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
1,100
1,100
['Alice Schoenauer-Sebag', 'Marc Schoenauer', 'Michèle Sebag']
1709.01427v1
When applied to training deep neural networks, stochastic gradient descent (SGD) often incurs steady progression phases, interrupted by catastrophic episodes in which loss and gradient norm explode. A possible mitigation of such events is to slow down the learning process. This paper presents a novel approach to contro...
Stochastic Gradient Descent: Going As Fast As Possible But Not Faster
2,017
http://arxiv.org/pdf/1709.01427v1
Title Stochastic Gradient Descent Going Fast Possible Faster Summary applied training deep neural network stochastic gradient descent SGD often incurs steady progression phase interrupted catastrophic episode loss gradient norm explode possible mitigation event slow learning process paper present novel approach control...
[-0.023533958941698074, 0.08653374761343002, 0.0063201566226780415, 0.009362902492284775, 0.03733007237315178, -0.012138701975345612, 0.008516749367117882, -0.003429237985983491, -0.029963333159685135, 0.04887547343969345, -0.0038244668394327164, -0.003739206586033106, -0.008749003522098064, 0.06556230038404465, -0.002...
1,101
1,101
['Heikki Arponen', 'Matti Herranen', 'Harri Valpola']
1709.02797v1
We prove an exact relationship between the optimal denoising function and the data distribution in the case of additive Gaussian noise, showing that denoising implicitly models the structure of data allowing it to be exploited in the unsupervised learning of representations. This result generalizes a known relationship...
On the exact relationship between the denoising function and the data distribution
2,017
http://arxiv.org/pdf/1709.02797v1
Title exact relationship denoising function data distribution Summary prove exact relationship optimal denoising function data distribution case additive Gaussian noise showing denoising implicitly model structure data allowing exploited unsupervised learning representation result generalizes known relationship 2 valid...
[-0.046098098158836365, 0.04847697541117668, 0.015318041667342186, 0.033663034439086914, 0.0303803663700819, -0.012547886930406094, 0.01051382441073656, 0.01687128096818924, -0.05102303624153137, 0.038276027888059616, 0.006631487514823675, 0.048245903104543686, 0.02472578175365925, 0.06531286239624023, 0.02057365700602...
1,102
1,102
['Lars Eidnes', 'Arild Nøkland']
1709.04054v3
We propose a simple extension to the ReLU-family of activation functions that allows them to shift the mean activation across a layer towards zero. Combined with proper weight initialization, this alleviates the need for normalization layers. We explore the training of deep vanilla recurrent neural networks (RNNs) with...
Shifting Mean Activation Towards Zero with Bipolar Activation Functions
2,017
http://arxiv.org/pdf/1709.04054v3
Title Shifting Mean Activation Towards Zero Bipolar Activation Functions Summary propose simple extension ReLUfamily activation function allows shift mean activation across layer towards zero Combined proper weight initialization alleviates need normalization layer explore training deep vanilla recurrent neural network...
[-0.014231669716536999, -0.0019325724570080638, -0.004295293241739273, 0.039573490619659424, 0.0019657553639262915, 0.00898071750998497, 0.053671982139348984, -0.018745196983218193, -0.005313708912581205, -0.006675517186522484, -0.009859907440841198, -0.028438277542591095, -0.007702076341956854, 0.05146991088986397, -0...
1,103
1,103
['Bingzhen Wei', 'Xu Sun', 'Xuancheng Ren', 'Jingjing Xu']
1709.05804v1
As traditional neural network consumes a significant amount of computing resources during back propagation, \citet{Sun2017mePropSB} propose a simple yet effective technique to alleviate this problem. In this technique, only a small subset of the full gradients are computed to update the model parameters. In this paper ...
Minimal Effort Back Propagation for Convolutional Neural Networks
2,017
http://arxiv.org/pdf/1709.05804v1
Title Minimal Effort Back Propagation Convolutional Neural Networks Summary traditional neural network consumes significant amount computing resource back propagation citetSun2017mePropSB propose simple yet effective technique alleviate problem technique small subset full gradient computed update model parameter paper ...
[0.020810863003134727, 0.05241648480296135, -0.016696369275450706, 0.05617378279566765, -0.005921851843595505, -0.03958631306886673, 0.02968607284128666, 0.02149210497736931, -0.023223748430609703, 0.0007350181112997234, -0.0015729524893686175, 0.02498076856136322, 0.025910042226314545, 0.014754088595509529, 0.04045772...
1,104
1,104
['Emmanuel Dufourq', 'Bruce A. Bassett']
1709.09161v1
Deep neural networks continue to show improved performance with increasing depth, an encouraging trend that implies an explosion in the possible permutations of network architectures and hyperparameters for which there is little intuitive guidance. To address this increasing complexity, we propose Evolutionary DEep Net...
EDEN: Evolutionary Deep Networks for Efficient Machine Learning
2,017
http://arxiv.org/pdf/1709.09161v1
Title EDEN Evolutionary Deep Networks Efficient Machine Learning Summary Deep neural network continue show improved performance increasing depth encouraging trend implies explosion possible permutation network architecture hyperparameters little intuitive guidance address increasing complexity propose Evolutionary DEep...
[0.002227277494966984, 0.08010131120681763, -0.020145077258348465, 0.052604977041482925, -0.014026503078639507, 0.0058843581937253475, -0.017270216718316078, 0.0015761401737108827, -0.04190029203891754, -0.006488995160907507, -0.04458775743842125, 0.047163452953100204, -0.04568575322628021, 0.059074562042951584, 0.0230...
1,105
1,105
['Aaron Tuor', 'Samuel Kaplan', 'Brian Hutchinson', 'Nicole Nichols', 'Sean Robinson']
1710.00811v2
Analysis of an organization's computer network activity is a key component of early detection and mitigation of insider threat, a growing concern for many organizations. Raw system logs are a prototypical example of streaming data that can quickly scale beyond the cognitive power of a human analyst. As a prospective fi...
Deep Learning for Unsupervised Insider Threat Detection in Structured Cybersecurity Data Streams
2,017
http://arxiv.org/pdf/1710.00811v2
Title Deep Learning Unsupervised Insider Threat Detection Structured Cybersecurity Data Streams Summary Analysis organization computer network activity key component early detection mitigation insider threat growing concern many organization Raw system log prototypical example streaming data quickly scale beyond cognit...
[0.04815596342086792, 0.05398675799369812, -0.010955262929201126, 0.01917029544711113, -0.03882159665226936, -0.0296748336404562, 0.0659121423959732, -0.0192261915653944, 0.015061609447002411, -0.01018973346799612, 0.003621893236413598, -0.007878278382122517, -0.01585996337234974, 0.12241838872432709, 0.023167828097939...
1,106
1,106
['Emanuele Sansone', 'Francesco G. B. De Natale']
1710.01013v1
Training feedforward neural networks with standard logistic activations is considered difficult because of the intrinsic properties of these sigmoidal functions. This work aims at showing that these networks can be trained to achieve generalization performance comparable to those based on hyperbolic tangent activations...
Training Feedforward Neural Networks with Standard Logistic Activations is Feasible
2,017
http://arxiv.org/pdf/1710.01013v1
Title Training Feedforward Neural Networks Standard Logistic Activations Feasible Summary Training feedforward neural network standard logistic activation considered difficult intrinsic property sigmoidal function work aim showing network trained achieve generalization performance comparable based hyperbolic tangent ac...
[-0.04957413673400879, -0.03195488825440407, 0.012159457430243492, -0.018573923036456108, 0.023631619289517403, -0.0013827010989189148, 0.059380196034908295, -0.014853348024189472, 0.004821982700377703, 0.005603186786174774, -0.0275629423558712, -0.009456124156713486, 0.006556310690939426, 0.09351392835378647, 0.037621...
1,107
1,107
['Brian DePasquale', 'Christopher J. Cueva', 'Kanaka Rajan', 'G. Sean Escola', 'L. F. Abbott']
1710.03070v1
Trained recurrent networks are powerful tools for modeling dynamic neural computations. We present a target-based method for modifying the full connectivity matrix of a recurrent network to train it to perform tasks involving temporally complex input/output transformations. The method introduces a second network during...
full-FORCE: A Target-Based Method for Training Recurrent Networks
2,017
http://arxiv.org/pdf/1710.03070v1
Title fullFORCE TargetBased Method Training Recurrent Networks Summary Trained recurrent network powerful tool modeling dynamic neural computation present targetbased method modifying full connectivity matrix recurrent network train perform task involving temporally complex inputoutput transformation method introduces ...
[0.01072996947914362, -0.007863006554543972, -0.021596841514110565, 0.00283726304769516, 0.005893248599022627, -0.01272099930793047, -0.010007153265178204, -0.04204770550131798, -0.06445682048797607, -0.017407167702913284, -0.036843154579401016, 0.0030421693809330463, 0.01374089065939188, 0.06351528316736221, 0.0261058...
1,108
1,108
['Alessandro Bay', 'Biswa Sengupta']
1710.04211v2
A widely studied non-deterministic polynomial time (NP) hard problem lies in finding a route between the two nodes of a graph. Often meta-heuristics algorithms such as $A^{*}$ are employed on graphs with a large number of nodes. Here, we propose a deep recurrent neural network architecture based on the Sequence-2-Seque...
StackSeq2Seq: Dual Encoder Seq2Seq Recurrent Networks
2,017
http://arxiv.org/pdf/1710.04211v2
Title StackSeq2Seq Dual Encoder Seq2Seq Recurrent Networks Summary widely studied nondeterministic polynomial time NP hard problem lie finding route two node graph Often metaheuristics algorithm employed graph large number node propose deep recurrent neural network architecture based Sequence2Sequence Seq2Seq model wid...
[0.009995982982218266, 0.01957784779369831, -0.009134890511631966, 0.052747149020433426, -0.02433199994266033, -0.003931952640414238, -0.0049291616305708885, -0.006745851133018732, -0.041256628930568695, -0.013062567450106144, 0.018196649849414825, -0.012484371662139893, 0.008527291007339954, 0.0737173929810524, 0.0181...
1,109
1,109
['Or Sharir', 'Amnon Shashua']
1710.04404v3
We present a novel tractable generative model that extends Sum-Product Networks (SPNs) and significantly boosts their power. We call it Sum-Product-Quotient Networks (SPQNs), whose core concept is to incorporate conditional distributions into the model by direct computation using quotient nodes, e.g. $P(A|B) = \frac{P(...
Sum-Product-Quotient Networks
2,017
http://arxiv.org/pdf/1710.04404v3
Title SumProductQuotient Networks Summary present novel tractable generative model extends SumProduct Networks SPNs significantly boost power call SumProductQuotient Networks SPQNs whose core concept incorporate conditional distribution model direct computation using quotient node eg PAB fracPABPB provide sufficient co...
[-0.016943322494626045, 0.0831017717719078, -0.016239991411566734, -0.005331403575837612, -0.04835599288344383, -0.055108051747083664, -0.005476513411849737, -0.04812392219901085, -0.06525160372257233, 0.03061613254249096, 0.000116072638775222, 0.01809966005384922, 0.0034743710421025753, 0.061793822795152664, 0.0432007...
1,110
1,110
['Filippo Maria Bianchi', 'Karl Øyvind Mikalsen', 'Robert Jenssen']
1710.07547v1
Clinical measurements collected over time are naturally represented as multivariate time series (MTS), which often contain missing data. An autoencoder can learn low dimensional vectorial representations of MTS that preserve important data characteristics, but cannot deal explicitly with missing data. In this work, we ...
Learning compressed representations of blood samples time series with missing data
2,017
http://arxiv.org/pdf/1710.07547v1
Title Learning compressed representation blood sample time series missing data Summary Clinical measurement collected time naturally represented multivariate time series MTS often contain missing data autoencoder learn low dimensional vectorial representation MTS preserve important data characteristic cannot deal expli...
[-0.03724183514714241, 0.03668700531125069, -0.0291629359126091, -0.025012625381350517, -0.012861729599535465, 0.03459732607007027, 0.04728443920612335, 0.0397295281291008, -0.03931954503059387, 0.044939033687114716, 0.07390826940536499, -0.029955096542835236, 0.009525634348392487, 0.08961577713489532, 0.01234323158860...
1,111
1,111
['Takashi Shinozaki']
1710.09574v1
In this study, we propose a novel deep neural network and its supervised learning method that uses a feedforward supervisory signal. The method is inspired by the human visual system and performs human-like association-based learning without any backward error propagation. The feedforward supervisory signal that produc...
Biologically Inspired Feedforward Supervised Learning for Deep Self-Organizing Map Networks
2,017
http://arxiv.org/pdf/1710.09574v1
Title Biologically Inspired Feedforward Supervised Learning Deep SelfOrganizing Map Networks Summary study propose novel deep neural network supervised learning method us feedforward supervisory signal method inspired human visual system performs humanlike associationbased learning without backward error propagation fe...
[-0.003377028275281191, 0.0023127535823732615, -0.008769068866968155, -0.007668983656913042, 0.008423716761171818, 0.018064914271235466, 0.04059820622205734, -0.02249019779264927, 0.06295769661664963, 0.030142875388264656, -0.06392107903957367, 0.07173925638198853, 0.0030380499083548784, 0.031097568571567535, 0.0141812...
1,112
1,112
['Zichao Long', 'Yiping Lu', 'Xianzhong Ma', 'Bin Dong']
1710.09668v2
In this paper, we present an initial attempt to learn evolution PDEs from data. Inspired by the latest development of neural network designs in deep learning, we propose a new feed-forward deep network, called PDE-Net, to fulfill two objectives at the same time: to accurately predict dynamics of complex systems and to ...
PDE-Net: Learning PDEs from Data
2,017
http://arxiv.org/pdf/1710.09668v2
Title PDENet Learning PDEs Data Summary paper present initial attempt learn evolution PDEs data Inspired latest development neural network design deep learning propose new feedforward deep network called PDENet fulfill two objective time accurately predict dynamic complex system uncover underlying hidden PDE model basi...
