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
1,200
1,200
['Wenlin Chen', 'James T. Wilson', 'Stephen Tyree', 'Kilian Q. Weinberger', 'Yixin Chen']
1506.04449v1
Convolutional neural networks (CNN) are increasingly used in many areas of computer vision. They are particularly attractive because of their ability to "absorb" great quantities of labeled data through millions of parameters. However, as model sizes increase, so do the storage and memory requirements of the classifier...
Compressing Convolutional Neural Networks
2,015
http://arxiv.org/pdf/1506.04449v1
Title Compressing Convolutional Neural Networks Summary Convolutional neural network CNN increasingly used many area computer vision particularly attractive ability absorb great quantity labeled data million parameter However model size increase storage memory requirement classifier present novel network architecture F...
[-0.04463288187980652, 0.06163084879517555, -0.012253152206540108, 0.05511300638318062, -0.0025499064940959215, -0.012141133658587933, 0.043387748301029205, 0.03678958863019943, -0.05649621784687042, 0.02461875230073929, -0.01154039055109024, 0.007315043825656176, -0.009642541408538818, 0.06366273015737534, 0.008259010...
1,201
1,201
['Jason Yosinski', 'Jeff Clune', 'Anh Nguyen', 'Thomas Fuchs', 'Hod Lipson']
1506.06579v1
Recent years have produced great advances in training large, deep neural networks (DNNs), including notable successes in training convolutional neural networks (convnets) to recognize natural images. However, our understanding of how these models work, especially what computations they perform at intermediate layers, h...
Understanding Neural Networks Through Deep Visualization
2,015
http://arxiv.org/pdf/1506.06579v1
Title Understanding Neural Networks Deep Visualization Summary Recent year produced great advance training large deep neural network DNNs including notable success training convolutional neural network convnets recognize natural image However understanding model work especially computation perform intermediate layer la...
[-0.0007392374100163579, 0.012455489486455917, -0.013493246398866177, 0.059293001890182495, 0.0012556114234030247, 0.007671872619539499, 0.042305462062358856, 0.009318207390606403, -0.04370226711034775, 0.02192402444779873, -0.028857452794909477, 0.036842938512563705, 0.00389855750836432, 0.07204461097717285, 0.0410939...
1,202
1,202
["Keiron O'Shea"]
1507.05053v1
Greedy Restrictive Boltzmann Machines yield an fairly low 0.72% error rate on the famous MNIST database of handwritten digits. All that was required to achieve this result was a high number of hidden layers consisting of many neurons, and a graphics card to greatly speed up the rate of learning.
Massively Deep Artificial Neural Networks for Handwritten Digit Recognition
2,015
http://arxiv.org/pdf/1507.05053v1
Title Massively Deep Artificial Neural Networks Handwritten Digit Recognition Summary Greedy Restrictive Boltzmann Machines yield fairly low 072 error rate famous MNIST database handwritten digit required achieve result high number hidden layer consisting many neuron graphic card greatly speed rate learning Authors 0 A...
[-0.006818361114710569, 0.03810611739754677, -0.030558491125702858, 0.0776267796754837, 0.0075163389556109905, -0.0014100931584835052, 0.07726393640041351, 0.007971227169036865, -0.012542841956019402, 0.02090977504849434, 0.00881256815046072, -0.003433726029470563, 0.018626883625984192, 0.05653011426329613, 0.005641664...
1,203
1,203
['Shuchang Zhou', 'Jia-Nan Wu']
1507.05775v2
In this paper we propose and study a technique to reduce the number of parameters and computation time in fully-connected layers of neural networks using Kronecker product, at a mild cost of the prediction quality. The technique proceeds by replacing Fully-Connected layers with so-called Kronecker Fully-Connected layer...
Compression of Fully-Connected Layer in Neural Network by Kronecker Product
2,015
http://arxiv.org/pdf/1507.05775v2
Title Compression FullyConnected Layer Neural Network Kronecker Product Summary paper propose study technique reduce number parameter computation time fullyconnected layer neural network using Kronecker product mild cost prediction quality technique proceeds replacing FullyConnected layer socalled Kronecker FullyConnec...
[0.012350041419267654, 0.06303177028894424, -0.0011156087275594473, 0.06900572776794434, 0.00630797166377306, 0.010950058698654175, 0.03933585807681084, 0.03983045741915703, -0.0337163545191288, 0.034862034022808075, -0.017430569976568222, -0.009826810099184513, 0.03917694836854935, 0.019877038896083832, 0.024440456181...
1,204
1,204
['David Held', 'Sebastian Thrun', 'Silvio Savarese']
1507.08286v1
Deep learning methods have typically been trained on large datasets in which many training examples are available. However, many real-world product datasets have only a small number of images available for each product. We explore the use of deep learning methods for recognizing object instances when we have only a sin...
Deep Learning for Single-View Instance Recognition
2,015
http://arxiv.org/pdf/1507.08286v1
Title Deep Learning SingleView Instance Recognition Summary Deep learning method typically trained large datasets many training example available However many realworld product datasets small number image available product explore use deep learning method recognizing object instance single training example per class sh...
[-0.03700682520866394, 0.0302085280418396, 0.07231617718935013, 0.059746719896793365, -0.007951096631586552, 0.015302322804927826, 0.06946630775928497, -0.009710571728646755, -0.008127566426992416, -0.02611890248954296, 0.00427075382322073, 0.04233561083674431, -0.01811395399272442, 0.06964488327503204, 0.0286777149885...
1,205
1,205
['D. Garcia-Gasulla', 'J. Béjar', 'U. Cortés', 'E. Ayguadé', 'J. Labarta', 'T. Suzumura', 'R. Chen']
1507.08818v6
Vector-space word representations obtained from neural network models have been shown to enable semantic operations based on vector arithmetic. In this paper, we explore the existence of similar information on vector representations of images. For that purpose we define a methodology to obtain large, sparse vector repr...
A Visual Embedding for the Unsupervised Extraction of Abstract Semantics
2,015
http://arxiv.org/pdf/1507.08818v6
Title Visual Embedding Unsupervised Extraction Abstract Semantics Summary Vectorspace word representation obtained neural network model shown enable semantic operation based vector arithmetic paper explore existence similar information vector representation image purpose define methodology obtain large sparse vector re...
[0.01893017813563347, 0.0445433147251606, -0.026213865727186203, 0.07263161242008209, -0.00889803096652031, 0.019370445981621742, 0.02691350318491459, 0.02719893679022789, -0.0438244603574276, -0.02506692335009575, 0.025498639792203903, 0.02854757569730282, 0.026448871940374374, -0.0012791512999683619, 0.04734561964869...
1,206
1,206
['Zhibin Liao', 'Gustavo Carneiro']
1508.00330v2
Deep feedforward neural networks with piecewise linear activations are currently producing the state-of-the-art results in several public datasets. The combination of deep learning models and piecewise linear activation functions allows for the estimation of exponentially complex functions with the use of a large numbe...
On the Importance of Normalisation Layers in Deep Learning with Piecewise Linear Activation Units
2,015
http://arxiv.org/pdf/1508.00330v2
Title Importance Normalisation Layers Deep Learning Piecewise Linear Activation Units Summary Deep feedforward neural network piecewise linear activation currently producing stateoftheart result several public datasets combination deep learning model piecewise linear activation function allows estimation exponentially ...
[-0.04846473038196564, 0.035292260348796844, -0.02174227498471737, 0.01446592714637518, 0.02305065281689167, 0.0014239975716918707, 0.09119880199432373, -0.0049864850006997585, -0.027966203168034554, 0.024907488375902176, -0.019258320331573486, -0.028377695009112358, -0.0119673702865839, 0.09649862349033356, 0.00079292...
1,207
1,207
['Pooya Khorrami', 'Tom Le Paine', 'Thomas S. Huang']
1510.02969v3
Despite being the appearance-based classifier of choice in recent years, relatively few works have examined how much convolutional neural networks (CNNs) can improve performance on accepted expression recognition benchmarks and, more importantly, examine what it is they actually learn. In this work, not only do we show...
Do Deep Neural Networks Learn Facial Action Units When Doing Expression Recognition?
2,015
http://arxiv.org/pdf/1510.02969v3
Title Deep Neural Networks Learn Facial Action Units Expression Recognition Summary Despite appearancebased classifier choice recent year relatively work examined much convolutional neural network CNNs improve performance accepted expression recognition benchmark importantly examine actually learn work show CNNs achiev...
[0.011459064669907093, 0.059427086263895035, -0.014143026433885098, 0.0437263585627079, 0.02084403857588768, 0.034001708030700684, 0.05372658371925354, 0.025187969207763672, -0.019521865993738174, 0.013856342062354088, -0.0031554359011352062, 0.015224515460431576, -0.0230898167937994, 0.070708729326725, 0.0646035298705...
1,208
1,208
['Jure Žbontar', 'Yann LeCun']
1510.05970v2
We present a method for extracting depth information from a rectified image pair. Our approach focuses on the first stage of many stereo algorithms: the matching cost computation. We approach the problem by learning a similarity measure on small image patches using a convolutional neural network. Training is carried ou...
Stereo Matching by Training a Convolutional Neural Network to Compare Image Patches
2,015
http://arxiv.org/pdf/1510.05970v2
Title Stereo Matching Training Convolutional Neural Network Compare Image Patches Summary present method extracting depth information rectified image pair approach focus first stage many stereo algorithm matching cost computation approach problem learning similarity measure small image patch using convolutional neural ...
[-0.0351475328207016, 0.05099889636039734, 0.017119554802775383, 0.0978991761803627, -0.024354256689548492, -0.015232251025736332, -0.0027228386607021093, -0.0017938062082976103, -0.04783998057246208, -0.006862140726298094, -0.014630568213760853, 0.029849525541067123, 0.05968781188130379, 0.07348904758691788, 0.0161718...
1,209
1,209
['Aleksandar Zlateski', 'Kisuk Lee', 'H. Sebastian Seung']
1510.06706v1
Convolutional networks (ConvNets) have become a popular approach to computer vision. It is important to accelerate ConvNet training, which is computationally costly. We propose a novel parallel algorithm based on decomposition into a set of tasks, most of which are convolutions or FFTs. Applying Brent's theorem to the ...
ZNN - A Fast and Scalable Algorithm for Training 3D Convolutional Networks on Multi-Core and Many-Core Shared Memory Machines
2,015
http://arxiv.org/pdf/1510.06706v1
Title ZNN Fast Scalable Algorithm Training 3D Convolutional Networks MultiCore ManyCore Shared Memory Machines Summary Convolutional network ConvNets become popular approach computer vision important accelerate ConvNet training computationally costly propose novel parallel algorithm based decomposition set task convolu...
[-0.0027172863483428955, -0.0008499150862917304, -0.0034768441691994667, 0.09231258183717728, -2.317625330761075e-05, 0.002218489535152912, 0.039021629840135574, -0.014027212746441364, -0.031511902809143066, 0.008283151313662529, -0.02940419502556324, 0.02358069457113743, 0.028758520260453224, 0.01312054879963398, 0.00...
1,210
1,210
['Vijay Badrinarayanan', 'Alex Kendall', 'Roberto Cipolla']
1511.00561v3
We present a novel and practical deep fully convolutional neural network architecture for semantic pixel-wise segmentation termed SegNet. This core trainable segmentation engine consists of an encoder network, a corresponding decoder network followed by a pixel-wise classification layer. The architecture of the encoder...
SegNet: A Deep Convolutional Encoder-Decoder Architecture for Image Segmentation
2,015
http://arxiv.org/pdf/1511.00561v3
Title SegNet Deep Convolutional EncoderDecoder Architecture Image Segmentation Summary present novel practical deep fully convolutional neural network architecture semantic pixelwise segmentation termed SegNet core trainable segmentation engine consists encoder network corresponding decoder network followed pixelwise c...
[-0.015823449939489365, -0.006357388570904732, 0.027102001011371613, 0.08989871293306351, -0.0036129863001406193, -0.009315085597336292, 0.021208327263593674, -0.03256989270448685, -0.04699869453907013, 0.019279733300209045, -0.007104781456291676, 0.08203595876693726, -0.02864234521985054, 0.027319790795445442, 0.00020...
1,211
1,211
['Yuke Zhu', 'Oliver Groth', 'Michael Bernstein', 'Li Fei-Fei']
1511.03416v4
We have seen great progress in basic perceptual tasks such as object recognition and detection. However, AI models still fail to match humans in high-level vision tasks due to the lack of capacities for deeper reasoning. Recently the new task of visual question answering (QA) has been proposed to evaluate a model's cap...
Visual7W: Grounded Question Answering in Images
2,015
http://arxiv.org/pdf/1511.03416v4
Title Visual7W Grounded Question Answering Images Summary seen great progress basic perceptual task object recognition detection However AI model still fail match human highlevel vision task due lack capacity deeper reasoning Recently new task visual question answering QA proposed evaluate model capacity deep image und...
[0.05107489973306656, -0.0074351998046040535, 0.004608576186001301, 0.061327192932367325, 0.006238750647753477, 0.014390667900443077, 0.004267389886081219, 0.018421735614538193, 0.018133854493498802, -0.034159690141677856, -0.02181611768901348, 0.021694866940379143, -0.005826032254844904, 0.0415089875459671, 0.02757996...
1,212
1,212
['Lionel Pibre', 'Pasquet Jérôme', 'Dino Ienco', 'Marc Chaumont']
1511.04855v2
Since the BOSS competition, in 2010, most steganalysis approaches use a learning methodology involving two steps: feature extraction, such as the Rich Models (RM), for the image representation, and use of the Ensemble Classifier (EC) for the learning step. In 2015, Qian et al. have shown that the use of a deep learning...
Deep learning is a good steganalysis tool when embedding key is reused for different images, even if there is a cover source-mismatch
2,015
http://arxiv.org/pdf/1511.04855v2
Title Deep learning good steganalysis tool embedding key reused different image even cover sourcemismatch Summary Since BOSS competition 2010 steganalysis approach use learning methodology involving two step feature extraction Rich Models RM image representation use Ensemble Classifier EC learning step 2015 Qian et al ...
