e-01
1.15921915e+00 -1.04443240e+00 -1.26375961e+00 1.58701241e-01
-9.54509974e-01 -1.21983504e+00 -4.62495863e-01 6.07664526e-01
-1.12796795e+00 -2.38015205e-01 -4.35638607e-01 -1.22202671e+00
8.66626650e-02 -1.39252996e+00 9.37357128e-01 2.04066873e-01
-3.88228685e-01 -9.95606422e-01 6.40957355e-01 4.79005694e-01
5.82800627e-01 2.79231757e-01 8.11327696e-01 -1.97791353e-01
-1.03122127e+00 2.00108349e-01 2.14047387e-01 -3.42266411e-02
-4.26445156e-01 2.99824715e-01 -1.01406181e+00 -4.00352061e-01
-6.46404386e-01 -8.30420554e-01 9.57306445e-01 3.57103765e-01
1.11862445e+00 -3.64482582e-01 -2.81658947e-01 2.98960298e-01
1.28575838e+00 4.25044805e-01 7.88837075e-01 3.22502732e-01
4.33060259e-01 4.41137671e-01 8.05512667e-01 3.55174094e-01
4.88304198e-01 6.97048247e-01 1.19991672e+00 -1.06721677e-01
1.26988769e-01 -5.05165637e-01 6.41102970e-01 8.18060264e-02
2.92349979e-02 -2.29533464e-01 -1.04698634e+00 4.90340054e-01
-2.32178426e+00 -1.09940851e+00 4.31441873e-01 2.04846764e+00
8.17373276e-01 5.35990655e-01 -2.78394092e-02 -6.17951393e-01
1.75702855e-01 4.21167195e-01 -1.28375304e+00 -6.39812052e-01
3.86098564e-01 1.46967158e-01 4.90301013e-01 8.15451026e-01
-1.19011819e+00 1.05766988e+00 7.10690832e+00 4.85666752e-01
-8.82817090e-01 1.26130953e-01 5.16973734e-01 -4.52497840e-01
-1.36799976e-01 4.23410386e-01 -5.20789027e-01 2.00672060e-01
1.07078469e+00 1.97423667e-01 7.19722986e-01 9.23045695e-01
2.89906591e-01 -5.17817795e-01 -1.36054230e+00 5.14561117e-01
-2.42771775e-01 -1.55762148e+00 -4.59143460e-01 9.65062380e-02
7.64544249e-01 5.19059718e-01 1.08455658e-01 8.26972067e-01
1.33042026e+00 -1.09122539e+00 1.14588344e+00 8.65677178e-01
4.01325703e-01 -8.61982286e-01 1.58180848e-01 6.53966784e-01
-6.66288316e-01 -3.27774435e-01 -1.09836966e-01 -2.97491252e-01
2.33727202e-01 -1.63638294e-01 -8.51127207e-01 -1.39676526e-01
4.57393527e-01 6.34814143e-01 -4.47325021e-01 8.03181648e-01
-2.55058616e-01 6.37226164e-01 -1.39397904e-01 -2.21997485e-01
5.64461529e-01 -1.54065207e-01 5.32937825e-01 8.76437902e-01
-3.29443246e-01 -9.69115049e-02 6.00307524e-01 1.29903877e+00
1.51636094e-01 -7.69793808e-01 -6.29411876e-01 -2.26383567e-01
3.84689897e-01 8.14996660e-01 -3.59080881e-01 -4.54998463e-01
-1.66012883e-01 8.53043556e-01 7.96212435e-01 6.91021085e-01
-9.08936858e-01 -1.78316038e-03 1.18541074e+00 -2.65416205e-01
2.39956483e-01 -4.80987728e-01 1.92687333e-01 -9.43856657e-01
-5.75409591e-01 -7.15436339e-01 1.53755262e-01 -1.04933012e+00
-1.06090224e+00 -7.05313683e-02 3.57200146e-01 -6.79627359e-01
-6.94882452e-01 -6.28068388e-01 -7.17105567e-01 7.11935759e-01
-1.36776626e+00 -1.02573156e+00 8.51439387e-02 4.07033056e-01
6.11351252e-01 2.91072228e-03 9.88461554e-01 -3.78417730e-01
-6.33202672e-01 5.27019612e-02 2.04244226e-01 2.04230472e-01
4.55208331e-01 -1.60488141e+00 7.69007444e-01 6.42106533e-01
-2.77532712e-02 5.58100581e-01 9.24510658e-01 -5.78812420e-01
-1.93760705e+00 -8.52919757e-01 1.46355122e-01 -7.98027992e-01
6.95863664e-01 -5.91094673e-01 -5.15254438e-01 9.93217170e-01
7.93570876e-01 4.17087108e-01 1.50065392e-01 1.78081706e-01
-2.51150995e-01 8.93143192e-02 -1.01531756e+00 8.16422820e-01
1.04128182e+00 -6.51560545e-01 -6.62947774e-01 3.87048423e-02
6.72957242e-01 -6.10382259e-01 -6.53755128e-01 1.23691171e-01
4.51931804e-01 -7.69773126e-01 1.09469712e+00 -1.05851305e+00
2.18676463e-01 -3.10833097e-01 9.32434481e-03 -1.62926555e+00
-5.88117898e-01 -6.28933847e-01 -6.20870709e-01 6.23130381e-01
4.88742471e-01 -5.03897667e-01 8.35602582e-01 8.92345488e-01
-2.61030734e-01 -8.41243446e-01 -8.89439881e-01 -1.44716695e-01
1.26871675e-01 -3.93077701e-01 3.00036758e-01 6.34555817e-01
-1.01527035e-01 5.02292514e-01 -2.16092914e-01 1.05295546e-01
7.33061910e-01 1.27204461e-02 7.90640056e-01 -9.64748383e-01
-3.96685511e-01 -1.11652963e-01 -7.72081390e-02 -1.00567627e+00
6.60367012e-01 -5.19261658e-01 3.83219272e-01 -1.94291234e+00
2.23604023e-01 -2.87449807e-01 -3.03034782e-01 6.28530562e-01
1.57676116e-02 -2.32068866e-01 5.64351857e-01 -2.47942120e-01
-1.15315032e+00 8.61608684e-01 1.53314114e+00 -5.64867735e-01
-1.18035534e-02 -3.44997078e-01 -3.16254228e-01 7.17860520e-01
7.30884373e-01 -2.18481854e-01 -8.55141938e-01 -6.48850143e-01
4.36780304e-01 2.70399302e-01 6.99405313e-01 -7.32909381e-01
4.32476014e-01 -8.16306353e-01 4.14006621e-01 -5.29803395e-01
9.42383349e-01 -7.95189440e-01 1.01903275e-01 6.80756807e-01
-9.41073060e-01 -7.76775032e-02 1.91214129e-01 9.17410254e-01
1.52979851e-01 1.40159860e-01 8.96799088e-01 -3.01481783e-01
-1.14423728e+00 3.19339991e-01 -9.04666662e-01 1.96286634e-01
9.85441446e-01 1.42232537e-01 -6.27655268e-01 -2.98429668e-01
-9.83504415e-01 6.19948864e-01 4.02615517e-01 5.72252154e-01
5.48935950e-01 -1.15484643e+00 -5.16482890e-01 7.39401355e-02
2.12072134e-01 3.03220242e-01 3.78734246e-02 3.39048028e-01
-4.83948886e-01 1.84182495e-01 -3.18440765e-01 -5.36953449e-01
-6.73882067e-01 6.44057989e-01 6.14329934e-01 -2.70254165e-01
-4.66429770e-01 4.15635854e-01 1.21333539e-01 -7.01810002e-01
4.67541605e-01 -3.14725667e-01 -1.56399667e-01 -1.70047402e-01
1.24201909e-01 4.42707181e-01 -2.95880377e-01 -2.84102082e-01
-2.03405276e-01 -5.42847812e-02 1.31931184e-02 -3.88420343e-01
1.30471897e+00 -4.01482768e-02 3.16880792e-01 6.97930634e-01
1.01963484e+00 -4.76841599e-01 -2.26444387e+00 7.56603330e-02
-6.95739314e-02 -3.42323720e-01 5.52739263e-01 -1.13156283e+00
-9.17046666e-01 8.93219173e-01 9.67787266e-01 -3.88183296e-01
3.64738375e-01 -8.14165547e-02 3.35109204e-01 9.28675830e-01
4.77523714e-01 -1.51804960e+00 2.66929716e-01 7.35866129e-01
8.54432106e-01 -1.57811189e+00 -2.28740051e-02 3.02412391e-01
-9.71118748e-01 5.99068582e-01 1.01950955e+00 -3.36426914e-01
5.65887630e-01 4.05613333e-01 1.81851253e-01 -2.43439868e-01
-1.64548874e+00 -3.50225240e-01 4.00570109e-02 6.96902514e-01
4.61269580e-02 1.08999729e-01 4.91984457e-01 -3.47376674e-01
1.76195174e-01 6.90693408e-02 8.55893418e-02 1.16130018e+00
-4.22489285e-01 -7.58409917e-01 -7.98360705e-02 2.06881687e-01
-3.69077101e-02 3.32589805e-01 -2.59973675e-01 7.98546553e-01
2.78951321e-02 9.31730390e-01 5.81264675e-01 2.48208404e-01
3.68377030e-01 -5.25053998e-04 5.96972227e-01 -6.53639495e-01
-4.07919705e-01 -1.83066100e-01 3.57590109e-01 -1.07007456e+00
-3.63788068e-01 -5.68622172e-01 -1.44202554e+00 -2.33618841e-01
9.29330010e-03 2.10969616e-02 5.64472795e-01 7.50945151e-01
4.98125315e-01 6.06933057e-01 2.69260705e-01 -1.11625850e+00
-6.60672247e-01 -7.15553761e-01 -7.81465904e-04 4.42297667e-01
6.33407474e-01 -9.78480339e-01 -1.67740598e-01 -2.97341719e-02]
|
[4.253448009490967, 1.498718500137329]
|
08b8d549-0a7a-4e8d-a264-83a226890042
|
q-tod-a-query-driven-task-oriented-dialogue
|
2210.07564
| null |
https://arxiv.org/abs/2210.07564v1
|
https://arxiv.org/pdf/2210.07564v1.pdf
|
Q-TOD: A Query-driven Task-oriented Dialogue System
|
Existing pipelined task-oriented dialogue systems usually have difficulties adapting to unseen domains, whereas end-to-end systems are plagued by large-scale knowledge bases in practice. In this paper, we introduce a novel query-driven task-oriented dialogue system, namely Q-TOD. The essential information from the dialogue context is extracted into a query, which is further employed to retrieve relevant knowledge records for response generation. Firstly, as the query is in the form of natural language and not confined to the schema of the knowledge base, the issue of domain adaption is alleviated remarkably in Q-TOD. Secondly, as the query enables the decoupling of knowledge retrieval from the generation, Q-TOD gets rid of the issue of knowledge base scalability. To evaluate the effectiveness of the proposed Q-TOD, we collect query annotations for three publicly available task-oriented dialogue datasets. Comprehensive experiments verify that Q-TOD outperforms strong baselines and establishes a new state-of-the-art performance on these datasets.
|
['Hua Wu', 'Shuqi Sun', 'Huang He', 'Fan Wang', 'Siqi Bao', 'Mengfei Song', 'Yingzhan Lin', 'Xin Tian']
|
2022-10-14
| null | null | null | null |
['task-oriented-dialogue-systems']
|
['natural-language-processing']
|
[-2.60630399e-01 3.57087076e-01 -1.33937582e-01 -2.55134076e-01
-1.18325162e+00 -8.77882183e-01 7.05449224e-01 3.23310792e-02
-7.22240388e-01 9.79395330e-01 6.03804111e-01 -2.49487534e-02
7.01193660e-02 -5.61486125e-01 -3.08659703e-01 -2.71311164e-01
3.40813339e-01 9.42534864e-01 6.91847742e-01 -9.06382740e-01
5.90040199e-02 -1.23643100e-01 -7.03076482e-01 5.03417850e-01
1.00690901e+00 8.11412036e-01 1.48133516e-01 6.17347181e-01
-2.89497674e-01 8.23988736e-01 -8.63283157e-01 -6.55901194e-01
-1.19557709e-01 -3.98383170e-01 -1.56363201e+00 3.83706973e-03
2.69749202e-02 -7.83402681e-01 -3.98382038e-01 6.23694718e-01
8.62227261e-01 2.92203963e-01 4.32647675e-01 -9.53311086e-01
-7.15340018e-01 6.11464202e-01 2.18758490e-02 6.17833389e-03
5.98706305e-01 2.69131601e-01 1.10111880e+00 -1.00328386e+00
9.24756825e-01 1.38308012e+00 2.79350996e-01 8.47285628e-01
-1.17716229e+00 -2.76237316e-02 1.04575276e-01 1.89052541e-02
-1.15832508e+00 -7.42699921e-01 7.80754566e-01 -2.01503396e-01
1.22811449e+00 1.38177529e-01 2.15051323e-01 1.15016723e+00
-3.20447117e-01 1.19153261e+00 6.46657526e-01 -3.99772018e-01
1.40237331e-01 2.73771942e-01 2.29069605e-01 4.67756629e-01
-3.32672924e-01 -3.66200507e-01 -9.88049567e-01 -3.93752247e-01
4.40809906e-01 -6.49031699e-01 -3.00078779e-01 -3.14077735e-01
-1.32784688e+00 7.90019572e-01 2.05982745e-01 5.34753576e-02
-2.87743121e-01 -3.84399980e-01 9.04024601e-01 4.91031528e-01
5.12319922e-01 7.64873087e-01 -6.88219965e-01 -4.15853798e-01
-4.21856165e-01 6.81105852e-01 1.27280152e+00 1.30473936e+00
6.16564393e-01 -4.07743573e-01 -8.55897605e-01 1.20953500e+00
-4.55183443e-03 3.53115022e-01 5.45284808e-01 -1.01976299e+00
7.83242941e-01 8.17172289e-01 4.96970564e-01 -4.60621148e-01
-3.99586380e-01 -5.29126152e-02 -5.94888866e-01 -8.06536257e-01
6.54445171e-01 -5.41578114e-01 -3.04420590e-01 1.65783393e+00
6.38524890e-01 -7.42054164e-01 5.88677168e-01 9.91772473e-01
1.13857448e+00 5.79578459e-01 1.35155097e-01 -2.26177320e-01
1.68348300e+00 -1.20549619e+00 -8.42455089e-01 -2.80326277e-01
8.29972029e-01 -6.16592169e-01 1.13459826e+00 1.83378339e-01
-1.02474368e+00 -3.66703868e-01 -5.52024245e-01 -5.13767898e-01
-1.23584569e-01 -4.07179892e-02 3.57499927e-01 1.35883123e-01
-9.88177240e-01 -9.90589056e-03 -3.22607994e-01 -5.02577722e-01
-1.02645911e-01 4.39848751e-02 -2.40616694e-01 -1.01582505e-01
-1.71142590e+00 9.18879211e-01 6.86804652e-01 8.09579641e-02
-8.84179413e-01 -4.04025674e-01 -6.81801617e-01 -2.76603736e-02
8.15657437e-01 -7.63192952e-01 2.02259326e+00 -5.08874118e-01
-1.78250659e+00 7.32342780e-01 -2.68947244e-01 -3.30276936e-01
7.28243887e-01 -4.79807556e-01 -1.60719976e-01 2.81647772e-01
2.60870516e-01 5.60489535e-01 5.74697554e-01 -8.69856775e-01
-6.86498582e-01 -2.86610156e-01 5.59756041e-01 6.32221699e-01
-4.46847320e-01 -2.46684570e-02 -8.77646863e-01 -3.52244228e-01
-2.35423788e-01 -7.75283754e-01 -9.12065282e-02 -3.70404273e-01
-6.54739499e-01 -9.62853849e-01 7.06828356e-01 -5.37211418e-01
1.34613264e+00 -2.09768844e+00 1.94564447e-01 -5.22571146e-01
1.29271284e-01 4.48507369e-01 -1.35487825e-01 8.95941436e-01
6.67465329e-01 -1.41427428e-01 -1.52269825e-01 -1.22855134e-01
2.66242534e-01 1.83411047e-01 -4.12279189e-01 -2.21574515e-01
4.20363456e-01 1.13207459e+00 -1.24621522e+00 -7.87754774e-01
-3.07085633e-01 -5.57552613e-02 -4.64218587e-01 7.16663957e-01
-8.70432913e-01 5.04956603e-01 -1.14050317e+00 4.33815867e-01
2.16282025e-01 -2.38374889e-01 2.31264517e-01 -9.82801840e-02
7.31075630e-02 7.59167254e-01 -5.18857121e-01 2.11687636e+00
-4.12666798e-01 2.46660888e-01 4.96452153e-02 -6.31946146e-01
8.37808311e-01 6.09633684e-01 9.77020785e-02 -9.42874789e-01
-2.93685943e-01 1.55080780e-01 -2.81238973e-01 -7.56148875e-01
1.00216508e+00 1.41033188e-01 -4.67596591e-01 6.61988020e-01
2.69132972e-01 -2.07183391e-01 2.58238137e-01 5.82213938e-01
9.66116250e-01 1.62072822e-01 1.90556392e-01 -1.79802239e-01
5.60190558e-01 4.77085322e-01 4.92564678e-01 7.66406834e-01
-3.19396615e-01 3.43715437e-02 5.90426087e-01 -2.95754999e-01
-7.05544174e-01 -7.72752166e-01 3.09166797e-02 1.72420359e+00
8.55085533e-03 -3.90359730e-01 -8.44953001e-01 -9.82593954e-01
-3.19575220e-02 6.36685550e-01 -2.72992134e-01 -2.78158247e-01
-6.32135332e-01 -5.22064030e-01 7.36561000e-01 3.51770788e-01
8.23146462e-01 -1.13056791e+00 -3.73349160e-01 6.43946826e-01
-7.29336262e-01 -1.33740699e+00 -7.32365429e-01 -1.17422298e-01
-6.73053265e-01 -9.61653233e-01 -8.03169668e-01 -7.16358840e-01
3.08131099e-01 3.62225801e-01 1.55472529e+00 -1.20022021e-01
3.34520727e-01 4.13586587e-01 -6.98204458e-01 -2.89675653e-01
-5.75634599e-01 5.93990326e-01 -2.24161640e-01 -1.59980103e-01
6.62554264e-01 -1.09999441e-01 -6.63202822e-01 4.71519798e-01
-7.33656943e-01 3.58734950e-02 4.42810863e-01 1.13239229e+00
2.74887592e-01 -3.87617230e-01 1.13690376e+00 -1.02672899e+00
1.24448466e+00 -3.87307703e-01 -3.99955720e-01 4.75947976e-01
-2.97175795e-01 2.78389752e-01 6.27683759e-01 -3.33310723e-01
-1.45248842e+00 7.81996250e-02 -6.59274608e-02 2.35016197e-01
4.68363799e-02 5.67374468e-01 -2.41663679e-01 4.13898438e-01
8.71973574e-01 3.95394474e-01 -3.02838206e-01 -6.79252386e-01
7.00946212e-01 1.09002399e+00 5.56192458e-01 -9.49243009e-01
4.59825546e-01 5.40391207e-02 -6.91101372e-01 -9.17545676e-01
-1.10649478e+00 -6.96008027e-01 -6.25262499e-01 -7.67468736e-02
6.94031775e-01 -9.57496464e-01 -6.81072772e-01 4.84874547e-01
-1.43681729e+00 -4.69411850e-01 -2.30011716e-01 6.73561245e-02
-5.38315356e-01 3.54651630e-01 -8.23527575e-01 -8.34302783e-01
-7.53279507e-01 -9.97448146e-01 1.15929770e+00 2.02064514e-01
-3.03490639e-01 -1.06328702e+00 2.10655183e-01 7.21928418e-01
3.40939641e-01 -3.59482020e-01 1.17977798e+00 -1.33754289e+00
-4.32329565e-01 -1.01326190e-01 -3.51591021e-01 1.24633178e-01
1.21597379e-01 -4.83161330e-01 -1.10514092e+00 -2.01675326e-01
-4.83304635e-02 -1.06333315e+00 6.49558365e-01 -3.40466172e-01
4.98372227e-01 -5.63583314e-01 -1.08854815e-01 -1.11713270e-02
7.41890371e-01 -1.28410710e-02 1.81729347e-01 3.50137651e-01
4.79217380e-01 9.28834498e-01 1.02860272e+00 3.82101148e-01
1.03866899e+00 1.02926385e+00 -8.50032791e-02 9.88960639e-02
7.16067702e-02 -5.16365469e-01 2.71599919e-01 9.07396853e-01
3.46158087e-01 -3.23867589e-01 -9.03022289e-01 8.23587358e-01
-2.14354491e+00 -5.25676489e-01 8.29267651e-02 2.01387262e+00
1.65165985e+00 8.82371888e-03 4.29357439e-01 -4.88918066e-01
5.15548468e-01 1.33855119e-01 -7.02336550e-01 -3.29004377e-01
-2.69905888e-02 -3.02074671e-01 -6.02399111e-02 4.33871746e-01
-1.00581622e+00 1.36515892e+00 6.04157400e+00 7.92748153e-01
-8.00980330e-01 1.74027905e-01 3.08354437e-01 1.40945137e-01
-2.89516579e-02 1.34889095e-03 -1.00886106e+00 1.13363184e-01
7.74548888e-01 -5.71259081e-01 9.96836051e-02 9.02784348e-01
5.79275116e-02 -1.14353180e-01 -1.34225690e+00 3.80979061e-01
-2.54992127e-01 -8.32909584e-01 1.81895956e-01 -2.34934539e-01
3.14487129e-01 -1.14743151e-02 -2.71052778e-01 7.76374936e-01
6.45374715e-01 -5.95999360e-01 4.13992912e-01 3.30868900e-01
7.21999466e-01 -4.44276810e-01 7.64168441e-01 6.71837330e-01
-6.88399255e-01 2.12693155e-01 -5.23859680e-01 1.50020212e-01
1.55988455e-01 1.96294606e-01 -1.50356710e+00 6.32336140e-01
3.87140334e-01 3.09057683e-01 -3.79573256e-01 5.00038981e-01
-4.10633296e-01 3.64982486e-01 -1.34659097e-01 -2.13931873e-01
3.97640407e-01 2.63703883e-01 5.68692744e-01 1.44557571e+00
-4.46013749e-01 2.66504943e-01 4.43059862e-01 6.98621035e-01
-5.04322231e-01 2.15133220e-01 -5.06786108e-01 -1.79799110e-01
7.88770556e-01 1.17449272e+00 5.19123226e-02 -3.81882727e-01
-4.35622096e-01 1.09080708e+00 6.33537412e-01 5.18953145e-01
-1.71495423e-01 -5.23744524e-01 4.25991446e-01 -4.20973271e-01
1.17785849e-01 -8.98077637e-02 3.74499470e-01 -1.24196982e+00
2.84659177e-01 -1.09341776e+00 5.84185779e-01 -3.96360457e-01
-1.51839626e+00 6.86550021e-01 3.33700441e-02 -9.59059060e-01
-8.16793025e-01 -1.96264580e-01 -1.71026647e-01 9.67069924e-01
-1.71860349e+00 -1.18516850e+00 -6.80914223e-02 8.12590957e-01
1.03578627e+00 -9.30849835e-02 1.12659943e+00 3.97098325e-02
-4.88806665e-01 6.69964075e-01 9.40753073e-02 6.64001524e-01
1.26761270e+00 -1.34659100e+00 6.52112961e-01 4.98429984e-01
-1.93479255e-01 5.96871376e-01 4.65272158e-01 -6.00000978e-01
-1.65453422e+00 -1.07265115e+00 1.12669098e+00 -7.70117164e-01
7.56567061e-01 -5.51299095e-01 -1.25542605e+00 3.73843610e-01
2.79579639e-01 -2.07130417e-01 6.24056697e-01 3.08724999e-01
-4.23889965e-01 -8.44863430e-02 -9.10763502e-01 4.55349654e-01
8.05769324e-01 -8.32188904e-01 -1.04005563e+00 6.65108740e-01
1.17491770e+00 -6.27955139e-01 -1.07925701e+00 6.95751160e-02
4.47316229e-01 -6.26026571e-01 7.62054026e-01 -1.04042065e+00
2.62523681e-01 8.92382562e-02 5.89068532e-02 -1.33763194e+00
6.09199516e-02 -9.58614588e-01 -2.80206323e-01 1.42765307e+00
6.51052177e-01 -3.15660834e-01 4.61732447e-01 1.03278971e+00
-7.38855228e-02 -5.54734528e-01 -1.09230793e+00 -6.36682749e-01
2.52202928e-01 8.22890103e-02 5.62207878e-01 8.91818523e-01
5.14814794e-01 1.15270054e+00 -3.66596043e-01 -1.79446891e-01
3.26858729e-01 1.65531322e-01 1.12163270e+00 -1.01825547e+00
-2.83342510e-01 -8.30183998e-02 3.11771333e-01 -1.61625469e+00
1.51427895e-01 -4.34251845e-01 4.00426567e-01 -1.48926616e+00
1.82172269e-01 -3.98044854e-01 -9.34962370e-03 4.94051307e-01
-5.33564568e-01 -2.95757324e-01 8.72940049e-02 4.79515642e-01
-1.19268131e+00 9.18879807e-01 1.47311544e+00 -1.80985466e-01
-6.07924104e-01 -4.13754657e-02 -6.96625471e-01 4.48481232e-01
5.24790406e-01 -2.96498358e-01 -6.58938527e-01 -7.84936368e-01
1.62212238e-01 6.35449231e-01 1.25524472e-04 -2.71262765e-01
5.72834909e-01 -1.92322224e-01 -2.72846401e-01 -2.59192020e-01
3.59955430e-01 -3.84105861e-01 -7.23192453e-01 3.91303226e-02
-6.48087502e-01 -2.02465415e-01 1.48878008e-01 5.12349784e-01
-3.18472266e-01 -1.88332751e-01 4.22135651e-01 -2.11452767e-01
-7.84127712e-01 7.85245076e-02 -2.95374155e-01 8.51979017e-01
5.07874370e-01 2.76920229e-01 -8.75397623e-01 -5.53513765e-01
-3.52789313e-01 9.03569877e-01 2.94540375e-01 5.26613951e-01
4.43327934e-01 -1.00957632e+00 -8.65847528e-01 -2.51555085e-01
7.08228111e-01 4.99999225e-01 2.23107576e-01 6.81178689e-01
-6.02666512e-02 9.33251798e-01 8.86829495e-02 -4.18938011e-01
-8.71656716e-01 4.71160591e-01 3.12824219e-01 -5.70897341e-01
-4.54701394e-01 9.08386946e-01 3.43081504e-01 -6.72559142e-01
3.38082254e-01 4.59268093e-02 -3.49657148e-01 1.90234974e-01
5.22921026e-01 -1.73970498e-02 1.10241346e-01 -3.91581833e-01
-1.56519517e-01 -6.51961789e-02 -6.52730525e-01 -3.18671763e-01
1.06134534e+00 -4.20397490e-01 -1.03508852e-01 4.15955335e-01
7.93557584e-01 -2.43188545e-01 -1.13679433e+00 -1.05029011e+00
4.04695302e-01 -1.00500017e-01 -2.15258494e-01 -1.19443953e+00
-4.57152814e-01 7.29204118e-01 -7.54024386e-02 4.02139246e-01
1.00434232e+00 6.82373345e-02 1.10391307e+00 1.22987378e+00
5.46811879e-01 -1.29226613e+00 2.27552310e-01 1.03000867e+00
9.83700514e-01 -1.28167009e+00 -2.77243167e-01 -2.01176733e-01
-1.06753588e+00 8.87186766e-01 8.92932594e-01 5.24591982e-01
-9.10644326e-03 -2.37408891e-01 3.22484136e-01 -2.42683813e-01
-1.18313169e+00 -2.61401564e-01 1.81359231e-01 5.92180669e-01
4.14175153e-01 -2.25960851e-01 -1.47142544e-01 9.07749653e-01
-1.78594530e-01 -3.54854688e-02 2.59689659e-01 1.05908489e+00
-5.00634670e-01 -1.19400382e+00 1.09959021e-01 6.21494837e-02
-3.81940395e-01 -1.54551238e-01 -1.01427579e+00 6.62750542e-01
-7.35919774e-01 1.28129601e+00 -4.31477308e-01 -2.36671701e-01
8.34707558e-01 3.71318549e-01 1.13422575e-03 -8.57633948e-01
-8.34614813e-01 -8.89963936e-03 7.52505481e-01 -4.93993163e-01
-2.62620628e-01 -1.96246713e-01 -1.13017702e+00 1.34216053e-02
-4.16751325e-01 7.71592617e-01 2.79428661e-01 9.59514201e-01
5.88441193e-01 2.35752419e-01 5.74300528e-01 -2.57039040e-01
-9.95916009e-01 -1.28521407e+00 -2.28079021e-01 3.80034804e-01
4.54677969e-01 -4.21726733e-01 1.80402264e-01 7.45098293e-03]
|
[12.249588966369629, 8.037284851074219]
|
b10c91f6-2ea3-47e3-8395-5c42ac67832a
|
rate-splitting-multiple-access-for-joint
|
2104.08180
| null |
https://arxiv.org/abs/2104.08180v1
|
https://arxiv.org/pdf/2104.08180v1.pdf
|
Rate-Splitting Multiple Access for Joint Radar-Communications with Low-Resolution DACs
|
In this paper, we introduce the design of a multi-antenna Joint Radar-Communication (JRC) system with Rate Splitting Multiple Access (RSMA) and low resolution Digital-to-Analog Converter (DAC) units. Using RSMA, the communication messages are split into private and common parts, then precoded and quantized before transmission. We use a problem formulation to design the JRC system with RSMA and low resolution DACs by maximizing communication sum-rate and the proximity of the resulting JRC waveform to an optimal radar beampattern under an average transmit power constraint. We solve the joint sum-rate maximization and beampattern error minimization problem using Alternating Direction Method of Multipliers (ADMM) method. The numerical results show that RSMA achieves a significantly higher sum-rate compared to Space Division Multiple Access (SDMA) while providing the same Normalized Mean Square Error (NMSE) for the designed radar beampattern.
|
['Christos Masouros', 'Bruno Clerckx', 'Aryan Kaushik', 'Onur Dizdar']
|
2021-04-16
| null | null | null | null |
['joint-radar-communication']
|
['robots']
|
[ 7.49351442e-01 -1.10060997e-01 -1.37326568e-01 -3.09112102e-01
-9.66354430e-01 -4.22445655e-01 4.89345372e-01 -5.63779056e-01
-3.01776379e-01 8.92471790e-01 3.01781595e-01 -4.82197911e-01
-7.72660136e-01 -8.34849954e-01 -9.23019499e-02 -6.87677026e-01
-4.40032333e-01 -3.01591549e-02 -5.20342469e-01 3.15588452e-02
1.64216414e-01 7.78187394e-01 -7.64951408e-01 -2.81476051e-01
7.70037889e-01 1.59319770e+00 -5.89557663e-02 1.07481050e+00
3.86777580e-01 5.37186384e-01 -8.35049033e-01 -1.46125585e-01
8.61449063e-01 -6.78438723e-01 1.15604751e-01 6.82128593e-02
5.07687271e-01 -6.23366773e-01 -6.71547771e-01 1.17442322e+00
6.55784786e-01 -1.20458312e-01 9.43815231e-01 -1.08264911e+00
-6.22056246e-01 2.90489912e-01 -8.97309840e-01 -1.22797832e-01
8.50781277e-02 -2.42337853e-01 7.97014952e-01 -6.74483657e-01
2.71031827e-01 1.31621492e+00 3.52608413e-01 1.18640080e-01
-1.38182938e+00 -1.05516386e+00 -6.39450967e-01 1.43328896e-02
-1.83240187e+00 -6.99595690e-01 3.85080159e-01 -1.81849614e-01
3.21492225e-01 4.75057036e-01 3.05790514e-01 3.43003929e-01
5.03604412e-01 1.12310484e-01 1.05043674e+00 -6.40115798e-01
4.76844698e-01 -2.40132809e-01 -1.14408702e-01 6.34062231e-01
8.21093142e-01 4.47303742e-01 -3.58834594e-01 -4.35371816e-01
1.16202748e+00 -2.65254021e-01 -6.63120449e-01 -7.56768063e-02
-1.25067449e+00 9.53109920e-01 1.07436262e-01 1.03864698e-02
-6.44157767e-01 5.73306739e-01 -6.90261185e-01 7.17718899e-01
-3.31419796e-01 3.07897180e-01 1.99943587e-01 4.99667019e-01
-1.15532196e+00 1.53995812e-01 9.81951058e-01 1.18166625e+00
3.24345708e-01 6.68441415e-01 -3.79250020e-01 7.06773520e-01
9.89382207e-01 1.72953224e+00 -1.72597781e-01 -1.12916052e+00
6.38499200e-01 -3.20705593e-01 3.23889285e-01 -9.85523999e-01
-7.69793466e-02 -9.75621700e-01 -1.26737201e+00 4.54452604e-01
3.08764964e-01 -7.61753261e-01 -5.85666120e-01 1.42552042e+00
-2.45486021e-01 -2.05190286e-01 6.27673209e-01 1.24396276e+00
-1.07439468e-02 1.19748986e+00 -5.60169280e-01 -8.53187799e-01
1.38468099e+00 -5.15327863e-02 -1.24498844e+00 -4.74222749e-01
-1.16070211e-01 -1.09068620e+00 -3.96661192e-01 5.71992874e-01
-1.02527940e+00 -1.93275452e-01 -1.75363779e+00 5.66774189e-01
4.63944137e-01 4.59350288e-01 2.70057142e-01 1.03153372e+00
-4.59888816e-01 7.58088008e-02 -6.87253997e-02 1.21116616e-01
2.09151775e-01 3.87337469e-02 3.16477895e-01 -1.49624333e-01
-9.83994842e-01 9.88095760e-01 -2.46981099e-01 5.25158131e-03
-5.21650374e-01 -7.47103453e-01 -4.65036303e-01 -2.47685686e-01
8.28851387e-03 -5.29784799e-01 1.02087903e+00 -3.43985647e-01
-1.72194517e+00 1.02466390e-01 2.54099905e-01 -9.29586291e-01
-8.05720240e-02 -6.53054658e-03 -1.02486372e+00 1.81608588e-01
-3.91803533e-01 2.61295170e-01 1.12732708e+00 -9.38175321e-01
-7.82407701e-01 -2.61543065e-01 -7.88820148e-01 -2.83532202e-01
4.74765152e-01 -3.50414634e-01 2.39159733e-01 -1.08149207e+00
4.68893141e-01 -6.54432476e-01 -4.03567642e-01 5.17167032e-01
-1.84572896e-03 6.08352602e-01 6.98538184e-01 -5.72407961e-01
1.55787349e+00 -2.20335317e+00 1.63445756e-01 7.17229724e-01
-3.54173750e-01 1.77415967e-01 -1.99173391e-01 3.49172205e-01
3.10283244e-01 -7.14004040e-01 -4.67472970e-01 3.23598713e-01
-7.37366900e-02 1.41715229e-01 -7.50330925e-01 9.65513766e-01
-1.91663891e-01 2.55195409e-01 -1.93679154e-01 -1.40733615e-01
-2.32706785e-01 1.15957893e-01 -2.32137516e-01 -7.99061581e-02
1.32544003e-02 1.74488891e-02 -4.65554446e-01 9.20625746e-01
1.22473359e+00 2.75778323e-01 2.76493251e-01 -4.91733193e-01
-3.53591502e-01 -2.06596464e-01 -1.67202580e+00 1.20948601e+00
-4.21156645e-01 8.65646541e-01 9.72224891e-01 -6.43054008e-01
1.50589287e+00 2.83305556e-01 1.83233157e-01 -1.02309847e+00
1.94128677e-01 4.91266489e-01 -2.48023234e-02 -7.24274898e-03
4.65444505e-01 -5.13850749e-01 -2.25917041e-01 6.85668886e-01
-2.06995562e-01 -1.31046623e-01 -2.01177746e-01 1.57380149e-01
9.16118026e-01 -7.37699687e-01 7.44793475e-01 -2.59829968e-01
6.29775941e-01 -2.77853191e-01 6.20723665e-01 6.74863219e-01
1.56546943e-02 3.04580405e-02 -2.06055388e-01 2.75457680e-01
-1.10117412e+00 -1.41861105e+00 -5.01638412e-01 8.37702602e-02
2.21701011e-01 2.33292654e-01 1.36786044e-01 3.70925128e-01
6.55710042e-01 9.66667771e-01 4.33230996e-01 1.38313666e-01
-3.29947799e-01 -6.25940204e-01 7.22506940e-01 -1.74229279e-01
6.39889777e-01 2.05625206e-01 -6.43214107e-01 5.74034810e-01
2.88776070e-01 -1.07113528e+00 -3.48420173e-01 -9.04381350e-02
-7.68256068e-01 -4.28743660e-01 -8.29631507e-01 -3.53500664e-01
4.79468971e-01 4.63184237e-01 7.95985088e-02 -9.41954553e-01
-6.02426410e-01 6.64281130e-01 -1.11958824e-01 -1.73234299e-01
-1.19864255e-01 -9.54253256e-01 2.69554317e-01 3.52789640e-01
-2.18685204e-03 -7.04519928e-01 -3.96734923e-01 2.20645159e-01
-5.12412190e-01 -2.49782220e-01 1.31166649e+00 4.07076925e-01
2.38453239e-01 -9.96337682e-02 8.45283687e-01 1.79470882e-01
6.29492581e-01 -2.94079900e-01 -1.44687450e+00 2.18910217e-01
-8.09706628e-01 3.53212088e-01 3.21268767e-01 -2.47234508e-01
-8.12290907e-01 1.84854686e-01 3.88827384e-01 -4.55063730e-02
4.48959917e-01 3.04723859e-01 -4.82127480e-02 -6.24416828e-01
6.78571463e-01 4.24538881e-01 4.91616964e-01 -3.08956355e-01
6.59875751e-01 1.29680073e+00 9.44779456e-01 -2.03556716e-01
1.33385360e+00 3.88986230e-01 6.54430211e-01 -1.20780599e+00
-6.36456907e-01 -5.45678101e-02 -8.25945567e-03 -3.12319756e-01
5.29155910e-01 -1.26245570e+00 -5.27048647e-01 -1.21078275e-01
-1.10968101e+00 3.12506884e-01 -4.46792766e-02 1.34495056e+00
-4.20180380e-01 4.16660994e-01 -1.08787857e-01 -1.35090518e+00
-6.32360220e-01 -3.15155149e-01 3.79259020e-01 8.69029909e-02
-6.43877238e-02 -2.56599754e-01 1.33893313e-02 1.06768169e-01
8.95430207e-01 3.43753219e-01 6.65108562e-01 2.82964706e-01
-1.32454479e+00 -3.87283862e-01 -3.93389761e-01 1.65076494e-01
-1.05763622e-01 -7.69596636e-01 -3.10351223e-01 -5.14724314e-01
-6.19038045e-02 1.92624837e-01 5.90786934e-01 6.43033981e-01
3.66662174e-01 -7.66705096e-01 -3.27253461e-01 6.99468553e-01
1.98971426e+00 4.06202018e-01 1.00084090e+00 -2.53431082e-01
-6.01397812e-01 -2.25959778e-01 6.71384752e-01 1.07817543e+00
-1.55323505e-01 6.42685652e-01 1.63774759e-01 6.16364062e-01
-9.71508622e-02 3.26662630e-01 6.01497233e-01 4.03852165e-01
3.06058377e-01 -1.98952630e-01 -8.89823586e-02 1.98145673e-01
-1.39878178e+00 -1.18355918e+00 -2.80443370e-01 2.45791793e+00
8.43666434e-01 -2.61064738e-01 -3.57952207e-01 8.37446898e-02
5.96188545e-01 1.58476084e-01 -1.77535519e-01 -4.10103172e-01
-1.57062352e-01 4.07763809e-01 1.57728493e+00 7.62022257e-01
-8.12473893e-01 -1.11789726e-01 6.32531357e+00 9.02200997e-01
-7.74370074e-01 -1.19630108e-02 -3.99492979e-01 -1.45295948e-01
-2.50131041e-01 -3.02060805e-02 -8.97768795e-01 2.94140100e-01
1.23106432e+00 -6.19094372e-01 5.82405090e-01 3.70253354e-01
-4.79730032e-03 -3.89005512e-01 -7.66645849e-01 1.39537573e+00
9.96447131e-02 -1.40318346e+00 -3.99797261e-02 4.19067949e-01
4.71599758e-01 -4.64679360e-01 8.38438328e-03 6.99298829e-02
3.33773166e-01 -7.47061610e-01 5.52269816e-01 1.09024644e+00
1.06316662e+00 -6.50436997e-01 1.84648946e-01 1.49374336e-01
-1.12424076e+00 -5.40191054e-01 -5.86297810e-01 -1.46438420e-01
5.20664334e-01 1.23197865e+00 -5.17426848e-01 6.60957336e-01
-2.45842069e-01 -1.05575025e-01 2.34447494e-01 1.30704784e+00
-2.83987448e-02 4.22196835e-01 -6.28398359e-01 -2.51531750e-01
1.42643094e-01 -6.85539901e-01 1.05897534e+00 1.09734654e+00
1.00797641e+00 9.22001839e-01 -1.04775287e-01 8.16775501e-01
2.08109319e-01 -4.73936349e-01 -4.47424680e-01 -3.32578793e-02
1.20621192e+00 1.01965952e+00 3.31238568e-01 -1.19062752e-01
-3.57918203e-01 5.83566725e-01 -7.90763199e-01 4.06099588e-01
-6.58051968e-01 -1.02912915e+00 8.17891419e-01 -1.70031860e-01
4.72209275e-01 -8.46000731e-01 -4.66624677e-01 -4.14118737e-01
-1.76184610e-01 -5.68522334e-01 3.14306855e-01 -5.97499251e-01
-1.11546159e+00 -6.94157481e-02 -3.61803621e-01 -1.88703644e+00
-3.45738158e-02 -2.54678756e-01 -2.44543582e-01 1.09034026e+00
-1.65740800e+00 -5.04197180e-01 6.50394037e-02 3.86258334e-01
-1.75481234e-02 -8.00500333e-01 7.91834652e-01 4.86552566e-01
1.04202807e-01 6.34017110e-01 6.25629425e-01 2.63001174e-02
5.73945582e-01 -5.50251722e-01 -4.76823211e-01 1.04422200e+00
-4.14281368e-01 3.20644081e-01 8.42475712e-01 -5.86999655e-01
-2.19689727e+00 -1.23644769e+00 7.32707798e-01 6.08515441e-01
6.90192759e-01 1.06302537e-01 -1.57599047e-01 3.67803186e-01
3.88029456e-01 -3.78467917e-01 9.11611080e-01 -5.93261182e-01
-2.94742674e-01 -5.15102565e-01 -1.44077671e+00 3.79621267e-01
3.86645287e-01 -2.09638178e-01 -7.36793876e-01 3.64105217e-02
2.75750399e-01 -4.35273610e-02 -1.03291333e+00 2.43433103e-01
8.98168266e-01 -1.34232238e-01 1.11808705e+00 2.69820958e-01
3.84306237e-02 -9.12274122e-01 -1.16758466e+00 -1.31194448e+00
-8.11910987e-01 -9.52894330e-01 -2.41491422e-01 9.29749370e-01
5.10577321e-01 -4.97651041e-01 2.76363343e-01 -9.86661613e-02
2.26745725e-01 -1.18866295e-01 -1.31189895e+00 -1.15412772e+00
-4.04263854e-01 -2.33416364e-01 1.12095825e-01 4.09729242e-01
2.49322712e-01 5.20837963e-01 -9.09018576e-01 1.00640607e+00
1.66516733e+00 4.14335936e-01 6.00774229e-01 -1.29114437e+00
-4.90372062e-01 -1.01780325e-01 -4.40644145e-01 -1.61332011e+00
-5.64627230e-01 -1.01902175e+00 -1.81059018e-02 -1.57380891e+00
-5.98234177e-01 -4.76490259e-01 2.35104471e-01 -1.31342426e-01
8.47688735e-01 -1.67415068e-02 2.11045578e-01 8.06828141e-02
-1.98567182e-01 7.38003790e-01 1.07372487e+00 -2.62799591e-01
3.05375922e-02 3.77783835e-01 -4.77689981e-01 2.08971761e-02
5.12440860e-01 -2.47725040e-01 1.42951429e-01 -2.09880441e-01
-2.49666404e-02 1.05707037e+00 3.36742222e-01 -1.47614932e+00
6.19240999e-01 -4.05354649e-01 5.53426862e-01 -9.88584101e-01
6.18276775e-01 -1.22947621e+00 4.13128883e-01 9.64014471e-01
-1.52518719e-01 -7.01778293e-01 -2.21363679e-01 1.01290500e+00
8.26562122e-02 -1.81778818e-01 1.30247748e+00 5.68356156e-01
-3.82875532e-01 -1.62210032e-01 -1.17226076e+00 -5.44334412e-01
1.12390101e+00 8.91940221e-02 -6.88613832e-01 -6.79243624e-01
-4.31584835e-01 5.25457144e-01 -3.69978011e-01 -9.71455947e-02
9.18071449e-01 -1.73638690e+00 -1.01890492e+00 4.99905407e-01
-1.78630993e-01 -8.81541431e-01 -4.58604023e-02 6.51216090e-01
-3.22426558e-01 8.91710043e-01 -2.70791978e-01 -3.28455657e-01
-1.11491489e+00 -2.91208863e-01 2.83304840e-01 3.46177042e-01
-4.27710950e-01 4.80913669e-01 -9.11901474e-01 1.76041201e-01
3.28316033e-01 1.56983603e-02 3.17020118e-01 -3.37098576e-02
9.61001337e-01 5.75450003e-01 -3.66482317e-01 -1.77131936e-01
-3.60130966e-01 7.77194262e-01 3.40372384e-01 -6.01016521e-01
9.62225974e-01 -2.04292089e-01 -3.01245265e-02 -3.84494036e-01
1.03989255e+00 2.37313390e-01 -8.21738839e-01 -5.61608493e-01
-2.71089405e-01 -8.53435695e-01 5.18944979e-01 -8.92871797e-01
-8.75140131e-01 2.61811614e-01 1.12806344e+00 -8.85167196e-02
1.05745137e+00 -4.20726091e-01 8.00348282e-01 8.17767262e-01
6.69910491e-01 -1.22073066e+00 -2.07720309e-01 4.09408391e-01
9.48865473e-01 -3.00846279e-01 8.09551358e-01 -1.11262068e-01
-3.65500093e-01 1.32608593e+00 -1.98891491e-01 -2.81145364e-01
8.71390820e-01 7.54962146e-01 -2.46747166e-01 -1.38788540e-02
-7.45291770e-01 -2.01218888e-01 -1.52557539e-02 7.20551193e-01
6.48086816e-02 3.60456765e-01 -8.89485240e-01 5.84235430e-01
-1.80089697e-01 6.67811334e-02 9.75972652e-01 8.42141032e-01
-1.44394708e+00 -8.26548100e-01 -1.23956203e+00 6.29452467e-01
5.23672923e-02 1.56286150e-01 8.71148184e-02 1.29075006e-01
-2.53533959e-01 1.26144350e+00 3.16890806e-01 -1.17528401e-01
4.14615721e-01 -1.53372154e-01 7.42449522e-01 6.31350353e-02
5.56808829e-01 2.63498843e-01 3.61774564e-01 -1.27932966e-01
-6.02480769e-02 -5.87836802e-01 -1.07243109e+00 -2.83971965e-01
-3.99944186e-01 1.87702015e-01 9.19598103e-01 5.62757194e-01
5.07853746e-01 5.04599512e-01 1.12780058e+00 3.91947338e-03
-1.19933808e+00 -5.71024597e-01 -1.13381875e+00 -7.51309395e-01
5.98228395e-01 -2.78431445e-01 -4.67684031e-01 -1.05449326e-01]
|
[6.405088424682617, 1.2365617752075195]
|
2d45279d-576e-470d-a015-27ae0850ac16
|
growing-and-serving-large-open-domain
|
2305.09464
| null |
https://arxiv.org/abs/2305.09464v1
|
https://arxiv.org/pdf/2305.09464v1.pdf
|
Growing and Serving Large Open-domain Knowledge Graphs
|
Applications of large open-domain knowledge graphs (KGs) to real-world problems pose many unique challenges. In this paper, we present extensions to Saga our platform for continuous construction and serving of knowledge at scale. In particular, we describe a pipeline for training knowledge graph embeddings that powers key capabilities such as fact ranking, fact verification, a related entities service, and support for entity linking. We then describe how our platform, including graph embeddings, can be leveraged to create a Semantic Annotation service that links unstructured Web documents to entities in our KG. Semantic annotation of the Web effectively expands our knowledge graph with edges to open-domain Web content which can be used in various search and ranking problems. Finally, we leverage annotated Web documents to drive Open-domain Knowledge Extraction. This targeted extraction framework identifies important coverage issues in the KG, then finds relevant data sources for target entities on the Web and extracts missing information to enrich the KG. Finally, we describe adaptations to our knowledge platform needed to construct and serve private personal knowledge on-device. This includes private incremental KG construction, cross-device knowledge sync, and global knowledge enrichment.
|
['Chiraag Sumanth', 'Theodoros Rekatsinas', 'Jeffrey Pound', 'Ali Mousavi', 'Umar Farooq Minhas', 'Yunyao Li', 'JP Lacerda', 'Ihab F. Ilyas']
|
2023-05-16
| null | null | null | null |
['knowledge-graph-embeddings', 'knowledge-graph-embeddings', 'fact-verification', 'entity-linking']
|
['graphs', 'methodology', 'natural-language-processing', 'natural-language-processing']
|
[-6.99805439e-01 8.32933843e-01 -7.86649346e-01 -1.34461731e-01
-8.64536107e-01 -1.11062765e+00 4.14682567e-01 7.64786124e-01
-1.28563181e-01 8.84294987e-01 6.51632667e-01 -1.10314831e-01
-6.20323241e-01 -1.32977855e+00 -8.28198016e-01 2.27329120e-01
-9.10684019e-02 9.37381387e-01 8.41162920e-01 -3.78837407e-01
-1.85083747e-01 1.97529018e-01 -1.17117214e+00 3.21204782e-01
7.87151337e-01 9.55057919e-01 -3.05896461e-01 1.72732085e-01
-7.29829192e-01 8.74307692e-01 -3.45668882e-01 -1.08538508e+00
1.66765869e-01 6.26863837e-01 -1.45124829e+00 -4.54909801e-01
4.66953367e-01 -1.02444239e-01 -5.23080409e-01 9.08419549e-01
4.73506331e-01 -4.25252281e-02 2.71639138e-01 -1.52579260e+00
-1.19678497e+00 1.15441096e+00 6.58032596e-02 5.12716956e-02
4.61696118e-01 -4.42665130e-01 1.42215776e+00 -6.89000428e-01
1.42782748e+00 7.53625333e-01 8.86619270e-01 2.04205111e-01
-4.91577893e-01 -4.18233514e-01 -1.01709887e-01 4.81742889e-01
-1.27856946e+00 -3.03616494e-01 5.03546178e-01 -3.05155933e-01
1.34803689e+00 -5.04962727e-02 8.40554476e-01 9.50010598e-01
-5.92427373e-01 6.26155674e-01 2.88898259e-01 -4.90538448e-01
9.53172743e-02 5.62483966e-01 6.46319926e-01 1.09529436e+00
1.13657951e+00 -6.59780443e-01 -8.15653622e-01 -4.66321260e-01
5.01729786e-01 -3.20533752e-01 -1.69242695e-01 -7.65594244e-01
-8.96491289e-01 5.66974282e-01 3.88292193e-01 8.87370557e-02
-4.25442547e-01 3.61413687e-01 5.77829957e-01 1.13283060e-01
4.34874803e-01 7.30328918e-01 -1.18464577e+00 4.37715789e-03
-2.84433216e-01 2.28774235e-01 1.39382207e+00 1.55397785e+00
1.05442941e+00 -5.25529504e-01 3.12161773e-01 8.52486551e-01
1.94181949e-01 2.76669323e-01 2.86367238e-01 -9.61254954e-01
5.78667283e-01 1.24279284e+00 2.14442983e-01 -8.62960815e-01
-4.54153657e-01 -3.26367468e-01 3.52422267e-01 -5.53996027e-01
2.46276289e-01 -6.78759962e-02 -5.83575904e-01 1.26818931e+00
8.43944311e-01 1.46143571e-01 5.08407354e-01 5.39268553e-01
1.44813347e+00 -5.03958128e-02 1.47145599e-01 4.11009938e-01
1.80276394e+00 -8.27187061e-01 -8.60845029e-01 1.89365242e-02
1.02357936e+00 -2.49611020e-01 8.04347634e-01 -1.89641133e-01
-7.78666735e-01 -4.12326641e-02 -8.86367142e-01 -4.27123636e-01
-1.53075147e+00 -3.34049523e-01 1.08988607e+00 5.17661095e-01
-1.10599101e+00 4.52332199e-01 -5.83170712e-01 -9.31326091e-01
7.53039360e-01 2.25034401e-01 -8.29926193e-01 -2.79790044e-01
-1.66855490e+00 9.05576944e-01 1.16780043e+00 -8.54936302e-01
-4.08805400e-01 -1.21107984e+00 -1.14297700e+00 1.50535643e-01
9.17352796e-01 -1.12463343e+00 9.31399703e-01 -1.27222940e-01
-5.60717404e-01 9.13883746e-01 2.66136020e-01 -4.19092327e-01
-3.60246152e-01 -1.66928604e-01 -1.10398340e+00 2.43376389e-01
5.33494234e-01 3.79799724e-01 1.74871057e-01 -1.09643495e+00
-9.33809102e-01 -5.64779580e-01 4.23724741e-01 4.98570465e-02
-8.70469689e-01 -1.85452010e-02 -9.86870885e-01 -2.02426001e-01
-1.29149944e-01 -2.94870853e-01 2.33380616e-01 -4.10280854e-01
-3.70380849e-01 -5.16725838e-01 9.87470090e-01 -9.47911382e-01
1.40650821e+00 -1.82024002e+00 -3.22748572e-01 4.59386647e-01
6.77756667e-01 -8.00328627e-02 1.42805457e-01 8.50035667e-01
1.90228894e-01 4.95236874e-01 4.95370835e-01 4.01870102e-01
4.11647469e-01 4.32097793e-01 -2.78183192e-01 -2.11234480e-01
1.59440473e-01 1.63595557e+00 -1.27113760e+00 -7.24293530e-01
-3.53861690e-01 1.97691172e-01 -5.16335785e-01 -3.33690315e-01
-6.56044900e-01 -4.20614779e-01 -9.13514078e-01 1.11904049e+00
3.12555641e-01 -8.43757570e-01 9.33368981e-01 -6.69365108e-01
3.29183161e-01 4.02143449e-01 -1.23290670e+00 1.76703012e+00
-3.18644881e-01 2.28981316e-01 -1.72254369e-01 -5.78371763e-01
6.75577819e-01 1.64585650e-01 6.83946371e-01 -3.31412494e-01
-1.29878059e-01 4.24962074e-01 -9.24532294e-01 -8.07412088e-01
9.28746223e-01 6.10543191e-01 -2.95726180e-01 2.93113500e-01
7.53878713e-01 1.98885217e-01 3.92771155e-01 9.13942695e-01
1.63680601e+00 2.19870493e-01 4.45776820e-01 -5.58510832e-02
6.67838976e-02 5.93242943e-01 3.58528703e-01 3.53366971e-01
1.55716062e-01 -4.53267068e-01 5.58194280e-01 -5.01877964e-01
-9.91326094e-01 -1.20543790e+00 -3.12342285e-03 1.10318625e+00
2.06161693e-01 -1.01370704e+00 -3.00670773e-01 -1.54652882e+00
8.14426482e-01 5.81463039e-01 -3.96133631e-01 -3.67955156e-02
-2.25458682e-01 -3.52251798e-01 7.92016983e-01 8.49352062e-01
2.55161881e-01 -7.72765219e-01 6.34771585e-02 1.97710395e-01
-2.53691107e-01 -1.51424241e+00 2.65263300e-02 4.53829207e-02
-3.33561122e-01 -1.81084323e+00 8.56177509e-02 -1.05314469e+00
5.56993961e-01 -5.63792177e-02 1.47909403e+00 -2.34744176e-01
-2.41164997e-01 1.13803220e+00 -6.34165287e-01 -4.24814612e-01
-9.58509669e-02 2.91859448e-01 1.06835008e-01 -6.19857967e-01
9.23206151e-01 -4.97788668e-01 -3.26283008e-01 1.68045610e-01
-9.32583690e-01 -5.36162257e-01 1.29376620e-01 4.50599968e-01
6.15022004e-01 5.11811018e-01 9.86445665e-01 -1.23321521e+00
6.60556257e-01 -1.03907704e+00 -5.74659407e-01 6.87300265e-01
-9.52069402e-01 2.49185607e-01 2.16685399e-01 1.23868547e-02
-1.05896890e+00 -1.96207717e-01 5.86483218e-02 -3.17124933e-01
2.02410802e-01 1.06529057e+00 -4.25200373e-01 -1.98474571e-01
8.86017263e-01 -2.99403638e-01 -4.64207947e-01 -7.46626794e-01
1.27218878e+00 7.03084290e-01 6.60670042e-01 -7.45878816e-01
9.13692951e-01 5.25046825e-01 -2.45854959e-01 -2.55658746e-01
-1.03515863e+00 -7.66410947e-01 -5.82754076e-01 1.81303173e-01
6.59587622e-01 -1.13962734e+00 -7.41425335e-01 -2.34490156e-01
-8.56516480e-01 -1.25663728e-01 -9.42242265e-01 3.13276350e-02
-1.47962451e-01 3.23189974e-01 -4.49003369e-01 -2.30777293e-01
-3.71073037e-01 -2.39532977e-01 7.99080253e-01 2.96348602e-01
5.67248762e-02 -1.44985390e+00 1.19992927e-01 7.73926973e-01
1.86165079e-01 1.36076510e-01 9.71336544e-01 -1.38575053e+00
-7.67239571e-01 -4.32874680e-01 -5.56741297e-01 -5.18384250e-03
1.61976293e-01 -2.66777575e-01 -6.24037683e-01 1.11595288e-01
-1.14824271e+00 -5.54525316e-01 4.88677710e-01 -2.60843128e-01
8.43120337e-01 -5.08231223e-01 -8.99751246e-01 5.95158160e-01
1.56735229e+00 -2.58445263e-01 3.72710586e-01 8.32310855e-01
9.58053112e-01 5.90568423e-01 5.67390800e-01 3.78156543e-01
1.00550580e+00 5.23958921e-01 2.71370262e-01 2.40989685e-01
-3.33293229e-01 -7.98405945e-01 -9.55626890e-02 6.81485355e-01
2.08613083e-01 -3.33369344e-01 -9.88652110e-01 1.12787426e+00
-1.92009723e+00 -9.19452310e-01 -2.35806927e-01 1.78855658e+00
1.02254653e+00 -1.51892807e-02 -7.57891312e-02 -4.00954872e-01
4.74061012e-01 -1.96773589e-01 -5.11760712e-01 1.69425845e-01
-2.49004349e-01 3.61756623e-01 1.17976367e+00 3.00268590e-01
-8.43595028e-01 1.46711946e+00 6.08610821e+00 7.73296356e-01
-2.99986660e-01 6.32670820e-01 -3.61787766e-01 1.47416309e-01
-7.84011483e-01 4.16283280e-01 -1.34553504e+00 2.12122455e-01
9.42598045e-01 -7.59306490e-01 3.21280599e-01 1.32882023e+00
-6.53967977e-01 4.71611321e-01 -8.35056484e-01 7.22729266e-01
-5.39229736e-02 -2.10359097e+00 1.42266452e-01 1.40025333e-01
7.20980287e-01 2.38875583e-01 -4.18205589e-01 6.27562761e-01
1.23959625e+00 -4.99954045e-01 1.81507275e-01 3.33647817e-01
1.16747832e+00 -5.99394143e-01 8.14500511e-01 -2.44009092e-01
-1.34767985e+00 -1.86544865e-01 -2.64285564e-01 6.49550200e-01
2.45600969e-01 8.30667436e-01 -1.31686497e+00 1.22143817e+00
7.85848618e-01 8.08947027e-01 -5.96733272e-01 6.21063113e-01
-4.76114720e-01 1.84290931e-01 -4.16043013e-01 1.03175893e-01
-1.94331333e-01 3.37860435e-01 3.90126616e-01 1.21417952e+00
1.47538170e-01 -1.17241986e-01 2.37604335e-01 6.08996749e-01
-8.24290991e-01 3.07967067e-01 -8.10463250e-01 -6.04979932e-01
1.01205087e+00 1.47036517e+00 -3.63516957e-01 -7.74561048e-01
-6.88519597e-01 9.06379521e-01 8.09838295e-01 3.74254197e-01
-6.51110113e-01 -8.36965621e-01 8.96183968e-01 3.48884642e-01
4.80330825e-01 -2.59823576e-02 2.85601139e-01 -1.46523941e+00
1.69950858e-01 -4.23356205e-01 1.28262258e+00 -9.01648641e-01
-1.48342240e+00 1.37373626e-01 1.04226425e-01 -5.57300150e-01
-1.07678004e-01 -6.34728789e-01 6.60356181e-03 4.58952278e-01
-1.76244211e+00 -1.68239748e+00 -4.47118104e-01 8.85287344e-01
-1.67840049e-01 -2.09187299e-01 8.68252575e-01 7.30318546e-01
-1.79849848e-01 5.62580705e-01 2.41272766e-02 5.12270749e-01
7.54483640e-01 -1.41256964e+00 6.24805093e-01 4.20634627e-01
2.97426760e-01 8.73729169e-01 5.24861701e-02 -1.41484499e+00
-1.57000089e+00 -1.34119582e+00 1.01509976e+00 -1.08753502e+00
1.41274130e+00 -2.72380143e-01 -7.76776314e-01 1.24430907e+00
-1.62761703e-01 4.97223556e-01 1.01112437e+00 7.42287993e-01
-9.07363176e-01 -9.22573879e-02 -1.31720924e+00 2.54256129e-01
1.54779017e+00 -6.15305424e-01 -8.01668763e-01 6.47635520e-01
1.34587348e+00 -2.65184581e-01 -1.62761497e+00 2.62408257e-01
5.22149384e-01 -5.72374612e-02 1.08932269e+00 -1.37271929e+00
1.10174417e-01 -4.97685522e-01 -2.10458577e-01 -1.16522694e+00
-3.96830410e-01 -3.59567523e-01 -9.82718229e-01 1.50920320e+00
7.08075106e-01 -7.85216868e-01 1.00191498e+00 6.96595430e-01
-2.50261724e-01 -3.56355190e-01 -6.54374421e-01 -9.54347014e-01
-4.50995326e-01 -4.04024273e-01 9.03818309e-01 1.59171343e+00
6.67115867e-01 1.99981391e-01 2.56510764e-01 7.01106548e-01
5.82421720e-01 2.10029352e-02 8.22514415e-01 -1.51060569e+00
-1.99148580e-01 8.35918784e-02 -8.18724871e-01 -5.26503801e-01
2.18089119e-01 -1.54315758e+00 -8.95051062e-01 -2.40310478e+00
1.70600593e-01 -7.38895237e-01 -4.01612133e-01 1.04571557e+00
9.65983272e-02 -5.51194325e-03 -3.74101818e-01 6.50399253e-02
-1.12850463e+00 1.10977858e-01 6.54660702e-01 -1.18861817e-01
-1.19959384e-01 -8.19154143e-01 -1.30017400e+00 4.96411681e-01
5.20669281e-01 -4.54316705e-01 -5.07623732e-01 -4.97246355e-01
8.84115040e-01 -4.99672711e-01 3.49773198e-01 -7.63838947e-01
5.03188610e-01 -1.74162891e-02 2.61540949e-01 -2.46297583e-01
1.44633740e-01 -9.14875269e-01 2.60373533e-01 -2.61350662e-01
-2.36883219e-02 -3.28006655e-01 3.15164089e-01 7.25489855e-01
-1.18965402e-01 -2.25585192e-01 1.43503830e-01 -4.05305654e-01
-1.38455904e+00 3.71284604e-01 2.83427685e-01 6.16059959e-01
1.11546350e+00 -2.60657426e-02 -1.08699358e+00 -1.61504447e-01
-9.45712149e-01 6.68207169e-01 5.71915686e-01 5.92060804e-01
3.37906271e-01 -1.48765731e+00 -2.70417958e-01 -1.66004434e-01
8.28581154e-01 -1.53160945e-01 2.06383258e-01 3.83686930e-01
-3.56655210e-01 3.77025753e-01 -2.82051321e-02 2.42226645e-01
-1.09439564e+00 8.13587010e-01 -8.10016785e-03 -4.49314654e-01
-6.31706178e-01 7.12316573e-01 -4.65650111e-01 -8.04106534e-01
6.31199181e-02 -4.98124107e-04 -2.34549567e-01 2.79354215e-01
5.16665757e-01 4.34021384e-01 3.64918143e-01 2.45292555e-03
-5.90807915e-01 1.21228799e-01 -2.48140171e-02 9.01866704e-02
1.52461898e+00 -1.65217236e-01 -3.25078249e-01 -6.21400066e-02
1.04868996e+00 6.39083147e-01 -4.12402242e-01 -4.47748780e-01
4.23070848e-01 -3.17055374e-01 -2.30181552e-02 -1.24848759e+00
-9.39845562e-01 -1.36688277e-01 -7.07054362e-02 3.84839624e-01
6.16040885e-01 7.66817331e-01 9.23361301e-01 7.64148295e-01
7.41782129e-01 -1.42109990e+00 -1.84931621e-01 4.67948049e-01
3.80276740e-01 -9.36253428e-01 9.38258767e-02 -8.21648598e-01
-5.32007813e-01 9.92493689e-01 7.27050185e-01 4.87670898e-01
8.52150083e-01 2.11981729e-01 -1.08572714e-01 -1.02150631e+00
-7.48715878e-01 -6.33254766e-01 1.78477094e-01 1.10518599e+00
-3.09095047e-02 -1.59797475e-01 -2.09792424e-02 1.08479416e+00
-1.70709670e-01 3.27670306e-01 3.48871559e-01 1.01349974e+00
-4.71617132e-01 -1.27206469e+00 1.36846930e-01 7.23774314e-01
-4.43695247e-01 -4.37642187e-01 -4.53311414e-01 8.35695386e-01
4.09335554e-01 6.03499472e-01 -2.40112156e-01 -4.16697830e-01
6.17550313e-01 5.32912970e-01 1.52519330e-01 -7.84626544e-01
-2.78515041e-01 -7.04317808e-01 9.98248041e-01 -7.07911551e-01
-9.51898168e-04 -2.47428611e-01 -1.51363540e+00 -3.96747440e-01
-4.89134848e-01 5.26532710e-01 7.51914382e-01 4.85987097e-01
1.17069507e+00 3.66623998e-01 -2.36485958e-01 3.08118910e-01
6.46271184e-02 -4.81504172e-01 -7.94192493e-01 5.80326676e-01
-4.94417161e-01 -7.12022424e-01 7.99959805e-03 1.90159023e-01]
|
[9.1265230178833, 8.140043258666992]
|
09bfe8de-66f6-437d-b419-07ebfdb9e8ed
|
multi-objective-reinforcement-learning-with
|
1406.3497
| null |
http://arxiv.org/abs/1406.3497v2
|
http://arxiv.org/pdf/1406.3497v2.pdf
|
Multi-objective Reinforcement Learning with Continuous Pareto Frontier Approximation Supplementary Material
|
This document contains supplementary material for the paper "Multi-objective
Reinforcement Learning with Continuous Pareto Frontier Approximation",
published at the Twenty-Ninth AAAI Conference on Artificial Intelligence
(AAAI-15). The paper is about learning a continuous approximation of the Pareto
frontier in Multi-Objective Markov Decision Problems (MOMDPs). We propose a
policy-based approach that exploits gradient information to generate solutions
close to the Pareto ones. Differently from previous policy-gradient
multi-objective algorithms, where n optimization routines are use to have n
solutions, our approach performs a single gradient-ascent run that at each step
generates an improved continuous approximation of the Pareto frontier. The idea
is to exploit a gradient-based approach to optimize the parameters of a
function that defines a manifold in the policy parameter space so that the
corresponding image in the objective space gets as close as possible to the
Pareto frontier. Besides deriving how to compute and estimate such gradient, we
will also discuss the non-trivial issue of defining a metric to assess the
quality of the candidate Pareto frontiers. Finally, the properties of the
proposed approach are empirically evaluated on two interesting MOMDPs.
|
['Matteo Pirotta', 'Marcello Restelli', 'Simone Parisi']
|
2014-06-13
| null | null | null | null |
['multi-objective-reinforcement-learning']
|
['methodology']
|
[-3.02292965e-02 6.62514195e-02 -2.58745313e-01 -1.28021568e-01
-9.57288563e-01 -5.39549828e-01 3.27097714e-01 3.21709394e-01
-7.52672136e-01 1.32314575e+00 1.33349448e-02 -3.10124218e-01
-7.01665282e-01 -5.63375175e-01 -7.86468446e-01 -8.32228959e-01
-8.71127993e-02 8.48730087e-01 -2.39451346e-03 -4.75285724e-02
7.41563499e-01 5.22971094e-01 -1.57905746e+00 6.97280914e-02
1.34300888e+00 9.52713788e-01 3.05336595e-01 9.48748171e-01
-4.42782678e-02 1.36626899e-01 -7.74621129e-01 -2.36381069e-01
4.18400019e-01 -4.39292699e-01 -9.93427813e-01 -1.13719799e-01
7.64030442e-02 -1.13997385e-01 4.65227187e-01 1.30368137e+00
4.08707142e-01 5.19767404e-01 9.86091971e-01 -1.32699490e+00
-2.68518835e-01 4.28596526e-01 -4.78228331e-01 1.40102983e-01
2.59895265e-01 2.11070418e-01 1.08301485e+00 -3.86066407e-01
6.12483859e-01 1.51895225e+00 2.99462825e-01 3.22658896e-01
-1.17834318e+00 -2.54974682e-02 1.62076011e-01 3.55158359e-01
-9.78954971e-01 -8.98502544e-02 5.41638970e-01 -3.43163669e-01
5.28293908e-01 1.93564907e-01 8.76449347e-01 8.27403307e-01
5.27894020e-01 9.24508989e-01 1.47528934e+00 -6.17973864e-01
8.68019819e-01 2.21332058e-01 -3.05781275e-01 7.84586310e-01
2.11331174e-01 4.94558245e-01 -2.11627990e-01 -3.80472958e-01
4.60081190e-01 -2.62080699e-01 -6.65639415e-02 -8.01624417e-01
-9.63569880e-01 1.00366724e+00 3.20191294e-01 3.45954835e-01
-9.03729379e-01 1.14028797e-01 -1.06288746e-01 3.02455425e-01
1.41954333e-01 9.53348994e-01 -2.95531094e-01 -2.81784981e-01
-9.53232229e-01 5.45697510e-01 9.58855152e-01 4.50322926e-01
8.88153434e-01 5.49346991e-02 -4.17414695e-01 3.82167459e-01
3.20046395e-01 4.60667133e-01 3.99537295e-01 -1.34157073e+00
3.99883479e-01 4.22660738e-01 9.56468344e-01 -5.93192637e-01
-3.24879646e-01 -5.83519995e-01 -1.87473193e-01 9.54460561e-01
6.64460182e-01 -6.84286773e-01 -5.71939468e-01 1.39797187e+00
5.59535801e-01 -2.38650188e-01 3.50669891e-01 9.73116517e-01
-1.75597817e-01 7.86756277e-01 -4.00815830e-02 -5.37439287e-01
7.71853983e-01 -8.64948988e-01 -4.40407842e-01 1.91002354e-01
3.60361278e-01 -5.75281143e-01 1.21572959e+00 4.30232495e-01
-9.98918474e-01 -2.36011252e-01 -9.93507504e-01 7.91966438e-01
-3.46000284e-01 1.57521561e-01 2.55274922e-01 5.12085021e-01
-9.73978579e-01 1.07678080e+00 -5.27230203e-01 -3.13583106e-01
3.72351795e-01 2.76406676e-01 3.26978177e-01 2.44620383e-01
-9.24292445e-01 1.27082109e+00 9.32277799e-01 -2.03313872e-01
-1.13892376e+00 -4.95329469e-01 -2.71237850e-01 2.18652010e-01
7.27639377e-01 -5.83816409e-01 1.22799957e+00 -1.47889829e+00
-2.06614804e+00 2.06238672e-01 1.05125494e-01 -4.74290341e-01
7.81108618e-01 -2.65275270e-01 -2.26587474e-01 2.89425969e-01
-2.52098199e-02 6.64101481e-01 1.08832610e+00 -1.47672570e+00
-1.18160045e+00 -2.56127089e-01 2.65017062e-01 6.38380885e-01
-3.55918050e-01 -8.19963291e-02 2.51005888e-01 -2.09365949e-01
-5.63599706e-01 -7.14154303e-01 -4.88236606e-01 -4.54023868e-01
-2.21377119e-01 -4.24250394e-01 6.25180185e-01 -3.66178423e-01
1.20172203e+00 -1.64288306e+00 4.26479846e-01 4.45365876e-01
-2.38946870e-01 1.80564776e-01 -2.13250637e-01 3.83094758e-01
4.30198997e-01 -3.58870951e-03 -2.59452701e-01 7.36382827e-02
1.10501267e-01 1.17720291e-01 -2.03090563e-01 5.14630675e-01
-4.98296581e-02 7.43742108e-01 -1.22142100e+00 -4.86170530e-01
-6.01273663e-02 -2.01627523e-01 -4.45574015e-01 5.66223674e-02
-7.38102853e-01 3.51941347e-01 -6.72614217e-01 5.45493007e-01
2.95120716e-01 -2.74181161e-02 2.03461662e-01 2.69690722e-01
-5.41927516e-01 -3.54547918e-01 -1.29905748e+00 1.35346615e+00
-2.47423962e-01 2.09995896e-01 -5.34491576e-02 -1.22355437e+00
9.73497450e-01 1.32860065e-01 6.20522559e-01 -3.36918712e-01
2.64586180e-01 3.81767601e-01 -1.03425629e-01 -2.99000174e-01
6.31425679e-01 -3.41389775e-01 1.40809715e-01 3.07593793e-01
3.81467491e-02 -1.91424653e-01 5.49610436e-01 -4.88802373e-01
6.94005966e-01 6.42212451e-01 3.35164219e-01 -6.60122395e-01
7.12557137e-01 3.36215138e-01 5.43103635e-01 9.34906423e-01
-2.88575202e-01 -3.29573788e-02 6.29852176e-01 -4.91875976e-01
-9.49146569e-01 -1.01005590e+00 2.43591800e-01 1.02249646e+00
2.72332281e-01 1.10992938e-01 -7.55781233e-01 -8.68819416e-01
7.50830621e-02 1.20242238e+00 -5.46653986e-01 -5.60496375e-02
-5.71143985e-01 -6.31977737e-01 8.34852904e-02 6.75961971e-02
3.14681202e-01 -1.08626020e+00 -1.20886278e+00 3.47096413e-01
-7.13498220e-02 -4.00295943e-01 -1.88133657e-01 2.01397300e-01
-1.19649875e+00 -1.01530099e+00 -1.13257945e+00 -5.33015072e-01
5.50605237e-01 -4.18459892e-01 7.93177485e-01 -5.05597711e-01
1.48463091e-02 4.24476504e-01 -1.16600007e-01 -6.19265556e-01
-6.46043539e-01 2.86892969e-02 5.98732978e-02 1.24579996e-01
1.01241522e-01 -4.04208899e-01 -4.66832012e-01 3.12215477e-01
-5.43604851e-01 -3.00348192e-01 7.88914025e-01 7.10947454e-01
8.36200356e-01 2.54420429e-01 7.42544472e-01 -1.84891820e-01
1.08411896e+00 -3.53920698e-01 -1.34493935e+00 6.05403602e-01
-9.05317962e-01 6.96954966e-01 1.11206007e+00 -5.72510481e-01
-1.03636801e+00 8.88809562e-02 3.43894958e-01 -5.83249450e-01
-1.28377154e-01 2.85286635e-01 1.58670098e-01 5.47729544e-02
6.89095616e-01 1.54429570e-01 1.37769386e-01 -4.78469014e-01
3.64737660e-01 3.40302408e-01 3.64676893e-01 -8.26962292e-01
6.23916745e-01 3.36582899e-01 4.16589707e-01 -5.26146531e-01
-7.85087228e-01 -3.39706302e-01 -4.25922632e-01 -5.93097985e-01
7.04164982e-01 -4.23294418e-02 -1.14221585e+00 6.91850632e-02
-8.75370204e-01 -3.22168022e-01 -7.22725272e-01 7.86953211e-01
-1.20238423e+00 1.28750339e-01 1.55241936e-01 -1.18323815e+00
-1.98956296e-01 -9.63360310e-01 5.10428965e-01 6.45440757e-01
2.64525637e-02 -1.07694840e+00 4.69755501e-01 -6.08792566e-02
1.96536660e-01 3.69321376e-01 1.10649848e+00 -6.25018656e-01
-2.55668283e-01 1.38106555e-01 2.69200116e-01 1.73952535e-01
-1.12731732e-01 -2.43967343e-02 -4.29433465e-01 -5.38796902e-01
-2.35650912e-02 -3.49655688e-01 6.27761424e-01 8.38345885e-01
8.13083947e-01 -5.85027695e-01 -3.59820485e-01 3.15819055e-01
1.76891303e+00 5.93292475e-01 3.01591158e-01 7.83837914e-01
-3.20228934e-02 5.61563909e-01 1.13004172e+00 6.31254673e-01
1.55188730e-02 4.43654776e-01 7.21060395e-01 3.91756266e-01
4.63706374e-01 -2.10392624e-01 4.96060103e-01 -6.77017644e-02
-3.06943625e-01 -3.53282690e-01 -8.04206192e-01 4.87128675e-01
-2.03546476e+00 -9.59301651e-01 4.92244154e-01 2.38193727e+00
3.86009604e-01 3.69523764e-02 6.48630619e-01 -2.24966511e-01
8.93170536e-01 -1.75147370e-01 -8.19663584e-01 -8.20618212e-01
-3.11690923e-02 2.14968435e-02 5.82601607e-01 8.26074600e-01
-1.06869328e+00 8.20931435e-01 6.71912909e+00 7.57192194e-01
-9.60679531e-01 -2.20810860e-01 5.00282288e-01 -1.86806455e-01
-3.14116836e-01 1.65765479e-01 -7.86768377e-01 4.03261989e-01
1.01962876e+00 -3.75465125e-01 8.04294467e-01 8.71500313e-01
4.15572882e-01 -4.99178350e-01 -7.39979625e-01 5.22280812e-01
-4.12064821e-01 -1.29996347e+00 1.22474834e-01 2.09324241e-01
1.01794851e+00 -3.24749410e-01 1.55399844e-01 1.76838651e-01
7.15575099e-01 -8.57367396e-01 7.60283470e-01 9.00738239e-01
3.93280029e-01 -1.33140850e+00 5.80400407e-01 4.80937719e-01
-6.88444197e-01 -8.07207167e-01 -5.34323871e-01 9.46312398e-02
3.08050532e-02 3.23910654e-01 -9.55873966e-01 7.09419847e-01
6.15658820e-01 1.03589356e-01 -1.20572150e-01 1.40568697e+00
-9.18504447e-02 3.46573055e-01 -2.50275433e-01 -8.09308648e-01
8.33817959e-01 -6.82065547e-01 1.14875472e+00 7.68059790e-01
5.27964473e-01 -3.20393384e-01 2.59172857e-01 9.02838409e-01
2.67331362e-01 2.92938024e-01 -3.59410733e-01 -2.51759231e-01
8.93781483e-02 1.08063555e+00 -8.06456268e-01 -2.46622443e-01
1.01537600e-01 8.58572841e-01 4.22292173e-01 5.35748541e-01
-6.81427360e-01 -5.45267582e-01 3.74894649e-01 -2.46313408e-01
5.43379724e-01 2.08948962e-02 -5.31522781e-02 -7.54938126e-01
-9.10931826e-02 -7.60132968e-01 5.59311390e-01 -4.15890634e-01
-9.02181923e-01 3.80710900e-01 2.53986955e-01 -1.20213044e+00
-5.54636121e-01 -5.47296524e-01 -5.91459215e-01 7.56841779e-01
-1.36158788e+00 -4.57140595e-01 2.44395480e-01 2.87218273e-01
3.19023907e-01 -3.93393993e-01 6.52427554e-01 -3.60406101e-01
-1.43490300e-01 2.09194720e-01 9.67528343e-01 -5.37888587e-01
2.99682885e-01 -1.53896034e+00 -4.60029334e-01 7.02566326e-01
-2.47747302e-01 1.19526379e-01 1.00231326e+00 -5.61223567e-01
-1.22551072e+00 -7.62599826e-01 4.65432554e-01 -8.94779339e-03
7.05964565e-01 4.41040605e-01 -3.58980387e-01 3.64096135e-01
2.85206676e-01 -5.40986657e-01 2.01214746e-01 -9.67568904e-02
5.39033651e-01 -2.90428817e-01 -1.30621362e+00 7.04395711e-01
5.73136330e-01 3.58986974e-01 -6.21025801e-01 1.89284280e-01
2.80638754e-01 -9.23717320e-02 -7.61125028e-01 2.67647922e-01
5.62679350e-01 -8.44099760e-01 7.52844334e-01 -1.02080107e+00
4.01640356e-01 -4.02311146e-01 -1.05021998e-01 -2.01589990e+00
-4.05179292e-01 -1.08675480e+00 -3.67240816e-01 7.27365792e-01
2.23502442e-01 -7.33629167e-01 7.64726102e-01 -1.94709495e-01
-9.37149525e-02 -1.25213075e+00 -1.18335760e+00 -1.13171172e+00
3.09743464e-01 3.99817705e-01 5.60964227e-01 2.61928290e-01
-1.72766343e-01 1.31681502e-01 -1.91001534e-01 9.08760205e-02
8.69728565e-01 5.53776622e-01 4.44264084e-01 -1.23786843e+00
-4.22508180e-01 -7.60314524e-01 -9.35144499e-02 -7.93150485e-01
1.63508505e-01 -5.99907279e-01 1.56031683e-01 -1.64726985e+00
-1.39709309e-01 -3.99653703e-01 -5.95797479e-01 1.98948495e-02
-1.39111593e-01 -6.13091171e-01 3.85674119e-01 9.21787620e-02
-8.11403513e-01 8.36277783e-01 1.43442225e+00 4.00063209e-02
-5.37964940e-01 5.11489630e-01 -7.53402233e-01 7.71503091e-01
9.99033391e-01 -6.03166223e-01 -3.34009022e-01 -8.19773506e-03
3.58933508e-02 3.75511765e-01 1.52907968e-01 -1.02690172e+00
-9.25171375e-02 -8.64250362e-01 4.52695906e-01 -6.28537536e-01
3.02908659e-01 -7.87386715e-01 -1.64659664e-01 9.15641427e-01
-4.23140049e-01 1.26895979e-01 6.10377342e-02 6.42731786e-01
2.93113161e-02 -7.98297763e-01 1.03552186e+00 -3.43182504e-01
-5.67982495e-01 -1.20263891e-02 -6.59510612e-01 9.88921598e-02
1.35567403e+00 -1.70233667e-01 -3.71676981e-02 -4.02085811e-01
-6.06783509e-01 6.10713243e-01 3.19594055e-01 9.12004709e-02
4.96891439e-01 -1.06425047e+00 -6.68063462e-01 -4.25713539e-01
-3.45404476e-01 -5.06252408e-01 -1.86432019e-01 6.63151324e-01
-3.75855029e-01 5.29399633e-01 -5.75679481e-01 -3.25201809e-01
-8.64067674e-01 7.36464143e-01 6.54879868e-01 -7.32007742e-01
-2.68579602e-01 4.57588494e-01 -3.59093904e-01 -3.34624350e-01
2.64086783e-01 -1.91163257e-01 -3.55040252e-01 1.39676705e-01
3.10736984e-01 9.67140198e-01 -4.47983831e-01 -1.09828137e-01
-1.98821470e-01 4.42954749e-01 2.15305850e-01 -6.75522685e-01
1.36494887e+00 -5.35260364e-02 1.13547847e-01 4.55258012e-01
9.50397670e-01 -2.47075275e-01 -1.51419389e+00 1.01285741e-01
3.85884970e-01 -3.67360324e-01 -3.35150994e-02 -1.14323926e+00
-4.45075035e-01 5.68560839e-01 8.34434867e-01 6.52725995e-02
1.08277094e+00 -3.86328787e-01 3.54522586e-01 7.61657178e-01
2.32481092e-01 -1.83338618e+00 3.04427862e-01 5.04039347e-01
9.29791629e-01 -8.68319452e-01 7.94063434e-02 5.10704577e-01
-8.13949704e-01 1.43482959e+00 4.85449523e-01 -1.95012629e-01
3.19458336e-01 -8.76881480e-02 -7.55031705e-02 3.15439850e-02
-5.47170341e-01 -2.42409796e-01 2.14915156e-01 6.50946975e-01
-2.89605886e-01 2.45100752e-01 -5.31555057e-01 -8.15077573e-02
-1.95850238e-01 5.82140200e-02 3.25801283e-01 1.06040394e+00
-1.02871954e+00 -1.02619660e+00 -7.47687042e-01 3.44410509e-01
-4.02318478e-01 2.73024291e-01 -2.24477485e-01 5.62024891e-01
-2.56553233e-01 9.74042833e-01 -1.25455260e-01 1.11317240e-01
7.91852400e-02 2.29779750e-01 6.66919470e-01 -1.00178033e-01
-4.07878488e-01 -8.66281912e-02 -3.52460556e-02 -5.76065660e-01
-2.77802199e-01 -6.91029847e-01 -1.28339708e+00 -9.68733802e-02
-7.61131058e-03 6.76051378e-01 8.54572356e-01 1.01052022e+00
2.53765225e-01 4.16428000e-01 8.71849298e-01 -7.72356808e-01
-1.21128225e+00 -6.11125231e-01 -4.86219019e-01 -6.01861961e-02
3.24013293e-01 -8.23996663e-01 -1.31109387e-01 -4.97527957e-01]
|
[4.299102306365967, 2.364431858062744]
|
66088790-96db-4124-9f09-3a534bcf867d
|
joint-engagement-classification-using-video
|
2212.14128
| null |
https://arxiv.org/abs/2212.14128v1
|
https://arxiv.org/pdf/2212.14128v1.pdf
|
Joint Engagement Classification using Video Augmentation Techniques for Multi-person Human-robot Interaction
|
Affect understanding capability is essential for social robots to autonomously interact with a group of users in an intuitive and reciprocal way. However, the challenge of multi-person affect understanding comes from not only the accurate perception of each user's affective state (e.g., engagement) but also the recognition of the affect interplay between the members (e.g., joint engagement) that presents as complex, but subtle, nonverbal exchanges between them. Here we present a novel hybrid framework for identifying a parent-child dyad's joint engagement by combining a deep learning framework with various video augmentation techniques. Using a dataset of parent-child dyads reading storybooks together with a social robot at home, we first train RGB frame- and skeleton-based joint engagement recognition models with four video augmentation techniques (General Aug, DeepFake, CutOut, and Mixed) applied datasets to improve joint engagement classification performance. Second, we demonstrate experimental results on the use of trained models in the robot-parent-child interaction context. Third, we introduce a behavior-based metric for evaluating the learned representation of the models to investigate the model interpretability when recognizing joint engagement. This work serves as the first step toward fully unlocking the potential of end-to-end video understanding models pre-trained on large public datasets and augmented with data augmentation and visualization techniques for affect recognition in the multi-person human-robot interaction in the wild.
|
['Hae Won Park', 'Cynthia Breazeal', 'Sharifa Alghowinem', 'Huili Chen', 'Yubin Kim']
|
2022-12-28
| null | null | null | null |
['face-swapping', 'video-understanding']
|
['computer-vision', 'computer-vision']
|
[ 1.80555329e-01 4.56637114e-01 2.25711703e-01 -5.98632693e-01
-3.86437029e-01 -3.15686792e-01 6.87251687e-01 7.97330216e-02
-8.56123120e-02 3.34877998e-01 4.38912004e-01 6.06014132e-01
1.57409146e-01 -3.14328134e-01 -5.44409931e-01 -5.99226773e-01
-3.32933128e-01 6.60708010e-01 -5.85391998e-01 -2.60850847e-01
-2.79986173e-01 3.60750705e-01 -1.74869096e+00 4.95889485e-01
5.15708208e-01 1.06791604e+00 -1.35122761e-01 8.91430914e-01
3.29579115e-01 1.33519995e+00 -4.28319216e-01 -3.76024872e-01
3.28558758e-02 -2.52131015e-01 -1.06799495e+00 3.45979154e-01
2.87129343e-01 -8.56478572e-01 -9.53320935e-02 5.28709531e-01
3.19688052e-01 1.75286397e-01 7.58449018e-01 -1.94972515e+00
-1.70464769e-01 4.58069474e-01 -5.82690597e-01 -3.97306144e-01
1.17809927e+00 3.56187433e-01 8.44759941e-01 -5.19123852e-01
5.57794929e-01 1.69123662e+00 5.39777398e-01 5.35787821e-01
-1.10376525e+00 -6.41279519e-01 1.83347911e-01 4.50603455e-01
-8.32614839e-01 -4.36846524e-01 7.93996692e-01 -7.01626778e-01
1.02037477e+00 5.35600372e-02 1.12094533e+00 1.69271159e+00
-3.75654280e-01 1.06710052e+00 8.39957952e-01 -2.75467038e-01
-3.13461274e-02 -1.37170494e-01 3.48755687e-01 7.89302289e-01
-4.22446787e-01 -1.74742356e-01 -7.40600169e-01 -2.98489749e-01
6.95245087e-01 -1.56464949e-01 -1.52795732e-01 -4.68378782e-01
-1.27962983e+00 5.90219200e-01 3.71538669e-01 2.14589760e-01
-7.85395086e-01 5.19344985e-01 7.94720232e-01 3.72913480e-01
4.74032074e-01 2.99440891e-01 -4.34641421e-01 -8.83887887e-01
-1.05646759e-01 2.02360809e-01 8.31205845e-01 8.48749399e-01
7.63311982e-01 -4.65692490e-01 -4.56300229e-02 8.99599254e-01
3.58730793e-01 3.01673532e-01 2.94611931e-01 -1.25862908e+00
1.43011093e-01 8.25411558e-01 1.67094439e-01 -1.06610894e+00
-7.69152343e-01 4.30795163e-01 -5.35017133e-01 4.07184094e-01
3.80801678e-01 -2.96871096e-01 -5.98131299e-01 2.23957705e+00
7.28970408e-01 1.56722039e-01 3.56775075e-01 9.51114118e-01
8.90329242e-01 5.38189292e-01 3.66337746e-01 -1.55657440e-01
1.48881721e+00 -1.27398765e+00 -7.61620998e-01 -3.19195479e-01
8.65175366e-01 -2.80314714e-01 9.92533863e-01 4.91659582e-01
-1.05140054e+00 -4.92723286e-01 -8.66346180e-01 -7.89112747e-02
-5.72565123e-02 2.54630297e-01 8.23362172e-01 1.89621314e-01
-8.51170182e-01 2.92612106e-01 -9.43594635e-01 -9.78016794e-01
2.83780545e-01 6.41112864e-01 -9.85687494e-01 5.88198453e-02
-8.08058202e-01 1.02792859e+00 -5.09900674e-02 -3.24288793e-02
-1.03384972e+00 -3.03464949e-01 -1.04510748e+00 4.33073984e-03
7.50222653e-02 -5.77489078e-01 1.44596219e+00 -1.77740431e+00
-1.64046586e+00 1.03393936e+00 1.16285697e-01 -1.91289738e-01
4.35960144e-01 -5.79917490e-01 1.04184195e-01 4.44124281e-01
3.11364401e-02 1.07675314e+00 6.96198761e-01 -1.27166891e+00
-3.01971644e-01 -7.21238196e-01 3.26570779e-01 7.37522244e-01
-3.82943630e-01 2.11108640e-01 -2.57100254e-01 -1.29221976e-01
7.17675760e-02 -1.16653645e+00 9.65161473e-02 1.78339869e-01
-1.28110647e-01 -2.44819164e-01 1.02933359e+00 -5.72119713e-01
4.23793048e-01 -2.28701019e+00 6.80036902e-01 -3.15748365e-03
3.19903672e-01 -3.79881226e-02 -3.15570176e-01 5.41633606e-01
-4.05705422e-01 -3.95364165e-01 7.97707066e-02 -8.87145996e-01
1.35222822e-01 1.74166530e-01 2.76199371e-01 5.47307312e-01
2.70748377e-01 6.70583785e-01 -1.03014803e+00 -4.04318780e-01
3.40724856e-01 4.91231203e-01 -3.66961211e-01 9.31788862e-01
-5.70364557e-02 8.16991568e-01 -2.89608181e-01 6.73826575e-01
2.88635820e-01 6.41902536e-02 2.47319862e-01 -1.78653717e-01
3.40563536e-01 -6.64969012e-02 -8.48546088e-01 1.72189867e+00
-5.98768651e-01 6.59233510e-01 5.26656866e-01 -7.11849868e-01
8.00422192e-01 6.69970930e-01 8.12359214e-01 -2.57845968e-01
3.41807544e-01 -2.47522905e-01 -1.45476878e-01 -9.22851980e-01
2.72856265e-01 1.20560126e-02 -2.86607802e-01 7.92069495e-01
2.43897542e-01 -1.00007437e-01 -1.84482560e-01 3.02362144e-01
1.34578359e+00 3.37495774e-01 4.63723361e-01 2.62162268e-01
2.22300127e-01 -2.00560734e-01 2.16737419e-01 2.54608482e-01
-7.43694484e-01 4.84649330e-01 8.30220044e-01 -3.94317955e-01
-6.20959520e-01 -7.78495669e-01 5.74416399e-01 1.57912815e+00
1.37365505e-01 -2.13032469e-01 -7.99251616e-01 -7.39009738e-01
-3.33646566e-01 5.14109313e-01 -9.37676549e-01 -3.79843801e-01
-2.55644172e-01 -3.57126683e-01 7.36758649e-01 4.40836519e-01
4.49475765e-01 -1.36312938e+00 -9.35839176e-01 -1.01637589e-02
-7.03074634e-01 -1.38519228e+00 1.26532242e-01 2.32875094e-01
-3.21201086e-01 -1.24314785e+00 -1.13922209e-01 -7.12212145e-01
6.11993551e-01 1.44458830e-01 1.08465779e+00 1.03937224e-01
-2.47544888e-02 1.39920652e+00 -6.43320620e-01 -3.86225224e-01
-5.49930453e-01 -3.52578163e-01 2.24908605e-01 1.50443614e-01
4.04013455e-01 -8.95207405e-01 -5.06041825e-01 3.29045117e-01
-5.29282808e-01 4.89337444e-01 4.46178287e-01 6.23601973e-01
-1.84885517e-01 -6.22484446e-01 4.46310073e-01 -2.49527141e-01
5.14100611e-01 -7.89667904e-01 3.79643738e-01 1.18954882e-01
1.85388193e-01 -4.63274062e-01 1.39234155e-01 -7.96486318e-01
-1.22087097e+00 5.44476688e-01 4.05494757e-02 -5.47330320e-01
-6.97830677e-01 2.02526808e-01 -2.70317882e-01 2.10562050e-01
5.54270327e-01 -3.90214980e-01 4.80268329e-01 8.73234347e-02
5.21440685e-01 7.27827489e-01 7.08001018e-01 -7.51496613e-01
1.66466653e-01 4.93765652e-01 -2.75730699e-01 -9.03955281e-01
-3.85585636e-01 -5.66258788e-01 -9.85741973e-01 -9.20200348e-01
1.16334844e+00 -1.28638268e+00 -1.15477180e+00 8.36985469e-01
-1.41556323e+00 -6.74146175e-01 3.38575020e-02 3.69218796e-01
-1.05501068e+00 2.63183653e-01 -8.30455065e-01 -1.05249679e+00
-3.92012089e-01 -1.07950270e+00 1.36951983e+00 -2.24735402e-02
-9.37361896e-01 -6.96173728e-01 1.87142462e-01 8.56114149e-01
-9.55137610e-02 6.94545805e-01 6.19668245e-01 -7.03385830e-01
4.64710593e-02 -2.12860480e-01 -1.60492226e-01 1.85521945e-01
1.24953590e-01 1.77677959e-01 -1.10711205e+00 1.29111871e-01
6.99495524e-02 -1.22413290e+00 7.43643045e-02 6.93002343e-02
5.82838416e-01 -3.03842038e-01 -1.99262381e-01 8.88122339e-03
6.09626293e-01 1.84262887e-01 6.97861850e-01 9.15242657e-02
9.47972178e-01 1.03010190e+00 1.09145010e+00 6.90650582e-01
8.91642094e-01 8.28755438e-01 7.85926163e-01 -7.73347169e-02
2.35078484e-01 1.99469831e-02 8.11984360e-01 2.96731114e-01
-7.69032836e-02 -6.73205256e-02 -8.50342989e-01 3.60402614e-01
-2.29853797e+00 -9.03270662e-01 -1.94360241e-01 1.55855846e+00
5.81851423e-01 -3.90869826e-01 4.16197419e-01 5.27542122e-02
4.39579248e-01 -2.60446854e-02 -5.61478496e-01 -7.94141591e-01
2.91278392e-01 -3.54582101e-01 -3.70748341e-01 1.77601010e-01
-1.13201332e+00 9.01527703e-01 5.18768024e+00 1.28496781e-01
-8.92525733e-01 2.75666490e-02 8.73311698e-01 -2.09568635e-01
3.49820584e-01 -5.33878565e-01 7.09175467e-02 -5.44342995e-02
7.25398481e-01 5.42062998e-01 5.47916412e-01 9.40447450e-01
4.50311482e-01 -4.11560267e-01 -1.72294271e+00 1.21910799e+00
2.68958867e-01 -3.25378060e-01 -5.24808764e-01 -2.98366904e-01
2.61901736e-01 -1.23713776e-01 -1.84462756e-01 5.30499637e-01
2.11910620e-01 -1.02779734e+00 7.31923997e-01 3.68454248e-01
5.86187243e-01 -4.39682156e-01 5.89171529e-01 2.79375404e-01
-9.63993669e-01 -2.50388056e-01 3.92981201e-01 -5.96485913e-01
1.08801946e-01 -8.79688486e-02 -1.09954512e+00 2.30427366e-02
8.89812231e-01 8.52273226e-01 -2.87678242e-01 5.07229418e-02
-1.71139568e-01 1.59392297e-01 -4.79563683e-01 -5.75150885e-02
8.21259543e-02 -1.88063562e-01 5.38641393e-01 1.10756588e+00
1.12660155e-01 4.70861733e-01 3.14696953e-02 7.22736299e-01
8.73402804e-02 -2.60300059e-02 -8.67871225e-01 -1.51236147e-01
2.16803581e-01 1.62634420e+00 -4.91247147e-01 -1.90129161e-01
-3.66247743e-01 1.46345723e+00 8.00868869e-01 1.57179222e-01
-1.06550646e+00 4.78373170e-01 1.13611531e+00 -3.47667277e-01
-1.71170294e-01 -2.41003811e-01 -7.05486163e-02 -8.62287164e-01
-5.88792041e-02 -1.06923342e+00 1.41777128e-01 -1.36286986e+00
-1.07789159e+00 3.84989977e-01 4.27246764e-02 -1.09649873e+00
-6.39638484e-01 -4.58831519e-01 -9.04846847e-01 1.88346073e-01
-4.56875086e-01 -1.80862379e+00 -1.00686979e+00 5.42265713e-01
4.80849415e-01 1.92453697e-01 1.11253893e+00 -3.11047509e-02
-7.15412140e-01 2.80906796e-01 -6.34212852e-01 9.94090922e-03
7.53107011e-01 -1.26159799e+00 -6.28547966e-02 2.32892632e-01
-1.27416179e-01 9.91032869e-02 7.93029547e-01 -4.43037093e-01
-1.47850633e+00 -7.71086752e-01 1.38185114e-01 -5.36357641e-01
6.24794602e-01 -4.93968129e-01 -6.69142187e-01 1.07860756e+00
3.53461534e-01 -3.60217482e-01 7.88023233e-01 4.08592582e-01
-2.21048430e-01 2.12279871e-01 -1.50463510e+00 7.39828646e-01
1.16016781e+00 -4.43544686e-01 -5.33685088e-01 2.24126533e-01
6.71446443e-01 -3.28033179e-01 -8.96273375e-01 3.96829993e-01
9.62565601e-01 -1.35749543e+00 7.79694259e-01 -3.74332428e-01
9.03089583e-01 1.27939329e-01 -2.22148299e-01 -1.39873946e+00
3.20600881e-03 -4.68399316e-01 -9.30717364e-02 1.49783528e+00
-1.61023483e-01 -1.83965102e-01 6.65930688e-01 1.13029206e+00
-1.37329027e-02 -7.76544273e-01 -6.80317521e-01 -1.62783952e-03
-3.68755221e-01 -6.33633971e-01 3.75341445e-01 9.98498857e-01
7.43129075e-01 6.15670264e-01 -4.95755464e-01 1.29802793e-01
1.65812582e-01 -4.05530304e-01 1.37072217e+00 -1.09815216e+00
-2.40231633e-01 -2.33722150e-01 -8.08093667e-01 -6.55551255e-01
4.70174134e-01 -4.15608495e-01 2.02303916e-01 -1.39380193e+00
4.89428461e-01 -2.65239831e-02 2.26011306e-01 5.89493454e-01
-7.38449469e-02 3.69395949e-02 1.67875767e-01 -1.53898485e-02
-8.75148177e-01 8.90644670e-01 8.74247730e-01 3.17152068e-02
-2.55054563e-01 -2.58568525e-01 -3.95996928e-01 1.08122134e+00
6.26886785e-01 -7.69341514e-02 -3.49426538e-01 -7.58311152e-02
-2.29246374e-02 3.84117305e-01 6.00504339e-01 -1.05520272e+00
3.63535527e-03 -9.91576165e-03 2.67973930e-01 -3.41718435e-01
9.53420877e-01 -1.02198517e+00 1.94880098e-01 2.24097282e-01
-3.23246807e-01 -1.32108271e-01 2.09924251e-01 3.91357362e-01
-7.60685056e-02 1.05615191e-01 6.54878795e-01 -2.42358800e-02
-6.64596796e-01 -1.25886099e-02 -6.73750937e-01 -4.91302073e-01
1.46602213e+00 -9.93316174e-02 -1.26320153e-01 -1.13896632e+00
-9.16573286e-01 5.26600242e-01 5.92507243e-01 5.30124247e-01
5.74669838e-01 -1.26432514e+00 -5.56380987e-01 8.00699368e-02
4.12378877e-01 -5.55688813e-02 3.70271266e-01 9.51407135e-01
-3.69527876e-01 -2.99376667e-01 -5.45477986e-01 -7.64092445e-01
-1.89270151e+00 1.51051402e-01 4.29860145e-01 -1.68249279e-01
-2.71702051e-01 9.10868168e-01 5.39587975e-01 -8.13127697e-01
4.67237949e-01 -2.12401941e-01 -4.21351433e-01 3.13549489e-01
3.37937713e-01 5.24462640e-01 -4.63751614e-01 -1.01885116e+00
-1.97542951e-01 2.28043392e-01 2.44422913e-01 -1.46784171e-01
1.36235535e+00 -4.14908558e-01 -3.24451268e-01 7.34176993e-01
1.31434262e+00 -6.36603296e-01 -1.27998495e+00 1.45620093e-01
-1.66261747e-01 -1.53612942e-01 -4.84661490e-01 -8.68228912e-01
-8.09432447e-01 6.90915346e-01 6.66272044e-01 1.37021080e-01
1.19475806e+00 3.00121754e-01 5.07704258e-01 5.51178694e-01
3.90298396e-01 -1.05080831e+00 8.44027877e-01 4.55066919e-01
1.31314015e+00 -1.67165625e+00 -2.13421494e-01 -4.48116213e-01
-1.08420193e+00 1.07359338e+00 1.21036112e+00 2.47841969e-01
3.07779491e-01 1.04502395e-01 2.44505271e-01 -5.31976759e-01
-1.01484430e+00 -9.91250649e-02 -1.41782925e-01 7.02361941e-01
4.55017507e-01 2.62698740e-01 3.09181154e-01 7.79936492e-01
-1.98411956e-01 -1.42441630e-01 6.66008651e-01 7.73017466e-01
-1.07646219e-01 -6.32977366e-01 -4.30979133e-01 3.47385019e-01
-1.59855112e-01 5.05046248e-01 -8.22140992e-01 7.18589962e-01
1.30211696e-01 1.30549300e+00 2.38930434e-01 -6.43267334e-01
4.32531416e-01 1.40349999e-01 5.83205938e-01 -6.08368993e-01
-7.99215317e-01 -1.67109653e-01 5.33613384e-01 -9.45024967e-01
-7.84761250e-01 -1.02988803e+00 -1.47343087e+00 -2.32089907e-01
3.35892402e-02 -3.84749651e-01 6.71937525e-01 1.08093655e+00
1.65118709e-01 2.49593616e-01 6.93713844e-01 -1.61868596e+00
2.69187316e-02 -1.43476677e+00 -3.27859491e-01 8.93630087e-01
2.32433990e-01 -7.95113444e-01 -3.89416426e-01 -2.25998685e-02]
|
[13.407364845275879, 2.2124762535095215]
|
bcdedc75-547f-40fa-a6bf-2dfe0af15b5c
|
question-rewriting-for-conversational
|
2004.14652
| null |
https://arxiv.org/abs/2004.14652v3
|
https://arxiv.org/pdf/2004.14652v3.pdf
|
Question Rewriting for Conversational Question Answering
|
Conversational question answering (QA) requires the ability to correctly interpret a question in the context of previous conversation turns. We address the conversational QA task by decomposing it into question rewriting and question answering subtasks. The question rewriting (QR) subtask is specifically designed to reformulate ambiguous questions, which depend on the conversational context, into unambiguous questions that can be correctly interpreted outside of the conversational context. We introduce a conversational QA architecture that sets the new state of the art on the TREC CAsT 2019 passage retrieval dataset. Moreover, we show that the same QR model improves QA performance on the QuAC dataset with respect to answer span extraction, which is the next step in QA after passage retrieval. Our evaluation results indicate that the QR model we proposed achieves near human-level performance on both datasets and the gap in performance on the end-to-end conversational QA task is attributed mostly to the errors in QA.