[-0.027787648141384125, 0.05331917852163315, -0.02440316043794155, 0.02564864233136177, 0.06089714914560318, -0.014410550706088543, 0.018484773114323616, -0.004308223724365234, 0.007820633240044117, 0.029921628534793854, 0.00661085732281208, 0.015485466457903385, -0.04757052659988403, 0.026699058711528778, 0.0388537906...
1,113
1,113
['Carlo Baldassi', 'Federica Gerace', 'Hilbert J. Kappen', 'Carlo Lucibello', 'Luca Saglietti', 'Enzo Tartaglione', 'Riccardo Zecchina']
1710.09825v2
Stochasticity and limited precision of synaptic weights in neural network models are key aspects of both biological and hardware modeling of learning processes. Here we show that a neural network model with stochastic binary weights naturally gives prominence to exponentially rare dense regions of solutions with a numb...
On the role of synaptic stochasticity in training low-precision neural networks
2,017
http://arxiv.org/pdf/1710.09825v2
Title role synaptic stochasticity training lowprecision neural network Summary Stochasticity limited precision synaptic weight neural network model key aspect biological hardware modeling learning process show neural network model stochastic binary weight naturally give prominence exponentially rare dense region soluti...
[-0.00040755729423835874, -0.020681282505393028, -0.026885297149419785, -0.0037511002738028765, 0.016790498048067093, -0.04613510146737099, 0.02433118037879467, -0.040235649794340134, -0.0019956338219344616, 0.02664707787334919, -0.00777424406260252, -0.034107375890016556, 0.02624696120619774, 0.03465999290347099, 0.03...
1,114
1,114
['Yucen Luo', 'Jun Zhu', 'Mengxi Li', 'Yong Ren', 'Bo Zhang']
1711.00258v1
The paper proposes an inductive semi-supervised learning method, called Smooth Neighbors on Teacher Graphs (SNTG). At each iteration during training, a graph is dynamically constructed based on predictions of the teacher model, i.e., the implicit self-ensemble of models. Then the graph serves as a similarity measure wi...
Smooth Neighbors on Teacher Graphs for Semi-supervised Learning
2,017
http://arxiv.org/pdf/1711.00258v1
Title Smooth Neighbors Teacher Graphs Semisupervised Learning Summary paper proposes inductive semisupervised learning method called Smooth Neighbors Teacher Graphs SNTG iteration training graph dynamically constructed based prediction teacher model ie implicit selfensemble model graph serf similarity measure respect r...
[0.0017343710642307997, -0.01670762151479721, 0.015250284224748611, 0.03593845292925835, -0.012864897958934307, -0.03826121613383293, 0.025093887001276016, -0.03726274147629738, 0.043266937136650085, 0.006753548048436642, -0.011827273294329643, 0.031377196311950684, -0.014431734569370747, -0.0007278169505298138, 0.0357...
1,115
1,115
['K. Youssef', 'Louis-S. Bouchard', 'K. Z. Haigh', 'H. Krovi', 'J. Silovsky', 'C. P. Vander Valk']
1711.01559v2
With the development and widespread use of wireless devices in recent years (mobile phones, Internet of Things, Wi-Fi), the electromagnetic spectrum has become extremely crowded. In order to counter security threats posed by rogue or unknown transmitters, it is important to identify RF transmitters not by the data cont...
Machine Learning Approach to RF Transmitter Identification
2,017
http://arxiv.org/pdf/1711.01559v2
Title Machine Learning Approach RF Transmitter Identification Summary development widespread use wireless device recent year mobile phone Internet Things WiFi electromagnetic spectrum become extremely crowded order counter security threat posed rogue unknown transmitter important identify RF transmitter data content tr...
[0.00015784108836669475, -0.011474218219518661, -0.007591120898723602, -0.019726943224668503, 0.002488064346835017, 0.013304205611348152, 0.0315057598054409, -0.03926815092563629, -0.06861262768507004, 0.00024393470084760338, 0.028411652892827988, 0.06351444125175476, -0.014467889443039894, -0.014063957147300243, 0.011...
1,116
1,116
['Anirudh Goyal', 'Nan Rosemary Ke', 'Surya Ganguli', 'Yoshua Bengio']
1711.02282v1
We propose a novel method to directly learn a stochastic transition operator whose repeated application provides generated samples. Traditional undirected graphical models approach this problem indirectly by learning a Markov chain model whose stationary distribution obeys detailed balance with respect to a parameteriz...
Variational Walkback: Learning a Transition Operator as a Stochastic Recurrent Net
2,017
http://arxiv.org/pdf/1711.02282v1
Title Variational Walkback Learning Transition Operator Stochastic Recurrent Net Summary propose novel method directly learn stochastic transition operator whose repeated application provides generated sample Traditional undirected graphical model approach problem indirectly learning Markov chain model whose stationary...
[-0.002988402033224702, 0.06210906803607941, -0.010815510526299477, -0.04608750715851784, -0.0063091604970395565, -0.03807089105248451, -0.0257234126329422, -0.02196257747709751, -0.044346198439598083, 0.01113633532077074, 0.01421776507049799, 0.012098885141313076, 0.00020008556020911783, 0.07421796768903732, 0.0446127...
1,117
1,117
['Anirbit Mukherjee', 'Amitabh Basu']
1711.03073v2
Motivated by the resurgence of neural networks in being able to solve complex learning tasks we undertake a study of high depth networks using ReLU gates which implement the function $x \mapsto \max\{0,x\}$. We try to understand the role of depth in such neural networks by showing size lowerbounds against such network ...
Lower bounds over Boolean inputs for deep neural networks with ReLU gates
2,017
http://arxiv.org/pdf/1711.03073v2
Title Lower bound Boolean input deep neural network ReLU gate Summary Motivated resurgence neural network able solve complex learning task undertake study high depth network using ReLU gate implement function x mapsto max0x try understand role depth neural network showing size lowerbounds network architecture parameter...
[-0.0410325787961483, -0.009115271270275116, -0.025697646662592888, 0.023753883317112923, -0.03578046336770058, -0.0540885254740715, 0.02829502336680889, -0.02082742005586624, 0.004541699308902025, -0.00540276151150465, 0.0170154832303524, 0.03335939720273018, -0.024952055886387825, 0.11586958169937134, 0.0259785484522...
1,118
1,118
['Jesse Engel', 'Matthew Hoffman', 'Adam Roberts']
1711.05772v2
Deep generative neural networks have proven effective at both conditional and unconditional modeling of complex data distributions. Conditional generation enables interactive control, but creating new controls often requires expensive retraining. In this paper, we develop a method to condition generation without retrai...
Latent Constraints: Learning to Generate Conditionally from Unconditional Generative Models
2,017
http://arxiv.org/pdf/1711.05772v2
Title Latent Constraints Learning Generate Conditionally Unconditional Generative Models Summary Deep generative neural network proven effective conditional unconditional modeling complex data distribution Conditional generation enables interactive control creating new control often requires expensive retraining paper ...
[-0.03133583813905716, 0.11304455995559692, -0.019776614382863045, 0.0005511343479156494, 0.0007693670922890306, -0.04503391310572624, 0.07050642371177673, -0.004905105568468571, -0.09045220166444778, 0.02283698134124279, -0.006459948141127825, -0.002421072917059064, -0.04089650511741638, 0.060738932341337204, 0.035294...
1,119
1,119
['Hesham Mostafa', 'Vishwajith Ramesh', 'Gert Cauwenberghs']
1711.06756v1
Error backpropagation is a highly effective mechanism for learning high-quality hierarchical features in deep networks. Updating the features or weights in one layer, however, requires waiting for the propagation of error signals from higher layers. Learning using delayed and non-local errors makes it hard to reconcile...
Deep supervised learning using local errors
2,017
http://arxiv.org/pdf/1711.06756v1
Title Deep supervised learning using local error Summary Error backpropagation highly effective mechanism learning highquality hierarchical feature deep network Updating feature weight one layer however requires waiting propagation error signal higher layer Learning using delayed nonlocal error make hard reconcile back...
[0.0012740854872390628, 0.02596929855644703, -0.005941873881965876, -0.006949312053620815, 0.026877105236053467, -0.008833497762680054, 0.053921595215797424, -0.005153717007488012, -0.005604002624750137, 0.012378048151731491, -0.03923448175191879, 0.022535841912031174, 0.03680374473333359, 0.036361996084451675, 0.02393...
1,120
1,120
['Eli David', 'H. Jaap van den Herik', 'Moshe Koppel', 'Nathan S. Netanyahu']
1711.06840v1
This paper demonstrates the use of genetic algorithms for evolving a grandmaster-level evaluation function for a chess program. This is achieved by combining supervised and unsupervised learning. In the supervised learning phase the organisms are evolved to mimic the behavior of human grandmasters, and in the unsupervi...
Simulating Human Grandmasters: Evolution and Coevolution of Evaluation Functions
2,017
http://arxiv.org/pdf/1711.06840v1
Title Simulating Human Grandmasters Evolution Coevolution Evaluation Functions Summary paper demonstrates use genetic algorithm evolving grandmasterlevel evaluation function chess program achieved combining supervised unsupervised learning supervised learning phase organism evolved mimic behavior human grandmaster unsu...
[0.020322106778621674, 0.020244529470801353, -0.028560908511281013, 0.012545092962682247, 0.011475012637674809, -0.009451094083487988, -0.035754527896642685, 0.0222161877900362, 0.03172442689538002, 0.0026927879080176353, 0.01810167171061039, 0.010651623830199242, -0.028525784611701965, 0.05312333628535271, 0.055465035...
1,121
1,121
['Eli David', 'Moshe Koppel', 'Nathan S. Netanyahu']
1711.06841v1
In this paper we demonstrate how genetic algorithms can be used to reverse engineer an evaluation function's parameters for computer chess. Our results show that using an appropriate expert (or mentor), we can evolve a program that is on par with top tournament-playing chess programs, outperforming a two-time World Com...
Expert-Driven Genetic Algorithms for Simulating Evaluation Functions
2,017
http://arxiv.org/pdf/1711.06841v1
Title ExpertDriven Genetic Algorithms Simulating Evaluation Functions Summary paper demonstrate genetic algorithm used reverse engineer evaluation function parameter computer chess result show using appropriate expert mentor evolve program par top tournamentplaying chess program outperforming twotime World Computer Che...
[0.02294021286070347, 0.02894601598381996, -0.037818681448698044, 0.03305109217762947, -0.024535035714507103, -0.046048056334257126, -0.047119878232479095, 0.023884175345301628, 0.0030101421289145947, -0.03153252974152565, 0.04147019237279892, 0.004094711504876614, 0.007820905186235905, 0.03704185038805008, 0.032689116...
1,122
1,122
['Eli David', 'Iddo Greental']
1711.07655v1
In recent years, deep learning methods applying unsupervised learning to train deep layers of neural networks have achieved remarkable results in numerous fields. In the past, many genetic algorithms based methods have been successfully applied to training neural networks. In this paper, we extend previous work and pro...
Genetic Algorithms for Evolving Deep Neural Networks
2,017
http://arxiv.org/pdf/1711.07655v1
Title Genetic Algorithms Evolving Deep Neural Networks Summary recent year deep learning method applying unsupervised learning train deep layer neural network achieved remarkable result numerous field past many genetic algorithm based method successfully applied training neural network paper extend previous work propos...
[-0.018242081627249718, 0.10377950221300125, -0.05786368250846863, 0.029986655339598656, -0.02349856309592724, -0.01100686751306057, -0.001891893451102078, 0.035150784999132156, 0.007364320103079081, 0.012979554943740368, 0.03234772011637688, 0.018739482387900352, -0.00653053866699338, 0.06489547342061996, 0.0189957562...
1,123
1,123
['Zuozhu Liu', 'Tony Q. S. Quek', 'Shaowei Lin']
1711.07732v1
The quest for biologically plausible deep learning is driven, not just by the desire to explain experimentally-observed properties of biological neural networks, but also by the hope of discovering more efficient methods for training artificial networks. In this paper, we propose a new algorithm named Variational Proba...
Variational Probability Flow for Biologically Plausible Training of Deep Neural Networks
2,017
http://arxiv.org/pdf/1711.07732v1
Title Variational Probability Flow Biologically Plausible Training Deep Neural Networks Summary quest biologically plausible deep learning driven desire explain experimentallyobserved property biological neural network also hope discovering efficient method training artificial network paper propose new algorithm named ...
[-0.002544934395700693, 0.004648725967854261, -0.02565760165452957, 0.007484641391783953, -0.034696727991104126, -0.022492263466119766, -0.005011906381696463, 0.004086244851350784, -0.08267068862915039, 0.033986885100603104, 0.02845451980829239, -0.011348098516464233, 0.008211774751543999, 0.05411528795957565, 0.054546...
1,124
1,124
['Eli David', 'Nathan S. Netanyahu']
1711.08336v2
This paper presents a novel deep learning based method for automatic malware signature generation and classification. The method uses a deep belief network (DBN), implemented with a deep stack of denoising autoencoders, generating an invariant compact representation of the malware behavior. While conventional signature...
DeepSign: Deep Learning for Automatic Malware Signature Generation and Classification
2,017
http://arxiv.org/pdf/1711.08336v2
Title DeepSign Deep Learning Automatic Malware Signature Generation Classification Summary paper present novel deep learning based method automatic malware signature generation classification method us deep belief network DBN implemented deep stack denoising autoencoders generating invariant compact representation malw...
[0.00925884023308754, 0.031365495175123215, -0.019826913252472878, 0.025607211515307426, -0.04124295711517334, -0.023198960348963737, 0.04058532416820526, -0.02736840210855007, 0.0460549034178257, -0.0254779364913702, 0.03489964082837105, -0.028506381437182426, 0.024820609018206596, 0.10969073325395584, 0.0621058531105...