[0.0007217373349703848, 0.03198404982686043, -0.004846740979701281, 0.06166112422943115, -0.06095695123076439, -0.033450644463300705, 0.08221230655908585, 0.033968161791563034, -0.015449803322553635, -0.011888159438967705, 0.03574667125940323, 0.04193343222141266, -0.0034147421829402447, 0.11006923764944077, 0.04086233...
1,213
1,213
['Ashesh Jain', 'Amir R. Zamir', 'Silvio Savarese', 'Ashutosh Saxena']
1511.05298v3
Deep Recurrent Neural Network architectures, though remarkably capable at modeling sequences, lack an intuitive high-level spatio-temporal structure. That is while many problems in computer vision inherently have an underlying high-level structure and can benefit from it. Spatio-temporal graphs are a popular tool for i...
Structural-RNN: Deep Learning on Spatio-Temporal Graphs
2,015
http://arxiv.org/pdf/1511.05298v3
Title StructuralRNN Deep Learning SpatioTemporal Graphs Summary Deep Recurrent Neural Network architecture though remarkably capable modeling sequence lack intuitive highlevel spatiotemporal structure many problem computer vision inherently underlying highlevel structure benefit Spatiotemporal graph popular tool imposi...
[-0.027727581560611725, 0.011094358749687672, 0.00748787447810173, 0.016364485025405884, -0.0006045926711522043, 0.0014146527973935008, 0.03624793142080307, -0.014842328615486622, -0.0188528373837471, -0.014680354855954647, 0.03022313304245472, -0.04048275202512741, 0.06506609171628952, 0.0663042962551117, 0.0269631296...
1,214
1,214
['Suraj Srinivas', 'R. Venkatesh Babu']
1511.05497v2
Deep neural networks with millions of parameters are at the heart of many state of the art machine learning models today. However, recent works have shown that models with much smaller number of parameters can also perform just as well. In this work, we introduce the problem of architecture-learning, i.e; learning the ...
Learning Neural Network Architectures using Backpropagation
2,015
http://arxiv.org/pdf/1511.05497v2
Title Learning Neural Network Architectures using Backpropagation Summary Deep neural network million parameter heart many state art machine learning model today However recent work shown model much smaller number parameter also perform well work introduce problem architecturelearning ie learning architecture neural ne...
[0.009205243550240993, 0.007152481470257044, -0.030795007944107056, 0.010060195811092854, 0.050279419869184494, -0.05333491414785385, 0.04178154468536377, -0.002634103875607252, -0.022237518802285194, 0.011714121326804161, -0.004620780702680349, 0.007795977406203747, 0.0006253026658669114, 0.034281179308891296, 0.05595...
1,215
1,215
['George Toderici', "Sean M. O'Malley", 'Sung Jin Hwang', 'Damien Vincent', 'David Minnen', 'Shumeet Baluja', 'Michele Covell', 'Rahul Sukthankar']
1511.06085v5
A large fraction of Internet traffic is now driven by requests from mobile devices with relatively small screens and often stringent bandwidth requirements. Due to these factors, it has become the norm for modern graphics-heavy websites to transmit low-resolution, low-bytecount image previews (thumbnails) as part of th...
Variable Rate Image Compression with Recurrent Neural Networks
2,015
http://arxiv.org/pdf/1511.06085v5
Title Variable Rate Image Compression Recurrent Neural Networks Summary large fraction Internet traffic driven request mobile device relatively small screen often stringent bandwidth requirement Due factor become norm modern graphicsheavy website transmit lowresolution lowbytecount image preview thumbnail part initial ...
[0.0018737510545179248, 0.028686663135886192, -0.019925139844417572, 0.042102329432964325, -0.015018021687865257, -0.018420454114675522, 0.04429816082119942, 0.02109697461128235, -0.0697450265288353, -0.020246099680662155, 0.01047572772949934, -0.018706675618886948, 0.0055426787585020065, 0.060231972485780716, 0.024832...
1,216
1,216
['Stefan Lee', 'Senthil Purushwalkam', 'Michael Cogswell', 'David Crandall', 'Dhruv Batra']
1511.06314v1
Convolutional Neural Networks have achieved state-of-the-art performance on a wide range of tasks. Most benchmarks are led by ensembles of these powerful learners, but ensembling is typically treated as a post-hoc procedure implemented by averaging independently trained models with model variation induced by bagging or...
Why M Heads are Better than One: Training a Diverse Ensemble of Deep Networks
2,015
http://arxiv.org/pdf/1511.06314v1
Title Heads Better One Training Diverse Ensemble Deep Networks Summary Convolutional Neural Networks achieved stateoftheart performance wide range task benchmark led ensemble powerful learner ensembling typically treated posthoc procedure implemented averaging independently trained model model variation induced bagging...
[-0.030866816639900208, 0.05782420560717583, -0.004393278621137142, 0.02829958312213421, 0.03423222899436951, 0.00753226550295949, 0.03818124160170555, -0.025428064167499542, -0.042238421738147736, 0.020469633862376213, -0.008617752231657505, -0.01240132749080658, 0.013209804892539978, 0.03680972754955292, 0.0112457703...
1,217
1,217
['Yani Ioannou', 'Duncan Robertson', 'Jamie Shotton', 'Roberto Cipolla', 'Antonio Criminisi']
1511.06744v3
We propose a new method for creating computationally efficient convolutional neural networks (CNNs) by using low-rank representations of convolutional filters. Rather than approximating filters in previously-trained networks with more efficient versions, we learn a set of small basis filters from scratch; during traini...
Training CNNs with Low-Rank Filters for Efficient Image Classification
2,015
http://arxiv.org/pdf/1511.06744v3
Title Training CNNs LowRank Filters Efficient Image Classification Summary propose new method creating computationally efficient convolutional neural network CNNs using lowrank representation convolutional filter Rather approximating filter previouslytrained network efficient version learn set small basis filter scratc...
[0.013833130709826946, 0.0065141646191477776, -0.0036438454408198595, 0.09312603622674942, 0.04233706742525101, -0.01859935373067856, 0.018550774082541466, 0.02881435863673687, 0.027457937598228455, -0.017694061622023582, -0.02891833893954754, -0.0006739930831827223, -0.00835569016635418, 0.01039261743426323, 0.0296271...
1,218
1,218
['Leslie N. Smith', 'Emily M. Hand', 'Timothy Doster']
1511.06951v1
We introduce the concept of dynamically growing a neural network during training. In particular, an untrainable deep network starts as a trainable shallow network and newly added layers are slowly, organically added during training, thereby increasing the network's depth. This is accomplished by a new layer, which we c...
Gradual DropIn of Layers to Train Very Deep Neural Networks
2,015
http://arxiv.org/pdf/1511.06951v1
Title Gradual DropIn Layers Train Deep Neural Networks Summary introduce concept dynamically growing neural network training particular untrainable deep network start trainable shallow network newly added layer slowly organically added training thereby increasing network depth accomplished new layer call DropIn DropIn ...
[0.007731180172413588, 0.05761919915676117, -0.04624626040458679, 0.021057158708572388, -5.801829684060067e-05, 0.006422598380595446, 0.0274411179125309, 0.0029014733154326677, -0.008146865293383598, 0.045804914087057114, 0.02467726357281208, 0.06007247790694237, -0.023067601025104523, 0.0986555814743042, 0.00203208439...
1,219
1,219
['Spyros Gidaris', 'Nikos Komodakis']
1511.07763v2
We propose a novel object localization methodology with the purpose of boosting the localization accuracy of state-of-the-art object detection systems. Our model, given a search region, aims at returning the bounding box of an object of interest inside this region. To accomplish its goal, it relies on assigning conditi...
LocNet: Improving Localization Accuracy for Object Detection
2,015
http://arxiv.org/pdf/1511.07763v2
Title LocNet Improving Localization Accuracy Object Detection Summary propose novel object localization methodology purpose boosting localization accuracy stateoftheart object detection system model given search region aim returning bounding box object interest inside region accomplish goal relies assigning conditional...
[0.006463375873863697, -0.017804117873311043, 0.006045058835297823, 0.0758776143193245, -0.011595061048865318, -0.02267525903880596, 0.015514788217842579, 0.01747891493141651, 0.014340521767735481, 0.006594728212803602, -0.027386587113142014, 0.03774680197238922, -0.009540103375911713, 0.047759633511304855, 0.000552312...
1,220
1,220
['Haibing Wu', 'Xiaodong Gu']
1512.00242v1
Recently, dropout has seen increasing use in deep learning. For deep convolutional neural networks, dropout is known to work well in fully-connected layers. However, its effect in convolutional and pooling layers is still not clear. This paper demonstrates that max-pooling dropout is equivalent to randomly picking acti...
Towards Dropout Training for Convolutional Neural Networks
2,015
http://arxiv.org/pdf/1512.00242v1
Title Towards Dropout Training Convolutional Neural Networks Summary Recently dropout seen increasing use deep learning deep convolutional neural network dropout known work well fullyconnected layer However effect convolutional pooling layer still clear paper demonstrates maxpooling dropout equivalent randomly picking ...
[0.005901362746953964, 0.0003758076927624643, -0.0017538379179313779, 0.043250299990177155, 0.02279280126094818, -0.01044152770191431, 0.0548686645925045, -0.012765503488481045, -0.013536915183067322, 0.010258102789521217, -0.007797673344612122, 0.027180422097444534, -0.008747848682105541, 0.05980594456195831, 0.012807...
1,221
1,221
['Edward Grant', 'Stephan Sahm', 'Mariam Zabihi', 'Marcel van Gerven']
1512.01289v2
Judgments about personality based on facial appearance are strong effectors in social decision making, and are known to have impact on areas from presidential elections to jury decisions. Recent work has shown that it is possible to predict perception of memorability, trustworthiness, intelligence and other attributes ...
Predicting and visualizing psychological attributions with a deep neural network
2,015
http://arxiv.org/pdf/1512.01289v2
Title Predicting visualizing psychological attribution deep neural network Summary Judgments personality based facial appearance strong effector social decision making known impact area presidential election jury decision Recent work shown possible predict perception memorability trustworthiness intelligence attribute ...
[0.02326471358537674, 0.09605322778224945, -0.04167664423584938, -0.01901482790708542, -0.0010711831273511052, 0.027963444590568542, 0.04361509531736374, -0.015402032062411308, -0.013899405486881733, 0.03444215655326843, -0.011700818315148354, -0.002669362584128976, 0.017736410722136497, 0.07271398603916168, 0.07164623...
1,222
1,222
['Haibing Wu', 'Xiaodong Gu']
1512.01400v1
Recently, dropout has seen increasing use in deep learning. For deep convolutional neural networks, dropout is known to work well in fully-connected layers. However, its effect in pooling layers is still not clear. This paper demonstrates that max-pooling dropout is equivalent to randomly picking activation based on a ...
Max-Pooling Dropout for Regularization of Convolutional Neural Networks
2,015
http://arxiv.org/pdf/1512.01400v1
Title MaxPooling Dropout Regularization Convolutional Neural Networks Summary Recently dropout seen increasing use deep learning deep convolutional neural network dropout known work well fullyconnected layer However effect pooling layer still clear paper demonstrates maxpooling dropout equivalent randomly picking activ...
[0.016897985711693764, -7.620026008225977e-05, 0.0044533731415867805, 0.031333647668361664, 0.0012570390244945884, -0.023930512368679047, 0.04078090190887451, -0.009809983894228935, -0.015115304850041866, 0.009812970645725727, 0.006984876934438944, 0.01681486889719963, -0.008236587047576904, 0.03719208762049675, 0.0213...
1,223
1,223
['Volodymyr Turchenko', 'Artur Luczak']
1512.01596v3
The development of a deep (stacked) convolutional auto-encoder in the Caffe deep learning framework is presented in this paper. We describe simple principles which we used to create this model in Caffe. The proposed model of convolutional auto-encoder does not have pooling/unpooling layers yet. The results of our exper...
Creation of a Deep Convolutional Auto-Encoder in Caffe
2,015
http://arxiv.org/pdf/1512.01596v3
Title Creation Deep Convolutional AutoEncoder Caffe Summary development deep stacked convolutional autoencoder Caffe deep learning framework presented paper describe simple principle used create model Caffe proposed model convolutional autoencoder poolingunpooling layer yet result experimental research show comparable ...
[-0.01935165748000145, 0.02691812813282013, -0.02918732911348343, 0.06044015288352966, 0.013700141571462154, -0.003019306343048811, 0.07921355962753296, -0.019282866269350052, -0.0870761051774025, 0.006803006865084171, -0.0018187519162893295, 0.020771779119968414, -0.00492615532130003, 0.060892872512340546, -0.00297425...
1,224
1,224
['Michael Maire', 'Takuya Narihira', 'Stella X. Yu']
1512.02767v2
Spectral embedding provides a framework for solving perceptual organization problems, including image segmentation and figure/ground organization. From an affinity matrix describing pairwise relationships between pixels, it clusters pixels into regions, and, using a complex-valued extension, orders pixels according to ...
Affinity CNN: Learning Pixel-Centric Pairwise Relations for Figure/Ground Embedding
2,015
http://arxiv.org/pdf/1512.02767v2
Title Affinity CNN Learning PixelCentric Pairwise Relations FigureGround Embedding Summary Spectral embedding provides framework solving perceptual organization problem including image segmentation figureground organization affinity matrix describing pairwise relationship pixel cluster pixel region using complexvalued ...
[-0.026018617674708366, 0.008031721226871014, -0.001296941307373345, 0.07457586377859116, -0.01237773708999157, -0.019648507237434387, -0.004271939396858215, -0.012492301873862743, -0.022054489701986313, 0.01679740473628044, -0.011167469434440136, 0.03816167265176773, -0.0008061739499680698, -0.00455609243363142, 0.028...
