Unnamed: 0.1
int64
0
41k
Unnamed: 0
int64
0
41k
author
stringlengths
9
1.39k
id
stringlengths
11
18
summary
stringlengths
25
3.66k
title
stringlengths
4
258
year
int64
1.99k
2.02k
arxiv_url
stringlengths
32
39
info
stringlengths
523
3.18k
embeddings
stringlengths
16.9k
17.1k
300
300
['Zhengli Zhao', 'Dheeru Dua', 'Sameer Singh']
1710.11342v2
Due to their complex nature, it is hard to characterize the ways in which machine learning models can misbehave or be exploited when deployed. Recent work on adversarial examples, i.e. inputs with minor perturbations that result in substantially different model predictions, is helpful in evaluating the robustness of th...
Generating Natural Adversarial Examples
2,017
http://arxiv.org/pdf/1710.11342v2
Title Generating Natural Adversarial Examples Summary Due complex nature hard characterize way machine learning model misbehave exploited deployed Recent work adversarial example ie input minor perturbation result substantially different model prediction helpful evaluating robustness model exposing adversarial scenario...
[0.025467297062277794, 0.0345236137509346, -0.02137664146721363, 0.03239176422357559, -0.019342325627803802, -0.021692601963877678, 0.06322627514600754, 0.017848998308181763, -0.00045189441880211234, -0.05543028563261032, 0.0008594866376370192, 0.01735883392393589, 0.0008117759716697037, 0.06891801208257675, 0.07641702...
301
301
['Xu Sun', 'Xuancheng Ren', 'Shuming Ma', 'Bingzhen Wei', 'Wei Li', 'Houfeng Wang']
1711.06528v1
We propose a simple yet effective technique to simplify the training and the resulting model of neural networks. In back propagation, only a small subset of the full gradient is computed to update the model parameters. The gradient vectors are sparsified in such a way that only the top-$k$ elements (in terms of magnitu...
Training Simplification and Model Simplification for Deep Learning: A Minimal Effort Back Propagation Method
2,017
http://arxiv.org/pdf/1711.06528v1
Title Training Simplification Model Simplification Deep Learning Minimal Effort Back Propagation Method Summary propose simple yet effective technique simplify training resulting model neural network back propagation small subset full gradient computed update model parameter gradient vector sparsified way topk element ...
[-0.014363215304911137, 0.04249928519129753, -0.023921480402350426, 0.047291845083236694, 0.023564763367176056, 0.003695698222145438, 0.03487322852015495, 0.036320097744464874, -0.03584415838122368, 0.002887630369514227, 0.0019521421054378152, 0.04438763111829758, 0.042764995247125626, 0.035919446498155594, 0.026562314...
302
302
['Abhishek Das', 'Samyak Datta', 'Georgia Gkioxari', 'Stefan Lee', 'Devi Parikh', 'Dhruv Batra']
1711.11543v2
We present a new AI task -- Embodied Question Answering (EmbodiedQA) -- where an agent is spawned at a random location in a 3D environment and asked a question ("What color is the car?"). In order to answer, the agent must first intelligently navigate to explore the environment, gather information through first-person ...
Embodied Question Answering
2,017
http://arxiv.org/pdf/1711.11543v2
Title Embodied Question Answering Summary present new AI task Embodied Question Answering EmbodiedQA agent spawned random location 3D environment asked question color car order answer agent must first intelligently navigate explore environment gather information firstperson egocentric vision answer question orange chal...
[0.06596985459327698, 0.012691699899733067, -0.019865935668349266, 0.00982577633112669, 0.013813847675919533, -6.300141831161454e-05, -0.020515792071819305, -0.017123965546488762, -0.007951822131872177, -0.048882368952035904, -0.0033281196374446154, 0.04057680442929268, 0.014527885243296623, 0.07339311391115189, 0.0211...
303
303
['Aishwarya Agrawal', 'Dhruv Batra', 'Devi Parikh', 'Aniruddha Kembhavi']
1712.00377v1
A number of studies have found that today's Visual Question Answering (VQA) models are heavily driven by superficial correlations in the training data and lack sufficient image grounding. To encourage development of models geared towards the latter, we propose a new setting for VQA where for every question type, train ...
Don't Just Assume; Look and Answer: Overcoming Priors for Visual Question Answering
2,017
http://arxiv.org/pdf/1712.00377v1
Title Dont Assume Look Answer Overcoming Priors Visual Question Answering Summary number study found today Visual Question Answering VQA model heavily driven superficial correlation training data lack sufficient image grounding encourage development model geared towards latter propose new setting VQA every question typ...
[0.051256436854600906, 0.05397116392850876, -0.019469494000077248, 0.03969557583332062, 0.013161051087081432, 0.014416893012821674, 0.0011420906521379948, 0.030919628217816353, 0.004973620176315308, -0.03810565546154976, -0.04670104756951332, 0.011278714053332806, -0.03848955035209656, 0.042882878333330154, 0.069922007...
304
304
['Jin-Hwa Kim', 'Devi Parikh', 'Dhruv Batra', 'Byoung-Tak Zhang', 'Yuandong Tian']
1712.05558v1
In this work, we propose a goal-driven collaborative task that contains vision, language, and action in a virtual environment as its core components. Specifically, we develop a collaborative `Image Drawing' game between two agents, called CoDraw. Our game is grounded in a virtual world that contains movable clip art ob...
CoDraw: Visual Dialog for Collaborative Drawing
2,017
http://arxiv.org/pdf/1712.05558v1
Title CoDraw Visual Dialog Collaborative Drawing Summary work propose goaldriven collaborative task contains vision language action virtual environment core component Specifically develop collaborative Image Drawing game two agent called CoDraw game grounded virtual world contains movable clip art object Two player Tel...
[0.0213069636374712, -0.03527986630797386, -0.03172720596194267, 0.015142887830734253, -0.04521238058805466, -0.003029627725481987, 0.03395911306142807, 0.03418860584497452, -0.03761177137494087, 0.003921459894627333, -0.058106403797864914, 0.008718117140233517, 0.00938714761286974, 0.07117760926485062, -0.000524366216...
305
305
['Sang-Woo Lee', 'Yu-Jung Heo', 'Byoung-Tak Zhang']
1802.03881v1
Goal-oriented dialogue has been paid attention for its numerous applications in artificial intelligence. To solve this task, deep learning and reinforcement learning have recently been applied. However, these approaches struggle to find a competent recurrent neural questioner, owing to the complexity of learning a seri...
Answerer in Questioner's Mind for Goal-Oriented Visual Dialogue
2,018
http://arxiv.org/pdf/1802.03881v1
Title Answerer Questioners Mind GoalOriented Visual Dialogue Summary Goaloriented dialogue paid attention numerous application artificial intelligence solve task deep learning reinforcement learning recently applied However approach struggle find competent recurrent neural questioner owing complexity learning series se...
[0.08786161243915558, 0.057616252452135086, -0.016296284273266792, 0.029011499136686325, 0.001160371582955122, 0.007916153408586979, 0.020862827077507973, 0.01133309118449688, -0.010481358505785465, -0.013166438788175583, -0.011685785837471485, -0.020735904574394226, -0.014125279150903225, 0.08974589407444, 0.021924840...
306
306
['Tolga Bolukbasi', 'Kai-Wei Chang', 'Joseph Wang', 'Venkatesh Saligrama']
1602.08761v2
We study the problem of structured prediction under test-time budget constraints. We propose a novel approach applicable to a wide range of structured prediction problems in computer vision and natural language processing. Our approach seeks to adaptively generate computationally costly features during test-time in ord...
Resource Constrained Structured Prediction
2,016
http://arxiv.org/pdf/1602.08761v2
Title Resource Constrained Structured Prediction Summary study problem structured prediction testtime budget constraint propose novel approach applicable wide range structured prediction problem computer vision natural language processing approach seek adaptively generate computationally costly feature testtime order r...
[0.031651295721530914, 0.025738857686519623, 0.02919025346636772, 0.060441844165325165, -0.011648801155388355, -0.02231590449810028, -0.01613144390285015, 0.05831962823867798, 0.048590101301670074, -0.07138843089342117, -0.004201329778879881, -0.008253581821918488, 0.018609730526804924, 0.09011174738407135, -0.05046676...
307
307
['Hongyuan Mei', 'Mohit Bansal', 'Matthew R. Walter']
1506.04089v4
We propose a neural sequence-to-sequence model for direction following, a task that is essential to realizing effective autonomous agents. Our alignment-based encoder-decoder model with long short-term memory recurrent neural networks (LSTM-RNN) translates natural language instructions to action sequences based upon a ...
Listen, Attend, and Walk: Neural Mapping of Navigational Instructions to Action Sequences
2,015
http://arxiv.org/pdf/1506.04089v4
Title Listen Attend Walk Neural Mapping Navigational Instructions Action Sequences Summary propose neural sequencetosequence model direction following task essential realizing effective autonomous agent alignmentbased encoderdecoder model long shortterm memory recurrent neural network LSTMRNN translates natural languag...
[0.024290094152092934, -0.0022293042857199907, 0.0014228236395865679, 0.036840006709098816, -0.017758684232831, -0.003951700404286385, -0.01378256268799305, -0.05063077062368393, 0.008958814665675163, -0.051902733743190765, -0.01914503239095211, -0.005800245329737663, 0.028867525979876518, 0.05201384425163269, 0.019872...
308
308
['Lili Mou', 'Zhengdong Lu', 'Hang Li', 'Zhi Jin']
1612.02741v4
Building neural networks to query a knowledge base (a table) with natural language is an emerging research topic in deep learning. An executor for table querying typically requires multiple steps of execution because queries may have complicated structures. In previous studies, researchers have developed either fully d...
Coupling Distributed and Symbolic Execution for Natural Language Queries
2,016
http://arxiv.org/pdf/1612.02741v4
Title Coupling Distributed Symbolic Execution Natural Language Queries Summary Building neural network query knowledge base table natural language emerging research topic deep learning executor table querying typically requires multiple step execution query may complicated structure previous study researcher developed ...
[0.03237861767411232, 0.0621078796684742, -0.007896753028035164, 0.016422171145677567, -0.02499024197459221, 0.01809725910425186, 0.026219798251986504, 0.0015843062428757548, 0.013832502998411655, -0.029324224218726158, -0.015943409875035286, 0.02600032649934292, -0.0275852270424366, 0.05248690024018288, -0.02991908974...
309
309
['Christian Napoli', 'Giuseppe Pappalardo', 'Emiliano Tramontana']
1409.8484v1
Due to the huge availability of documents in digital form, and the deception possibility raise bound to the essence of digital documents and the way they are spread, the authorship attribution problem has constantly increased its relevance. Nowadays, authorship attribution,for both information retrieval and analysis, h...
An agent-driven semantical identifier using radial basis neural networks and reinforcement learning
2,014
http://arxiv.org/pdf/1409.8484v1
Title agentdriven semantical identifier using radial basis neural network reinforcement learning Summary Due huge availability document digital form deception possibility raise bound essence digital document way spread authorship attribution problem constantly increased relevance Nowadays authorship attributionfor info...
[0.08474917709827423, 0.03390363231301308, -0.015425737015902996, 0.03638617321848869, -0.06699015945196152, -0.004816551227122545, 0.07219133526086807, -0.0009300599340349436, -0.037098102271556854, -0.03223426267504692, 0.003375340485945344, 0.04647037014365196, -0.01319155190140009, -0.021497365087270737, -0.0090126...
310
310
['Karla Stepanova', 'Matej Hoffmann', 'Zdenek Straka', 'Frederico B. Klein', 'Angelo Cangelosi', 'Michal Vavrecka']
1706.02490v1
Humans and animals are constantly exposed to a continuous stream of sensory information from different modalities. At the same time, they form more compressed representations like concepts or symbols. In species that use language, this process is further structured by this interaction, where a mapping between the senso...
Where is my forearm? Clustering of body parts from simultaneous tactile and linguistic input using sequential mapping
2,017
http://arxiv.org/pdf/1706.02490v1
Title forearm Clustering body part simultaneous tactile linguistic input using sequential mapping Summary Humans animal constantly exposed continuous stream sensory information different modality time form compressed representation like concept symbol specie use language process structured interaction mapping sensorimo...
[0.0074343751184642315, -0.02270730584859848, -0.017841124907135963, -0.012989522889256477, 0.009440958499908447, 0.009954162873327732, 0.015917371958494186, -0.015028340741991997, -0.016659880056977272, -0.02245429903268814, -0.042172789573669434, -0.03100765496492386, 0.054333820939064026, 0.07643438875675201, -0.003...
311
311
['Tara N. Sainath', 'Brian Kingsbury', 'Abdel-rahman Mohamed', 'George E. Dahl', 'George Saon', 'Hagen Soltau', 'Tomas Beran', 'Aleksandr Y. Aravkin', 'Bhuvana Ramabhadran']
1309.1501v3
Deep Convolutional Neural Networks (CNNs) are more powerful than Deep Neural Networks (DNN), as they are able to better reduce spectral variation in the input signal. This has also been confirmed experimentally, with CNNs showing improvements in word error rate (WER) between 4-12% relative compared to DNNs across a var...
Improvements to deep convolutional neural networks for LVCSR
2,013
http://arxiv.org/pdf/1309.1501v3
Title Improvements deep convolutional neural network LVCSR Summary Deep Convolutional Neural Networks CNNs powerful Deep Neural Networks DNN able better reduce spectral variation input signal also confirmed experimentally CNNs showing improvement word error rate WER 412 relative compared DNNs across variety LVCSR task ...
[0.01443859189748764, 0.02785332500934601, 0.02006520889699459, 0.079326331615448, 0.006320999935269356, -0.029197150841355324, 0.010000168345868587, 0.03634302318096161, -0.06231171265244484, 0.013957126997411251, -0.056182097643613815, -0.019267337396740913, 0.017107225954532623, 0.023459646850824356, 0.0020984876900...