1,125
1,125
['Eli David', 'H. Jaap van den Herik', 'Moshe Koppel', 'Nathan S. Netanyahu']
1711.08337v1
This paper demonstrates the use of genetic algorithms for evolving: 1) a grandmaster-level evaluation function, and 2) a search mechanism for a chess program, the parameter values of which are initialized randomly. The evaluation function of the program is evolved by learning from databases of (human) grandmaster games...
Genetic Algorithms for Evolving Computer Chess Programs
2,017
http://arxiv.org/pdf/1711.08337v1
Title Genetic Algorithms Evolving Computer Chess Programs Summary paper demonstrates use genetic algorithm evolving 1 grandmasterlevel evaluation function 2 search mechanism chess program parameter value initialized randomly evaluation function program evolved learning database human grandmaster game first organism evo...
[0.019404390826821327, 0.021222995594143867, -0.030976952984929085, 0.003095234977081418, 0.00010742327867774293, -0.021815089508891106, -0.057958345860242844, 0.012628355994820595, 0.01640373282134533, -0.010797156020998955, 0.03459586203098297, 0.0052337441593408585, -0.007963829673826694, 0.0452282652258873, 0.04607...
1,126
1,126
['Ishai Rosenberg', 'Guillaume Sicard', 'Eli David']
1711.09666v1
In recent years numerous advanced malware, aka advanced persistent threats (APT) are allegedly developed by nation-states. The task of attributing an APT to a specific nation-state is extremely challenging for several reasons. Each nation-state has usually more than a single cyber unit that develops such advanced malwa...
DeepAPT: Nation-State APT Attribution Using End-to-End Deep Neural Networks
2,017
http://arxiv.org/pdf/1711.09666v1
Title DeepAPT NationState APT Attribution Using EndtoEnd Deep Neural Networks Summary recent year numerous advanced malware aka advanced persistent threat APT allegedly developed nationstates task attributing APT specific nationstate extremely challenging several reason nationstate usually single cyber unit develops ad...
[0.032231904566287994, 0.08398722857236862, -0.02823079191148281, 0.018052946776151657, -0.038730423897504807, -0.03669818490743637, 0.11450744420289993, -0.04255962371826172, -0.014650348573923111, -0.00301318964920938, 0.03221747279167175, -0.008821929804980755, 0.022825969383120537, 0.06246566399931908, 0.0801397338...
1,127
1,127
['Eli David', 'Nathan S. Netanyahu', 'Lior Wolf']
1711.09667v1
We present an end-to-end learning method for chess, relying on deep neural networks. Without any a priori knowledge, in particular without any knowledge regarding the rules of chess, a deep neural network is trained using a combination of unsupervised pretraining and supervised training. The unsupervised training extra...
DeepChess: End-to-End Deep Neural Network for Automatic Learning in Chess
2,017
http://arxiv.org/pdf/1711.09667v1
Title DeepChess EndtoEnd Deep Neural Network Automatic Learning Chess Summary present endtoend learning method chess relying deep neural network Without priori knowledge particular without knowledge regarding rule chess deep neural network trained using combination unsupervised pretraining supervised training unsupervi...
[0.0037610246799886227, 0.012153133749961853, -0.026529457420110703, 0.038155969232320786, 0.008212012238800526, -0.01764289103448391, -0.01196859311312437, 0.005999785847961903, -0.022913457825779915, -0.0070060552097857, 0.02004295587539673, -0.025058090686798096, 0.025066092610359192, 0.02632308192551136, 0.05972419...
1,128
1,128
['Sebastian Urban', 'Marcus Basalla', 'Patrick van der Smagt']
1711.11059v1
We propose stochastic, non-parametric activation functions that are fully learnable and individual to each neuron. Complexity and the risk of overfitting are controlled by placing a Gaussian process prior over these functions. The result is the Gaussian process neuron, a probabilistic unit that can be used as the basic...
Gaussian Process Neurons Learn Stochastic Activation Functions
2,017
http://arxiv.org/pdf/1711.11059v1
Title Gaussian Process Neurons Learn Stochastic Activation Functions Summary propose stochastic nonparametric activation function fully learnable individual neuron Complexity risk overfitting controlled placing Gaussian process prior function result Gaussian process neuron probabilistic unit used basic building block p...
[-0.017173418775200844, 0.030412275344133377, -0.025017142295837402, 0.045846134424209595, 0.024670632556080818, -0.033847495913505554, 0.04858731850981712, -0.04409675672650337, -0.03505656123161316, 0.03333181515336037, -0.06188272312283516, 0.011672805063426495, 0.026436982676386833, 0.06290179491043091, 0.037364628...
1,129
1,129
['Nand Sharma']
1712.01977v1
The P300 event-related potential (ERP), evoked in scalp-recorded electroencephalography (EEG) by external stimuli, has proven to be a reliable response for controlling a BCI. The P300 component of an event related potential is thus widely used in brain-computer interfaces to translate the subjects' intent by mere thoug...
Single-trial P300 Classification using PCA with LDA, QDA and Neural Networks
2,017
http://arxiv.org/pdf/1712.01977v1
Title Singletrial P300 Classification using PCA LDA QDA Neural Networks Summary P300 eventrelated potential ERP evoked scalprecorded electroencephalography EEG external stimulus proven reliable response controlling BCI P300 component event related potential thus widely used braincomputer interface translate subject int...
[-0.03696279972791672, -0.009270417504012585, -0.013153167441487312, 0.02340158447623253, 0.03507130593061447, 0.03937353566288948, 0.017703216522932053, 0.0007825135253369808, 0.03180678188800812, -0.013781469315290451, 0.0017483846750110388, 0.0013999610673636198, 0.046338990330696106, 0.06322246789932251, 0.07187613...
1,130
1,130
['Haitao Zhao']
1712.04118v1
Principal component analysis (PCA) is largely adopted for chemical process monitoring and numerous PCA-based systems have been developed to solve various fault detection and diagnosis problems. Since PCA-based methods assume that the monitored process is linear, nonlinear PCA models, such as autoencoder models and kern...
Neural Component Analysis for Fault Detection
2,017
http://arxiv.org/pdf/1712.04118v1
Title Neural Component Analysis Fault Detection Summary Principal component analysis PCA largely adopted chemical process monitoring numerous PCAbased system developed solve various fault detection diagnosis problem Since PCAbased method assume monitored process linear nonlinear PCA model autoencoder model kernel princ...
[-0.019701054319739342, 0.023402836173772812, -0.040475692600011826, 0.06899537891149521, 0.04474571347236633, -0.0040658945217728615, -0.01721874438226223, 0.026618817821145058, 0.025129923596978188, -0.003830086439847946, 0.04195936769247055, 0.03842960298061371, -0.0035042930394411087, 0.04383981600403786, 0.0512229...
1,131
1,131
['Yoshihiro Nagano', 'Ryo Karakida', 'Masato Okada']
1712.04195v1
Deep generative models are reported to be useful in broad applications including image generation. Repeated inference between data space and latent space in these models can denoise cluttered images and improve the quality of inferred results. However, previous studies only qualitatively evaluated image outputs in data...
Concept Formation and Dynamics of Repeated Inference in Deep Generative Models
2,017
http://arxiv.org/pdf/1712.04195v1
Title Concept Formation Dynamics Repeated Inference Deep Generative Models Summary Deep generative model reported useful broad application including image generation Repeated inference data space latent space model denoise cluttered image improve quality inferred result However previous study qualitatively evaluated im...
[0.0032961489632725716, 0.060837626457214355, -0.04279022291302681, -0.006917293649166822, -0.0013383604818955064, 0.00887275394052267, 0.027103886008262634, -0.013472897000610828, -0.06423020362854004, 0.046813737601041794, 0.03775607794523239, -0.014390717260539532, 0.0035072958562523127, 0.06607209891080856, 0.05246...
1,132
1,132
['Noah Golowich', 'Alexander Rakhlin', 'Ohad Shamir']
1712.06541v2
We study the sample complexity of learning neural networks, by providing new bounds on their Rademacher complexity assuming norm constraints on the parameter matrix of each layer. Compared to previous work, these complexity bounds have improved dependence on the network depth, and under some additional assumptions, are...
Size-Independent Sample Complexity of Neural Networks
2,017
http://arxiv.org/pdf/1712.06541v2
Title SizeIndependent Sample Complexity Neural Networks Summary study sample complexity learning neural network providing new bound Rademacher complexity assuming norm constraint parameter matrix layer Compared previous work complexity bound improved dependence network depth additional assumption fully independent netw...
[0.003524190979078412, 0.01977686770260334, -0.02690802700817585, 0.0274517722427845, -0.013206768780946732, -0.030560728162527084, 0.014864535070955753, -0.01448987890034914, 0.010727527551352978, 0.04459585249423981, 0.0357276014983654, 0.0026398540940135717, 0.03675161302089691, 0.018321627750992775, 0.0106732007116...
1,133
1,133
['Tsendsuren Munkhdalai', 'Xingdi Yuan', 'Soroush Mehri', 'Adam Trischler']
1712.09926v2
We describe a mechanism by which artificial neural networks can learn rapid adaptation - the ability to adapt on the fly, with little data, to new tasks - that we call conditionally shifted neurons. We apply this mechanism in the framework of metalearning, where the aim is to replicate some of the flexibility of human ...
Rapid Adaptation with Conditionally Shifted Neurons
2,017
http://arxiv.org/pdf/1712.09926v2
Title Rapid Adaptation Conditionally Shifted Neurons Summary describe mechanism artificial neural network learn rapid adaptation ability adapt fly little data new task call conditionally shifted neuron apply mechanism framework metalearning aim replicate flexibility human learning machine Conditionally shifted neuron m...
[0.005249854642897844, -0.019846472889184952, -0.025360235944390297, -0.012409625574946404, -0.004113589879125357, 0.008987221866846085, 0.018232297152280807, 0.0014598462730646133, -0.051789071410894394, -0.01960615999996662, -0.023291006684303284, 0.028681369498372078, 0.008499734103679657, 0.05643950030207634, 0.025...
1,134
1,134
['Hokchhay Tann', 'Soheil Hashemi', 'Sherief Reda']
1801.07353v1
The recent success of Deep Neural Networks (DNNs) has drastically improved the state of the art for many application domains. While achieving high accuracy performance, deploying state-of-the-art DNNs is a challenge since they typically require billions of expensive arithmetic computations. In addition, DNNs are typica...
Flexible Deep Neural Network Processing
2,018
http://arxiv.org/pdf/1801.07353v1
Title Flexible Deep Neural Network Processing Summary recent success Deep Neural Networks DNNs drastically improved state art many application domain achieving high accuracy performance deploying stateoftheart DNNs challenge since typically require billion expensive arithmetic computation addition DNNs typically deploy...
[-0.019001200795173645, 0.03419560566544533, -0.025929687544703484, 0.03815295919775963, 0.01863759197294712, -0.018791353330016136, 0.04537772387266159, -0.028053848072886467, -0.03891840949654579, 0.003456718288362026, -0.00034248598967678845, 0.02721397764980793, 0.025742389261722565, 0.07948097586631775, -0.0093286...
1,135
1,135
['Hu Wang']
1801.09856v1
The artificial neural network shows powerful ability of inference, but it is still criticized for lack of interpretability and prerequisite needs of big dataset. This paper proposes the Rule-embedded Neural Network (ReNN) to overcome the shortages. ReNN first makes local-based inferences to detect local patterns, and t...
ReNN: Rule-embedded Neural Networks
2,018
http://arxiv.org/pdf/1801.09856v1
Title ReNN Ruleembedded Neural Networks Summary artificial neural network show powerful ability inference still criticized lack interpretability prerequisite need big dataset paper proposes Ruleembedded Neural Network ReNN overcome shortage ReNN first make localbased inference detect local pattern us rule based domain ...
[-0.008670496754348278, 0.04612550139427185, -0.04507697746157646, 0.022900013253092766, -0.024079469963908195, -0.011817583814263344, 0.04279576614499092, 0.02491512894630432, -0.0035166104789823294, 0.021882185712456703, 0.03424374759197235, 0.010227542370557785, 0.007281122263520956, 0.036688871681690216, -0.0003815...
1,136
1,136
['Akisato Kimura', 'Zoubin Ghahramani', 'Koh Takeuchi', 'Tomoharu Iwata', 'Naonori Ueda']
1802.03039v2
In this paper, we propose imitation networks, a simple but effective method for training neural networks with a limited amount of training data. Our approach inherits the idea of knowledge distillation that transfers knowledge from a deep or wide reference model to a shallow or narrow target model. The proposed method ...
Imitation networks: Few-shot learning of neural networks from scratch
2,018
http://arxiv.org/pdf/1802.03039v2
Title Imitation network Fewshot learning neural network scratch Summary paper propose imitation network simple effective method training neural network limited amount training data approach inherits idea knowledge distillation transfer knowledge deep wide reference model shallow narrow target model proposed method empl...
[0.009386753663420677, -0.0036415390204638243, 0.017495201900601387, 0.04016489163041115, 0.01671043410897255, -0.027781346812844276, 0.04019254073500633, 0.02911299094557762, -0.019321482628583908, -0.024630064144730568, -0.03135360777378082, 0.02040671370923519, -0.04812893643975258, -0.042363278567790985, 0.04167685...
1,137
1,137
['Wengong Jin', 'Regina Barzilay', 'Tommi Jaakkola']
1802.04364v2
We seek to automate the design of molecules based on specific chemical properties. In computational terms, this task involves continuous embedding and generation of molecular graphs. Our primary contribution is the direct realization of molecular graphs, a task previously approached by generating linear SMILES strings ...