1,225
1,225
['K. Eswaran', 'K. Damodhar Rao']
1512.04509v2
Recently an algorithm, was discovered, which separates points in n-dimension by planes in such a manner that no two points are left un-separated by at least one plane{[}1-3{]}. By using this new algorithm we show that there are two ways of classification by a neural network, for a large dimension feature space, both of...
On non-iterative training of a neural classifier
2,015
http://arxiv.org/pdf/1512.04509v2
Title noniterative training neural classifier Summary Recently algorithm discovered separate point ndimension plane manner two point left unseparated least one plane13 using new algorithm show two way classification neural network large dimension feature space noniterative deterministic demonstrate power method apply e...
[-0.04375074803829193, 0.015573645941913128, -0.025961928069591522, 0.08972742408514023, -0.02186267450451851, 0.014111526310443878, 0.05056721717119217, -0.013636123389005661, -0.033608149737119675, -0.02154446765780449, 0.006811062339693308, 0.005885583348572254, 0.035941798239946365, 0.017500808462500572, 0.03784235...
1,226
1,226
['Jiaji Huang', 'Qiang Qiu', 'Robert Calderbank', 'Guillermo Sapiro']
1512.06757v2
Deep neural networks have proved very successful in domains where large training sets are available, but when the number of training samples is small, their performance suffers from overfitting. Prior methods of reducing overfitting such as weight decay, Dropout and DropConnect are data-independent. This paper proposes...
GraphConnect: A Regularization Framework for Neural Networks
2,015
http://arxiv.org/pdf/1512.06757v2
Title GraphConnect Regularization Framework Neural Networks Summary Deep neural network proved successful domain large training set available number training sample small performance suffers overfitting Prior method reducing overfitting weight decay Dropout DropConnect dataindependent paper proposes new method GraphCon...
[-0.019845163449645042, 0.032285433262586594, -0.006354076322168112, 0.03249770775437355, -0.001878057257272303, -0.07909730076789856, 0.015089893713593483, 0.011589949950575829, 0.04447934031486511, 0.04493793472647667, 0.01331805158406496, 0.027606826275587082, 0.0083230622112751, 0.06991183757781982, 0.0275301113724...
1,227
1,227
['Jiuxiang Gu', 'Zhenhua Wang', 'Jason Kuen', 'Lianyang Ma', 'Amir Shahroudy', 'Bing Shuai', 'Ting Liu', 'Xingxing Wang', 'Li Wang', 'Gang Wang', 'Jianfei Cai', 'Tsuhan Chen']
1512.07108v6
In the last few years, deep learning has led to very good performance on a variety of problems, such as visual recognition, speech recognition and natural language processing. Among different types of deep neural networks, convolutional neural networks have been most extensively studied. Leveraging on the rapid growth ...
Recent Advances in Convolutional Neural Networks
2,015
http://arxiv.org/pdf/1512.07108v6
Title Recent Advances Convolutional Neural Networks Summary last year deep learning led good performance variety problem visual recognition speech recognition natural language processing Among different type deep neural network convolutional neural network extensively studied Leveraging rapid growth amount annotated da...
[0.05359747260808945, 0.024194877594709396, -0.0008037922088988125, 0.09627276659011841, -0.00257887109182775, -0.036091335117816925, 0.03494212031364441, 0.030311036854982376, -0.0066754515282809734, -0.038316793739795685, -0.07044162601232529, 0.0075162118300795555, 0.020437592640519142, 0.05179395526647568, 0.035933...
1,228
1,228
['Zhangyang Wang', 'Shiyu Chang', 'Florin Dolcos', 'Diane Beck', 'Ding Liu', 'Thomas S. Huang']
1601.04155v2
Image aesthetics assessment has been challenging due to its subjective nature. Inspired by the scientific advances in the human visual perception and neuroaesthetics, we design Brain-Inspired Deep Networks (BDN) for this task. BDN first learns attributes through the parallel supervised pathways, on a variety of selecte...
Brain-Inspired Deep Networks for Image Aesthetics Assessment
2,016
http://arxiv.org/pdf/1601.04155v2
Title BrainInspired Deep Networks Image Aesthetics Assessment Summary Image aesthetic assessment challenging due subjective nature Inspired scientific advance human visual perception neuroaesthetics design BrainInspired Deep Networks BDN task BDN first learns attribute parallel supervised pathway variety selected featu...
[0.024955784901976585, 0.08123048394918442, -0.014758794568479061, 0.010384960100054741, -0.0019827804062515497, 0.018979554995894432, 0.024620572105050087, -0.014311722479760647, -0.05005025491118431, 0.01838209107518196, -0.06827984005212784, 0.008559571579098701, 0.03436597064137459, 0.03679542616009712, 0.013159986...
1,229
1,229
['Weiguang Ding', 'Graham Taylor']
1602.07383v1
Monitoring the number of insect pests is a crucial component in pheromone-based pest management systems. In this paper, we propose an automatic detection pipeline based on deep learning for identifying and counting pests in images taken inside field traps. Applied to a commercial codling moth dataset, our method shows ...
Automatic Moth Detection from Trap Images for Pest Management
2,016
http://arxiv.org/pdf/1602.07383v1
Title Automatic Moth Detection Trap Images Pest Management Summary Monitoring number insect pest crucial component pheromonebased pest management system paper propose automatic detection pipeline based deep learning identifying counting pest image taken inside field trap Applied commercial codling moth dataset method s...
[-0.03675941005349159, 0.02926117181777954, 0.025709254667162895, 0.022765884175896645, -0.05537954345345497, -0.059473756700754166, 0.012641625478863716, -0.05560461804270744, 0.00659280875697732, -0.0009184268419630826, 0.028968779370188713, 0.048867370933294296, -0.019907567650079727, 0.061927732080221176, 0.0139992...
1,230
1,230
['Shumeet Baluja']
1603.04000v1
Typography is a ubiquitous art form that affects our understanding, perception, and trust in what we read. Thousands of different font-faces have been created with enormous variations in the characters. In this paper, we learn the style of a font by analyzing a small subset of only four letters. From these four letters...
Learning Typographic Style
2,016
http://arxiv.org/pdf/1603.04000v1
Title Learning Typographic Style Summary Typography ubiquitous art form affect understanding perception trust read Thousands different fontfaces created enormous variation character paper learn style font analyzing small subset four letter four letter learn two task first discrimination task given four letter new candi...
[0.025311928242444992, 0.04977130889892578, -0.030461637303233147, 0.009247307665646076, -0.003765208413824439, 0.012525367550551891, 0.04215028136968613, 0.0364910364151001, -0.015570483170449734, 0.006841660477221012, 0.008558154106140137, -0.03144262358546257, 0.0500129871070385, 0.05901293829083443, 0.0281207393854...
1,231
1,231
['Adedotun Akintayo', 'Kin Gwn Lore', 'Soumalya Sarkar', 'Soumik Sarkar']
1603.07839v1
This paper proposes an end-to-end convolutional selective autoencoder approach for early detection of combustion instabilities using rapidly arriving flame image frames. The instabilities arising in combustion processes cause significant deterioration and safety issues in various human-engineered systems such as land a...
Early Detection of Combustion Instabilities using Deep Convolutional Selective Autoencoders on Hi-speed Flame Video
2,016
http://arxiv.org/pdf/1603.07839v1
Title Early Detection Combustion Instabilities using Deep Convolutional Selective Autoencoders Hispeed Flame Video Summary paper proposes endtoend convolutional selective autoencoder approach early detection combustion instability using rapidly arriving flame image frame instability arising combustion process cause sig...
[-0.06691711395978928, 0.05354198068380356, -0.009279065765440464, 0.06643844395875931, 0.024544352665543556, 0.02912520430982113, 0.00240775546990335, 0.05960454046726227, -0.05433112382888794, -0.039916910231113434, 0.04682253673672676, -0.009627824649214745, -0.0168723426759243, 0.06770304590463638, 0.01484145317226...
1,232
1,232
['Markus Thom', 'Günther Palm']
1603.08367v1
Sparseness is a useful regularizer for learning in a wide range of applications, in particular in neural networks. This paper proposes a model targeted at classification tasks, where sparse activity and sparse connectivity are used to enhance classification capabilities. The tool for achieving this is a sparseness-enfo...
Sparse Activity and Sparse Connectivity in Supervised Learning
2,016
http://arxiv.org/pdf/1603.08367v1
Title Sparse Activity Sparse Connectivity Supervised Learning Summary Sparseness useful regularizer learning wide range application particular neural network paper proposes model targeted classification task sparse activity sparse connectivity used enhance classification capability tool achieving sparsenessenforcing pr...
[-0.03592899069190025, 0.022208554670214653, -0.015088890679180622, 0.039058420807123184, 0.04725993424654007, -0.03798478841781616, 0.03636178746819496, 0.021143300458788872, -0.05912933126091957, -0.01127125695347786, -0.010733835399150848, 0.01797039993107319, 0.03514299914240837, 0.02981664054095745, 0.041079726070...
1,233
1,233
['Gao Huang', 'Yu Sun', 'Zhuang Liu', 'Daniel Sedra', 'Kilian Weinberger']
1603.09382v3
Very deep convolutional networks with hundreds of layers have led to significant reductions in error on competitive benchmarks. Although the unmatched expressiveness of the many layers can be highly desirable at test time, training very deep networks comes with its own set of challenges. The gradients can vanish, the f...
Deep Networks with Stochastic Depth
2,016
http://arxiv.org/pdf/1603.09382v3
Title Deep Networks Stochastic Depth Summary deep convolutional network hundred layer led significant reduction error competitive benchmark Although unmatched expressiveness many layer highly desirable test time training deep network come set challenge gradient vanish forward flow often diminishes training time painful...
[-0.010279844515025616, 0.04579173028469086, -0.004896120633929968, 0.05674741044640541, -0.0073264408856630325, -0.0472818948328495, 0.042658377438783646, -0.016029395163059235, -0.04169977083802223, 0.014062102884054184, 0.026848286390304565, 0.008819473907351494, 0.0012567774392664433, 0.07068894803524017, -0.000841...
1,234
1,234
['Shuai Zheng', 'Abhinav Vishnu', 'Chris Ding']
1605.01369v2
Deep Learning is a very powerful machine learning model. Deep Learning trains a large number of parameters for multiple layers and is very slow when data is in large scale and the architecture size is large. Inspired from the shrinking technique used in accelerating computation of Support Vector Machines (SVM) algorith...
Accelerating Deep Learning with Shrinkage and Recall
2,016
http://arxiv.org/pdf/1605.01369v2
Title Accelerating Deep Learning Shrinkage Recall Summary Deep Learning powerful machine learning model Deep Learning train large number parameter multiple layer slow data large scale architecture size large Inspired shrinking technique used accelerating computation Support Vector Machines SVM algorithm screening techn...
[-0.018599096685647964, 0.028540238738059998, -0.02879328466951847, 0.039566945284605026, -0.007782716769725084, -0.028863022103905678, 0.01361249852925539, 0.012104012072086334, 0.01318956259638071, 0.0012450627982616425, -0.019846074283123016, -0.0011515641817823052, 0.010489356704056263, 0.03564151003956795, -0.0091...
1,235
1,235
['Philipp Gysel']
1605.06402v1
Convolutional neural networks (CNN) have achieved major breakthroughs in recent years. Their performance in computer vision have matched and in some areas even surpassed human capabilities. Deep neural networks can capture complex non-linear features; however this ability comes at the cost of high computational and mem...
Ristretto: Hardware-Oriented Approximation of Convolutional Neural Networks
2,016
http://arxiv.org/pdf/1605.06402v1
Title Ristretto HardwareOriented Approximation Convolutional Neural Networks Summary Convolutional neural network CNN achieved major breakthrough recent year performance computer vision matched area even surpassed human capability Deep neural network capture complex nonlinear feature however ability come cost high comp...
[-0.02497262880206108, 0.06187088042497635, -0.027119766920804977, 0.06563989073038101, -0.00484851049259305, -0.024940812960267067, 0.07125523686408997, -0.009898168966174126, -0.04892946407198906, 0.005995097570121288, 0.00806562602519989, 0.0040030754171311855, 0.011462477967143059, 0.03946921229362488, 0.0260490756...
1,236
1,236
['Milad Mohammadi', 'Subhasis Das']
1605.08512v1
It has been proven that transfer learning provides an easy way to achieve state-of-the-art accuracies on several vision tasks by training a simple classifier on top of features obtained from pre-trained neural networks. The goal of this work is to generate better features for transfer learning from multiple publicly av...
SNN: Stacked Neural Networks
2,016
http://arxiv.org/pdf/1605.08512v1
Title SNN Stacked Neural Networks Summary proven transfer learning provides easy way achieve stateoftheart accuracy several vision task training simple classifier top feature obtained pretrained neural network goal work generate better feature transfer learning multiple publicly available pretrained neural network end ...
[-0.04533090814948082, 0.03276192024350166, -0.00010578485671430826, 0.04264337942004204, 0.014198139309883118, -0.007219626568257809, 0.052011869847774506, -0.02789374627172947, -0.04731709882616997, -0.00361702055670321, -0.0900450050830841, 0.046305689960718155, 0.04552968218922615, 0.035725757479667664, 0.019604390...
1,237
1,237
['Tim Salimans', 'Ian Goodfellow', 'Wojciech Zaremba', 'Vicki Cheung', 'Alec Radford', 'Xi Chen']
1606.03498v1
We present a variety of new architectural features and training procedures that we apply to the generative adversarial networks (GANs) framework. We focus on two applications of GANs: semi-supervised learning, and the generation of images that humans find visually realistic. Unlike most work on generative models, our p...
Improved Techniques for Training GANs
2,016
http://arxiv.org/pdf/1606.03498v1
Title Improved Techniques Training GANs Summary present variety new architectural feature training procedure apply generative adversarial network GANs framework focus two application GANs semisupervised learning generation image human find visually realistic Unlike work generative model primary goal train model assigns...
[0.00790507160127163, 0.07754165679216385, -0.006464189384132624, 0.045290105044841766, -0.00022948422702029347, -0.004433211870491505, 0.028524931520223618, -0.0028470002580434084, -0.006867441348731518, 0.04633037745952606, -0.04948092997074127, 0.025544341653585434, -0.05395067483186722, 0.027009515091776848, 0.0715...