312
312
['Hao Wang', 'Naiyan Wang', 'Dit-Yan Yeung']
1409.2944v2
Collaborative filtering (CF) is a successful approach commonly used by many recommender systems. Conventional CF-based methods use the ratings given to items by users as the sole source of information for learning to make recommendation. However, the ratings are often very sparse in many applications, causing CF-based ...
Collaborative Deep Learning for Recommender Systems
2,014
http://arxiv.org/pdf/1409.2944v2
Title Collaborative Deep Learning Recommender Systems Summary Collaborative filtering CF successful approach commonly used many recommender system Conventional CFbased method use rating given item user sole source information learning make recommendation However rating often sparse many application causing CFbased meth...
[0.016514373943209648, 0.01339777559041977, 0.011778511106967926, 0.006345792673528194, -0.03892507404088974, -0.0008645325433462858, 0.052268750965595245, -0.007281020749360323, -0.016501009464263916, 0.008097927086055279, -0.08761310577392578, -0.004940424580127001, 0.0025139900390058756, 0.07491623610258102, -0.0362...
313
313
['Leila Arras', 'Franziska Horn', 'Grégoire Montavon', 'Klaus-Robert Müller', 'Wojciech Samek']
1606.07298v1
Layer-wise relevance propagation (LRP) is a recently proposed technique for explaining predictions of complex non-linear classifiers in terms of input variables. In this paper, we apply LRP for the first time to natural language processing (NLP). More precisely, we use it to explain the predictions of a convolutional n...
Explaining Predictions of Non-Linear Classifiers in NLP
2,016
http://arxiv.org/pdf/1606.07298v1
Title Explaining Predictions NonLinear Classifiers NLP Summary Layerwise relevance propagation LRP recently proposed technique explaining prediction complex nonlinear classifier term input variable paper apply LRP first time natural language processing NLP precisely use explain prediction convolutional neural network C...
[0.04382606968283653, 0.034422617405653, -0.023115500807762146, 0.04026142880320549, -0.01811833120882511, 0.002976194489747286, 0.009605631232261658, 0.030828626826405525, -0.044057056307792664, -0.05451709404587746, 0.03667408600449562, 0.032001424580812454, -0.00562599953263998, 0.04703965783119202, 0.00246474356390...
314
314
['Vasily Pestun', 'Yiannis Vlassopoulos']
1710.10248v2
We propose a new statistical model suitable for machine learning of systems with long distance correlations such as natural languages. The model is based on directed acyclic graph decorated by multi-linear tensor maps in the vertices and vector spaces in the edges, called tensor network. Such tensor networks have been ...
Tensor network language model
2,017
http://arxiv.org/pdf/1710.10248v2
Title Tensor network language model Summary propose new statistical model suitable machine learning system long distance correlation natural language model based directed acyclic graph decorated multilinear tensor map vertex vector space edge called tensor network tensor network previously employed effective numerical ...
[0.0006524427444674075, -0.01208573393523693, -0.03267817571759224, 0.030661992728710175, -0.04480453208088875, 0.01856502890586853, 0.004919448867440224, 0.013554394245147705, -0.011060710996389389, -0.015567171387374401, -0.008440441451966763, -0.06007058173418045, 0.011063058860599995, 0.016790524125099182, 0.033239...
315
315
['Vasily Pestun', 'John Terilla', 'Yiannis Vlassopoulos']
1711.01416v1
We propose a statistical model for natural language that begins by considering language as a monoid, then representing it in complex matrices with a compatible translation invariant probability measure. We interpret the probability measure as arising via the Born rule from a translation invariant matrix product state.
Language as a matrix product state
2,017
http://arxiv.org/pdf/1711.01416v1
Title Language matrix product state Summary propose statistical model natural language begin considering language monoid representing complex matrix compatible translation invariant probability measure interpret probability measure arising via Born rule translation invariant matrix product state Authors 0 Ahmed Osman W...
[0.01929464563727379, 0.024036217480897903, -0.05126729980111122, 0.029100241139531136, -0.05427022650837898, 0.011777141131460667, 0.03540840372443199, 0.009799971245229244, -0.03212626278400421, -0.055762164294719696, 0.026631193235516548, -0.04721926152706146, 0.041042253375053406, 0.06456483155488968, 0.01606570184...
316
316
['Tara N. Sainath', 'Lior Horesh', 'Brian Kingsbury', 'Aleksandr Y. Aravkin', 'Bhuvana Ramabhadran']
1309.1508v3
Hessian-free training has become a popular parallel second or- der optimization technique for Deep Neural Network training. This study aims at speeding up Hessian-free training, both by means of decreasing the amount of data used for training, as well as through reduction of the number of Krylov subspace solver iterati...
Accelerating Hessian-free optimization for deep neural networks by implicit preconditioning and sampling
2,013
http://arxiv.org/pdf/1309.1508v3
Title Accelerating Hessianfree optimization deep neural network implicit preconditioning sampling Summary Hessianfree training become popular parallel second der optimization technique Deep Neural Network training study aim speeding Hessianfree training mean decreasing amount data used training well reduction number Kr...
[-0.02776738442480564, 0.06034661456942558, -0.010230598039925098, 0.026364007964730263, 0.034973468631505966, -0.011620214208960533, 0.01766505092382431, 0.00014441912935581058, 0.004503607749938965, -0.017087500542402267, -0.046327415853738785, -0.03596096485853195, -0.004380589351058006, -0.033307358622550964, -0.00...
317
317
['Roberto Camacho Barranco', 'Laura M. Rodriguez', 'Rebecca Urbina', 'M. Shahriar Hossain']
1606.07496v1
While textual reviews have become prominent in many recommendation-based systems, automated frameworks to provide relevant visual cues against text reviews where pictures are not available is a new form of task confronted by data mining and machine learning researchers. Suggestions of pictures that are relevant to the ...
Is a Picture Worth Ten Thousand Words in a Review Dataset?
2,016
http://arxiv.org/pdf/1606.07496v1
Title Picture Worth Ten Thousand Words Review Dataset Summary textual review become prominent many recommendationbased system automated framework provide relevant visual cue text review picture available new form task confronted data mining machine learning researcher Suggestions picture relevant content review could s...
[0.06192536652088165, 0.02066613733768463, -0.0245407372713089, -0.012043209746479988, 0.0009616810129955411, 0.021104011684656143, 0.03494877740740776, 0.03164412081241608, -0.006985929794609547, -0.0476079061627388, -0.026939189061522484, 0.03745461255311966, 0.0016328382771462202, 0.13056330382823944, -0.00723130442...
318
318
['Matthias Scholz']
1204.0684v1
Linear principal component analysis (PCA) can be extended to a nonlinear PCA by using artificial neural networks. But the benefit of curved components requires a careful control of the model complexity. Moreover, standard techniques for model selection, including cross-validation and more generally the use of an indepe...
Validation of nonlinear PCA
2,012
http://arxiv.org/pdf/1204.0684v1
Title Validation nonlinear PCA Summary Linear principal component analysis PCA extended nonlinear PCA using artificial neural network benefit curved component requires careful control model complexity Moreover standard technique model selection including crossvalidation generally use independent test set fail applied n...
[-0.004732702858746052, 0.03236550837755203, -0.03606225922703743, 0.009588680230081081, 0.045808855444192886, 0.0013039211044088006, 0.004947184585034847, -0.015589880757033825, 0.0017258682055398822, 0.04779564589262009, 0.06112824007868767, 0.025742240250110626, 0.054516520351171494, 0.05529404431581497, 0.028260655...
319
319
['Ethan Fetaya', 'Ohad Shamir', 'Shimon Ullman']
1406.2602v1
We consider the problem of learning from a similarity matrix (such as spectral clustering and lowd imensional embedding), when computing pairwise similarities are costly, and only a limited number of entries can be observed. We provide a theoretical analysis using standard notions of graph approximation, significantly ...
Graph Approximation and Clustering on a Budget
2,014
http://arxiv.org/pdf/1406.2602v1
Title Graph Approximation Clustering Budget Summary consider problem learning similarity matrix spectral clustering lowd imensional embedding computing pairwise similarity costly limited number entry observed provide theoretical analysis using standard notion graph approximation significantly generalizing previous resu...
[-0.035050295293331146, -0.040311697870492935, -0.012310556136071682, 0.040419816970825195, -0.0392531156539917, -0.019893303513526917, -0.0028460719622671604, 0.025596272200345993, 0.062237318605184555, -0.00827780831605196, -0.01603635773062706, 0.03985195234417915, 0.016865190118551254, -0.007200275082141161, 0.0191...
320
320
['Shai Shalev-Shwartz', 'Yonatan Wexler', 'Amnon Shashua']
1109.0820v1
Multiclass prediction is the problem of classifying an object into a relevant target class. We consider the problem of learning a multiclass predictor that uses only few features, and in particular, the number of used features should increase sub-linearly with the number of possible classes. This implies that features ...
ShareBoost: Efficient Multiclass Learning with Feature Sharing
2,011
http://arxiv.org/pdf/1109.0820v1
Title ShareBoost Efficient Multiclass Learning Feature Sharing Summary Multiclass prediction problem classifying object relevant target class consider problem learning multiclass predictor us feature particular number used feature increase sublinearly number possible class implies feature shared several class describe ...
[-0.024203041568398476, 0.049770817160606384, -0.012178342789411545, 0.006836077198386192, -0.0008035636856220663, -0.022024810314178467, 0.06082786247134209, 0.016924966126680374, -0.03441485017538071, -0.039819348603487015, 0.012645447626709938, 0.006126794032752514, -0.02052612416446209, 0.07946121692657471, -0.0219...
321
321
['Nan Lin', 'Junhai Jiang', 'Shicheng Guo', 'Momiao Xiong']
1408.0204v1
Due to advances in sensors, growing large and complex medical image data have the ability to visualize the pathological change in the cellular or even the molecular level or anatomical changes in tissues and organs. As a consequence, the medical images have the potential to enhance diagnosis of disease, prediction of c...
Functional Principal Component Analysis and Randomized Sparse Clustering Algorithm for Medical Image Analysis
2,014
http://arxiv.org/pdf/1408.0204v1
Title Functional Principal Component Analysis Randomized Sparse Clustering Algorithm Medical Image Analysis Summary Due advance sensor growing large complex medical image data ability visualize pathological change cellular even molecular level anatomical change tissue organ consequence medical image potential enhance d...
[-0.002297246130183339, -0.005757798440754414, -0.03162936493754387, -0.0022903424687683582, 0.019461993128061295, 0.02461056225001812, 0.03363807499408722, 0.042401023209095, 0.03999677300453186, 0.06328421086072922, 0.014441088773310184, -0.02556086704134941, 0.04983000457286835, 0.03338916599750519, 0.01118294708430...
322
322
['Liwen Zhang', 'Subhransu Maji', 'Ryota Tomioka']
1503.01521v3
Similarity between objects is multi-faceted and it can be easier for human annotators to measure it when the focus is on a specific aspect. We consider the problem of mapping objects into view-specific embeddings where the distance between them is consistent with the similarity comparisons of the form "from the t-th vi...
Jointly Learning Multiple Measures of Similarities from Triplet Comparisons
2,015
http://arxiv.org/pdf/1503.01521v3
Title Jointly Learning Multiple Measures Similarities Triplet Comparisons Summary Similarity object multifaceted easier human annotator measure focus specific aspect consider problem mapping object viewspecific embeddings distance consistent similarity comparison form tth view object similar B C framework jointly learn...
[-0.027166103944182396, 0.041536830365657806, 0.021815259009599686, 0.029312513768672943, 0.005949103739112616, 0.036198556423187256, 0.03107762150466442, 0.0023092380724847317, 0.007253725081682205, -0.0360211506485939, -0.03544493019580841, 0.027011608704924583, -0.001517869415692985, -0.017213789746165276, 0.0248469...
323
323
['Andreas C. Damianou', 'Michalis K. Titsias', 'Neil D. Lawrence']
1409.2287v1
The Gaussian process latent variable model (GP-LVM) provides a flexible approach for non-linear dimensionality reduction that has been widely applied. However, the current approach for training GP-LVMs is based on maximum likelihood, where the latent projection variables are maximized over rather than integrated out. I...
Variational Inference for Uncertainty on the Inputs of Gaussian Process Models
2,014
http://arxiv.org/pdf/1409.2287v1
Title Variational Inference Uncertainty Inputs Gaussian Process Models Summary Gaussian process latent variable model GPLVM provides flexible approach nonlinear dimensionality reduction widely applied However current approach training GPLVMs based maximum likelihood latent projection variable maximized rather integrate...
[-0.01661471277475357, 0.0767054408788681, -0.0016049164114519954, 0.008719809353351593, 0.0033447102177888155, 0.0007342900498770177, 0.0016128488350659609, -0.02853258326649666, -0.07877147942781448, 0.030441954731941223, 0.04057313874363899, -0.006728356704115868, 0.01299707219004631, 0.08199609071016312, 0.02946901...
324
324
['Mehdi Mirza', 'Simon Osindero']
1411.1784v1
Generative Adversarial Nets [8] were recently introduced as a novel way to train generative models. In this work we introduce the conditional version of generative adversarial nets, which can be constructed by simply feeding the data, y, we wish to condition on to both the generator and discriminator. We show that this...
Conditional Generative Adversarial Nets
2,014
http://arxiv.org/pdf/1411.1784v1
Title Conditional Generative Adversarial Nets Summary Generative Adversarial Nets 8 recently introduced novel way train generative model work introduce conditional version generative adversarial net constructed simply feeding data wish condition generator discriminator show model generate MNIST digit conditioned class ...
[0.02483746036887169, 0.06669881194829941, 0.0025523377116769552, 0.036323923617601395, 0.007651783991605043, -0.009663864970207214, 0.05208682641386986, 0.01105070672929287, 0.017810167744755745, -0.014641670510172844, 0.016967596486210823, 0.0031483510974794626, -0.04884492978453636, 0.042640913277864456, 0.034193415...
325
325
['Krzysztof Chalupka', 'Pietro Perona', 'Frederick Eberhardt']
1412.2309v2
We provide a rigorous definition of the visual cause of a behavior that is broadly applicable to the visually driven behavior in humans, animals, neurons, robots and other perceiving systems. Our framework generalizes standard accounts of causal learning to settings in which the causal variables need to be constructed ...