Junction Tree Variational Autoencoder for Molecular Graph Generation
2,018
http://arxiv.org/pdf/1802.04364v2
Title Junction Tree Variational Autoencoder Molecular Graph Generation Summary seek automate design molecule based specific chemical property computational term task involves continuous embedding generation molecular graph primary contribution direct realization molecular graph task previously approached generating lin...
[0.0021462864242494106, 0.06322257965803146, -0.03349020332098007, -0.004740773234516382, -0.013205011375248432, -0.04205642268061638, 0.0017355246236547828, -0.013193567283451557, 0.05757036805152893, -0.0017536296509206295, 0.041107553988695145, 0.009035793133080006, -0.022332405671477318, 0.13015849888324738, 0.0650...
1,138
1,138
['Raghavendra Chalapathy', 'Aditya Krishna Menon', 'Sanjay Chawla']
1802.06360v1
We propose a one-class neural network (OC-NN) model to detect anomalies in complex data sets. OC-NN combines the ability of deep networks to extract progressively rich representation of data with the one-class objective of creating a tight envelope around normal data. The OC-NN approach breaks new ground for the follow...
Anomaly Detection using One-Class Neural Networks
2,018
http://arxiv.org/pdf/1802.06360v1
Title Anomaly Detection using OneClass Neural Networks Summary propose oneclass neural network OCNN model detect anomaly complex data set OCNN combine ability deep network extract progressively rich representation data oneclass objective creating tight envelope around normal data OCNN approach break new ground followin...
[-0.010964613407850266, 0.047546565532684326, 0.004046381451189518, 0.05215051770210266, -0.02251601591706276, 0.0250351931899786, 0.018512263894081116, 0.027146663516759872, -0.032383762300014496, 0.013866227120161057, 0.08955027908086777, 0.00839999970048666, -0.03316951170563698, 0.0783255398273468, -0.0179366581141...
1,139
1,139
['Alireza Bagheri', 'Osvaldo Simeone', 'Bipin Rajendran']
1802.08567v2
Classifiers trained using conventional empirical risk minimization or maximum likelihood methods are known to suffer dramatic performance degradations when tested over examples adversarially selected based on knowledge of the classifier's decision rule. Due to the prominence of Artificial Neural Networks (ANNs) as clas...
Adversarial Training for Probabilistic Spiking Neural Networks
2,018
http://arxiv.org/pdf/1802.08567v2
Title Adversarial Training Probabilistic Spiking Neural Networks Summary Classifiers trained using conventional empirical risk minimization maximum likelihood method known suffer dramatic performance degradation tested example adversarially selected based knowledge classifier decision rule Due prominence Artificial Neu...
[-0.02947329543530941, -0.0031885800417512655, -0.025176960974931717, 0.01962488889694214, 0.03315690904855728, -0.007650242652744055, 0.028072137385606766, -0.0126175032928586, 0.006367605645209551, -0.026162298396229744, -0.0292319618165493, 0.05352967977523804, 0.007382904179394245, 0.06434989720582962, 0.0350425280...
1,140
1,140
['Simone Scardapane', 'Steven Van Vaerenbergh', 'Danilo Comminiello', 'Aurelio Uncini']
1802.09405v1
Graph neural networks (GNNs) are a class of neural networks that allow to efficiently perform inference on data that is associated to a graph structure, such as, e.g., citation networks or knowledge graphs. While several variants of GNNs have been proposed, they only consider simple nonlinear activation functions in th...
Improving Graph Convolutional Networks with Non-Parametric Activation Functions
2,018
http://arxiv.org/pdf/1802.09405v1
Title Improving Graph Convolutional Networks NonParametric Activation Functions Summary Graph neural network GNNs class neural network allow efficiently perform inference data associated graph structure eg citation network knowledge graph several variant GNNs proposed consider simple nonlinear activation function layer...
[0.006082061678171158, 0.014817495830357075, 0.008025052025914192, 0.042837973684072495, -0.014543769881129265, -0.024829404428601265, -0.006329887080937624, 0.028335634618997574, 0.052533842623233795, -0.012339714914560318, -0.03793356940150261, 0.028873000293970108, 0.00017180068243760616, 0.05213920399546623, 0.0239...
1,141
1,141
['Chaoyue Wang', 'Chang Xu', 'Xin Yao', 'Dacheng Tao']
1803.00657v1
Generative adversarial networks (GAN) have been effective for learning generative models for real-world data. However, existing GANs (GAN and its variants) tend to suffer from training problems such as instability and mode collapse. In this paper, we propose a novel GAN framework called evolutionary generative adversar...
Evolutionary Generative Adversarial Networks
2,018
http://arxiv.org/pdf/1803.00657v1
Title Evolutionary Generative Adversarial Networks Summary Generative adversarial network GAN effective learning generative model realworld data However existing GANs GAN variant tend suffer training problem instability mode collapse paper propose novel GAN framework called evolutionary generative adversarial network E...
[0.010272204875946045, 0.09158603101968765, -0.030459903180599213, -0.002668011700734496, 0.0026398932095617056, 0.003979936242103577, 0.01650751382112503, 0.018956409767270088, -0.06062399595975876, 0.05409334599971771, 0.0051561021246016026, 0.018835460767149925, -0.042248327285051346, 0.027485426515340805, 0.0465993...
1,142
1,142
['Boyuan Chen', 'Harvey Wu', 'Warren Mo', 'Ishanu Chattopadhyay', 'Hod Lipson']
1803.00684v1
We introduce an automatic machine learning (AutoML) modeling architecture called Autostacker, which combines an innovative hierarchical stacking architecture and an Evolutionary Algorithm (EA) to perform efficient parameter search. Neither prior domain knowledge about the data nor feature preprocessing is needed. Using...
Autostacker: A Compositional Evolutionary Learning System
2,018
http://arxiv.org/pdf/1803.00684v1
Title Autostacker Compositional Evolutionary Learning System Summary introduce automatic machine learning AutoML modeling architecture called Autostacker combine innovative hierarchical stacking architecture Evolutionary Algorithm EA perform efficient parameter search Neither prior domain knowledge data feature preproc...
[-0.009350688196718693, 0.07142148911952972, -0.03556790575385094, -0.01868763566017151, 0.0034113661386072636, 0.002789134392514825, 0.008181674405932426, -0.005109881982207298, -0.021102387458086014, -0.02422032319009304, 0.016028722748160362, 0.020955143496394157, 0.00630628177896142, 0.08925792574882507, 0.01131764...
1,143
1,143
['Savitha Ramasamy', 'Kanagasabai Rajaraman', 'Pavitra Krishnaswamy', 'Vijay Chandrasekhar']
1803.02043v1
We propose a novel online learning algorithm for Restricted Boltzmann Machines (RBM), namely, the Online Generative Discriminative Restricted Boltzmann Machine (OGD-RBM), that provides the ability to build and adapt the network architecture of RBM according to the statistics of streaming data. The OGD-RBM is trained in...
Online Deep Learning: Growing RBM on the fly
2,018
http://arxiv.org/pdf/1803.02043v1
Title Online Deep Learning Growing RBM fly Summary propose novel online learning algorithm Restricted Boltzmann Machines RBM namely Online Generative Discriminative Restricted Boltzmann Machine OGDRBM provides ability build adapt network architecture RBM according statistic streaming data OGDRBM trained two phase 1 onl...
[-0.005910401698201895, 0.041006047278642654, -0.042627591639757156, 0.018145432695746422, 0.019718311727046967, 0.0001682284491835162, 0.03820082172751427, -0.000826661940664053, -0.02746398001909256, 0.009925426915287971, -0.012729462236166, 0.022809363901615143, -0.011376707814633846, 0.08152801543474197, -0.0140504...
1,144
1,144
['Konstantinos Pitas', 'Mike Davies', 'Pierre Vandergheynst']
1803.04239v1
Recent DNN pruning algorithms have succeeded in reducing the number of parameters in fully connected layers, often with little or no drop in classification accuracy. However, most of the existing pruning schemes either have to be applied during training or require a costly retraining procedure after pruning to regain c...
FeTa: A DCA Pruning Algorithm with Generalization Error Guarantees
2,018
http://arxiv.org/pdf/1803.04239v1
Title FeTa DCA Pruning Algorithm Generalization Error Guarantees Summary Recent DNN pruning algorithm succeeded reducing number parameter fully connected layer often little drop classification accuracy However existing pruning scheme either applied training require costly retraining procedure pruning regain classificat...
[-0.025678711012005806, 0.011933911591768265, -0.055938661098480225, 0.01918279379606247, 0.007981592789292336, -0.041590627282857895, 0.026866603642702103, 0.00506711658090353, -0.03171619400382042, 0.048783935606479645, -0.002492825035005808, -5.415475243353285e-06, 0.06292160600423813, 0.012489678338170052, -0.02443...
1,145
1,145
['Andy Brown', 'Aaron Tuor', 'Brian Hutchinson', 'Nicole Nichols']
1803.04967v1
Deep learning has recently demonstrated state-of-the art performance on key tasks related to the maintenance of computer systems, such as intrusion detection, denial of service attack detection, hardware and software system failures, and malware detection. In these contexts, model interpretability is vital for administ...
Recurrent Neural Network Attention Mechanisms for Interpretable System Log Anomaly Detection
2,018
http://arxiv.org/pdf/1803.04967v1
Title Recurrent Neural Network Attention Mechanisms Interpretable System Log Anomaly Detection Summary Deep learning recently demonstrated stateofthe art performance key task related maintenance computer system intrusion detection denial service attack detection hardware software system failure malware detection contex...
[0.0412030965089798, 0.010878362692892551, 0.003240384394302964, 0.04046930745244026, -0.039829518646001816, -0.021998239681124687, 0.029133031144738197, -0.02874932251870632, -0.030092651024460793, -0.062430016696453094, 0.05303765833377838, -0.01854141615331173, -0.013156484812498093, 0.08448047190904617, 0.023695535...
1,146
1,146
['Karlson Pfannschmidt', 'Pritha Gupta', 'Eyke Hüllermeier']
1803.05796v1
Object ranking is an important problem in the realm of preference learning. On the basis of training data in the form of a set of rankings of objects, which are typically represented as feature vectors, the goal is to learn a ranking function that predicts a linear order of any new set of objects. Current approaches co...
Deep architectures for learning context-dependent ranking functions
2,018
http://arxiv.org/pdf/1803.05796v1
Title Deep architecture learning contextdependent ranking function Summary Object ranking important problem realm preference learning basis training data form set ranking object typically represented feature vector goal learn ranking function predicts linear order new set object Current approach commonly focus ranking ...
[0.01705208793282509, 0.08603771775960922, -0.0036730323918163776, 0.015018611215054989, 0.019471652805805206, -0.023915840312838554, 0.009506791830062866, -0.03602812811732292, -0.013317014090716839, -0.034494925290346146, -0.016427036374807358, 0.013872172683477402, 0.01318166684359312, 0.055431634187698364, 0.002908...
1,147
1,147
['Pierre Sermanet', 'Soumith Chintala', 'Yann LeCun']
1204.3968v1
We classify digits of real-world house numbers using convolutional neural networks (ConvNets). ConvNets are hierarchical feature learning neural networks whose structure is biologically inspired. Unlike many popular vision approaches that are hand-designed, ConvNets can automatically learn a unique set of features opti...
Convolutional Neural Networks Applied to House Numbers Digit Classification
2,012
http://arxiv.org/pdf/1204.3968v1
Title Convolutional Neural Networks Applied House Numbers Digit Classification Summary classify digit realworld house number using convolutional neural network ConvNets ConvNets hierarchical feature learning neural network whose structure biologically inspired Unlike many popular vision approach handdesigned ConvNets a...
[-0.0295088067650795, 0.05542977899312973, -0.00955856405198574, 0.07740107923746109, -0.00398270133882761, 0.011271463707089424, 0.05721588432788849, -0.017523469403386116, -0.02410871721804142, -0.011100158095359802, -0.03473342955112457, 0.027583153918385506, 0.023189879953861237, 0.028961583971977234, 0.01630169339...
1,148
1,148
['Charles F. Cadieu', 'Ha Hong', 'Dan Yamins', 'Nicolas Pinto', 'Najib J. Majaj', 'James J. DiCarlo']
1301.3530v2
A key requirement for the development of effective learning representations is their evaluation and comparison to representations we know to be effective. In natural sensory domains, the community has viewed the brain as a source of inspiration and as an implicit benchmark for success. However, it has not been possible...
The Neural Representation Benchmark and its Evaluation on Brain and Machine
2,013
http://arxiv.org/pdf/1301.3530v2
Title Neural Representation Benchmark Evaluation Brain Machine Summary key requirement development effective learning representation evaluation comparison representation know effective natural sensory domain community viewed brain source inspiration implicit benchmark success However possible directly test representati...
[-0.0039588212966918945, 0.003302240977063775, -0.03360433876514435, 0.022029174491763115, -0.002450785366818309, 0.03976312652230263, 0.04049595072865486, -0.002576206112280488, 0.018751533702015877, -0.00841642264276743, -0.05868291109800339, 0.019487865269184113, 0.05903693288564682, 0.03654365986585617, 0.063618250...
1,149
1,149
['Sainbayar Sukhbaatar', 'Joan Bruna', 'Manohar Paluri', 'Lubomir Bourdev', 'Rob Fergus']
1406.2080v4
The availability of large labeled datasets has allowed Convolutional Network models to achieve impressive recognition results. However, in many settings manual annotation of the data is impractical; instead our data has noisy labels, i.e. there is some freely available label for each image which may or may not be accur...
Training Convolutional Networks with Noisy Labels
2,014
http://arxiv.org/pdf/1406.2080v4
Title Training Convolutional Networks Noisy Labels Summary availability large labeled datasets allowed Convolutional Network model achieve impressive recognition result However many setting manual annotation data impractical instead data noisy label ie freely available label image may may accurate paper explore perform...