1,238
1,238
['Kun He', 'Yan Wang', 'John Hopcroft']
1606.04801v2
To what extent is the success of deep visualization due to the training? Could we do deep visualization using untrained, random weight networks? To address this issue, we explore new and powerful generative models for three popular deep visualization tasks using untrained, random weight convolutional neural networks. F...
A Powerful Generative Model Using Random Weights for the Deep Image Representation
2,016
http://arxiv.org/pdf/1606.04801v2
Title Powerful Generative Model Using Random Weights Deep Image Representation Summary extent success deep visualization due training Could deep visualization using untrained random weight network address issue explore new powerful generative model three popular deep visualization task using untrained random weight con...
[-0.002128409221768379, 0.06832403689622879, -0.02799643576145172, 0.03999074921011925, -0.003075722139328718, 0.008872193284332752, 0.022194141522049904, -0.02055860497057438, -0.029334917664527893, 0.04254429414868355, -0.021199896931648254, 0.02316487953066826, -0.02739294059574604, 0.02755904011428356, 0.0155478017...
1,239
1,239
['Hengyuan Hu', 'Rui Peng', 'Yu-Wing Tai', 'Chi-Keung Tang']
1607.03250v1
State-of-the-art neural networks are getting deeper and wider. While their performance increases with the increasing number of layers and neurons, it is crucial to design an efficient deep architecture in order to reduce computational and memory costs. Designing an efficient neural network, however, is labor intensive ...
Network Trimming: A Data-Driven Neuron Pruning Approach towards Efficient Deep Architectures
2,016
http://arxiv.org/pdf/1607.03250v1
Title Network Trimming DataDriven Neuron Pruning Approach towards Efficient Deep Architectures Summary Stateoftheart neural network getting deeper wider performance increase increasing number layer neuron crucial design efficient deep architecture order reduce computational memory cost Designing efficient neural networ...
[-0.01882012188434601, 0.04524293541908264, -0.00882603507488966, 0.06347029656171799, 0.0020454837940633297, -0.04534580186009407, 0.025939280167222023, 0.021990377455949783, -0.06560501456260681, 0.04826058819890022, -0.02478449046611786, 0.024757875129580498, -0.010864495299756527, 0.05403487756848335, 0.03348536789...
1,240
1,240
['Christian Bailer', 'Kiran Varanasi', 'Didier Stricker']
1607.08064v3
Learning based approaches have not yet achieved their full potential in optical flow estimation, where their performance still trails heuristic approaches. In this paper, we present a CNN based patch matching approach for optical flow estimation. An important contribution of our approach is a novel thresholded loss for...
CNN-based Patch Matching for Optical Flow with Thresholded Hinge Embedding Loss
2,016
http://arxiv.org/pdf/1607.08064v3
Title CNNbased Patch Matching Optical Flow Thresholded Hinge Embedding Loss Summary Learning based approach yet achieved full potential optical flow estimation performance still trail heuristic approach paper present CNN based patch matching approach optical flow estimation important contribution approach novel thresho...
[-0.048210423439741135, 0.038108449429273605, 0.009967567399144173, 0.07074859738349915, -0.05375124514102936, -0.05146218091249466, 0.0552767850458622, 0.03851082921028137, -0.07135859876871109, -0.005664671305567026, 0.03291254863142967, 0.040145501494407654, 0.03029008023440838, 0.08703414350748062, 0.00930542778223...
1,241
1,241
['Priyadarshini Panda', 'Kaushik Roy']
1608.00611v1
One of the key challenges in machine learning is to design a computationally efficient multi-class classifier while maintaining the output accuracy and performance. In this paper, we present a tree-based classifier: Attention Tree (ATree) for large-scale image classification that uses recursive Adaboost training to con...
Attention Tree: Learning Hierarchies of Visual Features for Large-Scale Image Recognition
2,016
http://arxiv.org/pdf/1608.00611v1
Title Attention Tree Learning Hierarchies Visual Features LargeScale Image Recognition Summary One key challenge machine learning design computationally efficient multiclass classifier maintaining output accuracy performance paper present treebased classifier Attention Tree ATree largescale image classification us recu...
[0.0018988262163475156, 0.0030219191685318947, -0.02189488150179386, 0.05939989537000656, 0.030536331236362457, 0.019668323919177055, 0.013437477871775627, 0.014852151274681091, 0.0002752818400040269, -0.05663195252418518, -0.03965822979807854, -0.003996850922703743, 0.025973834097385406, 0.020865997299551964, 0.006839...
1,242
1,242
['Martin Simonovsky', 'Nikos Komodakis']
1608.02728v1
The focus of our work is speeding up evaluation of deep neural networks in retrieval scenarios, where conventional architectures may spend too much time on negative examples. We propose to replace a monolithic network with our novel cascade of feature-sharing deep classifiers, called OnionNet, where subsequent stages m...
OnionNet: Sharing Features in Cascaded Deep Classifiers
2,016
http://arxiv.org/pdf/1608.02728v1
Title OnionNet Sharing Features Cascaded Deep Classifiers Summary focus work speeding evaluation deep neural network retrieval scenario conventional architecture may spend much time negative example propose replace monolithic network novel cascade featuresharing deep classifier called OnionNet subsequent stage may add ...
[-0.025480329990386963, 0.08324232697486877, -0.001398551627062261, 0.07479862868785858, -0.045973315834999084, -0.012426000088453293, 0.07170754671096802, 0.017373129725456238, 0.013058639131486416, -0.009118038229644299, -0.02865036576986313, 0.038936637341976166, -0.05493044853210449, 0.06826525181531906, -0.0517177...
1,243
1,243
['Yiwen Guo', 'Anbang Yao', 'Yurong Chen']
1608.04493v2
Deep learning has become a ubiquitous technology to improve machine intelligence. However, most of the existing deep models are structurally very complex, making them difficult to be deployed on the mobile platforms with limited computational power. In this paper, we propose a novel network compression method called dy...
Dynamic Network Surgery for Efficient DNNs
2,016
http://arxiv.org/pdf/1608.04493v2
Title Dynamic Network Surgery Efficient DNNs Summary Deep learning become ubiquitous technology improve machine intelligence However existing deep model structurally complex making difficult deployed mobile platform limited computational power paper propose novel network compression method called dynamic network surger...
[-0.006695954594761133, 0.04581962525844574, -0.0478852316737175, -0.010604342445731163, -0.014581057243049145, -0.030758507549762726, 0.0069971345365047455, -0.007448140066117048, -0.08795372396707535, 0.019251277670264244, 0.02410922572016716, 0.013347893953323364, 0.05058887600898743, 0.07207558304071426, 0.00647721...
1,244
1,244
['Martin Simonovsky', 'Benjamín Gutiérrez-Becker', 'Diana Mateus', 'Nassir Navab', 'Nikos Komodakis']
1609.05396v1
Multimodal registration is a challenging problem in medical imaging due the high variability of tissue appearance under different imaging modalities. The crucial component here is the choice of the right similarity measure. We make a step towards a general learning-based solution that can be adapted to specific situati...
A Deep Metric for Multimodal Registration
2,016
http://arxiv.org/pdf/1609.05396v1
Title Deep Metric Multimodal Registration Summary Multimodal registration challenging problem medical imaging due high variability tissue appearance different imaging modality crucial component choice right similarity measure make step towards general learningbased solution adapted specific situation present metric bas...
[-0.0062970793806016445, 0.026432298123836517, -0.005144237540662289, -0.009815048426389694, -0.020571034401655197, 0.012395408004522324, 0.035873934626579285, 0.025956571102142334, -0.028962887823581696, 0.01889321208000183, -0.024078000336885452, -0.10039685666561127, 0.026016797870397568, 0.04106814041733742, 0.0533...
1,245
1,245
['Rasool Fakoor', 'Abdel-rahman Mohamed', 'Margaret Mitchell', 'Sing Bing Kang', 'Pushmeet Kohli']
1611.02261v4
We present a method to improve video description generation by modeling higher-order interactions between video frames and described concepts. By storing past visual attention in the video associated to previously generated words, the system is able to decide what to look at and describe in light of what it has already...
Memory-augmented Attention Modelling for Videos
2,016
http://arxiv.org/pdf/1611.02261v4
Title Memoryaugmented Attention Modelling Videos Summary present method improve video description generation modeling higherorder interaction video frame described concept storing past visual attention video associated previously generated word system able decide look describe light already looked described enables eff...
[0.0396299734711647, -0.007141186390072107, -0.00021379107784014195, 0.04924452304840088, 0.017194844782352448, -0.0006667841225862503, 0.01020939089357853, -0.012556235305964947, -0.03456626832485199, -0.0841597467660904, 0.0023134704679250717, -0.08068519830703735, 0.023881971836090088, 0.06138698756694794, 0.0318479...
1,246
1,246
['Jason Kuen', 'Xiangfei Kong', 'Gang Wang', 'Yap-Peng Tan']
1611.05552v5
Deluge Networks (DelugeNets) are deep neural networks which efficiently facilitate massive cross-layer information inflows from preceding layers to succeeding layers. The connections between layers in DelugeNets are established through cross-layer depthwise convolutional layers with learnable filters, acting as a flexi...
DelugeNets: Deep Networks with Efficient and Flexible Cross-layer Information Inflows
2,016
http://arxiv.org/pdf/1611.05552v5
Title DelugeNets Deep Networks Efficient Flexible Crosslayer Information Inflows Summary Deluge Networks DelugeNets deep neural network efficiently facilitate massive crosslayer information inflow preceding layer succeeding layer connection layer DelugeNets established crosslayer depthwise convolutional layer learnable...
[-0.048334017395973206, 0.02082235738635063, -0.023733941838145256, 0.015828952193260193, -0.018368760123848915, -0.017619770020246506, 0.07667647302150726, -0.018047496676445007, -0.050517670810222626, 0.020544596016407013, -0.020787661895155907, 0.017004692927002907, -0.023921053856611252, 0.06855510175228119, 0.0013...
1,247
1,247
['Jose M Alvarez', 'Mathieu Salzmann']
1611.06321v2
Nowadays, the number of layers and of neurons in each layer of a deep network are typically set manually. While very deep and wide networks have proven effective in general, they come at a high memory and computation cost, thus making them impractical for constrained platforms. These networks, however, are known to hav...
Learning the Number of Neurons in Deep Networks
2,016
http://arxiv.org/pdf/1611.06321v2
Title Learning Number Neurons Deep Networks Summary Nowadays number layer neuron layer deep network typically set manually deep wide network proven effective general come high memory computation cost thus making impractical constrained platform network however known many redundant parameter could thus principle replace...
[-0.0319480262696743, 0.03335307538509369, -0.017721883952617645, 0.05101701244711876, 0.03260786086320877, -0.03915951028466225, 0.03651270642876625, -0.04963458329439163, -0.02142641879618168, 0.030403725802898407, 0.0038869809359312057, 0.015116853639483452, -3.600350100896321e-05, 0.007221557199954987, 0.0531693510...
1,248
1,248
['Tsung-Wei Ke', 'Michael Maire', 'Stella X. Yu']
1611.07661v2
We propose a multigrid extension of convolutional neural networks (CNNs). Rather than manipulating representations living on a single spatial grid, our network layers operate across scale space, on a pyramid of grids. They consume multigrid inputs and produce multigrid outputs; convolutional filters themselves have bot...
Multigrid Neural Architectures
2,016
http://arxiv.org/pdf/1611.07661v2
Title Multigrid Neural Architectures Summary propose multigrid extension convolutional neural network CNNs Rather manipulating representation living single spatial grid network layer operate across scale space pyramid grid consume multigrid input produce multigrid output convolutional filter withinscale crossscale exte...
[0.02347070351243019, 0.040291331708431244, -0.02252259850502014, 0.06388084590435028, -0.0015876631950959563, 0.018364202231168747, 0.1017490029335022, -0.04520164057612419, 0.004124898929148912, 0.034628696739673615, -0.03358730301260948, 0.0078040133230388165, -0.015620511956512928, 0.052562206983566284, 0.068220809...
1,249
1,249
['Yanghao Li', 'Naiyan Wang', 'Jiaying Liu', 'Xiaodi Hou']
1701.01036v2
Neural Style Transfer has recently demonstrated very exciting results which catches eyes in both academia and industry. Despite the amazing results, the principle of neural style transfer, especially why the Gram matrices could represent style remains unclear. In this paper, we propose a novel interpretation of neural ...
Demystifying Neural Style Transfer
2,017
http://arxiv.org/pdf/1701.01036v2
Title Demystifying Neural Style Transfer Summary Neural Style Transfer recently demonstrated exciting result catch eye academia industry Despite amazing result principle neural style transfer especially Gram matrix could represent style remains unclear paper propose novel interpretation neural style transfer treating d...
[0.0018681659130379558, 0.041154734790325165, -0.019428160041570663, 0.03537003695964813, -0.016766520217061043, 0.005738838575780392, -0.0003907916834577918, 0.0451369434595108, -0.08303647488355637, -0.0006824033916927874, -0.05712631344795227, -0.017913484945893288, 0.05085799843072891, 0.02553568407893181, 0.055940...
1,250
1,250
['Volodymyr Turchenko', 'Eric Chalmers', 'Artur Luczak']
1701.04949v1
This paper presents the development of several models of a deep convolutional auto-encoder in the Caffe deep learning framework and their experimental evaluation on the example of MNIST dataset. We have created five models of a convolutional auto-encoder which differ architecturally by the presence or absence of poolin...
A Deep Convolutional Auto-Encoder with Pooling - Unpooling Layers in Caffe
2,017
http://arxiv.org/pdf/1701.04949v1
Title Deep Convolutional AutoEncoder Pooling Unpooling Layers Caffe Summary paper present development several model deep convolutional autoencoder Caffe deep learning framework experimental evaluation example MNIST dataset created five model convolutional autoencoder differ architecturally presence absence pooling unpo...