Visual Causal Feature Learning
2,014
http://arxiv.org/pdf/1412.2309v2
Title Visual Causal Feature Learning Summary provide rigorous definition visual cause behavior broadly applicable visually driven behavior human animal neuron robot perceiving system framework generalizes standard account causal learning setting causal variable need constructed microvariables prove Causal Coarsening Th...
[-0.0047796741127967834, 0.010913015343248844, -0.02355905994772911, -0.037079911679029465, -0.020145250484347343, 0.015469438396394253, 0.020491614937782288, 0.012045727111399174, 0.005402937065809965, -0.005775235127657652, 0.019823789596557617, 0.10863538831472397, -0.0016414948040619493, 0.08719237148761749, 0.0548...
326
326
['Behnam Neyshabur', 'Ryota Tomioka', 'Nathan Srebro']
1412.6614v4
We present experiments demonstrating that some other form of capacity control, different from network size, plays a central role in learning multilayer feed-forward networks. We argue, partially through analogy to matrix factorization, that this is an inductive bias that can help shed light on deep learning.
In Search of the Real Inductive Bias: On the Role of Implicit Regularization in Deep Learning
2,014
http://arxiv.org/pdf/1412.6614v4
Title Search Real Inductive Bias Role Implicit Regularization Deep Learning Summary present experiment demonstrating form capacity control different network size play central role learning multilayer feedforward network argue partially analogy matrix factorization inductive bias help shed light deep learning Authors 0 ...
[-0.032525043934583664, 0.006899933330714703, -0.025447964668273926, 0.012960067950189114, 0.007396212313324213, -0.0015194164589047432, 0.02669672481715679, -0.003807682543992996, -0.027569405734539032, 0.05250503495335579, -0.03671678900718689, -0.030321545898914337, -0.01712300255894661, 0.06209549680352211, 0.06958...
327
327
['Muhammad Ghifary', 'W. Bastiaan Kleijn', 'Mengjie Zhang', 'David Balduzzi']
1508.07680v1
The problem of domain generalization is to take knowledge acquired from a number of related domains where training data is available, and to then successfully apply it to previously unseen domains. We propose a new feature learning algorithm, Multi-Task Autoencoder (MTAE), that provides good generalization performance ...
Domain Generalization for Object Recognition with Multi-task Autoencoders
2,015
http://arxiv.org/pdf/1508.07680v1
Title Domain Generalization Object Recognition Multitask Autoencoders Summary problem domain generalization take knowledge acquired number related domain training data available successfully apply previously unseen domain propose new feature learning algorithm MultiTask Autoencoder MTAE provides good generalization per...
[-0.024374762549996376, 0.03168083727359772, -0.026710782200098038, 0.02303546667098999, 0.020019298419356346, 0.0238779429346323, 0.0904061496257782, -0.003569613676518202, -0.028121083974838257, -0.03811495378613472, -0.05400795489549637, -0.021186141297221184, -0.021273376420140266, 0.07549112290143967, 0.0344295278...
328
328
['John-Alexander M. Assael', 'Niklas Wahlström', 'Thomas B. Schön', 'Marc Peter Deisenroth']
1510.02173v2
Data-efficient reinforcement learning (RL) in continuous state-action spaces using very high-dimensional observations remains a key challenge in developing fully autonomous systems. We consider a particularly important instance of this challenge, the pixels-to-torques problem, where an RL agent learns a closed-loop con...
Data-Efficient Learning of Feedback Policies from Image Pixels using Deep Dynamical Models
2,015
http://arxiv.org/pdf/1510.02173v2
Title DataEfficient Learning Feedback Policies Image Pixels using Deep Dynamical Models Summary Dataefficient reinforcement learning RL continuous stateaction space using highdimensional observation remains key challenge developing fully autonomous system consider particularly important instance challenge pixelstotorqu...
[-0.003288840176537633, 0.01718503050506115, -0.008624612353742123, 0.03099151700735092, -0.009667797945439816, 0.008136320859193802, 0.024760344997048378, -0.004043809603899717, -0.055324871093034744, 0.010250277817249298, 0.00019719271222129464, 0.020453568547964096, -0.03006250038743019, 0.0509597547352314, 0.042944...
329
329
['Muhammad Ghifary', 'David Balduzzi', 'W. Bastiaan Kleijn', 'Mengjie Zhang']
1510.04373v2
This paper addresses classification tasks on a particular target domain in which labeled training data are only available from source domains different from (but related to) the target. Two closely related frameworks, domain adaptation and domain generalization, are concerned with such tasks, where the only difference ...
Scatter Component Analysis: A Unified Framework for Domain Adaptation and Domain Generalization
2,015
http://arxiv.org/pdf/1510.04373v2
Title Scatter Component Analysis Unified Framework Domain Adaptation Domain Generalization Summary paper address classification task particular target domain labeled training data available source domain different related target Two closely related framework domain adaptation domain generalization concerned task differ...
[-0.027125462889671326, -0.02980937995016575, -0.02665247768163681, 0.0053611998446285725, 0.03437906876206398, 0.014813361689448357, 0.08279822766780853, -0.0020676180720329285, -0.057248033583164215, -0.025099707767367363, -0.010397735051810741, -0.017613111063838005, 0.03673088923096657, 0.052566204220056534, -0.000...
330
330
['Zhao Kang', 'Chong Peng', 'Qiang Cheng']
1510.08971v1
Matrix rank minimization problem is in general NP-hard. The nuclear norm is used to substitute the rank function in many recent studies. Nevertheless, the nuclear norm approximation adds all singular values together and the approximation error may depend heavily on the magnitudes of singular values. This might restrict...
Robust Subspace Clustering via Tighter Rank Approximation
2,015
http://arxiv.org/pdf/1510.08971v1
Title Robust Subspace Clustering via Tighter Rank Approximation Summary Matrix rank minimization problem general NPhard nuclear norm used substitute rank function many recent study Nevertheless nuclear norm approximation add singular value together approximation error may depend heavily magnitude singular value might r...
[-0.0073812054470181465, 0.02380376122891903, -0.031976472586393356, 0.05247168987989426, 0.034567274153232574, 0.01644269749522209, 0.01957681030035019, 0.06488088518381119, 0.04088689759373665, 0.034004077315330505, 0.011404366232454777, 0.0055304719135165215, 0.017084890976548195, 0.007732492871582508, 0.02722683362...
331
331
['Amogh Gudi']
1512.00743v2
The human face constantly conveys information, both consciously and subconsciously. However, as basic as it is for humans to visually interpret this information, it is quite a big challenge for machines. Conventional semantic facial feature recognition and analysis techniques are already in use and are based on physiol...
Recognizing Semantic Features in Faces using Deep Learning
2,015
http://arxiv.org/pdf/1512.00743v2
Title Recognizing Semantic Features Faces using Deep Learning Summary human face constantly conveys information consciously subconsciously However basic human visually interpret information quite big challenge machine Conventional semantic facial feature recognition analysis technique already use based physiological he...
[0.01549519132822752, 0.006907382048666477, 0.010920019820332527, 0.05677114427089691, 0.016417087987065315, 0.05331089347600937, 0.014811868779361248, -0.0002497587993275374, 0.0006581523921340704, -0.011993249878287315, -0.017399411648511887, 0.012742520309984684, 0.016883907839655876, 0.0881194919347763, 0.008204822...
332
332
['Muhammad Ghifary', 'W. Bastiaan Kleijn', 'Mengjie Zhang', 'David Balduzzi', 'Wen Li']
1607.03516v2
In this paper, we propose a novel unsupervised domain adaptation algorithm based on deep learning for visual object recognition. Specifically, we design a new model called Deep Reconstruction-Classification Network (DRCN), which jointly learns a shared encoding representation for two tasks: i) supervised classification...
Deep Reconstruction-Classification Networks for Unsupervised Domain Adaptation
2,016
http://arxiv.org/pdf/1607.03516v2
Title Deep ReconstructionClassification Networks Unsupervised Domain Adaptation Summary paper propose novel unsupervised domain adaptation algorithm based deep learning visual object recognition Specifically design new model called Deep ReconstructionClassification Network DRCN jointly learns shared encoding representa...
[-0.0274956151843071, 0.05599018186330795, -0.002992043038830161, 0.05921749770641327, 0.014201276004314423, 0.01939086616039276, 0.03933119401335716, -0.010237772017717361, -0.04541082680225372, -0.03239825367927551, -0.045764897018671036, 0.01125722099095583, -0.00964750349521637, 0.05330250412225723, -0.007811758201...
333
333
['Po-Hsuan Chen', 'Xia Zhu', 'Hejia Zhang', 'Javier S. Turek', 'Janice Chen', 'Theodore L. Willke', 'Uri Hasson', 'Peter J. Ramadge']
1608.04846v1
Finding the most effective way to aggregate multi-subject fMRI data is a long-standing and challenging problem. It is of increasing interest in contemporary fMRI studies of human cognition due to the scarcity of data per subject and the variability of brain anatomy and functional response across subjects. Recent work o...
A Convolutional Autoencoder for Multi-Subject fMRI Data Aggregation
2,016
http://arxiv.org/pdf/1608.04846v1
Title Convolutional Autoencoder MultiSubject fMRI Data Aggregation Summary Finding effective way aggregate multisubject fMRI data longstanding challenging problem increasing interest contemporary fMRI study human cognition due scarcity data per subject variability brain anatomy functional response across subject Recent...
[-0.02918638475239277, 0.11094777286052704, -0.022762315347790718, 0.03536400571465492, 0.01663896255195141, 0.04999234899878502, 0.10483603924512863, 0.008796414360404015, -0.045401688665151596, 0.0467711016535759, -0.05497492849826813, -0.048663605004549026, 0.03624100238084793, 0.10474361479282379, 0.058560971170663...
334
334
['Yun Wang', 'Xu Chen', 'Peter J. Ramadge']
1608.06010v2
One way to solve lasso problems when the dictionary does not fit into available memory is to first screen the dictionary to remove unneeded features. Prior research has shown that sequential screening methods offer the greatest promise in this endeavor. Most existing work on sequential screening targets the context of ...
Feedback-Controlled Sequential Lasso Screening
2,016
http://arxiv.org/pdf/1608.06010v2
Title FeedbackControlled Sequential Lasso Screening Summary One way solve lasso problem dictionary fit available memory first screen dictionary remove unneeded feature Prior research shown sequential screening method offer greatest promise endeavor existing work sequential screening target context tuning parameter sele...
[0.0072308434173464775, 0.06638697534799576, -0.012301868759095669, -0.020716991275548935, 0.03417609632015228, 0.009785800240933895, 0.038112539798021317, 0.04079483821988106, 0.04703814163804054, -0.024393374100327492, 0.016419949010014534, 0.02853362262248993, 0.0038744346238672733, 0.046383678913116455, 0.023715449...
335
335
['Yun Wang', 'Peter J. Ramadge']
1608.06014v2
Recently dictionary screening has been proposed as an effective way to improve the computational efficiency of solving the lasso problem, which is one of the most commonly used method for learning sparse representations. To address today's ever increasing large dataset, effective screening relies on a tight region boun...
The Symmetry of a Simple Optimization Problem in Lasso Screening
2,016
http://arxiv.org/pdf/1608.06014v2
Title Symmetry Simple Optimization Problem Lasso Screening Summary Recently dictionary screening proposed effective way improve computational efficiency solving lasso problem one commonly used method learning sparse representation address today ever increasing large dataset effective screening relies tight region bound...
[-0.0346694178879261, 0.02484263852238655, -0.003674357198178768, 0.031780585646629333, 0.016243983060121536, 0.013051044195890427, 0.014753670431673527, 0.010811420157551765, 0.03335801139473915, -0.024480551481246948, 0.01261131465435028, 0.0030675269663333893, 0.0026696978602558374, 0.03679244965314865, 0.0142356306...
336
336
['Maxime Bucher', 'Stéphane Herbin', 'Frédéric Jurie']
1608.07441v1
Zero-Shot learning has been shown to be an efficient strategy for domain adaptation. In this context, this paper builds on the recent work of Bucher et al. [1], which proposed an approach to solve Zero-Shot classification problems (ZSC) by introducing a novel metric learning based objective function. This objective fun...
Hard Negative Mining for Metric Learning Based Zero-Shot Classification
2,016
http://arxiv.org/pdf/1608.07441v1
Title Hard Negative Mining Metric Learning Based ZeroShot Classification Summary ZeroShot learning shown efficient strategy domain adaptation context paper build recent work Bucher et al 1 proposed approach solve ZeroShot classification problem ZSC introducing novel metric learning based objective function objective fu...
[-0.06956709176301956, -0.04282480478286743, -0.005442819558084011, 0.011128395795822144, -0.0002812288876157254, 0.0005613146349787712, 0.023711450397968292, 0.03313487023115158, -0.02153163030743599, -0.03498203679919243, -0.02448679693043232, 8.437623182544485e-05, -0.009236202575266361, 0.014643154107034206, 0.0095...
337
337
['Xiang Xiang', 'Trac D. Tran']
1609.07042v4
In this paper, we deal with two challenges for measuring the similarity of the subject identities in practical video-based face recognition - the variation of the head pose in uncontrolled environments and the computational expense of processing videos. Since the frame-wise feature mean is unable to characterize the po...
Pose-Selective Max Pooling for Measuring Similarity
2,016
http://arxiv.org/pdf/1609.07042v4
Title PoseSelective Max Pooling Measuring Similarity Summary paper deal two challenge measuring similarity subject identity practical videobased face recognition variation head pose uncontrolled environment computational expense processing video Since framewise feature mean unable characterize pose diversity among fram...
[-0.03755255788564682, 0.02456127665936947, 0.0022203100379556417, 0.05058281868696213, 0.0008902945555746555, 0.04240700602531433, 0.05976670980453491, 0.005668295081704855, -0.009141220711171627, -0.00640525110065937, 0.02521650679409504, -0.029562024399638176, 0.028338605538010597, 0.01709176041185856, 0.02508774399...