|
['Zhucheng Tu', 'Raviteja Anantha', 'Svitlana Vakulenko', 'Shayne Longpre']
|
2020-04-30
| null | null | null | null |
['question-rewriting']
|
['natural-language-processing']
|
[ 3.01294953e-01 4.72341001e-01 6.94807172e-01 -4.61140692e-01
-1.76576495e+00 -1.10819530e+00 9.85381901e-01 6.08199872e-02
-3.78123224e-01 8.92204046e-01 9.72044766e-01 -7.30644405e-01
-1.04531296e-01 -5.50309539e-01 -5.18033981e-01 -1.05815284e-01
3.09163719e-01 9.14828479e-01 3.63591403e-01 -1.09035349e+00
1.97633058e-01 -2.23207027e-01 -1.33589685e+00 1.16917467e+00
1.06763768e+00 7.86879599e-01 -5.97204864e-02 1.42700040e+00
-4.12409425e-01 1.47488761e+00 -9.97214854e-01 -8.09404314e-01
-2.25843772e-01 -8.71346712e-01 -2.09746122e+00 -3.47717524e-01
4.83655661e-01 -2.64919370e-01 -1.31152123e-01 3.59962851e-01
3.78960371e-01 2.01546133e-01 4.17138308e-01 -9.42900896e-01
-6.71010137e-01 4.46891665e-01 6.35742605e-01 3.01335067e-01
1.29262900e+00 9.31437612e-02 1.41237700e+00 -7.55384028e-01
6.01005137e-01 1.36402559e+00 3.53469998e-01 8.22508693e-01
-7.70545721e-01 1.56944379e-01 -1.03218695e-02 6.87938273e-01
-7.54208505e-01 -4.84371394e-01 3.13595533e-01 -2.25972772e-01
1.30960786e+00 8.74221563e-01 1.95154458e-01 8.35649014e-01
1.20861322e-01 7.93286145e-01 9.16031778e-01 -5.64098835e-01
1.82306394e-01 -3.56526673e-01 7.66274393e-01 3.63657027e-01
-5.63800693e-01 -3.41801077e-01 -4.08684701e-01 -3.72078747e-01
-2.68785030e-01 -6.97093785e-01 -6.47313058e-01 1.59354925e-01
-1.05419683e+00 8.75224471e-01 2.15119153e-01 2.20750049e-01
-2.86127746e-01 -7.46175498e-02 5.60392499e-01 8.85672748e-01
2.61226147e-02 1.03987896e+00 -5.60577154e-01 -6.22457504e-01
-2.86074638e-01 7.02740490e-01 1.65732837e+00 6.52808964e-01
3.38432640e-01 -9.74552929e-01 -6.68997705e-01 9.63959694e-01
5.06829098e-02 5.81986845e-01 2.29759961e-01 -1.48922348e+00
9.16741729e-01 7.65922964e-01 5.77715814e-01 -5.77708006e-01
-2.45717123e-01 -2.12419149e-03 -1.84668764e-01 -5.99405646e-01
7.49094665e-01 -2.85317302e-01 -5.12089491e-01 1.52301085e+00
3.54032964e-01 -4.22624141e-01 6.61457360e-01 7.48893976e-01
1.46411467e+00 8.95648301e-01 3.76945734e-02 -9.45898369e-02
1.96630263e+00 -1.28817928e+00 -8.67594361e-01 -2.70532161e-01
6.89698040e-01 -9.14874196e-01 1.31045079e+00 1.12249367e-01
-1.12704480e+00 -3.06521386e-01 -8.21890891e-01 -7.92734504e-01
-1.84841797e-01 -2.60900706e-01 9.96533856e-02 6.35173261e-01
-1.19716120e+00 -1.20398030e-01 -1.46912843e-01 -2.99781740e-01
-3.16748917e-01 -6.47962233e-03 -6.30576313e-02 -4.50432301e-01
-1.78878784e+00 1.17100632e+00 -1.90633148e-01 1.87353998e-01
-7.65792072e-01 -5.04417777e-01 -9.07534003e-01 2.44145483e-01
4.10313487e-01 -1.08343542e+00 2.19916058e+00 -5.90160191e-01
-1.72642982e+00 7.15075135e-01 -6.53214574e-01 -5.31540811e-01
2.89423645e-01 -4.35570955e-01 -3.43979836e-01 5.53457260e-01
1.94370136e-01 5.04389822e-01 3.60271066e-01 -1.12409222e+00
-7.25011826e-01 -2.35414609e-01 9.99139547e-01 7.09185958e-01
4.72259194e-01 3.03847846e-02 -4.70817089e-01 2.24534601e-01
-9.04129222e-02 -9.48970497e-01 2.42588833e-01 -7.56165683e-01
-5.06754927e-02 -7.65484452e-01 5.79568863e-01 -1.14477479e+00
1.35469317e+00 -1.58627653e+00 1.23657785e-01 -2.27805212e-01
3.95082682e-02 1.78097010e-01 -3.46869349e-01 1.03928602e+00
1.78352952e-01 2.50226073e-02 -4.07366902e-01 -3.52297097e-01
1.87266707e-01 3.84368747e-01 -7.64744878e-01 -3.04413915e-01
1.72305405e-01 1.30753636e+00 -9.93309438e-01 -2.67152935e-01
-1.70929298e-01 -6.56798705e-02 -4.56279963e-01 6.37424827e-01
-7.13807940e-01 5.08903027e-01 -3.99688274e-01 2.82068282e-01
3.22634876e-01 -2.89827079e-01 4.25854139e-02 2.24993169e-01
2.33166799e-01 1.21981108e+00 -3.97973061e-01 1.75146759e+00
-6.50825679e-01 6.92148685e-01 1.97856709e-01 -4.86704081e-01
4.64278519e-01 7.25938976e-01 -1.49968743e-01 -1.02787089e+00
-2.32265428e-01 1.06901981e-01 1.08384505e-01 -8.19805682e-01
8.94298851e-01 -6.58627898e-02 -5.04205704e-01 7.34572411e-01
-1.66050822e-01 -6.32762969e-01 3.37915838e-01 5.43386519e-01
1.30517471e+00 -3.56705606e-01 9.18418095e-02 -1.38173550e-01
1.19203866e+00 4.51728016e-01 -1.25092179e-01 1.08236825e+00
-2.52656281e-01 6.36024654e-01 7.55060852e-01 -2.76904941e-01
-6.18446827e-01 -1.09098601e+00 2.19499946e-01 1.34496665e+00
6.72476813e-02 -4.72983152e-01 -1.00701344e+00 -9.50569451e-01
-3.82833183e-01 1.06270087e+00 -5.81829011e-01 1.14054330e-01
-8.76391053e-01 -2.24509135e-01 8.18398714e-01 5.93773834e-03
6.11782074e-01 -1.12465847e+00 -3.36469084e-01 2.40674317e-01
-1.33816099e+00 -1.26808274e+00 -4.91979182e-01 -5.40289164e-01
-3.81626785e-01 -1.53607678e+00 -5.82865357e-01 -7.64877319e-01
2.86978018e-02 2.23665982e-01 1.74033785e+00 2.90425986e-01
3.32369864e-01 1.13793910e+00 -8.72929871e-01 -3.19848806e-01
-8.24610651e-01 3.77159327e-01 -7.97664285e-01 -2.89120078e-01
4.05878097e-01 -1.65002584e-01 -8.63709569e-01 4.72752392e-01
-8.71984541e-01 -4.10078056e-02 1.47779323e-02 8.05407226e-01
-2.14156866e-01 -8.23703945e-01 1.02659261e+00 -7.22733021e-01
1.45529914e+00 -3.90302509e-01 -6.80391937e-02 7.64764786e-01
-1.38639212e-01 1.64931625e-01 4.52945441e-01 3.07763159e-01
-1.54557025e+00 -5.33255637e-01 -6.75891042e-01 8.12958062e-01
-3.95618007e-03 5.68916380e-01 -3.70242782e-02 2.59732425e-01
8.46924067e-01 2.82014132e-01 -1.38381757e-02 -1.62800252e-01
7.04914391e-01 9.62585628e-01 6.95848465e-01 -8.37401688e-01
2.22812131e-01 1.79138944e-01 -4.13154602e-01 -6.60564780e-01
-1.19886935e+00 -7.36102641e-01 -2.54882097e-01 -4.01131988e-01
1.07462680e+00 -7.69257247e-01 -1.19209933e+00 9.98881757e-02
-1.55634236e+00 -2.83336103e-01 -2.32068658e-01 -1.38429597e-01
-6.91258073e-01 8.09844196e-01 -7.26070583e-01 -8.06947708e-01
-8.77772272e-01 -1.01780689e+00 1.03675020e+00 1.79294810e-01
-6.59653783e-01 -9.02585089e-01 4.34438616e-01 1.47013807e+00
3.47448081e-01 -1.84152842e-01 1.27425098e+00 -8.45993340e-01
-4.86442715e-01 -2.80392412e-02 -2.40303520e-02 4.48719978e-01
-1.44548610e-01 -5.32246768e-01 -9.82466996e-01 2.09281649e-02
2.02408463e-01 -6.69811189e-01 5.65849185e-01 -3.26405883e-01
5.02090394e-01 -3.56955111e-01 3.31857353e-01 -4.36483175e-01
7.27054536e-01 2.13257805e-01 8.82185578e-01 1.82519689e-01
1.26065835e-01 8.56790781e-01 4.91609126e-01 -2.62306064e-01
1.19616842e+00 5.56430161e-01 1.60843760e-01 4.93026704e-01
-2.32192636e-01 -1.62646025e-01 3.34858656e-01 1.06253040e+00
2.80308634e-01 -4.59548801e-01 -1.07452977e+00 8.49066377e-01
-2.06069183e+00 -9.86776710e-01 -4.61790681e-01 1.89796853e+00
1.01034594e+00 -2.53513724e-01 -1.94398522e-01 -1.21418521e-01
2.40979955e-01 9.23328996e-02 -1.29398540e-01 -1.03677559e+00
-7.57114217e-02 4.36120123e-01 -5.41149676e-01 1.20103526e+00
-6.34128451e-01 7.74239480e-01 6.71671963e+00 4.76795584e-01
-3.09090644e-01 1.86854586e-01 3.70257169e-01 2.82008857e-01
-5.66170633e-01 1.21242613e-01 -4.35556203e-01 8.49916041e-02
1.20750988e+00 -6.23005927e-02 4.74846780e-01 2.70785749e-01
-1.02435229e-02 -5.07788777e-01 -1.36757636e+00 5.71938992e-01
3.28940451e-01 -1.28047931e+00 3.58872116e-01 -5.59747696e-01
5.53516388e-01 -1.18270576e-01 -3.36351216e-01 9.94449973e-01
4.25502062e-01 -1.15002203e+00 1.94080308e-01 5.39840877e-01
1.78488165e-01 -6.14414632e-01 1.12849975e+00 5.36700010e-01
-7.80206203e-01 -1.76752776e-01 9.13662985e-02 -4.51856673e-01
4.94913280e-01 9.67941154e-03 -1.21243644e+00 8.18909824e-01
4.76579696e-01 -1.93673000e-01 -6.43983603e-01 7.05424368e-01
-6.68273389e-01 7.59978056e-01 -7.64324665e-02 -3.36937517e-01
4.65161294e-01 -1.57765433e-01 6.58360422e-01 1.10742617e+00
-1.09954119e-01 6.86365247e-01 -2.06810683e-01 3.29229176e-01
-3.07538927e-01 1.48237646e-02 -8.10524374e-02 3.23048711e-01
4.24364150e-01 9.28580761e-01 2.60410160e-01 -5.14036238e-01
-2.80289084e-01 1.11592162e+00 4.04640228e-01 4.82412308e-01
-3.75053138e-01 -4.21954751e-01 5.17861784e-01 -4.11982566e-01
1.57552421e-01 -1.23623244e-01 -8.65372196e-02 -1.15808272e+00
5.02602994e-01 -1.45974994e+00 8.47611248e-01 -1.13695741e+00
-1.29490995e+00 7.79156327e-01 -1.23145051e-01 -6.33230686e-01
-9.38248158e-01 -3.34514022e-01 -6.43771172e-01 1.11278284e+00
-1.60892832e+00 -9.17165816e-01 -3.48291934e-01 6.37989104e-01
8.25507343e-01 4.46853906e-01 1.12378252e+00 9.08662900e-02
1.23007514e-01 3.95025223e-01 -1.10784434e-01 2.23398507e-01
7.49486327e-01 -1.38163531e+00 5.23225069e-01 6.33288443e-01
-1.15553187e-02 7.02272773e-01 7.96229303e-01 -2.35744372e-01
-1.37907672e+00 -5.40646851e-01 1.66409385e+00 -1.38187158e+00
5.37329495e-01 -2.86105961e-01 -1.07843578e+00 5.78903258e-01
7.65695333e-01 -7.99531221e-01 8.32312465e-01 3.74089688e-01
-4.12178755e-01 1.17253192e-01 -9.62922096e-01 6.58692598e-01
6.91342473e-01 -1.22727287e+00 -1.73299050e+00 4.56211060e-01
1.16747439e+00 -6.58786833e-01 -7.79215276e-01 3.59142512e-01
3.67633492e-01 -7.90990412e-01 9.47384655e-01 -9.25361395e-01
5.42367339e-01 -3.41833562e-01 -1.35276601e-01 -1.15375459e+00
2.54913598e-01 -7.00238168e-01 -8.92124549e-02 9.97461200e-01
8.50118041e-01 -4.36883867e-01 4.68339950e-01 1.01160979e+00
-2.14344308e-01 -4.87923622e-01 -1.36808050e+00 -2.41064727e-01
3.66998613e-01 -4.34778273e-01 6.45307481e-01 5.88537455e-01
4.41929936e-01 1.11002457e+00 1.02632701e-01 8.76368061e-02
-4.18452397e-02 3.97446841e-01 9.04851019e-01 -7.75690973e-01
-2.04509810e-01 -1.36701033e-01 3.23934764e-01 -1.67233634e+00
7.04574659e-02 -6.14803851e-01 4.38572168e-01 -2.11406302e+00
-6.87676296e-02 2.12866366e-01 4.42108393e-01 -3.94077152e-02
-6.12698138e-01 -1.89664379e-01 3.25420529e-01 8.62584785e-02
-1.09182656e+00 6.97103322e-01 1.41968918e+00 -1.43329963e-01
-1.70480862e-01 2.34433040e-01 -9.46155250e-01 3.05551678e-01
5.06036997e-01 -1.35860192e-02 -5.57357371e-01 -7.66572416e-01
5.97387612e-01 7.39739537e-01 2.73524344e-01 -6.31742299e-01
5.19094050e-01 1.74498513e-01 -3.68470758e-01 -6.23126328e-01
5.50539613e-01 -3.64113659e-01 -5.16840875e-01 1.89908639e-01
-7.95650125e-01 3.78168702e-01 1.61280081e-01 5.38998485e-01
-6.09431446e-01 -2.93525606e-01 1.12877592e-01 -1.64852545e-01
-4.17467773e-01 -5.03702402e-01 -8.42450261e-01 7.18406439e-01
4.19300854e-01 1.78664431e-01 -8.27040851e-01 -1.24203992e+00
-8.73655140e-01 9.64719832e-01 -2.67684162e-01 6.66712940e-01
3.89506280e-01 -8.93596768e-01 -1.11015868e+00 -6.35833204e-01
4.00032818e-01 7.55854025e-02 4.52803522e-01 4.80933815e-01
-5.13440371e-01 1.14295959e+00 3.92793864e-01 -5.05499363e-01
-1.38726032e+00 3.24055925e-02 6.31432652e-01 -8.51182461e-01
9.10308771e-03 7.85041809e-01 -8.91100317e-02 -1.04882467e+00
3.77342626e-02 -4.42577094e-01 -6.29325628e-01 8.84721950e-02
8.36585104e-01 3.25395554e-01 5.29160321e-01 -3.51660520e-01
-2.43214861e-01 1.85218543e-01 -1.72524661e-01 -5.73212147e-01
5.76822758e-01 -6.46000683e-01 -3.82147223e-01 2.52606153e-01
1.14008760e+00 1.52609101e-03 -5.08598387e-01 -3.67297113e-01
2.72500068e-01 -6.11076178e-03 -6.19684458e-01 -1.59740150e+00
1.36064202e-01 7.61622190e-01 1.63350999e-02 6.53699696e-01
8.43505681e-01 2.23201677e-01 1.15610254e+00 1.10903597e+00
1.84750855e-01 -9.33367670e-01 8.08392540e-02 1.44930995e+00
1.52621567e+00 -1.15851939e+00 -3.50237221e-01 -4.46124315e-01
-7.34105468e-01 9.41333115e-01 4.70620006e-01 2.83435017e-01
2.64390651e-03 -5.28511167e-01 4.99298692e-01 -4.57868308e-01
-1.06117988e+00 -2.11744085e-01 5.59328496e-01 3.14991415e-01
3.32472444e-01 -1.07425135e-02 -5.57324529e-01 6.63868010e-01
-7.60570347e-01 -2.50919491e-01 4.87354815e-01 8.99479091e-01
-4.69653815e-01 -1.00630248e+00 -3.52463424e-01 3.31654958e-02
-4.49566960e-01 -3.83707494e-01 -1.10549045e+00 6.44156098e-01
-5.61096251e-01 2.00121307e+00 5.45348637e-02 -4.18466963e-02
7.61832058e-01 6.03055298e-01 3.65505666e-01 -5.61292410e-01
-1.33931971e+00 -9.09984887e-01 1.19043660e+00 -5.32010496e-01
-3.07667017e-01 -4.13784057e-01 -1.15594220e+00 -1.85442880e-01
-8.36135149e-02 1.01153588e+00 3.82414877e-01 1.48994172e+00
6.47261143e-01 2.43527830e-01 5.00075519e-01 2.23282099e-01
-6.95851684e-01 -1.10584128e+00 4.55145180e-01 6.28609002e-01
7.11825907e-01 2.11282000e-01 -4.29172367e-01 -5.76952398e-02]
|
[11.973424911499023, 7.991094589233398]
|
e8acfa73-d4d0-47b3-a797-38553418e8e0
|
two-languages-are-better-than-one-bilingual
| null | null |
https://aclanthology.org/2022.coling-1.176
|
https://aclanthology.org/2022.coling-1.176.pdf
|
Two Languages Are Better than One: Bilingual Enhancement for Chinese Named Entity Recognition
|
Chinese Named Entity Recognition (NER) has continued to attract research attention. However, most existing studies only explore the internal features of the Chinese language but neglect other lingual modal features. Actually, as another modal knowledge of the Chinese language, English contains rich prompts about entities that can potentially be applied to improve the performance of Chinese NER. Therefore, in this study, we explore the bilingual enhancement for Chinese NER and propose a unified bilingual interaction module called the Adapted Cross-Transformers with Global Sparse Attention (ACT-S) to capture the interaction of bilingual information. We utilize a model built upon several different ACT-Ss to integrate the rich English information into the Chinese representation. Moreover, our model can learn the interaction of information between bilinguals (inter-features) and the dependency information within Chinese (intra-features). Compared with existing Chinese NER methods, our proposed model can better handle entities with complex structures. The English text that enhances the model is automatically generated by machine translation, avoiding high labour costs. Experimental results on four well-known benchmark datasets demonstrate the effectiveness and robustness of our proposed model.
|
['Jian Wang', 'Hongfei Lin', 'Yuanyuan Sun', 'Zhizheng Wang', 'Zhihao Yang', 'Jinzhong Ning']
| null | null | null | null |
coling-2022-10
|
['chinese-named-entity-recognition']
|
['natural-language-processing']
|
[-4.16425675e-01 -3.65090102e-01 -1.31653339e-01 -4.73449260e-01
-6.18094325e-01 -6.07858658e-01 5.51465392e-01 -1.67212814e-01
-8.36089194e-01 8.08988750e-01 8.19355249e-01 -3.20577651e-01
3.44554514e-01 -7.58567870e-01 -4.98319536e-01 -4.63483781e-01
4.50324655e-01 1.51611552e-01 -9.73545983e-02 -4.68164414e-01
1.59960493e-01 2.37297282e-01 -7.55047262e-01 2.29081467e-01
1.68207943e+00 5.11756241e-01 5.37000299e-01 -2.12695509e-01
-6.11528635e-01 7.74754047e-01 -4.24204201e-01 -6.02822483e-01
-1.14017297e-02 -5.16474366e-01 -9.32637513e-01 -2.93314397e-01
-3.37037385e-01 -1.23452194e-01 -3.63344848e-01 1.30125105e+00
6.57084167e-01 9.42013264e-02 3.04181695e-01 -6.01839244e-01
-1.04930890e+00 1.18778002e+00 -4.39047217e-01 2.59888083e-01
2.57705480e-01 -2.02275783e-01 8.07240069e-01 -1.18501520e+00
7.40624309e-01 9.37770486e-01 6.08769119e-01 3.88656408e-01
-5.10825396e-01 -8.96268427e-01 3.33442181e-01 4.11789119e-01
-1.56938767e+00 -2.40179136e-01 6.32330894e-01 8.35064352e-02
9.32640254e-01 2.81937979e-02 2.99424022e-01 8.23077977e-01
-2.37787683e-02 9.64815736e-01 1.07058048e+00 -5.08412004e-01
-3.21809232e-01 4.30144846e-01 2.05692515e-01 4.46211219e-01
2.01005727e-01 -1.23249799e-01 -1.65225253e-01 3.49234015e-01
6.75218523e-01 1.03401616e-01 -4.16126311e-01 3.09857666e-01
-1.42481005e+00 6.10090137e-01 5.89377701e-01 1.11773574e+00
-4.32449341e-01 -4.49030459e-01 4.77100343e-01 3.28427479e-02
3.75708312e-01 4.23383296e-01 -8.77534449e-01 -1.69979244e-01
-5.71235776e-01 -4.93817836e-01 8.32045674e-01 1.45253587e+00
9.20039654e-01 9.33500451e-06 -1.81519926e-01 8.45475376e-01
2.09616452e-01 5.96770108e-01 8.34985077e-01 -2.13769078e-01
8.69666874e-01 8.91364932e-01 -1.46185318e-02 -9.22077179e-01
-4.33966875e-01 -6.50175512e-01 -1.03504276e+00 -9.12435353e-01
-1.17498681e-01 -7.58241415e-01 -5.83044827e-01 1.73684955e+00
2.20414519e-01 7.18844980e-02 3.08039874e-01 7.94357419e-01
1.02015972e+00 8.14023972e-01 2.83694208e-01 -8.85997117e-02
1.41847730e+00 -1.21549606e+00 -1.07287300e+00 -3.14364702e-01
1.01816511e+00 -1.00583982e+00 7.55442739e-01 -2.78184503e-01
-6.07321560e-01 -6.32014871e-01 -7.28498101e-01 -2.54755646e-01
-8.04851413e-01 7.43948102e-01 7.49329329e-01 6.20855272e-01
-7.17890859e-01 2.55067647e-01 -7.63456583e-01 -4.34269130e-01
4.65383567e-02 5.05001962e-01 -6.15746140e-01 -3.84208500e-01
-1.81730020e+00 1.06442297e+00 8.74350905e-01 7.27223396e-01
-3.00668180e-01 -3.81991506e-01 -1.11641777e+00 2.23224282e-01
4.37957048e-01 -2.94025689e-01 7.78992116e-01 -9.06299174e-01
-1.29289412e+00 3.07444781e-01 -3.96531940e-01 1.29571423e-01
1.04129221e-02 -5.80948234e-01 -8.71394396e-01 -1.75409198e-01
3.59144509e-01 3.81539792e-01 -1.12850122e-01 -9.73726451e-01
-6.19028091e-01 -2.70429492e-01 4.70721759e-02 6.03334427e-01
-5.59326887e-01 5.00605464e-01 -8.31499755e-01 -7.28768349e-01
-8.63239765e-02 -6.58340037e-01 -4.38083082e-01 -9.32307303e-01
-5.40425897e-01 -2.16240063e-01 4.83075917e-01 -1.04429984e+00
1.68321955e+00 -2.01448250e+00 -3.95821594e-02 5.11894515e-03
-1.82632923e-01 4.62482065e-01 -3.72450978e-01 7.65539408e-01
-3.27150859e-02 5.43268144e-01 -2.48528764e-01 2.18282826e-03
-9.54649448e-02 9.94021595e-02 8.80531743e-02 1.28821865e-01
4.83715683e-01 1.16382742e+00 -8.77552688e-01 -7.09457219e-01
1.14088878e-02 5.98018527e-01 -4.09335673e-01 2.93472320e-01
4.38866258e-01 6.50100350e-01 -9.20242250e-01 4.87274617e-01
9.93522346e-01 -1.42808408e-01 3.95893693e-01 -4.28485364e-01
-5.38486481e-01 4.67439622e-01 -1.06886709e+00 1.74413514e+00
-6.44695878e-01 4.94577773e-02 -1.32944241e-01 -7.59050488e-01
8.96367610e-01 4.53592420e-01 1.29534319e-01 -9.05692816e-01
2.38004401e-01 3.63455653e-01 3.60071249e-02 -4.98346448e-01
5.39375663e-01 -6.21627159e-02 -1.56885728e-01 2.86710829e-01
2.89563566e-01 5.31764805e-01 3.05813760e-01 2.36596227e-01
7.56320298e-01 2.99403876e-01 5.56229770e-01 -4.04305726e-01
9.59705412e-01 3.94010963e-03 9.90267396e-01 3.65155965e-01
-8.16952530e-03 1.92781582e-01 1.70042902e-01 -1.82177931e-01
-7.87737131e-01 -3.56255800e-01 -1.60830885e-01 1.03564215e+00
4.20292795e-01 -5.69839537e-01 -9.83754694e-01 -1.08479142e+00
-7.11693525e-01 5.87668419e-01 -4.20321196e-01 -2.67636329e-02
-9.94099081e-01 -9.22653317e-01 6.80410624e-01 8.47609699e-01
9.70564783e-01 -1.38282526e+00 1.64520040e-01 3.28371078e-01
-6.89327180e-01 -1.36529624e+00 -1.01274788e+00 1.17043473e-01
-7.04797924e-01 -7.84540057e-01 -7.76895046e-01 -1.24347198e+00
8.59688044e-01 2.00131655e-01 9.87087607e-01 1.73131466e-01
2.35026911e-01 9.85511094e-02 -8.53175402e-01 -1.19229980e-01
2.99692247e-02 5.12765706e-01 -1.38336763e-01 7.44690076e-02
8.99738371e-01 -7.52052814e-02 -3.35729480e-01 2.87613273e-01
-9.71714497e-01 1.56698301e-01 1.15531433e+00 1.07971466e+00
3.83608609e-01 1.53292239e-01 4.86750096e-01 -1.17023766e+00
4.76926833e-01 -7.73477554e-01 -1.85518190e-01 6.34605944e-01
-3.18936437e-01 3.45895141e-01 9.17349696e-01 -2.54666448e-01
-1.82176018e+00 -3.49344835e-02 -3.00521582e-01 2.23245651e-01
-2.78319746e-01 1.11173773e+00 -8.94175231e-01 9.39422920e-02
1.70020182e-02 5.81123710e-01 -8.12765479e-01 -7.70026684e-01
4.60177809e-01 7.78098106e-01 2.62801290e-01 -8.65930736e-01
6.22101188e-01 5.01872003e-02 -4.84728575e-01 -6.49070442e-01
-8.97054672e-01 -5.05174339e-01 -1.10410750e+00 2.41491556e-01
1.03568792e+00 -1.23430157e+00 -3.92620981e-01 4.64950979e-01
-1.48530126e+00 1.99138522e-01 3.16443115e-01 8.53579462e-01
1.96924403e-01 4.89818007e-01 -9.45801735e-01 -4.78303522e-01
-3.69122326e-01 -1.07678950e+00 9.29832399e-01 5.93793511e-01
3.21329027e-01 -1.15969169e+00 8.65000933e-02 6.85494542e-02
5.23794711e-01 -2.67287791e-01 7.95873880e-01 -9.85759318e-01
-5.04484534e-01 8.86769593e-03 -5.47606528e-01 2.83983409e-01
2.74840772e-01 -5.49850821e-01 -5.33496678e-01 -2.99879480e-02
3.46360728e-02 -2.08671410e-02 7.00852334e-01 -1.75202727e-01
7.74452865e-01 -3.22842121e-01 -4.19656247e-01 5.47667205e-01
1.61463273e+00 3.29558164e-01 6.46420121e-01 4.00152028e-01
1.10965586e+00 5.43536305e-01 6.30096495e-01 2.06242189e-01
8.23831499e-01 3.28403801e-01 -1.52268216e-01 -4.97864187e-01
2.23865941e-01 -4.31874543e-01 5.42641044e-01 1.71168923e+00
-3.33845139e-01 2.15795487e-02 -1.02758694e+00 5.56223691e-01
-1.67993057e+00 -7.58542418e-01 -1.38433874e-01 1.68291998e+00
1.04589224e+00 -3.24183136e-01 -6.01027429e-01 -5.73460758e-01
1.13671732e+00 -1.55705944e-01 -2.98757195e-01 -2.10757144e-02
-4.46197480e-01 3.48539352e-01 5.01126885e-01 1.46027997e-01
-1.19074619e+00 1.39182913e+00 4.71377611e+00 1.00710845e+00
-9.69291747e-01 2.51571149e-01 4.46256399e-01 7.70557940e-01
-5.82168937e-01 2.98215389e-01 -1.23218715e+00 4.86148685e-01
6.95688367e-01 -3.55069578e-01 3.42862219e-01 6.40457571e-01
9.95739698e-02 4.16881859e-01 -7.61977136e-01 6.32726073e-01
1.64081991e-01 -8.59243035e-01 3.05424899e-01 3.66245471e-02
9.38894689e-01 1.38662830e-01 -4.13310468e-01 7.23540425e-01
3.38337630e-01 -7.81722605e-01 3.88070971e-01 6.57052994e-01
8.32694411e-01 -8.79820406e-01 1.40231860e+00 4.15461123e-01
-1.67096615e+00 7.83473328e-02 -5.27822495e-01 2.07679123e-01
2.96025127e-01 3.14158916e-01 -2.05498099e-01 1.30580795e+00
5.63721299e-01 1.02097023e+00 -7.65125513e-01 8.45028579e-01
-5.93317568e-01 5.35028815e-01 -2.06031278e-01 -1.55626759e-01
4.63887602e-01 -5.16806066e-01 8.85665491e-02 1.57229817e+00
4.98956323e-01 4.00264710e-01 2.37040907e-01 7.39233732e-01
-2.56321996e-01 9.29205835e-01 -4.96249229e-01 -3.86473060e-01
5.60083389e-01 1.37246561e+00 -5.16986132e-01 -3.82793665e-01
-9.14715230e-01 1.05351114e+00 7.16200590e-01 5.36875546e-01
-7.70988643e-01 -8.67315531e-01 8.88108760e-02 -5.70161998e-01
5.42870045e-01 -2.04811722e-01 -1.20551392e-01 -1.91450226e+00
-1.52773065e-02 -9.45512891e-01 3.66906166e-01 -4.60828125e-01
-1.50195396e+00 1.02298236e+00 -3.76973659e-01 -1.28074658e+00
2.30633125e-01 -4.93125767e-01 -6.87900066e-01 1.10629761e+00
-1.57149947e+00 -1.60734785e+00 7.00985640e-02 7.35042393e-01
5.04096925e-01 -4.48120665e-03 7.66856313e-01 7.03924179e-01
-9.70960796e-01 6.72203779e-01 1.32039517e-01 8.88970017e-01
8.29069853e-01 -1.05395389e+00 2.54461527e-01 1.19306290e+00
2.46674567e-01 1.43257630e+00 -6.97487444e-02 -8.43141437e-01
-1.22988725e+00 -1.04973900e+00 1.63008857e+00 -1.82914078e-01
5.48281252e-01 -2.34583139e-01 -8.62799644e-01 8.65828574e-01
6.94072664e-01 -2.54345268e-01 9.56579745e-01 3.29664350e-01
-1.60421968e-01 -5.79923652e-02 -6.02050364e-01 6.08720720e-01
9.68311310e-01 -6.07275069e-01 -8.00902426e-01 -7.54112303e-02
8.91949117e-01 -2.35059053e-01 -1.00532329e+00 4.38800186e-01
2.29408622e-01 -5.36198974e-01 4.21645164e-01 -7.25717664e-01
2.94467449e-01 -4.60644603e-01 -4.22431491e-02 -1.25681710e+00
-4.80133682e-01 -2.49129772e-01 3.85320812e-01 1.70249140e+00
6.35324597e-01 -5.14363587e-01 1.16508983e-01 5.64811766e-01
-3.57651234e-01 -5.11720538e-01 -6.29847169e-01 -4.49128717e-01
1.87302366e-01 -1.44868076e-01 9.56149578e-01 1.44443786e+00
3.04289609e-01 7.42395043e-01 -4.19456184e-01 2.34308735e-01
1.76859513e-01 6.42802641e-02 2.81071126e-01 -9.19293463e-01
6.73537478e-02 -2.03577533e-01 3.85763124e-02 -1.25992799e+00
4.35868889e-01 -1.04801774e+00 -2.02368461e-02 -1.50031352e+00
5.70250869e-01 -5.00285685e-01 -5.92722893e-01 6.63467586e-01
-7.69160748e-01 -1.22979134e-01 1.82820022e-01 8.04712251e-02
-7.88705170e-01 8.82348955e-01 1.32092309e+00 1.18461259e-01
-4.02928777e-02 -4.56195921e-01 -1.00436604e+00 5.63044429e-01
5.37196279e-01 -5.41488171e-01 8.39356929e-02 -9.01977718e-01
2.84156501e-01 -1.26942724e-01 -3.40369463e-01 -7.37554908e-01
5.74117243e-01 -7.07760006e-02 4.94850755e-01 -5.96346140e-01
-3.09046030e-01 -9.93534565e-01 -1.48588747e-01 1.83967084e-01
-1.51195481e-01 3.46885651e-01 1.14663579e-01 3.85214180e-01
-6.63034022e-01 -2.26449922e-01 3.00727904e-01 -4.50396448e-01
-8.86467934e-01 4.54105705e-01 -1.33406833e-01 2.34360933e-01
7.02458918e-01 2.69949287e-01 -2.81461746e-01 -1.03320099e-01
-4.78797823e-01 5.55380106e-01 1.64195314e-01 5.52290082e-01
3.45580786e-01 -1.62931025e+00 -6.98198497e-01 2.17407331e-01
1.09420836e-01 -2.31351554e-01 5.18200278e-01 8.98966789e-01
-4.17608589e-01 6.93111062e-01 -1.88291281e-01 9.53768846e-03
-7.38699317e-01 7.20682561e-01 1.31402299e-01 -6.68127716e-01
-2.45964274e-01 4.68697041e-01 3.87071431e-01 -9.18897927e-01
-2.55922288e-01 1.89809240e-02 -9.36004102e-01 -1.00027636e-01
4.60588545e-01 1.96368337e-01 -1.21562943e-01 -1.08456564e+00
-4.89739597e-01 6.58395231e-01 -3.40907276e-01 -8.47121254e-02
1.31343830e+00 -3.77094537e-01 -6.31635845e-01 1.38387397e-01
1.12737799e+00 4.50375199e-01 -4.75046009e-01 -4.53527957e-01
3.61648709e-01 2.96400907e-03 -2.18643650e-01 -8.52087796e-01
-1.11813080e+00 1.00327134e+00 1.83502123e-01 -3.67165059e-01
1.23889875e+00 -1.04321711e-01 1.02470315e+00 6.09961987e-01
5.80181599e-01 -1.17493272e+00 -6.06189847e-01 9.79593217e-01
4.78609532e-01 -1.30180418e+00 -4.03864920e-01 -4.98354703e-01
-9.36503351e-01 1.04023492e+00 9.33129907e-01 1.71356022e-01
8.30112875e-01 1.04244471e-01 3.44551116e-01 8.05432051e-02
-3.15724045e-01 -5.15819192e-01 4.42725956e-01 3.07819426e-01
8.70250285e-01 -1.42647490e-01 -1.05884433e+00 1.27633786e+00
6.91695809e-02 -1.30149096e-01 1.92630470e-01 7.86692262e-01
-6.95710480e-02 -1.55521858e+00 4.24135476e-02 8.45550448e-02
-7.08386123e-01 -8.79581630e-01 -2.64256984e-01 7.39530027e-01
4.56443965e-01 1.01035905e+00 -2.85071194e-01 -3.08456868e-01
2.82474399e-01 -5.92497848e-02 -7.14440867e-02 -7.21685648e-01
-8.73322725e-01 2.19688073e-01 9.73549262e-02 -4.40133899e-01
-7.00964034e-01 -4.41505730e-01 -1.28549337e+00 9.61712450e-02
-7.86448538e-01 7.90502906e-01 5.78723192e-01 1.20721686e+00
5.37268817e-01 3.41468394e-01 8.44908059e-01 -2.79853910e-01
-1.97459981e-01 -1.27750969e+00 -5.05854130e-01 3.01982582e-01
-2.76458174e-01 -2.91290104e-01 -1.33295566e-01 -4.80905324e-02]
|
[9.811388969421387, 9.747503280639648]
|
e62af787-7e45-4f27-9a72-bf355eec49ec
|
conviformers-convolutionally-guided-vision
|
2208.08900
| null |
https://arxiv.org/abs/2208.08900v2
|
https://arxiv.org/pdf/2208.08900v2.pdf
|
Conviformers: Convolutionally guided Vision Transformer
|
Vision transformers are nowadays the de-facto choice for image classification tasks. There are two broad categories of classification tasks, fine-grained and coarse-grained. In fine-grained classification, the necessity is to discover subtle differences due to the high level of similarity between sub-classes. Such distinctions are often lost as we downscale the image to save the memory and computational cost associated with vision transformers (ViT). In this work, we present an in-depth analysis and describe the critical components for developing a system for the fine-grained categorization of plants from herbarium sheets. Our extensive experimental analysis indicated the need for a better augmentation technique and the ability of modern-day neural networks to handle higher dimensional images. We also introduce a convolutional transformer architecture called Conviformer which, unlike the popular Vision Transformer (ConViT), can handle higher resolution images without exploding memory and computational cost. We also introduce a novel, improved pre-processing technique called PreSizer to resize images better while preserving their original aspect ratios, which proved essential for classifying natural plants. With our simple yet effective approach, we achieved SoTA on Herbarium 202x and iNaturalist 2019 dataset.
|
['Ivań Felipe Rodríguez', 'Thomas Serre', 'Thomas Fel', 'Mohit Vaishnav']
|
2022-08-17
| null | null | null | null |
['fine-grained-image-classification']
|
['computer-vision']
|
[ 3.68837327e-01 -3.16850305e-01 2.79311866e-01 -1.14480175e-01
-1.81049153e-01 -8.24945152e-01 4.62936014e-01 3.48367691e-02
-2.13905036e-01 5.60334384e-01 -2.96987504e-01 -4.41119522e-01
-3.34697098e-01 -1.17886364e+00 -4.04241204e-01 -6.11071110e-01
2.09864944e-01 2.97972798e-01 3.21886063e-01 -1.98867410e-01
3.60998005e-01 8.96103680e-01 -1.87096214e+00 3.34691167e-01
9.23362613e-01 1.20534873e+00 6.62300110e-01 6.65949702e-01
-4.95162934e-01 4.74912792e-01 -6.26592696e-01 -4.20641422e-01
2.19434723e-01 6.02530092e-02 -1.06471312e+00 3.36298853e-01
6.59670711e-01 -1.99213326e-01 -3.21323471e-03 1.32115173e+00
2.89010018e-01 -2.89460897e-01 7.55901515e-01 -1.21159136e+00
-1.09812176e+00 4.45275158e-01 -6.86757982e-01 2.57156193e-01
-9.64374244e-02 -6.05222173e-02 6.83047831e-01 -9.46361542e-01
2.90107995e-01 1.19184303e+00 1.04503381e+00 2.69486696e-01
-1.10998595e+00 -5.10694563e-01 -6.43869117e-03 5.35598636e-01
-1.44655621e+00 -1.07879341e-01 4.44250196e-01 -5.86608171e-01
6.17628157e-01 3.72278184e-01 6.39800012e-01 6.14813030e-01
3.32212299e-01 2.43310928e-01 1.33609736e+00 -3.91889930e-01
2.98680216e-02 2.49030106e-02 7.47210681e-02 9.15123761e-01
3.01962346e-01 -7.69209117e-02 -1.31264208e-02 1.45231336e-01
1.19142580e+00 3.60967785e-01 -3.69397849e-01 -2.94072330e-01
-1.29184651e+00 7.53994465e-01 8.12314749e-01 6.47765994e-01
-4.58189100e-01 -3.01535279e-01 4.69293714e-01 1.25780880e-01
2.91327417e-01 6.60858691e-01 -4.00185376e-01 1.56042323e-01
-9.09362972e-01 -3.58325452e-01 4.80150580e-01 8.97287548e-01
6.70421362e-01 2.02136114e-01 -8.92341956e-02 9.41919446e-01
-2.21624643e-01 2.87409931e-01 7.39493549e-01 -7.32970595e-01
-1.58878550e-01 8.56980979e-01 -3.23348880e-01 -1.15694761e+00
-2.97525823e-01 -4.73433882e-01 -1.43013275e+00 4.11463827e-01
4.45912659e-01 5.81169248e-01 -1.37389314e+00 1.18347466e+00
5.96733950e-02 -4.16265763e-02 -2.63786018e-01 5.76192737e-01
1.00240183e+00 4.17407334e-01 1.55495405e-01 9.95276645e-02
1.83056021e+00 -7.45526969e-01 -4.67819333e-01 4.51971684e-03
1.97725415e-01 -9.50769782e-01 1.31397438e+00 3.75057220e-01
-8.65878224e-01 -7.82189429e-01 -1.18538880e+00 -3.95388976e-02
-1.07594323e+00 3.82564604e-01 8.53762090e-01 5.47668576e-01
-1.13126588e+00 8.98054421e-01 -5.57968676e-01 -6.96882904e-01
6.87373519e-01 2.79664248e-01 -7.26920545e-01 2.24756047e-01
-7.66043603e-01 9.22223270e-01 5.59483528e-01 2.60613739e-01
-7.53276765e-01 -8.65740597e-01 -6.16852045e-01 4.44693714e-01
1.31876245e-01 -4.38602716e-01 1.17771590e+00 -7.45656669e-01
-1.40992379e+00 1.14338946e+00 2.94763416e-01 -3.58001918e-01
1.34634972e-01 2.89854139e-01 -8.20717067e-02 1.87391788e-01
9.33115780e-02 6.72044456e-01 9.87493396e-01 -9.99078453e-01
-9.52188313e-01 -5.93618572e-01 7.90264904e-02 -1.69057876e-01
-7.26972640e-01 -2.96427041e-01 -2.11039737e-01 -7.25733161e-01
1.26797110e-01 -9.23396111e-01 1.29157240e-02 4.48099911e-01
-1.92637205e-01 -1.20411091e-01 1.26356983e+00 -4.76009399e-01
7.33810842e-01 -2.01882648e+00 8.74584317e-02 -2.29464829e-01
6.22078121e-01 8.16060543e-01 -2.27727145e-01 3.70885693e-02
-4.24087912e-01 1.49537668e-01 -1.37474492e-01 1.55426413e-01
-2.19398409e-01 1.65562868e-01 -5.00092983e-01 3.38054925e-01
1.11733027e-01 9.34714496e-01 -6.44101858e-01 -5.20020485e-01
6.78266406e-01 5.51827073e-01 -1.30855486e-01 3.20823789e-02
2.48002470e-01 7.41739571e-02 -3.17830205e-01 9.76788282e-01
9.38238442e-01 -4.40039068e-01 -2.73290813e-01 -7.85315692e-01
-2.98397809e-01 -3.41861784e-01 -8.17633510e-01 1.32167876e+00
-4.91791666e-01 6.88871622e-01 2.28450090e-01 -1.13305557e+00
1.02775311e+00 1.85378328e-01 3.63781482e-01 -4.92041796e-01
2.56950736e-01 2.13440090e-01 -2.02626646e-01 -4.78058169e-03
5.35443306e-01 -1.24022953e-01 1.34869784e-01 -1.15641944e-01
1.26888663e-01 -7.24406362e-01 2.30939031e-01 -2.03039452e-01
7.12628365e-01 -3.07087209e-02 6.94635332e-01 -6.79641604e-01
6.63481593e-01 2.33623788e-01 3.89551163e-01 4.49543238e-01
-4.76508290e-01 5.62491953e-01 8.44499469e-02 -8.84928584e-01
-1.01691806e+00 -8.75249922e-01 -3.44215512e-01 9.55112517e-01
1.09419785e-02 -3.50707203e-01 -7.12333322e-01 -5.50343394e-01
-8.77955705e-02 9.52472389e-02 -8.01826894e-01 -9.68106091e-02
-1.68306932e-01 -8.40153933e-01 6.58409238e-01 5.86602986e-01
1.02877748e+00 -1.02372050e+00 -6.97999597e-01 3.22123468e-02
8.84834770e-03 -1.04668438e+00 -2.88296044e-01 5.83736718e-01
-7.15509713e-01 -1.17668700e+00 -8.62620413e-01 -1.05963409e+00
5.80665171e-01 7.31235087e-01 1.14259839e+00 4.66405302e-02
-7.32026279e-01 -1.31670713e-01 -3.68072420e-01 -5.28497159e-01
-5.33176549e-02 2.33091727e-01 -1.20564319e-01 -4.64697301e-01
3.16941619e-01 -6.45716488e-01 -5.00485837e-01 3.56345594e-01
-1.01920819e+00 8.63858014e-02 6.61357403e-01 1.07889616e+00
7.38992810e-01 5.49840868e-01 3.33742231e-01 -7.76855707e-01
5.78493476e-01 1.32079780e-01 -7.53861010e-01 3.76493394e-01
-4.80010599e-01 3.45210657e-02 1.07681060e+00 -4.99669731e-01
-8.07080984e-01 1.11900404e-01 -1.21907301e-01 -5.20954311e-01
-2.11317718e-01 2.78126210e-01 -4.04761173e-02 -6.31282985e-01
7.39823341e-01 3.08757395e-01 -2.72274137e-01 -4.54714596e-01
2.17969239e-01 8.57350349e-01 8.52454364e-01 -2.68827528e-01
9.10223544e-01 4.60034072e-01 1.88146830e-01 -1.03625262e+00
-8.41071844e-01 -2.46717915e-01 -8.85774672e-01 9.07039195e-02
8.15515459e-01 -5.77036440e-01 -1.12459493e+00 8.45549405e-01
-8.80361319e-01 1.27433613e-02 -5.31418264e-01 1.01914564e-02
-3.78523499e-01 4.14040983e-01 -5.19674659e-01 -1.94691107e-01
-7.32358456e-01 -1.16358888e+00 1.24635625e+00 5.67273378e-01
2.65597165e-01 -7.98382998e-01 -2.86120743e-01 1.80603728e-01
7.41803110e-01 1.61095098e-01 9.57337379e-01 -1.58440009e-01
-2.83219129e-01 -1.20495386e-01 -8.36462557e-01 1.84800848e-01
7.41160572e-01 3.55917543e-01 -1.11383986e+00 -2.93785870e-01
-1.20455809e-01 -2.46282220e-01 8.30099702e-01 3.71690959e-01
1.83676124e+00 1.14495352e-01 -3.53324711e-01 1.03255606e+00
1.46892917e+00 1.93312898e-01 6.25038624e-01 5.12241662e-01
9.21667516e-01 4.05938059e-01 5.29809892e-01 2.95483917e-01
1.34098083e-01 5.31269073e-01 5.87509155e-01 -5.81186950e-01
-2.39999518e-01 -7.13787600e-02 -3.74070346e-01 5.76257825e-01
-2.90825725e-01 -3.65991965e-02 -7.20097005e-01 5.15060902e-01
-1.44318211e+00 -9.06662285e-01 7.62332380e-02 1.98900306e+00
6.78065538e-01 -9.63516980e-02 -1.50496870e-01 4.33120936e-01
8.05383623e-01 -9.55281109e-02 -5.64215124e-01 -4.54283357e-01
-1.46819368e-01 4.28000897e-01 7.77897060e-01 1.52090192e-01
-1.30951571e+00 1.20289516e+00 6.86664915e+00 1.08983338e+00
-1.56411529e+00 -3.91957462e-01 5.65376818e-01 5.86167872e-01
2.53587306e-01 -4.30034578e-01 -7.18174517e-01 2.16246530e-01
6.36058331e-01 -1.18173435e-01 2.40291864e-01 1.02615535e+00
-1.84190035e-01 7.34188706e-02 -7.70786464e-01 1.09535480e+00
-2.06592277e-01 -1.56075370e+00 3.04283530e-01 5.78191541e-02
2.57360488e-01 -2.38837421e-01 1.44096380e-02 2.15973914e-01
3.95507246e-01 -1.20987189e+00 4.29641247e-01 1.15151465e-01
1.16338038e+00 -5.86393654e-01 6.90168738e-01 1.32326204e-02
-1.77951074e+00 -1.31734163e-01 -8.93569529e-01 9.94208157e-02
-4.69372094e-01 5.68645179e-01 -8.69765997e-01 5.99947751e-01
1.07874572e+00 7.77898431e-01 -9.24221337e-01 8.61850858e-01
2.09058642e-01 -2.78782845e-03 -1.16852410e-01 2.32324004e-01
2.23333389e-01 -1.67291954e-01 1.51124522e-01 1.00318420e+00
5.25669098e-01 -1.18980020e-01 1.29679173e-01 5.04054964e-01
7.24579096e-02 -1.95395067e-01 -5.70524275e-01 -3.64176899e-01
4.59690243e-01 1.80503392e+00 -1.20829725e+00 -3.60231608e-01
-2.56135464e-01 1.12021470e+00 2.99973279e-01 -2.70813763e-01
-6.38335943e-01 -6.83478534e-01 5.72324932e-01 -8.36292282e-03
6.06407404e-01 -9.64523107e-02 -2.71656275e-01 -1.17876172e+00
-3.37920696e-01 -9.59118128e-01 4.65208948e-01 -8.49860787e-01
-1.30570114e+00 1.02781796e+00 -2.31633738e-01 -1.24898839e+00
-3.54279466e-02 -1.15896022e+00 -3.78687024e-01 8.11087370e-01
-1.57929981e+00 -1.53732228e+00 -8.97466540e-01 7.97403038e-01
6.91782832e-01 -7.74694830e-02 1.26311159e+00 3.27084452e-01
-4.44728225e-01 4.05407339e-01 3.79452258e-02 -6.45376220e-02
5.72441161e-01 -1.49774051e+00 2.33685002e-01 6.69972897e-01
-5.50211221e-03 4.13493812e-01 6.28852785e-01 -2.11895347e-01
-1.19737315e+00 -1.24041820e+00 5.86618543e-01 -1.68501865e-02
6.86328471e-01 -1.96939409e-01 -9.35035646e-01 4.77265567e-01
7.41803721e-02 1.01228587e-01 5.86336851e-01 -3.95706594e-02
-4.55361724e-01 -2.28469446e-01 -1.41950130e+00 6.09704077e-01
8.28711450e-01 -6.07305050e-01 -5.77660084e-01 3.15813720e-01
4.55956310e-01 -1.33649573e-01 -1.06872213e+00 5.15643656e-01
6.65232241e-01 -8.40050757e-01 1.16375256e+00 -1.70858353e-01
2.89260000e-01 -5.43267190e-01 -2.21747696e-01 -1.55128956e+00
-9.78265584e-01 -2.14539900e-01 4.44656610e-01 1.09584534e+00
-8.73284340e-02 -5.67821443e-01 8.07821929e-01 4.73385006e-02
-9.51233879e-02 -3.94338876e-01 -5.36037028e-01 -7.25544751e-01
1.44862548e-01 1.29025057e-01 8.13229680e-01 8.47298205e-01
-1.90857679e-01 3.72105509e-01 -7.54381642e-02 1.18131109e-01
5.09107172e-01 5.80488503e-01 4.53430176e-01 -1.61536252e+00
5.48361801e-02 -5.52759707e-01 -7.99315274e-01 -5.72824717e-01
-1.47634357e-01 -6.53522551e-01 3.44377458e-02 -1.48275983e+00
3.10705274e-01 -1.44060850e-01 -1.18947454e-01 7.41927207e-01
-1.20352902e-01 7.50519395e-01 2.85139561e-01 3.15039456e-01
1.83365494e-02 4.54120845e-01 1.53499436e+00 -3.44993651e-01
3.39291282e-02 -8.70577097e-02 -9.29680228e-01 9.64113832e-01
8.50867271e-01 -2.89619584e-02 -4.64701205e-01 -3.58236045e-01
-3.86011332e-01 -3.62112880e-01 3.90869230e-01 -1.21096814e+00
1.05884053e-01 -1.49827868e-01 7.89486647e-01 -8.40280592e-01
4.05636281e-01 -1.07044291e+00 2.50731945e-01 6.20731175e-01
4.14670818e-02 1.82769597e-01 5.38857937e-01 4.35067788e-02
-2.76376456e-01 -1.93962634e-01 1.29915333e+00 -3.53063643e-01
-1.02915239e+00 4.47389394e-01 -4.52450335e-01 -5.44837952e-01
1.20024860e+00 -5.19878089e-01 -5.16332686e-01 1.69995666e-01
-5.57282627e-01 -2.37740502e-01 5.06976306e-01 3.70220453e-01
5.01548111e-01 -1.14989626e+00 -5.61785936e-01 3.47958177e-01
2.37016305e-01 -7.59254396e-02 1.77264035e-01 3.41639906e-01
-8.80495548e-01 7.39985347e-01 -1.09448862e+00 -7.62945831e-01
-1.57537270e+00 7.91282177e-01 4.84862357e-01 -3.88769567e-01
-6.77229404e-01 8.35428715e-01 6.40068412e-01 -4.70953465e-01
6.36224151e-02 -5.85939825e-01 -5.21464944e-01 1.54341877e-01
6.88415945e-01 2.24974826e-01 4.50718731e-01 -4.98450667e-01
-4.42145616e-01 8.12964201e-01 -2.59305269e-01 5.46610475e-01
1.52243376e+00 -7.18988851e-02 -2.24500120e-01 1.49770781e-01
8.60517621e-01 -3.38104010e-01 -1.07478249e+00 -1.52328685e-02
-2.38591835e-01 -6.53552294e-01 4.01873469e-01 -6.02912784e-01
-1.30372548e+00 9.01455224e-01 9.41385627e-01 7.38202929e-01
1.53047335e+00 -2.80640662e-01 4.91403580e-01 3.73587787e-01
4.09075975e-01 -7.13095307e-01 -2.49579608e-01 5.73878825e-01
8.82819414e-01 -1.09585249e+00 -2.74508540e-02 -7.02546418e-01
-2.00350150e-01 1.27704453e+00 6.37532830e-01 -6.72145635e-02
6.48687243e-01 5.51177382e-01 -3.43964063e-02 -3.92354876e-01
-4.44055021e-01 -3.26147497e-01 1.62024826e-01 1.03378844e+00
5.13669848e-01 2.51942188e-01 9.49373245e-02 1.65986523e-01
-5.06224453e-01 2.42972791e-01 4.99037951e-01 8.81111920e-01
-7.10048676e-01 -8.89393687e-01 -6.23750389e-01 4.79156315e-01
-3.80729020e-01 -1.67567894e-01 -5.07319272e-01 6.74401104e-01
2.12822020e-01 7.49775767e-01 1.70662805e-01 -4.84626770e-01
4.42485839e-01 -3.41983378e-01 6.06627405e-01 -4.05782223e-01
-5.40214956e-01 -2.66469687e-01 -3.07545781e-01 -4.26562458e-01
-2.78292090e-01 -1.60397559e-01 -6.97078526e-01 -7.80166328e-01
-3.90461057e-01 -3.78709994e-02 7.41884828e-01 7.87961006e-01
2.59931922e-01 8.86297524e-01 5.43808043e-01 -1.05629373e+00
-4.33934957e-01 -9.11006868e-01 -7.58647621e-01 7.66022354e-02
2.39966929e-01 -6.98495388e-01 -1.89347714e-01 3.86488378e-01]
|
[9.633136749267578, 2.0327093601226807]
|
9ce12623-fa83-49cf-8d44-50a161b949c3
|
exploiting-feature-diversity-for-make-up
|
2208.06179
| null |
https://arxiv.org/abs/2208.06179v1
|
https://arxiv.org/pdf/2208.06179v1.pdf
|
Exploiting Feature Diversity for Make-up Temporal Video Grounding
|
This technical report presents the 3rd winning solution for MTVG, a new task introduced in the 4-th Person in Context (PIC) Challenge at ACM MM 2022. MTVG aims at localizing the temporal boundary of the step in an untrimmed video based on a textual description. The biggest challenge of this task is the fi ne-grained video-text semantics of make-up steps. However, current methods mainly extract video features using action-based pre-trained models. As actions are more coarse-grained than make-up steps, action-based features are not sufficient to provide fi ne-grained cues. To address this issue,we propose to achieve fi ne-grained representation via exploiting feature diversities. Specifically, we proposed a series of methods from feature extraction, network optimization, to model ensemble. As a result, we achieved 3rd place in the MTVG competition.
|
['Ruizhi Qiao', 'Chen Wu', 'Sunan He', 'Taian Guo', 'Wei Wen', 'Xiujun Shu']
|
2022-08-12
| null | null | null | null |
['video-grounding']
|
['computer-vision']
|
[ 3.32145989e-01 -2.82442003e-01 -3.93416375e-01 -4.31867003e-01
-7.71886826e-01 -4.64243650e-01 8.28455210e-01 5.72197400e-02
-5.53534567e-01 6.06675923e-01 9.21063840e-01 2.05086365e-01
1.33116450e-02 -2.10472167e-01 -5.65492511e-01 -2.06169456e-01
-7.84457400e-02 4.55234908e-02 1.03434347e-01 -2.53377974e-01
4.82405007e-01 2.32288480e-01 -1.62779677e+00 9.70633805e-01
4.07181472e-01 9.27815616e-01 1.31277487e-01 7.62845278e-01
-1.10768802e-01 9.02556241e-01 -5.10424972e-01 -2.38514602e-01
1.96984276e-01 -5.13043702e-01 -8.24941516e-01 2.98490912e-01
8.59522343e-01 -2.95383096e-01 -5.02428472e-01 7.17108130e-01
2.68187374e-01 4.62463886e-01 8.98712337e-01 -1.22341692e+00
-1.49448305e-01 7.64012635e-01 -4.11054313e-01 4.36086059e-01
7.18333244e-01 4.39292677e-02 1.21483803e+00 -8.45360875e-01
9.26081181e-01 1.25222504e+00 5.45531571e-01 8.69264007e-01
-7.63665318e-01 -5.12929522e-02 8.08115780e-01 7.06260502e-01
-1.39316714e+00 -4.75575835e-01 6.76181316e-01 -6.67040884e-01
1.09361315e+00 4.53260183e-01 7.81853974e-01 1.39799118e+00
5.13537936e-02 1.16570425e+00 8.02643955e-01 -3.76174182e-01
4.89053018e-02 -2.80534476e-01 1.95937619e-01 6.78813577e-01
8.59186053e-03 -2.69227147e-01 -8.46743703e-01 -3.19336876e-02
6.63852930e-01 2.26307616e-01 -2.28623033e-01 -2.69814339e-02
-1.35660088e+00 5.12902796e-01 1.71808258e-01 5.96881926e-01
-4.79816735e-01 3.17673832e-01 4.23709452e-01 -9.69552472e-02
1.69044495e-01 4.12841558e-01 -4.64419007e-01 -6.91183686e-01
-1.36870694e+00 5.58295131e-01 7.04198599e-01 9.34605777e-01
6.44688904e-01 -2.55143285e-01 -7.43613601e-01 5.23605168e-01
1.70643106e-01 5.91211729e-02 3.74955744e-01 -1.05589962e+00
8.50579560e-01 6.45902812e-01 2.32244164e-01 -1.00090420e+00
-4.38746005e-01 -9.69781950e-02 -5.91866672e-01 -3.04054707e-01
4.58824664e-01 -1.11092761e-01 -1.01162720e+00 1.62049162e+00
-3.03239003e-02 3.89056534e-01 -4.50681776e-01 1.08774388e+00
6.85410798e-01 5.46772599e-01 3.76231790e-01 -8.92203674e-02
1.43104410e+00 -1.22488308e+00 -6.51150882e-01 -3.46763194e-01
7.18834162e-01 -3.81501019e-01 8.26896250e-01 3.54605079e-01
-9.85067725e-01 -6.46459579e-01 -1.18438935e+00 4.64528240e-03
-4.31972414e-01 2.26912707e-01 3.41816366e-01 4.81222510e-01
-8.12507570e-01 6.98190510e-01 -7.40193009e-01 -4.85034436e-01
4.75679994e-01 1.69441506e-01 -4.36588079e-01 -8.37732255e-02
-1.14051330e+00 7.25835681e-01 3.70705098e-01 -1.73578281e-02
-6.63405895e-01 -3.72333229e-01 -7.72824526e-01 3.64789646e-03
7.36223698e-01 -7.24489450e-01 1.11126566e+00 -9.75418091e-01
-1.13939273e+00 9.08068776e-01 -5.76859772e-01 -6.84428692e-01
3.92763108e-01 -5.06423652e-01 -4.31748301e-01 3.95986170e-01
1.24478281e-01 4.95283306e-01 1.10100949e+00 -8.67484689e-01
-1.04166174e+00 -3.40458453e-01 1.96427628e-01 2.04968601e-01
-3.76520962e-01 2.86016643e-01 -7.43218780e-01 -7.66164660e-01
-5.66335162e-04 -7.70811975e-01 -2.93661267e-01 -3.45238298e-01
-4.19897437e-01 -3.29507560e-01 6.80394232e-01 -8.63556385e-01
1.84871900e+00 -2.10061669e+00 3.88420284e-01 -7.96863884e-02
4.50780392e-01 3.67767900e-01 -5.19255735e-03 5.68446934e-01
2.23919034e-01 4.33525503e-01 1.28059208e-01 -5.73575616e-01
2.53633589e-01 -7.02998564e-02 -1.45708323e-01 2.53466219e-01
1.07910730e-01 1.02019775e+00 -7.38292396e-01 -7.14814901e-01
2.73226738e-01 4.31682378e-01 -5.20789623e-01 -4.89230305e-02
-3.67550731e-01 4.42309052e-01 -7.12654114e-01 5.77239394e-01
2.19411343e-01 -3.01637739e-01 2.03464732e-01 -3.43386918e-01
-1.46109939e-01 1.55165538e-01 -9.98990774e-01 1.84537077e+00
-2.06681699e-01 6.31065726e-01 -2.25423470e-01 -9.75417435e-01
3.65318537e-01 3.91327083e-01 6.42600775e-01 -5.12507260e-01
4.76017632e-02 -6.58722669e-02 -2.74086416e-01 -8.08555126e-01
7.19059944e-01 3.72056693e-01 -4.47939634e-01 1.26122922e-01
6.01050332e-02 4.83586967e-01 5.13173521e-01 2.29687646e-01
1.14172220e+00 5.08109391e-01 6.26386642e-01 1.99884758e-03
6.31836951e-01 -1.75482631e-01 5.88480949e-01 9.84187126e-01
-5.03276289e-01 1.08415556e+00 3.70202929e-01 -6.14065707e-01
-6.07061565e-01 -6.87138736e-01 4.95490372e-01 1.37730229e+00
-5.47223203e-02 -1.23118472e+00 -1.01864421e+00 -1.02576518e+00
-2.95488864e-01 4.83457804e-01 -8.55738819e-01 3.39464508e-02
-1.17463124e+00 -2.27111176e-01 4.19350535e-01 7.30975568e-01
5.91974556e-01 -9.73331809e-01 -7.20826447e-01 3.10728192e-01
-7.83857107e-01 -1.43347371e+00 -9.58709717e-01 -1.09013513e-01
-6.15956664e-01 -1.07677031e+00 -7.24826872e-01 -3.68423045e-01
3.18095922e-01 2.63010949e-01 1.06452930e+00 5.50553761e-02
-4.20534045e-01 5.98153293e-01 -5.68403006e-01 -2.31614739e-01
2.07647055e-01 6.15722910e-02 -3.56324054e-02 4.00391161e-01
6.63647354e-01 -3.49798888e-01 -7.56388664e-01 2.90185750e-01
-4.62989807e-01 4.99559119e-02 4.12341297e-01 5.21426558e-01
5.28697014e-01 -7.36240372e-02 2.75919855e-01 -5.50979793e-01
3.85860741e-01 -8.88647810e-02 -1.48484930e-01 4.61488783e-01
-1.85087427e-01 -4.74204309e-02 5.02940595e-01 -4.06187236e-01
-1.00901163e+00 1.67644516e-01 -3.42928097e-02 -4.05960739e-01
-3.36547762e-01 4.51530993e-01 -2.85470128e-01 3.33113015e-01
5.76766968e-01 3.70306760e-01 -6.02498651e-01 -4.87611353e-01
3.36921632e-01 4.71101254e-01 5.02921045e-01 -6.33951902e-01
3.79812330e-01 4.81874853e-01 -2.72279070e-03 -8.59453321e-01
-1.07875514e+00 -9.17775333e-01 -9.15197790e-01 -4.88428056e-01
1.31879365e+00 -9.35233414e-01 -6.15932643e-01 2.91134298e-01
-1.12738633e+00 -1.52394429e-01 -1.39588282e-01 4.49938059e-01
-7.86535144e-01 5.20225346e-01 -4.14685309e-01 -6.94087207e-01
-8.01334605e-02 -8.92823279e-01 1.20924842e+00 2.12109759e-01
-5.51915884e-01 -8.12343955e-01 -1.28467917e-01 7.51596749e-01
2.18330830e-01 2.75588781e-01 5.43468952e-01 -6.84207380e-01
-7.13418961e-01 -3.63188058e-01 -6.17172532e-02 1.27625257e-01
5.27415164e-02 1.57012455e-02 -8.03100348e-01 -6.12071389e-03
-1.67554751e-01 -6.35484383e-02 1.20610499e+00 5.46112120e-01
1.37118804e+00 -2.93694377e-01 -6.24057293e-01 5.47818005e-01
9.87683952e-01 3.62654813e-02 6.07740760e-01 4.30786133e-01
8.61496627e-01 5.08113325e-01 9.12008226e-01 6.50264263e-01
4.18240786e-01 1.12340403e+00 1.36500150e-01 6.08212054e-01
-2.85121948e-01 -5.69111347e-01 3.64772975e-01 2.85800189e-01
-7.79587805e-01 -3.83874506e-01 -8.53823125e-01 4.79193300e-01
-2.29959536e+00 -1.47207344e+00 5.43583781e-02 1.93718171e+00
2.87687391e-01 2.25323796e-01 3.88478845e-01 -4.40834612e-02
6.11654639e-01 4.66503173e-01 -1.06615171e-01 3.43067423e-02
-7.63649866e-02 -2.18986422e-01 9.73208025e-02 1.44525081e-01
-1.56363571e+00 1.07740664e+00 6.06736326e+00 9.70783114e-01
-8.90070736e-01 1.24614716e-01 3.01300734e-01 -3.42520505e-01
1.01912834e-01 -1.21543102e-01 -1.35389972e+00 4.88824427e-01
8.81851196e-01 -8.39179233e-02 1.72758847e-01 5.23926795e-01
5.76282561e-01 3.03470660e-02 -1.37916696e+00 1.33780217e+00
4.63760138e-01 -1.52947140e+00 1.68016106e-01 9.54199955e-02
6.99109614e-01 -2.43851423e-01 -1.97619408e-01 4.70853508e-01
-2.85786867e-01 -1.12155795e+00 8.30583453e-01 7.38289773e-01
5.86836874e-01 -5.88223755e-01 3.78797174e-01 4.41887736e-01
-1.52450693e+00 -2.27118149e-01 -1.67018771e-01 -7.17563406e-02
4.70935404e-01 1.52092785e-01 -6.19683981e-01 5.52991867e-01
6.29081011e-01 1.09653616e+00 -5.10983646e-01 1.02426875e+00
-2.79913843e-01 6.38228893e-01 -2.39837095e-02 -5.65549359e-03
3.78763676e-01 1.99956879e-01 6.83812678e-01 1.59767294e+00
2.41755307e-01 2.47722313e-01 4.43292350e-01 3.42865348e-01
-1.29652664e-01 -7.76397064e-02 -4.12417263e-01 -3.03871095e-01
1.83022708e-01 1.21758020e+00 -6.29226685e-01 -4.61266249e-01
-4.61994380e-01 1.22056985e+00 2.19705418e-01 4.90556329e-01
-9.67564166e-01 -3.60150486e-02 5.90483904e-01 2.69655943e-01
4.66606706e-01 -3.28438878e-01 -1.20732442e-01 -1.49939954e+00
9.08976123e-02 -8.63246202e-01 5.46266556e-01 -4.25480753e-01
-1.01904333e+00 6.43037260e-01 1.99022472e-01 -1.39877141e+00
-7.09047198e-01 -5.89494348e-01 -5.45031488e-01 6.17764711e-01
-1.04502356e+00 -1.34177196e+00 -3.92499745e-01 6.45279408e-01
9.53757107e-01 -1.46880597e-01 6.70101166e-01 2.72428989e-01
-4.92267251e-01 6.15280390e-01 -3.04339677e-01 2.32718602e-01
6.37743473e-01 -1.04147518e+00 4.92023975e-01 9.71455872e-01
2.84819990e-01 6.97724223e-01 6.77673817e-01 -6.73818529e-01
-1.27970707e+00 -1.20990276e+00 1.28172028e+00 -7.21478343e-01
4.10596579e-01 -3.73598933e-01 -5.15074372e-01 9.45019722e-01
-5.45183802e-03 -1.27046695e-02 5.85968375e-01 2.97543526e-01
-4.95674372e-01 1.91411301e-01 -7.85835087e-01 8.38608801e-01
1.49003851e+00 -6.06012583e-01 -8.10765266e-01 1.65861040e-01
4.31942403e-01 -1.31919578e-01 -6.34068966e-01 2.00490773e-01
6.13695800e-01 -8.51295829e-01 1.14269602e+00 -8.83470297e-01
5.51702201e-01 -1.84581801e-01 -5.23667932e-01 -1.02440667e+00
-3.97070080e-01 -7.24307299e-01 -4.33861345e-01 1.00725567e+00
2.51007438e-01 5.45186773e-02 9.39431131e-01 6.29294097e-01
-1.90673992e-01 -6.75957918e-01 -8.04768384e-01 -7.82777905e-01
-4.16572005e-01 -5.42232931e-01 4.18767452e-01 6.67344391e-01
1.87617347e-01 1.87064260e-01 -8.53655756e-01 -1.90618664e-01
5.79816699e-01 -6.83888793e-02 6.58847988e-01 -1.06435406e+00
-4.05820251e-01 -5.12888849e-01 -6.32976651e-01 -1.36999261e+00
3.32664810e-02 -6.52273774e-01 -1.31059196e-02 -1.69221187e+00
4.18704897e-01 2.77568191e-01 -3.94917160e-01 3.95733416e-01
-2.85643458e-01 1.19205751e-01 6.63229585e-01 8.44453350e-02
-1.24621117e+00 2.00439245e-01 9.89556432e-01 -2.60151237e-01
-1.79778248e-01 1.28705114e-01 -5.17916322e-01 9.25545454e-01
8.01265597e-01 -1.61452860e-01 -2.92337209e-01 -3.09556127e-01
2.21283007e-02 9.88007486e-02 4.78752673e-01 -1.17439067e+00
3.50845665e-01 -4.76389825e-01 4.02405202e-01 -7.84530461e-01
8.07031929e-01 -5.82177579e-01 2.16410998e-02 9.52750817e-02
-4.64954466e-01 -3.64787191e-01 -2.52949260e-02 6.46626651e-01
-3.87039423e-01 -2.11161986e-01 2.65055329e-01 -4.12289858e-01
-1.33406842e+00 3.98621261e-01 -5.72900772e-01 2.39410345e-02
9.21230197e-01 -4.55522597e-01 -3.13393027e-01 -4.04686809e-01
-1.16327775e+00 1.13034934e-01 9.07733291e-02 6.58118367e-01
7.29151905e-01 -1.18843877e+00 -7.00625837e-01 -1.57193586e-01
3.15661728e-01 -5.52541971e-01 6.60552263e-01 8.27223718e-01
-7.65110478e-02 7.29754031e-01 -2.24778801e-01 -4.92687017e-01
-1.56433356e+00 5.41008830e-01 2.34977603e-01 -4.24467057e-01
-6.63389862e-01 8.64817858e-01 1.85027584e-01 5.84949069e-02
2.97156423e-01 -1.28503129e-01 -5.92181802e-01 2.05059797e-01
9.86581922e-01 5.07442296e-01 -1.61061332e-01 -7.52086818e-01
-5.02098918e-01 7.31771588e-01 -2.48422444e-01 -1.78046897e-01
1.20522451e+00 -3.91166776e-01 2.15231046e-01 3.73420924e-01
1.14439797e+00 -2.36126557e-01 -1.45476460e+00 -1.80086911e-01
2.61557043e-01 -4.92857695e-01 -1.00343153e-01 -8.53534639e-01
-7.02164769e-01 8.67440403e-01 2.90740758e-01 -1.24855340e-01
1.03240609e+00 -7.25673838e-03 7.54256129e-01 4.33347642e-01
4.81927931e-01 -1.32571483e+00 2.03239650e-01 6.14859641e-01
1.00146914e+00 -1.13362157e+00 -2.44261790e-02 -3.96150500e-01
-7.05441117e-01 1.18186831e+00 7.27625847e-01 1.42529190e-01
4.97070730e-01 -1.39526814e-01 -2.17898354e-01 -1.59437239e-01
-8.53705943e-01 -4.29228008e-01 8.28020096e-01 4.76733059e-01
4.05710518e-01 5.68281189e-02 -4.30781960e-01 1.06215990e+00
9.36877951e-02 2.26851493e-01 2.72313058e-01 1.10815001e+00
-7.07413852e-01 -1.02241480e+00 -1.73402607e-01 5.59959710e-01
-6.37623429e-01 1.36354370e-02 -4.49098647e-01 6.62756979e-01
2.69202560e-01 1.15097451e+00 -2.77909338e-01 -6.90619528e-01
3.52575898e-01 1.79324239e-01 6.53616130e-01 -5.65705180e-01
-5.85882187e-01 1.65151477e-01 4.41548616e-01 -1.05733478e+00
-5.73930323e-01 -8.45690310e-01 -1.13259768e+00 -1.09698705e-01
3.53034109e-01 -7.97204003e-02 6.18631661e-01 1.20505428e+00
4.53805000e-01 4.88145232e-01 2.31689945e-01 -1.07229567e+00
-3.45024139e-01 -8.60322177e-01 -4.08256114e-01 5.72918892e-01
1.79072738e-01 -5.67747772e-01 -2.44964182e-01 4.18872803e-01]
|
[8.445093154907227, 0.5660225749015808]
|
4da52274-ea85-4a6e-8de7-9c1fc4fb9903
|
the-generic-holdout-preventing-false
|
1809.05596
| null |
http://arxiv.org/abs/1809.05596v1
|
http://arxiv.org/pdf/1809.05596v1.pdf
|
The Generic Holdout: Preventing False-Discoveries in Adaptive Data Science
|
Adaptive data analysis has posed a challenge to science due to its ability to
generate false hypotheses on moderately large data sets. In general, with
non-adaptive data analyses (where queries to the data are generated without
being influenced by answers to previous queries) a data set containing $n$
samples may support exponentially many queries in $n$. This number reduces to
linearly many under naive adaptive data analysis, and even sophisticated
remedies such as the Reusable Holdout (Dwork et. al 2015) only allow
quadratically many queries in $n$.
In this work, we propose a new framework for adaptive science which
exponentially improves on this number of queries under a restricted yet
scientifically relevant setting, where the goal of the scientist is to find a
single (or a few) true hypotheses about the universe based on the samples. Such
a setting may describe the search for predictive factors of some disease based
on medical data, where the analyst may wish to try a number of predictive
models until a satisfactory one is found.
Our solution, the Generic Holdout, involves two simple ingredients: (1) a
partitioning of the data into a exploration set and a holdout set and (2) a
limited exposure strategy for the holdout set. An analyst is free to use the
exploration set arbitrarily, but when testing hypotheses against the holdout
set, the analyst only learns the answer to the question: "Is the given
hypothesis true (empirically) on the holdout set?" -- and no more information,
such as "how well" the hypothesis fits the holdout set. The resulting scheme is
immediate to analyze, but despite its simplicity we do not believe our method
is obvious, as evidenced by the many violations in practice.
Our proposal can be seen as an alternative to pre-registration, and allows
researchers to get the benefits of adaptive data analysis without the problems
of adaptivity.
|
['Jarosław Błasiok', 'Preetum Nakkiran']
|
2018-09-14
| null | null | null | null |
['holdout-set']
|
['computer-vision']
|
[ 1.82936653e-01 2.76631445e-01 -1.86215729e-01 -2.61865169e-01
-7.56884933e-01 -8.41295302e-01 3.12218010e-01 4.31709379e-01
-6.44945025e-01 7.29385138e-01 -1.78721413e-01 -7.05122292e-01
-7.01192498e-01 -1.07009554e+00 -8.17682445e-01 -7.99394488e-01
-1.14966281e-01 8.96797121e-01 2.24803641e-01 -1.20837763e-02
4.77228284e-01 4.30418134e-01 -1.55248928e+00 -1.74611703e-01
9.06263709e-01 7.62352169e-01 6.28178939e-02 7.21925914e-01
-1.08339358e-02 3.68847102e-01 -4.16367322e-01 -5.13424635e-01
4.08014834e-01 -4.70857322e-01 -8.13104093e-01 -1.57002926e-01
2.11750135e-01 -2.52854899e-02 1.16557099e-01 1.19504225e+00
3.84921700e-01 1.50625691e-01 3.66141856e-01 -1.28527069e+00
-2.45250881e-01 5.59890747e-01 -2.71323711e-01 2.08037704e-01
2.98518360e-01 3.63353640e-01 1.00755179e+00 -6.69677079e-01
7.45536625e-01 9.97630477e-01 5.67819417e-01 3.25416028e-01
-1.50093579e+00 -7.79290199e-01 2.26384737e-02 7.79756904e-02
-1.38909292e+00 -4.64765847e-01 2.77247280e-01 -4.43481356e-01
4.74085003e-01 7.51440644e-01 6.83607221e-01 7.63506651e-01
4.43027280e-02 7.15133324e-02 1.03784370e+00 -3.17720324e-01
8.08745384e-01 3.96660328e-01 2.20538303e-01 5.74278712e-01
5.80110729e-01 2.45191067e-01 -5.54774821e-01 -7.73094177e-01
5.93879223e-01 5.54767251e-02 -1.24444589e-01 -2.51117349e-01
-8.23243678e-01 1.01396811e+00 8.91078487e-02 1.96876988e-01
-3.62399161e-01 -2.20276490e-01 1.07485764e-01 4.45556790e-01
3.04042727e-01 1.03028440e+00 -6.97680354e-01 -4.47968766e-02
-8.80896568e-01 6.70103133e-01 8.79235506e-01 5.72035789e-01
7.43870735e-01 -5.57921946e-01 2.71129906e-02 3.54767621e-01
1.82367653e-01 3.84497494e-01 3.77009660e-01 -9.90352333e-01
-1.77236535e-02 6.93692446e-01 3.97571087e-01 -8.78806472e-01
-3.83547693e-01 -3.66359711e-01 -8.14296961e-01 2.07735211e-01
8.51690233e-01 -1.16348431e-01 -6.27291143e-01 2.04549026e+00
6.81767344e-01 -5.94292991e-02 -2.49391392e-01 6.26487672e-01
2.76518703e-01 3.79160821e-01 -2.05852240e-02 -6.13878310e-01
1.43662262e+00 -1.08495943e-01 -4.32809055e-01 -1.08372204e-01
5.89984953e-01 -3.90953720e-01 1.20892167e+00 7.84274638e-01
-1.18324876e+00 -2.64458537e-01 -8.75220776e-01 5.80787808e-02
-3.69102240e-01 -6.10226810e-01 6.76933169e-01 6.11548305e-01
-9.16187048e-01 4.80681151e-01 -6.73069954e-01 -2.62207121e-01
4.36197072e-01 4.71009284e-01 -1.50492251e-01 -1.38252705e-01
-1.18446743e+00 6.02819264e-01 3.42177927e-01 -1.66157737e-01
-6.72562063e-01 -8.45469236e-01 -4.16497171e-01 2.70325094e-01
9.43054676e-01 -8.43360722e-01 7.78070152e-01 -7.91737795e-01
-8.88713419e-01 8.29249680e-01 -3.66005808e-01 -3.84467959e-01
7.89166689e-01 5.97442761e-02 -9.42169577e-02 -3.18928175e-02
2.05827758e-01 1.57446742e-01 5.97151041e-01 -8.02298129e-01
-5.11781096e-01 -7.41065145e-01 2.20142305e-02 -1.98707372e-01
-9.97541100e-02 1.11806765e-01 -3.50138396e-01 -3.70876789e-01
2.32012644e-01 -9.95450139e-01 -7.01574743e-01 2.62647867e-01
-4.25013363e-01 -4.60229844e-01 4.05309528e-01 -2.19029084e-01
1.35574877e+00 -2.06951666e+00 -1.79723635e-01 6.39747202e-01
4.65189338e-01 -6.38203993e-02 1.60890535e-01 1.64864540e-01
-1.70126244e-01 5.04210293e-01 -3.91958386e-01 1.62862375e-01
-1.25880003e-01 -1.16058281e-02 -3.59749168e-01 5.05042851e-01
-1.01124883e-01 6.43493831e-01 -9.07077849e-01 -3.31991106e-01
-2.38826305e-01 -1.53512001e-01 -7.88017213e-01 2.55178750e-01
-4.10151124e-01 1.65310293e-01 -5.68933189e-01 4.56420094e-01
7.51947582e-01 -5.82803428e-01 2.77441025e-01 3.39527935e-01
-2.40510151e-01 1.14445902e-01 -1.43725097e+00 1.15002382e+00
2.09304988e-01 1.76524013e-01 2.43204497e-02 -1.25954080e+00
6.84742868e-01 2.08983943e-01 4.71728295e-01 -2.98801005e-01
-7.91558400e-02 3.04554135e-01 2.26581246e-01 -6.23386741e-01
1.56217530e-01 -5.57912409e-01 -7.64784589e-02 5.58070481e-01
-3.74832004e-01 5.94978742e-02 8.99531618e-02 4.11114395e-02
1.42420101e+00 -3.27662796e-01 5.31058192e-01 -3.88583571e-01
1.71278611e-01 2.14686245e-01 7.51288235e-01 1.28579700e+00
-6.60442282e-03 3.91841352e-01 8.80641639e-01 -4.51053411e-01
-1.02721786e+00 -1.08033586e+00 -4.35404718e-01 8.74645650e-01
-1.08255185e-01 -3.84686202e-01 -5.06076753e-01 -6.71285033e-01
1.00800596e-01 6.68231010e-01 -9.34453785e-01 -6.06702194e-02
-3.13348293e-01 -8.73503983e-01 1.77318901e-01 2.75755059e-02
2.88251251e-01 -7.50878572e-01 -8.38827968e-01 4.28874865e-02
8.16284399e-03 -4.45197612e-01 -2.85266668e-01 2.67027706e-01
-8.96734178e-01 -1.36627948e+00 -1.45788819e-01 -1.08809069e-01
6.66106761e-01 -3.18839438e-02 9.60237682e-01 2.28086233e-01
-3.34599286e-01 1.04233459e-01 -1.00963712e-01 -7.27424681e-01
-5.29619396e-01 -1.98860452e-01 2.05715507e-01 -1.25113130e-01
5.74281156e-01 -4.98046964e-01 -7.40755498e-01 3.69426519e-01
-1.06992614e+00 -3.84088635e-01 5.54181337e-01 7.40383089e-01
6.77939832e-01 3.13106596e-01 7.30158150e-01 -1.40493894e+00
4.12982494e-01 -9.29778755e-01 -7.92001605e-01 4.14984047e-01
-1.03967023e+00 1.75559729e-01 6.38593435e-01 -7.26662159e-01
-5.70065200e-01 -1.48894891e-01 7.68668354e-02 -3.23481113e-01
-4.75372392e-04 7.83093452e-01 -2.45305240e-01 2.90999889e-01
1.04545975e+00 6.17185459e-02 1.45618156e-01 -5.41559279e-01
8.56913701e-02 3.95214826e-01 3.60823125e-01 -5.19251823e-01
7.37971067e-01 4.06372607e-01 1.72992155e-01 -6.53090477e-01
-7.61440039e-01 -1.97700158e-01 -3.38891655e-01 1.65544644e-01
6.43694043e-01 -4.10427213e-01 -9.85569239e-01 2.09672190e-02
-6.85070693e-01 -1.18497849e-01 -7.86071956e-01 4.62700129e-01
-3.97436887e-01 1.80998579e-01 1.06567658e-01 -1.03053749e+00
-4.10814397e-02 -1.01907587e+00 4.89877820e-01 1.83583558e-01
-6.07641578e-01 -9.00348246e-01 6.80612549e-02 3.18146735e-01
4.42862421e-01 3.75224978e-01 1.23376882e+00 -1.25590324e+00
-6.74541235e-01 -3.69656175e-01 2.44797692e-01 -1.98559374e-01
5.91559112e-02 -4.02211957e-02 -9.24659729e-01 -4.54947293e-01
3.98860723e-01 -2.48176277e-01 4.97748226e-01 4.80055839e-01
1.58346963e+00 -7.56221354e-01 -4.55869466e-01 4.04557884e-01
1.27926588e+00 2.82831997e-01 4.76490378e-01 2.97688216e-01
8.92899092e-03 6.96202278e-01 4.80441570e-01 6.56824768e-01
4.34713773e-02 4.26051050e-01 2.25088328e-01 5.89707270e-02
5.15515864e-01 -1.49181634e-01 -1.22726066e-02 1.86068714e-02
5.99369407e-02 -1.86982825e-01 -8.64635468e-01 4.20972377e-01
-1.65570593e+00 -9.85477865e-01 -2.23781858e-02 2.88794732e+00
1.12254393e+00 3.45792800e-01 2.73157477e-01 1.07956361e-02
4.97257888e-01 -3.80525708e-01 -1.18631983e+00 -1.83603734e-01
-8.54661837e-02 3.02689642e-01 3.57604176e-01 5.17351747e-01
-6.59786880e-01 3.73159081e-01 6.77899122e+00 8.37277353e-01
-9.35237646e-01 -6.39082640e-02 9.18419600e-01 -3.23398560e-01
-6.50654852e-01 4.24847066e-01 -7.09477425e-01 4.93718356e-01
1.20129156e+00 -6.33712292e-01 3.38779777e-01 7.18494594e-01
2.69289106e-01 -5.13075233e-01 -1.45648646e+00 6.18510962e-01
-2.73054302e-01 -1.34020603e+00 -1.10504322e-01 5.63869953e-01
2.85063267e-01 -2.57401317e-01 1.13563575e-01 1.30769596e-01
6.06055558e-01 -1.25635755e+00 4.74710017e-01 7.72472680e-01
7.85646498e-01 -5.81015408e-01 4.19463307e-01 8.51406515e-01
-5.13217509e-01 -1.62524238e-01 -3.17347884e-01 -8.30128714e-02
-3.86846542e-01 7.27153540e-01 -9.62560058e-01 3.33576500e-01
8.05647254e-01 -6.64284974e-02 -5.58707774e-01 9.44196641e-01
3.06173980e-01 8.21469665e-01 -7.00664461e-01 -1.75735857e-02
-2.38331005e-01 -1.96500331e-01 6.59634531e-01 6.86611593e-01
4.01089281e-01 4.13269937e-01 -7.12139346e-03 1.09276986e+00
9.27496552e-02 1.32053390e-01 -6.92326009e-01 -4.39914390e-02
7.19966650e-01 9.25252140e-01 -6.12358987e-01 -3.29345167e-01
-1.32629275e-01 3.31953853e-01 2.37354934e-01 2.11940482e-01
-2.30131358e-01 -1.78288475e-01 4.04729068e-01 5.05260646e-01
3.07988506e-02 7.71955103e-02 -4.84890670e-01 -9.70511198e-01
-1.01480126e-01 -1.15956938e+00 1.01472294e+00 -5.48260629e-01
-1.31798553e+00 9.59321484e-02 2.42765412e-01 -8.58983934e-01
-4.12817359e-01 -4.39704001e-01 -2.79850751e-01 9.54157174e-01
-8.10605466e-01 -4.13461089e-01 1.04844071e-01 5.11464059e-01
2.51472354e-01 1.25732169e-01 8.00925314e-01 -1.71437666e-01
-3.88852745e-01 6.49980724e-01 1.81930766e-01 -3.11503917e-01
6.10293746e-01 -1.21853626e+00 -5.80111146e-02 7.55042970e-01
-1.25656933e-01 1.09237194e+00 1.12491703e+00 -8.03365648e-01
-1.42953002e+00 -6.75650060e-01 9.70523834e-01 -7.31202006e-01
6.18941963e-01 -4.54317123e-01 -1.29604828e+00 6.24054015e-01
-3.83889049e-01 -5.56657044e-03 9.21438336e-01 4.34427470e-01
-1.61048591e-01 -1.40313104e-01 -1.50096035e+00 5.48508108e-01
8.05482924e-01 -5.85728213e-02 -4.44791317e-01 3.94219667e-01
5.70656657e-01 -1.10184506e-01 -8.01896632e-01 2.89408207e-01
6.13053858e-01 -9.13589180e-01 7.03133523e-01 -1.01214254e+00
2.61587441e-01 -1.63407341e-01 -1.34897053e-01 -8.93344402e-01
-2.96479255e-01 -7.91976452e-01 1.12614624e-01 9.36993122e-01
5.28003156e-01 -8.28072011e-01 9.10068929e-01 1.37032688e+00
4.81410056e-01 -9.25353706e-01 -1.17979848e+00 -5.07243872e-01
2.76555538e-01 -5.73415279e-01 6.77156508e-01 1.14186084e+00
-1.10424735e-01 1.42927513e-01 -9.92021896e-03 3.77587587e-01
6.52262568e-01 2.63583422e-01 9.09089923e-01 -1.53585517e+00
-6.21629596e-01 -3.01826566e-01 -1.51026705e-02 -7.67768025e-01
-4.69257355e-01 -7.57824421e-01 -1.07359886e-01 -9.07405138e-01
4.97598350e-01 -9.00758386e-01 -2.12681487e-01 4.45032090e-01
-4.40596938e-01 -2.90256172e-01 -1.32264361e-01 3.71145785e-01
-3.11356008e-01 3.02701257e-02 8.86652708e-01 2.82976985e-01
-4.08947796e-01 2.26837650e-01 -1.18301404e+00 7.12792575e-01
5.60425997e-01 -6.72590315e-01 -4.63198572e-01 2.24854231e-01
7.59774864e-01 2.37746224e-01 4.50057834e-01 -3.81636441e-01
4.00065154e-01 -6.22778654e-01 3.74671072e-01 -3.11192423e-01
-2.95720529e-02 -6.57662630e-01 5.53123653e-01 6.50135994e-01
-5.66518962e-01 6.06093109e-02 2.78313644e-02 5.71367800e-01
3.36311489e-01 -4.71749902e-01 7.01337397e-01 -3.77921492e-01
-3.00698932e-02 3.87107313e-01 -2.10585549e-01 2.93933213e-01
9.47730064e-01 -1.22548185e-01 -3.78453672e-01 -3.38448226e-01
-8.46454620e-01 3.92495245e-01 5.67035079e-01 -2.26213355e-02
5.03811479e-01 -8.18324745e-01 -7.85362065e-01 3.21204841e-01
3.67812486e-03 3.47730100e-01 2.18618393e-01 8.39709759e-01
-7.36260712e-02 1.58995703e-01 3.86715382e-01 -4.77416188e-01
-1.11804914e+00 8.94084454e-01 3.17814112e-01 -3.00412387e-01
-5.15613794e-01 8.19373131e-01 4.23504323e-01 -1.61765948e-01
3.44627276e-02 -1.93169653e-01 1.52343735e-01 1.33838981e-01
6.16872013e-01 3.91316622e-01 -1.83455810e-01 2.75770612e-02
-3.46450388e-01 1.31826103e-01 -1.98409915e-01 -1.94489986e-01
1.45922852e+00 -1.52354511e-02 -2.64303744e-01 7.54455149e-01
8.71360302e-01 3.99407476e-01 -8.54131877e-01 -3.81367892e-01
7.37455636e-02 -6.44411147e-01 -2.65525013e-01 -9.23930526e-01
-4.98063743e-01 4.33436126e-01 5.02196074e-01 6.63791478e-01
1.16135466e+00 2.84972787e-01 1.49443969e-01 4.90062177e-01
1.85216293e-01 -8.47155631e-01 -1.52701333e-01 -4.36499082e-02
8.89430344e-01 -1.23412216e+00 3.25019121e-01 -6.34582564e-02
-2.73209363e-01 9.08485472e-01 4.54140455e-01 5.57273999e-02
6.91334367e-01 1.36456698e-01 -2.34170511e-01 -5.67822456e-01
-1.08494329e+00 2.16190323e-01 1.04255475e-01 1.58073083e-01
4.03993316e-02 -5.74474633e-02 -4.24408615e-01 7.19318569e-01
-6.12640679e-01 1.08944856e-01 4.55499530e-01 7.11850584e-01
-5.83013594e-01 -1.06131375e+00 -6.83128178e-01 8.88684034e-01
-5.13936937e-01 -2.58340333e-02 -3.90467376e-01 8.94325912e-01
3.66645008e-01 9.05156791e-01 1.43321648e-01 -4.97484617e-02
2.72602320e-01 1.95055053e-01 1.33056477e-01 -6.42578185e-01
-4.64276254e-01 1.19972907e-01 -1.87892988e-01 -4.70994920e-01
-1.30066305e-01 -9.57886875e-01 -1.07500017e+00 -3.87465417e-01
-3.04605722e-01 5.30552983e-01 3.66021544e-01 1.02029192e+00
3.96687806e-01 -7.92157426e-02 7.15366781e-01 5.49743436e-02
-8.90958488e-01 -7.82269835e-01 -6.34139836e-01 3.52040708e-01
4.92160887e-01 -3.94349605e-01 -5.96175969e-01 -4.02048789e-02]
|
[7.73859977722168, 4.7084641456604]
|
aa697d43-4a5c-4189-b8d3-a6c0c6cdef30
|
humans-need-not-label-more-humans-occlusion
|
2210.03686
| null |
https://arxiv.org/abs/2210.03686v1
|
https://arxiv.org/pdf/2210.03686v1.pdf
|
Humans need not label more humans: Occlusion Copy & Paste for Occluded Human Instance Segmentation
|
Modern object detection and instance segmentation networks stumble when picking out humans in crowded or highly occluded scenes. Yet, these are often scenarios where we require our detectors to work well. Many works have approached this problem with model-centric improvements. While they have been shown to work to some extent, these supervised methods still need sufficient relevant examples (i.e. occluded humans) during training for the improvements to be maximised. In our work, we propose a simple yet effective data-centric approach, Occlusion Copy & Paste, to introduce occluded examples to models during training - we tailor the general copy & paste augmentation approach to tackle the difficult problem of same-class occlusion. It improves instance segmentation performance on occluded scenarios for "free" just by leveraging on existing large-scale datasets, without additional data or manual labelling needed. In a principled study, we show whether various proposed add-ons to the copy & paste augmentation indeed contribute to better performance. Our Occlusion Copy & Paste augmentation is easily interoperable with any models: by simply applying it to a recent generic instance segmentation model without explicit model architectural design to tackle occlusion, we achieve state-of-the-art instance segmentation performance on the very challenging OCHuman dataset. Source code is available at https://github.com/levan92/occlusion-copy-paste.
|
['Minhoe Hur', 'Dezhao Huang', 'Evan Ling']
|
2022-10-07
| null | null | null | null |
['human-instance-segmentation']
|
['computer-vision']
|
[ 3.53485048e-01 4.82956201e-01 -8.47461149e-02 -4.68563139e-01
-7.35081732e-01 -2.85851300e-01 6.90084577e-01 4.90599088e-02
-6.46832466e-01 5.59569180e-01 -1.40500933e-01 -2.06745118e-01
1.71085209e-01 -4.27850962e-01 -8.57765138e-01 -4.77021873e-01
2.45698243e-01 9.15160000e-01 7.70176768e-01 -2.23940492e-01
9.39263310e-03 4.41934347e-01 -1.66234493e+00 2.58141011e-01
8.38206887e-01 6.57929480e-01 1.61063701e-01 7.90525436e-01
-1.25542104e-01 4.60214406e-01 -6.05615556e-01 -5.00073850e-01
7.14283347e-01 -6.18161261e-02 -8.37831318e-01 3.64181936e-01
8.84635806e-01 -2.98213303e-01 1.13714049e-02 5.88494003e-01
6.50060534e-01 8.25379938e-02 4.65707868e-01 -1.24582040e+00
-8.82652551e-02 2.81168848e-01 -8.49838674e-01 2.00846002e-01
1.40560195e-01 4.79102433e-01 7.61790872e-01 -7.56894290e-01
8.10604692e-01 1.33891237e+00 8.32214594e-01 6.22243583e-01
-1.36161780e+00 -3.69661301e-01 4.47099745e-01 2.23341465e-01
-1.04118276e+00 -4.42302555e-01 7.38495767e-01 -4.93173420e-01
8.86482358e-01 3.79964024e-01 7.00993717e-01 1.03232765e+00
-3.57848942e-01 1.28381705e+00 1.11129630e+00 -5.25777996e-01
2.14463044e-02 3.75476867e-01 3.04237187e-01 4.50969130e-01
3.11436087e-01 1.10457197e-01 -7.36825094e-02 1.68903664e-01
5.79165876e-01 -1.31192476e-01 -2.45798558e-01 -7.55081415e-01
-1.04724121e+00 8.04842949e-01 6.54688835e-01 2.52535641e-01
-1.92698091e-01 1.25968859e-01 3.98721814e-01 -9.14348010e-03
6.84296370e-01 5.30242741e-01 -7.47368455e-01 1.45832419e-01
-1.23750699e+00 5.74773729e-01 7.13069379e-01 9.83268321e-01
7.38114178e-01 -2.69134998e-01 -3.30840558e-01 7.99138188e-01
1.36279225e-01 1.86022341e-01 2.76722670e-01 -9.93378997e-01
5.65615058e-01 6.74819589e-01 1.77198455e-01 -6.97014034e-01
-7.01608956e-01 -6.99212432e-01 -4.80307370e-01 3.09456199e-01
8.53803456e-01 -8.70466679e-02 -1.29211962e+00 1.58379948e+00
8.68067682e-01 2.29397789e-01 -3.23550433e-01 9.30881858e-01
7.85222948e-01 2.98427492e-01 2.63093174e-01 1.52606398e-01
1.55850732e+00 -1.41907573e+00 -4.75119263e-01 -6.53160036e-01
8.31453860e-01 -7.04052567e-01 1.20207989e+00 3.70512187e-01
-1.07372224e+00 -6.35676384e-01 -9.85050261e-01 -2.35141590e-01
-4.08398628e-01 1.94991350e-01 6.60598338e-01 8.03977191e-01
-8.79555106e-01 4.68745470e-01 -7.38671005e-01 -4.05002028e-01
7.82010496e-01 3.55772823e-01 -4.54622746e-01 -2.22171113e-01
-6.79763436e-01 9.94463265e-01 3.84624600e-01 2.03452975e-01
-6.31658196e-01 -6.54571891e-01 -7.86892533e-01 -2.90238857e-01
9.06462550e-01 -7.83826590e-01 1.41410601e+00 -8.83178353e-01
-1.13072276e+00 1.04650235e+00 -2.89013386e-01 -7.14656651e-01
1.16301501e+00 -5.52418888e-01 2.20590100e-01 4.10261042e-02
1.01627640e-01 1.15738142e+00 8.03383887e-01 -1.60034251e+00
-5.90744793e-01 -2.81250328e-01 2.99210042e-01 1.08849198e-01
9.81234945e-03 -6.06604442e-02 -7.10245788e-01 -5.73700309e-01
4.85841325e-03 -9.86606956e-01 -5.81039667e-01 5.73182032e-02
-4.91291553e-01 -2.63069868e-01 1.14595568e+00 -6.96840286e-01
8.58841360e-01 -1.85753644e+00 3.93047463e-03 -3.09102908e-02
1.82667822e-01 8.48306060e-01 -2.05678895e-01 1.76774561e-01
1.56444266e-01 2.25377157e-01 -7.44952679e-01 -9.30030823e-01
8.88715610e-02 3.47887456e-01 9.69935115e-03 4.07633901e-01
5.21538377e-01 1.14221752e+00 -6.61331952e-01 -7.17260718e-01
5.73415279e-01 6.62115693e-01 -7.30788171e-01 1.67163074e-01
-5.20141363e-01 7.48624206e-01 -2.13859737e-01 5.10588050e-01
7.90285230e-01 4.12219726e-02 -1.58768147e-01 1.03350550e-01
8.23047459e-02 -1.19300326e-02 -1.42317307e+00 1.64186704e+00
-3.40109766e-01 6.62053704e-01 1.37200072e-01 -1.04020464e+00
5.80478370e-01 1.68067575e-01 1.96321249e-01 -6.17475986e-01
2.30070397e-01 2.48857588e-01 8.91614631e-02 -5.80009699e-01
3.21630001e-01 -8.77520964e-02 1.59655094e-01 5.58255985e-02
-5.49299419e-02 -2.85321057e-01 4.67685431e-01 1.85867935e-01
9.53490973e-01 6.36754692e-01 3.13760072e-01 -7.04675168e-02
5.06307602e-01 1.99222073e-01 5.35404146e-01 9.58218575e-01
-4.18192655e-01 9.41970289e-01 3.94547850e-01 -4.59984481e-01
-1.14266074e+00 -5.47030389e-01 -3.34516674e-01 1.04652178e+00
-1.39403343e-03 -2.34092236e-01 -1.21981061e+00 -8.53642404e-01
-2.05915332e-01 5.25603235e-01 -5.72732747e-01 3.28044683e-01
-8.50704491e-01 -7.99189031e-01 4.46268708e-01 6.54245079e-01
6.19967163e-01 -1.26110208e+00 -9.13695633e-01 3.38666856e-01
-5.22058308e-02 -1.36315298e+00 -7.76806399e-02 1.80321887e-01
-7.19621956e-01 -1.17940688e+00 -9.25333798e-01 -5.75715065e-01
6.12199187e-01 1.25842646e-01 1.25229454e+00 4.64932263e-01
-7.24000275e-01 2.76618749e-01 -3.50676000e-01 -6.30645156e-01
-1.27934933e-01 2.69302934e-01 -3.55903804e-01 -1.33764639e-01
2.20585793e-01 -4.17083174e-01 -8.09430659e-01 4.81685847e-01
-9.51626122e-01 1.46865457e-01 5.92306852e-01 5.60140729e-01
5.74443221e-01 -3.11450392e-01 4.76100206e-01 -1.29184484e+00
2.25885689e-01 -5.53058684e-02 -5.17288506e-01 -4.66864929e-02
-2.71910816e-01 -1.26592308e-01 3.23773772e-01 -4.56039101e-01
-9.63550508e-01 4.41123754e-01 -4.58425343e-01 -3.28089058e-01
-7.73776829e-01 -7.61808828e-02 -2.56132692e-01 -5.66089340e-02
7.61453271e-01 -2.97618926e-01 -2.36148998e-01 -6.85587108e-01
7.38337517e-01 4.71312910e-01 5.27770102e-01 -5.51607966e-01
8.73896718e-01 6.99073136e-01 -1.11384161e-01 -7.15205550e-01
-9.59242463e-01 -8.27026665e-01 -1.10555279e+00 -2.73371339e-02
1.00967181e+00 -7.40716159e-01 -2.79802412e-01 3.93393159e-01
-1.22648168e+00 -7.15274513e-01 -5.09791493e-01 8.21997151e-02
-4.98732150e-01 3.86613280e-01 -3.26262206e-01 -9.82091248e-01
-6.99507520e-02 -1.17711008e+00 1.35786247e+00 1.03302523e-01
-1.62376598e-01 -8.20714653e-01 -9.05242413e-02 6.86992586e-01
3.76717567e-01 5.25056243e-01 3.73278677e-01 -7.78425694e-01
-6.18439734e-01 -4.83342379e-01 -3.14596295e-01 3.98448378e-01
-2.93194979e-01 -1.50535867e-01 -1.28903627e+00 -2.46989369e-01
-8.85222182e-02 -2.71382213e-01 1.06781340e+00 3.85063767e-01
1.12550616e+00 -2.03098118e-01 -4.84085023e-01 4.72646803e-01
1.16943073e+00 -1.89081907e-01 7.20180988e-01 4.53599453e-01
9.52532768e-01 9.53380764e-01 6.40752792e-01 8.76373723e-02
3.99628937e-01 1.01300871e+00 4.37917411e-01 -6.20044649e-01
-5.52550316e-01 1.93041973e-02 -1.68752179e-01 2.01799288e-01
-1.03102036e-01 -2.70768911e-01 -9.96719599e-01 7.47721076e-01
-2.05522895e+00 -7.03903675e-01 -5.92839003e-01 2.02368975e+00
6.73722088e-01 4.41462457e-01 3.99329484e-01 3.23219538e-01
6.44389153e-01 1.25905931e-01 -3.53745759e-01 -1.77051142e-01
4.66444008e-02 2.50755519e-01 3.98808628e-01 6.50517285e-01
-1.35969317e+00 1.12467372e+00 5.35764933e+00 8.18991840e-01
-7.94146299e-01 4.80319709e-01 6.83319807e-01 -1.79339036e-01
8.49827006e-02 1.51367992e-01 -9.46388483e-01 2.69675553e-01
6.04394913e-01 5.45273125e-01 -8.16338211e-02 1.02190304e+00
2.70156682e-01 -3.87150705e-01 -1.11720693e+00 6.87312126e-01
3.48415449e-02 -9.80708003e-01 -3.72996002e-01 8.68411586e-02
7.16065526e-01 3.43863852e-02 -8.46675336e-02 4.07406300e-01
1.01330699e-02 -9.21515882e-01 7.10494578e-01 2.42489472e-01
2.89649040e-01 -4.00584728e-01 9.38606679e-01 5.71651876e-01
-1.04375815e+00 8.25158358e-02 -3.26736838e-01 -1.86696157e-01
4.73497361e-01 7.64688015e-01 -9.82499957e-01 4.37484294e-01
6.01606071e-01 3.59084874e-01 -1.01220417e+00 1.41056895e+00
-4.68352884e-01 4.74723041e-01 -5.49034476e-01 3.81892860e-01
3.05030465e-01 9.25600007e-02 6.07617259e-01 1.46104276e+00
-1.94634363e-01 -1.50727630e-01 3.48946601e-01 6.73837066e-01
1.54768694e-02 1.39170349e-01 -4.69666570e-01 6.04687870e-01
8.89644101e-02 1.36834800e+00 -1.07022023e+00 -6.35037422e-01
-1.96529329e-01 8.63950729e-01 4.47054833e-01 2.48855650e-01
-9.36867118e-01 -1.20017096e-01 2.17218205e-01 5.61360240e-01
4.09343779e-01 -1.07854508e-01 -3.90074313e-01 -9.23829257e-01
3.71789128e-01 -8.11460495e-01 1.81418806e-01 -5.06046891e-01
-1.19466364e+00 5.80219090e-01 3.13106596e-01 -8.99267972e-01
-1.26614720e-01 -6.42561972e-01 -5.81498146e-01 5.49691617e-01
-1.71577835e+00 -1.34836459e+00 -4.13502336e-01 3.72745693e-01
8.72256935e-01 3.43404949e-01 3.55093420e-01 4.18282866e-01
-6.67228222e-01 4.78148252e-01 -4.88951415e-01 2.09534512e-04
5.77475429e-01 -1.42970979e+00 6.02573395e-01 1.03543353e+00
3.63635510e-01 3.68434340e-01 8.67389381e-01 -6.16293192e-01
-6.73909783e-01 -1.12227404e+00 6.52293503e-01 -8.85789633e-01
2.03835800e-01 -7.10445464e-01 -1.01707947e+00 7.44139552e-01
1.23398826e-01 3.01003724e-01 1.48623914e-01 9.34159681e-02
-2.85970867e-01 2.65259951e-01 -1.32745051e+00 5.90302646e-01
1.27019334e+00 3.88744287e-02 -6.56345904e-01 6.71757102e-01
8.37202728e-01 -5.11920869e-01 -4.72712874e-01 6.44668937e-01
2.70215362e-01 -1.13761640e+00 1.10888934e+00 -5.72443545e-01
2.31790319e-01 -4.22046959e-01 1.29278764e-01 -8.75326157e-01
7.49187618e-02 -6.01105213e-01 -1.84688076e-01 1.23116136e+00
4.57012147e-01 -7.13396668e-01 9.03572321e-01 6.87157810e-01
-3.29687357e-01 -9.36625957e-01 -1.05914044e+00 -8.58714879e-01
1.33030415e-01 -6.84873104e-01 4.29253280e-01 7.25054562e-01
-4.89272743e-01 1.24881521e-01 -3.79015535e-01 8.60515833e-02
5.52936733e-01 -2.39760354e-01 1.37700021e+00 -1.23871231e+00
-5.02342582e-01 -4.20501590e-01 -4.80862230e-01 -1.02759576e+00
-7.38652349e-02 -7.24708915e-01 1.98679075e-01 -1.94325054e+00
8.39937478e-02 -7.03051627e-01 1.62004247e-01 6.88261151e-01
-4.82431710e-01 5.59930205e-01 5.85537434e-01 2.30377540e-01
-7.84636080e-01 2.83202976e-01 1.27031708e+00 -5.62456660e-02
-2.85336673e-01 1.91665977e-01 -5.44097364e-01 9.35564697e-01
8.26972008e-01 -5.98269343e-01 -2.72201538e-01 -4.26980615e-01
-2.17033148e-01 -3.95226657e-01 6.52253330e-01 -1.30841291e+00
-1.69825554e-02 2.18366161e-01 2.81196803e-01 -4.75580513e-01
5.41157842e-01 -8.64019871e-01 -1.93669964e-02 5.02549529e-01
-1.45737082e-01 -1.74951643e-01 3.85188878e-01 5.73721528e-01
8.15944523e-02 -4.03803319e-01 7.80332088e-01 -2.92001784e-01
-7.10142136e-01 2.24303275e-01 1.14849038e-01 1.58526570e-01
1.15680158e+00 -3.83548021e-01 -3.56583446e-01 -4.20848392e-02
-9.24929440e-01 2.77795732e-01 5.30742586e-01 4.81567442e-01
1.52485535e-01 -6.86391413e-01 -8.14995110e-01 6.33350760e-03
-7.85799045e-03 5.35283744e-01 2.00248003e-01 1.10496128e+00
-4.74411339e-01 3.48822474e-01 1.09324865e-01 -9.55056667e-01
-1.25010860e+00 6.14237368e-01 3.61015856e-01 -5.24314642e-01
-8.57404113e-01 9.82125759e-01 3.80784750e-01 -6.47218227e-01
2.84752309e-01 -3.57980907e-01 -5.44906966e-02 6.99525326e-02
3.75006795e-01 2.33113527e-01 3.28469962e-01 -5.75633109e-01
-3.00029546e-01 6.47939682e-01 -2.18895927e-01 -1.01314977e-01
1.26516271e+00 3.33721451e-02 2.92265683e-01 2.16384664e-01
8.91462743e-01 -1.24960087e-01 -1.58771729e+00 -1.08218484e-01
7.95898288e-02 -5.50984323e-01 -2.23692819e-01 -8.29469085e-01
-1.11089385e+00 1.08800316e+00 7.22521186e-01 1.77005842e-01
8.16606939e-01 2.54896581e-01 6.59649968e-01 1.99429423e-01
3.45271349e-01 -9.94088411e-01 -1.81214940e-02 2.48522103e-01
7.84274757e-01 -1.56991374e+00 1.83871731e-01 -7.29095042e-01
-5.71370482e-01 5.28651059e-01 8.59748304e-01 -1.18496284e-01
2.49946117e-01 1.07309423e-01 2.08144635e-01 -2.98947543e-01
-3.80186111e-01 -6.88155890e-01 3.51356715e-01 8.30797613e-01
3.42336684e-01 1.72551554e-02 -4.62394714e-01 2.47278631e-01
-2.22722977e-01 -1.82386622e-01 2.93701679e-01 1.03465199e+00
-4.79537755e-01 -1.29983068e+00 -5.54226577e-01 4.57537442e-01
-3.46324325e-01 9.18099657e-02 -3.85055214e-01 1.25430226e+00
5.39362729e-01 9.16174650e-01 -7.83878416e-02 8.33528489e-02
6.22137666e-01 1.84374720e-01 4.16853428e-01 -8.22245836e-01
-7.93204784e-01 1.97071820e-01 1.33595452e-01 -6.25252843e-01
-6.15683496e-01 -6.90162659e-01 -1.12788785e+00 1.08164720e-01
-5.68711579e-01 -2.73687690e-01 7.64954805e-01 1.06301200e+00
2.51119435e-01 7.38314509e-01 -9.92463622e-03 -1.45216668e+00
-1.43420130e-01 -1.10651386e+00 -5.39668351e-02 7.20614910e-01
2.61206806e-01 -8.34829330e-01 -3.11363488e-01 -3.70879397e-02]
|
[9.325706481933594, 0.34449198842048645]
|
b0f2ab6b-35b8-4316-95cc-7c239611adf6
|
improving-fine-grain-segmentation-via
|
2210.03879
| null |
https://arxiv.org/abs/2210.03879v2
|
https://arxiv.org/pdf/2210.03879v2.pdf
|
Improving Data-Efficient Fossil Segmentation via Model Editing
|
Most computer vision research focuses on datasets containing thousands of images of commonplace objects. However, many high-impact datasets, such as those in medicine and the geosciences, contain fine-grain objects that require domain-expert knowledge to recognize and are time-consuming to collect and annotate. As a result, these datasets contain few labeled images, and current machine vision models cannot train intensively on them. Originally introduced to correct large-language models, model-editing techniques in machine learning have been shown to improve model performance using only small amounts of data and additional training. Using a Mask R-CNN to segment ancient reef fossils in rock sample images, we present a two-part paradigm to improve fossil segmentation with few labeled images: we first identify model weaknesses using image perturbations and then mitigate those weaknesses using model editing. Specifically, we apply domain-informed image perturbations to expose the Mask R-CNN's inability to distinguish between different classes of fossils and its inconsistency in segmenting fossils with different textures. To address these shortcomings, we extend an existing model-editing method for correcting systematic mistakes in image classification to image segmentation with no additional labeled data needed and show its effectiveness in decreasing confusion between different kinds of fossils. We also highlight the best settings for model editing in our situation: making a single edit using all relevant pixels in one image (vs. using multiple images, multiple edits, or fewer pixels). Though we focus on fossil segmentation, our approach may be useful in other similar fine-grain segmentation problems where data is limited.