[-0.026916606351733208, 0.016761701554059982, -0.02285112999379635, 0.0665423721075058, 0.026855766773223877, -0.0014077763771638274, 0.06978847831487656, -0.018267475068569183, -0.07678340375423431, 0.010368173941969872, -0.018519587814807892, 0.04988749697804451, -0.004826694261282682, 0.04490241780877113, -0.0072960...
1,251
1,251
['Mete Ozay', 'Takayuki Okatani']
1701.06123v2
Recent advances in optimization methods used for training convolutional neural networks (CNNs) with kernels, which are normalized according to particular constraints, have shown remarkable success. This work introduces an approach for training CNNs using ensembles of joint spaces of kernels constructed using different ...
Optimization on Product Submanifolds of Convolution Kernels
2,017
http://arxiv.org/pdf/1701.06123v2
Title Optimization Product Submanifolds Convolution Kernels Summary Recent advance optimization method used training convolutional neural network CNNs kernel normalized according particular constraint shown remarkable success work introduces approach training CNNs using ensemble joint space kernel constructed using dif...
[0.0029807635582983494, -0.002941632643342018, 0.0029629445634782314, 0.09010506421327591, 0.011009071953594685, 0.010367858223617077, 0.023977885022759438, -0.019275924190878868, -0.0011275589931756258, 0.03760818764567375, -0.02370557188987732, -0.019632741808891296, -0.010397171601653099, 0.021992480382323265, 0.047...
1,252
1,252
['Di Xie', 'Jiang Xiong', 'Shiliang Pu']
1703.01827v3
Deep neural network is difficult to train and this predicament becomes worse as the depth increases. The essence of this problem exists in the magnitude of backpropagated errors that will result in gradient vanishing or exploding phenomenon. We show that a variant of regularizer which utilizes orthonormality among diff...
All You Need is Beyond a Good Init: Exploring Better Solution for Training Extremely Deep Convolutional Neural Networks with Orthonormality and Modulation
2,017
http://arxiv.org/pdf/1703.01827v3
Title Need Beyond Good Init Exploring Better Solution Training Extremely Deep Convolutional Neural Networks Orthonormality Modulation Summary Deep neural network difficult train predicament becomes worse depth increase essence problem exists magnitude backpropagated error result gradient vanishing exploding phenomenon ...
[-0.0031870584934949875, 0.08209974318742752, -0.0010925524402409792, 0.06085912510752678, 0.05100148171186447, -0.011209608986973763, 0.012526903301477432, 0.0075141736306250095, -0.06407284736633301, 0.03482397645711899, -0.06920332461595535, 0.0718940794467926, -0.011172013357281685, -0.006591996643692255, 0.0229009...
1,253
1,253
['Tharindu Fernando', 'Simon Denman', 'Aaron McFadyen', 'Sridha Sridharan', 'Clinton Fookes']
1703.04706v1
In the domain of sequence modelling, Recurrent Neural Networks (RNN) have been capable of achieving impressive results in a variety of application areas including visual question answering, part-of-speech tagging and machine translation. However this success in modelling short term dependencies has not successfully tra...
Tree Memory Networks for Modelling Long-term Temporal Dependencies
2,017
http://arxiv.org/pdf/1703.04706v1
Title Tree Memory Networks Modelling Longterm Temporal Dependencies Summary domain sequence modelling Recurrent Neural Networks RNN capable achieving impressive result variety application area including visual question answering partofspeech tagging machine translation However success modelling short term dependency su...
[0.0025645678397268057, 0.041818492114543915, -0.014846110716462135, 0.032753944396972656, -0.03767929598689079, -0.0027936280239373446, -0.008567067794501781, -0.032484278082847595, -0.0760747417807579, -0.03437585011124611, 0.034454211592674255, -0.06219428405165672, 0.029815437272191048, 0.06776227056980133, -0.0109...
1,254
1,254
['Timo Hackel', 'Nikolay Savinov', 'Lubor Ladicky', 'Jan D. Wegner', 'Konrad Schindler', 'Marc Pollefeys']
1704.03847v1
This paper presents a new 3D point cloud classification benchmark data set with over four billion manually labelled points, meant as input for data-hungry (deep) learning methods. We also discuss first submissions to the benchmark that use deep convolutional neural networks (CNNs) as a work horse, which already show re...
Semantic3D.net: A new Large-scale Point Cloud Classification Benchmark
2,017
http://arxiv.org/pdf/1704.03847v1
Title Semantic3Dnet new Largescale Point Cloud Classification Benchmark Summary paper present new 3D point cloud classification benchmark data set four billion manually labelled point meant input datahungry deep learning method also discus first submission benchmark use deep convolutional neural network CNNs work horse...
[-0.02490987256169319, 0.012011982500553131, 0.036000609397888184, 0.07820869982242584, 0.02205965854227543, 0.025714172050356865, 0.03594175726175308, -0.05654573440551758, -0.06215629354119301, 0.011109132319688797, -0.02479170821607113, 0.07157041132450104, 0.009195130318403244, 0.04298366233706474, 0.02099038474261...
1,255
1,255
['Justin Lazarow', 'Long Jin', 'Zhuowen Tu']
1704.07820v1
We study unsupervised learning by developing introspective generative modeling (IGM) that attains a generator using progressively learned deep convolutional neural networks. The generator is itself a discriminator, capable of introspection: being able to self-evaluate the difference between its generated samples and th...
Introspective Generative Modeling: Decide Discriminatively
2,017
http://arxiv.org/pdf/1704.07820v1
Title Introspective Generative Modeling Decide Discriminatively Summary study unsupervised learning developing introspective generative modeling IGM attains generator using progressively learned deep convolutional neural network generator discriminator capable introspection able selfevaluate difference generated sample...
[0.019453367218375206, 0.07172452658414841, -0.02202647365629673, 0.07391197979450226, -0.01537074614316225, 0.008400761522352695, 0.05157385393977165, -0.02893260307610035, -0.01746879145503044, 0.027252964675426483, -0.0040261270478367805, 0.011639992706477642, -0.04287824407219887, 0.013733461499214172, 0.0148847941...
1,256
1,256
['Fabrizio Pedersoli', 'George Tzanetakis', 'Andrea Tagliasacchi']
1705.07175v2
There are many applications scenarios for which the computational performance and memory footprint of the prediction phase of Deep Neural Networks (DNNs) needs to be optimized. Binary Neural Networks (BDNNs) have been shown to be an effective way of achieving this objective. In this paper, we show how Convolutional Neu...
Espresso: Efficient Forward Propagation for BCNNs
2,017
http://arxiv.org/pdf/1705.07175v2
Title Espresso Efficient Forward Propagation BCNNs Summary many application scenario computational performance memory footprint prediction phase Deep Neural Networks DNNs need optimized Binary Neural Networks BDNNs shown effective way achieving objective paper show Convolutional Neural Networks CNNs implemented using b...
[-0.016200853511691093, 0.028504690155386925, -0.018677322193980217, 0.04408161714673042, -0.027021383866667747, -0.002883989829570055, 0.07190146297216415, 0.003577472409233451, -0.028449367731809616, -0.0029134268406778574, 0.03664013743400574, 0.003072368912398815, -0.004900366999208927, 0.08810308575630188, 0.00768...
1,257
1,257
['Haojin Yang', 'Martin Fritzsche', 'Christian Bartz', 'Christoph Meinel']
1705.09864v1
Binary Neural Networks (BNNs) can drastically reduce memory size and accesses by applying bit-wise operations instead of standard arithmetic operations. Therefore it could significantly improve the efficiency and lower the energy consumption at runtime, which enables the application of state-of-the-art deep learning mo...
BMXNet: An Open-Source Binary Neural Network Implementation Based on MXNet
2,017
http://arxiv.org/pdf/1705.09864v1
Title BMXNet OpenSource Binary Neural Network Implementation Based MXNet Summary Binary Neural Networks BNNs drastically reduce memory size access applying bitwise operation instead standard arithmetic operation Therefore could significantly improve efficiency lower energy consumption runtime enables application stateo...
[-0.05149267986416817, 0.04199634864926338, -0.01881209760904312, 0.05497327446937561, -0.01935211569070816, -0.028668515384197235, 0.07538139820098877, 0.010077118873596191, -0.05978301167488098, 0.00014186944463290274, -0.007982340641319752, 0.005640218034386635, -0.016821784898638725, 0.07750771939754486, 0.01979602...
1,258
1,258
['Jean-Charles Vialatte', 'Vincent Gripon', 'Gilles Coppin']
1706.02684v3
We propose a simple and generic layer formulation that extends the properties of convolutional layers to any domain that can be described by a graph. Namely, we use the support of its adjacency matrix to design learnable weight sharing filters able to exploit the underlying structure of signals in the same fashion as f...
Learning Local Receptive Fields and their Weight Sharing Scheme on Graphs
2,017
http://arxiv.org/pdf/1706.02684v3
Title Learning Local Receptive Fields Weight Sharing Scheme Graphs Summary propose simple generic layer formulation extends property convolutional layer domain described graph Namely use support adjacency matrix design learnable weight sharing filter able exploit underlying structure signal fashion image proposed formu...
[0.030468454584479332, 0.0005635989364236593, 0.011782009154558182, 0.06126868352293968, -0.01075859647244215, -0.052926141768693924, 0.05323818698525429, 0.034196887165308, 0.06526892632246017, 0.048263028264045715, -0.019077030941843987, 0.0367056280374527, -0.02867376059293747, 0.06673018634319305, 0.010753354988992...
1,259
1,259
['Yuntao Chen', 'Naiyan Wang', 'Zhaoxiang Zhang']
1707.01220v2
We have witnessed rapid evolution of deep neural network architecture design in the past years. These latest progresses greatly facilitate the developments in various areas such as computer vision and natural language processing. However, along with the extraordinary performance, these state-of-the-art models also brin...
DarkRank: Accelerating Deep Metric Learning via Cross Sample Similarities Transfer
2,017
http://arxiv.org/pdf/1707.01220v2
Title DarkRank Accelerating Deep Metric Learning via Cross Sample Similarities Transfer Summary witnessed rapid evolution deep neural network architecture design past year latest progress greatly facilitate development various area computer vision natural language processing However along extraordinary performance stat...
[0.002645194763317704, 0.021841926500201225, -0.015017774887382984, 0.06543312966823578, -0.0023301439359784126, 0.025294283404946327, 0.007461649365723133, 0.029021771624684334, 0.002284357091411948, -0.024278344586491585, -0.034518204629421234, -0.011612228117883205, -0.0004274074744898826, -0.013436558656394482, 0.0...
1,260
1,260
['Surat Teerapittayanon', 'Bradley McDanel', 'H. T. Kung']
1709.01686v1
Deep neural networks are state of the art methods for many learning tasks due to their ability to extract increasingly better features at each network layer. However, the improved performance of additional layers in a deep network comes at the cost of added latency and energy usage in feedforward inference. As networks...
BranchyNet: Fast Inference via Early Exiting from Deep Neural Networks
2,017
http://arxiv.org/pdf/1709.01686v1
Title BranchyNet Fast Inference via Early Exiting Deep Neural Networks Summary Deep neural network state art method many learning task due ability extract increasingly better feature network layer However improved performance additional layer deep network come cost added latency energy usage feedforward inference netwo...
[-0.014560464769601822, -0.004123300313949585, -0.021990522742271423, 0.01897454634308815, 0.010444418527185917, -0.006409340538084507, 0.03740128129720688, -0.004397416952997446, -0.015668360516428947, 0.033885929733514786, -0.002063640160486102, 0.04671047255396843, 0.012329970486462116, 0.040317635983228683, -0.0013...
1,261
1,261
['Lars Hertel', 'Erhardt Barth', 'Thomas Käster', 'Thomas Martinetz']
1710.02286v1
Recognizing objects in natural images is an intricate problem involving multiple conflicting objectives. Deep convolutional neural networks, trained on large datasets, achieve convincing results and are currently the state-of-the-art approach for this task. However, the long time needed to train such deep networks is a...
Deep Convolutional Neural Networks as Generic Feature Extractors
2,017
http://arxiv.org/pdf/1710.02286v1
Title Deep Convolutional Neural Networks Generic Feature Extractors Summary Recognizing object natural image intricate problem involving multiple conflicting objective Deep convolutional neural network trained large datasets achieve convincing result currently stateoftheart approach task However long time needed train ...
[0.055276669561862946, 0.08531878143548965, 0.02434695139527321, 0.08176027983427048, -0.006701956037431955, -0.0053156595677137375, 0.03818340599536896, 0.00768188014626503, -0.03675878793001175, -0.020023290067911148, -0.05037746950984001, 0.020229918882250786, -0.0051667108200490475, 0.05279255658388138, 0.004422547...
1,262
1,262
['Amir R. Zamir', 'Tilman Wekel', 'Pulkit Argrawal', 'Colin Weil', 'Jitendra Malik', 'Silvio Savarese']
1710.08247v1
Though a large body of computer vision research has investigated developing generic semantic representations, efforts towards developing a similar representation for 3D has been limited. In this paper, we learn a generic 3D representation through solving a set of foundational proxy 3D tasks: object-centric camera pose ...
Generic 3D Representation via Pose Estimation and Matching
2,017
http://arxiv.org/pdf/1710.08247v1
Title Generic 3D Representation via Pose Estimation Matching Summary Though large body computer vision research investigated developing generic semantic representation effort towards developing similar representation 3D limited paper learn generic 3D representation solving set foundational proxy 3D task objectcentric c...
[0.0035579355899244547, 0.008825414814054966, 0.04246417060494423, 0.07153753936290741, 0.018051359802484512, 0.026438191533088684, -0.0189497172832489, 0.0006842162692919374, -0.02422534115612507, -0.06660788506269455, -0.032758746296167374, 0.016573769971728325, 0.020808732137084007, 0.02202565036714077, 0.0511442534...