338
338
['Shehroz S. Khan', 'Babak Taati']
1610.03761v3
A fall is an abnormal activity that occurs rarely, so it is hard to collect real data for falls. It is, therefore, difficult to use supervised learning methods to automatically detect falls. Another challenge in using machine learning methods to automatically detect falls is the choice of engineered features. In this p...
Detecting Unseen Falls from Wearable Devices using Channel-wise Ensemble of Autoencoders
2,016
http://arxiv.org/pdf/1610.03761v3
Title Detecting Unseen Falls Wearable Devices using Channelwise Ensemble Autoencoders Summary fall abnormal activity occurs rarely hard collect real data fall therefore difficult use supervised learning method automatically detect fall Another challenge using machine learning method automatically detect fall choice eng...
[-0.07382499426603317, -0.013315945863723755, -0.04300755262374878, 0.025670094415545464, 0.055513981729745865, -0.007175646722316742, 0.024653678759932518, -0.001905350130982697, -0.00956262368708849, -0.030558494850993156, 0.08103613555431366, 0.03480686992406845, 0.03531024605035782, 0.07011061161756516, 0.038253977...
339
339
['Jure Sokolic', 'Raja Giryes', 'Guillermo Sapiro', 'Miguel R. D. Rodrigues']
1610.04574v3
This paper studies the generalization error of invariant classifiers. In particular, we consider the common scenario where the classification task is invariant to certain transformations of the input, and that the classifier is constructed (or learned) to be invariant to these transformations. Our approach relies on fa...
Generalization Error of Invariant Classifiers
2,016
http://arxiv.org/pdf/1610.04574v3
Title Generalization Error Invariant Classifiers Summary paper study generalization error invariant classifier particular consider common scenario classification task invariant certain transformation input classifier constructed learned invariant transformation approach relies factoring input space product base space s...
[-0.0070000262930989265, 0.025901639834046364, -0.0019966927357017994, 0.009904432110488415, -0.006227373145520687, 0.01137002557516098, 0.029324378818273544, 0.00957962404936552, -0.05482788756489754, -0.015758013352751732, 0.04085661843419075, 0.01983199454843998, -0.005337031092494726, 0.01314003299921751, 0.0235902...
340
340
['Seyed-Mohsen Moosavi-Dezfooli', 'Alhussein Fawzi', 'Omar Fawzi', 'Pascal Frossard']
1610.08401v3
Given a state-of-the-art deep neural network classifier, we show the existence of a universal (image-agnostic) and very small perturbation vector that causes natural images to be misclassified with high probability. We propose a systematic algorithm for computing universal perturbations, and show that state-of-the-art ...
Universal adversarial perturbations
2,016
http://arxiv.org/pdf/1610.08401v3
Title Universal adversarial perturbation Summary Given stateoftheart deep neural network classifier show existence universal imageagnostic small perturbation vector cause natural image misclassified high probability propose systematic algorithm computing universal perturbation show stateoftheart deep neural network hig...
[-0.02164355292916298, 0.03978942707180977, -0.02774822898209095, 0.014536814764142036, -0.02070380188524723, -0.01888200268149376, 0.012238597497344017, -0.008295377716422081, -0.01625387743115425, 0.0015916165430098772, 0.005666621029376984, -0.0056359716691076756, -0.004936543758958578, 0.023915022611618042, 0.07806...
341
341
['Xiang Xiang', 'Trac D. Tran']
1701.03102v1
Limited annotated data available for the recognition of facial expression and action units embarrasses the training of deep networks, which can learn disentangled invariant features. However, a linear model with just several parameters normally is not demanding in terms of training data. In this paper, we propose an el...
Linear Disentangled Representation Learning for Facial Actions
2,017
http://arxiv.org/pdf/1701.03102v1
Title Linear Disentangled Representation Learning Facial Actions Summary Limited annotated data available recognition facial expression action unit embarrasses training deep network learn disentangled invariant feature However linear model several parameter normally demanding term training data paper propose elegant li...
[-0.04392202943563461, 0.0781177431344986, -0.0037950545083731413, 0.05190938338637352, 0.018013276159763336, 0.07468525320291519, 0.01124053169041872, 0.013821710832417011, 0.018731869757175446, -0.04951053857803345, -0.008052118122577667, 0.002034249948337674, -0.013612786307930946, 0.10846571624279022, 0.03620795533...
342
342
['Jan Hendrik Metzen', 'Tim Genewein', 'Volker Fischer', 'Bastian Bischoff']
1702.04267v2
Machine learning and deep learning in particular has advanced tremendously on perceptual tasks in recent years. However, it remains vulnerable against adversarial perturbations of the input that have been crafted specifically to fool the system while being quasi-imperceptible to a human. In this work, we propose to aug...
On Detecting Adversarial Perturbations
2,017
http://arxiv.org/pdf/1702.04267v2
Title Detecting Adversarial Perturbations Summary Machine learning deep learning particular advanced tremendously perceptual task recent year However remains vulnerable adversarial perturbation input crafted specifically fool system quasiimperceptible human work propose augment deep neural network small detector subnet...
[0.009992479346692562, 0.06535062938928604, -0.02379775047302246, 0.05759544298052788, 0.0002230452955700457, -0.01433858834207058, 0.02848367765545845, -0.007302304729819298, -0.00942637212574482, -0.029911896213889122, -0.009395372122526169, 0.036857280880212784, 0.0065127466805279255, 0.02342786267399788, 0.05856072...
343
343
['Zhiming Zhou', 'Han Cai', 'Shu Rong', 'Yuxuan Song', 'Kan Ren', 'Weinan Zhang', 'Yong Yu', 'Jun Wang']
1703.02000v7
Class labels have been empirically shown useful in improving the sample quality of generative adversarial nets (GANs). In this paper, we mathematically study the properties of the current variants of GANs that make use of class label information. With class aware gradient and cross-entropy decomposition, we reveal how ...
Activation Maximization Generative Adversarial Nets
2,017
http://arxiv.org/pdf/1703.02000v7
Title Activation Maximization Generative Adversarial Nets Summary Class label empirically shown useful improving sample quality generative adversarial net GANs paper mathematically study property current variant GANs make use class label information class aware gradient crossentropy decomposition reveal class label ass...
[-0.007207329850643873, 0.09048110246658325, 0.004069111775606871, 0.015610373578965664, 0.020884061232209206, -0.01440146192908287, 0.0439172200858593, 0.01177096925675869, -0.024044563993811607, 0.011028407141566277, -0.03259025141596794, 0.012897496111690998, -0.014729304239153862, 0.02782369963824749, 0.04626112803...
344
344
['Ruth Fong', 'Andrea Vedaldi']
1704.03296v3
As machine learning algorithms are increasingly applied to high impact yet high risk tasks, such as medical diagnosis or autonomous driving, it is critical that researchers can explain how such algorithms arrived at their predictions. In recent years, a number of image saliency methods have been developed to summarize ...
Interpretable Explanations of Black Boxes by Meaningful Perturbation
2,017
http://arxiv.org/pdf/1704.03296v3
Title Interpretable Explanations Black Boxes Meaningful Perturbation Summary machine learning algorithm increasingly applied high impact yet high risk task medical diagnosis autonomous driving critical researcher explain algorithm arrived prediction recent year number image saliency method developed summarize highly co...
[-0.0006938378792256117, -0.02330688014626503, -0.016262996941804886, 0.026938078925013542, -0.0102374954149127, 0.023205429315567017, 0.020958315581083298, 0.042662255465984344, -0.03826913610100746, -0.008606581017374992, 0.07605795562267303, 0.035809777677059174, 0.004910981748253107, 0.0569586418569088, 0.021658621...
345
345
['Hong Zhao']
1704.06885v1
Though the deep learning is pushing the machine learning to a new stage, basic theories of machine learning are still limited. The principle of learning, the role of the a prior knowledge, the role of neuron bias, and the basis for choosing neural transfer function and cost function, etc., are still far from clear. In ...
A General Theory for Training Learning Machine
2,017
http://arxiv.org/pdf/1704.06885v1
Title General Theory Training Learning Machine Summary Though deep learning pushing machine learning new stage basic theory machine learning still limited principle learning role prior knowledge role neuron bias basis choosing neural transfer function cost function etc still far clear paper present general theoretical ...
[0.012812108732759953, -0.01811564527451992, -0.033388908952474594, 0.05551278963685036, 0.019576644524931908, -0.003350692568346858, 0.0046074045822024345, 0.00016276097449008375, -0.016794392839074135, 0.00841854140162468, 0.010855269618332386, 0.001528909895569086, 0.022682560607790947, 0.04476592689752579, 0.040381...
346
346
['Yotam Hechtlinger', 'Purvasha Chakravarti', 'Jining Qin']
1704.08165v1
This paper introduces a generalization of Convolutional Neural Networks (CNNs) from low-dimensional grid data, such as images, to graph-structured data. We propose a novel spatial convolution utilizing a random walk to uncover the relations within the input, analogous to the way the standard convolution uses the spatia...
A Generalization of Convolutional Neural Networks to Graph-Structured Data
2,017
http://arxiv.org/pdf/1704.08165v1
Title Generalization Convolutional Neural Networks GraphStructured Data Summary paper introduces generalization Convolutional Neural Networks CNNs lowdimensional grid data image graphstructured data propose novel spatial convolution utilizing random walk uncover relation within input analogous way standard convolution ...
[0.03470451012253761, 0.039715345948934555, -0.007747909985482693, 0.010968849994242191, -0.008142500184476376, -0.01994643546640873, 0.025686796754598618, 0.02761830948293209, 0.047427188605070114, 0.03826754167675972, 0.04872293770313263, 0.03953459486365318, 0.0011169301578775048, 0.05940108373761177, 0.054836910218...
347
347
['Matthias Hein', 'Maksym Andriushchenko']
1705.08475v2
Recent work has shown that state-of-the-art classifiers are quite brittle, in the sense that a small adversarial change of an originally with high confidence correctly classified input leads to a wrong classification again with high confidence. This raises concerns that such classifiers are vulnerable to attacks and ca...
Formal Guarantees on the Robustness of a Classifier against Adversarial Manipulation
2,017
http://arxiv.org/pdf/1705.08475v2
Title Formal Guarantees Robustness Classifier Adversarial Manipulation Summary Recent work shown stateoftheart classifier quite brittle sense small adversarial change originally high confidence correctly classified input lead wrong classification high confidence raise concern classifier vulnerable attack call question ...
[0.006188260857015848, 0.04354482144117355, -0.013780206441879272, 0.004464882891625166, 0.01014008466154337, -0.04665730521082878, 0.0035088288132101297, -0.0023654901888221502, -0.0012604391667991877, -0.027245502918958664, 0.05376371368765831, 0.04295532405376434, 0.007749683689326048, 0.053151994943618774, 0.067649...
348
348
['Alhussein Fawzi', 'Seyed-Mohsen Moosavi-Dezfooli', 'Pascal Frossard', 'Stefano Soatto']
1705.09552v1
The goal of this paper is to analyze the geometric properties of deep neural network classifiers in the input space. We specifically study the topology of classification regions created by deep networks, as well as their associated decision boundary. Through a systematic empirical investigation, we show that state-of-t...
Classification regions of deep neural networks
2,017
http://arxiv.org/pdf/1705.09552v1
Title Classification region deep neural network Summary goal paper analyze geometric property deep neural network classifier input space specifically study topology classification region created deep network well associated decision boundary systematic empirical investigation show stateoftheart deep net learn connected...
[-0.020803967490792274, 0.011388389393687248, -0.06109420582652092, 0.05848638340830803, -0.03571015223860741, -0.02050507441163063, 0.02685772068798542, -0.03100939467549324, -0.012953218072652817, 0.0019034185679629445, 0.002322569489479065, 0.04575680196285248, 0.010220460593700409, 0.041609954088926315, 0.043501451...
349
349
['Seyed-Mohsen Moosavi-Dezfooli', 'Alhussein Fawzi', 'Omar Fawzi', 'Pascal Frossard', 'Stefano Soatto']
1705.09554v1
Deep networks have recently been shown to be vulnerable to universal perturbations: there exist very small image-agnostic perturbations that cause most natural images to be misclassified by such classifiers. In this paper, we propose the first quantitative analysis of the robustness of classifiers to universal perturba...
Analysis of universal adversarial perturbations
2,017
http://arxiv.org/pdf/1705.09554v1
Title Analysis universal adversarial perturbation Summary Deep network recently shown vulnerable universal perturbation exist small imageagnostic perturbation cause natural image misclassified classifier paper propose first quantitative analysis robustness classifier universal perturbation draw formal link robustness u...
[-0.006989311892539263, 0.007897590287029743, -0.027320239692926407, 0.018769506365060806, -0.02717568166553974, -0.01606050319969654, 0.010597921907901764, -0.03947088122367859, -0.009403351694345474, -0.005333918612450361, 0.013914406299591064, 0.013099141418933868, -0.0457087904214859, 0.029455699026584625, 0.065333...
350
350
['Yunus Saatchi', 'Andrew Gordon Wilson']
1705.09558v3
Generative adversarial networks (GANs) can implicitly learn rich distributions over images, audio, and data which are hard to model with an explicit likelihood. We present a practical Bayesian formulation for unsupervised and semi-supervised learning with GANs. Within this framework, we use stochastic gradient Hamilton...
Bayesian GAN
2,017
http://arxiv.org/pdf/1705.09558v3
Title Bayesian GAN Summary Generative adversarial network GANs implicitly learn rich distribution image audio data hard model explicit likelihood present practical Bayesian formulation unsupervised semisupervised learning GANs Within framework use stochastic gradient Hamiltonian Monte Carlo marginalize weight generator...
[-0.02308654971420765, 0.08587949723005295, -0.019761765375733376, 0.004068419802933931, 0.02050447277724743, -0.004102040082216263, 0.04403464123606682, -0.020683765411376953, -0.06667576730251312, 0.03175995126366615, -0.07144792377948761, 0.027946999296545982, -0.0347418487071991, 0.02525789849460125, 0.063532970845...