[0.04108227789402008, 0.05968485400080681, 0.01706288568675518, 0.09314297139644623, 0.0622330866754055, 0.01650608330965042, 0.030579162761569023, 0.010779866948723793, 0.0009578403551131487, -0.011014712043106556, -0.05105092003941536, 0.04187460243701935, -0.04658769071102142, -0.01153375580906868, 0.027952730655670...
1,150
1,150
['Spyros Gidaris', 'Nikos Komodakis']
1505.01749v3
We propose an object detection system that relies on a multi-region deep convolutional neural network (CNN) that also encodes semantic segmentation-aware features. The resulting CNN-based representation aims at capturing a diverse set of discriminative appearance factors and exhibits localization sensitivity that is es...
Object detection via a multi-region & semantic segmentation-aware CNN model
2,015
http://arxiv.org/pdf/1505.01749v3
Title Object detection via multiregion semantic segmentationaware CNN model Summary propose object detection system relies multiregion deep convolutional neural network CNN also encodes semantic segmentationaware feature resulting CNNbased representation aim capturing diverse set discriminative appearance factor exhibi...
[0.007372119929641485, 0.030670158565044403, 0.019754108041524887, 0.08465336263179779, -0.0007036687457002699, -0.0014927758602425456, 0.01239309273660183, -0.02302706055343151, -0.03840603306889534, -0.03290083259344101, 0.011282364837825298, 0.06122457981109619, -0.020796116441488266, 0.0012578039895743132, -0.01285...
1,151
1,151
['Yanhai Gan', 'Jun Liu', 'Junyu Dong', 'Guoqiang Zhong']
1505.03703v1
In this paper, we propose a novel unsupervised deep learning model, called PCA-based Convolutional Network (PCN). The architecture of PCN is composed of several feature extraction stages and a nonlinear output stage. Particularly, each feature extraction stage includes two layers: a convolutional layer and a feature po...
A PCA-Based Convolutional Network
2,015
http://arxiv.org/pdf/1505.03703v1
Title PCABased Convolutional Network Summary paper propose novel unsupervised deep learning model called PCAbased Convolutional Network PCN architecture PCN composed several feature extraction stage nonlinear output stage Particularly feature extraction stage includes two layer convolutional layer feature pooling layer...
[-0.02677331492304802, 0.048957161605358124, -0.0133145647123456, 0.06440284848213196, 0.019574133679270744, 0.013990684412419796, 0.0576404333114624, -0.024195870384573936, 0.0013291968498378992, 0.0467040129005909, -0.012295362539589405, -0.0005471096956171095, 0.003021844895556569, 0.10288002341985703, 0.02170361764...
1,152
1,152
['Nikolaos Karianakis', 'Jingming Dong', 'Stefano Soatto']
1505.06795v2
We conduct an empirical study to test the ability of Convolutional Neural Networks (CNNs) to reduce the effects of nuisance transformations of the input data, such as location, scale and aspect ratio. We isolate factors by adopting a common convolutional architecture either deployed globally on the image to compute cla...
An Empirical Evaluation of Current Convolutional Architectures' Ability to Manage Nuisance Location and Scale Variability
2,015
http://arxiv.org/pdf/1505.06795v2
Title Empirical Evaluation Current Convolutional Architectures Ability Manage Nuisance Location Scale Variability Summary conduct empirical study test ability Convolutional Neural Networks CNNs reduce effect nuisance transformation input data location scale aspect ratio isolate factor adopting common convolutional arch...
[0.0468648299574852, 0.044710274785757065, -0.036668311804533005, 0.008009655401110649, -0.03680485859513283, -0.016945745795965195, 0.09149753302335739, -0.01071076188236475, -0.0353146456182003, 0.010362102650105953, -0.011211677454411983, -0.0037144862581044436, 0.01688903011381626, 0.009227074682712555, 0.011847059...
1,153
1,153
['Xundong Wu', 'Yong Wu', 'Yong Zhao']
1604.03058v5
We trained Binarized Neural Networks (BNNs) on the high resolution ImageNet ILSVRC-2102 dataset classification task and achieved a good performance. With a moderate size network of 13 layers, we obtained top-5 classification accuracy rate of 84.1 % on validation set through network distillation, much better than previo...
Binarized Neural Networks on the ImageNet Classification Task
2,016
http://arxiv.org/pdf/1604.03058v5
Title Binarized Neural Networks ImageNet Classification Task Summary trained Binarized Neural Networks BNNs high resolution ImageNet ILSVRC2102 dataset classification task achieved good performance moderate size network 13 layer obtained top5 classification accuracy rate 841 validation set network distillation much bet...
[-0.002541116438806057, 0.016054600477218628, -0.01673029363155365, 0.05816510692238808, 0.005526314489543438, -0.002218130510300398, 0.05388500913977623, 0.029977304860949516, -0.030868738889694214, 0.03593028709292412, 0.013777765445411205, -0.012954327277839184, 0.0376349575817585, 0.021037405356764793, -0.011757368...
1,154
1,154
['Farahnaz Ahmed Wick', 'Michael L. Wick', 'Marc Pomplun']
1604.04125v1
Humans perceive their surroundings in great detail even though most of our visual field is reduced to low-fidelity color-deprived (e.g. dichromatic) input by the retina. In contrast, most deep learning architectures are computationally wasteful in that they consider every part of the input when performing an image proc...
Filling in the details: Perceiving from low fidelity images
2,016
http://arxiv.org/pdf/1604.04125v1
Title Filling detail Perceiving low fidelity image Summary Humans perceive surroundings great detail even though visual field reduced lowfidelity colordeprived eg dichromatic input retina contrast deep learning architecture computationally wasteful consider every part input performing image processing task Yet human vi...
[0.014998704195022583, 0.034216172993183136, -0.011359835974872112, 0.026553815230727196, -0.03567114099860191, -0.011959845200181007, 0.021425774320960045, 0.018946614116430283, -0.06814071536064148, 0.01758870668709278, 0.02882254868745804, -0.013187600299715996, 0.04372801259160042, 0.07052610069513321, 0.0125433243...
1,155
1,155
['Youngbin Park', 'Sungphill Moon', 'Il Hong Suh']
1604.04528v1
Kinect skeleton tracker is able to achieve considerable human body tracking performance in convenient and a low-cost manner. However, The tracker often captures unnatural human poses such as discontinuous and vibrated motions when self-occlusions occur. A majority of approaches tackle this problem by using multiple Kin...
Tracking Human-like Natural Motion Using Deep Recurrent Neural Networks
2,016
http://arxiv.org/pdf/1604.04528v1
Title Tracking Humanlike Natural Motion Using Deep Recurrent Neural Networks Summary Kinect skeleton tracker able achieve considerable human body tracking performance convenient lowcost manner However tracker often capture unnatural human pose discontinuous vibrated motion selfocclusions occur majority approach tackle ...
[0.0316491462290287, 0.0404246523976326, 0.008870125748217106, 0.07162043452262878, 0.009143626317381859, -0.01760951057076454, -0.023010095581412315, -0.0016386572970077395, -0.015485485084354877, -0.018613839522004128, -0.010070974938571453, -0.004958989564329386, 0.031186124309897423, 0.02513994835317135, 0.04224587...
1,156
1,156
['Yueying Kao', 'Ran He', 'Kaiqi Huang']
1604.04970v3
Human beings often assess the aesthetic quality of an image coupled with the identification of the image's semantic content. This paper addresses the correlation issue between automatic aesthetic quality assessment and semantic recognition. We cast the assessment problem as the main task among a multi-task deep model, ...
Deep Aesthetic Quality Assessment with Semantic Information
2,016
http://arxiv.org/pdf/1604.04970v3
Title Deep Aesthetic Quality Assessment Semantic Information Summary Human being often ass aesthetic quality image coupled identification image semantic content paper address correlation issue automatic aesthetic quality assessment semantic recognition cast assessment problem main task among multitask deep model argue ...
[0.017322996631264687, 0.08146955817937851, -0.009402932599186897, 0.06560926139354706, -0.018968982622027397, 0.011272359639406204, 0.029344549402594566, 0.005888975225389004, -0.0552033968269825, -0.007680512499064207, -0.045550137758255005, 0.01375756599009037, 0.04708356782793999, 0.005729892756789923, 0.0031730278...
1,157
1,157
['Xichuan Zhou', 'Shengli Li', 'Kai Qin', 'Kunping Li', 'Fang Tang', 'Shengdong Hu', 'Shujun Liu', 'Zhi Lin']
1604.06154v1
Deep neural networks are state-of-the-art models for understanding the content of images, video and raw input data. However, implementing a deep neural network in embedded systems is a challenging task, because a typical deep neural network, such as a Deep Belief Network using 128x128 images as input, could exhaust Gig...
Deep Adaptive Network: An Efficient Deep Neural Network with Sparse Binary Connections
2,016
http://arxiv.org/pdf/1604.06154v1
Title Deep Adaptive Network Efficient Deep Neural Network Sparse Binary Connections Summary Deep neural network stateoftheart model understanding content image video raw input data However implementing deep neural network embedded system challenging task typical deep neural network Deep Belief Network using 128x128 ima...
[-0.030481301248073578, 0.044509630650281906, -0.018535345792770386, 0.09910361468791962, 0.013808433897793293, -0.028728630393743515, 0.054738849401474, 0.013895130716264248, -0.009262732230126858, -0.0022414277773350477, 0.0010786964558064938, 0.011214299127459526, 0.021869424730539322, 0.04535343125462532, 0.0202816...
1,158
1,158
['Tung Nguyen', 'Kazuki Mori', 'Ruck Thawonmas']
1604.07904v1
In this paper, we present a novel approach that uses deep learning techniques for colorizing grayscale images. By utilizing a pre-trained convolutional neural network, which is originally designed for image classification, we are able to separate content and style of different images and recombine them into a single im...
Image Colorization Using a Deep Convolutional Neural Network
2,016
http://arxiv.org/pdf/1604.07904v1
Title Image Colorization Using Deep Convolutional Neural Network Summary paper present novel approach us deep learning technique colorizing grayscale image utilizing pretrained convolutional neural network originally designed image classification able separate content style different image recombine single image propos...
[0.005037536844611168, 0.046755556017160416, 0.02543296478688717, 0.06628739088773727, 0.016366271302103996, 0.02372201904654503, 0.02275553159415722, 0.008643696084618568, -0.07339019328355789, -0.009657555259764194, -0.05347876623272896, 4.652081406675279e-05, 0.03374161198735237, 0.041877325624227524, -0.02242477051...
1,159
1,159
['Ragav Venkatesan', 'Baoxin Li']
1604.08220v1
Modern computer vision is all about the possession of powerful image representations. Deeper and deeper convolutional neural networks have been built using larger and larger datasets and are made publicly available. A large swath of computer vision scientists use these pre-trained networks with varying degrees of succe...
Diving deeper into mentee networks
2,016
http://arxiv.org/pdf/1604.08220v1
Title Diving deeper mentee network Summary Modern computer vision possession powerful image representation Deeper deeper convolutional neural network built using larger larger datasets made publicly available large swath computer vision scientist use pretrained network varying degree success various task Even though tr...
[-0.013223097659647465, 0.04181468114256859, -0.024073075503110886, 0.02593091130256653, -0.01546950452029705, -0.010036279447376728, 0.07521811872720718, 0.0015826351009309292, -0.033967193216085434, 0.014024985022842884, -0.0024407438468188047, 0.041446369141340256, -0.009952100925147533, 0.05126949027180672, 0.03475...
1,160
1,160
['Théodore Bluche']
1604.08352v1
Offline handwriting recognition systems require cropped text line images for both training and recognition. On the one hand, the annotation of position and transcript at line level is costly to obtain. On the other hand, automatic line segmentation algorithms are prone to errors, compromising the subsequent recognition...
Joint Line Segmentation and Transcription for End-to-End Handwritten Paragraph Recognition
2,016
http://arxiv.org/pdf/1604.08352v1
Title Joint Line Segmentation Transcription EndtoEnd Handwritten Paragraph Recognition Summary Offline handwriting recognition system require cropped text line image training recognition one hand annotation position transcript line level costly obtain hand automatic line segmentation algorithm prone error compromising ...
[-0.014315525069832802, 0.02320125699043274, 0.029534975066781044, 0.06339122354984283, -0.051685355603694916, -0.012972481548786163, 0.06986070424318314, 0.013262067921459675, 0.004646047484129667, -0.0359012745320797, 0.0075047994032502174, -0.0072057489305734634, 0.0552641823887825, 0.015586316585540771, 0.012282567...
1,161
1,161
['Shu Kong', 'Zhuolin Jiang', 'Qiang Yang']
1401.5535v2
We now know that mid-level features can greatly enhance the performance of image learning, but how to automatically learn the image features efficiently and in an unsupervised manner is still an open question. In this paper, we present a very efficient mid-level feature learning approach (MidFea), which only involves s...
Learning Mid-Level Features and Modeling Neuron Selectivity for Image Classification
2,014
http://arxiv.org/pdf/1401.5535v2
Title Learning MidLevel Features Modeling Neuron Selectivity Image Classification Summary know midlevel feature greatly enhance performance image learning automatically learn image feature efficiently unsupervised manner still open question paper present efficient midlevel feature learning approach MidFea involves simp...
[-0.007802170235663652, 0.041989076882600784, -0.01772521249949932, 0.011236567981541157, 0.003773581236600876, 0.039561886340379715, 0.03018280118703842, 0.020399829372763634, 0.018079638481140137, -0.04010763391852379, -0.004959286656230688, 0.009344859048724174, -0.0008382834494113922, 0.0351761132478714, 0.00714441...
1,162
1,162
['Wenbin Li', 'Mario Fritz']
1408.2938v1
The recent progress in sparse coding and deep learning has made unsupervised feature learning methods a strong competitor to hand-crafted descriptors. In computer vision, success stories of learned features have been predominantly reported for object recognition tasks. In this paper, we investigate if and how feature l...