1,263
1,263
['Wentao Zhu', 'Xiang Xiang', 'Trac D. Tran', 'Gregory D. Hager', 'Xiaohui Xie']
1710.09288v2
Mass segmentation provides effective morphological features which are important for mass diagnosis. In this work, we propose a novel end-to-end network for mammographic mass segmentation which employs a fully convolutional network (FCN) to model a potential function, followed by a CRF to perform structured learning. Be...
Adversarial Deep Structured Nets for Mass Segmentation from Mammograms
2,017
http://arxiv.org/pdf/1710.09288v2
Title Adversarial Deep Structured Nets Mass Segmentation Mammograms Summary Mass segmentation provides effective morphological feature important mass diagnosis work propose novel endtoend network mammographic mass segmentation employ fully convolutional network FCN model potential function followed CRF perform structur...
[0.005084969103336334, 0.0478687658905983, -0.023579174652695656, 0.0025526713579893112, 0.021077895537018776, -0.006411607842892408, 0.039174992591142654, -0.0019371509552001953, -0.02620006911456585, 0.08844754844903946, 0.017531489953398705, -0.013856551609933376, -0.042832471430301666, 0.0659298375248909, -0.006094...
1,264
1,264
['Abram L. Friesen', 'Pedro Domingos']
1710.11573v2
As neural networks grow deeper and wider, learning networks with hard-threshold activations is becoming increasingly important, both for network quantization, which can drastically reduce time and energy requirements, and for creating large integrated systems of deep networks, which may have non-differentiable componen...
Deep Learning as a Mixed Convex-Combinatorial Optimization Problem
2,017
http://arxiv.org/pdf/1710.11573v2
Title Deep Learning Mixed ConvexCombinatorial Optimization Problem Summary neural network grow deeper wider learning network hardthreshold activation becoming increasingly important network quantization drastically reduce time energy requirement creating large integrated system deep network may nondifferentiable compon...
[-0.024745939299464226, 0.048497576266527176, -0.010489068925380707, 0.04372246935963631, 0.00037308523315005004, -0.03732449933886528, 0.02477993257343769, 0.010313551872968674, -0.03590583801269531, -0.007668009493499994, -0.01979215256869793, -0.025559019297361374, -0.00286197941750288, 0.035007186233997345, 0.01519...
1,265
1,265
['Clemens Rosenbaum', 'Tim Klinger', 'Matthew Riemer']
1711.01239v2
Multi-task learning (MTL) with neural networks leverages commonalities in tasks to improve performance, but often suffers from task interference which reduces the benefits of transfer. To address this issue we introduce the routing network paradigm, a novel neural network and training algorithm. A routing network is a ...
Routing Networks: Adaptive Selection of Non-linear Functions for Multi-Task Learning
2,017
http://arxiv.org/pdf/1711.01239v2
Title Routing Networks Adaptive Selection Nonlinear Functions MultiTask Learning Summary Multitask learning MTL neural network leverage commonality task improve performance often suffers task interference reduces benefit transfer address issue introduce routing network paradigm novel neural network training algorithm r...
[-0.03320518508553505, 0.018011756241321564, -0.039724837988615036, 0.0017999497940763831, -0.016830507665872574, -0.04531371220946312, 0.06524404138326645, -0.04493222385644913, -0.03968140855431557, -0.03363380953669548, -0.048899583518505096, 0.024483533576130867, -0.013164801523089409, 0.04224236309528351, 0.025904...
1,266
1,266
['Asit Mishra', 'Debbie Marr']
1711.05852v1
Deep learning networks have achieved state-of-the-art accuracies on computer vision workloads like image classification and object detection. The performant systems, however, typically involve big models with numerous parameters. Once trained, a challenging aspect for such top performing models is deployment on resourc...
Apprentice: Using Knowledge Distillation Techniques To Improve Low-Precision Network Accuracy
2,017
http://arxiv.org/pdf/1711.05852v1
Title Apprentice Using Knowledge Distillation Techniques Improve LowPrecision Network Accuracy Summary Deep learning network achieved stateoftheart accuracy computer vision workload like image classification object detection performant system however typically involve big model numerous parameter trained challenging as...
[-0.012383731082081795, 0.03359812870621681, -0.004361429251730442, 0.07009915262460709, 0.008439279161393642, -0.025960611179471016, 0.012676875106990337, 0.038575172424316406, -0.05150830000638962, 0.015350639820098877, -0.00023165768652688712, 0.022468218579888344, 0.016864120960235596, 0.035549335181713104, -0.0147...
1,267
1,267
['Assaf Shocher', 'Nadav Cohen', 'Michal Irani']
1712.06087v1
Deep Learning has led to a dramatic leap in Super-Resolution (SR) performance in the past few years. However, being supervised, these SR methods are restricted to specific training data, where the acquisition of the low-resolution (LR) images from their high-resolution (HR) counterparts is predetermined (e.g., bicubic ...
"Zero-Shot" Super-Resolution using Deep Internal Learning
2,017
http://arxiv.org/pdf/1712.06087v1
Title ZeroShot SuperResolution using Deep Internal Learning Summary Deep Learning led dramatic leap SuperResolution SR performance past year However supervised SR method restricted specific training data acquisition lowresolution LR image highresolution HR counterpart predetermined eg bicubic downscaling without distra...
[-0.005256193690001965, 0.014045723713934422, 0.018430551514029503, 0.028767792508006096, 0.0009444840252399445, -0.048796772956848145, 0.02405684068799019, 0.03092619962990284, -0.05667911469936371, 0.018793640658259392, 0.025771131739020348, 0.04578088968992233, -0.002007536357268691, -0.017837952822446823, 0.0489461...
1,268
1,268
['Brian Kenji Iwana', 'Seiichi Uchida']
1712.06530v2
In this paper, we propose a method of improving Convolutional Neural Networks (CNN) by determining the optimal alignment of weights and inputs using dynamic programming. Conventional CNNs convolve learnable shared weights, or filters, across the input data. The filters use a linear matching of weights to inputs using a...
Dynamic Weight Alignment for Convolutional Neural Networks
2,017
http://arxiv.org/pdf/1712.06530v2
Title Dynamic Weight Alignment Convolutional Neural Networks Summary paper propose method improving Convolutional Neural Networks CNN determining optimal alignment weight input using dynamic programming Conventional CNNs convolve learnable shared weight filter across input data filter use linear matching weight input u...
[-0.017075086012482643, 0.05346205458045006, -0.00879612471908331, 0.025621332228183746, 0.008343745023012161, 0.014775816351175308, 0.049178797751665115, 0.017555607482790947, -0.01668686419725418, 0.05007995665073395, 0.00695795426145196, -0.0374453030526638, 0.04209280386567116, 0.037956222891807556, 0.0169405825436...
1,269
1,269
['Timothy Rozario', 'Troy Long', 'Mingli Chen', 'Weiguo Lu', 'Steve Jiang']
1801.00096v1
Data cleaning consumes about 80% of the time spent on data analysis for clinical research projects. This is a much bigger problem in the era of big data and machine learning in the field of medicine where large volumes of data are being generated. We report an initial effort towards automated patient data cleaning usin...
Towards automated patient data cleaning using deep learning: A feasibility study on the standardization of organ labeling
2,017
http://arxiv.org/pdf/1801.00096v1
Title Towards automated patient data cleaning using deep learning feasibility study standardization organ labeling Summary Data cleaning consumes 80 time spent data analysis clinical research project much bigger problem era big data machine learning field medicine large volume data generated report initial effort towar...
[0.03695407509803772, 0.0910504013299942, -0.014825277961790562, -0.022756880149245262, 0.013462974689900875, 0.021876422688364983, 0.04678391292691231, -0.0018635698361322284, -0.023861782625317574, 0.0483638234436512, -0.0020824761595577, -0.05180240795016289, 0.013292522169649601, 0.07247072458267212, -0.03056285902...
1,270
1,270
['Mohammad Javad Shafiee', 'Brendan Chwyl', 'Francis Li', 'Rongyan Chen', 'Michelle Karg', 'Christian Scharfenberger', 'Alexander Wong']
1801.05387v1
The computational complexity of leveraging deep neural networks for extracting deep feature representations is a significant barrier to its widespread adoption, particularly for use in embedded devices. One particularly promising strategy to addressing the complexity issue is the notion of evolutionary synthesis of dee...
StressedNets: Efficient Feature Representations via Stress-induced Evolutionary Synthesis of Deep Neural Networks
2,018
http://arxiv.org/pdf/1801.05387v1
Title StressedNets Efficient Feature Representations via Stressinduced Evolutionary Synthesis Deep Neural Networks Summary computational complexity leveraging deep neural network extracting deep feature representation significant barrier widespread adoption particularly use embedded device One particularly promising st...
[-0.0016200004611164331, -0.006138931959867477, -0.05016482248902321, 0.05396905168890953, 0.020888006314635277, 0.01014507282525301, 0.049316197633743286, 0.01982908509671688, -0.015409830026328564, 0.010732641443610191, 0.0002856180944945663, 0.016164004802703857, -0.01144458819180727, -0.013783845119178295, 0.017129...
1,271
1,271
['Wentao Zhu', 'Chaochun Liu', 'Wei Fan', 'Xiaohui Xie']
1801.09555v1
In this work, we present a fully automated lung computed tomography (CT) cancer diagnosis system, DeepLung. DeepLung consists of two components, nodule detection (identifying the locations of candidate nodules) and classification (classifying candidate nodules into benign or malignant). Considering the 3D nature of lun...
DeepLung: Deep 3D Dual Path Nets for Automated Pulmonary Nodule Detection and Classification
2,018
http://arxiv.org/pdf/1801.09555v1
Title DeepLung Deep 3D Dual Path Nets Automated Pulmonary Nodule Detection Classification Summary work present fully automated lung computed tomography CT cancer diagnosis system DeepLung DeepLung consists two component nodule detection identifying location candidate nodule classification classifying candidate nodule b...
[-0.022021811455488205, 0.00792370829731226, -0.03821546584367752, 0.014682930894196033, 0.00035765624488703907, 0.004730533808469772, 0.015938475728034973, 0.010848598554730415, -0.020664917305111885, -0.0013650734908878803, 0.006906967144459486, -0.012424222193658352, -0.011567771434783936, 0.03137766569852829, 0.022...
1,272
1,272
['Yang Hu', 'Guihua Wen', 'Huiqiang Liao', 'Changjun Wang', 'Dan Dai', 'Zhiwen Yu', 'Jun Zhang']
1802.02203v2
The tongue image is an important physical information of human, it is of great importance to the diagnosis and treatment in clinical medicine. Herbal prescriptions are simple, noninvasive and low side effects, and are widely applied in China. Researches on automatic construction technology of herbal prescription based ...
Automatic construction of Chinese herbal prescription from tongue image via CNNs and auxiliary latent therapy topics
2,018
http://arxiv.org/pdf/1802.02203v2
Title Automatic construction Chinese herbal prescription tongue image via CNNs auxiliary latent therapy topic Summary tongue image important physical information human great importance diagnosis treatment clinical medicine Herbal prescription simple noninvasive low side effect widely applied China Researches automatic ...
[0.06061870977282524, 0.032625459134578705, -0.01484221126884222, 0.001112616853788495, 0.010239568538963795, -0.001036180998198688, 0.03783392161130905, 0.037268463522195816, -0.020599544048309326, 0.011865121312439442, 0.015518378466367722, -0.01854405365884304, 0.042677294462919235, 0.04356662184000015, 0.0364194624...
1,273
1,273
['Yoojin Choi', 'Mostafa El-Khamy', 'Jungwon Lee']
1802.02271v1
Compression of deep neural networks (DNNs) for memory- and computation-efficient compact feature representations becomes a critical problem particularly for deployment of DNNs on resource-limited platforms. In this paper, we investigate lossy compression of DNNs by weight quantization and lossless source coding for mem...
Universal Deep Neural Network Compression
2,018
http://arxiv.org/pdf/1802.02271v1
Title Universal Deep Neural Network Compression Summary Compression deep neural network DNNs memory computationefficient compact feature representation becomes critical problem particularly deployment DNNs resourcelimited platform paper investigate lossy compression DNNs weight quantization lossless source coding memor...
[-0.05824408307671547, 0.049109093844890594, -0.005034924950450659, 0.03527325764298439, -0.008865266107022762, -0.005506171379238367, 0.039409030228853226, 0.0197879858314991, -0.08290362358093262, 0.03960421308875084, 0.010879616253077984, 0.011784383095800877, 0.016462275758385658, -0.005858412478119135, 0.011204366...
1,274
1,274
['Emanuele Sansone', 'Quoc-Tin Phan', 'Francesco G. B. De Natale']
1802.03505v1
Estimating the true density in high-dimensional feature spaces is a well-known problem in machine learning. This work shows that it is possible to formulate the optimization problem as a minimization and use the representational power of neural networks to learn very complex densities. A theoretical bound on the estima...
Plummer Autoencoders
2,018
http://arxiv.org/pdf/1802.03505v1
Title Plummer Autoencoders Summary Estimating true density highdimensional feature space wellknown problem machine learning work show possible formulate optimization problem minimization use representational power neural network learn complex density theoretical bound estimation error given dealing finite number sample...
[-0.01979243941605091, 0.07196786999702454, -0.03691814839839935, 0.013316767290234566, 0.0035808829125016928, -0.026110144332051277, 0.04829821735620499, 0.0013687587343156338, -0.05493529513478279, 0.038091544061899185, 0.020084815099835396, -0.008838819339871407, 0.0027800099924206734, 0.07562094926834106, 0.0615407...
1,275
1,275
['Jeff Zhang', 'Tianyu Gu', 'Kanad Basu', 'Siddharth Garg']
1802.04657v2
Due to their growing popularity and computational cost, deep neural networks (DNNs) are being targeted for hardware acceleration. A popular architecture for DNN acceleration, adopted by the Google Tensor Processing Unit (TPU), utilizes a systolic array based matrix multiplication unit at its core. This paper deals with...
Analyzing and Mitigating the Impact of Permanent Faults on a Systolic Array Based Neural Network Accelerator
2,018
http://arxiv.org/pdf/1802.04657v2
Title Analyzing Mitigating Impact Permanent Faults Systolic Array Based Neural Network Accelerator Summary Due growing popularity computational cost deep neural network DNNs targeted hardware acceleration popular architecture DNN acceleration adopted Google Tensor Processing Unit TPU utilizes systolic array based matri...