|
['Ruth Fong', 'Adam Maloof', 'Ryan Manzuk', 'Indu Panigrahi']
|
2022-10-08
| null | null | null | null |
['model-editing']
|
['natural-language-processing']
|
[ 6.49725020e-01 2.86478907e-01 2.30396003e-01 -3.56129944e-01
-7.29394853e-01 -9.03920710e-01 4.55595642e-01 8.03183690e-02
-7.23003983e-01 4.99852598e-01 -2.48659864e-01 -4.11804974e-01
2.13813320e-01 -8.53282452e-01 -1.07469153e+00 -5.59953868e-01
2.31216431e-01 6.92525029e-01 5.28398454e-01 -3.53098549e-02
5.19794762e-01 5.87273300e-01 -1.52565026e+00 2.59883732e-01
8.06814194e-01 5.17988026e-01 4.34300601e-01 6.51630521e-01
-3.78758103e-01 4.51192707e-01 -8.12176168e-01 -3.02525520e-01
7.02167332e-01 -2.74225026e-01 -9.51563418e-01 2.51228541e-01
9.04832900e-01 -2.24905118e-01 4.20652702e-02 1.26371706e+00
3.73839170e-01 -2.12683037e-01 7.23088980e-01 -8.81751597e-01
-8.21108282e-01 5.35915196e-01 -7.38794744e-01 3.17724973e-01
-2.06040263e-01 4.86228168e-01 6.39543474e-01 -6.78091466e-01
5.96971929e-01 1.26185429e+00 9.39040959e-01 6.94756448e-01
-1.63031960e+00 -5.83397508e-01 6.30871803e-02 -6.64093345e-02
-1.37686658e+00 -3.48999918e-01 4.43439394e-01 -8.35309923e-01
8.21092486e-01 3.51580769e-01 6.78535044e-01 7.85497665e-01
-6.89520687e-02 3.58552724e-01 1.21044207e+00 -4.60921943e-01
2.23684013e-01 -7.90673047e-02 7.41390139e-02 7.08757520e-01
3.26376438e-01 1.06591165e-01 -3.02565783e-01 -1.18719309e-03
1.01858854e+00 -1.51926294e-01 -1.92469910e-01 1.53061137e-01
-1.53842223e+00 6.40917003e-01 2.88267165e-01 1.39000326e-01
-1.30871922e-01 1.95378974e-01 1.95308447e-01 2.28286505e-01
6.27201319e-01 9.26132977e-01 -6.48088336e-01 2.81334847e-01
-1.22392845e+00 1.30927339e-01 5.25928020e-01 7.61896670e-01
1.12783635e+00 2.14917213e-01 3.89396064e-02 9.94736791e-01
4.61580195e-02 5.10268986e-01 4.60736483e-01 -1.11458945e+00
1.20876312e-01 6.17466569e-01 -2.03202292e-01 -9.01015937e-01
-4.53261107e-01 -2.02565148e-01 -7.76581764e-01 4.66769487e-01
7.49639392e-01 1.70889258e-01 -1.53264785e+00 1.50859177e+00
2.76680619e-01 1.87732175e-01 -1.64917946e-01 8.11306119e-01
6.58362329e-01 4.41565692e-01 1.79313377e-01 2.67208070e-01
1.41753507e+00 -7.23486066e-01 -1.03560694e-01 -6.06673539e-01
6.37332022e-01 -8.30349267e-01 1.46575117e+00 3.84261012e-01
-8.17781329e-01 -5.19719362e-01 -9.35927451e-01 -1.47070110e-01
-6.50687099e-01 -6.04617000e-02 5.34962356e-01 5.89296222e-01
-9.89775419e-01 8.00365865e-01 -7.74799824e-01 -5.05276680e-01
6.16265714e-01 3.25908273e-01 -4.12775517e-01 1.43686026e-01
-7.06198692e-01 9.81398582e-01 4.91698444e-01 4.93944250e-02
-1.03816462e+00 -1.03108430e+00 -7.54258275e-01 -1.22477338e-01
3.61673832e-01 -5.72226226e-01 1.25905359e+00 -1.28609180e+00
-9.46599305e-01 1.38445747e+00 1.29898146e-01 -4.09248352e-01
6.57537758e-01 6.49313554e-02 -5.56457043e-02 1.08120024e-01
3.34259152e-01 1.06592500e+00 8.06953549e-01 -1.48223031e+00
-6.94212615e-01 -3.85383338e-01 -4.48015779e-02 -1.87069610e-01
-9.86889303e-02 -3.09209824e-02 -5.49475670e-01 -1.06936574e+00
1.77763298e-01 -8.39644432e-01 -4.13632065e-01 4.85147864e-01
-2.59332567e-01 2.00089067e-01 6.62059128e-01 -7.26294339e-01
7.27762699e-01 -2.00251532e+00 4.87365527e-03 2.02760369e-01
9.89021510e-02 2.71970809e-01 -3.65267336e-01 -1.04731098e-01
1.09561663e-02 6.42821848e-01 -1.02454376e+00 -1.85739249e-01
-3.93282533e-01 7.39070833e-01 -1.88392743e-01 4.27557439e-01
3.71342838e-01 6.10181749e-01 -7.76941955e-01 -7.89348722e-01
2.61118501e-01 2.89231956e-01 -6.30635202e-01 3.21791042e-03
-5.52360833e-01 5.32139003e-01 -7.40744025e-02 8.45161259e-01
6.94134831e-01 -2.06994057e-01 -5.80047593e-02 -2.20105141e-01
-3.08690459e-01 -2.38588095e-01 -1.05927241e+00 1.55341339e+00
-2.75316834e-01 5.62486768e-01 6.50820732e-02 -9.94469941e-01
7.01611161e-01 -9.08977389e-02 2.99829721e-01 -7.11641669e-01
-1.55195460e-01 3.39127719e-01 7.14344624e-03 -5.52476764e-01
4.58181709e-01 -3.99036258e-01 1.98466346e-01 4.63340849e-01
-2.02101469e-01 -6.36991858e-01 3.15800041e-01 8.55519325e-02
8.93947423e-01 1.64455131e-01 1.19573593e-01 -4.40095842e-01
1.30519450e-01 4.04232711e-01 7.40652740e-01 1.03283167e+00
-4.64191511e-02 1.14533865e+00 6.81061596e-02 -6.51969492e-01
-1.43896210e+00 -7.61326730e-01 -1.36844501e-01 1.02303267e+00
1.31223291e-01 2.17075814e-02 -9.09339726e-01 -5.79649031e-01
7.13447332e-02 5.17252088e-01 -8.44780087e-01 -1.37363538e-01
-6.80285752e-01 -1.20572567e+00 1.04508567e+00 5.30054808e-01
3.81453156e-01 -1.27893913e+00 -7.88914382e-01 7.83364251e-02
-7.20199803e-03 -7.99615920e-01 -4.47781980e-01 3.45249772e-01
-8.53072703e-01 -1.24605978e+00 -8.04610133e-01 -8.73526573e-01
1.13523424e+00 1.85038030e-01 1.12958157e+00 6.21837676e-01
-6.02026403e-01 2.09193602e-01 -3.25937808e-01 -4.07422662e-01
-5.70487440e-01 -5.34867123e-02 -9.32672098e-02 -2.54761636e-01
1.17318913e-01 -3.46118093e-01 -3.99875253e-01 3.60237807e-01
-1.41630471e+00 1.82679251e-01 6.66568398e-01 8.54974389e-01
6.73336625e-01 -2.28885546e-01 2.46145859e-01 -1.19199944e+00
1.93994105e-01 -2.48014942e-01 -6.40446544e-01 5.03850341e-01
-5.06418347e-01 2.36186519e-01 5.86654007e-01 -5.97114146e-01
-9.12360728e-01 2.74840027e-01 -1.56854838e-01 -2.16125980e-01
-3.65337968e-01 4.22576755e-01 3.22117098e-02 -3.58111084e-01
9.97610033e-01 2.11167440e-01 -1.36082172e-01 -6.97629988e-01
2.18830064e-01 5.63186228e-01 8.99103403e-01 -8.70845973e-01
8.44100773e-01 7.02854276e-01 -2.21963316e-01 -9.50149536e-01
-6.86104357e-01 -2.25554779e-01 -7.72841752e-01 -1.65134549e-01
8.70996892e-01 -7.55033493e-01 -2.27012292e-01 6.25520587e-01
-1.06619430e+00 -6.41458094e-01 -4.60134447e-01 2.42218658e-01
-3.14023346e-01 7.08863497e-01 -4.93894339e-01 -3.46627802e-01
-2.93161005e-01 -1.22170770e+00 1.23935819e+00 3.05996031e-01
-5.44964857e-02 -7.03383446e-01 -4.25689965e-02 2.98993409e-01
2.91421294e-01 1.14963874e-01 1.07650757e+00 -5.51403344e-01
-4.83057708e-01 1.31262138e-01 -3.89363140e-01 4.16761696e-01
3.17002684e-01 4.62305307e-01 -9.94563580e-01 -9.60653871e-02
-3.42423707e-01 -1.65642813e-01 1.05800414e+00 2.36411139e-01
1.57018363e+00 -3.23792100e-02 -1.83047786e-01 7.63820887e-01
1.24045038e+00 2.87049800e-01 7.36173749e-01 6.59180641e-01
9.30601954e-01 6.98115766e-01 3.32609087e-01 2.32017174e-01
1.82452276e-01 4.25826788e-01 4.38922137e-01 -5.29321611e-01
-2.37550974e-01 -1.58882607e-02 4.77631912e-02 3.09764504e-01
-1.78524092e-01 -3.01080272e-02 -1.19458485e+00 8.00998509e-01
-1.60922515e+00 -8.19704831e-01 -2.14775696e-01 2.17778015e+00
1.11441612e+00 -3.37334275e-02 -6.35153055e-02 3.14470171e-03
8.62398207e-01 -1.54229999e-01 -7.02607989e-01 -2.71189064e-02
-4.94863510e-01 3.35693598e-01 9.59558606e-01 4.68549967e-01
-1.09299433e+00 1.31217504e+00 7.10741425e+00 7.08602250e-01
-1.17991519e+00 -1.52722254e-01 7.22499430e-01 1.17462672e-01
-4.16277468e-01 1.93738624e-01 -4.96070772e-01 4.52784866e-01
4.48414147e-01 2.28240237e-01 5.20424485e-01 7.35367060e-01
2.29270101e-01 -2.26229519e-01 -1.09514689e+00 9.49162781e-01
2.15144549e-02 -1.51236165e+00 2.73075223e-01 4.49731648e-02
6.46567881e-01 -1.02078812e-02 -2.87847966e-01 -7.78963268e-02
5.05651116e-01 -1.03893101e+00 9.78916466e-01 4.54743028e-01
6.32203877e-01 -4.05619740e-01 4.44416642e-01 2.98683077e-01
-8.86732519e-01 2.10019991e-01 -6.30713999e-01 -9.46519673e-02
-1.34810498e-02 5.59574008e-01 -7.39063382e-01 2.00830698e-01
1.07346761e+00 5.76782942e-01 -1.06967652e+00 1.05514657e+00
1.90191399e-02 5.92487216e-01 -5.50856888e-01 5.14666259e-01
-1.17436042e-02 -2.21066207e-01 1.91467971e-01 1.31996739e+00
2.56851315e-01 4.09982875e-02 1.37373000e-01 8.56227100e-01
-1.06500700e-01 -9.11807194e-02 -4.94779348e-01 -2.75068101e-03
4.41682130e-01 1.05874503e+00 -1.32441795e+00 -6.85276508e-01
-3.69179487e-01 9.26238835e-01 2.48358145e-01 1.14859045e-01
-6.13813221e-01 -1.94023669e-01 6.30386055e-01 1.58918068e-01
2.07238078e-01 -1.70708746e-01 -7.92449832e-01 -1.19600904e+00
-1.48282513e-01 -1.11953926e+00 3.77607822e-01 -7.73988187e-01
-1.37646782e+00 3.25039804e-01 5.39843440e-02 -1.01715279e+00
2.44625598e-01 -7.06815660e-01 -4.02177036e-01 5.27543783e-01
-1.39151669e+00 -1.36760592e+00 -3.72763067e-01 4.25473720e-01
6.62730515e-01 1.21288933e-01 5.84581792e-01 4.67525423e-01
-4.23687756e-01 3.94893885e-01 5.95984533e-02 3.25573266e-01
8.72379422e-01 -1.35675693e+00 7.00439751e-01 9.80411947e-01
1.72334984e-01 6.52841926e-01 8.05623889e-01 -8.42383265e-01
-9.84980285e-01 -1.15561807e+00 4.48250115e-01 -3.17650616e-01
4.38482046e-01 -2.55397767e-01 -1.23650026e+00 9.04431343e-01
-1.20094046e-01 2.71998607e-02 5.38605273e-01 -3.47362071e-01
-6.08522832e-01 8.38176683e-02 -1.27699554e+00 6.80783868e-01
9.86052513e-01 -3.91045600e-01 -6.07862234e-01 4.69885021e-01
4.31847781e-01 -3.07286561e-01 -7.89375126e-01 5.11067927e-01
6.13924205e-01 -6.43747389e-01 1.13645852e+00 -6.82499230e-01
4.57181692e-01 -8.70440364e-01 -1.58458248e-01 -1.20659935e+00
-6.65084273e-02 -7.83455074e-02 6.77614868e-01 1.14886808e+00
5.58884323e-01 -4.01800692e-01 7.88776815e-01 8.04360926e-01
-4.71780717e-01 -4.23168428e-02 -7.77329981e-01 -7.74942815e-01
3.63267303e-01 -3.25955957e-01 5.20721018e-01 1.22853827e+00
-6.37598574e-01 -2.02026069e-01 -5.29282510e-01 3.63289803e-01
6.35823727e-01 6.91870321e-03 9.42427993e-01 -1.35888922e+00
-2.67942667e-01 -4.89753187e-01 -2.53227383e-01 -7.47556269e-01
-4.01405506e-02 -7.59449780e-01 5.10110974e-01 -1.39458001e+00
2.43233562e-01 -5.45795023e-01 1.81781739e-01 9.96326029e-01
-2.54747748e-01 7.39960313e-01 1.52326480e-01 4.41094071e-01
-5.15684411e-02 6.07110299e-02 1.02370131e+00 -5.20007670e-01
1.42120808e-01 -4.17972326e-01 -6.89171851e-01 9.79111910e-01
6.89337611e-01 -8.85737002e-01 4.36313078e-02 -9.10702407e-01
1.03752188e-01 -5.27595997e-01 6.25326812e-01 -8.65936697e-01
5.89725971e-02 -4.48643804e-01 4.93086100e-01 -2.93143034e-01
1.49807855e-02 -8.69450808e-01 5.14763057e-01 7.11078763e-01
-2.19718635e-01 -8.89233649e-02 2.56146371e-01 3.49773228e-01
-6.20774413e-03 -4.44650978e-01 1.24876583e+00 -8.38387728e-01
-9.81897712e-01 1.16824761e-01 -5.37750721e-01 5.17622009e-03
8.80377769e-01 -4.45819527e-01 -4.83031929e-01 1.96219742e-01
-7.13396430e-01 1.27281621e-01 1.08976471e+00 2.35849142e-01
4.53672945e-01 -7.14484870e-01 -6.71442151e-01 2.12075919e-01
1.63692668e-01 5.13329208e-01 2.09695473e-01 3.98723662e-01
-1.22138381e+00 -1.77070782e-01 -4.25250113e-01 -7.50584602e-01
-1.43711495e+00 3.47557396e-01 4.92375880e-01 8.25845543e-03
-6.75264895e-01 9.66121018e-01 4.08481121e-01 -7.08943784e-01
6.31177500e-02 -5.39800525e-01 3.14361602e-02 6.67757019e-02
4.55949932e-01 5.17949201e-02 1.30186215e-01 -6.27687275e-01
-3.71439636e-01 9.68081176e-01 -1.65171206e-01 5.89851104e-02
1.55864167e+00 8.30048993e-02 -1.24690875e-01 1.76286474e-01
7.49573588e-01 -1.77982062e-01 -1.39181328e+00 -1.02761760e-01
2.18012422e-01 -5.87018549e-01 -2.01132134e-01 -7.36469090e-01
-1.25883889e+00 8.78381014e-01 6.75674617e-01 7.72935227e-02
1.06560946e+00 1.60075113e-01 4.09258574e-01 4.28447664e-01
2.90924460e-01 -1.20821857e+00 -3.92374359e-02 3.82622778e-01
7.74030447e-01 -1.47087824e+00 2.06402555e-01 -1.66198969e-01
-5.26526511e-01 1.12530303e+00 7.70673573e-01 7.34256729e-02
4.04054850e-01 3.62131625e-01 4.72299188e-01 -2.91407257e-01
-1.57501072e-01 -3.30239952e-01 -4.06624041e-02 7.76268005e-01
2.57531047e-01 -1.63358063e-01 -1.29695073e-01 1.35058969e-01
-2.61509001e-01 -4.69301119e-02 6.88350201e-01 1.11229265e+00
-5.98519564e-01 -1.12983024e+00 -7.06429064e-01 6.37202263e-01
-4.68806833e-01 -2.76824564e-01 -6.20388329e-01 7.08150744e-01
4.26742464e-01 7.14062870e-01 2.33753249e-01 -2.04662353e-01
3.62471431e-01 -1.05291747e-01 3.83385956e-01 -9.06054497e-01
-8.35331261e-01 -2.31673504e-04 -1.05044805e-01 -2.33828768e-01
-7.49178588e-01 -6.14779234e-01 -1.24332893e+00 -2.70533621e-01
-1.87309951e-01 -1.49641678e-01 5.93726695e-01 1.10464776e+00
1.75414383e-01 3.83145064e-01 -2.05642395e-02 -1.11611176e+00
-1.48980960e-01 -9.19235229e-01 -5.40703893e-01 5.97709775e-01
2.63809562e-01 -4.93479669e-01 -4.60574895e-01 7.86625743e-01]
|
[9.607952117919922, 0.962203860282898]
|
c879b720-83c1-4ec2-af66-c7c6231e7b92
|
blurry-video-frame-interpolation
|
2002.12259
| null |
https://arxiv.org/abs/2002.12259v1
|
https://arxiv.org/pdf/2002.12259v1.pdf
|
Blurry Video Frame Interpolation
|
Existing works reduce motion blur and up-convert frame rate through two separate ways, including frame deblurring and frame interpolation. However, few studies have approached the joint video enhancement problem, namely synthesizing high-frame-rate clear results from low-frame-rate blurry inputs. In this paper, we propose a blurry video frame interpolation method to reduce motion blur and up-convert frame rate simultaneously. Specifically, we develop a pyramid module to cyclically synthesize clear intermediate frames. The pyramid module features adjustable spatial receptive field and temporal scope, thus contributing to controllable computational complexity and restoration ability. Besides, we propose an inter-pyramid recurrent module to connect sequential models to exploit the temporal relationship. The pyramid module integrates a recurrent module, thus can iteratively synthesize temporally smooth results without significantly increasing the model size. Extensive experimental results demonstrate that our method performs favorably against state-of-the-art methods.
|
['Zhiyong Gao', 'Wenbo Bao', 'Li Chen', 'Xiongkuo Min', 'Wang Shen', 'Guangtao Zhai']
|
2020-02-27
|
blurry-video-frame-interpolation-1
|
http://openaccess.thecvf.com/content_CVPR_2020/html/Shen_Blurry_Video_Frame_Interpolation_CVPR_2020_paper.html
|
http://openaccess.thecvf.com/content_CVPR_2020/papers/Shen_Blurry_Video_Frame_Interpolation_CVPR_2020_paper.pdf
|
cvpr-2020-6
|
['video-enhancement']
|
['computer-vision']
|
[ 3.57864141e-01 -5.19439042e-01 -1.49044562e-02 -9.85404626e-02
-3.86503518e-01 -2.37941474e-01 3.06358457e-01 -7.14247584e-01
-3.17766070e-01 8.56511295e-01 4.08439845e-01 -1.47767529e-01
1.61689058e-01 -5.39348364e-01 -5.90322137e-01 -6.66136980e-01
2.49553800e-01 -9.09123540e-01 4.42869693e-01 -5.24981273e-03
3.21153194e-01 2.32026592e-01 -1.32823789e+00 4.87054259e-01
1.25670362e+00 7.88066089e-01 7.90646911e-01 7.58837640e-01
2.16060057e-01 1.27647769e+00 -2.98279256e-01 -1.16354696e-01
2.28963479e-01 -6.71839118e-01 -4.38293695e-01 4.75477785e-01
3.47602874e-01 -8.30664992e-01 -8.39333594e-01 1.13968563e+00
3.60511273e-01 1.53594062e-01 2.16457829e-01 -6.51106238e-01
-1.19168675e+00 2.59507358e-01 -9.33306813e-01 4.92485374e-01
3.99818927e-01 4.36522365e-01 4.61835295e-01 -9.35430765e-01
4.79791939e-01 1.31555474e+00 5.13324976e-01 4.88873184e-01
-1.31766522e+00 -6.11429870e-01 2.92560905e-01 3.19547892e-01
-1.20303738e+00 -6.02621675e-01 7.08633900e-01 -1.93551183e-01
5.27602255e-01 3.09237450e-01 5.02415001e-01 6.03860497e-01
2.86674291e-01 5.87614417e-01 1.18081129e+00 -2.38898113e-01
-1.64178312e-02 -4.45048571e-01 -2.17763945e-01 5.15947640e-01
2.00020999e-01 2.87016153e-01 -4.40765202e-01 1.89744741e-01
1.54136789e+00 2.35734671e-01 -9.90701616e-01 1.38307855e-01
-1.18742907e+00 1.90504342e-01 3.43869984e-01 2.95925528e-01
-4.71652269e-01 2.55413353e-01 1.54413149e-01 1.62629828e-01
5.20979643e-01 1.88874394e-01 -2.31533140e-01 3.15859430e-02
-1.23621035e+00 7.81799033e-02 1.50890276e-01 9.21552062e-01
7.14275539e-01 2.49918953e-01 -5.43883502e-01 9.62820470e-01
3.14955145e-01 3.23088855e-01 4.47623223e-01 -1.32358968e+00
5.43802857e-01 4.88029458e-02 4.45267141e-01 -9.09446776e-01
1.85106397e-01 -8.45493525e-02 -1.19135737e+00 2.28042185e-01
2.16980144e-01 -2.02344522e-01 -9.05901611e-01 1.50422895e+00
5.00461794e-02 7.14003503e-01 -2.91533936e-02 1.27343237e+00
6.06895924e-01 9.40089107e-01 7.19250599e-03 -6.37727439e-01
1.47915053e+00 -1.21976924e+00 -1.11756408e+00 -1.39550805e-01
-7.47990385e-02 -1.10237801e+00 8.65305841e-01 1.86883435e-01
-1.72931063e+00 -9.94356215e-01 -1.07510996e+00 -3.43405396e-01
4.46505666e-01 2.11453810e-01 4.05508220e-01 4.00838822e-01
-1.16603863e+00 5.75274885e-01 -7.57176936e-01 3.54855627e-01
3.34780186e-01 1.76504672e-01 -5.27838469e-02 -3.32684606e-01
-1.23237574e+00 6.80979192e-01 3.06775302e-01 2.93655276e-01
-6.67024553e-01 -8.59569311e-01 -1.00259089e+00 1.42442986e-01
1.51901662e-01 -1.05319548e+00 1.14621484e+00 -9.58793461e-01
-1.78090239e+00 2.18287304e-01 -7.20694900e-01 -3.85808170e-01
6.55759335e-01 -3.74709338e-01 -4.68928158e-01 2.08632514e-01
-1.94210798e-01 7.08719015e-01 1.40442824e+00 -1.32873416e+00
-7.70885468e-01 1.97959408e-01 4.85393703e-02 3.39631617e-01
-2.66256362e-01 6.48278818e-02 -7.97801375e-01 -1.10331059e+00
9.41782221e-02 -5.43200493e-01 -4.49406087e-01 -1.73078403e-02
5.46290912e-02 3.76801878e-01 1.01298320e+00 -1.03584337e+00
1.73384559e+00 -2.36585593e+00 1.36793211e-01 -4.99320656e-01
3.85242552e-01 4.37705904e-01 -1.29485354e-01 -6.03253804e-02
-6.56707585e-02 -1.95026398e-02 -3.67971122e-01 -2.64591008e-01
-5.56598485e-01 6.51746541e-02 -5.51428080e-01 3.42590064e-01
3.10920686e-01 8.88476431e-01 -9.15896356e-01 -3.60749006e-01
6.06100559e-01 8.18662703e-01 -7.71145225e-01 3.11571240e-01
9.27357972e-02 5.26940703e-01 -1.96926072e-01 3.75723720e-01
1.21399844e+00 -3.38047624e-01 8.47117901e-02 -5.51692069e-01
-4.50214773e-01 4.23402861e-02 -9.82002318e-01 1.63904774e+00
-6.39654338e-01 8.74865174e-01 1.93640485e-01 -3.53581667e-01
8.36189032e-01 2.61092365e-01 2.13402912e-01 -5.96663415e-01
-7.53067136e-02 1.29180819e-01 -2.16694608e-01 -4.74646121e-01
9.53065097e-01 -3.78764644e-02 4.82231557e-01 -7.09181428e-02
-3.79549801e-01 -4.44892682e-02 2.02519730e-01 -1.41600549e-01
8.44554961e-01 3.54420155e-01 1.28640771e-01 -8.84923860e-02
9.34610426e-01 -5.55342317e-01 6.45711720e-01 4.24318880e-01
-2.09090978e-01 1.14785945e+00 1.24644250e-01 -3.59427363e-01
-1.20062172e+00 -1.14312184e+00 6.73467070e-02 6.74597859e-01
7.71376789e-01 -2.99983680e-01 -8.26861262e-01 -4.97988611e-02
-4.20492411e-01 3.38713080e-01 -2.32316136e-01 -7.71537796e-02
-8.82539690e-01 -5.69701016e-01 2.58765608e-01 3.73432279e-01
1.05531168e+00 -8.12704146e-01 -5.76822221e-01 4.26513344e-01
-6.45483911e-01 -1.17122769e+00 -1.11450410e+00 -4.22367603e-01
-9.89183843e-01 -7.47509420e-01 -1.22486913e+00 -1.03986812e+00
7.90497482e-01 8.71288657e-01 8.70194495e-01 1.13754105e-02
-1.28144935e-01 -2.26922825e-01 -1.50496334e-01 3.64090353e-01
-2.79289007e-01 -3.72169077e-01 -1.20671958e-01 1.91419795e-01
-1.64841935e-01 -6.68335974e-01 -1.09608793e+00 5.12028992e-01
-1.29150283e+00 6.81751251e-01 6.00047290e-01 9.24396694e-01
4.49960381e-01 4.88391891e-02 2.96996325e-01 -4.17259127e-01
7.68097401e-01 -6.56976625e-02 -7.43079662e-01 2.08378434e-01
-3.15239519e-01 6.28394401e-03 8.11798453e-01 -5.87953150e-01
-1.71595502e+00 -1.33215249e-01 1.71225429e-01 -7.48939276e-01
4.61228602e-02 3.64102945e-02 -2.49505043e-02 -9.61892530e-02
3.56211036e-01 5.29348075e-01 -1.97618946e-01 -4.38554436e-01
4.36774969e-01 6.39654517e-01 9.15254354e-01 -2.83750415e-01
7.77293503e-01 5.79216063e-01 -2.46026739e-01 -7.06110239e-01
-4.90933210e-01 -2.00280070e-01 -3.11243594e-01 -3.18151385e-01
8.83238435e-01 -1.44469917e+00 -6.69447303e-01 7.19645798e-01
-1.37289083e+00 -3.19231093e-01 2.47596912e-02 5.51663756e-01
-5.35626411e-01 7.51250625e-01 -1.25974584e+00 -4.74949390e-01
-2.64923245e-01 -1.29375196e+00 8.28298509e-01 5.96488178e-01
1.04603186e-01 -6.87580287e-01 -3.40888202e-01 2.75581270e-01
6.90826893e-01 -1.45350173e-01 2.87176788e-01 7.81627297e-01
-1.11969686e+00 4.65604484e-01 -7.78003633e-01 4.17045921e-01
6.70507967e-01 -1.87685549e-01 -8.80072236e-01 -3.42298955e-01
3.42022687e-01 3.01008314e-01 1.01913798e+00 7.61800528e-01
1.34018755e+00 -4.62354451e-01 -4.82767001e-02 9.10935044e-01
1.37564325e+00 3.93444717e-01 1.25164878e+00 2.39838555e-01
8.48226190e-01 1.44080386e-01 6.10643148e-01 4.71631795e-01
2.27890015e-01 7.18972921e-01 -1.22340500e-01 -4.05981243e-01
-6.30571306e-01 -1.85883313e-01 5.17940879e-01 8.55053604e-01
-3.97970527e-01 -1.00679286e-01 -3.31085861e-01 4.75295991e-01
-1.91903222e+00 -1.25675273e+00 -9.05420482e-02 1.81182861e+00
1.21699131e+00 -7.49406368e-02 -3.79689425e-01 -6.25678226e-02
1.01067889e+00 5.52324235e-01 -3.90640974e-01 -9.09251049e-02
-1.59976736e-01 -5.70571702e-03 5.35921633e-01 8.19870830e-01
-1.01157653e+00 9.58173573e-01 6.31946325e+00 1.05732048e+00
-1.10996664e+00 -1.37346517e-02 9.59154606e-01 -8.32917988e-02
-3.61712664e-01 4.13246192e-02 -5.78408599e-01 8.65828574e-01
5.79877317e-01 -2.36207709e-01 6.94199264e-01 4.21489805e-01
8.78383815e-01 -1.57544494e-01 -7.47653604e-01 1.27101564e+00
-8.08334127e-02 -1.56282723e+00 7.77240768e-02 -9.16458592e-02
1.00461257e+00 -5.22717297e-01 1.78808734e-01 -1.60145372e-01
1.15687586e-01 -8.52379203e-01 6.70016527e-01 7.81000495e-01
9.66070950e-01 -5.72648406e-01 3.48896801e-01 4.85612303e-02
-1.42981350e+00 2.33733337e-02 -3.69729012e-01 -1.58822909e-01
5.82643032e-01 7.25775540e-01 -1.57916900e-02 6.24147832e-01
8.10675085e-01 9.90313470e-01 -3.27542871e-01 1.11003482e+00
-2.47673899e-01 2.96108037e-01 1.35801464e-01 5.22309005e-01
-2.52954978e-02 -4.67776179e-01 4.11227465e-01 1.27493906e+00
4.24055398e-01 5.06509006e-01 -1.48053527e-01 8.51438582e-01
-5.94508350e-02 -4.91826892e-01 -2.06449717e-01 5.28839052e-01
3.85982752e-01 1.02677441e+00 -3.01452130e-01 -3.91586602e-01
-5.24767995e-01 1.46103406e+00 -1.28989801e-01 7.05624878e-01
-1.12654948e+00 -4.79092091e-01 8.65551770e-01 -6.98244497e-02
6.22417033e-01 -3.57338369e-01 -4.97179329e-01 -1.56153202e+00
2.24988744e-01 -7.77958393e-01 -1.53668582e-01 -1.14734173e+00
-9.58261490e-01 5.97923398e-01 -2.85900205e-01 -1.62315500e+00
-3.80321443e-02 -1.16211891e-01 -3.88850093e-01 1.14804661e+00
-1.98803031e+00 -7.49525487e-01 -5.91191411e-01 6.48008525e-01
1.01225674e+00 2.65448093e-01 4.82923277e-02 5.53060591e-01
-4.16827500e-01 2.55360186e-01 1.30072102e-01 -1.09859943e-01
8.45339239e-01 -6.89583421e-01 5.06469071e-01 1.36177516e+00
-4.97210920e-01 7.21855104e-01 6.11849010e-01 -6.44796848e-01
-1.11117172e+00 -1.22733176e+00 6.17599487e-01 5.74673973e-02
3.97402048e-01 5.75228147e-02 -1.19729543e+00 2.70614058e-01
4.13160980e-01 2.59249777e-01 8.08796957e-02 -6.84795320e-01
-2.25842431e-01 -2.84450859e-01 -9.26731110e-01 1.02949107e+00
9.94087636e-01 -5.47309220e-01 -4.65260625e-01 -2.26095304e-01
9.93346691e-01 -6.23038888e-01 -7.47555673e-01 4.43468273e-01
3.95558387e-01 -1.34319949e+00 1.23594105e+00 1.41886905e-01
1.03920186e+00 -9.10442173e-01 1.81526765e-01 -1.17768633e+00
-8.11631083e-01 -9.78497148e-01 -3.52947235e-01 1.29345739e+00
-4.37369645e-02 -4.46835756e-01 3.93089086e-01 6.36703908e-01
-1.63136840e-01 -5.40725708e-01 -5.87702274e-01 -6.14851654e-01
-4.83827919e-01 2.48104073e-02 3.71109992e-01 7.55792022e-01
-1.20145589e-01 2.37294231e-02 -9.30830598e-01 2.20792755e-01
6.32155240e-01 1.57938495e-01 4.78222877e-01 -4.34811473e-01
-3.59459102e-01 -3.78797203e-01 -5.61792962e-02 -1.86854303e+00
-2.81818390e-01 -4.15369682e-02 2.16237560e-01 -1.35779655e+00
1.91197127e-01 -1.19970605e-01 -2.44257659e-01 2.89524514e-02
-6.71003401e-01 2.76968777e-01 4.10753042e-01 5.53426147e-01
-3.28518212e-01 5.73715270e-01 1.80944991e+00 7.40646049e-02
-4.23699349e-01 -3.08600515e-01 -6.34692192e-01 6.81606233e-01
5.93803465e-01 2.99191140e-02 -4.84823316e-01 -8.62821400e-01
-4.35966730e-01 4.53373373e-01 4.19793040e-01 -1.02418208e+00
2.93313056e-01 -4.26436633e-01 6.38648629e-01 -4.20068711e-01
3.35299373e-01 -6.82592392e-01 3.96469355e-01 4.45745647e-01
-2.94788122e-01 5.12750410e-02 2.08224386e-01 7.61269748e-01
-5.25387883e-01 8.92264675e-03 1.12321103e+00 -7.27429288e-03
-8.87348175e-01 2.83033997e-01 -4.10975337e-01 -2.23617062e-01
9.96821761e-01 -2.97688276e-01 -4.28230315e-01 -4.03589040e-01
-3.61443937e-01 2.19863076e-02 7.22735047e-01 5.97082138e-01
9.23278987e-01 -1.23558998e+00 -7.35755563e-01 2.69838303e-01
-5.32446802e-01 -6.60744458e-02 7.42528439e-01 7.70901859e-01
-7.68700123e-01 2.50858605e-01 -3.76758814e-01 -5.00785172e-01
-1.23012125e+00 6.70261383e-01 2.46795386e-01 -1.19867027e-01
-6.20252311e-01 6.60167456e-01 6.81974769e-01 5.09120047e-01
-9.77572948e-02 -5.23401797e-01 -7.66146034e-02 -3.86272609e-01
1.05803525e+00 3.84923190e-01 -4.13402736e-01 -3.86436224e-01
-2.63695531e-02 6.58204377e-01 -2.55161941e-01 -1.53046459e-01
1.13542867e+00 -8.06646585e-01 -1.32026106e-01 -1.64185278e-02
9.27000821e-01 -2.06577927e-02 -2.03036976e+00 -1.63799807e-01
-4.22111750e-01 -1.05957425e+00 1.53170064e-01 -3.68292183e-01
-1.22957242e+00 6.26654565e-01 5.49854338e-01 3.77935618e-02
1.73663318e+00 -4.89274532e-01 9.90118384e-01 -2.55146593e-01
1.11496747e-01 -7.40568221e-01 7.86473528e-02 4.14679825e-01
7.38595068e-01 -9.31479514e-01 1.20074652e-01 -8.23725760e-01
-5.79058766e-01 1.13586605e+00 6.66867793e-01 -1.94624886e-01
2.14876860e-01 3.85994196e-01 -9.75413844e-02 5.16850114e-01
-7.99007058e-01 -6.15763627e-02 3.57187659e-01 5.09510219e-01
6.01931870e-01 -3.00781906e-01 -5.02657473e-01 2.70166844e-01
3.45200509e-01 5.89889288e-01 8.14547837e-01 5.69315851e-01
-4.89051431e-01 -8.88727546e-01 -5.63108444e-01 -4.46750224e-02
-4.94851410e-01 -5.27822316e-01 4.58567232e-01 2.06972793e-01
4.18572761e-02 1.11510551e+00 8.65120813e-02 -1.70697272e-01
1.40666261e-01 -4.97326314e-01 3.86320621e-01 -1.15639362e-02
-3.14949095e-01 4.76654470e-01 -1.79040894e-01 -4.58206713e-01
-6.19712353e-01 -3.52129728e-01 -8.93463790e-01 -4.29457545e-01
-2.26449698e-01 -1.67207822e-01 7.07344115e-02 6.75572634e-01
5.16758025e-01 8.23165715e-01 6.94908321e-01 -1.14895749e+00
-6.75350353e-02 -7.52968967e-01 -3.79341871e-01 3.39978009e-01
7.36906469e-01 -2.07764626e-01 -3.37293237e-01 7.88024247e-01]
|
[11.220183372497559, -2.215895652770996]
|
34bf95a2-85f0-46ec-ac13-6a9a1c883f5e
|
a-robustness-evaluation-framework-for
| null | null |
https://aclanthology.org/2022.argmining-1.16
|
https://aclanthology.org/2022.argmining-1.16.pdf
|
A Robustness Evaluation Framework for Argument Mining
|
Standard practice for evaluating the performance of machine learning models for argument mining is to report different metrics such as accuracy or F1. However, little is usually known about the model’s stability and consistency when deployed in real-world settings. In this paper, we propose a robustness evaluation framework to guide the design of rigorous argument mining models. As part of the framework, we introduce several novel robustness tests tailored specifically to argument mining tasks. Additionally, we integrate existing robustness tests designed for other natural language processing tasks and re-purpose them for argument mining. Finally, we illustrate the utility of our framework on two widely used argument mining corpora, UKP topic-sentences and IBM Debater Evidence Sentence. We argue that our framework should be used in conjunction with standard performance evaluation techniques as a measure of model stability.
|
['Oana Cocarascu', 'Matteo Fortier', 'Mehmet Sofi']
| null | null | null | null |
argmining-acl-2022-10
|
['argument-mining']
|
['natural-language-processing']
|
[ 2.96278358e-01 5.78319430e-01 -6.22994840e-01 -3.59842330e-01
-9.99466062e-01 -7.95122862e-01 1.02989447e+00 1.01319456e+00
-5.94574034e-01 8.49769115e-01 5.37571847e-01 -1.13152742e+00
-5.44922590e-01 -6.97954059e-01 -6.69343829e-01 -1.97157547e-01
1.00783510e-02 4.26323891e-01 2.57475197e-01 -1.47157833e-01
6.37081683e-01 1.98407263e-01 -1.47982359e+00 4.33319926e-01
9.27481830e-01 7.15317130e-01 -5.48763990e-01 6.13291800e-01
1.62023634e-01 8.50205421e-01 -7.27246344e-01 -8.34193289e-01
1.28029659e-01 -2.96752125e-01 -1.36463606e+00 -4.04069692e-01
2.06228703e-01 -1.26648024e-02 2.96259940e-01 7.84979820e-01
3.74409705e-01 2.04672012e-02 6.46352828e-01 -1.19193017e+00
-1.67064771e-01 1.42137909e+00 -1.87695950e-01 5.33444762e-01
4.76856291e-01 -8.50694254e-03 1.49788642e+00 -3.00268114e-01
8.37027550e-01 1.38215494e+00 7.03276932e-01 7.39278868e-02
-1.22062433e+00 -3.87800276e-01 5.11829317e-01 1.69792458e-01
-2.95377463e-01 -6.40996218e-01 7.45142937e-01 -1.53418347e-01
9.27283764e-01 5.03706098e-01 4.36639667e-01 1.22771251e+00
2.71844622e-02 8.99435937e-01 1.48451030e+00 -7.67450571e-01
2.49676451e-01 1.93560079e-01 6.86119020e-01 3.07243437e-01
8.59394193e-01 3.88914347e-02 -4.80038106e-01 -7.61246085e-01
2.82429129e-01 -9.52723742e-01 1.90688834e-01 -3.14094931e-01
-8.96722496e-01 1.30389726e+00 -7.25213513e-02 3.00481200e-01
-1.46690965e-01 -3.00424807e-02 7.75760293e-01 5.22105455e-01
8.00080001e-01 5.73437810e-01 -7.62165606e-01 -3.80218714e-01
-7.52923012e-01 8.05109560e-01 1.08384597e+00 3.06959391e-01
2.44138576e-02 -4.25946176e-01 -1.82164535e-01 8.24864507e-01
5.22191525e-01 5.04120253e-02 3.27621609e-01 -1.13186717e+00
6.82635725e-01 6.63092315e-01 1.94876075e-01 -8.41407120e-01
-4.95576560e-01 -3.32699567e-01 -7.50905424e-02 1.65621080e-02
6.41908884e-01 -9.29321721e-02 -2.14341953e-01 1.81404865e+00
5.45732021e-01 -2.28376836e-01 2.83939868e-01 4.33688462e-01
6.75815463e-01 7.70010874e-02 3.25263947e-01 -3.60473543e-01
1.23492205e+00 -6.68792665e-01 -4.21894580e-01 -1.81733787e-01
1.12776732e+00 -8.62256348e-01 1.32965064e+00 3.62974554e-01
-1.41702104e+00 -1.18755244e-01 -9.70979095e-01 -2.43638232e-02
-1.18082434e-01 -3.20306480e-01 8.60543549e-01 8.02791953e-01
-3.45952034e-01 7.04755902e-01 -6.45992815e-01 -2.04731882e-01
3.22832227e-01 2.79234257e-02 -1.15826689e-01 3.92200530e-01
-1.16657460e+00 1.17554903e+00 4.35030997e-01 -3.31620276e-01
-3.64187837e-01 -3.51276547e-01 -8.22164834e-01 -6.42892495e-02
4.56558049e-01 -6.31964326e-01 1.56556118e+00 -7.32739568e-01
-1.12521994e+00 9.78017449e-01 7.74442777e-02 -7.57633626e-01
7.86940932e-01 -5.42823255e-01 -2.59906203e-01 -1.27690226e-01
3.55021030e-01 -1.32070825e-01 4.61814314e-01 -1.12207103e+00
-6.21386826e-01 -3.09564888e-01 5.31530619e-01 -7.53407180e-03
-1.48891374e-01 5.63976407e-01 2.05854088e-01 -7.14662075e-01
-4.14886102e-02 -7.59739459e-01 -8.92250165e-02 -4.42432404e-01
-2.99730599e-01 -6.95402920e-01 5.51116765e-01 -2.72062004e-01
1.59331489e+00 -1.77677310e+00 -2.74465501e-01 4.11424190e-01
-3.50069344e-01 2.22542465e-01 -8.63511637e-02 4.13466096e-01
-2.48316675e-01 4.29858923e-01 -3.19156915e-01 -5.43647818e-02
1.31032422e-01 2.52492458e-01 -4.67315435e-01 5.08899152e-01
2.57320464e-01 9.34168339e-01 -7.06669271e-01 -6.94068909e-01
-3.83045226e-02 -4.66214530e-02 -6.87113285e-01 -1.99551046e-01
-3.64734471e-01 -1.69451442e-03 -4.91548330e-01 5.65181017e-01
1.93079576e-01 -2.59292871e-01 4.63301986e-01 1.87002510e-01
-1.06673464e-01 1.27497661e+00 -1.11455071e+00 1.27200067e+00
-1.19537964e-01 4.32712823e-01 3.75587270e-02 -1.31887305e+00
6.72178924e-01 1.48876518e-01 -3.75014544e-02 -6.97805166e-01
2.83301353e-01 2.54769772e-01 3.29466671e-01 -3.55447799e-01
2.86228716e-01 -1.61923598e-02 -5.97072579e-02 1.19659674e+00
-4.44063306e-01 -9.52353105e-02 6.70484126e-01 1.37728751e-01
1.04231203e+00 1.48873433e-01 5.99756539e-01 -6.23784482e-01
6.07036650e-01 2.51515985e-01 3.42053205e-01 9.59744871e-01
-9.44606736e-02 1.74987584e-01 8.68098676e-01 -5.30002832e-01
-1.06532764e+00 -7.56332397e-01 -7.03882456e-01 1.15586829e+00
-3.46695930e-01 -6.33268774e-01 -6.21907532e-01 -1.27588928e+00
1.29422203e-01 9.36963499e-01 -7.33762622e-01 8.86859894e-02
-8.92601013e-01 -1.16118908e+00 8.25433791e-01 4.72302198e-01
1.53932229e-01 -1.12726736e+00 -1.10499346e+00 3.12496752e-01
-3.58413249e-01 -8.12228382e-01 3.85168791e-01 3.49663109e-01
-1.07136703e+00 -1.71987033e+00 2.09389418e-01 -3.41829926e-01
1.17541015e-01 -8.67980123e-02 1.46319544e+00 6.87601745e-01
1.44489884e-01 2.67508537e-01 -6.26221001e-01 -7.77714849e-01
-7.99004436e-01 1.89327404e-01 -3.11051942e-02 -7.77425587e-01
3.26599568e-01 -3.18074346e-01 -3.20122272e-01 2.54021764e-01
-9.95460391e-01 -3.67270470e-01 3.15434188e-01 9.46964800e-01
1.54799104e-01 -1.31733760e-01 7.46627212e-01 -1.29664850e+00
1.42913866e+00 -5.58736324e-01 -3.55790585e-01 4.57486898e-01
-1.18819439e+00 1.66507512e-01 1.44105092e-01 -5.27967751e-01
-1.01303899e+00 -9.06353056e-01 -2.39254534e-01 6.45878851e-01
-1.95781011e-02 9.37693477e-01 1.74338445e-01 2.44109064e-01
9.29841757e-01 -6.59362316e-01 9.82347652e-02 -4.33333784e-01
3.62669498e-01 5.68952560e-01 1.76677704e-01 -1.26532233e+00
7.52878845e-01 3.81106138e-01 -2.14604735e-01 -5.50196707e-01
-1.09365690e+00 -1.65715843e-01 -2.07216397e-01 1.37726918e-01
5.68993650e-02 -5.34268856e-01 -7.21096933e-01 -1.20987207e-01
-9.06885207e-01 -5.57898402e-01 -2.90558755e-01 4.29519176e-01
-6.59527600e-01 6.82399690e-01 -6.18415773e-01 -9.54616785e-01
-5.19436002e-01 -8.13101709e-01 8.13200295e-01 -3.56756411e-02
-1.00494635e+00 -1.26129353e+00 4.30483580e-01 6.39682233e-01
9.45291966e-02 2.78572649e-01 1.24625826e+00 -1.20798504e+00
2.44766206e-01 -1.74774781e-01 -5.25698960e-02 2.04857528e-01
-1.72886148e-01 3.01906526e-01 -8.50672007e-01 -9.16602015e-02
2.33771265e-01 -6.41280830e-01 8.95300388e-01 5.29887319e-01
8.86248946e-01 -5.89341283e-01 -1.31615490e-01 3.93870212e-02
9.37435567e-01 -2.30854720e-01 4.37638283e-01 1.25517154e+00
-1.18828453e-01 1.05276322e+00 1.05297661e+00 3.50296855e-01
2.37638637e-01 4.37815398e-01 2.54597098e-01 1.74818605e-01
3.00346792e-01 -2.91682035e-01 2.88332939e-01 2.92609245e-01
-1.53133035e-01 -2.91510552e-01 -9.63333905e-01 5.10533452e-01
-2.15252090e+00 -1.07194233e+00 -2.51999021e-01 2.01790977e+00
8.27784240e-01 8.75384510e-01 4.93930995e-01 7.40850091e-01
2.55916327e-01 1.29799336e-01 -1.61483511e-01 -9.46850240e-01
-3.20644975e-01 4.15665239e-01 6.15868270e-02 5.81906796e-01
-1.16883039e+00 7.36987352e-01 7.27599669e+00 5.88053226e-01
-5.10894656e-01 -1.37011670e-02 7.65546799e-01 -5.42562269e-02
-6.06793702e-01 5.60281754e-01 -4.56962764e-01 1.02113314e-01
1.13224578e+00 -2.23852232e-01 -2.90272802e-01 8.15069854e-01
2.34521836e-01 -2.17550561e-01 -1.10748291e+00 2.01719895e-01
-1.49498969e-01 -1.35448790e+00 -5.61990142e-02 7.45904371e-02
4.48745579e-01 -2.03239158e-01 -2.30235849e-02 2.96265066e-01
7.06561029e-01 -1.11604416e+00 8.39066684e-01 -2.07205504e-01
1.03543885e-01 -6.10423326e-01 8.50575566e-01 3.77170384e-01
-5.61604917e-01 -3.12663823e-01 -1.30958065e-01 -5.79167187e-01
2.53139198e-01 5.81744432e-01 -7.61124909e-01 5.90874970e-01
6.29848838e-01 4.32112187e-01 -6.02629006e-01 7.30601668e-01
-5.17379045e-01 1.15516007e+00 -3.55470121e-01 -5.68666942e-02
3.56294572e-01 -6.43191040e-02 5.67105711e-01 1.20339692e+00
-3.46072137e-01 3.61127071e-02 -1.75471842e-01 5.28830767e-01
6.02157377e-02 3.79653990e-01 -6.15210831e-01 2.21019741e-02
6.05950832e-01 1.07388854e+00 -8.37269247e-01 -3.23154777e-01
-3.39377314e-01 2.88632736e-02 5.00592172e-01 -5.43847270e-02
-7.67723203e-01 1.47512525e-01 4.94585752e-01 2.63827890e-01
-2.22077459e-01 3.94646749e-02 -6.93369508e-01 -9.39044714e-01
3.56195807e-01 -1.40148497e+00 9.22709525e-01 -2.78941035e-01
-1.41441250e+00 3.19001287e-01 5.23542821e-01 -7.10520804e-01
-4.98224080e-01 -6.80842578e-01 -8.93948495e-01 4.10025597e-01
-1.46734858e+00 -1.11099851e+00 3.38119090e-01 2.37270683e-01
3.49632084e-01 4.09672521e-02 7.48549759e-01 -2.31993869e-01
-5.49941540e-01 5.98219693e-01 -2.45438397e-01 1.02606148e-01
6.88367724e-01 -1.10477424e+00 6.43898606e-01 7.99315274e-01
3.02189738e-01 8.86932075e-01 9.93221164e-01 -7.01958716e-01
-8.20062339e-01 -4.70439732e-01 9.62102354e-01 -8.46245468e-01
1.01482105e+00 1.86982781e-01 -9.36117113e-01 6.74750865e-01
1.12494648e-01 -5.40577471e-01 8.21703315e-01 9.52281415e-01
-5.13108909e-01 3.19231510e-01 -1.12823117e+00 4.96069163e-01
9.80991781e-01 -3.25515121e-01 -1.21208882e+00 3.51394057e-01
5.08236825e-01 -2.43860930e-01 -8.81852865e-01 8.97165179e-01
8.39673460e-01 -1.02352059e+00 9.32786942e-01 -1.26064038e+00
5.73844194e-01 1.26041293e-01 -1.48077831e-01 -8.56083155e-01
-6.15644455e-02 -5.42835116e-01 -1.29566789e-01 1.28198850e+00
8.73219311e-01 -7.36910045e-01 6.98914826e-01 8.38729322e-01
8.30612406e-02 -8.33834231e-01 -9.24288511e-01 -5.35298049e-01
6.38716102e-01 -7.68368900e-01 5.27457178e-01 1.01215792e+00
3.26684505e-01 4.84917521e-01 1.98945463e-01 -2.00902864e-01
6.84343815e-01 2.79806405e-01 9.74223375e-01 -1.60975945e+00
-2.52650291e-01 -7.30910957e-01 2.51204371e-02 -4.48166549e-01
5.23169816e-01 -7.00912595e-01 -5.43544471e-01 -1.36271060e+00
2.90672958e-01 -6.40246391e-01 -9.39550996e-02 3.13701570e-01
-3.95550758e-01 -1.81121141e-01 1.11160530e-02 1.64811641e-01
-4.33268994e-01 2.21432984e-01 6.60165787e-01 1.15982547e-01
-1.96939334e-01 1.72190472e-01 -1.03277695e+00 9.45398986e-01
1.17973483e+00 -6.05265379e-01 -4.02425081e-01 -2.78854549e-01
6.68287337e-01 -3.49621892e-01 3.77404481e-01 -5.02976894e-01
-4.62795198e-02 -3.46474349e-01 7.36653507e-02 -3.60428005e-01
-2.76418567e-01 -4.19675201e-01 -2.09341124e-01 4.57026452e-01
-6.77261412e-01 5.08381248e-01 2.11853787e-01 3.88931066e-01
-1.51968896e-01 -5.89783490e-01 5.55634022e-01 -3.77160832e-02
3.92555399e-03 -3.81822556e-01 -5.48972264e-02 5.52621424e-01
8.23346913e-01 -8.94180909e-02 -7.20002711e-01 -1.76697031e-01
-3.53388339e-01 2.32079640e-01 4.95111972e-01 4.08649296e-01
2.73399979e-01 -8.17517102e-01 -1.09825516e+00 -2.35028490e-01
1.41917437e-01 -2.22315535e-01 -3.85510623e-01 8.06379557e-01
-3.70795459e-01 5.07370830e-01 1.62152395e-01 -2.34124243e-01
-1.56511092e+00 4.40184027e-01 -1.12708583e-01 -7.21439898e-01
-5.52964866e-01 4.80518430e-01 -3.88248771e-01 -4.01653558e-01
2.15541035e-01 -4.36171532e-01 -3.08905363e-01 9.48162973e-02
3.88008654e-01 4.11904216e-01 1.63508430e-01 -3.18623871e-01
-4.19796228e-01 6.08246177e-02 -1.29223928e-01 -4.85372365e-01
1.44255054e+00 3.09356526e-02 -2.56836385e-01 4.33910549e-01
4.85061735e-01 4.00644034e-01 -5.75447261e-01 -1.09418243e-01
7.08718300e-01 -1.02748707e-01 -3.51864249e-01 -8.31318498e-01
-2.04554573e-01 2.77230978e-01 4.44199052e-03 5.71467102e-01
7.41522074e-01 1.55260757e-01 1.08314388e-01 5.82059801e-01
8.56771786e-03 -1.48511720e+00 -2.21173644e-01 4.67677176e-01
7.91348040e-01 -1.20408928e+00 4.94350195e-01 -3.11914235e-01
-5.43802440e-01 8.38862836e-01 2.33550265e-01 -4.42124046e-02
5.52634776e-01 2.84018189e-01 8.85707512e-02 -4.23693985e-01
-1.14876115e+00 4.95615229e-02 1.81915373e-01 2.72604942e-01
9.66001928e-01 8.23602527e-02 -1.17063928e+00 6.72330379e-01
-7.59686410e-01 -4.12869006e-01 3.52842897e-01 1.06146669e+00
-4.18941677e-01 -1.48962021e+00 -3.72700483e-01 6.05440974e-01
-9.68788266e-01 -7.96261057e-02 -8.28283429e-01 1.26782084e+00
-4.18547630e-01 1.48290944e+00 -1.67242661e-01 1.23682218e-02
3.31439227e-01 3.12642664e-01 4.86922979e-01 -4.61696953e-01
-9.84516382e-01 -1.89806804e-01 1.02096629e+00 -4.22959924e-01
-9.77112830e-01 -1.01187897e+00 -9.32928562e-01 -5.13652444e-01
-5.77868521e-01 5.94765961e-01 4.74235147e-01 1.12654006e+00
-4.73033674e-02 3.17104869e-02 1.68092817e-01 4.75546196e-02
-8.37711453e-01 -1.17678452e+00 -2.74716429e-02 7.38400400e-01
5.19158095e-02 -8.28338802e-01 -5.08729935e-01 -3.79782677e-01]
|
[9.534215927124023, 9.600945472717285]
|
203bce20-cdee-4b88-9f10-12ff419e17fa
|
unsupervised-learning-of-multi-frame-optical
| null | null |
http://openaccess.thecvf.com/content_ECCV_2018/html/Joel_Janai_Unsupervised_Learning_of_ECCV_2018_paper.html
|
http://openaccess.thecvf.com/content_ECCV_2018/papers/Joel_Janai_Unsupervised_Learning_of_ECCV_2018_paper.pdf
|
Unsupervised Learning of Multi-Frame Optical Flow with Occlusions
|
Learning optical flow with neural networks is hampered by the need for obtaining training data with associated ground truth. Unsupervised learning is a promising direction, yet the performance of current unsupervised methods is still limited. In particular, the lack of proper occlusion handling in commonly used data terms constitutes a major source of error. While most optical flow methods process pairs of consecutive frames, more advanced occlusion reasoning can be realized when considering multiple frames. In this paper, we propose a framework for unsupervised learning of optical flow and occlusions over multiple frames. More specifically, we exploit the minimal configuration of three frames to strengthen the photometric loss and explicitly reason about occlusions. We demonstrate that our multi-frame, occlusion-sensitive formulation outperforms existing unsupervised two-frame methods and even produces results on par with some fully supervised methods.
|
['Michael Black', 'Anurag Ranjan', 'Joel Janai', 'Fatma Guney', 'Andreas Geiger']
|
2018-09-01
| null | null | null |
eccv-2018-9
|
['occlusion-handling']
|
['computer-vision']
|
[ 1.22252032e-01 -2.15361789e-01 -4.91958141e-01 -3.94468874e-01
-3.41312617e-01 -4.93428171e-01 5.69306552e-01 4.14237529e-02
-4.03506279e-01 1.12201965e+00 3.25252444e-01 -1.52497441e-02
-1.98188335e-01 -4.93728489e-01 -5.67586660e-01 -7.25289643e-01
9.28555131e-02 6.56287670e-02 5.54901883e-02 1.02912381e-01
2.06941932e-01 6.68867946e-01 -1.53323305e+00 -1.65648177e-01
7.70200431e-01 8.54234874e-01 -3.93462554e-02 6.14651084e-01
-2.17135340e-01 1.25601304e+00 -3.81850153e-01 -2.57143378e-01
4.33312714e-01 -3.54344934e-01 -9.37902927e-01 5.70230305e-01
1.00171614e+00 -7.62672067e-01 -4.49685186e-01 9.37835574e-01
2.87287146e-01 5.97340226e-01 3.35360765e-01 -1.26461828e+00
-2.30736494e-01 1.45772859e-01 -3.72378051e-01 3.42242628e-01
2.87112981e-01 2.04274341e-01 1.14579082e+00 -7.36949325e-01
9.37609494e-01 1.14910054e+00 4.27260786e-01 4.71818388e-01
-1.41462648e+00 -7.97838122e-02 2.46623904e-01 2.48724774e-01
-9.45698380e-01 -7.43297577e-01 1.05462742e+00 -7.12311149e-01
7.44090140e-01 -2.23183129e-02 6.75287843e-01 9.32497203e-01
-1.44227922e-01 8.23974252e-01 8.44300449e-01 -5.04917681e-01
1.78811222e-01 -2.58145809e-01 2.14629501e-01 8.76930714e-01
4.26951021e-01 2.71122485e-01 -7.97863603e-01 1.20233819e-01
9.53282356e-01 1.04883639e-02 -4.39418316e-01 -7.18432307e-01
-1.15054989e+00 7.29960322e-01 2.97457278e-01 2.36068770e-01
-2.64398754e-01 1.81872487e-01 2.84966916e-01 3.76225784e-02
6.80034339e-01 4.27571297e-01 -3.66444349e-01 -2.11470157e-01
-1.24116027e+00 1.86577469e-01 8.02619815e-01 8.49896252e-01
1.37403083e+00 3.76737982e-01 8.87685828e-03 5.03169119e-01
2.23473147e-01 1.56785116e-01 -1.47201242e-02 -1.69479370e+00
2.18707517e-01 3.21183413e-01 2.72068202e-01 -1.28386092e+00
-3.29335392e-01 -2.77203232e-01 -7.23826170e-01 4.46448714e-01
9.33053970e-01 -2.14546919e-01 -5.32307863e-01 1.74761558e+00
2.84952581e-01 5.25081158e-01 9.59296376e-02 1.01472783e+00
5.31850219e-01 3.33973497e-01 -1.54507458e-02 -6.65558159e-01
8.10239196e-01 -1.19576228e+00 -9.51208115e-01 -2.35103577e-01
4.58377123e-01 -8.56548488e-01 5.67324400e-01 3.53836358e-01
-1.19947457e+00 -7.58306265e-01 -8.90640140e-01 -4.69692796e-01
-1.15484536e-01 1.60220936e-02 1.01612127e+00 5.93948543e-01
-9.98475850e-01 6.59266710e-01 -1.02092731e+00 -2.83539355e-01
5.24404645e-01 3.18020731e-01 -5.16843915e-01 -1.68918520e-01
-8.01898777e-01 7.93501735e-01 3.11823219e-01 4.17216301e-01
-6.41388714e-01 -7.49440312e-01 -1.14825785e+00 -1.94739029e-01
3.38919580e-01 -7.61101842e-01 1.00672436e+00 -9.91170704e-01
-1.57191265e+00 6.61990702e-01 -5.09489417e-01 -6.29051745e-01
7.52554774e-01 -6.01391554e-01 -2.39917496e-03 6.31115258e-01
-1.09946348e-01 9.13811684e-01 1.00542796e+00 -1.17204809e+00
-4.51410681e-01 6.18430562e-02 5.57987034e-01 1.62405387e-01
-3.13611805e-01 -2.08273321e-01 -2.99642652e-01 -6.25668347e-01
8.68229717e-02 -7.69501567e-01 -2.66297638e-01 3.33627701e-01
-3.20745170e-01 -6.01275675e-02 9.11850333e-01 -3.16622972e-01
8.89536977e-01 -1.90400577e+00 1.60430148e-01 -1.65647388e-01
2.85188019e-01 3.60154152e-01 7.55955577e-02 9.27891582e-02
-4.63281535e-02 -2.40662694e-01 -3.08876634e-01 -7.43940949e-01
-1.96639508e-01 6.47904694e-01 -2.85799176e-01 7.13974953e-01
4.34835851e-01 7.14857936e-01 -1.19734991e+00 -7.41476238e-01
7.20429718e-01 7.51248598e-01 -8.73860002e-01 2.00954989e-01
-1.83881342e-01 9.83181000e-01 -2.78976142e-01 4.67886209e-01
4.97015804e-01 -3.07078838e-01 3.58437896e-02 -3.95912498e-01
-2.85985231e-01 3.08719456e-01 -1.33995938e+00 2.13627434e+00
-1.52122051e-01 1.08876300e+00 -1.86330557e-01 -1.12739921e+00
6.22023702e-01 3.65139663e-01 1.00783134e+00 -1.44399911e-01
1.35831222e-01 7.46453926e-03 -1.81474864e-01 -8.02451491e-01
5.27998269e-01 -7.64676183e-02 6.19090438e-01 4.19606298e-01
2.76270002e-01 -8.88605192e-02 6.99551702e-01 1.77473336e-01
8.03691983e-01 6.70681596e-01 1.57476515e-01 -2.79980987e-01
7.48261154e-01 -6.66557252e-02 9.78761017e-01 7.29733706e-01
-5.89913011e-01 7.53752112e-01 3.59298378e-01 -7.33125806e-01
-8.95408452e-01 -8.45749259e-01 -2.00305134e-01 7.64527977e-01
2.10227698e-01 -4.28093284e-01 -7.08881438e-01 -6.05244637e-01
-1.63880840e-01 2.06814140e-01 -3.94416422e-01 1.20334268e-01
-7.71569848e-01 -5.96341491e-01 2.89798051e-01 3.89730513e-01
6.28237367e-01 -9.17228997e-01 -6.98681474e-01 1.70205161e-01
-5.26691616e-01 -1.65551150e+00 -3.70333105e-01 -1.04775310e-01
-1.07800388e+00 -1.33584738e+00 -4.56716210e-01 -5.99532902e-01
8.16259265e-01 3.70254189e-01 1.14868557e+00 2.04838559e-01
-3.06253493e-01 5.09677231e-01 -9.14410800e-02 3.79657634e-02
-6.72253519e-02 -1.23610355e-01 8.04233104e-02 3.33641529e-01
1.66960359e-01 -6.61555171e-01 -7.31859326e-01 3.64339873e-02
-9.72508609e-01 -8.49423930e-03 1.29747778e-01 7.98258126e-01
5.27408421e-01 -6.60799220e-02 2.25373507e-01 -8.62837613e-01
2.64496747e-02 8.80703181e-02 -6.98027790e-01 1.97574906e-02
-3.93480331e-01 2.70811975e-01 4.78457630e-01 -3.41294199e-01
-1.38762164e+00 3.23560804e-01 1.43114567e-01 -5.72408080e-01
-4.73442942e-01 2.02333406e-01 -1.15552284e-01 -3.59039158e-01
5.22047341e-01 -2.31444120e-01 2.16381308e-02 -3.36555272e-01
4.79245514e-01 -7.60781318e-02 6.32928014e-01 -7.63883710e-01
7.35423684e-01 9.60296988e-01 2.83442110e-01 -8.52781892e-01
-1.10810351e+00 -6.31859899e-01 -1.09096098e+00 -4.63322550e-01
8.15806270e-01 -7.50045002e-01 -5.84063649e-01 3.98978770e-01
-1.40300536e+00 -2.51674116e-01 -4.00093496e-01 9.32347178e-01
-7.33402848e-01 7.49323964e-01 -7.20290065e-01 -7.75215805e-01
9.94461179e-02 -1.35231566e+00 7.07291424e-01 1.27731279e-01
-1.87280208e-01 -1.44311309e+00 3.14797536e-02 4.67578948e-01
1.92452773e-01 2.77935237e-01 2.20395356e-01 -1.85450524e-01
-9.05538738e-01 1.39664710e-02 -2.19787553e-01 5.49265683e-01
4.47709948e-01 3.58845234e-01 -1.08673692e+00 -2.18181089e-01
1.39128879e-01 -3.92643273e-01 9.04390633e-01 6.25678778e-01
9.25369859e-01 -2.54504532e-02 1.34714693e-01 6.88651979e-01
1.41260076e+00 -1.45699173e-01 6.20487392e-01 2.56585419e-01
9.49840367e-01 8.93831432e-01 3.24638069e-01 3.62705529e-01
2.52554983e-01 3.80267143e-01 3.77237976e-01 -6.87840730e-02
-3.52325112e-01 1.87705718e-02 1.45071909e-01 8.75343025e-01
-4.27252144e-01 -3.69415805e-02 -6.24503791e-01 4.92737204e-01
-1.87252498e+00 -1.06588912e+00 -3.15028220e-01 2.11797380e+00
7.73916841e-01 1.22272454e-01 8.35521333e-03 4.62646008e-01
6.25074029e-01 4.90671158e-01 -2.97746658e-01 1.55139893e-01
-3.67428601e-01 1.87131777e-01 3.41525972e-01 1.04624081e+00
-1.28054047e+00 1.00371432e+00 6.73607874e+00 2.49875352e-01
-9.90470529e-01 -5.68019971e-02 4.50427353e-01 -9.80228409e-02
-7.98215866e-02 4.73237991e-01 -7.80446589e-01 1.34403378e-01
4.04004872e-01 1.72799155e-01 2.10700557e-01 4.44779009e-01
5.48970580e-01 -3.21554095e-01 -1.25428545e+00 1.09859586e+00
1.53773442e-01 -1.26731992e+00 2.80385502e-02 3.47306691e-02
1.11518991e+00 -2.59619743e-01 -1.82850778e-01 -3.12018752e-01
1.01760969e-01 -6.89728260e-01 4.56030071e-01 6.61807418e-01
3.18548024e-01 -3.63736808e-01 3.56629759e-01 -1.21475654e-02
-9.74867642e-01 1.70946926e-01 -3.24676573e-01 -4.03476238e-01
4.99011219e-01 1.01941431e+00 -7.46033788e-02 6.09393954e-01
4.05068249e-01 1.31039488e+00 -3.11990649e-01 1.22040772e+00
-4.41568792e-01 4.46736634e-01 -3.96417201e-01 4.56220388e-01
3.12214226e-01 -4.20025170e-01 6.33479714e-01 1.00074828e+00
-7.54316747e-02 -9.58999693e-02 3.71553004e-01 6.22745991e-01
-2.65083816e-02 6.77670166e-02 -6.17092490e-01 2.52853096e-01
9.82294828e-02 1.27518165e+00 -6.24767363e-01 -5.63480258e-01
-7.66175449e-01 7.10140347e-01 3.86561155e-01 8.19216073e-01
-5.38840115e-01 7.89305791e-02 7.82294393e-01 -2.88207203e-01
6.48306608e-02 -6.18487716e-01 -2.13317245e-01 -1.65180433e+00
1.31373763e-01 -4.21073049e-01 2.47616529e-01 -7.08119094e-01
-1.19054389e+00 2.17239231e-01 -5.53271659e-02 -1.23578489e+00
-3.33529294e-01 -6.97895646e-01 -3.57369840e-01 4.36111063e-01
-2.03423190e+00 -8.54967713e-01 -4.23724949e-01 8.06427479e-01
5.93105137e-01 1.17537998e-01 5.56687713e-01 5.55842638e-01
-6.65087044e-01 1.96493864e-01 -2.00035855e-01 3.23651642e-01
8.55078399e-01 -1.22746968e+00 -1.43962890e-01 1.22067559e+00
5.24776876e-01 6.62062049e-01 8.69846642e-01 -3.22457314e-01
-1.19908392e+00 -7.75819778e-01 9.74293113e-01 -2.01218084e-01
7.09370136e-01 7.85279498e-02 -9.79490280e-01 7.87495553e-01
2.63701826e-01 5.72885513e-01 5.10267198e-01 -5.56545481e-02
-3.22091758e-01 -2.68782705e-01 -7.42694020e-01 5.37034810e-01
1.02424157e+00 -7.69795299e-01 -4.62871850e-01 3.56897920e-01
4.18529510e-01 -4.14009035e-01 -7.61137366e-01 3.11050892e-01
5.35292029e-01 -1.32541394e+00 1.13087571e+00 -6.86458468e-01
4.45073515e-01 -3.96742493e-01 -3.98033857e-02 -9.31303680e-01
4.55640033e-02 -9.13793564e-01 -3.63104880e-01 1.19874322e+00
-5.17784581e-02 -5.18718839e-01 1.09359896e+00 8.61452460e-01
-4.47321422e-02 -1.96861625e-01 -5.92053115e-01 -7.25923896e-01
-6.29088730e-02 -6.54815316e-01 9.50311273e-02 1.12167871e+00
-1.65890530e-01 -4.36553136e-02 -7.02486932e-01 7.21081123e-02
1.08696246e+00 1.78421587e-02 8.37094128e-01 -1.22834015e+00
-1.83910519e-01 -3.86129647e-01 -5.46943545e-01 -1.28534889e+00
6.13876045e-01 -5.17444968e-01 1.16596416e-01 -1.30867207e+00
-1.72161356e-01 -1.85331166e-01 -2.58354276e-01 2.74555027e-01
-1.79792315e-01 4.77991104e-01 1.49651080e-01 3.17744493e-01
-6.19245231e-01 4.23696935e-01 1.43287301e+00 -5.73651902e-02
-1.07350275e-01 -2.93606460e-01 -1.64303273e-01 9.68740821e-01
9.09956455e-01 -1.93048209e-01 -5.10905921e-01 -7.01641381e-01
-2.55211480e-02 -5.07874005e-02 5.91241360e-01 -9.80227828e-01
4.59820449e-01 -2.62115091e-01 2.41826013e-01 -4.08186257e-01
2.95640618e-01 -7.23951221e-01 -6.09449372e-02 1.38932467e-01
-3.96036386e-01 -1.57749802e-01 -1.20497227e-01 5.60036540e-01
-4.68506992e-01 -3.81145418e-01 6.70584738e-01 -1.49464399e-01
-8.28725994e-01 4.12423223e-01 -1.56075507e-01 1.83282226e-01
5.36609054e-01 -2.24366039e-01 -1.46858662e-01 -2.43622273e-01
-7.58758843e-01 -7.21156076e-02 2.97567427e-01 2.91702181e-01
4.60068613e-01 -1.19143820e+00 -4.01654959e-01 2.44115755e-01
-1.01473674e-01 2.96586245e-01 2.67888606e-02 9.75776970e-01
-8.43772769e-01 3.44261020e-01 -2.51452923e-01 -7.26825118e-01
-8.93246055e-01 4.42266166e-01 2.79758453e-01 -1.10300601e-01
-7.78197289e-01 5.67798257e-01 1.72834650e-01 5.04282489e-02
4.05795962e-01 -1.50142342e-01 -1.39886767e-01 1.08907200e-01
5.18208504e-01 5.52710176e-01 -7.34576136e-02 -9.10152912e-01
-1.20987847e-01 6.78580940e-01 4.07812744e-02 -5.08531332e-02
1.12517226e+00 -4.02027518e-01 -3.10369432e-01 4.33911085e-01
1.32444179e+00 -7.74929672e-02 -1.88035226e+00 -3.97279561e-01
7.37082213e-02 -6.92310631e-01 8.20597038e-02 -1.07312247e-01
-1.49058390e+00 1.02581382e+00 2.36629680e-01 5.77526204e-02
1.00650239e+00 -4.42643702e-01 7.01021850e-01 4.78329659e-01
1.34692073e-01 -1.08480048e+00 1.79705486e-01 5.12641311e-01
2.83519328e-01 -1.53223574e+00 2.53912389e-01 -5.81302702e-01
6.29649963e-03 1.34348416e+00 4.61073339e-01 -3.08465362e-01
6.65831625e-01 1.47982687e-02 1.89209193e-01 -2.37992741e-02
-4.17638779e-01 -5.77557206e-01 3.73217881e-01 5.87883115e-01
4.98524368e-01 -4.63618279e-01 -2.83015341e-01 -3.60777438e-01
-5.54703781e-03 1.74011648e-01 5.59798598e-01 1.08983922e+00
-1.78419441e-01 -1.30316794e+00 -2.55253911e-01 -6.74527660e-02
-6.91043794e-01 -7.14677796e-02 7.19324499e-02 7.10467517e-01
2.22441822e-01 1.10370851e+00 7.06779584e-02 2.89646149e-01
3.97966988e-02 3.66467983e-02 7.93696702e-01 -3.30472946e-01
-1.54835925e-01 3.37966859e-01 2.60581088e-04 -5.90246856e-01
-1.53042865e+00 -7.29960322e-01 -1.06447184e+00 -3.29519689e-01
-2.12288857e-01 -3.80007699e-02 3.13804865e-01 1.23299336e+00
3.69944870e-02 3.88522387e-01 4.46497053e-01 -1.13520563e+00
1.12980917e-01 -6.15822315e-01 -3.94381374e-01 8.96602452e-01
8.51377308e-01 -9.83185768e-01 -6.75999820e-01 8.34687054e-01]
|
[8.803592681884766, -1.782378911972046]
|
62e2c1d9-49c2-48de-925e-ffbcd1832eca
|
kagnet-knowledge-aware-graph-networks-for
|
1909.02151
| null |
https://arxiv.org/abs/1909.02151v1
|
https://arxiv.org/pdf/1909.02151v1.pdf
|
KagNet: Knowledge-Aware Graph Networks for Commonsense Reasoning
|
Commonsense reasoning aims to empower machines with the human ability to make presumptions about ordinary situations in our daily life. In this paper, we propose a textual inference framework for answering commonsense questions, which effectively utilizes external, structured commonsense knowledge graphs to perform explainable inferences. The framework first grounds a question-answer pair from the semantic space to the knowledge-based symbolic space as a schema graph, a related sub-graph of external knowledge graphs. It represents schema graphs with a novel knowledge-aware graph network module named KagNet, and finally scores answers with graph representations. Our model is based on graph convolutional networks and LSTMs, with a hierarchical path-based attention mechanism. The intermediate attention scores make it transparent and interpretable, which thus produce trustworthy inferences. Using ConceptNet as the only external resource for Bert-based models, we achieved state-of-the-art performance on the CommonsenseQA, a large-scale dataset for commonsense reasoning.
|
['Bill Yuchen Lin', 'Xiang Ren', 'Xinyue Chen', 'Jamin Chen']
|
2019-09-04
|
kagnet-knowledge-aware-graph-networks-for-1
|
https://aclanthology.org/D19-1282
|
https://aclanthology.org/D19-1282.pdf
|
ijcnlp-2019-11
|
['knowledge-base-question-answering']
|
['natural-language-processing']
|
[ 2.90298700e-01 1.04351544e+00 -3.58502008e-02 -3.41728359e-01
-2.78361320e-01 -3.59530091e-01 6.53643191e-01 3.22282940e-01
3.22104506e-02 6.00606680e-01 4.32392627e-01 -6.69009864e-01
-2.64847111e-02 -1.56455028e+00 -9.02921975e-01 2.59943604e-01
3.04627746e-01 8.71912003e-01 3.42710078e-01 -6.38920844e-01
3.23079020e-01 -1.31712900e-02 -1.18416357e+00 7.30001152e-01
1.38594747e+00 1.01056337e+00 -1.90646783e-01 4.94860113e-01
-7.31128514e-01 2.23136806e+00 -6.70970857e-01 -1.36809254e+00
-4.32172358e-01 -6.72520161e-01 -1.55651426e+00 -4.59607601e-01
4.02272761e-01 -4.22528237e-01 -6.31407082e-01 1.38112354e+00
-8.62445831e-02 2.91901082e-01 4.48711038e-01 -1.46025515e+00
-1.82001376e+00 1.33610582e+00 3.63998294e-01 1.44139990e-01
8.11973035e-01 2.54368603e-01 1.46618700e+00 -3.75085771e-01
6.67188108e-01 1.63475919e+00 5.52886188e-01 6.12342358e-01
-8.63142252e-01 -4.11632240e-01 -5.36412410e-02 1.01454115e+00
-8.58794808e-01 -1.80806443e-01 1.12718296e+00 -1.37477323e-01
1.53269947e+00 3.20515215e-01 1.02492869e+00 1.20860958e+00
4.21995461e-01 7.97963798e-01 9.68209803e-01 -2.38451794e-01
3.63129079e-01 -3.36908430e-01 4.85586643e-01 1.17000747e+00
3.63067091e-01 -2.05406308e-01 -4.61732417e-01 -1.38629630e-01
6.95654154e-01 -4.14206199e-02 -3.13809246e-01 -2.05497786e-01
-1.23226368e+00 1.12962008e+00 1.31152713e+00 3.46913606e-01
-4.98714328e-01 7.01574922e-01 4.40835208e-01 3.37393701e-01
2.61752278e-01 8.60690475e-01 -1.37317643e-01 2.08407700e-01
-3.89871120e-01 1.04965515e-01 1.14450300e+00 9.30826068e-01
4.39644039e-01 9.67849866e-02 -4.27218586e-01 3.89299572e-01
3.77365798e-01 6.55157208e-01 4.06160653e-01 -1.20412052e+00
6.29459798e-01 1.29270852e+00 -3.79462808e-01 -1.55088079e+00
-2.82445282e-01 -3.32941145e-01 -6.51585877e-01 -3.52545679e-01
8.67227688e-02 2.65050679e-01 -6.27945721e-01 1.71711767e+00
1.66622445e-01 2.49004409e-01 3.34528983e-01 9.99791324e-01
1.57158220e+00 3.27247113e-01 2.66132891e-01 1.52503163e-01
1.46768034e+00 -8.94745409e-01 -9.32722330e-01 -5.35337865e-01
5.15629411e-01 1.45500496e-01 1.39971972e+00 -3.21721174e-02
-1.06358659e+00 -2.30315715e-01 -1.07755971e+00 -8.55356097e-01
-7.16253400e-01 -4.61440951e-01 8.65915954e-01 -4.33781669e-02
-1.20192552e+00 6.31468356e-01 -3.81257951e-01 -4.19985622e-01
9.09145534e-01 -1.99069574e-01 -4.01331298e-02 -2.28190571e-01
-2.00130200e+00 1.38361979e+00 9.99919474e-01 1.75719678e-01
-7.41619825e-01 -5.09694815e-01 -1.31840897e+00 4.47298735e-01
8.74085784e-01 -1.30030632e+00 1.06436276e+00 -6.77464902e-01
-1.31066620e+00 1.18752182e+00 1.14530204e-02 -6.99211299e-01
2.93435484e-01 -4.98121418e-02 -6.50174081e-01 2.96763629e-01
2.04857677e-01 2.89696276e-01 7.08645582e-01 -1.10235810e+00
1.66032851e-01 -4.00670201e-01 7.39312768e-01 -7.32261091e-02
1.10728450e-01 -4.13342953e-01 5.72776683e-02 -3.13124180e-01
2.36347854e-01 -4.12643313e-01 1.27096415e-01 -1.97897986e-01
-1.01248467e+00 -5.04315436e-01 4.02665913e-01 -8.98860037e-01
1.09816933e+00 -1.51855278e+00 2.35256404e-01 4.41114604e-02
1.04248691e+00 -1.15098627e-02 -5.75762093e-02 3.32619786e-01
-2.00322140e-02 2.27128208e-01 -2.91057408e-01 1.89105183e-01
5.31508923e-01 4.06974554e-01 -8.08788955e-01 -3.19653362e-01
3.57066572e-01 1.85487926e+00 -1.58898067e+00 -6.16353691e-01
3.04784179e-02 1.09825812e-01 -5.03304362e-01 2.83125997e-01
-8.07252526e-01 -4.12768900e-01 -5.50203145e-01 4.33384866e-01
3.76571506e-01 -7.50128627e-01 2.25275904e-01 -3.07238042e-01
8.91869724e-01 7.00718105e-01 -3.18856001e-01 1.68848670e+00
-4.05251712e-01 6.46811903e-01 -5.75654447e-01 -8.16859066e-01
9.88327026e-01 -3.60107777e-04 -6.01534963e-01 -8.10525775e-01
2.33340055e-01 1.60496999e-02 -3.04307300e-03 -6.79379225e-01
4.92392838e-01 -5.83800673e-01 -2.11347193e-01 4.43984747e-01
2.15375319e-01 -7.53201425e-01 7.69424019e-03 1.01935387e+00
1.31860352e+00 1.46328226e-01 5.93399346e-01 -9.30395424e-02
4.02374357e-01 2.30897516e-01 -9.99448076e-03 6.33541465e-01
-1.49788350e-01 -6.35483414e-02 9.72685099e-01 -5.41979253e-01
-5.38725853e-01 -1.51998568e+00 4.60293561e-01 8.91572058e-01
1.25566214e-01 -4.32136625e-01 -7.05970407e-01 -8.49459887e-01
1.88791826e-01 1.75041294e+00 -8.47654104e-01 -7.87904441e-01
-2.00447664e-01 2.06563741e-01 7.30717659e-01 5.93348086e-01
9.20821667e-01 -1.83221591e+00 -5.86510837e-01 1.90818712e-01
-7.07519770e-01 -1.26613438e+00 6.86980635e-02 -2.77094170e-02
-6.96048141e-01 -1.53431213e+00 3.48968357e-01 -3.78505707e-01
4.15932178e-01 -1.67443022e-01 1.88987780e+00 5.84588170e-01
6.76390901e-02 3.89878541e-01 -3.05341870e-01 -3.70703161e-01
-6.32347226e-01 -4.29114282e-01 -5.00161231e-01 -4.39124912e-01
8.48129511e-01 -7.75864184e-01 -2.99118280e-01 -3.84824276e-01
-9.08636510e-01 1.77960798e-01 1.46731846e-02 5.86763322e-01
2.63123512e-01 -2.64576703e-01 6.28612280e-01 -1.09730303e+00
1.20039010e+00 -6.96005404e-01 -1.36964977e-01 4.66002345e-01
-2.92383105e-01 2.64188677e-01 1.01764476e+00 1.04698300e-01
-1.05385053e+00 -8.99466634e-01 1.52652755e-01 -4.70628679e-01
2.49131754e-01 8.62204432e-01 -3.08886468e-01 3.21977258e-01
9.56104934e-01 2.86672235e-01 -1.54988110e-01 3.20957482e-01
1.09355938e+00 3.18278223e-01 9.38293159e-01 -8.04498672e-01
7.19061196e-01 2.64603198e-01 2.85989675e-03 -2.82793790e-01
-1.68118012e+00 1.48012370e-01 -4.35878217e-01 -8.02029297e-02
1.01836312e+00 -4.68064904e-01 -1.15487838e+00 -6.24567270e-02
-1.57184315e+00 -4.52291518e-01 -5.03541708e-01 -1.41669035e-01
-7.99055934e-01 2.96344787e-01 -8.85543287e-01 -4.89718705e-01
-7.20365107e-01 -4.19513881e-01 7.23319650e-01 -4.61193323e-02
-4.65340167e-01 -1.34742296e+00 -2.26186171e-01 9.47561562e-01
4.46489960e-01 5.99119782e-01 1.43225873e+00 -9.99664128e-01
-7.28399277e-01 8.68769139e-02 -7.15463936e-01 2.47601986e-01
-1.55006021e-01 -3.29615742e-01 -9.98256385e-01 3.36319357e-01
-1.19003514e-02 -8.65688145e-01 1.02792001e+00 -1.25108345e-04
1.38624954e+00 -6.83123708e-01 -6.82577863e-02 3.39151025e-01
1.24893129e+00 -2.81672031e-01 1.09081936e+00 1.77970901e-01
9.51519608e-01 3.16142559e-01 9.65504795e-02 2.14980483e-01
1.10541236e+00 -9.36182886e-02 8.25172424e-01 2.52627403e-01
-2.61769533e-01 -8.69052231e-01 1.24497656e-02 7.55730808e-01
-1.98310614e-01 -1.36573732e-01 -1.07288241e+00 5.66664338e-01
-1.86712348e+00 -1.52630281e+00 -2.59820640e-01 1.38353479e+00
1.00018120e+00 1.36981130e-01 -4.89775926e-01 -1.05598196e-01
7.04123199e-01 3.32031608e-01 -8.38113904e-01 -6.39177799e-01
-1.00576676e-01 3.43863159e-01 -3.21326137e-01 7.03341544e-01
-5.56701481e-01 1.41497588e+00 5.43032503e+00 5.33703148e-01
-4.96767640e-01 5.44180423e-02 2.32926145e-01 1.59187958e-01
-9.17565525e-01 7.95715153e-02 2.47427896e-01 1.80905268e-01
8.24752450e-01 -5.02297819e-01 7.79260218e-01 9.12757158e-01
-4.22659248e-01 2.98112899e-01 -1.36087072e+00 8.28693867e-01
3.01971316e-01 -1.67666686e+00 6.39107227e-01 -4.68793899e-01
4.34098572e-01 2.64237970e-02 -2.83484787e-01 8.91896725e-01
8.60158265e-01 -1.29736817e+00 7.52747834e-01 8.31282020e-01
6.23742282e-01 -6.13159537e-01 7.64259636e-01 2.42381185e-01
-9.01226759e-01 -1.00715905e-01 -6.60773873e-01 -3.81703496e-01
1.15303900e-02 6.65797353e-01 -7.44943619e-01 6.22827709e-01
2.87465990e-01 6.08513355e-01 -6.39207125e-01 2.45967343e-01
-1.26523185e+00 4.70444798e-01 1.99364215e-01 -4.09818947e-01
1.44700453e-01 -1.74604561e-02 3.42260927e-01 9.44474518e-01
1.09486341e-01 6.18140578e-01 -1.27902046e-01 1.92254221e+00
-6.97608411e-01 -3.49329561e-01 -7.26979911e-01 -4.90847319e-01
6.07360899e-01 1.04615116e+00 -3.21578294e-01 -8.20284307e-01
3.09282653e-02 1.17799747e+00 1.10385227e+00 4.09131706e-01
-1.05466914e+00 -6.21256292e-01 2.20684305e-01 -6.50858060e-02
-2.92511135e-02 2.07454279e-01 -1.95990220e-01 -1.47227681e+00
4.74641882e-02 -7.54393995e-01 7.21590340e-01 -1.68843520e+00
-1.66160011e+00 6.17656827e-01 -2.19944194e-01 -2.99297988e-01
-3.23469967e-01 -6.54889524e-01 -9.72421944e-01 7.26093769e-01
-1.49538255e+00 -1.46383917e+00 -5.95175326e-01 8.53095174e-01
-3.57973203e-02 1.32278889e-01 1.01916444e+00 -6.66835308e-01
-6.69223070e-02 1.73420861e-01 -7.34297156e-01 4.81658936e-01
1.26260817e-01 -1.51397061e+00 7.84674644e-01 5.40614724e-01
1.26164585e-01 1.07796288e+00 5.67348301e-01 -9.17196095e-01
-1.37671387e+00 -1.11072838e+00 1.14438438e+00 -1.01268435e+00
1.31021583e+00 -1.18051894e-01 -1.28996396e+00 1.15853918e+00
3.47815990e-01 1.77901480e-02 7.49174953e-01 3.36643338e-01
-9.85825896e-01 4.27481532e-01 -1.26870024e+00 7.70966351e-01
1.65734589e+00 -1.18359518e+00 -1.82188833e+00 4.85782593e-01
1.51726234e+00 -4.42501545e-01 -8.32713962e-01 -1.49139985e-01
-5.40687293e-02 -9.51187134e-01 7.77495205e-01 -1.21829236e+00
1.26489890e+00 -2.21826926e-01 -1.43362820e-01 -1.59979522e+00
-4.74393100e-01 -3.94034177e-01 -7.06435859e-01 7.83913195e-01
4.13734138e-01 -7.25872159e-01 3.94122690e-01 8.66632342e-01
-2.80770093e-01 -6.04505301e-01 -8.63204598e-01 -5.09875894e-01
3.09245149e-03 -6.10920548e-01 1.01262259e+00 1.38076568e+00
8.00743639e-01 9.16467249e-01 2.94094592e-01 -5.09187803e-02
8.53905797e-01 5.33241987e-01 2.74877995e-01 -1.34112239e+00
-1.89951673e-01 -4.71967995e-01 -4.23016548e-01 -5.83677590e-01
8.91093373e-01 -1.59356999e+00 -2.86282718e-01 -2.43338251e+00
4.33836818e-01 3.85067344e-01 -2.12621361e-01 7.62337327e-01
-4.47874278e-01 -2.23966613e-01 3.10549945e-01 -3.13379616e-01
-9.98833299e-01 6.58496559e-01 1.63800693e+00 -5.27067840e-01
3.10251147e-01 -8.16225171e-01 -1.06289423e+00 7.94709861e-01
7.46063292e-01 -9.91561785e-02 -7.44535208e-01 -5.98971725e-01
8.47372830e-01 1.10752568e-01 1.08650875e+00 -6.52858555e-01
3.68717551e-01 -3.29656035e-01 3.09493661e-01 -2.52824783e-01
2.58913875e-01 -6.75458491e-01 -3.00824285e-01 5.04551649e-01
-6.16590738e-01 -1.06037825e-01 2.03367054e-01 5.36328077e-01
-1.88357174e-01 9.33048949e-02 4.15187567e-01 -3.68811935e-01
-9.08586264e-01 1.02272205e-01 3.73002857e-01 7.50571847e-01
6.31396770e-01 3.71909365e-02 -1.08814478e+00 -6.96360886e-01
-7.72595167e-01 3.00385863e-01 2.50902891e-01 2.67264962e-01
1.09644091e+00 -1.54727447e+00 -6.85250819e-01 -3.80166680e-01
4.59165722e-01 -1.32485673e-01 3.59586269e-01 2.64902145e-01
-6.03887796e-01 4.47247952e-01 -2.72051722e-01 -3.03075952e-03
-5.16051710e-01 9.56832409e-01 4.60902244e-01 -2.25714147e-01
-6.50521100e-01 9.81912017e-01 -9.37236920e-02 -7.66534150e-01
-3.77586544e-01 -8.54251564e-01 -2.37073496e-01 -3.72341573e-01
4.85071808e-01 1.61349669e-01 -2.40700036e-01 -3.08841825e-01
-4.13098037e-01 7.41689578e-02 3.33140671e-01 3.59414607e-01
1.05146098e+00 2.41919741e-01 -8.74380350e-01 3.82943213e-01
9.00439024e-01 -4.14710641e-01 -4.51272517e-01 -3.49061459e-01
-4.77869734e-02 -3.61466080e-01 -7.38413334e-02 -1.20680082e+00
-7.02651262e-01 8.93142998e-01 -6.48106873e-01 5.34024656e-01
6.93872094e-01 4.56221074e-01 9.66958702e-01 1.06141889e+00
4.50504780e-01 -7.91751564e-01 4.26079452e-01 9.86942470e-01
1.50407612e+00 -1.12807453e+00 -2.01147333e-01 -4.28732812e-01
-8.20667446e-01 1.13618064e+00 8.37961495e-01 -1.48065865e-01
5.54781780e-03 -3.59563649e-01 -2.58631080e-01 -8.85891259e-01
-8.83651555e-01 -1.62911654e-01 4.06491011e-01 8.27821136e-01
3.71964127e-01 4.92067128e-01 1.94426581e-01 1.10485840e+00
-7.92596519e-01 1.78363159e-01 5.28444827e-01 2.83804536e-01
-6.42336130e-01 -1.64482579e-01 -9.14091244e-02 6.47431791e-01
1.31222457e-01 -6.87443256e-01 -1.04266036e+00 7.89981425e-01
-2.14796394e-01 1.12445378e+00 -1.34170279e-01 -3.38725001e-01
4.79497433e-01 3.15195978e-01 6.99504673e-01 -5.41866481e-01
-4.55647320e-01 -1.31092787e+00 5.41705072e-01 -1.14120889e+00
-6.22869469e-02 2.35331029e-01 -2.01000762e+00 -8.81781399e-01
-6.63268343e-02 8.64887461e-02 9.67123806e-02 1.32127976e+00
3.99439275e-01 9.27595019e-01 -1.51983276e-01 8.95701870e-02
-6.68080986e-01 -8.44956040e-01 -4.84550476e-01 8.99702489e-01
1.42418854e-02 -4.79901195e-01 -4.07488853e-01 -6.44302741e-02]
|
[9.968637466430664, 8.041258811950684]
|
e5e978fb-22d1-4dc2-bd33-4ab5e1156f2f
|
towards-ground-truth-for-single-image
|
2206.10779
| null |
https://arxiv.org/abs/2206.10779v2
|
https://arxiv.org/pdf/2206.10779v2.pdf
|
Not Just Streaks: Towards Ground Truth for Single Image Deraining
|
We propose a large-scale dataset of real-world rainy and clean image pairs and a method to remove degradations, induced by rain streaks and rain accumulation, from the image. As there exists no real-world dataset for deraining, current state-of-the-art methods rely on synthetic data and thus are limited by the sim2real domain gap; moreover, rigorous evaluation remains a challenge due to the absence of a real paired dataset. We fill this gap by collecting a real paired deraining dataset through meticulous control of non-rain variations. Our dataset enables paired training and quantitative evaluation for diverse real-world rain phenomena (e.g. rain streaks and rain accumulation). To learn a representation robust to rain phenomena, we propose a deep neural network that reconstructs the underlying scene by minimizing a rain-robust loss between rainy and clean images. Extensive experiments demonstrate that our model outperforms the state-of-the-art deraining methods on real rainy images under various conditions. Project website: https://visual.ee.ucla.edu/gt_rain.htm/.