Learning Multi-Scale Representations for Material Classification
2,014
http://arxiv.org/pdf/1408.2938v1
Title Learning MultiScale Representations Material Classification Summary recent progress sparse coding deep learning made unsupervised feature learning method strong competitor handcrafted descriptor computer vision success story learned feature predominantly reported object recognition task paper investigate feature ...
[-0.05626474320888519, 0.017522213980555534, -0.0003610796411521733, 0.10350881516933441, 0.009545763954520226, 0.025959372520446777, 0.06340019404888153, 0.03212714195251465, -0.034482184797525406, 0.02562752366065979, -0.03621735796332359, 0.02508546970784664, -0.0023223471362143755, 0.04681791737675667, 0.0150165967...
1,163
1,163
['Nitish Srivastava', 'Elman Mansimov', 'Ruslan Salakhutdinov']
1502.04681v3
We use multilayer Long Short Term Memory (LSTM) networks to learn representations of video sequences. Our model uses an encoder LSTM to map an input sequence into a fixed length representation. This representation is decoded using single or multiple decoder LSTMs to perform different tasks, such as reconstructing the i...
Unsupervised Learning of Video Representations using LSTMs
2,015
http://arxiv.org/pdf/1502.04681v3
Title Unsupervised Learning Video Representations using LSTMs Summary use multilayer Long Short Term Memory LSTM network learn representation video sequence model us encoder LSTM map input sequence fixed length representation representation decoded using single multiple decoder LSTMs perform different task reconstructi...
[-0.052686020731925964, 0.011461062356829643, -0.0034623087849467993, 0.055924344807863235, 0.005129434168338776, 0.0169371347874403, 0.03400232642889023, 0.011858956888318062, -0.01387422438710928, -0.0677715465426445, -0.01668103225529194, -0.04760347306728363, -0.012540929019451141, 0.04347721114754677, 0.0083717759...
1,164
1,164
['Adam W. Harley', 'Alex Ufkes', 'Konstantinos G. Derpanis']
1502.07058v1
This paper presents a new state-of-the-art for document image classification and retrieval, using features learned by deep convolutional neural networks (CNNs). In object and scene analysis, deep neural nets are capable of learning a hierarchical chain of abstraction from pixel inputs to concise and descriptive represe...
Evaluation of Deep Convolutional Nets for Document Image Classification and Retrieval
2,015
http://arxiv.org/pdf/1502.07058v1
Title Evaluation Deep Convolutional Nets Document Image Classification Retrieval Summary paper present new stateoftheart document image classification retrieval using feature learned deep convolutional neural network CNNs object scene analysis deep neural net capable learning hierarchical chain abstraction pixel input ...
[0.013508636504411697, 0.050493694841861725, 0.0162911806255579, 0.04685450345277786, -0.006317938677966595, 0.018621070310473442, 0.044260263442993164, 0.04868004471063614, -0.015194610692560673, -0.02569856494665146, -0.05465009808540344, 0.025278130546212196, -0.018749767914414406, 0.029190093278884888, -0.016847314...
1,165
1,165
['Tejas D. Kulkarni', 'Will Whitney', 'Pushmeet Kohli', 'Joshua B. Tenenbaum']
1503.03167v4
This paper presents the Deep Convolution Inverse Graphics Network (DC-IGN), a model that learns an interpretable representation of images. This representation is disentangled with respect to transformations such as out-of-plane rotations and lighting variations. The DC-IGN model is composed of multiple layers of convol...
Deep Convolutional Inverse Graphics Network
2,015
http://arxiv.org/pdf/1503.03167v4
Title Deep Convolutional Inverse Graphics Network Summary paper present Deep Convolution Inverse Graphics Network DCIGN model learns interpretable representation image representation disentangled respect transformation outofplane rotation lighting variation DCIGN model composed multiple layer convolution deconvolution ...
[0.0011044723214581609, 0.029420452192425728, 0.015137768350541592, 0.04677632823586464, -0.016094936057925224, 0.004330192226916552, 0.0075656576082110405, -0.007865700870752335, -0.054606929421424866, 0.02428392693400383, -0.013117609545588493, 0.05206291005015373, -0.002153700217604637, 0.04476996883749962, 0.052979...
1,166
1,166
['Mark D. McDonnell', 'Tony Vladusich']
1503.04596v3
We present a neural network architecture and training method designed to enable very rapid training and low implementation complexity. Due to its training speed and very few tunable parameters, the method has strong potential for applications requiring frequent retraining or online training. The approach is characteriz...
Enhanced Image Classification With a Fast-Learning Shallow Convolutional Neural Network
2,015
http://arxiv.org/pdf/1503.04596v3
Title Enhanced Image Classification FastLearning Shallow Convolutional Neural Network Summary present neural network architecture training method designed enable rapid training low implementation complexity Due training speed tunable parameter method strong potential application requiring frequent retraining online tra...
[0.021685756742954254, 0.006532217375934124, -0.028852449730038643, 0.06513015180826187, 0.035899143666028976, -0.021469274535775185, 0.06414470821619034, 0.016379402950406075, 0.003170014824718237, -0.0132228909060359, -0.032111331820487976, 0.059331174939870834, 0.022099893540143967, 0.04688040167093277, 0.0408718585...
1,167
1,167
['Pedro Tabacof', 'Julia Tavares', 'Eduardo Valle']
1612.00155v1
We investigate adversarial attacks for autoencoders. We propose a procedure that distorts the input image to mislead the autoencoder in reconstructing a completely different target image. We attack the internal latent representations, attempting to make the adversarial input produce an internal representation as simila...
Adversarial Images for Variational Autoencoders
2,016
http://arxiv.org/pdf/1612.00155v1
Title Adversarial Images Variational Autoencoders Summary investigate adversarial attack autoencoders propose procedure distorts input image mislead autoencoder reconstructing completely different target image attack internal latent representation attempting make adversarial input produce internal representation simila...
[0.01628081500530243, 0.06574033945798874, -0.05422608554363251, 0.030601464211940765, -0.03613877668976784, -0.0053708916530013084, 0.0425930917263031, -0.015294991433620453, -0.053283099085092545, -0.041203636676073074, 0.0025408079382032156, 0.018495548516511917, 0.019637271761894226, 0.06512095779180527, 0.08778548...
1,168
1,168
['Jonathan A. Cox']
1612.00891v2
The significant computational costs of deploying neural networks in large-scale or resource constrained environments, such as data centers and mobile devices, has spurred interest in model compression, which can achieve a reduction in both arithmetic operations and storage memory. Several techniques have been proposed ...
Parameter Compression of Recurrent Neural Networks and Degradation of Short-term Memory
2,016
http://arxiv.org/pdf/1612.00891v2
Title Parameter Compression Recurrent Neural Networks Degradation Shortterm Memory Summary significant computational cost deploying neural network largescale resource constrained environment data center mobile device spurred interest model compression achieve reduction arithmetic operation storage memory Several techni...
[-0.012158039957284927, -0.005493543576449156, 0.009676381014287472, 0.036406707018613815, 0.015727955847978592, -0.03607922047376633, 0.02531939186155796, -0.005473319906741381, -0.09070512652397156, 0.011097518727183342, -0.02061234414577484, 0.0011853245086967945, 0.04986757040023804, -0.012282567098736763, 0.017201...
1,169
1,169
['Venkataraman Santhanam', 'Vlad I. Morariu', 'Larry S. Davis']
1612.03268v1
We present a Deep Convolutional Neural Network architecture which serves as a generic image-to-image regressor that can be trained end-to-end without any further machinery. Our proposed architecture: the Recursively Branched Deconvolutional Network (RBDN) develops a cheap multi-context image representation very early o...
Generalized Deep Image to Image Regression
2,016
http://arxiv.org/pdf/1612.03268v1
Title Generalized Deep Image Image Regression Summary present Deep Convolutional Neural Network architecture serf generic imagetoimage regressor trained endtoend without machinery proposed architecture Recursively Branched Deconvolutional Network RBDN develops cheap multicontext image representation early using efficie...
[-0.005882944446057081, 0.013864055275917053, -0.023329321295022964, 0.05903022736310959, 0.016891922801733017, -0.037013668566942215, 0.06298796087503433, 0.01651635393500328, -0.02697465755045414, 0.030836792662739754, -0.01484741736203432, 0.034815721213817596, 0.009596115909516811, 0.02482055313885212, 0.0074264579...
1,170
1,170
['Forrest Iandola']
1612.06519v1
In recent years, the research community has discovered that deep neural networks (DNNs) and convolutional neural networks (CNNs) can yield higher accuracy than all previous solutions to a broad array of machine learning problems. To our knowledge, there is no single CNN/DNN architecture that solves all problems optimal...
Exploring the Design Space of Deep Convolutional Neural Networks at Large Scale
2,016
http://arxiv.org/pdf/1612.06519v1
Title Exploring Design Space Deep Convolutional Neural Networks Large Scale Summary recent year research community discovered deep neural network DNNs convolutional neural network CNNs yield higher accuracy previous solution broad array machine learning problem knowledge single CNNDNN architecture solves problem optima...
[0.028512990102171898, 0.06864992529153824, -0.026708688586950302, 0.033479638397693634, -0.028080279007554054, -0.028865713626146317, 0.07930532842874527, -0.007254811469465494, -0.05030093342065811, 0.007052831817418337, -0.021591760218143463, 0.02593742311000824, 0.0005915407673455775, 0.055755604058504105, 0.020467...
1,171
1,171
['Abhishek Pandey', 'Anjna Jayant Deen', 'Rajeev Pandey']
1311.6881v1
The improvement of the accuracy of image query retrieval used image classification technique. Image classification is well known technique of supervised learning. The improved method of image classification increases the working efficiency of image query retrieval. For the improvements of classification technique we us...
Color and Shape Content Based Image Classification using RBF Network and PSO Technique: A Survey
2,013
http://arxiv.org/pdf/1311.6881v1
Title Color Shape Content Based Image Classification using RBF Network PSO Technique Survey Summary improvement accuracy image query retrieval used image classification technique Image classification well known technique supervised learning improved method image classification increase working efficiency image query re...
[0.020809000357985497, -0.010857206769287586, 0.0010451372945681214, 0.05201631411910057, -0.01726602576673031, 0.0074258605018258095, -0.010359138250350952, 0.05050471052527428, -0.06310081481933594, -0.01555678527802229, -0.010094397701323032, 0.01655256375670433, 0.019511396065354347, 0.0025938311591744423, -0.03409...
1,172
1,172
['Eren Golge', 'Pinar Duygulu']
1312.4384v1
We attack the problem of learning concepts automatically from noisy web image search results. Going beyond low level attributes, such as colour and texture, we explore weakly-labelled datasets for the learning of higher level concepts, such as scene categories. The idea is based on discovering common characteristics sh...
Rectifying Self Organizing Maps for Automatic Concept Learning from Web Images
2,013
http://arxiv.org/pdf/1312.4384v1
Title Rectifying Self Organizing Maps Automatic Concept Learning Web Images Summary attack problem learning concept automatically noisy web image search result Going beyond low level attribute colour texture explore weaklylabelled datasets learning higher level concept scene category idea based discovering common chara...
[-0.0011109424522146583, -0.022579725831747055, -0.013296166434884071, 0.056584328413009644, 0.022614015266299248, 0.04146831855177879, -0.004697971045970917, 0.017650315538048744, 0.0022688941098749638, -0.013240000233054161, -0.0003524446510709822, 0.02909974940121174, -0.017047520726919174, 0.05096779391169548, -0.0...
1,173
1,173
['Alexey Dosovitskiy', 'Jost Tobias Springenberg', 'Thomas Brox']
1312.5242v3
When deep learning is applied to visual object recognition, data augmentation is often used to generate additional training data without extra labeling cost. It helps to reduce overfitting and increase the performance of the algorithm. In this paper we investigate if it is possible to use data augmentation as the main ...
Unsupervised feature learning by augmenting single images
2,013
http://arxiv.org/pdf/1312.5242v3
Title Unsupervised feature learning augmenting single image Summary deep learning applied visual object recognition data augmentation often used generate additional training data without extra labeling cost help reduce overfitting increase performance algorithm paper investigate possible use data augmentation main comp...
[-0.005550989881157875, 0.035060539841651917, 0.013478443026542664, 0.056349270045757294, 0.013700556010007858, -0.009848319925367832, 0.0743454247713089, -0.025407887995243073, 0.01433628797531128, 0.0034055348951369524, 0.02288675121963024, 0.05241881310939789, 0.01651272550225258, 0.06427125632762909, 0.051534991711...
1,174
1,174
['Ti Wang', 'Daniel L. Silver']
1312.6171v2
This paper presents an unsupervised multi-modal learning system that learns associative representation from two input modalities, or channels, such that input on one channel will correctly generate the associated response at the other and vice versa. In this way, the system develops a kind of supervised classification ...
Learning Paired-associate Images with An Unsupervised Deep Learning Architecture
2,013
http://arxiv.org/pdf/1312.6171v2
Title Learning Pairedassociate Images Unsupervised Deep Learning Architecture Summary paper present unsupervised multimodal learning system learns associative representation two input modality channel input one channel correctly generate associated response vice versa way system develops kind supervised classification ...
[-0.03601371869444847, 0.07168442755937576, -0.030961880460381508, 0.05667264759540558, 0.0034195459447801113, 0.013864521868526936, 0.06111782789230347, -0.01510024257004261, -0.03012257255613804, -0.01575295999646187, -0.08435498178005219, -0.008494822308421135, 0.0052413479425013065, 0.05364517495036125, 0.032898988...
1,175
1,175
['Arjun Jain', 'Jonathan Tompson', 'Mykhaylo Andriluka', 'Graham W. Taylor', 'Christoph Bregler']
1312.7302v6
This paper introduces a new architecture for human pose estimation using a multi- layer convolutional network architecture and a modified learning technique that learns low-level features and higher-level weak spatial models. Unconstrained human pose estimation is one of the hardest problems in computer vision, and our...