351
351
['Emily Denton', 'Vighnesh Birodkar']
1705.10915v1
We present a new model DrNET that learns disentangled image representations from video. Our approach leverages the temporal coherence of video and a novel adversarial loss to learn a representation that factorizes each frame into a stationary part and a temporally varying component. The disentangled representation can ...
Unsupervised Learning of Disentangled Representations from Video
2,017
http://arxiv.org/pdf/1705.10915v1
Title Unsupervised Learning Disentangled Representations Video Summary present new model DrNET learns disentangled image representation video approach leverage temporal coherence video novel adversarial loss learn representation factorizes frame stationary part temporally varying component disentangled representation u...
[-0.04752199724316597, 0.05794048309326172, 0.012145565822720528, 0.058078933507204056, -0.018604572862386703, -0.01233634352684021, 0.012698106467723846, -0.006449037231504917, -0.018665088340640068, 0.003239475656300783, -0.006050102412700653, 0.016986601054668427, -0.006373200099915266, 0.07627525180578232, 0.009441...
352
352
['Yujia Li', 'Alexander Schwing', 'Kuan-Chieh Wang', 'Richard Zemel']
1706.06216v1
Generative adversarial nets (GANs) are a promising technique for modeling a distribution from samples. It is however well known that GAN training suffers from instability due to the nature of its maximin formulation. In this paper, we explore ways to tackle the instability problem by dualizing the discriminator. We sta...
Dualing GANs
2,017
http://arxiv.org/pdf/1706.06216v1
Title Dualing GANs Summary Generative adversarial net GANs promising technique modeling distribution sample however well known GAN training suffers instability due nature maximin formulation paper explore way tackle instability problem dualizing discriminator start linear discriminator case conjugate duality provides m...
[-0.0033893149811774492, 0.05406666174530983, -0.03474172204732895, 0.028156612068414688, 0.03303755074739456, -0.0018474722746759653, -0.01774638518691063, 0.0013656294904649258, -0.035452473908662796, 0.021841276437044144, -0.02383692003786564, -0.0471636988222599, -0.047075945883989334, -0.011637123301625252, 0.0765...
353
353
['Eunhee Kang', 'Jaejun Yoo', 'Jong Chul Ye']
1707.09938v2
Model based iterative reconstruction (MBIR) algorithms for low-dose X-ray CT are computationally expensive. To address this problem, we recently proposed the world-first deep convolutional neural network (CNN) for low-dose X-ray CT and won the second place in 2016 AAPM Low-Dose CT Grand Challenge. However, some of the ...
Wavelet Residual Network for Low-Dose CT via Deep Convolutional Framelets
2,017
http://arxiv.org/pdf/1707.09938v2
Title Wavelet Residual Network LowDose CT via Deep Convolutional Framelets Summary Model based iterative reconstruction MBIR algorithm lowdose Xray CT computationally expensive address problem recently proposed worldfirst deep convolutional neural network CNN lowdose Xray CT second place 2016 AAPM LowDose CT Grand Chal...
[-0.005884728394448757, 0.08598175644874573, 0.0025116964243352413, 0.047985225915908813, 0.011937020346522331, 0.01774267852306366, 0.018186010420322418, 0.025364933535456657, -0.014553194865584373, 0.0772213265299797, 0.009773721918463707, 0.03725576400756836, 0.0014747347449883819, -0.002036175923421979, 0.022917762...
354
354
['Chuhang Zou', 'Ersin Yumer', 'Jimei Yang', 'Duygu Ceylan', 'Derek Hoiem']
1708.01648v1
The success of various applications including robotics, digital content creation, and visualization demand a structured and abstract representation of the 3D world from limited sensor data. Inspired by the nature of human perception of 3D shapes as a collection of simple parts, we explore such an abstract shape represe...
3D-PRNN: Generating Shape Primitives with Recurrent Neural Networks
2,017
http://arxiv.org/pdf/1708.01648v1
Title 3DPRNN Generating Shape Primitives Recurrent Neural Networks Summary success various application including robotics digital content creation visualization demand structured abstract representation 3D world limited sensor data Inspired nature human perception 3D shape collection simple part explore abstract shape ...
[-0.026282045990228653, 0.011064124293625355, -0.002964281477034092, 0.03498736023902893, -0.014073737896978855, -0.0253475159406662, -0.004067199770361185, -0.04992694780230522, -0.08539851009845734, 0.02635801210999489, 0.0007618836243636906, 0.0014090482145547867, 0.011729761026799679, 0.07452578097581863, 0.0408184...
355
355
['Zhiming Zhou', 'Weinan Zhang', 'Jun Wang']
1708.01729v2
In this article, we mathematically study several GAN related topics, including Inception score, label smoothing, gradient vanishing and the -log(D(x)) alternative. --- An advanced version is included in arXiv:1703.02000 "Activation Maximization Generative Adversarial Nets". Please refer Section 6 in 1703.02000 for de...
Inception Score, Label Smoothing, Gradient Vanishing and -log(D(x)) Alternative
2,017
http://arxiv.org/pdf/1708.01729v2
Title Inception Score Label Smoothing Gradient Vanishing logDx Alternative Summary article mathematically study several GAN related topic including Inception score label smoothing gradient vanishing logDx alternative advanced version included arXiv170302000 Activation Maximization Generative Adversarial Nets Please ref...
[-0.007836745120584965, 0.05415726453065872, 0.00969302374869585, 0.05750613287091255, 0.0044973078183829784, -0.02896241657435894, 0.021426258608698845, 0.009973583742976189, -0.044108401983976364, 0.037905432283878326, 0.0024502864107489586, -0.016263682395219803, -0.01615750789642334, 0.02052057534456253, 0.04730376...
356
356
['Kai Arulkumaran', 'Marc Peter Deisenroth', 'Miles Brundage', 'Anil Anthony Bharath']
1708.05866v2
Deep reinforcement learning is poised to revolutionise the field of AI and represents a step towards building autonomous systems with a higher level understanding of the visual world. Currently, deep learning is enabling reinforcement learning to scale to problems that were previously intractable, such as learning to p...
A Brief Survey of Deep Reinforcement Learning
2,017
http://arxiv.org/pdf/1708.05866v2
Title Brief Survey Deep Reinforcement Learning Summary Deep reinforcement learning poised revolutionise field AI represents step towards building autonomous system higher level understanding visual world Currently deep learning enabling reinforcement learning scale problem previously intractable learning play video gam...
[0.001934720086865127, 0.018902035430073738, -0.0207810178399086, 0.024373536929488182, -0.006007027346640825, -0.008741470053792, 0.022436074912548065, -0.02780217118561268, -0.061189599335193634, 0.015256786718964577, -0.01792782172560692, 0.004106141161173582, 0.006279111839830875, 0.06601067632436752, -0.0008085505...
357
357
['Caiwen Ding', 'Siyu Liao', 'Yanzhi Wang', 'Zhe Li', 'Ning Liu', 'Youwei Zhuo', 'Chao Wang', 'Xuehai Qian', 'Yu Bai', 'Geng Yuan', 'Xiaolong Ma', 'Yipeng Zhang', 'Jian Tang', 'Qinru Qiu', 'Xue Lin', 'Bo Yuan']
1708.08917v1
Large-scale deep neural networks (DNNs) are both compute and memory intensive. As the size of DNNs continues to grow, it is critical to improve the energy efficiency and performance while maintaining accuracy. For DNNs, the model size is an important factor affecting performance, scalability and energy efficiency. Weig...
CirCNN: Accelerating and Compressing Deep Neural Networks Using Block-CirculantWeight Matrices
2,017
http://arxiv.org/pdf/1708.08917v1
Title CirCNN Accelerating Compressing Deep Neural Networks Using BlockCirculantWeight Matrices Summary Largescale deep neural network DNNs compute memory intensive size DNNs continues grow critical improve energy efficiency performance maintaining accuracy DNNs model size important factor affecting performance scalabil...
[-0.01497843861579895, 0.0228132251650095, -0.016226397827267647, 0.06257771700620651, 0.004769204184412956, -0.0003510774113237858, 0.060953062027692795, 0.006617339793592691, -0.005156007129698992, 0.024105293676257133, -0.02313764952123165, -0.01523275300860405, 0.015319127589464188, 0.003590119071304798, 0.01766840...
358
358
['Cătălina Cangea', 'Petar Veličković', 'Pietro Liò']
1709.00572v1
We propose two multimodal deep learning architectures that allow for cross-modal dataflow (XFlow) between the feature extractors, thereby extracting more interpretable features and obtaining a better representation than through unimodal learning, for the same amount of training data. These models can usefully exploit c...
XFlow: 1D-2D Cross-modal Deep Neural Networks for Audiovisual Classification
2,017
http://arxiv.org/pdf/1709.00572v1
Title XFlow 1D2D Crossmodal Deep Neural Networks Audiovisual Classification Summary propose two multimodal deep learning architecture allow crossmodal dataflow XFlow feature extractor thereby extracting interpretable feature obtaining better representation unimodal learning amount training data model usefully exploit c...
[-0.04427507147192955, 0.014811825007200241, -0.01971091330051422, 0.04842028394341469, 0.007753276266157627, -0.028473369777202606, 0.06911830604076385, -0.01682870276272297, -0.04300621524453163, -0.0021564385388046503, -0.12860608100891113, -0.002302121836692095, 0.01866266317665577, 0.06615891307592392, 0.028456566...
359
359
['Kun ho Kim', 'Oisin Mac Aodha', 'Pietro Perona']
1710.01691v2
Low dimensional embeddings that capture the main variations of interest in collections of data are important for many applications. One way to construct these embeddings is to acquire estimates of similarity from the crowd. However, similarity is a multi-dimensional concept that varies from individual to individual. Ex...
Context Embedding Networks
2,017
http://arxiv.org/pdf/1710.01691v2
Title Context Embedding Networks Summary Low dimensional embeddings capture main variation interest collection data important many application One way construct embeddings acquire estimate similarity crowd However similarity multidimensional concept varies individual individual Existing model learning embeddings crowd ...
[-0.03537415713071823, 0.06395672261714935, -0.007716844789683819, 0.07160276174545288, 0.008155053481459618, 0.007829046808183193, 0.04133368283510208, -0.00821962021291256, -0.03246283158659935, -0.010783545672893524, -0.029003407806158066, 0.027255209162831306, 0.002010752446949482, 0.006190483458340168, 0.068160429...
360
360
['Garrett B. Goh', 'Charles Siegel', 'Abhinav Vishnu', 'Nathan O. Hodas', 'Nathan Baker']
1710.02238v2
The meteoric rise of deep learning models in computer vision research, having achieved human-level accuracy in image recognition tasks is firm evidence of the impact of representation learning of deep neural networks. In the chemistry domain, recent advances have also led to the development of similar CNN models, such ...
How Much Chemistry Does a Deep Neural Network Need to Know to Make Accurate Predictions?
2,017
http://arxiv.org/pdf/1710.02238v2
Title Much Chemistry Deep Neural Network Need Know Make Accurate Predictions Summary meteoric rise deep learning model computer vision research achieved humanlevel accuracy image recognition task firm evidence impact representation learning deep neural network chemistry domain recent advance also led development simila...
[0.00770224817097187, 0.02630946971476078, -0.025033870711922646, 0.006477885879576206, 0.021847710013389587, -0.024519821628928185, 0.006756324786692858, 0.0017183911986649036, 0.04864497855305672, 0.02938755787909031, -0.009336560033261776, 0.013640801422297955, -0.040505681186914444, 0.05479798838496208, 0.036912634...
361
361
['Abhishek Kumar', 'Prasanna Sattigeri', 'Avinash Balakrishnan']
1711.00848v2
Disentangled representations, where the higher level data generative factors are reflected in disjoint latent dimensions, offer several benefits such as ease of deriving invariant representations, transferability to other tasks, interpretability, etc. We consider the problem of unsupervised learning of disentangled rep...
Variational Inference of Disentangled Latent Concepts from Unlabeled Observations
2,017
http://arxiv.org/pdf/1711.00848v2
Title Variational Inference Disentangled Latent Concepts Unlabeled Observations Summary Disentangled representation higher level data generative factor reflected disjoint latent dimension offer several benefit ease deriving invariant representation transferability task interpretability etc consider problem unsupervised...
[-0.013651109300553799, 0.13964641094207764, -0.014092152938246727, 0.04212312772870064, -0.01712259091436863, 0.02894134446978569, -0.0025744237937033176, -0.014820554293692112, -0.019702458754181862, 0.021822955459356308, 0.0009461340378038585, 0.010021849535405636, 0.025552650913596153, 0.060200463980436325, 0.03280...
362
362
['Stanisław Jastrzębski', 'Zachary Kenton', 'Devansh Arpit', 'Nicolas Ballas', 'Asja Fischer', 'Yoshua Bengio', 'Amos Storkey']
1711.04623v1
We study the properties of the endpoint of stochastic gradient descent (SGD). By approximating SGD as a stochastic differential equation (SDE) we consider the Boltzmann-Gibbs equilibrium distribution of that SDE under the assumption of isotropic variance in loss gradients. Through this analysis, we find that three fact...
Three Factors Influencing Minima in SGD
2,017
http://arxiv.org/pdf/1711.04623v1
Title Three Factors Influencing Minima SGD Summary study property endpoint stochastic gradient descent SGD approximating SGD stochastic differential equation SDE consider BoltzmannGibbs equilibrium distribution SDE assumption isotropic variance loss gradient analysis find three factor learning rate batch size variance ...
[-0.03010372631251812, -0.06074045971035957, -0.015608815476298332, 0.004908312577754259, 0.015153571963310242, -0.05740569531917572, 0.02932601235806942, 0.008304344490170479, -0.04873552918434143, 0.055902350693941116, 0.02767437882721424, 0.006699562072753906, -0.020996060222387314, 0.03697074204683304, -0.003464875...