|
['Achuta Kadambi', 'Alex Wong', 'Stefano Soatto', 'Suya You', 'Celso de Melo', 'Chethan Chinder Chandrappa', 'Arnold Pfahnl', 'Akira Suzuki', 'Ethan Yang', 'Howard Zhang', 'Yunhao Ba']
|
2022-06-22
| null | null | null | null |
['single-image-deraining']
|
['computer-vision']
|
[-1.09054856e-02 -5.70571363e-01 4.31282908e-01 -6.97746158e-01
-9.66627419e-01 -3.70965213e-01 1.67206451e-01 -5.44881463e-01
-5.60314879e-02 1.18262172e+00 -5.24156131e-02 -2.01141626e-01
1.97864264e-01 -8.72751713e-01 -9.80512083e-01 -1.06231439e+00
-3.41607720e-01 1.30851135e-01 -8.84587318e-02 -4.55543220e-01
-2.00482592e-01 5.78341663e-01 -1.70287395e+00 3.22113857e-02
1.48220623e+00 5.05835593e-01 5.56428730e-01 8.39445710e-01
1.74915463e-01 8.00942719e-01 -6.69400394e-01 1.31936848e-01
5.87166309e-01 -7.26312280e-01 -1.96457103e-01 2.03994721e-01
9.92631257e-01 -6.50831223e-01 -4.34750706e-01 1.12901318e+00
7.62884617e-01 -5.81397600e-02 2.60013759e-01 -9.13654685e-01
-8.51784170e-01 -4.26864848e-02 -7.29511082e-01 5.83198428e-01
-4.76894118e-02 4.96050358e-01 6.88786745e-01 -1.15679896e+00
5.31264365e-01 1.18839467e+00 5.62739849e-01 3.31219763e-01
-1.16760898e+00 -7.19211280e-01 1.58408821e-01 2.21373037e-01
-1.11068499e+00 -5.65153420e-01 4.20192778e-01 -2.70942271e-01
3.78903866e-01 4.37650204e-01 6.77301824e-01 1.12243092e+00
7.35076219e-02 6.53343499e-01 1.66871333e+00 -3.26989561e-01
1.26502946e-01 -1.39663056e-01 2.33919367e-01 5.14835656e-01
6.42260909e-01 4.56165135e-01 -1.95342392e-01 3.39511067e-01
7.22186744e-01 2.46460363e-01 -9.22036588e-01 -2.00352892e-01
-8.00601244e-01 6.98412299e-01 7.75625229e-01 -7.37336837e-03
-4.26557600e-01 -1.87408015e-01 -2.02762052e-01 7.23835707e-01
8.55588615e-01 3.32606643e-01 -4.27012205e-01 4.28695619e-01
-1.27506912e+00 7.13941276e-01 7.71418452e-01 5.86841285e-01
9.59118009e-01 4.29743648e-01 -1.11734211e-01 8.86431038e-01
3.10016125e-01 1.52102017e+00 -1.29922882e-01 -9.48847890e-01
5.92967153e-01 -8.70360807e-02 6.37500703e-01 -8.28656077e-01
-2.87834048e-01 -4.80665267e-01 -1.09570050e+00 7.90013015e-01
4.32137132e-01 -1.49126381e-01 -1.28215826e+00 1.52190030e+00
1.84728846e-01 5.09992659e-01 2.50026673e-01 1.36507678e+00
9.01074946e-01 8.07950914e-01 -3.25517476e-01 -5.14529407e-01
9.60156739e-01 -9.60105956e-01 -9.47715580e-01 -4.98191088e-01
-7.40009919e-02 -7.27692842e-01 1.31342185e+00 3.23794872e-01
-8.81243825e-01 -4.92573321e-01 -1.02132964e+00 2.16312572e-01
-1.85138986e-01 -1.11521289e-01 3.96332651e-01 2.83645272e-01
-8.48746717e-01 5.96671224e-01 -7.24692881e-01 -3.95724535e-01
4.26431805e-01 -3.19492787e-01 -2.16981694e-01 -7.57518232e-01
-1.39966071e+00 1.06474197e+00 -1.17749318e-01 8.95029247e-01
-1.40591311e+00 -9.56568360e-01 -7.98254967e-01 -2.13834479e-01
-5.48918210e-02 -6.80701971e-01 9.19079125e-01 -1.20286918e+00
-1.03292406e+00 8.85498643e-01 -1.48450255e-01 -4.20474172e-01
4.61069703e-01 -7.80188501e-01 -7.08173633e-01 2.09368780e-01
1.99658647e-02 1.47549689e-01 1.33266485e+00 -2.11531949e+00
-4.26557302e-01 -3.28820407e-01 6.68548048e-02 3.22127551e-01
2.90458053e-01 -1.83585003e-01 -2.31201530e-01 -7.39810646e-01
-1.23616807e-01 -8.85573924e-01 -2.80061215e-01 2.40659207e-01
-2.31629670e-01 9.19805229e-01 9.93421614e-01 -1.06979871e+00
9.87972319e-01 -1.99199665e+00 -4.50381339e-02 -2.28522584e-01
3.23999196e-01 7.01205432e-01 -5.04096508e-01 2.69273698e-01
-1.22713871e-01 -1.22886732e-01 -8.39047492e-01 -5.07853627e-01
-2.68674403e-01 6.87103033e-01 -6.84870541e-01 8.52741897e-01
2.48119980e-01 5.17322540e-01 -8.35955977e-01 -2.44169101e-01
2.94280529e-01 7.35517383e-01 -4.34110947e-02 9.63832438e-01
-2.25028932e-01 6.38149142e-01 -1.72760606e-01 9.04315710e-01
1.54232693e+00 1.12925194e-01 -1.29703641e-01 -7.25494921e-02
-6.25315532e-02 -1.00065619e-01 -1.03410161e+00 1.21816170e+00
-6.81679845e-01 7.41136730e-01 6.02022529e-01 -6.21669352e-01
9.85940397e-01 9.22018476e-03 -1.99233994e-01 -9.48942959e-01
-2.32811570e-01 3.32933486e-01 -2.88693428e-01 -9.58882391e-01
3.14629614e-01 -4.12667006e-01 5.82821429e-01 1.70295537e-01
-2.79437512e-01 -4.57345575e-01 4.63079894e-03 1.63398713e-01
9.89843905e-01 7.76043534e-02 -1.67760119e-01 -1.79283604e-01
4.32033129e-02 -2.21197098e-01 7.57532477e-01 8.27111721e-01
-3.93848568e-01 1.30472267e+00 -1.35465428e-01 -6.10570073e-01
-1.00046217e+00 -1.40750241e+00 -2.77495861e-01 7.06806183e-01
4.07788813e-01 4.30468649e-01 -5.00459254e-01 -4.48417544e-01
1.05083875e-01 6.21807039e-01 -8.32199574e-01 2.89389670e-01
-6.15920365e-01 -1.57744849e+00 3.40547413e-01 6.36282563e-02
7.03198850e-01 -1.32247698e+00 -4.00435239e-01 -7.87069499e-02
-4.03160334e-01 -1.24026859e+00 -8.74742940e-02 1.68836549e-01
-9.48750556e-01 -1.31145048e+00 -6.10456884e-01 -5.00302553e-01
5.93226969e-01 9.02800560e-01 1.74546361e+00 1.45149603e-01
-5.75854301e-01 2.13364027e-02 -6.39347017e-01 -3.16155702e-01
-2.33009025e-01 -6.28820777e-01 -1.88174099e-01 -1.52444050e-01
-6.82813227e-02 -8.53881359e-01 -9.25607562e-01 8.34550038e-02
-1.15954220e+00 -1.36075079e-01 6.06785536e-01 9.26490247e-01
5.94644427e-01 -1.28498971e-01 3.64434749e-01 -9.18508053e-01
3.07942182e-01 -7.21319079e-01 -7.03920186e-01 1.29365742e-01
-4.04344827e-01 -1.94515109e-01 4.47993606e-01 5.73091432e-02
-1.34095597e+00 -2.14503139e-01 1.20693587e-01 -4.96032327e-01
-3.10751230e-01 3.48091096e-01 -2.30893344e-01 3.42423772e-03
7.97630668e-01 2.73391664e-01 -1.77581221e-01 -6.18954718e-01
5.16349137e-01 3.21546435e-01 8.24212372e-01 -2.73495495e-01
1.36965299e+00 7.44161546e-01 -2.02009007e-01 -1.01710927e+00
-1.36323273e+00 -3.19773942e-01 -3.18057954e-01 -1.66038081e-01
6.38137877e-01 -1.49995553e+00 -2.46492997e-02 8.75062525e-01
-1.04099500e+00 -7.08948135e-01 -1.94957286e-01 3.67824823e-01
-2.17017695e-01 4.16586727e-01 -5.19791663e-01 -8.73872817e-01
-7.13992715e-01 -6.80737972e-01 9.84945476e-01 3.70976120e-01
5.23806274e-01 -7.17489660e-01 4.71050650e-01 2.73914605e-01
7.60626078e-01 6.42566919e-01 2.27435037e-01 4.75959778e-01
-8.81061137e-01 1.98490709e-01 -4.94450569e-01 7.40726352e-01
3.84977609e-01 2.76084840e-01 -1.07996583e+00 -7.06286252e-01
2.18940631e-01 -4.95793998e-01 1.29241121e+00 5.30134618e-01
7.88154900e-01 -2.25593284e-01 1.69489682e-01 1.19298232e+00
1.86768055e+00 -2.90749907e-01 1.09149826e+00 5.30558050e-01
6.37321293e-01 3.11069399e-01 8.58932972e-01 4.53826398e-01
2.12973624e-01 3.28882784e-01 1.01419604e+00 -6.64990962e-01
-4.28694487e-01 2.64161050e-01 4.00985748e-01 6.29931390e-01
-1.85483560e-01 -5.59157252e-01 -6.05846763e-01 9.04435396e-01
-1.60981119e+00 -1.13453233e+00 -3.10615003e-01 2.24007630e+00
9.05988991e-01 -1.89517170e-01 -5.34927607e-01 -9.40786228e-02
5.26987910e-01 6.75281465e-01 -5.37659347e-01 1.07789621e-01
-6.92250490e-01 4.78626877e-01 6.23204947e-01 7.68023074e-01
-1.29322588e+00 7.99910009e-01 5.69858789e+00 3.16235840e-01
-1.21409154e+00 8.52913111e-02 4.74167377e-01 -2.27219298e-01
-2.77833045e-01 -2.23668948e-01 -6.64146304e-01 5.45881748e-01
8.27094376e-01 2.34074309e-01 5.26730239e-01 3.91717702e-01
8.11324716e-01 -2.91999817e-01 -3.89170527e-01 8.56221378e-01
2.91903075e-02 -9.98954058e-01 -1.93247542e-01 -4.15976852e-01
1.00328684e+00 6.08658671e-01 -6.35055378e-02 2.22712666e-01
6.42739177e-01 -1.16043580e+00 3.82861644e-01 8.98391068e-01
7.85535395e-01 -2.72198856e-01 7.62568355e-01 -2.39615086e-02
-7.89186954e-01 2.17285097e-01 -4.45379019e-01 1.19856605e-02
1.49883360e-01 1.51231170e+00 -3.24868858e-01 6.99204206e-01
1.23353791e+00 9.32000875e-01 -4.83065754e-01 1.14421165e+00
-7.40298808e-01 9.91508305e-01 -3.40506613e-01 8.47853601e-01
-5.67675419e-02 -5.19022405e-01 6.44468784e-01 1.44753492e+00
3.74745548e-01 2.08935335e-01 -4.01046686e-03 7.21801341e-01
-1.52488708e-01 -3.93645376e-01 -7.71857738e-01 3.45460832e-01
4.75823134e-01 1.16962743e+00 -9.80264470e-02 -1.58148840e-01
-1.98813140e-01 1.08198345e+00 1.90820962e-01 8.34488034e-01
-9.52813148e-01 -3.32895696e-01 1.35065150e+00 -3.22241336e-02
4.78718430e-01 -3.52874815e-01 -2.90783882e-01 -1.39151335e+00
3.08912843e-01 -1.19161248e+00 -8.68174210e-02 -9.97428000e-01
-1.45847893e+00 8.50343645e-01 -3.64884049e-01 -1.38298583e+00
2.68151253e-01 -2.43222162e-01 -9.06168997e-01 1.10849249e+00
-2.50936866e+00 -1.02158177e+00 -1.26400614e+00 5.63323736e-01
4.65730101e-01 1.74943835e-01 7.19985425e-01 5.10658085e-01
-6.88885391e-01 6.78117946e-02 6.63706481e-01 -1.88283905e-01
1.16950464e+00 -1.35432303e+00 4.12212640e-01 1.34992135e+00
-1.56477958e-01 1.79048464e-01 1.36550605e+00 -2.85560906e-01
-1.42465568e+00 -1.58136308e+00 6.34312630e-01 -2.52762735e-01
4.90809381e-01 -3.01186025e-01 -1.42774153e+00 5.35605609e-01
2.34197170e-01 6.73750043e-01 6.45463392e-02 -1.99158415e-01
-5.61941445e-01 -5.31125069e-01 -1.12854815e+00 3.44726235e-01
8.58421981e-01 -3.31603110e-01 -5.02908409e-01 5.56234896e-01
7.49347627e-01 -4.46160764e-01 -4.78752226e-01 6.36020899e-01
3.46740454e-01 -1.47194684e+00 9.38753188e-01 -2.76834279e-01
8.15332413e-01 -6.79971457e-01 -4.25806314e-01 -1.84906089e+00
-8.27129409e-02 -2.49643236e-01 -3.08757365e-01 1.04943860e+00
1.49493471e-01 -5.21434546e-01 3.24311316e-01 -8.31419677e-02
-1.87871292e-01 -3.56927872e-01 -5.85429013e-01 -8.33210051e-01
-2.20104735e-02 -8.15056041e-02 3.62307191e-01 9.84212637e-01
-9.90111470e-01 7.07523664e-03 -8.41074765e-01 8.55885148e-01
1.07850683e+00 7.52027631e-01 9.47274029e-01 -8.74011695e-01
-2.15302214e-01 1.32066160e-01 1.69915900e-01 -7.91907549e-01
-4.43592072e-02 -2.34463945e-01 7.61745512e-01 -1.69023979e+00
1.29738748e-01 -5.30871630e-01 -2.03119293e-01 4.02068049e-01
-6.72222733e-01 4.24965113e-01 2.26391941e-01 3.71706933e-01
-4.34380621e-01 1.06906807e+00 1.47934949e+00 -1.46081835e-01
-8.34271982e-02 -1.08275346e-01 -4.21972841e-01 6.10820711e-01
1.00669122e+00 -6.91879213e-01 -1.55961186e-01 -9.73854244e-01
-7.91804418e-02 8.11428875e-02 4.93012458e-01 -1.11310875e+00
-4.80110794e-01 -3.89799058e-01 1.94885448e-01 -3.42252463e-01
3.58652920e-01 -5.98811984e-01 5.68173639e-02 2.50155002e-01
8.02415684e-02 -2.25202590e-01 1.48740649e-01 5.97686768e-01
-5.75695097e-01 2.92271703e-01 1.39871550e+00 -1.05108216e-01
-6.50579333e-01 4.33729827e-01 -6.70908168e-02 2.58217245e-01
5.50293624e-01 2.14694366e-01 -8.66454661e-01 -5.37974417e-01
-5.91988564e-01 4.21890974e-01 5.00363350e-01 3.03002357e-01
7.02722967e-01 -8.03378999e-01 -1.27666473e+00 2.52678573e-01
8.69340599e-02 1.14058360e-01 4.76170242e-01 6.70389533e-01
-8.56067717e-01 -3.08959126e-01 -3.63165349e-01 -3.06243092e-01
-1.38669348e+00 2.63101429e-01 6.64147973e-01 -3.45838219e-01
-9.12235320e-01 6.99978411e-01 4.10309851e-01 -4.88919944e-01
6.11080304e-02 -3.41628641e-01 6.47662580e-02 -2.50614345e-01
8.33952487e-01 1.09849595e-01 2.43237466e-01 -3.07168007e-01
-1.24003977e-01 4.23249871e-01 1.92425981e-01 3.73572022e-01
1.72047603e+00 -4.30515558e-01 -1.69480860e-01 3.83657336e-01
8.90983403e-01 -1.56909287e-01 -1.64109457e+00 -2.51298398e-01
-6.56884730e-01 -9.45487797e-01 1.15480609e-01 -9.71905291e-01
-1.66880214e+00 8.85392070e-01 1.09703386e+00 -2.50421315e-02
1.49531341e+00 -4.85004485e-01 8.73256803e-01 5.66883147e-01
9.78822336e-02 -5.73209107e-01 -1.34054899e-01 7.90071070e-01
1.30687916e+00 -1.72509921e+00 1.48570269e-01 -3.67751896e-01
-5.59829772e-01 7.91649163e-01 6.01347804e-01 -4.95672703e-01
7.70015419e-01 4.31798369e-01 8.11104298e-01 -2.43783668e-01
-5.56915283e-01 -4.83057588e-01 -4.06167924e-01 7.22106874e-01
2.33363464e-01 2.75122851e-01 -1.83094010e-01 1.36741266e-01
1.70334339e-01 2.66698837e-01 8.35103869e-01 9.82262790e-01
-7.72267997e-01 -6.79863691e-01 -9.27053452e-01 3.66377294e-01
-1.95065811e-01 -3.71204734e-01 2.74017096e-01 7.53115714e-01
1.87013760e-01 9.98686314e-01 -2.27498993e-01 4.69236225e-02
4.52795178e-01 -4.97051209e-01 4.83920336e-01 -3.14221680e-01
-1.76374093e-01 -3.20182055e-01 -5.07880636e-02 -6.30515873e-01
-7.58507788e-01 -5.33740759e-01 -7.62871146e-01 -4.19716746e-01
-3.78065966e-02 7.52411932e-02 4.36704427e-01 7.37810194e-01
2.61941016e-01 4.70377475e-01 9.69490111e-01 -1.39645505e+00
-3.38331163e-01 -1.10055649e+00 -1.25578880e+00 5.89516163e-01
1.10086155e+00 -6.35124564e-01 -7.65477180e-01 2.39731073e-01]
|
[10.907051086425781, -3.239934206008911]
|
d428af48-766b-478d-890e-d0817b5b4252
|
computational-efficient-deep-neural-network
|
2011.12082
| null |
https://arxiv.org/abs/2011.12082v2
|
https://arxiv.org/pdf/2011.12082v2.pdf
|
Computational efficient deep neural network with difference attention maps for facial action unit detection
|
In this paper, we propose a computational efficient end-to-end training deep neural network (CEDNN) model and spatial attention maps based on difference images. Firstly, the difference image is generated by image processing. Then five binary images of difference images are obtained using different thresholds, which are used as spatial attention maps. We use group convolution to reduce model complexity. Skip connection and $\text{1}\times \text{1}$ convolution are used to ensure good performance even if the network model is not deep. As an input, spatial attention map can be selectively fed into the input of each block. The feature maps tend to focus on the parts that are related to the target task better. In addition, we only need to adjust the parameters of classifier to train different numbers of AU. It can be easily extended to varying datasets without increasing too much computation. A large number of experimental results show that the proposed CEDNN is obviously better than the traditional deep learning method on DISFA+ and CK+ datasets. After adding spatial attention maps, the result is better than the most advanced AU detection method. At the same time, the scale of the network is small, the running speed is fast, and the requirement for experimental equipment is low.
|
['Meichen Liu', 'Kejun Wang', 'Chenhui Wang', 'Jing Chen']
|
2020-11-24
| null | null | null | null |
['action-unit-detection', 'facial-action-unit-detection']
|
['computer-vision', 'computer-vision']
|
[ 5.47193829e-03 -4.61123139e-01 2.02640012e-01 -2.11647213e-01
1.54216483e-03 4.25538011e-02 -8.38654190e-02 -2.32498527e-01
-8.57766390e-01 3.57795149e-01 -2.11442158e-01 -1.01444893e-01
2.08513923e-02 -1.19961846e+00 -4.78014499e-01 -8.14666569e-01
2.47782305e-01 -1.58887893e-01 8.36001992e-01 -2.17384085e-01
3.01461011e-01 4.14903224e-01 -1.45726681e+00 2.49687225e-01
8.43454361e-01 1.43266058e+00 8.57276022e-01 2.95938849e-01
-1.41972527e-01 6.94568217e-01 -8.81540835e-01 1.46275088e-01
2.96363086e-01 -4.43534344e-01 -5.03043413e-01 -6.20110519e-02
-5.04163280e-02 -5.78720033e-01 -4.97886568e-01 1.32329893e+00
8.57611895e-01 1.67628556e-01 4.92926925e-01 -8.67134690e-01
-8.35540056e-01 3.40482086e-01 -7.21900463e-01 6.51175320e-01
-3.18473130e-01 2.05117509e-01 5.34821391e-01 -9.17786360e-01
1.92426331e-02 1.04664922e+00 4.76902246e-01 4.52184081e-01
-7.14883685e-01 -8.73018324e-01 2.16075405e-01 6.23090446e-01
-1.56849122e+00 1.00413598e-01 7.31998026e-01 -2.04255074e-01
6.54931843e-01 1.73180401e-01 6.15659595e-01 5.48670352e-01
1.27854839e-01 9.20673370e-01 6.49564803e-01 -3.98100108e-01
-9.51487273e-02 1.03301018e-01 1.63383126e-01 6.71748102e-01
1.92813635e-01 -1.39823645e-01 1.06654257e-01 3.71087968e-01
1.03773916e+00 5.27572691e-01 -5.66408515e-01 -2.65914630e-02
-1.12231135e+00 8.16353798e-01 1.08103120e+00 5.99683285e-01
-4.19083983e-01 -3.55560668e-02 5.24298012e-01 1.23136044e-01
2.08610684e-01 6.08348139e-02 -3.53186131e-01 1.71827480e-01
-5.58144212e-01 -1.45136505e-01 5.57114780e-02 8.32938552e-01
8.18862379e-01 1.91483777e-02 -2.16484979e-01 1.09866393e+00
-1.92692913e-02 1.75936460e-01 9.16675210e-01 -4.64561909e-01
4.98117208e-01 8.67005646e-01 -9.96457115e-02 -1.18818581e+00
-5.22622347e-01 -5.19388020e-01 -1.16072845e+00 4.27582830e-01
2.97656268e-01 -3.44012946e-01 -9.04214621e-01 1.26051033e+00
8.01460817e-02 1.80400554e-02 -2.75461958e-03 1.21676791e+00
1.08838832e+00 1.04851067e+00 -6.40088841e-02 -1.04085661e-01
1.42698276e+00 -1.18751335e+00 -7.15436876e-01 -4.93356526e-01
6.50042057e-01 -6.63140833e-01 1.35071385e+00 4.47667271e-01
-8.80578101e-01 -1.06015408e+00 -1.30417728e+00 -9.38056558e-02
-4.48595375e-01 7.15551078e-01 4.74056482e-01 2.34398022e-01
-7.40549445e-01 4.06331122e-01 -4.96579766e-01 -1.48258612e-01
5.04316151e-01 3.96776199e-01 -7.73792202e-03 -9.87001359e-02
-1.46850729e+00 7.37474322e-01 8.62969697e-01 4.54535574e-01
-5.12208939e-01 -7.92632997e-02 -8.22478533e-01 2.67648697e-01
2.77490437e-01 -7.55704939e-02 1.08010781e+00 -1.17076862e+00
-1.20381415e+00 3.49060953e-01 1.05692409e-01 -1.01702109e-01
3.90818447e-01 -5.26860356e-02 -3.78681511e-01 6.17507696e-02
7.93329328e-02 7.78205216e-01 6.08748972e-01 -7.47221947e-01
-8.92925620e-01 -6.79391474e-02 1.41288461e-02 3.43387634e-01
-6.80517495e-01 2.13611916e-01 -8.04043949e-01 -6.85674131e-01
2.78143108e-01 -5.47033370e-01 -1.42715901e-01 1.50950670e-01
-6.62799971e-03 -3.66603136e-01 1.16959715e+00 -6.11841023e-01
1.43620670e+00 -2.43463969e+00 2.93615535e-02 1.22465581e-01
4.17708158e-02 6.25870883e-01 2.83277743e-02 -1.98685542e-01
-9.27409008e-02 -9.45793465e-02 -1.79263443e-01 2.32122764e-01
-4.73147243e-01 9.80559140e-02 1.74387008e-01 2.54647911e-01
1.09643631e-01 6.25315368e-01 -6.39017999e-01 -7.02173293e-01
3.16161394e-01 1.65454805e-01 -2.68534422e-01 2.34045878e-01
7.02227876e-02 9.30818841e-02 -4.18092221e-01 2.67482787e-01
8.18117619e-01 -1.74056470e-01 -4.39451873e-01 -3.83961827e-01
-2.56582499e-01 -3.00940070e-02 -1.37005675e+00 1.31851399e+00
-3.94828975e-01 6.04915500e-01 -7.67534748e-02 -1.07949591e+00
1.24152875e+00 3.34456861e-02 4.40180255e-03 -1.02492893e+00
7.54570305e-01 1.76966973e-02 3.46299142e-01 -7.33979702e-01
3.42055298e-02 1.54111624e-01 6.94832355e-02 4.05314535e-01
-2.53047585e-01 2.94104576e-01 1.45858169e-01 -1.87114134e-01
5.64299881e-01 -1.35340616e-01 1.19028017e-01 -2.85911828e-01
7.42339969e-01 1.06716594e-02 7.25355804e-01 3.78909260e-01
-1.96978182e-01 5.81402242e-01 3.65337193e-01 -6.17460430e-01
-7.63468862e-01 -4.98865247e-01 -2.42430463e-01 1.22122777e+00
6.08759880e-01 7.94264451e-02 -8.97712529e-01 -7.04468012e-01
-3.79640967e-01 2.97194541e-01 -7.68998206e-01 -4.73440051e-01
-7.00044811e-01 -9.06940460e-01 2.88313210e-01 9.64853585e-01
1.24155033e+00 -1.57839596e+00 -7.94354916e-01 2.23213792e-01
-5.73735200e-02 -6.83918059e-01 -5.98949373e-01 1.47529267e-04
-7.97158599e-01 -8.53416085e-01 -1.05748224e+00 -1.56548047e+00
8.18191826e-01 4.70019341e-01 3.94845009e-01 3.59525055e-01
-3.55840087e-01 -5.51493347e-01 -4.01283741e-01 -6.32327080e-01
1.56437397e-01 2.92965304e-02 -1.20183088e-01 7.46699274e-02
6.45235181e-01 -1.86371416e-01 -7.96977520e-01 4.57917333e-01
-9.32910264e-01 8.88259560e-02 8.35936308e-01 9.89186883e-01
4.59119201e-01 2.96414286e-01 4.82102394e-01 -3.47740620e-01
7.60040462e-01 -1.28835052e-01 -6.58098638e-01 1.21292740e-01
-2.33364537e-01 -1.11035116e-01 8.22029710e-01 -7.20897257e-01
-9.49077129e-01 5.17682126e-03 -2.13364512e-01 -4.53421950e-01
-1.72060475e-01 3.86611164e-01 -1.89171836e-01 -4.44530882e-02
6.79858863e-01 4.39374834e-01 1.40465170e-01 -4.90906924e-01
-8.69228542e-02 1.08973444e+00 4.66661543e-01 5.67344315e-02
2.97633380e-01 1.36380196e-01 -4.13501799e-01 -5.49956799e-01
-6.15083218e-01 -4.72500920e-02 -5.69354117e-01 -3.83622833e-02
1.11659813e+00 -7.57626474e-01 -6.22570992e-01 7.79978871e-01
-1.17536616e+00 -3.95089328e-01 2.48579457e-01 6.47908330e-01
7.22276866e-02 3.12392414e-01 -9.31347013e-01 -3.59619558e-01
-6.32376134e-01 -1.24083686e+00 5.58136225e-01 7.23615825e-01
1.75827742e-01 -6.01342857e-01 -5.18072128e-01 -2.39907458e-01
3.23446393e-01 -1.74334437e-01 6.89424276e-01 -3.08711469e-01
-3.00907671e-01 -3.47138345e-01 -7.33656168e-01 6.14018619e-01
2.57989585e-01 3.33185084e-02 -7.21997857e-01 -1.30721048e-01
1.87564254e-01 -1.22960478e-01 9.76548135e-01 3.70588571e-01
1.82740426e+00 -3.39603007e-01 -4.82625961e-01 5.18283129e-01
1.22871470e+00 7.90923178e-01 8.67273211e-01 5.43523014e-01
7.67858088e-01 5.87659299e-01 9.23065186e-01 1.52992234e-01
-2.21577361e-02 4.73988950e-01 5.50120831e-01 -6.19286478e-01
7.62148798e-02 2.41247550e-01 4.99945842e-02 5.38571835e-01
-5.67823015e-02 4.55121212e-02 -7.93728709e-01 6.08696282e-01
-1.66791081e+00 -1.05571544e+00 -3.52551490e-01 2.03561568e+00
6.80537879e-01 2.57729411e-01 7.12878928e-02 3.93386126e-01
9.67195034e-01 2.94770468e-02 -4.68144268e-01 -3.42702776e-01
7.05254897e-02 2.32957721e-01 3.50642174e-01 1.16849557e-01
-1.15372062e+00 7.89716780e-01 5.08221149e+00 1.04460740e+00
-1.53115797e+00 1.20390400e-01 7.73962975e-01 -8.66580084e-02
2.98223138e-01 -4.42626715e-01 -7.69654512e-01 8.95726740e-01
1.75799698e-01 1.56094536e-01 1.92446128e-01 1.03981447e+00
3.34478542e-02 -2.00722709e-01 -6.33953989e-01 1.26475048e+00
8.38125050e-02 -1.04831421e+00 -2.53455788e-01 -2.52696037e-01
4.81450140e-01 2.22633518e-02 1.00728281e-01 2.64415205e-01
-1.00133844e-01 -8.74221087e-01 6.19804561e-01 2.26071388e-01
8.38588417e-01 -9.15997922e-01 1.13772571e+00 5.20524979e-01
-1.30144715e+00 -3.60368162e-01 -8.90424252e-01 -3.10726672e-01
-3.41636539e-01 3.44263017e-01 -3.62925798e-01 2.68430829e-01
1.20328522e+00 5.45673311e-01 -5.73964834e-01 1.10892582e+00
-3.74177039e-01 2.16677785e-01 -4.26193327e-01 -5.50721765e-01
5.08456290e-01 -1.81947425e-01 -2.05143392e-01 1.24201977e+00
5.32587528e-01 4.41626072e-01 -1.45715009e-02 6.53177738e-01
-4.50094603e-02 2.75773019e-01 -4.75317746e-01 5.00012398e-01
3.93155903e-01 1.32204068e+00 -6.54120266e-01 -5.53055406e-01
-4.59484428e-01 1.24143875e+00 3.27598423e-01 1.84180632e-01
-9.82884943e-01 -1.20770144e+00 2.57983029e-01 -1.99400738e-01
6.46216035e-01 2.74175480e-02 -2.27051616e-01 -7.65511751e-01
-9.44143219e-04 -5.06460607e-01 5.46098173e-01 -1.01300025e+00
-9.86019969e-01 8.07959199e-01 -2.46142253e-01 -1.50601459e+00
1.94601879e-01 -8.46201897e-01 -1.02794421e+00 1.15604174e+00
-1.17035413e+00 -4.66848850e-01 -7.92097092e-01 6.49595618e-01
4.70149338e-01 3.59616168e-02 4.73649591e-01 4.87279803e-01
-1.01167345e+00 7.29310632e-01 1.31538749e-01 6.42007828e-01
4.72501904e-01 -6.96837306e-01 8.68229792e-02 9.39350486e-01
-2.43705422e-01 3.19582522e-01 1.61539316e-01 -3.62265855e-01
-5.39986551e-01 -1.19014025e+00 5.16746700e-01 1.48914605e-01
1.92130283e-01 -2.97077537e-01 -1.16207385e+00 5.01831055e-01
2.09013626e-01 2.83133984e-01 2.08273530e-01 -3.81455421e-01
1.80023521e-01 -4.33051169e-01 -1.06057525e+00 6.11428499e-01
9.12416577e-01 -4.80261333e-02 -5.65782845e-01 1.67483866e-01
6.87141240e-01 -5.61252892e-01 -6.14571512e-01 5.05261123e-01
2.55130559e-01 -8.98841202e-01 7.35903561e-01 -1.15630835e-01
3.34679693e-01 -6.53916776e-01 2.07163602e-01 -1.15201032e+00
-7.35488713e-01 7.85932317e-02 3.82362038e-01 9.70422149e-01
3.54167670e-01 -7.32053876e-01 6.66617751e-01 1.85266182e-01
-2.49951899e-01 -9.63369846e-01 -4.87822145e-01 -5.51803052e-01
-1.07768305e-01 -1.58296213e-01 7.15086222e-01 8.80385220e-01
-2.11890310e-01 5.23550630e-01 -2.10235119e-01 3.22594821e-01
1.16557866e-01 1.07786573e-01 4.11549062e-01 -1.11916888e+00
-1.28068477e-01 -5.50137460e-01 -4.18283939e-01 -1.28775799e+00
-4.81081039e-01 -4.09690350e-01 6.80152848e-02 -1.73901057e+00
3.20108756e-02 -7.00174212e-01 -5.86133540e-01 8.01093519e-01
-5.09520710e-01 4.86194223e-01 -1.98474899e-02 2.41741613e-01
-3.30146015e-01 6.07997060e-01 1.39022160e+00 -2.39411190e-01
-2.93111801e-01 -7.03029856e-02 -5.99037588e-01 8.16007912e-01
1.19155967e+00 -2.16407254e-01 -2.98423141e-01 -7.76849627e-01
-3.15013587e-01 -2.50760317e-01 2.00995639e-01 -1.29831398e+00
3.56963903e-01 7.41241649e-02 9.31705117e-01 -5.78055024e-01
1.51161104e-01 -8.59011829e-01 -3.02573532e-01 8.61465514e-01
-2.68259868e-02 2.76131257e-02 4.48957562e-01 7.40346313e-02
-1.83553919e-01 -6.51547253e-01 9.45033193e-01 -1.97063446e-01
-1.16710305e+00 4.25886810e-01 -2.89128542e-01 -3.00367236e-01
1.41816032e+00 -4.25371021e-01 -1.97185144e-01 -1.73203483e-01
-5.76173186e-01 3.13333333e-01 1.47213161e-01 3.83617282e-01
7.62173295e-01 -1.50098860e+00 -5.42632043e-01 3.94357681e-01
8.61342177e-02 4.91748691e-01 5.76258898e-01 7.66462386e-01
-8.70874584e-01 2.42793962e-01 -4.23071444e-01 -5.71459532e-01
-1.44043684e+00 8.44928622e-01 5.88140309e-01 1.78421229e-01
-5.34116864e-01 1.19362843e+00 5.08065343e-01 -7.52237588e-02
2.82999754e-01 -2.91037649e-01 -5.53283095e-01 -1.55755848e-01
9.84183788e-01 2.65935689e-01 -1.83568895e-02 -4.09996659e-01
-3.10788125e-01 6.80925488e-01 -3.92763048e-01 2.94741988e-01
1.13512075e+00 1.14920363e-01 -1.54432207e-01 6.47121295e-02
1.26592743e+00 -2.68868148e-01 -1.40145767e+00 -2.95362920e-01
-6.27200663e-01 -3.30397934e-01 1.74348846e-01 -6.13010347e-01
-1.40071785e+00 1.37977397e+00 9.59637880e-01 2.68098265e-01
1.45696330e+00 -4.87807505e-02 8.13588023e-01 1.45651028e-01
1.21428013e-01 -1.09226787e+00 1.51632383e-01 4.37448382e-01
8.52699757e-01 -1.03237593e+00 -8.68939608e-02 -2.64205158e-01
-6.92308843e-01 1.13235533e+00 1.27754772e+00 -4.05197293e-01
3.74741226e-01 2.00278565e-01 1.62336543e-01 -7.10413456e-02
-1.07748665e-01 -2.72518039e-01 6.01898991e-02 4.67824221e-01
1.99618280e-01 -3.01801860e-01 -4.97936517e-01 6.51014984e-01
2.04858966e-02 -1.56553403e-01 2.59859473e-01 8.27646375e-01
-8.45218778e-01 -6.94678485e-01 -4.36580986e-01 5.59432745e-01
-4.74980742e-01 -2.06493810e-01 -9.94389281e-02 8.07607949e-01
6.08560145e-01 7.81402111e-01 5.37942290e-01 -6.91353798e-01
3.68113637e-01 -4.10843402e-01 2.36641020e-01 -3.52085680e-01
-3.27317804e-01 -3.70048173e-02 -3.90267283e-01 -2.46200979e-01
-2.40620393e-02 -2.63209820e-01 -1.66419506e+00 -2.20306411e-01
-5.78078210e-01 2.49742806e-01 1.31863683e-01 7.79184461e-01
3.16170961e-01 8.30172122e-01 5.81301928e-01 -7.76096284e-01
-1.82066724e-01 -1.33397627e+00 -2.99145907e-01 1.86860055e-01
-3.06348726e-02 -6.02272511e-01 -1.14478052e-01 -2.86776453e-01]
|
[9.225232124328613, -0.5527082681655884]
|
2cf784e4-2ceb-48e6-9c1f-e5d44c89a9ad
|
interpreting-black-box-predictions-using
|
1810.10118
| null |
http://arxiv.org/abs/1810.10118v1
|
http://arxiv.org/pdf/1810.10118v1.pdf
|
Interpreting Black Box Predictions using Fisher Kernels
|
Research in both machine learning and psychology suggests that salient
examples can help humans to interpret learning models. To this end, we take a
novel look at black box interpretation of test predictions in terms of training
examples. Our goal is to ask `which training examples are most responsible for
a given set of predictions'? To answer this question, we make use of Fisher
kernels as the defining feature embedding of each data point, combined with
Sequential Bayesian Quadrature (SBQ) for efficient selection of examples. In
contrast to prior work, our method is able to seamlessly handle any sized
subset of test predictions in a principled way. We theoretically analyze our
approach, providing novel convergence bounds for SBQ over discrete candidate
atoms. Our approach recovers the application of influence functions for
interpretability as a special case yielding novel insights from this
connection. We also present applications of the proposed approach to three use
cases: cleaning training data, fixing mislabeled examples and data
summarization.
|
['Oluwasanmi Koyejo', 'Joydeep Ghosh', 'Been Kim', 'Rajiv Khanna']
|
2018-10-23
| null | null | null | null |
['data-summarization']
|
['miscellaneous']
|
[ 4.98290539e-01 6.75177515e-01 -2.97504216e-01 -7.87475288e-01
-8.21274102e-01 -5.87140739e-01 4.80153292e-01 4.95139331e-01
-3.26095670e-01 9.30080414e-01 5.12010492e-02 -3.36362481e-01
-6.79119110e-01 -6.33519471e-01 -9.62322891e-01 -6.66704178e-01
1.47622108e-01 6.06244385e-01 1.56872913e-01 1.05549932e-01
6.86319828e-01 4.77506280e-01 -1.79934406e+00 3.66306305e-01
1.21435833e+00 7.77367830e-01 5.59870787e-02 4.62320238e-01
1.08381778e-01 5.20425677e-01 -4.24074262e-01 -6.64357960e-01
1.56711712e-01 -3.61972719e-01 -9.39975441e-01 2.66402841e-01
4.74080473e-01 -3.48317139e-02 6.76113844e-01 1.14268482e+00
2.59058654e-01 2.62418777e-01 1.07701886e+00 -1.22865498e+00
-7.76549995e-01 5.27620852e-01 -3.07132781e-01 1.87750861e-01
2.64074177e-01 -1.02576204e-01 1.33746541e+00 -1.07847631e+00
4.70129102e-01 1.00433600e+00 6.39522851e-01 5.08807123e-01
-1.42829669e+00 -2.41145089e-01 2.05516800e-01 2.85903066e-01
-1.05054307e+00 -4.82132286e-01 6.70048952e-01 -3.35999399e-01
6.42096102e-01 6.49449468e-01 5.20494878e-01 9.19188023e-01
-5.10178208e-02 8.64133656e-01 1.04411232e+00 -8.18265617e-01
6.28982723e-01 8.35136414e-01 6.18351221e-01 5.38951218e-01
6.77591324e-01 -9.21882614e-02 -5.62288702e-01 -3.71087521e-01
4.75351214e-01 -6.99326843e-02 -2.08047628e-01 -4.92532700e-01
-8.51115525e-01 9.70403612e-01 2.08591104e-01 -9.62682366e-02
-3.16809207e-01 -6.90733865e-02 5.28150611e-02 1.45234600e-01
5.91418564e-01 6.27130210e-01 -6.47932112e-01 3.17742079e-01
-7.18418181e-01 4.42678779e-01 8.91768277e-01 9.28575695e-01
8.23577702e-01 -2.73354650e-01 -1.76990226e-01 7.86989450e-01
5.49096823e-01 3.01809877e-01 2.94816941e-01 -1.05612576e+00
3.07596296e-01 5.25247753e-01 3.56308132e-01 -7.57300079e-01
-2.40484431e-01 -6.59808099e-01 -2.96108931e-01 1.44175977e-01
5.08936703e-01 -2.40131710e-02 -5.56174338e-01 1.54804134e+00
6.51102722e-01 1.08603977e-01 1.20506972e-01 7.04097807e-01
3.19129825e-01 2.38943502e-01 2.77881846e-02 -5.03837407e-01
1.24749482e+00 -6.36371374e-01 -5.45992911e-01 -1.33020401e-01
6.32873833e-01 -4.89011884e-01 1.07504976e+00 6.80027425e-01
-9.73362267e-01 -4.63029295e-01 -8.25746477e-01 -9.70281735e-02
-2.15149805e-01 3.34742457e-01 6.50446832e-01 7.13694394e-01
-7.82291591e-01 8.12685668e-01 -4.87661928e-01 -1.66931495e-01
6.41305029e-01 4.71979946e-01 -2.53794521e-01 8.05141181e-02
-7.61171281e-01 9.35410798e-01 4.96292591e-01 1.52880579e-01
-5.26548386e-01 -7.74026453e-01 -6.97952509e-01 1.30087838e-01
5.69457769e-01 -8.01617563e-01 1.31169426e+00 -1.29447222e+00
-1.07403409e+00 7.27203667e-01 -5.28807998e-01 -6.71327591e-01
4.13773984e-01 -5.73111773e-01 1.45995058e-02 1.08183824e-01
1.96871042e-01 5.18101633e-01 9.50618565e-01 -1.40832269e+00
-6.02430880e-01 -4.82812554e-01 4.13179621e-02 1.02973916e-01
-1.94228247e-01 -2.12620661e-01 1.90890715e-01 -4.06246990e-01
2.69390374e-01 -6.41333759e-01 -3.41130286e-01 -1.94568872e-01
-5.44870675e-01 -4.37815040e-01 2.95792729e-01 -3.92878532e-01
1.19817388e+00 -1.93389547e+00 1.57540068e-01 4.30980951e-01
2.19117761e-01 3.02014332e-02 3.99129272e-01 3.10615033e-01
-1.50690138e-01 3.08971912e-01 -3.52688670e-01 -3.35213423e-01
2.61071205e-01 1.47248119e-01 -6.99606180e-01 4.20608252e-01
6.19784534e-01 7.23784506e-01 -8.09892476e-01 -4.99507874e-01
2.12706923e-01 2.35867009e-01 -7.86812544e-01 1.20209016e-01
-3.22354525e-01 1.09454088e-01 -5.74611843e-01 5.68398535e-01
6.64226949e-01 -2.64768571e-01 1.29071400e-01 -3.28420475e-02
4.64301743e-02 1.92588046e-01 -1.33799982e+00 1.10203171e+00
-7.45809376e-02 2.40733430e-01 -3.53059798e-01 -1.24059641e+00
8.38989437e-01 3.92743535e-02 1.32199004e-02 -8.70484114e-02
-6.04787879e-02 3.87324452e-01 -3.30579542e-02 -7.08422184e-01
2.99157530e-01 -4.37059730e-01 1.32271722e-01 4.72512156e-01
1.50521234e-01 1.70451760e-01 -6.18950874e-02 -8.17260668e-02
8.25721323e-01 2.51869649e-01 7.39840508e-01 -4.64207560e-01
4.65571433e-01 7.98142478e-02 5.64984500e-01 9.61304724e-01
2.47350838e-02 5.71041822e-01 6.72272265e-01 -5.53381920e-01
-1.01281559e+00 -1.05741370e+00 -4.32883710e-01 9.32013214e-01
-9.71754417e-02 -2.99016595e-01 -8.53004158e-01 -8.95832360e-01
1.24461330e-01 1.35234153e+00 -9.77747142e-01 -1.60187647e-01
-1.42412499e-01 -8.26782405e-01 5.73978834e-02 2.95277625e-01
8.06064233e-02 -1.06207848e+00 -1.00826991e+00 -6.53035343e-02
1.32476032e-01 -4.39753413e-01 1.31993992e-02 6.27314568e-01
-1.11347139e+00 -1.28037405e+00 -2.05070451e-01 -4.55631703e-01
9.22157288e-01 1.93014577e-01 1.09730971e+00 -8.43437612e-02
-1.33542493e-01 3.88537884e-01 -5.10659516e-01 -7.64285862e-01
-3.64282638e-01 -8.11926275e-02 1.95278004e-02 7.86883458e-02
7.54554927e-01 -5.26015520e-01 -5.33850133e-01 7.53398538e-02
-9.23484147e-01 -3.06188241e-02 7.23123074e-01 9.11809206e-01
7.24325120e-01 -7.70430937e-02 7.31432259e-01 -1.51345479e+00
7.80277908e-01 -7.54129589e-01 -4.11888361e-01 6.14726365e-01
-8.58858883e-01 3.92660379e-01 7.79800653e-01 -3.38857412e-01
-9.85557139e-01 2.83816550e-02 4.14398536e-02 -1.90456554e-01
-4.63667780e-01 3.82614225e-01 -1.09236456e-01 2.01852098e-01
9.57036316e-01 1.06814422e-01 -1.71246395e-01 -4.04153436e-01
3.60824198e-01 5.43193877e-01 1.21652752e-01 -8.29949677e-01
6.35940254e-01 4.40631419e-01 8.97687450e-02 -7.34467328e-01
-1.16575897e+00 -3.80160481e-01 -7.83007145e-01 -1.50944307e-01
5.80089092e-01 -4.16556954e-01 -8.19451094e-01 -5.21502435e-01
-1.10804820e+00 1.43714666e-01 -7.10971653e-01 4.75116611e-01
-8.37322235e-01 2.54882306e-01 1.05102398e-01 -1.29772568e+00
-1.85851648e-01 -9.66124535e-01 1.21935558e+00 2.55258709e-01
-4.44955111e-01 -1.11011970e+00 4.31566238e-02 3.42545837e-01
-1.42215326e-01 1.99964121e-01 1.01661646e+00 -1.33951616e+00
-4.84368414e-01 -2.89003223e-01 6.97740093e-02 5.06510198e-01
-1.17072485e-01 -1.31668016e-01 -1.32726240e+00 4.48658317e-03
3.87522578e-01 -3.59519631e-01 8.84523392e-01 5.77376664e-01
1.18399966e+00 -4.48626608e-01 -1.82505190e-01 2.76373714e-01
1.37651098e+00 -1.50806606e-02 4.26084995e-01 2.25943133e-01
3.38759422e-01 1.03580046e+00 9.75992203e-01 5.99026859e-01
3.67072225e-02 4.28652018e-01 4.41139787e-01 3.00082564e-01
5.26256502e-01 -2.22175956e-01 2.47414187e-01 2.18385383e-01
-1.82866499e-01 -4.83423024e-02 -7.05450535e-01 3.84430915e-01
-2.09993958e+00 -1.03494346e+00 -2.06080243e-01 2.51920581e+00
5.05660236e-01 1.97190881e-01 5.42530557e-03 3.08713257e-01
5.19263268e-01 -3.35788012e-01 -4.90111500e-01 -4.15898323e-01
2.30287597e-01 3.12846154e-01 2.32104883e-01 6.71305001e-01
-1.00544012e+00 5.39314389e-01 6.85233545e+00 8.02018702e-01
-6.33088946e-01 -6.21942170e-02 7.22841561e-01 1.83098558e-02
-7.51977921e-01 3.35076332e-01 -1.03026021e+00 2.25888357e-01
9.75270927e-01 -7.13437051e-02 2.55260438e-01 1.04544258e+00
3.28348011e-01 -3.12227756e-01 -1.39043045e+00 5.68191767e-01
1.81235984e-01 -1.15152681e+00 1.73017263e-01 1.89552289e-02
4.77992147e-01 -6.03237927e-01 2.31824860e-01 1.11548088e-01
1.10581256e-01 -1.08569801e+00 7.67983258e-01 6.18442714e-01
2.47137710e-01 -8.15070152e-01 6.32976830e-01 5.69521487e-01
-4.32518333e-01 -3.04368287e-01 -6.45957291e-01 -3.22917074e-01
-3.28885883e-01 8.52499962e-01 -1.42212534e+00 6.03263617e-01
3.01153272e-01 5.72552860e-01 -6.53770566e-01 1.03294504e+00
-3.35439712e-01 8.07190239e-01 -2.00923875e-01 -2.34483644e-01
-5.96498065e-02 -2.76676595e-01 4.57998127e-01 9.50303257e-01
2.37433091e-01 9.80021991e-03 -2.69320309e-01 1.27566457e+00
2.55890131e-01 2.46916965e-01 -7.03383625e-01 2.33743981e-01
4.76902872e-01 1.05856574e+00 -8.21884453e-01 -2.62420505e-01
-1.37934864e-01 7.71836221e-01 6.52251661e-01 3.79771858e-01
-7.17070639e-01 -6.68916330e-02 3.60060900e-01 9.02683362e-02
3.70680362e-01 2.30741024e-01 -5.00008047e-01 -1.14065015e+00
2.79089004e-01 -7.51206398e-01 2.55788475e-01 -6.89914405e-01
-1.22827315e+00 9.01318714e-02 3.11844021e-01 -1.26296771e+00
-4.23153728e-01 -7.62531817e-01 -5.64587355e-01 8.66258800e-01
-1.40976846e+00 -8.88508618e-01 1.71529889e-01 2.06737611e-02
6.19544983e-01 6.41150773e-02 9.84959662e-01 -2.36826316e-01
-4.52959776e-01 4.45779473e-01 2.75246859e-01 -3.69592279e-01
6.45638287e-01 -1.61528265e+00 6.58224747e-02 7.42332697e-01
3.22908998e-01 1.06629384e+00 1.15806389e+00 -6.06320202e-01
-1.19473016e+00 -9.37551498e-01 9.26444352e-01 -8.05731773e-01
2.36752033e-01 -2.05594227e-01 -9.19365227e-01 7.11120903e-01
-1.60991773e-01 -5.06490879e-02 1.20533288e+00 4.78154331e-01
-1.53348893e-01 -1.00454599e-01 -1.29271853e+00 4.80318040e-01
7.99633265e-01 -2.30498865e-01 -9.33026612e-01 4.58972603e-01
5.05103052e-01 -7.67236054e-02 -7.33283341e-01 3.38378906e-01
4.76312667e-01 -1.34474075e+00 8.17754090e-01 -9.56513226e-01
5.95817983e-01 -3.35365862e-01 -1.59617618e-01 -1.22475851e+00
-1.76633820e-01 -3.18984121e-01 -3.46699022e-02 1.04377413e+00
6.02401316e-01 -5.15859187e-01 8.39511693e-01 9.04822171e-01
-2.79845923e-01 -1.07031870e+00 -8.33968639e-01 -3.08281898e-01
-6.15705736e-03 -6.50717318e-01 2.90682107e-01 7.98534870e-01
9.43896547e-02 2.69854218e-01 -2.46362135e-01 3.91917616e-01
8.57514679e-01 2.14859620e-01 6.32954240e-01 -1.36390746e+00
-6.00235641e-01 -1.20153487e-01 -2.62228042e-01 -7.61265695e-01
5.89823872e-02 -8.75395238e-01 -4.52709831e-02 -1.25429928e+00
3.04024577e-01 -2.71618783e-01 -1.73615873e-01 3.11633438e-01
-3.63732100e-01 -5.27449735e-02 -1.50806606e-02 9.59457383e-02
-5.84173977e-01 3.38505983e-01 9.15799320e-01 2.62540251e-01
-8.95201564e-02 3.51685047e-01 -1.13573539e+00 8.39729011e-01
6.98157489e-01 -4.92877573e-01 -6.86727047e-01 -2.16990396e-01
5.81828833e-01 -1.94150075e-01 5.49199581e-01 -6.26409471e-01
8.91780853e-02 -3.31639737e-01 5.74322164e-01 -3.37250620e-01
8.80721882e-02 -9.81232345e-01 1.81510463e-01 2.64274776e-01
-6.70124471e-01 -2.45539188e-01 5.02007417e-02 7.68448651e-01
-7.08883330e-02 -8.63888741e-01 4.68083024e-01 -8.48468840e-02
-3.77786726e-01 -3.35770994e-01 -2.36632768e-02 -1.33701665e-02
1.13029993e+00 -3.76021296e-01 -1.45500541e-01 -1.45445652e-02
-8.92292798e-01 1.37130812e-01 3.17024857e-01 -1.04630716e-01
8.16369236e-01 -1.15993488e+00 -5.42805433e-01 3.11602533e-01
1.37148453e-02 2.31884286e-01 1.06073767e-01 8.71062338e-01
-3.06719989e-01 4.21393752e-01 9.85036194e-02 -5.42604387e-01
-1.10086203e+00 5.26008844e-01 1.60694897e-01 -2.77483389e-02
-2.46218011e-01 7.65005171e-01 1.90252081e-01 -3.68187577e-01
1.13781363e-01 -5.52842438e-01 -3.07470649e-01 5.14170043e-02
5.79505980e-01 3.99742126e-01 -1.84646230e-02 -2.84129292e-01
-2.02581704e-01 3.65911007e-01 -2.45826006e-01 -8.83073434e-02
1.45956516e+00 -7.25667318e-03 8.09312332e-03 9.18097317e-01
9.37601328e-01 8.54406580e-02 -1.17617965e+00 5.15140891e-02
2.69855589e-01 -4.69864756e-01 -4.33777809e-01 -8.05427015e-01
-1.73773333e-01 7.90725052e-01 2.39851594e-01 4.94036674e-01
1.05957615e+00 1.37270078e-01 -9.35248956e-02 6.27732158e-01
4.99161296e-02 -1.01453233e+00 -9.63385925e-02 8.56420994e-02
7.57679999e-01 -1.45190251e+00 1.53563574e-01 -6.41074479e-01
-6.80876434e-01 1.33647072e+00 3.10577005e-01 -2.86841333e-01
4.10924345e-01 -3.66067648e-01 -3.60617191e-01 -2.13923454e-01
-9.28534925e-01 -1.40716480e-02 6.16550148e-01 4.59973007e-01
5.17350435e-01 3.18057165e-02 -4.45878744e-01 8.30769539e-01
-2.05527380e-01 -2.71325782e-02 4.63971287e-01 7.87309110e-01
-8.40089023e-01 -8.91797364e-01 -4.68584299e-01 7.91321516e-01
-3.19337726e-01 -1.88277557e-01 -4.87571120e-01 7.72396386e-01
2.95834929e-01 7.50514984e-01 -2.52890229e-01 -3.19053173e-01
3.03857893e-01 4.42873687e-01 5.05390346e-01 -9.40138817e-01
-3.44886154e-01 1.11380080e-02 -3.06377355e-02 -2.29632378e-01
-5.85748613e-01 -9.19566214e-01 -1.17267406e+00 1.59431651e-01
-5.53322375e-01 5.48313558e-01 5.16599357e-01 1.19512880e+00
2.14132488e-01 3.62522602e-01 4.21467543e-01 -4.62549359e-01
-1.07740915e+00 -8.46207380e-01 -5.56801617e-01 5.02362072e-01
3.56048107e-01 -6.43204570e-01 -6.33629560e-01 3.77241187e-02]
|
[8.792282104492188, 5.716201305389404]
|
bc029ef1-9bb4-4e61-a10b-4cba400c4a8a
|
water-filling-an-efficient-algorithm-for
|
1904.09763
| null |
https://arxiv.org/abs/1904.09763v2
|
https://arxiv.org/pdf/1904.09763v2.pdf
|
Water-Filling: An Efficient Algorithm for Digitized Document Shadow Removal
|
In this paper, we propose a novel algorithm to rectify illumination of the digitized documents by eliminating shading artifacts. Firstly, a topographic surface of an input digitized document is created using luminance value of each pixel. Then the shading artifact on the document is estimated by simulating an immersion process. The simulation of the immersion process is modeled using a novel diffusion equation with an iterative update rule. After estimating the shading artifacts, the digitized document is reconstructed using the Lambertian surface model. In order to evaluate the performance of the proposed algorithm, we conduct rigorous experiments on a set of digitized documents which is generated using smartphones under challenging lighting conditions. According to the experimental results, it is found that the proposed method produces promising illumination correction results and outperforms the results of the state-of-the-art methods.
|
['Changick Kim', 'Seungjun Jung', 'Muhammad Abul Hasan']
|
2019-04-22
| null | null | null | null |
['shadow-removal']
|
['computer-vision']
|
[ 7.20952749e-01 -4.68836099e-01 7.39669681e-01 -2.31581867e-01
-2.02234119e-01 -5.45920253e-01 6.35034800e-01 -3.20854455e-01
-1.79580107e-01 5.68028629e-01 3.70020643e-02 -4.42315713e-02
1.58602431e-01 -6.18447065e-01 -4.44110274e-01 -7.07330465e-01
5.10977685e-01 -9.05842483e-02 2.89967000e-01 1.09744906e-01
8.23273718e-01 4.87653166e-01 -1.68403125e+00 9.00301039e-02
1.30122888e+00 7.65406191e-01 4.97763038e-01 7.22679734e-01
-3.40214640e-01 2.41569474e-01 -9.29422021e-01 -2.90339649e-01
2.43821546e-01 -3.45846444e-01 -2.46381819e-01 5.73482633e-01
3.86827976e-01 -5.30589402e-01 9.10568908e-02 1.24571753e+00
3.73163104e-01 1.62759468e-01 7.93545723e-01 -5.52989364e-01
-8.27988684e-01 -4.63309914e-01 -9.98699903e-01 -6.31934032e-02
5.20208001e-01 -6.82467148e-02 2.27995306e-01 -1.05223143e+00
7.30702162e-01 1.13725233e+00 5.74717283e-01 1.83201045e-01
-1.06413639e+00 -3.40012789e-01 -1.39215887e-01 1.15023382e-01
-1.46471679e+00 -1.55472174e-01 1.08139575e+00 -7.36242309e-02
5.25517344e-01 3.54986638e-01 6.77256107e-01 3.66982728e-01
6.85420811e-01 5.30638158e-01 1.61059308e+00 -7.12219477e-01
3.66833657e-01 3.88162076e-01 3.26904446e-01 6.10025346e-01
2.85843998e-01 -4.00147855e-01 -2.13464066e-01 -2.66090602e-01
6.50277197e-01 -2.07681879e-02 -4.83153909e-01 -7.47931972e-02
-5.64258754e-01 1.49309784e-01 1.00331493e-01 2.38954395e-01
-4.46544677e-01 -1.16562292e-01 -2.49374751e-02 -1.64972633e-01
5.65885544e-01 -2.30668843e-01 2.49503642e-01 6.67750761e-02
-1.11806381e+00 -9.78689119e-02 7.13405669e-01 6.84304655e-01
5.46810806e-01 7.59848282e-02 2.37761199e-01 8.38700950e-01
6.54072523e-01 7.68058479e-01 1.09087288e-01 -8.76092374e-01
1.36921853e-01 5.89202821e-01 6.15871847e-01 -1.07256413e+00
1.04534484e-01 -5.79523705e-02 -6.30521417e-01 6.45947039e-01
2.51280069e-01 -5.01499698e-02 -8.95843863e-01 8.13426435e-01
5.44058263e-01 3.52943182e-01 2.12161347e-01 8.33985150e-01
5.41044235e-01 1.13364494e+00 -4.80178356e-01 -6.56879306e-01
1.10840416e+00 -6.97907269e-01 -1.35491824e+00 3.03202391e-01
-3.87379825e-02 -1.19012201e+00 1.32862902e+00 9.76277232e-01
-1.18657219e+00 -5.05977154e-01 -1.23835373e+00 -1.31677032e-01
-3.20356153e-02 3.61543268e-01 1.93528801e-01 9.94003773e-01
-1.11602116e+00 2.34973669e-01 -5.34474909e-01 -4.17019010e-01
1.06969818e-01 3.81332040e-02 1.81071058e-01 -2.61307091e-01
-6.40757680e-01 7.77559698e-01 -4.17845160e-01 3.54858398e-01
-6.17635489e-01 -4.09492791e-01 -2.29328990e-01 -7.39755332e-02
-1.36628270e-01 -2.85715878e-01 8.90856624e-01 -1.18480933e+00
-2.12549067e+00 7.13931203e-01 -6.49590135e-01 1.71373844e-01
6.84868932e-01 -2.52656490e-01 -4.51247126e-01 1.56165928e-01
-2.71221757e-01 -2.06896931e-01 8.88265073e-01 -1.93416297e+00
-3.42314005e-01 -4.20191139e-01 -2.80318052e-01 4.98385787e-01
-2.67125964e-01 -7.18371943e-02 -6.45331264e-01 -2.66754538e-01
1.50953382e-01 -6.77968442e-01 1.60437942e-01 2.43115023e-01
-3.16833049e-01 2.51120478e-01 1.15890431e+00 -8.20129275e-01
1.15366018e+00 -2.44834709e+00 -3.00249636e-01 5.55981338e-01
-3.45318645e-01 2.39520863e-01 9.73283723e-02 4.84597772e-01
4.95059758e-01 -2.54851252e-01 -5.22051990e-01 -5.13500988e-01
-3.45663935e-01 4.85889763e-02 -3.97759110e-01 6.81396782e-01
-4.91378963e-01 1.48744196e-01 -4.45174038e-01 -3.66774291e-01
2.74665952e-01 1.07491934e+00 1.87349468e-02 1.22973874e-01
4.43308130e-02 3.24430346e-01 -3.04994017e-01 6.43521786e-01
1.39741397e+00 2.94751704e-01 8.66971463e-02 -1.99161127e-01
-5.81472576e-01 -2.71285355e-01 -1.53523242e+00 1.29005396e+00
-4.61237937e-01 6.29180729e-01 2.53993720e-01 -1.05024584e-01
1.36811650e+00 8.98390710e-02 -9.29753669e-03 -6.45317614e-01
1.36795446e-01 3.99028331e-01 -4.75694776e-01 -5.31704485e-01
7.24073708e-01 3.06178238e-02 7.42107213e-01 4.49507713e-01
-8.37548316e-01 -5.42606533e-01 -4.60217707e-02 1.17335171e-01
7.72497535e-01 1.86622828e-01 -2.25755781e-01 -4.59610373e-01
9.30852115e-01 -5.82022816e-02 1.14856035e-01 5.01332879e-01
8.30580071e-02 6.01453543e-01 -2.53380891e-02 -1.99733585e-01
-9.89568472e-01 -1.16623926e+00 -3.22959542e-01 2.37110168e-01
6.80239499e-01 1.64402500e-01 -1.32269931e+00 -3.63686075e-03
-1.73949942e-01 9.49233830e-01 -2.68039763e-01 3.60002160e-01
-2.89857149e-01 -8.45066071e-01 1.43421263e-01 -3.73071805e-02
9.68564630e-01 -8.08401346e-01 -5.83279312e-01 7.79392794e-02
8.89010727e-02 -8.00446808e-01 -3.32355678e-01 -4.83348578e-01
-1.06450093e+00 -9.86257315e-01 -8.78580868e-01 -9.50737357e-01
1.14259863e+00 6.85102403e-01 5.51637352e-01 3.21683139e-01
-4.68418717e-01 5.88688612e-01 -1.66638717e-01 -3.99780661e-01
-9.25965533e-02 -5.89913011e-01 -2.48041570e-01 3.51593345e-01
2.71128923e-01 -3.28117520e-01 -9.64924335e-01 2.34674484e-01
-1.05759788e+00 6.15809076e-02 2.35198170e-01 3.00154775e-01
7.57384062e-01 6.16412580e-01 1.04083128e-01 -7.85238504e-01
1.07024610e+00 7.80197419e-03 -1.00058615e+00 2.14645579e-01
-5.52546263e-01 -3.89194936e-01 5.32884955e-01 -3.77041489e-01
-1.93790948e+00 2.78881378e-02 2.60518789e-01 2.42793351e-01
-6.21917509e-02 7.13570863e-02 -1.28285736e-01 -2.74030089e-01
2.71025062e-01 4.98536825e-01 -9.03434306e-02 -5.66801965e-01
1.34383440e-01 1.04789340e+00 7.31123388e-01 -2.10638657e-01
6.87310100e-01 1.12063348e+00 4.17988338e-02 -1.33850908e+00
-2.03498840e-01 -1.59458071e-01 -4.02351528e-01 -6.91867411e-01
8.23611081e-01 -3.79615486e-01 -6.86008990e-01 9.84667003e-01
-1.32016253e+00 -1.97505459e-01 1.08372778e-01 2.94295698e-01
-1.11005366e-01 6.62726164e-01 -4.89677668e-01 -1.38548291e+00
-4.49879915e-01 -1.03926003e+00 9.74782646e-01 6.27652466e-01
1.39236122e-01 -9.75223899e-01 3.06153774e-01 3.76950145e-01
4.07196701e-01 -8.92068539e-03 8.11185360e-01 4.04629618e-01
-7.30408788e-01 -1.61051363e-01 -2.80132473e-01 3.21897864e-01
4.41672087e-01 6.76243842e-01 -1.16311491e+00 -1.06074385e-01
5.16415238e-01 3.84448946e-01 4.68828380e-01 4.42509055e-01
9.26295877e-01 2.01792896e-01 -1.58238187e-01 5.89844286e-01
1.90265501e+00 5.20931184e-01 1.08899415e+00 5.40919662e-01
3.05524677e-01 4.02135909e-01 6.47750974e-01 5.65499306e-01
3.45552415e-01 2.70837873e-01 3.78742158e-01 -3.07916820e-01
-2.14068592e-01 1.03195190e-01 2.74108589e-01 8.19600165e-01
-1.14828415e-01 -6.74519539e-01 -6.37821734e-01 3.07662874e-01
-1.26476383e+00 -6.12082243e-01 -9.54249620e-01 2.18398046e+00
6.06436431e-01 -6.55760765e-02 -6.09152734e-01 3.63743514e-01
8.44434202e-01 5.25553226e-02 -2.77003318e-01 -8.86375487e-01
-1.99033067e-01 9.83009115e-02 3.26730579e-01 8.80332112e-01
-4.84078586e-01 7.11670935e-01 7.21131802e+00 2.84582615e-01
-1.20470047e+00 -6.38954565e-02 3.66858780e-01 1.76589131e-01
-6.52162254e-01 -7.94226974e-02 -6.43057048e-01 6.58719420e-01
4.35607821e-01 -4.64921482e-02 5.57105184e-01 3.47913980e-01
7.26882339e-01 -9.17115986e-01 -3.77630264e-01 8.07606578e-01
4.00650442e-01 -5.99488616e-01 5.94870038e-02 -1.52179912e-01
1.10296547e+00 -4.47945535e-01 2.54739761e-01 -6.36005878e-01
-3.55898254e-02 -5.50466418e-01 3.30763638e-01 9.93264914e-01
6.21289670e-01 -5.82702696e-01 6.87101305e-01 2.57621378e-01
-8.70015800e-01 2.78265953e-01 -4.00497049e-01 -1.61822721e-01
3.96488816e-01 9.25637960e-01 -4.78614479e-01 1.75643668e-01
5.24221361e-01 2.45169029e-01 -4.35143560e-01 1.24899685e+00
-3.16767931e-01 5.56486607e-01 -4.01198834e-01 -1.50718004e-01
3.14960144e-02 -9.33265150e-01 3.65378559e-01 1.17494953e+00
6.38754547e-01 3.44242066e-01 -5.43646574e-01 8.47702324e-01
2.42926180e-02 3.65686029e-01 -5.46920776e-01 4.74016935e-01
2.73879826e-01 1.15246844e+00 -7.73480058e-01 -3.85717660e-01
-4.91157591e-01 1.47715652e+00 -3.66202354e-01 8.01430106e-01
-7.00278938e-01 -5.34607410e-01 -6.15143329e-02 1.57028392e-01
2.63958308e-03 -2.76837885e-01 -6.63012981e-01 -7.46691108e-01
2.68219084e-01 -4.76491243e-01 -3.47251534e-01 -1.52037680e+00
-7.74331927e-01 3.52391869e-01 -3.73116046e-01 -9.74682748e-01
6.16365016e-01 -4.34929788e-01 -9.44660902e-01 1.10615146e+00
-1.58530796e+00 -6.59765661e-01 -7.37299502e-01 4.16821390e-01
6.60235941e-01 1.04471952e-01 6.04373634e-01 1.49813399e-01
-3.04737687e-01 -1.04906708e-01 9.23471272e-01 -7.84336567e-01
6.57921374e-01 -1.00954568e+00 1.12117410e-01 1.03881574e+00
-4.21946704e-01 5.05364478e-01 8.31846654e-01 -9.58160698e-01
-1.57539034e+00 -6.18489921e-01 6.53808415e-01 -1.33852780e-01
1.91575035e-01 -2.08326131e-01 -1.16909719e+00 1.95573032e-01
5.96875429e-01 -3.50859165e-01 3.62863392e-01 -7.61988938e-01
1.99022278e-01 -1.90178543e-01 -1.56907260e+00 6.52374864e-01
5.01930356e-01 -3.55602324e-01 -4.50619608e-01 1.48115367e-01
-4.02511619e-02 -3.28085870e-01 -5.32044947e-01 -1.79191295e-03
8.10687900e-01 -1.03752947e+00 5.82370758e-01 6.89061582e-01
1.71213523e-01 -7.03380346e-01 -1.00174829e-01 -1.19154751e+00
-1.04806058e-01 -5.03126919e-01 2.12022245e-01 1.39661610e+00
2.11362869e-01 -7.02099800e-01 6.46152377e-01 7.86117792e-01
1.36268243e-01 -3.24897557e-01 -4.82718706e-01 -3.20522130e-01
-5.44392228e-01 2.15339698e-02 3.54744196e-01 5.18111050e-01
-4.19791758e-01 -1.35700703e-01 -2.90456533e-01 4.21725154e-01
1.02238762e+00 -5.40174805e-02 7.20139325e-01 -1.11989546e+00
1.69005290e-01 1.76232308e-01 -7.48870382e-03 -1.10584784e+00
-1.56796277e-01 -2.60812700e-01 8.97804275e-02 -1.99766612e+00
3.10971886e-01 -1.91442817e-02 2.55452748e-02 -3.99517983e-01
-2.50610858e-01 3.28470409e-01 -2.86633545e-03 2.41345495e-01
-1.22455239e-01 3.82166147e-01 1.38996172e+00 -3.30555774e-02
-3.84550065e-01 -1.46255746e-01 -2.68059075e-01 8.84481668e-01
7.07637548e-01 -1.69162706e-01 -6.62230968e-01 -8.10524523e-01
-3.25268395e-02 -2.93332368e-01 1.54897511e-01 -8.45131457e-01
1.00711182e-01 3.95124749e-04 5.08989096e-01 -9.48892355e-01
4.24444020e-01 -1.12593317e+00 3.51607054e-01 3.81492734e-01
1.75708532e-02 8.10194295e-03 1.47576006e-02 5.86006641e-01
2.00653188e-02 -5.56445420e-01 8.88884187e-01 1.46740913e-01
-3.38759899e-01 -4.46739197e-01 -6.22285247e-01 -4.83374119e-01
9.84297276e-01 -7.48112261e-01 -3.59304398e-01 -3.43015403e-01
-1.10398650e-01 -1.83648348e-01 9.77182806e-01 -9.77570117e-02
9.54689920e-01 -1.01029027e+00 -3.29113752e-01 4.25137550e-01
-5.05369127e-01 -2.39607722e-01 2.37980977e-01 6.30227685e-01
-1.15111017e+00 -1.03556760e-01 1.02187907e-02 -4.69290972e-01
-1.61070597e+00 9.72720012e-02 9.37842205e-02 2.05192998e-01
-5.98544598e-01 4.29188758e-01 9.04255062e-02 1.96605995e-02
1.82942793e-01 -1.82449430e-01 -1.21649608e-01 -4.28645551e-01
6.05560899e-01 8.80264759e-01 2.76838173e-03 -5.22506356e-01
-2.62021631e-01 1.22572148e+00 7.99209997e-02 -5.23760557e-01
1.27029288e+00 -5.23945332e-01 -2.08034351e-01 3.76987159e-01
9.10606503e-01 4.71423686e-01 -1.32655966e+00 1.71335757e-01
-4.01189923e-01 -9.57075238e-01 3.02964985e-01 -8.54773879e-01
-8.15456390e-01 7.86338806e-01 1.02177227e+00 1.02913886e-01
1.37311053e+00 -8.50786388e-01 7.10624397e-01 2.16341451e-01
2.76339501e-01 -1.44208658e+00 -1.05400227e-01 2.24240601e-01
7.60160208e-01 -8.20670962e-01 3.46308291e-01 -5.48903167e-01
-6.30227923e-01 1.18386555e+00 2.92897463e-01 -3.35146546e-01
4.93119419e-01 5.99223614e-01 3.07171881e-01 -8.54411870e-02
-3.70582759e-01 3.41449112e-01 -5.16570918e-02 5.67564130e-01
4.70976114e-01 -3.87257874e-01 -7.79128969e-01 9.92752314e-02
1.90424621e-01 3.81307870e-01 9.57167923e-01 9.65045214e-01
-7.10211337e-01 -6.96507990e-01 -9.34754431e-01 9.95534211e-02
-4.36179072e-01 -1.66514609e-02 -5.33398151e-01 4.82085317e-01
-9.15470123e-02 1.10032523e+00 7.97286183e-02 2.06269354e-01
4.74516958e-01 -1.38349950e-01 4.56383407e-01 -1.00649215e-01
-3.18975687e-01 3.33358616e-01 -1.33979872e-01 -2.07919866e-01
-5.49077928e-01 -5.64433336e-01 -1.50063133e+00 -1.71980843e-01
-5.16405702e-01 2.06110179e-02 1.23478353e+00 4.89511251e-01
1.51661873e-01 4.49388504e-01 8.64235461e-01 -7.82082796e-01
-8.55108872e-02 -6.69798374e-01 -8.49680543e-01 4.06863153e-01
1.13156103e-01 -4.21240956e-01 -7.44390845e-01 3.89923662e-01]
|
[10.646778106689453, -2.8209621906280518]
|
f350c785-4da8-43d0-9e72-36163f738d6d
|
egocentric-deep-multi-channel-audio-visual
|
2201.01928
| null |
https://arxiv.org/abs/2201.01928v1
|
https://arxiv.org/pdf/2201.01928v1.pdf
|
Egocentric Deep Multi-Channel Audio-Visual Active Speaker Localization
|
Augmented reality devices have the potential to enhance human perception and enable other assistive functionalities in complex conversational environments. Effectively capturing the audio-visual context necessary for understanding these social interactions first requires detecting and localizing the voice activities of the device wearer and the surrounding people. These tasks are challenging due to their egocentric nature: the wearer's head motion may cause motion blur, surrounding people may appear in difficult viewing angles, and there may be occlusions, visual clutter, audio noise, and bad lighting. Under these conditions, previous state-of-the-art active speaker detection methods do not give satisfactory results. Instead, we tackle the problem from a new setting using both video and multi-channel microphone array audio. We propose a novel end-to-end deep learning approach that is able to give robust voice activity detection and localization results. In contrast to previous methods, our method localizes active speakers from all possible directions on the sphere, even outside the camera's field of view, while simultaneously detecting the device wearer's own voice activity. Our experiments show that the proposed method gives superior results, can run in real time, and is robust against noise and clutter.
|
['Vamsi Krishna Ithapu', 'Calvin Murdock', 'Hao Jiang']
|
2022-01-06
| null |
http://openaccess.thecvf.com//content/CVPR2022/html/Jiang_Egocentric_Deep_Multi-Channel_Audio-Visual_Active_Speaker_Localization_CVPR_2022_paper.html
|
http://openaccess.thecvf.com//content/CVPR2022/papers/Jiang_Egocentric_Deep_Multi-Channel_Audio-Visual_Active_Speaker_Localization_CVPR_2022_paper.pdf
|
cvpr-2022-1
|
['active-speaker-localization', 'audio-visual-active-speaker-detection']
|
['audio', 'computer-vision']
|
[ 9.47398972e-03 -2.14256257e-01 2.54747957e-01 4.99928594e-02
-8.68249297e-01 -4.52205747e-01 1.93660796e-01 -3.77262652e-01
-1.86538965e-01 5.01338065e-01 4.95898813e-01 1.20759688e-01
2.73239881e-01 -3.73255834e-02 -3.64945054e-01 -7.22571671e-01
4.16587889e-02 -1.04802333e-01 3.43274027e-01 1.27320573e-01
-1.55956164e-01 3.62901479e-01 -2.01588297e+00 2.11967304e-01
3.86655241e-01 9.75338638e-01 3.48854363e-01 9.95024264e-01
2.67289400e-01 6.71827614e-01 -1.11478543e+00 1.06874511e-01
-2.37443402e-01 -5.82293719e-02 -2.66969725e-02 4.50227022e-01
6.40262961e-01 -5.63948452e-01 -5.06515503e-01 9.00269330e-01
1.07111096e+00 2.37131029e-01 -8.90047699e-02 -1.23002732e+00
-2.02000782e-01 -2.64702946e-01 -5.30313671e-01 2.98744231e-01
1.21578085e+00 1.54997200e-01 4.48662221e-01 -1.01361823e+00
2.72270799e-01 1.32355130e+00 6.76140726e-01 7.04074144e-01
-7.82442451e-01 -4.92953867e-01 2.82517791e-01 2.49157310e-01
-1.38900793e+00 -1.11528397e+00 9.72643018e-01 -3.12238365e-01
6.62982106e-01 6.77998543e-01 7.92220473e-01 1.58780253e+00
-5.76945171e-02 8.63920867e-01 6.86753631e-01 -3.26690435e-01
4.07945812e-01 2.88482189e-01 -1.33671924e-01 3.48179251e-01
-1.04721583e-01 -1.38142094e-01 -9.48068976e-01 -4.14179176e-01
5.46411097e-01 2.26318642e-01 -8.53958726e-01 -4.64824885e-01
-1.41108108e+00 2.59816706e-01 1.39542416e-01 4.54024583e-01
-4.40910786e-01 -4.89575006e-02 2.47333512e-01 -5.68372943e-02
6.06944203e-01 9.61965919e-02 -1.30803749e-01 -5.44112086e-01
-6.44111574e-01 -6.12082146e-02 8.26751232e-01 7.16233373e-01
-8.08309950e-03 -2.31844395e-01 4.92703468e-02 1.00963080e+00
5.70395172e-01 6.44464254e-01 2.35730946e-01 -8.72852385e-01
5.01537681e-01 2.07830563e-01 4.42507595e-01 -8.94865036e-01
-4.34314013e-01 -4.42780733e-01 -3.72680575e-01 2.16533735e-01
2.62712777e-01 -5.29185653e-01 -2.37200215e-01 1.57000434e+00
7.98801422e-01 5.08841872e-01 -3.14923704e-01 1.17259037e+00
8.26631129e-01 4.60628837e-01 -3.70956242e-01 -8.02176237e-01
1.49659705e+00 -8.51232529e-01 -1.40235364e+00 -4.77343619e-01
1.59083411e-01 -8.23901057e-01 1.26212382e+00 5.53852499e-01
-1.06371105e+00 -6.33810818e-01 -1.12263775e+00 1.71375424e-01
3.30837034e-02 2.97908723e-01 3.95091265e-01 8.37710619e-01
-1.12409341e+00 -2.86516130e-01 -9.92648065e-01 -4.73028272e-01
4.90960255e-02 3.35482210e-01 -3.05403918e-01 -7.83645958e-02
-7.54595637e-01 4.50199038e-01 -5.94722629e-01 6.09401524e-01
-6.22361898e-01 -2.60048896e-01 -7.85387456e-01 -1.24994509e-01
6.31716251e-01 -6.74468994e-01 1.58796608e+00 -9.00615871e-01
-1.74360263e+00 4.91041034e-01 -6.69616520e-01 8.05558264e-02
6.49606645e-01 -7.32027531e-01 -7.62623131e-01 9.45982561e-02
-3.67641896e-02 8.88533220e-02 8.61731589e-01 -1.16457701e+00
-4.62641537e-01 -7.63433695e-01 3.75006115e-03 5.68751276e-01
-5.32498717e-01 3.60297441e-01 -7.49989390e-01 -3.22819114e-01
1.04899622e-01 -9.02817905e-01 2.19601586e-01 3.59498024e-01
-2.74536014e-01 -2.48439405e-02 1.16345668e+00 -5.99092662e-01
1.10468602e+00 -2.44923639e+00 -1.67293578e-01 -2.88334012e-01
3.66604000e-01 3.88122439e-01 3.27678531e-01 8.23538378e-02
2.88431972e-01 -5.72860301e-01 2.44537920e-01 -9.09611046e-01
-2.09671780e-01 -1.58718511e-01 4.20233309e-02 8.46269429e-01
-5.64292848e-01 4.20023382e-01 -8.54576886e-01 -2.62346983e-01
4.57883894e-01 1.07103240e+00 -3.73081118e-01 4.52679932e-01
4.48343426e-01 7.80008972e-01 -1.09886989e-01 7.17162848e-01
6.28162980e-01 2.05958914e-02 3.18214148e-02 1.54962823e-01
-3.45225260e-02 3.74430478e-01 -1.52294040e+00 1.56075609e+00
-8.47558141e-01 1.14441049e+00 9.02726650e-01 -3.28234792e-01
5.68452954e-01 7.67907441e-01 2.86418676e-01 -5.54313838e-01
9.69999377e-03 4.21088487e-02 -2.89196670e-01 -1.01084971e+00
4.14805889e-01 2.30380952e-01 8.45484510e-02 2.50396430e-01
-3.58570278e-01 5.81398726e-01 -5.88517725e-01 -1.41035795e-01
1.23814428e+00 -4.16126996e-01 2.27090001e-01 3.18787754e-01
5.26360631e-01 -9.78696585e-01 4.29976463e-01 5.44524491e-01
-5.03149450e-01 6.13457680e-01 3.84477563e-02 -1.69297177e-02
-2.92215377e-01 -1.24244177e+00 2.44471788e-01 1.30763555e+00
2.65604854e-01 -4.57005471e-01 -7.45168924e-01 -5.22158623e-01
-3.99926662e-01 3.21271718e-01 -3.60483408e-01 3.55767049e-02
-5.76768875e-01 -3.58157367e-01 2.31975377e-01 5.57633221e-01
5.30143499e-01 -8.26552629e-01 -6.38218641e-01 1.46699712e-01
-6.46484971e-01 -1.36868632e+00 -7.88916409e-01 -5.82247794e-01
-3.60862076e-01 -9.49563682e-01 -1.00520039e+00 -6.79737210e-01
5.45783460e-01 9.48452592e-01 8.00484836e-01 -3.34123552e-01
-3.23419482e-01 9.76693094e-01 8.79632309e-02 -4.50740695e-01
-3.03154029e-02 -5.66475511e-01 3.77507955e-01 4.45284247e-01
2.92364359e-01 -3.90524179e-01 -9.68359590e-01 5.77049017e-01
-1.32193908e-01 -4.46998954e-01 -9.13900957e-02 4.69625443e-01
1.00636013e-01 1.63625777e-01 3.69054377e-01 -3.02656561e-01
7.20016062e-01 -2.20409974e-01 -2.00574055e-01 -4.18975949e-02
1.92469165e-01 -9.25592840e-01 1.87399179e-01 -7.68127084e-01
-1.24189878e+00 2.91832268e-01 -2.14270443e-01 -4.43863869e-01
-5.54660738e-01 -1.86348960e-01 -7.89059758e-01 1.87378734e-01
6.37530208e-01 -1.55532241e-01 -2.81963614e-03 -6.17798269e-01
-2.98933368e-02 1.42497432e+00 6.31924093e-01 2.85261542e-01
2.30325177e-01 7.99201071e-01 -5.02132475e-01 -1.56380355e+00
-4.02903914e-01 -9.87659276e-01 -3.97888035e-01 -7.32350707e-01
6.73567235e-01 -1.26954913e+00 -1.00663900e+00 7.35120237e-01
-1.34335446e+00 7.02503771e-02 2.11199105e-01 7.75187135e-01
-1.82387426e-01 5.55250287e-01 -2.54316658e-01 -1.58774674e+00
-1.11814842e-01 -1.30445707e+00 1.58307242e+00 2.04829752e-01
-4.52801943e-01 -8.26172233e-01 7.53519824e-03 7.35727370e-01
3.36782008e-01 5.56579195e-02 -1.50092915e-01 -5.80875129e-02
-3.24108213e-01 -4.96735483e-01 3.91945243e-01 1.66272014e-01
5.28013825e-01 -3.71395320e-01 -1.63461137e+00 -4.24687177e-01
4.95410711e-01 1.68314829e-01 2.77974576e-01 6.68589592e-01
8.51025462e-01 -5.01685858e-01 -5.14466643e-01 3.01367640e-01
5.41807830e-01 3.27729106e-01 5.46987593e-01 -2.11598739e-01
8.83577943e-01 7.26582646e-01 6.28492355e-01 4.91965890e-01
2.66571790e-01 1.28675270e+00 4.77330476e-01 -4.17534471e-01
-1.74903199e-01 1.91743914e-02 7.95389712e-01 4.93023634e-01
1.53121248e-01 -5.28398871e-01 -5.62642336e-01 4.22043234e-01
-1.76812625e+00 -8.16185653e-01 -1.41643360e-01 2.52958059e+00
1.59985423e-01 -3.10410745e-02 4.44261819e-01 4.43359315e-01
9.66266215e-01 1.37388498e-01 -5.26594937e-01 4.79063317e-02
1.18514404e-01 -2.94907957e-01 -4.60832529e-02 6.41453385e-01
-1.08580863e+00 3.46451163e-01 6.21095562e+00 2.67265350e-01
-1.33205223e+00 4.06856298e-01 -1.44271003e-02 -7.99560547e-01
1.97953314e-01 -7.09252656e-01 -5.68425357e-01 5.60344696e-01
6.95079327e-01 6.95159137e-01 3.33549291e-01 9.64725196e-01
5.92697918e-01 -2.66188562e-01 -1.09753871e+00 1.74583006e+00
5.67751229e-01 -6.69360340e-01 -1.02058434e+00 3.78675282e-01
1.22326717e-01 -1.14478134e-01 6.80155084e-02 -1.11686617e-01
-4.38422352e-01 -6.77389979e-01 6.29799724e-01 3.76752019e-01
6.10873640e-01 -5.19974351e-01 4.88223910e-01 4.75724906e-01
-1.19807732e+00 -2.98535973e-01 7.71618038e-02 -2.93059409e-01
4.47661877e-01 5.82877755e-01 -1.00039351e+00 -3.39604586e-01
9.03302073e-01 3.34686309e-01 -3.76345932e-01 1.16394854e+00
-2.62895823e-01 4.25510794e-01 -3.73249352e-01 -1.61135003e-01
-5.40486932e-01 3.36761415e-01 1.14098334e+00 1.00218904e+00
4.37120497e-01 -1.11167543e-01 1.44743863e-02 3.92437875e-01
2.70100515e-02 -3.87123637e-02 -9.32161510e-01 3.08772177e-01
6.28664970e-01 9.91543472e-01 -5.09433627e-01 -7.53789581e-03
-5.06992996e-01 1.25796413e+00 -2.72825956e-01 5.65634906e-01
-7.96497643e-01 -3.45734686e-01 9.26985502e-01 3.54636610e-01
1.11818776e-01 -5.39527595e-01 8.66393000e-02 -1.22043896e+00
6.70215964e-01 -8.64842832e-01 1.19629495e-01 -1.01126826e+00
-6.49418771e-01 4.95321423e-01 -3.46962661e-01 -1.48460770e+00
-3.14576894e-01 -4.96096939e-01 -6.79418921e-01 5.24585903e-01
-7.60343492e-01 -7.91121423e-01 -5.64620256e-01 7.74697304e-01
8.76369953e-01 -2.09779814e-01 8.20339262e-01 6.92212284e-01
-6.36520207e-01 8.55526268e-01 1.59766167e-01 -1.43386215e-01
7.94318080e-01 -1.01508522e+00 2.45474383e-01 8.65252852e-01
2.48542175e-01 6.28243446e-01 9.27389383e-01 -1.69094697e-01
-1.61113310e+00 -5.00451684e-01 8.79576266e-01 -6.89970672e-01
2.55456448e-01 -1.05130291e+00 -6.05679095e-01 4.18882757e-01
8.86423700e-03 2.38332838e-01 9.11967039e-01 4.00303751e-01
4.10835966e-02 -2.86954314e-01 -1.00318003e+00 5.27713597e-01
1.16474271e+00 -8.06029797e-01 -4.61408764e-01 3.20232719e-01
3.61765891e-01 -5.68000555e-01 -2.54665732e-01 -4.99651162e-03
9.04036641e-01 -1.12186348e+00 1.02245271e+00 1.80183411e-01
-5.17036974e-01 -3.21584612e-01 -4.14397642e-02 -1.12840104e+00
-1.52663207e-02 -1.05932629e+00 -5.55824757e-01 1.26720631e+00
3.11600622e-02 -7.11069822e-01 5.10187089e-01 4.23291206e-01
-8.19501281e-02 -3.75034630e-01 -1.40710676e+00 -4.22372341e-01
-1.11179388e+00 -7.54221439e-01 4.25877362e-01 6.47699714e-01
2.74267435e-01 6.17693126e-01 -6.60917878e-01 4.40932602e-01
4.23311740e-01 -3.81136626e-01 1.06768870e+00 -1.29628646e+00
-4.42643106e-01 4.74611409e-02 -5.21410227e-01 -1.38288212e+00
-6.72214776e-02 1.77530915e-01 2.55772024e-01 -1.46454084e+00
-2.83715993e-01 3.37284915e-02 5.47065735e-02 1.32642267e-02
-1.53292473e-02 3.54651093e-01 -6.54479116e-02 1.19222857e-01
-7.99376607e-01 5.45151234e-01 9.67170298e-01 -6.48517460e-02
-6.01973116e-01 6.68037891e-01 -6.26091897e-01 9.96766508e-01
2.61816978e-01 2.04647109e-02 -5.74040949e-01 -3.83388102e-01
2.01996416e-01 2.57108331e-01 6.52102768e-01 -1.13491893e+00
4.81819838e-01 2.76991457e-01 3.09256554e-01 -5.68181038e-01
1.17033672e+00 -9.20130312e-01 1.00760795e-01 1.62223339e-01
1.79995981e-03 -1.22806840e-01 1.20694198e-01 7.93254554e-01
5.13039418e-02 3.17099631e-01 5.21138310e-01 -1.40832830e-02
-2.91718543e-01 -1.18578993e-01 -7.21419573e-01 -3.67645353e-01
7.69702375e-01 -3.82396728e-01 -2.75887966e-01 -8.15246105e-01
-9.79505777e-01 -2.50522997e-02 2.89144427e-01 8.13021541e-01
5.99201500e-01 -1.26924038e+00 -2.12454543e-01 3.78010243e-01
3.90901454e-02 -1.05639309e-01 5.56705415e-01 9.70200181e-01
-9.42956582e-02 4.39019203e-01 3.29393923e-01 -1.04630113e+00
-1.91981196e+00 3.61744970e-01 5.37309468e-01 4.71113414e-01
-7.33613193e-01 9.56753492e-01 4.77322489e-01 -1.18573658e-01
9.27317500e-01 -2.38481075e-01 -3.25694859e-01 1.82514265e-01
1.11205268e+00 7.69834757e-01 2.39022821e-01 -9.72444952e-01
-5.62572122e-01 5.42156756e-01 1.40737846e-01 -3.80856156e-01
8.54945242e-01 -8.42620194e-01 3.83191645e-01 6.46129668e-01
1.23445833e+00 6.59984171e-01 -1.26711643e+00 -3.12271446e-01
-5.99091053e-01 -7.89896965e-01 3.15919280e-01 -4.84140813e-01
-8.98257673e-01 1.14683449e+00 1.20149398e+00 2.74597555e-01
1.14175630e+00 1.93561539e-01 6.80015743e-01 2.38215715e-01
3.86158049e-01 -9.65826154e-01 3.11251014e-01 -1.14296284e-03
9.65493500e-01 -1.12384307e+00 -2.96539307e-01 -6.74204528e-01
-5.73413908e-01 7.39091218e-01 5.61558008e-01 6.09963834e-01
6.07836723e-01 2.58301646e-01 4.27797377e-01 -1.42288968e-01
-5.51262021e-01 -2.22889170e-01 2.21708879e-01 9.72908914e-01
3.09940010e-01 -8.93047675e-02 4.30843621e-01 4.87972021e-01
-2.48248532e-01 -2.66480446e-01 2.74621636e-01 9.72865820e-01
-2.61296779e-01 -3.22414547e-01 -9.26722467e-01 -1.39417499e-01
-5.77006757e-01 5.73214054e-01 -5.93300462e-01 3.73990715e-01
1.49466977e-01 1.60201836e+00 2.31680498e-01 -3.32704604e-01
6.90181732e-01 3.76053900e-02 4.34361637e-01 -4.63399112e-01
-3.28671306e-01 7.30491400e-01 1.30858839e-01 -6.44684076e-01
-4.12728578e-01 -8.92066300e-01 -7.86424339e-01 5.28073385e-02
-5.26458740e-01 -1.58505458e-02 8.36090982e-01 9.15983498e-01
5.84349096e-01 7.52458274e-01 8.63167286e-01 -1.20276713e+00
-5.40786609e-02 -1.08593774e+00 -6.07565582e-01 2.52443701e-02
9.81123030e-01 -8.40921104e-01 -6.16800010e-01 -1.74914345e-01]
|
[14.536835670471191, 5.204346179962158]
|
19b75955-8c43-4422-a2e0-a297974e7ad4
|
analysis-of-face-detection-face-landmarking
|
2207.06478
| null |
https://arxiv.org/abs/2207.06478v1
|
https://arxiv.org/pdf/2207.06478v1.pdf
|
Analysis of face detection, face landmarking, and face recognition performance with masked face images
|
Face recognition has become an essential task in our lives. However, the current COVID-19 pandemic has led to the widespread use of face masks. The effect of wearing face masks is currently an understudied issue. The aim of this paper is to analyze face detection, face landmarking, and face recognition performance with masked face images. HOG and CNN face detectors are used for face detection in combination with 5-point and 68-point face landmark predictors and VGG16 face recognition model is used for face recognition on masked and unmasked images. We found that the performance of face detection, face landmarking, and face recognition is negatively impacted by face masks
|
['Ožbej Golob']
|
2022-06-03
| null | null | null | null |
['face-detection']
|
['computer-vision']
|
[-4.93511604e-03 -1.54337779e-01 6.60222545e-02 -5.61984420e-01
-2.31193915e-01 -5.76925874e-01 3.11096996e-01 -4.01204407e-01
-4.12157416e-01 2.82505095e-01 -2.72385199e-02 -4.69178110e-02
3.00896615e-01 -4.73016024e-01 -4.25304472e-01 -6.27624035e-01
-1.94174424e-01 6.01998158e-02 -2.27948785e-01 1.04758069e-01
1.98262393e-01 1.30668318e+00 -1.65555668e+00 3.42957526e-01
5.56676649e-02 1.13228452e+00 -2.88894624e-01 5.30461073e-01
2.85716742e-01 5.63533232e-02 -4.88073677e-01 -3.48046213e-01
5.15408218e-01 -1.01147041e-01 -1.67913303e-01 2.57589936e-01
7.03327656e-01 -7.62848020e-01 -2.20427930e-01 7.05108345e-01
6.78939760e-01 -1.63689539e-01 5.55743814e-01 -1.12174225e+00
-5.38216710e-01 -2.74933487e-01 -8.50549102e-01 4.58758086e-01
3.49798530e-01 2.36220285e-01 -6.10436536e-02 -1.42105818e+00
3.23872417e-01 1.44432831e+00 7.16929436e-01 8.18378568e-01
-8.08883011e-01 -1.15131378e+00 -2.91357428e-01 3.09838858e-02
-1.92206764e+00 -1.12138951e+00 3.76310885e-01 -5.07795632e-01
9.74546671e-01 6.21885480e-03 3.08518797e-01 6.42491281e-01
4.10905212e-01 -2.56456226e-01 1.03255904e+00 -4.08459187e-01
-2.81685919e-01 3.41090053e-01 -2.77199805e-01 1.16886699e+00
3.56298834e-01 5.24970353e-01 -4.53281254e-01 -1.75081655e-01
9.40720499e-01 9.23938677e-02 -8.56431872e-02 4.26378042e-01
-2.68845558e-01 8.20164979e-01 2.89261848e-01 2.35706195e-01
-3.97647947e-01 -1.28985405e-01 6.04264140e-02 -2.18661167e-02
6.13657176e-01 -1.71213493e-01 -2.77810195e-03 2.93024927e-01
-1.05167663e+00 -2.41658762e-01 3.29778105e-01 4.46477920e-01
6.99428320e-01 1.72665268e-01 -2.43227214e-01 7.11300790e-01
5.32834351e-01 7.99641788e-01 1.82408497e-01 -5.50066233e-01
-1.17535144e-01 7.75171518e-01 -1.23273931e-01 -1.22923923e+00
-5.44630051e-01 3.17304470e-02 -5.58592975e-01 3.17976266e-01
2.88796157e-01 -3.16469640e-01 -1.16663873e+00 1.19293535e+00
4.41387028e-01 3.11551929e-01 -3.76929402e-01 8.26682746e-01
9.36657310e-01 2.09202781e-01 1.27852440e-01 -1.81272939e-01
1.53295171e+00 -3.25278044e-01 -6.62259042e-01 -2.12348059e-01
1.88354045e-01 -8.16239536e-01 4.01175827e-01 -3.20454091e-02
-7.12007761e-01 -6.08265340e-01 -7.83212185e-01 -8.16426873e-02
-4.60073709e-01 3.89820665e-01 1.53784752e-01 1.50742972e+00
-1.25497866e+00 9.86374691e-02 -5.17282546e-01 -5.21123230e-01
1.03140616e+00 8.80347371e-01 -9.16228235e-01 -3.98448594e-02
-5.31222045e-01 1.22413373e+00 -2.13087857e-01 4.50337261e-01
-1.00254834e+00 -3.50803077e-01 -8.18617761e-01 -4.65583466e-02
-3.83284763e-02 8.17802697e-02 5.84332168e-01 -7.06898808e-01
-1.04642045e+00 1.54895329e+00 -6.49138629e-01 -9.33313146e-02
2.01622024e-01 1.48620874e-01 -7.37307489e-01 2.09603578e-01
-1.96459487e-01 7.81903386e-01 1.33615875e+00 -8.12722206e-01
-4.01791215e-01 -9.20358360e-01 -5.85447669e-01 7.99135119e-02
-5.90920039e-02 7.90531158e-01 -1.20168619e-01 -1.89040229e-01
-8.76396075e-02 -8.09410691e-01 3.37844819e-01 1.33620903e-01
-1.98912904e-01 -1.14870086e-01 1.37962937e+00 -1.05118322e+00
6.70390129e-01 -2.32208252e+00 -7.23057389e-01 2.71304876e-01
3.74856740e-02 4.50139433e-01 -1.39191300e-01 1.56369619e-02
-7.53330663e-02 2.47364670e-01 1.17140137e-01 -4.73940223e-01
-3.09715092e-01 -1.09906808e-01 -2.64059097e-01 9.97803211e-01
3.12121123e-01 8.80041718e-01 -1.44292653e-01 -4.46914941e-01
3.27060044e-01 1.01153743e+00 -2.97621429e-01 1.94944367e-01
2.97223359e-01 4.89967853e-01 5.93199134e-02 1.23427963e+00
9.92842376e-01 1.71515226e-01 8.87139738e-02 -4.61680740e-02
-1.43965989e-01 -2.67915189e-01 -8.39792311e-01 6.03732646e-01
1.77468032e-01 6.90962374e-01 3.93886000e-01 -2.45071039e-01
1.07545459e+00 5.94273329e-01 6.53330758e-02 -4.80973989e-01
4.89732653e-01 2.47822683e-02 1.56914778e-02 -5.51322520e-01
1.10212958e-03 -1.95133299e-01 4.73877966e-01 2.21548766e-01
-1.75408408e-01 2.76102215e-01 -3.88893604e-01 -4.26323175e-01
4.93071556e-01 -1.09018341e-01 3.07179779e-01 -3.33912700e-01
4.30745333e-01 -3.98274064e-01 2.50084996e-01 1.72130898e-01
-6.03276253e-01 7.03321934e-01 2.30895758e-01 -6.23813212e-01
-4.87432510e-01 -8.13643336e-01 -3.61790776e-01 9.86065567e-01
-3.50969017e-01 2.21360326e-01 -9.14521813e-01 -7.98144281e-01
3.83086294e-01 1.49086058e-01 -8.14413965e-01 -9.10060480e-02
-4.97622341e-01 -6.31085575e-01 7.65662193e-01 4.76722151e-01
5.32568157e-01 -1.19561434e+00 -6.70348704e-01 -3.35053205e-01
4.41138029e-01 -9.76774335e-01 -6.79828346e-01 -3.10554594e-01
-5.48016489e-01 -1.32018209e+00 -5.02145708e-01 -9.40995634e-01
1.02773976e+00 3.07659686e-01 3.83281559e-01 5.18386722e-01
-8.52011323e-01 3.43741417e-01 9.26751196e-02 -6.38915658e-01
4.33652066e-02 -3.20716918e-01 5.25158882e-01 4.88793224e-01
7.53956974e-01 -7.25584328e-02 -8.55163574e-01 4.65592355e-01
-5.87148488e-01 -7.73776174e-01 2.82361567e-01 3.20443004e-01
2.43019298e-01 -1.09811880e-01 6.09696150e-01 -5.01801550e-01
3.73801649e-01 -2.60367513e-01 -7.46616364e-01 1.33188039e-01
-4.09787327e-01 -7.44601667e-01 8.62914473e-02 -1.38170704e-01
-9.79566693e-01 3.02933127e-01 -5.42273410e-02 -4.96850938e-01
-3.73512357e-01 -2.50338376e-01 -1.79067224e-01 -7.42906153e-01
6.02691472e-01 -7.86331296e-02 4.94425446e-01 -4.07818705e-01
-2.39491731e-01 9.14984047e-01 3.32645208e-01 9.90289599e-02
7.34390676e-01 8.64132345e-01 1.26899570e-01 -1.46167099e+00
-4.14656878e-01 -3.62373143e-01 -7.45260835e-01 -4.27716970e-01
1.09357142e+00 -7.98121512e-01 -1.19916236e+00 5.13967812e-01
-1.07246280e+00 4.11159024e-02 3.86584014e-01 2.55604357e-01
1.92269236e-01 1.47524491e-01 -3.19089800e-01 -1.17713869e+00
-4.70358074e-01 -1.07011807e+00 9.57946301e-01 3.88351887e-01
-1.22931898e-01 -5.77221870e-01 -2.41187930e-01 4.31249261e-01
5.05144000e-01 2.29790956e-01 4.76204157e-01 -4.06417280e-01
-3.63327444e-01 -6.48765087e-01 -5.17255902e-01 3.53087187e-01
5.92265725e-01 2.38533646e-01 -1.57865286e+00 -4.08520043e-01
9.05246064e-02 -5.72965257e-02 5.96844435e-01 5.90301573e-01
9.80443180e-01 -1.67225108e-01 -5.68261027e-01 6.52892530e-01
1.12716770e+00 6.58567667e-01 7.51889348e-01 -3.87916952e-01
4.86988872e-01 9.57471848e-01 -2.18109135e-02 1.61878273e-01
-3.12420949e-02 4.82796460e-01 2.13516265e-01 -4.05216515e-02
-1.94876641e-01 -5.75197414e-02 1.57917842e-01 -1.73386097e-01
-1.10787041e-01 2.20458712e-02 -1.06061554e+00 1.90220252e-01
-9.79056120e-01 -8.64302099e-01 2.73122549e-01 2.05612421e+00
1.06631584e-01 -4.49393749e-01 2.20668882e-01 -1.21748209e-01
1.01448762e+00 -1.45366922e-01 -3.42050344e-01 -2.19107926e-01
1.42278867e-02 4.20871586e-01 4.46310043e-01 4.31064069e-01
-1.12792313e+00 8.85569036e-01 7.72230196e+00 2.39272162e-01
-1.43599212e+00 1.80216119e-01 1.11415493e+00 -1.83779985e-01
7.00205028e-01 -4.78693128e-01 -1.23309028e+00 4.87070709e-01
6.50430739e-01 3.64658684e-01 5.29148698e-01 6.64375842e-01
1.91011801e-01 -1.54113576e-01 -7.96828568e-01 1.17255092e+00
4.12618577e-01 -9.67308223e-01 -2.05122292e-01 2.70076483e-01
2.73425579e-01 -2.84552455e-01 5.39294124e-01 2.75795795e-02
-5.48126936e-01 -1.75563598e+00 1.54532224e-01 3.86623979e-01
1.39272177e+00 -6.37729108e-01 5.98197699e-01 -8.92159417e-02
-1.03596234e+00 -1.55439585e-01 -1.37862504e-01 -2.09670201e-01
-9.49794650e-02 -8.42960328e-02 -1.21113670e+00 -3.95489663e-01
6.55823886e-01 -5.41398451e-02 -5.36783576e-01 8.48963320e-01
1.88596454e-02 4.61718231e-01 -3.36049616e-01 1.98055476e-01
-1.24186590e-01 7.13333413e-02 1.42747477e-01 1.06130910e+00
2.95269161e-01 2.66899556e-01 -6.59293681e-02 7.40663469e-01
-3.58450711e-01 2.95885336e-02 -8.48070741e-01 -1.45299986e-01
4.20488924e-01 1.39866829e+00 -8.87940288e-01 1.07285611e-01
-5.27963817e-01 7.46704221e-01 8.25788230e-02 3.04988593e-01
-5.39015889e-01 -2.77498513e-02 8.66243720e-01 5.86611271e-01
2.15264246e-01 -3.22476216e-02 -3.68014842e-01 -4.98349100e-01
-4.51809056e-02 -6.35020137e-01 4.35047984e-01 -3.72180462e-01
-8.76546562e-01 5.35937190e-01 -7.19277933e-02 -4.60770905e-01
4.96538989e-02 -8.94755960e-01 -7.55867958e-01 1.34243095e+00
-1.27357531e+00 -1.00579131e+00 -4.17455196e-01 4.77881759e-01
1.12669095e-01 -3.96876305e-01 9.49124575e-01 3.19871843e-01
-7.77575791e-01 8.32267940e-01 -3.88297379e-01 4.85231131e-01
4.93352085e-01 -1.73679322e-01 3.99457574e-01 6.38231635e-01
-2.22298831e-01 6.80695355e-01 9.36292112e-02 -1.01000571e+00
-1.29321384e+00 -1.15046680e+00 9.08988118e-01 -6.95349216e-01
-1.19659282e-01 -3.84496778e-01 -7.66208291e-01 8.93508375e-01
-1.18240103e-01 2.33042449e-01 5.75637817e-01 -4.26659942e-01
-3.55307162e-01 -2.57226795e-01 -1.92588925e+00 3.40754867e-01
7.96577275e-01 -7.55190790e-01 -1.06140897e-01 3.49538207e-01
-1.06563019e-02 -1.09594956e-01 -2.62366712e-01 5.53330243e-01
9.38032389e-01 -7.53403187e-01 8.89338791e-01 -3.56703103e-01
-2.35771269e-01 -2.04747289e-01 -1.53607264e-01 -5.92409253e-01
-2.74536759e-01 -4.33015734e-01 2.43941814e-01 1.07105446e+00
1.40147597e-01 -9.67681646e-01 9.49762285e-01 8.05915177e-01
3.84115994e-01 -4.10500109e-01 -1.32245326e+00 -5.23289323e-01
-2.84423411e-01 2.22354889e-01 4.48730826e-01 7.00023055e-01
-2.89209485e-01 -9.13963243e-02 -1.76617920e-01 5.44133484e-01
5.58991611e-01 -2.96389729e-01 3.13736767e-01 -9.26964402e-01
4.32585984e-01 -4.37373310e-01 -6.56260967e-01 -1.23021211e-02
9.49584618e-02 -7.69805849e-01 -2.78450936e-01 -1.09122539e+00
5.27358018e-02 5.57572246e-02 6.61008880e-02 5.73176861e-01
1.30170524e-01 1.04627216e+00 1.16355419e-01 -2.14474201e-01
2.35610828e-01 8.01693499e-02 7.95579731e-01 1.72886774e-01
-9.21141058e-02 -1.84702441e-01 -5.05629599e-01 6.06662214e-01
8.85250449e-01 -2.96305776e-01 8.04521516e-02 -1.53753504e-01
-5.08825898e-01 -7.77877644e-02 6.55551851e-01 -9.93065417e-01
1.88657820e-01 9.69664380e-02 1.27203906e+00 -4.20295149e-01
7.58450210e-01 -7.08710253e-01 2.28691489e-01 6.42891467e-01
3.72476667e-01 2.14748755e-01 6.58016205e-01 -9.06418711e-02
2.16869444e-01 -8.16664919e-02 1.20098162e+00 -4.98200916e-02
-4.17937994e-01 5.15287936e-01 -2.42543742e-01 -5.31862736e-01
1.27623785e+00 -5.05765498e-01 -3.65067720e-01 -1.74509630e-01
-6.17399275e-01 -9.50897411e-02 4.58281636e-01 4.43750620e-01
9.87645328e-01 -9.98015106e-01 -8.42310369e-01 1.00692248e+00
-2.47566909e-01 -6.31223381e-01 3.53898227e-01 6.88205004e-01
-6.20119095e-01 6.14176333e-01 -4.35998589e-01 -4.44900721e-01
-1.75767910e+00 5.74572742e-01 4.94258344e-01 7.60005474e-01
-3.13790113e-01 1.16381705e+00 2.01518893e-01 -3.61752920e-02
3.57970625e-01 2.58948833e-01 -1.90548241e-01 1.42727032e-01
8.30619991e-01 6.52953625e-01 1.07170559e-01 -1.10807157e+00
-7.56810844e-01 7.08880305e-01 -2.45179698e-01 2.15997517e-01
6.91001594e-01 8.77505690e-02 -3.33971173e-01 -1.91286206e-01
1.16778696e+00 -7.01099858e-02 -9.52632666e-01 3.12454104e-01
-2.00612336e-01 -6.88360751e-01 -4.82195169e-02 -8.01064432e-01
-1.11026084e+00 9.27886069e-01 1.23157728e+00 -2.32942417e-01
9.79864180e-01 -5.90286106e-02 2.48422444e-01 -1.16638735e-01
3.05950373e-01 -7.33375371e-01 -2.57255435e-01 2.95184880e-01
1.03963196e+00 -1.45439506e+00 -1.58440337e-01 -5.29147208e-01
-2.88379192e-01 9.02640224e-01 7.73620129e-01 -2.79199537e-02
8.91397417e-01 3.62640679e-01 3.37826788e-01 -4.89958704e-01
-2.33677834e-01 -1.05339527e-01 5.72886467e-01 7.06690788e-01
4.56077814e-01 1.54590636e-01 1.18399963e-01 2.42436677e-01
1.03628859e-02 -6.01609834e-02 -8.73201042e-02 8.16634893e-01
-4.93721962e-01 -5.31049252e-01 -7.74108231e-01 7.52880156e-01
-6.65907919e-01 2.26011556e-02 -6.24184728e-01 6.90930486e-01
3.26997042e-01 1.08858931e+00 2.77401179e-01 -3.10299695e-01
1.34757414e-01 6.38502419e-01 5.62247515e-01 -8.29837024e-01
-6.71332240e-01 -2.06590921e-01 -2.89386600e-01 -4.60818499e-01
-1.07385144e-02 -4.12619770e-01 -9.10897315e-01 -4.17188793e-01
-1.52652338e-01 -3.22881520e-01 8.52795303e-01 7.02446938e-01
6.04203284e-01 -1.19953342e-01 5.01605988e-01 -8.29491019e-01
-3.40840578e-01 -1.07672620e+00 -8.10924113e-01 -1.34555802e-01
5.65600634e-01 -8.09055328e-01 -4.38877910e-01 2.64886580e-02]
|
[13.321727752685547, 0.799431562423706]
|
6854ae08-3cc8-4231-a265-31a320d10188
|
csdr-bert-a-pre-trained-scientific-dataset
|
2301.12700
| null |
https://arxiv.org/abs/2301.12700v3
|
https://arxiv.org/pdf/2301.12700v3.pdf
|
CSDR-BERT: a pre-trained scientific dataset match model for Chinese Scientific Dataset Retrieval
|
As the number of open and shared scientific datasets on the Internet increases under the open science movement, efficiently retrieving these datasets is a crucial task in information retrieval (IR) research. In recent years, the development of large models, particularly the pre-training and fine-tuning paradigm, which involves pre-training on large models and fine-tuning on downstream tasks, has provided new solutions for IR match tasks. In this study, we use the original BERT token in the embedding layer, improve the Sentence-BERT model structure in the model layer by introducing the SimCSE and K-Nearest Neighbors method, and use the cosent loss function in the optimization phase to optimize the target output. Our experimental results show that our model outperforms other competing models on both public and self-built datasets through comparative experiments and ablation implementations. This study explores and validates the feasibility and efficiency of pre-training techniques for semantic retrieval of Chinese scientific datasets.