[0.018174592405557632, -0.010011626407504082, -0.04175625368952751, 0.04996875301003456, 0.0658169835805893, 0.020216185599565506, 0.07176627963781357, -0.00827297754585743, 0.010033844970166683, -0.02260962687432766, 0.02354070171713829, 0.03112345188856125, -0.0018009681953117251, 0.04219643399119377, 0.0370676666498...
1,276
1,276
['Xingyu Liu', 'Jeff Pool', 'Song Han', 'William J. Dally']
1802.06367v1
Convolutional Neural Networks (CNNs) are computationally intensive, which limits their application on mobile devices. Their energy is dominated by the number of multiplies needed to perform the convolutions. Winograd's minimal filtering algorithm (Lavin, 2015) and network pruning (Han et al., 2015) can reduce the opera...
Efficient Sparse-Winograd Convolutional Neural Networks
2,018
http://arxiv.org/pdf/1802.06367v1
Title Efficient SparseWinograd Convolutional Neural Networks Summary Convolutional Neural Networks CNNs computationally intensive limit application mobile device energy dominated number multiplies needed perform convolution Winograds minimal filtering algorithm Lavin 2015 network pruning Han et al 2015 reduce operation...
[-0.006850548554211855, 0.07805029302835464, 0.0282461978495121, 0.0728176161646843, 0.01069050282239914, -0.039613474160432816, 0.07376556843519211, 0.006438352167606354, -0.044668786227703094, 0.024298977106809616, -0.01772988773882389, -0.04599148407578468, -0.008649812079966068, 0.010818030685186386, 0.047760773450...
1,277
1,277
['Vinay Uday Prabhu', 'Nishant Desai', 'John Whaley']
1802.06927v1
In this paper, we would like to disseminate a serendipitous discovery involving Lyapunov exponents of a 1-D time series and their use in serving as a filtering defense tool against a specific kind of deep adversarial perturbation. To this end, we use the state-of-the-art CleverHans library to generate adversarial pertu...
On Lyapunov exponents and adversarial perturbation
2,018
http://arxiv.org/pdf/1802.06927v1
Title Lyapunov exponent adversarial perturbation Summary paper would like disseminate serendipitous discovery involving Lyapunov exponent 1D time series use serving filtering defense tool specific kind deep adversarial perturbation end use stateoftheart CleverHans library generate adversarial perturbation standard Conv...
[-0.022910982370376587, 0.018659869208931923, -0.02031935565173626, 0.015665875747799873, -0.019329702481627464, -0.016183625906705856, 0.00815452728420496, 0.018171528354287148, -0.031016558408737183, 0.008201711811125278, 0.04307468235492706, -0.08097264915704727, -0.04538678005337715, 0.06527532637119293, 0.06715340...
1,278
1,278
['Tal Ben-Nun', 'Torsten Hoefler']
1802.09941v1
Deep Neural Networks (DNNs) are becoming an important tool in modern computing applications. Accelerating their training is a major challenge and techniques range from distributed algorithms to low-level circuit design. In this survey, we describe the problem from a theoretical perspective, followed by approaches for i...
Demystifying Parallel and Distributed Deep Learning: An In-Depth Concurrency Analysis
2,018
http://arxiv.org/pdf/1802.09941v1
Title Demystifying Parallel Distributed Deep Learning InDepth Concurrency Analysis Summary Deep Neural Networks DNNs becoming important tool modern computing application Accelerating training major challenge technique range distributed algorithm lowlevel circuit design survey describe problem theoretical perspective fo...
[-0.014139722101390362, 0.04417242109775543, -0.016048695892095566, 0.07408139109611511, -0.03066244162619114, -0.01148309838026762, 0.08359061926603317, -0.05428282544016838, -0.01868341863155365, -0.012694623321294785, -0.045549653470516205, -0.024997875094413757, 0.016991129145026207, 0.022708186879754066, -0.042600...
1,279
1,279
['Kagan Tumer', 'Joydeep Ghosh']
cs/9905013v1
Integrating the outputs of multiple classifiers via combiners or meta-learners has led to substantial improvements in several difficult pattern recognition problems. In the typical setting investigated till now, each classifier is trained on data taken or resampled from a common data set, or (almost) randomly selected ...
Robust Combining of Disparate Classifiers through Order Statistics
1,999
http://arxiv.org/pdf/cs/9905013v1
Title Robust Combining Disparate Classifiers Order Statistics Summary Integrating output multiple classifier via combiners metalearners led substantial improvement several difficult pattern recognition problem typical setting investigated till classifier trained data taken resampled common data set almost randomly sele...
[0.0033243349753320217, 0.03700825572013855, -0.01981152780354023, -0.018016992136836052, -0.0012344983406364918, 0.033205606043338776, 0.06795665621757507, -0.00480225682258606, -0.01868198625743389, -0.045117784291505814, 0.021949773654341698, -0.014264748431742191, 0.09334788471460342, -0.012700811959803104, -0.0323...
1,280
1,280
['Frédéric Bastien', 'Yoshua Bengio', 'Arnaud Bergeron', 'Nicolas Boulanger-Lewandowski', 'Thomas Breuel', 'Youssouf Chherawala', 'Moustapha Cisse', 'Myriam Côté', 'Dumitru Erhan', 'Jeremy Eustache', 'Xavier Glorot', 'Xavier Muller', 'Sylvain Pannetier Lebeuf', 'Razvan Pascanu', 'Salah Rifai', 'Francois Savard', 'Guill...
1009.3589v1
Recent theoretical and empirical work in statistical machine learning has demonstrated the importance of learning algorithms for deep architectures, i.e., function classes obtained by composing multiple non-linear transformations. Self-taught learning (exploiting unlabeled examples or examples from other distributions)...
Deep Self-Taught Learning for Handwritten Character Recognition
2,010
http://arxiv.org/pdf/1009.3589v1
Title Deep SelfTaught Learning Handwritten Character Recognition Summary Recent theoretical empirical work statistical machine learning demonstrated importance learning algorithm deep architecture ie function class obtained composing multiple nonlinear transformation Selftaught learning exploiting unlabeled example exa...
[-0.01655230112373829, 0.02031043730676174, -0.016390619799494743, 0.06834777444601059, -0.00966436043381691, 0.006898415740579367, 0.10378711670637131, 0.015818534418940544, 0.0017315736040472984, 0.012750939466059208, -0.01459258608520031, 0.026064295321702957, 0.037962015718221664, 0.041975442320108414, 0.0291551947...
1,281
1,281
['Holger R. Roth', 'Le Lu', 'Ari Seff', 'Kevin M. Cherry', 'Joanne Hoffman', 'Shijun Wang', 'Jiamin Liu', 'Evrim Turkbey', 'Ronald M. Summers']
1406.2639v1
Automated Lymph Node (LN) detection is an important clinical diagnostic task but very challenging due to the low contrast of surrounding structures in Computed Tomography (CT) and to their varying sizes, poses, shapes and sparsely distributed locations. State-of-the-art studies show the performance range of 52.9% sensi...
A New 2.5D Representation for Lymph Node Detection using Random Sets of Deep Convolutional Neural Network Observations
2,014
http://arxiv.org/pdf/1406.2639v1
Title New 25D Representation Lymph Node Detection using Random Sets Deep Convolutional Neural Network Observations Summary Automated Lymph Node LN detection important clinical diagnostic task challenging due low contrast surrounding structure Computed Tomography CT varying size pose shape sparsely distributed location ...
[-0.008637003600597382, -0.017441168427467346, -0.02512010559439659, -0.002739807590842247, -0.04122290387749672, -0.016685549169778824, 0.03773622214794159, 0.010982673615217209, 0.02463010884821415, 0.041838034987449646, 0.020850876346230507, -0.016511712223291397, -0.030121080577373505, 0.008731517009437084, -0.0184...
1,282
1,282
['Sijin Li', 'Zhi-Qiang Liu', 'Antoni B. Chan']
1406.3474v1
We propose an heterogeneous multi-task learning framework for human pose estimation from monocular image with deep convolutional neural network. In particular, we simultaneously learn a pose-joint regressor and a sliding-window body-part detector in a deep network architecture. We show that including the body-part dete...
Heterogeneous Multi-task Learning for Human Pose Estimation with Deep Convolutional Neural Network
2,014
http://arxiv.org/pdf/1406.3474v1
Title Heterogeneous Multitask Learning Human Pose Estimation Deep Convolutional Neural Network Summary propose heterogeneous multitask learning framework human pose estimation monocular image deep convolutional neural network particular simultaneously learn posejoint regressor slidingwindow bodypart detector deep netwo...
[-0.02100444585084915, 0.02864932082593441, -0.036122579127550125, 0.029148230329155922, 0.04041983187198639, 0.020600976422429085, 0.015538412146270275, -0.014263544231653214, 0.02987072430551052, -0.035333264619112015, -0.05488698184490204, -0.06873107701539993, 0.03127753734588623, 0.007236036472022533, 0.0603905878...
1,283
1,283
['Alexey Dosovitskiy', 'Philipp Fischer', 'Jost Tobias Springenberg', 'Martin Riedmiller', 'Thomas Brox']
1406.6909v2
Deep convolutional networks have proven to be very successful in learning task specific features that allow for unprecedented performance on various computer vision tasks. Training of such networks follows mostly the supervised learning paradigm, where sufficiently many input-output pairs are required for training. Acq...
Discriminative Unsupervised Feature Learning with Exemplar Convolutional Neural Networks
2,014
http://arxiv.org/pdf/1406.6909v2
Title Discriminative Unsupervised Feature Learning Exemplar Convolutional Neural Networks Summary Deep convolutional network proven successful learning task specific feature allow unprecedented performance various computer vision task Training network follows mostly supervised learning paradigm sufficiently many inputo...
[-0.02972966991364956, 0.014566835947334766, 0.0014093855861574411, 0.08222246915102005, 0.024936173111200333, -0.006173251196742058, 0.06945673376321793, -0.002376290038228035, 0.010993259027600288, -0.008254345506429672, -0.010138756595551968, 0.057871926575899124, -0.018555255606770515, 0.044171079993247986, 0.02166...
1,284
1,284
['Geoffrey E. Hinton', 'Nitish Srivastava', 'Alex Krizhevsky', 'Ilya Sutskever', 'Ruslan R. Salakhutdinov']
1207.0580v1
When a large feedforward neural network is trained on a small training set, it typically performs poorly on held-out test data. This "overfitting" is greatly reduced by randomly omitting half of the feature detectors on each training case. This prevents complex co-adaptations in which a feature detector is only helpful...
Improving neural networks by preventing co-adaptation of feature detectors
2,012
http://arxiv.org/pdf/1207.0580v1
Title Improving neural network preventing coadaptation feature detector Summary large feedforward neural network trained small training set typically performs poorly heldout test data overfitting greatly reduced randomly omitting half feature detector training case prevents complex coadaptations feature detector helpfu...
[-0.022551847621798515, 0.04460768401622772, 0.028782280161976814, 0.0554269477725029, 0.005728363059461117, -0.0026410461869090796, 0.03652056306600571, -0.014345422387123108, -0.026584817096590996, -0.007680885028094053, -0.058417920023202896, 0.01764410361647606, 0.030286256223917007, 0.0943911224603653, 0.003108060...
1,285
1,285
['Lou Marvin Caraig']
1207.5774v3
The Sparse Distributed Memory proposed by Pentii Kanerva (SDM in short) was thought to be a model of human long term memory. The architecture of the SDM permits to store binary patterns and to retrieve them using partially matching patterns. However Kanerva's model is especially efficient only in handling random data. ...
A New Training Algorithm for Kanerva's Sparse Distributed Memory
2,012
http://arxiv.org/pdf/1207.5774v3
Title New Training Algorithm Kanervas Sparse Distributed Memory Summary Sparse Distributed Memory proposed Pentii Kanerva SDM short thought model human long term memory architecture SDM permit store binary pattern retrieve using partially matching pattern However Kanervas model especially efficient handling random data...
[0.028168728575110435, -0.004952354356646538, 0.00010467105312272906, 0.02855062298476696, -0.02128901705145836, -0.00922656524926424, 0.040516674518585205, 0.032775506377220154, -0.038389988243579865, -0.028713220730423927, 0.018169773742556572, -0.04427427798509598, -0.0016448211390525103, 0.03755858168005943, 0.0093...
1,286
1,286
['Xiangyu Zhang', 'Jianhua Zou', 'Kaiming He', 'Jian Sun']
1505.06798v2
This paper aims to accelerate the test-time computation of convolutional neural networks (CNNs), especially very deep CNNs that have substantially impacted the computer vision community. Unlike previous methods that are designed for approximating linear filters or linear responses, our method takes the nonlinear units ...
Accelerating Very Deep Convolutional Networks for Classification and Detection
2,015
http://arxiv.org/pdf/1505.06798v2
Title Accelerating Deep Convolutional Networks Classification Detection Summary paper aim accelerate testtime computation convolutional neural network CNNs especially deep CNNs substantially impacted computer vision community Unlike previous method designed approximating linear filter linear response method take nonlin...
[-0.01299706194549799, 0.049706313759088516, -0.013436630368232727, 0.13289044797420502, 0.0028946702368557453, -0.006387717556208372, 0.0325813889503479, 0.02892559953033924, 0.012740431353449821, -0.02126402221620083, 0.012752329930663109, 0.044696491211652756, -0.030155330896377563, 0.05201643705368042, 0.0209901910...
1,287
1,287
['Lei Wang', 'Baochang Zhang']
1505.06800v1
This paper proposes boosting-like deep learning (BDL) framework for pedestrian detection. Due to overtraining on the limited training samples, overfitting is a major problem of deep learning. We incorporate a boosting-like technique into deep learning to weigh the training samples, and thus prevent overtraining in the ...