363
363
['Paweł Liskowski', 'Wojciech Jaśkowski', 'Krzysztof Krawiec']
1711.06583v1
Achieving superhuman playing level by AlphaGo corroborated the capabilities of convolutional neural architectures (CNNs) for capturing complex spatial patterns. This result was to a great extent due to several analogies between Go board states and 2D images CNNs have been designed for, in particular translational invar...
Learning to Play Othello with Deep Neural Networks
2,017
http://arxiv.org/pdf/1711.06583v1
Title Learning Play Othello Deep Neural Networks Summary Achieving superhuman playing level AlphaGo corroborated capability convolutional neural architecture CNNs capturing complex spatial pattern result great extent due several analogy Go board state 2D image CNNs designed particular translational invariance relativel...
[-0.004043870139867067, 0.045120298862457275, -0.05259804427623749, 0.06701121479272842, 0.009386658668518066, -0.02731923572719097, 0.0049227941781282425, -0.0018305566627532244, -0.04357238486409187, 0.025756599381566048, -0.018135692924261093, -0.002616130979731679, -0.0027867441531270742, 0.059946559369564056, 0.07...
364
364
['Jaejun Yoo', 'Sohail Sabir', 'Duchang Heo', 'Kee Hyun Kim', 'Abdul Wahab', 'Yoonseok Choi', 'Seul-I Lee', 'Eun Young Chae', 'Hak Hee Kim', 'Young Min Bae', 'Young-wook Choi', 'Seungryong Cho', 'Jong Chul Ye']
1712.00912v1
Can artificial intelligence (AI) learn complicated non-linear physics? Here we propose a novel deep learning approach that learns non-linear photon scattering physics and obtains accurate 3D distribution of optical anomalies. In contrast to the traditional black-box deep learning approaches to inverse problems, our dee...
Deep Learning Can Reverse Photon Migration for Diffuse Optical Tomography
2,017
http://arxiv.org/pdf/1712.00912v1
Title Deep Learning Reverse Photon Migration Diffuse Optical Tomography Summary artificial intelligence AI learn complicated nonlinear physic propose novel deep learning approach learns nonlinear photon scattering physic obtains accurate 3D distribution optical anomaly contrast traditional blackbox deep learning approa...
[0.015387685038149357, -0.026141151785850525, -0.03536622226238251, -0.014373610727488995, -0.017385335639119148, -0.01864778995513916, 0.011410445906221867, 0.011899719946086407, -0.08374179899692535, 0.022748271003365517, 0.04173847660422325, -0.0041262502782046795, -0.022999459877610207, 0.016142629086971283, 0.0176...
365
365
['Garrett B. Goh', 'Charles Siegel', 'Abhinav Vishnu', 'Nathan O. Hodas']
1712.02734v2
With access to large datasets, deep neural networks (DNN) have achieved human-level accuracy in image and speech recognition tasks. However, in chemistry, data is inherently small and fragmented. In this work, we develop an approach of using rule-based knowledge for training ChemNet, a transferable and generalizable de...
Using Rule-Based Labels for Weak Supervised Learning: A ChemNet for Transferable Chemical Property Prediction
2,017
http://arxiv.org/pdf/1712.02734v2
Title Using RuleBased Labels Weak Supervised Learning ChemNet Transferable Chemical Property Prediction Summary access large datasets deep neural network DNN achieved humanlevel accuracy image speech recognition task However chemistry data inherently small fragmented work develop approach using rulebased knowledge trai...
[-0.01637466624379158, 0.09591254591941833, -0.015093043446540833, 0.01855919323861599, 0.022989120334386826, -0.03170975670218468, 0.04311203956604004, 0.013675128109753132, 0.04681979492306709, 0.0271495059132576, -0.05752216652035713, 0.023094531148672104, -0.04586184397339821, 0.049026962369680405, 0.01768498681485...
366
366
['Yeo Hun Yoon', 'Shujaat Khan', 'Jaeyoung Huh', 'Jong Chul Ye']
1712.06096v2
In portable, three dimensional, and ultra-fast ultrasound (US) imaging systems, there is an increasing need to reconstruct high quality images from a limited number of RF data from receiver (Rx) or scan-line (SC) sub-sampling. However, due to the severe side lobe artifacts from RF sub-sampling, the standard beam-former...
Deep Learning in RF Sub-sampled B-mode Ultrasound Imaging
2,017
http://arxiv.org/pdf/1712.06096v2
Title Deep Learning RF Subsampled Bmode Ultrasound Imaging Summary portable three dimensional ultrafast ultrasound US imaging system increasing need reconstruct high quality image limited number RF data receiver Rx scanline SC subsampling However due severe side lobe artifact RF subsampling standard beamformer often pr...
[0.0008131703943945467, 0.05844822898507118, -0.02203519642353058, -0.026201285421848297, 0.022137360647320747, 0.01236604992300272, 0.045752402395009995, 0.03850363940000534, -0.07252732664346695, 0.04685826599597931, -0.0035056264605373144, -0.07232800126075745, -0.003909144084900618, 0.04674515500664711, -0.00812863...
367
367
['Yoseob Han', 'Jawook Gu', 'Jong Chul Ye']
1712.10248v2
Interior tomography for the region-of-interest (ROI) imaging has advantages of using a small detector and reducing X-ray radiation dose. However, standard analytic reconstruction suffers from severe cupping artifacts due to existence of null space in the truncated Radon transform. Existing penalized reconstruction meth...
Deep Learning Interior Tomography for Region-of-Interest Reconstruction
2,017
http://arxiv.org/pdf/1712.10248v2
Title Deep Learning Interior Tomography RegionofInterest Reconstruction Summary Interior tomography regionofinterest ROI imaging advantage using small detector reducing Xray radiation dose However standard analytic reconstruction suffers severe cupping artifact due existence null space truncated Radon transform Existin...
[-0.03072839044034481, 0.03370160236954689, -0.01693260483443737, 0.005782017018646002, -0.020232217386364937, 0.005034395959228277, -0.019472967833280563, 0.06936164945363998, 0.013841195963323116, 0.0629819855093956, -0.04035986587405205, -0.022958718240261078, 0.010653545148670673, 0.047053512185811996, 0.0118306698...
368
368
['Yoseob Han', 'Jingu Kang', 'Jong Chul Ye']
1801.01258v1
For homeland and transportation security applications, 2D X-ray explosive detection system (EDS) have been widely used, but they have limitations in recognizing 3D shape of the hidden objects. Among various types of 3D computed tomography (CT) systems to address this issue, this paper is interested in a stationary CT u...
Deep Learning Reconstruction for 9-View Dual Energy CT Baggage Scanner
2,018
http://arxiv.org/pdf/1801.01258v1
Title Deep Learning Reconstruction 9View Dual Energy CT Baggage Scanner Summary homeland transportation security application 2D Xray explosive detection system EDS widely used limitation recognizing 3D shape hidden object Among various type 3D computed tomography CT system address issue paper interested stationary CT u...
[-0.04104342311620712, 0.032573625445365906, -0.005146138835698366, 0.051213718950748444, 0.013152572326362133, 0.0104643814265728, 0.03190971165895462, 0.019044918939471245, -0.054176948964595795, 0.0614003948867321, 0.01349288783967495, 0.0028527413960546255, 0.006450768560171127, 0.10132794082164764, 0.0114539684727...
369
369
['Jayanta K Dutta', 'Jiayi Liu', 'Unmesh Kurup', 'Mohak Shah']
1801.08577v1
Deep learning has shown promising results on many machine learning tasks but DL models are often complex networks with large number of neurons and layers, and recently, complex layer structures known as building blocks. Finding the best deep model requires a combination of finding both the right architecture and the co...
Effective Building Block Design for Deep Convolutional Neural Networks using Search
2,018
http://arxiv.org/pdf/1801.08577v1
Title Effective Building Block Design Deep Convolutional Neural Networks using Search Summary Deep learning shown promising result many machine learning task DL model often complex network large number neuron layer recently complex layer structure known building block Finding best deep model requires combination findin...
[0.020713958889245987, 0.06414908915758133, -0.00911646243184805, 0.07008805871009827, 0.019468775019049644, -0.03916212543845177, 0.05745573341846466, 0.03063204325735569, -0.0124319763854146, 0.014917802065610886, -0.011679022572934628, -0.0071962615475058556, -0.03560882434248924, 0.045701056718826294, 0.01874051615...
370
370
['Haque Ishfaq', 'Assaf Hoogi', 'Daniel Rubin']
1802.04403v1
Deep metric learning has been demonstrated to be highly effective in learning semantic representation and encoding information that can be used to measure data similarity, by relying on the embedding learned from metric learning. At the same time, variational autoencoder (VAE) has widely been used to approximate infere...
TVAE: Triplet-Based Variational Autoencoder using Metric Learning
2,018
http://arxiv.org/pdf/1802.04403v1
Title TVAE TripletBased Variational Autoencoder using Metric Learning Summary Deep metric learning demonstrated highly effective learning semantic representation encoding information used measure data similarity relying embedding learned metric learning time variational autoencoder VAE widely used approximate inference...
[-0.019333014264702797, 0.03620387986302376, -0.011517582461237907, 0.03062313050031662, 0.00433266069740057, 0.024699347093701363, 0.004918583203107119, 0.0010207797167822719, -0.04832843318581581, -0.01274857483804226, 0.006990343797951937, 0.00701321242377162, -0.02643158845603466, 0.08037950843572617, 0.05721693858...
371
371
['Nick Haber', 'Damian Mrowca', 'Li Fei-Fei', 'Daniel L. K. Yamins']
1802.07442v1
Infants are experts at playing, with an amazing ability to generate novel structured behaviors in unstructured environments that lack clear extrinsic reward signals. We seek to mathematically formalize these abilities using a neural network that implements curiosity-driven intrinsic motivation. Using a simple but ecolo...
Learning to Play with Intrinsically-Motivated Self-Aware Agents
2,018
http://arxiv.org/pdf/1802.07442v1
Title Learning Play IntrinsicallyMotivated SelfAware Agents Summary Infants expert playing amazing ability generate novel structured behavior unstructured environment lack clear extrinsic reward signal seek mathematically formalize ability using neural network implement curiositydriven intrinsic motivation Using simple...
[0.007545554544776678, 0.010510986670851707, -0.035660646855831146, 0.01321401633322239, 0.02706293947994709, -0.02532416209578514, -0.01706690713763237, -0.03375625237822533, -0.020416108891367912, 0.006275697145611048, -0.06261751055717468, 0.07055346667766571, -0.045897942036390305, 0.07986181229352951, 0.0372456461...
372
372
['Nick Haber', 'Damian Mrowca', 'Li Fei-Fei', 'Daniel L. K. Yamins']
1802.07461v1
Infants are experts at playing, with an amazing ability to generate novel structured behaviors in unstructured environments that lack clear extrinsic reward signals. We seek to replicate some of these abilities with a neural network that implements curiosity-driven intrinsic motivation. Using a simple but ecologically ...
Emergence of Structured Behaviors from Curiosity-Based Intrinsic Motivation
2,018
http://arxiv.org/pdf/1802.07461v1
Title Emergence Structured Behaviors CuriosityBased Intrinsic Motivation Summary Infants expert playing amazing ability generate novel structured behavior unstructured environment lack clear extrinsic reward signal seek replicate ability neural network implement curiositydriven intrinsic motivation Using simple ecologi...
[0.0014127821195870638, 0.01567617617547512, -0.03436420485377312, 0.00010565984848653898, 0.015486599877476692, -0.00438009575009346, -0.043718114495277405, -0.016360478475689888, -0.04642879217863083, 0.02967933937907219, -0.035631969571113586, 0.02661474049091339, -0.018195616081357002, 0.0818972960114479, 0.0488295...
373
373
['Emily Denton', 'Rob Fergus']
1802.07687v2
Generating video frames that accurately predict future world states is challenging. Existing approaches either fail to capture the full distribution of outcomes, or yield blurry generations, or both. In this paper we introduce an unsupervised video generation model that learns a prior model of uncertainty in a given en...
Stochastic Video Generation with a Learned Prior
2,018
http://arxiv.org/pdf/1802.07687v2
Title Stochastic Video Generation Learned Prior Summary Generating video frame accurately predict future world state challenging Existing approach either fail capture full distribution outcome yield blurry generation paper introduce unsupervised video generation model learns prior model uncertainty given environment Vi...
[-0.01858837902545929, 0.07819383591413498, 0.04127709940075874, -0.027748022228479385, -0.005655407905578613, -0.03293720632791519, -0.008636883459985256, 0.019356105476617813, -0.04649140685796738, -0.030510656535625458, 0.061888791620731354, -0.02407054416835308, -0.014147943817079067, 0.07956864684820175, 0.0528176...
374
374
['Weifeng Ge', 'Sibei Yang', 'Yizhou Yu']
1802.09129v1
Supervised object detection and semantic segmentation require object or even pixel level annotations. When there exist image level labels only, it is challenging for weakly supervised algorithms to achieve accurate predictions. The accuracy achieved by top weakly supervised algorithms is still significantly lower than ...
Multi-Evidence Filtering and Fusion for Multi-Label Classification, Object Detection and Semantic Segmentation Based on Weakly Supervised Learning
2,018
http://arxiv.org/pdf/1802.09129v1
Title MultiEvidence Filtering Fusion MultiLabel Classification Object Detection Semantic Segmentation Based Weakly Supervised Learning Summary Supervised object detection semantic segmentation require object even pixel level annotation exist image level label challenging weakly supervised algorithm achieve accurate pre...
[0.03029974363744259, -0.01569221541285515, 0.030196411535143852, 0.06377570331096649, 0.007281121332198381, -0.0022118433844298124, 0.009080919437110424, -0.033626068383455276, 0.029017699882388115, 0.006745519116520882, -0.05035201460123062, 0.06610722839832306, -0.012566382996737957, 0.030714480206370354, 0.02697962...
375
375
['Quynh Nguyen', 'Mahesh Mukkamala', 'Matthias Hein']
1803.00094v1
In the recent literature the important role of depth in deep learning has been emphasized. In this paper we argue that sufficient width of a feedforward network is equally important by answering the simple question under which conditions the decision regions of a neural network are connected. It turns out that for a cl...