|
['XiaoFeng Wang', 'Jian Wang', 'Xunxun Gu', 'Meng Wang', 'Yingfei Wang', 'Jianping Liu', 'Xintao Chu']
|
2023-01-30
| null | null | null | null |
['semantic-retrieval']
|
['natural-language-processing']
|
[ 1.79574471e-02 -8.15386549e-02 -1.94270283e-01 -4.09592897e-01
-1.12929583e+00 -6.16384149e-01 6.85045421e-01 3.45273286e-01
-9.54603970e-01 5.16827404e-01 3.40440094e-01 -2.87171453e-01
-3.69700730e-01 -6.67198837e-01 -5.71344316e-01 -2.58713812e-01
1.26754433e-01 7.26729929e-01 1.00305721e-01 -2.06571296e-01
5.43568075e-01 3.04707378e-01 -1.37801015e+00 4.35758561e-01
1.15561259e+00 8.00066710e-01 6.31081581e-01 5.56285322e-01
-5.46474993e-01 3.15521628e-01 -5.19737244e-01 -5.10208845e-01
1.94012195e-01 1.91084202e-02 -1.05412161e+00 -9.65070009e-01
2.38952145e-01 1.15930296e-01 -1.20708607e-01 8.40890586e-01
1.03701818e+00 1.80901870e-01 4.03271884e-01 -5.65887988e-01
-1.08285928e+00 8.82994652e-01 -3.45047861e-01 2.94654518e-01
5.09626754e-02 -2.32900247e-01 1.37321150e+00 -8.38606596e-01
7.91272819e-01 1.35797012e+00 3.37979674e-01 6.61510646e-01
-1.02669394e+00 -6.39969409e-01 -4.53077331e-02 4.37793881e-01
-1.47555327e+00 -3.86452407e-01 4.88642216e-01 -4.65009771e-02
1.23437560e+00 4.67799902e-01 1.90409005e-01 8.49726737e-01
2.30726227e-02 8.49591672e-01 6.55385017e-01 -8.94447148e-01
3.05940881e-02 3.37234408e-01 2.81985343e-01 3.13612789e-01
8.07653964e-02 -1.35630772e-01 -3.90041113e-01 -4.20090705e-01
1.62648693e-01 -6.31155446e-02 -1.02195799e-01 -5.30294962e-02
-1.16582167e+00 8.25276315e-01 5.72155178e-01 5.43174028e-01
-1.53755248e-01 -1.05889566e-01 3.75159621e-01 4.92329538e-01
7.60895610e-01 1.18310893e+00 -6.87974095e-01 3.20061147e-02
-1.00549889e+00 3.22446734e-01 7.99412668e-01 8.82998228e-01
6.01167858e-01 -7.81358123e-01 -7.45434701e-01 1.29631758e+00
3.90885532e-01 2.90992081e-01 8.69010746e-01 -7.54098475e-01
5.42287767e-01 7.26475358e-01 -1.60056129e-01 -5.05604029e-01
-2.41430044e-01 -6.92172289e-01 -5.09223163e-01 -6.09576285e-01
-6.43623397e-02 1.33171290e-01 -8.78701985e-01 1.62628138e+00
2.40461305e-01 9.56644937e-02 2.16290906e-01 8.91835093e-01
1.00966883e+00 7.29529500e-01 4.76878881e-01 1.77416533e-01
1.34016716e+00 -1.25202882e+00 -7.22133875e-01 -2.04685092e-01
1.03994310e+00 -1.03202200e+00 1.30214930e+00 -1.23576835e-01
-9.55135286e-01 -5.18269300e-01 -9.35360372e-01 -6.09140456e-01
-7.96449721e-01 2.74355471e-01 5.99130690e-01 2.08497897e-01
-1.11729980e+00 7.48703182e-01 -5.22472084e-01 -3.12072515e-01
3.23084086e-01 2.10299596e-01 -1.79039970e-01 -4.22460109e-01
-1.59757602e+00 9.50873852e-01 4.12159204e-01 4.72616442e-02
-4.46258992e-01 -1.09778810e+00 -5.70215523e-01 6.13708317e-01
1.20222926e-01 -8.84558022e-01 8.88902843e-01 -4.99291450e-01
-1.53200924e+00 9.83672023e-01 -7.53996223e-02 -2.38824010e-01
2.68317640e-01 -5.08692503e-01 -3.42756480e-01 -4.78956178e-02
1.45088788e-02 8.36044371e-01 3.12820494e-01 -8.80043209e-01
-2.51954615e-01 -3.67160559e-01 1.28838569e-01 3.08198690e-01
-7.89501190e-01 4.14268970e-01 -8.88029158e-01 -5.06008267e-01
-1.91614836e-01 -8.11503053e-01 -3.36806029e-01 -2.40878686e-01
-2.72758186e-01 -7.03906417e-01 3.97589773e-01 -6.45907879e-01
1.56839490e+00 -2.11111569e+00 3.33442718e-01 2.48492792e-01
-1.91803854e-02 4.53536004e-01 -6.75530791e-01 5.12827814e-01
8.33291411e-02 3.56093258e-01 -5.80045804e-02 -4.28234518e-01
4.68073338e-02 -4.04081456e-02 -3.26477885e-01 -3.73320766e-02
4.66374829e-02 1.12733316e+00 -8.77622247e-01 -6.97251081e-01
-2.77268469e-01 4.47354466e-01 -6.50014162e-01 4.17727619e-01
-3.09576780e-01 1.84917629e-01 -8.39620709e-01 3.90806198e-01
5.56758821e-01 -6.99025095e-01 6.46596998e-02 -9.40741040e-03
-1.67517066e-01 6.67238355e-01 -7.65479386e-01 2.31993198e+00
-6.22604311e-01 1.95878088e-01 -1.66340783e-01 -8.46105874e-01
6.31677687e-01 2.51637965e-01 3.06812614e-01 -1.03617179e+00
-2.44010434e-01 3.87617886e-01 -1.30398139e-01 -5.18664896e-01
5.54458618e-01 4.50412273e-01 -7.10385069e-02 5.47367394e-01
8.48061293e-02 4.99215024e-03 3.42107028e-01 5.47007442e-01
1.10547686e+00 3.07228804e-01 -1.87844291e-01 -4.89140093e-01
7.06387281e-01 4.93183434e-02 3.34533364e-01 9.49832022e-01
3.41280907e-01 5.60084879e-01 8.20247680e-02 -2.97758400e-01
-9.27368581e-01 -6.55346990e-01 -4.16558862e-01 1.36456800e+00
-5.58547024e-03 -5.42463839e-01 -6.76657438e-01 -5.88701546e-01
1.50064319e-01 6.18611693e-01 -3.68849963e-01 -4.89810824e-01
-5.96064031e-01 -8.94897878e-01 4.18429166e-01 7.87162408e-02
2.91675359e-01 -1.05154383e+00 -1.47975773e-01 1.43706262e-01
-3.32438648e-01 -1.05564654e+00 -4.65831667e-01 6.53556138e-02
-8.67714882e-01 -6.30694926e-01 -8.80593717e-01 -7.73621678e-01
5.25961101e-01 3.17972124e-01 1.24889541e+00 2.57200241e-01
-6.13368094e-01 6.32019788e-02 -4.99018520e-01 -5.90799689e-01
-1.34291887e-01 7.48191357e-01 -2.60183156e-01 -4.21783686e-01
6.87037528e-01 -2.19026640e-01 -7.16519952e-01 -6.13377653e-02
-1.04440010e+00 1.11006252e-01 7.89714634e-01 8.51084948e-01
5.38447797e-01 -5.17399728e-01 7.03253627e-01 -9.72841442e-01
8.41739416e-01 -4.81282443e-01 -6.62576675e-01 9.51287806e-01
-1.12942934e+00 5.12614787e-01 4.59099591e-01 -2.04654008e-01
-9.82133090e-01 -4.96637046e-01 -5.57009168e-02 -4.40653831e-01
2.58658946e-01 9.24239814e-01 -4.27899621e-02 4.07461170e-03
5.79483449e-01 -1.54017620e-02 -9.10891294e-02 -1.14210463e+00
5.23040771e-01 1.07775283e+00 -2.22096741e-02 -6.56538606e-01
7.03358114e-01 1.64295420e-01 -3.31911623e-01 -5.09181857e-01
-1.17178249e+00 -7.44735420e-01 -3.75144303e-01 2.92335927e-01
7.07273424e-01 -1.15157115e+00 -4.88333970e-01 -5.71623736e-04
-1.26148665e+00 1.24198996e-01 -1.37674555e-01 5.73038578e-01
1.12920571e-02 3.40985894e-01 -7.07444966e-01 -3.56689095e-01
-9.59838271e-01 -1.13321066e+00 1.27760267e+00 3.60252291e-01
-1.31226420e-01 -8.97844493e-01 3.86353105e-01 5.35302401e-01
9.55562174e-01 -6.67128265e-01 1.26491749e+00 -9.55413640e-01
-5.86945295e-01 -2.49268308e-01 -4.19027865e-01 2.98559546e-01
-1.99863285e-01 -3.48364860e-01 -1.07787371e+00 -3.58941168e-01
-3.52328479e-01 -4.93798107e-01 1.15532589e+00 1.13259651e-01
1.65659869e+00 -7.63597190e-02 -4.79175478e-01 7.61054337e-01
1.39021647e+00 -2.72971690e-01 5.75299203e-01 7.23634422e-01
5.11941314e-01 6.49027705e-01 5.04376650e-01 1.52642414e-01
3.97975445e-01 6.80265486e-01 -4.02537622e-02 -3.77138257e-02
-1.73104122e-01 -2.31805190e-01 2.80799647e-03 1.06810856e+00
9.51604024e-02 -2.60985881e-01 -8.91666174e-01 4.04513389e-01
-1.73275459e+00 -7.94428051e-01 3.28090876e-01 2.30983806e+00
1.40112162e+00 -2.32879072e-01 -6.56091869e-01 -5.53947806e-01
3.69364142e-01 -3.42588797e-02 -4.19573188e-01 -3.46255690e-01
-2.58275539e-01 6.34138703e-01 4.76723462e-01 4.98630613e-01
-8.48771453e-01 1.25575781e+00 6.27417660e+00 1.17940736e+00
-9.86517370e-01 1.79072320e-01 6.71151876e-01 -1.92559093e-01
-7.12711930e-01 1.79198563e-01 -1.11802006e+00 4.13127422e-01
1.15383828e+00 -2.73872018e-01 4.48356986e-01 5.67288339e-01
7.03272689e-03 2.14710310e-01 -1.07678902e+00 6.54736161e-01
7.86917433e-02 -1.49758780e+00 3.96370500e-01 -2.43519947e-01
5.15650690e-01 3.95856440e-01 -1.23965383e-01 6.96362495e-01
3.09489965e-01 -9.90154088e-01 2.66158462e-01 5.93798637e-01
7.00138032e-01 -3.96570235e-01 7.84989357e-01 2.91559309e-01
-5.63886285e-01 -3.55438255e-02 -6.36913002e-01 3.56521130e-01
-1.60619557e-01 7.11514473e-01 -3.80616039e-01 7.18941450e-01
8.55008662e-01 6.94135845e-01 -7.78416574e-01 1.05698311e+00
-9.34650972e-02 5.24191082e-01 -3.23194504e-01 -1.20598935e-01
1.72914430e-01 -3.23120773e-01 4.56350118e-01 1.26630735e+00
2.08252892e-01 -1.51924863e-01 -1.70882687e-01 1.03651249e+00
-4.56401676e-01 4.06116724e-01 -2.21745834e-01 -3.20542485e-01
4.20237929e-01 1.46203113e+00 -4.55844514e-02 -2.82363296e-01
-4.05624509e-01 8.19293380e-01 7.62316942e-01 4.39312875e-01
-5.56317687e-01 -8.60464156e-01 5.95991552e-01 -2.41269156e-01
1.15307182e-01 5.90443760e-02 -1.24858938e-01 -1.27059293e+00
1.79044724e-01 -8.08524251e-01 7.18016505e-01 -5.85307300e-01
-1.33489037e+00 5.93216538e-01 -7.62491152e-02 -8.26402366e-01
-1.69110090e-01 -3.14953059e-01 -4.01722401e-01 1.24020827e+00
-2.26596951e+00 -1.08006525e+00 1.08626060e-01 2.82081366e-01
3.26794326e-01 -1.94618240e-01 1.14112890e+00 6.25983357e-01
-6.56684697e-01 8.34419668e-01 7.83575594e-01 4.02090922e-02
1.07883656e+00 -8.02904844e-01 2.53301233e-01 3.95042866e-01
1.23763248e-01 1.13326621e+00 4.37766284e-01 -4.06441092e-01
-1.60502088e+00 -7.83368707e-01 1.51626372e+00 -3.81640851e-01
5.90187073e-01 -6.27117276e-01 -1.06550229e+00 2.88487971e-01
3.22189629e-01 -1.55004099e-01 8.86576772e-01 5.45037389e-01
-4.10680115e-01 -2.52191842e-01 -7.71384716e-01 4.76223528e-01
9.31431293e-01 -5.78102350e-01 -7.20444560e-01 5.59411824e-01
1.09693921e+00 -1.66075647e-01 -1.02808380e+00 5.36220133e-01
5.15661597e-01 -1.23392068e-01 1.30043650e+00 -1.07906103e+00
5.63808024e-01 9.98919681e-02 1.46229923e-01 -1.29733586e+00
-4.07096744e-01 -4.84278411e-01 2.96334513e-02 1.20403230e+00
7.73513615e-01 -3.69527131e-01 3.86443198e-01 7.73178577e-01
-1.39653191e-01 -7.39544988e-01 -8.75403225e-01 -6.28325343e-01
3.87135655e-01 1.51173264e-01 5.58867931e-01 1.04518008e+00
1.19691595e-01 4.93592352e-01 1.70101404e-01 -1.77780315e-01
3.72815877e-01 5.27108133e-01 4.06170219e-01 -1.32869172e+00
-2.97434390e-01 -4.94941831e-01 1.34488374e-01 -1.02081001e+00
4.99200702e-01 -1.35326505e+00 -2.52729446e-01 -1.56797981e+00
5.54433107e-01 -9.40736175e-01 -9.14845109e-01 4.90867972e-01
-7.08288789e-01 -2.25360379e-01 8.84763524e-03 5.27966261e-01
-8.27441752e-01 7.16360211e-01 1.04830444e+00 -1.58296451e-01
-7.81041980e-02 -3.72048199e-01 -7.89870381e-01 4.10908237e-02
5.43899715e-01 -7.73464143e-01 -2.12815911e-01 -1.03391361e+00
5.84863842e-01 -4.34178531e-01 -5.81191573e-03 -6.75631762e-01
3.33332956e-01 1.01723522e-01 1.22306295e-01 -1.80115193e-01
7.04897121e-02 -6.49635017e-01 -3.29659283e-01 2.48511612e-01
-1.02294695e+00 4.67848778e-02 4.02328283e-01 4.22499001e-01
-2.78624803e-01 -4.10786241e-01 4.65017289e-01 -1.71675786e-01
-4.63082910e-01 1.97825104e-01 2.15670317e-01 2.16781393e-01
5.68183661e-01 4.09890771e-01 -4.30551440e-01 1.87416151e-01
-2.09149927e-01 7.54797578e-01 1.08174235e-01 8.64668906e-01
4.05038804e-01 -1.05043340e+00 -9.20418441e-01 1.46836802e-01
5.33069193e-01 -9.07802954e-03 1.98529705e-01 6.11047924e-01
-3.26559335e-01 9.37086523e-01 1.14750013e-01 -3.04694772e-01
-1.12474906e+00 3.94590735e-01 -1.79099031e-02 -9.28952634e-01
-3.83589953e-01 9.76738691e-01 4.86257523e-02 -8.50129724e-01
3.36036265e-01 2.46400252e-01 -5.09475887e-01 -1.85657829e-01
6.55275226e-01 1.35382816e-01 2.96739042e-01 -5.17159663e-02
-1.94617167e-01 2.86446750e-01 -6.38539076e-01 -3.82085033e-02
1.54031897e+00 -1.49984926e-01 -5.81830263e-01 2.23110884e-01
1.58887017e+00 -1.44829810e-01 -3.21021825e-01 -7.56381035e-01
2.85102099e-01 -3.25815499e-01 4.80423212e-01 -1.16430926e+00
-8.22866261e-01 8.10945451e-01 5.58073044e-01 -3.89000088e-01
8.52408230e-01 1.63174406e-01 7.93386817e-01 7.58116841e-01
1.14919178e-01 -1.36495304e+00 -2.87588775e-01 7.27485895e-01
6.38553023e-01 -1.26502752e+00 6.02176301e-02 -9.64033455e-02
-2.73442060e-01 8.24231505e-01 5.92204094e-01 2.27819592e-01
7.01425910e-01 -6.55783415e-02 1.77476645e-01 -2.50961512e-01
-1.01674008e+00 -2.81994343e-02 5.47759891e-01 -1.42133504e-01
9.17431772e-01 -2.29017317e-01 -7.55604208e-01 5.18382013e-01
-4.69414610e-03 1.50878415e-01 -1.57628268e-01 9.99779701e-01
-3.12206179e-01 -1.54554355e+00 4.48579639e-02 5.19820631e-01
-8.11822295e-01 -7.62335122e-01 -3.81435931e-01 3.08337122e-01
-4.51460928e-01 7.32927918e-01 2.74764933e-02 -7.81209543e-02
2.43182078e-01 4.20385391e-01 1.91711143e-01 -6.74696207e-01
-9.07908559e-01 -2.52746373e-01 -7.33371973e-02 -6.46472454e-01
-1.68268651e-01 -2.61716247e-01 -1.05693221e+00 -4.66082655e-02
-6.19584084e-01 7.27510035e-01 1.09502637e+00 7.77153671e-01
1.15514112e+00 5.23892581e-01 5.02041936e-01 -3.55016112e-01
-1.11038494e+00 -1.27488112e+00 -1.32615164e-01 5.23813486e-01
-1.50354624e-01 -2.26992533e-01 -3.57520163e-01 -2.83193707e-01]
|
[11.330060958862305, 7.793327331542969]
|
bbd0eedf-d542-4d9d-ac43-03348c21251c
|
graph-driven-generative-models-for
|
1911.08709
| null |
https://arxiv.org/abs/1911.08709v1
|
https://arxiv.org/pdf/1911.08709v1.pdf
|
Graph-Driven Generative Models for Heterogeneous Multi-Task Learning
|
We propose a novel graph-driven generative model, that unifies multiple heterogeneous learning tasks into the same framework. The proposed model is based on the fact that heterogeneous learning tasks, which correspond to different generative processes, often rely on data with a shared graph structure. Accordingly, our model combines a graph convolutional network (GCN) with multiple variational autoencoders, thus embedding the nodes of the graph i.e., samples for the tasks) in a uniform manner while specializing their organization and usage to different tasks. With a focus on healthcare applications (tasks), including clinical topic modeling, procedure recommendation and admission-type prediction, we demonstrate that our method successfully leverages information across different tasks, boosting performance in all tasks and outperforming existing state-of-the-art approaches.
|
['Zhe Gan', 'Wenlin Wang', 'Lawrence Carin', 'Qian Yang', 'Liqun Chen', 'Bai Li', 'Wenqi Wang', 'Hongteng Xu', 'Guoyin Wang']
|
2019-11-20
| null | null | null | null |
['type-prediction']
|
['computer-code']
|
[-8.68424922e-02 3.46657932e-01 -1.55334488e-01 -1.34296253e-01
-4.80118454e-01 -3.07457358e-01 7.07594395e-01 2.12739244e-01
4.43329476e-02 5.72094262e-01 5.68961084e-01 -1.79963231e-01
-1.76838800e-01 -1.00757372e+00 -7.53226161e-01 -8.70154738e-01
2.72676975e-01 8.20888340e-01 -1.42631814e-01 1.11838788e-01
-2.89825201e-01 -1.05355971e-01 -9.78923738e-01 2.31102839e-01
9.49699998e-01 6.20414138e-01 3.79962176e-01 5.39843559e-01
-8.76319781e-02 9.49102283e-01 -6.15297854e-01 -8.56669903e-01
-1.65721729e-01 -5.07545590e-01 -5.86814523e-01 2.82662362e-01
1.13458954e-01 4.63801958e-02 -5.80078185e-01 8.43329549e-01
4.30727363e-01 1.55175552e-01 1.06600940e+00 -1.28076077e+00
-1.16144764e+00 7.88884819e-01 -1.13432929e-01 -1.77429408e-01
-1.45088956e-01 -3.94389033e-02 1.25388598e+00 -5.22424519e-01
8.36113513e-01 1.02077544e+00 6.18612945e-01 6.08110964e-01
-1.54091442e+00 -3.91547889e-01 3.84533703e-01 4.63350443e-03
-1.33074975e+00 6.40267134e-02 9.26712513e-01 -7.90486097e-01
6.19361460e-01 -1.35975763e-01 8.75057638e-01 1.97419846e+00
7.39752829e-01 8.81025434e-01 6.25130057e-01 2.36569464e-01
4.82437849e-01 -5.16598970e-02 2.56872088e-01 7.43426561e-01
5.99937558e-01 -4.82752770e-01 -3.61592978e-01 -3.29306006e-01
7.47513831e-01 6.94773138e-01 -2.69018531e-01 -7.64595151e-01
-1.11850595e+00 1.12433577e+00 3.35506499e-01 4.35959995e-01
-7.96086371e-01 2.30793551e-01 3.75916719e-01 -1.07153468e-02
5.07379711e-01 8.59175250e-02 -7.70792887e-02 2.21411303e-01
-7.89322853e-01 2.73821294e-01 9.92029428e-01 1.12204051e+00
7.36175776e-01 1.71114728e-01 -8.31145585e-01 5.07806480e-01
5.18051267e-01 2.78813839e-01 5.89062512e-01 -3.71476054e-01
3.15799773e-01 7.30283439e-01 -3.32471579e-01 -9.12942350e-01
-4.93425786e-01 -9.42491055e-01 -1.27977395e+00 -5.57728887e-01
-1.42497858e-02 -4.10818249e-01 -1.05375099e+00 1.97535563e+00
2.48104125e-01 5.85227072e-01 1.80101216e-01 4.17086333e-01
1.13513064e+00 4.38594788e-01 5.91737449e-01 9.13401619e-02
1.27804279e+00 -1.07263112e+00 -7.90931761e-01 -1.55891925e-01
4.33942527e-01 -2.02237427e-01 8.35501552e-01 3.42353463e-01
-8.97560239e-01 -6.66473031e-01 -8.36866677e-01 -2.49795809e-01
-2.64461815e-01 8.15539606e-05 6.54862285e-01 2.78887123e-01
-1.25878048e+00 5.24575114e-01 -1.04853499e+00 -4.22025502e-01
5.23121238e-01 -5.74499555e-02 -3.15819904e-02 -2.03468233e-01
-9.72316086e-01 3.27838600e-01 4.10138726e-01 -2.09874496e-01
-1.37988329e+00 -8.27783883e-01 -8.67045879e-01 5.40928960e-01
5.31949103e-01 -1.78266597e+00 8.63690138e-01 -4.01164830e-01
-1.39542043e+00 4.53788340e-01 5.14027029e-02 -5.28039038e-01
5.26459754e-01 -1.22825146e-01 -3.09343457e-01 -1.60578728e-01
-1.07115991e-01 1.80259854e-01 1.23495042e+00 -1.32942891e+00
-2.25692838e-01 -3.48081648e-01 -1.46163896e-01 2.60792188e-02
-5.25755525e-01 -6.24490201e-01 -6.97674334e-01 -7.45767176e-01
-2.67871171e-01 -1.06595039e+00 -3.59962761e-01 -7.59104848e-01
-6.97203338e-01 -4.95208114e-01 6.52650177e-01 -4.93481666e-01
1.46133196e+00 -2.03158855e+00 9.31952596e-01 7.69832870e-04
1.05347168e+00 -1.39737308e-01 -4.01018262e-02 7.04607964e-01
3.11361849e-01 2.41790518e-01 -2.11122945e-01 -9.50077772e-01
2.08448142e-01 4.32942748e-01 -6.19334355e-02 3.04460138e-01
4.05616276e-02 1.52051365e+00 -1.10614681e+00 -3.69748652e-01
1.00514881e-01 7.29633331e-01 -6.45499766e-01 5.23970306e-01
-2.58518398e-01 6.60184801e-01 -6.94421589e-01 2.73842961e-01
1.85363427e-01 -1.07465672e+00 6.43981457e-01 -2.44961217e-01
5.62923372e-01 1.52209863e-01 -9.42475379e-01 1.95165789e+00
-6.30869150e-01 1.80506304e-01 -1.24437653e-01 -1.17178607e+00
6.20605350e-01 6.25888348e-01 7.39635706e-01 -5.74471476e-03
1.72907427e-01 -1.14971280e-01 -1.36579931e-01 -4.34682727e-01
3.43128800e-01 9.75710247e-03 -2.02815518e-01 5.80716133e-01
7.65026867e-01 1.35752961e-01 -3.24768871e-02 4.48997349e-01
1.11969090e+00 6.90241829e-02 4.65396106e-01 -3.02680969e-01
-7.19415620e-02 -4.66427118e-01 4.53692526e-01 1.02414000e+00
1.21442946e-02 4.48593318e-01 7.21892059e-01 -3.84629160e-01
-9.85681117e-01 -1.22357345e+00 1.37809202e-01 9.35996652e-01
-1.92029506e-01 -6.98788941e-01 -7.79022813e-01 -8.89907539e-01
2.01299936e-01 5.76693118e-01 -1.15083551e+00 -3.83300483e-01
-2.49197900e-01 -8.43505859e-01 1.99759081e-01 6.52597010e-01
-2.99597438e-02 -1.12151480e+00 -1.90124825e-01 4.23911244e-01
8.22591558e-02 -1.20135641e+00 -2.08404616e-01 6.04881272e-02
-9.96410787e-01 -9.62444425e-01 -1.02198863e+00 -6.11626267e-01
4.38001871e-01 1.49021327e-01 1.71551132e+00 -1.69624865e-01
-1.00005427e-02 9.11074460e-01 -2.76956439e-01 -4.28320289e-01
-5.98671317e-01 6.11375988e-01 -2.99596667e-01 3.97849977e-01
1.48786128e-01 -7.13541210e-01 -7.27454901e-01 -2.85156965e-01
-1.08784735e+00 2.18943432e-01 5.87884665e-01 1.04751134e+00
6.83037877e-01 -3.64335865e-01 5.41836381e-01 -1.64425457e+00
9.37293887e-01 -1.10008931e+00 -1.13034502e-01 3.96974921e-01
-8.49265456e-01 9.52583179e-03 6.78323448e-01 -4.04461026e-01
-7.75858998e-01 -5.12537897e-01 2.57436514e-01 -9.94276166e-01
-2.68375754e-01 7.13771820e-01 -8.38127285e-02 4.89262313e-01
4.78854209e-01 3.66055161e-01 -9.79902223e-02 -3.99736673e-01
5.75075924e-01 2.71243066e-01 1.75535277e-01 -3.62639874e-01
6.71968281e-01 3.80612254e-01 7.47826174e-02 -6.31305695e-01
-8.71676266e-01 -3.34248930e-01 -4.07294095e-01 -2.34718062e-03
1.29812562e+00 -1.15269256e+00 -5.80888271e-01 4.33774859e-01
-1.03974164e+00 -3.85718912e-01 -2.42799327e-01 5.70325971e-01
-5.56812465e-01 1.14116482e-01 -9.18770671e-01 -5.18127799e-01
-4.89994556e-01 -1.20928609e+00 1.47663438e+00 1.76686063e-01
-6.06069043e-02 -1.61026955e+00 3.74656707e-01 1.92459658e-01
4.37692106e-01 4.15974647e-01 1.19565940e+00 -9.67690587e-01
-7.11685658e-01 9.80722066e-03 7.73328841e-02 1.94515973e-01
3.12471747e-01 -2.42676780e-01 -9.53377485e-01 -5.63339889e-01
-5.94550557e-02 -1.01000875e-01 1.03203022e+00 8.99017334e-01
1.30016768e+00 -1.62323222e-01 -5.90752780e-01 8.61832500e-01
1.40068221e+00 -1.74668968e-01 4.88731563e-01 -1.15249693e-01
1.39229715e+00 4.16200161e-01 -3.58082771e-01 4.77720469e-01
7.39638269e-01 4.59425360e-01 4.45390582e-01 -2.49607161e-01
-1.60331912e-02 -3.71722490e-01 1.27447754e-01 1.14093661e+00
-2.97930241e-01 -6.99943483e-01 -7.70297348e-01 5.67739844e-01
-2.20218778e+00 -8.48066986e-01 8.45009312e-02 1.97624481e+00
4.23716187e-01 -1.26458883e-01 1.54253021e-01 -5.91272533e-01
5.93688071e-01 5.45159638e-01 -5.94832897e-01 -5.18346168e-02
2.00664520e-01 3.52401286e-01 1.09666340e-01 1.59903184e-01
-1.05478001e+00 8.20767045e-01 6.10366821e+00 5.34459412e-01
-9.35614169e-01 4.07939076e-01 5.45581877e-01 4.22263518e-02
-6.52455747e-01 -3.71773571e-01 -5.09950459e-01 5.38035393e-01
1.08420980e+00 -3.71589452e-01 3.30820799e-01 9.56358910e-01
-1.37351096e-01 7.70771503e-01 -1.27207494e+00 8.50805640e-01
1.83948159e-01 -1.45838428e+00 3.86033952e-01 2.50615358e-01
1.00886965e+00 1.74506485e-01 3.17481101e-01 6.17174804e-01
8.32720459e-01 -9.76847649e-01 4.82374758e-01 6.85099006e-01
2.50200033e-01 -4.61885720e-01 7.09900856e-01 1.21445209e-01
-1.21943128e+00 1.57176405e-01 -2.14003518e-01 5.48294902e-01
1.68782592e-01 6.15306377e-01 -9.13235486e-01 1.11309838e+00
6.03986859e-01 1.19991589e+00 -3.71622443e-01 6.72032714e-01
-3.00868779e-01 7.83276379e-01 4.62156206e-01 1.45915942e-02
2.06282079e-01 -5.26029289e-01 4.88420576e-01 1.21287513e+00
4.41225320e-01 -3.86259943e-01 5.48790693e-01 1.20527112e+00
-3.90997976e-01 3.59395593e-01 -8.45438421e-01 -2.79045224e-01
6.53573945e-02 1.37234652e+00 -4.68077064e-01 -6.12410903e-01
-7.67591655e-01 1.05202389e+00 4.46636707e-01 7.55975604e-01
-8.84164751e-01 8.77302662e-02 5.94874203e-01 -6.47718981e-02
5.85345745e-01 -1.75115675e-01 -5.14950678e-02 -1.47768521e+00
-2.01715097e-01 -5.81030905e-01 5.54351091e-01 -4.60504085e-01
-1.74611104e+00 7.79768646e-01 -1.63885131e-01 -1.12204432e+00
-5.08274257e-01 -3.80979568e-01 -5.34483254e-01 9.87208009e-01
-1.52176034e+00 -1.52842712e+00 -4.86480415e-01 8.81220281e-01
3.73196423e-01 -2.72466004e-01 8.74534190e-01 2.19806880e-01
-8.08255613e-01 4.81645823e-01 1.92242175e-01 6.94828182e-02
5.57838082e-01 -1.37545133e+00 4.15154308e-01 6.42556071e-01
4.62871999e-01 8.95798862e-01 2.84101278e-01 -7.58944094e-01
-1.33435333e+00 -1.53981876e+00 6.77031696e-01 -4.60099787e-01
7.50788331e-01 -5.88568389e-01 -1.03021717e+00 1.07941544e+00
4.19795245e-01 -2.36149907e-01 1.24419522e+00 5.43413162e-01
-3.07106853e-01 2.63579339e-01 -5.85553110e-01 5.79131484e-01
1.08211958e+00 -5.54569483e-01 -3.59341770e-01 5.21315932e-01
1.02107537e+00 -3.57831538e-01 -1.23548007e+00 1.01255432e-01
2.82354295e-01 -7.41592467e-01 8.41919422e-01 -1.23948371e+00
7.40329683e-01 1.76620752e-01 -1.06547929e-01 -1.64980173e+00
-8.75955641e-01 -6.12508476e-01 -7.42663205e-01 8.05616021e-01
3.46799493e-01 -6.42265677e-01 5.72153747e-01 3.30261618e-01
-4.18547451e-01 -8.34899604e-01 -6.33346915e-01 -4.35829580e-01
-4.54460345e-02 -1.47077024e-01 7.33938158e-01 1.04725051e+00
-4.89164352e-01 6.49520159e-01 -6.19246066e-01 1.67122439e-01
6.18835807e-01 4.05490667e-01 7.28633225e-01 -1.78739071e+00
-7.93830037e-01 -4.79656249e-01 -4.07874912e-01 -6.60911500e-01
4.60556209e-01 -1.16138041e+00 -3.79906744e-01 -1.91920078e+00
5.45552254e-01 -1.54421061e-01 -6.74278975e-01 4.89970982e-01
-7.14444458e-01 -3.54222991e-02 1.91554353e-01 3.05572659e-01
-6.59681559e-01 7.41526186e-01 1.24539852e+00 -2.62313068e-01
-3.07976186e-01 3.83764021e-02 -9.65557516e-01 3.90804976e-01
4.67664689e-01 -4.27072287e-01 -6.82335079e-01 -5.88575542e-01
1.64670572e-01 7.38562942e-02 3.01188916e-01 -8.87030542e-01
2.39473674e-02 -4.62255590e-02 1.69719025e-01 -2.58618817e-02
9.60207731e-02 -6.15367651e-01 5.16863704e-01 4.39403892e-01
-3.44659597e-01 5.22166416e-02 -1.04769714e-01 1.25214195e+00
-1.42119139e-01 1.17774688e-01 3.34791422e-01 -2.52745152e-01
-3.59510243e-01 9.52698708e-01 -2.20561568e-02 1.64964899e-01
1.05903292e+00 4.13927555e-01 -3.15049767e-01 -4.47268009e-01
-1.11797345e+00 3.55954189e-03 3.87170643e-01 6.04937673e-01
3.84445876e-01 -1.28255808e+00 -9.23805475e-01 4.09247428e-02
3.11632335e-01 1.22348443e-01 6.98428512e-01 7.78395593e-01
-8.39153081e-02 3.00406933e-01 -7.49132782e-03 -7.68211722e-01
-7.12213635e-01 8.54967177e-01 1.67262301e-01 -9.81186032e-01
-7.95159638e-01 4.82108742e-01 7.14239657e-01 -1.73692778e-01
-8.11864510e-02 -2.89612859e-01 -3.91852587e-01 2.03359395e-01
5.27189560e-02 1.73074394e-01 1.80231988e-01 -2.34884277e-01
8.22268128e-02 2.51855791e-01 -1.14404798e-01 2.28156149e-01
1.46638799e+00 6.94610700e-02 6.27389774e-02 7.26163030e-01
1.13111997e+00 -2.18599468e-01 -1.03670800e+00 -4.58801359e-01
-5.33572793e-01 1.71562687e-01 1.43472627e-01 -2.81566441e-01
-1.32592475e+00 8.27856481e-01 1.36729002e-01 4.91653919e-01
8.09436321e-01 3.25583816e-01 7.18350053e-01 1.39785796e-01
3.88060838e-01 -7.07829475e-01 1.36426896e-01 1.47470608e-01
5.96173108e-01 -1.07769692e+00 -1.44024387e-01 -2.20713884e-01
-9.30184305e-01 8.25408936e-01 2.71970630e-01 -2.43720293e-01
8.28364611e-01 -8.27098042e-02 -2.78496265e-01 -4.83970165e-01
-8.59060526e-01 -5.46527430e-02 6.35510206e-01 5.48360586e-01
4.71597463e-01 5.54364860e-01 -2.65538931e-01 9.77421165e-01
1.59684345e-01 4.02177610e-02 3.35487515e-01 6.23687387e-01
1.48564920e-01 -1.10877621e+00 2.79949963e-01 7.84811854e-01
-3.84145319e-01 -1.73722103e-01 -1.20689772e-01 6.93776309e-01
-1.37484461e-01 6.29952788e-01 -4.58460227e-02 -4.72208142e-01
1.61789626e-01 4.14759994e-01 2.84996063e-01 -9.90521193e-01
-6.51265323e-01 2.23127946e-01 -4.06018466e-01 -5.26467860e-01
-3.64104509e-01 -4.98758882e-01 -5.87318778e-01 -1.79037068e-03
1.60673752e-01 1.06979370e-01 2.74815947e-01 7.85337090e-01
8.86691630e-01 1.31179059e+00 3.99742782e-01 -7.10812509e-01
-4.55806285e-01 -1.06973290e+00 -9.74529028e-01 8.79730701e-01
7.10643157e-02 -8.54104340e-01 -1.82681143e-01 1.38229579e-01]
|
[7.347444534301758, 6.202232360839844]
|
32dbef74-9982-4dbe-a49a-e9f17c77ad72
|
carl-a-benchmark-for-contextual-and-adaptive
|
2110.02102
| null |
https://arxiv.org/abs/2110.02102v2
|
https://arxiv.org/pdf/2110.02102v2.pdf
|
CARL: A Benchmark for Contextual and Adaptive Reinforcement Learning
|
While Reinforcement Learning has made great strides towards solving ever more complicated tasks, many algorithms are still brittle to even slight changes in their environment. This is a limiting factor for real-world applications of RL. Although the research community continuously aims at improving both robustness and generalization of RL algorithms, unfortunately it still lacks an open-source set of well-defined benchmark problems based on a consistent theoretical framework, which allows comparing different approaches in a fair, reliable and reproducibleway. To fill this gap, we propose CARL, a collection of well-known RL environments extended to contextual RL problems to study generalization. We show the urgent need of such benchmarks by demonstrating that even simple toy environments become challenging for commonly used approaches if different contextual instances of this task have to be considered. Furthermore, CARL allows us to provide first evidence that disentangling representation learning of the states from the policy learning with the context facilitates better generalization. By providing variations of diverse benchmarks from classic control, physical simulations, games and a real-world application of RNA design, CARL will allow the community to derive many more such insights on a solid empirical foundation.
|
['Marius Lindauer', 'Frank Hutter', 'Bodo Rosenhahn', 'André Biedenkapp', 'Frederik Schubert', 'Theresa Eimer', 'Carolin Benjamins']
|
2021-10-05
| null | null | null | null |
['physical-simulations']
|
['miscellaneous']
|
[ 2.05700427e-01 -2.17484906e-01 -3.27553600e-01 3.29212286e-02
-6.91900134e-01 -8.65143478e-01 8.16376030e-01 2.44002789e-02
-5.47342300e-01 1.32180107e+00 -7.97445551e-02 -3.09234411e-01
-3.70100588e-01 -4.79038388e-01 -6.66703939e-01 -1.00282669e+00
-4.03497487e-01 5.38986206e-01 2.59722501e-01 -7.34941065e-01
2.99265772e-01 5.49185812e-01 -1.63720083e+00 -8.79385546e-02
5.80367029e-01 2.89814442e-01 2.39960760e-01 6.38276279e-01
3.18102866e-01 5.43449700e-01 -4.96710032e-01 1.28523722e-01
4.92754042e-01 -8.83526742e-01 -8.48670065e-01 -2.72741467e-01
-4.99042235e-02 1.29625991e-01 -1.32585624e-02 9.92609203e-01
7.79210031e-01 4.43868548e-01 2.84355104e-01 -1.21628428e+00
-3.30700576e-01 4.49534774e-01 -2.53537208e-01 -2.26005097e-03
4.99695271e-01 6.00798488e-01 1.20009911e+00 4.74374257e-02
8.61252010e-01 1.23119307e+00 6.10926628e-01 9.27767217e-01
-1.41680527e+00 -5.51680386e-01 1.50573120e-01 7.99577907e-02
-8.31098497e-01 -2.92085886e-01 3.93596411e-01 -4.29782234e-02
7.91230142e-01 3.20207059e-01 7.40257263e-01 1.35802984e+00
1.10143073e-01 5.20600021e-01 1.63145077e+00 -3.08384717e-01
5.64700007e-01 -2.08314598e-01 -1.26019135e-01 4.27904785e-01
2.16426566e-01 5.81300199e-01 -3.27503115e-01 -1.63952276e-01
6.94686770e-01 -2.90615052e-01 -3.20706785e-01 -8.52587700e-01
-1.28887320e+00 7.08314419e-01 3.14760566e-01 5.03288329e-01
-1.29035085e-01 3.70495707e-01 5.20214796e-01 6.21436954e-01
-1.63175285e-01 9.71904874e-01 -5.84278047e-01 -6.06327534e-01
-4.65954870e-01 7.52455294e-01 9.51876521e-01 7.50639796e-01
6.44126475e-01 -2.68994644e-02 1.74855411e-01 4.02499646e-01
-4.31464426e-02 2.94467151e-01 4.09951717e-01 -1.26798594e+00
3.01698707e-02 2.65675902e-01 4.42514658e-01 -6.45994008e-01
-7.41885483e-01 -5.89597106e-01 -5.26441395e-01 5.77311039e-01
9.64388430e-01 -1.38938040e-01 -4.75204945e-01 2.20946002e+00
2.88602382e-01 1.57647714e-01 2.27159709e-01 9.84694242e-01
6.12846725e-02 5.48345625e-01 -1.41515389e-01 -3.39567840e-01
1.02326465e+00 -5.97443581e-01 -3.59699488e-01 -2.38400489e-01
6.18367970e-01 -5.02607286e-01 1.33646095e+00 7.35198855e-01
-9.89962816e-01 -1.83134496e-01 -1.09487605e+00 2.12483212e-01
-2.95088500e-01 -6.14697158e-01 1.01115990e+00 5.82692266e-01
-9.42050278e-01 9.46286738e-01 -8.07451367e-01 -8.18223238e-01
1.10009678e-01 5.80753446e-01 -4.47524816e-01 -8.92887078e-03
-1.24871254e+00 1.26020670e+00 1.90123856e-01 -9.60078984e-02
-1.00949419e+00 -5.50267518e-01 -4.27406967e-01 -1.91112950e-01
8.31702113e-01 -6.73004091e-01 1.43074572e+00 -8.88745844e-01
-1.85423076e+00 5.41516125e-01 2.39795089e-01 -4.42578614e-01
6.27668858e-01 4.27554883e-02 3.01162228e-02 -2.99563646e-01
-1.62400052e-01 3.38667125e-01 5.03517866e-01 -1.20502114e+00
-3.74194086e-01 -4.14457291e-01 3.76506984e-01 1.22642532e-01
2.42760643e-01 -1.46186218e-01 1.20050229e-01 -3.64362001e-01
-2.82701641e-01 -1.18090844e+00 -6.09983325e-01 -3.38206589e-01
-2.61560887e-01 -8.23955610e-02 5.50468922e-01 1.85087591e-01
7.42671132e-01 -1.78103209e+00 5.81039131e-01 -5.25521673e-02
1.27877057e-01 2.42442206e-01 -3.91008288e-01 9.78732109e-01
-1.24514133e-01 1.33637547e-01 -1.55215502e-01 1.23407423e-01
2.67198920e-01 5.95811546e-01 -3.43807727e-01 6.51792109e-01
-3.84757482e-03 8.71876836e-01 -1.33881915e+00 5.19987866e-02
1.45105183e-01 2.65229076e-01 -7.95479417e-01 4.45450209e-02
-6.19465232e-01 8.10490131e-01 -5.98681509e-01 7.85247833e-02
1.14219651e-01 -1.06551059e-01 4.54842091e-01 3.96938771e-01
-2.26690963e-01 3.93789351e-01 -1.20700347e+00 1.80410492e+00
-3.81976157e-01 4.14744496e-01 -3.15451575e-03 -1.30404019e+00
5.06478310e-01 1.13813475e-01 5.87899268e-01 -7.16241777e-01
1.81335598e-01 1.54411480e-01 4.78243798e-01 -4.34248775e-01
3.22100490e-01 -5.41627228e-01 -2.84116447e-01 6.44967139e-01
-1.32046714e-01 -4.94526953e-01 3.86913747e-01 -6.85174763e-02
1.41569829e+00 4.71862644e-01 5.15599012e-01 -4.37779397e-01
2.85201043e-01 6.77637607e-02 6.31481349e-01 1.10146880e+00
-4.91899282e-01 2.84222841e-01 7.48835206e-01 -5.42219102e-01
-1.11340129e+00 -1.01822281e+00 -2.58304887e-02 1.21371949e+00
7.10870326e-02 -5.04924953e-01 -7.31724441e-01 -4.85701442e-01
3.23745348e-02 6.55754805e-01 -8.28875363e-01 -2.57916301e-01
-5.83006203e-01 -9.97525573e-01 6.20166659e-01 1.38365284e-01
2.08245188e-01 -1.28755093e+00 -9.45937157e-01 2.30223835e-01
2.69656062e-01 -8.81762147e-01 -6.27082288e-02 4.69355285e-01
-6.67957604e-01 -1.29986072e+00 -4.12511945e-01 -4.06922042e-01
5.98694980e-02 1.81888700e-01 1.16438925e+00 3.30733716e-01
-3.93500000e-01 5.47556758e-01 -2.78494090e-01 -2.72632062e-01
-8.47315311e-01 9.15277079e-02 1.79935962e-01 -6.80822968e-01
2.53314041e-02 -8.05754483e-01 -4.10307139e-01 4.04193223e-01
-1.01530671e+00 -1.41924456e-01 4.63512868e-01 1.05825794e+00
2.23968357e-01 -1.25453621e-01 8.40020537e-01 -7.02582955e-01
8.70100021e-01 -4.17847544e-01 -8.22678566e-01 9.22783613e-02
-7.00777352e-01 4.86447990e-01 9.00951028e-01 -4.53965187e-01
-5.43467581e-01 -1.10237263e-01 -8.29763934e-02 3.20807397e-01
-1.75119922e-01 3.70140642e-01 -7.16171637e-02 -3.56500000e-02
8.39482784e-01 3.59424412e-01 2.70712346e-01 -1.61019444e-01
6.50300682e-01 1.20092221e-01 2.19365612e-01 -1.06602645e+00
8.19502771e-01 2.44876459e-01 2.78302968e-01 -8.12014759e-01
-5.42968690e-01 -5.74514344e-02 -3.05980325e-01 6.45405948e-02
5.71841240e-01 -3.11554790e-01 -1.13839507e+00 1.67875290e-01
-7.36776173e-01 -8.61540139e-01 -5.17149806e-01 2.02848017e-01
-1.22617817e+00 3.91841710e-01 -4.92375970e-01 -7.10418165e-01
3.62145901e-01 -1.55739868e+00 7.34166384e-01 1.16003439e-01
-1.59256727e-01 -9.43072915e-01 6.39125466e-01 5.42535400e-03
6.09265685e-01 2.08915979e-01 1.04165483e+00 -6.49584115e-01
-5.06483555e-01 1.22944891e-01 2.76537925e-01 1.20159708e-01
1.66278675e-01 7.67531544e-02 -9.05993760e-01 -4.53925610e-01
-3.32531333e-02 -6.67854786e-01 6.82559788e-01 1.06541082e-01
9.63743687e-01 8.23871419e-03 -1.35161355e-01 2.86906123e-01
1.33567441e+00 3.06941748e-01 7.07284033e-01 5.36149263e-01
1.83116615e-01 4.34119433e-01 5.03707409e-01 4.28067923e-01
2.10511804e-01 8.44770968e-01 4.86998439e-01 6.06344119e-02
2.46491358e-01 -2.94621289e-02 4.30603296e-01 4.68513399e-01
-3.41390699e-01 -2.24953771e-01 -7.85691500e-01 -2.82840915e-02
-2.01396704e+00 -1.31535816e+00 3.61431777e-01 2.46801710e+00
1.08137369e+00 1.80994138e-01 4.72200304e-01 2.03200281e-01
4.51088160e-01 1.26578793e-01 -7.63329268e-01 -3.80068153e-01
-1.06126014e-02 3.51906687e-01 2.24527389e-01 4.76751655e-01
-7.39222646e-01 1.06471598e+00 6.94233418e+00 8.11549306e-01
-1.25006092e+00 -1.17630109e-01 3.59459668e-01 -9.52702090e-02
-5.05223460e-02 1.07847072e-01 -4.12159055e-01 1.81472912e-01
9.27562475e-01 -1.95928246e-01 1.10239649e+00 6.88974500e-01
4.05958563e-01 -2.24330023e-01 -1.32740903e+00 6.70869470e-01
-3.60513151e-01 -1.28045893e+00 -3.92866164e-01 1.19922757e-01
5.68026364e-01 1.49975404e-01 1.39479741e-01 6.18094265e-01
9.23922420e-01 -1.23120391e+00 3.87028098e-01 2.95523494e-01
3.15065831e-01 -6.86913192e-01 3.62093896e-01 5.32033443e-01
-6.11289799e-01 -1.73223704e-01 -4.04203773e-01 -4.66482908e-01
-2.54352599e-01 1.38980657e-01 -7.78166771e-01 4.78874683e-01
3.18186671e-01 6.99838459e-01 -2.45740429e-01 1.02871907e+00
-1.57010987e-01 4.98235852e-01 -2.11997315e-01 -4.35953975e-01
4.91302460e-01 -2.87989825e-01 5.48858047e-01 8.73466015e-01
1.22658981e-04 -1.63979025e-03 2.26438105e-01 7.08917558e-01
1.42599225e-01 6.75631762e-02 -9.74950135e-01 -2.64820576e-01
4.49672677e-02 1.20027471e+00 -8.80877316e-01 7.64009356e-02
-2.13652521e-01 5.75432301e-01 4.98257667e-01 4.25964683e-01
-9.50070918e-01 -8.14494267e-02 1.04827750e+00 -2.16717139e-01
5.92417158e-02 -5.50927043e-01 1.10397190e-01 -1.19340813e+00
-3.20753396e-01 -1.53978658e+00 1.31826907e-01 -5.61462343e-01
-1.17597103e+00 2.88855523e-01 -8.58090669e-02 -9.84452784e-01
-4.46799546e-01 -6.49836540e-01 -3.76864880e-01 4.47428137e-01
-1.39709806e+00 -5.26317894e-01 7.48084858e-02 4.88319963e-01
3.18080693e-01 -6.07338827e-03 9.34767723e-01 -8.27685148e-02
-7.44895995e-01 2.79369831e-01 4.04826850e-01 -3.43044460e-01
8.55818391e-01 -1.37445688e+00 2.97061682e-01 4.40569013e-01
2.88217932e-01 7.95699239e-01 1.19191527e+00 -2.44542614e-01
-2.00343776e+00 -5.39514303e-01 6.17084280e-02 -6.59197330e-01
8.84475827e-01 -6.19326830e-01 -6.37941599e-01 5.40318727e-01
1.40749052e-01 -3.78749669e-02 3.19554061e-01 3.85607690e-01
-3.78508329e-01 1.04305297e-02 -9.14164305e-01 9.73035634e-01
1.22255874e+00 -3.93927932e-01 -5.25441408e-01 4.33821321e-01
5.06385386e-01 -3.65936965e-01 -7.16054618e-01 2.79945731e-01
6.76047146e-01 -1.31469691e+00 9.12546694e-01 -1.08435059e+00
2.09023461e-01 -3.51728201e-01 -1.42482802e-01 -1.55504632e+00
-4.39017043e-02 -1.07425117e+00 2.72767216e-01 8.30640256e-01
3.70939612e-01 -7.68358886e-01 7.70593345e-01 3.02459747e-01
-7.47939758e-03 -9.11236644e-01 -9.16842461e-01 -9.97911870e-01
6.49851561e-01 -6.50959790e-01 4.71077621e-01 9.07337964e-01
3.94999474e-01 3.61816406e-01 -3.59487683e-01 -1.55187786e-01
2.98118860e-01 1.71095103e-01 9.98202860e-01 -1.03929400e+00
-7.24210143e-01 -7.81866789e-01 -5.19589841e-01 -9.04043794e-01
2.22623333e-01 -8.30123007e-01 3.29544216e-01 -1.18527794e+00
2.83604920e-01 -6.97884142e-01 -3.96630436e-01 4.45134819e-01
-9.12550464e-02 5.59617691e-02 2.84174114e-01 -5.95940687e-02
-1.00000322e+00 5.67625523e-01 1.49331427e+00 1.27201349e-01
-1.58332214e-01 2.91284211e-02 -7.50133097e-01 5.62619627e-01
9.76267397e-01 -4.83281046e-01 -6.53291523e-01 3.89298913e-03
5.94655156e-01 1.03463814e-01 2.93749213e-01 -1.00371206e+00
-3.57536860e-02 -6.57478511e-01 -7.85136875e-03 1.53669998e-01
2.05889717e-01 -5.95975876e-01 -6.55271858e-02 8.02925110e-01
-6.65215075e-01 1.11173756e-01 2.96531349e-01 6.32415771e-01
3.56551677e-01 -1.23271830e-01 8.02828252e-01 -3.63382936e-01
-5.07888496e-01 1.50196671e-01 -4.98614788e-01 6.23431683e-01
1.05992401e+00 2.92261858e-02 -4.75523353e-01 -5.38803220e-01
-7.33447790e-01 1.46692052e-01 9.35145080e-01 2.75176138e-01
9.99871939e-02 -7.10063040e-01 -4.50129956e-01 -9.27395932e-03
-6.01786412e-02 -3.67235869e-01 -1.33717895e-01 8.49819541e-01
-3.11607689e-01 3.91270638e-01 -4.55890059e-01 -4.52545911e-01
-9.37930405e-01 7.83992112e-01 5.29333055e-01 -3.78563464e-01
-5.73076367e-01 3.71446669e-01 1.37543857e-01 -5.06370962e-01
1.71474591e-01 -5.88080525e-01 2.35034972e-01 -2.72994101e-01
4.65059370e-01 1.45342097e-01 -5.72479218e-02 -1.86526030e-01
-3.08863819e-01 4.35521692e-01 1.57922238e-01 -2.14939862e-01
1.49062073e+00 8.93815160e-02 1.09681167e-01 5.40426970e-01
6.02498949e-01 1.08830109e-01 -1.55551767e+00 1.38889059e-01
3.23336512e-01 -5.96033186e-02 -2.81441152e-01 -8.67582500e-01
-6.64583862e-01 8.36558342e-01 3.99941981e-01 4.98362988e-01
9.69852388e-01 -1.54694885e-01 3.69533181e-01 9.22065854e-01
8.86941075e-01 -8.73364866e-01 2.07345173e-01 6.40317798e-01
8.28275144e-01 -1.10176218e+00 6.46945462e-02 5.72821610e-02
-5.07554173e-01 1.07221365e+00 2.46604919e-01 -3.19876015e-01
1.84053034e-01 4.36296582e-01 -5.14516383e-02 4.96525206e-02
-1.05123568e+00 -4.09132868e-01 -2.91140854e-01 8.11661482e-01
4.76796627e-01 -1.60708010e-01 -3.27508360e-01 2.63362825e-01
-2.32677490e-01 -2.41131466e-02 7.27662504e-01 1.01490521e+00
-5.74992537e-01 -1.78329527e+00 -1.08727321e-01 3.22967954e-02
-3.25750262e-01 1.35752633e-01 -3.78023565e-01 1.41764975e+00
-1.86779633e-01 6.94395900e-01 -5.76265574e-01 -2.42179349e-01
2.14239523e-01 9.31473225e-02 1.01461530e+00 -5.83720744e-01
-6.87557578e-01 -2.51032710e-01 1.15162812e-01 -8.43127072e-01
-4.98875380e-01 -8.15509319e-01 -1.37006140e+00 -4.30447727e-01
-1.15459770e-01 3.85075271e-01 6.25376225e-01 1.12682474e+00
2.69625455e-01 3.98324102e-01 3.55537415e-01 -9.90134060e-01
-9.87557411e-01 -5.01692235e-01 -5.06405532e-01 4.16082948e-01
3.99182200e-01 -8.92321944e-01 -2.85064399e-01 -3.96534532e-01]
|
[4.039078712463379, 1.7787187099456787]
|
0bf3e59c-1a9e-4f19-a73d-de50db6049b2
|
adaptive-streaming-perception-using-deep
|
2106.05665
| null |
https://arxiv.org/abs/2106.05665v2
|
https://arxiv.org/pdf/2106.05665v2.pdf
|
Learning Runtime Decisions for Adaptive Real-Time Perception
|
Real-time perception requires planned resource utilization. Computational planning in real-time perception is governed by two considerations -- accuracy and latency. There exist run-time decisions (e.g. choice of input resolution) that induce tradeoffs affecting performance on a given hardware, arising from intrinsic (content, e.g. scene clutter) and extrinsic (system, e.g. resource contention) characteristics. Earlier runtime execution frameworks employed rule-based decision algorithms and operated with a fixed algorithm latency budget to balance these concerns, which is sub-optimal and inflexible. We propose Chanakya, a learned approximate execution framework that naturally derives from the streaming perception paradigm, to automatically learn decisions induced by these tradeoffs instead. Chanakya is trained via novel rewards balancing accuracy and latency implicitly, without approximating either objectives. Chanakya simultaneously considers intrinsic and extrinsic context, and predicts decisions in a flexible manner. Chanakya, designed with low overhead in mind, outperforms state-of-the-art static and dynamic execution policies on public datasets on both server GPUs and edge devices.
|
['Aditya Singh', 'Vaibhav Balloli', 'Tanuja Ganu', 'Akshay Nambi', 'Anurag Ghosh']
|
2021-06-10
| null | null | null | null |
['real-time-object-detection']
|
['computer-vision']
|
[ 3.50580871e-01 -1.85693145e-01 -3.26456368e-01 -4.93234128e-01
-5.72155595e-01 -5.84052920e-01 5.79475820e-01 2.20220745e-01
-6.56587362e-01 3.06867301e-01 2.15121627e-01 -4.65446830e-01
-1.43875197e-01 -7.56155133e-01 -7.36615539e-01 -5.97383678e-01
-1.83042347e-01 4.17715788e-01 4.63120788e-01 -4.92458791e-02
5.30896664e-01 2.16475710e-01 -2.02193570e+00 5.16804874e-01
7.33520567e-01 1.24339819e+00 4.47240472e-01 1.31537235e+00
6.74710050e-02 1.09066832e+00 -4.85875189e-01 3.69930901e-02
5.03396988e-01 -6.77930042e-02 -4.94052112e-01 -7.96810687e-02
3.01345378e-01 -4.62240249e-01 4.79907617e-02 7.08973825e-01
4.26638871e-01 -9.02583748e-02 2.90964633e-01 -1.29051137e+00
-2.11867556e-01 6.35507345e-01 -4.73667294e-01 5.31497955e-01
2.48993617e-02 9.21213746e-01 9.53467607e-01 -3.45288634e-01
2.99626917e-01 1.01337910e+00 2.48161271e-01 2.72703648e-01
-1.53234267e+00 -4.34990525e-01 6.44873798e-01 1.83934540e-01
-1.11334848e+00 -8.10010433e-01 2.30847791e-01 -4.96883333e-01
1.45322406e+00 4.02559727e-01 5.90056121e-01 1.29471099e+00
6.94077849e-01 5.98036706e-01 1.27322245e+00 -4.33065653e-01
9.27406311e-01 -6.82129487e-02 -3.75558473e-02 2.94155121e-01
3.31308297e-03 4.81638223e-01 -9.67912018e-01 -3.01851988e-01
7.43723035e-01 -4.73611802e-01 -1.21766761e-01 -3.19788128e-01
-1.17156744e+00 4.24967647e-01 9.84333642e-03 -2.15358660e-01
-3.97719145e-01 5.47101557e-01 7.55422950e-01 3.83172959e-01
-4.34807763e-02 7.25294888e-01 -7.51943827e-01 -7.10049033e-01
-8.77236664e-01 2.98533708e-01 7.43617833e-01 1.07378066e+00
6.77351773e-01 -6.44072741e-02 -5.99285722e-01 4.02880907e-01
9.28667486e-02 4.28396136e-01 4.11244214e-01 -1.23477840e+00
3.70779276e-01 1.43031165e-01 4.21482444e-01 -7.89610803e-01
-6.49236500e-01 -5.35143971e-01 -3.57917994e-01 6.72520459e-01
5.79400122e-01 -1.02244653e-01 -9.65802014e-01 1.89779925e+00
3.63555551e-01 2.46834546e-01 2.17389408e-02 1.40240812e+00
-7.41401613e-02 5.28948188e-01 3.95712793e-01 -1.11749738e-01
1.41987693e+00 -1.05106986e+00 -2.88994431e-01 -4.96327966e-01
4.34643567e-01 -7.82782555e-01 1.56058693e+00 9.03780282e-01
-1.19530416e+00 -5.86694241e-01 -1.16251099e+00 9.00077075e-02
9.23242792e-02 1.98151737e-01 7.79161572e-01 6.46264076e-01
-1.13505065e+00 5.98637104e-01 -1.43339086e+00 -1.77122325e-01
-5.02469167e-02 5.00739157e-01 5.87186635e-01 3.20298851e-01
-5.57148993e-01 6.92506433e-01 3.05293739e-01 -1.29607290e-01
-9.67058063e-01 -7.43022203e-01 -2.22125664e-01 1.77037254e-01
1.13996947e+00 -9.30774629e-01 1.80198944e+00 -1.30183077e+00
-2.04939818e+00 6.21109307e-01 2.30013087e-01 -7.31852293e-01
6.74048662e-01 -3.07554811e-01 -3.04964930e-01 -1.71988785e-01
-9.61311981e-02 5.39315701e-01 1.06464171e+00 -9.89483833e-01
-8.12373340e-01 -3.43845099e-01 5.14337003e-01 4.70881075e-01
3.76321040e-02 -1.61968186e-01 -4.51901793e-01 -2.15553999e-01
-2.42149562e-01 -1.04962015e+00 -5.58342457e-01 -1.41377121e-01
-2.32261345e-01 2.09549423e-02 2.61241555e-01 -3.52597125e-02
1.17070913e+00 -1.98271263e+00 -4.74785939e-02 1.20984428e-01
1.48384050e-01 -3.79573703e-02 -8.92440826e-02 1.95539549e-01
6.60078764e-01 -2.42211670e-01 2.49262527e-01 -2.30352014e-01
1.25626385e-01 3.09281051e-01 -4.35193866e-01 1.45956039e-01
1.09579585e-01 3.59578222e-01 -1.12809026e+00 -2.99946129e-01
1.94287166e-01 7.89386570e-04 -9.65968311e-01 3.08538139e-01
-8.18866372e-01 3.58937085e-01 -5.51896870e-01 6.51634216e-01
4.87032086e-01 -4.05399948e-01 4.14342880e-01 -1.32978231e-01
-6.20042861e-01 5.03669918e-01 -1.19880867e+00 1.81507814e+00
-9.80862916e-01 3.95428419e-01 2.52286136e-01 -3.41551721e-01
5.15950024e-01 -1.07523300e-01 4.94592004e-02 -1.10136223e+00
2.92220592e-01 3.15944016e-01 2.24271268e-01 -4.51624304e-01
9.52152550e-01 5.09295046e-01 5.68068847e-02 3.22631210e-01
-3.26395065e-01 6.23557642e-02 -3.58499624e-02 1.24786280e-01
1.50195634e+00 5.34383237e-01 4.31344539e-01 -5.72878897e-01
-8.51078033e-02 4.35536057e-01 7.12199390e-01 1.19132411e+00
-2.70361871e-01 3.57800364e-01 6.17050469e-01 -6.59772813e-01
-1.17034435e+00 -8.69584680e-01 -1.07399084e-01 1.76267922e+00
4.25028086e-01 -6.56907082e-01 -7.52923012e-01 -2.49234721e-01
-9.61183906e-02 1.03873384e+00 -6.53951883e-01 -1.33982152e-01
-5.06966352e-01 -5.06718934e-01 1.24592938e-01 5.88718235e-01
1.39093280e-01 -9.77683604e-01 -1.93823588e+00 6.68779075e-01
3.08366865e-01 -1.29635286e+00 -6.00366950e-01 5.17685235e-01
-5.88579416e-01 -7.67069817e-01 1.88587800e-01 3.59025449e-01
2.90166199e-01 4.14779723e-01 1.67379057e+00 5.54924877e-03
-4.76698160e-01 5.48793852e-01 -3.51346135e-01 -4.14521784e-01
-3.12747478e-01 5.27164899e-02 1.82323009e-01 -2.02918723e-01
1.21413976e-01 -7.00258017e-01 -1.00135040e+00 3.29492390e-01
-7.55007923e-01 6.19355023e-01 6.57775998e-01 8.24303508e-01
8.56396735e-01 -1.36662826e-01 -7.85077736e-03 -9.50120866e-01
5.19964755e-01 -4.01272625e-01 -1.10717702e+00 2.91926563e-01
-9.68492806e-01 2.87880868e-01 8.96323860e-01 -7.66760230e-01
-1.03582096e+00 -3.24709644e-03 2.55573750e-01 -5.96895754e-01
-3.29058915e-01 8.35723728e-02 -9.13551152e-02 2.18636423e-01
9.03933942e-01 7.30805025e-02 -5.20514429e-01 1.17866080e-02
3.24781895e-01 4.31879431e-01 5.72845161e-01 -1.30505729e+00
-1.22626103e-01 4.26422328e-01 -8.41312930e-02 -5.13003826e-01
-8.59401226e-01 -2.96536267e-01 -1.49814218e-01 -2.69788504e-01
7.06166267e-01 -8.82105470e-01 -1.08843470e+00 2.97351658e-01
-9.72724795e-01 -9.00516450e-01 -3.46255779e-01 2.46508345e-01
-8.87914896e-01 -4.54647750e-01 -3.30782235e-01 -9.28849399e-01
-2.56494552e-01 -1.51133609e+00 1.29732442e+00 2.30671763e-01
-2.10085720e-01 -4.02873933e-01 -1.52371880e-02 6.92644492e-02
7.55306542e-01 1.88544780e-01 5.58256924e-01 -2.28472620e-01
-1.07159436e+00 5.74710369e-01 -4.02085811e-01 -1.43818080e-01
-3.95876020e-01 1.71232089e-01 -1.17367899e+00 -5.12513638e-01
-2.01109629e-02 -3.67508769e-01 5.20400167e-01 4.11277711e-01
1.51617312e+00 -5.79553604e-01 -1.24013484e-01 8.65732789e-01
1.67018640e+00 2.46867105e-01 3.62471700e-01 6.07126176e-01
4.15162325e-01 3.12114567e-01 8.19284737e-01 1.03898466e+00
5.74994981e-01 1.02548575e+00 7.95614779e-01 3.28650951e-01
1.37377784e-01 3.04024667e-02 4.24487889e-01 3.31548393e-01
-1.65964574e-01 -2.42187321e-01 -1.09837568e+00 2.57537454e-01
-2.07839799e+00 -6.88700080e-01 1.68816954e-01 2.50006557e+00
7.50170946e-01 7.78430343e-01 2.74916500e-01 -2.64105499e-01
-1.63140818e-02 2.42218655e-02 -1.23858845e+00 -9.08950686e-01
2.34162733e-01 3.32984626e-02 1.03212130e+00 2.89113343e-01
-6.81924522e-01 9.60753560e-01 5.78548050e+00 7.70824730e-01
-1.36947536e+00 1.17587954e-01 8.21184754e-01 -6.26152873e-01
-2.87552267e-01 2.14799568e-01 -8.47289681e-01 5.79674959e-01
1.45669317e+00 -1.05080463e-01 8.50657403e-01 1.04836261e+00
4.25467670e-01 -4.01639760e-01 -1.41308486e+00 9.17265475e-01
-4.84518141e-01 -1.35562575e+00 -3.58272195e-01 2.73108453e-01
4.10229921e-01 2.72440016e-01 1.61726028e-01 3.80314916e-01
6.36535645e-01 -9.64476585e-01 1.40426660e+00 4.59827691e-01
7.93150067e-01 -6.65816903e-01 3.29269141e-01 4.31730449e-01
-1.05109489e+00 -2.90996760e-01 -2.08535388e-01 -4.49179828e-01
-9.91514027e-02 6.35865629e-01 -6.82583451e-01 3.05320434e-02
7.99872100e-01 -2.83276271e-02 -2.86869943e-01 7.30835915e-01
1.26369044e-01 6.84040666e-01 -2.70158499e-01 -1.78504005e-01
4.79080468e-01 -2.93658599e-02 5.84667087e-01 1.39520288e+00
2.02552062e-02 4.56334829e-01 6.96615338e-01 7.85721540e-01
4.96326238e-01 -1.28471032e-01 -7.03189000e-02 1.96539521e-01
7.73181856e-01 1.13252413e+00 -9.18855369e-01 -3.08791071e-01
-2.56649196e-01 8.21137846e-01 4.64523017e-01 1.29410297e-01
-9.96432781e-01 3.44977021e-01 1.14217126e+00 3.21181595e-01
9.44575071e-02 -5.01027405e-01 -5.83428383e-01 -1.04885340e+00
-7.44966865e-02 -1.05680323e+00 2.11268112e-01 -4.96920079e-01
-8.51045549e-01 7.27873087e-01 -7.13204443e-02 -1.18922508e+00
-1.73875704e-01 -6.20173693e-01 -4.10578012e-01 6.18373275e-01
-1.46351373e+00 -8.65606606e-01 -4.51868355e-01 1.95940420e-01
7.69564986e-01 1.81300625e-01 6.74582839e-01 -1.13277487e-01
-5.72246075e-01 6.54705405e-01 -2.39842430e-01 -7.57654488e-01
6.03034496e-01 -1.35114992e+00 6.26020551e-01 7.48114765e-01
-2.22147852e-01 3.09043258e-01 1.13677502e+00 -2.71411598e-01
-1.82848966e+00 -7.68909276e-01 6.89716190e-02 -3.05314541e-01
6.24265194e-01 -3.51974130e-01 -6.80730343e-01 2.72233784e-01
7.14563578e-02 7.01419637e-02 5.59040070e-01 3.09923202e-01
-4.49203432e-01 -8.56615901e-02 -9.50207651e-01 9.38806891e-01
1.29101217e+00 -2.43947357e-01 1.71636820e-01 2.31541485e-01
9.49355781e-01 -1.06964052e+00 -5.54501712e-01 1.79441139e-01
7.61578202e-01 -1.41224170e+00 7.39257932e-01 -5.05022168e-01
3.03356379e-01 -2.99706757e-01 -5.04459679e-01 -1.04363847e+00
-4.96401310e-01 -6.99184060e-01 -3.52068752e-01 7.33865142e-01
1.85743406e-01 -4.19748574e-01 7.87796736e-01 8.26750696e-01
-2.81226188e-01 -1.06859648e+00 -7.89110482e-01 -8.11493456e-01
-4.88772541e-01 -7.46532440e-01 9.48232174e-01 4.28553432e-01
-3.82479161e-01 2.56724447e-01 -2.54598767e-01 5.46481311e-01
5.97336352e-01 2.56604910e-01 8.93779397e-01 -6.62703633e-01
-9.08521056e-01 -6.31169379e-01 -1.37155324e-01 -1.11539388e+00
-3.14166158e-01 -2.14370504e-01 2.54260659e-01 -7.42610335e-01
-2.15573218e-02 -9.85510826e-01 -4.95143682e-01 4.02692974e-01
-8.18607509e-02 -4.73459512e-01 3.65499258e-01 3.92116338e-01
-8.72173309e-01 3.99963349e-01 9.99783754e-01 3.38468105e-01
-6.18196428e-01 -3.67999300e-02 -7.26006031e-01 6.79432929e-01
7.93057501e-01 -1.46397784e-01 -5.64734638e-01 -9.27485585e-01
4.09754455e-01 4.50710237e-01 1.95126668e-01 -1.35962057e+00
4.13846999e-01 -8.75870168e-01 2.66574733e-02 -2.46169046e-01
3.55198950e-01 -7.85744905e-01 4.58687395e-02 2.75118977e-01
-5.47592700e-01 3.30895364e-01 2.38607064e-01 7.10463345e-01
3.98579925e-01 2.70184457e-01 7.41436958e-01 -2.89286599e-02
-1.02803433e+00 2.80455351e-01 -2.85134405e-01 1.23802766e-01
1.05831027e+00 -3.36085081e-01 -4.85409498e-01 1.07899681e-01
-5.31485379e-01 4.62970048e-01 6.85664952e-01 3.74885589e-01
2.53135592e-01 -7.78678894e-01 -6.12664402e-01 2.20762938e-02
2.79294372e-01 6.13669585e-03 3.31393301e-01 6.15789890e-01
-4.38339591e-01 1.72992602e-01 -3.36328328e-01 -8.85086179e-01
-9.87901747e-01 5.82230330e-01 2.52433687e-01 -5.90655804e-01
-5.15664995e-01 7.94945538e-01 1.43031046e-01 -5.41253798e-02
2.31394589e-01 -6.51853919e-01 3.03393275e-01 -3.82654279e-01
6.21764421e-01 4.03672814e-01 2.63165891e-01 3.89231257e-02
-3.70026082e-01 1.25070333e-01 -1.46596760e-01 -2.34852716e-01
8.76891792e-01 -2.16896668e-01 3.68800700e-01 4.47150141e-01
3.93145055e-01 -8.84813294e-02 -2.00284648e+00 -1.72263682e-01
-4.32964154e-02 -7.77164042e-01 4.86363053e-01 -1.23199713e+00
-5.98293900e-01 4.29745555e-01 7.43888140e-01 1.20853528e-01
1.50656533e+00 -3.98597479e-01 6.83021069e-01 9.96142328e-02
9.82076168e-01 -1.38672280e+00 9.08914283e-02 3.82208884e-01
5.28710306e-01 -1.14727247e+00 -1.22768365e-01 -2.86186486e-01
-6.29139066e-01 1.07115662e+00 1.19386232e+00 -9.96428952e-02
1.49647996e-01 9.17127967e-01 5.17008603e-02 8.84703025e-02
-1.63416803e+00 -6.72781467e-02 -2.04148609e-02 3.61793041e-01
2.20772937e-01 7.12947309e-01 -7.82602653e-02 5.65048218e-01
-3.39281946e-01 1.34203043e-02 5.52406490e-01 1.13167393e+00
-4.30728078e-01 -9.37112868e-01 -5.38703024e-01 5.79138458e-01
-1.49489909e-01 -2.51081381e-02 2.22607702e-01 4.96439576e-01
3.88799906e-01 9.12910044e-01 4.62597966e-01 -5.55623829e-01
2.38930821e-01 -6.92994595e-01 5.02385974e-01 -6.78618968e-01
-8.08185101e-01 3.95664200e-02 1.69192836e-01 -1.22102416e+00
6.33396730e-02 -6.97180986e-01 -1.19171476e+00 -3.50096554e-01
-1.12845577e-01 -1.87006593e-01 8.96651804e-01 7.73124397e-01
8.55007708e-01 9.89284992e-01 4.31539744e-01 -1.10106373e+00
-9.69677866e-01 -3.29751998e-01 -2.51820683e-01 -6.48132786e-02
3.33261073e-01 -5.45897067e-01 -1.20400945e-02 -1.92131609e-01]
|
[5.235624313354492, 2.93346905708313]
|
c7c5252f-0bdc-47e6-a4fa-05f556c702ae
|
multi-scale-local-temporal-similarity-fusion
|
2107.12762
| null |
https://arxiv.org/abs/2107.12762v1
|
https://arxiv.org/pdf/2107.12762v1.pdf
|
Multi-Scale Local-Temporal Similarity Fusion for Continuous Sign Language Recognition
|
Continuous sign language recognition (cSLR) is a public significant task that transcribes a sign language video into an ordered gloss sequence. It is important to capture the fine-grained gloss-level details, since there is no explicit alignment between sign video frames and the corresponding glosses. Among the past works, one promising way is to adopt a one-dimensional convolutional network (1D-CNN) to temporally fuse the sequential frames. However, CNNs are agnostic to similarity or dissimilarity, and thus are unable to capture local consistent semantics within temporally neighboring frames. To address the issue, we propose to adaptively fuse local features via temporal similarity for this task. Specifically, we devise a Multi-scale Local-Temporal Similarity Fusion Network (mLTSF-Net) as follows: 1) In terms of a specific video frame, we firstly select its similar neighbours with multi-scale receptive regions to accommodate different lengths of glosses. 2) To ensure temporal consistency, we then use position-aware convolution to temporally convolve each scale of selected frames. 3) To obtain a local-temporally enhanced frame-wise representation, we finally fuse the results of different scales using a content-dependent aggregator. We train our model in an end-to-end fashion, and the experimental results on RWTH-PHOENIX-Weather 2014 datasets (RWTH) demonstrate that our model achieves competitive performance compared with several state-of-the-art models.
|
['Xiaohui Hu', 'Bin Wang', 'Jianwei Cui', 'Mengyi Zhao', 'Yao Du', 'Zhi Cui', 'Pan Xie']
|
2021-07-27
| null | null | null | null |
['sign-language-recognition']
|
['computer-vision']
|
[ 9.72395539e-02 -8.20320189e-01 -1.45468712e-01 -5.85527956e-01
-7.53700018e-01 -3.20228159e-01 4.85319704e-01 -5.68683207e-01
-6.61878288e-01 1.77224323e-01 7.90352881e-01 8.88995528e-02
-6.16755597e-02 -5.09827197e-01 -5.57360411e-01 -8.43606055e-01
-1.67960301e-02 -5.15485220e-02 6.74585879e-01 -1.75762728e-01
6.15918823e-02 4.04815316e-01 -1.72899449e+00 6.89233303e-01
8.44909132e-01 1.12524521e+00 2.90381581e-01 5.88795960e-01
-1.42231673e-01 9.36876774e-01 -3.32549423e-01 -2.31307730e-01
4.10079151e-01 -5.26906967e-01 -5.42602539e-01 2.52029449e-01
1.01428115e+00 -7.94711530e-01 -6.61068559e-01 1.04888797e+00
5.57640731e-01 4.22038049e-01 4.87312794e-01 -9.77254629e-01
-7.22236693e-01 2.08126470e-01 -3.67916912e-01 2.83126235e-01
1.40380755e-01 5.28346896e-01 1.01078081e+00 -9.14328516e-01
8.24625969e-01 1.40870225e+00 4.36189294e-01 5.22273779e-01
-6.99263215e-01 -7.50038326e-01 6.53843462e-01 4.29666251e-01
-1.17280424e+00 -3.77559453e-01 6.90166295e-01 -9.91613790e-02
7.93300629e-01 -1.20500252e-02 9.29166436e-01 1.16331816e+00
1.39198601e-01 1.20041740e+00 1.14211488e+00 -1.23069711e-01
9.94272456e-02 -8.57775986e-01 -6.89685866e-02 7.73774624e-01
-2.07372010e-01 9.42965895e-02 -6.77018642e-01 1.76251248e-01
8.89678538e-01 3.31895262e-01 -2.87292629e-01 -2.21410453e-01
-1.40596378e+00 5.86275041e-01 6.64754331e-01 4.72483635e-01
-5.89122117e-01 3.90901119e-01 4.04425085e-01 3.77843112e-01
3.05010676e-01 -2.43658110e-01 -2.98115283e-01 -9.96033847e-02
-9.84750032e-01 2.39641935e-01 2.87162930e-01 6.51210010e-01
4.78500277e-01 -5.14682494e-02 -3.80413949e-01 8.13349605e-01
4.22900945e-01 4.23602670e-01 6.50716066e-01 -1.01983607e+00
5.04722297e-01 2.58828193e-01 -1.94651052e-01 -5.78160107e-01
-1.74032420e-01 -9.75553319e-02 -8.12149704e-01 3.92375141e-01
4.38341618e-01 1.06316485e-01 -1.60264468e+00 1.62911868e+00
1.00940771e-01 8.00634205e-01 -3.86497355e-03 1.46792102e+00
9.45208311e-01 4.95024443e-01 4.73642975e-01 1.95506681e-02
1.37598991e+00 -1.06784141e+00 -6.29515707e-01 3.72776203e-02
2.18495667e-01 -8.21361601e-01 1.01603770e+00 1.56028926e-01
-1.04995084e+00 -5.97736239e-01 -7.19246030e-01 -3.03765804e-01
-4.13360804e-01 1.80821314e-01 4.46570158e-01 2.64219623e-02
-1.06651831e+00 3.77324879e-01 -1.09912574e+00 -4.74381387e-01
3.33329856e-01 3.43200676e-02 -3.38115901e-01 -3.35253179e-01
-1.07312596e+00 7.38992512e-01 2.30656400e-01 3.89824003e-01
-6.80266321e-01 -4.05784428e-01 -9.57128525e-01 -2.29571879e-01
2.28691101e-01 -7.91038990e-01 1.14331090e+00 -9.87060130e-01
-1.68632936e+00 7.22813308e-01 -5.93650758e-01 -4.69629198e-01
6.62892163e-01 -1.58456802e-01 -4.92314249e-01 2.49841288e-01
-6.25386536e-02 8.08376610e-01 1.08192599e+00 -1.00527561e+00
-1.15741420e+00 -6.04575016e-02 1.74485654e-01 4.06346887e-01
-9.20479223e-02 5.50574481e-01 -8.70317519e-01 -1.07514775e+00
3.65492165e-01 -8.67977262e-01 -1.15536906e-01 2.02530459e-01
1.55294433e-01 -4.54313844e-01 1.04391384e+00 -8.71608377e-01
1.17911434e+00 -2.18805766e+00 2.66656578e-01 1.43465295e-01
1.09519228e-01 5.36856532e-01 -3.81261975e-01 4.43607643e-02
2.24512056e-01 -2.94243753e-01 -3.80859405e-01 -6.04841053e-01
3.33876163e-02 4.39213336e-01 -3.73914093e-01 4.35646147e-01
1.55100048e-01 1.03004336e+00 -1.05640030e+00 -5.46474040e-01
6.63591266e-01 7.26202428e-01 -5.94814718e-01 1.31235331e-01
-2.31707573e-01 4.29801047e-01 -5.16573548e-01 9.96401608e-01
5.03643751e-01 -1.97532311e-01 -4.38614562e-02 -5.54302394e-01
-2.55510151e-01 2.61629909e-01 -1.05526698e+00 2.02584648e+00
-3.78030807e-01 7.04656720e-01 -1.81571394e-01 -7.33429134e-01
7.03690469e-01 1.70571178e-01 6.80010557e-01 -8.30870986e-01
1.19701102e-01 4.12661284e-01 -1.32985279e-01 -6.03996158e-01
2.87049174e-01 4.39724000e-03 -9.61767584e-02 2.02981621e-01
1.29324660e-01 1.20933622e-01 3.07405114e-01 -1.14203215e-01
1.09019899e+00 3.07841539e-01 1.10173076e-01 2.48563543e-01
5.91202497e-01 -4.39358056e-01 7.27585673e-01 5.46753764e-01
-6.25638366e-01 9.25852060e-01 -4.81273159e-02 -8.10423791e-01
-7.48812437e-01 -1.30267191e+00 1.25682831e-01 1.22204971e+00
3.23768407e-01 -1.65699676e-01 -3.25726897e-01 -7.77683258e-01
-8.78059864e-03 3.02696317e-01 -5.31987607e-01 7.63558745e-02
-1.11864710e+00 -3.58348787e-01 4.19350564e-01 6.75916851e-01
1.02649069e+00 -1.35769463e+00 -7.84639597e-01 3.10769469e-01
-4.91748154e-01 -1.19094729e+00 -1.08798504e+00 -1.46797910e-01
-4.86426830e-01 -7.86160290e-01 -1.09942150e+00 -1.16562116e+00
3.89690429e-01 3.58207494e-01 6.88784957e-01 -2.94064377e-02
-7.36293569e-02 1.77785665e-01 -6.82371795e-01 -2.89602997e-03
-2.29025278e-02 -3.82338852e-01 -5.15088737e-02 3.74043316e-01
5.68745077e-01 -5.47870398e-01 -9.70874608e-01 3.36188227e-01
-1.04925096e+00 -2.64473110e-02 7.30516076e-01 8.59837949e-01
8.17924559e-01 -2.57763296e-01 3.26575577e-01 -1.81647111e-02
5.89860201e-01 3.29700522e-02 -4.73095834e-01 5.89577198e-01
-9.01971161e-02 7.50934258e-02 5.40653765e-01 -5.80028713e-01
-1.05002677e+00 9.98589247e-02 -1.82164386e-01 -8.75151753e-01
-2.80701578e-01 3.69489312e-01 -3.19752507e-02 -9.54515338e-02
9.03334990e-02 5.43588459e-01 -3.13725621e-02 -4.78632927e-01
6.07344329e-01 5.76592505e-01 9.18715477e-01 -4.08525079e-01
5.65057635e-01 6.93756580e-01 -3.38429272e-01 -4.94211197e-01
-8.14045906e-01 -6.12329185e-01 -6.57459378e-01 -4.20578778e-01
1.27509034e+00 -9.63113606e-01 -6.74475312e-01 7.44224012e-01
-9.99901175e-01 -4.25430089e-01 -2.91869015e-01 7.42606103e-01
-6.24492645e-01 5.63657343e-01 -6.94229960e-01 -4.76222962e-01
-2.11548984e-01 -1.26015675e+00 1.43571889e+00 3.74391258e-01
-8.80122483e-02 -5.57414174e-01 1.92674045e-02 3.35580230e-01
4.26838040e-01 2.46289566e-01 3.83521408e-01 -6.76129907e-02
-7.02606380e-01 1.04614221e-01 -4.15090501e-01 5.34326851e-01
4.27876174e-01 -1.91626698e-02 -6.85745835e-01 -3.50220621e-01
-4.17743891e-01 -2.75042593e-01 1.35806108e+00 6.67933166e-01
1.18245947e+00 -1.93260089e-01 6.75198659e-02 8.86775136e-01
1.11464620e+00 1.77253962e-01 5.78441858e-01 2.80635864e-01
6.15590215e-01 2.91622043e-01 6.05292916e-01 2.14905456e-01
6.14179492e-01 7.11957097e-01 1.81113437e-01 -1.18750215e-01
-7.09354341e-01 -3.19726318e-01 5.86943269e-01 8.14589918e-01
-4.47328895e-01 -1.55089825e-01 -4.73678797e-01 6.31121814e-01
-2.16406417e+00 -1.28121591e+00 4.13176715e-01 1.96567500e+00
7.01332271e-01 -1.28304094e-01 2.20515266e-01 -1.21382758e-01
5.39140999e-01 6.14946842e-01 -6.08313203e-01 -5.69768511e-02
-4.49446499e-01 2.04965934e-01 4.55810905e-01 3.11402559e-01
-1.35323513e+00 1.24342132e+00 5.62220669e+00 8.16547692e-01
-1.38621676e+00 -8.93110875e-03 2.10968971e-01 -4.04822826e-01
1.08375154e-01 -1.45236894e-01 -5.25214136e-01 6.47362769e-01
4.75373775e-01 1.68672398e-01 5.41236639e-01 4.78498697e-01
4.83693242e-01 8.29835385e-02 -7.80323863e-01 9.15030658e-01
2.54493356e-01 -1.29128075e+00 2.79221117e-01 -1.97546761e-02
8.81344140e-01 4.26395059e-01 4.91245352e-02 2.57250190e-01
5.47570705e-01 -7.00198829e-01 9.10408676e-01 8.31568420e-01
7.38365829e-01 -6.03673279e-01 7.33931601e-01 -3.41102779e-02
-1.82870674e+00 -1.78196758e-01 -1.19021617e-01 2.59184688e-01
5.26059031e-01 7.38690719e-02 -2.08970919e-01 5.04744351e-01
1.09714246e+00 1.05693996e+00 -3.42440069e-01 1.25289285e+00
-2.25584045e-01 4.79776084e-01 -4.72681880e-01 1.92394108e-01
6.36934698e-01 -1.39966652e-01 3.98918450e-01 1.30444479e+00
4.36831206e-01 3.81515563e-01 3.34262639e-01 4.29056764e-01
-1.44032976e-02 -1.72153547e-01 -2.86119461e-01 2.87476599e-01
1.88281596e-01 7.69748569e-01 -5.78704953e-01 -6.04001701e-01
-7.53544688e-01 1.32845080e+00 4.39761430e-02 6.35128081e-01
-7.64341295e-01 -1.85313269e-01 1.14600706e+00 -2.80820638e-01
7.15688884e-01 -3.99531603e-01 2.71023046e-02 -1.45990300e+00
3.11292529e-01 -8.52662623e-01 7.21279144e-01 -8.23023856e-01
-1.53850126e+00 5.11832714e-01 -2.14776233e-01 -1.79332912e+00
-3.04507315e-01 -5.09959221e-01 -5.70069313e-01 7.76023388e-01
-1.89181292e+00 -1.54568398e+00 -3.68170083e-01 1.15307009e+00
8.44351947e-01 6.44395649e-02 4.27763939e-01 3.86548847e-01
-1.07479438e-01 6.99722409e-01 -5.23535945e-02 4.84886885e-01
9.31557178e-01 -9.65957046e-01 3.43850315e-01 1.16347980e+00
2.74954289e-01 5.19331455e-01 3.11869681e-01 -7.00076759e-01
-1.02380621e+00 -1.30625117e+00 1.20057225e+00 -2.61423737e-01
6.91224575e-01 8.45646486e-02 -8.48693252e-01 5.92645645e-01
6.98985159e-02 4.92464244e-01 2.13055566e-01 -4.04800832e-01
-6.30253911e-01 -1.89724997e-01 -8.88721466e-01 8.26708317e-01
1.52181816e+00 -7.89837480e-01 -8.81822228e-01 1.38225734e-01
6.49843097e-01 -4.91306126e-01 -7.21040785e-01 6.34077072e-01
1.00928950e+00 -9.77409542e-01 1.07206571e+00 -4.63289261e-01
2.75080562e-01 -7.35330462e-01 -3.60745043e-01 -1.07057095e+00
-2.49568090e-01 -5.52466512e-01 -6.15400597e-02 8.62438619e-01
-1.51016086e-01 -5.19163787e-01 5.86836040e-01 4.71225411e-01
-2.73914814e-01 -6.78653717e-01 -1.16587257e+00 -9.03220236e-01
-1.76632538e-01 -5.91709673e-01 6.64214492e-01 5.48079431e-01
-2.96211630e-01 -2.43940234e-01 -5.58909178e-01 2.15106145e-01
6.41855955e-01 4.84934181e-01 6.58856273e-01 -8.66314471e-01
-4.98269089e-02 -8.00763190e-01 -6.36984289e-01 -1.51894605e+00
2.02988848e-01 -6.16612554e-01 3.25160086e-01 -1.63138580e+00
-6.63571358e-02 -1.06964633e-01 -7.05390871e-01 6.63438559e-01
-2.18888685e-01 3.13220263e-01 3.97013754e-01 4.83226269e-01
-8.08920801e-01 7.12106705e-01 1.42719877e+00 -2.39164397e-01
-1.82867169e-01 -1.08128168e-01 -9.10682082e-02 8.06160808e-01
4.34124082e-01 -4.55775224e-02 -1.36199340e-01 -6.55452430e-01
-5.56482613e-01 -9.00405869e-02 6.29237771e-01 -1.01151955e+00
4.54405904e-01 -3.22519094e-01 1.79686353e-01 -8.52630556e-01
4.03000951e-01 -8.87536407e-01 -1.63604915e-01 2.69384891e-01
-3.26646566e-01 -8.62259790e-02 -2.00926319e-01 6.14208937e-01
-6.62146091e-01 4.23360437e-01 8.55531991e-01 -1.29052093e-02
-1.34790719e+00 7.56861091e-01 -4.33047533e-01 -2.49475643e-01
7.00513124e-01 -1.96751043e-01 -2.05933824e-01 -3.98225397e-01
-5.83151340e-01 2.82422811e-01 3.19557995e-01 8.70593727e-01
8.57976317e-01 -1.47061968e+00 -6.79479599e-01 2.99046308e-01
2.93299258e-01 -5.14230654e-02 4.10995334e-01 7.19250143e-01
-3.79633367e-01 3.08496058e-01 -3.26694995e-01 -7.24368989e-01
-1.39468181e+00 1.09037176e-01 3.73378515e-01 -1.18990973e-01
-1.09969246e+00 9.68756318e-01 2.11117491e-01 -1.80083364e-01
4.77292806e-01 -1.00427926e+00 -1.34958342e-01 -1.26802608e-01
6.56929910e-01 -1.15866989e-01 -1.32242218e-01 -9.53299403e-01
-3.38273078e-01 1.00739622e+00 -4.64540459e-02 -1.71792328e-01
1.32338440e+00 -2.04925030e-01 2.01677829e-01 1.22216851e-01
1.36883807e+00 -3.85607392e-01 -1.78514063e+00 -8.34486544e-01
-2.54735827e-01 -6.52981281e-01 -1.46299265e-02 -6.65641963e-01
-1.37650216e+00 6.11307144e-01 6.70069814e-01 -4.04128164e-01
1.61609006e+00 1.77958068e-02 1.23710477e+00 3.13227087e-01
3.10614824e-01 -1.17053592e+00 1.35307163e-01 6.97278082e-01
9.40656364e-01 -1.30431521e+00 -3.43401581e-01 1.65688340e-03
-6.42327845e-01 1.21405447e+00 4.90059823e-01 -3.17558289e-01
6.55580044e-01 4.12925519e-02 3.95043820e-01 -8.02903175e-02
-5.64824343e-01 -7.29116142e-01 5.74794352e-01 3.61818105e-01
1.51203305e-01 2.30595469e-02 -3.76946002e-01 2.17633903e-01
-9.33910385e-02 3.74681771e-01 -7.64712468e-02 9.34074163e-01
-3.69799465e-01 -9.12171543e-01 -2.27005422e-01 2.96009481e-01
4.17247862e-02 -9.03229713e-02 -2.01606885e-01 5.98520756e-01
3.92345726e-01 9.34933841e-01 1.51211157e-01 -4.59238052e-01
4.22563344e-01 1.39442652e-01 4.29001004e-01 -8.65093768e-02
-4.70669955e-01 3.79770756e-01 -1.55408472e-01 -8.46139908e-01
-8.58339608e-01 -9.04724300e-01 -1.38011050e+00 -1.42014846e-01
1.46093473e-01 -5.20141780e-01 3.24605376e-01 1.20557690e+00
1.57533452e-01 4.93463844e-01 3.56440544e-01 -9.85514104e-01
-3.06691647e-01 -8.19023371e-01 -4.17482764e-01 7.37925947e-01
5.87111831e-01 -6.66974187e-01 -2.93700963e-01 2.70512581e-01]
|
[9.21903133392334, -6.490591049194336]
|
922e2c1b-a95b-460c-ba7c-422d164b0687
|
large-scale-unsupervised-semantic
|
2106.03149
| null |
https://arxiv.org/abs/2106.03149v3
|
https://arxiv.org/pdf/2106.03149v3.pdf
|
Large-scale Unsupervised Semantic Segmentation
|
Empowered by large datasets, e.g., ImageNet, unsupervised learning on large-scale data has enabled significant advances for classification tasks. However, whether the large-scale unsupervised semantic segmentation can be achieved remains unknown. There are two major challenges: i) we need a large-scale benchmark for assessing algorithms; ii) we need to develop methods to simultaneously learn category and shape representation in an unsupervised manner. In this work, we propose a new problem of large-scale unsupervised semantic segmentation (LUSS) with a newly created benchmark dataset to help the research progress. Building on the ImageNet dataset, we propose the ImageNet-S dataset with 1.2 million training images and 50k high-quality semantic segmentation annotations for evaluation. Our benchmark has a high data diversity and a clear task objective. We also present a simple yet effective method that works surprisingly well for LUSS. In addition, we benchmark related un/weakly/fully supervised methods accordingly, identifying the challenges and possible directions of LUSS. The benchmark and source code is publicly available at https://github.com/LUSSeg.