Boosting-like Deep Learning For Pedestrian Detection
2,015
http://arxiv.org/pdf/1505.06800v1
Title Boostinglike Deep Learning Pedestrian Detection Summary paper proposes boostinglike deep learning BDL framework pedestrian detection Due overtraining limited training sample overfitting major problem deep learning incorporate boostinglike technique deep learning weigh training sample thus prevent overtraining ite...
[-0.027217471972107887, -0.026003912091255188, 0.002112295478582382, 0.0628073662519455, 0.00659519387409091, 0.025538448244333267, -0.015554523095488548, 0.0038489762227982283, -0.006982299033552408, 0.013168951496481895, 0.014449158683419228, 0.04842018336057663, -0.03582747280597687, -0.02246512472629547, -0.0138990...
1,288
1,288
['Jyothi Korra']
1604.02376v2
In Computer Vision, problem of identifying or classifying the objects present in an image is called Object Categorization. It is a challenging problem, especially when the images have clutter background, occlusions or different lighting conditions. Many vision features have been proposed which aid object categorization...
Finding Optimal Combination of Kernels using Genetic Programming
2,016
http://arxiv.org/pdf/1604.02376v2
Title Finding Optimal Combination Kernels using Genetic Programming Summary Computer Vision problem identifying classifying object present image called Object Categorization challenging problem especially image clutter background occlusion different lighting condition Many vision feature proposed aid object categorizat...
[-0.009314346127212048, 0.019023843109607697, -0.018202193081378937, 0.06400197744369507, -0.00837255921214819, -0.0026058570947498083, 0.023166514933109283, 0.05722185969352722, 0.03262889012694359, -0.026188919320702553, 0.03877796232700348, 0.05200709030032158, 0.01111442968249321, 0.011155807413160801, 0.0175313204...
1,289
1,289
['Biswajit Paria', 'Vikas Reddy', 'Anirban Santara', 'Pabitra Mitra']
1604.02646v3
The success of deep neural networks is mostly due their ability to learn meaningful features from the data. Features learned in the hidden layers of deep neural networks trained in computer vision tasks have been shown to be similar to mid-level vision features. We leverage this fact in this work and propose the visual...
Visualization Regularizers for Neural Network based Image Recognition
2,016
http://arxiv.org/pdf/1604.02646v3
Title Visualization Regularizers Neural Network based Image Recognition Summary success deep neural network mostly due ability learn meaningful feature data Features learned hidden layer deep neural network trained computer vision task shown similar midlevel vision feature leverage fact work propose visualization regul...
[-0.012037497945129871, 0.04560335725545883, -0.02106517180800438, 0.04609900712966919, 0.023573758080601692, -0.015860341489315033, 0.015994740650057793, 0.02769344486296177, -0.00286907353438437, 0.05008585751056671, 0.02630486525595188, 0.09598731249570847, 0.008211609907448292, -0.004103078041225672, 0.007778150960...
1,290
1,290
['Ashley Prater']
1604.03073v2
Reservoir computing is a recently introduced machine learning paradigm that has been shown to be well-suited for the processing of spatiotemporal data. Rather than training the network node connections and weights via backpropagation in traditional recurrent neural networks, reservoirs instead have fixed connections an...
Reservoir computing for spatiotemporal signal classification without trained output weights
2,016
http://arxiv.org/pdf/1604.03073v2
Title Reservoir computing spatiotemporal signal classification without trained output weight Summary Reservoir computing recently introduced machine learning paradigm shown wellsuited processing spatiotemporal data Rather training network node connection weight via backpropagation traditional recurrent neural network r...
[-0.05331907421350479, 0.010840771719813347, -0.03555891290307045, -0.01076823566108942, 0.0038990764878690243, 0.017771175131201744, 0.04048176109790802, -0.034472402185201645, 0.010651816613972187, 0.0266720037907362, 0.019613485783338547, -0.0010804618941619992, 0.040080476552248, 0.022253913804888725, 0.04988390579...
1,291
1,291
['Jason Kuen', 'Kian Ming Lim', 'Chin Poo Lee']
1604.04144v1
Visual representation is crucial for a visual tracking method's performances. Conventionally, visual representations adopted in visual tracking rely on hand-crafted computer vision descriptors. These descriptors were developed generically without considering tracking-specific information. In this paper, we propose to l...
Self-taught learning of a deep invariant representation for visual tracking via temporal slowness principle
2,016
http://arxiv.org/pdf/1604.04144v1
Title Selftaught learning deep invariant representation visual tracking via temporal slowness principle Summary Visual representation crucial visual tracking method performance Conventionally visual representation adopted visual tracking rely handcrafted computer vision descriptor descriptor developed generically witho...
[-0.049022823572158813, 0.06272765249013901, -0.0016849592793732882, 0.057420335710048676, 0.02984175831079483, -0.01895485259592533, 0.03547830507159233, 0.028954308480024338, -0.031147433444857597, -0.01848485693335533, 0.05986462160944939, -0.020572742447257042, -0.024902770295739174, 0.06879204511642456, 0.01768510...
1,292
1,292
['Jiang Wang', 'Yi Yang', 'Junhua Mao', 'Zhiheng Huang', 'Chang Huang', 'Wei Xu']
1604.04573v1
While deep convolutional neural networks (CNNs) have shown a great success in single-label image classification, it is important to note that real world images generally contain multiple labels, which could correspond to different objects, scenes, actions and attributes in an image. Traditional approaches to multi-labe...
CNN-RNN: A Unified Framework for Multi-label Image Classification
2,016
http://arxiv.org/pdf/1604.04573v1
Title CNNRNN Unified Framework Multilabel Image Classification Summary deep convolutional neural network CNNs shown great success singlelabel image classification important note real world image generally contain multiple label could correspond different object scene action attribute image Traditional approach multilab...
[0.04776640981435776, 0.0533481165766716, 0.014942086301743984, 0.0747450441122055, 0.004805758595466614, 0.00595345301553607, 0.04863865673542023, 0.006765434984117746, -0.016401587054133415, -0.03206595778465271, -0.03871645778417587, 0.048140451312065125, -0.03995596989989281, 0.004954193253070116, 0.006143094506114...
1,293
1,293
['Markus Thom', 'Matthias Rapp', 'Günther Palm']
1604.04767v1
Learning dictionaries suitable for sparse coding instead of using engineered bases has proven effective in a variety of image processing tasks. This paper studies the optimization of dictionaries on image data where the representation is enforced to be explicitly sparse with respect to a smooth, normalized sparseness m...
Efficient Dictionary Learning with Sparseness-Enforcing Projections
2,016
http://arxiv.org/pdf/1604.04767v1
Title Efficient Dictionary Learning SparsenessEnforcing Projections Summary Learning dictionary suitable sparse coding instead using engineered base proven effective variety image processing task paper study optimization dictionary image data representation enforced explicitly sparse respect smooth normalized sparsenes...
[-0.04228078946471214, 0.02037239819765091, 0.004286458715796471, 0.04913000017404556, 0.008266383782029152, -0.014589275233447552, 0.0052305907011032104, 0.028183577582240105, -0.02306833676993847, 0.01894480548799038, 0.06478360295295715, 0.019637849181890488, 0.005339419934898615, 0.0883985087275505, -0.010233661159...
1,294
1,294
['Mariusz Bojarski', 'Davide Del Testa', 'Daniel Dworakowski', 'Bernhard Firner', 'Beat Flepp', 'Prasoon Goyal', 'Lawrence D. Jackel', 'Mathew Monfort', 'Urs Muller', 'Jiakai Zhang', 'Xin Zhang', 'Jake Zhao', 'Karol Zieba']
1604.07316v1
We trained a convolutional neural network (CNN) to map raw pixels from a single front-facing camera directly to steering commands. This end-to-end approach proved surprisingly powerful. With minimum training data from humans the system learns to drive in traffic on local roads with or without lane markings and on highw...
End to End Learning for Self-Driving Cars
2,016
http://arxiv.org/pdf/1604.07316v1
Title End End Learning SelfDriving Cars Summary trained convolutional neural network CNN map raw pixel single frontfacing camera directly steering command endtoend approach proved surprisingly powerful minimum training data human system learns drive traffic local road without lane marking highway also operates area unc...
[-0.0052204192616045475, -0.0005378538044169545, 0.0010341499000787735, 0.09203584492206573, 0.004228456411510706, 0.00496628088876605, 0.03637706860899925, 0.00038848735857754946, -0.031071534380316734, -0.024753548204898834, 0.01998056285083294, 0.08151568472385406, -0.03609565272927284, 0.06427893042564392, 0.032730...
1,295
1,295
['Wen Wang', 'Zhen Cui', 'Hong Chang', 'Shiguang Shan', 'Xilin Chen']
1402.2031v1
The comparison of heterogeneous samples extensively exists in many applications, especially in the task of image classification. In this paper, we propose a simple but effective coupled neural network, called Deeply Coupled Autoencoder Networks (DCAN), which seeks to build two deep neural networks, coupled with each ot...
Deeply Coupled Auto-encoder Networks for Cross-view Classification
2,014
http://arxiv.org/pdf/1402.2031v1
Title Deeply Coupled Autoencoder Networks Crossview Classification Summary comparison heterogeneous sample extensively exists many application especially task image classification paper propose simple effective coupled neural network called Deeply Coupled Autoencoder Networks DCAN seek build two deep neural network cou...
[-0.03552306443452835, -0.0017497153021395206, -0.02156813256442547, 0.06486256420612335, 0.020600484684109688, 0.014657450839877129, 0.050341397523880005, -0.01938673108816147, 0.0218329019844532, -0.021376438438892365, -0.08015791326761246, 0.01679808273911476, 0.02498188242316246, 0.020889148116111755, 0.00358513859...
1,296
1,296
['S. K. Katiyar', 'P. V. Arun']
1405.6137v1
The improvements in spectral and spatial resolution of the satellite images have facilitated the automatic extraction and identification of the features from satellite images and aerial photographs. An automatic object extraction method is presented for extracting and identifying the various objects from satellite imag...
An enhanced neural network based approach towards object extraction
2,014
http://arxiv.org/pdf/1405.6137v1
Title enhanced neural network based approach towards object extraction Summary improvement spectral spatial resolution satellite image facilitated automatic extraction identification feature satellite image aerial photograph automatic object extraction method presented extracting identifying various object satellite im...
[0.026542624458670616, 0.022746378555893898, 0.01991499401628971, 0.07399605959653854, -0.023753676563501358, -0.009220050647854805, -0.015764430165290833, -0.035527877509593964, -0.014567505568265915, 0.0048294151201844215, 0.03618008643388748, 0.06346581876277924, 0.030375467613339424, 0.03418363630771637, 0.03978973...
1,297
1,297
['Mohammad Ali Keyvanrad', 'Mohammad Mehdi Homayounpour']
1408.3264v7
Nowadays, this is very popular to use the deep architectures in machine learning. Deep Belief Networks (DBNs) are deep architectures that use stack of Restricted Boltzmann Machines (RBM) to create a powerful generative model using training data. DBNs have many ability like feature extraction and classification that are...
A brief survey on deep belief networks and introducing a new object oriented toolbox (DeeBNet)
2,014
http://arxiv.org/pdf/1408.3264v7
Title brief survey deep belief network introducing new object oriented toolbox DeeBNet Summary Nowadays popular use deep architecture machine learning Deep Belief Networks DBNs deep architecture use stack Restricted Boltzmann Machines RBM create powerful generative model using training data DBNs many ability like featu...
[-0.03950696066021919, 0.04416142776608467, -0.016896501183509827, 0.054647836834192276, -0.011401049792766571, -0.0042555853724479675, 0.07369701564311981, -0.024179836735129356, 0.016666268929839134, 0.01472508069127798, -0.025297628715634346, 0.017359817400574684, 0.009478696621954441, 0.06041313335299492, 0.0070667...
1,298
1,298
['Sébastien Ouellet']
1408.3750v1
The goal of the present study is to explore the application of deep convolutional network features to emotion recognition. Results indicate that they perform similarly to other published models at a best recognition rate of 94.4%, and do so with a single still image rather than a video stream. An implementation of an a...
Real-time emotion recognition for gaming using deep convolutional network features
2,014
http://arxiv.org/pdf/1408.3750v1
Title Realtime emotion recognition gaming using deep convolutional network feature Summary goal present study explore application deep convolutional network feature emotion recognition Results indicate perform similarly published model best recognition rate 944 single still image rather video stream implementation affe...
[0.0020066150464117527, 0.017236609011888504, -0.02321830578148365, 0.07818114012479782, 0.04029898717999458, 0.025645524263381958, 0.018791958689689636, -0.005568698048591614, -0.007751898840069771, -0.007275332231074572, -0.016296900808811188, -0.013765068724751472, -0.02998919039964676, 0.07420498877763748, 0.030264...
1,299
1,299
['Yangqing Jia', 'Evan Shelhamer', 'Jeff Donahue', 'Sergey Karayev', 'Jonathan Long', 'Ross Girshick', 'Sergio Guadarrama', 'Trevor Darrell']
1408.5093v1
Caffe provides multimedia scientists and practitioners with a clean and modifiable framework for state-of-the-art deep learning algorithms and a collection of reference models. The framework is a BSD-licensed C++ library with Python and MATLAB bindings for training and deploying general-purpose convolutional neural net...
Caffe: Convolutional Architecture for Fast Feature Embedding
2,014
http://arxiv.org/pdf/1408.5093v1
Title Caffe Convolutional Architecture Fast Feature Embedding Summary Caffe provides multimedia scientist practitioner clean modifiable framework stateoftheart deep learning algorithm collection reference model framework BSDlicensed C library Python MATLAB binding training deploying generalpurpose convolutional neural ...
[0.0014900341629981995, 0.03233158588409424, 0.007082968018949032, 0.07052892446517944, 0.013149814680218697, -0.03736178204417229, 0.04942554980516434, -0.002487603574991226, -0.07673437148332596, -0.01635831408202648, -0.009302343241870403, 0.0036060784477740526, -0.013153237290680408, 0.08294370025396347, 0.02912564...