Neural Networks Should Be Wide Enough to Learn Disconnected Decision Regions
2,018
http://arxiv.org/pdf/1803.00094v1
Title Neural Networks Wide Enough Learn Disconnected Decision Regions Summary recent literature important role depth deep learning emphasized paper argue sufficient width feedforward network equally important answering simple question condition decision region neural network connected turn class activation function inc...
[0.0016157060163095593, 0.06191333010792732, -0.02980755642056465, 0.038863662630319595, -0.003247797256335616, -0.021014919504523277, 0.07946360111236572, -0.019299747422337532, -0.020289214327931404, -0.004484830889850855, -0.006770009640604258, 0.016767330467700958, 0.008010574616491795, 0.02744484692811966, 0.02263...
376
376
['Chengliang Yang', 'Anand Rangarajan', 'Sanjay Ranka']
1803.02544v2
We develop three efficient approaches for generating visual explanations from 3D convolutional neural networks (3D-CNNs) for Alzheimer's disease classification. One approach conducts sensitivity analysis on hierarchical 3D image segmentation, and the other two visualize network activations on a spatial map. Visual chec...
Visual Explanations From Deep 3D Convolutional Neural Networks for Alzheimer's Disease Classification
2,018
http://arxiv.org/pdf/1803.02544v2
Title Visual Explanations Deep 3D Convolutional Neural Networks Alzheimers Disease Classification Summary develop three efficient approach generating visual explanation 3D convolutional neural network 3DCNNs Alzheimers disease classification One approach conduct sensitivity analysis hierarchical 3D image segmentation t...
[-0.001759025384671986, 0.022373871877789497, 0.0005091670900583267, 0.051700640469789505, 0.025821156799793243, 0.00948801264166832, 0.051337163895368576, -1.5511152014369145e-05, -0.0009512273827567697, 0.03166145086288452, 0.02257706969976425, 0.022253500297665596, 0.03461276739835739, 0.03744889050722122, 0.0280076...
377
377
['Pavel Izmailov', 'Dmitrii Podoprikhin', 'Timur Garipov', 'Dmitry Vetrov', 'Andrew Gordon Wilson']
1803.05407v1
Deep neural networks are typically trained by optimizing a loss function with an SGD variant, in conjunction with a decaying learning rate, until convergence. We show that simple averaging of multiple points along the trajectory of SGD, with a cyclical or constant learning rate, leads to better generalization than conv...
Averaging Weights Leads to Wider Optima and Better Generalization
2,018
http://arxiv.org/pdf/1803.05407v1
Title Averaging Weights Leads Wider Optima Better Generalization Summary Deep neural network typically trained optimizing loss function SGD variant conjunction decaying learning rate convergence show simple averaging multiple point along trajectory SGD cyclical constant learning rate lead better generalization conventi...
[-0.02773691713809967, 0.02495228312909603, 0.008313159458339214, 0.036607641726732254, 0.02793952263891697, -0.0017876147758215666, 0.06204995885491371, -0.0396859273314476, -0.0422554649412632, 0.07349628955125809, 0.03743846341967583, -0.04685478284955025, 0.03435065224766731, 0.030554182827472687, 0.039004039019346...
378
378
['Abdulrahman Oladipupo Ibraheem']
1412.6749v1
By drawing on ideas from optimisation theory, artificial neural networks (ANN), graph embeddings and sparse representations, I develop a novel technique, termed SENNS (Sparse Extraction Neural NetworkS), aimed at addressing the feature extraction problem. The proposed method uses (preferably deep) ANNs for projecting i...
SENNS: Sparse Extraction Neural NetworkS for Feature Extraction
2,014
http://arxiv.org/pdf/1412.6749v1
Title SENNS Sparse Extraction Neural NetworkS Feature Extraction Summary drawing idea optimisation theory artificial neural network ANN graph embeddings sparse representation develop novel technique termed SENNS Sparse Extraction Neural NetworkS aimed addressing feature extraction problem proposed method us preferably ...
[-0.0012898645363748074, 0.04691426083445549, -0.014183642342686653, 0.08530305325984955, -0.0022745367605239153, -0.0011008928995579481, 0.06389493495225906, 0.026457104831933975, -0.01797901839017868, -0.009211565367877483, -0.00664258049800992, -0.014513140544295311, 0.051773786544799805, 0.026841364800930023, 0.024...
379
379
['Dougal J. Sutherland', 'Hsiao-Yu Tung', 'Heiko Strathmann', 'Soumyajit De', 'Aaditya Ramdas', 'Alex Smola', 'Arthur Gretton']
1611.04488v4
We propose a method to optimize the representation and distinguishability of samples from two probability distributions, by maximizing the estimated power of a statistical test based on the maximum mean discrepancy (MMD). This optimized MMD is applied to the setting of unsupervised learning by generative adversarial ne...
Generative Models and Model Criticism via Optimized Maximum Mean Discrepancy
2,016
http://arxiv.org/pdf/1611.04488v4
Title Generative Models Model Criticism via Optimized Maximum Mean Discrepancy Summary propose method optimize representation distinguishability sample two probability distribution maximizing estimated power statistical test based maximum mean discrepancy MMD optimized MMD applied setting unsupervised learning generati...
[0.013605509884655476, 0.07093748450279236, -0.03573538362979889, 0.02927105501294136, -0.013702447526156902, -0.004465721547603607, 0.05487830191850662, -0.011813649907708168, -0.03715183958411217, 0.006023547146469355, 0.022856414318084717, -0.01222627516835928, -0.0026135826483368874, 0.009016846306622028, 0.0426554...
380
380
['Jiequn Han', 'Weinan E']
1611.07422v1
Many real world stochastic control problems suffer from the "curse of dimensionality". To overcome this difficulty, we develop a deep learning approach that directly solves high-dimensional stochastic control problems based on Monte-Carlo sampling. We approximate the time-dependent controls as feedforward neural networ...
Deep Learning Approximation for Stochastic Control Problems
2,016
http://arxiv.org/pdf/1611.07422v1
Title Deep Learning Approximation Stochastic Control Problems Summary Many real world stochastic control problem suffer curse dimensionality overcome difficulty develop deep learning approach directly solves highdimensional stochastic control problem based MonteCarlo sampling approximate timedependent control feedforwa...
[-0.03404943272471428, 0.016462624073028564, -0.009168632328510284, 0.01569977216422558, 0.03265073150396347, -0.028827466070652008, 0.0308519396930933, 0.012387428432703018, -0.023935584351420403, 0.02761777676641941, -0.005317831877619028, -0.019688762724399567, -0.03616824373602867, 0.09597254544496536, 0.0452575609...
381
381
['Marwin H. S. Segler', 'Thierry Kogej', 'Christian Tyrchan', 'Mark P. Waller']
1701.01329v1
In de novo drug design, computational strategies are used to generate novel molecules with good affinity to the desired biological target. In this work, we show that recurrent neural networks can be trained as generative models for molecular structures, similar to statistical language models in natural language process...
Generating Focussed Molecule Libraries for Drug Discovery with Recurrent Neural Networks
2,017
http://arxiv.org/pdf/1701.01329v1
Title Generating Focussed Molecule Libraries Drug Discovery Recurrent Neural Networks Summary de novo drug design computational strategy used generate novel molecule good affinity desired biological target work show recurrent neural network trained generative model molecular structure similar statistical language model...
[0.03883090987801552, 0.023383507505059242, -0.0018499937141314149, 0.007632397580891848, 0.002424982376396656, -0.029439885169267654, 0.032943740487098694, 0.002812813501805067, 0.001298057846724987, -0.011611754074692726, 0.012721366249024868, 0.00701055396348238, 0.02611383982002735, 0.07596792280673981, 0.062137685...
382
382
['Matthias Plappert', 'Rein Houthooft', 'Prafulla Dhariwal', 'Szymon Sidor', 'Richard Y. Chen', 'Xi Chen', 'Tamim Asfour', 'Pieter Abbeel', 'Marcin Andrychowicz']
1706.01905v2
Deep reinforcement learning (RL) methods generally engage in exploratory behavior through noise injection in the action space. An alternative is to add noise directly to the agent's parameters, which can lead to more consistent exploration and a richer set of behaviors. Methods such as evolutionary strategies use param...
Parameter Space Noise for Exploration
2,017
http://arxiv.org/pdf/1706.01905v2
Title Parameter Space Noise Exploration Summary Deep reinforcement learning RL method generally engage exploratory behavior noise injection action space alternative add noise directly agent parameter lead consistent exploration richer set behavior Methods evolutionary strategy use parameter perturbation discard tempora...
[-0.0003265595296397805, 0.047623660415410995, 0.0010571572929620743, -0.02193429134786129, 0.014544806443154812, -0.002184385433793068, -0.035086486488580704, -0.03803787752985954, -0.06285911798477173, 0.028673654422163963, 0.015436917543411255, 0.03805789723992348, -0.04189801216125488, 0.04146125540137291, 0.018569...
383
383
['El Mahdi El Mhamdi', 'Rachid Guerraoui', 'Sebastien Rouault']
1707.08167v2
With the development of neural networks based machine learning and their usage in mission critical applications, voices are rising against the \textit{black box} aspect of neural networks as it becomes crucial to understand their limits and capabilities. With the rise of neuromorphic hardware, it is even more critical ...
On The Robustness of a Neural Network
2,017
http://arxiv.org/pdf/1707.08167v2
Title Robustness Neural Network Summary development neural network based machine learning usage mission critical application voice rising textitblack box aspect neural network becomes crucial understand limit capability rise neuromorphic hardware even critical understand neural network distributed system tolerates fail...
[-0.026057636365294456, 0.03605284169316292, -0.02108803763985634, 0.025893907994031906, 0.003046888392418623, -0.05413280799984932, 0.011631691828370094, -0.04342258721590042, 0.0014416167978197336, 0.007316091563552618, 0.044314607977867126, 0.02957012690603733, -0.0011027463478967547, 0.06523576378822327, 0.04307548...
384
384
['Jiaxin Shi', 'Jianfei Chen', 'Jun Zhu', 'Shengyang Sun', 'Yucen Luo', 'Yihong Gu', 'Yuhao Zhou']
1709.05870v1
In this paper we introduce ZhuSuan, a python probabilistic programming library for Bayesian deep learning, which conjoins the complimentary advantages of Bayesian methods and deep learning. ZhuSuan is built upon Tensorflow. Unlike existing deep learning libraries, which are mainly designed for deterministic neural netw...
ZhuSuan: A Library for Bayesian Deep Learning
2,017
http://arxiv.org/pdf/1709.05870v1
Title ZhuSuan Library Bayesian Deep Learning Summary paper introduce ZhuSuan python probabilistic programming library Bayesian deep learning conjoins complimentary advantage Bayesian method deep learning ZhuSuan built upon Tensorflow Unlike existing deep learning library mainly designed deterministic neural network sup...
[-0.01321567315608263, 0.04119647294282913, -0.007770095020532608, -0.015688182786107063, 0.009080410934984684, 0.0002429285377729684, 0.04519704729318619, -0.015076905488967896, -0.038274284452199936, -0.004509477876126766, 0.017884528264403343, -0.011379523202776909, 0.03480071201920509, 0.09155195206403732, 0.012622...
385
385
['Konstantinos Chatzilygeroudis', 'Jean-Baptiste Mouret']
1709.06917v2
The most data-efficient algorithms for reinforcement learning in robotics are model-based policy search algorithms, which alternate between learning a dynamical model of the robot and optimizing a policy to maximize the expected return given the model and its uncertainties. Among the few proposed approaches, the recent...
Using Parameterized Black-Box Priors to Scale Up Model-Based Policy Search for Robotics
2,017
http://arxiv.org/pdf/1709.06917v2
Title Using Parameterized BlackBox Priors Scale ModelBased Policy Search Robotics Summary dataefficient algorithm reinforcement learning robotics modelbased policy search algorithm alternate learning dynamical model robot optimizing policy maximize expected return given model uncertainty Among proposed approach recentl...
[0.011151357553899288, -0.012564446777105331, 0.001024615135975182, -0.06330064684152603, 0.008877766318619251, -0.0006621857755817473, 0.014855976216495037, -0.004244039300829172, -0.028440549969673157, -0.009114322252571583, -0.04701918363571167, -0.018953688442707062, -0.012030098587274551, 0.07456639409065247, 0.02...
386
386
['Rémi Pautrat', 'Konstantinos Chatzilygeroudis', 'Jean-Baptiste Mouret']
1709.06919v2
One of the most interesting features of Bayesian optimization for direct policy search is that it can leverage priors (e.g., from simulation or from previous tasks) to accelerate learning on a robot. In this paper, we are interested in situations for which several priors exist but we do not know in advance which one fi...
Bayesian Optimization with Automatic Prior Selection for Data-Efficient Direct Policy Search
2,017
http://arxiv.org/pdf/1709.06919v2
Title Bayesian Optimization Automatic Prior Selection DataEfficient Direct Policy Search Summary One interesting feature Bayesian optimization direct policy search leverage prior eg simulation previous task accelerate learning robot paper interested situation several prior exist know advance one fit best current situat...
[-0.00995427742600441, 0.028338341042399406, -0.012279587797820568, -0.004892553668469191, 0.01670580916106701, -0.006485891528427601, 0.032854028046131134, 0.020987290889024734, -0.007547059096395969, -0.0314612090587616, 0.006708703003823757, 0.0399947352707386, 0.0047004008665680885, 0.032445043325424194, 0.01308621...
387
387
['Thiago Serra', 'Christian Tjandraatmadja', 'Srikumar Ramalingam']
1711.02114v2
In this paper, we study the representational power of deep neural networks (DNN) that belong to the family of piecewise-linear (PWL) functions, based on PWL activation units such as rectifier or maxout. We investigate the complexity of such networks by studying the number of linear regions of the PWL function. Typicall...