|
['ShangHua Gao', 'Philip Torr', 'Junwei Han', 'Ming-Ming Cheng', 'Ming-Hsuan Yang', 'Zhong-Yu Li']
|
2021-06-06
| null | null | null | null |
['unsupervised-semantic-segmentation']
|
['computer-vision']
|
[ 3.36079478e-01 1.90191761e-01 -3.11405689e-01 -4.92523521e-01
-9.73573506e-01 -6.35604799e-01 2.07449570e-01 -2.20109731e-01
-5.44453681e-01 5.93171775e-01 9.19807926e-02 1.96308754e-02
5.17334379e-02 -5.94625056e-01 -7.43248701e-01 -6.95643365e-01
2.40413278e-01 6.69347882e-01 6.13984227e-01 6.00968003e-02
2.16354817e-01 2.17440929e-02 -1.62522316e+00 1.96882039e-01
9.44997728e-01 1.21819854e+00 4.80324149e-01 5.04429102e-01
-2.11531147e-01 5.56367457e-01 -4.20209765e-01 -1.34634897e-01
3.08762908e-01 -4.34298128e-01 -1.27736580e+00 2.53191769e-01
5.70415497e-01 -5.76873869e-02 1.58656657e-01 1.14540851e+00
4.75958198e-01 5.00910655e-02 5.15216053e-01 -1.35674632e+00
-4.08385873e-01 7.28089809e-01 -5.17234206e-01 -7.20900074e-02
-1.56254709e-01 1.14557385e-01 1.08762288e+00 -7.06255317e-01
8.55485916e-01 9.12550390e-01 6.60753250e-01 6.61258221e-01
-9.39318359e-01 -5.02286673e-01 8.94627273e-02 1.04297422e-01
-1.39186776e+00 -3.09286714e-01 6.82268858e-01 -4.92419392e-01
5.77700555e-01 2.94157416e-01 4.82948750e-01 1.01084173e+00
-5.31984150e-01 1.18361187e+00 1.28599668e+00 -2.36908972e-01
3.29050481e-01 -7.45740458e-02 4.27020192e-01 6.89744234e-01
1.53618366e-01 -2.10880458e-01 -2.27184951e-01 2.01261431e-01
6.17346108e-01 -5.88019453e-02 1.07737944e-01 -4.43988204e-01
-1.45575655e+00 8.10330927e-01 4.99204189e-01 4.21710044e-01
-4.81120832e-02 8.14625397e-02 4.78009731e-01 1.16740189e-01
5.30609429e-01 5.66018522e-01 -9.57500577e-01 -1.99475557e-01
-1.00579226e+00 9.86658186e-02 8.62303495e-01 9.77398276e-01
9.34380233e-01 -1.02092288e-01 -4.40259092e-02 1.20101309e+00
1.14117250e-01 3.60317975e-01 5.87265134e-01 -1.15125608e+00
2.71396399e-01 7.59313285e-01 -2.48493418e-01 -6.88369870e-01
-5.01909554e-01 -3.31062794e-01 -9.37460601e-01 -9.48221684e-02
5.22559762e-01 -1.30181955e-02 -1.11182642e+00 1.59119403e+00
3.68188351e-01 4.64640588e-01 1.15997307e-01 9.41693544e-01
1.19059038e+00 3.42897594e-01 1.38839483e-01 1.07652873e-01
1.27629137e+00 -1.48766744e+00 -5.49419880e-01 -3.21109802e-01
6.76973164e-01 -6.25651598e-01 1.28812003e+00 2.36323729e-01
-7.55780041e-01 -6.05866492e-01 -8.34295869e-01 -5.17930686e-02
-7.17016101e-01 2.40089700e-01 8.58317018e-01 5.09399056e-01
-1.16694260e+00 3.62440199e-01 -8.88895988e-01 -6.30856156e-01
7.94752359e-01 2.80395716e-01 -2.39905432e-01 5.86732873e-04
-9.77397859e-01 3.79010499e-01 8.40922296e-01 -4.81739007e-02
-8.44684362e-01 -4.76481766e-01 -8.79678488e-01 -2.63689399e-01
7.42929578e-01 -4.46370006e-01 1.23215735e+00 -1.21100509e+00
-1.31746829e+00 1.20559382e+00 1.52455807e-01 -3.94353241e-01
3.85540664e-01 -1.55201241e-01 -1.94835410e-01 3.20122063e-01
5.31688631e-01 1.23121786e+00 5.20496547e-01 -1.50112808e+00
-6.18814051e-01 -5.32251656e-01 -1.01978123e-01 7.28709027e-02
-5.47637820e-01 8.85806754e-02 -9.66014624e-01 -7.99598873e-01
2.06030220e-01 -9.22195077e-01 -5.47238529e-01 -2.91343778e-01
-7.82867610e-01 -3.17865252e-01 7.72046864e-01 -4.75374520e-01
1.12391114e+00 -2.08017111e+00 9.41182002e-02 -6.25331327e-02
3.20332855e-01 3.24146509e-01 -3.04305822e-01 1.18089616e-01
1.14737324e-01 2.88027525e-01 -8.45995784e-01 -3.57658893e-01
-3.76324914e-02 3.63815784e-01 -1.13988414e-01 1.56397268e-01
1.68875888e-01 1.32417905e+00 -9.76672590e-01 -7.37433493e-01
7.94815347e-02 -2.51275040e-02 -4.97349799e-01 2.62022972e-01
-3.83542836e-01 6.73303425e-01 -6.29094303e-01 1.06785035e+00
5.86225986e-01 -6.28920615e-01 -1.22284308e-01 -2.10144579e-01
-3.18733491e-02 -1.58636183e-01 -1.10598576e+00 2.07821703e+00
7.68599585e-02 4.97758716e-01 1.80829000e-02 -1.46484494e+00
7.44425595e-01 1.22147445e-02 7.07179427e-01 -6.55345440e-01
6.41008466e-02 4.47121829e-01 -2.81442821e-01 -5.29075384e-01
5.00256360e-01 1.10469975e-01 -3.46524209e-01 2.55131751e-01
3.63274515e-01 -2.83183038e-01 6.01106763e-01 2.84498185e-01
9.91278410e-01 1.50485039e-01 2.33684741e-02 -5.40266812e-01
2.84532249e-01 4.07720417e-01 7.18265355e-01 7.79841125e-01
-3.86352360e-01 8.77666056e-01 5.47985554e-01 -3.20359349e-01
-8.86254966e-01 -9.50537682e-01 -2.95274347e-01 1.15472436e+00
4.66885805e-01 -5.48141539e-01 -1.16587234e+00 -8.86093140e-01
-6.40677959e-02 1.34931251e-01 -6.12757266e-01 2.82070965e-01
-3.18496376e-01 -8.88722181e-01 7.44469106e-01 7.45350420e-01
8.24390292e-01 -1.36378622e+00 -2.57304490e-01 -5.81575185e-02
-2.99012482e-01 -1.51833534e+00 -2.91074902e-01 4.93899249e-02
-9.71769154e-01 -1.35803747e+00 -7.49707580e-01 -1.17342305e+00
8.15672696e-01 1.43195063e-01 1.33606577e+00 2.46290453e-02
-4.10681963e-01 4.50216830e-01 -5.91780484e-01 -3.54893982e-01
-1.50633112e-01 4.84611064e-01 -2.02073351e-01 -1.13233708e-01
2.22737759e-01 -3.18923503e-01 -6.87130630e-01 7.28639960e-01
-1.07302988e+00 3.68484020e-01 5.46322465e-01 5.96199334e-01
1.05857241e+00 -3.28655392e-02 8.64615500e-01 -1.27793181e+00
1.90570012e-01 -5.71363807e-01 -6.09042943e-01 1.53592572e-01
-5.70433438e-01 -2.82882780e-01 5.30430794e-01 -3.10996380e-02
-8.95227730e-01 2.55561769e-01 -4.47988838e-01 -7.48161227e-02
-6.42164052e-01 4.46809798e-01 -2.07586646e-01 -4.94658155e-03
5.00954509e-01 8.08669329e-02 -5.40129207e-02 -8.40477705e-01
5.86952269e-01 8.16566885e-01 6.03874683e-01 -7.65989184e-01
6.67691410e-01 6.12124562e-01 -4.30510223e-01 -9.23584819e-01
-1.26238954e+00 -8.86495709e-01 -9.77232397e-01 2.75210235e-02
1.21776259e+00 -9.81837213e-01 -1.85430080e-01 8.77922177e-01
-7.66532063e-01 -9.42091763e-01 -4.40120727e-01 1.49518892e-01
-6.97476268e-01 4.89285499e-01 -6.11241937e-01 -2.19861835e-01
-3.10586363e-01 -1.17318416e+00 1.25759804e+00 2.87417978e-01
1.10227063e-01 -1.02073264e+00 4.73028086e-02 1.02655411e+00
2.86137491e-01 3.08537453e-01 2.22040012e-01 -7.28955925e-01
-6.40517712e-01 1.16916470e-01 -5.46806037e-01 7.37467110e-01
8.28484073e-02 -6.74756020e-02 -9.01910901e-01 -2.28829488e-01
-2.62070477e-01 -8.84474456e-01 1.18344235e+00 3.24477047e-01
1.85709667e+00 -4.16277424e-02 -2.24940687e-01 8.82785320e-01
1.35907388e+00 -1.87415287e-01 4.94648725e-01 3.81472558e-01
1.14349365e+00 4.17126209e-01 7.83745050e-01 7.75197744e-02
5.68173468e-01 3.74427289e-01 2.54220128e-01 -3.88748854e-01
-3.82926792e-01 -9.22159106e-02 -3.25673074e-02 1.26717103e+00
-2.25696251e-01 -8.69025663e-02 -1.16999936e+00 6.52612507e-01
-1.97697914e+00 -5.49129784e-01 -2.05884695e-01 1.81564796e+00
9.39769566e-01 1.28148451e-01 1.63655072e-01 2.89743505e-02
6.93439901e-01 1.96364850e-01 -7.30644166e-01 2.14237094e-01
-2.48178557e-01 2.30962649e-01 5.15685081e-01 1.07200891e-01
-1.57436371e+00 1.49106753e+00 6.20003843e+00 1.11653900e+00
-8.90849113e-01 3.02240789e-01 9.21495855e-01 3.32103908e-01
-1.20038778e-01 -5.46635278e-02 -8.62095892e-01 4.81823921e-01
7.52896369e-01 2.24659488e-01 9.66296792e-02 9.73675013e-01
-7.65429363e-02 -8.02451447e-02 -6.47258043e-01 9.10202384e-01
6.98928609e-02 -1.27761376e+00 -7.10807666e-02 -6.19472340e-02
1.23601019e+00 6.12028301e-01 -1.25336573e-01 3.36302817e-01
3.07915747e-01 -8.84632051e-01 5.16817331e-01 8.52555111e-02
1.07316983e+00 -3.19386333e-01 6.93293750e-01 3.75980467e-01
-1.35307074e+00 1.55236438e-01 -3.69999498e-01 1.34174854e-01
3.19032967e-02 5.39020002e-01 -2.92750746e-01 5.58837175e-01
8.63027036e-01 1.17032981e+00 -9.07479167e-01 1.12955952e+00
-4.13125247e-01 1.04763758e+00 -3.83258969e-01 3.22302938e-01
4.41377759e-01 -3.89044553e-01 1.80357322e-01 1.33207369e+00
-4.75338101e-02 -4.42669876e-02 6.65149033e-01 7.00665832e-01
-2.89532334e-01 3.12552631e-01 -3.10063899e-01 -2.99205095e-01
1.74214676e-01 1.54984188e+00 -1.40927219e+00 -4.77281541e-01
-4.04360324e-01 1.03631711e+00 3.88308674e-01 4.56766069e-01
-6.67767227e-01 -1.34479091e-01 4.92784262e-01 -1.39594123e-01
1.56816006e-01 -1.87624097e-01 -4.04970556e-01 -1.37092733e+00
-1.58018708e-01 -7.70942330e-01 6.68108284e-01 -6.19767249e-01
-1.39020550e+00 4.38938618e-01 -1.52162090e-01 -8.84616792e-01
1.90721259e-01 -6.53889954e-01 -3.52463782e-01 1.75451174e-01
-1.77391529e+00 -1.39203715e+00 -5.96761882e-01 6.36066973e-01
8.21165264e-01 -1.61844030e-01 7.45942652e-01 4.37327504e-01
-7.10446060e-01 4.50088769e-01 1.77243412e-01 5.46853781e-01
6.91841960e-01 -1.55620897e+00 5.00072837e-01 6.17841184e-01
3.74130160e-01 1.30264103e-01 2.12877169e-01 -5.81708789e-01
-1.19504356e+00 -1.36965692e+00 4.67214286e-01 -6.87608480e-01
7.71744430e-01 -4.82353985e-01 -8.26749980e-01 6.60491645e-01
-3.36745754e-02 3.93668056e-01 7.30314970e-01 7.97651857e-02
-3.11578482e-01 2.02467348e-02 -1.02180743e+00 2.74609029e-01
1.37803984e+00 -3.31625640e-01 -4.82578635e-01 5.94128489e-01
9.47572410e-01 -4.64368254e-01 -9.15783763e-01 6.88176751e-01
2.34956667e-01 -8.98251176e-01 1.02888358e+00 -5.57902694e-01
5.83774507e-01 -1.77644745e-01 -2.44186923e-01 -1.01162136e+00
4.03149389e-02 -3.26524884e-01 7.17383549e-02 1.34949827e+00
3.25973451e-01 -7.37446547e-01 9.86899614e-01 2.80805349e-01
-3.69789094e-01 -8.96136522e-01 -5.74562907e-01 -9.24349487e-01
2.26828963e-01 -5.45352101e-01 3.34467828e-01 1.06172204e+00
-3.98972809e-01 1.74862087e-01 -1.08880453e-01 -1.02945723e-01
9.03997064e-01 3.21725518e-01 7.61226594e-01 -1.24569356e+00
-6.20029904e-02 -5.36751747e-01 -4.82222795e-01 -1.12189293e+00
2.98575729e-01 -1.11618650e+00 1.44707173e-01 -1.67159402e+00
6.54656231e-01 -6.48093879e-01 -5.35935342e-01 7.52805293e-01
-2.68457234e-01 7.45338619e-01 5.56327440e-02 4.09323871e-01
-1.37310398e+00 5.18936872e-01 1.39386725e+00 -1.19539827e-01
2.75642276e-02 -1.37192145e-01 -8.15435112e-01 9.30669665e-01
1.21553421e+00 -2.67686039e-01 -4.33343858e-01 -4.47040588e-01
-1.25471726e-01 -4.66132134e-01 3.57841164e-01 -9.35717344e-01
1.51143298e-01 -1.21527836e-01 -7.14859962e-02 -6.30023837e-01
-6.43323660e-02 -5.01291931e-01 -2.13708550e-01 1.33170068e-01
-3.06850940e-01 -5.39663017e-01 8.24778378e-02 4.76315469e-01
-6.10228240e-01 -1.40727639e-01 8.35273087e-01 -4.03843254e-01
-1.37038088e+00 6.35638356e-01 1.18087769e-01 6.40908599e-01
9.72912848e-01 -1.29817605e-01 -1.55236065e-01 1.38721928e-01
-7.36546934e-01 5.24633825e-01 6.30932808e-01 6.58147275e-01
3.81148070e-01 -1.18555868e+00 -6.49181366e-01 9.85463802e-03
4.36152458e-01 4.57288593e-01 1.89038545e-01 6.84808314e-01
-5.23710072e-01 3.60107571e-01 -1.24913923e-01 -7.93499529e-01
-1.02441418e+00 3.00436050e-01 -1.32583315e-02 -1.41709581e-01
-5.78544557e-01 1.00843060e+00 3.91977519e-01 -9.79599178e-01
1.76584616e-01 -1.79647639e-01 -3.03964108e-01 1.01501986e-01
3.20704281e-01 1.51265919e-01 -6.40139580e-02 -7.12104976e-01
-2.17027873e-01 7.95580685e-01 1.44917995e-01 2.66533464e-01
1.56690598e+00 -1.15314856e-01 -3.43908817e-01 4.48291063e-01
1.29429030e+00 -3.66877556e-01 -1.35075927e+00 -2.90679246e-01
3.61169755e-01 -3.24786901e-01 -7.81751126e-02 -8.01104248e-01
-1.45000076e+00 6.70023441e-01 4.71673697e-01 2.16864616e-01
1.14255691e+00 4.89810556e-01 1.02252054e+00 3.47188473e-01
4.57591861e-01 -1.47276020e+00 2.71456629e-01 6.03081226e-01
6.08922958e-01 -1.78428102e+00 -2.72454560e-01 -6.73632264e-01
-7.17060387e-01 7.60985732e-01 7.69442677e-01 3.30621004e-02
7.00240433e-01 2.07087681e-01 2.71550506e-01 -5.11493027e-01
-1.76840469e-01 -7.26864040e-01 3.30183536e-01 4.77401793e-01
3.38784516e-01 2.22909123e-01 -2.30282620e-01 8.52077782e-01
-2.05706805e-01 8.23788419e-02 1.56806514e-01 7.01730311e-01
-5.17806113e-01 -1.14643431e+00 -1.37038589e-01 5.49120486e-01
-3.41283321e-01 -1.98061224e-02 -4.69226986e-01 5.44769347e-01
1.68774888e-01 8.30952168e-01 -8.24219137e-02 -2.12129742e-01
1.25499174e-01 4.51120688e-03 2.00891376e-01 -8.48680198e-01
-1.81255654e-01 7.08533973e-02 -1.30457086e-02 -8.67275178e-01
-6.71044648e-01 -5.28580368e-01 -1.49075234e+00 2.25970775e-01
-1.62020072e-01 8.55484009e-02 7.42813230e-01 8.58349383e-01
3.91389012e-01 4.03687000e-01 3.83996099e-01 -6.78583205e-01
-2.53100097e-01 -8.53910863e-01 -6.87612534e-01 8.53519142e-01
-1.94587588e-01 -5.20056546e-01 -2.81613082e-01 3.28016639e-01]
|
[9.592916488647461, 0.7916719317436218]
|
cae80d87-e120-419d-a7c4-3e85fa040676
|
knowledge-assembly-semi-supervised-multi-task
|
2306.08839
| null |
https://arxiv.org/abs/2306.08839v1
|
https://arxiv.org/pdf/2306.08839v1.pdf
|
Knowledge Assembly: Semi-Supervised Multi-Task Learning from Multiple Datasets with Disjoint Labels
|
In real-world scenarios we often need to perform multiple tasks simultaneously. Multi-Task Learning (MTL) is an adequate method to do so, but usually requires datasets labeled for all tasks. We propose a method that can leverage datasets labeled for only some of the tasks in the MTL framework. Our work, Knowledge Assembly (KA), learns multiple tasks from disjoint datasets by leveraging the unlabeled data in a semi-supervised manner, using model augmentation for pseudo-supervision. Whilst KA can be implemented on any existing MTL networks, we test our method on jointly learning person re-identification (reID) and pedestrian attribute recognition (PAR). We surpass the single task fully-supervised performance by $4.2\%$ points for reID and $0.9\%$ points for PAR.
|
['Tae-hoon Kim', 'Minhyeong Yu', 'Philipp Benz', 'Federica Spinola']
|
2023-06-15
| null | null | null | null |
['person-re-identification', 'pedestrian-attribute-recognition', 'multi-task-learning']
|
['computer-vision', 'computer-vision', 'methodology']
|
[ 1.44979700e-01 1.67693496e-01 -6.92985430e-02 -7.39174962e-01
-1.04943097e+00 -5.85598588e-01 6.26711547e-01 1.09052323e-01
-8.42188597e-01 1.04045594e+00 -2.95242891e-02 4.01438549e-02
1.10105224e-01 -5.88994384e-01 -1.05168879e+00 -4.35318559e-01
2.92022854e-01 8.39879990e-01 5.92268296e-02 1.80255398e-01
-2.66015500e-01 2.48230979e-01 -1.55614185e+00 5.19747436e-01
8.59201491e-01 6.64658785e-01 -8.31565410e-02 5.75069487e-01
-2.58572698e-01 7.03272045e-01 -4.26601052e-01 -9.99582350e-01
3.71721536e-01 1.03233337e-01 -1.00688386e+00 1.09170623e-01
9.44919586e-01 -1.35573983e-01 -8.37994888e-02 7.89942861e-01
5.16535103e-01 1.46419063e-01 6.76894605e-01 -1.63956869e+00
-5.32998264e-01 5.78339577e-01 -6.47238433e-01 -1.71358794e-01
2.68226773e-01 4.89726327e-02 6.61903322e-01 -9.97960150e-01
2.63592660e-01 1.24274802e+00 9.40133393e-01 6.80838823e-01
-1.60037160e+00 -8.54625523e-01 2.05314055e-01 1.75761327e-01
-1.50344610e+00 -7.67716408e-01 5.19191563e-01 -4.83564883e-01
9.15531337e-01 -1.88424870e-01 9.48126167e-02 1.48173511e+00
-5.61041355e-01 1.02939451e+00 1.30957639e+00 -4.85721648e-01
-1.27166152e-01 3.41676205e-01 4.17246580e-01 5.88889956e-01
3.65832448e-01 -1.45209208e-02 -7.07795918e-01 -1.54758438e-01
7.28142858e-01 5.08021601e-02 2.99377590e-01 -3.80343825e-01
-1.16144598e+00 4.70924050e-01 1.22113571e-01 -2.81691015e-01
-8.64090249e-02 1.44068256e-01 5.34528971e-01 4.18917716e-01
4.85565662e-01 2.42076993e-01 -7.59357035e-01 6.78632110e-02
-8.13190460e-01 2.92515486e-01 7.36138940e-01 1.31927884e+00
1.21626902e+00 -1.11431122e-01 -1.56373858e-01 1.03954065e+00
6.78024963e-02 5.28117120e-01 1.28940448e-01 -8.33038151e-01
5.31696141e-01 6.31954670e-01 3.89144033e-01 -2.46333063e-01
-4.03810948e-01 -1.88970879e-01 -9.52241123e-01 1.39190555e-01
7.66285002e-01 -4.16692525e-01 -1.05283451e+00 1.88876271e+00
2.18717515e-01 3.37873757e-01 1.58431053e-01 3.64669234e-01
7.73769975e-01 4.87776138e-02 5.54104567e-01 1.73825890e-01
1.29483426e+00 -1.18732476e+00 -3.08883429e-01 -6.14892006e-01
7.91744828e-01 -6.03589952e-01 9.98114586e-01 2.35894039e-01
-8.23728502e-01 -9.90968525e-01 -6.79684281e-01 -1.28687978e-01
-4.77560252e-01 4.45255548e-01 5.12860835e-01 7.95253515e-01
-1.02462423e+00 1.43760651e-01 -6.35757744e-01 -3.48031640e-01
6.40572309e-01 4.81116831e-01 -8.52684915e-01 -4.07085806e-01
-7.47984290e-01 9.92145896e-01 5.57116330e-01 -1.15252763e-01
-9.28893209e-01 -6.81173623e-01 -1.04443038e+00 -1.16276830e-01
5.41313112e-01 -8.02356541e-01 1.14991176e+00 -7.38399386e-01
-1.04478931e+00 1.22956419e+00 -2.89908916e-01 -4.81694907e-01
7.28784800e-01 -6.73574746e-01 -4.28625107e-01 -2.34488651e-01
3.75829756e-01 9.13458467e-01 7.88523912e-01 -1.54605269e+00
-7.04447389e-01 -4.00417268e-01 5.49921347e-03 3.62252183e-02
-2.05311164e-01 1.73933692e-02 -4.62177664e-01 -6.13945305e-01
-4.03367788e-01 -1.01040757e+00 -2.49945894e-01 -7.87368342e-02
-5.13135612e-01 -3.95297945e-01 6.76823676e-01 -6.56562388e-01
6.97239816e-01 -1.98884726e+00 -2.11843755e-02 1.24661354e-02
1.85361877e-01 3.81889671e-01 -3.66069168e-01 1.24559864e-01
6.53778911e-02 3.80027294e-02 -2.95238286e-01 -1.13155198e+00
-7.91191906e-02 4.19803500e-01 -9.05278325e-03 2.19393045e-01
2.16917515e-01 1.07246482e+00 -6.90354407e-01 -5.83170772e-01
1.00806542e-01 2.53819138e-01 -2.34734789e-01 3.12585354e-01
-9.85147879e-02 6.38879478e-01 -1.05026975e-01 7.69789994e-01
7.15443909e-01 -2.47208029e-01 1.85035989e-01 -2.74340779e-01
-2.47562863e-02 -2.55952835e-01 -1.38677537e+00 1.89045537e+00
-4.74510968e-01 4.36131060e-01 1.65227190e-01 -1.07041156e+00
9.43274379e-01 2.72313774e-01 4.24807698e-01 -4.02708948e-01
-3.91805023e-01 1.64161578e-01 -3.85034770e-01 -3.87428075e-01
2.60890901e-01 1.09838217e-01 -3.37733358e-01 6.52634263e-01
3.96533847e-01 4.67005312e-01 1.09300576e-01 1.07317895e-01
9.81839836e-01 4.47931647e-01 2.54378527e-01 -1.01814441e-01
4.87871021e-01 3.42969485e-02 7.35068500e-01 1.11489928e+00
-2.55360812e-01 4.29259658e-01 -8.83254632e-02 -7.42694676e-01
-1.34361613e+00 -1.20758569e+00 7.82993902e-03 1.46038520e+00
-7.09933564e-02 -2.11007833e-01 -6.22333288e-01 -1.08123636e+00
3.43847513e-01 5.04534841e-01 -4.28655267e-01 1.72490150e-01
-6.69803202e-01 -8.51528823e-01 9.67695594e-01 9.03397918e-01
8.75680447e-01 -9.16988313e-01 -9.26567093e-02 2.81280190e-01
-2.49821380e-01 -1.60612643e+00 -2.66903907e-01 1.50505766e-01
-3.56640756e-01 -1.27050090e+00 -5.82558393e-01 -7.08149433e-01
8.21463645e-01 2.78196454e-01 1.21662521e+00 -7.03028738e-02
-2.53237396e-01 6.28397942e-01 -2.85454899e-01 -6.53799415e-01
-3.13915193e-01 2.71892995e-01 5.10816932e-01 2.45606765e-01
7.52448022e-01 -5.80039620e-01 -1.99922875e-01 5.83077371e-01
-4.97542143e-01 2.80672699e-01 7.72056937e-01 8.65175724e-01
5.77133179e-01 -2.24429369e-01 9.19775188e-01 -1.15404844e+00
8.37654024e-02 -2.85187542e-01 -4.12166148e-01 5.85535109e-01
-4.53442395e-01 1.38619170e-01 4.61833090e-01 -5.14609396e-01
-1.20381737e+00 5.02220988e-01 6.24957234e-02 -4.53095883e-01
-7.73884177e-01 1.70655385e-01 -3.44956875e-01 -2.09086880e-01
7.00096548e-01 1.84964076e-01 -1.39909998e-01 -9.47941005e-01
5.00091374e-01 6.54658437e-01 8.62006426e-01 -8.91289353e-01
9.68733430e-01 5.13391316e-01 -1.59345716e-01 -5.17320693e-01
-1.33068335e+00 -4.73531127e-01 -1.30446875e+00 -1.05386540e-01
9.02985573e-01 -1.37884235e+00 -8.07431698e-01 6.73953831e-01
-1.09073067e+00 -5.66414177e-01 -1.30400032e-01 3.41761440e-01
-4.76338327e-01 4.10783142e-01 -3.24682981e-01 -7.97297239e-01
-3.48280042e-01 -8.57254565e-01 9.48248386e-01 6.59588799e-02
-1.57222986e-01 -8.43236387e-01 -2.12675691e-01 8.80506694e-01
3.17137599e-01 1.71053573e-01 7.00403750e-01 -1.03686857e+00
-4.30894047e-01 -1.98122755e-01 -5.24638057e-01 3.79496992e-01
2.22374201e-01 -4.96283650e-01 -1.32046068e+00 -4.44935739e-01
-5.12327373e-01 -8.73575449e-01 1.03923821e+00 9.99259725e-02
1.20604491e+00 -1.88443214e-01 -5.44678688e-01 4.69902307e-01
1.17927980e+00 -3.21961015e-01 3.41563165e-01 4.46653873e-01
1.16085136e+00 7.10768282e-01 3.94610554e-01 2.11188972e-01
1.02420187e+00 7.68774986e-01 8.94067138e-02 -3.14140588e-01
-3.04889560e-01 -3.36240560e-01 2.71426737e-01 1.87904164e-01
-1.32169694e-01 8.13376829e-02 -1.29161024e+00 5.97166598e-01
-2.21826649e+00 -8.61115873e-01 -1.25931501e-01 2.24836159e+00
9.52727139e-01 8.99433866e-02 4.06635165e-01 -2.48998813e-02
8.52290928e-01 -2.21288159e-01 -6.65153921e-01 1.25254005e-01
-2.38959044e-01 1.40728441e-03 5.84042847e-01 3.46007258e-01
-1.51954913e+00 1.04898477e+00 6.79024124e+00 6.18413866e-01
-4.29055601e-01 2.38446385e-01 6.14202619e-01 -7.46999830e-02
1.67338401e-01 -1.33008137e-01 -1.21607685e+00 3.51961762e-01
8.49964380e-01 1.61923841e-02 2.22414747e-01 9.60136831e-01
-2.12901577e-01 -2.41795182e-01 -1.42583752e+00 1.32238698e+00
1.33702099e-01 -1.05225348e+00 1.66657582e-01 -5.71616367e-02
6.92875624e-01 6.29702285e-02 -1.71908572e-01 6.22951806e-01
8.62502217e-01 -1.11279809e+00 4.91965592e-01 5.18821776e-01
1.08577943e+00 -7.09320366e-01 7.51731634e-01 5.61926782e-01
-1.37965131e+00 -1.41584575e-01 -1.48565561e-01 5.63675398e-03
2.56250560e-01 5.35602748e-01 -7.18197346e-01 7.24524915e-01
8.71544302e-01 6.27307415e-01 -8.81577194e-01 9.36922491e-01
-2.60861665e-01 3.40118289e-01 -3.12164724e-01 6.72834933e-01
-1.71535611e-01 2.39778440e-02 1.84782550e-01 1.42554557e+00
8.11045116e-04 -2.88588911e-01 7.86991358e-01 4.81005669e-01
-3.61352831e-01 -1.86661825e-01 -6.90109134e-01 3.73970538e-01
6.39184713e-01 1.10856962e+00 -2.61646926e-01 -5.79387605e-01
-7.11117506e-01 1.32627928e+00 7.45219350e-01 3.73797774e-01
-6.34191453e-01 -6.24016039e-02 6.68510973e-01 -1.41033009e-01
-3.09223719e-02 -3.58688414e-01 -3.77352774e-01 -1.31398118e+00
2.69412220e-01 -5.93517125e-01 7.24585474e-01 -5.51599801e-01
-1.95179594e+00 3.35522830e-01 2.17068344e-01 -9.35943067e-01
-2.64019489e-01 -4.99363482e-01 -2.57381529e-01 1.09748363e+00
-1.79297721e+00 -2.04660940e+00 -5.28819740e-01 8.61406147e-01
5.35299897e-01 -4.20966238e-01 1.07385767e+00 5.14921725e-01
-8.09049308e-01 9.41827774e-01 -1.71840459e-01 5.57481170e-01
1.29141700e+00 -1.39022124e+00 7.50423551e-01 7.51805007e-01
2.09004298e-01 3.50179762e-01 2.47775391e-01 -6.45876110e-01
-9.55835700e-01 -1.41914141e+00 9.39656734e-01 -9.77070570e-01
3.71806294e-01 -6.06662929e-01 -9.34167266e-01 1.22747266e+00
-3.02585829e-02 2.84594953e-01 1.12899852e+00 4.65875477e-01
-8.30344021e-01 -1.43547177e-01 -1.23799384e+00 2.70622790e-01
1.36751008e+00 -7.25277245e-01 -5.17511725e-01 2.48854786e-01
4.80663657e-01 -3.38108212e-01 -9.12910581e-01 3.61060172e-01
4.36525017e-01 -6.09610975e-01 1.25623500e+00 -8.54243815e-01
-6.30713850e-02 -3.63994122e-01 -1.50610507e-01 -1.17219520e+00
-2.55450457e-01 -3.90324265e-01 -1.84324279e-01 1.51913989e+00
4.99995470e-01 -6.26026750e-01 9.07518804e-01 1.07879829e+00
-8.31116214e-02 -1.20043971e-01 -9.98249292e-01 -1.05820668e+00
6.66574389e-02 -4.72040206e-01 5.77537417e-01 1.11688960e+00
-4.79172230e-01 4.31691587e-01 -7.66816914e-01 3.55762124e-01
1.21840394e+00 -9.15920138e-02 1.16344607e+00 -1.60327947e+00
-1.51334554e-01 -5.00810507e-04 -2.12275907e-01 -8.10263336e-01
5.01255214e-01 -9.98184204e-01 -2.81991243e-01 -1.39644122e+00
5.49799740e-01 -8.69272709e-01 -2.68248737e-01 1.13543439e+00
-3.87569577e-01 3.44087869e-01 2.27591962e-01 4.32286084e-01
-7.87574828e-01 2.90928125e-01 5.81627488e-01 -1.31509066e-01
-1.14602387e-01 1.08850583e-01 -7.20721722e-01 7.73614287e-01
8.08361888e-01 -4.57236469e-01 -2.12615341e-01 -8.43814254e-01
-1.78823210e-02 -4.02670234e-01 7.11451113e-01 -1.18203008e+00
4.78443503e-01 -3.14464048e-02 6.91981733e-01 -5.07093310e-01
4.52179521e-01 -7.57395864e-01 1.31019473e-01 -1.25574604e-01
-2.91774452e-01 1.74270198e-02 3.67512524e-01 6.34347618e-01
-6.53343424e-02 9.85934138e-02 7.15982795e-01 -2.63537437e-01
-9.59024131e-01 3.68873388e-01 2.44682338e-02 -6.18769750e-02
1.02988875e+00 -2.53098637e-01 -4.77052540e-01 -4.42676879e-02
-1.05188322e+00 6.30797088e-01 4.92933363e-01 3.35534573e-01
3.90874863e-01 -1.38517761e+00 -1.00786567e+00 2.41662711e-01
4.20769751e-01 2.33105287e-01 1.80045918e-01 4.91063058e-01
1.52550235e-01 1.93851501e-01 -3.44769597e-01 -4.87585366e-01
-1.35955644e+00 5.07300079e-01 2.39664063e-01 -2.60551035e-01
-5.24671257e-01 8.18732262e-01 -6.11110404e-02 -8.46844554e-01
3.13868761e-01 4.87285465e-01 -1.53324664e-01 1.41137928e-01
5.62913954e-01 4.80086714e-01 -1.07079595e-01 -5.94457090e-01
-3.09314668e-01 3.64601254e-01 -3.82732689e-01 -1.97724089e-01
1.33339036e+00 -1.74646571e-01 2.76064612e-02 3.26559067e-01
9.16576982e-01 -3.01692933e-01 -1.46900749e+00 -6.60832465e-01
3.10532361e-01 -3.03597271e-01 -4.83324647e-01 -1.08102930e+00
-6.27560318e-01 7.02549577e-01 4.67155039e-01 -2.08886713e-01
8.47557187e-01 1.85546935e-01 5.80839872e-01 7.77175069e-01
7.25439668e-01 -1.21654809e+00 1.03829063e-01 3.49536419e-01
4.68876243e-01 -1.79958832e+00 -7.54035488e-02 -5.02522111e-01
-7.60397971e-01 8.70320261e-01 1.12958372e+00 2.60985702e-01
4.25158679e-01 2.99598634e-01 7.88965598e-02 4.43344638e-02
-5.04258573e-01 -4.44729388e-01 1.29849732e-01 9.38808560e-01
2.29322582e-01 8.02683979e-02 4.33309287e-01 6.32202208e-01
8.72959644e-02 8.78714621e-02 3.40479106e-01 1.02262115e+00
-1.08995184e-01 -1.52267826e+00 -4.23778564e-01 6.76290452e-01
-2.70675510e-01 -9.07587260e-02 -3.25847387e-01 5.47932327e-01
2.32725888e-01 1.00444090e+00 5.33565804e-02 -4.07368898e-01
3.99334073e-01 4.68165159e-01 4.60005134e-01 -7.82006621e-01
-5.93539655e-01 -2.78837353e-01 4.14721072e-01 -3.11296493e-01
-7.03094482e-01 -6.67994022e-01 -8.29717278e-01 -2.81825304e-01
8.36589262e-02 -9.69759822e-02 3.55330706e-01 1.03523135e+00
4.81461406e-01 2.79178321e-02 3.80918831e-01 -7.55754769e-01
-1.55929476e-01 -9.63256121e-01 -3.06647390e-01 6.35082603e-01
3.31099451e-01 -8.08004677e-01 -6.17970191e-02 4.02490705e-01]
|
[14.747150421142578, 1.0500743389129639]
|
3568a57b-86bd-4fc9-b379-14ac47c3ec12
|
render-and-compare-cross-view-6-dof
|
2302.06287
| null |
https://arxiv.org/abs/2302.06287v1
|
https://arxiv.org/pdf/2302.06287v1.pdf
|
Render-and-Compare: Cross-View 6 DoF Localization from Noisy Prior
|
Despite the significant progress in 6-DoF visual localization, researchers are mostly driven by ground-level benchmarks. Compared with aerial oblique photography, ground-level map collection lacks scalability and complete coverage. In this work, we propose to go beyond the traditional ground-level setting and exploit the cross-view localization from aerial to ground. We solve this problem by formulating camera pose estimation as an iterative render-and-compare pipeline and enhancing the robustness through augmenting seeds from noisy initial priors. As no public dataset exists for the studied problem, we collect a new dataset that provides a variety of cross-view images from smartphones and drones and develop a semi-automatic system to acquire ground-truth poses for query images. We benchmark our method as well as several state-of-the-art baselines and demonstrate that our method outperforms other approaches by a large margin.
|
['Maojun Zhang', 'Yu Liu', 'Rouwan Wu', 'Juelin Zhu', 'Yuxiang Liu', 'Xiaoya Cheng', 'Shen Yan']
|
2023-02-13
| null | null | null | null |
['visual-localization']
|
['computer-vision']
|
[ 1.68929957e-02 -3.65592182e-01 -1.17292576e-01 -4.78916198e-01
-1.19566453e+00 -1.27721965e+00 5.97759426e-01 -2.71573097e-01
-3.67547750e-01 5.14980614e-01 3.00248321e-02 -1.80862799e-01
2.27696270e-01 -4.96514827e-01 -9.38717782e-01 -3.43828827e-01
7.30742700e-03 3.89003158e-01 4.78920579e-01 -1.85471058e-01
1.78835586e-01 6.24826789e-01 -1.41651928e+00 -2.43997797e-01
7.35945642e-01 9.15651381e-01 1.47332937e-01 7.52156258e-01
7.19433725e-01 2.14324147e-01 -2.59442657e-01 -3.07963550e-01
7.51886427e-01 -1.25192523e-01 -5.40233731e-01 3.48816127e-01
1.25299418e+00 -6.53928697e-01 -2.81798065e-01 1.09923840e+00
6.42861426e-01 1.20531283e-01 2.39418954e-01 -1.36073601e+00
-1.80684045e-01 -2.32792437e-01 -7.75134444e-01 -3.85103375e-02
8.99638355e-01 2.02823058e-01 9.19678748e-01 -8.83542538e-01
6.70479715e-01 1.00102103e+00 1.08921337e+00 -2.46861931e-02
-1.01316226e+00 -4.89323199e-01 4.35276568e-01 -3.67761940e-01
-1.73278356e+00 -3.67680818e-01 4.61327642e-01 -5.55507600e-01
8.35617304e-01 9.32584610e-03 6.32485747e-01 1.24294841e+00
-2.40793720e-01 4.82918978e-01 9.76238370e-01 -1.65252194e-01
5.51128983e-02 -1.34783939e-01 -4.03949291e-01 9.44053173e-01
4.28787500e-01 7.19206557e-02 -5.21937847e-01 -2.53882438e-01
7.80279458e-01 1.60218015e-01 -5.13074458e-01 -1.05771446e+00
-1.59481001e+00 4.43230242e-01 6.83074355e-01 -2.90402800e-01
-2.19757646e-01 3.62997770e-01 -1.27573848e-01 -6.34906739e-02
4.21741903e-01 5.74876845e-01 -3.79694492e-01 7.85995275e-02
-1.26068771e+00 3.68562222e-01 5.48514128e-01 1.47240520e+00
9.91023540e-01 -2.64096409e-01 1.86459079e-01 2.53410637e-01
3.95302802e-01 9.84671533e-01 -1.91007759e-02 -1.00084674e+00
7.04892397e-01 6.37622118e-01 6.14633620e-01 -1.22880709e+00
-2.66774237e-01 -5.00509739e-01 -4.27912176e-01 -7.69891292e-02
4.48779017e-01 -1.37261093e-01 -8.70027959e-01 1.39717901e+00
6.27312303e-01 2.17364579e-01 -1.31109998e-01 1.17704368e+00
5.58081329e-01 4.05265063e-01 -4.64737445e-01 4.19275790e-01
1.16451705e+00 -1.05048347e+00 -2.58183092e-01 -7.57960975e-01
5.29206336e-01 -7.89594054e-01 1.00732386e+00 3.95580828e-01
-6.75461173e-01 -4.81949627e-01 -1.30600071e+00 -2.40599766e-01
-3.77273381e-01 6.47763014e-01 4.49106306e-01 3.76651525e-01
-1.21607053e+00 1.08362682e-01 -9.74665344e-01 -7.78133214e-01
1.01861618e-01 2.56861240e-01 -7.91545272e-01 -2.65039861e-01
-7.29873300e-01 6.98088884e-01 5.97316734e-02 -3.95571403e-02
-1.10069108e+00 -5.59072137e-01 -1.07107937e+00 -4.04366970e-01
7.38586605e-01 -9.32217956e-01 1.22052717e+00 -4.51786578e-01
-1.18637347e+00 9.76659119e-01 -1.66478887e-01 -4.06706035e-01
7.32166886e-01 -4.93071049e-01 9.95825008e-02 1.91857457e-01
4.75332648e-01 8.11975539e-01 6.41052067e-01 -1.48046672e+00
-7.92764008e-01 -4.42883343e-01 4.69046026e-01 4.20465380e-01
8.42937827e-02 -3.29491049e-01 -1.01104069e+00 -4.70049024e-01
7.37691164e-01 -1.35455334e+00 -2.63620466e-01 8.92541781e-02
-5.04893124e-01 5.14904320e-01 7.08732486e-01 -4.22927201e-01
9.18381333e-01 -1.89800096e+00 2.22455040e-02 2.88359541e-02
1.13146931e-01 -2.34936476e-01 1.13803156e-01 5.98624170e-01
1.78962335e-01 1.64152831e-02 2.69442890e-03 -7.71190941e-01
6.34660050e-02 -8.61917883e-02 -6.00543976e-01 9.05950069e-01
-6.69591948e-02 7.12181687e-01 -1.16656852e+00 -2.83010781e-01
4.58568245e-01 4.44797575e-01 -7.02455521e-01 2.58892864e-01
-4.87165488e-02 5.49251378e-01 -3.32082719e-01 1.18200469e+00
9.61995423e-01 -3.61579806e-01 -1.96175147e-02 -3.38523269e-01
-1.60375968e-01 2.29137316e-01 -1.20495594e+00 2.23125005e+00
-2.86590785e-01 5.70412934e-01 1.31716773e-01 -3.43453646e-01
6.28546834e-01 -1.50115773e-01 3.44565034e-01 -1.12764649e-01
1.95372608e-02 9.79918167e-02 -5.97320676e-01 -1.44570842e-01
8.64027023e-01 4.35806453e-01 -2.83718437e-01 -4.43818420e-02
7.90844113e-02 -6.92106903e-01 -1.76339582e-01 2.00748250e-01
1.07696557e+00 5.88713288e-01 4.48385626e-01 -7.62827471e-02
3.38745385e-01 2.84491181e-01 2.67688245e-01 7.10883737e-01
-4.35535470e-03 1.17144775e+00 9.75897387e-02 -4.26310897e-01
-6.89271510e-01 -1.08028483e+00 1.41199723e-01 6.85238719e-01
6.79122269e-01 -7.80654490e-01 -8.30709755e-01 -8.35376143e-01
9.93527323e-02 2.02831507e-01 -4.79123682e-01 3.04019988e-01
-1.82664260e-01 -5.87131083e-01 5.52741289e-01 4.57332939e-01
9.62493956e-01 2.20167972e-02 -8.75217736e-01 -2.60990202e-01
-3.87909710e-01 -1.66094422e+00 -6.00160301e-01 -7.28660226e-02
-6.87937140e-01 -1.21520591e+00 -6.38254821e-01 -3.10308605e-01
7.24479377e-01 9.13181961e-01 1.19606006e+00 -1.36044189e-01
-3.96715924e-02 6.69511080e-01 -3.01496297e-01 -3.01891744e-01
3.21274787e-01 3.70224774e-01 3.18647563e-01 -4.90665548e-02
-5.47340401e-02 -3.98396969e-01 -8.34020436e-01 6.50651395e-01
-5.09935260e-01 -7.46970698e-02 6.08327985e-01 4.66314107e-01
9.73813415e-01 -3.44332248e-01 -3.41546863e-01 -3.78699958e-01
4.23399769e-02 -3.83907855e-01 -1.30011380e+00 1.08001910e-01
-5.09048104e-01 -2.63835192e-01 2.40858912e-01 -4.46151383e-02
-3.98282617e-01 7.77668178e-01 2.25770101e-01 -7.68489540e-01
-2.13697597e-01 2.54782468e-01 -2.77575552e-01 -6.58411384e-01
6.30657673e-01 1.50341098e-03 -4.26107645e-01 -2.84308285e-01
4.35156882e-01 4.03559268e-01 7.89138258e-01 -4.07283157e-01
1.24214530e+00 8.16777050e-01 8.90174955e-02 -7.25140333e-01
-1.12152648e+00 -7.92191863e-01 -9.07098949e-01 -1.50512680e-01
8.97070229e-01 -1.53071487e+00 -5.28022528e-01 1.76739693e-01
-1.15007424e+00 -2.45374128e-01 1.15133487e-01 6.37587965e-01
-5.58079004e-01 4.64677036e-01 6.59794314e-03 -6.47310734e-01
9.60214064e-02 -1.47160041e+00 1.87852108e+00 -6.88082278e-02
1.21030405e-01 -6.12262189e-01 3.43107551e-01 3.17584485e-01
1.08913826e-02 5.38604677e-01 -2.47109607e-01 -6.18062764e-02
-1.28517461e+00 -5.79969525e-01 -1.86411276e-01 1.61388386e-02
-6.29680753e-02 -1.44797519e-01 -1.10962033e+00 -5.86144745e-01
-2.80729681e-01 -3.34595561e-01 7.07062125e-01 3.26677740e-01
7.38451779e-01 -1.63327996e-02 -6.15895212e-01 1.35125268e+00
1.53646898e+00 -3.59338671e-01 3.68349493e-01 6.53369725e-01
9.02816951e-01 2.71341950e-01 1.00356317e+00 2.90201664e-01
8.79454017e-01 9.85510945e-01 9.79028881e-01 -2.55183250e-01
2.27329075e-01 -7.29314923e-01 1.97310150e-01 1.01139545e-01
-2.13043392e-01 -5.26415288e-01 -1.07528639e+00 5.61712384e-01
-1.78823936e+00 -6.20387673e-01 6.28387975e-03 2.41827965e+00
1.91690177e-01 -2.42810741e-01 3.91139090e-03 -3.55596691e-01
4.81754094e-01 3.64329249e-01 -3.87922585e-01 7.18956172e-01
-4.26191166e-02 -5.33291936e-01 1.28686643e+00 6.37093484e-01
-1.58141518e+00 1.33252180e+00 6.73529482e+00 3.15064877e-01
-1.07490659e+00 -1.07569225e-01 1.59860864e-01 -1.27558395e-01
1.10002637e-01 3.59413177e-01 -1.10664964e+00 1.25443473e-01
4.10924017e-01 4.41841006e-01 4.59993452e-01 1.15937805e+00
-9.92010757e-02 -4.27597731e-01 -1.03690386e+00 1.46338165e+00
3.51095945e-01 -1.24347472e+00 -2.39039555e-01 4.13441300e-01
1.11523628e+00 6.02816939e-01 -7.33117983e-02 -7.57345408e-02
4.75502670e-01 -7.66806245e-01 9.14801061e-01 3.02005023e-01
9.42223012e-01 -4.41318244e-01 7.47754633e-01 3.48634154e-01
-1.45462704e+00 2.41805643e-01 -3.04843187e-01 -7.97898397e-02
2.38555223e-01 2.35494316e-01 -8.43878448e-01 8.63157988e-01
9.59552228e-01 1.00050569e+00 -1.02474153e+00 1.11137104e+00
-3.88203263e-01 2.00775087e-01 -5.78538239e-01 3.61092597e-01
4.76950407e-01 -2.72138685e-01 5.14068425e-01 8.25410903e-01
7.02640593e-01 4.47363080e-03 6.10894680e-01 6.16807759e-01
-1.52516142e-01 -2.72559226e-01 -1.21044254e+00 2.34989539e-01
6.26706719e-01 1.36661589e+00 -8.06443870e-01 -2.03692704e-01
-3.70497972e-01 1.27940464e+00 9.41876993e-02 3.57227117e-01
-1.08497894e+00 -2.02085048e-01 8.34994972e-01 4.67466623e-01
2.41701186e-01 -6.99100494e-01 1.93452433e-01 -1.46887755e+00
2.15491131e-01 -7.89400756e-01 2.64282059e-02 -1.10930073e+00
-8.96144688e-01 8.14496219e-01 2.93457687e-01 -1.72573280e+00
-3.76852959e-01 -4.83136356e-01 -6.76152259e-02 6.83002710e-01
-1.52897239e+00 -1.56566691e+00 -8.97758007e-01 5.36211669e-01
2.85199106e-01 5.53257018e-02 4.96600062e-01 1.73792616e-01
-1.38936117e-01 3.97684038e-01 -5.00748344e-02 1.56136766e-01
9.74425137e-01 -1.17228103e+00 7.03752756e-01 1.33699584e+00
3.58231276e-01 7.57249653e-01 5.91247976e-01 -6.94664896e-01
-1.67969048e+00 -1.13787365e+00 5.21614015e-01 -1.27757907e+00
6.22378588e-01 -7.92108119e-01 -1.35083631e-01 1.10981774e+00
5.55863902e-02 3.75478715e-01 2.21348722e-02 -1.40692472e-01
-2.10738882e-01 -6.85015768e-02 -8.37687016e-01 4.97182995e-01
1.37224638e+00 -6.16269588e-01 -3.13041180e-01 4.52633560e-01
6.71471059e-01 -1.11216450e+00 -5.01459956e-01 4.58353370e-01
5.15789211e-01 -1.22461355e+00 1.40725493e+00 3.74285504e-02
-1.50010034e-01 -9.63107169e-01 -7.00144112e-01 -1.25128388e+00
6.65083304e-02 -7.66052485e-01 2.64406025e-01 1.00279486e+00
7.61664361e-02 -5.67928791e-01 8.59874129e-01 1.92576021e-01
-2.72956669e-01 -4.13679361e-01 -6.55946612e-01 -9.04469550e-01
-7.25738168e-01 -4.43366647e-01 6.47189140e-01 7.79889405e-01
-6.02236211e-01 3.22028250e-01 -5.81816137e-01 9.85441625e-01
6.45636797e-01 2.22834855e-01 1.52730215e+00 -1.04460752e+00
-1.07929811e-01 1.25837818e-01 -6.69900239e-01 -1.64477587e+00
-6.45379126e-02 -5.34865260e-01 2.84333229e-01 -1.55120373e+00
-1.42444789e-01 -1.73024371e-01 3.89317483e-01 3.61066788e-01
1.95373893e-02 6.75102651e-01 -2.31625396e-03 2.86936820e-01
-9.86388147e-01 3.92616391e-01 8.01172435e-01 6.41773045e-02
-5.68252541e-02 -8.63366723e-02 -4.41802710e-01 9.87285435e-01
3.92181665e-01 -3.37336481e-01 -6.15571082e-01 -7.37626910e-01
4.54158187e-01 -1.47262961e-01 8.53863239e-01 -1.35584235e+00
3.74208182e-01 2.84569766e-02 4.56222653e-01 -1.11266577e+00
6.80154741e-01 -1.01432097e+00 1.54315308e-01 1.08927384e-01
1.50201231e-01 6.30612969e-01 9.45346281e-02 8.12903821e-01
-3.09455067e-01 1.95893750e-01 4.29016531e-01 -2.31530651e-01
-8.13165724e-01 5.61519325e-01 8.46288726e-02 2.85020709e-01
9.82216179e-01 -3.92403215e-01 -2.68135250e-01 -6.16985917e-01
-1.85532480e-01 3.20075840e-01 1.28327274e+00 2.93951690e-01
4.07581776e-01 -1.14104342e+00 -2.06689313e-01 3.11717778e-01
6.17126465e-01 4.33966786e-01 -1.03234932e-01 9.19234514e-01
-9.51038778e-01 4.53695178e-01 1.73412055e-01 -1.01690471e+00
-1.14468837e+00 4.37954664e-01 4.24571335e-01 -1.72027834e-02
-3.58410090e-01 7.45828569e-01 2.49049500e-01 -7.62694955e-01
2.08059803e-01 -7.58704603e-01 3.61824036e-01 -1.53185442e-01
3.45207214e-01 2.45664269e-01 1.30131364e-01 -9.41266060e-01
-7.35929310e-01 1.17631555e+00 5.19563377e-01 -2.81414837e-01
1.07127988e+00 -4.84093845e-01 9.76055935e-02 1.88298538e-01
1.04224288e+00 4.26508635e-01 -1.65866828e+00 1.22528225e-01
-2.06946939e-01 -9.07577932e-01 1.79188535e-01 -5.16072154e-01
-7.15202630e-01 7.90613234e-01 6.75082445e-01 -7.15183020e-02
8.64220142e-01 8.33371431e-02 3.75612676e-01 8.00704181e-01
9.56548512e-01 -8.34521472e-01 5.03998995e-02 5.52364171e-01
7.37429023e-01 -1.63242924e+00 3.95398676e-01 -5.94337225e-01
-4.84223098e-01 7.47307241e-01 5.75978041e-01 -3.61138344e-01
2.60397047e-01 1.28505260e-01 1.36241689e-01 -2.62026727e-01
-1.95284650e-01 -4.53629911e-01 5.07020056e-01 7.04130173e-01
9.65047106e-02 -2.46262997e-01 5.02526164e-01 2.94532180e-01
-4.11633968e-01 -9.43160653e-02 2.18013421e-01 8.81229877e-01
-1.49503350e-01 -8.55564296e-01 -5.44697702e-01 -2.16837421e-01
-1.72317803e-01 -1.76051915e-01 -7.37453282e-01 1.23042238e+00
9.15823057e-02 8.52866828e-01 -2.64705747e-01 -6.66184902e-01
4.07256901e-01 -4.14518356e-01 5.84073842e-01 -5.99487722e-01
-1.50117993e-01 1.71768680e-01 4.31954265e-02 -1.13539457e+00
-4.64312524e-01 -6.71754062e-01 -7.55417705e-01 -4.94585298e-02
-3.94909233e-01 -1.47205263e-01 7.71842480e-01 6.21578217e-01
6.05971873e-01 1.25575721e-01 4.05511379e-01 -1.59423757e+00
-5.04739821e-01 -6.18906379e-01 -2.18593359e-01 -1.19338036e-02
6.35097027e-01 -8.96139979e-01 -4.23877358e-01 -3.04240757e-03]
|
[7.6434783935546875, -2.276855230331421]
|
d537fd33-c45c-4f3f-8066-d34033939d9e
|
domain-specific-language-model-pretraining
|
2007.15779
| null |
https://arxiv.org/abs/2007.15779v6
|
https://arxiv.org/pdf/2007.15779v6.pdf
|
Domain-Specific Language Model Pretraining for Biomedical Natural Language Processing
|
Pretraining large neural language models, such as BERT, has led to impressive gains on many natural language processing (NLP) tasks. However, most pretraining efforts focus on general domain corpora, such as newswire and Web. A prevailing assumption is that even domain-specific pretraining can benefit by starting from general-domain language models. In this paper, we challenge this assumption by showing that for domains with abundant unlabeled text, such as biomedicine, pretraining language models from scratch results in substantial gains over continual pretraining of general-domain language models. To facilitate this investigation, we compile a comprehensive biomedical NLP benchmark from publicly-available datasets. Our experiments show that domain-specific pretraining serves as a solid foundation for a wide range of biomedical NLP tasks, leading to new state-of-the-art results across the board. Further, in conducting a thorough evaluation of modeling choices, both for pretraining and task-specific fine-tuning, we discover that some common practices are unnecessary with BERT models, such as using complex tagging schemes in named entity recognition (NER). To help accelerate research in biomedical NLP, we have released our state-of-the-art pretrained and task-specific models for the community, and created a leaderboard featuring our BLURB benchmark (short for Biomedical Language Understanding & Reasoning Benchmark) at https://aka.ms/BLURB.
|
['Hoifung Poon', 'Xiaodong Liu', 'Yu Gu', 'Tristan Naumann', 'Naoto Usuyama', 'Jianfeng Gao', 'Hao Cheng', 'Robert Tinn', 'Michael Lucas']
|
2020-07-31
| null | null | null | null |
['participant-intervention-comparison-outcome', 'continual-pretraining', 'drug-drug-interaction-extraction', 'pico']
|
['medical', 'methodology', 'natural-language-processing', 'natural-language-processing']
|
[ 1.35527447e-01 2.41231665e-01 -4.28933799e-01 -6.09462440e-01
-9.73935246e-01 -4.93375599e-01 2.95027345e-01 3.28615040e-01
-8.38936388e-01 9.00026977e-01 4.30813074e-01 -5.26568055e-01
1.14331916e-01 -5.14892697e-01 -7.68481314e-01 -3.42636257e-01
7.48142526e-02 7.23775804e-01 -1.09128594e-01 -2.03310475e-01
-3.30705166e-01 4.48262244e-01 -5.80585539e-01 6.31515980e-01
6.43341064e-01 5.32791436e-01 4.64347564e-02 4.06717241e-01
-3.52865994e-01 8.06907833e-01 -4.63331312e-01 -5.30784369e-01
-7.23152831e-02 -1.76510662e-01 -1.21493924e+00 -3.88339758e-01
8.65930468e-02 -1.24472283e-01 -2.74444878e-01 8.51276994e-01
7.72512972e-01 6.68939054e-02 6.17855072e-01 -9.78937328e-01
-7.86576629e-01 9.45668101e-01 -3.27958226e-01 3.39213043e-01
-8.36869776e-02 1.93756342e-01 1.07557464e+00 -7.64193356e-01
8.55868459e-01 1.03990293e+00 1.01823545e+00 1.02138019e+00
-1.29253781e+00 -7.66005874e-01 5.52338064e-02 -1.08110286e-01
-1.32590485e+00 -6.00928664e-01 5.56504786e-01 -3.09780180e-01
1.34756613e+00 -4.15200777e-02 9.70160589e-02 1.48514807e+00
3.45517904e-01 1.21446013e+00 8.72896314e-01 -4.52324361e-01
7.57190660e-02 1.22590428e-02 5.08940101e-01 5.86437464e-01
3.82966310e-01 -9.57534313e-02 -2.45688230e-01 -3.54513198e-01
5.63858271e-01 -4.14265394e-02 -2.53194541e-01 -1.84399277e-01
-1.36151254e+00 7.51217008e-01 3.54758590e-01 6.99638844e-01
-3.94969076e-01 6.41152114e-02 6.16075695e-01 2.58752614e-01
5.11946201e-01 8.29558671e-01 -1.03111684e+00 -6.53774887e-02
-9.50097501e-01 -5.35588861e-02 1.05378306e+00 8.08019698e-01
5.52270114e-01 -2.49411985e-01 -1.70900196e-01 1.19400394e+00
1.53873175e-01 1.69486851e-01 7.11355269e-01 -6.46578908e-01
4.49400157e-01 5.09988904e-01 -2.69349545e-01 -6.59265935e-01
-7.53179312e-01 -5.54849267e-01 -1.24233818e+00 -3.82166147e-01
5.05569041e-01 -6.38757706e-01 -1.21203971e+00 1.97838223e+00
3.62987295e-02 2.04263777e-01 3.15119445e-01 5.06636620e-01
1.13247120e+00 4.63306665e-01 6.53366745e-01 1.02191076e-01
1.58475876e+00 -9.47929561e-01 -7.05270588e-01 -6.06332302e-01
1.03334737e+00 -5.45879066e-01 8.75616193e-01 3.93176824e-01
-8.68174672e-01 -3.76235276e-01 -6.86487138e-01 -4.31108594e-01
-6.73270166e-01 1.10660233e-01 7.71593273e-01 4.48421210e-01
-1.09425139e+00 5.19627273e-01 -1.13331878e+00 -7.52428591e-01
6.28289819e-01 3.68800908e-01 -5.30682206e-01 -3.76595169e-01
-1.44953322e+00 1.02955353e+00 5.96441805e-01 -2.53743585e-02
-9.79707837e-01 -1.08743477e+00 -7.82961011e-01 1.07281461e-01
3.08248818e-01 -1.02131128e+00 1.35638463e+00 -7.34589696e-01
-1.14426696e+00 1.29060805e+00 3.10382042e-02 -7.15810597e-01
3.34082842e-01 -2.96350896e-01 -4.71323997e-01 -2.36685332e-02
1.89903378e-02 1.02605462e+00 1.57000050e-01 -8.77165735e-01
-3.03689480e-01 -3.50078158e-02 -7.96869248e-02 -4.54463735e-02
-4.79050159e-01 1.86096355e-01 -4.92934585e-01 -6.51909053e-01
-4.97395366e-01 -7.14957118e-01 -5.38277864e-01 -1.18046701e-01
-5.18077970e-01 -4.39455420e-01 1.63369685e-01 -6.86801434e-01
1.01492357e+00 -2.15734172e+00 -3.03762913e-01 -1.99925989e-01
3.62434596e-01 4.00039047e-01 -4.13590908e-01 4.08845812e-01
-2.85913140e-01 4.28104997e-01 -2.60897517e-01 -4.60775018e-01
-7.74709955e-02 4.10375357e-01 -1.59631386e-01 2.90268332e-01
3.78078222e-01 1.08047056e+00 -1.01577067e+00 -5.55542886e-01
-2.16117918e-01 5.64672828e-01 -6.66431308e-01 7.91673213e-02
-2.65470237e-01 4.22615498e-01 -6.10525548e-01 5.62816322e-01
3.66988778e-01 -7.76408792e-01 2.49688759e-01 -1.97235897e-01
2.44192198e-01 6.17735445e-01 -5.67512453e-01 1.97783446e+00
-3.74614269e-01 3.30792516e-01 1.50799185e-01 -1.26305819e+00
6.42605960e-01 6.14079475e-01 6.77104652e-01 -3.30306739e-01
8.33736658e-02 8.01160336e-02 1.08507693e-01 -3.43429923e-01
1.62210077e-01 -4.13864106e-01 -1.38241619e-01 2.76834190e-01
5.47200084e-01 2.00565279e-01 2.23400876e-01 2.28971586e-01
1.47129619e+00 -1.38806388e-01 7.06941307e-01 -3.93429667e-01
2.70571560e-01 3.37545007e-01 7.11994529e-01 7.87552893e-01
-4.77241337e-01 4.31339771e-01 3.55575055e-01 -5.20303011e-01
-7.07751870e-01 -7.74543881e-01 -3.73695880e-01 1.22969806e+00
-4.75136995e-01 -4.29099947e-01 -6.09350622e-01 -1.04774320e+00
-2.89686192e-02 6.81042731e-01 -6.71826124e-01 -1.48839206e-01
-5.14037132e-01 -1.26068687e+00 1.02976084e+00 6.45354390e-01
4.12598044e-01 -1.19098067e+00 -1.63442403e-01 3.39189589e-01
-2.10537627e-01 -1.35463321e+00 -3.66890788e-01 7.07324624e-01
-9.96941030e-01 -9.08529222e-01 -9.41204190e-01 -9.53921020e-01
6.31729841e-01 -1.75372586e-01 1.55516839e+00 -2.09399089e-02
-2.54914194e-01 3.04753363e-01 -2.54511803e-01 -6.60943031e-01
-5.73264837e-01 4.78763580e-01 -1.77560061e-01 -5.14416873e-01
5.52605152e-01 -3.53917360e-01 -5.25132716e-01 1.61719531e-01
-9.08689082e-01 1.75394014e-01 9.40727234e-01 1.05919075e+00
6.29922807e-01 -2.00859681e-01 7.63945997e-01 -1.35157502e+00
7.17169166e-01 -6.90222740e-01 -3.75637189e-02 4.06643659e-01
-4.84482557e-01 8.57413188e-02 6.22362912e-01 -4.09956187e-01
-9.37764764e-01 -6.40286282e-02 -7.12993264e-01 -4.12204303e-02
-5.64580917e-01 9.72480774e-01 -1.20672084e-01 1.65679082e-01
9.85876501e-01 1.20069921e-01 -3.01774800e-01 -7.55597711e-01
3.92318964e-01 6.05857313e-01 4.63096797e-01 -7.51890838e-01
2.90077031e-01 2.62048513e-01 -4.77216065e-01 -7.85860300e-01
-1.16303706e+00 -7.32108295e-01 -5.66424131e-01 5.78056395e-01
1.07336760e+00 -1.17606831e+00 -2.96898663e-01 2.15319768e-01
-1.11951816e+00 -8.09801996e-01 -1.88676655e-01 4.26349133e-01
-2.31264070e-01 2.20946297e-01 -1.03991449e+00 -1.96510002e-01
-4.80149657e-01 -9.34562683e-01 9.41916406e-01 -2.54001692e-02
-5.06853223e-01 -1.42276335e+00 3.28040719e-01 4.15495366e-01
3.64521116e-01 1.20811731e-01 1.10963154e+00 -1.33268535e+00
5.32160215e-02 2.03670710e-02 -1.92099750e-01 4.31620955e-01
1.62794486e-01 -5.07053912e-01 -1.02009821e+00 -2.96369225e-01
-2.04623237e-01 -6.34851992e-01 9.27045465e-01 3.43566030e-01
1.14429247e+00 -1.82322875e-01 -7.70602107e-01 6.81352496e-01
1.31122172e+00 4.01681429e-03 3.66180807e-01 2.32301593e-01
6.03988409e-01 2.86071897e-01 2.04168603e-01 3.57604027e-02
4.55505699e-01 2.20554084e-01 -1.45827740e-01 -5.11244535e-01
-1.33464530e-01 -1.38090625e-01 1.74435303e-01 7.53052294e-01
2.41497368e-01 -4.25115019e-01 -1.54270506e+00 7.91072428e-01
-1.50422823e+00 -5.03992021e-01 3.08174998e-01 1.70648086e+00
1.50768697e+00 1.05018631e-01 -8.99886563e-02 -4.95356530e-01
4.68195647e-01 -3.79435569e-02 -6.15316808e-01 -2.08724990e-01
-4.22257930e-02 3.75485152e-01 4.23883617e-01 1.40277743e-01
-1.30505645e+00 1.14641786e+00 6.67313576e+00 9.40591753e-01
-1.25328732e+00 3.44889104e-01 9.85757947e-01 1.20979309e-01
-1.54226169e-01 -4.13091689e-01 -1.00468147e+00 2.09976524e-01
1.36325097e+00 -3.18252951e-01 5.51095605e-02 9.85707343e-01
-1.15031712e-02 5.05313575e-01 -1.43729067e+00 8.46972644e-01
-9.07237008e-02 -1.46994960e+00 8.87430161e-02 6.47895262e-02
6.83502734e-01 8.82046461e-01 -2.52008468e-01 8.25109184e-01
9.00568604e-01 -1.20635438e+00 6.71740249e-02 1.35390431e-01
7.83675969e-01 -3.16686690e-01 9.89895821e-01 4.41375256e-01
-6.67998314e-01 2.69376844e-01 -5.50872862e-01 2.63832659e-01
1.88923389e-01 9.37360704e-01 -1.11278379e+00 5.15429974e-01
5.26077867e-01 7.75370061e-01 -4.31915760e-01 9.60501194e-01
-2.11165994e-01 9.78507102e-01 -3.35455596e-01 1.80330887e-01
2.84258425e-01 3.97613645e-01 4.65936810e-02 1.73406243e+00
-3.68776992e-02 2.58957893e-01 3.15148771e-01 7.47535825e-01
-7.93252885e-01 3.03584337e-01 -5.63527167e-01 -5.73490798e-01
2.30355695e-01 1.26984966e+00 -6.68618917e-01 -6.21965170e-01
-6.00147605e-01 7.07381845e-01 6.72960222e-01 3.96355212e-01
-7.68214226e-01 -9.78477821e-02 6.77133083e-01 -1.31400138e-01
1.56855043e-02 -2.01873124e-01 -4.31874037e-01 -1.27747226e+00
-6.03731811e-01 -1.10563457e+00 7.62351096e-01 -5.53039014e-01
-1.71955836e+00 6.60210192e-01 -1.18415833e-01 -7.31102169e-01
-2.09298059e-02 -8.80838454e-01 -3.92685413e-01 7.61946321e-01
-1.71579885e+00 -1.17010236e+00 1.54073969e-01 5.55755138e-01
4.94142234e-01 -4.77676168e-02 1.11930513e+00 5.74741483e-01
-6.34991646e-01 7.23736763e-01 9.61485133e-02 6.95689321e-01
1.23533225e+00 -1.18926895e+00 5.52174747e-01 5.66486418e-01
3.76077205e-01 1.16480577e+00 3.88259262e-01 -6.41035676e-01
-1.12848735e+00 -1.32476377e+00 1.29272950e+00 -6.23620629e-01
8.14802647e-01 -4.40480947e-01 -9.29392636e-01 1.00533450e+00
2.35565558e-01 1.89244762e-01 1.16039300e+00 6.44267321e-01
-2.72162646e-01 5.24952337e-02 -1.08607507e+00 4.56659019e-01
1.01294434e+00 -5.35042763e-01 -8.99987757e-01 6.80980742e-01
7.62213171e-01 -3.83483559e-01 -1.16451633e+00 4.69550967e-01
4.15645927e-01 -2.90213972e-01 1.07227004e+00 -1.24039471e+00
5.80900490e-01 1.99371681e-01 1.17323443e-01 -1.34360170e+00
-4.88066554e-01 -4.00104374e-01 2.05738872e-01 1.22416520e+00
8.14844787e-01 -5.95352232e-01 8.45516503e-01 8.15037131e-01
-3.58021438e-01 -7.78183937e-01 -5.99156797e-01 -6.92709148e-01
6.31370485e-01 -5.42624772e-01 1.54455706e-01 1.36963904e+00
1.42524168e-01 6.29071116e-01 -1.40332803e-01 2.06942037e-02
1.74804643e-01 -2.55785823e-01 5.13591945e-01 -1.14991069e+00
-3.39575976e-01 -2.79931366e-01 -6.14558831e-02 -1.05975389e+00
3.90188873e-01 -1.27639818e+00 2.23253384e-01 -1.77921522e+00
5.04039645e-01 -4.97058570e-01 -8.67731094e-01 1.24951935e+00
-3.00286949e-01 2.04924569e-01 -3.59991305e-02 1.75591901e-01
-7.15051949e-01 4.46815342e-02 1.14345837e+00 -2.53165752e-01
-1.73143838e-02 -3.44685912e-01 -1.07578874e+00 8.11622500e-01
8.52732062e-01 -7.26456225e-01 -2.32718438e-01 -8.74525189e-01
2.35370725e-01 -3.99544761e-02 2.72928104e-02 -8.48377705e-01
1.92109570e-01 -1.22785643e-01 3.40024143e-01 -1.79234982e-01
-1.11055095e-02 -5.57496428e-01 -1.74226224e-01 4.67041910e-01
-6.43558860e-01 -2.57222980e-01 6.06339633e-01 3.76634538e-01
-2.71914631e-01 -9.81314406e-02 7.38767743e-01 -3.85083526e-01
-7.56425440e-01 4.18755084e-01 -2.51807898e-01 5.41926801e-01
4.70244825e-01 2.92986810e-01 -5.02979279e-01 -1.16320699e-01
-9.79674637e-01 4.09485340e-01 1.28950536e-01 2.56317556e-01
2.38431528e-01 -9.39386427e-01 -9.41315413e-01 -3.15470695e-02
1.50374591e-01 1.11275733e-01 3.27004820e-01 9.77357864e-01
-4.04588461e-01 7.76419938e-01 -7.47940019e-02 -4.59791809e-01
-1.05804110e+00 6.28524005e-01 2.35716969e-01 -9.62581396e-01
-5.92645943e-01 1.15482342e+00 6.83423460e-01 -7.66058564e-01
1.83104217e-01 -7.62389839e-01 -1.17307417e-01 -2.09909961e-01
5.30239642e-01 -2.91333407e-01 2.40704000e-01 -1.15767233e-01
-6.51444256e-01 -2.76887510e-02 -3.62649739e-01 1.57164857e-01
1.60308290e+00 2.17982516e-01 9.44353566e-02 3.07485402e-01
1.26439703e+00 -1.15936309e-01 -8.38415980e-01 -2.92770922e-01
3.07476193e-01 4.03391361e-01 6.92834035e-02 -1.37495077e+00
-9.29538310e-01 9.57096517e-01 2.78097451e-01 -2.34050080e-01
9.61888969e-01 1.04340762e-01 9.54649508e-01 9.33646619e-01
3.87646288e-01 -8.01913857e-01 -2.02808335e-01 7.54797161e-01
6.87762558e-01 -1.31929195e+00 -1.21707022e-01 -2.44266734e-01
-6.45186663e-01 8.19997311e-01 4.87316966e-01 7.99658000e-02
8.25537682e-01 5.67212522e-01 2.60405451e-01 -1.78253725e-01
-8.38161349e-01 -7.39936307e-02 3.15902591e-01 6.31893635e-01
1.05253065e+00 2.46400312e-02 -2.67331272e-01 1.14495802e+00
-5.95371015e-02 5.71898460e-01 1.54040247e-01 8.60772014e-01
-1.06741846e-01 -1.41358852e+00 1.39130279e-03 4.96498555e-01
-1.02257228e+00 -6.53281152e-01 -3.38141173e-01 9.00315046e-01
1.40414312e-01 6.91027939e-01 -2.45253712e-01 1.06026502e-02
2.13758484e-01 4.64433104e-01 2.72720277e-01 -1.03389096e+00
-6.93547130e-01 -3.48234214e-02 4.74236876e-01 -4.94272530e-01
-2.46928781e-01 -3.08330476e-01 -1.39979219e+00 4.63398499e-03
-1.22421822e-02 3.65132391e-01 3.94491166e-01 9.60249424e-01
6.39435470e-01 7.06900179e-01 -3.67780417e-01 -2.38570958e-01
-5.18070757e-01 -1.03584671e+00 -3.36120844e-01 4.33025837e-01
1.90948382e-01 -3.91776800e-01 -8.32762495e-02 2.87733972e-01]
|
[8.631025314331055, 8.674267768859863]
|
599137ec-22de-409d-9d74-88ad61ae6b94
|
few-shot-novel-concept-learning-for-semantic
| null | null |
https://aclanthology.org/2021.findings-emnlp.177
|
https://aclanthology.org/2021.findings-emnlp.177.pdf
|
Few-Shot Novel Concept Learning for Semantic Parsing
|
Humans are capable of learning novel concepts from very few examples; in contrast, state-of-the-art machine learning algorithms typically need thousands of examples to do so. In this paper, we propose an algorithm for learning novel concepts by representing them as programs over existing concepts. This way the concept learning problem is naturally a program synthesis problem and our algorithm learns from a few examples to synthesize a program representing the novel concept. In addition, we perform a theoretical analysis of our approach for the case where the program defining the novel concept over existing ones is context-free. We show that given a learned grammar-based parser and a novel production rule, we can augment the parser with the production rule in a way that provably generalizes. We evaluate our approach by learning concepts in the semantic parsing domain extended to the few-shot novel concept learning setting, showing that our approach significantly outperforms end-to-end neural semantic parsers.
|
['Dan Roth', 'Osbert Bastani', 'Soham Dan']
| null | null | null | null |
findings-emnlp-2021-11
|
['novel-concepts']
|
['reasoning']
|
[ 5.35081267e-01 6.68853700e-01 -9.47408378e-02 -6.69512630e-01
-8.05502355e-01 -6.41728163e-01 4.92566615e-01 7.41081715e-01
-5.26408851e-01 5.82384706e-01 -9.87150520e-02 -3.62122476e-01
1.80769116e-01 -1.28538132e+00 -1.26170766e+00 -4.53454375e-01
-1.87525302e-01 6.95477724e-01 4.94730473e-01 -1.79463461e-01
3.80180150e-01 6.26969635e-02 -1.88515687e+00 3.80243331e-01
9.34467971e-01 2.62213975e-01 6.04317427e-01 9.71470952e-01
-7.01013267e-01 3.91662747e-01 -5.70834816e-01 -3.24708343e-01
8.34148675e-02 -7.05912828e-01 -9.82482076e-01 -1.60986170e-01
5.53465664e-01 8.75963867e-02 2.07712948e-01 1.11582100e+00
-9.88634601e-02 5.85893095e-01 2.44316176e-01 -1.15072215e+00
-7.02103317e-01 9.04773891e-01 -6.54300600e-02 7.54558342e-03
5.79809248e-01 1.08194299e-01 1.32303452e+00 -7.12666631e-01
9.91133571e-01 1.34747350e+00 4.06045675e-01 1.03434956e+00
-1.53063512e+00 -4.31040585e-01 3.56252640e-01 2.12435108e-02
-7.25504637e-01 5.21266572e-02 6.88323557e-01 -2.96234667e-01
1.28777862e+00 -2.05983207e-01 4.79093999e-01 8.37681413e-01
-8.98990110e-02 7.70381093e-01 9.35073912e-01 -1.00360942e+00
6.56774402e-01 2.63807088e-01 4.91610825e-01 9.81498718e-01
2.98926443e-01 3.52833807e-01 -2.97626823e-01 -2.52526313e-01
1.85190752e-01 -1.81420855e-02 3.39793652e-01 -6.14175498e-01
-8.30663562e-01 1.31626320e+00 2.31114566e-01 5.70513070e-01
2.62987055e-02 4.94653225e-01 7.28232920e-01 5.06813943e-01
2.16980815e-01 6.95281327e-01 -6.53324723e-01 3.63212153e-02
-6.69007897e-01 6.41704500e-01 1.02051985e+00 1.22677827e+00
1.11990130e+00 -1.02435842e-01 2.77618170e-01 5.75825036e-01
-1.30861863e-01 2.66866148e-01 6.58519030e-01 -8.95011067e-01
2.46276170e-01 4.30404812e-01 -2.67747402e-01 -4.76626664e-01
-2.82994628e-01 -3.30314189e-02 9.26043242e-02 1.77126363e-01
1.53518990e-01 -2.57585287e-01 -6.88710630e-01 2.08377099e+00
2.16909483e-01 2.00806916e-01 5.20009816e-01 7.62397870e-02
4.40851659e-01 8.31473827e-01 5.18343031e-01 -4.24395293e-01
1.01969445e+00 -9.73380804e-01 -1.78170055e-01 -5.87881804e-01
1.07768250e+00 -9.50183794e-02 1.43820250e+00 2.81982839e-01
-7.59567082e-01 -8.34029555e-01 -9.75313246e-01 2.18258165e-02
-5.19765198e-01 -3.68640751e-01 9.60233331e-01 6.24384880e-01
-9.11617875e-01 8.13374460e-01 -5.36529243e-01 -9.04547155e-01
5.63886583e-01 1.09586835e-01 -2.78073788e-01 -3.91320020e-01
-7.98586428e-01 6.99639201e-01 1.28885567e+00 -8.36912930e-01
-9.48210180e-01 -7.15128124e-01 -1.12941909e+00 3.01122904e-01
7.10646629e-01 -8.93059850e-01 1.71453440e+00 -1.50574934e+00
-1.15221703e+00 1.01381719e+00 -3.56267005e-01 -7.75516748e-01
-9.45285410e-02 -7.61538148e-02 -2.25118697e-01 3.38189870e-01
3.61553341e-01 7.41431594e-01 7.39913106e-01 -1.22816992e+00
-1.04244149e+00 -2.99219370e-01 7.22050846e-01 -1.13817476e-01
3.89511138e-02 -4.05542785e-03 3.46882194e-01 -6.07870519e-01
-1.07618324e-01 -8.79643321e-01 -3.83805901e-01 1.03192769e-01
1.75804704e-01 -5.62026143e-01 6.87305689e-01 -1.04954295e-01
7.94514298e-01 -2.28152966e+00 1.30709827e-01 -2.36579813e-02
-8.38475376e-02 5.96724264e-02 -2.19841212e-01 3.84398848e-01
-1.49848565e-01 1.18000403e-01 -8.46700311e-01 5.45134395e-02
1.22838125e-01 6.60472572e-01 -5.32691181e-01 -1.86300963e-01
4.07372087e-01 9.05994117e-01 -1.46435869e+00 -2.91776001e-01
8.47746711e-03 -3.09684545e-01 -1.00497043e+00 3.48154396e-01
-1.23404682e+00 3.65956090e-02 -5.53053439e-01 2.97059089e-01
2.85072267e-01 7.72033781e-02 4.41291958e-01 5.92836916e-01
3.39954019e-01 1.19249113e-01 -9.14384723e-01 2.24589729e+00
-9.37159777e-01 5.87276936e-01 -4.38308239e-01 -1.55578804e+00
1.16379416e+00 8.96596164e-02 -2.05664247e-01 -2.05703840e-01
-1.67361632e-01 2.65284896e-01 -6.06689565e-02 -8.13307583e-01
2.90228993e-01 -1.01028836e+00 -4.19836819e-01 5.51113844e-01
5.70178509e-01 -3.17050129e-01 4.49973524e-01 4.50866014e-01
1.11533141e+00 3.29263568e-01 5.67343950e-01 -4.04243231e-01
4.51022059e-01 4.41412330e-01 4.57232952e-01 1.18698394e+00
6.77171126e-02 5.45190684e-02 6.23058915e-01 -5.61283529e-01
-9.28081453e-01 -1.26661944e+00 4.01383907e-01 1.53033710e+00
-2.55696207e-01 -6.36150539e-01 -8.86271715e-01 -1.00149345e+00
-7.79558122e-02 1.13770258e+00 -5.17940402e-01 -2.52610058e-01
-8.38145792e-01 -2.43775547e-01 3.32310110e-01 6.87834084e-01
2.59790361e-01 -1.27717733e+00 -1.13654053e+00 5.27726710e-01
2.01676041e-01 -9.79891837e-01 -6.38687611e-03 4.10860181e-01
-1.35129690e+00 -1.30262208e+00 -1.09976299e-01 -1.35688341e+00
7.60273576e-01 -1.25146866e-01 1.28065050e+00 2.42268801e-01
-4.05290365e-01 5.93180716e-01 -5.45725584e-01 -4.64561671e-01
-9.78803456e-01 -3.83744128e-02 -2.24564090e-01 -7.09023237e-01
6.84483469e-01 -7.75989115e-01 2.32637182e-01 -8.20175886e-01
-1.16013074e+00 2.82176528e-02 4.76025611e-01 1.05956769e+00
3.42400581e-01 3.09057504e-01 1.01539016e+00 -1.69587576e+00
4.51092422e-01 -5.96189737e-01 -6.91219628e-01 4.02025133e-01
-3.97710949e-01 6.54632330e-01 1.07635784e+00 -4.25664485e-01
-1.40365720e+00 6.07146800e-01 -7.32915327e-02 1.43754587e-01
-6.49222016e-01 7.33441949e-01 -2.12315336e-01 1.91804111e-01
1.09220958e+00 2.51934826e-01 -3.54677886e-01 -3.50345045e-01
9.39399064e-01 1.15227342e-01 7.68652737e-01 -1.41730499e+00
8.94469619e-01 2.98529953e-01 -5.68555109e-02 -6.15884304e-01
-8.52110445e-01 -1.54328391e-01 -8.28400970e-01 3.75646740e-01
6.72954619e-01 -6.35271788e-01 -2.15156883e-01 -1.23890407e-01
-1.43571329e+00 -3.14922780e-01 -6.74867690e-01 6.12731501e-02
-9.99057353e-01 2.29088068e-01 -8.94247442e-02 -7.00690269e-01
-6.38627112e-02 -8.44926119e-01 7.32592344e-01 3.29188019e-01
-3.33110660e-01 -1.05148804e+00 3.03016335e-01 -2.37814948e-01
7.58453086e-02 5.49394011e-01 1.82730007e+00 -1.07683623e+00
-5.43826640e-01 9.10549611e-03 2.39894584e-01 3.72529447e-01
-5.03615662e-02 -2.86959797e-01 -1.06020534e+00 -1.07136644e-01
2.89538465e-02 -3.55217725e-01 8.48457932e-01 4.64762002e-02
1.19144702e+00 -3.86502683e-01 -4.66001391e-01 1.84202969e-01
1.84241569e+00 3.58436227e-01 3.30877870e-01 1.46972045e-01
3.27711642e-01 8.39158475e-01 6.79107726e-01 4.52761335e-04
7.43728802e-02 1.52232543e-01 -6.90091550e-02 7.10251689e-01
2.51286209e-01 -4.75363135e-01 2.60185659e-01 2.22414061e-01
3.89013767e-01 2.87292153e-01 -9.92569268e-01 9.34513271e-01
-1.78216147e+00 -9.25553024e-01 5.32722950e-01 1.98570418e+00
1.02758527e+00 4.18094307e-01 1.28794998e-01 -1.79879010e-01
7.68817663e-01 -1.29127875e-01 -3.79079372e-01 -9.85481262e-01
1.10844783e-01 1.18877625e+00 -1.18541233e-01 6.06031716e-01
-9.68850672e-01 1.34482598e+00 6.25841379e+00 4.71457303e-01
-6.50664985e-01 2.58636892e-01 6.16678223e-02 3.49687077e-02
-4.64490205e-01 5.47743499e-01 -5.68122387e-01 -4.10878845e-02
1.28620851e+00 -7.11649120e-01 3.56167823e-01 1.34560299e+00
-4.74369287e-01 -2.38459826e-01 -1.69061768e+00 5.41356683e-01
1.83837444e-01 -1.47605860e+00 2.59943962e-01 -6.65795743e-01
7.35947073e-01 -1.55651808e-01 -4.09466118e-01 8.36495817e-01
5.90944111e-01 -7.62017906e-01 3.65301669e-01 1.05527200e-01
4.24731076e-01 -9.68763173e-01 1.68275490e-01 4.39556688e-01
-1.00259376e+00 -5.61209440e-01 -4.63310987e-01 -3.10992777e-01
-1.19025491e-01 4.74909037e-01 -7.27754056e-01 3.17810357e-01
3.48865300e-01 6.90777898e-01 -5.97895503e-01 7.59967864e-01
-6.62709475e-01 4.19639885e-01 1.24142272e-03 -2.04939887e-01
3.17856818e-01 9.86458585e-02 4.11119252e-01 1.44726825e+00
3.24033767e-01 3.73365849e-01 2.96182841e-01 1.30161786e+00
-2.90411860e-01 -1.14609092e-01 -1.00751233e+00 -2.43968159e-01
5.05776227e-01 8.92704427e-01 -8.05359840e-01 -8.45967531e-01
-7.29894817e-01 8.91084731e-01 4.62842166e-01 2.78731078e-01
-3.38113815e-01 -8.07294965e-01 4.08419967e-01 -2.60202765e-01
5.29830933e-01 3.93115096e-02 -5.84153198e-02 -1.05611992e+00
2.57462412e-01 -8.21419716e-01 6.27747774e-01 -8.59957039e-01
-1.24971080e+00 3.46015394e-01 2.63247281e-01 -5.10493040e-01
-6.57791018e-01 -7.98306465e-01 -9.57713664e-01 6.05325997e-01
-1.29011297e+00 -8.73044908e-01 -1.03299608e-02 1.84272334e-01
9.69745576e-01 -2.17490986e-01 1.27966225e+00 -2.95254737e-01
3.01311582e-01 3.28617007e-01 -1.71028480e-01 -1.87380775e-03
2.63175398e-01 -1.40423119e+00 8.47694099e-01 1.00996184e+00
3.82069528e-01 8.96530092e-01 8.26692164e-01 -4.47269678e-01
-1.31181300e+00 -1.16525531e+00 9.68767881e-01 -3.72705907e-01
6.62943602e-01 -5.20104170e-01 -1.19108593e+00 1.02702928e+00
-1.05387636e-01 7.30199218e-02 6.88016832e-01 2.59039402e-01
-9.27286208e-01 1.89444020e-01 -1.24891210e+00 3.90707493e-01
1.05395901e+00 -7.78956473e-01 -1.47548079e+00 4.29765910e-01
1.29840255e+00 -1.45533413e-01 -4.03675497e-01 -1.08562997e-02
2.87651151e-01 -5.84091246e-01 7.36195326e-01 -1.21373665e+00
7.50713587e-01 -1.47923261e-01 -2.96332955e-01 -1.45822370e+00
1.52509600e-01 -4.23968971e-01 -2.06923753e-01 9.50990081e-01
3.75880867e-01 -4.90272999e-01 7.09040046e-01 3.78333390e-01
-2.63063431e-01 -4.56431597e-01 -7.68667996e-01 -1.19255805e+00
6.95251167e-01 -5.68860233e-01 4.83402938e-01 7.17131257e-01
4.77645963e-01 5.32651842e-01 3.81856889e-01 -3.49062607e-02
6.29873872e-01 4.68959659e-01 6.14864349e-01 -1.39267993e+00
-8.81515503e-01 -2.51944780e-01 -3.67643803e-01 -7.46063411e-01
7.94375479e-01 -1.21866751e+00 3.86660069e-01 -1.21726012e+00
4.86281335e-01 -4.29999888e-01 4.20839600e-02 4.30467814e-01
-2.81013459e-01 -5.67570865e-01 2.32886165e-01 -4.33782607e-01
-4.20710027e-01 8.76639932e-02 6.49251580e-01 -3.20130825e-01
-2.23384351e-01 -2.80936211e-01 -1.02675617e+00 7.46805727e-01
8.55963349e-01 -9.52227056e-01 -6.36800706e-01 -3.64401311e-01
3.49182934e-01 -1.30823389e-01 4.92707640e-01 -9.04197395e-01
2.20372468e-01 -2.78215468e-01 -2.02704862e-01 8.00766051e-02
-2.39565015e-01 -6.54715776e-01 -3.08825046e-01 6.67459786e-01
-7.45137334e-01 -1.58539563e-01 4.03079003e-01 8.08917403e-01
-9.51905623e-02 -9.49183345e-01 8.76344562e-01 -7.39868343e-01
-1.30976307e+00 -7.01545253e-02 -3.70876431e-01 5.64327955e-01
1.00758421e+00 5.91313802e-02 -1.93990141e-01 7.50824139e-02
-8.74019146e-01 -6.39137626e-02 4.90140676e-01 5.26269317e-01
5.92899621e-01 -9.04067338e-01 -4.41514373e-01 1.64705366e-01
7.61480927e-01 5.97049445e-02 -6.75396845e-02 -1.49811223e-01
-5.85488975e-01 2.03346044e-01 -1.20474458e-01 -3.80960137e-01
-1.17626095e+00 1.06322753e+00 1.82173073e-01 3.35616954e-02
-7.85918295e-01 7.67159462e-01 3.66163850e-01 -8.26899409e-01
2.05827832e-01 -5.50463438e-01 2.03850687e-01 -5.42907059e-01
6.67636693e-01 -1.77497655e-01 -2.34512553e-01 -5.26870377e-02
-2.15281453e-02 3.39777321e-01 -2.04566181e-01 -9.88723710e-02
1.48417163e+00 2.41975740e-01 -1.96376801e-01 5.48821986e-01
1.29808640e+00 -4.64586735e-01 -8.70370924e-01 -5.54565787e-01
7.01729298e-01 -4.31598991e-01 -5.86058736e-01 -7.26612985e-01
-4.39829797e-01 9.91187632e-01 4.36925441e-01 -2.10560516e-01
1.16084635e+00 5.23256421e-01 7.64043272e-01 1.30952859e+00
6.15425944e-01 -1.16837907e+00 3.10129732e-01 7.04334855e-01
4.14134234e-01 -1.20055878e+00 -2.39716142e-01 -4.16749656e-01
-2.60766238e-01 1.42612994e+00 5.30253172e-01 -5.08589745e-01
2.46460110e-01 9.68452394e-02 -5.00956237e-01 -2.22347662e-01
-8.23893130e-01 -1.11658283e-01 -2.49544069e-01 1.08398521e+00
2.64907718e-01 1.27571151e-01 -2.19953433e-01 8.69016945e-01
-2.34403655e-01 1.30921409e-01 8.76712322e-01 1.50448442e+00
-1.00170386e+00 -1.43789804e+00 1.61874473e-01 3.92361432e-01
-1.74460739e-01 -3.28483135e-01 -1.81661844e-01 7.90112734e-01
2.31871754e-01 6.58466637e-01 4.69677076e-02 5.57794273e-02
2.61750162e-01 6.82827055e-01 8.27894807e-01 -1.68090630e+00
-3.91416430e-01 -5.30773818e-01 7.84919932e-02 -4.31368291e-01
-5.85395575e-01 -5.71968794e-01 -1.86741650e+00 6.26115054e-02
-5.04038520e-02 3.15068960e-01 5.66121340e-01 1.20136893e+00
1.19157219e-02 4.39033478e-01 3.74246567e-01 -4.12955955e-02
-5.10272145e-01 -5.27287364e-01 -4.80037272e-01 4.70705777e-01
2.57399648e-01 -4.07596737e-01 -2.46249631e-01 6.17517412e-01]
|
[10.505653381347656, 8.979100227355957]
|
8856f9a7-f8aa-4f6b-bb3a-34d65dab23d9
|
an-analysis-of-annotated-corpora-for-emotion
| null | null |
https://aclanthology.org/C18-1179
|
https://aclanthology.org/C18-1179.pdf
|
An Analysis of Annotated Corpora for Emotion Classification in Text
|
Several datasets have been annotated and published for classification of emotions. They differ in several ways: (1) the use of different annotation schemata (e. g., discrete label sets, including joy, anger, fear, or sadness or continuous values including valence, or arousal), (2) the domain, and, (3) the file formats. This leads to several research gaps: supervised models often only use a limited set of available resources. Additionally, no previous work has compared emotion corpora in a systematic manner. We aim at contributing to this situation with a survey of the datasets, and aggregate them in a common file format with a common annotation schema. Based on this aggregation, we perform the first cross-corpus classification experiments in the spirit of future research enabled by this paper, in order to gain insight and a better understanding of differences of models inferred from the data. This work also simplifies the choice of the most appropriate resources for developing a model for a novel domain. One result from our analysis is that a subset of corpora is better classified with models trained on a different corpus. For none of the corpora, training on all data altogether is better than using a subselection of the resources. Our unified corpus is available at http://www.ims.uni-stuttgart.de/data/unifyemotion.
|
['Laura-Ana-Maria Bostan', 'Roman Klinger']
|
2018-08-01
|
an-analysis-of-annotated-corpora-for-emotion-1
|
https://aclanthology.org/C18-1179
|
https://aclanthology.org/C18-1179.pdf
|
coling-2018-8
|
['cross-corpus']
|
['computer-vision']
|
[ 3.19415554e-02 -4.81600761e-02 -2.40005150e-01 -7.81103909e-01
-2.64670432e-01 -7.45539367e-01 6.01517320e-01 5.55447102e-01
-4.33204055e-01 7.44665623e-01 2.38774866e-01 8.17507058e-02
-1.97362006e-01 -5.05095303e-01 -4.84793857e-02 -4.82650936e-01
1.57383934e-01 5.82565486e-01 -1.17049776e-01 -3.66034716e-01
2.23716453e-01 4.44104999e-01 -1.97414815e+00 4.98962075e-01
4.86725241e-01 1.06563389e+00 -1.95349008e-02 2.90177494e-01
-2.66928881e-01 7.01607466e-01 -5.90802670e-01 -5.81553817e-01
1.79781858e-02 -7.19871938e-01 -1.18756497e+00 -2.83353124e-02
-6.97912946e-02 1.92025140e-01 3.15215498e-01 8.18490863e-01
3.65674198e-01 6.13169819e-02 5.83095789e-01 -1.45662653e+00
-3.57738853e-01 6.27741992e-01 4.39782478e-02 -1.33670256e-01
5.72534204e-01 -8.09386820e-02 1.01919723e+00 -5.19876420e-01
9.07875359e-01 1.02981758e+00 4.89419520e-01 7.23425031e-01
-1.12342978e+00 -5.61945558e-01 3.26471925e-02 1.46659762e-01
-1.28368878e+00 -5.63408971e-01 9.24733579e-01 -4.44077551e-01
1.02366793e+00 5.71051478e-01 6.99889123e-01 1.34898734e+00
-2.35493034e-01 3.50297272e-01 1.64366603e+00 -6.11036837e-01
1.97469115e-01 8.14619064e-01 2.75013685e-01 8.59583095e-02
2.33929396e-01 -2.99019307e-01 -2.76659787e-01 -1.55084953e-01
1.92639217e-01 -4.77455556e-01 -2.69103438e-01 -2.89324045e-01
-9.21080649e-01 9.20419455e-01 -2.05890760e-01 9.95341659e-01
-1.06602289e-01 -6.03879035e-01 9.59433615e-01 6.59315526e-01
6.85927212e-01 7.01197445e-01 -8.04442465e-01 -5.85615158e-01
-9.19935167e-01 2.54297018e-01 1.36836112e+00 5.79398215e-01
7.68483937e-01 -1.03677228e-01 4.25135732e-01 1.06602752e+00
1.57110661e-01 -1.14859566e-01 6.82268500e-01 -8.55499804e-01
1.39886409e-01 5.45705140e-01 2.04341948e-01 -1.00155735e+00
-7.36577094e-01 -1.35886088e-01 -4.14115667e-01 1.91355541e-01
6.74718082e-01 -2.68997014e-01 -3.15799326e-01 1.82909238e+00
1.64622977e-01 -3.36351246e-01 3.39646161e-01 8.86422515e-01
1.03267503e+00 3.18449020e-01 1.34153545e-01 -4.50695038e-01
1.48932755e+00 -5.47518492e-01 -9.62285280e-01 -8.54786485e-02
1.10033619e+00 -9.70928013e-01 1.13282204e+00 8.29511225e-01
-9.64455724e-01 -3.74655992e-01 -9.35985863e-01 -4.99099120e-02
-9.87657487e-01 3.16510238e-02 8.07458699e-01 9.83039618e-01
-9.09047246e-01 5.49068272e-01 -6.24870062e-01 -8.65880132e-01
-8.11449289e-02 1.67841405e-01 -4.15338695e-01 3.35171133e-01
-1.45613003e+00 1.26799512e+00 6.78920746e-01 -2.36701876e-01
1.32130414e-01 -3.80034834e-01 -8.72142375e-01 -3.09259713e-01
1.87512487e-01 -1.77460432e-01 1.11123168e+00 -1.63218629e+00
-1.50785899e+00 1.38214183e+00 1.33451238e-01 -3.63801539e-01
3.79169554e-01 3.27224657e-02 -7.07909465e-01 -4.49221693e-02
-6.20718747e-02 5.58772326e-01 1.08401857e-01 -1.25086224e+00
-4.69941348e-01 -4.94795173e-01 5.54900169e-02 9.47918966e-02
-4.24527645e-01 6.16998553e-01 -1.72038347e-01 -5.80213070e-01
-3.18017483e-01 -9.09696221e-01 6.68993518e-02 -5.02536714e-01
-3.74612340e-04 -2.87807554e-01 6.35693312e-01 -5.22067547e-01
1.42820859e+00 -2.22378039e+00 1.78748116e-01 -1.13937221e-02
-2.30557308e-01 1.12842664e-01 2.02061921e-01 8.69710863e-01
-4.00206059e-01 4.69848007e-01 -1.62240282e-01 -4.04487818e-01
2.07695931e-01 5.25053322e-01 -2.78367430e-01 3.95155966e-01
1.35043666e-01 3.09675097e-01 -7.53952861e-01 -6.52107000e-01
1.70375362e-01 3.92692208e-01 -2.50777185e-01 -7.82351345e-02
2.55346354e-02 3.50064784e-01 -2.86516070e-01 4.69415009e-01
4.41209227e-01 2.32369408e-01 3.43666494e-01 -1.23192772e-01
-3.93636882e-01 4.85645175e-01 -1.25369143e+00 1.62811410e+00
-5.36560953e-01 6.56495154e-01 -4.69151065e-02 -1.23835933e+00
1.24837410e+00 7.26786315e-01 7.38134742e-01 -3.49549890e-01
4.41482365e-01 4.28540796e-01 1.51472107e-01 -5.38982928e-01
6.64283872e-01 -2.54296333e-01 -3.73446792e-01 3.26976091e-01
3.16907942e-01 -2.45083898e-01 6.09697223e-01 -2.33004749e-01
5.67644000e-01 1.78780213e-01 5.01644969e-01 -2.74542302e-01
4.11646456e-01 1.01878703e-01 6.01072431e-01 3.60200763e-01
-3.09133291e-01 4.72404957e-01 8.33796918e-01 -3.63192558e-01
-9.29193020e-01 -4.47759211e-01 -7.66425252e-01 8.82410765e-01
-3.09367150e-01 -8.10660720e-01 -6.19498432e-01 -3.90182823e-01
-3.73288095e-01 9.59248066e-01 -6.91214919e-01 1.02085643e-01
-1.83408096e-01 -7.42758036e-01 7.59109557e-01 1.77467138e-01
9.90585759e-02 -1.29477119e+00 -9.18505073e-01 1.01582296e-01
-3.52525383e-01 -1.07004440e+00 5.14879882e-01 5.47637343e-01
-8.25000048e-01 -9.20964479e-01 -8.93730596e-02 -5.01516640e-01
2.09624082e-01 -3.65717679e-01 1.26637018e+00 8.85757059e-02
1.11476161e-01 3.50187659e-01 -8.88057947e-01 -7.55146921e-01
-5.58876932e-01 1.70186028e-01 1.41154472e-02 -1.45355344e-01
8.15687120e-01 -3.93076211e-01 -9.82197374e-03 2.70063639e-01
-1.03684330e+00 -1.27766088e-01 1.93972751e-01 5.72931707e-01
3.01647425e-01 1.31038547e-01 7.11679816e-01 -1.05332959e+00
8.34097981e-01 -7.59193063e-01 -1.79249927e-01 2.61650272e-02
-6.94061100e-01 -3.05799127e-01 5.91540456e-01 -3.23997736e-01
-9.95798409e-01 -7.98324645e-02 -3.06155682e-01 -1.17187388e-01
-8.24725449e-01 7.59976387e-01 -9.39817503e-02 3.17457825e-01
6.01766646e-01 -1.43240288e-01 -4.51913141e-02 -5.64768136e-01
1.32593229e-01 9.11781967e-01 2.07813561e-01 -7.07918525e-01
1.60618961e-01 1.76962271e-01 -2.89515942e-01 -8.03005576e-01
-6.30938172e-01 -4.75337654e-01 -9.65391040e-01 -3.26991439e-01
8.21224570e-01 -6.16153836e-01 -3.56434762e-01 2.67209619e-01
-8.57001841e-01 -3.91755730e-01 -3.37369770e-01 6.36281431e-01
-5.37517905e-01 2.89673775e-01 -4.98404801e-01 -9.99442756e-01
-8.43257904e-02 -9.68695581e-01 6.73987508e-01 8.10601860e-02
-1.02124763e+00 -1.14170742e+00 2.37827852e-01 4.71059829e-01
1.97762519e-01 5.94504774e-01 7.16615975e-01 -1.17289853e+00
5.88419259e-01 -3.16260755e-01 2.73619711e-01 4.74759936e-01
2.26512149e-01 4.23430234e-01 -1.07496393e+00 -1.51952179e-02
3.32577527e-01 -7.53757358e-01 4.23102468e-01 -1.46474957e-01
1.01328933e+00 -5.88957258e-02 -7.07908869e-02 1.29278436e-01
1.39351773e+00 4.39092577e-01 6.32235765e-01 6.53181434e-01
1.45099089e-01 1.13943756e+00 1.08606565e+00 5.70007741e-01
3.26534122e-01 8.10240328e-01 2.47574553e-01 1.35269046e-01
3.30988675e-01 2.80799776e-01 4.06196713e-01 7.69149959e-01
-2.40885720e-01 -2.20190555e-01 -9.61372256e-01 5.80021262e-01
-1.71957147e+00 -9.78973150e-01 -3.08869421e-01 2.14557791e+00
9.16908681e-01 1.37641639e-01 4.68034804e-01 5.11586785e-01
3.93814236e-01 1.47441238e-01 4.96905530e-03 -1.06762230e+00
-1.51590988e-01 3.43722224e-01 4.34281006e-02 3.76978666e-01
-9.02996123e-01 7.38974273e-01 5.77483892e+00 5.45005500e-01
-1.34445000e+00 3.45268920e-02 5.42451799e-01 -1.31041333e-01
-1.23805694e-01 6.69258386e-02 -4.70075995e-01 5.24274886e-01
1.33709502e+00 -9.00933892e-02 3.80760461e-01 8.34350467e-01
2.33027712e-01 -1.11316092e-01 -1.15202582e+00 8.88570368e-01
2.39385590e-01 -6.63884938e-01 -3.49763632e-01 -7.80822411e-02
1.74039930e-01 4.69327010e-02 -2.31300160e-01 4.22708571e-01
-1.22871682e-01 -9.89084184e-01 9.33257520e-01 3.79616976e-01
4.55506712e-01 -7.19275713e-01 9.53007162e-01 2.05278918e-01
-6.40595257e-01 1.33065328e-01 -1.15020059e-01 -5.89582860e-01
-5.59635200e-02 4.34116632e-01 -4.37795907e-01 8.09174061e-01
9.35640216e-01 5.97143292e-01 -5.40275335e-01 5.54895103e-01
-5.69302328e-02 7.08858252e-01 -3.31233531e-01 -2.48939797e-01
8.97887796e-02 -3.66165787e-01 4.29574758e-01 1.60970390e+00
1.56966493e-01 -4.24599089e-02 1.25383005e-01 6.37901008e-01
3.49300504e-01 6.65409684e-01 -5.40331721e-01 -2.21422389e-01
3.77808422e-01 1.65935051e+00 -9.78387415e-01 -3.31951350e-01
-6.83540165e-01 5.87596297e-01 2.45974198e-01 5.69725297e-02
-7.40311146e-01 -4.82504696e-01 5.82598686e-01 4.38219160e-02
-2.98603121e-02 -1.30075216e-02 -3.83886456e-01 -1.17753708e+00
-3.10155675e-02 -1.03903830e+00 6.90222800e-01 -9.76888239e-01
-1.32314539e+00 9.27837253e-01 4.62492615e-01 -1.05046308e+00
-4.18710113e-01 -6.79861426e-01 -3.19580078e-01 7.43395567e-01
-1.11311555e+00 -9.14059460e-01 -3.18091035e-01 5.36449432e-01
2.67752945e-01 1.51552424e-01 1.30896616e+00 4.24865752e-01
-6.45729601e-01 2.61480600e-01 -2.87518054e-01 7.59411380e-02
1.15037668e+00 -1.28310895e+00 -3.43514144e-01 3.29978526e-01
1.84635460e-01 3.43932837e-01 1.06534660e+00 -3.74977648e-01
-8.21181893e-01 -5.16723871e-01 1.19960558e+00 -5.42389810e-01
8.18307817e-01 -4.05170470e-01 -1.03606331e+00 7.01946020e-01
6.77815259e-01 -3.59091759e-01 1.19943941e+00 5.79617023e-01
-2.10199878e-01 8.32900181e-02 -1.34183586e+00 4.22469616e-01
5.96670091e-01 -2.92611599e-01 -7.08726108e-01 1.43374354e-01
2.29850844e-01 -3.40133429e-01 -1.35419929e+00 2.92569160e-01
5.96108794e-01 -1.22351706e+00 3.85812610e-01 -6.74126148e-01
5.52674353e-01 3.88825163e-02 -2.68963605e-01 -1.34655893e+00
-1.14336804e-01 -2.15480283e-01 4.07500654e-01 1.88477910e+00
4.40498531e-01 -8.20652902e-01 3.56066257e-01 8.38820457e-01
-1.93164870e-01 -8.31170499e-01 -6.78640246e-01 -4.87690002e-01
2.88418323e-01 -8.79021168e-01 5.95419765e-01 1.40740502e+00
5.30525982e-01 3.22157234e-01 -2.07045943e-01 -3.68801683e-01
-4.92271595e-02 2.08016951e-02 7.41031408e-01 -1.51022434e+00
-1.39082568e-02 -6.41072035e-01 -4.95909870e-01 -7.71570206e-02
2.69312978e-01 -7.72942603e-01 -3.46121788e-01 -1.36358368e+00
-8.15386921e-02 -5.97061813e-01 -3.44613284e-01 6.63182676e-01
2.80906200e-01 2.76542783e-01 1.80234835e-01 1.64095372e-01
-3.00663888e-01 9.00591761e-02 7.59800375e-01 3.85675490e-01
-3.17066789e-01 -2.93088913e-01 -8.91885519e-01 9.42888558e-01
1.36404061e+00 -4.96436983e-01 -4.13767487e-01 -4.95203957e-02
4.85409975e-01 -7.99092576e-02 2.05102488e-01 -8.69015694e-01
-2.81818867e-01 -2.13895082e-01 1.73995167e-01 -2.55815685e-01
5.31488240e-01 -9.07271147e-01 3.97706479e-01 1.00370482e-01
-4.06549752e-01 1.25528937e-02 3.50434244e-01 -6.78009093e-02
-4.24844474e-01 -6.99725389e-01 6.85452759e-01 -2.91274339e-01
-6.74075544e-01 -3.80381197e-01 -4.95509267e-01 7.96504170e-02
1.00021541e+00 -4.23752695e-01 -1.59493070e-02 -3.46397698e-01
-9.42414761e-01 -2.81268060e-01 8.65615785e-01 6.44872189e-01
-6.31560162e-02 -1.18297625e+00 -6.23471022e-01 -1.02260396e-01
3.72675180e-01 -4.10838932e-01 1.05485432e-01 1.06377208e+00
-2.39005655e-01 4.62596804e-01 -6.43274844e-01 -2.21301153e-01
-1.45600927e+00 5.67235947e-01 1.80249721e-01 -3.99700731e-01
-1.88774198e-01 2.35467464e-01 -3.02764505e-01 -4.87430841e-01
-5.93851395e-02 -3.50039303e-01 -5.92009485e-01 6.50639117e-01
1.87120855e-01 1.28028616e-01 1.79630771e-01 -1.13979709e+00
-3.95755589e-01 1.88226834e-01 1.06196821e-01 -2.13103056e-01
1.36369503e+00 -1.05410300e-01 -4.07431245e-01 1.11007786e+00
1.08430386e+00 1.72864899e-01 -5.42049766e-01 2.02346638e-01
1.70343310e-01 -3.69716197e-01 -7.57292435e-02 -8.69329154e-01
-7.94330359e-01 4.55090195e-01 3.45170796e-01 7.39230454e-01
1.35610330e+00 2.97582783e-02 2.00333446e-01 8.40152204e-02
2.85607308e-01 -1.31980014e+00 -3.75387520e-01 4.30731237e-01
1.01125085e+00 -1.11759174e+00 2.04959977e-03 -3.59198511e-01
-1.05734837e+00 1.39745152e+00 4.13850427e-01 1.31385744e-01
5.69302857e-01 3.45020890e-01 3.79825264e-01 -3.90125901e-01
-8.97666514e-01 -3.45646501e-01 1.76242683e-02 5.19151270e-01
1.13112628e+00 2.95359232e-02 -1.05804718e+00 1.07069707e+00
-5.46358705e-01 8.60506669e-02 6.40074313e-01 9.37692583e-01
-1.32655546e-01 -1.53234506e+00 -3.69847953e-01 4.14704591e-01
-8.37567627e-01 1.87099546e-01 -9.85862732e-01 1.24650657e+00
4.47871268e-01 1.24730015e+00 6.92198351e-02 -3.69393796e-01
3.43593627e-01 5.10682583e-01 4.10279214e-01 -6.62911236e-01
-7.93844163e-01 -7.32394606e-02 6.95497155e-01 -2.00798795e-01
-1.04592156e+00 -1.09280372e+00 -1.02322423e+00 -2.07769677e-01
-2.61001825e-01 4.68277425e-01 7.33214557e-01 9.43098247e-01
1.01921014e-01 2.68851876e-01 2.97894776e-01 -7.58348644e-01
-1.23206280e-01 -1.13049448e+00 -7.31464088e-01 6.99699521e-01
-3.84622395e-01 -7.56617308e-01 -4.61703092e-01 1.10263333e-01]
|
[12.66522216796875, 6.388296604156494]
|
07137853-1de4-4509-94b6-8874859d0d73
|
rendering-nighttime-image-via-cascaded-color
|
2204.08970
| null |
https://arxiv.org/abs/2204.08970v2
|
https://arxiv.org/pdf/2204.08970v2.pdf
|
Rendering Nighttime Image Via Cascaded Color and Brightness Compensation
|
Image signal processing (ISP) is crucial for camera imaging, and neural networks (NN) solutions are extensively deployed for daytime scenes. The lack of sufficient nighttime image dataset and insights on nighttime illumination characteristics poses a great challenge for high-quality rendering using existing NN ISPs. To tackle it, we first built a high-resolution nighttime RAW-RGB (NR2R) dataset with white balance and tone mapping annotated by expert professionals. Meanwhile, to best capture the characteristics of nighttime illumination light sources, we develop the CBUnet, a two-stage NN ISP to cascade the compensation of color and brightness attributes. Experiments show that our method has better visual quality compared to traditional ISP pipeline, and is ranked at the second place in the NTIRE 2022 Night Photography Rendering Challenge for two tracks by respective People's and Professional Photographer's choices. The code and relevant materials are avaiable on our website: https://njuvision.github.io/CBUnet.