Learning Human Pose Estimation Features with Convolutional Networks
2,013
http://arxiv.org/pdf/1312.7302v6
Title Learning Human Pose Estimation Features Convolutional Networks Summary paper introduces new architecture human pose estimation using multi layer convolutional network architecture modified learning technique learns lowlevel feature higherlevel weak spatial model Unconstrained human pose estimation one hardest pro...
[0.0074235908687114716, 0.0371863953769207, -0.011357709765434265, 0.05753004923462868, 0.02945040538907051, 0.0190549548715353, 0.01681080274283886, 0.007177891209721565, 0.05160966143012047, -0.058856405317783356, -0.07568042725324631, -0.045783631503582, 0.032997600734233856, 0.0075264498591423035, 0.064253054559230...
1,176
1,176
['Marius Leordeanu', 'Rahul Sukthankar']
1404.2903v1
We propose a general multi-class visual recognition model, termed the Classifier Graph, which aims to generalize and integrate ideas from many of today's successful hierarchical recognition approaches. Our graph-based model has the advantage of enabling rich interactions between classes from different levels of interpr...
Thoughts on a Recursive Classifier Graph: a Multiclass Network for Deep Object Recognition
2,014
http://arxiv.org/pdf/1404.2903v1
Title Thoughts Recursive Classifier Graph Multiclass Network Deep Object Recognition Summary propose general multiclass visual recognition model termed Classifier Graph aim generalize integrate idea many today successful hierarchical recognition approach graphbased model advantage enabling rich interaction class differ...
[-0.005787791218608618, 0.003931772895157337, 0.012955086305737495, 0.0660126730799675, -0.0074519501067698, 0.020293347537517548, 0.009947988204658031, 0.01825726591050625, 0.04649331793189049, -0.045332975685596466, 0.02396870218217373, 0.01844165101647377, -0.02790277823805809, 0.03639392554759979, -0.00043625530088...
1,177
1,177
['Tsung-Han Chan', 'Kui Jia', 'Shenghua Gao', 'Jiwen Lu', 'Zinan Zeng', 'Yi Ma']
1404.3606v2
In this work, we propose a very simple deep learning network for image classification which comprises only the very basic data processing components: cascaded principal component analysis (PCA), binary hashing, and block-wise histograms. In the proposed architecture, PCA is employed to learn multistage filter banks. It...
PCANet: A Simple Deep Learning Baseline for Image Classification?
2,014
http://arxiv.org/pdf/1404.3606v2
Title PCANet Simple Deep Learning Baseline Image Classification Summary work propose simple deep learning network image classification comprises basic data processing component cascaded principal component analysis PCA binary hashing blockwise histogram proposed architecture PCA employed learn multistage filter bank fo...
[-0.03839147835969925, 0.06020686402916908, -0.024995213374495506, 0.05370697006583214, 0.013671415857970715, 0.024674443528056145, 0.08246254175901413, -0.020163416862487793, 0.014533517882227898, 0.029217323288321495, 0.015147311612963676, -0.017543042078614235, -0.001590044586919248, 0.0821528285741806, 0.0175614263...
1,178
1,178
['Dong Yi', 'Zhen Lei', 'Stan Z. Li']
1407.4979v1
Various hand-crafted features and metric learning methods prevail in the field of person re-identification. Compared to these methods, this paper proposes a more general way that can learn a similarity metric from image pixels directly. By using a "siamese" deep neural network, the proposed method can jointly learn the...
Deep Metric Learning for Practical Person Re-Identification
2,014
http://arxiv.org/pdf/1407.4979v1
Title Deep Metric Learning Practical Person ReIdentification Summary Various handcrafted feature metric learning method prevail field person reidentification Compared method paper proposes general way learn similarity metric image pixel directly using siamese deep neural network proposed method jointly learn color feat...
[-0.012252047657966614, 0.06511197984218597, -0.01900240033864975, 0.03464969992637634, -0.02371952310204506, 0.022564392536878586, 0.04104078188538551, -0.010611026547849178, -0.0336814820766449, 0.05045405030250549, 0.02100403420627117, -0.036398403346538544, 0.02817046456038952, 0.0048348912969231606, 0.019943296909...
1,179
1,179
['Yin Zheng', 'Yu-Jin Zhang', 'Hugo Larochelle']
1409.3970v3
Topic modeling based on latent Dirichlet allocation (LDA) has been a framework of choice to deal with multimodal data, such as in image annotation tasks. Another popular approach to model the multimodal data is through deep neural networks, such as the deep Boltzmann machine (DBM). Recently, a new type of topic model c...
A Deep and Autoregressive Approach for Topic Modeling of Multimodal Data
2,014
http://arxiv.org/pdf/1409.3970v3
Title Deep Autoregressive Approach Topic Modeling Multimodal Data Summary Topic modeling based latent Dirichlet allocation LDA framework choice deal multimodal data image annotation task Another popular approach model multimodal data deep neural network deep Boltzmann machine DBM Recently new type topic model called Do...
[0.021732155233621597, 0.023018820211291313, -0.012090014293789864, 0.03589092940092087, -0.002502666786313057, 0.01629704236984253, 0.02150529995560646, 0.012978853657841682, -0.032115768641233444, -0.07282567024230957, -0.03021206706762314, 0.004710216075181961, -0.0284023005515337, 0.05617635324597359, 0.01252533029...
1,180
1,180
['Arjun Jain', 'Jonathan Tompson', 'Yann LeCun', 'Christoph Bregler']
1409.7963v1
In this work, we propose a novel and efficient method for articulated human pose estimation in videos using a convolutional network architecture, which incorporates both color and motion features. We propose a new human body pose dataset, FLIC-motion, that extends the FLIC dataset with additional motion features. We ap...
MoDeep: A Deep Learning Framework Using Motion Features for Human Pose Estimation
2,014
http://arxiv.org/pdf/1409.7963v1
Title MoDeep Deep Learning Framework Using Motion Features Human Pose Estimation Summary work propose novel efficient method articulated human pose estimation video using convolutional network architecture incorporates color motion feature propose new human body pose dataset FLICmotion extends FLIC dataset additional m...
[-0.0042267958633601665, -0.007699322421103716, -0.012377479113638401, 0.04815269634127617, 0.04338251054286957, 0.02711505815386772, -0.0002202142059104517, 0.037404615432024, -0.02440672554075718, -0.0016269233310595155, -0.021811574697494507, -0.06368982046842575, 0.041435737162828445, -0.005415136460214853, 0.03875...
1,181
1,181
['Yichong Xu', 'Tianjun Xiao', 'Jiaxing Zhang', 'Kuiyuan Yang', 'Zheng Zhang']
1411.6369v1
Even though convolutional neural networks (CNN) has achieved near-human performance in various computer vision tasks, its ability to tolerate scale variations is limited. The popular practise is making the model bigger first, and then train it with data augmentation using extensive scale-jittering. In this paper, we pr...
Scale-Invariant Convolutional Neural Networks
2,014
http://arxiv.org/pdf/1411.6369v1
Title ScaleInvariant Convolutional Neural Networks Summary Even though convolutional neural network CNN achieved nearhuman performance various computer vision task ability tolerate scale variation limited popular practise making model bigger first train data augmentation using extensive scalejittering paper propose sca...
[-0.01994854398071766, -0.02219575084745884, -0.02670830301940441, 0.0791030302643776, 0.0478089265525341, 0.02220674604177475, 0.0777878686785698, 0.013726456090807915, -0.0064918785355985165, 0.011106901802122593, -0.02796213887631893, 0.003878897288814187, -0.022491510957479477, 0.06539906561374664, 0.02948929928243...
1,182
1,182
['Andrea Vedaldi', 'Karel Lenc']
1412.4564v3
MatConvNet is an implementation of Convolutional Neural Networks (CNNs) for MATLAB. The toolbox is designed with an emphasis on simplicity and flexibility. It exposes the building blocks of CNNs as easy-to-use MATLAB functions, providing routines for computing linear convolutions with filter banks, feature pooling, and...
MatConvNet - Convolutional Neural Networks for MATLAB
2,014
http://arxiv.org/pdf/1412.4564v3
Title MatConvNet Convolutional Neural Networks MATLAB Summary MatConvNet implementation Convolutional Neural Networks CNNs MATLAB toolbox designed emphasis simplicity flexibility expose building block CNNs easytouse MATLAB function providing routine computing linear convolution filter bank feature pooling many manner M...
[0.018049243837594986, -0.031610436737537384, -0.010005714371800423, 0.08874107897281647, 0.009966826997697353, -0.01072793547064066, 0.059664323925971985, 0.00903765857219696, -0.012324524112045765, 0.016597742214798927, -0.010806288570165634, 0.06111716851592064, -0.005820184014737606, 0.0485830157995224, 0.020606249...
1,183
1,183
['Brian Cheung', 'Jesse A. Livezey', 'Arjun K. Bansal', 'Bruno A. Olshausen']
1412.6583v4
Deep learning has enjoyed a great deal of success because of its ability to learn useful features for tasks such as classification. But there has been less exploration in learning the factors of variation apart from the classification signal. By augmenting autoencoders with simple regularization terms during training, ...
Discovering Hidden Factors of Variation in Deep Networks
2,014
http://arxiv.org/pdf/1412.6583v4
Title Discovering Hidden Factors Variation Deep Networks Summary Deep learning enjoyed great deal success ability learn useful feature task classification le exploration learning factor variation apart classification signal augmenting autoencoders simple regularization term training demonstrate standard deep architectu...
[-0.0153499199077487, 0.0714932233095169, -0.042611587792634964, 0.056749917566776276, 0.028401462361216545, 0.03299066051840782, 0.07660507410764694, 0.011725285090506077, -0.05483688786625862, 0.025959905236959457, 0.02009459212422371, -0.0007515234174206853, 0.030084634199738503, 0.09917834401130676, -9.331139881396...
1,184
1,184
['Tom Le Paine', 'Pooya Khorrami', 'Wei Han', 'Thomas S. Huang']
1412.6597v4
Convolutional neural networks perform well on object recognition because of a number of recent advances: rectified linear units (ReLUs), data augmentation, dropout, and large labelled datasets. Unsupervised data has been proposed as another way to improve performance. Unfortunately, unsupervised pre-training is not use...
An Analysis of Unsupervised Pre-training in Light of Recent Advances
2,014
http://arxiv.org/pdf/1412.6597v4
Title Analysis Unsupervised Pretraining Light Recent Advances Summary Convolutional neural network perform well object recognition number recent advance rectified linear unit ReLUs data augmentation dropout large labelled datasets Unsupervised data proposed another way improve performance Unfortunately unsupervised pre...
[-0.010164584033191204, 0.05411393195390701, 0.0008122712024487555, 0.05860473960638046, 0.031957127153873444, -0.00696803443133831, 0.06982054561376572, 0.014374001882970333, -3.954367639380507e-05, 0.0065709673799574375, -0.038988012820482254, 0.062486350536346436, -0.03653137758374214, 0.01823771931231022, 0.0044616...
1,185
1,185
['Martin Kiefel', 'Varun Jampani', 'Peter V. Gehler']
1412.6618v3
This paper presents a convolutional layer that is able to process sparse input features. As an example, for image recognition problems this allows an efficient filtering of signals that do not lie on a dense grid (like pixel position), but of more general features (such as color values). The presented algorithm makes u...
Permutohedral Lattice CNNs
2,014
http://arxiv.org/pdf/1412.6618v3
Title Permutohedral Lattice CNNs Summary paper present convolutional layer able process sparse input feature example image recognition problem allows efficient filtering signal lie dense grid like pixel position general feature color value presented algorithm make use permutohedral lattice data structure permutohedral ...
[-0.03173922002315521, 0.05348813161253929, -0.03253931552171707, 0.0744873657822609, -0.03268434479832649, 0.008095175959169865, 0.027235040441155434, 0.03400423005223274, -0.025968344882130623, -0.009324859827756882, -0.07330909371376038, 0.0066679599694907665, -0.02638312242925167, -0.00828834343701601, 0.0202187839...
1,186
1,186
['Jost Tobias Springenberg', 'Alexey Dosovitskiy', 'Thomas Brox', 'Martin Riedmiller']
1412.6806v3
Most modern convolutional neural networks (CNNs) used for object recognition are built using the same principles: Alternating convolution and max-pooling layers followed by a small number of fully connected layers. We re-evaluate the state of the art for object recognition from small images with convolutional networks,...
Striving for Simplicity: The All Convolutional Net
2,014
http://arxiv.org/pdf/1412.6806v3
Title Striving Simplicity Convolutional Net Summary modern convolutional neural network CNNs used object recognition built using principle Alternating convolution maxpooling layer followed small number fully connected layer reevaluate state art object recognition small image convolutional network questioning necessity ...
[-0.0062371729873120785, 0.03633156418800354, 0.004972409922629595, 0.10651324689388275, -0.012324887327849865, -0.017668358981609344, 0.049999576061964035, -0.005501391366124153, -0.02760699763894081, 0.02110479213297367, -0.02075655199587345, 0.05893290787935257, 0.0009062927565537393, 0.03987820819020271, 0.01978810...
1,187
1,187
['Jun Yuan', 'Bingbing Ni', 'Ashraf A. Kassim']
1412.6885v1
The Convolutional Neural Network (CNN) has achieved great success in image classification. The classification model can also be utilized at image or patch level for many other applications, such as object detection and segmentation. In this paper, we propose a whole-image CNN regression model, by removing the full conn...
Half-CNN: A General Framework for Whole-Image Regression
2,014
http://arxiv.org/pdf/1412.6885v1
Title HalfCNN General Framework WholeImage Regression Summary Convolutional Neural Network CNN achieved great success image classification classification model also utilized image patch level many application object detection segmentation paper propose wholeimage CNN regression model removing full connection layer trai...