Bounding and Counting Linear Regions of Deep Neural Networks
2,017
http://arxiv.org/pdf/1711.02114v2
Title Bounding Counting Linear Regions Deep Neural Networks Summary paper study representational power deep neural network DNN belong family piecewiselinear PWL function based PWL activation unit rectifier maxout investigate complexity network studying number linear region PWL function Typically PWL function DNN seen l...
[-0.050812914967536926, 0.04456111043691635, -0.04539470374584198, 0.06889405846595764, 0.018647756427526474, -0.03307289630174637, 0.053659118711948395, -0.04094891995191574, -0.011029049754142761, 0.03817381337285042, 0.009715624153614044, -0.005379774607717991, -0.011343705467879772, 0.06885793060064316, 0.045825649...
388
388
['Guillaume Bellec', 'David Kappel', 'Wolfgang Maass', 'Robert Legenstein']
1711.05136v4
Neuromorphic hardware tends to pose limits on the connectivity of deep networks that one can run on them. But also generic hardware and software implementations of deep learning run more efficiently for sparse networks. Several methods exist for pruning connections of a neural network after it was trained without conne...
Deep Rewiring: Training very sparse deep networks
2,017
http://arxiv.org/pdf/1711.05136v4
Title Deep Rewiring Training sparse deep network Summary Neuromorphic hardware tends pose limit connectivity deep network one run also generic hardware software implementation deep learning run efficiently sparse network Several method exist pruning connection neural network trained without connectivity constraint pres...
[-0.043236829340457916, 0.016069017350673676, -0.0027782388497143984, 0.04958239570260048, 0.008396126329898834, -0.05497850477695465, 0.00010714112431742251, -0.016810746863484383, -0.05139242112636566, -0.010272297076880932, -0.042071785777807236, 0.02903454750776291, 0.0350470244884491, 0.03272346034646034, 0.054298...
389
389
['Artit Wangperawong', 'Kettip Kriangchaivech', 'Austin Lanari', 'Supui Lam', 'Panthong Wangperawong']
1801.03143v1
To compare entities of differing types and structural components, the artificial neural network paradigm was used to cross-compare structural components between heterogeneous documents. Trainable weighted structural components were input into machine-learned activation functions of the neurons. The model was used for m...
Comparing heterogeneous entities using artificial neural networks of trainable weighted structural components and machine-learned activation functions
2,018
http://arxiv.org/pdf/1801.03143v1
Title Comparing heterogeneous entity using artificial neural network trainable weighted structural component machinelearned activation function Summary compare entity differing type structural component artificial neural network paradigm used crosscompare structural component heterogeneous document Trainable weighted s...
[0.03217822685837746, 0.026238176971673965, -0.025483589619398117, 0.0047994875349104404, -0.006483330857008696, 0.01590096764266491, 0.03214715048670769, 0.030115464702248573, -0.017625799402594566, -0.04611966386437416, -0.04700959846377373, -0.02767229452729225, 0.040130794048309326, 0.04134885221719742, -0.00246275...
390
390
['Adrien Baranes', 'Pierre-Yves Oudeyer']
1301.4862v1
We introduce the Self-Adaptive Goal Generation - Robust Intelligent Adaptive Curiosity (SAGG-RIAC) architecture as an intrinsi- cally motivated goal exploration mechanism which allows active learning of inverse models in high-dimensional redundant robots. This allows a robot to efficiently and actively learn distributi...
Active Learning of Inverse Models with Intrinsically Motivated Goal Exploration in Robots
2,013
http://arxiv.org/pdf/1301.4862v1
Title Active Learning Inverse Models Intrinsically Motivated Goal Exploration Robots Summary introduce SelfAdaptive Goal Generation Robust Intelligent Adaptive Curiosity SAGGRIAC architecture intrinsi cally motivated goal exploration mechanism allows active learning inverse model highdimensional redundant robot allows ...
[0.01569799706339836, -0.019697241485118866, -0.010390009731054306, -0.04707631096243858, 0.019959047436714172, 0.0067192609421908855, -0.008895752020180225, -0.052416108548641205, -0.023944152519106865, -0.005776107311248779, -0.06151485815644264, 0.05318199843168259, -0.0376877561211586, 0.05663970857858658, 0.006373...
391
391
['Peter Ondruska', 'Julie Dequaire', 'Dominic Zeng Wang', 'Ingmar Posner']
1604.05091v2
In this work we present a novel end-to-end framework for tracking and classifying a robot's surroundings in complex, dynamic and only partially observable real-world environments. The approach deploys a recurrent neural network to filter an input stream of raw laser measurements in order to directly infer object locati...
End-to-End Tracking and Semantic Segmentation Using Recurrent Neural Networks
2,016
http://arxiv.org/pdf/1604.05091v2
Title EndtoEnd Tracking Semantic Segmentation Using Recurrent Neural Networks Summary work present novel endtoend framework tracking classifying robot surroundings complex dynamic partially observable realworld environment approach deploys recurrent neural network filter input stream raw laser measurement order directl...
[-0.0054203178733587265, -0.001586581813171506, 0.032583948224782944, 0.06681782752275467, 0.008075389079749584, -0.026314599439501762, 0.0012463745661079884, -0.06645060330629349, -0.04777192696928978, -0.017123231664299965, 0.03182670474052429, 0.04510796442627907, -0.04645577073097229, 0.03661247342824936, -0.014234...
392
392
['Peter Ondruska', 'Ingmar Posner']
1602.00991v2
This paper presents to the best of our knowledge the first end-to-end object tracking approach which directly maps from raw sensor input to object tracks in sensor space without requiring any feature engineering or system identification in the form of plant or sensor models. Specifically, our system accepts a stream of...
Deep Tracking: Seeing Beyond Seeing Using Recurrent Neural Networks
2,016
http://arxiv.org/pdf/1602.00991v2
Title Deep Tracking Seeing Beyond Seeing Using Recurrent Neural Networks Summary paper present best knowledge first endtoend object tracking approach directly map raw sensor input object track sensor space without requiring feature engineering system identification form plant sensor model Specifically system accepts st...
[-0.040110498666763306, 0.020441696047782898, 0.03217256814241409, 0.0626005306839943, -0.010437419638037682, -0.030604690313339233, -0.00931694358587265, -0.04911627620458603, -0.008815133012831211, -0.015731601044535637, 0.02961554564535618, 0.012684076093137264, 0.0022867396473884583, 0.08908965438604355, 0.01761854...
393
393
['William Lotter', 'Gabriel Kreiman', 'David Cox']
1605.08104v5
While great strides have been made in using deep learning algorithms to solve supervised learning tasks, the problem of unsupervised learning - leveraging unlabeled examples to learn about the structure of a domain - remains a difficult unsolved challenge. Here, we explore prediction of future frames in a video sequenc...
Deep Predictive Coding Networks for Video Prediction and Unsupervised Learning
2,016
http://arxiv.org/pdf/1605.08104v5
Title Deep Predictive Coding Networks Video Prediction Unsupervised Learning Summary great stride made using deep learning algorithm solve supervised learning task problem unsupervised learning leveraging unlabeled example learn structure domain remains difficult unsolved challenge explore prediction future frame video...
[-0.02492854744195938, 0.03656262159347534, -0.0023122630082070827, 0.03793924301862717, 0.02912059612572193, 0.008039752021431923, 0.009802541695535183, 0.01682428829371929, -0.0555635504424572, 0.022038007155060768, 0.03866586089134216, -0.0014181542210280895, 0.015415824018418789, 0.08241347223520279, 0.041331782937...
394
394
['Martin Engelcke', 'Dushyant Rao', 'Dominic Zeng Wang', 'Chi Hay Tong', 'Ingmar Posner']
1609.06666v2
This paper proposes a computationally efficient approach to detecting objects natively in 3D point clouds using convolutional neural networks (CNNs). In particular, this is achieved by leveraging a feature-centric voting scheme to implement novel convolutional layers which explicitly exploit the sparsity encountered in...
Vote3Deep: Fast Object Detection in 3D Point Clouds Using Efficient Convolutional Neural Networks
2,016
http://arxiv.org/pdf/1609.06666v2
Title Vote3Deep Fast Object Detection 3D Point Clouds Using Efficient Convolutional Neural Networks Summary paper proposes computationally efficient approach detecting object natively 3D point cloud using convolutional neural network CNNs particular achieved leveraging featurecentric voting scheme implement novel convo...
[-0.010530673898756504, -0.00397380068898201, 0.03137296810746193, 0.08886940777301788, -0.00474140839651227, -0.0034435151610523462, 0.012546230107545853, -0.039406031370162964, -0.03957242891192436, 0.02292507514357567, 0.008108907379209995, 0.05722877010703087, -0.008607571944594383, 0.06548508256673813, 0.035875830...
395
395
['Naveen Kodali', 'Jacob Abernethy', 'James Hays', 'Zsolt Kira']
1705.07215v5
We propose studying GAN training dynamics as regret minimization, which is in contrast to the popular view that there is consistent minimization of a divergence between real and generated distributions. We analyze the convergence of GAN training from this new point of view to understand why mode collapse happens. We hy...
On Convergence and Stability of GANs
2,017
http://arxiv.org/pdf/1705.07215v5
Title Convergence Stability GANs Summary propose studying GAN training dynamic regret minimization contrast popular view consistent minimization divergence real generated distribution analyze convergence GAN training new point view understand mode collapse happens hypothesize existence undesirable local equilibrium non...
[-0.04972640424966812, 0.07576105743646622, -0.020075885578989983, 0.007751937955617905, 0.036929573863744736, -0.014320901595056057, -0.007350770756602287, 0.008607647381722927, -0.06422300636768341, 0.0783543586730957, -0.006085442379117012, -0.030311310663819313, -0.020794961601495743, 0.007470865733921528, 0.056025...
396
396
['YuXuan Liu', 'Abhishek Gupta', 'Pieter Abbeel', 'Sergey Levine']
1707.03374v1
Imitation learning is an effective approach for autonomous systems to acquire control policies when an explicit reward function is unavailable, using supervision provided as demonstrations from an expert, typically a human operator. However, standard imitation learning methods assume that the agent receives examples of...
Imitation from Observation: Learning to Imitate Behaviors from Raw Video via Context Translation
2,017
http://arxiv.org/pdf/1707.03374v1
Title Imitation Observation Learning Imitate Behaviors Raw Video via Context Translation Summary Imitation learning effective approach autonomous system acquire control policy explicit reward function unavailable using supervision provided demonstration expert typically human operator However standard imitation learnin...
[0.015244332142174244, 0.020528141409158707, 0.016690772026777267, -0.00498945452272892, -0.014369165524840355, -0.00951950903981924, 0.032931145280599594, 0.016603803262114525, -0.026855409145355225, -0.03213277459144592, -0.016663720831274986, 0.02962314523756504, -0.027660749852657318, 0.028468824923038483, 0.017942...
397
397
['Vamsi K. Ithapu', 'Sathya Ravi', 'Vikas Singh']
1506.03412v3
Unsupervised pretraining and dropout have been well studied, especially with respect to regularization and output consistency. However, our understanding about the explicit convergence rates of the parameter estimates, and their dependence on the learning (like denoising and dropout rate) and structural (like depth and...
Convergence rates for pretraining and dropout: Guiding learning parameters using network structure
2,015
http://arxiv.org/pdf/1506.03412v3
Title Convergence rate pretraining dropout Guiding learning parameter using network structure Summary Unsupervised pretraining dropout well studied especially respect regularization output consistency However understanding explicit convergence rate parameter estimate dependence learning like denoising dropout rate stru...
[-0.022588271647691727, 0.05241832509636879, -0.01163649931550026, 0.012558311223983765, 0.01884423941373825, -0.02940197102725506, 0.014257178641855717, 0.0019173461478203535, -0.042971834540367126, 0.020522965118288994, -0.014107691124081612, 0.02639881707727909, -0.008156240917742252, 0.09559813886880875, 0.00765846...
398
398
['Zhuolin Jiang', 'Yaming Wang', 'Larry Davis', 'Walt Andrews', 'Viktor Rozgic']
1602.01168v2
Deep Convolutional Neural Networks (CNN) enforces supervised information only at the output layer, and hidden layers are trained by back propagating the prediction error from the output layer without explicit supervision. We propose a supervised feature learning approach, Label Consistent Neural Network, which enforces...
Learning Discriminative Features via Label Consistent Neural Network
2,016
http://arxiv.org/pdf/1602.01168v2
Title Learning Discriminative Features via Label Consistent Neural Network Summary Deep Convolutional Neural Networks CNN enforces supervised information output layer hidden layer trained back propagating prediction error output layer without explicit supervision propose supervised feature learning approach Label Consi...
[0.020428013056516647, 0.04332013055682182, -0.0005980939022265375, 0.03674156591296196, 0.031040387228131294, -0.0040699574165046215, 0.050413019955158234, -0.005540117155760527, 0.012124375440180302, -0.029857371002435684, -0.041967373341321945, 0.0609920397400856, -0.05401777848601341, 0.026943597942590714, -0.01131...
399
399
['Hamid Dadkhahi', 'Marco F. Duarte', 'Benjamin Marlin']
1606.08282v3
This paper proposes an out-of-sample extension framework for a global manifold learning algorithm (Isomap) that uses temporal information in out-of-sample points in order to make the embedding more robust to noise and artifacts. Given a set of noise-free training data and its embedding, the proposed framework extends t...
Out-of-Sample Extension for Dimensionality Reduction of Noisy Time Series
2,016
http://arxiv.org/pdf/1606.08282v3
Title OutofSample Extension Dimensionality Reduction Noisy Time Series Summary paper proposes outofsample extension framework global manifold learning algorithm Isomap us temporal information outofsample point order make embedding robust noise artifact Given set noisefree training data embedding proposed framework exte...
[-0.05783146992325783, 0.03087870217859745, 0.016298944130539894, 0.0324084535241127, 0.02472299337387085, 0.04794783517718315, 0.006237872876226902, 0.01697060652077198, 0.028118178248405457, 0.0300690196454525, 0.07045266032218933, 0.03230315446853638, 0.004669292829930782, 0.006983146071434021, 0.042533956468105316,...