|
['Zhan Ma', 'Si Yi', 'Zhihao LI']
|
2022-04-19
| null | null | null | null |
['tone-mapping']
|
['computer-vision']
|
[ 2.95307845e-01 -4.82849628e-01 2.29035795e-01 -6.24990761e-01
-7.03889549e-01 -6.71240687e-01 4.18131679e-01 -7.49476194e-01
-3.87510091e-01 5.79130828e-01 1.31684929e-01 -5.44206321e-01
4.23590727e-02 -4.85090464e-01 -5.66053689e-01 -6.92614317e-01
3.16478521e-01 -3.23092103e-01 -7.54186977e-03 -3.79844695e-01
-2.64361538e-02 6.24662042e-01 -1.62434602e+00 1.52917922e-01
9.64327872e-01 1.34519756e+00 2.26569206e-01 9.97292280e-01
2.65738696e-01 8.49620402e-01 -4.11606848e-01 -5.42142868e-01
9.25528407e-01 -2.89925426e-01 -2.65721619e-01 -1.37777939e-01
9.97445166e-01 -7.25597382e-01 -5.05746603e-01 1.22923505e+00
7.24063873e-01 2.73159385e-01 1.27550706e-01 -1.20804155e+00
-6.55602217e-01 -1.84152678e-01 -2.02194378e-01 4.78144437e-01
-3.44791412e-01 6.59059405e-01 8.02399397e-01 -7.68333137e-01
2.48399153e-01 5.81046939e-01 8.16038251e-01 2.28349030e-01
-9.01348114e-01 -7.46992111e-01 -1.95225686e-01 4.91054535e-01
-1.03916860e+00 -7.49415815e-01 7.86981225e-01 -1.20035373e-01
6.18603647e-01 5.50631166e-01 7.50410974e-01 1.08184063e+00
-9.56377834e-02 2.74107158e-01 1.70767188e+00 -6.74328059e-02
4.66567203e-02 -1.49757206e-01 -4.12525684e-01 3.73774081e-01
-1.82404593e-01 2.42830381e-01 -5.67754388e-01 3.37788761e-01
7.90107608e-01 -3.73639613e-02 -6.04339063e-01 2.49311417e-01
-7.26337731e-01 3.02301884e-01 8.47146094e-01 -3.04853082e-01
-2.83078700e-01 1.89421237e-01 -6.13694936e-02 1.94260433e-01
7.31401563e-01 3.70824605e-01 -5.97299635e-01 -2.84656644e-01
-9.05552208e-01 -1.29121855e-01 2.89770216e-01 7.41259456e-01
7.34684587e-01 1.55631647e-01 5.77120706e-02 1.18205059e+00
5.05595654e-03 8.33137274e-01 -1.48520574e-01 -1.66703737e+00
2.61819720e-01 2.29177386e-01 2.32204989e-01 -9.13779855e-01
-5.06650627e-01 -3.63660067e-01 -9.18868124e-01 4.58559036e-01
2.66718566e-01 -2.38480166e-01 -8.24845910e-01 1.15358233e+00
5.24950214e-02 4.11028080e-02 -2.93495238e-01 1.50845063e+00
8.89189184e-01 7.41008162e-01 -1.45980105e-01 6.98165240e-05
1.27761912e+00 -9.29573178e-01 -4.82810676e-01 -4.10957694e-01
-2.19903290e-01 -1.00259411e+00 1.27968931e+00 6.07896268e-01
-1.09648526e+00 -6.38906479e-01 -9.48436856e-01 -5.21551669e-01
-3.27313870e-01 5.72261691e-01 7.45420396e-01 7.10868299e-01
-1.27494884e+00 6.02879882e-01 -7.20027208e-01 -2.97687829e-01
4.74025905e-01 -1.82108074e-01 1.15672573e-01 -3.30863535e-01
-1.20160306e+00 7.93273807e-01 -2.66190588e-01 8.86233985e-01
-9.25013185e-01 -9.06719446e-01 -4.60914820e-01 -2.56948918e-01
1.85193613e-01 -4.29495335e-01 1.19287157e+00 -1.08668756e+00
-1.41060817e+00 8.29733491e-01 -6.95023686e-02 -3.32332164e-01
7.47350693e-01 -2.93816417e-01 -8.41089845e-01 3.79389912e-01
-5.27410284e-02 6.40626192e-01 9.71036673e-01 -1.10934818e+00
-5.37708223e-01 -1.16047315e-01 2.29332522e-01 4.71149445e-01
-3.25075328e-01 3.08744133e-01 -7.89712906e-01 -4.07774180e-01
-1.19153425e-01 -7.75657654e-01 -1.23686440e-01 5.66231728e-01
-3.13918263e-01 3.69243324e-01 7.27631688e-01 -1.13848472e+00
5.97100198e-01 -2.24022651e+00 -6.20931983e-01 -1.60927951e-01
1.67563766e-01 3.83364856e-01 -2.89892647e-02 -4.12762202e-02
-9.10551026e-02 -1.82121351e-01 -2.70998091e-01 -3.78324449e-01
6.24778913e-03 1.64553896e-01 -4.43757266e-01 6.29378200e-01
2.27046639e-01 7.89704680e-01 -6.84371829e-01 1.34299582e-04
6.20551527e-01 8.05236220e-01 -1.01747870e-01 2.56791085e-01
-1.22534402e-01 4.95070159e-01 1.64520741e-01 8.82062554e-01
9.12053525e-01 -2.12567985e-01 -1.80394314e-02 -7.40129054e-01
-6.32591546e-01 2.41354868e-01 -8.75285506e-01 1.44852328e+00
-5.80541253e-01 1.40599465e+00 3.72547328e-01 -6.76754341e-02
9.22090411e-01 -1.67669624e-01 1.90754354e-01 -1.22502911e+00
1.88829362e-01 2.74434924e-01 -2.96183348e-01 -5.57995975e-01
7.18978047e-01 1.06625766e-01 4.62457061e-01 1.07368469e-01
-2.86640465e-01 -1.92577943e-01 2.70610768e-02 -6.09959289e-02
9.04078484e-01 3.04429591e-01 -3.52843434e-01 2.24973466e-02
2.14012951e-01 3.90927158e-02 6.81471407e-01 6.59686923e-01
-5.10591149e-01 1.20356488e+00 1.14835754e-01 -5.64506531e-01
-1.36497116e+00 -1.12080264e+00 -1.59323141e-01 1.02429509e+00
-1.74581278e-02 -1.06023205e-02 -5.26450992e-01 1.09687731e-01
-4.83299673e-01 5.59867322e-01 -3.08079541e-01 2.00000674e-01
-3.99532288e-01 -9.24782634e-01 5.19652784e-01 3.45454186e-01
1.19285691e+00 -8.81635666e-01 -7.81844497e-01 -3.19631785e-01
-5.07625818e-01 -1.35860801e+00 -2.56573290e-01 1.08269937e-01
-3.01653922e-01 -1.11263537e+00 -7.20881522e-01 -1.54997036e-01
4.77652818e-01 6.77089036e-01 1.19261837e+00 -1.57068327e-01
-8.33696246e-01 2.48968303e-01 -2.74331391e-01 -4.98508453e-01
-4.23454084e-02 -1.63108453e-01 -6.43401071e-02 8.47491100e-02
3.47607359e-02 -4.91297275e-01 -1.29591572e+00 3.60305667e-01
-7.97263026e-01 2.62902439e-01 5.37587881e-01 2.20702425e-01
3.34967941e-01 3.15420181e-01 -1.10167667e-01 -6.07678652e-01
2.84628749e-01 -6.79120123e-02 -1.17097557e+00 3.06134038e-02
-6.05646849e-01 -7.79160857e-01 6.08735979e-01 1.17754683e-01
-1.54042184e+00 -5.03242016e-03 -2.08938569e-01 -5.11650920e-01
-3.72574031e-01 -6.31115511e-02 -2.69382447e-01 -4.31602746e-01
7.58727372e-01 1.30678087e-01 -2.76485443e-01 -5.05310953e-01
3.00118178e-01 8.11575532e-01 1.07189798e+00 -2.62977391e-01
9.47349548e-01 9.11820948e-01 -1.35782771e-02 -8.63656104e-01
-1.05658555e+00 -5.22160232e-01 -2.96825677e-01 -7.70581245e-01
9.20287371e-01 -1.30423677e+00 -7.94722557e-01 7.72971392e-01
-8.71076286e-01 -8.33718956e-01 -1.37470528e-01 2.42737144e-01
-1.94874451e-01 1.00263588e-01 -5.65138042e-01 -8.49564910e-01
-4.21277195e-01 -9.44167376e-01 8.82672727e-01 6.70902252e-01
3.94806266e-01 -5.69380045e-01 -1.30153596e-01 6.95568204e-01
7.60360599e-01 8.44507962e-02 4.55193579e-01 5.69853187e-01
-7.95235515e-01 8.10651928e-02 -1.01981294e+00 8.99332464e-01
-9.85140949e-02 2.35817537e-01 -1.65540779e+00 6.66268542e-02
5.18705435e-02 -1.61832884e-01 8.35293472e-01 6.26141012e-01
1.37613094e+00 -1.41386002e-01 5.12787580e-01 1.32417345e+00
1.65347672e+00 -9.85720977e-02 1.05495131e+00 7.38972068e-01
9.12734866e-01 6.05245233e-01 4.58551556e-01 5.20508647e-01
4.82025951e-01 5.34053743e-01 7.37193167e-01 -9.21383917e-01
-5.21247149e-01 5.88997789e-02 2.83420354e-01 4.40623730e-01
-3.57871801e-01 3.14978622e-02 -8.45455825e-01 2.86181182e-01
-1.32945299e+00 -6.79147065e-01 -6.12432837e-01 1.97980666e+00
7.48627126e-01 -1.84201747e-01 -7.91612566e-02 -2.50864297e-01
5.61363280e-01 4.02842820e-01 -6.81426167e-01 -2.34509379e-01
-7.96385765e-01 -9.29588545e-03 1.30648541e+00 1.06093556e-01
-1.14716089e+00 6.95082903e-01 5.77329731e+00 5.82041144e-01
-1.30759609e+00 2.20145762e-01 9.25028205e-01 -3.69399548e-01
-1.49826063e-02 7.08891777e-03 -4.39215809e-01 5.47113419e-01
1.02667916e+00 4.25537914e-01 1.10091281e+00 5.41384876e-01
8.20345640e-01 -1.75679460e-01 -2.27737531e-01 1.20443797e+00
1.84921235e-01 -1.27061439e+00 -5.90849221e-01 -6.70557246e-02
8.65014911e-01 8.18903983e-01 3.72427285e-01 -1.55404568e-01
3.08950752e-01 -7.74810195e-01 6.35618508e-01 8.23441744e-01
1.08671379e+00 -5.79660416e-01 3.94220203e-01 -1.60399362e-01
-9.09528017e-01 -2.28906259e-01 -6.61629021e-01 -1.33451834e-01
3.23648959e-01 6.71811700e-01 -3.98202956e-01 5.23965001e-01
1.37895560e+00 8.47370565e-01 -9.76068795e-01 1.29201770e+00
-6.90534055e-01 6.46876752e-01 -2.45569482e-01 5.22354543e-01
5.60393333e-02 -4.87270415e-01 2.97560722e-01 8.91716599e-01
1.89946428e-01 2.78478652e-01 -3.08035195e-01 8.50534558e-01
-2.12755173e-01 -4.55870420e-01 -1.70421377e-01 2.40142927e-01
7.83452094e-02 1.96025264e+00 -5.43362141e-01 -1.52890244e-02
-2.86858380e-01 1.09854150e+00 -3.16805214e-01 7.05539584e-01
-1.00661302e+00 -3.65943313e-01 9.41788316e-01 4.71116863e-02
-1.14682078e-01 -3.41000766e-01 -3.62491190e-01 -9.22918618e-01
2.14685276e-01 -7.49555588e-01 2.51819175e-02 -1.91457665e+00
-1.32953942e+00 7.10705340e-01 -3.81973624e-01 -1.40328240e+00
5.45582116e-01 -9.97655869e-01 -6.02574229e-01 1.01709569e+00
-2.03955531e+00 -1.40976894e+00 -9.68897820e-01 6.99929297e-01
4.76971567e-01 1.60418049e-01 4.71932054e-01 7.27109313e-01
-7.22369313e-01 -1.58997567e-03 3.09435844e-01 8.27582181e-02
8.85959744e-01 -1.11575389e+00 5.90275884e-01 1.38694966e+00
-2.25619212e-01 3.84005129e-01 7.31224954e-01 -3.24623615e-01
-1.44371808e+00 -1.42605615e+00 4.07149136e-01 -3.72297943e-01
7.49534845e-01 -3.33842784e-01 -5.79513431e-01 3.99244338e-01
2.62426198e-01 2.02541500e-01 3.36848855e-01 -1.94748655e-01
-3.19458276e-01 -6.95410669e-01 -7.86495507e-01 7.30964780e-01
9.06576753e-01 -8.58903766e-01 1.78975850e-01 6.54128075e-01
5.26542068e-01 -4.21738923e-01 -6.45307243e-01 3.67371261e-01
6.77593946e-01 -1.36426616e+00 1.11835992e+00 1.57545626e-01
6.17614627e-01 -7.16822088e-01 -2.11394608e-01 -1.10375381e+00
-5.04317731e-02 -6.48146033e-01 4.92708504e-01 1.24522650e+00
2.10551724e-01 -7.01616168e-01 5.50520957e-01 9.70099211e-01
-4.17433202e-01 -2.58198947e-01 -6.68526411e-01 -6.73341632e-01
-5.69287956e-01 -6.80912137e-01 4.73859996e-01 8.17609668e-01
-1.09461701e+00 -2.10705206e-01 -7.48296022e-01 5.03552079e-01
9.03081596e-01 4.46951017e-02 8.01066399e-01 -9.61646438e-01
-2.00058401e-01 -3.37737888e-01 5.55419251e-02 -6.47101045e-01
-2.69921809e-01 -4.55514163e-01 2.44889632e-01 -1.72358310e+00
-8.15825537e-02 -4.19636577e-01 -2.48090014e-01 4.02268887e-01
-1.97141450e-02 9.97084022e-01 2.63164192e-01 4.25320029e-01
-6.04391932e-01 3.89528275e-01 1.10620034e+00 5.50059006e-02
1.56661481e-01 -2.08453149e-01 -6.11029267e-01 6.60901129e-01
9.69643593e-01 -6.14942685e-02 -1.34387493e-01 -9.25645709e-01
5.79178095e-01 -3.48090529e-01 9.98023093e-01 -1.14948356e+00
2.51439631e-01 -1.38656244e-01 8.40602577e-01 -4.81224209e-01
6.82249367e-01 -7.39206612e-01 5.13740480e-01 -3.17755938e-02
-8.75988081e-02 -4.70020398e-02 1.84567556e-01 1.02634043e-01
4.88422289e-02 2.69068360e-01 8.41735542e-01 -2.22235575e-01
-1.01367128e+00 3.16776723e-01 -2.81642191e-02 -7.41637796e-02
2.51851767e-01 -1.86382219e-01 -1.11468863e+00 -4.31733102e-01
-2.19570488e-01 4.91034128e-02 7.52033532e-01 2.01786116e-01
4.74022061e-01 -9.04517293e-01 -5.11131823e-01 1.46122575e-01
9.59400758e-02 -1.80239275e-01 8.43798876e-01 1.04155219e+00
-1.02426004e+00 7.29388697e-03 -4.66407418e-01 -4.18005794e-01
-9.25540030e-01 -1.13481045e-01 5.12241304e-01 4.30468112e-01
-6.94351494e-01 9.71188307e-01 -1.14169806e-01 -4.10874546e-01
-1.05424769e-01 -1.40319273e-01 2.88359709e-02 -1.11337021e-01
5.99609792e-01 7.10564733e-01 3.28392178e-01 -5.59984505e-01
-6.52797520e-02 4.34049189e-01 4.66532975e-01 1.54958265e-02
1.58696640e+00 -6.15626395e-01 -3.41905802e-01 4.12045836e-01
1.16948843e+00 -1.17636837e-01 -1.80988550e+00 1.32525921e-01
-6.28874123e-01 -9.12245691e-01 4.50559765e-01 -1.14262760e+00
-1.33549225e+00 8.41168344e-01 1.04480016e+00 8.69050547e-02
1.80250108e+00 -4.87942666e-01 8.93196225e-01 1.91515297e-01
7.23655969e-02 -1.20865262e+00 -3.76855344e-01 3.41297507e-01
7.76147425e-01 -1.38891613e+00 1.08017378e-01 -3.22601467e-01
-6.43348396e-01 1.09910393e+00 5.67211092e-01 1.15185842e-01
2.05883041e-01 1.30683839e-01 8.09696913e-01 -2.67397035e-02
-4.48635101e-01 -3.99689317e-01 3.07639986e-01 6.07766449e-01
9.14953053e-02 5.82767986e-02 3.52283120e-01 3.62060480e-02
-4.18118447e-01 -6.17599562e-02 5.29316723e-01 3.92426968e-01
-2.09911212e-01 -5.52218378e-01 -3.88055980e-01 4.70466942e-01
-3.51389885e-01 -4.82478589e-01 -1.46903574e-01 3.24829370e-01
2.06041485e-01 1.03340042e+00 1.17782675e-01 -1.89549416e-01
2.72609681e-01 -4.99780655e-01 9.39808637e-02 -4.04368825e-02
-4.80148762e-01 1.84451908e-01 2.66737044e-01 -9.65226591e-01
-3.55462432e-01 -4.58190918e-01 -5.51628411e-01 -3.38538021e-01
4.58295085e-02 -4.70838845e-01 1.32953787e+00 3.39784890e-01
3.39733094e-01 5.43304026e-01 7.22040594e-01 -1.16581380e+00
1.38678849e-01 -5.65404713e-01 -6.96762383e-01 1.65893048e-01
3.98223698e-01 -3.20545405e-01 -6.39403880e-01 2.57161409e-01]
|
[10.741889953613281, -2.498948097229004]
|
1effe3b7-cab2-480d-bd41-6600c2c5dcaa
|
system-log-parsing-a-survey
|
2212.14277
| null |
https://arxiv.org/abs/2212.14277v1
|
https://arxiv.org/pdf/2212.14277v1.pdf
|
System Log Parsing: A Survey
|
Modern information and communication systems have become increasingly challenging to manage. The ubiquitous system logs contain plentiful information and are thus widely exploited as an alternative source for system management. As log files usually encompass large amounts of raw data, manually analyzing them is laborious and error-prone. Consequently, many research endeavors have been devoted to automatic log analysis. However, these works typically expect structured input and struggle with the heterogeneous nature of raw system logs. Log parsing closes this gap by converting the unstructured system logs to structured records. Many parsers were proposed during the last decades to accommodate various log analysis applications. However, due to the ample solution space and lack of systematic evaluation, it is not easy for practitioners to find ready-made solutions that fit their needs. This paper aims to provide a comprehensive survey on log parsing. We begin with an exhaustive taxonomy of existing log parsers. Then we empirically analyze the critical performance and operational features for 17 open-source solutions both quantitatively and qualitatively, and whenever applicable discuss the merits of alternative approaches. We also elaborate on future challenges and discuss the relevant research directions. We envision this survey as a helpful resource for system administrators and domain experts to choose the most desirable open-source solution or implement new ones based on application-specific requirements.
|
['Fabio Pianese', 'Chung Shue Chen', 'Myriana Rifai', 'Gabriele Castellano', 'Han Qiu', 'Tianzhu Zhang']
|
2022-12-29
| null | null | null | null |
['log-parsing']
|
['computer-code']
|
[-6.86332881e-02 -3.08769763e-01 -3.95403832e-01 -3.42743218e-01
-7.30558753e-01 -8.15240324e-01 7.19768181e-02 6.26038432e-01
-4.75116372e-02 4.42017525e-01 -1.48926735e-01 -7.64601946e-01
2.54111644e-03 -6.32943153e-01 -1.31415054e-01 -1.64000750e-01
-3.90539646e-01 4.41304773e-01 3.69561613e-01 -1.33644715e-01
4.80438799e-01 4.52348351e-01 -1.69410574e+00 2.95209438e-02
8.63921881e-01 9.01128054e-01 3.67740005e-01 5.73142827e-01
-3.98775280e-01 8.95820320e-01 -7.61212528e-01 -2.09832996e-01
2.24189103e-01 -5.03214300e-01 -9.57600534e-01 1.60684004e-01
-7.17358813e-02 -3.42564285e-01 -1.75203085e-01 1.00901616e+00
3.30914587e-01 -1.14189401e-01 -1.69254094e-02 -1.38465667e+00
-4.42693681e-01 7.06179023e-01 -5.21828711e-01 5.11655271e-01
6.36943817e-01 -5.30906059e-02 1.06668806e+00 -5.18836439e-01
4.15833563e-01 5.60459077e-01 2.87743330e-01 -3.47129814e-02
-9.68397975e-01 -4.54929352e-01 1.22529253e-01 4.83253837e-01
-1.29898846e+00 -6.50918186e-01 7.66473711e-01 -3.50296557e-01
1.30197668e+00 5.69463253e-01 2.95498163e-01 6.18329167e-01
3.66144449e-01 6.26029789e-01 8.95744383e-01 -6.41606450e-01
1.95134014e-01 4.08706725e-01 9.00171459e-01 4.42140937e-01
7.32710779e-01 -4.45183009e-01 -4.68876630e-01 -4.83349741e-01
4.52469140e-01 6.73002228e-02 -2.49605998e-01 -3.19730222e-01
-7.72326469e-01 4.74125057e-01 -1.71538606e-01 5.18311322e-01
-1.66094720e-01 -3.21099132e-01 5.26714444e-01 5.36803365e-01
6.93494752e-02 3.89348269e-01 -4.28881913e-01 -7.80449688e-01
-9.09248650e-01 -3.47864404e-02 1.33553946e+00 1.20888710e+00
8.31309736e-01 -1.84434906e-01 3.66137236e-01 6.41219437e-01
3.06921750e-01 6.88290149e-02 4.95383739e-01 -5.33962905e-01
5.22695243e-01 7.52860963e-01 -9.30982456e-02 -1.07843173e+00
-2.95482814e-01 -1.24914169e-01 -6.46315694e-01 -1.29804537e-01
2.84701288e-01 2.39202216e-01 -1.84263721e-01 1.12124062e+00
3.76847596e-03 -3.34395468e-01 -3.28935444e-01 4.59238112e-01
3.25568140e-01 5.77672243e-01 -8.55882391e-02 -7.77074516e-01
1.57868207e+00 -7.42773056e-01 -8.26475918e-01 -3.95460576e-01
5.89291990e-01 -1.11359489e+00 1.39852571e+00 2.71527469e-01
-9.71256912e-01 -2.51883268e-01 -1.03718984e+00 1.95507050e-01
-1.46438420e-01 -3.63518409e-02 7.53053188e-01 8.77971768e-01
-8.17212701e-01 4.13408220e-01 -1.50841200e+00 -6.69184148e-01
-7.14152530e-02 3.75987232e-01 -1.73305720e-02 2.69714057e-01
-6.35347962e-01 6.34382486e-01 3.98024231e-01 -1.87511388e-02
-1.74667314e-01 -2.06094906e-01 -6.46649241e-01 4.43020999e-01
8.15476537e-01 -2.10819542e-01 1.72223163e+00 -2.33913019e-01
-1.46248162e+00 5.38072288e-01 -2.64340907e-01 -1.53231487e-01
9.18934420e-02 -2.23270029e-01 -4.79109854e-01 -7.55843818e-02
2.05150135e-02 -5.68703473e-01 3.57834041e-01 -8.90376270e-01
-6.21114612e-01 -2.79057711e-01 1.17504865e-01 -9.69014838e-02
-9.42121625e-01 6.05985641e-01 -7.09102690e-01 -3.11322302e-01
-1.07358498e-02 -7.20600665e-01 -1.20839648e-01 -6.85109913e-01
-5.58532298e-01 -1.81901187e-01 6.65789843e-01 -3.12833518e-01
2.35073018e+00 -2.06608868e+00 -3.45302999e-01 4.72819731e-02
4.14531916e-01 2.30601519e-01 2.30131671e-01 1.11066985e+00
4.31171209e-02 3.81572187e-01 -9.49590951e-02 -3.10567260e-01
-1.68113783e-02 3.77216488e-02 -5.32114327e-01 1.90240756e-01
-2.13906392e-01 5.84727705e-01 -6.45635664e-01 -8.32333088e-01
2.50954360e-01 -5.43995425e-02 -3.51271391e-01 5.97109914e-01
-2.63531469e-02 7.25473687e-02 -6.76580250e-01 9.35671926e-01
5.10334492e-01 -8.57024014e-01 4.60409671e-01 4.10455726e-02
-3.72315168e-01 5.23542583e-01 -1.13453567e+00 1.12841034e+00
-4.63153094e-01 5.17136812e-01 9.62586105e-02 -7.79659867e-01
7.44916439e-01 2.17401728e-01 7.67557025e-01 -5.53972125e-01
2.64501572e-01 3.18017513e-01 1.13316607e-02 -6.90503657e-01
6.83310032e-01 9.61335003e-02 -4.30658758e-01 1.03216386e+00
-4.21081781e-01 2.51598477e-01 6.56047225e-01 2.16984510e-01
1.64145184e+00 -2.77337193e-01 9.21968699e-01 8.38349238e-02
6.44613147e-01 7.88217261e-02 5.35330832e-01 4.87595201e-01
-5.19438326e-01 3.04992259e-01 7.48173535e-01 -3.19588602e-01
-6.51654363e-01 -8.32963645e-01 -1.54933378e-01 1.17716610e+00
7.51743913e-02 -1.08574307e+00 -6.55949712e-01 -6.58135891e-01
-3.16197336e-01 3.90236795e-01 1.85162406e-02 1.56451359e-01
-8.24147046e-01 -9.07329142e-01 3.21238905e-01 6.06691360e-01
3.62541288e-01 -1.36247051e+00 -9.50227857e-01 5.38381219e-01
-3.78331214e-01 -1.27409613e+00 -3.47139806e-01 3.15188259e-01
-9.93678391e-01 -1.29842925e+00 6.22435510e-02 -6.98793650e-01
4.27041292e-01 7.38767505e-01 1.18016207e+00 4.61730421e-01
-3.86240423e-01 2.02229097e-01 -7.66827703e-01 -1.26624063e-01
-5.91320515e-01 4.08288240e-01 1.83522552e-01 -2.26068109e-01
7.67334044e-01 -8.23906243e-01 -3.27484608e-01 4.20645535e-01
-9.20149088e-01 -1.88480884e-01 6.46363378e-01 4.22737479e-01
3.90694857e-01 3.33371907e-01 5.13529301e-01 -1.27568841e+00
1.06630099e+00 -5.44593513e-01 -7.77663171e-01 3.64308804e-01
-1.20986915e+00 6.66662678e-02 1.14520764e+00 -4.05540802e-02
-9.64065611e-01 -1.87275320e-01 -2.56278992e-01 1.12962015e-01
-2.99658746e-01 8.23736966e-01 -2.17274472e-01 1.52491614e-01
5.86840034e-01 3.90319973e-02 -1.10807881e-01 -6.72051191e-01
-1.00303590e-01 1.14965701e+00 4.08260405e-01 -6.40323162e-01
7.65004516e-01 2.26170734e-01 -5.88010073e-01 -8.14215958e-01
-6.37383580e-01 -7.39712358e-01 -5.33960760e-01 1.51967421e-01
2.70963132e-01 -1.40257388e-01 -8.55035007e-01 2.87075311e-01
-1.11558068e+00 -1.62149027e-01 -1.10562675e-01 4.10497747e-02
-4.87024873e-01 8.94034386e-01 -9.05514300e-01 -6.76095366e-01
-4.00648296e-01 -1.23407292e+00 6.81444645e-01 1.75093547e-01
-5.18905163e-01 -7.70785749e-01 3.08126122e-01 3.01647514e-01
6.54303849e-01 -6.67379498e-02 1.06528747e+00 -7.22327173e-01
-9.18721795e-01 -6.74759507e-01 -3.68634641e-01 8.20253342e-02
5.10891199e-01 1.04219079e-01 -9.10078168e-01 -4.23294902e-01
1.87164098e-01 -6.31775111e-02 1.64328992e-01 -9.81792808e-02
1.25851035e+00 -2.96069920e-01 -4.50638801e-01 4.35783952e-01
1.36634851e+00 4.81604367e-01 4.72418636e-01 5.60215414e-01
4.97078776e-01 4.63636994e-01 7.24922776e-01 7.65606165e-01
3.06793094e-01 5.18353939e-01 2.55747169e-01 4.99984264e-01
3.86800945e-01 -2.09074274e-01 5.14653981e-01 1.54108715e+00
2.57224515e-02 -5.94722211e-01 -9.89225686e-01 3.39823306e-01
-1.75719380e+00 -7.27613628e-01 -2.14092076e-01 2.30943441e+00
6.70488238e-01 5.33098280e-01 1.17261194e-01 4.26620901e-01
5.04883230e-01 6.94764182e-02 -3.35841715e-01 -2.14457497e-01
3.76189291e-01 1.89010993e-01 2.56175399e-01 8.58312026e-02
-7.82079756e-01 6.91098094e-01 6.48704815e+00 5.12105048e-01
-8.93448114e-01 2.47059693e-03 1.00769863e-01 1.82275578e-01
-8.37344490e-03 5.07933021e-01 -9.52040017e-01 5.33139944e-01
1.53699672e+00 -6.69519067e-01 2.62325555e-01 8.97801518e-01
2.45411888e-01 -3.62800866e-01 -1.33652818e+00 1.11429358e+00
-2.79080987e-01 -1.15381086e+00 -3.22113454e-01 4.84193236e-01
2.92506348e-02 -1.27662435e-01 -3.02203745e-01 4.02665555e-01
-1.95036810e-02 -7.53210366e-01 3.86826396e-01 3.92011441e-02
7.25411713e-01 -3.05239707e-01 7.37005889e-01 5.57226598e-01
-1.61328733e+00 -9.76738036e-02 -2.92660028e-01 -6.09092414e-01
6.72073543e-01 4.69528466e-01 -5.26726663e-01 4.26849425e-01
7.71535099e-01 5.39928794e-01 -7.18938112e-01 1.12151968e+00
9.81279016e-02 9.00410235e-01 -4.15606707e-01 -7.78674036e-02
-2.28046313e-01 -1.67306527e-01 2.91497231e-01 1.05602050e+00
2.47249886e-01 -2.91572977e-02 4.54143852e-01 5.02294600e-01
1.08967349e-01 4.02351052e-01 -5.15012026e-01 -5.43961227e-01
7.82285929e-01 1.31268680e+00 -1.22100997e+00 -1.86412930e-01
-8.43939543e-01 6.57149434e-01 2.45952383e-01 5.60698174e-02
-6.34309232e-01 -6.98463142e-01 6.25105977e-01 4.42833751e-01
-2.08179817e-01 -5.85415423e-01 -4.29865927e-01 -1.27337408e+00
4.70059425e-01 -9.39929068e-01 5.43382287e-01 -1.38374746e-01
-1.22732949e+00 1.04018450e+00 1.67747438e-01 -1.41824794e+00
-6.49195075e-01 -5.21868408e-01 -5.24458528e-01 5.03324389e-01
-1.21002173e+00 -4.90144610e-01 -5.78819811e-01 1.49879843e-01
6.17341578e-01 -4.95364740e-02 9.48463798e-01 5.97278893e-01
-8.86101127e-01 6.65009320e-01 1.54907450e-01 1.70678440e-02
8.59625995e-01 -1.08635879e+00 5.67551017e-01 8.84558976e-01
8.28861371e-02 1.14585185e+00 7.44064331e-01 -5.23638487e-01
-1.76509798e+00 -5.46093941e-01 9.70543504e-01 -7.19841242e-01
9.84716892e-01 -3.87356162e-01 -1.36966574e+00 8.65198791e-01
2.70850211e-01 -1.35232091e-01 8.89819384e-01 2.71741897e-01
-1.70124292e-01 -1.28088832e-01 -5.44012964e-01 3.21028113e-01
8.63286793e-01 -5.28254211e-01 -2.99748093e-01 1.23251438e-01
5.50096869e-01 -1.75858319e-01 -8.56840551e-01 4.72873747e-02
3.37048680e-01 -1.35853624e+00 4.32633936e-01 -3.35468739e-01
-3.20435613e-02 -1.41533315e-01 -1.04658343e-02 -6.00419164e-01
-9.92174000e-02 -1.05932200e+00 -3.96594793e-01 1.39999151e+00
7.15491548e-02 -7.74411678e-01 8.96392465e-01 6.55330896e-01
-2.18315154e-01 -7.04784393e-01 -4.01059002e-01 -1.01943302e+00
-4.04162914e-01 -7.64210641e-01 5.32841980e-01 7.71113336e-01
7.30766714e-01 5.56178749e-01 -5.62550016e-02 -7.59329880e-04
4.07437503e-01 5.97728550e-01 9.92199481e-01 -1.36076808e+00
-4.76389408e-01 -5.04144549e-01 -2.74481624e-01 -1.43052220e+00
-6.23148680e-02 -7.01305211e-01 -5.62098362e-02 -1.40031803e+00
3.53177458e-01 -6.52902305e-01 -1.55253366e-01 5.53527713e-01
8.08234811e-02 -3.27976197e-02 1.60564035e-01 7.38496900e-01
-6.83576703e-01 1.36556998e-01 5.83220959e-01 2.68587321e-01
-7.17491210e-01 5.37250519e-01 -1.05344105e+00 6.45424664e-01
7.70420849e-01 -4.99127835e-01 -4.92861629e-01 -2.56485373e-01
5.50270855e-01 4.13999408e-01 -8.53620693e-02 -8.99849474e-01
4.19858247e-01 -3.87154460e-01 -4.19985145e-01 -5.91521859e-01
-1.89041078e-01 -9.51441526e-01 1.19602114e-01 1.35700390e-01
2.67342985e-01 5.12752652e-01 1.11000478e-01 5.42042315e-01
-5.60795426e-01 -4.05051976e-01 3.83248806e-01 6.80538127e-04
-8.43800724e-01 3.71266276e-01 -6.10230207e-01 3.00273094e-02
1.14767754e+00 -4.17926550e-01 -2.35234052e-01 -3.81340384e-01
-3.89226198e-01 2.45959267e-01 8.26919138e-01 4.28454965e-01
3.86869848e-01 -7.70845890e-01 -1.15731351e-01 4.19372350e-01
3.84165615e-01 -2.00140551e-01 4.20029042e-03 7.24604428e-01
-7.20560014e-01 6.71682000e-01 -1.15525119e-01 -3.84331584e-01
-1.32724583e+00 7.33911693e-01 -1.56072751e-01 -6.92754984e-01
-6.33869112e-01 3.25141102e-01 -6.04318976e-02 1.71028227e-01
2.03196302e-01 -2.83496499e-01 -1.46064092e-03 -1.41147664e-02
6.31139517e-01 4.88409668e-01 4.27925110e-01 -2.17802271e-01
-4.52347338e-01 2.52484947e-01 -3.53408635e-01 2.50519633e-01
1.30506933e+00 -5.15296757e-01 -4.03958470e-01 5.96237957e-01
1.11887717e+00 1.53789580e-01 -7.71152496e-01 -2.42730081e-01
5.38244843e-01 -5.17709851e-01 -3.45026702e-01 -1.62348613e-01
-7.46700764e-01 8.52905273e-01 1.91856652e-01 1.03028250e+00
1.53780937e+00 9.47455615e-02 9.70445037e-01 4.16179627e-01
9.96100724e-01 -9.26940501e-01 2.37116776e-02 5.01542270e-01
2.16321185e-01 -1.12278259e+00 9.08071697e-02 -7.46190846e-01
-1.57477885e-01 1.22259879e+00 6.63044989e-01 1.85866460e-01
7.97378480e-01 6.82272732e-01 2.90693790e-02 -2.18416959e-01
-7.04992533e-01 1.98995788e-03 -1.91399589e-01 3.87687325e-01
9.61920083e-01 -1.08337656e-01 -6.26448095e-01 8.71073663e-01
-3.51600021e-01 -1.13943890e-01 9.66881037e-01 1.68042445e+00
-6.58925712e-01 -1.78391027e+00 -4.02176559e-01 7.03658283e-01
-4.13444161e-01 1.12589397e-01 -2.73783296e-01 7.81174898e-01
-5.34812391e-01 1.12017417e+00 -5.98077327e-02 -4.52410996e-01
4.51131135e-01 6.92194998e-02 1.24305077e-01 -1.00570881e+00
-4.61696416e-01 -5.27731003e-03 -8.33366141e-02 -5.98686993e-01
5.07552773e-02 -7.26779580e-01 -1.33699405e+00 -6.92162514e-01
-4.93621975e-01 3.26754719e-01 4.17790383e-01 6.90125108e-01
5.07831216e-01 5.28647900e-01 5.82938373e-01 -3.37351114e-01
-8.49657059e-01 -8.94749105e-01 -6.78962171e-01 1.27798155e-01
1.20416582e-01 -5.81715226e-01 -1.95031211e-01 3.84724885e-01]
|
[7.997892379760742, 6.886466026306152]
|
af6b4218-339a-46e8-bdd7-6ec548cd0ca2
|
urvos-unified-referring-video-object
| null | null |
https://www.ecva.net/papers/eccv_2020/papers_ECCV/html/2327_ECCV_2020_paper.php
|
https://www.ecva.net/papers/eccv_2020/papers_ECCV/papers/123600205.pdf
|
URVOS: Unified Referring Video Object Segmentation Network with a Large-Scale Benchmark
|
We propose a unified referring video object segmentation network (URVOS). URVOS takes a video and a referring expression as inputs, and estimates the {object masks} referred by the given language expression in the whole video frames. Our algorithm addresses the challenging problem by performing language-based object segmentation and mask propagation jointly using a single deep neural network with a proper combination of two attention models. In addition, we construct the first large-scale referring video object segmentation dataset called Refer-Youtube-VOS. We evaluate our model on two benchmark datasets including ours and demonstrate the effectiveness of the proposed approach. The dataset is released at \url{https://github.com/skynbe/Refer-Youtube-VOS}.
|
['Joon-Young Lee', 'Seonguk Seo', 'Bohyung Han']
| null | null | null | null |
eccv-2020-8
|
['one-shot-visual-object-segmentation', 'referring-expression-segmentation', 'referring-video-object-segmentation']
|
['computer-vision', 'computer-vision', 'computer-vision']
|
[-1.35456428e-01 -4.27313149e-02 -5.21968484e-01 -4.00174737e-01
-1.06134403e+00 -5.47928572e-01 2.01467425e-01 -6.06698632e-01
-4.09354925e-01 4.32589918e-01 1.45121804e-02 -3.78600806e-02
3.87293071e-01 -3.84581596e-01 -1.03967655e+00 -1.74367771e-01
4.15499151e-01 2.35623419e-01 4.42363262e-01 2.40430281e-01
2.42561027e-01 3.42763782e-01 -1.06629694e+00 1.36941046e-01
5.43330967e-01 1.44802880e+00 3.48812819e-01 8.65228653e-01
-2.32776716e-01 1.23870671e+00 -5.94917715e-01 -5.16259372e-01
2.60259658e-01 -5.26157975e-01 -1.16913033e+00 4.34655160e-01
5.46044707e-01 -5.88049591e-01 -8.46909285e-01 1.24657357e+00
2.01341137e-01 3.70504260e-01 2.91962147e-01 -1.57365549e+00
-1.14710557e+00 6.90806150e-01 -5.77728152e-01 6.44043267e-01
3.52603495e-01 3.90726507e-01 8.99937272e-01 -8.22747886e-01
8.71489108e-01 1.26529682e+00 1.55660093e-01 7.57930636e-01
-6.15948260e-01 -6.56979442e-01 5.81701159e-01 3.01249981e-01
-1.72980094e+00 -4.29591507e-01 8.10491085e-01 -4.84271497e-01
6.40715480e-01 1.77055806e-01 5.74337363e-01 1.02538121e+00
-3.40674967e-01 1.52959955e+00 4.88998353e-01 1.62663385e-01
1.15872756e-01 -1.60918474e-01 2.66130179e-01 6.94926620e-01
-2.20481262e-01 -6.56710267e-01 -3.46506208e-01 1.07381009e-01
8.88399363e-01 1.78615794e-01 -6.64689720e-01 5.18650562e-02
-1.01986146e+00 6.45137370e-01 4.63644296e-01 2.39920512e-01
-3.88670355e-01 8.93224478e-01 3.98048043e-01 -2.79757101e-02
6.24568701e-01 7.29446113e-03 -3.71926695e-01 -3.36230248e-01
-1.03328562e+00 1.52348727e-01 5.09532571e-01 1.68606937e+00
4.02701437e-01 5.89341447e-02 -2.89469689e-01 6.93817019e-01
4.17500257e-01 3.18199992e-01 4.12734933e-02 -1.66079867e+00
5.03529787e-01 3.08818609e-01 2.86560804e-01 -7.55775869e-01
5.89974299e-02 3.59243378e-02 -3.70155394e-01 -4.05785888e-01
3.52548629e-01 -2.52348244e-01 -1.01811647e+00 1.80031025e+00
3.14849257e-01 4.83995348e-01 -1.52155682e-01 1.30046511e+00
1.48384356e+00 8.24404716e-01 3.05380702e-01 -7.72443339e-02
1.29602337e+00 -1.39422631e+00 -9.66219187e-01 -6.97180107e-02
4.13631320e-01 -3.99351746e-01 1.12897086e+00 1.13467090e-01
-1.40398777e+00 -5.22685707e-01 -7.29886532e-01 -5.45887291e-01
-2.73644537e-01 1.38903722e-01 3.80479604e-01 2.02548236e-01
-1.09760034e+00 2.96780527e-01 -7.81407237e-01 -3.74321580e-01
1.07781959e+00 4.08302426e-01 -5.08759283e-02 -2.98717655e-02
-9.78484035e-01 1.40086949e-01 3.28320712e-01 1.99294731e-01
-9.34323430e-01 -3.72559935e-01 -8.74860585e-01 3.47239785e-02
8.81685197e-01 -6.63710356e-01 1.47848177e+00 -1.32139695e+00
-1.30209684e+00 1.16692328e+00 -5.41962445e-01 -3.47861558e-01
7.18534112e-01 -5.52884281e-01 -1.03630386e-01 5.31607270e-01
3.42301160e-01 1.23958731e+00 6.58934832e-01 -1.38236749e+00
-5.29232562e-01 -1.22488640e-01 5.24720132e-01 -1.88044664e-02
7.48303086e-02 5.40722072e-01 -1.72202110e+00 -7.03528941e-01
7.05989227e-02 -5.97363770e-01 6.00829301e-03 1.12839334e-01
-8.03339422e-01 -6.50329888e-01 8.80605459e-01 -7.12988973e-01
1.31236148e+00 -2.27673960e+00 2.89451212e-01 -5.10520697e-01
4.10097033e-01 1.74981639e-01 -2.60966629e-01 -6.35755137e-02
4.99312691e-02 6.46891713e-01 -2.94727296e-01 -3.74962777e-01
7.39322379e-02 1.35918066e-01 -2.30706692e-01 3.55272561e-01
1.40086412e-01 1.36369646e+00 -7.25835860e-01 -7.90524662e-01
-9.94560644e-02 3.71402085e-01 -4.96071696e-01 5.03681362e-01
-6.88197970e-01 3.95090699e-01 -8.95322561e-01 1.02879441e+00
4.87875402e-01 -4.67510283e-01 -2.75943696e-01 -1.52890429e-01
1.99342266e-01 -2.22338811e-01 -8.71990561e-01 1.74931419e+00
-1.93442125e-02 8.45593035e-01 9.67443511e-02 -8.05116057e-01
6.60413384e-01 3.11775416e-01 6.44707680e-01 -5.14951468e-01
7.58635223e-01 8.70943442e-02 -5.71052432e-01 -9.95735407e-01
4.37669486e-01 4.28802043e-01 -1.49928972e-01 3.25773209e-01
3.51422697e-01 -6.34781122e-02 3.78200680e-01 6.27048314e-01
7.63895273e-01 6.28826261e-01 1.22151725e-01 -7.85329752e-03
5.54444909e-01 -3.26759703e-02 6.95901155e-01 7.85775244e-01
-7.05323577e-01 8.37401628e-01 9.17530358e-01 -3.27853262e-01
-7.48507917e-01 -8.08693945e-01 1.37630120e-01 1.12259758e+00
5.72519720e-01 -3.12532872e-01 -1.10563922e+00 -9.34838772e-01
-1.59780741e-01 6.84825957e-01 -6.10626101e-01 1.75921604e-01
-5.25978982e-01 -1.21342689e-01 4.44684923e-01 7.59676218e-01
7.71043718e-01 -1.44806683e+00 -4.18040603e-01 -8.95673037e-02
-3.87307614e-01 -1.78098750e+00 -1.06504869e+00 -2.30641693e-01
-4.88914549e-01 -1.23503458e+00 -9.34656262e-01 -8.17290306e-01
5.29504478e-01 1.36161283e-01 1.17442834e+00 3.23762476e-01
9.92607847e-02 7.92983949e-01 -3.53197455e-01 -1.19102441e-01
-5.03761927e-04 6.46414459e-02 -3.28093290e-01 1.46487236e-01
6.24573052e-01 -2.27952227e-01 -5.95704973e-01 2.24153578e-01
-9.73831177e-01 -2.58340891e-02 -5.37530109e-02 3.18303615e-01
9.93729532e-01 -5.35880327e-01 5.22244632e-01 -7.44747400e-01
3.71274352e-01 -8.58662486e-01 -7.26132274e-01 3.45064074e-01
3.25916499e-01 -5.39794028e-01 2.96237737e-01 -3.25153708e-01
-6.41760528e-01 1.40966237e-01 -2.52897024e-01 -1.11044550e+00
-5.08146942e-01 1.14644982e-01 -4.96757269e-01 1.03902213e-01
-1.26211107e-01 1.44577771e-01 -4.83714342e-01 -4.62769985e-01
4.95277941e-01 8.20713043e-01 8.59510481e-01 -5.77151418e-01
4.06834394e-01 4.42948878e-01 -5.96933246e-01 -6.40548706e-01
-1.12270677e+00 -5.39340138e-01 -8.49638104e-01 -4.02609289e-01
1.48222482e+00 -1.03043628e+00 -1.08004975e+00 3.59626919e-01
-1.58957112e+00 -7.19031036e-01 -3.26304197e-01 1.90494314e-01
-7.21090019e-01 3.75558645e-01 -8.32247853e-01 -8.37776065e-01
-2.60435075e-01 -1.37717497e+00 1.27252591e+00 4.50805962e-01
-4.79730219e-02 -8.69996130e-01 -6.16167068e-01 6.25609636e-01
6.19764579e-03 2.97513068e-01 3.19778353e-01 -8.46420586e-01
-1.15775073e+00 -7.51110166e-02 -7.61493027e-01 2.58602649e-01
-3.44757885e-02 1.61288336e-01 -8.39769304e-01 2.48086184e-01
-1.50426537e-01 -4.73805100e-01 1.06285107e+00 5.55417657e-01
1.73048723e+00 -1.05272010e-01 -1.84898898e-01 9.57546711e-01
1.36972654e+00 5.97215712e-01 3.46349031e-01 2.20497206e-01
1.06389940e+00 1.99887007e-01 6.77613258e-01 3.33492637e-01
6.28274798e-01 6.69000506e-01 6.21098220e-01 1.29735246e-01
-2.14665338e-01 -1.95265040e-02 1.21865258e-01 7.53741384e-01
4.02570702e-03 -9.06643391e-01 -8.06037307e-01 8.35776031e-01
-2.00907278e+00 -7.75695384e-01 -8.99673700e-02 1.52304173e+00
5.00673831e-01 -2.33154967e-02 1.26642123e-01 -5.73823571e-01
8.82891476e-01 3.74267787e-01 -9.45334256e-01 -2.18224183e-01
-7.16275275e-02 -1.37923688e-01 3.18524718e-01 4.24147874e-01
-1.41699982e+00 1.58506644e+00 5.73365736e+00 7.62003183e-01
-1.01268280e+00 5.07479250e-01 9.95948672e-01 -3.40652257e-01
-1.51730165e-01 -2.55928099e-01 -7.59976804e-01 5.09493947e-01
7.00568855e-01 -1.46516725e-01 4.71488595e-01 9.25673366e-01
3.75171065e-01 -1.85778856e-01 -9.90651965e-01 1.15114045e+00
3.54637533e-01 -1.38286209e+00 4.60195318e-02 -3.81551534e-01
6.45386159e-01 3.23753566e-01 -6.67780042e-02 1.45390809e-01
8.56363997e-02 -7.29832411e-01 1.23355734e+00 5.87887585e-01
9.50225651e-01 -3.80879909e-01 7.49780595e-01 -1.85658962e-01
-1.32244730e+00 1.53060883e-01 -7.71798147e-03 2.89734811e-01
6.03262901e-01 -5.47046475e-02 -4.60692495e-02 3.05750340e-01
1.00341928e+00 9.92564917e-01 -4.14257705e-01 9.81004775e-01
-2.52044886e-01 7.03287244e-01 -3.41095388e-01 8.77544656e-02
5.38173378e-01 -2.69856960e-01 4.70092565e-01 1.12707269e+00
9.46622565e-02 5.79756677e-01 3.37209165e-01 1.42488074e+00
-8.27888608e-01 -9.68923420e-03 -2.84452438e-01 -3.21191937e-01
5.68459213e-01 1.07279444e+00 -1.08348036e+00 -7.19789505e-01
-6.06910229e-01 1.17726326e+00 2.48893172e-01 8.59348118e-01
-1.31106603e+00 -2.49716237e-01 6.44995093e-01 -6.03299923e-02
6.38117909e-01 -2.13108093e-01 3.25890258e-02 -1.21083438e+00
6.40073940e-02 -5.25113165e-01 3.38978082e-01 -1.37059307e+00
-1.06041443e+00 6.95106626e-01 -1.75161136e-03 -8.39235544e-01
2.12731287e-01 -4.79194790e-01 -5.83640933e-01 6.29243195e-01
-1.42429364e+00 -1.03960967e+00 -5.33997297e-01 7.63421595e-01
1.03660583e+00 2.22641360e-02 9.83222798e-02 4.93259937e-01
-1.01849401e+00 4.00241286e-01 -4.45142686e-01 6.13756120e-01
3.57008785e-01 -1.00820100e+00 6.23234451e-01 8.54692221e-01
1.78675026e-01 4.43482816e-01 3.50490093e-01 -7.43364155e-01
-1.29465032e+00 -1.34984136e+00 5.66212535e-01 -5.39901257e-01
7.86302209e-01 -3.19522500e-01 -8.68977785e-01 1.26150787e+00
4.92276937e-01 4.96857882e-01 3.11911911e-01 -6.41647518e-01
-1.47429947e-02 3.79644811e-01 -1.00927162e+00 6.48871541e-01
1.39270735e+00 -6.50815308e-01 -4.18494642e-01 6.46922469e-01
1.33453596e+00 -7.33271956e-01 -7.28672028e-01 3.96655276e-02
1.33331686e-01 -6.92691028e-01 6.79201066e-01 -9.36349988e-01
5.93697608e-01 -2.52252489e-01 -4.59606498e-01 -4.18119431e-01
1.29370734e-01 -6.87724888e-01 -2.15422601e-01 1.50320899e+00
1.27480477e-01 -3.39311808e-01 7.36500025e-01 8.16413343e-01
-2.12834612e-01 -1.06254876e+00 -7.25963056e-01 -6.07735217e-01
-2.14397460e-01 -8.62691164e-01 4.66935784e-01 7.95153439e-01
-5.31729937e-01 4.47439328e-02 -4.10262764e-01 1.42585978e-01
3.97397012e-01 5.98058701e-02 7.33498216e-01 -5.84394455e-01
-8.96820351e-02 -4.74967808e-01 -3.49248886e-01 -1.63213718e+00
5.59394956e-01 -8.30276310e-01 9.56398845e-02 -1.66077256e+00
3.04839015e-01 1.60778109e-02 -1.82768792e-01 5.51275015e-01
-3.44745636e-01 5.41282296e-01 6.37026489e-01 2.36198246e-01
-1.48389137e+00 5.34076333e-01 1.27897286e+00 -2.10631162e-01
-1.54741317e-01 -1.22331709e-01 -6.21648610e-01 9.37514365e-01
7.30211198e-01 -5.46199262e-01 -1.22000746e-01 -7.93522120e-01
-2.04447985e-01 2.49558136e-01 5.25122285e-01 -6.55229211e-01
2.08403945e-01 -1.21178202e-01 -1.81333497e-02 -7.33214915e-01
3.90361011e-01 -6.40853763e-01 -8.97556543e-02 2.94882548e-03
-4.47556287e-01 5.40651381e-02 1.72312245e-01 5.06317019e-01
-3.71437728e-01 -2.25556359e-01 5.80794394e-01 -1.89725548e-01
-1.13701606e+00 7.62078464e-01 -2.26914287e-01 5.79318345e-01
1.17747021e+00 -2.13339522e-01 -2.06369489e-01 -5.82417130e-01
-1.01175594e+00 6.36566520e-01 3.55988979e-01 6.42272353e-01
6.67561114e-01 -1.27558255e+00 -3.53789300e-01 -3.45730782e-01
-3.18984762e-02 1.56962484e-01 3.46681923e-01 9.36729014e-01
-6.75319970e-01 3.76183063e-01 9.30421725e-02 -6.44629121e-01
-1.12671256e+00 8.13595653e-01 4.78013366e-01 4.57513571e-01
-6.08286023e-01 1.39689565e+00 3.95648986e-01 -7.55145401e-02
4.44272727e-01 -3.79198819e-01 -2.21736789e-01 -1.49162337e-01
3.07395011e-01 2.95844879e-02 -7.09045410e-01 -1.05705583e+00
-3.15113515e-01 7.33864963e-01 1.72264844e-01 1.43190935e-01
1.09682930e+00 -5.09073973e-01 -2.76766345e-02 5.82176268e-01
1.44384515e+00 -3.13016891e-01 -1.47807574e+00 -2.97143102e-01
2.29934789e-02 -4.85925406e-01 3.66024976e-03 -2.21228868e-01
-1.73351812e+00 7.35270858e-01 1.94778904e-01 9.28851813e-02
1.12088418e+00 5.21852434e-01 1.17728567e+00 2.11215898e-01
1.34539872e-01 -8.77371907e-01 7.16226026e-02 2.91143566e-01
9.57729697e-01 -1.34077322e+00 -1.10186301e-01 -6.10505521e-01
-8.90657306e-01 8.76846194e-01 8.15584540e-01 -2.74725288e-01
6.53596461e-01 1.66689157e-01 2.36569121e-01 -3.76458973e-01
-5.87994516e-01 -5.38083792e-01 2.05407962e-01 1.45224959e-01
2.84496844e-01 -3.14630538e-01 -2.24561065e-01 9.14379895e-01
3.04754257e-01 4.30134833e-01 5.33562720e-01 7.09705114e-01
-2.04972073e-01 -7.29036868e-01 -2.12933302e-01 2.59016156e-01
-8.38040233e-01 -6.16018474e-02 -5.81738293e-01 9.13943946e-01
1.19777963e-01 7.68648744e-01 2.51752883e-01 -1.36058703e-01
2.89688438e-01 1.66327074e-01 1.37316555e-01 -4.35409039e-01
-5.00779986e-01 3.36127162e-01 -9.19709578e-02 -9.92374122e-01
-7.81081617e-01 -5.24539232e-01 -1.58857059e+00 -7.98772648e-02
2.98096910e-02 1.05498187e-01 2.62506783e-01 1.01765776e+00
4.28906441e-01 7.87868679e-01 2.88610131e-01 -1.00054002e+00
1.80180997e-01 -5.96985936e-01 -5.88097632e-01 5.33139944e-01
2.03029096e-01 -4.62531447e-01 -2.04562247e-01 4.13079828e-01]
|
[9.475698471069336, 0.32337403297424316]
|
e00f500c-df34-40cb-858c-2c84042c8344
|
sface-sigmoid-constrained-hypersphere-loss-1
|
2205.12010
| null |
https://arxiv.org/abs/2205.12010v1
|
https://arxiv.org/pdf/2205.12010v1.pdf
|
SFace: Sigmoid-Constrained Hypersphere Loss for Robust Face Recognition
|
Deep face recognition has achieved great success due to large-scale training databases and rapidly developing loss functions. The existing algorithms devote to realizing an ideal idea: minimizing the intra-class distance and maximizing the inter-class distance. However, they may neglect that there are also low quality training images which should not be optimized in this strict way. Considering the imperfection of training databases, we propose that intra-class and inter-class objectives can be optimized in a moderate way to mitigate overfitting problem, and further propose a novel loss function, named sigmoid-constrained hypersphere loss (SFace). Specifically, SFace imposes intra-class and inter-class constraints on a hypersphere manifold, which are controlled by two sigmoid gradient re-scale functions respectively. The sigmoid curves precisely re-scale the intra-class and inter-class gradients so that training samples can be optimized to some degree. Therefore, SFace can make a better balance between decreasing the intra-class distances for clean examples and preventing overfitting to the label noise, and contributes more robust deep face recognition models. Extensive experiments of models trained on CASIA-WebFace, VGGFace2, and MS-Celeb-1M databases, and evaluated on several face recognition benchmarks, such as LFW, MegaFace and IJB-C databases, have demonstrated the superiority of SFace.
|
['Dongchao Wen', 'Xian Li', 'Dongyue Zhao', 'Jiani Hu', 'Weihong Deng', 'Yaoyao Zhong']
|
2022-05-24
|
sface-sigmoid-constrained-hypersphere-loss
|
https://ieeexplore.ieee.org/document/9318547
|
https://ieeexplore.ieee.org/document/9318547
|
ieee-transactions-on-image-processing-2021-1
|
['robust-face-recognition']
|
['computer-vision']
|
[-2.16111302e-01 -1.71382234e-01 -2.84430720e-02 -8.63954425e-01
-2.90071338e-01 -6.95219263e-03 3.85717392e-01 -3.78435582e-01
-4.20749277e-01 5.39330184e-01 -1.46369830e-01 4.90187705e-02
-3.42287153e-01 -9.07393634e-01 -5.49035549e-01 -9.13488269e-01
-6.76406845e-02 1.87703550e-01 -6.05208799e-02 -1.69985220e-01
1.63760588e-01 6.26250088e-01 -1.47670376e+00 1.17769480e-01
1.03723872e+00 1.29972720e+00 -1.72769010e-01 -5.19510470e-02
-2.08529234e-01 2.68816501e-01 -7.43621051e-01 -5.27050912e-01
3.06477576e-01 -2.01190129e-01 -3.04617107e-01 1.60589769e-01
5.25789499e-01 -1.26076773e-01 -4.41575199e-01 1.38208663e+00
5.95354795e-01 9.34718698e-02 6.68401182e-01 -1.39565206e+00
-8.95476639e-01 7.17613697e-02 -6.82042003e-01 8.87942240e-02
-1.79118186e-01 6.58740550e-02 5.57958901e-01 -1.09001911e+00
2.79949844e-01 1.57497764e+00 6.29352450e-01 6.82607293e-01
-9.22896504e-01 -1.03126085e+00 1.46661058e-01 3.15291613e-01
-1.72470498e+00 -5.71494639e-01 6.88383758e-01 -1.87787011e-01
6.62537336e-01 3.42055649e-01 2.33526126e-01 8.51674676e-01
8.06716979e-02 4.91019964e-01 1.09726942e+00 -2.22489119e-01
2.92222369e-02 2.74609298e-01 6.63396195e-02 8.25708628e-01
2.49324426e-01 1.61218345e-01 -2.78694719e-01 -2.06146929e-02
8.18538904e-01 2.40306541e-01 -5.65871894e-01 -2.79632777e-01
-8.41660857e-01 8.54024291e-01 5.80815375e-01 2.27263495e-01
-1.49857581e-01 -4.72131103e-01 3.91692609e-01 2.87763804e-01
4.29320425e-01 -3.25569473e-02 -2.17349574e-01 2.48362854e-01
-6.34263158e-01 -1.89129144e-01 5.02264559e-01 8.44757974e-01
8.51169288e-01 1.34762257e-01 -2.64090568e-01 1.44192934e+00
5.72402477e-01 6.46199882e-01 6.05391622e-01 -4.48396087e-01
5.76883793e-01 6.91296220e-01 -2.11619154e-01 -1.40091681e+00
-2.32476801e-01 -5.28788447e-01 -1.06018221e+00 3.60860765e-01
4.06281739e-01 2.61622127e-02 -1.06935561e+00 1.76539040e+00
4.43415463e-01 2.37704054e-01 2.55906116e-02 1.02871323e+00
9.32781756e-01 6.20760322e-01 -3.31812468e-03 -3.18191081e-01
1.16404891e+00 -8.85125875e-01 -8.31877470e-01 -1.10558145e-01
1.91548422e-01 -6.61896110e-01 1.25196850e+00 3.75215322e-01
-7.59500027e-01 -6.30204439e-01 -1.22871852e+00 1.89462572e-01
-4.21446860e-01 3.13939065e-01 4.41068202e-01 7.64292300e-01
-6.22321367e-01 6.39460802e-01 -6.26137614e-01 3.00589744e-02
9.32150185e-01 4.74939167e-01 -5.68744600e-01 -1.95111319e-01
-1.04217196e+00 6.47914052e-01 2.58920610e-01 5.91567338e-01
-7.79444516e-01 -5.43883443e-01 -5.64565361e-01 8.82582888e-02
2.78872281e-01 -3.70038450e-02 5.98932207e-01 -9.71088707e-01
-1.40342903e+00 9.26424444e-01 1.06753446e-01 1.33032218e-01
4.41838205e-01 1.34150699e-01 -9.72983778e-01 -5.50869443e-02
-1.73792735e-01 6.21150136e-01 9.55862105e-01 -1.21178424e+00
-4.13996428e-01 -8.58295858e-01 -1.90704852e-01 1.67647153e-01
-1.02266753e+00 -6.39378503e-02 -5.06637514e-01 -6.54553473e-01
7.85432532e-02 -6.00167871e-01 2.10204601e-01 1.91788867e-01
-3.56159538e-01 -3.46911430e-01 1.10975921e+00 -4.84830558e-01
1.08414960e+00 -2.23298645e+00 -1.49526849e-01 2.57784694e-01
6.92920163e-02 7.64974415e-01 -4.97977316e-01 -2.20686734e-01
-1.60543308e-01 1.94884408e-02 -3.56198758e-01 -2.25661740e-01
-1.43718183e-01 3.28638762e-01 -1.36720106e-01 5.42163789e-01
3.75013173e-01 5.91429830e-01 -6.42255127e-01 -3.70986044e-01
1.04658805e-01 7.75344074e-01 -2.60610133e-01 1.33248478e-01
9.25782323e-02 -2.26835646e-02 -4.30661201e-01 6.94391906e-01
1.26284790e+00 -6.37807474e-02 -1.61343157e-01 -3.13811719e-01
1.90041885e-01 -1.05730593e-01 -1.35746026e+00 1.14166534e+00
-2.55892545e-01 3.84084374e-01 3.10677022e-01 -1.25815618e+00
1.32199550e+00 -1.63807794e-02 3.26664686e-01 -1.00016665e+00
2.16632247e-01 1.85282111e-01 -8.75134543e-02 -4.73672599e-01
-9.99185741e-02 -2.79534340e-01 5.58703005e-01 -7.40837380e-02
-8.40174556e-02 3.90277296e-01 -2.35582646e-02 -2.86672503e-01
3.38317990e-01 -1.82921916e-01 -6.54719099e-02 -5.14633417e-01
8.87584090e-01 -5.72318792e-01 8.72531354e-01 1.92915425e-01
-3.97731036e-01 7.31970727e-01 4.17531908e-01 -5.08755922e-01
-5.88485658e-01 -8.55729699e-01 -6.48741007e-01 6.89415336e-01
3.08167368e-01 -1.40872478e-01 -9.51465786e-01 -1.03341186e+00
2.24939764e-01 2.22304583e-01 -7.62611806e-01 -4.68898296e-01
-5.52015245e-01 -1.18406916e+00 5.08064270e-01 4.69448447e-01
8.71635079e-01 -1.11779320e+00 1.23428060e-02 -9.19365585e-02
1.66497827e-01 -8.46087933e-01 -7.82089114e-01 -1.42143399e-01
-7.67365813e-01 -1.26621187e+00 -7.15905011e-01 -9.93115485e-01
9.08388019e-01 3.30058545e-01 8.29864144e-01 4.40121800e-01
-4.27253574e-01 -7.14018047e-02 -1.58334136e-01 -3.24880868e-01
4.66969348e-02 -3.26216012e-01 1.57010704e-01 4.07906801e-01
7.00560629e-01 -4.24107969e-01 -7.14926839e-01 7.99754798e-01
-8.87108326e-01 -4.52941775e-01 3.74533147e-01 1.20046580e+00
6.10314310e-01 1.42970115e-01 7.85964251e-01 -5.48822522e-01
5.25825858e-01 -3.77846271e-01 -6.41835988e-01 5.47827661e-01
-9.68038261e-01 -2.38544837e-01 8.38582218e-01 -5.52375317e-01
-9.75004435e-01 -3.54418874e-01 -1.99924871e-01 -6.25597596e-01
1.27361700e-01 2.11440325e-01 -6.82637513e-01 -4.52795595e-01
5.88260472e-01 2.45655105e-01 2.06657439e-01 -3.56121689e-01
-2.51628403e-02 8.21608007e-01 2.82584310e-01 -5.16858339e-01
6.86839283e-01 4.57125545e-01 1.27217304e-02 -8.04853797e-01
-5.22329688e-01 -1.77120358e-01 -3.36153328e-01 -9.31462571e-02
4.48314399e-01 -6.83489501e-01 -8.95215929e-01 8.14602673e-01
-8.06951880e-01 -1.59341563e-02 -1.39927924e-01 4.61328685e-01
4.36617248e-02 4.99443144e-01 -4.68818098e-01 -6.30035937e-01
-4.84947175e-01 -1.33924639e+00 8.30657303e-01 7.18954742e-01
5.44835627e-01 -8.82918239e-01 -3.51123810e-01 2.16403201e-01
5.20646214e-01 1.62522435e-01 7.15184093e-01 -5.21770895e-01
-3.81128818e-01 -1.65116787e-01 -5.84324062e-01 8.82174194e-01
4.56174821e-01 7.75272325e-02 -1.01941597e+00 -5.00289857e-01
1.74258038e-01 -4.87372339e-01 8.23534667e-01 1.93919852e-01
1.48587430e+00 -3.72241259e-01 -2.70438373e-01 1.12863088e+00
1.23284161e+00 3.48217636e-01 9.85752702e-01 3.15043658e-01
6.75178170e-01 5.96519113e-01 5.92779756e-01 2.72889405e-01
2.33145207e-01 7.09651470e-01 4.22088355e-01 -1.04972668e-01
-8.51245821e-02 -1.49991050e-01 1.24409646e-01 6.49392962e-01
9.43879113e-02 6.70029968e-03 -5.98923206e-01 1.83283463e-01
-1.53025925e+00 -8.22034180e-01 3.40822339e-02 2.35675883e+00
8.71858656e-01 4.25182283e-02 -4.18387763e-02 6.62936941e-02
7.21212864e-01 1.67562187e-01 -7.64200151e-01 2.55397763e-02
-4.22675997e-01 9.48020667e-02 2.09988505e-01 4.11485702e-01
-1.09202445e+00 7.05418468e-01 5.51370478e+00 1.28843653e+00
-1.55369306e+00 -5.76664461e-03 1.13783944e+00 -5.49565181e-02
-1.83109909e-01 -3.35793287e-01 -1.00080585e+00 7.54493594e-01
4.97421861e-01 9.55438614e-02 6.13971472e-01 9.71731782e-01
-6.30743355e-02 4.61600840e-01 -7.76080787e-01 1.37218750e+00
1.75070539e-01 -1.08817923e+00 6.20497353e-02 1.24174155e-01
5.95782757e-01 -8.56078491e-02 1.90108731e-01 4.43151772e-01
-2.93267727e-01 -1.32256818e+00 4.30365115e-01 4.73135829e-01
9.39289391e-01 -9.00334895e-01 5.83013713e-01 2.31861591e-01
-1.15606809e+00 -1.35972485e-01 -8.92108679e-01 4.06673700e-01
-4.48391974e-01 6.69533491e-01 -3.15434158e-01 4.08607215e-01
8.57913733e-01 6.77748740e-01 -5.25667071e-01 1.00689340e+00
-5.67442551e-02 3.91629547e-01 -3.85952502e-01 -5.19616641e-02
8.51563662e-02 -7.25389957e-01 2.68541247e-01 1.03246319e+00
3.13791007e-01 1.66177467e-01 3.32197174e-02 9.86262023e-01
-3.60563487e-01 2.84465075e-01 -1.63056597e-01 -2.73990026e-03
6.30674064e-01 1.44080162e+00 -4.90790904e-01 -1.34180561e-01
-3.98181349e-01 7.68512249e-01 4.55851406e-01 4.44209844e-01
-8.41940284e-01 -7.11503506e-01 8.19076300e-01 4.72478792e-02
2.04806790e-01 7.15677142e-02 -2.67663956e-01 -1.20992064e+00
3.87597203e-01 -1.15653205e+00 4.80796516e-01 -3.66475403e-01
-1.35946143e+00 8.89112532e-01 -3.41803312e-01 -9.72346425e-01
3.85588109e-01 -8.77382398e-01 -7.24467754e-01 9.53568757e-01
-1.81578648e+00 -8.99173498e-01 -6.75510526e-01 8.23039293e-01
1.55364767e-01 -4.94494110e-01 6.10935509e-01 7.58234441e-01
-9.74497318e-01 1.15686345e+00 2.80821800e-01 2.95327574e-01
7.83927381e-01 -8.03083956e-01 9.44659021e-03 4.74709630e-01
-3.50901075e-02 7.35441864e-01 2.11805299e-01 -3.65658820e-01
-1.27798271e+00 -1.12247646e+00 4.25818503e-01 -1.19054332e-01
1.57596275e-01 -3.14828962e-01 -1.29303634e+00 3.73061180e-01
-2.45748505e-01 5.31632066e-01 4.96965528e-01 2.51009036e-02
-6.00490630e-01 -8.12505186e-01 -1.43378234e+00 3.37453902e-01
1.18149364e+00 -4.35313761e-01 -3.17555517e-01 6.10948384e-01
2.79024959e-01 -2.10029438e-01 -9.77674007e-01 8.22423697e-01
4.28564191e-01 -1.00025535e+00 1.05512559e+00 -7.37444878e-01
1.88772932e-01 -3.49498659e-01 -9.22497511e-02 -1.34452033e+00
-2.04564452e-01 -3.04878771e-01 4.44139838e-02 1.39238310e+00
7.90031254e-02 -9.76031780e-01 8.65302145e-01 4.59679186e-01
-2.02349752e-01 -1.24068081e+00 -1.08913112e+00 -8.80815327e-01
1.16565429e-01 7.23687559e-02 9.40765917e-01 1.07762218e+00
-3.20120424e-01 -2.89121419e-02 -3.11788231e-01 1.42234027e-01
7.29549825e-01 -3.39024253e-02 6.11863554e-01 -1.23663461e+00
6.29058704e-02 -6.66472495e-01 -4.10255224e-01 -9.36018407e-01
1.56804383e-01 -8.52747858e-01 -2.59639248e-02 -1.02578437e+00
1.69754356e-01 -7.84360826e-01 -5.62157452e-01 6.33667171e-01
-3.62474680e-01 2.95195878e-01 -8.17839801e-02 3.52034181e-01
-2.82297820e-01 1.14473355e+00 1.55348146e+00 -2.51233995e-01
7.80856833e-02 -2.46644229e-01 -6.53399467e-01 5.20561934e-01
5.26446521e-01 -2.46172860e-01 -4.73848075e-01 -5.45938849e-01
-4.29242045e-01 -2.17304304e-01 2.19955996e-01 -8.57546628e-01
1.75113082e-01 -2.78104752e-01 6.61467433e-01 -2.03536615e-01
3.88652295e-01 -7.02079535e-01 -1.63589895e-01 2.91244537e-01
-6.53197095e-02 -3.07903141e-01 2.29037479e-01 4.52286124e-01
-4.14610267e-01 -7.67796785e-02 1.17833984e+00 1.80264041e-01
-5.35686195e-01 8.85848343e-01 4.78918314e-01 4.90135439e-02
1.00835621e+00 -4.55483228e-01 -4.68124777e-01 -7.05687776e-02
-3.50339413e-01 4.45265532e-01 2.81713843e-01 7.27027118e-01
9.03205872e-01 -1.69640458e+00 -8.14879358e-01 7.87171066e-01
1.15410335e-01 -5.94264828e-02 3.79719257e-01 6.49718046e-01
-3.07893932e-01 2.72716582e-01 -3.95948857e-01 -5.86924195e-01
-1.24519193e+00 4.79463369e-01 8.02097499e-01 6.58359155e-02
-4.32974428e-01 1.14291263e+00 3.37822378e-01 -6.24986827e-01
5.07660866e-01 1.07772909e-01 -2.22783327e-01 -7.01894313e-02
9.08595026e-01 2.98916817e-01 2.56901205e-01 -8.19898784e-01
-5.60608268e-01 6.91281974e-01 -2.59581447e-01 6.10886872e-01
1.26565218e+00 1.51534285e-02 -2.09859490e-01 -1.98097318e-01
1.53970599e+00 -1.81072414e-01 -1.33528459e+00 -1.43459201e-01
-2.02216119e-01 -8.21211457e-01 8.08815286e-02 -7.69679308e-01
-1.62964451e+00 9.71978664e-01 9.77989197e-01 1.05999291e-01
1.31012082e+00 -5.17952383e-01 8.06791008e-01 3.05092752e-01
1.67843536e-01 -1.17040360e+00 3.34876537e-01 4.02997881e-01
9.93989885e-01 -1.36573935e+00 -1.69248506e-01 -4.58233356e-01
-4.39058751e-01 1.25153553e+00 9.67070758e-01 -3.67787071e-02
7.93315470e-01 1.50938064e-01 1.21404134e-01 -1.43369928e-01
-3.41633946e-01 3.21074426e-01 5.26374638e-01 5.58129907e-01
3.29106838e-01 8.18859041e-02 -2.00125396e-01 7.13656127e-01
9.32460502e-02 1.94860119e-02 -8.04165453e-02 4.38315839e-01
-5.13947070e-01 -1.04024899e+00 -5.28653920e-01 4.56101298e-01
-2.20585003e-01 1.03891104e-01 -9.14409533e-02 8.46213996e-01
2.88636148e-01 9.32322145e-01 1.74239918e-03 -5.29913366e-01
4.55713779e-01 -2.06520751e-01 3.44891220e-01 -2.49144688e-01
-2.09893510e-01 -5.60293980e-02 -3.00581008e-01 -4.12615985e-01
-3.84766832e-02 -2.60373414e-01 -1.28161204e+00 -3.62739265e-01
-5.82677424e-01 2.58906662e-01 5.19647598e-01 8.07256460e-01
4.54521418e-01 3.24211359e-01 1.01067698e+00 -3.89214426e-01
-1.16354382e+00 -9.66521800e-01 -6.65276527e-01 6.86549664e-01
1.16856545e-01 -8.46072376e-01 -5.22803128e-01 -3.41323256e-01]
|
[13.182503700256348, 0.7688756585121155]
|
78de69a4-8706-4f3a-96d3-13ae6b9dd717
|
can-predicate-argument-relationships-be
| null | null |
https://aclanthology.org/2021.law-1.5
|
https://aclanthology.org/2021.law-1.5.pdf
|
Can predicate-argument relationships be extracted from UD trees?
|
In this paper we investigate the possibility of extracting predicate-argument relations from UD trees (and enhanced UD graphs). Con- cretely, we apply UD parsers on an En- glish question answering/semantic-role label- ing data set (FitzGerald et al., 2018) and check if the annotations reflect the relations in the resulting parse trees, using a small number of rules to extract this information. We find that 79.1% of the argument-predicate pairs can be found in this way, on the basis of Ud- ify (Kondratyuk and Straka, 2019). Error anal- ysis reveals that half of the error cases are at- tributable to shortcomings in the dataset. The remaining errors are mostly due to predicate- argument relations not being extractible algo- rithmically from the UD trees (requiring se- mantic reasoning to be resolved). The parser itself is only responsible for a small portion of errors. Our analysis suggests a number of improvements to the UD annotation schema: we propose to enhance the schema in four ways, in order to capture argument-predicate relations. Additionally, we propose improve- ments regarding data collection for question answering/semantic-role labeling data.
|
['Stergios Chatzikyriakidis', 'Jean-Philippe Bernardy', 'Adam Ek']
| null | null | null | null |
emnlp-law-dmr-2021-11
|
['semantic-role-labeling']
|
['natural-language-processing']
|
[ 1.12519152e-01 1.09074914e+00 -7.04586357e-02 -3.80594492e-01
-9.37307835e-01 -1.11641777e+00 4.21948165e-01 7.18697608e-01
-2.31733978e-01 8.46834958e-01 4.28286135e-01 -7.39597797e-01
-5.16504288e-01 -1.02994776e+00 -8.47849131e-01 -2.91759707e-02
4.58341897e-01 7.42921591e-01 4.97326314e-01 -4.48798805e-01
5.63439801e-02 1.81648016e-01 -1.67923450e+00 4.81479645e-01
9.03865218e-01 7.64943838e-01 -1.05082348e-01 6.41893506e-01
-4.61989522e-01 1.23848915e+00 -7.36446857e-01 -8.06060791e-01
-3.59628461e-02 -4.51328486e-01 -1.63869500e+00 -2.26313800e-01
4.93719935e-01 -4.85074148e-02 -1.00202993e-01 8.32257748e-01
1.49750650e-01 8.93002935e-03 4.66036350e-01 -1.29404271e+00
-5.77429116e-01 8.23424757e-01 1.12819165e-01 3.69266272e-01
7.48592257e-01 -4.16860938e-01 1.71711302e+00 -5.94322383e-01
1.00383556e+00 1.23286319e+00 4.80627596e-01 5.58416843e-01
-1.08224952e+00 -3.95117342e-01 -1.37592833e-02 5.94070077e-01
-8.19960713e-01 -2.79734790e-01 5.21711588e-01 -3.74616742e-01
1.33473897e+00 5.27365446e-01 1.97147548e-01 7.07687795e-01
-3.80653620e-01 6.75063670e-01 9.66152847e-01 -1.01761794e+00
-8.14924948e-03 1.32132936e-02 7.31070340e-01 6.93844736e-01
5.52806258e-01 -2.36979753e-01 -2.98158884e-01 -4.27488238e-01
3.41942728e-01 -7.38101304e-01 -2.71505974e-02 -1.05253495e-01
-7.42092371e-01 7.80143321e-01 1.91605911e-01 4.71971303e-01
-3.03399771e-01 -2.04781651e-01 4.64784056e-01 3.28382641e-01
2.24915281e-01 6.91103101e-01 -1.05579495e+00 -2.27824360e-01
-1.75205395e-01 4.11226839e-01 1.01533318e+00 8.40479851e-01
8.35880220e-01 -4.91656125e-01 -9.09055024e-02 9.98783231e-01
2.75059462e-01 -3.69050987e-02 4.64292735e-01 -1.27101433e+00
6.12221241e-01 1.06999958e+00 1.56373709e-01 -7.55084157e-01
-5.33151627e-01 9.26505625e-02 -6.12704456e-03 -3.96116376e-01
8.64489794e-01 -6.04894347e-02 -6.69589400e-01 1.64792204e+00
6.26363873e-01 -3.02224278e-01 3.33427727e-01 5.20120621e-01
1.24331903e+00 3.16400111e-01 3.83715957e-01 7.83324689e-02
2.00738215e+00 -7.15237617e-01 -8.75811040e-01 -3.50682199e-01
1.24386001e+00 -5.51769435e-01 1.14197016e+00 3.50316092e-02
-8.35664451e-01 -2.28369221e-01 -7.98831761e-01 -3.75470579e-01
-4.46615934e-01 5.39897047e-02 6.63339138e-01 4.70180154e-01
-6.23341680e-01 3.68351817e-01 -4.14868176e-01 -5.41364431e-01
1.09202676e-01 1.06593728e-01 -4.50088441e-01 -2.19875015e-02
-1.51999307e+00 1.01849735e+00 7.23146379e-01 -2.00849205e-01
-1.24331713e-01 -6.91365898e-01 -1.20439839e+00 1.14724338e-01
7.67498076e-01 -4.95567381e-01 1.51283777e+00 -5.55104911e-01
-9.39225435e-01 1.21728206e+00 -3.70206892e-01 -3.92763913e-01
8.81878585e-02 -1.37043986e-02 -4.01577652e-01 3.33006799e-01
5.52037716e-01 1.95092887e-01 1.79966792e-01 -1.16067636e+00
-9.67458069e-01 -5.86567521e-01 5.70436478e-01 6.22788221e-02
2.14894842e-02 4.03301656e-01 -1.79494202e-01 -4.42749023e-01
5.29394567e-01 -7.99565792e-01 2.01714396e-01 -5.41112840e-01
-3.31481010e-01 -8.64709258e-01 6.15192950e-01 -9.75946844e-01
1.47942352e+00 -1.70418549e+00 -2.56368250e-01 4.02277820e-02
2.23322570e-01 2.73304701e-01 -8.14006105e-03 5.58875561e-01
-4.96123165e-01 4.31104690e-01 -3.50618511e-01 5.12844250e-02
1.77638069e-01 8.15594137e-01 -3.55810344e-01 -9.22354311e-02
3.50321025e-01 1.00957394e+00 -7.83274114e-01 -4.21707988e-01
-3.82784829e-02 -3.51192653e-01 -5.16570449e-01 3.45943272e-01
-6.04052007e-01 7.32419118e-02 -3.15983146e-01 6.15127325e-01
4.77361739e-01 -6.78213015e-02 4.66190398e-01 -2.93691363e-03
-5.53347990e-02 1.25163198e+00 -1.22035217e+00 1.06525564e+00
-3.52704853e-01 5.86340070e-01 -9.25933495e-02 -1.21766305e+00
7.42567122e-01 3.06329072e-01 3.19242299e-01 -7.58806586e-01
-4.96719405e-03 4.45332557e-01 2.18478739e-01 -9.50709581e-01
5.27544081e-01 -9.86120328e-02 -1.90268993e-01 3.55883896e-01
2.12061733e-01 4.24590185e-02 6.38440132e-01 1.99527889e-01
1.24655616e+00 2.35326171e-01 6.51836932e-01 -2.23504588e-01
7.38752961e-01 6.77870512e-01 7.41762877e-01 5.91352046e-01
-1.35185167e-01 4.55152899e-01 1.09225380e+00 -2.99894392e-01
-8.48636925e-01 -6.12129748e-01 -2.92348862e-01 1.21365619e+00
-2.39062235e-01 -8.69089663e-01 -8.07689548e-01 -1.15391588e+00
-2.13141561e-01 1.14102447e+00 -6.19109869e-01 -3.35136540e-02
-8.36692512e-01 -4.65461344e-01 9.38485384e-01 5.02729058e-01
1.96694314e-01 -1.18832135e+00 -7.20565915e-01 4.48782831e-01
-6.55641079e-01 -1.45374060e+00 4.47328568e-01 5.70228398e-01
-6.37798011e-01 -1.65470016e+00 2.63377905e-01 -7.44143546e-01
4.37940776e-01 -1.60878778e-01 1.36859262e+00 3.89073938e-01
1.12713188e-01 3.69201243e-01 -6.77090228e-01 -3.67491543e-01
-7.27025032e-01 5.33097982e-02 -4.50869113e-01 -6.32003307e-01
7.74460435e-01 -2.94073671e-01 -9.62523296e-02 2.94163525e-01
-7.49980211e-01 -3.57911199e-01 2.19952926e-01 7.31322646e-01
6.32960320e-01 6.95213825e-02 4.06789333e-01 -1.43802476e+00
4.59759086e-01 -6.33712411e-01 -4.37937260e-01 3.95215094e-01
-5.80322921e-01 3.75431776e-01 5.46646237e-01 -5.35506830e-02
-1.11383927e+00 -2.33001351e-01 -7.77538836e-01 2.16056317e-01
-4.83176768e-01 4.71919239e-01 -3.91948313e-01 3.88443261e-01
8.17145109e-01 -3.42223555e-01 -3.33358020e-01 -5.98914921e-01
3.26202810e-01 5.14338017e-01 4.25809324e-01 -1.23211539e+00
5.10487497e-01 1.09949410e-01 -1.68147460e-01 -4.67275769e-01
-1.42868447e+00 -5.84004998e-01 -4.68717754e-01 3.00244153e-01
7.93933153e-01 -5.77127099e-01 -6.91366673e-01 4.56083752e-03
-1.29582012e+00 -1.86816394e-01 -4.85128939e-01 1.96675807e-01
-4.36344385e-01 4.29052413e-01 -7.91486382e-01 -6.71914577e-01
-1.54082477e-01 -7.92990208e-01 9.03831720e-01 6.25859722e-02
-6.04336143e-01 -8.44408274e-01 -3.08953170e-02 7.02362061e-01
-1.84238702e-01 1.29119217e-01 1.57794976e+00 -1.27196431e+00
-9.23087746e-02 3.40232849e-02 -3.40436608e-01 1.56560689e-01
7.79501721e-02 -1.30659372e-01 -1.02231336e+00 4.09338772e-01
6.18719086e-02 -1.31328925e-01 3.81175399e-01 2.47570872e-02
1.01208234e+00 -4.73209858e-01 -5.66115007e-02 -1.18728071e-01
1.17156518e+00 1.01540126e-01 6.52461767e-01 7.71340132e-01
5.55054307e-01 1.06833887e+00 8.13507080e-01 -9.85671431e-02
1.00456524e+00 5.86173594e-01 3.68430972e-01 4.04704571e-01
-2.47505337e-01 -6.99397683e-01 9.07645673e-02 4.09438908e-01
1.03727020e-01 -4.21381630e-02 -1.10244179e+00 7.93634236e-01
-2.01951575e+00 -7.92337775e-01 -9.99771237e-01 1.78518367e+00
9.25710857e-01 2.13026762e-01 1.08699039e-01 3.27728122e-01
6.05918169e-01 -1.75636008e-01 -1.17624074e-01 -5.85444331e-01
-2.05869064e-01 4.50563639e-01 3.08987647e-01 6.87357962e-01
-7.49923527e-01 1.17084885e+00 6.09515572e+00 4.76093382e-01
-3.08726639e-01 1.63842127e-01 1.54975533e-01 6.44526184e-01
-5.92116594e-01 4.53545719e-01 -8.85648966e-01 9.33267772e-02
1.16634524e+00 1.47539955e-02 1.49274036e-01 8.55358839e-01
-2.71645337e-01 -3.48001033e-01 -1.01695895e+00 5.55601180e-01
1.95154753e-02 -1.23950565e+00 -1.63441986e-01 -1.18707225e-01
1.15589589e-01 -3.29334885e-01 -6.60348952e-01 6.61434531e-01
4.42212343e-01 -6.64085209e-01 9.40260530e-01 5.41206114e-02
4.03911531e-01 -4.93874937e-01 9.48275328e-01 4.87791896e-01
-9.43495631e-01 -1.08323365e-01 -3.96558762e-01 -3.87704879e-01
-1.03776731e-01 5.43914676e-01 -8.92412424e-01 7.82208920e-01
9.44228649e-01 2.66816318e-01 -7.53498018e-01 4.77684259e-01
-9.55206156e-01 1.00051188e+00 -3.75457376e-01 -5.09727001e-03
5.41405566e-03 -1.97945029e-01 4.74071831e-01 1.01849687e+00
9.06238109e-02 4.42843795e-01 -2.27315083e-01 7.57763863e-01
-1.68426543e-01 1.52867734e-01 -3.65272909e-01 -7.44419321e-02
5.80951512e-01 1.15022051e+00 -5.99520862e-01 -4.62002844e-01
-3.77708554e-01 7.42405176e-01 7.44325876e-01 7.07085580e-02
-6.66377783e-01 -2.00104058e-01 6.21033311e-01 1.53496772e-01
2.06330657e-01 -8.34762678e-03 -4.16873664e-01 -1.08416367e+00
3.83036345e-01 -1.01179481e+00 1.08882022e+00 -8.80766928e-01
-1.24165905e+00 4.70368624e-01 1.54470041e-01 -5.68187237e-01
-5.24892271e-01 -7.32932568e-01 -1.60958976e-01 8.21458578e-01
-1.66608143e+00 -1.07031679e+00 -6.61283433e-02 2.41509557e-01
2.40594029e-01 4.36819792e-01 1.13700664e+00 2.09863037e-01
-3.96053851e-01 3.78316462e-01 -6.49677873e-01 2.83999264e-01
6.27177775e-01 -1.40143895e+00 4.10364240e-01 9.24107313e-01
2.08269283e-01 7.42429137e-01 8.05120826e-01 -6.19461298e-01
-1.08529818e+00 -6.96450114e-01 1.68637609e+00 -9.25463498e-01
9.25791800e-01 -3.36717635e-01 -1.33812714e+00 9.44533706e-01
1.30773455e-01 2.61503272e-04 8.37983489e-01 4.61260974e-01
-5.55263877e-01 3.53685558e-01 -1.27147293e+00 2.19519258e-01
1.18746746e+00 -6.88360810e-01 -1.35262275e+00 2.78593451e-01
9.62395668e-01 -6.14727020e-01 -9.09996152e-01 3.73378277e-01
6.95988685e-02 -8.49585235e-01 6.20659471e-01 -1.09803414e+00
4.80095863e-01 -3.87515098e-01 -3.27562779e-01 -9.85263348e-01
-3.56508754e-02 -3.19374382e-01 -2.88455188e-01 1.64121532e+00
7.30966926e-01 -7.44231820e-01 6.72254264e-01 1.03183031e+00
-3.11575234e-01 -4.74446505e-01 -9.02448714e-01 -5.90931296e-01
1.65593952e-01 -8.87719035e-01 7.23801732e-01 1.11287284e+00
1.03912778e-01 7.13117719e-01 4.16210920e-01 2.60327876e-01
1.83348954e-01 6.82430044e-02 5.41621208e-01 -1.41261601e+00
-3.01720381e-01 -1.30536407e-02 3.74603346e-02 -7.98606873e-01
5.39593518e-01 -9.54191685e-01 -1.06208667e-01 -2.04093313e+00
-2.94705451e-01 -8.16542506e-01 2.30914116e-01 1.00029111e+00
-3.17866385e-01 -1.02619514e-01 8.80430192e-02 1.90149590e-01
-4.23450798e-01 1.56785518e-01 8.59938383e-01 1.47351637e-01
-2.92255543e-02 -1.98965058e-01 -1.00980377e+00 8.99634242e-01
7.69863248e-01 -6.83630824e-01 -8.04359615e-02 -3.94189239e-01
6.29869521e-01 4.65460820e-03 2.40639463e-01 -3.72974634e-01
-1.18192136e-02 -6.89374804e-02 -1.36540830e-01 -3.50747019e-01
-5.56324748e-03 -8.39637399e-01 -1.09147564e-01 2.02497363e-01
-1.91029668e-01 1.02357328e-01 3.23836327e-01 2.20324874e-01
-4.06712890e-01 -9.26447988e-01 3.47633481e-01 -1.49410158e-01
-9.45660949e-01 -3.88873965e-01 -3.96854579e-01 5.02368331e-01
9.05503571e-01 -2.97614783e-02 -7.06649125e-01 6.41173869e-02
-6.98368430e-01 1.54743761e-01 2.27084607e-01 3.43575805e-01
-2.54370440e-02 -1.03849006e+00 -4.90432888e-01 -5.02597205e-02
4.98099059e-01 1.05872914e-01 7.28152841e-02 4.68986452e-01
-5.60933650e-01 6.07938886e-01 1.09374702e-01 -1.53115327e-02
-1.19671047e+00 3.07889998e-01 2.55955249e-01 -5.14818907e-01
-5.54106295e-01 6.84116006e-01 -2.82979965e-01 -9.76620197e-01
-1.56634063e-01 -4.64085907e-01 -4.83837485e-01 2.73264468e-01
1.96169809e-01 3.07429075e-01 5.85603118e-01 -6.40228748e-01
-4.87456560e-01 1.93006039e-01 1.73297733e-01 2.15002149e-02
1.32908654e+00 -8.68616551e-02 -5.95439255e-01 1.97259679e-01
9.60809708e-01 3.62151414e-01 -3.98214757e-01 -2.54676163e-01
6.92298114e-01 -2.62788713e-01 -4.10278499e-01 -1.03440309e+00
-4.63634610e-01 5.04874527e-01 -6.29017353e-02 7.19712377e-01
7.72345662e-01 6.67337418e-01 6.97589755e-01 3.97869766e-01
2.48761058e-01 -1.16271496e+00 -4.69789326e-01 9.03030515e-01
6.57728553e-01 -1.07954729e+00 -2.62166798e-01 -9.63516593e-01
-4.87058997e-01 1.13243914e+00 8.19029450e-01 5.18153682e-02
2.79025227e-01 1.01018727e-01 -2.52021514e-02 -5.70454240e-01
-7.53688335e-01 -6.43573999e-01 2.54729122e-01 5.95023036e-01
5.97605467e-01 7.33654201e-02 -9.13490117e-01 9.89750564e-01
-6.79760575e-01 -5.39836526e-01 7.77932167e-01 9.29446459e-01
-3.29687119e-01 -1.39272523e+00 -5.51073909e-01 4.00963426e-01
-6.15655184e-01 -8.33667666e-02 -6.25836313e-01 9.65187311e-01
2.32259393e-01 1.32918549e+00 9.76594687e-02 8.21810495e-03
7.85878599e-01 3.78694028e-01 5.21961272e-01 -1.00274348e+00
-5.64122140e-01 -3.57221544e-01 9.94547963e-01 -5.20971537e-01
-5.09829044e-01 -6.45594358e-01 -1.79848480e+00 1.01448419e-02
-3.46471429e-01 4.04597521e-01 5.65579057e-01 1.39773953e+00
3.89957875e-01 6.54006064e-01 -6.51172400e-02 2.76744008e-01
-3.20091903e-01 -8.61406267e-01 -1.45184904e-01 5.82443714e-01
-1.67201847e-01 -6.78188741e-01 -3.96031827e-01 -1.63387023e-02]
|
[10.10348129272461, 9.083333015441895]
|
3590178b-e881-4b98-8142-11dd038f5fb0
|
efficient-transformer-based-method-for-remote
|
2103.00208
| null |
https://arxiv.org/abs/2103.00208v3
|
https://arxiv.org/pdf/2103.00208v3.pdf
|
Remote Sensing Image Change Detection with Transformers
|
Modern change detection (CD) has achieved remarkable success by the powerful discriminative ability of deep convolutions. However, high-resolution remote sensing CD remains challenging due to the complexity of objects in the scene. Objects with the same semantic concept may show distinct spectral characteristics at different times and spatial locations. Most recent CD pipelines using pure convolutions are still struggling to relate long-range concepts in space-time. Non-local self-attention approaches show promising performance via modeling dense relations among pixels, yet are computationally inefficient. Here, we propose a bitemporal image transformer (BIT) to efficiently and effectively model contexts within the spatial-temporal domain. Our intuition is that the high-level concepts of the change of interest can be represented by a few visual words, i.e., semantic tokens. To achieve this, we express the bitemporal image into a few tokens, and use a transformer encoder to model contexts in the compact token-based space-time. The learned context-rich tokens are then feedback to the pixel-space for refining the original features via a transformer decoder. We incorporate BIT in a deep feature differencing-based CD framework. Extensive experiments on three CD datasets demonstrate the effectiveness and efficiency of the proposed method. Notably, our BIT-based model significantly outperforms the purely convolutional baseline using only 3 times lower computational costs and model parameters. Based on a naive backbone (ResNet18) without sophisticated structures (e.g., FPN, UNet), our model surpasses several state-of-the-art CD methods, including better than four recent attention-based methods in terms of efficiency and accuracy. Our code is available at https://github.com/justchenhao/BIT\_CD.