[0.00815102830529213, 0.03162198141217232, -0.019845256581902504, 0.11706286668777466, 0.01054694876074791, 0.009896422736346722, 0.02459181845188141, 0.022482501342892647, -0.028426798060536385, 0.004469461273401976, 0.03586346283555031, 0.006647802423685789, -0.03321422263979912, 0.040269967168569565, 0.0553890131413...
1,188
1,188
['Soumik Sarkar', 'Vivek Venugopalan', 'Kishore Reddy', 'Michael Giering', 'Julian Ryde', 'Navdeep Jaitly']
1412.7007v3
Occlusion edges in images which correspond to range discontinuity in the scene from the point of view of the observer are an important prerequisite for many vision and mobile robot tasks. Although they can be extracted from range data however extracting them from images and videos would be extremely beneficial. We trai...
Occlusion Edge Detection in RGB-D Frames using Deep Convolutional Networks
2,014
http://arxiv.org/pdf/1412.7007v3
Title Occlusion Edge Detection RGBD Frames using Deep Convolutional Networks Summary Occlusion edge image correspond range discontinuity scene point view observer important prerequisite many vision mobile robot task Although extracted range data however extracting image video would extremely beneficial trained deep con...
[-0.009697893634438515, 0.013901393860578537, 0.028795752674341202, 0.050891898572444916, -0.03288666903972626, -0.03235531970858574, 0.004082001745700836, 0.008668274618685246, -0.02408420667052269, -0.00983634777367115, 0.06335771828889847, 0.058174487203359604, 0.009800058789551258, 0.05620647966861725, -0.047545660...
1,189
1,189
['Matthieu Courbariaux', 'Yoshua Bengio', 'Jean-Pierre David']
1412.7024v5
Multipliers are the most space and power-hungry arithmetic operators of the digital implementation of deep neural networks. We train a set of state-of-the-art neural networks (Maxout networks) on three benchmark datasets: MNIST, CIFAR-10 and SVHN. They are trained with three distinct formats: floating point, fixed poin...
Training deep neural networks with low precision multiplications
2,014
http://arxiv.org/pdf/1412.7024v5
Title Training deep neural network low precision multiplication Summary Multipliers space powerhungry arithmetic operator digital implementation deep neural network train set stateoftheart neural network Maxout network three benchmark datasets MNIST CIFAR10 SVHN trained three distinct format floating point fixed point ...
[-0.054257698357105255, 0.01694902591407299, -0.015204568393528461, -0.020901653915643692, 0.024385927245020866, -0.020834170281887054, 0.07310999929904938, -0.014716248959302902, -0.1146274134516716, -0.012073386460542679, -0.005043957382440567, -0.029618732631206512, 0.04252206161618233, 0.014858455397188663, 0.03239...
1,190
1,190
['Liang-Chieh Chen', 'George Papandreou', 'Iasonas Kokkinos', 'Kevin Murphy', 'Alan L. Yuille']
1412.7062v4
Deep Convolutional Neural Networks (DCNNs) have recently shown state of the art performance in high level vision tasks, such as image classification and object detection. This work brings together methods from DCNNs and probabilistic graphical models for addressing the task of pixel-level classification (also called "s...
Semantic Image Segmentation with Deep Convolutional Nets and Fully Connected CRFs
2,014
http://arxiv.org/pdf/1412.7062v4
Title Semantic Image Segmentation Deep Convolutional Nets Fully Connected CRFs Summary Deep Convolutional Neural Networks DCNNs recently shown state art performance high level vision task image classification object detection work brings together method DCNNs probabilistic graphical model addressing task pixellevel cla...
[0.003917040769010782, 0.012304132804274559, 0.023919017985463142, 0.08240780979394913, -0.002242246875539422, -0.011498323641717434, 0.02763756364583969, -0.02591523714363575, -0.03951062262058258, 0.02086465433239937, 0.02629050798714161, 0.0646139532327652, -0.01955905184149742, 0.013432687148451805, -0.031654633581...
1,191
1,191
['Xingchao Peng', 'Baochen Sun', 'Karim Ali', 'Kate Saenko']
1412.7122v4
Crowdsourced 3D CAD models are becoming easily accessible online, and can potentially generate an infinite number of training images for almost any object category.We show that augmenting the training data of contemporary Deep Convolutional Neural Net (DCNN) models with such synthetic data can be effective, especially ...
Learning Deep Object Detectors from 3D Models
2,014
http://arxiv.org/pdf/1412.7122v4
Title Learning Deep Object Detectors 3D Models Summary Crowdsourced 3D CAD model becoming easily accessible online potentially generate infinite number training image almost object categoryWe show augmenting training data contemporary Deep Convolutional Neural Net DCNN model synthetic data effective especially real tra...
[-0.009264468215405941, 0.018649863079190254, 0.014483417384326458, 0.09390662610530853, 0.0031473583076149225, -0.022199178114533424, 0.020702403038740158, -0.05152898281812668, -0.08301170915365219, -0.006656216457486153, 0.0012119581224396825, 0.042374540120363235, -0.018137751147150993, 0.05643996223807335, 0.00756...
1,192
1,192
['Chelsea Finn', 'Lisa Anne Hendricks', 'Trevor Darrell']
1412.7155v4
Recently, nested dropout was proposed as a method for ordering representation units in autoencoders by their information content, without diminishing reconstruction cost. However, it has only been applied to training fully-connected autoencoders in an unsupervised setting. We explore the impact of nested dropout on the...
Learning Compact Convolutional Neural Networks with Nested Dropout
2,014
http://arxiv.org/pdf/1412.7155v4
Title Learning Compact Convolutional Neural Networks Nested Dropout Summary Recently nested dropout proposed method ordering representation unit autoencoders information content without diminishing reconstruction cost However applied training fullyconnected autoencoders unsupervised setting explore impact nested dropou...
[-0.010077380575239658, 0.06220291182398796, -0.027559589594602585, 0.08705859631299973, 0.0036577200517058372, 0.022866129875183105, 0.06511367112398148, 0.014984921552240849, -0.018148789182305336, -0.02266945131123066, -0.031215855851769447, 0.013403204269707203, 0.02143993228673935, 0.05536461994051933, 0.023879420...
1,193
1,193
['Wentao Zhu', 'Jun Miao', 'Laiyun Qing']
1501.06115v2
Extreme learning machine (ELM) is an extremely fast learning method and has a powerful performance for pattern recognition tasks proven by enormous researches and engineers. However, its good generalization ability is built on large numbers of hidden neurons, which is not beneficial to real time response in the test pr...
Constrained Extreme Learning Machines: A Study on Classification Cases
2,015
http://arxiv.org/pdf/1501.06115v2
Title Constrained Extreme Learning Machines Study Classification Cases Summary Extreme learning machine ELM extremely fast learning method powerful performance pattern recognition task proven enormous research engineer However good generalization ability built large number hidden neuron beneficial real time response te...
[-0.00016215168579947203, -0.031990692019462585, -0.03844224289059639, -0.013342687860131264, 0.015807954594492912, 0.004834046121686697, 0.016038676723837852, 0.0105202104896307, -0.013344862498342991, 0.001643502153456211, 0.009671161882579327, 0.050289250910282135, 0.002268573036417365, 0.04191475734114647, 7.405393...
1,194
1,194
['Jyh-Jing Hwang', 'Tyng-Luh Liu']
1504.01989v1
We address the problem of contour detection via per-pixel classifications of edge point. To facilitate the process, the proposed approach leverages with DenseNet, an efficient implementation of multiscale convolutional neural networks (CNNs), to extract an informative feature vector for each pixel and uses an SVM class...
Pixel-wise Deep Learning for Contour Detection
2,015
http://arxiv.org/pdf/1504.01989v1
Title Pixelwise Deep Learning Contour Detection Summary address problem contour detection via perpixel classification edge point facilitate process proposed approach leverage DenseNet efficient implementation multiscale convolutional neural network CNNs extract informative feature vector pixel us SVM classifier accompl...
[-0.05085853859782219, -0.0011577006662264466, 0.010026364587247372, 0.08999267220497131, -0.028274793177843094, -0.035862162709236145, 0.03042386658489704, 0.00872362032532692, -0.03147992119193077, 0.030692221596837044, 0.015475289896130562, 0.04983939230442047, -0.021221330389380455, 0.07726418972015381, -0.02419399...
1,195
1,195
['Guosheng Hu', 'Yongxin Yang', 'Dong Yi', 'Josef Kittler', 'William Christmas', 'Stan Z. Li', 'Timothy Hospedales']
1504.02351v1
Deep learning, in particular Convolutional Neural Network (CNN), has achieved promising results in face recognition recently. However, it remains an open question: why CNNs work well and how to design a 'good' architecture. The existing works tend to focus on reporting CNN architectures that work well for face recognit...
When Face Recognition Meets with Deep Learning: an Evaluation of Convolutional Neural Networks for Face Recognition
2,015
http://arxiv.org/pdf/1504.02351v1
Title Face Recognition Meets Deep Learning Evaluation Convolutional Neural Networks Face Recognition Summary Deep learning particular Convolutional Neural Network CNN achieved promising result face recognition recently However remains open question CNNs work well design good architecture existing work tend focus report...
[0.02643747627735138, 0.048500701785087585, -0.012422620318830013, 0.07418139278888702, 0.0014857214409857988, 0.032821156084537506, 0.07385306805372238, -0.018928691744804382, 0.004802605602890253, 0.029135094955563545, -0.011722677387297153, 0.004523654468357563, 0.009986836463212967, 0.07897753268480301, 0.030757561...
1,196
1,196
['Sergey Zagoruyko', 'Nikos Komodakis']
1504.03641v1
In this paper we show how to learn directly from image data (i.e., without resorting to manually-designed features) a general similarity function for comparing image patches, which is a task of fundamental importance for many computer vision problems. To encode such a function, we opt for a CNN-based model that is trai...
Learning to Compare Image Patches via Convolutional Neural Networks
2,015
http://arxiv.org/pdf/1504.03641v1
Title Learning Compare Image Patches via Convolutional Neural Networks Summary paper show learn directly image data ie without resorting manuallydesigned feature general similarity function comparing image patch task fundamental importance many computer vision problem encode function opt CNNbased model trained account ...
[-0.04542417451739311, 0.05120190978050232, -0.006027920637279749, 0.10113733261823654, -0.005309062544256449, -0.016205277293920517, 0.019385285675525665, 0.022859914228320122, -0.01735495775938034, -0.0008695810101926327, 0.011831983923912048, -0.0012180284829810262, -0.01644665189087391, 0.04412601888179779, 0.05506...
1,197
1,197
['Jimmy Ba', 'Kevin Swersky', 'Sanja Fidler', 'Ruslan Salakhutdinov']
1506.00511v2
One of the main challenges in Zero-Shot Learning of visual categories is gathering semantic attributes to accompany images. Recent work has shown that learning from textual descriptions, such as Wikipedia articles, avoids the problem of having to explicitly define these attributes. We present a new model that can class...
Predicting Deep Zero-Shot Convolutional Neural Networks using Textual Descriptions
2,015
http://arxiv.org/pdf/1506.00511v2
Title Predicting Deep ZeroShot Convolutional Neural Networks using Textual Descriptions Summary One main challenge ZeroShot Learning visual category gathering semantic attribute accompany image Recent work shown learning textual description Wikipedia article avoids problem explicitly define attribute present new model ...
[0.02694452926516533, 0.045447058975696564, 0.015129673294723034, 0.06988035887479782, 0.008176092058420181, -0.0017006839625537395, 0.0053497664630413055, 0.02700401470065117, -0.03376204892992973, -0.0780453011393547, -0.0039454493671655655, -0.01081142295151949, 0.0072234575636684895, 0.02006075531244278, 0.02587523...
1,198
1,198
['Leslie N. Smith']
1506.01186v6
It is known that the learning rate is the most important hyper-parameter to tune for training deep neural networks. This paper describes a new method for setting the learning rate, named cyclical learning rates, which practically eliminates the need to experimentally find the best values and schedule for the global lea...
Cyclical Learning Rates for Training Neural Networks
2,015
http://arxiv.org/pdf/1506.01186v6
Title Cyclical Learning Rates Training Neural Networks Summary known learning rate important hyperparameter tune training deep neural network paper describes new method setting learning rate named cyclical learning rate practically eliminates need experimentally find best value schedule global learning rate Instead mon...
[-0.02338424324989319, 0.025197019800543785, -0.016003090888261795, 0.03918524459004402, -0.021108048036694527, -0.000512154249008745, 0.01673538237810135, -0.023098839446902275, 0.007070645689964294, 0.03280239924788475, 0.008422127924859524, -0.0058524347841739655, -0.02127740904688835, 0.012290284968912601, -0.03565...
1,199
1,199
['Song Han', 'Jeff Pool', 'John Tran', 'William J. Dally']
1506.02626v3
Neural networks are both computationally intensive and memory intensive, making them difficult to deploy on embedded systems. Also, conventional networks fix the architecture before training starts; as a result, training cannot improve the architecture. To address these limitations, we describe a method to reduce the s...
Learning both Weights and Connections for Efficient Neural Networks
2,015
http://arxiv.org/pdf/1506.02626v3
Title Learning Weights Connections Efficient Neural Networks Summary Neural network computationally intensive memory intensive making difficult deploy embedded system Also conventional network fix architecture training start result training cannot improve architecture address limitation describe method reduce storage c...
[-0.01703660935163498, 0.018625807017087936, -0.024971889331936836, 0.05597003549337387, 0.03314291685819626, -0.03909572586417198, 0.03673790395259857, -0.022958016023039818, -0.020091870799660683, 0.017056426033377647, 0.017175737768411636, 0.03421682119369507, 0.020746098831295967, 0.029795898124575615, 0.0259987786...