|
['Zhenwei Shi', 'Zipeng Qi', 'Hao Chen']
|
2021-02-27
| null | null | null | null |
['building-change-detection-for-remote-sensing']
|
['miscellaneous']
|
[ 4.07011241e-01 -4.68757778e-01 1.68353036e-01 -4.26061720e-01
-7.73483515e-01 -5.40724456e-01 7.74293542e-01 1.69107635e-02
-3.39748681e-01 4.41536069e-01 3.64856362e-01 -1.80357605e-01
-1.84806213e-02 -1.16370893e+00 -8.93951833e-01 -8.58813345e-01
-9.53436457e-03 -7.06770048e-02 3.27245802e-01 -2.80095875e-01
-5.80304787e-02 3.59072894e-01 -1.69090426e+00 3.89011919e-01
8.54149938e-01 1.12792349e+00 6.86782658e-01 4.83667284e-01
-1.37522161e-01 6.48739696e-01 -4.33362126e-02 3.35623138e-02
3.66602540e-01 -3.97245079e-01 -6.73491836e-01 -3.64417210e-02
6.15512848e-01 -4.80204850e-01 -4.60143954e-01 1.19255245e+00
4.30471629e-01 8.67912471e-02 2.15760514e-01 -6.50923729e-01
-9.20967877e-01 3.62063676e-01 -5.22071898e-01 1.78301007e-01
-4.82482091e-02 1.17297605e-01 1.25071537e+00 -1.17996156e+00
4.85832363e-01 1.13647544e+00 8.28038633e-01 2.10530296e-01
-1.29253364e+00 -7.30363667e-01 4.53459710e-01 4.61190015e-01
-1.57222867e+00 -2.17437491e-01 7.30120838e-01 -3.96609366e-01
1.17888784e+00 2.18645066e-01 8.50045025e-01 7.99043357e-01
1.96894426e-02 6.74926579e-01 1.21981955e+00 -2.29235828e-01
1.97041467e-01 -3.56736451e-01 -2.02486202e-01 6.93508446e-01
1.55053511e-02 5.90797886e-02 -5.27769804e-01 1.45305440e-01
8.29842985e-01 4.00683075e-01 -3.78087223e-01 -1.04994029e-01
-1.31784511e+00 8.41461360e-01 1.08687627e+00 5.17495394e-01
-5.52568614e-01 5.62894106e-01 -4.18828640e-05 1.80044807e-02
7.61061668e-01 2.26017222e-01 -5.29375255e-01 1.92354262e-01
-9.13511455e-01 2.52900153e-01 2.54367441e-01 7.21935511e-01
1.18384171e+00 -2.22297907e-01 -2.57482857e-01 8.31026077e-01
2.47634932e-01 6.52602911e-01 2.03634501e-01 -7.86750317e-01
2.60431319e-01 3.63269120e-01 5.80957974e-04 -1.01939023e+00
-2.31739625e-01 -6.02518976e-01 -9.32667196e-01 2.82350536e-02
-3.52863334e-02 2.12895587e-01 -1.18933976e+00 1.82350206e+00
2.82064587e-01 4.69021231e-01 -2.14175224e-01 8.47415030e-01
6.00976706e-01 1.02846611e+00 1.72571331e-01 7.64264986e-02
1.41747093e+00 -9.16033149e-01 -5.27242601e-01 -4.75500792e-01
3.44010293e-01 -5.13803840e-01 1.03257751e+00 -2.31237113e-02
-8.11315358e-01 -5.10165870e-01 -9.10218418e-01 -3.50286365e-01
-5.97156048e-01 1.85380038e-02 7.01728404e-01 1.52568460e-01
-1.14574814e+00 6.65601015e-01 -9.21596646e-01 -5.09769917e-01
6.88152015e-01 -5.77503070e-02 -1.00102976e-01 -2.77572721e-01
-1.28686154e+00 6.61814868e-01 2.78991163e-01 2.33877406e-01
-9.89403844e-01 -9.65881705e-01 -9.48997617e-01 1.76265925e-01
1.82859004e-01 -7.31311679e-01 1.22513127e+00 -1.09382784e+00
-1.26789463e+00 7.14233220e-01 -3.43782216e-01 -3.27634037e-01
2.98212826e-01 -2.60974616e-01 -3.96300107e-01 2.46259958e-01
3.01845163e-01 9.55451667e-01 7.95535922e-01 -1.24433208e+00
-1.00851893e+00 -1.96733832e-01 3.52835596e-01 3.72851044e-01
-1.82092488e-01 -8.43490809e-02 -4.71591949e-01 -8.20192814e-01
4.84615803e-01 -7.29782999e-01 -1.86972350e-01 6.55172408e-01
-6.75003007e-02 4.84383889e-02 7.62021780e-01 -5.37744761e-01
1.11500525e+00 -2.28111362e+00 -9.78205577e-02 -5.73063008e-02
2.81152874e-01 2.80187696e-01 -2.24287555e-01 4.07625854e-01
-8.24906453e-02 1.60041153e-01 -7.29290903e-01 -1.88829571e-01
-3.43237855e-02 4.05345768e-01 -4.38943475e-01 5.31051278e-01
4.08791512e-01 1.09323549e+00 -1.19251525e+00 -1.29387110e-01
4.69119191e-01 6.32651687e-01 -6.58716083e-01 8.86063129e-02
-4.48826522e-01 3.22502762e-01 -4.17662293e-01 6.63703144e-01
8.57225537e-01 -3.21095407e-01 2.50230655e-02 -3.85059446e-01
-5.87037385e-01 5.76667070e-01 -8.75466049e-01 1.90026140e+00
-5.93042076e-01 8.38559806e-01 -6.41986355e-02 -1.08446121e+00
6.59343839e-01 -6.61763027e-02 4.28693593e-01 -1.01076460e+00
-2.80917376e-01 3.22696686e-01 -2.81581789e-01 -3.71930003e-01
6.40264452e-01 -2.85266787e-01 1.71735808e-01 9.20914933e-02
-1.72375143e-01 -1.64225802e-01 -2.66286612e-01 -4.57778089e-02
9.77025628e-01 3.45372826e-01 2.88408846e-01 -2.80991584e-01
2.31714800e-01 1.07765846e-01 5.53963006e-01 7.85649300e-01
-9.59950686e-02 7.60279357e-01 -1.84223764e-02 -4.48238164e-01
-9.62402165e-01 -1.07358885e+00 -3.53437394e-01 1.04436421e+00
3.09577703e-01 -3.18624645e-01 -3.18072975e-01 -2.49609128e-01
2.73661539e-02 4.49615657e-01 -6.09283388e-01 2.62715686e-02
-5.16317785e-01 -8.08939934e-01 5.33458829e-01 5.15388429e-01
1.09627366e+00 -7.74229765e-01 -6.15047097e-01 4.62888181e-01
-3.82347226e-01 -1.06543016e+00 -1.82313800e-01 2.40260363e-01
-6.62993848e-01 -6.47916973e-01 -6.25467300e-01 -7.26326168e-01
3.63314450e-01 9.25705492e-01 9.19054568e-01 -1.59749463e-01
-3.18596929e-01 1.17490947e-01 -4.74337459e-01 -1.83009863e-01
1.57455772e-01 -1.09624997e-01 -3.77661854e-01 5.86516969e-02
4.04607296e-01 -7.42574751e-01 -1.13452184e+00 4.39772494e-02
-1.15970230e+00 3.25377643e-01 5.67297220e-01 9.55341578e-01
8.10281217e-01 1.34913743e-01 1.58382133e-01 -5.74477434e-01
-1.91161614e-02 -6.26796365e-01 -4.65664357e-01 1.60575479e-01
-5.27605772e-01 1.17262028e-01 3.52204323e-01 -2.17062384e-01
-1.17777979e+00 8.79246667e-02 -2.64652699e-01 -2.14000463e-01
-1.89531088e-01 7.78590798e-01 -5.06151542e-02 -4.64466102e-02
5.94970405e-01 4.65850055e-01 -4.18105662e-01 -6.24426484e-01
6.02246463e-01 6.11882806e-01 6.03092074e-01 -4.49382186e-01
8.32852602e-01 1.03017294e+00 -2.26793170e-01 -7.81625926e-01
-1.08507049e+00 -6.29522443e-01 -5.21204948e-01 -3.38868075e-03
8.56392384e-01 -1.42172766e+00 -3.00741363e-02 8.36098015e-01
-1.13280737e+00 -4.39753592e-01 -3.02214295e-01 3.44868302e-01
-2.80284792e-01 3.13398182e-01 -4.86510545e-01 -4.22015280e-01
-2.58953273e-01 -7.23609805e-01 1.33776903e+00 1.30387694e-01
3.29924673e-01 -8.99099767e-01 4.80863042e-02 -4.24936302e-02
7.56392598e-01 2.77045548e-01 8.74162614e-01 1.01335146e-01
-8.78345847e-01 2.30783165e-01 -7.08448768e-01 2.62971520e-01
4.32546914e-01 -1.58660457e-01 -1.28751588e+00 -1.88422367e-01
-1.36551633e-01 -5.05408123e-02 1.38466215e+00 5.10794997e-01
1.42866218e+00 -2.90630996e-01 -3.02168489e-01 1.00558126e+00
1.86468422e+00 -6.59018829e-02 7.89708078e-01 4.03304785e-01
8.97478104e-01 3.77682477e-01 5.41490078e-01 5.97662628e-01
6.87024772e-01 6.09905005e-01 7.01305151e-01 -4.15318936e-01
-4.45458561e-01 -2.58352339e-01 3.90069783e-01 6.06377482e-01
-9.26237106e-02 -2.50364721e-01 -9.01224792e-01 1.00527513e+00
-1.84839392e+00 -1.28567457e+00 -1.49753451e-01 1.91011012e+00
1.01170886e+00 -1.96565837e-01 -4.66653854e-01 -2.14190736e-01
5.95354199e-01 7.02485442e-01 -7.03948736e-01 1.24698043e-01
-3.95478874e-01 4.88989979e-01 6.78515017e-01 4.82013136e-01
-1.32563639e+00 1.21526635e+00 5.31783009e+00 8.10915232e-01
-1.39176810e+00 3.43943626e-01 3.36428255e-01 -6.75802156e-02
-5.31637847e-01 1.56453595e-01 -6.92103684e-01 2.13632151e-01
5.70628226e-01 1.36095315e-01 5.51972747e-01 5.65312982e-01
3.29830140e-01 -1.47148564e-01 -8.00270081e-01 9.63129103e-01
1.34644331e-02 -1.51850986e+00 1.30272344e-01 -1.68027356e-01
9.36655521e-01 7.01428413e-01 1.41438991e-01 1.97553262e-01
2.84916162e-01 -8.57098043e-01 1.02878964e+00 4.99963552e-01
1.01149118e+00 -2.77639449e-01 3.70926619e-01 -1.62355080e-02
-1.71211934e+00 -7.91500136e-02 -5.59774220e-01 -2.53050268e-01
-1.20369950e-02 7.70898283e-01 -3.86277825e-01 6.82651222e-01
1.02224910e+00 1.31607103e+00 -4.20498729e-01 8.10681999e-01
-2.54691213e-01 6.01145864e-01 -4.17514116e-01 3.08988392e-01
6.82323873e-01 -7.66297877e-02 3.80701482e-01 1.34433675e+00
6.00903273e-01 3.63868743e-01 -1.96210947e-02 1.24993193e+00
-1.61186568e-02 -2.51887739e-01 -4.97343034e-01 4.99871653e-03
5.31631351e-01 1.20133531e+00 -3.21218014e-01 -3.10034961e-01
-7.03741193e-01 1.21069181e+00 2.21559450e-01 3.71700168e-01
-9.09079432e-01 -3.60277206e-01 1.15199399e+00 6.28508851e-02
7.76363432e-01 -3.57641131e-01 -1.71887949e-01 -1.19384038e+00
1.95393160e-01 -5.62345207e-01 2.04825968e-01 -1.06825650e+00
-1.21747494e+00 4.62106705e-01 -8.75976235e-02 -1.39302576e+00
2.67037243e-01 -5.40545583e-01 -3.74792546e-01 9.86843944e-01
-2.20799828e+00 -1.52698529e+00 -7.57346213e-01 7.50199199e-01
6.19341969e-01 4.73243237e-01 6.92127049e-01 3.23181361e-01
-3.23990881e-01 5.50338775e-02 4.61991251e-01 2.02660728e-03
6.40956104e-01 -1.14431036e+00 6.49802625e-01 1.12919283e+00
7.35762492e-02 4.42133754e-01 3.79487157e-01 -5.26933491e-01
-1.22883272e+00 -1.67727637e+00 1.02670228e+00 -1.38184085e-01
8.98023069e-01 -3.78588349e-01 -1.07230771e+00 5.39389074e-01
3.53524722e-02 2.92605042e-01 3.31472516e-01 -1.73471823e-01
-6.39626741e-01 -3.37142915e-01 -8.97306323e-01 6.10438406e-01
1.45514107e+00 -1.09219611e+00 -4.94440943e-01 3.16780657e-01
9.31065857e-01 -2.77560025e-01 -6.48100197e-01 3.57863605e-01
5.37988722e-01 -9.91258323e-01 1.03066015e+00 -1.69598967e-01
6.23382390e-01 -5.88396847e-01 -6.09430671e-01 -1.23796904e+00
-8.18992317e-01 -2.08173901e-01 1.63444027e-01 1.09590077e+00
7.12124556e-02 -7.19240010e-01 1.89807817e-01 2.10568473e-01
-3.64158869e-01 -5.14663041e-01 -9.37125921e-01 -7.49731064e-01
1.97302878e-01 -6.62367344e-01 8.48950565e-01 1.13321483e+00
-4.53201175e-01 1.13565035e-01 -1.76630944e-01 6.15960836e-01
3.99763525e-01 5.38129985e-01 1.89306393e-01 -1.00292778e+00
-1.45346880e-01 -5.07141888e-01 -2.68124044e-01 -1.40707970e+00
-7.90834874e-02 -9.62851644e-01 3.12442005e-01 -1.73431206e+00
3.41532260e-01 -7.18202710e-01 -4.26021755e-01 8.35647702e-01
-2.51631767e-01 3.45736861e-01 5.75880660e-03 3.31751883e-01
-3.09049517e-01 8.19687605e-01 1.07933676e+00 -4.01726305e-01
-2.24482808e-02 -5.79126418e-01 -5.56904495e-01 4.83357579e-01
7.87931740e-01 -5.21780491e-01 -1.20375708e-01 -1.00145650e+00
3.29897940e-01 -4.59108502e-01 8.74050379e-01 -1.03876436e+00
1.93356529e-01 -2.69632369e-01 1.72759414e-01 -6.58175111e-01
3.14723164e-01 -7.76814580e-01 3.86307448e-01 3.61372828e-01
-2.00436532e-01 -2.83924013e-01 1.19912595e-01 6.20424509e-01
-3.12028319e-01 1.66527972e-01 8.29767227e-01 -2.61168182e-01
-1.08220434e+00 6.46190464e-01 -2.74662197e-01 -2.44549617e-01
6.21849000e-01 -1.32768974e-01 -5.94188273e-01 -2.23814085e-01
-2.66818374e-01 2.57284399e-02 4.95015115e-01 3.56987655e-01
5.46854675e-01 -1.30388737e+00 -7.92014778e-01 1.56739235e-01
3.99596989e-01 2.31372908e-01 5.04486501e-01 7.16442227e-01
-5.88722527e-01 3.31517994e-01 -1.59889132e-01 -7.80786693e-01
-8.17652702e-01 3.02648067e-01 6.22855008e-01 4.18783277e-02
-8.84457886e-01 8.68894279e-01 6.76109612e-01 -2.37110749e-01
-2.87839711e-01 -7.57314801e-01 2.13378012e-01 3.10353730e-02
6.06493771e-01 2.12474503e-02 7.87476748e-02 -6.75849438e-01
-5.15796483e-01 8.35570514e-01 1.52141333e-01 -5.12005948e-02
1.62138581e+00 -2.72943974e-01 -2.14395776e-01 2.77025193e-01
1.24746120e+00 -2.31495902e-01 -1.64627576e+00 -8.17427635e-01
-3.05334151e-01 -6.72984302e-01 5.27736187e-01 -7.50420928e-01
-1.17605817e+00 1.03664935e+00 8.28930438e-01 -2.30796020e-02
1.48005414e+00 1.82966664e-01 7.34341323e-01 3.75609189e-01
3.15076619e-01 -8.62971127e-01 -7.50168711e-02 7.94587970e-01
9.10150588e-01 -1.26331544e+00 -3.21987793e-02 -3.51632565e-01
-3.33111554e-01 9.34685469e-01 3.13832879e-01 -8.60521868e-02
8.42660606e-01 -1.21299168e-02 -2.05538534e-02 -3.96279901e-01
-6.63407445e-01 -7.73290634e-01 2.21280858e-01 5.55809319e-01
2.24248394e-01 1.76835567e-01 -7.63901696e-02 2.39913002e-01
8.55738074e-02 -6.10786900e-02 2.09282160e-01 9.02772367e-01
-6.77957058e-01 -8.04040730e-01 -1.35358632e-01 2.04953149e-01
-2.36241847e-01 -6.42396033e-01 -8.25289860e-02 4.70462710e-01
4.47402447e-01 7.92299747e-01 2.96230763e-01 -4.17339712e-01
1.08392559e-01 -2.06828311e-01 2.08820850e-01 -6.02310359e-01
-4.93495643e-01 1.21434808e-01 -9.64660421e-02 -8.38508129e-01
-9.23038960e-01 -8.63031447e-01 -1.25106442e+00 -3.24940175e-01
-1.00664817e-01 -4.20081466e-01 5.72770298e-01 9.03633356e-01
5.44791996e-01 4.63429868e-01 7.86486685e-01 -8.55129182e-01
-2.27472082e-01 -9.49417233e-01 -5.62275171e-01 2.42357463e-01
6.26956046e-01 -6.03251934e-01 -1.62960052e-01 1.27496645e-01]
|
[9.677633285522461, -1.3343946933746338]
|
569db09b-db76-4cfa-a48c-0f40eaf0d3c1
|
multilingual-offensive-lexicon-annotated-with
| null | null |
https://openreview.net/forum?id=WCBAn7V584l
|
https://openreview.net/pdf?id=WCBAn7V584l
|
Multilingual offensive lexicon annotated with contextual information
|
Online hate speech and offensive comments detection is not a trivial research problem since pragmatic (contextual) factors influence what is considered offensive. Moreover, offensive terms are hardly found in classical lexical resources such as wordnets, sentiment, and emotion lexicons. In this paper, we embrace the challenges and opportunities of the area and introduce the first multilingual offensive lexicon (MOL), which is composed of 1,000 explicit and implicit pejorative terms and expressions annotated with contextual information. The terms and expressions were manually extracted by a specialist from Instagram abusive comments originally written in Portuguese and manually translated by American English, Latin American Spanish, African French, and German native speakers. Each expression was annotated by three different annotators, producing high human inter-annotator agreement. Accordingly, this resource provides a new perspective to explore abusive language detection.
|
['Anonymous']
|
2021-11-16
| null | null | null |
acl-arr-november-2021-11
|
['abusive-language']
|
['natural-language-processing']
|
[ 7.38107637e-02 1.77805915e-01 -5.54838181e-01 9.03189834e-03
-5.15079558e-01 -1.14954066e+00 5.90262592e-01 7.18844295e-01
-5.94363868e-01 8.93249154e-01 5.35063624e-01 3.92764546e-02
4.51222777e-01 -2.57266313e-01 1.49626955e-01 -3.09594482e-01
3.90868276e-01 8.97178501e-02 -2.06783995e-01 -5.71904480e-01
5.60100615e-01 3.17549586e-01 -9.95438993e-01 3.20165277e-01
1.17897189e+00 6.23152971e-01 -3.03104043e-01 2.29516968e-01
-3.25521082e-01 1.28222454e+00 -1.07234728e+00 -1.29975033e+00
-1.55859321e-01 -4.87283677e-01 -8.30688298e-01 -1.25447080e-01
2.10867509e-01 -7.04173520e-02 1.42457739e-01 1.54980993e+00
4.85129356e-01 -5.26039563e-02 4.91528809e-01 -8.05314302e-01
-9.12814200e-01 7.77693808e-01 -5.43301582e-01 3.53985518e-01
5.47504067e-01 1.01279110e-01 1.17517734e+00 -9.08603489e-01
1.15904224e+00 1.12401044e+00 4.85864639e-01 5.88316500e-01
-9.10378993e-01 -7.70652831e-01 3.56867574e-02 2.42435887e-01
-1.32813454e+00 -2.09106311e-01 9.94709253e-01 -8.85377884e-01
8.50474119e-01 2.87699223e-01 8.32860887e-01 1.79271770e+00
-1.93016723e-01 4.70607400e-01 1.19803607e+00 -5.56028247e-01
4.10516113e-02 9.20129240e-01 2.32802972e-01 3.41435194e-01
2.40365595e-01 -5.45878172e-01 -8.07345390e-01 -6.69244528e-01
-9.21954438e-02 -4.38348800e-01 -3.65851611e-01 3.30302507e-01
-8.02281976e-01 1.31495655e+00 6.16972670e-02 9.29564536e-01
-2.41677850e-01 -4.46085662e-01 1.05619359e+00 5.28601110e-01
9.42833304e-01 9.37832117e-01 -1.89340681e-01 -5.61917067e-01
-5.68691730e-01 -6.40807254e-03 1.12533998e+00 5.43483555e-01
3.66477519e-01 -3.28672379e-02 2.19160944e-01 1.32522821e+00
1.37260482e-01 6.32113397e-01 6.12684488e-01 -3.43760848e-01
1.93679556e-01 6.89810991e-01 1.40489474e-01 -1.75473928e+00
-2.73964822e-01 -9.88283455e-02 -2.51609206e-01 -2.53880382e-01
1.42289937e-01 -4.83695418e-01 -1.90848574e-01 1.51774931e+00
3.13395530e-01 -6.76356375e-01 -1.20910309e-01 9.80482817e-01
7.86581576e-01 5.71440160e-01 5.86682200e-01 -6.85329974e-01
1.62921107e+00 -6.91960216e-01 -1.23784053e+00 -3.20593536e-01
8.48663926e-01 -1.27633166e+00 1.14245689e+00 5.80462992e-01
-8.83702338e-01 1.10740475e-01 -1.04947281e+00 -2.14298666e-01
-9.36800539e-01 -1.85233012e-01 2.29845390e-01 9.17951584e-01
-4.01457191e-01 1.94081053e-01 -8.04177448e-02 -5.21564662e-01
3.16025317e-01 -3.12119275e-01 -4.99886304e-01 4.81161088e-01
-1.63275480e+00 1.48113942e+00 9.05751511e-02 -1.61274597e-01
-3.74027252e-01 -5.28460562e-01 -8.73877108e-01 -6.57548189e-01
5.77004135e-01 2.75669694e-01 1.10894728e+00 -1.62416387e+00
-1.57451248e+00 1.76482248e+00 5.87141067e-02 -5.62538169e-02
3.17159772e-01 -6.40212476e-01 -1.06479967e+00 2.21665278e-01
3.08872640e-01 -1.62220880e-01 9.40581262e-01 -8.87692451e-01
-8.01926181e-02 -3.77836585e-01 1.02849096e-01 1.49655402e-01
-8.06535363e-01 1.11747026e+00 3.77141356e-01 -1.04543722e+00
-5.45145094e-01 -6.49361968e-01 2.70646214e-01 -2.57598609e-01
-6.07058346e-01 -2.47099921e-01 6.71624124e-01 -1.03301966e+00
1.68942797e+00 -2.36701465e+00 1.83575928e-01 7.77995959e-02
4.34482485e-01 5.39891064e-01 4.47255433e-01 7.88436830e-01
6.66321367e-02 5.09649098e-01 -1.16741017e-01 6.94619343e-02
3.02774727e-01 1.54987136e-02 -3.66299719e-01 7.39654183e-01
-2.43979748e-02 6.62793338e-01 -1.22828090e+00 -6.55308366e-01
-5.39471954e-03 4.75717723e-01 -2.53204346e-01 3.48077640e-02
5.37694581e-02 3.38155091e-01 -4.96271521e-01 8.22817206e-01
2.10519850e-01 1.61079824e-01 3.50599319e-01 2.88340691e-02
-4.74276364e-01 5.63864052e-01 -4.76780713e-01 9.35522616e-01
-5.55665195e-01 9.51487601e-01 3.87711972e-01 -2.83694237e-01
1.14294064e+00 3.80314916e-01 -2.45042127e-02 -4.83689487e-01
6.89259768e-01 7.05043375e-01 -3.47811123e-03 -7.10871160e-01
7.05347061e-01 -5.85604250e-01 -5.30839503e-01 3.59823257e-01
-2.31713578e-02 -2.00833663e-01 1.87981412e-01 2.16358185e-01
7.29586422e-01 -1.88924775e-01 1.01580739e+00 -3.24011713e-01
8.36761236e-01 2.97594577e-01 4.19315249e-01 1.73225939e-01
-7.39021897e-01 -3.18368711e-02 8.96490335e-01 -4.98994052e-01
-8.00648630e-01 -4.97152925e-01 -3.81723851e-01 1.46148479e+00
-5.65900542e-02 -7.47626126e-01 -7.68066943e-01 -7.02319384e-01
-3.30814987e-01 8.95041883e-01 -6.85877383e-01 1.67425945e-02
-3.27796012e-01 -5.42057395e-01 6.67321205e-01 -1.82711557e-01
1.56055212e-01 -1.26730013e+00 -6.17310107e-01 3.12818825e-01
-3.69668186e-01 -1.28910649e+00 -2.72161663e-01 7.33942091e-02
1.52510628e-01 -1.12418234e+00 -3.39446962e-01 -3.98047000e-01
1.04331039e-01 -1.62302300e-01 9.83851671e-01 9.77729037e-02
-3.27191919e-01 -3.74269515e-01 -8.39297056e-01 -7.26498365e-01
-7.65862405e-01 2.45983824e-01 9.85732377e-02 4.77240235e-02
1.02611923e+00 -3.72903228e-01 4.55675237e-02 -1.80244640e-01
-9.53969896e-01 -4.73550022e-01 5.04341125e-02 4.49194223e-01
1.26909167e-02 -7.52545416e-01 5.19759536e-01 -1.28989899e+00
1.02075660e+00 -7.28115439e-01 -5.06385341e-02 -1.65310726e-01
-1.46867976e-01 -6.30961120e-01 8.50726366e-01 -5.52040339e-01
-9.61962759e-01 -5.86088896e-01 -2.88467020e-01 -7.08599389e-02
-2.24244609e-01 4.46344465e-01 1.57987494e-02 3.85487042e-02
1.05758929e+00 -4.33274448e-01 -2.10702956e-01 -4.54026133e-01
3.75774384e-01 1.01119137e+00 2.24170834e-01 -4.27513629e-01
7.72598207e-01 1.92245573e-01 -7.13449419e-01 -1.15216362e+00
-1.53297603e+00 -6.88842416e-01 -8.31097662e-01 -4.93912965e-01
1.05082774e+00 -7.82114744e-01 -6.06461346e-01 3.98586869e-01
-1.32045567e+00 2.86538582e-02 -9.71970484e-02 1.88739777e-01
8.49799067e-02 3.52232724e-01 -9.10948694e-01 -7.48115182e-01
-6.14760935e-01 -7.56550968e-01 6.08357787e-01 3.71568948e-02
-1.09234166e+00 -1.11164272e+00 4.68550652e-01 3.69496524e-01
3.42326880e-01 7.16168702e-01 8.12814176e-01 -1.13550878e+00
9.32055533e-01 -3.72532099e-01 -1.09819651e-01 3.51489365e-01
8.49396512e-02 2.62826234e-01 -8.77897799e-01 1.71184912e-01
1.54090777e-01 -9.75425899e-01 2.04767600e-01 -6.40132785e-01
3.73797446e-01 -8.73276114e-01 -3.61617357e-02 2.64895614e-02
1.29979146e+00 3.95212471e-02 3.27686399e-01 5.88785708e-01
6.36508524e-01 9.38219309e-01 4.56814945e-01 8.39361548e-01
9.24261287e-02 5.33873320e-01 1.76877245e-01 2.42542833e-01
3.93873125e-01 -2.77987629e-01 8.23058724e-01 1.17633617e+00
7.49253556e-02 -4.81423140e-02 -9.35712576e-01 6.35497808e-01
-1.29504085e+00 -8.45957220e-01 -4.26080197e-01 1.42316580e+00
1.16872907e+00 -4.60749529e-02 2.45369896e-01 -3.42536718e-02
8.28071952e-01 6.29626334e-01 -4.51604323e-03 -1.18896639e+00
-3.66268069e-01 2.16457590e-01 1.79844514e-01 6.78286791e-01
-1.15123308e+00 1.34977055e+00 5.64197826e+00 9.49935734e-01
-1.20903695e+00 5.10151148e-01 3.65219474e-01 -2.28902385e-01
-2.10220948e-01 -2.35856757e-01 -5.81159294e-01 4.84602481e-01
8.70103121e-01 -3.57857794e-01 2.82393754e-01 1.21040833e+00
1.04455158e-01 1.06658436e-01 -5.19600868e-01 7.43593216e-01
6.97796822e-01 -9.13517594e-01 -3.22785676e-01 4.96499659e-03
6.54794931e-01 8.38012435e-03 -5.80911078e-02 3.01519424e-01
2.88736731e-01 -1.00858128e+00 9.03758287e-01 5.47151663e-04
7.05263436e-01 -7.71660388e-01 1.01884341e+00 2.01594219e-01
-3.65316570e-01 2.03980461e-01 -1.80149972e-01 -1.96417078e-01
3.70041251e-01 5.93411207e-01 -4.12814826e-01 -3.24561186e-02
5.97631991e-01 7.96743035e-01 -6.13886833e-01 1.12173676e-01
-8.80337179e-01 6.68841124e-01 -2.33041421e-02 -6.24894738e-01
5.22188902e-01 -3.20608795e-01 9.05093789e-01 1.75346446e+00
-2.58114845e-01 1.76856533e-01 1.92411959e-01 6.10767424e-01
-1.64802954e-01 9.61076319e-01 -8.26359510e-01 -6.55628860e-01
3.41093153e-01 1.77651942e+00 -5.61485052e-01 -3.04699212e-01
-4.51508611e-01 9.73137140e-01 7.71196187e-01 -3.51348072e-02
-7.40983069e-01 -4.77302104e-01 7.82066107e-01 1.83135942e-01
-3.67053509e-01 1.67658031e-01 -1.60995692e-01 -1.20126200e+00
-1.93168268e-01 -1.26640630e+00 3.96010131e-01 -4.58361715e-01
-1.88879633e+00 9.43022907e-01 -1.42588511e-01 -8.59909773e-01
-2.17538118e-01 -8.16472113e-01 -1.67322353e-01 6.11181080e-01
-1.14500022e+00 -8.89533460e-01 -6.86812624e-02 3.07351589e-01
2.22945556e-01 6.65699169e-02 9.97932434e-01 4.14782137e-01
-7.23745704e-01 3.77598166e-01 -4.01247531e-01 2.88919419e-01
1.01150858e+00 -8.86761427e-01 -2.29439273e-01 5.39738357e-01
-1.22729294e-01 6.53704703e-01 1.07094550e+00 -6.92046523e-01
-5.41251242e-01 -6.29789829e-01 1.46448219e+00 -5.97178698e-01
1.78272974e+00 -3.19308400e-01 -1.04622793e+00 5.71618617e-01
6.93135917e-01 -2.86734611e-01 1.52014887e+00 9.70904380e-02
-8.41415584e-01 7.46287107e-01 -1.20512629e+00 7.95734763e-01
9.44930136e-01 -8.43119383e-01 -8.15612614e-01 7.25690782e-01
5.11964858e-01 -3.69483471e-01 -8.70841980e-01 -7.75768906e-02
3.46262902e-01 -6.92175388e-01 3.17365140e-01 -8.32512796e-01
7.32996881e-01 1.61783591e-01 -1.09777696e-01 -1.29924321e+00
8.67877807e-03 -8.86294603e-01 7.88837969e-02 1.32329547e+00
4.69111532e-01 -7.06725359e-01 -5.87901250e-02 4.28119391e-01
-1.31579265e-01 -4.15386707e-01 -9.11608875e-01 -2.43955910e-01
3.51702183e-01 -4.46514577e-01 3.69444564e-02 1.72879326e+00
9.83363748e-01 7.31684923e-01 -5.21269977e-01 -5.61511755e-01
1.12005025e-01 -3.00639197e-02 5.20342648e-01 -1.24883187e+00
1.03111058e-01 -7.28986561e-01 -3.91444981e-01 -3.56373876e-01
6.00452065e-01 -8.67443860e-01 -2.34244093e-01 -7.57840157e-01
3.08841497e-01 9.27080810e-02 3.40229243e-01 5.15389502e-01
-1.88703671e-01 7.00330317e-01 2.87906826e-01 2.92142838e-01
-5.48598647e-01 3.48260015e-01 9.73869920e-01 1.06340714e-01
-2.49382347e-01 -6.92633390e-01 -9.80281115e-01 1.40379226e+00
8.97702575e-01 -7.53251255e-01 1.78507820e-01 9.85384881e-02
8.31291974e-01 -6.05039656e-01 8.64202306e-02 -2.31972083e-01
-2.26934522e-01 -4.63000774e-01 -1.74910650e-01 -8.39513540e-03
3.87192041e-01 -5.41439056e-01 -1.49491683e-01 3.31026763e-01
-4.29689109e-01 1.11152932e-01 1.02688961e-01 1.06505029e-01
-4.48026836e-01 -7.75542796e-01 1.11211860e+00 -3.51946414e-01
-5.17658830e-01 -2.74724275e-01 -8.49197626e-01 6.91141546e-01
1.29198980e+00 1.26824072e-02 -6.98627889e-01 -5.08823156e-01
-6.58658922e-01 -4.13502872e-01 7.45528281e-01 4.28558618e-01
2.30303869e-01 -9.53907609e-01 -8.43624473e-01 -3.72187525e-01
5.41421354e-01 -9.51415777e-01 2.46935487e-02 9.51841414e-01
-6.14910960e-01 7.25074187e-02 -1.96851388e-01 2.76204228e-01
-1.15327787e+00 6.89260721e-01 -1.66304857e-02 -1.54688463e-01
-4.69772220e-01 6.93981767e-01 -2.24121511e-01 -3.72235119e-01
-2.80596048e-01 4.99362558e-01 -6.34771287e-01 7.80809581e-01
7.21310258e-01 2.10772321e-01 -2.29474291e-01 -1.65575516e+00
-4.74806577e-01 2.56993860e-01 -1.66148379e-01 -2.18164742e-01
9.89449859e-01 -9.93118212e-02 -7.63588727e-01 9.00493979e-01
1.38449466e+00 9.26956356e-01 -2.25473046e-02 -9.12980586e-02
3.45819026e-01 -7.02053487e-01 -3.25136960e-01 -8.26390088e-01
-6.01818025e-01 6.09754980e-01 -2.91207284e-01 7.28892088e-01
5.63032091e-01 1.55574039e-01 8.19254160e-01 7.28508607e-02
1.62005484e-01 -1.54103363e+00 2.79043794e-01 7.77275980e-01
1.08310819e+00 -9.78833973e-01 -1.00250334e-01 -7.62760878e-01
-1.10682750e+00 1.04025292e+00 7.27582932e-01 -1.19682200e-01
6.37428105e-01 2.13099092e-01 6.52684629e-01 -6.10472739e-01
-5.77888131e-01 -1.17457062e-01 2.21603051e-01 4.09853101e-01
1.03926361e+00 1.09609127e-01 -1.30124938e+00 8.08822632e-01
-6.17678046e-01 -6.05719328e-01 6.32938862e-01 7.16267765e-01
-3.07441711e-01 -9.23330307e-01 -2.46485934e-01 1.01212233e-01
-1.28591120e+00 -2.77789086e-01 -1.46298933e+00 8.93215001e-01
2.18786582e-01 9.82380509e-01 -2.26723194e-01 -3.39803129e-01
1.83942273e-01 1.64608449e-01 -4.02530171e-02 -7.61843920e-01
-1.36001670e+00 -1.65669676e-02 7.20749676e-01 -2.30207905e-01
-5.42548835e-01 -5.85253179e-01 -8.52648675e-01 -4.47580516e-01
-3.08900923e-01 4.01787311e-01 5.65605700e-01 1.07482958e+00
7.89608881e-02 -3.17424275e-02 5.61367810e-01 -3.17408472e-01
-2.03251317e-01 -1.25578797e+00 -5.08017480e-01 6.26845360e-01
7.35879838e-02 -5.47060728e-01 -7.51870513e-01 -1.21079557e-01]
|
[8.777693748474121, 10.5396728515625]
|
b9e32a5a-7e20-4275-8259-6b8edd385e20
|
reinforcement-learning-with-imbalanced
| null | null |
https://aclanthology.org/2020.findings-emnlp.202
|
https://aclanthology.org/2020.findings-emnlp.202.pdf
|
Reinforcement Learning with Imbalanced Dataset for Data-to-Text Medical Report Generation
|
Automated generation of medical reports that describe the findings in the medical images helps radiologists by alleviating their workload. Medical report generation system should generate correct and concise reports. However, data imbalance makes it difficult to train models accurately. Medical datasets are commonly imbalanced in their finding labels because incidence rates differ among diseases; moreover, the ratios of abnormalities to normalities are significantly imbalanced. We propose a novel reinforcement learning method with a reconstructor to improve the clinical correctness of generated reports to train the data-to-text module with a highly imbalanced dataset. Moreover, we introduce a novel data augmentation strategy for reinforcement learning to additionally train the model on infrequent findings. From the perspective of a practical use, we employ a Two-Stage Medical Report Generator (TS-MRGen) for controllable report generation from input images. TS-MRGen consists of two separated stages: an image diagnosis module and a data-to-text module. Radiologists can modify the image diagnosis module results to control the reports that the data-to-text module generates. We conduct an experiment with two medical datasets to assess the data-to-text module and the entire two-stage model. Results demonstrate that the reports generated by our model describe the findings in the input image more correctly.
|
['Keigo Nakamura', 'Tomoko Ohkuma', 'Motoki Taniguchi', 'Yuki Tagawa', 'Norihisa Nakano', 'Ryuji Kano', 'Tomoki Taniguchi', 'Yohei Momoki', 'Ryota Ozaki', 'Toru Nishino']
|
2020-11-01
| null | null | null |
findings-of-the-association-for-computational
|
['medical-report-generation']
|
['medical']
|
[ 3.09382200e-01 4.58363175e-01 -2.47710064e-01 -7.82537162e-01
-1.12458718e+00 -2.41753280e-01 8.50436091e-02 4.26281333e-01
-1.59480393e-01 8.13061714e-01 9.90925133e-02 -4.41446453e-01
9.90642142e-03 -9.69822526e-01 -6.29699767e-01 -4.56538558e-01
1.76528007e-01 6.93020701e-01 -5.47688454e-02 2.22622350e-01
2.81282634e-01 3.97857651e-02 -1.38934529e+00 7.03360438e-01
1.11067915e+00 7.74549186e-01 2.84448147e-01 1.11171269e+00
1.68433283e-02 1.41351259e+00 -1.03580761e+00 -4.05974895e-01
5.21838069e-02 -8.88528347e-01 -5.71328998e-01 5.84691823e-01
1.69850856e-01 -4.78100926e-01 -1.58122301e-01 7.78530002e-01
6.99591935e-01 -2.68399149e-01 7.08501518e-01 -1.27058625e+00
-7.35297799e-01 6.20451331e-01 -7.90911436e-01 3.57524484e-01
3.52246404e-01 4.64118481e-01 5.26136458e-01 -4.53632385e-01
8.37710440e-01 8.15622509e-01 3.37177575e-01 6.75794840e-01
-8.23325276e-01 -6.96819067e-01 1.11934043e-01 6.37004822e-02
-1.08484519e+00 -1.65377691e-01 6.11075282e-01 -5.86009681e-01
6.41528726e-01 5.02297342e-01 5.84123492e-01 6.85119271e-01
5.12982011e-01 6.51929975e-01 9.63268161e-01 -1.41307727e-01
2.89664716e-01 2.06709176e-01 -1.39943853e-01 9.03701365e-01
5.60192585e-01 -5.91759346e-02 -3.65500838e-01 -9.89919528e-02
8.63123953e-01 1.14203997e-01 -2.02341959e-01 7.43003935e-02
-1.34253359e+00 9.07139301e-01 4.48386222e-01 9.50090364e-02
-6.81105316e-01 -5.75943366e-02 4.59236979e-01 2.57343829e-01
3.58628571e-01 5.04047394e-01 -1.00590855e-01 3.24241906e-01
-9.38515007e-01 4.83145505e-01 6.15558147e-01 9.75545287e-01
2.27734998e-01 -1.16538592e-01 -6.93848372e-01 7.69040108e-01
9.09336731e-02 5.51951051e-01 1.05730402e+00 -5.66379607e-01
8.54659498e-01 8.63414764e-01 9.78331268e-02 -1.14900482e+00
-6.21599555e-01 -7.14703798e-01 -1.10803783e+00 -3.96149568e-02
-1.57040641e-01 -2.77677208e-01 -1.06910312e+00 1.36985111e+00
4.40207988e-01 -2.15198129e-01 1.97694808e-01 1.10965610e+00
1.15145063e+00 5.21789432e-01 1.32968098e-01 -5.74981511e-01
1.45108557e+00 -9.25862551e-01 -1.01082385e+00 -4.69896160e-02
1.11157417e+00 -6.70557916e-01 8.44199598e-01 2.69270718e-01
-1.38645685e+00 -7.00973928e-01 -1.07396543e+00 1.98731914e-01
4.54344228e-02 7.85752356e-01 5.03149867e-01 2.01260298e-01
-7.70615518e-01 1.31168351e-01 -7.91897833e-01 8.02043602e-02
4.90781248e-01 3.33480775e-01 -3.11726272e-01 -5.53983748e-02
-9.87695217e-01 8.95076036e-01 6.22213960e-01 -7.27398396e-02
-7.62341440e-01 -7.82289147e-01 -1.07735717e+00 -1.52477518e-01
4.32840228e-01 -1.10103607e+00 1.64839864e+00 -7.86522150e-01
-1.03221571e+00 1.16031063e+00 1.01058915e-01 -3.49219024e-01
7.13429689e-01 4.61432040e-01 -4.92782712e-01 3.24547470e-01
4.49997008e-01 7.61003375e-01 6.98366284e-01 -1.22851276e+00
-8.80144179e-01 -1.60201237e-01 -9.99043733e-02 1.81550100e-01
2.53533483e-01 -4.30727392e-01 -2.92815149e-01 -1.02963293e+00
1.73806921e-01 -8.25893283e-01 -5.39716303e-01 -1.96353689e-01
-5.67738652e-01 -1.06172964e-01 3.47058654e-01 -6.44653320e-01
1.40894938e+00 -1.95423114e+00 -4.39482838e-01 1.26604900e-01
4.57871497e-01 2.37025544e-02 3.78691480e-02 -6.50181994e-03
-5.30374169e-01 2.39594169e-02 -3.72958809e-01 -2.92634070e-01
-4.42425698e-01 2.08044440e-01 -9.60797593e-02 1.31219164e-01
7.99484909e-01 7.23094642e-01 -9.72738028e-01 -8.94544244e-01
-1.96270645e-03 -8.90050828e-02 -7.61200011e-01 9.21332419e-01
-7.59294778e-02 5.62644124e-01 -5.67377746e-01 5.28029263e-01
6.75801098e-01 -6.64798081e-01 2.71786153e-01 -3.10373098e-01
1.33777857e-01 3.45611513e-01 -1.00547194e+00 1.36562610e+00
-5.83312869e-01 -1.46937743e-02 -4.97922271e-01 -9.44535375e-01
9.89424706e-01 5.71677804e-01 5.36842465e-01 -8.27040851e-01
2.68023200e-02 1.42448336e-01 4.62289363e-01 -1.09309852e+00
4.74417478e-01 -4.05172765e-01 -8.58680531e-02 9.03277755e-01
-8.10851231e-02 -4.82103020e-01 5.43471694e-01 4.10237372e-01
1.27513707e+00 -2.85450369e-01 4.11083102e-01 4.48531866e-01
5.40396452e-01 2.68295288e-01 6.67265594e-01 7.18826532e-01
-1.13672882e-01 9.15395617e-01 6.85061276e-01 -5.32499790e-01
-1.02150321e+00 -8.81582439e-01 6.24347217e-02 5.40810227e-01
-1.35520279e-01 -2.14367837e-01 -6.03004396e-01 -1.14714825e+00
-1.98075041e-01 5.95075846e-01 -7.96649814e-01 -4.23204184e-01
-3.86234760e-01 -1.01234591e+00 1.65126055e-01 6.41872466e-01
4.14555073e-01 -1.39287853e+00 -8.64347398e-01 2.91504323e-01
-4.30922538e-01 -1.08531559e+00 -6.81171656e-01 8.29936936e-02
-9.24753129e-01 -1.20744812e+00 -6.73115373e-01 -7.68128872e-01
1.39903271e+00 -1.07777685e-01 1.35345197e+00 4.38482374e-01
-4.31211472e-01 -3.86951561e-03 -5.59719503e-01 -8.19194794e-01
-1.03952038e+00 1.71966165e-01 -2.74479628e-01 -1.92882046e-01
2.85781622e-02 -9.03866142e-02 -7.96672523e-01 1.33877531e-01
-1.32563448e+00 5.42759359e-01 9.75222230e-01 8.75581026e-01
8.49701762e-01 4.07801755e-02 8.68007779e-01 -1.32506049e+00
6.97147608e-01 -5.23263752e-01 -2.94791430e-01 2.14015767e-01
-6.16650581e-01 2.85581976e-01 6.11935198e-01 -2.50630200e-01
-1.03291357e+00 1.51687652e-01 -2.28058100e-01 -2.37543583e-01
-9.44184735e-02 5.23299098e-01 2.06744224e-01 6.30884230e-01
6.00129008e-01 3.64905208e-01 4.94299419e-02 2.47424886e-01
-1.34211540e-01 7.65525281e-01 6.29765809e-01 -3.38783860e-02
6.17774904e-01 1.89447209e-01 -2.21185833e-01 1.58279330e-01
-1.20853686e+00 -2.90123045e-01 -2.39569440e-01 -2.19788879e-01
6.88843489e-01 -8.64494145e-01 -6.37518227e-01 1.46832494e-02
-1.08369398e+00 4.13897745e-02 -4.63374585e-01 5.90233147e-01
-7.62252629e-01 -2.53690451e-01 -5.47302008e-01 -5.31863987e-01
-7.37187207e-01 -1.45098150e+00 9.75177824e-01 2.55535573e-01
-4.03571039e-01 -6.97234750e-01 1.27377629e-01 3.59009176e-01
-2.04766802e-02 4.33842093e-01 1.03266680e+00 -6.05414927e-01
-4.14566517e-01 -6.03252836e-02 -2.31805459e-01 1.56698450e-01
4.29598302e-01 1.13925776e-02 -7.27189720e-01 -7.00612292e-02
9.01506990e-02 -2.69706339e-01 7.03487039e-01 4.78324741e-01
1.60264158e+00 -5.14313698e-01 -1.97416097e-01 3.53150249e-01
1.21191740e+00 5.00920355e-01 4.41245347e-01 7.08205476e-02
5.16143322e-01 6.00707650e-01 8.96465540e-01 4.66549188e-01
7.30947137e-01 2.50672936e-01 4.02660549e-01 -4.35407490e-01
-2.24320993e-01 -2.70273447e-01 -5.31668775e-02 1.08878756e+00
1.90568045e-01 -1.32787719e-01 -7.05030859e-01 6.79122388e-01
-1.74929452e+00 -8.10767233e-01 4.19790894e-02 1.86647367e+00
1.45657313e+00 2.92459667e-01 6.59905374e-02 2.62876183e-01
5.81017792e-01 -1.81445301e-01 -5.62361181e-01 -4.88039702e-01
4.53057200e-01 1.80442035e-02 2.76514471e-01 2.09002197e-01
-8.25012326e-01 3.13241720e-01 6.10633230e+00 3.53451282e-01
-1.30450702e+00 3.07811052e-02 1.19971681e+00 -1.43727511e-01
-4.40432400e-01 -6.36750102e-01 -6.07998610e-01 6.28194749e-01
7.98952699e-01 -4.51400653e-02 -4.70811278e-01 8.35069835e-01
4.90584463e-01 -3.65624160e-01 -1.27721846e+00 1.05760825e+00
1.38061225e-01 -1.60199642e+00 5.71975827e-01 -2.04418316e-01
8.20370555e-01 -7.18063653e-01 3.70807871e-02 2.60979474e-01
2.65431821e-01 -8.60096574e-01 5.13939619e-01 7.71000326e-01
1.10066271e+00 -8.96739483e-01 1.04211855e+00 2.40705192e-01
-6.83155060e-01 7.20161898e-03 -2.64145374e-01 2.60673463e-01
-2.97253509e-03 8.96109998e-01 -1.75225437e+00 7.65410364e-01
2.53626138e-01 5.14012635e-01 -6.08638823e-01 8.57627690e-01
-2.14025632e-01 4.62895215e-01 2.80976474e-01 1.10885285e-01
-5.47837466e-02 1.98948398e-01 1.26577869e-01 1.22239494e+00
2.94140607e-01 3.57379049e-01 2.69882172e-01 9.74237800e-01
-3.62461329e-01 1.90944090e-01 -4.84843612e-01 -4.25333604e-02
1.87439039e-01 1.40096378e+00 -8.56219530e-01 -6.96557403e-01
-1.09605193e-01 5.77450633e-01 1.78954199e-01 -1.30189136e-01
-9.08588886e-01 -2.89374560e-01 4.49798554e-02 4.20248210e-01
-3.93143781e-02 4.15027022e-01 -5.92449188e-01 -8.23728204e-01
1.98394611e-01 -1.05048859e+00 6.39188528e-01 -9.97101665e-01
-1.16359711e+00 8.39880705e-01 -3.45084637e-01 -1.60312009e+00
-6.72677755e-01 -2.43694082e-01 -5.43022037e-01 5.90444982e-01
-1.54387629e+00 -8.29959214e-01 -6.10009611e-01 3.43591213e-01
6.63501620e-01 -1.88564524e-01 7.84573197e-01 2.59249300e-01
-5.05924940e-01 7.73878932e-01 -6.75795913e-01 4.12691921e-01
8.14253449e-01 -1.36414135e+00 1.87633395e-01 5.35559416e-01
-3.48897755e-01 2.85791606e-01 4.51849729e-01 -8.27336311e-01
-8.15569282e-01 -1.44079769e+00 6.49189770e-01 -2.29277194e-01
9.39141661e-02 1.62651658e-01 -8.02524865e-01 5.53360283e-01
3.72715248e-03 7.55276307e-02 8.99806559e-01 -5.59266865e-01
1.07591756e-01 -2.12825820e-01 -1.39634573e+00 4.56975400e-01
5.63166082e-01 -4.39387597e-02 -5.78430891e-01 4.49799120e-01
9.61209893e-01 -9.77308691e-01 -9.16627645e-01 4.96125519e-01
3.00332665e-01 -7.02056706e-01 4.85491484e-01 -5.89732051e-01
1.19213331e+00 -4.23207581e-01 3.65007937e-01 -1.62790835e+00
-9.30093974e-02 -4.35402840e-02 3.59616987e-02 8.83895755e-01
6.20980203e-01 -3.19329411e-01 6.48656607e-01 2.61068314e-01
-1.94792584e-01 -1.10587454e+00 -4.14257854e-01 5.06101176e-02
-4.16302323e-01 -2.88566083e-01 7.64835000e-01 1.00868118e+00
-7.30099156e-02 2.35041708e-01 2.36495398e-03 1.93693370e-01
2.28372842e-01 2.05098063e-01 5.44415474e-01 -7.18688726e-01
-3.33068013e-01 -1.46581396e-01 -2.00462878e-01 -5.11960685e-01
-3.81975025e-01 -8.60805273e-01 2.33249113e-01 -1.89169502e+00
5.07656813e-01 -6.26192570e-01 -1.32412627e-01 5.52101493e-01
-7.76347458e-01 2.79961020e-01 5.15551493e-03 1.37679070e-01
-6.10465288e-01 2.00223088e-01 1.89971101e+00 -3.72040689e-01
-2.59285599e-01 3.25832486e-01 -9.61068332e-01 5.50159097e-01
6.66074634e-01 -9.27373648e-01 -5.22407293e-01 -2.26842374e-01
5.15490294e-01 5.66158891e-01 9.24803764e-02 -1.02306259e+00
9.30246115e-02 -1.56223461e-01 6.80333078e-01 -1.00508654e+00
-3.41356605e-01 -5.59843183e-01 -1.62179485e-01 8.56250465e-01
-6.03578329e-01 5.15984178e-01 1.35640457e-01 1.24280050e-01
-4.52679336e-01 -1.99359119e-01 6.70567811e-01 -5.35266399e-01
-4.92775701e-02 3.12845796e-01 -3.68864298e-01 3.89149413e-02
1.19809568e+00 2.14306358e-02 -2.88964629e-01 -4.97405440e-01
-8.29013705e-01 4.78787512e-01 -1.93787947e-01 4.01241332e-01
9.07596231e-01 -1.30562484e+00 -1.17114496e+00 2.77745485e-01
4.24405783e-01 5.17796218e-01 5.68184018e-01 9.04591203e-01
-7.68490136e-01 2.72646636e-01 -1.05880521e-01 -7.33799994e-01
-1.09114504e+00 6.01103902e-01 4.78550196e-01 -1.04878080e+00
-3.15364987e-01 6.12828016e-01 3.46221954e-01 -5.86514294e-01
-4.20017308e-03 -7.48150706e-01 -2.17267513e-01 7.04756826e-02
8.77036572e-01 -9.81602222e-02 4.52209949e-01 -1.20257072e-01
-1.05744883e-01 1.08112276e-01 -6.32451534e-01 3.29012014e-02
1.34051800e+00 1.37236372e-01 1.85250193e-02 2.48958409e-01
7.63628066e-01 -2.36159340e-01 -7.92888582e-01 5.70174381e-02
-3.35696965e-01 -1.38449803e-01 -1.22349337e-01 -1.11414587e+00
-1.33332694e+00 5.60660601e-01 5.53647637e-01 2.76862681e-01
1.37683344e+00 -4.65038568e-02 6.91960156e-01 4.63225245e-02
-5.91119155e-02 -1.03764451e+00 5.23432910e-01 -1.06161267e-01
1.14654732e+00 -1.40225196e+00 6.42096773e-02 -4.31752861e-01
-1.11533034e+00 9.85631347e-01 1.03076696e+00 -9.08006579e-02
3.02838534e-01 5.85085928e-01 4.25320715e-01 -1.77077368e-01
-1.02677584e+00 5.96354045e-02 2.02051625e-01 6.02740884e-01
5.79461277e-01 2.06526741e-01 -5.82885981e-01 7.20485032e-01
-5.45068443e-01 2.99873948e-01 9.65177536e-01 1.01636004e+00
-1.53236791e-01 -8.70296061e-01 -7.49114335e-01 8.93222630e-01
-5.66653073e-01 1.20126411e-01 4.51055914e-02 5.93745410e-01
3.98434192e-01 9.43521917e-01 3.43876153e-01 -4.96986777e-01
6.22608185e-01 -2.01899439e-01 3.81681651e-01 -1.00199866e+00
-7.07934260e-01 -1.81660801e-01 -5.26311025e-02 -3.27439368e-01
-4.64757979e-01 -3.92004520e-01 -1.60240388e+00 -1.78015418e-03
-1.91420063e-01 2.13741049e-01 6.31961524e-01 9.05469537e-01
3.99998665e-01 1.49508035e+00 9.25264537e-01 -5.26578203e-02
-5.00133872e-01 -1.22510076e+00 -3.77455086e-01 5.84752381e-01
5.73864698e-01 -4.77104098e-01 2.06129014e-01 5.65936089e-01]
|
[15.04706859588623, -1.3892362117767334]
|
b3217278-5ddd-42d6-829d-e59224a02a3b
|
fairness-for-workers-who-pull-the-arms-an
|
2303.00799
| null |
https://arxiv.org/abs/2303.00799v1
|
https://arxiv.org/pdf/2303.00799v1.pdf
|
Fairness for Workers Who Pull the Arms: An Index Based Policy for Allocation of Restless Bandit Tasks
|
Motivated by applications such as machine repair, project monitoring, and anti-poaching patrol scheduling, we study intervention planning of stochastic processes under resource constraints. This planning problem has previously been modeled as restless multi-armed bandits (RMAB), where each arm is an intervention-dependent Markov Decision Process. However, the existing literature assumes all intervention resources belong to a single uniform pool, limiting their applicability to real-world settings where interventions are carried out by a set of workers, each with their own costs, budgets, and intervention effects. In this work, we consider a novel RMAB setting, called multi-worker restless bandits (MWRMAB) with heterogeneous workers. The goal is to plan an intervention schedule that maximizes the expected reward while satisfying budget constraints on each worker as well as fairness in terms of the load assigned to each worker. Our contributions are two-fold: (1) we provide a multi-worker extension of the Whittle index to tackle heterogeneous costs and per-worker budget and (2) we develop an index-based scheduling policy to achieve fairness. Further, we evaluate our method on various cost structures and show that our method significantly outperforms other baselines in terms of fairness without sacrificing much in reward accumulated.
|
['Milind Tambe', 'Susobhan Ghosh', 'Paula Rodriguez Diaz', 'Jackson A. Killian', 'Arpita Biswas']
|
2023-03-01
| null | null | null | null |
['multi-armed-bandits']
|
['miscellaneous']
|
[ 4.59262848e-01 3.60320032e-01 -8.21903884e-01 -8.87838677e-02
-6.28571630e-01 -3.63074809e-01 3.17628592e-01 6.88976124e-02
-5.44991851e-01 1.05531394e+00 1.54769540e-01 -5.48271477e-01
-6.83900177e-01 -5.69456279e-01 -6.01388872e-01 -8.42587411e-01
1.32660389e-01 1.13931215e+00 -2.84695197e-02 2.33086482e-01
1.28112689e-01 6.48011565e-01 -9.42253828e-01 4.98467386e-02
4.79301214e-01 9.00092542e-01 8.64270255e-02 7.06201255e-01
2.52134621e-01 1.07473540e+00 -6.82590246e-01 -3.83909345e-01
4.25567895e-01 -1.75966799e-01 -1.15543032e+00 4.79521811e-01
-3.57696235e-01 -5.00309885e-01 -3.17113757e-01 5.70737123e-01
4.77370799e-01 5.27389884e-01 7.84877360e-01 -1.57770252e+00
-2.76869357e-01 1.02989352e+00 -1.12638152e+00 2.33729139e-01
-2.49668062e-02 1.91566929e-01 1.08570182e+00 1.69554770e-01
4.17359360e-02 1.57919216e+00 1.23282038e-02 7.72033513e-01
-1.63418472e+00 -4.33590561e-01 5.19836843e-01 -1.72676459e-01
-6.20146513e-01 -3.16172510e-01 2.21734598e-01 -3.30922216e-01
8.09711158e-01 3.82679880e-01 3.09235334e-01 8.05736899e-01
-2.57832883e-03 1.11822748e+00 1.07325315e+00 -7.51727283e-01
4.36433196e-01 -3.57734263e-01 3.35765451e-01 2.37512127e-01
2.77956069e-01 -4.13828567e-02 -4.04228717e-01 -5.75017393e-01
7.68558145e-01 3.72930735e-01 -1.32230148e-01 -3.38141143e-01
-9.67588782e-01 8.74232233e-01 -2.68533289e-01 -2.00420231e-01
-9.86308336e-01 6.72312677e-01 3.97516549e-01 1.07987903e-01
7.34211862e-01 8.54171515e-02 -1.89042941e-01 1.63833916e-01
-7.95851111e-01 3.34898889e-01 8.18384111e-01 1.21587551e+00
4.30397660e-01 -3.22937369e-01 -1.32542551e+00 7.82225311e-01
5.44376560e-02 7.36096382e-01 -1.62616596e-01 -1.41067696e+00
7.57799327e-01 -7.65856951e-02 1.01080847e+00 -5.07821023e-01
-4.46922064e-01 -3.29318047e-01 -8.65040123e-01 1.40506566e-01
4.01558012e-01 -3.01571518e-01 -7.58348048e-01 1.84505558e+00
1.81941137e-01 -4.23830077e-02 -3.84533674e-01 8.57791185e-01
-2.95420945e-01 5.02308846e-01 1.15095377e-01 -7.74078906e-01
1.42309713e+00 -1.48587584e+00 -7.57443309e-01 -1.90800041e-01
1.31300256e-01 -5.23801446e-01 8.16183925e-01 4.20628995e-01
-1.53558564e+00 1.32670999e-01 -3.57504427e-01 5.59943259e-01
4.62999970e-01 -2.80236065e-01 6.28907323e-01 9.80749786e-01
-6.88276172e-01 3.54210615e-01 -9.35299754e-01 4.63575944e-02
6.14246786e-01 4.55609232e-01 2.06165507e-01 -2.69786745e-01
-5.89251518e-01 7.53019154e-01 -6.16226867e-02 -1.15993164e-01
-1.42338908e+00 -6.74253166e-01 -3.16780061e-01 5.57572842e-01
1.29504502e+00 -9.62038040e-01 2.00767970e+00 -6.60601556e-01
-1.60055304e+00 4.19837117e-01 -1.57202139e-01 -5.17998636e-01
7.36286819e-01 -5.09918854e-02 3.96124750e-01 6.72477484e-02
2.03908458e-01 6.78905621e-02 6.49509609e-01 -1.36041045e+00
-1.06437743e+00 -3.49582106e-01 5.10168493e-01 2.97450453e-01
-2.34019935e-01 4.80941236e-01 -2.62482762e-01 -3.84321749e-01
-4.52310443e-01 -1.15680242e+00 -6.37604237e-01 -8.89303982e-01
-6.37087882e-01 -2.17936784e-01 -2.24355161e-02 -3.15384328e-01
1.21981454e+00 -1.77996552e+00 1.20238267e-01 2.84621596e-01
-4.72456105e-02 -2.07093820e-01 -1.37821108e-01 5.60447037e-01
3.76018196e-01 2.70723552e-01 -5.22535555e-02 -8.71772349e-01
5.40682554e-01 4.55253243e-01 -4.06030178e-01 6.29179716e-01
-3.27428013e-01 5.30643344e-01 -8.55680227e-01 -2.98283368e-01
-9.12017375e-02 -4.90920484e-01 -3.19143444e-01 3.61075759e-01
-3.51511896e-01 3.37386996e-01 -5.74330151e-01 5.63635528e-01
5.88054538e-01 -1.93962410e-01 1.46099210e-01 8.59278083e-01
1.13143638e-01 -1.23137563e-01 -1.21644890e+00 8.56670320e-01
-7.10853577e-01 -1.19948626e-01 4.88290608e-01 -1.16759348e+00
2.06310600e-01 4.45343256e-01 6.36325896e-01 -2.78161705e-01
-2.96826009e-02 -1.76143453e-01 -2.73506403e-01 -2.46595010e-01
4.78147119e-01 -3.43921185e-01 -2.03378931e-01 1.20510316e+00
-3.61432999e-01 -3.50749157e-02 2.63865620e-01 2.56558388e-01
1.35400808e+00 -3.57958019e-01 3.05851251e-01 -1.35616750e-01
4.92405146e-02 -2.78880298e-01 7.25283444e-01 1.39596999e+00
-3.12637061e-01 1.28190160e-01 9.62650597e-01 -2.17313021e-01
-8.70348990e-01 -7.37141311e-01 4.19480741e-01 1.75504649e+00
2.24244520e-01 5.19820869e-01 -5.09235084e-01 -7.43545413e-01
3.74936551e-01 9.08690274e-01 -5.05261540e-01 3.71306866e-01
-3.65548909e-01 -9.17520344e-01 1.36171550e-01 3.56789827e-01
1.50027692e-01 -1.02589333e+00 -8.96951854e-01 3.99114728e-01
-1.41702846e-01 -9.67224836e-01 -1.06772280e+00 1.59992650e-01
-5.12162209e-01 -9.04846132e-01 -9.97759581e-01 -1.27638981e-01
7.58844912e-01 6.11261904e-01 1.07837653e+00 -2.31683567e-01
-4.85005192e-02 7.78720796e-01 -9.42341089e-02 -8.95327687e-01
-8.00879002e-02 1.34463027e-01 5.52252531e-02 4.07356508e-02
-1.19699530e-01 -1.78571671e-01 -7.09205806e-01 6.86937869e-01
-1.08527350e+00 6.30375892e-02 4.28716868e-01 9.90242600e-01
4.47551876e-01 2.55734950e-01 7.86491215e-01 -9.84951496e-01
9.25147593e-01 -3.49392742e-01 -6.93742156e-01 7.35626578e-01
-4.36430454e-01 -1.22439153e-01 2.88674593e-01 -5.77023268e-01
-1.43911660e+00 -2.23911166e-01 8.19050252e-01 -8.84403884e-02
3.02122980e-02 1.14779554e-01 -6.25465810e-02 3.61354619e-01
1.38665065e-01 -3.34050953e-01 -2.75271805e-03 -1.47890300e-01
1.16730556e-01 6.99415565e-01 2.17579946e-01 -1.28519785e+00
2.53971487e-01 4.64558333e-01 3.40166658e-01 -6.49281144e-02
-1.08248270e+00 -5.81887662e-01 1.27838939e-01 -2.08321422e-01
3.81777972e-01 -5.53708375e-01 -1.70238602e+00 2.08555758e-01
-1.15649867e+00 -7.92824805e-01 -3.58508319e-01 4.25957382e-01
-1.18192995e+00 2.29170591e-01 -8.00763607e-01 -1.79286218e+00
-3.18713278e-01 -1.14845812e+00 9.22287703e-01 2.31503680e-01
8.55268538e-02 -5.36128581e-01 -6.48186207e-02 7.34659612e-01
3.45563471e-01 -4.42507751e-02 8.86521876e-01 -3.82821441e-01
-7.04669118e-01 -2.44212374e-02 -5.02314046e-02 1.86625376e-01
-1.13798305e-01 -4.59798187e-01 -3.63832355e-01 -5.33158779e-01
-1.53893784e-01 -4.71395284e-01 6.90869868e-01 8.06359947e-01
1.40894902e+00 -6.17973387e-01 -4.82754052e-01 -1.59783944e-01
9.56919968e-01 3.81679147e-01 1.37669042e-01 4.55663025e-01
1.60593763e-02 8.78336847e-01 8.13205361e-01 1.00927711e+00
1.97573289e-01 7.81256318e-01 6.80081069e-01 -4.37213816e-02
6.45309925e-01 7.43561015e-02 2.11947680e-01 -1.87463760e-01
-5.38450122e-01 -7.34864950e-01 -8.91645193e-01 6.66215003e-01
-2.55322361e+00 -1.15273952e+00 3.38737786e-01 2.48574233e+00
7.50786841e-01 1.50040388e-01 6.48612201e-01 -1.29847988e-01
8.92581642e-01 -1.41984850e-01 -7.12949038e-01 -7.10297287e-01
2.51712233e-01 3.52439657e-02 1.18056774e+00 4.78863299e-01
-8.92666876e-01 5.16119421e-01 6.24573851e+00 9.38356876e-01
-1.20728321e-01 3.43569785e-01 1.21991110e+00 -7.66589761e-01
8.02578181e-02 -5.50399311e-02 -8.35483015e-01 4.71021265e-01
9.66265798e-01 -2.72246361e-01 9.84295368e-01 7.71886706e-01
5.97494185e-01 -3.79131526e-01 -1.12941086e+00 5.29477954e-01
-3.97063971e-01 -9.61084247e-01 -5.23747027e-01 3.95145148e-01
1.10241675e+00 -4.43631619e-01 6.71131536e-02 3.68402719e-01
1.27205515e+00 -1.00997567e+00 6.86931372e-01 6.44509077e-01
4.16087002e-01 -1.41247320e+00 8.35401475e-01 6.83655500e-01
-5.12261689e-01 -5.08578062e-01 -2.95780927e-01 -2.81994164e-01
6.37640476e-01 6.88370466e-01 -6.72611654e-01 5.84704757e-01
4.53291327e-01 -3.39014769e-01 5.65608025e-01 1.02940631e+00
-2.21475556e-01 6.96888328e-01 -9.24500078e-02 1.55029640e-01
2.10397944e-01 -1.80349514e-01 2.30222389e-01 1.00451708e+00
5.38523011e-02 2.61836022e-01 7.85112739e-01 4.04700279e-01
-1.56517029e-01 -4.13879603e-02 -1.14234954e-01 8.34768564e-02
5.44542134e-01 1.03543520e+00 -8.25484395e-01 -3.14697146e-01
-3.16907555e-01 6.57834291e-01 2.80358493e-01 5.03640532e-01
-8.60920370e-01 -4.99351099e-02 7.32613027e-01 -8.96836817e-02
1.64909542e-01 3.22090536e-02 -4.90777016e-01 -4.65840131e-01
-2.70043612e-01 -7.37517416e-01 4.41599131e-01 -3.91349554e-01
-1.50118613e+00 -1.52993742e-02 3.98212969e-01 -3.49657565e-01
-1.55480579e-01 -9.96793434e-02 -4.59208339e-01 1.14047754e+00
-1.46803415e+00 -1.06190801e+00 3.40944856e-01 2.87801623e-01
5.18575132e-01 -9.94021520e-02 4.64289218e-01 6.56002685e-02
-9.35308754e-01 3.37289572e-01 2.41326541e-01 -5.65307915e-01
3.87735963e-01 -1.29645479e+00 -4.52078320e-02 6.36382580e-01
-4.42852169e-01 3.69008929e-01 7.86288500e-01 -5.17090619e-01
-1.10611248e+00 -9.04488266e-01 3.61917078e-01 -1.40324607e-01
5.25881708e-01 -1.77254379e-01 -2.26678684e-01 8.70415807e-01
2.01175809e-01 -1.23065718e-01 6.74495816e-01 3.97401392e-01
3.40093523e-01 -5.77939227e-02 -1.26170945e+00 6.07689500e-01
1.12364852e+00 2.08267309e-02 -5.56348972e-02 7.27859020e-01
8.54438007e-01 -3.70941818e-01 -6.06563628e-01 2.23331481e-01
3.45430762e-01 -5.08528233e-01 6.47223175e-01 -1.06301427e+00
2.71980911e-01 2.44233087e-01 7.79564008e-02 -1.34704542e+00
-6.07167900e-01 -1.15477788e+00 -1.79058462e-01 1.27849984e+00
3.74656588e-01 -5.65561354e-01 8.82461309e-01 1.28171599e+00
-7.03820661e-02 -6.89441502e-01 -1.02235842e+00 -1.05565584e+00
-1.16289034e-01 -2.33505055e-01 9.97471988e-01 3.76896173e-01
-1.15730716e-02 1.19557962e-01 -1.00299752e+00 1.69849038e-01
6.67495310e-01 2.42690846e-01 8.20795238e-01 -7.82491505e-01
-8.71999741e-01 -4.83929068e-01 5.75469971e-01 -1.01521516e+00
5.00089526e-01 -4.23116297e-01 4.84354764e-01 -1.59491026e+00
9.04672682e-01 -6.24434054e-01 -4.59418386e-01 7.99184501e-01
-1.22553267e-01 -2.81014919e-01 4.80395168e-01 2.32874945e-01
-1.01126480e+00 4.27756727e-01 1.18078494e+00 -2.65236378e-01
-2.67199069e-01 9.80661094e-01 -8.32199216e-01 3.13022554e-01
7.38322854e-01 -7.02665448e-01 -3.58684212e-01 -4.05049175e-01
1.67031691e-03 7.45663285e-01 2.29724906e-02 -1.75488770e-01
1.30726367e-01 -1.09411478e+00 -3.51095378e-01 -4.06927496e-01
1.13193877e-01 -8.19414258e-01 2.53777921e-01 6.09609425e-01
-7.72970855e-01 -2.55587064e-02 -2.92184055e-01 1.14101315e+00
3.83432478e-01 -5.86880267e-01 6.25308037e-01 -2.11578682e-01
2.14735836e-01 5.23237467e-01 -5.90403974e-01 -4.77114618e-02
1.42816186e+00 3.61841708e-01 -5.39595366e-01 -5.70657909e-01
-5.04404724e-01 8.60179663e-01 2.08341658e-01 -1.15797952e-01
-9.71976966e-02 -8.92536819e-01 -6.90944493e-01 -5.74178874e-01
-1.71407789e-01 1.69347480e-01 4.54117775e-01 8.35961699e-01
3.65424715e-03 6.44901931e-01 2.02718955e-02 -2.33659193e-01
-1.29006529e+00 8.66268754e-01 -7.50605389e-02 -1.09155822e+00
-1.06086418e-01 6.56466424e-01 2.62023091e-01 9.33225751e-02
5.36904156e-01 1.76340193e-02 2.31665388e-01 -4.99899499e-02
6.02624595e-01 9.64133203e-01 -2.42681295e-01 -4.10339944e-02
-5.43389022e-02 -2.19891578e-01 -2.04299301e-01 -5.61628401e-01
1.40564418e+00 -1.48414075e-01 -2.82242000e-01 8.59524757e-02
1.49340272e-01 -1.63390085e-01 -1.42015135e+00 -6.47640705e-01
1.91711038e-01 -7.56869316e-01 -1.72870662e-02 -7.53251612e-01
-9.75336730e-01 3.42780173e-01 1.14998491e-02 5.44437945e-01
1.08347642e+00 -3.83406095e-02 5.10898173e-01 1.64320961e-01
7.77453899e-01 -1.31706679e+00 1.97003350e-01 3.32133383e-01
6.12162411e-01 -1.08384383e+00 4.63010296e-02 -2.76411653e-01
-9.20857847e-01 4.67335045e-01 3.42926681e-01 2.13080332e-01
2.62745678e-01 1.99886113e-01 -6.95204675e-01 2.10944459e-01
-9.87424374e-01 -3.61094594e-01 -8.77266973e-02 4.43805754e-01
1.23266326e-02 5.86300611e-01 -5.36653876e-01 7.48968661e-01
3.85335684e-01 8.95511881e-02 6.80363834e-01 1.22861004e+00
-4.76665527e-01 -1.19134688e+00 -8.13161194e-01 7.45174289e-01
-9.34202373e-01 2.14827687e-01 1.17632017e-01 2.85593510e-01
-2.45436966e-01 1.38877916e+00 1.81137864e-02 5.73421955e-01
4.17380393e-01 -1.22403495e-01 4.50245887e-01 -1.06020248e+00
-6.30499542e-01 2.09869444e-01 1.15917534e-01 -1.98896810e-01
-4.28465009e-01 -5.75919747e-01 -7.20706701e-01 -5.26050091e-01
-4.27633673e-01 2.21068576e-01 5.13636529e-01 9.48525429e-01
1.02542780e-01 8.20366561e-01 9.77857947e-01 -9.54063237e-01
-1.32415831e+00 -9.96908069e-01 -1.03040051e+00 1.82325095e-01
2.00954884e-01 -8.08163285e-01 -1.28095791e-01 -3.17956388e-01]
|
[4.454455852508545, 3.1877329349517822]
|
fefa1af0-8af0-46b2-afa9-261453c7e8e7
|
an-improved-neural-baseline-for-temporal
|
1909.00429
| null |
https://arxiv.org/abs/1909.00429v1
|
https://arxiv.org/pdf/1909.00429v1.pdf
|
An Improved Neural Baseline for Temporal Relation Extraction
|
Determining temporal relations (e.g., before or after) between events has been a challenging natural language understanding task, partly due to the difficulty to generate large amounts of high-quality training data. Consequently, neural approaches have not been widely used on it, or showed only moderate improvements. This paper proposes a new neural system that achieves about 10% absolute improvement in accuracy over the previous best system (25% error reduction) on two benchmark datasets. The proposed system is trained on the state-of-the-art MATRES dataset and applies contextualized word embeddings, a Siamese encoder of a temporal common sense knowledge base, and global inference via integer linear programming (ILP). We suggest that the new approach could serve as a strong baseline for future research in this area.
|
['Qiang Ning', 'Dan Roth', 'Sanjay Subramanian']
|
2019-09-01
|
an-improved-neural-baseline-for-temporal-1
|
https://aclanthology.org/D19-1642
|
https://aclanthology.org/D19-1642.pdf
|
ijcnlp-2019-11
|
['temporal-relation-extraction']
|
['natural-language-processing']
|
[ 3.45260948e-02 2.57426873e-02 -6.12397969e-01 -5.48203945e-01
-8.30466509e-01 -3.12258124e-01 9.17332470e-01 3.34641576e-01
-8.65591168e-01 8.08897376e-01 5.25108695e-01 -2.08585098e-01
4.65182960e-02 -9.80907500e-01 -7.26830900e-01 -4.62762028e-01
-2.15472341e-01 5.76160908e-01 4.25108045e-01 -2.39558622e-01
1.26904681e-01 3.27608526e-01 -1.27883291e+00 2.82540560e-01
4.71562326e-01 8.26325119e-01 -5.99589907e-02 4.43141818e-01
-1.10457622e-01 1.00978959e+00 -3.82206053e-01 -6.00268066e-01
2.64342930e-02 -3.59444916e-01 -9.30734515e-01 -5.68789721e-01
3.63880277e-01 -2.90261716e-01 -7.05503762e-01 7.87221193e-01
3.47493142e-01 6.02954865e-01 4.65401709e-01 -1.23533702e+00
-1.07402873e+00 7.44538128e-01 -2.69971877e-01 6.89034343e-01
2.11591780e-01 -1.79145411e-01 1.39187849e+00 -7.66875863e-01
7.99028456e-01 1.15917623e+00 3.28488052e-01 5.69713950e-01
-1.19418526e+00 -4.21753883e-01 2.62594461e-01 8.87702525e-01
-1.40794706e+00 -2.50506669e-01 6.76930904e-01 -2.23394096e-01
1.67532122e+00 -7.42365718e-02 6.15202725e-01 1.36125553e+00
2.98475683e-01 8.63479137e-01 7.77761042e-01 -3.72812182e-01
4.48519379e-01 -1.45263970e-01 3.26318353e-01 6.22602642e-01
3.76553424e-02 2.59971231e-01 -8.34771097e-01 1.84313789e-01
5.69683433e-01 -2.26145089e-01 -6.60726577e-02 -5.00825793e-02
-1.38169062e+00 9.38275814e-01 5.77778876e-01 5.57643712e-01
-3.91246825e-01 3.42204541e-01 6.00087583e-01 2.66041487e-01
7.17200220e-01 5.85325778e-01 -6.61699533e-01 -5.85576594e-01
-9.88852262e-01 4.06061620e-01 7.02552676e-01 4.45323259e-01
4.03910279e-01 3.77230300e-03 -1.53465560e-02 6.64247632e-01
7.77250007e-02 1.14652552e-01 6.46567464e-01 -8.36819172e-01
6.55309141e-01 6.56756997e-01 7.71217495e-02 -1.24952865e+00
-3.84428501e-01 -1.63383394e-01 -6.19475543e-01 -1.92074299e-01
5.26427388e-01 -1.53905496e-01 -7.26710796e-01 2.10727835e+00
1.18496910e-01 6.88540399e-01 2.42298529e-01 9.03177500e-01
4.37247694e-01 1.19421351e+00 3.65829021e-02 -1.82365075e-01
1.19917715e+00 -1.16137385e+00 -8.96821499e-01 -4.15060282e-01
4.25719291e-01 -3.90363157e-01 9.50287640e-01 4.86859590e-01
-1.21550405e+00 -4.51362044e-01 -1.12389231e+00 -4.53918278e-01
-9.18250203e-01 6.51237518e-02 8.13527405e-01 1.19866811e-01
-1.00563383e+00 6.84583902e-01 -1.33877099e+00 -6.06724501e-01
1.84930712e-01 4.91374917e-02 -4.11163896e-01 -3.10252234e-02
-1.68648314e+00 1.32583117e+00 5.91592729e-01 2.19995245e-01
-7.46176302e-01 -7.51011610e-01 -1.10994184e+00 -6.52114376e-02
4.73299712e-01 -4.27635759e-01 1.08837819e+00 -5.91711700e-01
-1.44012749e+00 8.47684741e-01 -3.60078067e-01 -9.93651390e-01
1.32818133e-01 -6.37193382e-01 -7.51295805e-01 8.45777914e-02
8.57376754e-02 6.71103597e-01 4.42450374e-01 -6.23405397e-01
-4.89076346e-01 -2.15960950e-01 6.63898885e-02 -7.57321864e-02
-5.22440851e-01 3.01978737e-01 -3.98018628e-01 -8.73447657e-01
-2.86889076e-01 -8.15649152e-01 -3.24267417e-01 -5.01641557e-02
-5.87411560e-02 -6.55806482e-01 5.35908520e-01 -8.15611839e-01
1.57954192e+00 -1.94986594e+00 2.39492506e-01 -1.17808551e-01
-4.10390735e-01 3.55061531e-01 -3.19366187e-01 6.06655180e-01
-2.50508100e-01 5.78175895e-02 -3.29013020e-01 -2.16677010e-01
8.87933150e-02 3.86607558e-01 -5.49419105e-01 3.20760489e-01
4.12553072e-01 1.00943065e+00 -1.18406224e+00 -2.91611135e-01
1.24958336e-01 3.43978763e-01 -3.95624310e-01 1.26696438e-01
-3.86453778e-01 -1.34920433e-01 -3.08871239e-01 2.88514942e-01
1.80174515e-01 -1.69518188e-01 2.37547889e-01 -2.08755448e-01
-8.08701962e-02 7.02135980e-01 -1.07681406e+00 1.99196351e+00
-5.87265909e-01 8.18912923e-01 -6.54580772e-01 -1.39667714e+00
8.45088184e-01 4.84793514e-01 4.70973134e-01 -8.95792842e-01
-3.60447913e-02 -1.09227091e-01 1.11653805e-02 -6.94869339e-01
5.93350530e-01 -1.71925187e-01 -2.85319269e-01 4.43776160e-01
1.86925441e-01 -3.39747183e-02 5.38665116e-01 3.01699549e-01
1.24804401e+00 2.25028366e-01 5.71268976e-01 -7.82417357e-02
4.28163528e-01 1.19831316e-01 7.49987006e-01 4.48805034e-01
-3.89288694e-01 4.35404181e-01 5.14187336e-01 -8.43290329e-01
-1.03352249e+00 -1.21553493e+00 1.43902272e-01 1.00137269e+00
-2.55296584e-02 -6.59013152e-01 -4.03248727e-01 -7.37297595e-01
-2.73082912e-01 1.03632057e+00 -7.96945870e-01 -2.16886029e-01
-9.56461787e-01 -7.22920537e-01 7.98315465e-01 1.02903223e+00
4.20310050e-01 -1.27516103e+00 -4.77975339e-01 4.78497237e-01
-2.50541270e-01 -1.36882257e+00 -3.82823169e-01 1.78967789e-01
-7.45571613e-01 -8.91406000e-01 -2.76366770e-01 -7.91295111e-01
2.14174166e-01 -7.97154084e-02 1.29460812e+00 -4.38837677e-01
-2.17343166e-01 -1.69240981e-01 -4.34865415e-01 -3.17965031e-01
5.04213832e-02 4.95755188e-02 2.29199916e-01 -1.06716387e-01
6.87576652e-01 -5.71329296e-01 -3.32736999e-01 1.00893021e-01
-9.09840047e-01 -2.11912006e-01 2.03276336e-01 9.08185065e-01
3.98542523e-01 1.81550041e-01 8.42588782e-01 -6.03324831e-01
8.13345492e-01 -5.34565091e-01 -5.77432930e-01 2.38421455e-01
-4.45473522e-01 3.99076462e-01 7.41278291e-01 -3.75198752e-01
-1.25157082e+00 -1.84870720e-01 -2.53186196e-01 -2.36340210e-01
-1.64415136e-01 9.96900856e-01 1.82317674e-01 6.74479663e-01
6.30576253e-01 1.45601958e-01 -3.24088007e-01 -2.44090512e-01
7.44299948e-01 3.35599273e-01 6.60221040e-01 -7.45967567e-01
5.09954572e-01 4.65099543e-01 -2.84400761e-01 -5.09519637e-01
-1.22602332e+00 -5.00529468e-01 -7.30190516e-01 1.88940987e-01
1.16989255e+00 -1.00911665e+00 -1.94550022e-01 2.16067165e-01
-1.30675614e+00 -5.66096008e-01 -2.49278784e-01 8.01550567e-01
-4.76821840e-01 1.32939294e-01 -7.73682773e-01 -5.87509274e-01
-1.86414108e-01 -7.23928332e-01 7.89425373e-01 1.69256508e-01
-5.49679160e-01 -1.17017758e+00 3.32537383e-01 3.60688597e-01
3.74556094e-01 2.08792582e-01 8.11095417e-01 -6.89499974e-01
-4.68610376e-01 -1.65815219e-01 -1.70800880e-01 3.55208516e-01
6.81343377e-02 2.12617263e-01 -7.01488853e-01 3.75800282e-02
-3.03902149e-01 -6.56126797e-01 1.04924500e+00 2.72919923e-01
1.14735866e+00 -2.17998654e-01 -2.75329173e-01 1.90569147e-01
1.27593684e+00 4.51903611e-01 6.59009457e-01 3.71656924e-01
2.83578426e-01 3.73090863e-01 7.32157707e-01 3.96226764e-01
7.40661919e-01 8.12477827e-01 1.32584527e-01 3.02207649e-01
-1.27916545e-01 -3.21827203e-01 6.53188884e-01 9.89001989e-01
-1.67451367e-01 -4.38060313e-01 -9.55288887e-01 1.22715175e+00
-2.09681821e+00 -1.22802770e+00 1.19648315e-01 1.81962609e+00
1.11176038e+00 2.80441791e-01 -1.31826952e-01 2.87775155e-02
3.52735281e-01 5.50227523e-01 -3.79891157e-01 -7.66859055e-01
8.51821527e-02 5.50478578e-01 1.55150563e-01 4.43678945e-01
-1.26467800e+00 1.23490286e+00 6.45506191e+00 7.84493327e-01
-1.09522712e+00 2.24761218e-01 4.31638330e-01 -1.30769059e-01
1.29670277e-01 -3.00968066e-02 -7.21298933e-01 3.33710194e-01
1.32369304e+00 -2.59586930e-01 4.90159899e-01 7.41136074e-01
2.70117849e-01 -4.30774316e-02 -1.25805318e+00 8.09308350e-01
2.73517787e-01 -1.39582014e+00 -1.00962587e-01 -3.15685689e-01
7.53220797e-01 1.28980711e-01 -2.03392893e-01 5.53158700e-01
5.58544278e-01 -9.70805645e-01 5.76134562e-01 5.43103814e-01
4.10803258e-01 -7.09550798e-01 7.79223561e-01 3.15041333e-01
-1.37219083e+00 2.40186945e-01 -2.66712844e-01 -4.22226250e-01
5.07863343e-01 6.25680625e-01 -7.57352173e-01 5.22660136e-01
6.40736401e-01 9.36072469e-01 -3.43540579e-01 7.20793366e-01
-6.51546359e-01 8.99577618e-01 -3.93065035e-01 -2.49200359e-01
4.39828515e-01 1.11834556e-01 3.82489800e-01 1.28260422e+00
1.88343123e-01 1.59060121e-01 2.56590426e-01 8.02827001e-01
-1.27717957e-01 -5.26914410e-02 -6.24082029e-01 -4.88908678e-01
3.83233011e-01 1.03321409e+00 -5.39564848e-01 -4.70568776e-01
-7.12093055e-01 8.97448778e-01 7.05081820e-01 2.51784235e-01
-1.03807712e+00 -3.48774552e-01 7.66963303e-01 -3.50971937e-01
2.65666544e-01 -5.34384310e-01 -2.64616251e-01 -1.42334640e+00
2.35018618e-02 -6.30277336e-01 7.66764224e-01 -6.96664393e-01
-1.52857888e+00 4.87651408e-01 1.48499802e-01 -7.93821394e-01
-4.63088661e-01 -6.57571018e-01 -6.87651217e-01 6.28769279e-01
-1.47512436e+00 -9.41844702e-01 3.33013646e-02 5.34506440e-01
7.98911929e-01 -1.32675827e-01 8.83209705e-01 4.67408717e-01
-7.56806076e-01 5.20714343e-01 -2.04908267e-01 3.91554236e-01
7.53814220e-01 -1.26299298e+00 5.42974949e-01 1.14048600e+00
4.93849933e-01 7.66912162e-01 5.11308968e-01 -4.46651340e-01
-1.24152529e+00 -1.22208774e+00 1.54495454e+00 -4.23900664e-01
1.22873092e+00 -3.77217561e-01 -9.44361150e-01 9.24194932e-01
6.51055992e-01 1.54792920e-01 6.53089643e-01 3.62875044e-01
-5.31447232e-01 -2.59868890e-01 -8.11963916e-01 8.32852900e-01
1.00194240e+00 -6.27200723e-01 -1.15711772e+00 1.01282492e-01
8.66930127e-01 -2.47784957e-01 -1.04589689e+00 2.94095874e-01
2.90085673e-01 -5.11555135e-01 9.99158859e-01 -7.68272400e-01
7.81618476e-01 -2.22655043e-01 -1.91586867e-01 -1.47150481e+00
-2.35816836e-01 -3.02096575e-01 -4.16057616e-01 1.22546661e+00
5.98090887e-01 -5.24313509e-01 5.71830571e-01 5.73476851e-01
-2.49709580e-02 -8.74439836e-01 -9.58248615e-01 -8.69332373e-01
1.79251075e-01 -7.94960856e-01 3.20888817e-01 1.20241356e+00
3.52750033e-01 4.58578348e-01 -4.57818091e-01 1.68546587e-01
5.41561879e-02 2.50225216e-01 4.24831957e-01 -7.50522435e-01
-1.44496456e-01 -4.42454785e-01 -3.68798703e-01 -9.08917785e-01
6.46253169e-01 -9.56200659e-01 2.21364759e-02 -1.65048635e+00
1.76603608e-02 -1.75420031e-01 -5.91368020e-01 8.37166667e-01
-2.16234177e-01 1.19637184e-01 -5.00364751e-02 -4.15772825e-01
-6.02979541e-01 6.92710280e-01 8.75227451e-01 -3.26671392e-01
-8.04314911e-02 -3.70435089e-01 -4.92080986e-01 6.75473511e-01
8.99119556e-01 -5.15837669e-01 -5.03678679e-01 -5.33384681e-01
4.61040467e-01 1.08280554e-01 3.14192176e-01 -9.02061045e-01
4.51302290e-01 -5.67148030e-01 -1.49741005e-02 -5.06030738e-01
5.90279341e-01 -6.55171335e-01 -2.30513379e-01 2.02513024e-01
-6.47624135e-01 3.89319658e-01 3.59845370e-01 5.33656061e-01
-7.27365971e-01 -1.17429188e-02 5.36619127e-01 -1.17201909e-01
-1.20690215e+00 1.94160491e-01 -2.51149148e-01 3.76448005e-01
1.08385742e+00 2.62284994e-01 -2.85412967e-01 -2.06449658e-01
-7.20259309e-01 1.97753996e-01 -5.62136061e-02 7.86366105e-01
6.50237918e-01 -1.41413784e+00 -6.50631189e-01 -8.69374499e-02
1.20776720e-01 -1.68175265e-01 5.19128665e-02 5.34482837e-01
-3.25441211e-01 5.66793740e-01 4.18968312e-02 -1.06300615e-01
-9.70761180e-01 6.99373126e-01 7.72776604e-02 -7.10057676e-01
-5.74856818e-01 8.81304085e-01 -2.87031829e-01 -3.11552316e-01
2.15582788e-01 -6.56246066e-01 -1.45900935e-01 2.72531718e-01
6.81551933e-01 2.88542211e-01 3.27486433e-02 -3.56312454e-01
-6.24328613e-01 4.26965922e-01 -6.01651035e-02 -3.67056012e-01
1.63545239e+00 1.31819159e-01 -1.76335171e-01 7.26760983e-01
1.34488165e+00 -2.54472107e-01 -9.25481856e-01 -3.93935770e-01
2.47097656e-01 -3.04114610e-01 1.26925468e-01 -7.79042184e-01
-9.86975312e-01 1.01577318e+00 2.40428761e-01 2.10561812e-01
1.06212306e+00 7.63195828e-02 9.32075918e-01 6.42683506e-01
2.96058595e-01 -1.26312292e+00 2.03099802e-01 8.36765409e-01
8.61698270e-01 -1.27695704e+00 -1.05821095e-01 5.24524190e-02
-7.22570062e-01 1.19345677e+00 6.80488169e-01 -4.21287686e-01
7.90028632e-01 3.19945455e-01 -4.01891805e-02 -1.17526457e-01
-1.31724036e+00 -3.22877795e-01 3.41561139e-01 2.32989237e-01
6.02679551e-01 4.89898548e-02 -5.97008467e-01 6.09559357e-01
-1.39882550e-01 2.81398356e-01 2.73574263e-01 7.65069366e-01
-1.06691189e-01 -1.16989350e+00 2.25669935e-01 2.49220267e-01
-6.26346111e-01 -2.93655008e-01 -5.69448769e-02 7.29873300e-01
1.70438141e-01 9.59016263e-01 1.47547975e-01 -2.00164452e-01
2.91821361e-01 2.11643487e-01 3.29654574e-01 -4.62924689e-01
-3.20841521e-01 -5.00780344e-01 4.22489673e-01 -8.55623066e-01
-6.83376551e-01 -7.11547554e-01 -1.42135155e+00 -5.49863838e-02
6.25714138e-02 6.88639516e-03 4.47210252e-01 1.01043391e+00
3.34322810e-01 6.27166390e-01 1.92909628e-01 -5.31685114e-01
-2.85615742e-01 -8.29352498e-01 -1.95112720e-01 6.10672653e-01
9.09519009e-03 -6.75040603e-01 -9.63426605e-02 2.53788620e-01]
|
[9.207568168640137, 9.203386306762695]
|
a82b9f5f-efb5-48c0-96d5-84ee8a28e28b
|
box-aware-feature-enhancement-for-single
|
2108.04728
| null |
https://arxiv.org/abs/2108.04728v2
|
https://arxiv.org/pdf/2108.04728v2.pdf
|
Box-Aware Feature Enhancement for Single Object Tracking on Point Clouds
|
Current 3D single object tracking approaches track the target based on a feature comparison between the target template and the search area. However, due to the common occlusion in LiDAR scans, it is non-trivial to conduct accurate feature comparisons on severe sparse and incomplete shapes. In this work, we exploit the ground truth bounding box given in the first frame as a strong cue to enhance the feature description of the target object, enabling a more accurate feature comparison in a simple yet effective way. In particular, we first propose the BoxCloud, an informative and robust representation, to depict an object using the point-to-box relation. We further design an efficient box-aware feature fusion module, which leverages the aforementioned BoxCloud for reliable feature matching and embedding. Integrating the proposed general components into an existing model P2B, we construct a superior box-aware tracker (BAT). Experiments confirm that our proposed BAT outperforms the previous state-of-the-art by a large margin on both KITTI and NuScenes benchmarks, achieving a 15.2% improvement in terms of precision while running ~20% faster.
|
['Shuguang Cui', 'Zhen Li', 'Wei zhang', 'Weibing Zhao', 'Jiantao Gao', 'Xu Yan', 'Chaoda Zheng']
|
2021-08-10
| null |
http://openaccess.thecvf.com//content/ICCV2021/html/Zheng_Box-Aware_Feature_Enhancement_for_Single_Object_Tracking_on_Point_Clouds_ICCV_2021_paper.html
|
http://openaccess.thecvf.com//content/ICCV2021/papers/Zheng_Box-Aware_Feature_Enhancement_for_Single_Object_Tracking_on_Point_Clouds_ICCV_2021_paper.pdf
|
iccv-2021-1
|
['3d-single-object-tracking']
|
['computer-vision']
|
[-7.97678903e-02 -4.21975434e-01 -1.50806829e-01 -2.44089544e-01
-9.72566545e-01 -7.70861149e-01 7.45846510e-01 1.61839709e-01
-2.72301853e-01 3.10968697e-01 -1.23320799e-02 7.55611286e-02
-1.28314435e-01 -6.35120034e-01 -6.11365438e-01 -5.86822569e-01
9.55020636e-02 4.65866745e-01 6.77264988e-01 8.90517607e-02
1.45892873e-01 9.60915387e-01 -1.59538317e+00 -3.96935523e-01
6.54794455e-01 1.51596379e+00 1.25808552e-01 1.24217086e-01
-1.61470147e-03 -9.87887289e-03 -4.33324724e-01 -4.02699143e-01
7.35581458e-01 2.12955117e-01 6.35147691e-02 1.40986964e-02
9.40912902e-01 -3.98160666e-01 -4.04857785e-01 9.47261572e-01
4.78766501e-01 -1.66850332e-02 4.39836800e-01 -1.25674260e+00
3.00889835e-02 1.72866657e-01 -9.47998941e-01 -5.80920205e-02
3.70362371e-01 3.02641660e-01 9.16915298e-01 -1.19352114e+00
6.34445310e-01 1.29920232e+00 8.18342686e-01 2.54081309e-01
-1.18150342e+00 -1.11661470e+00 3.10158610e-01 3.85743682e-03
-1.73068655e+00 -4.39428836e-01 8.03422511e-01 -3.13346207e-01
5.06442845e-01 1.94817945e-01 8.83037090e-01 6.88909590e-01
1.36446774e-01 5.70737958e-01 8.84613574e-01 -5.91532933e-03
2.13910881e-02 -7.94571638e-02 -2.87552383e-02 5.99526644e-01
6.91855252e-01 4.64935541e-01 -5.46962500e-01 -2.98583448e-01
7.02545345e-01 3.15440178e-01 -2.56663710e-01 -1.15646839e+00
-1.31954849e+00 6.46765709e-01 9.12714303e-01 -4.85248491e-02
-3.23220491e-01 3.29533160e-01 1.19709589e-01 -2.18145475e-01
3.93332928e-01 5.74776456e-02 -4.25392063e-03 -1.02312908e-01
-1.04758835e+00 5.33202052e-01 3.86467159e-01 1.14964569e+00
8.21766078e-01 -3.95478792e-02 -2.38324821e-01 1.86001718e-01
6.61324620e-01 1.00168061e+00 -2.59132087e-01 -6.80763662e-01
4.59733576e-01 8.58871818e-01 3.53296578e-01 -1.14586425e+00
-3.24313939e-01 -7.33624279e-01 -4.76142973e-01 3.99335742e-01
3.17577928e-01 2.07899645e-01 -7.44498491e-01 1.49609995e+00
1.02048039e+00 4.16746229e-01 -2.10527554e-01 1.00746763e+00
7.37089932e-01 2.54449576e-01 -9.96120796e-02 6.46360964e-03
1.46144414e+00 -7.35134900e-01 -4.70729232e-01 -2.14283496e-01
4.63284016e-01 -8.30064714e-01 3.94207716e-01 4.35245186e-02
-7.79518306e-01 -5.30554414e-01 -1.15393686e+00 1.02075569e-01
-1.38989642e-01 2.77737230e-01 5.28570414e-01 6.17333710e-01
-5.22964835e-01 4.23625618e-01 -1.00417387e+00 -3.86388719e-01
6.89551473e-01 4.42359537e-01 -5.18977463e-01 -2.41972849e-01
-4.68214154e-01 8.81693840e-01 2.40881145e-01 1.08231314e-01
-7.91352808e-01 -1.16396081e+00 -9.77591157e-01 -1.35849163e-01
6.70549989e-01 -6.78550541e-01 1.00761592e+00 9.38697159e-02
-1.08907712e+00 6.02477252e-01 -3.03866774e-01 -5.17413735e-01
6.07220173e-01 -5.35014391e-01 -8.37913379e-02 6.88391924e-02
-5.22183217e-02 7.40503311e-01 8.40285957e-01 -1.31256771e+00
-8.14986169e-01 -7.97850788e-01 -1.52492434e-01 1.69929743e-01
-8.93229526e-03 -1.37570962e-01 -6.45767152e-01 -5.76025009e-01
4.19781387e-01 -1.05777335e+00 -6.43453971e-02 7.65344918e-01
-2.03416169e-01 -2.04317778e-01 1.27931201e+00 -2.77183920e-01
1.00335658e+00 -2.29518127e+00 -1.21263817e-01 2.33915895e-01
5.32331467e-01 2.75164157e-01 1.35705501e-01 3.02103311e-01
3.95082057e-01 -3.46838593e-01 -7.06898272e-02 -5.11441290e-01
1.77097544e-01 -1.20406765e-02 -3.60617757e-01 8.85696948e-01
3.56325537e-01 9.88337934e-01 -8.28613818e-01 -6.68938816e-01
4.87559110e-01 7.41450071e-01 -3.80097598e-01 1.62222534e-01
-9.79220271e-02 3.21237803e-01 -7.73848832e-01 1.05306876e+00
1.11470795e+00 7.48066455e-02 -2.34984130e-01 -5.08932412e-01
-4.14895833e-01 1.81346670e-01 -1.32476604e+00 1.84155834e+00
-9.08346996e-02 4.42633301e-01 4.67216708e-02 -3.92540902e-01
1.30580711e+00 -8.27365518e-02 7.45390058e-01 -3.62463653e-01
2.40858525e-01 2.55845338e-01 -2.25755777e-02 2.82900810e-01
5.46618879e-01 2.36964658e-01 -1.94391869e-02 1.63104117e-01
-2.65355945e-01 -2.69762933e-01 1.35011245e-02 7.37924874e-02
1.13274360e+00 5.20913661e-01 4.68242198e-01 -2.12883025e-01
4.96377021e-01 1.22645639e-01 6.64749205e-01 5.04194260e-01
-3.77795994e-01 5.51035106e-01 -1.04546748e-01 -2.63979524e-01
-7.21202195e-01 -1.08338177e+00 -3.99981737e-01 3.93885732e-01
5.65745234e-01 -5.76994240e-01 -3.88990134e-01 -6.28711224e-01
6.28863335e-01 2.28802145e-01 -4.69006598e-01 -1.37928501e-01
-6.63030624e-01 -1.04555257e-01 3.15990269e-01 7.31364667e-01
5.44646263e-01 -3.35535765e-01 -1.08624327e+00 1.45348102e-01
1.96944281e-01 -1.26171660e+00 -5.29277503e-01 -2.69677453e-02
-9.92779255e-01 -1.04952002e+00 -5.53700387e-01 -1.67937890e-01
4.48674381e-01 8.52968454e-01 7.22952068e-01 6.27382919e-02
-3.50627929e-01 3.61163676e-01 -3.50498050e-01 -4.64644015e-01
2.59364039e-01 -1.99678376e-01 2.02395618e-01 -2.01588608e-02
3.88770342e-01 -3.88585001e-01 -7.64944911e-01 4.45620775e-01
-5.14231861e-01 -1.28340900e-01 6.79772258e-01 4.92499650e-01
7.71233499e-01 -2.86975563e-01 -5.18443026e-02 -1.81966573e-01
-1.57836497e-01 -2.29373187e-01 -1.04581320e+00 7.58458450e-02
-3.86721224e-01 8.83047190e-03 2.00433247e-02 -4.55385208e-01
-6.95557177e-01 5.83774984e-01 2.00874656e-01 -8.92393351e-01
9.10893083e-02 1.74011990e-01 -3.23024690e-01 -7.39238799e-01
3.46569091e-01 2.87295002e-02 -1.01117224e-01 -7.03512847e-01
4.72628593e-01 3.42724413e-01 6.69028163e-01 -5.00596583e-01
1.64147127e+00 8.88007164e-01 4.40590799e-01 -4.75190222e-01
-8.71229529e-01 -8.30764174e-01 -8.82009387e-01 -3.17467451e-01
4.95882571e-01 -1.11427927e+00 -9.43608403e-01 1.22244574e-01
-1.06541097e+00 3.59016031e-01 -3.38985056e-01 6.00645006e-01
-3.48698020e-01 4.27937448e-01 1.06800422e-01 -8.97640467e-01
-4.06644702e-01 -1.10235476e+00 1.68310678e+00 3.52620631e-01
1.89981103e-01 -4.66913670e-01 1.30148605e-01 4.32938263e-02
2.72345632e-01 4.90655929e-01 2.52766728e-01 -5.44938862e-01
-1.13852990e+00 -5.22770941e-01 -5.72584093e-01 -2.74007231e-01
2.48305593e-02 1.64746895e-01 -9.31170940e-01 -4.98247862e-01
-1.72206253e-01 1.14729322e-01 7.50965893e-01 1.40048623e-01
6.73113406e-01 3.45687896e-01 -9.12279129e-01 7.45396197e-01
1.47828400e+00 -8.80793780e-02 3.18167895e-01 2.80821323e-01
5.46683311e-01 3.25930625e-01 1.40152276e+00 5.78503966e-01
3.90745074e-01 1.23904264e+00 7.93892741e-01 2.16254741e-01
-3.17750603e-01 -3.98664951e-01 3.38428408e-01 3.89949352e-01
1.22530364e-01 2.12126300e-01 -7.65110016e-01 4.29866612e-01
-1.80579376e+00 -7.20904052e-01 -9.79973525e-02 2.47405267e+00
4.23169315e-01 1.87325299e-01 1.80680096e-01 -8.61014053e-02
6.91644847e-01 1.71360552e-01 -5.71986854e-01 5.60747445e-01
5.63969500e-02 9.85012501e-02 6.21332169e-01 1.84526905e-01
-1.24570870e+00 8.98052990e-01 5.24440002e+00 8.88200164e-01
-1.05062628e+00 -3.01842997e-03 -2.70842642e-01 -1.42733127e-01
4.16842923e-02 2.99122959e-01 -1.41877794e+00 2.89400637e-01
5.50530434e-01 -4.49480057e-01 -1.66150883e-01 1.05393052e+00
1.20562920e-02 -1.23884693e-01 -1.11028433e+00 1.06351566e+00
6.14928082e-02 -1.20983362e+00 -1.67847261e-01 3.76923293e-01
4.45736140e-01 1.54572502e-02 -1.67677682e-02 2.00761721e-01
-9.33029689e-03 -5.09767890e-01 9.35312748e-01 4.58636314e-01
5.84482491e-01 -5.78867078e-01 5.59926391e-01 2.61759520e-01
-2.04246116e+00 3.87643948e-02 -5.61918795e-01 2.49228746e-01
1.91290826e-01 6.13615751e-01 -7.31382251e-01 9.45326149e-01
7.14761555e-01 8.76407623e-01 -6.66841030e-01 1.77924538e+00
2.30644438e-02 1.76841900e-01 -8.05668771e-01 1.93211302e-01
1.66572601e-01 -6.29960448e-02 8.69587004e-01 8.78140986e-01
6.34402752e-01 9.99797881e-02 5.72443902e-01 8.97019982e-01
2.99219415e-02 4.98451404e-02 -6.03357553e-01 1.81473106e-01
8.43539238e-01 1.58854342e+00 -6.61247015e-01 -1.21676281e-01
-3.98845106e-01 3.31254631e-01 1.41285330e-01 -1.85791016e-01
-1.03515637e+00 -1.52655095e-01 9.96159077e-01 1.31106883e-01
7.62683511e-01 -4.52539474e-01 -4.64825369e-02 -1.03211510e+00
3.41758877e-01 -5.26289105e-01 1.40538275e-01 -6.12264037e-01
-1.03870082e+00 6.39400065e-01 9.41999331e-02 -1.81119192e+00
-8.85367095e-02 -5.46437263e-01 -3.52696925e-01 7.39757061e-01
-1.67748499e+00 -1.61094928e+00 -7.86612034e-01 5.15816092e-01
1.45338103e-01 2.10328728e-01 4.29331332e-01 3.21792006e-01
-2.76655167e-01 4.05998260e-01 -1.43908441e-01 -9.23983231e-02
6.98605835e-01 -8.92092764e-01 4.54783440e-01 8.46842289e-01
3.69673610e-01 6.60211384e-01 5.79582930e-01 -1.01358664e+00
-1.96876061e+00 -1.13230908e+00 4.31159407e-01 -6.24594748e-01
6.90495968e-01 -5.05325019e-01 -7.52051413e-01 4.85490680e-01
-3.50251943e-01 3.86802822e-01 2.01967314e-01 -2.79742032e-01
-4.90234762e-01 -3.83126229e-01 -1.10765839e+00 4.54776198e-01
1.34527874e+00 -2.19742507e-01 -6.21478617e-01 7.96981975e-02
8.54544401e-01 -7.91679263e-01 -1.00245404e+00 8.79971921e-01
7.24563003e-01 -7.14640081e-01 1.32005775e+00 -1.45611271e-01
-3.19385856e-01 -8.75511408e-01 -3.07339102e-01 -8.48804653e-01
-2.07818627e-01 -5.37748456e-01 -4.30745959e-01 1.21422839e+00
-2.34160468e-01 -4.30999964e-01 9.96977568e-01 3.07402998e-01
-1.58232048e-01 -7.18170285e-01 -1.31522584e+00 -1.08272374e+00
-4.98835474e-01 -5.88540852e-01 8.12544882e-01 3.49265039e-01
-4.70042169e-01 -1.68073565e-01 -2.12512285e-01 4.39234823e-01
1.07343709e+00 5.66943645e-01 1.30950904e+00 -1.69377029e+00
2.44752169e-01 -4.03935075e-01 -8.09023261e-01 -1.36753082e+00
-1.29777402e-01 -6.60500467e-01 1.92438096e-01 -1.21910858e+00
1.64274290e-01 -6.53121352e-01 -5.59033379e-02 2.92224824e-01
-1.50919661e-01 5.13092756e-01 5.86202145e-01 2.13514477e-01
-6.51238382e-01 8.81548643e-01 1.06834304e+00 -3.34571339e-02
4.02559191e-02 1.42654434e-01 -4.70722139e-01 6.08969450e-01
2.28943035e-01 -6.42450631e-01 1.95288420e-01 -2.46392041e-01
-3.83196950e-01 -2.66345173e-01 7.07045317e-01 -1.35823584e+00
4.59613234e-01 5.59987314e-02 4.12349582e-01 -1.43951058e+00
9.25564229e-01 -1.41818964e+00 2.89744258e-01 5.09095550e-01
2.63733387e-01 1.58691294e-02 3.87785673e-01 7.91854382e-01
-8.47999305e-02 -7.14985430e-02 5.86442351e-01 3.33656996e-01
-6.38539374e-01 7.00180650e-01 3.96151453e-01 -3.10294032e-01
1.31218529e+00 -4.73289728e-01 -4.20698166e-01 -1.16912663e-01
-1.74212724e-01 3.04478139e-01 8.18243027e-01 4.47426111e-01
7.20621288e-01 -1.60034370e+00 -4.91669357e-01 2.51906335e-01
4.32299048e-01 7.30616301e-02 1.29477307e-01 1.07640076e+00
-2.13036641e-01 6.51216388e-01 -1.18387900e-01 -1.16418862e+00
-1.44982398e+00 4.81330216e-01 1.21551856e-01 -7.48253390e-02
-8.67625296e-01 4.97159153e-01 3.24946225e-01 -1.61963597e-01
3.68793070e-01 -4.71333742e-01 1.17731810e-01 -1.27581611e-01
5.53609729e-01 2.69747555e-01 -5.44007914e-03 -9.09993172e-01
-6.81173921e-01 1.30564022e+00 6.37755916e-02 1.39776081e-01
1.16614366e+00 -5.10324091e-02 2.88109720e-01 2.58530900e-02
8.26060534e-01 2.94725686e-01 -1.64607227e+00 -6.66566551e-01
9.07854810e-02 -1.09527397e+00 1.28896877e-01 -4.15229410e-01
-9.97827530e-01 8.61451030e-01 7.82550037e-01 -2.09133416e-01
8.66752386e-01 7.88450763e-02 7.24558532e-01 4.15641069e-01
7.34236360e-01 -3.60312432e-01 -1.16297640e-01 1.77912652e-01
9.38960493e-01 -1.20639896e+00 4.86161739e-01 -8.13230276e-01
-2.12181434e-01 1.00653648e+00 5.93389332e-01 -2.07119808e-01
4.53510702e-01 4.37801957e-01 -1.40953019e-01 -2.72811055e-01
-4.49725956e-01 -5.21481514e-01 7.30372190e-01 6.74760401e-01
-1.46177724e-01 -1.90051243e-01 1.60354504e-03 4.18902457e-01
-4.98598255e-02 -2.82203466e-01 -1.71739608e-01 1.00597656e+00
-6.88435316e-01 -1.08825004e+00 -7.10883677e-01 1.96193099e-01
3.92131805e-02 2.28617415e-01 -3.18839639e-01 1.17444825e+00
-4.59448732e-02 6.16811991e-01 7.77662769e-02 -3.56154412e-01
5.10743618e-01 -2.10853711e-01 5.94193339e-01 -5.18043041e-01
-5.44458032e-01 2.75539070e-01 -2.51527995e-01 -1.07235241e+00
-5.62942326e-01 -8.72778594e-01 -1.11463940e+00 -2.07725257e-01
-7.01368272e-01 -1.21363834e-01 9.29119051e-01 7.78681040e-01
5.52761078e-01 1.12635903e-01 4.77286875e-01 -1.21760249e+00
-8.00162017e-01 -6.97168529e-01 -2.99165100e-01 2.41738960e-01
1.77370265e-01 -1.51623321e+00 -1.28803805e-01 -4.68276083e-01]
|
[6.635537624359131, -2.3381707668304443]
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.